166-Application IE Method in Modern Agro-Ecological Park Planning
Uploaded by
justcham2806166-Application IE Method in Modern Agro-Ecological Park Planning
Uploaded by
justcham2806Ershi Qi
Jiang Shen
Runliang Dou Editors
The 19th International
Conference on
Industrial Engineering
and Engineering
Management
Assistive Technology of Industrial
Engineering
The 19th International Conference on Industrial
Engineering and Engineering Management
Ershi Qi Jiang Shen Runliang Dou
• •
Editors
The 19th International
Conference on Industrial
Engineering and Engineering
Management
Assistive Technology of Industrial
Engineering
123
Editors
Ershi Qi Runliang Dou
Jiang Shen College of Management and Economics
Industrial Engineering Institution of CME Tianjin University
Tianjin Tianjin
People’s Republic of China People’s Republic of China
ISBN 978-3-642-38390-8 ISBN 978-3-642-38391-5 (eBook)
DOI 10.1007/978-3-642-38391-5
Springer Heidelberg New York Dordrecht London
Library of Congress Control Number: 2013937993
Ó Springer-Verlag Berlin Heidelberg 2013
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the
work. Duplication of this publication or parts thereof is permitted only under the provisions of
the Copyright Law of the Publisher’s location, in its current version, and permission for use must
always be obtained from Springer. Permissions for use may be obtained through RightsLink at the
Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)
Contents
1 A Bayesian Learning Approach for Making Procurement
Policies Under Price Uncertainty. . . . . . . . . . . . . . . . . . . . . . . . 1
Zhi-xue Xie and Li Zheng
2 A Class of Robust Solution for Linear Bilevel Programming . . . 11
Bo Liu, Bo Li and Yan Li
3 A Comparison of the Modified Likelihood-Ratio-Test-Based
Shewhart and EWMA Control Charts for Monitoring
Binary Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Chao Yin, Yihai He, Zhen Shen and Chun-hui Wu
4 A Decision-Making Model of Price Control
for Administering Authority . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Chong-yi Jing
5 A Method for Multiple Attribute Decision Making without
Weight Information but with Preference Information
on Alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Yun-fei Li
6 A Prediction of the Container Throughput of Jiujiang Port
Based on Grey System Theory . . . . . . . . . . . . . . . . . . . . . . . . . 51
Yan Du
7 A Rapid Data Processing and Assessing Model
for ‘‘Scenario-Response’’ Types Natural Disaster
Emergency Alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Daji Ergu, Gang Kou and Yong Zhang
v
vi Contents
8 A Research on Value of Individual Human Capital
of High-Tech Enterprises Based on the BP Neural
Network Algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Xiu-fen Li and Ping Zhang
9 A Stochastic Programming Model for Evaluating Real
Options in Wind Power Investment Projects . . . . . . . . . . . . . . . 81
Han Qin and L. K. Chu
10 Action Recognition Based on Hierarchical Model. . . . . . . . . . . . 93
Yang-yang Wang, Yang Liu and Jin Xu
11 Adaptive Ant Colony Algorithm Researching
in Cloud Computing Routing Resource Scheduling . . . . . . . . . . 101
Zhi-gao Chen
12 An Empirical Analysis on Guangdong Industrial Energy
Intensity Based on Panel Data. . . . . . . . . . . . . . . . . . . . . . . . . . 109
Xin-dong Hao
13 An Empirical Research on the Ability of Sustainable
Development for Coal Resource Exhausted Cities . . . . . . . . . . . 115
Bing Zhang
14 An Extensive Innovation Procedure to Quality Function
Deployment for Product Design. . . . . . . . . . . . . . . . . . . . . . . . . 125
Chang-tzuoh Wu, Jyh-rong Chou and Chang-shiann Wu
15 Analysis on Grey Relation of Labor Export Mechanism’s
Influence Factors in Poverty-Stricken Areas . . . . . . . . . . . . . . . 137
Shan-ping Wang and Yi Zhou
16 Analysis on Trend of Research and Development
Intensity in China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Bin Huang and Lu-cheng Huang
17 Application of IE Method in Modern Agro-Ecological
Park Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Li Yue
18 Application of Improved Grey Prediction Model
for Chinese Petroleum Consumption . . . . . . . . . . . . . . . . . . . . . 171
Ying Ma and Meng Sun
Contents vii
19 Applying an Integrated SOM Model on Studying
Corporate Governance Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Jen-Ying Shih
20 Applying Constraint Satisfaction Methods in Four-Bar
Linkage Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Gang Chen
21 Appraisal of Estate Enterprise’s Social Responsibility
Based on ANP and Fuzzy Theory . . . . . . . . . . . . . . . . . . . . . . . 199
Chun-hua Wu
22 Bottleneck Detection Method Based on Production Line
Information for Semiconductor Manufacturing System . . . . . . . 209
Xiao-yu Yu, Fei Qiao and Yu-min Ma
23 Clinical Decision Support Model of Heart Disease Diagnosis
Based on Bayesian Networks and Case-Based Reasoning . . . . . . 219
Man Xu and Jiang Shen
24 Construction of Project Management Classification
Frame and Classification Controlling Model Based
on Project Portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Chao Yang and Fa-jie Wei
25 Product Material Combination Image Method
Based on GEP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Ke Su and Sheng-li Kong
26 D2-Index: A Dynamic Index Method for Querying XML
and Semi-Structured Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Yin Zhang, Hua Zhou, Zhi-hong Liang, Jun-hui Liu, Yun Liao,
Peng Duan and Zhen-li He
27 Data Preprocessing in Web Usage Mining . . . . . . . . . . . . . . . . . 257
Xiang-ying Li
28 Design Knowledge Reduction Approach Based
on Rough Sets of HCI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Qing Xue, Qi-qi Yin, Li-ying Feng and Min-xia Liu
29 Diagnostic Testing and Analysis of CAN Data Bus Based
on the Sagitar Power Transmission System . . . . . . . . . . . . . . . . 279
Wen Fang
viii Contents
30 Effectiveness to Driving by Interbeds in Channel Sand
of River Delta System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Lun Zhao, Shu-qin Wang, Zi-fei Fan, Lei Zhang and Ben-wei Li
31 Effectiveness Valuation of Electronic Countermeasure
on Ground Air Defense and Anti-missile . . . . . . . . . . . . . . . . . . 301
Yong-ling Yan, Zhi-feng Zhang, Qing-bo Zhang and Tao Dong
32 Energy Consumption and Economic Growth:
Cointegration and Granger Causality Test. . . . . . . . . . . . . . . . . 313
Jing-wei Liang and Zhou Liu
33 Estimation of Lead Time in the RFID-Enabled Real-Time
Shopfloor Production with a Data Mining Model . . . . . . . . . . . . 321
Ray Y. Zhong, George Q. Huang, Qing-yun Dai and Tao Zhang
34 Evaluation of Green Residence Using Integrated Structural
Equation Model with Fuzzy Theory. . . . . . . . . . . . . . . . . . . . . . 333
Michael H. Hu, M. Y. Ku, C. K. Liao and P. Y. Ding
35 Evaluation of Population Age Structure Model Using
Grey Clustering Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
Bao-ping Chen
36 Evaluation of Recycle Level of Qaidam Salt Lake Circular
Economy with Intuitionistic Fuzzy Entropy Measures . . . . . . . . 353
Fei Feng, Yu Liu, Jian Zhang, Nan Wang and Xiao-hui Xing
37 Evolution Analysis of Standardization Production Behavior
in GI Agricultural Product Enterprise Cluster. . . . . . . . . . . . . . 361
Zhi-fang Li and Tong Chen
38 Forecasting of Vehicle Capacity Based on BP Neural Network . . . 369
Ya-qin An, Ya-jun Wang and Wen-wei Gao
39 Fuzzy Data Mining with TOPSIS for Fuzzy Multiple
Criteria Decision Making Problems. . . . . . . . . . . . . . . . . . . . . . 377
Chin-yao Low and Sung-nung Lin
40 Knowledge Discovery from Granule Features Mining . . . . . . . . 391
Jian-hong Luo, Xi-yong Zhu and Xiao-jun Wang
Contents ix
41 Multivariable Forecasting Model Based on Trend Grey
Correlation Degree and its Application . . . . . . . . . . . . . . . . . . . 403
Li-ping Fu, Lu-yu Wang and Juan Han
42 Partner Selection About the PPP Reclaimed Water Project
Based on Extension Evaluation Method . . . . . . . . . . . . . . . . . . . 411
Xu-kuo Gao and Xiao-hu Chen
43 Personnel BDAR Ability Assessment Model Based
on Bayesian Stochastic Assessment Method . . . . . . . . . . . . . . . . 419
Zhi-feng You, Tian-bin Liu, Ning Ding and Kai-xuan Cui
44 Power Control of Cellular System in CDMA Technology
Integrated with SIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
Pan-dong Zhang and Juan-juan Min
45 Research of Embedded Intelligent Decision Support System
for Natural Disasters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
Jiang Shen, Tao Li and Man Xu
46 Research on Classifying Interim Products in Ship
Manufacturing Based on Clustering Algorithm . . . . . . . . . . . . . 449
Lin Gong, Yan Song, Hao Lin and Zi-xu Chen
47 Research on Financial Accounting Information Disclosure
of China’s Social Security Fund Based on Game Analysis . . . . . 459
Sha-sha Dai, Ke Pan and Yuan-yuan Dai
48 Research on the Multi-Target Tracking Algorithm
Based on Double FPGA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 471
Xu-dong Liu
49 A Study of the Effect of Xi’an Subway on the Rent
of Office Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
Ying Wang and Wen-jia Wei
50 Urban–Rural Income Gap in China: Evolutionary Trend
and Influencing Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
Cun-gui Li
51 Sensitivity Analysis on Optimized Sampling for Sealing
Performance of GVTP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
Long Yu, Shu-rong Yu and Wen-li Yang
x Contents
52 Study on Application of Logistic Curve Fitting
and Forecast from Inbound Tourist Market . . . . . . . . . . . . . . . 509
Wei-qi Tan
53 Study on Decision Mechanism Choosing by Cost Model
for Projectized Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . 519
Hua-ming Zhang
54 Study on Impact of Investor’s Herd Behavior
on Corporation Investment Behavior. . . . . . . . . . . . . . . . . . . . . 527
Hai-ming Wang and De-ming Zeng
55 Study on the Evaluation Framework of Enterprise
Informationization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535
Qing-wen Yuan, Shu-wei Yu, Yuan-yuan Huo and Dan Li
56 Supplier Selection Model Based on the Interval
Grey Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543
Zhi-yong Zhang and Sheng Wu
57 The Accuracy Improvement of Numerical Conformal
Mapping Using the Modified Gram-Schmidt Method . . . . . . . . . 555
Yi-bin Lu, De-an Wu, Ying-zi Wang and Sha-sha Zheng
58 The Application of AHP in Biotechnology Industry
with ERP KSF Implementation . . . . . . . . . . . . . . . . . . . . . . . . 565
Ming-Lang Wang, H. F. Lin and K. W. Wang
59 Application of Gray Correlation Method in the Employment
Analysis of Private Enterprises . . . . . . . . . . . . . . . . . . . . . . . . . 575
Bao-ping Chen
60 The Design of Three-Dimensional Model for the Economic
Evaluation of the Coal Enterprise Informationization . . . . . . . . 583
Qing-wen Yuan and Shu-wei Yu
61 The Effects Decomposition of Carbon Emission Changes
and Countermeasures in Shandong Province:
Based on De Bruyn’s Factors Decomposition Model. . . . . . . . . . 591
Guo-hua Wu, Jin-sheng Hou and Lin Wu
Contents xi
62 The Empirical Research of the Causality Relationship
Between CO2 Emissions Intensity, Energy Consumption
Structure, Energy Intensity and Industrial
Structure in China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 601
Tao Zhao and Xiao-song Ren
63 The Evaluation of China Construction Industry
Sustainable Development on DEA Model . . . . . . . . . . . . . . . . . . 611
Peng-Yang Liu, Jian-Ping Yang and Fan-Fang Wen
64 The Prediction of Customer Retention Costs Based
on Time Series Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
Fan Yu, Ji-fang Yang and Ai-wu Cheng
65 The Research on the Location Selection of the Bank
Outlets Based on Triangular Fuzzy Analytic
Hierarchy Process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
Yang Han and Fa-shan Dai
66 The Study of Sino-Russian Trade Forecasting Based
on the Improved Grey Prediction Model . . . . . . . . . . . . . . . . . . 637
Zhen-zhong Zhang, Shuang Liu and Li-xia Tian
67 The Role of Preference and Emotion in Environmental
Risk Perception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
Charlene Xie, Yang Liu, Shengxiang She and Dixi Song
68 The Model Research on Risk Control . . . . . . . . . . . . . . . . . . . . 653
Qing-hai Zhang
69 TOPSIS Based Power-Saving Plans Choice
for Manufacturing Enterprise . . . . . . . . . . . . . . . . . . . . . . . . . . 661
Dong-sheng Wang and Kuan-ming Zheng
70 Research on Central Control DDSS System
for Fund Portfolio Management . . . . . . . . . . . . . . . . . . . . . . . . 669
Cheng Hu and Er-shi Qi
71 The Evaluation and Application of Residential Structure
System Based on Matter-Element Model . . . . . . . . . . . . . . . . . . 677
Sen Yu and Xiang-ju Liu
xii Contents
72 Logistic Financial Crisis Early-Warning Model Subjoining
Nonfinancial Indexes for Listed Companies . . . . . . . . . . . . . . . . 685
Shao-fang Ding, Ying-chao Hou and Pei-pei Hou
73 Evaluation Research on Logistics Development of the Yangtze
River Port Based on the Principal Component Analysis . . . . . . . 697
Gao Fei
74 A Game Analysis of New Technical Equipment Procurement . . . 703
Ai-hua Zhang, Zhi-wei Yang, Qiong Zhang and Hong-wei Fu
75 Constructing Performance Measurement Indicators
in the Government’ Information Unit in Taiwan:
Using Balanced Scorecard and Fuzzy Analytic
Hierarchy Process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 709
Yi-Hui Liang
76 A Decision Model and Its Algorithm for Vehicle Routing
Problem to Meet Emergencies. . . . . . . . . . . . . . . . . . . . . . . . . . 717
Xing Liu, Min Zhong and Ya-hong Ma
77 A Solution to Optimize Enterprise Business
and Operation Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 727
Xue-wu Chang, Xiao-yuan Ji and Jian-xin Zhou
78 An Approach with Nested Partition for Resource-Constrained
Project Scheduling Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 737
Zhen-yuan Liu and Wen-min Yu
79 An Approximate Dynamic Programming Approach
for Computing Base Stock Levels . . . . . . . . . . . . . . . . . . . . . . . 749
Hong-zhi He
80 An Inventory Model Considering Credit Cost and Demand
Rate Varying with Credit Value . . . . . . . . . . . . . . . . . . . . . . . . 759
Zi-quan Long and Ran Gao
81 An M/M/1 Queue System with Single Working Vacation
and Impatient Customers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 767
Xiao-ming Yu, De-an Wu, Lei Wu, Yi-bing Lu
and Jiang-yan Peng
82 Analysis of Bank Queueing Based on Operations Research . . . . 777
Liang Li, Jia-lin Wu and Jin-xiang Ding
Contents xiii
83 Application of DEA-Malmquist Index in Analyzing
Chinese Banking’s Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . 789
Man Ding, Chao-qun Ma, Zhong-bao Zhou and Wen-bin Liu
84 The Improvement on R. G. Bland’s Method . . . . . . . . . . . . . . . 799
Yu-bo Liao
85 Influence Mechanism of Lean Production to Manufacturing
Enterprises’ Competitiveness. . . . . . . . . . . . . . . . . . . . . . . . . . . 805
Hong-liang Zhang and Zhan-wen Niu
86 Mobile Device User Research in Different Usage Situation . . . . . 815
Wei Liu and Jiu-zhou Li
87 Optimal Enterprise Cash Management Under Uncertainty. . . . . 825
Xiao-yun Wei and Li-yan Han
88 Problem Analysis and Optimizing of Setting Service
Desks in Supermarket Based on M/M/C Queuing System. . . . . . 833
Chun-feng Chai
89 Proposed a Novel Group Scheduling Problem in a Cellular
Manufacturing System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 843
Y. Gholipour-Kanani, N. Aghajani, R. Tavakkoli-Moghaddam
and S. Sadinejad
90 Regional Eco-Environment Optimization Based on Multiple
Parallel Sub-Systems’ Efficiency . . . . . . . . . . . . . . . . . . . . . . . . 855
Li Wang and Ning Li
91 Research on Brand Strategy to Small and Medium-Sized
Enterprises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865
Xin-zhu Li
92 Research on Information Mining About Priority Weight
of Linguistic Judgment Matrix . . . . . . . . . . . . . . . . . . . . . . . . . 873
Cai-feng Li
93 Research on the Project Hierarchy Scheduling for Domestic
Automobile Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 883
Peng Jia, Qi Gao, Zai-ming Jia, Hui Hou and Yun Wang
xiv Contents
94 SVM-Based Multi-Sensor Information Fusion Technology
Research in the Diesel Engine Fault Diagnosis . . . . . . . . . . . . . . 891
Jian-xin Lv, Jia Jia and Chun-ming Zhang
95 The Purchase House Choice Research Based on the Analytic
Hierarchy Process (AHP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 897
Zhi-hong Sun, Lu Pan, Yan-yan Wang and Da-hu Zhang
96 The Relations Tracking Method of Establishing Reachable
Matrix in ISM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 903
Xue-feng He and Yue-wu Jing
97 The Selection of the Regional Center City Under
the Policy of Expanding Domestic Demand . . . . . . . . . . . . . . . . 913
Jian-wei Cheng and Juan Shang
98 A Research on Mine Pressure Monitoring Data Analysis
and Forecast Expert System of Fully Mechanized Coal Face . . . 925
Hong-bing Qiao, Hai-long Xu, Ao-shuang Pang, Chang-dong Zhou
and Yi-lun Wang
99 The U-Shaped Relationship Between Corporate Social
Disclosure and Corporate Performance: Evidence
from Taiwan’s Electronics Industry. . . . . . . . . . . . . . . . . . . . . . 935
Chin-Shien Lin, Ruei-Yuan Chang and Van Thac Dang
100 Empirical Study on the Five-Dimensional Influencing
Factors of Entrepreneurial Performance . . . . . . . . . . . . . . . . . . 945
Xin Lan
101 A Multistart Local Search Heuristic for Knapsack Problem. . . . 957
Geng Lin
102 Heterogeneity of Institutional Investors and Investment Effects:
Empirical Evidence from Chinese Securities Market . . . . . . . . . 963
Ying Jin
103 Integer Linear Programming Model and Greedy Algorithm
for Camping Along the Big Long River Problem . . . . . . . . . . . . 973
Zhen-ping Li and Xiao-dong Huang
104 Research on End Distribution Path Problem
of Dairy Cold Chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 983
Zhen-ping Li and Shan Wang
Contents xv
105 Improved Evolutionary Strategy Genetic Algorithm
for Nonlinear Programming Problems . . . . . . . . . . . . . . . . . . . . 993
Hui-xia Zhu, Fu-lin Wang, Wen-tao Zhang and Qian-ting Li
106 Simulation and Optimization of a Kind of Manufacturing
and Packing Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1005
Chun-you Li
107 Equilibrium and Optimization to the Unequal Game
of Capital-Labor Interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1015
Ming-liang Wang and Yu Lu
108 Innovative and Entrepreneurship Education
in Underdeveloped Western Regions of China . . . . . . . . . . . . . 1023
Chang-jiang Lu, Yan Feng and De-wen Chen
109 Network-Based Optimal Design for International
Shipping System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1031
Er-shi Qi, Lian-yu Zhu and Meng-wei Yu
110 A Dynamic Analytic Approach to Study on the Interaction
Between Product Innovation and Process Innovation
of the Equipment Manufacturing Enterprises . . . . . . . . . . . . . 1037
Ting Wang, Ying Wang, Jing Liu and Yang Gao
111 A Multi-agent Simulation System Considering Psychological
Stress for Fire Evacuation. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1047
Fan-xing Meng, Qin-lin Cai and Wei Zhang
112 A Multi-Granularity Model for Energy Consumption
Simulation and Control of Discrete Manufacturing System . . . 1055
Jun-feng Wang, Shi-qi Li and Ji-hong Liu
113 Analysis on System Archetype of High College
and University Competitiveness Based on
Hierarchical Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1065
Li-qing Li and Ying-ting Yu
114 Analysis on the Operation Effects of Logistics Park Based
on BP Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1073
Jun Luo
xvi Contents
115 Application of Ant Colony Algorithm on Secondary
Cooling Optimization of Continuous Slab . . . . . . . . . . . . . . . . 1081
Ji-yun Li and Hong-xing Pei
116 Application of the Catastrophe Progression Method
in Employment Options for Beijing, Shanghai
and Guangzhou . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1089
Qun Yuan, Ting Chen and Yang Gao
117 Cooperation Relationship Analysis of Research Teams Based
on Social Network Analysis and Importance Measures. . . . . . . 1099
Zheng-sheng Han and Zhi-qiang Cai
118 Establishment of Construction Standard System Based
on the Complex System Theory . . . . . . . . . . . . . . . . . . . . . . . . 1109
Zhi Sun and Shou-jian Zhang
119 Location Selection of Coal Bunker Based
on Particle Swarm Optimization Algorithm . . . . . . . . . . . . . . . 1121
Qing-an Cui and Jing-jing Shen
120 Mechanism of Firm’s Response to Innovation Policy
in Industrial Cluster: Based on Echo Model . . . . . . . . . . . . . . 1129
Yong-an Zhang and Chen-guang Li
121 Modeling and Simulation of Troubleshooting Process
for Automobile Based on Petri Net and Flexsim. . . . . . . . . . . . 1141
Wei-zhen Liao, Kai-hu Hou, Yu-jie Zheng and Xiao-ling He
122 Modeling and Simulation of Wartime Casualty
Surgical Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1155
Kai Zhang, Rui-chang Wu, Yun-dou Wang, Xiao-feng Zhang
and Hai-jian Du
123 Modeling of Shipboard Aircraft Operational Support Process
Based on Hierarchical Timed Colored Petri-Net . . . . . . . . . . . 1167
Ting Wang, Bo-ping Xiao, Lin Ma and Yan-kun Tian
124 Modeling and Simulation of a Just-in-Time Flexible
Manufacturing System Using Petri Nets . . . . . . . . . . . . . . . . . 1179
Yue Cui and Yan-hong Wang
Contents xvii
125 Numerical Simulation of External-Compression Supersonic
Inlet Flow Fields. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1187
Ping Wang, Hong-wei Wang, Si-dong Wei, Xue-shan Liu,
Qing-guo Zhang and Xin Hua
126 Ontology-Based Multi-Enterprise Heterogeneous Model
Fusion Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1195
Hong-xiu Wang
127 Outpatient Scheduling in Highly Constrained Environments:
A Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1203
Xiao-dan Wu, Mohammad T. Khasawneh, Juan Hao
and Zhan-ting Gao
128 Realization of 3D Reconstruction of CAD Model Based
on Slicing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1215
Ming Li and Quan-qing Li
129 Recommender System Based ‘‘Scenario-Response’’
Types Post-Disaster Emergency Supplies Planning . . . . . . . . . . 1225
Gang Kou, Xin Zhao and Daji Ergu
130 Research on the Simulation Case of Traffic Accident . . . . . . . . 1235
Chao Wei, Xiang Gao and Miao-xin Nie
131 Regional Brand Development Model Under the Perspective
of System Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1241
Lai-bin Wang
132 Research on Bayesian Estimation of Time-Varying Delay. . . . . 1251
Meng Wang, Ying Liu and Ji-wang Zhang
133 Research on Design and Analysis Platform
for Modular Spacecraft. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1263
Xiao-wen Zeng, Zhao-xia He and Hao Luo
134 Research on H-Point of Driver Based on Human Physical
Dimensions of Chinese People . . . . . . . . . . . . . . . . . . . . . . . . . 1273
Lin-lin Sun, Fan-sen Kong, Duo-nian Yu, Bao Lan
and Xiang-bei Wang
135 Research on Modeling Framework of Product Service System
Based on Model Driven Architecture. . . . . . . . . . . . . . . . . . . . 1283
Xin-zhi Zhao and Xin Cai
xviii Contents
136 Research on the Civil Aircraft Customer Service System
Simulation Based on SD Model . . . . . . . . . . . . . . . . . . . . . . . . 1291
Jun Xu and Cui-xia Bi
137 Research on the Modeling Method of Wargaming
for Equipment Support on Computer . . . . . . . . . . . . . . . . . . . 1301
Xiao-ming Du, Gui-qi Wang, Ping Gu and Lu Gao
138 Risk Sharing Proportion of Cooperation Between the Banks
and Guarantee Agencies Based on Elman Neural Network. . . . 1309
Jun Liang and Qiang Mei
139 Simulation Analysis on Effect of the Orifice on Injection
Performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1317
Yu-lan Li, Xiang-bi An and Da-hai Jiang
140 Simulation and Optimization Analysis of a Wharf System
Based on Flexsim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1325
Na-qing Lin and Xiao-yan Zhai
141 Simulation Design of Piezoelectric Cantilever Beam Applied
on Railway Track . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1335
Hui Zheng and Fengjing Zheng
142 Simulation of Scenic Spots Dynamic Pricing Based
on Revenue Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1345
Zhi-ping Duan, Shu-lian Yang and Fu-ping Zhang
143 Study of Adaptive Noise Cancellation Controller . . . . . . . . . . . 1353
Cui-jian Zhao and Su-jing Sun
144 Study of Cost-Time-Quality in Project Failure Risk
Assessment Basedon Monte Carlo Simulation . . . . . . . . . . . . . 1361
Xing Pan and Zi-ling Xin
145 Study on Innovation and Practice of Independent Auto
Companies Lean Digital Factory Building . . . . . . . . . . . . . . . . 1369
Yu-chun Wang, Li-fang Wang, Ze-yong Xu, Zhen-kui Li
and Feng-qi Wang
146 The Application of Kernel Estimation in Analysis
of Crime Hot Spots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1379
Yan-yan Wang, Zhi-hong Sun, Lu Pan, Ting Wang
and Da-hu Zhang
Contents xix
147 The Research of Full-Bridge and Current Double Rectifier
Switched-Mode Power Supply for Vehicle . . . . . . . . . . . . . . . . 1387
Yi-lin Yin, Rui-peng Chen and Xi-ming Cheng
148 The Research of Industrial Optimization of Beijing CBD
Ribbon Based on Fitness Function Mode . . . . . . . . . . . . . . . . . 1401
Youliang Zhang and Gang Zong
149 A Study on the User Acceptance Model of SNS Websites
Based TAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1409
Dan Jin and Mei-mei Zhou
150 Augmented Reality Based Factory Model
Comparison Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1421
Wei-wei Sun, Jian-feng Lu and De-zhong Li
151 An Evaluation Method Based on Cloud Model
for the Credit of High-Tech SMEs . . . . . . . . . . . . . . . . . . . . . 1427
Guo-qiang Zhou, Xue-qing Wang, Rui Liu and Li-guo Sun
152 The Structural Optimum Design of Erected Circular
Medicine-Chest Based on Non-Intervention Motion . . . . . . . . . 1437
Zhi-qiang Zhang, Chao Yun, Xiang-quan Liu and Li-yong Wang
153 Application of Noise Estimator with Limited Memory Index
on Flexure Compensation of Rapid Transfer Alignment. . . . . . 1447
Wei-dong Zhou and Yu-ren Ji
154 A Cumulative SaaS Service Evolution Model Based
on Expanded Pi Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1457
Jun He, Tong Li and De-hai Zhang
155 Permanence and Extinction of Periodic Delay Predator–Prey
System with Two Predators and Stage Structure for Prey . . . . 1469
Wei-wei Zheng and Er-dong Han
156 Design of Military Logistics Management System Based
on Internet of Things Technology . . . . . . . . . . . . . . . . . . . . . . 1481
Xue-jiang Wei, Jian-fei Feng, Shao-juan Feng and Yu-zhu Zhai
157 Simulation and Optimization of the Steel Enterprise
Raw Material System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1489
Lin-wei Xu, Pei-qing Wang, Shang-lun Chen and Xing-li Zhong
xx Contents
158 Comprehensive Evaluation and Optimizing
for Boarding Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1499
Da-wei Sun, Xia-yang Zheng, Zi-jun Chen and Hong-min Wang
159 Modeling for Crime Busting . . . . . . . . . . . . . . . . . . . . . . . . . . 1507
Da-wei Sun, Xia-yang Zheng, Zi-jun Chen and Hong-min Wang
160 Personalized Emotion Model Based on Support
Vector Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1519
Jin-bin Wu and Wan-sen Wang
161 Research on Decision-Making Behavior Test System for Top
Management Team Based on Simulation Environment. . . . . . . 1527
Xue-ying Hong, Zhu-chao Yu, Zhu Wang and Yang Jiang
162 An Exploratory System Dynamics Model of Business
Model Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1535
Xiao-lin Shao and Ju-wei Shi
163 An Empirical Study on the Influence of Independent
Directors on Fund Performance in China Fund
Management Company . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1545
Tiao-yan Hui and Fei-lan Lu
164 Research on Optimal Enterprise Contribution of Hunan
Province Based on OLG Model . . . . . . . . . . . . . . . . . . . . . . . . 1557
Ni Yang
165 Comprehensive Experiment Design of Production
Logistics Based on CDIO . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1567
Ying-de Li and Xiu-ju Lan
166 Improved Grey Forecasting Model for Taiwan’s
Green GDP Accounting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1575
Shin-li Lu, Ching-I Lin and Shih-hung Tai
167 Simulation Research of the Fuzzy Torque Control for Hybrid
Electrical Vehicle Based on ADVISOR . . . . . . . . . . . . . . . . . . 1585
Bo-jun Zhang, Yu Wang and Jia-tian Guo
168 Vulnerability Analysis and Assessment System
of Natural Disaster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1593
Jiang Shen, Jing Huang, Tao Li and Man Xu
Contents xxi
169 Application of Actuarial Model in the Forecast of Maternity
Insurance Fund’s Revenue and Expenditure:
A Case Study of Tianjin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1601
Li-ping Fu, Jun Liu, Xue-zhen Chu and Jin-li Fan
170 Study on Process Reengineering of Medium and Small
Coal Machine Manufacture Enterprises. . . . . . . . . . . . . . . . . . 1611
Jing-wen An and Zhi-qiang Zhang
171 The Increase and Decrease Connecting Potential Analysis
on Urban and Rural Residential Land of Tianjin. . . . . . . . . . . 1623
Guo-bin Lin and Shuo Hao
172 Study on Re-Evaluation of Technological Innovation Efficiency
Based on the C2R Improvement Model in Zhongguancun
High-Tech Enterprises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1637
Jing-wen An, Sen Zhang and Guang-lin Sui
173 Research on the Strategic Management of Small
and Medium Enterprises in China. . . . . . . . . . . . . . . . . . . . . . 1649
Xin Zhu and Ying Li
Chapter 1
A Bayesian Learning Approach
for Making Procurement Policies
Under Price Uncertainty
Zhi-xue Xie and Li Zheng
Abstract In this paper we consider a procurement problem under purchase price
uncertainty, which is the case encountered by companies who purchase from spot
markets with fluctuating prices. We develop a procurement model by introducing
the dynamics of information revelation via Bayesian learning, derive its optimal
solution and identify some thresholds to improve purchase timing decisions. Using
historical spot price data of crudes oils, we verify the effectiveness of proposed
policies compared to the current policy of Chinese oil refineries, and find the
Bayesian learning model does perform well—billions of dollars could be saved
over the past several years.
Keywords Bayesian Price uncertainty Procurement management Purchase
timing
1.1 Introduction
Making procurement policies turns to be more and more challenging for manu-
facturers dealing with commodities with violent price fluctuations, such as agri-
cultural products, precious metals, mineral and energy resources. The timing
decision is a particularly hard choice for them; procurement made too early will
cause unnecessary inventory holding costs and miss the opportunity to purchase at
a possibly lower price in the future, while it made too late may squander the
chance to purchase at an earlier cheaper price.
Z. Xie (&) L. Zheng
Department of Industrial Engineering, Tsinghua University, Beijing, China
e-mail: [email protected]
L. Zheng
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_1,
Ó Springer-Verlag Berlin Heidelberg 2013
2 Z. Xie and L. Zheng
This research is motivated by crude oil procurement decision-making problems
faced by an oil refinery in China Petroleum & Chemical Corporation. Under
current policy of importing crude oils from international market, the refinery is
highly exposed to huge price risk, in that its orders are just evenly distributed along
the planning horizons so the settlement prices are generally dependent on the spot
prices during corresponding periods; it is actually playing as a speculator in the
international crude oil market. Given the fact that crude oil procurement cost
accounts for over 85 % of the overall costs of an oil refinery Jiang (2010); Yu and
Fang (2005), obliviously this is not a pleasant situation.
Total cost of crude oil procurement consists of four components, the spot price
of some marker crude, an adjustment factor, freight, and miscellaneous charges,
but the contributions of the latter three are relatively small and less important
compared to high and volatile crude spot prices (we will discuss about the freight
issue a little more in the modeling part); hence it is reasonable for this research to
focus on the oil price. There are two kinds of contracts in physical transactions,
term contract and spot contract. The term contract in crude procurement practice,
however, is quite different from commonly known long-term fixed-price contract
in other supply chains; the unit price of purchased crude oil under a term contract
is determined on the basis of spot prices in a certain period around its lifting time
rather than settled at the time of ordering, the same as under a spot contract.
Focused on the spot price issue, we do not distinguish between these two types of
contracts hereafter.
Oil prices have historically tended to be notoriously difficult to predict, resulted
from a combination of fundamentals and market behaviors, financial factors such as
oil futures prices and exchange rates, as well as economic events, military affairs
and politics. This complexity in price uncertainty, together with the fact crude oil is
the most influential commodity, makes our study more significant. And this
research, though aimed at Chinese refineries, provides potential benefits to world-
wide manufactures in commodity industries that bear some resemblance with it.
The remainder of this paper is organized as follows. In Sect. 1.2, the relevant
literature is reviewed. The procurement problem is modeled in Sect. 1.3, and the
analytical solutions are provided. Next in Sect. 1.4, real-world data are used to test
the effectiveness of proposed decision policies. Conclusions are given in Sect. 1.5
with some important directions for future research.
1.2 Literature Review
There are two lines of literature related to this research, procurement models under
purchase price uncertainty, and operations management studies involving infor-
mation updating.
1 A Bayesian Learning Approach for Making Procurement 3
1.2.1 Procurement Under Purchase Price Uncertainty
As many as procurement and inventory management researches in operations
management and supply chain management literature, they generally assume
constant or known purchase prices and work on fulfilling stochastic demands;
studies on stochastic purchase price uncertainty are limited. Fabian et al. (1959) is
probably the first inventory research that considers price uncertainty under sto-
chastic demand and Golabi (1985) addresses a problem with random ordering
prices and deterministic demands; but they both assume known and independent
distribution function for future prices, which are not applicable to commodity
prices. Kalymon (1971) studies an inventory model in which the purchase price
follows a Markovian process, (Yi J, Scheller-Wolf A Dual sourcing from a regular
supplier and a spot Market. unpublished) considers a multiple sourcing problem
with a fixed-price supply mode and a stochastic price mode, and lately Berling and
Martínez-de-Albéniz (2011) develops an effective procedure to characterize
optimal base-stock levels based on earlier structural results. A work closely related
to ours is Li and Kouvelis (1999) because we assume the same demands pattern
and price processes as one of their cases does. However, former inventory control
models all assume spot price value becomes known before ordering decision is
made to obtain some price-dependent policies. Our study is unique in modeling the
difference between spot price at the time of ordering and that at purchasing due to
industrial practices, and we also incorporate information revelation dynamics
within the procurement problem.
1.2.2 Information Update in Operations Management
Several authors have studied information acquisition and forecast updates in
procurement problems. (Gaur V, Seshadri S, Subrahmanyam MG Optimal timing
of inventory decisions with price uncertainty, unpublished) Considers an inventory
timing decision with forecasts updating for correlated demand and selling price,
but deterministic purchase cost. Gurnani and Tang (1999) proposes a two period
procurement model for seasonal product with uncertain demand and purchase
price, but only the demand is updated upon market signals between two selling
instants. (Secomandi N, Kekre S Commodity procurement with demand forecast
and forward price updates, unpublished) Derives the optimal procurement policy
under correlated demand forecast and forward price updates in the presence of
forward and spot markets. The existing literature generally focuses on the trade-off
between more accurate forecasts on demand and (potentially) higher costs, while
our problem is featured by deterministic demands and independent uncertain
purchase prices.
Bayesian method is not popular in procurement models, but has been exten-
sively used in inventory models [Scarf (1959); Karlin (1960); Azoury (1985)].
4 Z. Xie and L. Zheng
However, they are all about uncertain demands which distribution is periodically
updated based on newly obtained demand observations. Inspiringly, Miller et al.
(2005) incorporates Bayesian learning within a process design and capacity
planning problem and identifies a threshold to improve decision-making; we adopt
their framework and apply it to our procurement problem.
1.3 The Procurement Models
Consider an oil refinery faced with multiple dynamic but deterministic demands
for crude oils, which must be met and will be purchased at spot prices, and
amounts arrived before demanded trigger inventory holding costs. The assumption
of known demands pattern is realistic in this problem based on three reasons. First,
various kinds of oil products could serve several purposes such as generating
power, providing energies to automobiles and airplanes, etc., consequently the
overall oil demands arising from both life and industry do not fluctuate violently,
especially compared with the volatility of crude oil price. Second, since there are
industry regulations in Chinese oil markets to ensure relatively steady supply of oil
products, major refineries usually report their production plans yearly to the
administration and stick to them after being approved. Third, process industries
such as oil refining, characterized by continuous production process that couldn’t
afford to be disturbed too much, generally operate according to the determined
production plans out of economic considerations.
1.3.1 Basics of Cost Modeling
As researchers studying on the spot price of commodities did, we assume that the
spot price of each unit (e.g., barrel) of the crude oil, St, follows a geometric
Brownian motion with drift l and volatility r
dSt =St ¼ l dt þ r dWt ; ð1:1Þ
where Wt is a standard Weiner process. Note the point in time that matters is when
payment is made, not when order is placed. The terms on valuation time periods
are negotiable in practice, so for simplicity, the purchasing price is represented by
the spot price at delivery in following derivations and by the average of spot prices
during the lead time in empirical studies.
As inventory holding cost consists of cost of capital, cost of storage, insurance,
breakage, and many other items, among which cost of capital is the most signif-
icant component, we assume the inventory holding cost (per unit per time) is
charged at a fixed proportion, 0 \ h \ 1, to the corresponding purchasing cost.
That is,
1 A Bayesian Learning Approach for Making Procurement 5
ht ¼ h St : ð1:2Þ
In our problem, modeling in this way is more reasonable than commonly used
stationary or independent setting, because the cost of capital is particularly high for
the oil industry, which is capital-intensive.
What’s more, it is realistic to assume a lead time, say, one period, i.e., the
ordered quantity in current period will be delivered and paid in next period. Such a
setting also emphasizes the difference between the spot price observed when the
crude oils are ordered and the price to pay when they are received. Assume the
time length of one period is Dt.
1.3.2 Purchasing Timing Problem
The refinery’s decision is to determine when and how many units to purchase each
time to minimize the total expected costs over the planning horizon, in other
words, the refinery would behave risk-neutrally when making procurement deci-
sions. Risk-aversion modeling can be found in Martínez-de-Albéniz and Simchi-
Levi (2006), and it could be incorporated in the model by adjusting the drift term.
In this procurement problem, the refinery has a series of choices on ordering
timings for each quantity demanded and every single period in the planning
horizon is an option. Say there are n demands in the planning horizon, because
supply lead times are fixed, order i will always be used to serve the ith demand,
which arrival time and quantity are denoted by Ti and Di. We want to find an
optimal series of arrival times of each ordered quantity, denoting the ith order’s
delivery time as ti (thus its ordering time is ti - Dt), to obtain a minimization as
X
n
min Ef erti ½Di Sti þ Di hti ðTi ti Þg; ð1:3Þ
ti
i¼1
where r is an appropriate discount rate. The procurement problem can be solved
independently, as shown in Li and Kouvelis (1999) and Miller et al. (2005), for
each demand as
min Eferti Di Sti ½1 þ hðTi ti Þg: ð1:4Þ
ti
1.3.3 A Passive Decision Model
Let F (t, St) be the minimum expected discounted total cost if the current time is t,
the current unit price is St, and the firm has not made the purchase before. The
Bellman equation is
6 Z. Xie and L. Zheng
n o
F ðt; St Þ ¼ min erðtþDtÞ D E½StþDt jSt ½1 þ h ðT t DtÞ; E½F ðt þ Dt; StþDt ÞjSt ;
ð1:5Þ
where the first and second terms in braces on the right side of Eq. (1.5) are the
‘‘payoff’’ functions of termination and continuation, respectively. The objective
function is F (0, S0) where S0 is the spot price of the material at time 0; the
boundary condition is
F ðT Dt; STDt Þ ¼ erT D E½ST jSTDt : ð1:6Þ
Now we are ready to derive the optimal timing decision, and a passive decision
rule can be obtained as follows (we will come back to the meaning of ‘‘passive’’
later).
Property 1 The optimal timing decision is to order at time 0 (to purchase at time
Dt ) if
eðlrÞ ðTDtÞ hðT DtÞ 1 [ 0; ð1:7Þ
and to order at time T-Dt (to purchase at time T) otherwise. Accordingly, the
optimal value of objective function (minimum cost) can be derived in a straight-
forward way.
Note that the optimal purchasing decision for the stochastic case is the same as
if the price volatility is zero, which makes sense since we are optimizing on the
expectation level where the stochastic problem reduces to have the similar prop-
erty as in the deterministic case. Also as interpreted by Li and Kouvelis (1999),
because observing the price path of a geometric Brownian motion gives us no new
information about the distribution of price changes of the material in the future, the
optimal timing for the firm’s purchase can be determined at time 0.
However, this ‘‘expected scenario’’ analysis makes an implicit assumption that
the decision-maker acts passively following the decision made at the beginning of
planning horizon, which can be problematic in the actual marketplace character-
ized by change and uncertainty, because the realization of spot prices will probably
be quite different from what the decision-maker expected initially. Therefore, the
passive decision models works well only when the price trend remains the same
along the planning horizon, which situation is rare for crude oils in this decade.
Note the drift l at time 0 is actually what we have estimated without any future
spot prices; then a natural thought is, if some future price realizations suggest a
change in price trend, we don’t have to stick to the former decisions. Such flex-
ibility of management serves as the central idea of real options literature. To
address this passive management problem, we introduce Bayesian learning to
actively make use of new information; that is the ‘‘active’’ decision model.
1 A Bayesian Learning Approach for Making Procurement 7
1.3.4 An Active Decision Model
Suppose we obtain at time 0 an estimate of future spot price that follows a log-
normal distribution; its logarithm is normally distributed with an unknown mean l ~
and a known variance r. We may regard l ~ as a random variable and learn more
about it under Bayes rule. Presume the prior belief is normally distributed as
N (l00, r00). Spot prices between period 0 and period 1 are observed as time
evolves from 0 to Dt, denoted by s1k. There could be more than one price real-
izations (k [ 1) between consecutive decision points in time, because the spot
price process evolves in a continuous pattern from which we can obtain several
discrete realizations; for instance, we make ordering decisions per month while
obtain price information per day. Then, as shown in, (Fink D A compendium of
conjugate priors, unpublished) the posterior distribution of l ~ is also normal with
mean and standard variance
pffiffiffiffiffi
l01 ¼ a1 l00 þ ð1 a1 Þs1 ; r01 ¼ a1 r00
ð1:8Þ
a1 ¼ r2 ðr2 þ kr200 Þ;
where s1 is the mean statistic of logarithms of spot price data.
Now we are most concerned with whether there is any opportunity to improve
the passive decision made at time 0 when additional information is gathered.
Property 2 If the purchasing decision at time 0 is to order at time T-Dt , then at
period 1 the threshold of the newly gathered information about price realizations
that reverses the prior decision is
r þ ln½hðT2DtÞþ1
2
r2 a1 l00
s1 ¼ T2Dt
ð1:9Þ
1 a1
where T-2Dt [ 0. In other words, we should advance the ordering time to now as
long as s1 [ s*1, and hold on to the prior decision otherwise.
If the prior decision remains unchanged at period 1, then after new information
is obtained at later periods, similar procedure should be followed. And it is
straightforward to extend Property 2 to later periods.
1.4 Case Applications
Now we apply the proposed models to the procurement problems faced by Chinese
refineries, and the empirical studies are based on real-world data in crude oil spot
market. The benchmark model is the current policies of Chinese refineries, under
which the orders are almost evenly distributed along the planning horizons.
8 Z. Xie and L. Zheng
1.4.1 Data and Procedure
Our data set comes from public data source (Energy Information Administration)
and consists of daily spot prices (USD per barrel) of WTI from 1986/1/2 to 2011/6/
30 and those of Brent from 1988/1/4 to 2011/6/30. As for the quantity of crude oils
China has imported from the international market, there are yearly volume data
from 1994 to 2009, summing up to approximately 10.165 billion barrels.
We set half a month as the length of one period, which is also the length of lead
time and decision making interval, and two months (four periods) as the length of
one planning horizon. Taking case ‘‘demand for WTI at 2008/4/30’’ as an example,
a typical data processing procedure is as follows:
(1) Set 2008/3/1 as time 0, 2008/4/30 as time T (N = 4), target demand to be
served equals to 1; annualized risk-free rate equals to 3 %, which is approx-
imately the one-year Treasury bond rate; annualized cost of capital as per-
centage to occupied capital equals to 36 %, which is relatively high because
the oil industry is capital-intensive, as emphasized before.
(2) Use historical data, WTI’s spot prices from 2008/1/1 to 2008/2/29, to estimate
l00 and r at time 0; here we estimate from data of the past two months because
we want to make decision for future two months. Make passive decision
according to Property 1.
(3) For period 1 to N-1, if the order has not been placed before, consider spot
prices data in each month as additional information; calculate parameters and
statistics to make active decision according to Property 2.
(4) Use real-world price data to statistically assess the effectiveness of proposed
decision models compared to the benchmark, in which the costs are calculated
by taking average of the costs along the planning horizons.
1.4.2 Model Assessment
Faced with spot prices data from 1986/1988 to 2011 and volume data from 1994 to
2009, we perform rough yet reasonable calculations for the total costs by multi-
plying average unit purchase cost over 24/26 years by aggregate quantities of
16 years. We compare the proposed models to the current policy, and the results
are summarized in Table 1.1.
Table 1.1 Effectiveness of proposed models
Marker crude Total procurement costsa
Current policy Passive model Active model
WTI 369.36 370.59 364.74
Brent 374.94 377.49 371.97
a
In billions of US dollars
1 A Bayesian Learning Approach for Making Procurement 9
Surprisingly, the passive model costs more than the current policy, perhaps
because the oil price trend in the real world changes so frequently so it’s no good
for the refinery to act passively following the decision made at the beginning of
planning horizon. However, the active model did save a lot of money for the
refinery. Note all the numbers are in billions of US dollars; although they are not
accurate, improvements are still considerable.
1.5 Conclusion
In this paper, we develop procurement policies, particularly for Chinese oil
refineries, under purchase price uncertainty. Faced with difference between spot
price at the time of ordering and that at purchasing, we formulate a procurement
model and derive its optimal solution; but the model seems to be problematic when
price trend changes. To address this weakness, we then incorporate Bayesian
learning within the decision making process and identify a threshold of newly
observed spot prices to improve the procurement timing decisions. Finally, we
empirically validate the effectiveness of these proposed policies using public spot
prices data for WTI from 1986 to 2011 and Brent from 1988 to 2011. Now we
come to the conclusions that the order should be made at time 0 as suggested by
the passive model only if the decision-maker has enough evidences to believe the
oil price will increase at a relatively high rate throughout the planning horizon,
otherwise the active model is worth trying, which is practically effective under not
too long planning horizons.
The intrinsic logic of the Bayesian approach is that the thought process of
judging oil prices to be high or low is a natural Bayesian process: when you see a
realized price, you may think it as in high or low level and guess it will go up or
down in the future; when the next period comes, you will see whether your judg-
ment is right. If you have some opportunity to revise former decisions, the chance of
your making good decisions would be bigger. This is the basic idea of the active
model, and it indeed works well as proved by real data. However, our procurement
model needs extensions to deal with some more issues. One significant direction of
expending the current problem is to include an inventory capacity according to the
actual constraint, then the model will be more applicable under longer planning
horizons. Another direction is to address the problems of optimal hedging portfolio
and its interaction with operations management, because financial hedging may be
taken into consideration by Chinese oil companies in the future.
Acknowledgments This study is supported in part by the National Natural Science Foundation
of China (Grant No. 70771058/60834004), and the 863 Program of China (2008AA04Z102).
The authors thank H. S. Deng, X. F. Li, A. B. Pang, and G. P. Xiao from the China Petroleum
& Chemical Corporation for the detailed information on refinery procurement operations. The
authors also thank Prof. C. S. Park from Auburn University for his patient instructions on
Bayesian learning approach.
10 Z. Xie and L. Zheng
References
Azoury KS (1985) Bayes solution to dynamic inventory models under unknown demand
distribution. Manage Sci 31(9):1150–1160
Berling P, Martínez-de-Albéniz V (2011) Optimal inventory policies when purchase price and
demand are stochastic. Oper Res 59(1):109–124
Fabian T, Fisher JL, Sasieni MW, Yardeni A (1959) Purchasing raw material on a fluctuating
market. Oper Res 7(1):107–122
Fink D A compendium of conjugate priors, unpublished
Gaur V, Seshadri S, Subrahmanyam MG Optimal timing of inventory decisions with price
uncertainty, unpublished
Golabi K (1985) Optimal inventory policies when ordering prices are random. Oper Res
33(3):575–588
Gurnani H, Tang C (1999) Optimal ordering decisions with uncertain cost and demand forecast
updating. Manage Sci 45(10):1456–1462
Jiang XY (2010) The risk analysis of imported crude oil valuation (Chinese). Petrol Petrochem
Today 18(6):41–44
Kalymon BA (1971) Stochastic prices in a single-item inventory purchasing model. Oper Res
19(6):1434–1458
Karlin S (1960) Dynamic inventory policy with varying stochastic demands. Manage Sci
6(3):231–258
Li C, Kouvelis P (1999) Flexible and risk-sharing supply contracts under price uncertainty.
Manage Sci 45(10):1378–1398
Martínez-de-Albéniz V, Simchi-Levi D (2006) Mean-variance trade-offs in supply contracts. Nav
Res Logist 53(7):603–616
Miller LT, Park CS (2005) A learning real options framework with application to process design
and capacity planning. Prod Oper Manage 14(1):5–20
Scarf H (1959) Bayes solutions of the statistical inventory problem. Ann Math Stat
30(2):490–508
Secomandi N, Kekre S Commodity procurement with demand forecast and forward price
updates, unpublished
Yi J, Scheller-Wolf A Dual sourcing from a regular supplier and a spot Market. unpublished
Yu C, Fang J (2005) Optimization and control over the purchasing costs of imported crude oil (In
Chinese). Int Petrol Econ 13(8):44–46
Chapter 2
A Class of Robust Solution for Linear
Bilevel Programming
Bo Liu, Bo Li and Yan Li
Abstract Under the way of the centralized decision-making, the linear bi-level
programming (BLP) whose coefficients are supposed to be unknown but bounded
in box disturbance set is studied. Accordingly, a class of robust solution for linear
BLP is defined, and the original uncertain BLP was converted to the deterministic
triple level programming, then a solving process is proposed for the robust solu-
tion. Finally, a numerical example is shown to demonstrate the effectiveness and
feasibility of the algorithm.
Keywords Box disturbance Linear bilevel programming Robust optimization
Robust solution
2.1 Introduction
Bilevel programming (BLP) is the model with leader-follower hierarchical
structure, which makes the parameter optimization problems as the constraints
(Dempe 2002). In its decision framework, the upper level programming is con-
nected with not only the decision variables in its level but also with the optimal
solution in the lower level programming, while the optimal solution in the lower
lever programming is affected by decision variables in the upper level
B. Liu B. Li Y. Li
School of Management, Tianjin University,
Tianjin, China
B. Liu (&)
School of Information Science and Technology, Shihezi
University, Xinjiang, China
e-mail: [email protected]
Y. Li
School of Science, Shihezi University, Xinjiang, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 11
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_2,
Ó Springer-Verlag Berlin Heidelberg 2013
12 B. Liu et al.
programming. Due to the leader-follower hierarchical structure problems widely
exist in the realistic decision-making environment, the scholars have been paying
great attention to BLP and have brought about good results on the theory and
algorithms (Bialas and Karwan 1982; Fortuny-Amat and McCarl 1981; Mathieu
et al. 1994; Lai 1996). Some degree of uncertainty exists in realistic decision-
making environment, such as the inevitable error of measuring instrument in data
collection, incompleteness in data information, the approximate handle for the
model and other factors; hence it is necessary to study on the uncertain Bilevel
programming. For the uncertainty problem, the fuzzy optimization and stochastic
optimization have been applied widely. However, it is difficult for decision-makers
to give the precise distribution functions or membership functions which are
required in above methods. Thus, the robust optimization become an important
method, because it can seek for the best solution for the uncertain input without
considering the parameter distribution of uncertain parameters and is immune from
the uncertain data (Soyster 1973).
For the uncertain BLP, the definition of robust solution is influenced by the
dependent degree of the upper and lower levels in the decision-making process.
When the dependent degree is relative independence, the robust solution to the
uncertain BLP is defined by the way of the decentralized decision-making (Li and
Du 2011); when the dependent degree is relative dependence, the robust solution
to the uncertain BLP is defined by the way of the centralized decision-making, that
is, when the lower level seeks its own robust solution, it considers the influence to
the robust solution of the upper level firstly. In the paper, the latter case will be
discussed, and the coefficients of BLP are supposed to be unknown but bounded in
box disturbance set. By the transform of the uncertain model, the robust solution of
BLP is obtained. Finally, a numerical example is shown to demonstrate the
effectiveness and feasibility of the algorithm.
2.2 The Definition of Robust BLP
2.2.1 The Model and the Definition
In this paper we consider Linear BLP formulated as follows:
min Fðx; yÞ ¼ cT1 x þ d1T y
x
s:t: where y solves
min f ðx; yÞ ¼ cT2 x þ d2T y ð2:1Þ
y
s:t: Ax þ By h
x; y 0
2 A Class of Robust Solution for Linear Bilevel Programming 13
In model (2.1),
x 2 Rm1 ; y 2 Rn1 ; cl 2 Rm1 ; dl 2 Rn1 ; l 2 f1; 2g;
A 2 Rrm ; B 2 Rrn ; h 2 Rr1 ;
there is some uncertainty or variation in the parameters c1 ; d1 ; c2 ; d2 ; A; B; h. Let
ðc1 ; d1 ; c2 ; d2 ; A; B; hÞ 2 l, l is a given uncertainty set in Box disturbance as
follows:
8 9
> cli ¼ c þ ðuc Þ ; ðuc Þ ðuc Þ ðuc Þ >
>
> li l i l i l i l i >
>
>
> >>
>
> d lj ¼ d þ ð u Þ
dl j ; ð u Þ
dl j ð u Þ
dl j ð u Þ
dl j >
>
>
> lj >
>
>
> >
>
< aki ¼ aki þ ðuA Þki ; ðuA Þki ðuA Þki ðuA Þki >
>
=
l :¼ ðcl ; d l ; A; B; hÞ bkj ¼ bkj þ ðuB Þkj ; ðuB Þkj ðuB Þkj ðuB Þkj ð2:2Þ
>
> >
>
>
>
hk ¼ hk þ ð uh Þ k ; ð uh Þ k ð uh Þ k ð uh Þ k >
>
>
> >
>
>
> >
>
>
> l ¼ f1; 2g; i 2 f 1; . . .; mg; >
>
>
> >
>
: j 2 f1; . . .; ng; k ¼ f1; . . .; r g: ;
For l ¼ f1; 2g; i 2 f1; . . .; mg; j 2 f1; . . .; ng; k ¼ f1; . . .; r g, cli ; dlj ; aki ; dkj ; hk
are the given data, and ðucl Þi ; ðudl Þj ; ðuA Þki ; ðuB Þkj ; ðuh Þk are the given nonnegative
data.
Under the way of the centralized decision-making, the robust solution of
uncertain BLP (1) is defined as follows:
Definition 1
(1) Constraint region of the linear BLP (1):
X ¼ fðx; yÞjAx þ By h; x; y 0; ðA; B; hÞ 2 lg
(2) Feasible set for the follower for each fixed x
XðxÞ ¼ fyjAx þ By h; x; y 0; ðA; B; hÞ 2 lg
(3) Follower’s rational reaction set for each fixed x
( ( ))
cT2 x þ d2T y; y 2 XðxÞ;
MðxÞ ¼ yy 2 arg min
ðA; B; hÞ 2 l
(4) Inducible region:
IR ¼ fðx; yÞjðx; yÞ 2 X; y 2 MðxÞg:
14 B. Liu et al.
Definition 2 Let
( )
cT1 x þ d1T y t
m n
F :¼ ðx; y; tÞ 2 R R R
ðx; yÞ; ðc1 ; d1 Þ 2 l
The programming
minftjðx; y; tÞ 2 F g ð2:3Þ
x;y;t
is defined as robust counterpart of uncertain linear BLP(1); F is defined as the
robust feasible set of uncertain linear BLP(1).
2.2.2 The Transform of Uncertain BLP Model
Under the way of the centralized decision-making, based on the original idea of
robust optimization that the objective function can get the optimal solution even in
the worst and uncertain situation, the transform theorem can be described as
followings:
Theorem The robust linear BLP (1) with its coefficients unknown but bounded in
box disturbance set l is equivalent to Model (2.4) with certain coefficients as
followings:
X
m n
X
min Fðx; yÞ ¼ c1i þ lc1 i xi þ
d1j þ ld1 j yj
x
i¼1 j¼1
s:t: where d2 solves
Xm n
X
max c1i þ lc1 i xi þ
d1j þ ld1 j yj
d2
i¼1 j¼1
s:t: d2j ðud2 Þj
d2j d2j þ ðud2 Þj ; j ¼ f1; . . .; ng;
ð2:4Þ
where y solves
min f ðx; yÞ ¼ d2T y
y
X
m n
X
s:t: aki ðlA Þki xi þ bkj ðlB Þkj yj hk þ ðlh Þk ;
i¼1 j¼1
k ¼ f1; . . .; r g
x; y 0:
Proof (1) Firstly, the constraint region X of the linear BLP (1) is transformed into
the certain region. Consider the constraint region of the linear BLP (1):
2 A Class of Robust Solution for Linear Bilevel Programming 15
X ¼ fðx; yÞjAx þ By h; x; y 0; ðA; B; hÞ 2 lg
According to the process of the transformation (Lobo et al. 1998), we can
obtain
Ax þ By h; ðA; B; hÞ 2 l
8 9
> aki ¼ a þ ðuA Þ ; >
>
> ki ki >
>
>
> >
>
>
> ð u Þ
ðu Þ ð u Þ
; >
>
>
> A ki A ki A ki >
>
>
> >
>
>
> b ¼ b
þ ð u Þ ; >
>
>
> kj kj B kj >
>
>
> >
>
>
<X >
=
m X n ðuB Þkj ðuB Þkj ðuB Þkj ;
, 0 min aki xi þ
bkj yj hk
lA ;lB ;lh >
> hk ¼ hk þ ðuh Þk ; >
>
>
> i¼1 j¼1
>
>
>
> >
>
>
> ð u Þ
ð u Þ ð u Þ
; >
>
>
> h k h k h k >
>
>
> >
>
>
> i 2 f1; . . .; mg; j 2 f1; . . .; ng; > >
>
> >
>
>
> >
>
: k 2 f1; . . .; rg ;
Xm Xn
, 0 aki xi þ bkj yj hk
i¼1 j¼1
8 9
> ðuA Þki ðuA Þki ðuA Þki >
>
> >
>
>
> >
>
>
> ðuB Þkj ðuB Þkj ðuB Þkj >
>
>
<X >
=
m X n
þ min ðuA Þki xi þ
ðuB Þkj yj ðuh Þk ðuh Þk ðuh Þk ðuh Þk
lA ;lB ;lh >
> >
>
>
> i¼1 j¼1
i 2 f1; . . .; mg; j 2 f1; . . .; ng; >
>
>
> >
>
>
> >
>
: k 2 f1; . . .; rg ;
X m X n
x;y 0
, 0 aki ðuA Þki xi þ bkj ðuB Þkj yj hk þ ðuh Þk
8k2f1;...;rg
i¼1 j¼1
X
m n
X
, aki ðuA Þki xi þ bkj ðuB Þkj yj hk þ ðuh Þk ; k 2 f1; . . .; rg
i¼1 j¼1
ð2:5Þ
So the linear BLP (1) is transformed into the model (2.6) as followings:
16 B. Liu et al.
min Fðx; yÞ ¼ cT1 x þ d1T y
x
s:t: where y solves
min f ðx; yÞ ¼ cT2 x þ d2T y
y
X
m n
X ð2:6Þ
s:t: aki ðuA Þki xi þ bkj ðuB Þkj yj hk þ ðuh Þk
i¼1 j¼1
k ¼ f1; . . .; r g
x; y 0
(2) Next, according the equivalent form (Lobo et al. 1998)
min t
min f ðxÞ x;t
x , s:t: f ðxÞ t
s:t: x 2 D
x2D
and the K-T method, the model (2.6) can be transform-ed into the model (2.7) (Li
and Du 2011):
min Fðx; yÞ ¼ t
x;t
s:t: cT1 x þ d1T y t
where y solves
min f ðx; yÞ ¼ cT2 x þ d2T y
y ð2:7Þ
X
m n
X
s:t: aki ðuA Þki xi þ bkj ðuB Þkj yj hk þ ðuh Þk
i¼1 j¼1
k ¼ f1; . . .; r g
x; y 0
(3) Similar to the transformation (2.5),
m
x;y 0 P n
P
cT1 x þ d1T y t; ðc1 ; d1 Þ 2 l , c1i þ lc1 i xi þ
d1j þ ld1 j yj t
i¼1 j¼1
So the model (2.7) can be transformed to the model (2.8) as follows:
2 A Class of Robust Solution for Linear Bilevel Programming 17
X
m n
X
min Fðx; yÞ ¼ c1i þ lc1 i xi þ
d1j þ ld1 j yj
x;t
i¼1 j¼1
s:t: where y solves
min f ðx; yÞ ¼ cT2 x þ d2T y
y
ð2:8Þ
X
m n
X
s:t: aki ðuA Þki xi þ bkj ðuB Þkj yj hk þ ðuh Þk ;
i¼1 j¼1
k 2 f1; . . .; rg;
x; y 0
(4) Next, because the optimal solution of BLP (1) is not influenced by the value of
c2 , we only consider how to choose the value of d2 . Based on the original idea of
robust optimization, the model (2.8) is transformed into the model (2.4) above.
2.3 Solving Process of the Model
The deterministic triple level programming (2.4) can be written as the following
programming (2.9) by the K-T method.
X
m n
X
min max Fðx; yÞ ¼ c1i þ lc1 i xi þ
d1j þ ld1 j yj
x d2 ;y;u;v
i¼1 j¼1
X
r
s:t: d2j ¼ uk bkj ðuB Þkj þ vj
k¼1
d2j ðud2 Þj d2j d2j
þ ðud2 Þj
" #
m
X Xn
uk akj ðlA Þ xi þ bkj ðlB Þkj yj ðhk þ ðlh Þk Þ ¼ 0 ð2:9Þ
i¼j j¼1
vj yj ¼ 0
X m n
X
aki ðlA Þki xi þ bkj ðlB Þkj yj hk þ ðlh Þk ;
i¼1 j¼1
j ¼ f1; . . .; ng; k ¼ f1; . . .; r g;
x; y; u; v 0:
According to the literature (Wang 2010), the model (2.9) can be transformed
into the model (2.10) as follows
18 B. Liu et al.
min t
x;d2 ;y;u;v
X m n
X
s:t: c1i þ lc1 i xi þ
d1j þ ld1 j yj t
i¼1 j¼1
X
r
d2j ¼ uk bkj ðuB Þkj þ vj ;
k¼1
d2j ðud2 Þj d2j d2j
þ ðud2 Þj ;
" #
X
m Xn ð2:10Þ
uk aki ðlA Þki xi þ bkj ðlB Þkj yj hk þ ðlh Þk ¼ 0;
i¼1 j¼1
vj yj ¼ 0;
X m n
X
aki ðlA Þki xi þ bkj ðlB Þkj yj hk þ ðlh Þk ;
i¼1 j¼1
j ¼ f1; . . .; ng; k ¼ f1; . . .; r g;
x; y; u; v 0:
By introducing a large constant M, the model (2.10) above can be transformed
into a mixed integer programming as follows (Fortuny-Amat and McCarl 1981):
min t
x;d2 ;y;u;v;t;w
X m n
X
s:t: c1i þ lc1 i xi þ
d1j þ ld1 j yj t
i¼1 j¼1
X
r
d2j ¼ uk bkj ðuB Þkj þ vj ;
k¼1
d2j ðud2 Þj d2j d2j
þ ðud2 Þj ;
yj Mtj ;
vj M 1 tj ;
uk Mwk ;
Xm n
X
aki ðlA Þki xi þ bkj ðlB Þkj yj hk þ ðlh Þk M ð1 wk Þ
i¼1 j¼1
X
m n h
X i
aki ðlA Þki xi þ bkj ðlB Þkj yj hk þ ðlh Þk ;
i¼1 j¼1
j ¼ f1; . . .; ng; k ¼ f1; . . .; r g; tj 2 f0; 1g; wk 2 f0; 1g;
x; y; u; v 0:
ð2:11Þ
The model (2.11) can be solved by the software Lingo 9.0
2 A Class of Robust Solution for Linear Bilevel Programming 19
2.4 A Numerical Example
We give a numerical example to demonstrate the proposed approach as follows:
min F ¼ c11 x þ d11 y1 þ d12 y2
x
s:t: 2:5 x 8
where y1 ; y2 solve
min f ¼ d21 y1 þ d22 y2
y1 ;y2
s:t: a11 x1 þ b11 y1 þ b12 y2 h1
a21 x1 þ b21 y1 þ b22 y2 h2
a31 x1 þ b31 y1 þ b32 y2 h3
a41 x1 þ b41 y1 þ b42 y2 h4
y1 0;
y2 0
where a11 ¼ 0; b21 ¼ 1; a31 ¼ 0; a41 ¼ 0; b41 ¼ 0; b42 ¼ 1:
And the others are the uncertain data, the given variables and disturbances are
c11 ¼ 1:5; d11
¼ 1:5; d12 ¼ 2;
d21 ¼ 1:5; d22 ¼ 3:5;
b11 ¼ 1:75; b12 ¼ 1:15; h1 ¼ 2:5;
a21 ¼ 3:5; b22 ¼ 1:5; h4 ¼ 5:75:
h2 ¼ 11; b31 ¼ 3:5; b32 ¼ 1:25; h3 ¼ 23;
ðuc1 Þ1 ¼ 0:5; ðud1 Þ1 ¼ 0:5; ðud1 Þ2 ¼ 3; ðud2 Þ1 ¼ 0:5; ðud2 Þ2 ¼ 0:5;
ðub Þ11 ¼ 0:25; ðub Þ12 ¼ 0:15; ðuh Þ1 ¼ 0:5;
ðua Þ21 ¼ 0:5; ðub Þ22 ¼ 0:5; ðuh Þ2 ¼ 1;
ðub Þ31 ¼ 0:5; ðub Þ32 ¼ 0:25; ðuh Þ3 ¼ 1; ðuh Þ4 ¼ 0:25:
According to the theorem and these data above, robust model transformed is
demonstrated as
20 B. Liu et al.
min t
x;y;u;v;g;z
s:t: 2x y1 þ y2 t
2:5 x 8;
y2 5:5;
1:5y1 1:3y2 3;
4x þ y1 2y2 10;
4y1 1:5y2 22;
1 1:5u1 þ u2 4u3 þ v1 2;
4 1:3u1 2u2 1:5u3 u4 þ v2 3;
y1 Mg1 ; y2 Mg2 ;
v1 M ð1 g1 Þ; v2 M ð1 g2 Þ;
uk Mzk ; k ¼ 1; 2; 3; 4:
1:5y1 1:3y2 3 ð1 z1 Þ;
4x þ y1 2y2 þ 10 ð1 z2 Þ;
4y1 1:5y2 þ 22 ð1 z3 Þ;
y2 þ 5:5 ð1 z4 Þ;
x; y; u; v; g; z 0;
g1 ; g2 2 f0:1g; zk 2 f0:1g; k ¼ 1; 2; 3; 4:
By the software Lingo 9.0, the robust solution is obtained as follows:
ðx; y1 ; y2 Þ ¼ ð2:5201; 4:6443; 2:2819Þ;
The robust optimal value is Fmin ¼ 2:6779.
2.5 Conclusion and Future Work
Under the way of the centralized decision-making, a class of robust solution for
uncertain linear BLP is defined, which expands further the application of BLP in
different circumstances. And based on the original idea of robust optimization, the
uncertain BLP was converted to the deterministic triple level programming. The
solving process is proposed to obtain the robust solution of uncertain linear BLP.
Finally, a numerical example is shown to demonstrate the effectiveness and fea-
sibility of the algorithm.
2 A Class of Robust Solution for Linear Bilevel Programming 21
References
Bialas WF, Karwan MH (1982) On two-level optimization. IEEE Trans Autom control AC-
27(1):211–214
Dempe S (2002) Foundations of bilevel programming. Kluwer Academic Publisher, Boston
Fortuny-Amat J, McCarl B (1981). A representation and economic interpretation of two-level
programming problem. J Oper Res Soc 32:783–792
Fortuny-Amat J, McCarl BA (1981) Representation and economic interpretation of two-level
programming problem. J Oper Res Soc 32(7): 83–792
Lai YJ (1996) Hierarchical optimization: a satisfactory solution. Fuzzy Sets and Syst 77:321–335
Li Y, Du G (2011) Robust linear bilevel programming under ellipsoidal uncertainty. Syst Eng
11:96–100
Lobo MS, Vandenberghe L, Boyd S, Lebret H (1998) Application of second-order cone
programming. Linear Algebra Appl 284:193–228
Mathieu R, Pittard L, Anandalingam G (1994) Genetic algorithm based approach to bi-level
linear programming. Oper Res 28:1–21
Soyster AL (1973) Convex programming with set-inclusive constraints and applications to
inexact linear programming. Oper Res 21:1154–1157
Wang J (2010) Research on the methods of interval linear bi-level programming, pp 54–55.
Tianjin University, Tianjin
Chapter 3
A Comparison of the Modified
Likelihood-Ratio-Test-Based Shewhart
and EWMA Control Charts
for Monitoring Binary Profiles
Chao Yin, Yihai He, Zhen Shen and Chun-hui Wu
Abstract Profile monitoring is used to check and evaluate the stability of the
functional relationship between a response variable and one or more explanatory
variables known as profile over time. Many studies assume that the response
variable follows a continuous and normal distribution, while in fact it could be
discrete, for example binary profiles. However, at present, there are few researches
in this field. Based on an in-control binary dataset, this paper uses the logistic
regression model to estimate the parameters in Phase I. And in Phase II, we apply
bi-sectional search method to modifying the UCL’s calculation of the likelihood-
ratio-test-based Shewhart and EWMA control charts. Moreover, according to the
estimated parameters, ARL’s performances of the two modified control charts
under different parameters’ deviation are compared.
Keywords Binary profile Logistic regression model Bi-sectional search ARL
3.1 Introduction
In many applications, the quality process is characterized and summarized by a
functional relationship between a response variable and one or more explanatory
variables, which is known as profile. Many studies have been done here, for
example, according to the different types of profiles, Refs. Kim et al. (2003),
Mahmoud and Woodall (2004), Zou et al. (2007) and Refs. Kazemzadeh et al.
(2009), Amiri et al. (2009) respectively discussed the characteristics of linear
profiles and polynomial profiles, nonlinear profiles’ issues were talking about in
Refs. Eric et al. (2009), Yu et al. (2006), Williams et al. (2007). However, in the
C. Yin (&) Y. He Z. Shen C. Wu
School of Reliability and Systems Engineering,
Beijing University of Aeronautics and Astronautics, BUAA, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 23
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_3,
Springer-Verlag Berlin Heidelberg 2013
24 C. Yin et al.
aforementioned researches, they often assume the response variable follows a
continuous distribution but not discrete, such as binary profiles. Meanwhile, there
are some papers discussing the binary process in profile monitoring, Ref. (Yeh
et al. 2009) proposed 5 T2 based methods to monitor binary profiles in Phase I,
Refs. Koosha and Amiri (2011a, b), Koosha and Amiri (2012) conducted resear-
ches on other Phase I issues among binary profiles, such as autocorrelation
problems between profiles, effects of various types of profiles’ link function and so
on. However, they put more emphasis on Phase I analysis while few considerations
were made for Phase II process. This paper compares the performances of two
Phase II monitoring schemes based on the estimated parameters in Phase I. The
rest of this paper is structured as follows: Sect. 3.2 introduces a method called
logistic regression model, which is used to estimate parameters in Phase I. The
parameters estimating process is achieved in Sect. 3.3 based on an in-control
binary dataset. Sections 3.4 and 3.5 propose two Phase II monitoring modified
schemes. Section 3.6 compares the performances of these two control charts
according to estimated parameters in Sect. 3.3. Conclusion is explained in the last
section.
3.2 Logistic Regression Model
Suppose there are n independent profiles, in each profile, mi experiments will be
conducted and the explanatory variable with p dimensions could be denoted by
T
Xi ¼ xi1 ; xi2 ; . . .; xip , where i ¼ 1; 2; . . .; n and j ¼ 1; 2; . . .; p, meanwhile, Yi is
the response variable following a Bernoulli distribution with a probability of
success pi . To describe the link function in binary profiles, the logistic regression
model is applied:
pi
log ¼ b1 xi1 þ b2 xi2 þ . . . þ bp xip ð3:1Þ
1 pi
where b ¼ ðb1 ; b2 ; . . .; bp ÞT is the model’s parameter that needs to be estimated,
normally, xi1 will be assumed to be 1 and b1 serves as the intercept of the model.
Yeh et al. (2009) applied a Maximum Likelihood Estimation (M.L.E.) method to
estimate the parameters and got an updated estimate of b:
^ ^ ^
b ¼ ðX T W XÞ1 X T W q ð3:2Þ
where
^ ^ ^ ^ ^ ^ ^
W ¼ diagfm1 p1 ð1 p1 Þ; m2 p2 ð1 p2 Þ; . . .; mn pn ð1 pn Þg,
^ ^ ^ ^ ^ ^
q ¼ gi þðW Þ1 ðy lÞ ¼ X b þðW Þ1 ðy lÞ, and
^ ^ ^ ^
l ¼ ðm1 p; m2 p; . . .; mn pÞT :
1 2 n
3 A Comparison of the Modified Likelihood-Ratio-Test-Based Shewhart and EWMA 25
Still, an iterative algorithm was used here to obtain a more accurate value of b
according to (Yeh et al. 2009), steps were as listed below.
^ ^
(1) Initialize the estimate of b ¼ b0 , and b0 can be acquired by the ordinary least
squares (O.L.S.) estimation. Set i = 0;
^ ^ ^ ^ ^
(2) According to bi , calculate gi , pi , li , W i ;
^ ^ ^ ^
(3) Calculate qi ¼ gi þðW i Þ1 ðy li Þ;
^ ^ ^ ^
(4) Update the estimate of b by calculating bðiþ1Þ ¼ ðX T W i XÞ1 X T W i qi , and set
i = i ? 1;
^ ^ ^
l l1 l1
(5) Repeat steps (3.2) through (3.4) for l times, until b b =b a,
where is the Euclidean norm and a is a sufficiently small constant (a is equal to
^ ^
105 here). Then b ¼ bl is the desired estimator of b.
3.3 Parameters Estimation
We use an in-control dataset to complete the parameters estimation process. In this
case, 10 independent binary profiles are generated and the number of observations
in each level of the explanatory variable is set as 120, the values of x ranges from
2500, 2700, 2900,…, 4300. To describe the functional relationship between the
explanatory and response variable, the first-order model is set as:
logðp=½1 pÞ ¼ b0 þ b1 logðxÞ ð3:3Þ
where p is the probability of success in the binary process, and b ¼ ðb0 ; b1 ÞT is the
parameters that needs to be estimated. According to the methods displayed pre-
viously, we use Matlab 7.1.0.246 to finish the calculations and the result of the
O.L.S. initialization of b is valued as:
^
b0 ¼ ð42:0575; 5:1709ÞT ð3:4Þ
and Fig. 3.1 as listed below is showing that the effect of the estimation is fitting
well.
After the calculation of the iterative algorithm running in Matlab, the final
result could be obtained:
^ ^
b ¼ bl ¼ ð42:0537; 5:1704ÞT ð3:5Þ
26 C. Yin et al.
Fig. 3.1 Performance of the
ordinary least squares
estimation
and a ¼ 8:0889 109 \\105 .
In the end, we need to judge whether the binary dataset satisfies the second-
order model or not, thus logðp=½1 pÞ ¼ b0 þ b1 logðxÞ þ b2 ðlogðxÞÞ2 . The
deviance statistic turns out to be :
2:7954 109 \\v21;0:05 ¼ 3:841
which shows the second-order model is not fit to apply for the binary dataset and
the first-order model we set here is suitable.
3.4 Phase II Likelihood Ratio Test for Binary Profiles
The binary profiles’
P functional relationship between x and Y can be concluded as
p
log 1p ¼ b1 þ pj¼2 bj xj ; j ¼ 1; 2; ; p. Zhou (2008) proposed a likelihood ratio
test method to evaluate the model parameters b in Phase II, and the likelihood ratio
test statistic k will be applied in the control charts process we discussed.
H0 : Y Bernoulli(pðx)) H1 : Y Bernoulli(p ðx)) ð3:6Þ
P
p
exp b1 þ j¼2
bj x j
Where pðx) ¼ Pp
1þexp b1 þ bx
j¼2 j j
Pp
exp b1 þ b x
j¼2 2 2
and p ðx) ¼ Pp .
1þexp b1 þ b x
j¼2 2 2
3 A Comparison of the Modified Likelihood-Ratio-Test-Based Shewhart and EWMA 27
After the calculation process explained in Zhu (2008), the likelihood ratio test
statistic can be defined as:
k ¼ ‘H1 ‘H0 ð3:7Þ
where ‘H1 and ‘H0 are the log-likelihoods, and k will be compared with a critical
value that is determined by a pre-defined type-I error rate a, if k is greater than the
critical value, then it means at least one parameter has a shift.
3.5 The Modified Control Charts
Zhou and Lin (2010) proposed a Shewhart-type LRT-based control chart to
monitor the generalized linear profiles; we can call it Shewhart.LRT control chart.
During its process, 2k will have a distribution of v2 with k degrees of freedom
under H0 , k is the total number of parameters in the profile function model. To
monitor binary dataset, the LCL will be set as 0, and because of the binary
distribution increasing the complexity of calculating the UCL of Shewhart.LRT
control chart, a modification applying the bi-sectional search method can be made
to calculate it.
(1) Based on the pre-defined type-I error rate a, set the needed in-control average
run length (ARL). Then select a temporary value for the lower and upper limit
of UCLl and UCLu, calculate the corresponding ARLl and ARLu respectively,
and it needs to satisfy the condition that ARLl \ ARLIC \ ARLu.
(2) Let UCLtemp ¼ UCLl þUCL
2
u
, calculate ARLtemp. And if ARLtemp [ ARLIC, set
UCLu = UCLtemp and ARLu = ARLtemp. Otherwise set UCLl = UCLtemp.
(3) If the absolute value of ‘‘ARLu-ARLl’’ is less than a pre-defined threshold,
here the value is settled as 1, then the desired UCL ¼ UCLl þUCL
2
u
, otherwise go
back to step (2).
Yeh et al. (2004) proposed a likelihood-ratio-based EWMA control chart to
monitor multivariate normal processes, also, we can call it EWMA.LRT control
chart. To monitor binary dataset, the EWMA statistic can be defined as:
EWMAi ¼ h 2ki þ ð1 hÞ EWMAi1 ð3:8Þ
where h is the smoothing constant, and EWMA0 is set as k. Similarly, LCL is
assumed as 0, for the UCL of EWMA.LRT modification control chart, also apply
the bi-sectional search method to calculate it.
28 C. Yin et al.
3.6 Performances of the Modified Control Charts
in Phase II
In Phase I, the fitted logistic regression model has been set: logðp=½1 pÞ ¼
^
b0 þ b1 logðxÞ, and the obtained parameters is valued as: b ¼ ðb0 ; b1 ÞT ¼
ð42:0537; 5:1704ÞT .
In Phase II, we generate deviations on parameters b0 and b1 to evaluate the
ARL’s performances of the modified Shewhart.LRT and EWMA.LRT control
charts and to investigate which parameter is more sensitive under given deviations.
The pre-defined Type I error rate (a) is set as 0.005 which will yield an in control
ARL as 200, and for EWMA.LRT control chart, the smoothing constant h is
assumed as 0.2. Results of ARL’s values of two modification control charts are
both based on 10,000 Monte-Carlo simulations, results are as shown below in
Table 3.1.
In Table 3.1, we define that ARL1 and ARL2 separately represent the ARL’s
values of Shewhart.LRT and EWMA.LRT control charts, and we can see that with
the parameters’ deviation becoming larger, the ARL’s result of the modified
control charts are getting more and more smaller, and the ARL’s results of
EWMA.LRT control chart are smaller than Shewhart.LRT control chart under the
same level of parameter’s deviation, which shows that it has a better performance
than Shewhart.LRT control chart to monitor the binary process here.
Meanwhile, between the parameters b0 andb1 , b1 is more sensitive than b0
when a deviation happens, which can be seen obviously from the ARL’s result.
Table 3.1 only shows the performance of the increasing shift of deviation
parameters; moreover, similar results happen in the decreasing shift of deviation
parameters which is not shown here.
Table 3.1 Arl’s performance of the modified control charts under parameters’ deviation
Db0 ARL1 ARL2 Db1 ARL1 ARL2
0 201.443 200.632 0 202.206 20.468
0.002 173.541 152.457 0.002 93.632 80.653
0.006 149.675 107.453 0.006 27.184 24.339
0.01 115.058 84.445 0.01 9.547 6.542
0.014 98.246 65.994 0.014 4.429 3.875
0.018 82.882 41.123 0.018 2.460 2.087
0.022 70.253 33.682 0.022 1.650 1.456
0.026 60.620 22.546 0.026 1.278 0.897
0.03 52.553 18.578 0.03 1.095 0.786
0.034 45.207 12.125 0.034 1.043 0.743
0.038 38.162 11.557 0.038 1.001 0.725
0.042 35.511 8.546 0.042 1.002 0.708
0.046 29.484 6.078 0.046 1.000 0.704
0.05 25.953 5.271 0.05 0.998 0.698
3 A Comparison of the Modified Likelihood-Ratio-Test-Based Shewhart and EWMA 29
3.7 Conclusion
In this paper, the ARL’s performances of the modified Shewhart.LRT and
EWMA.LRT control charts are compared, and the result shows that the latter
scheme has a better effect on monitoring the binary process and parameter b1 is
more sensitive than b0 when a deviation happens according to the estimated
parameters. The further researches could be focused on proposing new methods to
estimate the parameters in Phase I and apply other types of control charts to
monitor Phase II process in binary profiles.
References
Amiri A, Jensen W, Kazemzadeh RB (2009) A case study on monitoring polynomial profiles in
the automotive industry. Qual Reliab Eng Int 26:509–520
Eric C, Joseph J, Simpson J (2009) Statistical process monitoring of nonlinear profiles using
wavelets. J Qual Technol 41:198–212
Kazemzadeh RB, Noorossana R, Amiri A (2009) Monitoring polynomial profiles in quality
control applications. Int J Adv Manuf Technol 42:703–712
Kim K, Mahmoud MA, Woodall WH (2003) On the monitoring of linear profiles. J Qual Technol
35:317–328
Koosha M, Amiri A (2011a) The effect of neglecting autocorrelation on the performance of T2
control charts in monitoring of logistic profiles. In: ICQR 2011, pp 264–267
Koosha M, Amiri A (2011b) The effect of link function on the monitoring of logistic regression
profiles. In: WCE 2011, pp 326–328
Koosha M, Amiri A (2012) Generalized linear mixed model for monitoring autocorrelated
logistic regression profiles. Int J Adv Manuf Technol 64:487–495
Mahmoud MA, Woodall WH (2004) Phase I analysis of linear profiles with calibration
applications. Technometrics 46:380–391
Williams JD, Woodall WH, Birch JB (2007) Statistical monitoring of nonlinear product and
process quality profiles. Qual Reliab Eng Int 23:925–941
Yeh AB, Huwang LC, Wu YF (2004) A likelihood ratio based EWMA control chart for
monitoring variability of multivariate normal processes. IIE Trans 36:865–879
Yeh AB, Huwang LC, Li YM (2009) Profile monitoring for a binary response. IIE Trans
41:931–941
Yu D, Li Z, Zhou SY (2006) Phase I analysis for monitoring nonlinear profiles in manufacturing
processes. J Qual Technol 38:199–216
Zhou JJ, Lin D (2010) Monitoring the slopes of linear profiles. Qual Eng 22:1–12
Zhu JJ (2008) Essays on monitoring profile data. The Pennsylvania State University
Zou CL, Fugee T, Wang ZJ (2007) Monitoring general linear profiles using multivariate
exponentially weighted moving average schemes. Technometrics 49:395–408
Chapter 4
A Decision-Making Model of Price
Control for Administering Authority
Chong-yi Jing
Abstract There are few quantitative studies on decision making of air ticket price
control problem. In this paper we establish a game decision making model of price
control for government by introducing two new factors: consumer’s surplus (the
public welfares) and the passenger load rate (LR). We get some interesting con-
clusions from modeling and discussion. The administering authority, CAAC (Civil
Aviation Administration of China) is inclined to ignore the public welfares when
setting a higher control price and the airlines are always inclined to disobey the
control price of CAAC for achieving a higher passenger load rate and strength-
ening the competition edge. As a whole, the optimal strategy of CAAC is to set an
inter-zone control price and the optimal strategy of airlines is to self determinate a
price between the inter-zone prices. The reason of decision dissonance is that the
cost evaluation of ticket pricing for the two players has tremendous difference.
Keywords Air ticket Decision making Game theory Price control
4.1 Introduction
There are many studies on air ticket control, which are almost limited to quali-
tative analysis. Some use natural monopoly theory (Zhang 2005; Liu 2006) and
some use welfare economics Liu (2002) to analyze the problem, and the result is
almost similar that the government should release control to air ticket price and be
unnecessary to intervene in operation or management of airlines (Liu 2002). Li and
Deng (2003), Min and Yang (2003) conclude that the civil aviation of China has
many pertinacious problems such as ticket price simplification, cut-throat
C. Jing (&)
Aviation Transportation Management School, Civil Aviation Flight
University of China, Guanghan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 31
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_4,
Ó Springer-Verlag Berlin Heidelberg 2013
32 C. Jing
competition and so on basing on an economics point of view, and analyze the
reasons from all conceivable angles Li and Deng (2003), Min and Yang (2003).
And there are also some studies using game theory to analyze the price control of
CAAC to airlines (Mei et al. 2006; Yang and Zhang 2002), and the conclusions are
that the control could cause lose of society welfare and ticket pricing should be
marketization (Mei et al. 2006; Yang and Zhang 2002). Another train of thought is
to advocate control, Zipping (Kang and Du 2006) approve of moderate control
(Kang and Du 2006), Qiu (2001) points out that releasing control should be
implemented gradually through comparing civil aviation development course of
China with that of USA (Qiu 2001). Han (2000) indicates that the government
should take some long-term measures to manage the special industry (Han 2000).
In a word, there’re few quantitative studies on the problem and even in some
quantitative literatures there are also some advancements to deserve promoting, for
example, CAAC just emphasizes the economic revenue maximization (Mei et al.
2006; Yang and Zhang 2002). However, CAAC, as a government department, has
a very important society function, which is protecting social welfare from being
damaged. On the other hand, it is probably not appropriate to suppose that the
airlines in China get into price war just for achieving economic profit maximi-
zation. The airlines may attach much more importance to increasing passenger
load factor, market share, and strengthening their competitive edge. In reality, both
CAAC and airlines might fall into prisoner’s dilemma easily, and it’s very difficult
to get away the vicious circle (Wang 2004; Yang 2002). For these problems, we
introduce consumer’s surplus factor and passenger load rate factor into our game
model to establish a ticket price control model for government.
4.2 Modeling
4.2.1 Hypothesis for Modeling
Hypothesis1: the game process is repetitive and limited, and the information for
each other is imperfect. (For example, cost, intension and so on).
Hypothesis2: the CAAC and airlines in China have some common interests and
close relations although the former is supervisor of the industry (for example, at
aspects of finance and political achievements).
Hypothesis3: at present in China, the general ticket demand is price inelastic.
4.2.2 Construct Price Control Model
As one player of the game, CAAC has two pure strategies: one is ‘‘Consider the
public welfare’’ (hereinafter called ‘‘Consider’’), the lower price is p1 and the
4 A Decision-Making Model of Price Control 33
Fig. 4.1 Extensive game
i de
between CAAC and airlines Ab )
Airlines ,p 1
(ß
der
o nsi Disob
e y
P01=P02
C (1-ß,
p1 p01 )
a
CAAC
ey
1-a ob
Dis ß,p 02)
p2
Ignor
e (1-
(ß,p )
Airlines 2
Abid
e
probability is a; another strategy is ‘‘Ignore the public welfare’’ (hereinafter called
‘‘Ignore’’), the higher price is p2 ðp2 [ p1 Þ, and the probability is 1 a. Airlines as
another player of the game also have two pure strategies: one is ‘‘Abide the price
control’’ (hereinafter called ‘‘Abide’’), the price is p1 and the probability is b;
another strategy is ‘‘Disobey the price control’’ (hereinafter called ‘‘Disobey’’), the
price they set will be p0i ði ¼ 1; 2Þ that means the self-determinate price of airlines
on condition that the control price is pi ði ¼ 1; 2Þ; and the probability is 1 b.
Figure 4.1 shows the players’ game decision tree with imperfect information,
which means airlines could not distinguish whether the control price contains the
public welfare or not, and CAAC also could not identify airlines’ cost information
and competition intention easily (shown as the broken line in Fig. 4.1).
The following is other some denotations used herein below.
TS Social welfare equals to sum of CS and ES.TS = CS ? ES
CS Public welfare, also called customers’ surplus
ES Enterprise welfare, also called enterprise surplus
f(p) Inverse demand function when the price is P, meanwhile, the equilibrium
demand can be defined as d(p)k, here d denotes conversion operator, k
denotes price elasticity (absolute value)
LR(p) Load rate of airlines when the ticket price is p, it’s decreasing function of
price p. It is supposed to be reasonable that airlines increase the passenger
load rate and strengthen competitive edge by cutting down price
TCL Transportation Capacity Limitation of airlines in a given period, it keeps
stable during a determinate period of time and irrelevant with price
c1 The unit cost of airlines estimated by CAAC
c2 The unit cost of airlines estimated by airlines themselves
Figure 4.2 is payoff matrix of the two players. We define their payoff functions
as bellows:
pCAAC ðCAÞ: The payoff of CAAC when CAAC considers the public welfare and
airlines abide the price control. Essentially, the payoff equals to TS, which should
R dðp Þk
contain CS and ES. And CS ¼ 0 1 f ðpÞdp, here d denotes derivative symbol,
ES ¼ ðp1 c1 Þ dðp1 Þk , so the payoff of CAAC is
34 C. Jing
Abide airlines Disobey
Consider
π CAAC (CA) πailines (CA) π CAAC (CD ) π ailines (CD )
CAAC
π CAAC ( IA) π ailines ( IA) π CAAC ( ID ) π ailines ( ID )
Ignore
Fig. 4.2 Payoff matrix of game players
Z dðp1 Þk
pCAAC ðCAÞ ¼ f ðpÞdp þ ðp1 c1 Þ dðp1 Þk
0
pairlines ðCAÞ: The payoff of airlines when airlines abide the price control and
CAAC considers the public welfare. Under the circumstances, the payoff of air-
lines is.
pairlines ðCAÞ ¼ ðp1 c2 Þ dðp1 Þk
pCAAC ðCDÞ: The payoff of CAAC when CAAC considers the public welfare and
airlines disobey the price control. p01 is self-determinate price of airlines on
condition that the control price of CAAC is p1 . Like pCAAC ðCAÞ, the payoff equals
to TS, which should contain CS and ES. So
Z dðp01 Þk
pCAAC ðCDÞ ¼ f ðpÞdp þ ðp01 c1 Þ dðp01 Þk
0
pairlines ðCDÞ: The payoff of airlines when airlines disobey the price control and
CAAC considers the public welfare. The payoff of airlines is
pairlines ðCDÞ ¼ ðp01 c2 Þ LRðp01 Þ TCL
pCAAC ðIAÞ: The payoff of CAAC when CAAC ignores the public welfare and
airlines abide the price control. Then the payoff just contains ES, so
pCAAC ðIAÞ ¼ ðp2 c1 Þ dðp2 Þk
pairlines ðIAÞ: The payoff of airlines when airlines abide the price control and CAAC
ignores the public welfare. The payoff of airlines is
pairlines ðIAÞ ¼ ðp2 c2 Þ dðp2 Þk
pCAAC ðIDÞ: The payoff of CAAC when CAAC ignores the public welfare and
airlines disobey the price control. p02 is self-determinate price of airlines on
condition that the control price of CAAC is p2 . The payoff of CAAC is.
pCAAC ðIDÞ ¼ ðp02 c1 Þ dðp02 Þk
4 A Decision-Making Model of Price Control 35
pairlines ðIDÞ: The payoff of airlines when airlines disobey the price control and
CAAC ignores the public welfare. The payoff of airlines is.
pairlines ðIDÞ ¼ ðp02 c2 Þ LRðp02 Þ TCL
4.2.3 Analyzing and Calculating for Model
According to the model above, the mixed strategy of CAAC is h1 ¼ ða; 1 aÞ,
and the mixed strategy of airlines is h2 ¼ ðb; 1 bÞ. The payoff function of
CAAC can be expressed as.
2 Z k
! 3
dðp1 Þ
6b f ðpÞdp þ ðp1 c1 Þ dðp1 Þk þ 7
6 0 7
6 7
l1 ðh1 ; h2 Þ ¼ a6 ! 7
6 Z dðp01 Þk 7
4 k 5
ð1 bÞ f ðpÞdp þ ðp01 c1 Þ dðp01 Þ ð4:1Þ
0
2 3
b ðp2 c1 Þðp2 Þk þ
6 7
þ ð 1 aÞ 4 5
k
ð1 bÞ ðp02 c1 Þ dðp02 Þ
The first order condition of (4.1) is
Z !
dðp1 Þk
ol1 k
¼b f ðpÞdp þ ðp1 c1 Þ dðp1 Þ
oa 0
Z !
dðp01 Þk
k
þ ð1 bÞ f ðpÞdp þ ðp01 c1 Þ dðp01 Þ
0
b ðp2 c1 Þ dðp2 Þk ð1 bÞ ðp02 c1 Þ dðp02 Þk
¼0
ol1
So we can get b from oa ¼0
R dðp Þk
ðp02 c1 Þ dðp02 Þk 0 01 f ðpÞdp ðp01 c1 Þ dðp01 Þk
b¼ Z dðp1 Þk Z dðp01 Þk ð4:2Þ
k
f ðpÞdp þ ðp1 c1 Þ dðp1 Þ f ðpÞdp
0 0
ðp01 c1 Þ dðp01 Þk ðp2 c1 Þ dðp2 Þk þ ðp02 c1 Þ dðp02 Þk
The payoff function of airlines can be expressed as:
36 C. Jing
2 3
a ðp1 c2 Þðp1 Þk þ
l2 ðh1 ; h2 Þ ¼ b4 5
k
ð1 aÞðp2 c2 Þ dðp2 Þ ð4:3Þ
" #
aðp01 c2 Þ LRðp01 Þþ
þ ð1 bÞ
ð1 aÞðp02 c2 Þðp02 Þ TCL
The first order condition of (4.3) is
ol2
¼ aðp1 c2 Þ dðp1 Þk þ ð1 aÞðp2 c2 Þ dðp2 Þk
ob
aðp01 c2 Þ LRðp01 Þ TCL ð1 aÞðp02 c2 Þ LRðp02 Þ TCL ¼ 0
ol2
So we can get a from ob ¼0
ðp02 c2 Þ LRðp02 Þ TCL ðp2 c2 Þ dðp2 Þk
a¼
ðp1 c2 Þ dðp1 Þ ðp2 c2 Þ dðp2 Þk ðp01 c2 Þ LRðp01 Þ TCL þ ðp02 c2 Þ LRðp02 Þ TCL
k
ð4:4Þ
According to Hypothesis1, the game is limited and imperfect information,
airlines cannot distinguish whether the control price is set basing on public welfare
or not, and CAAC also cannot pry about cost information and competition
intention of airlines easily. In a limited period, CAAC sets a control price, if
airlines intend to disobey, they are inclined to set self-determination price as p0
uniformly, so p01 ¼ p02 ¼ p0 , which can be substituted into (4.2) and (4.4), then
we can get new a and b.
R dðp0 Þk
0 f ðpÞdp
b¼R R dðp0 Þk
dðp1 Þk
0 f ðpÞdp þ ðp1 c1 Þ dðp1 Þk 0 f ðpÞdp ðp2 c1 Þ dðp2 Þk
ð4:5Þ
ðp0 c2 Þ LRðp0 Þ TCL ðp2 c2 Þ dðp2 Þk
a¼ ð4:6Þ
ðp1 c2 Þ dðp1 Þk ðp2 c2 Þ dðp2 Þk
4.3 Discussion
4.3.1 The Discussion of Optimal Decision of CAAC
Make some conversion to (4.5), b can be rewritten as:
4 A Decision-Making Model of Price Control 37
,2 k k R dðp1 Þk 3
ðp 2 c 1 Þ dðp 2 Þ ðp 1 c 1 Þ dðp 1 Þ f ðpÞdp
b ¼ 1 41 þ R dðp0 Þk
0 5
0 f ðpÞdp
, because 0 b 1, so
R dðp1 Þk
ðp2 c1 Þ dðp2 Þk ðp1 c1 Þ dðp1 Þk 0 f ðpÞdp
R dðp0 Þk 0;
0 f ðpÞdp
R dðp1 Þk
and therefore ðp2 c1 Þ dðp2 Þk ðp1 c1 Þ dðp1 Þk 0 f ðpÞdp;
and here.
Z dðp1 Þk
k k
ðp2 c1 Þ dðp2 Þ ðp1 c1 Þ dðp1 Þ þ f ðpÞdp ð4:7Þ
0
The left part of (4.7) is pCAAC ðIAÞ, while the right part of (4.7) is pCAAC ðCAÞ, so
(4.7) is pCAAC ðIAÞ pCAAC ðCAÞ, CAAC has no motivation to improve and con-
sider the public welfare, that is to say, CAAC is apt to maintain a higher industry
price unilaterally rather than decreasing the control price for the public. Actually,
CAAC has transferred the public welfare to enterprises and civil aviation industry
by pricing at a higher level (Shaffer 2001). According to Hypothesis2, CAAC and
airlines in China have many common interests and close relations, not only the
finance aspects, but also the politics achievements of CAAC have to rely on
development and stability of the civil aviation industry. At this point of view, the
result is corresponding to Hypothesis2, and also corresponds with the reality.
4.3.2 The Discussion of Optimal Decision of Airlines
Because the ticket demand is price inelastic (Hypothesis3), and p1 \p2 , for
parameter a, it is supposed to follow the restrictions as bellow:
8
>
> ðp0 c2 Þ LRðp0 Þ TCL ðp2 c2 Þ dðp2 Þk
>
> a ¼
< ðp1 c2 Þ dðp1 Þk ðp2 c2 Þ dðp2 Þk
ð4:8Þ
> ðp1 c2 Þ dðp1 Þk ðp2 c2 Þ dðp2 Þk \0
>
>
>
:
0a1
Referring to (4.8), we get the calculation result:
ðp1 c2 Þ dðp1 Þk ðp0 c2 Þ LRðp0 Þ TCL ðp2 c2 Þ dðp2 Þk ð4:9Þ
The middle part of (4.9) is pairlines ðCDÞ, and also pairlines ðIDÞ, the left part of
(4.9) is pairlines ðCAÞ and the right part of (4.9) is pairlines ðIAÞ. So (4.9) can also be
rewritten as:
38 C. Jing
pairlines ðCAÞ pairlines ðCDÞ ¼ pairlines ðIDÞ pairlines ðIAÞ
Airlines cannot observe whether CAAC has considered the public welfare or
not when setting control price, and obviously the self-determination price p0 could
produce more profits than the lower control price p1 could, and theoretically
speaking, abiding a higher control price p2 could get more profits than self
determining price p0 could, but there may be two reasons at least to make them
depart from the ‘‘optimization path’’: one is that airlines cannot stand a very low
passenger load rate because of a higher price, which is a big waste of resources;
another is that when a competitor reduce the price, the others have to fallow the
decreasing price strategy, or they may suffer much more lost, which results from
product homogeneity of air transportation in China.
4.3.3 The Discussion of Decision Dissonance of Two Players
Basing on the discussion A, the optimal strategy payoff of CAAC is ðp2 c1 Þ dðp2 Þk ,
o½ðp2 c1 Þdðp2 Þk k
of which the first order condition is op2 ¼ 0, so p2 ¼ 1þk c1 ; Basing on the
discussion B, the optimal strategy payoff of airlines is ðp0 c2 Þ LRðp0 Þ TCL, of
which the first order condition is o½ðp0 c2 ÞLRðp
op0
0 ÞTCL
¼ 0, so p0 ¼ c2 LRLR 0
0 , LR is the
derivative of LRðpÞ at p ¼ p0 , LRðpÞ is decreasing function of price p, so LR0 \0,
LR
simultaneously 0\LRðpÞ 1, so p0 ¼ c2 LR 0 [ c2 . Then the restriction
condition is
8
> k
>
> p ¼ c
> 2 1þk 1
< k
LR ; so c1 [ c2 ; because
>
> p 0 ¼ c 2 0
[ c 2 1þk
>
> LR
:
p2 [ p0
k
0\ 1þk \1, the calculation result is c1 [ c2 . CAAC sets the control price basing
on the cost evaluation of c1 , which is usually the average cost of civil aviation
industry. While airlines self determinate competitive price basing on the cost
evaluation of c2 , which is probably the margin cost of the company. The tre-
mendous difference between c1 and c2 causes decision dissonance of the two
players. So as long as p0 [ c2 , airlines always have the motivation of disobeying
and cutting down control price.
4 A Decision-Making Model of Price Control 39
4.4 Decision Making
For CAAC, the optimal strategy is setting a higher control price without considering
the public welfares because it has many underlying common interests with airlines
and the civil aviation industry, for example, at aspects of finance and political
achievements. But airlines are always inclined to disobey the control price and get
into price war. The reason is that airlines in China have to face furious competitions
after the so called ‘‘deregulation’’, while the property rights of leading airlines still
belong to administrative departments. For this kind of natural monopoly industry of
civil aviation, fixed costs are very high while the margin costs are very low, so for
airlines there always has been pressure and space of cutting down price. And we
have proved that when the self-determination price p0 is lower than the control price
p2 , airlines could become more competitive such as getting a higher passenger load
rate, a bigger market share and so on. However, when p0 \c2 , (it is possible because
of the unusual structure of property rights) the air transport market will be in disorder
and the whole industry welfares will be seriously damaged.
And then CAAC has to set a lower boundary for the control price to prevent this
phenomenon form happening, so that the self-determination price of airlines
cannot be under the lower boundary of control price on any account.
p1 is a lower control price when CAAC considers the public welfares, when
p0 [ p1 , the airlines can get excess profit, or else they will go into red. So the
optimal strategy for CAAC is setting an inter-zone control price, for example
½p1 ; p2 , and the optimal strategy for airlines is self-pricing between p1 and p2 .
Basing on the decisions above, CAAC could get balance among the industry
welfares, healthy development of the market and the public welfares, airlines
could get balance between the profits and competitive edges. Lastly it is necessary
to note that the final decision makings of the two players are both based on the
present special structure of property rights of civil aviation in China, which could
not be reformed or changed in a short time period.
4.5 Conclusion
Unlike many qualitative analyses on air ticket price control problem, we build a
quantitative price control model based on game theory by introducing two new
decision factors: CS (customer’s surplus, also called the public welfares) and LR
(passenger load rate). Through modeling and discussing, we get some interesting
results and conclusions: CAAC is inclined to ignore the public welfares when
setting the control price of air ticket, which may be a higher price p2 , whereas,
airlines are always inclined to disobey the control price of CAAC for achieving a
higher passenger load rate and strengthening the competition edge, which may be a
self-determination price p0 , and p1 \p0 \p2 , where p1 is a lower control price
when CAAC considers the public welfares.
40 C. Jing
At present, main airlines of China have very special structure of property rights,
that is to say airlines face dual pressures of administrative regulations and market
competitions, which may cause airline’s dumping price for market shares and
competition edge, even may lead to p0 \c2 , so CAAC has to define a lower
boundary for the control price to prevent this kind of vicious competition behavior.
As a result, CAAC sets an inter-zone control price for achieving to keep enough
industry welfares, ensure healthy development of the civil aviation industry and
consider moderate public welfares. And when p0 2 ½p1 ; p2 , airlines can get bal-
ance between the revenues and competition edge, and CAAC can also get balance
between the industry welfares and the public welfares.
The reason of decision dissonance between the two players is tremendous
difference of cost evaluation for air ticket pricing (c1 [ c2 , shown as discussion C).
c1 is the cost evaluation of CAAC, usually an average cost of the whole industry,
and c2 is the cost evaluation of airlines, usually a margin cost of the product. In
terms of economics and competitions, as long as p0 [ c2 , airlines are always
inclined to disobey and decrease the control price for a higher passenger load rate,
i.e. for competitive edge.
Acknowledgments We thank all the sponsors for their financial support. And especially,
Chongyi Jing Author thanks his dear friend Dr Bin Zhou for his thoughtful comments, sugges-
tions and helpful data resources, which greatly enhanced the accomplishment of this paper.
References
Han B (2000) A review on the price control of China’ s civil airplane ticket. Sci Technol Rev
1:52–54 (in Chinese)
Kang Z, Du W (2006) Control and re-control of civil aviation transport fare in our country.
Northeast Financ Univ Trans 3(5):16–18 (in Chinese)
Li X, Deng J (2003) The reform of price management system of civil aviation based on price
theory. China Civ Aviat Sch Trans 21(2):26–31 (in Chinese)
Liu J (2002) Game analysis and reform suggestion for air ticket price. China Price 11:26–29 (in
Chinese)
Liu J (2006) The control and competition of natural monopoly industry. Search Truth 2:181–182
(in Chinese)
Mei H, Zhu J, Wang X (2006) Study of revenue management pricing based on game analysis.
Forecasting 25(6):45–49 (in Chinese)
Min Z, Yang X (2003) Industry organization analysis on price evolvement of civil aviation
enterprises. ShanXi Financ Univ Trans 3:57–61 (in Chinese)
Qiu Y (2001) A comparison analysis of deregulating price control between China and USA and
the control policy choice of China. Special-zone Econ 4:35–36 (in Chinese)
Shaffer S (2001) Deregulation: theory and practice. J Econ Bus 53(6):107–109
Wang R (2004) Game explanation on price competition of airlines. Technoecon Manag Res
3:81–82 (in Chinese)
Yang S, Zhang X (2002) A game theoretical analysis of marketable air fare. Commer Res
12:10–13 (in Chinese)
Zhang L (2005) Control or deregulation of air transportation price. China Sci Technol Inf
12(21):113–123 (in Chinese)
Chapter 5
A Method for Multiple Attribute Decision
Making without Weight Information
but with Preference Information
on Alternatives
Yun-fei Li
Abstract A kind of uncertain multiple attribute decision making problems with
preference information on alternatives is studied, in which the information about
the attribute weights is unknown and the attribute values are interval numbers.
Based on the deviation degree between the comprehensive attribute value of
alternatives and the preference of decision-maker for alternatives, an optimum
model is constructed and a simple formula for obtaining the attribute weights is
given, then a new method to get the priorities of alternatives is presented.
An example is given to show the application about the method.
Keywords Deviation degree
Interval number Multiple attribute decision
making Preference information
5.1 Introduction
Multiple attribute decision making (MADM) is an important area in modern
decision theory. The methods for the MADM with complete weight information
have been widely studied (Cheng 1987; Chen and Zhao 1990; Hwang and Yoon
1981). However, in fact, the decision maker may have uncertain knowledge about
the attribute weight information and preference information on alternatives. So, the
research on the MADM with preference information on alternatives has important
theoretical significance and practical value.
Park and Kim (1997), Kim and Choi (1999), Kim and Ahn (1999) have
researched the MADM problem with incomplete weight information. Based on
introducing the preference degree, Gao (2000) has studied the MADM problem
Y. Li (&)
School of Mathematics and Information, China West Normal University, Nanchong, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 41
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_5,
Ó Springer-Verlag Berlin Heidelberg 2013
42 Y. Li
with incomplete weight information expressed by interval number and preference
information on alternatives. Under the situations where the decision maker has
preference information on alternatives, which takes the form of reciprocal
judgement matrix and complementary judgement matrix, Xu (2004) has estab-
lished two objective programming models to research the MADM problem under
partial weight information expressed by interval number. Based on a projection
method, Xu (2004) has studied the MADM problem with incomplete weight
information and preference information expressed by real numbers on alternatives.
Jiang and Fan (2005) have constructed an objective programming model to
analyze the MADM problem with partial weight information and preference
information expressed by interval numbers on alternatives. Based on the grey
relational coefficients between objective preference and subjective preference, Wei
and Wei (2008) have proposed a method to discuss the MADM problem with
interval numbers and preference information on alternatives. Fan et al. (2008) have
researched the MADM problem without weight information and the decision
maker expresses his preference information on alternatives by a fuzzy relation.
Based on the research of Xu (2003) has constructed an optimal model by using a
linear translation function to study the MADM problem, in which the weight
information is completely unknown, the attribute values are real numbers and the
decision maker has preference on alternatives. However, from the above resear-
ches, we can find few researches on the MADM problem, in which weight
information is completely unknown, the attribute values are expressed by interval
numbers and the decision maker has preference information expressed by interval
numbers on alternatives. The aim of this paper is to establish an optimal model to
solve the above problem.
5.2 Preliminaries
In the following, we will introduce some important concepts and algorithms about
interval numbers (Xu and Da 2003).
Let ~a ¼ ½a ; aþ ¼ fxja x aþ ; a ; aþ 2 Rg, ~a is an interval number.
Specially, if a ¼ aþ , ~a is a real number.
The algorithms related to interval numbers are following: if ~a ¼ ½a ; aþ and
~
b ¼ ½b ; bþ , b 0, then
a¼~
(1) ~ b if and only if a ¼ b ; aþ ¼ bþ ,
aþ~
(2) ~ b ¼ ½a þ b ; aþ þ bþ ,
a ¼ ½ba ; baþ , specially, if b ¼ 0, then b~
(3) b~ a ¼ 0.
Let X ¼ fX1 ; X2 ; ; Xn g ðn 2Þ be the set of alternatives and U ¼
fu1 ; u2 ; ; um g ðm 2Þ be the set of attributes. Suppose the decision maker gives
5 A Method for Multiple Attribute Decision Making 43
his subjective preference information on alternative Xi 2 X by ~hi ¼ ½h þ
i ; hi
þ ~ þ
ð0 hi hi 1Þ. Let A ¼ ð~ aij Þnm be the decision matrix, ~aij ¼ ½aij ; aij is the
attribute value of alternative Xi with respect to the attribute uj , i ¼ 1; 2; ; n;
j ¼ 1; 2; ; m.
In order to avoid the influence of different dimensions on the decision making
results, the decision matrix A ~ ¼ ð~ aij Þnm should to be normalized into the
dimensionless decision matrix R ~ ¼ ð~rij Þnm ; in which ~rij ¼ ½rij ; rijþ ¼ ftj0 rij
t rijþ 1g ði ¼ 1; 2; ; n j ¼ 1; 2; ; mÞ. Therefore following proportion
transformations is employed (Goh et al. 2003):
If ~aij is for the benefit, then
a ij aþ ij
rij ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; rijþ ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5:1Þ
P þ 2
n P 2
n
ðaij Þ ðaij Þ
i¼1 i¼1
If ~
aij is for the cost, then
. .
1 aþ ij 1 a ij
rij ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; rijþ ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5:2Þ
P n . P n .
2 2
ð1 a ij Þ ð1 aþ ij Þ
i¼1 i¼1
In order to compare the similarity degree of two interval numbers and realize
the order of alternatives, we will introduce the concepts about deviation degree and
possibility degree of interval number.
Definition 1 (Xu and Da 2003): Suppose
a ¼ ½a ; aþ ; ~
~ b ¼ ½b ; bþ
are two interval numbers, let
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bÞ ¼ ~
a; ~
dð~ b ¼ ðb a Þ2 þ ðbþ aþ Þ2
a~
a and ~
be the deviation degree between ~ b.
Definition 2 (Xu and Da 2003): Suppose ~a ¼ ½a ; aþ ; ~b ¼ ½b ; bþ
are two interval numbers and l~a ¼ a a ; l~b ¼ bþ b ; let
þ
minfl~a þ l~b ; maxðaþ b ; 0Þg
a~
pð~ bÞ ¼ ð5:3Þ
l~a þ l~b
a~
be the possibility degree of ~ a~
b and ~ b be the order relationship between ~a
~
and b.
In this paper, we will research how to rank Xi ði ¼ 1; 2; ; nÞ according to the
decision matrix A~ and the preference information ~hi ð~hi ¼ ½h ; hþ Þ.
i i
44 Y. Li
5.3 Model and Method
According to the normalized matrix R ~ ¼ ð~rij Þ , the attributes weight vector
nm
T
w ¼ ðw1 ; w2 ; ; wm Þ and the algorithms related to interval numbers, the com-
prehensive attribute value of Xi is
X
m Xm X
m
~zi ¼ ~rij wj ¼ ½ rij wj ; rijþ wj ði ¼ 1; 2; ; nÞ ð5:4Þ
j¼1 j¼1 j¼1
P
m
where wj is the weight of the attribute uj and wj ¼ 1ðwj 0 j ¼ 1; 2; ; mÞ:
j¼1
The preference information of the decision maker is subjective judgment to the
comprehensive attribute values of alternatives. But because of the limitations of
many real factors, there are deviations between the preference information of the
decision maker and the comprehensive attribute values of alternatives. In view of
the rationality of the decision making, the attributes weight vector w ¼ ðw1 ;
w2 ; ; wm ÞT will minimize the total deviation between the preference informa-
tion of the decision maker and the comprehensive attribute values of alternatives.
So, we will give the following optimal model:
8 X n Xn X m Xm
>
>
> minDðwÞ ¼
> di2 ð~zi ; ~
hi Þ ¼ ½ð rij wj h
i Þ2 þ ð rijþ wj hþ 2
i Þ
< i¼1 i¼1 j¼1 j¼1
>
> X
m
>
> wj ¼ 1 wj 0 j ¼ 1; 2; ; m
: s:t
j¼1
Where
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX X
u m m
hi Þ ¼ tð
di ð~zi ; ~ rij wj h i Þ 2
þ ð rijþ wj hþ i Þ
2
j¼1 j¼1
ði ¼ 1; 2; ; nÞ ð5:5Þ
be the deviation degree between the subjective preference information of Xi and
the comprehensive attribute values of Xi .
We construct the Lagrange function:
X
n X
m X
m X
m
Lðw; kÞ ¼ ½ð rij wj h
i Þ 2
þ ð r þ
w
ij j h þ 2
i Þ þ 2kð wj 1Þ ð5:6Þ
i¼1 j¼1 j¼1 j¼1
oLðw;kÞ
Let owk ¼0 ðk ¼ 1; 2; ; mÞ, we have
5 A Method for Multiple Attribute Decision Making 45
X
n Xm X
m
½ð rij wj h
i Þr
ik þ ð rijþ wj hþ þ
i Þrik þ k ¼ 0
i¼1 j¼1 j¼1
ðk ¼ 1; 2; ; mÞ ð5:7Þ
i.e.
X
m X n X
n
½ ðrij rik þrijþ rikþ Þwj ¼ ðh þ þ
i rik þ hi rik Þ k
j¼1 i¼1 i¼1
ðk ¼ 1; 2; ; mÞ ð5:8Þ
Let em ¼ ð1; 1; ; 1ÞT ; g ¼ ðg1 ; g2 ; ; gm ÞT ; Q ¼ ðqkj Þmm ,
Where
X
n
gk ¼ ðh þ þ
i rik þ hi rik Þ ðk ¼ 1; 2; ; mÞ ð5:9Þ
i¼1
X
n
qkj ¼ ðrij rik þrijþ rikþ Þ ðk; j ¼ 1; 2; ; mÞ ð5:10Þ
i¼1
Then (8) will transform to the matrix formation
Qw ¼ g kem ð5:11Þ
Theorem 1 The matrix Q is a positive definite matrix.
Proof From (5.10), we know qkj ¼ qjk ðk; j ¼ 1; 2; ; mÞ, i.e. the matrix Q is a
symmetric matrix. Suppose Y ¼ ðy1 ; y2 ; ; ym ÞT is a nonzero vector, then
m X
X m X n
Y T QY ¼ ½ ðrij rik þrijþ rikþ Þyk yj
k¼1 j¼1 i¼1
n X
X m n X
X m X
m
¼ ½ðrik Þ2 þðrikþ Þ2 y2k þ ðrij rik þrijþ rikþ Þyk yj
i¼1 k¼1 i¼1 j¼1 k¼1
k6¼j
X
n X
m X
m
¼ ½ð rik yk Þ2 þð rikþ yk Þ2 [ 0
i¼1 k¼1 k¼1
So, the matrix Q is a positive definite matrix.
From the theorem 1, we know the matrix Q1 is existent, so
w ¼ Q1 ðg kem Þ ð5:12Þ
46 Y. Li
Because eTm w ¼ 1, so, we can conclude
eTm Q1 g 1
k¼ ð5:13Þ
eTm Q1 em
So,
eTm Q1 g 1
w ¼ Q1 ðg em Þ ð5:14Þ
eTm Q1 em
The optimal weight vector w ¼ ðw1 ; w2 ; ; wm ÞT can be calculated by (5.14),
and then the comprehensive attribute values ~zi ði ¼ 1; 2; ; nÞ of all the alter-
natives can be calculated by (5.4). Because ~zi ði ¼ 1; 2; ; nÞ are still interval
numbers, it is inconvenient to rank the alternatives, so we will calculate the
possibility degree of ~zi ði ¼ 1; 2; ; nÞ by using formula (5.3) and establish the
possibility degree matrix P ¼ ðpil Þnn , where pil ¼ pð~zi ~zl Þði; l ¼ 1; 2; ; nÞ.
The matrix P ¼ ðpil Þnn is a complementary judgement matrix, the priority
vector x ¼ ðx1 ; x2 ; ; xn ÞT of P can be given by using the following formula
(Da and Xu 2002)
!
1 X n
n
xi ¼ pil þ 1 ði ¼ 1; 2; ; nÞ ð5:15Þ
nðn 1Þ l¼1 2
And further, the best alternative will be given if we rank all the alternatives
based on the components of the priority vector x ¼ ðx1 ; x2 ; ; xn ÞT .
Based on the above discussion, we develop a new method to analyze the
MADM problem, in which weight information is unknown, the attribute values are
expressed by interval numbers and the decision maker has preference information
expressed by interval numbers on alternatives. There are six steps in the new
method:
Step 1. Let X ¼ fX1 ; X2 ; ; Xn g ðn 2Þ be the set of alternatives, U ¼ fu1 ;
u2 ; ; um g ðm 2Þ be the set of attributes and A ~ ¼ ð~aij Þ
nm be the decision matrix
þ
where ~ aij ¼ ½aij ; a ij is the attribute value of alternative Xi with respect to the
attribute uj ði ¼ 1; 2; ; n; j ¼ 1; 2; ; mÞ;
Step 2. According to (5.1) and (5.2), the decision matrix A ~ ¼ ð~aij Þ
nm nor-
~
malized into the dimensionless decision matrix R ¼ ð~rij Þnm ;
Step 3. According to (5.14), we obtain the optimal weight vector w ¼ ðw1 ;
w2 ; ; wm ÞT ;
Step 4. Utilize (5.4) to get the comprehensive attribute values ~zi ði ¼ 1;
2; ; nÞ;
Step 5. Utilize (5.3) to get the possibility degree of ~zi ði ¼ 1; 2; ; nÞ and
establish the possibility degree matrix P ¼ ðpil Þnn ;
Step 6. We calculate the priority vector x ¼ ðx1 ; x2 ;
5 A Method for Multiple Attribute Decision Making 47
; xn ÞT of the possibility degree matrix P ¼ ðpil Þnn and rank all the alter-
natives based on the components of the priority vector x ¼ ðx1 ; x2 ; ; xn ÞT ,
then get the best alternative.
5.4 Illustrative Example
When the decision maker select cadres, on one hand, he wants to select capable
cadres, on the other hand, he also wants to select his preferred cadres. Hence, there
is preference information on alternatives. Now, suppose some company faces how
to select cadres. Firstly, the company built an index system with six attributes: u1 -
morality, u2 -attitude to work, u3 -style of work, u4 -levels of culture and
knowledge structure, u5 -ability of leadership and u6 - ability of innovation; sec-
ondly, the company determines five candidates xi ði ¼ 1; 2; ; 5Þ based on the
recommendation and evaluation of the masses and statistical treatment. Because
the evaluation results about the same candidate are different, so, the attribute
values after statistical treatment are expressed by interval numbers, as listed in the
following Table 5.1:
To select the best cadre, the six steps are included:
Step 1. We utilize (5.1), (5.2) to transform the decision matrix A ~ into the
dimensionless decision matrix
0 1
½0:378; 0:405 ½0:394; 0:414 ½0:398; 0:423 ½0:407; 0:432 ½0:394; 0:410 ½0:415; 0:437
B C
B ½0:394; 0:429 ½0:389; 0:410 ½0:394; 0:415 ½0:394; 0:415 ½0:411; 0:438 ½0:394; 0:419 C
B C
R¼B
~
B ½0:395; 0:420 ½0:377; 0:396 ½0:408; 0:433 ½0:408; 0:433 ½0:386; 0:410 ½0:408; 0:424 C
C
B C
@ ½0:385; 0:405 ½0:413; 0:433 ½0:385; 0:410 ½0:390; 0:414 ½0:395; 0:419 ½0:417; 0:433 A
½0:384; 0:410 ½0:402; 0:414 ½0:402; 0:414 ½0:407; 0:419 ½0:402; 0:414 ½0:380; 0:391
Step 2. Suppose the decision maker’s subjective preference value (after nor-
malized) on the five candidates xi ði ¼ 1; 2; ; 5Þ as follows:
~h1 ¼ ½0:3; 0:5; ~
h2 ¼ ½0:5; 0:6; ~h3 ¼ ½0:3; 0:4;
~
h4 ¼ ½0:4; 0:6; ~
h5 ¼ ½0:4; 0:5
Utilize (5.14) to get the optimal weight vector:
~
Table 5.1 Decision matrix A
u1 u2 u3 u4 u5 u6
x1 [0.85,0.90] [0.90,0.92] [0.91,0.94] [0.93,0.96] [0.90,0.91] [0.95,0.97]
x2 [0.90,0.95] [0.89,0.91] [0.90,0.92] [0.90,0.92] [0.94,0.97] [0.90,0.93]
x3 [0.88,0.91] [0.84,0.86] [0.91,0.94] [0.91,0.94] [0.86,0.89] [0.91,0.92]
x4 [0.93,0.96] [0.91,0.93] [0.85,0.88] [0.86,0.89] [0.87,0.90] [0.92,0.93]
x5 [0.86,0.86] [0.90,0.92] [0.90,0.95] [0.91,0.93] [0.90,0.92] [0.85,0.87]
48 Y. Li
w ¼ ð0:1639; 0:1658; 0:1666; 0:1681; 0:1674; 0:1682ÞT
Step 3. Utilize (5.4) to get the comprehensive attribute values of the five can-
didates xi ði ¼ 1; 2; ; 5Þ as follows:
~z1 ¼ ½0:3978; 0:4202; ~z2 ¼ ½0:3960; 0:4210;
~z3 ¼ ½0:3970; 0:4194; ~z4 ¼ ½0:3975; 0:4190;
~z5 ¼ ½0:3962; 0:4103
Step 4. Utilize (5.3) to calculate the possibility degree of ~zi ði ¼ 1; 2; ; 5Þ and
built the possibility degree matrix:
0 1
0:5 0:5105 0:5179 0:5171 0:6575
B 0:4895 0:5 0:5063 0:5054 0:6343 C
B C
P¼B B 0:4821 0:4937 0:5 0:4989 0:6356 C C
@ 0:4829 0:4946 0:5011 0:5 0:6405 A
0:3425 0:3657 0:3644 0:3595 0:5
Step 5. Utilize (5.15) to get the priority vector of the possibility degree matrix P;
x ¼ ð0:2102; 0:2068; 0:2055; 0:2059; 0:1716ÞT
Step 6. Rank all the alternatives xi ði ¼ 1; 2; ; 5Þ and select the best one
according to the xi ði ¼ 1; 2; ; 5Þ:
x1 x2 x4 x3 x5
Thus the most desirable candidate is x1 .
5.5 Conclusions
In this paper, a new method is proposed to discuss the MADM problem with
expert’s preference information expressed by interval numbers on alternatives, in
which the attribute weights information is unknown completely and the attribute
values are interval numbers. Considering the preference information of the
decision maker is subjective judgment to the comprehensive attribute values of
alternatives, hence the weight vector of attributes will minimize the total deviation
between the preference information values of the decision maker and the com-
prehensive attribute values of alternatives. In order to get the attribute weights, in
according to the above idea, we have established an optimal model based on the
deviation degree between the comprehensive attribute values of alternatives and
the preference information values of the decision maker. Solving the above model,
we can obtain the attribute weights and we utilize the priority vector of the
possibility degree matrix to compare the comprehensive attribute values of the
5 A Method for Multiple Attribute Decision Making 49
alternatives, and then rank the alternatives. The method is practical and effective
because it organically combines the subjective information and objective infor-
mation. Finally, an illustrative example is given to show the application of the
method.
References
Chen SY, Zhao YQ (1990) Fuzzy optimum theory and model (in Chinese). Fuzzy Syst Math
4(2):87–91
Cheng T (1987) Decision analysis. Science Press, Beijing
Da QL, Xu ZS (2002) Singule-objective optimization model in uncertain multi-attribute decision
making (in Chinese). J Syst Eng 17(1):50–55
Fan ZP, Ma J, Zhang Q (2008) An approach to multiple attribute decision making based on fuzzy
preference information on alternatives. Fuzzy Sets Syst 131(1):101–106
Gao FJ (2000) Multiple attribute decision making on plans with alternative preference under
incomplete information (in Chinese). Syst Eng Theory Pract 20(4):94–97
Goh CH, Tung YCA, Cheng CH (2003) A revised weighted sum decision model for robot
selection. Comput Ind Eng 30(2):193–199
Hwang CL, Yoon K (1981) Multiple attribute decision making: methods and applications.
Springer, New York
Jiang YP, Fan ZP (2005) Method for multiple attribute decision making with attribute interval
numbers and preference information on alternatives (in Chinese). Syst Eng Electr
27(2):250–252
Kim SH, Ahn BS (1999) Interactive group decision making procedure under incomplete
information. Eur J Oper Res 116(3):498–507
Kim SH, Choi SH, Kim JK (1999) An interactive procedure for multiple attribute group decision
making with incomplete information: range-based approach. Eur J Oper Res 118(1):139–152
Park KS, Kim SH (1997) Tools for interactive multi-attribute decision making with incompletely
identified information. Eur J Oper Res 98(1):111–123
Wei GW, Wei Y (2008) Model of grey relational analysis for interval multiple attribute decision
making with preference information on alternatives (in Chinese). Chin J Manage Sci
16(1):158–162
Xu ZS (2003) A method for multiple attribute decision making without weight information but
with preference information on alternatives (in Chinese). Syst Eng Theory Pract
23(12):100–103
Xu ZS (2004a) Method for multi-attribute decision making with reference information on
alternatives under partial weight information (in Chinese). Control Decis 19(1):85–88
Xu ZS (2004b) Projection method for uncertain multi-attribute decision making with preference
information on alternatives. Int J Inf Technol Decis Mak 3(3):429–434
Xu ZS, Da QL (2003) Possibility degree method for ranking interval numbers and its application
(in Chinese). J Syst Eng 18(1):67–70
Chapter 6
A Prediction of the Container Throughput
of Jiujiang Port Based on Grey System
Theory
Yan Du
Abstract Container handling system can be regarded as a grey system. Since the
grey GM (1,1) prediction model is applicable to small samples of data, the port
container throughout of Jiujiang port in the next 5 years can be predicted on the
basis of the relative data from 2006 to 2011. According to the predicted result,
feasible suggestions and available measures may be given to the development of
Jiujiang port, and scientific data may also be provided for it.
Keywords Container throughput GM (1,1) model Jiujiang port Prediction
6.1 Introduction
A port is served as the joint and shipping center of the land, sea and air. It plays an
important role in the entire logistics network. Meanwhile, as an integrated transport
hub and the main distribution center of import and export, the port’s economic
growth take an inevitably great position in the urban economy and regional devel-
opment. Jiujiang port is the only port connecting Yangtze River and seas in Jiangxi
Province, which has rather obvious advantage. The international container cargoes
in Jiangxi are mainly transported to the coastal ports through Jiujiang port. As a
feeder port of Shanghai port, the cargoes are first shipped from Jiujiang port to
Shanghai and then exported to the other cities and countries in the world. In 2008, the
two ports established a closer cooperation. The Shanghai Port provided the Jiujiang
Port with the advanced management experience, ideas and other sorts of resources.
A new Jiujiang port is growing bigger and better with the rapid development of
Jiangxi’s economy, its role is increasingly evident in recent years (Du 2011).
Y. Du (&)
BusinessSchool, JiujiangUniversity, Jiujiang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 51
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_6,
Ó Springer-Verlag Berlin Heidelberg 2013
52 Y. Du
The throughput prediction of the container has extremely important influence
on the future of a container port. The method of the prediction consists of quali-
tative and quantitative method. Qualitative prediction method includes Delphi
Method, the Subjective Probability Method and Supply Survey Method. Quanti-
tative prediction method includes Regression Analysis, Exponential Smoothing,
Gary Prediction Method, the Neural Network Method, Elastic Prediction Method
and Combined Prediction Method. The above models vary a lot in the scope of
application, complexity and accuracy. Therefore, specific analysis of the specific
objectives is required (Chen and Gao 2008; Zhou 2007).
Several factors influence a port’s container throughput, such as the economic
development of the hinterland, the port’s operation, its infrastructure, market
demand and so on (Liu and Wang 2005). The relationship among them is com-
paratively complicated and changeable, which is quite in line with the charac-
teristics of the grey system. In this paper, the grey prediction model is applied to
predict Jiujiang port’s container throughput based on the data in the past 6 years.
6.2 The Gray Prediction Model of Container Throughput
The Grey System Theory was built by Professor Deng Julong, a famous scholar in
China in 1982. It is a new method to study a problem of less data, poor information
and uncertainty. Gray System Theory specializes studying in the uncertain system,
which is a small sample of poor information by extracting valuable information
through the development of some of the known information to realize correct
description and effective monitoring and control system operation behavior and
evolution (Deng 2002). In this theory, ‘‘black’’ means information is unknown,
‘‘white’’ indicates information is completely clear, and ‘‘grey’’ describes some of
the information is clear and some information is not clear. Accordingly, the system
whose information is unknown is called a BlackSystem. The system whose
information is completely clear is called the WhiteSystem. The system between
them is called a Gray System (Liu and Xie 2008).
Gray System Model has a broad range of applications because there are no
special requirements and restrictions on the observed experimental data. Currently,
the most widely used Grey Prediction Model is GM (1,1) model, which contains a
variable, first derivative to predict the data.
The GM (1,1) model is based on a random time of the original sequence,
formed by the cumulative chronological time sequence showing the law can be
used first-order linear differential equations to approximate. In this paper, the grey
prediction GM (1,1) model, namely, the first-order linear differential equation
model in a variable model prediction is applied hereby.
6 A Prediction of the Container Throughput of Jiujiang Port 53
6.2.1 GM (1,1) model
The gray system modeling is the use of discrete time series data to establish the
approximate continuous differential equation model. In this process, the Accu-
mulated Generating Operator (AGO) is the basic means. Its generating function is
the basis of gray modeling and predictable (Liu and Deng 1999).
Equipped with the original sequence:
n o
xð0Þ ðtÞ ¼ xð0Þ ð1Þ; xð0Þ ð2Þ; . . .. . .; xð0Þ ðnÞ
(n is the length of time series of samples)
Coefficient column accumulated generating the AGO incremental series:
n o
xð1Þ ðtÞ ¼ xð1Þ ð1Þ; xð1Þ ð2Þ; . . .. . .; xð1Þ ðnÞ
P
And xð1Þ ðkÞ ¼ ki¼1 xð0Þ ðiÞ K ¼ 1; 2; . . .; n
xð1Þ ðkÞ GM (1,1) model for the albino form of the differential equation as
follows:
dxð1Þ
þ axð1Þ ¼ u ð6:1Þ
dt
a, u: Undetermined parameters
Differential to replace the differential, differential convenient variable trans-
formed into
xð0Þ ð2Þ 1=2 xð1Þ ð1Þ þ xð1Þ ð2Þ 1
xð0Þ ð3Þ 1=2 xð1Þ ð2Þ þ xð1Þ ð3Þ 1 a
... ¼ ...
þ
. . . u
ð6:2Þ
xð0Þ ðnÞ 1=2 xð1Þ ðn 1Þ þ xð1Þ ðnÞ 1
Abbreviated as follows:
YN ¼ B a ð6:3Þ
Using the least square method, you can get the solution of the equations are:
a 1
a ¼ ¼ BT B BT YN ð6:4Þ
u
Into the original differential equations, you can have the formula:
h uiak u
xð1Þ ðk þ 1Þ ¼ xð0Þ ð1Þ þ ð6:5Þ
a a
54 Y. Du
And then by the formula
xð0Þ ðkÞ ¼ xð1Þ ðkÞ xð1Þ ðk 1Þ to restore:
Series xð0Þ ðkÞ is the calculated value of the original series.
6.2.2 Inspection of the Model
In order to determine whether it is a reasonable and qualified model, a variety of
inspection is required. Only the model which goes through the test can be used as
predictors. A posteriori differential is used here to test the model commonly (Wang
2006; Xiong and Xu 2005).
The steps are as follows:
(1) Seek the mean of raw data and residuals
1 Xn
x ¼ xð 0 Þ ð k Þ ð6:6Þ
n k¼1
1 Xn
q¼
qð k Þ ð6:7Þ
n k¼1
(2) Find the variance of the raw data and residuals:
1 Xn h ð0Þ i2
s21 ¼ x ð k Þ
x ð6:8Þ
n k¼1
1 Xn
s22 ¼ ½qðkÞ q2 ð6:9Þ
n k¼1
(3) The calculation of a posteriori differential ratio c and the small error proba-
bility p:
s2
c¼ ð6:10Þ
s1
p ¼ pf j qð k Þ
qj\0:6745s1 g ð6:11Þ
Accuracy of the model is determined by c and p. A posteriori differential ratio c
as small as possible (The small c show that s2 is small and s1 is big. The variance
of residual is small. The variance of the raw data is big, Description of residual is
more concentrated, Small swings. Raw data is more scattered, large swings. So, if
you want to simulate effect, requires that s2 small as possible compared to s1 ).
Small error probability is the bigger the better. Accuracy level is in Table 6.1.
6 A Prediction of the Container Throughput of Jiujiang Port 55
Table 6.1 A posteriori Prediction accuracy P c
differential checklist
Good [0.95 \0.35
Qualified [0.80 \0.40
Reluctantly [0.70 \0.45
Failure B0.70 C0.45
6.3 A Prediction of the Container Throughput
of Jiujiang Port
6.3.1 Calculation
The container throughput of Jiujiang Port in recent years is in Table 6.2. It shows
that from 2006 to 2011, the container throughput of Jiujiang Port is growing
rapidly overall, only in 2008 declined compared to 2007 because of the interna-
tional financial crisis. In 2009, the volume grew again and has maintained a good
momentum of sustained growth in 2010 and 2011 (Shao 2010).
We can see from Table 6.2:
xð0Þ ¼ ð7:38; 8:80; 8:09; 10:07; 12:06; 14:00Þ
Cumulative, xð0Þ time accumulated generating operation sequence,
xð1Þ ¼ ð7:38; 16:18; 24:27; 34:34; 46:40; 60:40Þ
ð1Þ
Modeling dx ð1Þ
dt þ ax ¼ u
Calculate a and u using the least squares method, to determine the parameter
vector:
1
a ¼ ða; uÞT ¼ BT B BT YN
1=2 xð1Þ ð1Þ þ xð1Þ ð2Þ 1 11:78
1=2 xð1Þ ð2Þ þ xð1Þ ð3Þ 20:225 11
1
B ¼ 1=2 xð1Þ ð3Þ þ xð1Þ ð4Þ 1 ¼ 29:305 1
ð1Þ
1=2 x ð4Þ þ xð1Þ ð5Þ 1 40:37 1
ð1Þ 53:4 1
1=2 x ð5Þ þ xð1Þ ð6Þ 1
Table 6.2 Container throughput and cumulative data table of Jiujiang port over the years (Unit:
Million TEU)
Years 2006 2007 2008 2008 2010 2011
Throughput 7.38 8.80 8.09 10.07 12.06 14.00
Cumulative 7.38 16.18 24.27 34.34 46.40 60.40
TEU Twenty-foot Equivalent Unit
56 Y. Du
Table 6.3 The actual values and predicted values (Unit: Million TEU)
Years 2006 2007 2008 2009 2010 2011
Actual value 7.38 8.80 8.09 10.07 12.06 14.00
Predictive value 7.38 7.80 8.98 10.35 11.92 13.72
Year 2012 2013 2014 2015 2016
Actual value – – – – –
Predictive value 18.22 20.99 24.18 27.85 32.08
YN ¼ ½8:80; 8:09; 10:07; 12:06; 14:00
1
B; YN into a ¼ ða; uÞT ¼ BT B BT YN ;
Get a ¼ ½0:1414 6:2195T
(With the help of MATLAB) (Lee and Ruan 2013; Shi and Teng 2004)
Take parameter vector a into the differential equation:
dxð1Þ
0:1414xð1Þ ¼ 6:2195
dt
Get prediction model as follows:
h ui u
xð1Þ ðk þ 1Þ ¼ xð0Þ ð1Þ eak þ
a a
0:1414k
¼ 51:37 e 43:99
And then by the formula:
xð0Þ ðkÞ ¼ xð1Þ ðkÞ xð1Þ ðk 1Þ to restore
Thus available the container throughput predictive value of Jiujiang port from
2006 to 2011 (Table 6.3), forecast and actual values is close.
6.3.2 Test the Accuracy of the Model
Posteriori difference inspection carried out for the model: The mean of the original
data for: x ¼ 10:067; The mean of the residuals is: q ¼ 0:0416; The variance of
the original data for: s21 ¼ 4:47; s1 ¼ 2:114; Variance of the residuals: s22 ¼ 0:337;
s2 ¼ 0:58; the ratio of a posteriori difference: c ¼ ss21 ¼ 0:27; Small probability of
error: p = 1.
According to Table 6.1, p = 1 [ 0.95, c ¼ 0:27, it can be judged that the
prediction is comparatively accurate. Therefore, this prediction model can predict
the container throughput of Jiujiang Port in the next few years, which is shown in
Table 6.3.
6 A Prediction of the Container Throughput of Jiujiang Port 57
The calculated results show that Gray System Theory is more accurate to
predict the container throughput of Jiujiang port. The calculations and statistics are
very close. From the predicted results can be seen, the growth rate of the container
throughput of Jiujiang port in the next 5 years is fast and stable. It can be inferred
that the container transport of Jiujiang port will be greatly developed. The basic
data of the model is established on the basis of the 2006–2011 objective statistics.
In 2012 and the coming years, there are some unknown factors, which are difficult
to determine and affect the economic environment and the development of
logistics industry. Objectively speaking, the model also has a certain degree of
uncertainty, but as the development of the industry prediction, it has a relatively
reliable reference.
6.4 Suggestions to Jiujiang Port
Taking the development needs of the Jiujiang port, container throughput prediction
and the actual situation of Jiujiang port into account, specific measures for the
development of port logistics to Jiujiang port are provided as following.
6.4.1 Strengthening the Policy Guidance and Support
of the Government of the Port Logistics
During the ‘‘Eleventh Five-Year’’ period, make good use of the national imple-
mentation of the riverside development strategy to promote the construction of the
Yangtze golden waterway and to speed up port construction and port opening to
create the Jiujiang logistics distribution center, which will become the focus of
infrastructure investment in Jiangxi Province in order to enhance the Jiangxi
external the level of openness.
6.4.2 Paying More Attention to the Construction
of Supporting Transportation Network
A port’s development requires a rational and scientific layout and construction.
Therefore, full attention should be given supporting the construction of integrated
transport network of railways, highways, waterways and other modes of trans-
portation, which can be brought into full play of the advantage of port throughput.
Dock construction is the leading part to strengthen the railways, highways sup-
porting system (Zheng 2011; Jiao and Huang 2008).
58 Y. Du
6.4.3 Enhancing Service Awareness and Improving
the Management Level
Establishing the concept of logistics services, providing customers with safe,
cheap, fast and punctual container transportation and logistics services is the key to
form the mutual trust and to achieve a win–win situation. The port is able to
provide customers with a multi-functional container logistics services to meet the
logistics needs of the personalized customers. Great efforts must be made to train
and preserve all kinds of talents, who are required for the development of the port.
Improving the status and responsibility distribution system, giving full play to
the enthusiasm of the staff of the enterprise is also needed to form a good corporate
culture and provide a strong backing for the further development of Jiujiang port.
6.4.4 Expanding the Economic Hinterland to Promote
the Establishment of Cooperation Mechanisms
An economic hinterland is an area which the economic center can be touched and
can promote their economic development. Under the concept of modern logistics,
these factors which affect the development of the port are growing, such as ports
and the hinterland, the degree of value-added services capacity. After the opening
of the Beijing-Kowloon Railway, the hinterland economy of Jiujiang port will be
even broader. Jiujiang port is not only responsible for loading and unloading tasks
for the pillar industries transit in Jiangxi Province but also gradually expands to
inland provinces such as Hubei, Sichuan and Henan. In the strategy of opening up
along the river, Jiujiang must strengthen the cooperation of the adjacent regional
cities inside and outside the province and promote the establishment of coopera-
tion mechanisms (Yanbing and Liang 2003).
In summary, Jiujiang port should integrate existing resources to strengthen
infrastructure construction, to accelerate the chain structures of port container
logistics services, and to improve the efficiency of service operations to container
stations and operational management level. The goal is to provide a powerful
guarantee for the rapid development of Jiujiang port container transport.
References
Chen T, Gao Q (2008) The study on the factors influencing on the container throughput of the
port. Wuhan Univ Technol (Information and Management Engineering) 30(6):991–994+1003
Deng J (2002) The basis of the gray theory. Huazhong University of Science and Technology
Press, Wuhan, pp 100–118, 361–369
Du Y (2011) The studying on the issues of the development of port logistics distriparksin
Jiujiang, Nanchang University, 2011
6 A Prediction of the Container Throughput of Jiujiang Port 59
Jiao N, Huang G (2008) The container throughput forecast of Dalian port. World Shipping 6:7–9
Lee G, Ruan X (2007) Economic forecasting and decision-making and MATLAB. Tsinghua
University Press, Beijing
Liu S, Deng J (1999) GM (1,1) coding for exponential series. J Gray Syst 2:147–152
Liu M, Wang D (2005) Port cargo throughput forecasting method. Waterw Eng 2005 (3):53–56
Liu S, Xie N (2008) Gray system theory and its application. Science Press, Beijing, pp 66–78
Shao Z (2010) A prediction of the container throughput based on grey system theory—take
Wuhan port as an example. Port Econ 12:40–42
Shi B, Teng G (2004) The tutorial of the math example of MATLAB. China Railway Publishing
House, Beijing
Sun Y, Cheng L (2003) Gray prediction model in port container throughput forecast. Port
9:36–38
Wang F (2006) Improvement on unequal interval gray forest model. Fuzzy Inf Eng 6(1):118–123
Xiong H, Xu H (2005) Grey Control. National Defence Industry Press, Beijing, pp 32–35
Zheng Y (2011) The gray prediction of container throughput of Ningbo—Zhoushan port. China
Water Transp 11:2
Zhou Y (2007) The application of grey forecast model in the logistics needs. Modern logistics,
2007(11):59–61
Chapter 7
A Rapid Data Processing and Assessing
Model for ‘‘Scenario-Response’’ Types
Natural Disaster Emergency Alternatives
Daji Ergu, Gang Kou and Yong Zhang
Abstract In the processes of assessing the emergency alternatives of
‘‘scenario-response’’ types natural disaster by Analytic Hierarchy (Network)
Process (AHP/ANP), the elements or data of the scenario itself, the real-time data
and the trend factors of the evolution of ‘‘scenario-response’’ types natural disaster
emergencies etc. are usually inconsistent and intangible, which increase the dif-
ficulty of emergency alternatives assessment and delay the speed of emergency
response. Therefore, in this paper, a logarithm mean induced bias matrix
(LMIBM) model is proposed to quickly process the inconsistent data of ‘‘scenario-
response’’ type’s natural disaster when AHP/ANP is used to assess the natural
disaster emergency alternatives and evolution trend factors of natural disaster
emergency accidents. Two numerical examples are used to illustrate the proposed
model, and the results show that LMIBM can quickly identify the inconsistent
natural disaster data and improve the speed of emergency alternatives assessment
and natural disaster response by AHP/ANP.
Keywords AHP/ANP Data processing Emergency alternatives assessment
Logarithm mean induced bias matrix Natural disaster
D. Ergu (&) G. Kou Y. Zhang
School of Management and Economics, University of Electronic Science and Technology of
China, Chengdu, China
e-mail: [email protected]
G. Kou
e-mail: [email protected]
Y. Zhang
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 61
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_7,
Ó Springer-Verlag Berlin Heidelberg 2013
62 D. Ergu et al.
7.1 Introduction
In recent years, a number of studies on unexpected natural disaster emergency
alternatives assessment from different perspectives have been conducted by many
scholars since natural disasters have caused significant economic, social, financial,
property and infrastructure damages as well as tragic loss of human lives world-
wide (Xu and Lu 2009; Zhong et al. 2010; Wei et al. 2004; Bina and Edwards
2009). Quick assessment and scenario-response are very important for disaster
managers to save lives and reduce the property losses in unexpected natural
disaster emergency management (Tsai and Chen 2010, 2011). Therefore, ‘‘sce-
nario-response’’ type’s unexpected natural disaster emergency management is
becoming a new hot research topic.
However, the involved attributes of ‘‘scenario-response’’ types natural disasters
are different and usually intangible, which need to be quantified into quantitative
values. The elements or data of the scenario itself, the real-data and the trend
factors of the evolution of ‘‘scenario-response’’ type’s natural disaster emergencies
etc. are also usually inconsistent and intangible, which increase the difficulty of
emergency alternatives assessment and delay the speed of emergency response. In
addition, emergency management is regarded as a complex multi-objective opti-
mization problem, e.g. selecting the best emergency response alternatives or
emergency recovery alternatives and reasonably allocates relief resources etc.,
(Tufekci and Wallace 1998).
Furthermore, the unconventional characteristic of natural disaster evolution law
increases the difficulty of quick assessment and response. The pairwise comparison
technique, displayed as a positive reciprocal matrix in the Analytic Hierarchy
Process (AHP) (Saaty 1994) and Analytic Network Process (ANP) (Levy and Taji
2007), is a well-established and widely used technique to deal with the intangible
disaster attributes and assess the natural disaster emergency alternatives. However,
the consistency issue in a positive reciprocal matrix is a big challenge that the
emergency managers are facing, and it has been extensively studied over the past
few decades, e.g. Koczkodaj and Szarek developed distance-based inconsistency
reduction algorithms for pairwise comparisons (Koczkodaj and Szarek 2010; Cao
et al. 2008) proposed a heuristic approach to modify the inconsistent comparisons
in analytic hierarchy process, Li and Ma employed a Gower plot to detect ordinal
and cardinal inconsistencies (Li and Ma 2007). Ergu et al. (2011b) proposed an
induced bias matrix (IBM) model to improve the consistency ratio and further
extended it to process data consistency in emergency management (Ergu et al.
2012) and in risk assessment and decision analysis (Ergu et al. 2011a). Based on
the proposed induced bias matrix (IBM) model, in this paper, a logarithm mean
induced bias matrix (LMIBM) model is proposed to process the inconsistent
disaster data and simplify the observed numbers in order to improve the speed of
disaster assessment and response when AHP/ANP is used.
The rest of this paper is organized as follows. The next section presents the
theorems, corollary as well as the identifying processes of LMIBM. Two
7 A Rapid Data Processing and Assessing Model for ‘‘Scenario-Response’’ 63
numerical examples are used to test the proposed model in Sect. 7.3. Section 7.4
concludes the paper.
7.2 Logarithm Mean Induced Bias Matrix (LMIBM)
7.2.1 The Theorem of LMIBM
Definition 7.1 Matrix A = [aij]n 9 n is said to be perfectly cardinal consistency if
aij = aikakj holds for all i, j and k, where aij [ 0 and aij = 1/aji for all i, j, and k.
Theorem 7.1 The logarithm mean induced bias matrix (LMIBM) C should be a
zero matrix if matrix A is perfectly consistent, that is,
8 Q
< C ¼ 1n log
A log A ¼ 0
1
Qn ð7:1Þ
: cij ¼ n log aik akj log aij ¼ 0
k¼1
Proof Since matrix A is perfectly consistent, namely, aik akj ¼ aij holds for all i,
j and k, we have
1 Yn
cij ¼ log aik akj log aij
n k¼1
1
¼ log anij log aij ¼ 0
n
Therefore, all values in matrix C are zeroes, and matrix C is a zero matrix if
matrix A is perfectly consistent. h
Obviously, the above model can be transformed to the following models in terms
of the properties of logarithm function:
8 p ffiffiffiffiffiffiffiffiffi
Q
< C ¼ log
> n
s þ log AT ¼ 0
Affiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
n Q
n ð7:2Þ
>
: cij ¼ log aik akj þ log aji ¼ 0
k¼1
8 p ffiffiffiffiffiffiffiffiffi
Q
< C ¼ log
> n
AT ¼ 0 !
sAffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n Q
n ð7:3Þ
>
: cij ¼ log aik akj aji ¼ 0
k¼1
Theorem 7.2 The logarithm mean induced bias matrix (LMIBM) C is an anti
symmetric matrix if matrix A is inconsistent, that is,
64 D. Ergu et al.
C ¼ C T
ð7:4Þ
cij ¼ cji
Proof By formula (7.1) and the reciprocal property, we have
1 Yn
1 Yn
1 1
cji ¼ log ajk aki log aji ¼ log log
n k¼1
n a a
k¼1 kj ik
aij
!
1 Yn
¼ log akj aik þ log aij ¼ cij
n k¼1
h
Corollary 7.1 There must be some inconsistent entries in the logarithm mean
induced bias matrix (LMIBM) C deviating from zero if the matrix A is inconsistent.
Especially, any row or column of matrix C contains at least one non-zero entry.
Proof by contradiction: Assume all entries in matrix C are zeroes even if the
judgment matrix A is inconsistent, that is, aik akj 6¼ aij holds for some i, j and k, but
cij ¼ 0; ði; j ¼ 1; . . .; nÞ, namely
1 Yn
cij ¼ log aik akj log aij ¼ 0
n k¼1
We can get
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n
Yn
n
Yn
aij ¼ aik akj ¼ ail alj
k¼1 l¼1
Since alk = 1/akl, we can obtain
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n
Yn
n
Yn
n
Yn
aij ¼ ail alk akl alj ¼ ail alk akl alj
l¼1 l¼1 l¼1
) aij ¼ aik akj
The result contradicts the previous assumption that aij 6¼ aik akj for some j and k,
and cij ¼ 0. Therefore, any row or column of matrix C contains at least one non-
zero entry. h
Based on Corollary 7.1, the most inconsistent data in matrix A can be identified
by observing the largest value in the logarithm mean induced bias matrix C.
According to Theorem 7.2, there are two equal absolute largest values in matrix C
since the values are anti symmetric, therefore, one can either observe the absolute
largest value above or below the main diagonal in the matrix C, which simplify the
observing number of times to n(n - 1)/2 and improve the speed of disaster
assessment.
7 A Rapid Data Processing and Assessing Model for ‘‘Scenario-Response’’ 65
7.2.2 The Processes of Inconsistency Identification
by LMIBM
In this section, the processes of inconsistency identification by LMIBM are pro-
posed based on above Theorems and Corollary, including three steps:
Step 1: Compute a column matrix L and a row matrix R, as shown in the two
edges of matrix A.
L¼
0 1
a11 a1i a1n
a1j Q
n
B .. .. .. C a1k
B C
B . . . C k¼1
B C Q
n
B ai1 aii aij ain C
B C aik
B C k¼1
A¼BB .
.. .
.. C
C Qn
B C ajk
B .. C
B aj1 aji ajj . ajn C k¼1
B C
B C Qn
@ A ank
k¼1
an1 ani anj ann
Yn Y
n Y
n Y
n
R¼ ak1 ; aki ; akj ; akn
k¼1 k¼1 k¼1 k¼1
where
8 n T
>
> Q Q
n Q
n Q
n
>
<L¼ a1k ; aik ; ; ajk ; ; ank
k¼1 k¼1 k¼1 k¼1
n ð7:5Þ
>
> Q Qn Q
n Q
n
>
:R ¼ ak1 ; aki ; akj ; ; akn
k¼1 k¼1 k¼1 k¼1
Step 2: Compute logarithm mean induced bias matrix (LMIBM) C by formula,
8 Q
< C ¼ 1n log
A n log A
Q ð7:6Þ
: cij ¼ 1n log aik akj log aij
k¼1
where
!
Y Y
n
A¼LR¼ aik akj
k¼1 nn
Step 3: Observe the absolute largest values above or below the main diagonal
of matrix C, here denoted as cmax
ij , then we can easily identify the corresponding aij
in matrix A as the most inconsistent data. If there are other entries whose values
are close to cmax
ij , they can be identified as the most inconsistent entries.
66 D. Ergu et al.
7.2.3 The Processes of Inconsistency Adjustment by LMIBM
In above section, the most inconsistent data is identified. In the following, the
estimating formula of identified inconsistent data is derived to adjust the identified
inconsistent data. Assume aij is the identified inconsistent data, which is corre-
sponding to the cmax
ij in matrix C, we have
1 Yn
cmax
ij ¼ log aik akj log aij
n k¼1
1 Yn
¼ log a2ij aik akj log aij
n lk¼16¼i;j
1
¼ log a2ij ~
an2
ij log aij
n
We can get
ncmax
ij ¼ log a2ij ~an2
ij =aij
p ffiffiffiffiffiffiffiffiffiffiffiffiffi
max
ncmax
ij
ð7:7Þ
n2
)~aij ¼ aij 10ncij ¼ aij 10 n2
Therefore, the identified inconsistent can be estimated by formula (7.7).
7.3 Illustrative Examples
Assume a decision maker needs to quickly assess the disaster degree of four places
attacked by earthquake with respect to the indirect economic loss to make a
‘‘scenario-response’’ types relief resource allocation, the 4 4 matrix A with
CR = 0.173 [ 0.1 used in Ergu et al. (2011b) and Liu (1999) is assumed to be the
collected judgment matrix by emergency expert in this paper, that is,
0 1
1 1=9 3 1=5
B 9 1 5 2 C
A¼B @ 1=3 1=5 1 1=2 A
C
5 1=2 2 1
Apply the LMIBM to this matrix.
7.3.1 Step I: Inconsistency Identification
Step 1: Compute the column matrix L and the row matrix R by formula (7.5),
7 A Rapid Data Processing and Assessing Model for ‘‘Scenario-Response’’ 67
L ¼ ð 0:0667 90 0:0333 5 ÞT
,
R ¼ ð 15 0:0111 30 0:2 Þ
Step 2: Compute LMIBM C by formula (7.6), we get
1
C ¼ logðL RÞ log A
4
0 1
0 0:1717 0:4019 0:2302
B 0:1717 0 0:1589 0:0128 C
B C
¼B C
@ 0:4019 0:1589 0 0:2430 A
0:2302 0:0128 0:2430 0
Step 3: Identify the absolute largest value cmax
either above the main diagonal
ij
in matrix C or below it. Here the data below the main diagonal are used, and we
obtain that cmax
ij ¼ cmax
31 ¼ 0:4019, then the corresponding element a31 in matrix A
is regarded as the most inconsistent entry.
7.3.2 Step II: Inconsistency Adjustment
Applying formula (7.7) to estimate the possible proper value of a31, we get
4 1
a31 ¼ a31 1042c31 ¼ 1020:4019 ¼ 2:1217 2
~
3
Therefore, the identified inconsistent data and its estimated value are the same
as the ones in (Ergu et al. 2011b) and (Liu 1999), whose CR = 0.0028 \ 0.1, but
the number of observed entries is reduced to 6 entries.
In the following, an 8 9 8 pair-wise comparison matrix A introduced in Cao
et al. (2008), Ergu et al. (2011b) and Xu and We (1999) is introduced to test the
proposed model for matrix with high order.
2 3
1 5 3 7 6 6 1=3 1=4
6 1=5 1 1=3 5 3 3 1=5 1=7 7
6 7
6 1=3 3 1 6 3 4 6 1=5 7
6 7
6 1=7 1=5 1=6 1 1=3 1=4 1=7 1=8 7
A¼6 6 7
6 1=6 1=3 1=3 3 1 1=2 1=5 1=6 7 7
6 1=6 1=3 1=4 4 2 1 1=5 1=6 7
6 7
4 3 5 1=6 7 5 5 1 1=2 5
4 7 5 8 6 6 2 1
Applying the LMIBM to this matrix.
68 D. Ergu et al.
7.3.3 Step I: Inconsistency Identification
Step 1: The column matrix L and the row matrix R by formula (7.5) are,
L ¼ ð 315 0:0857 86:4 0 0:0009 0:0037 218:75 80640 Þ
R ¼ ð 0:0032 11:6667 0:0116 141120 1080 270 0:0046 0Þ
Step 2: Compute LMIBM C by formula (7.6), we get
1
C ¼ logðL RÞ log A
8
0 1
0 0:2533 0:4069 0:1109 0:0867 0:1619 0:4969 0:3010
B C
B 0:2533 0 0:1017 0:1886 0:2313 0:3066 0:2731 0:0984 C
B C
B C
B 0:4069 0:1017 0 0:1076 0:1441 0:0561 0:8286 0:3277 C
B C
B C
B 0:1109 0:1886 0:1076 0 0:2126 0:2623 0:0911 0:3539 C
B C
¼B C
B 0:0867 0:2313 0:1441 0:2126 0 0:2258 0:0273 0:2143 C
B C
B C
B 0:1619 0:3066 0:0561 0:2623 0:2258 0 0:1026 0:1391 C
B C
B C
B 0:4969 0:2731 0:8286 0:0911 0:0273 0:1026 0 C
0:0198 A
@
0:3010 0:0984 0:3277 0:3539 0:2143 0:1391 0:0198 0
Step 3: The absolute largest value cmax
ij above the main diagonal in matrix C is
max
c37 ¼ 0:8286, then the corresponding element a37 in matrix A is regarded as the
most inconsistent entry.
7.3.4 Step II: Inconsistency Adjustment
Applying formula (7.7) to estimate the possible proper value of a37, we get
8
a37 ¼ a37 1082c37 ¼ 6108=6ð0:8286Þ ¼ 0:4714 1=2
~
The identified inconsistent data are the same as the ones in Cao et al. (2008),
Ergu et al. (2011b) and Xu and We (1999), whose CR = 0.0828 \ 0.1, but the
number of observed entries is reduced to 28 entries instead of 56.
7.4 Conclusion
In this paper, a logarithm mean induced bias matrix (LMIBM) is proposed to
quickly process the inconsistent disaster data when AHP and ANP are used to
assess the ‘‘scenario-response’’ type’s natural disaster emergency alternatives. The
7 A Rapid Data Processing and Assessing Model for ‘‘Scenario-Response’’ 69
processes of inconsistency identification of LMIBM and estimating formula are
proposed and derived. Two numerical examples are used to illustrate the proposed
model. Since LMIBM is not only based on the original matrix and independent to
the way of deriving the priority weights, but also can reduce the observed number
of the induced bias data, therefore, the proposed model can speed up the processes
of inconsistent disaster data identification in a judgment matrix and improve the
speed of ‘‘scenario-response’’ types disaster emergency alternatives assessment
and response.
Acknowledgments This research has been supported by grants from the talent research plan of
Foxconn Technology Group (11F81210001).
References
Bina O, Edwards B (2009) Social capital and business giving to charity following a natural
disaster: an empirical assessment. J Socio-Econ 38:601–607
Cao D, Leung LC, Law JS (2008) Modifying inconsistent comparison matrix in analytic
hierarchy process: a heuristic approach. Decis Support Syst 44:944–953
Tsai C-H, Chen C-W (2010) An earthquake disaster management mechanism based on risk
assessment information for the tourism industry—a case study from the island of Taiwan.
Tour Manag 31:470–481
Ergu D, Kou G, Shi Y, Shi Y (2011a) Analytic network process in risk assessment and decision
analysis. Comput Oper Res. doi:10.1016/j.cor.2011.03.005 (in press)
Ergu D, Kou G, Peng Y, Shi Y (2011b) A simple method to improve the consistency ratio of the
pair-wise comparison matrix in ANP. Eur J Oper Res 213(1):246–259
Ergu D, Kou G, Peng Y, Shi Y, Li F (2012) Data consistency in emergency management. Int J
Comput Commun Control 7(3):451–459
Xu J, Lu Y (2009) Meta-synthesis pattern of analysis and assessment of earthquake disaster
system. SETP 29(11):1–18
Levy JK, Taji K (2007) Group decision support for hazards planning and emergency
management: a Group Analytic Network Process (GANP) approach. Math Comput Model
46(7–8):906–917
Koczkodaj WW, Szarek SJ (2010) On distance-based inconsistency reduction algorithms for
pairwise comparisons. Logic J IGPL 18(6):859–869
Zhong L, Liu L, Liu Y (2010) Natural disaster risk assessment of grain production in Dongting
Lake area, China. Agric Agric Sci Procedia 1:24–32
Li H, Ma L (2007) Detecting and adjusting ordinal and cardinal inconsistencies through a
graphical and optimal approach in AHP models. Comput Oper Res 34(3):780–798
Saaty TL (1994) How to make a decision: the analytic hierarchy process. Interfaces 24:19–43
Tsai C-H, Chen C-W (2011) The establishment of a rapid natural disaster risk assessment model
for the tourism industry. Tour Manag 32:158–171
Tufekci S, Wallace WA (1998) The emerging area of emergency management and engineering.
IEEE Trans Eng Manag 45(2):103–105
Liu W (1999) A new method of rectifying judgment matrix. Syst Eng: Theory Pract 6:30–34 (in
Chinese)
Wei Y-M, Fana Y, Lua C, Tsai H-T (2004) The assessment of vulnerability to natural disasters in
China by using the DEA method. Environ Impact Assess Rev 24:427–439
Xu Z, We C (1999) A consistency improving method in the analytic hierarchy process. Eur J
Oper Res 116:443–449
Chapter 8
A Research on Value of Individual Human
Capital of High-Tech Enterprises Based
on the BP Neural Network Algorithm
Xiu-fen Li and Ping Zhang
Abstract Based on the review of relevant research, this paper combines with the
characteristics of individual human capital of high-tech enterprises and constructs
a comprehensive evaluation index system including four first-level indicators and
sixteen secondary indicators. Using the BP algorithm of neural network model,
based on thirty equipment manufacturing enterprises, this paper also establishes an
evaluation model of the value of individual human capital for the chief engineer.
It holds that although there exist many approaches to evaluating individual human
capital value, many problems still remain unresolved, such as: too broad concept
definition, a lack of understanding of the specificity of evaluation object, a single
evaluation method and inadequacy of practicability or improper model design, and
among others the critical problem in the evaluation of human capital value lies in
the fact that the system cannot be truly applied to practice.
Keywords BP neural network algorithm High-tech enterprises Human capital
8.1 Introduction
Discussion on value of managers has been derived from the practice of early
human activities. The famous ancient Greek philosopher Plato has divided people
(according to their social values) in his Utopia into three classes, deeming it was
determined by the talent of man’s character. Ancient Roman slave-owners and
thinkers Gato and Varro focused on the selection criteria of the chamberlain in On
Agriculture, while in the eighteenth century, the father of ‘‘free economy’’ Adam
Smith has already regarded human as a form of capital. According to Neo-classical
X. Li P. Zhang (&)
Department of Management, Northwest University for Nationalities,
Lanzhou, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 71
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_8,
Springer-Verlag Berlin Heidelberg 2013
72 X. Li and P. Zhang
economics, capital refers specifically to ‘‘useful’’ production items, such as
machines and buildings, and the human is seen as two parts: ordinary labor and
entrepreneur’s talent (Guo 2003). To Marx, the concept of capital is not an urgent
problem to be resolved, rather, it is ‘‘but a ridiculous problem derived from the
cumbersome economics itself.’’ Marx explained the definition of labor (or laboring
ability), namely, ‘‘the human body, living exists in the human body, whenever a
person the production of a use value, the use of physical and mental sum of’’
(Marx 1975).
Modern human capital theory is usually recognized as being from Schultz
(1990), who believe that human capital is an embodiment of productive capacity,
skills and knowledge on the workers and one of the forms of capital as well.
Evaluation of the value of human capital initiates from the American Accounting
scientist Flamholtz (1986). Schultz (1961) and Becker (1962) laid a measure
system of human capital from the methodology. Currently, The currency measure
system in terms of the value of human capital can be divided into two kinds:
methods adopted from the aspects of the cost, such as historical cost, replacement
cost method, cost of investment in education (Zhang 2007); methods adopted from
the aspects of the value measurement, such as random pay, future earnings, dis-
counted future wage adjusted discounted, non-purchased goodwill method, eco-
nomic value measurement method, workers’ rights, producers’ rights, options
method (Yang and Kong 2005).
Particularity of the high-tech enterprises human capital, namely, the high
proportion of knowledge staff, is determinant for the type of capital in exerting
lasting and significant impact on business growth (Dai 2006). To tap, utilize and
stabilize human capital, to improve the composition of human capital aggregates,
thus forming an advantage of human capital in enhancing the core competitiveness
of enterprises, a scientific evaluation must be done to the value of human capital. A
study of the assessment of high-tech human capital is of great significance in terms
of scientifically determining the human capital status, attracting, retaining and
training talents, and exerting advantage of high-tech enterprises human capital.
8.2 Methodology
8.2.1 Conformation of the Index System
Based on the latest research results from domestic and foreign scholars (Fu 2007),
combined with the characteristics of individual human capital of high-tech
enterprises, the present paper holds that the value of human capital in this sort of
enterprises includes four parts: the existing value of human capital, the value of the
contribution of human capital, the ongoing costs and opportunity cost of human
capital. Their relationship is as follows:
8 A Research on Value of Individual Human Capital 73
the total value of Human capital ¼ the existing value of human capital
þ the potential value of human capital
replacement costs of human capital
opportunity cost of human capital
ð8:1Þ
Among them: the existing value of human capital is from the external (or past)
factors of the enterprises, determined by the staff’s physiological conditions,
abilities, education status, work experience and the previous contributions to the
enterprise. The potential value of human capital stands for the present value of the
expected contribution value of employees in the enterprise. The ongoing costs of
human capital stands for the must investment in human capital making contribu-
tion continually, including the expenditure of the staff’s recruitment, training, and
physiology, and so on. The opportunity cost of human capital is the maximum
possible gains brought by other options based on the same cost. Evaluation index
system of high-tech enterprise human capital value is shown in Tables 8.1, 8.2.
Table 8.1 Evaluation index system
The target layer First-level indicators Second-level indicators
Human capital Existing value of human Education
value capital Work experience
Professional knowledge and skills
Age
Gender
State of health
Assessment of the previous contribution
Potential value of human Improvement of the enterprise’s social
capital reputation
Average growth rate of the general output
value
Improvement of the collaboration of
teamwork
Ongoing costs of human Total cost of the recruitment
capital Total cost of investment in training
Vacation, entertainment costs
Wage income
Other income
Opportunity cost of human Average contribution of similar personnel
capital
74 X. Li and P. Zhang
Table 8.2 Quantified description of related indicators
The existing value of human capital
Education: divided into three level, represented respectively by the dummy variable ‘‘-1’’, ‘‘0’’,
‘‘1’’
Work experience: service year of same type work 9 1 ? service year of different types of
work 9 0.5
Professional knowledge and skills: divided into three level, represented respectively by the
dummy variables ‘‘-1’’, ‘‘0’’ and ‘‘1’’
Age: divided into two levels, represented respectively by the dummy variables ‘‘0’’ and ‘‘1’’
Gender: represented respectively by the dummy variables ‘‘0’’, ‘‘1’’
Health status: divided into two levels, represented respectively by the dummy variables ‘‘0’’ and
‘‘1’’
Evaluation of previous contribution: staff’s previous average economic contribution
Potential value of human capital
Improvement of the enterprise’s social reputation: represented by the dummy variable ‘‘0’’ and
‘‘1’’
Average growth rate of the general output value: the contribution of the staff should at least not
below the average level, represented by the enterprise’s average growth rate of total output
value of the latest three years
Improvement of the collaboration of teamwork: divided into three levels, represented respectively
by dummy variables ‘‘-1’’, ‘‘0’’ and ‘‘1’’
Ongoing costs of human capital
Total cost of the recruitment: covers various recruiting costs
Total cost of investment in training: covers various personnel training costs
Vacation, entertainment costs: represented by the gains from assumed normal work
Wage income: covers various wage earnings
Other income: covers various non-salary earnings
Opportunity cost of human capital
Average contribution of similar personnel: It’s a conservative estimate, i.e., the greatest value of
contribution from other personnel, should at least not be less than the average economic
contribution
8.2.2 The Evaluation Approaches of Bp Neural Network
At present, the neural network model has been developed into many types, of
which the most representative is the multi-layer perception neural network. In
1985, Rume jhart proposed the error back-pass learning algorithm, realizing
Minsky’s multi-layer network assumption, which can be shown in (Cheng et al.
2008) (Fig. 8.1).
BP neural network algorithm not only has the input layer nodes and output layer
nodes, but also has one or more hidden layer contacts. For the input signal, it
should first of all pass forward to the hidden layer nodes, after the role of function,
then pass the output signal at the hidden layer nodes to the output layer nodes, and
finally give the output. The S-shaped function is usually chosen as the activation
function at the node.
8 A Research on Value of Individual Human Capital 75
Fig. 8.1 Neural network
algorithm structure
Such as:
1
f ðXÞ ¼ ;
1 þ ex=Q
where Q is the Sigmoid parameter adjusting the form of activation function. The
learning process of the algorithm consists of forward propagation and back prop-
agation. In the forward propagation process, the input information from the input
once the hidden layer, processing layer by layer, and transferring to the output layer.
State of each layer of neurons will influence the state of next layer of neurons. If the
output layer fails to receive the desired output, then it will switch to back propa-
gation, returning the error signal along the connection channel, and by modifying
the right values of all layers of neurons, making the minimum error signal.
Set an arbitrary network containing n nodes, each node is of Sigmoid type. Set
the network has only one output y, the output of any node i Oi, and set N samples
(Xk, Yk) (k = 1, 2, 3,…, N), to a particular input Xk, the network output Yk of
node i for Oik, then the input of node j is:
And the error function is defined as
1X N 2
E¼ yk y1k ;
2 k¼1
where y1k is the actual output of the network, define
1 2 oEk P
Ek ¼ yk yk ; &jk ¼ onetjk ; netjk ¼ Wij Oik ,
i
oEk oEk onetjk oEk
And Ojk ¼ f netjk , so, oW ij
¼ onetjk oWij
¼ onet jk
Oik ¼ &jk Oik
_
When j is the output node, Ojk ¼ y k ,
_
oEk oy k
¼ ðyk y k Þf 0 ðnetjk Þ
_
djk ¼ _
oyk onet jk
If j is not the output node, then:
oEk oEk oOjk oEk 0
djk ¼ ¼ ¼ f ðnetjk Þ
onetjk oOjk onetjk oOjk
76 X. Li and P. Zhang
oEk X oEk onetmk
¼
oOjk m
onetmk oOjk
X oEk o X
¼ Wmi Oik
m
onetmk oOjk i
X oEk X
¼ Wmj
m
onet i
X
¼ dmk Wmj
m
Thus:
8 X
>
> djk ¼ f 0 ðnetjk Þ dmk Wmj
<
m
>
> oE
: k ¼ dmk Oik
oWik
The above algorithm indicates that the BP model converts the I/O problem of a
set of samples to a nonlinear optimization problem, reaching an effective
dimension reduction effect.
In the human capital evaluation index system of high-tech enterprises, indi-
vidual variables and overall enterprise variables, the relative indicators and
absolute indicators are taken into consideration, besides, dummy variables are also
introduced. By adopting BP artificial neural network approach and standardizing
the data, a dimensionless comprehensive result of evaluation can be drawn. In the
process of learning about Network model itself, network parameters should be
constantly adjusted until gaining desired output to meet the requirements.
8.3 Results
The Original empirical data come from the survey questionnaire of ‘‘the techno-
logical innovation power of Lanzhou city in 2006’’. The survey had 148 valid
questionnaires returned, and the choosing of industry was based on the criteria for
the classification of ‘‘high-tech industry’’ from the classification categories of
high-tech industry released in 2002 by National Bureau of Statistics. To make the
sample data strong in comparability, the present paper selected 30 equipment
manufacturing enterprises, established for the chief engineer an evaluation model
of individual human capital value.
Normalize the 30 groups of sample observations the data between 0 and 1, 15 of
which as model building data, and the rest 15 as model test data. The input
parameters of BP neural network model are: input nodes being 17, the intermediate
layer being 1, the middle layer nodes being 12, the minimum training rate being
0.1, the dynamic parameters being 0.6, the Sigmoid parameter being 0.9,
8 A Research on Value of Individual Human Capital 77
Table 8.3 Final fitting state of model
The fitting of the modeling samples The fitting of the testing samples
Number Expert ratings Model ratings Error Expert ratings Model ratings Error
1 0.9091 0.9100 0.0009 0.9697 0.9698 0.0001
2 0.9697 0.9688 0.0009 0.9194 0.9190 0.0004
3 0.9394 0.9399 0.0005 0.9683 0.9693 0.001
4 0.9495 0.9499 0.0004 0.9545 0.9544 0.0001
5 0.9293 0.9300 0.0007 0.9191 0.9159 0.0032
6 0.9394 0.9400 0.0006 0.9094 0.9084 0.0010
7 0.9697 0.9684 0.0013 0.9698 0.9696 0.0002
8 0.9394 0.9400 0.0006 0.9694 0.9698 0.0004
9 0.9495 0.9500 0.0005 0.9693 0.9697 0.0004
10 0.9697 0.9683 0.0014 0.9694 0.9676 0.0018
11 0.9293 0.9301 0.0008 0.9672 0.9675 0.0003
12 0.899 0.9012 0.0022 0.9194 0.9197 0.0003
13 0.9394 0.9401 0.0007 0.9693 0.9671 0.0022
14 0.9697 0.9680 0.0017 0.9699 0.9684 0.0015
15 0.9596 0.9600 0.0004 0.9694 0.9698 0.0004
permissible error being 0.001, and the maximum number of iterations being 1000.
After 280 iterations, the error of neural network comes close to zero, the fitting
residuals of the model are about 0.0001677, and thus the training result is desir-
able. Weights of each node in the Model are shown in Table 8.2.
The final fitting state of the model is shown in Table 8.3. The average error of
the 15 groups of modeling samples is 0.000979, while the average error of the 15
groups of testing samples is 0.000887. The learning result of the whole model is
desirable, and also fits the testing samples well. Therefore, the above model can be
used for the chief engineer in other similar enterprises to conduct the evaluation on
the individual human capital value.
8.4 Conclusion
At present, there are many approaches in terms of the study of human capital
value, mainly for the following three reasons:
First, the definition of the concept of human capital value is overly broad. It is
commonly believed that human capital value is derived from labor and creativity,
but in reality the embodiment of these factors is different and specific. As is often
the case, the newly recruited talents with high education and high titles have not
yet created new value, but they are still seen as part of increase of the stock of the
enterprise’ human capital. In addition, there is a relatively high discrepancy
between the potential value of human capital (or expected value) and the actual
value.
78 X. Li and P. Zhang
Second, the object of the evaluation of human capital value has its particularity.
While some argue that general human capital evaluation criteria can be estab-
lished, but due to the variety of evaluation object, the conclusions are quite dif-
ferent. From the point of view of scientific research, the conclusion simply from
induction will cause the ‘‘problem of induction’’ (Liang 2007). Owing to the fact
that the testing data fail to meet the demands of ‘‘random sampling assumes’’ or
‘‘law of large numbers’’, and ‘‘the central limit theorem’’, those research trying to
draw the value of general human capital can mostly lead to a generalization
problem (Wooldridge 2003). But the systematic analysis of non-equilibrium ran-
dom variable regression conducted by Clive Grange (1974) shows that, regardless
of the causal relationship between the random variables, the more unstable the
variables are, the better the fitting of the regression equation will be.
Third, part of the evaluation approaches to the human capital value is too
simple, and its practicability is not enough or the model design is unreasonable,
thus it can not be widely adopted in practice because of many problems, such as
inadequate of valid indicators, lack of uniqueness of the overall model set,
insufficiency of the statistical test of correlative relations, and so forth (Li 2008).
In view of the above problems, this paper holds that the value of human capital
evaluation should first define the scope of the study, analyze the characteristics of
the evaluation object in order to establish an evaluation index system and construct
an evaluation model.
Acknowledgments This paper is supported by the Humanities and Social Science Fund of the
Ministry of Education (09XJC790014; 11XJC630008).
References
Becker GS (1962) Investment in human capital: a theoretical analysis. J Polit Econ 70(5):9–49
Cheng L, Liu ZW, Ai XY (2008) An empirical analysis on the influencing factors of human
resource value. Acc Res 1:26–32
Dai CL (2006) Equity allocation and measurement of human capital for hi-tech enterprises.
J Wuhan Univ Technol (Inf Manag Eng) 28(12), pp 150–154
Flamholtz EG (1986) Human resource accounting. Shanghai Translation Publishing Company,
Shanghai, pp 23–27
Fu X (2007) Management contribution measurement on human capital of entrepreneurs.
Collected Essays Financ Econ 1:82–88
Granger C, Newbold P (1974) Spurious regressions in econometrics. J Econom 2:111–120
Guo XG (2003) The history of management theory. China Economic Publishing House Press,
Beijing, pp 9–10
Li ZN (2008) The specification of population regression models in econometric application
studies. Econ Res J 8:136–144
Liang ZY (2007) Study on the measurement of human capital value. Stat Decis 5:166–168
Marx K (1975) Capital. The Commercial Press, Beijing, p 34
Schultz TW (1961) Investment in human capital. Am Econ Rev 51(1):1–17
Schultz TW (1990) Human capital investment and urban competitiveness. Beijing Economy
Institute Press, Beijing, p 113
8 A Research on Value of Individual Human Capital 79
Wooldridge JM (2003) Introductory econometrics. South Western Press, pp 5–7
Yang L, Kong XX (2005) The analysis of the abroad measurement model of human capital value.
J Shandong Inst Light Ind (Nat Sci edn) 19(2), pp. 57–60, 66
Zhang QT (2007) Calculating model and means to divide the profit of human capital property
value based on current profit. Acc Res 1:23–29
Chapter 9
A Stochastic Programming Model
for Evaluating Real Options in Wind
Power Investment Projects
Han Qin and L. K. Chu
Abstract This study is concerned with the evaluation of wind power projects
under the Clean Development Mechanism (CDM), not only for the purpose of
CDM verification, but also for the financing of the project. A real options model is
developed in this paper to evaluate the investment decisions on wind power
project. The model obtains the real value of the project and determines the optimal
time to invest wind power project. Stochastic programming is employed to eval-
uate the real options model, and a scenario tree, generated by path-based sampling
method and LHS discretization, is constructed to approximate the original sto-
chastic program.
Keywords Clean Development Mechanism (CDM) Stochastic programming
Real options Wind power investment
9.1 Introduction
Environmental issues have received global attention and carbon trading is con-
sidered to be a primary solution to reduce greenhouse gas (GHG) emissions. As a
developing country, China is mainly involved in the Clean Development Mech-
anism (CDM), which is one of the instruments created by the Kyoto Protocol to
facilitate carbon trading, and China’s corporations related to environmental
investments are facing the decision problems under the uncertainty of carbon
price.
According to the statistics of international wind energy committee, over 80 %
of carbon emissions come from the energy industry, of which 40 % is due to the
H. Qin (&) L. K. Chu
Department of Industrial and Manufacturing Systems and Engineering, The University of
Hong Kong, Hong Kong, Hong Kong SAR
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 81
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_9,
Ó Springer-Verlag Berlin Heidelberg 2013
82 H. Qin and L. K. Chu
generation of electricity. Wind energy holds huge potential in emission reduction
since it produces no greenhouse gas. Despite the fact that the production of a
windmill itself consumes a significant amount of energy (about 3–6 months of
power generated by the plant) and hence causing carbon emission, the average life
length of wind turbine can last 20 years or so. Since the implementation of CDM,
wind power in China gets substantial development. In 2008, there were 314 wind
power projects in China, ranking first in the world, amidst the world total of 647.
However, recent wind power project applications have suffered a major setback
due to policy changes of the CDM Executive Board (CDM EB). In 2010, up to 14
wind power projects from China were rejected by the CDM EB. For a wind power
project to be approved, the applicant has to show that the project can effectively
reduce carbon emission and that the project will be profitable—the argument of
‘additionality’. It is mainly the latter requirement that rule out projects that need
funding from CDM. Also, a project can qualify for CDM funding only if no other
private investment has been sought and secured. The CDM EB will not approve an
application if the applicant is unable to produce a valid revenue/profit model for
evaluation by the CDM EB. Therefore, it is interesting to study the real value of
wind power projects under CDM mechanism, not only for the purpose of CDM
verification, but also for the financing of the project.
In fact, the revenue of a CDM wind power project comes from two parts:
income of electricity and earning from Certified Emission Reductions (CERs). The
uncertainty of carbon price makes the evaluation of wind power projects more
difficult. Low net present value (NPV) means little reward for investors. However,
if the strategic value and options value could be considered also, the real value
would be different from traditional NPV.
Traditionally, investment decisions are evaluated by means of the discounted
cash flow method (DCF) in which the NPV of an investment is often used.
However, DCF often tends to underestimate the value of high technology projects
which may actually be viable, mainly because of the conservatively high risk rate
assumed at the beginning of the investment. Moreover, DCF is not an adequate
methodology as the situation requiring strategic flexibility, where investment
decisions can, at some cost, be deferred to some proper future date. Another
drawback of DCF is that it cannot deal with the possible varying cost of GHG
emission credits.
In reality, corporations have the option to defer investment to some future date.
One tool that can prove beneficial in this type of investment environment is the use
of real options. This approach treats options at different stages as part of its overall
decision making process. A real options model is developed in this paper to
evaluate the investment decisions on wind power projects under the CDM
mechanism. The model determines the optimal time to invest wind power project
using real options. Stochastic programming is employed to evaluate the real
options embedded in the wind power project investment.
9 A Stochastic Programming Model 83
9.2 Literature Review
9.2.1 Environmental Investment Methods
As organisations are becoming increasingly environmentally conscious, the related
investment decisions have become an important consideration for management.
Some studies have discussed how environmental investments could benefit
organisations (Nehrt 1996; Porter and van der Linde 1995; Bonifant et al. 1995).
Nevertheless, many studies seem to investigate the issues in an empirical or
conceptual way, and do not incorporate any quantitative analysis. On the other
hand, some quantitative techniques have been proposed, such as stochastic
dynamic optimisation (Birge and Rosa 1996), mixed integer programming
(Mondschein and Schilkrut 1997), activity based costing (Presley and Sarkis
1994), and data envelopment analysis (Sarkis 1999). Although these models are
advanced and mathematically complex, the most popular technique used by
organisations still is the DCF method, which is essentially a cost-benefit analysis
based on the time value of money. The DCF method, which is mainly based on the
evaluation NPV, is simple and practical. However, DCF largely has ignored the
option to defer an investment. It thus seems that the dynamic option value
embedded in the options approach, which can be very useful in analysing some
investments, has not been adequately explored in assessing the true value of a
technology project.
9.2.2 Real Options Analysis
The fundamental hypothesis of the traditional DCF method is that future cash
flows are static and certain, and management does not need to rectify an invest-
ment strategy in face of changing circumstances (Myers 1977). Nevertheless, this
is inconsistent with the real-life situations. In practice, corporations often face
many uncertainties and risks, and the management has to deal address such
uncertainties proactively. This means that the issues of operating flexibility time
strategy have to be considered (Donaldson et al. 1983).
Other than DCF, real options analysis can be used to deal with investment options
and managerial flexibility. The real options method has its roots in financial options,
and it is the application and development of financial options in the field of real assets
investment. It is generally agreed that the pioneering work in real options was due to
Myers (1977). The work suggested that, although corporation investments do not
possess of forms like contract as financial options do, investments under a situation
of high uncertainty still have characteristics which are similar to financial options.
84 H. Qin and L. K. Chu
Therefore, the options pricing method can be used to evaluate the investments.
Subsequently, Myers and Turnbull (1977) suggested a ‘Growth Options’ for cor-
poration investment opportunity. Kester (1984) further developed Myers’ research,
and argued that even a project with an unfavourable NPV could also have investment
value if the manager had the flexibility to defer investment until a later date.
After three decades of development, the theory of real options has become an
important branch and a popular research topic. By examining the different
managerial flexibility embedded in real options were able to divide real options
into seven categories: Option to Defer, Option to Alter Operating Scale, Option to
Abandon, Option to Switch, and so on. Other types of options have also been
proposed and studied by researchers (O’Brien et al. 2003; Sing 2002).
The real options theory has been applied to investment problems in many
different fields including biotechnology, natural resources, research and develop-
ment, securities evaluation, corporation strategy, technology and so on (Miller and
Park 2002). It has also been applied to a gas company in Britain, leading to the
conclusion that certain projects are not economically feasible unless they are
permitted to have a faster price rise (Sarkis and Tamarkin 2005). Qin and Chu
(2011) developed a real options model to determine the optimal time to invest in
Carbon Capture and Storage (CCS) project in China, and analysed the behaviour
of China green technology business. It took the emission credits cost into con-
sideration and verified the real options analysis by evaluating environmental
related investments.
9.2.3 Stochastic Programming
Stochastic programming is a popular modelling framework for decision making
problems under uncertainty in a variety of disciplines. Applications can be found
in fields such as portfolio optimisation (Cariño et al. 1998; Cariño and Ziemba
1998) and power generation (Dentcheva 1998). Benefited from the well developed
optimisation techniques in mathematical programming, stochastic programming is
insensitive to the stochastic processes and constraints, and thus can easily be
applied to complex decision making problems including valuating compound real
options with path dependency. de Neufville (2004) employed stochastic mixed
integer programming to manage path dependency and interdependency of com-
pound real options embedded in a water resources planning problem.
In this paper, stochastic programming is chosen for solving the wind power
investment problem for the reason that it can easily model the complexity of the
portfolio of the real options embedded regardless of the stochastic processes of
uncertain factors.
9 A Stochastic Programming Model 85
9.3 Real Options Model
A license to invest in a wind power plant is a real option, which means that the
investor has the right, but not the obligation, to exercise the payment of investment
to obtain the revenue from the project. Facing the uncertain future and the irre-
versibility of decisions, the investor would value the opportunity to wait and get
more information about future conditions. The objective of this real options model
is to obtain an optimal timing of a wind power investment decision problem under
carbon price uncertainty.
9.3.1 The Deterministic Model
Different from the traditional NPV method, the real options analysis considers not
only whether to invest, but also when to invest. The opportunity to invest in a wind
power plant could be taken as an option, whose value is given by the following
equation:
ROV ¼ maxðNPVo Þ NPVs ð9:1Þ
where NPVs is the static NPV of the wind power project investment without
considering any management flexibility, i.e., investment is made at stage t = 1.
NPVo is the investment with timing options, i.e., the investor could delay the
investment to the optimal time. Let T be the numbers of the planning periods,
normally it is around 20 years as the franchise periods of wind power project in
China is 25 years, and it takes several years to construct the power plant. The
decision variable is denoted as Xt, where Xt is a 0–1 variable. If the investor
exercises the investment at stage t, Xt = 1. Otherwise Xt = 0. t 2 f1; . . .; T g.
1; to invest
Xt ¼ t 2 f1; . . .; T g
0; not to invest
Under the CDM mechanism, the revenues due to the investment in wind power
come from two sources. The first is derived from the income of the electricity
generated and sold (RE). RE is deterministic because the feed-in tariff in China is
determined in the bidding. The other revenue is made due to the income of CERs
(RC), which is uncertain because the carbon price varies from time to time. The
revenue of investment R (per kWh) in wind power is the sum of these two parts:
R = RE ? RC. In the proposed real options model, we will focus on the income
of CERs.
If the wind power plant produces Et (tonne) CERs per unit electricity output every
year, and all the CERs are tradable in the carbon trading market. Then, the income
from CERs sales is At * Et. Denote by It the initial investment, and the construction
86 H. Qin and L. K. Chu
will take k years, Qt as the annual generating capacity (kWh) of the wind power
plant, Ct as the unit cost of generating output, NPVo can be expressed as follow:
!
XT
1t XT
pt þ At Et ct
t t
NPVo ¼ X t1 þ Q ð9:2Þ
t¼1 1 þ cf K¼tþk ð1 þ cÞk1
where
It Initial investment
Qt Annual generating capacity
Ct Unit cost of generating output
Et Annual CERs emission
k Construction period
c Discount rate
cf Risk-free rate.
To obtain the real options value, we need to maximise the NPVo. The decision
problem can be described as:
max NPVo ð9:3aÞ
s.t.
X
T
X t ¼ 1; 8t 2 f1; . . .; T g ð9:3bÞ
t¼1
X t ¼ 0 or 1; 8t 2 f1; . . .; T g ð9:3cÞ
9.3.2 The Stochastic Model
The revenue of investment R (per kWh) in wind power is uncertain because of the
uncertainty of the revenue from CERs (RC), which can change considerably
depending on factors such as environment policy legislation, cost of alternative
fuels, product market demand and so on. Assume the price of CERs At shifts
stochastically over time represented by the geometric Brownian motions:
dAt ¼ lAt dt þ rAt dWt ð9:4Þ
where
At Price of CERs at the end of time period t
Wt The standard Brownian motions process
l Carbon price drift
r Carbon price volatility.
9 A Stochastic Programming Model 87
(T, 1) P1
(t, 1) (T-1, 1)
(2, 1) (T, 2) P2
( A11 ,1)
( Ant , Pnt )
(1, 1) (T, nT)
(2, n2) t T-1
(t, n ) (T-1, n )
(T,nT+1)
(2, N2)
(t, Nt) (T-1, NT-1)
(T, NT) PS
Fig. 9.1 Scenario tree
The stochastic process is assumed to be exogenous and independent of the
decision variables in the deterministic model. In addition, the process is considered
to follow some discrete distributions. A scenario tree is used to represent this type
of discrete stochastic process.
Figure 9.1 shows a scenario tree representing the evolution paths of the sto-
chastic variable At. The nodes in the tree are the states of At. A decision is made at
each node so as to choose an option to exercise. Over the planning horizon, each
particular scenario is reflected by the path from the root node to the particular
ending node. At stage t, nodes are indexed as (t, n), the number of nodes is Nt,
where 2 ft; . . .; T g and n 2 ft; . . .; N t g. Carbon price on node (t, n) is expressed as
Ptn , and the unconditional probability is expressed as Ptn . Thus, a scenario can be
expressed as the vector of ½1; n2 ; . . .; nt ; . . .nT , with probability PPTn . Some of the
variables and parameters used in the deterministic model should be modified as
follows:
1; to exercise
Xnt ¼ t 2 f1; . . .; T g; n 2 1; . . .; N T
0; not to exercise
Atn The price of CERs at the end of time period t node n; t 2 f1; . . .; T g; n 2 f1; . . .; N t g
PTn Unconditional probability of stochastic event Atn ; t 2 f1; . . .; T g; n 2 f1; . . .; N t g
88 H. Qin and L. K. Chu
Based on the notations above, NPVo can be modified as follows:
!
X
T
I t X
T
pt þ Atn Et ct
NPVo ¼ Ptn Xnt t1 þ Q
t
ð9:5Þ
t¼1 1 þ cf k¼tþk ð1 þ cÞk1
and the decision problem can be expressed as:
max NPVo ð9:6aÞ
s.t.
X
T
Xnt ¼ 1; 8t 2 f1; . . .; T g; n 2 f1; . . .; N t g ð9:6bÞ
t¼1
Xnt ¼ 0 or 1; 8t 2 f1; . . .; T g; n 2 f1; . . .; N t g ð9:6cÞ
9.3.3 Scenario Generation
Given the assumption that the prices of CERs Atn shift stochastically over time
represented by the geometric Brownian motions, which is expressed in Eq. (9.4),
the sample paths of Atn can be obtained. Figure 9.2 shows the result of 50 sample
paths with parameter values in Table 9.1.
In general, continuous distributions cannot be incorporated directly in sto-
chastic programs because of the infinite number of scenarios that need to be
examined. For this reason, discrete distributions are generally used to approximate
the continuous distribution of stochastic parameters. The quality of scenario tree,
which consists of such discretized distributions, is vital to the accuracy of the final
solutions.
50 Sample Path
140
CERs price (yuan)
120
100
80
60
40
0 2 4 6 8 10 12 14 16 18 20
Stage t
Fig. 9.2 Sample paths of CERs price
9 A Stochastic Programming Model 89
Table 9.1 Value of parameters
Notation Parameter Unit Value Data source
It Initial investment Yuan 2141129600 Wind Power station
Qt Annual generating capacity kWh 5919000 In Huitengshile
Ct Unit cost Yuan/kWh 0.1976 Inner Mongolia
k Construction length Year 2
T Operation period Year 20
Et Annual CERs T 51429.6
A11 Price of CERs Yuan 50 Estimated
l Carbon price drift 0.6
r Carbon price volatility 0.18
c Discount rate 0.15
cf Risk free rate 0.05
Various methods have been proposed to generate scenario tree. These can be
classified into external sampling and internal sampling methods. External sam-
pling methods consist of optimal discretization sampling, conditional sampling,
moment-matching, and path-based sampling. Internal sampling is actually a sto-
chastic program solving algorithm with scenarios to be sampled during the solu-
tion procedure. Commonly used internal sampling methods are stochastic
decomposition, stochastic quasi-gradient, EVPI-based sequential important sam-
pling (Chu et al. 2005).
Equation (9.4) describes the price of CERs Atn as a known stochastic process. For
the sake of simplicity, path-based sampling is applied to generate the scenario tree,
and Latin Hypercube Sampling (LHS) is employed in this study to discretize Atn .
9.4 Case Study
In order to investigate the application of stochastic programming in the evaluation
of real options, a case study is carried out. Data are collected from a wind power
station in Huitengshile Inner Mongolia China, which began its construction in
2005 and went into operation in 2008 (Zhu 2011). The data is shown in Table 9.1.
We sampled 500 scenarios by LHS. The NPV0 is found to be 4.278 million
Yuan at year 11, whereas the static NPV is 2.045 million Yuan. The difference
between them is the value of real options embedded in the wind power project
investment. If the strategic value and options value are not considered, the real
value would be underestimated. However, the result also shows that investor tends
to delay the exercise of wind power investment until the price of CERs increases to
certain level.
90 H. Qin and L. K. Chu
9.5 Conclusion
Stochastic programming is a powerful technique for solving decision making
problems under uncertainty. In this study, a real options model is developed to
evaluate the investment decisions on wind power projects under the CDM. A
scenario tree, generated by path-based sampling method and LHS discretization, is
constructed to approximate the original stochastic program.
This study has considered only the option to delay in the investment of wind
power projects. However, there are other important types of options to consider
during the operation of wind power projects, especially under the CDM and
China’s franchise clause for power station operation. The proposed model can be
extended to incorporate abandon options. In addition, multi-players can also be
considered to study the optimal bidding price of electricity.
References
Birge JR, Rosa CH (1996) Incorporating investment uncertainty into green-house policy models.
Energy J 17:79–90
Bonifant BC, Arnold MB, Long FJ (1995) Gaining competitive advantage through environmental
investments. Bus Horiz 38:37–47
Cariño DR, Ziemba WT (1998) Formulation of the Russell-Yasuda Kasai financial planning
model. Oper Res 46:433–449
Cariño DR, Myers DH, Ziemba WT (1998) Concepts, technical issues, and uses of the Russell-
Yasuda Kasai financial planning model. Oper Res 46:450–462
Chu LK, Xu YH, Mak KL (2005) Evaluation of interacting real options in RFID technology
investment: a multistage stochastic integer programming model. Presented at the 14th
international conference on management of technology, Austria
de Neufville TWR (2004) Building real options into physical systems with stochastic mixed
integer programming. Presented at the 8th international conference on real options, Montreal
Dentcheva WRD (1998) Optimal power generation under uncertainty via stochastic program-
ming. In: Marti K, Kail P (eds) Stochastic programming methods and technical applications.
Springer, Berlin, pp 22–56
Donaldson G, Lorsch J (1983) Decision making at the top: the shaping of strategic direction.
Basic Books, New York
Kester WC (1984) Today’s options for tomorrow’s growth. Harv Bus Rev 62(2):153–160
Miller LT, Park CS (2002) Decision making under uncertainty-real options to the rescue? Eng
Econ 47:105
Mondschein SV, Schilkrut A (1997) Optimal investment policies for pollution control in the
copper industry. Interfaces 27:69–87
Myers SC (1977) Determinants of corporate borrowing. J Finan Econ 5(2):147–176
Myers SC, Turnbull SM (1977) Capital budgeting, and the capital asset pricing model: good news
and bad news. J Finan 32:321–333
Nehrt C (1996) Timing and intensity effects of environmental investments. Strateg Manag J
17:535–547
O’Brien JP, Folta TB, Douglas RJ, Timothy B (2003) A real options perspective on entrepreneurial
entry in the face of uncertainty. Manag Decis Econ 24:515–544
9 A Stochastic Programming Model 91
Porter ME, van der Linde C (1995) Green and competitive: ending the stalemate. Harv Bus Rev
73:120–134
Presley A, Sarkis J (1994) An activity based strategic justification mehodology for ecm
technology. J Environ Conscious Des Manuf 3:5–17
Qin H, Chu LK (2011) Real options model for valuating China greentech investments. In: 7th
international conference on digital enterprise technology, Greece, pp 134–141
Sarkis J (1999) A methodological framework for evaluating environmentally conscious
manufacturing programs. Comput Ind Eng 36:793–810
Sarkis J, Tamarkin M (2005) Real options analysis for ‘‘green trading’’: the case of greenhouse
gases. Eng Econ 50:273–294
Sing TF (2002) Time to build options in construction processes. Constr Manag Econ 20:119–131
Zhu L (2011) Energy investment modeling and its applications under the background of energy
security. Doctor, Management Science and Engineering, University of Science and
Technology of China
Chapter 10
Action Recognition Based on Hierarchical
Model
Yang-yang Wang, Yang Liu and Jin Xu
Abstract The feature representation of human actions is one of the important
factors which influence the recognition accuracy of actions. Usually the recogni-
tion accuracy is higher, when the feature simultaneously includes both appearance
and motion information. However the dimensions of the feature space is high, and
this leads to high computational cost. To overcome this problem, we propose a
hierarchical model for action recognition. In the first hierarchy, we adopt box
features to divide the actions into two classes, according to whether or not legs are
all almost stayed in a static place. In the second hierarchy, we construct different
structure of motion feature descriptors to represent different kinds of actions, and
use nearest neighbor classifier to obtain the final classification results. Experiments
on the Weizmann dataset demonstrate the effectiveness of the proposed method.
Keywords Action recognition Box feature Hierarchical model Motion
feature
10.1 Introduction
Human action recognition is widely applied in the fields of surveillance, image and
video retrieval and so on. Recently many researches have been done for the
recognition, however, it still is a challenging problem due to diverse variations of
human body motions, for example, occlusion, scale, individual appearance and
individual motion fashion. To achieve good recognition performance, a good
representation with rich appearance and motion information is of vital. Usually if a
feature includes not only still appearance but also dynamic motion information, the
Y. Wang (&) Y. Liu J. Xu
College of Automation, Shenyang Aerospace University, Shenyang,
People’s Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 93
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_10,
Springer-Verlag Berlin Heidelberg 2013
94 Y. Wang et al.
feature dimension is high, and it is more discriminative to classify the actions. But
this leads to high computational costs. To overcome this problem, we propose a
hierarchical model for action recognition using multi-features, and show its per-
formance compared to other exist algorithms.
The rest of this paper is organized as follows. Section 10.2 describes related
work. Section 10.3 elaborates the details of our model. Section 10.4 reports the
experiment results. Conclusion is given in Sect. 10.5.
10.2 Related Work
Many works have been proposed for action recognition, and some reviews
(Moeslund et al. 2006; Poppe 2010; Ji and Liu 2010) have provided a detailed
description of the action recognition framework. Here, we only focus on the action
recognition that has related to our work.
Features based on shape or appearance is traditional representation in videos
analysis. This kind of representation is usually related to global properties of
human. First human is separated from the background, which is called box. And
then all kinds of descriptors based on silhouette or edges are described. Deng et al.
(2010) compute the silhouettes of the human body, and extracts points with equal
interval along the silhouette. Through constructing the 3D DAISY descriptor of
each point, the space–time shape of an action is obtained. Bobick and Davis (2001)
also make use of the silhouette, but they employ motion energy images (MEI) and
motion history image (MHI) to describe the actions. Ji and Liu (2009) present
contour shaper feature to describe multi-view silhouette images. In Nater et al.
(2010) a signed distance transform is applied to the box with fixed size. In
Weinland and Boyer (2008) silhouette templates are used to match. But the rec-
ognition rate may be reflected by the accurate degree of the localization of the box,
and the effect of background subtraction.
Another popular feature representation is about motion information. Messing
et al. (2009) use the velocity histories of tracked key points to recognize the
actions. Laptev et al. (2008) first use HOF descriptor to represent local motion
information. Recently the combination of shape and motion features has received
more attention. Lin et al. (2009) construct a shape-motion descriptor, and model a
shape-motion prototype tree using hierarchical k-means clustering. Klaser et al.
(2008) make use of HOG3D descriptor to combine motion and shape information
simultaneously. Although these descriptors include more rich information, and the
recognition accuracy is improved, the dimensions of the descriptors are bigger
than those of the descriptors with single shape or motion information. A trade-off
between computation complexity and accuracy should be considered. Therefore,
we propose a hierarchical model, in the first stage, a coarse classification is made
according to box features, and then different descriptors are designed to different
class, this can reduce the dim of the feature vectors.
10 Action Recognition Based on Hierarchical Model 95
10.3 Proposed Method
There are five main steps to our approach: first, preprocessing video to achieve the
human box; second, extracting global box features; third, preliminary dividing the
actions into two different classes based on box features; forth, computing
respective motion features for the actions which belong to different classes; and
fifth, recognizing actions using nearest neighbor classifier and voting algorithm. In
the following, we describe each step in turn.
10.3.1 Preprocess
We start from a video of k frames which are described in a RGB space. For feature
extraction and recognition, we use background subtraction and filtering to segment
the foreground object, which is the box of a person in each frame of an action
sequence. In our paper, the motion direction of the same action is all aligned to the
same direction, e.g. for the action ‘running’, the direction is appointed from left to
right, Fig. 10.1b, c. Furthermore, the resulting silhouettes are converted to a binary
representation. Examples of the box are illustrated in Fig. 10.1, and the repre-
sentation of the boxes of a video S is
S ¼ fB1 ; B2 ; . . .. . .; Bk g: ð10:1Þ
10.3.2 Coarse Classification Based on Box Feature
For each box Bi (i = 1,…, k), a three-dimensional box feature vector is computed,
i.e. Fbi = [cxi, cyi, ri], here cxi, cyi, ri denote the coordinates of the center, aspect
ratio of the box. According to the variance of coordinates and aspect ratio in an
image sequences, we can easily divided the simple actions into two classes. Class 1
includes the actions which the motions mainly focus on the leg movements, such
as jogging, running, walking, skipping and so on. And the actions which legs are
all almost stayed in a static place are all belong to the Class 2, such as bending,
waving.
10.3.3 Local Weighted-Motion Feature
Having classified the actions into two classes, we can construct different weighted-
motion feature according to their particular characteristic.
96 Y. Wang et al.
Fig. 10.1 Example of the
box: a a frame after
background subtraction,
b extracting the box,
c aligned the motion direction
of the box
(1) Box normalization. The size and the position of the box in each frame is
different, after getting the box feature Fbi, we normalized the boxes to equal
size (80 9 80 in our case) while maintaining aspect ratio.
(2) The division of subregions. When a person is in a state of motion in a video
sequence, his body is continually varying in time series. For the actions which
belong to Class 1, the movements of its upper limb and lower limb are more
important. Whereas for the actions in Class 2, e.g. running, the variances in
subregions of legs are more important. In order to exactly describe the
10 Action Recognition Based on Hierarchical Model 97
Fig. 10.2 The division of
subregions for Class 1
variation, the different division of subregions is adopted to different action
classes. First, for all kinds of actions, each normalized box is divided into
n square subregions. And then for Class 1, on the basis of the division men-
tioned above, for the first and last row of the box, we use a further division, as
shown in Fig. 10.2. And for Class 2, for the second and third row of the box,
each subregion is divided into 4 square grids, as shown in Fig. 10.3.
(3) Weighted-motion feature representation. Optical flow can be made to describe
the human motion. We calculate optical flow using Horn and Schunck algo-
rithm (Horn and Schunck 1981). To each pixel in each subregion, the velocity
of optical flow V (x, y, Vx, Vy) is computed, where (x, y) represents the image
location of V, and (Vx, Vy) are the x and y components of the velocity
V. Besides, for reducing the influences of noise dynamic backgrounds, we use
a Gauss filter to smooth the subregions before calculating optical flow.
According to the (Vx, Vy), the magnitude and orientation of the V is computed
Fig. 10.3 The division of
subregions for Class 2
98 Y. Wang et al.
M Vx ; V y ¼ Vx ; Vy ; ð10:2Þ
O Vx ; Vy ¼ arctan Vy Vx : ð10:3Þ
O(Vx, Vy) is divided into eight equal sectors in polar coordinates, and the M(Vx, Vy)
can be regarded as the weights. The weighted-motion histogram’s x-axis reflects
the eight orientations. The histogram’s y-axis shows the contribution within each
orientation. And for each subregion we achieve a histogram.
Further, the subregions with grids are more important than others, in order to
precisely represent actions, for each grid in a subregion a single motion histogram
is computed. Therefore, the number of histogram is 8 ? 8 9 4 = 40.
10.3.4 Final Classification Based on Weighted-Motion
Feature
Having got the features of each frame of an image sequence, we respectively
classify two different classes of actions. Although the features of respective class
are different, the classifier is the same.
NNC (Nearest Neighbor Classifier) is used to recognize the action categories.
For the fth frame of an unknown video Su, we get a distance vector Df = [d1,
d2,…, dq], here dj (j = 1,…, q) denotes the distance between the features of the fth
frame in Su and the features of each frame in training set, q denotes the number of
all of frames in the training set. The minimum distance in Df is picked up, and the
label of the corresponding training frame is assigned to the fth frame of Su. And
finally each frame of Su gets a label; the final action label of Su is determined by
voting algorithm.
10.4 Experiments
To evaluate the performance of our algorithm, the popular benchmark dataset—
Weizmann dataset is used in action recognition. The Weizmann dataset (Blank
et al. 2005) consists of 90 videos of nine actors performing ten different actions.
These actions are walking, running, jumping, siding, and bending, one-hand
waving, two-hands waving, jacking, jumping in place and skipping. Evaluations
were done with a leave-one-out cross-validation.
Our recognition result is shown in Table 10.1 and compared with Lin’ method
(Lin et al. 2009) and Scovanner’s method (Scovanner et al. 2007). In addition, the
confusion matrix is shown in Fig. 10.4. The data on the diagonal line present the
numbers of the correct identification, and other data in the graph are the numbers
of the misclassification. The average recognition rate of our approach is 96.7 %.
10 Action Recognition Based on Hierarchical Model 99
Table 10.1 Results using Descriptor Average precision (%)
different features on
Weizmann dataset Shape-motion tree (Lin et al. 2009) 94.44
3D SIFT (Scovanner et al. 2007) 82.6
Ours 96.7
Fig. 10.4 Confusion matrix
for Weizmann dataset
Besides we exchange the motion feature structure of the two classes. When the
actions in Class 2 use the feature structure of Class 1, as shown in Fig. 10.5, the
average accuracy is 93.4 %. When the actions in Class 1 use the feature structure
of Class 2, the influence of recognition rate is very small. The results prove that
our method is effective.
Fig. 10.5 Confusion matrix
for Class 2. The form of the
feature representation is used
Class 1’s
100 Y. Wang et al.
10.5 Conclusion
We have presented a novel hierarchical feature representation. First, the actions
are divided into two classes based on the box feature. And then for the different
classes of actions, the different motion feature structure is designed. This method is
clearly useful for action recognition. However background segmentation and the
alignment of boxes are required in the method. One future research direction
would be to improve the robustness of feature descriptors in scenes where boxes
can not be obtained reliably.
References
Blank M, Gorelick L, Shechtman E, Irani M, Basri R (2005) In: Proceedings of 10th IEEE
conference on computer vision, ICCV’05, Beijing, China, pp 1395–1402
Bobick AF, Davis JW (2001) The recognition of human movement using temporal templates.
IEEE Trans Pattern Anal Mach Intell 23(3):257–267
Deng C, Cao X, Liu H, Chen J (2010) A global spatio-temporal representation for action
recognition. In: Proceedings of 20th conference on pattern recognition, ICPR’10, Istanbul,
Turkey, pp 1816–1819
Horn BKP, Schunck BG (1981) Determining optical flow. Artif Intell 17(1–3):185–203
Ji X, Liu H (2009) View-invariant human action recognition using exemplar-based hidden
markov models. In: Proceedings of 2nd conference on intelligent robotics and applications,
ICIRA’09, Singapore, pp 78–89
Ji X, Liu H (2010) Advances in view-invariant human motion analysis: a review. IEEE Trans
Syst Man Cybern Part C Appl Rev 40(1):13–24
Klaser A, Marszalek M, Schmid C (2008) A spatio-temporal descriptor based on 3D-gradients. In:
Proceedings of 19th British machine vision conference, BMVC’08, Leeds, UK, pp 995–1004
Laptev I, Marszalek M, Schmid C, Rozenfeld B (2008) Learning realistic human actions from
movies. In: Proceedings of 21st IEEE conference on computer vision and pattern recognition,
CVPR’08, Anchorage, AK, pp 1–8
Lin Z, Jiang Z, Davis LS (2009) Recognizing actions by shape-motion prototype trees. In:
Proceedings of 12th IEEE conference on computer vision, ICCV’09, Kyoto, Japan, pp 444–451
Messing R, Pal C, Kautz H (2009) Activity recognition using the velocity histories of tracked
keypoints. In: Proceedings of 12th IEEE conference on computer vision, ICCV’09, Kyoto,
Japan, pp 104–111
Moeslund TB, Hilton A, Kruer V (2006) A survey of advances in vision-based human motion
capture and analysis. Comput Vis Image Underst 104(2–3):90–126
Nater F, Grabner H, Van Gool L (2010) Exploiting simple hierarchies for unsupervised human
behavior analysis. In: Proceedings of 23rd IEEE conference on computer vision and pattern
recognition, CVPR’10, San Francisco, CA, pp 2014–2021
Poppe R (2010) A survey on vision-based human action recognition. Image Vis Comput
28(6):976–990
Scovanner P, Ali S, Shah M (2007) A 3-dimensional sift descriptor and its application to action
recognition. In: Proceedings of 15th conference on multimedia, MM’07, Augsburg, Germany,
pp 357–360
Weinland D, Boyer E (2008) Action recognition using exemplar-based embedding. In:
Proceedings of 21st IEEE conference on computer vision and pattern recognition, CVPR’08,
Anchorage, AK, pp 1–7
Chapter 11
Adaptive Ant Colony Algorithm
Researching in Cloud Computing Routing
Resource Scheduling
Zhi-gao Chen
Abstract Cloud computing has been regarded as one of the most important
planning projects in the future, the technique will be beneficial to thousands
enterprises in our country. The advantages of Cloud service depend on efficient,
fast running network conditions. At present, under the condition of limited
bandwidth in our country, studying fast and efficient routing mechanism is nec-
essary, according to which Scheduling resource with the maximum capacity of a
network node. Therefore, in this paper, the parameters of network capacity was
increased as the threshold in each node to route adaptively, the shortest path can be
found quickly on the traditional ant algorithm, and also the network capacity of
nodes on the path can be adjusted accordingly. As the experimental result shown,
the congestion of data on the critical path can effectively avoid by this method.
Keywords Ant colony algorithm Cloud computing Pheromone
11.1 Introduction
Cloud computing is a distributed processing, parallel processing and grid com-
puting, that is stored in the PC, mobile phones and other devices on the wealth of
information and processor resources are concentrated together collaborative work,
great the expansion of IT capacity, a calculation method to provide services to
external customers (Asterisk 2010) (XUE Jing 2010). In 2011, cloud computing
has been treated as the key project in the twelfth five-year plan by our government,
the importance and significance have no question count (Proceedings of the ninth
international symposium on distributed computing and applications to business,
Z. Chen (&)
Hu Nan vocational institute of science and technology,
Changsha, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 101
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_11,
Ó Springer-Verlag Berlin Heidelberg 2013
102 Z. Chen
engineering and science 2010). At present, our country is in the growth stage of
cloud computing (Proceedings of 2010 third international symposium on knowl-
edge acquisition and modeling (KAM 2010) 2010) (HAN Bing 2010), over the
next 10 years, many IT enterprise and the small and medium-sized enterprises will
share hundreds of billions of dollars of ‘‘cloud computing cake’’ in the future of
our country (Nezamabadi-pour et al. 2006).
While there are some disadvantages in the cloud computing industry of our
country. Cloud computing needs to run in high-speed network to exert its
advantages (Satyanarayanan 2001), although China has been used widely the
broadband technology, but net is still less than other countries. Cloud computing
needs to run in high-speed network to play its advantages, Broadband technology
has been applied in our country widely, but the speed of the network is still lower
than other countries. It is the premise of cloud computing construction in china that
Operation cloud services needs a set of efficient and secure routing mechanism on
Cloud Computing to achieve efficient resource scheduling and storage, on the basis
of limited bandwidth; on the other hand, along with the expansion of cloud
computing network and increase of cloud services, higher requirements are put
forward to our existing network bandwidth and the resource scheduling, therefore,
the research and application of the cloud computation efficient routing mechanism
can improve the efficiency of routing and scheduling speed resources, as a result
the cloud service will run more efficiently in the existing network infrastructure
and improving fully on the return rate of investment on cloud computing infra-
structure, Reference and basis of China’s construction of cloud services platform is
provided better, the significant on the progress of our society and the pulling and
development of national industry is greatly.
11.2 Cloud Computing Infrastructure and Ant Algorithm
11.2.1 Introduced Cloud Computing Environment
At present, cloud computing infrastructure technology is mainly the Google non
open source system of GFS and hadoop technology for GFS open source imple-
mentation of the HDFS (Fetterly et al. 2010). Hadoop is a reliable, efficient,
scalable software framework, and can handle large amounts of distribution data. It
can make use of cluster technology processing PB data in the cheap personal
computer with parallel manner, and can be used freely because of its own JAVA
language framework (Asterisk 2010). The core part of Hadoop HDFS and Map
Reduce use Master/slave structure, the Hadoop system is unified into the two layer
structure. A HDFS cluster consists of a Name Node and a plurality of Data Node.
Name Node is the main server, responsible for the management of the file system
name space and client access to files, Data node is the slave server, which dis-
tributed on each physical node in a cluster usually, responsible for memory
11 Adaptive Ant Colony Algorithm 103
management on their physical nodes. In the internal node, files are usually divided
into one or more blocks of data the blocks of data stored in a set of Data Node (IP
Multimedia Subsystem (IMS) 2009). Name Node execution operation about the
file system name space, such as open, closed, the rename operation, and also
decided to map data block from Name Node to Data Node. Data Node is
responsible for processing of read and write requests of customers, also performs a
data block in accordance with the Name Node instruction.
11.2.2 Ant Algorithm
The ant algorithm (Ant Colony Algorithm) is proposed by the Italian scholar
Marco Dorigo in 1992, a parallel and efficient evolutionary algorithm (Guo et al.
2010). The algorithm is a probabilistic technique for the simulation of ant foraging
process in nature which is formed to find the optimal path in the graph, the core
idea is: ants will left a ‘‘pheromone’’ chemical substances in the path searching for
food (Hayden 2009), these ‘‘pheromone’’ can provide heuristic information where
selected on walking routes, for the follow-up ants to find food as the constant
updating of pheromone, optimal path can be found from the nest to the food in a
relatively short period of time. The algorithm has the advantages of high parallel,
convergence speed, and has Gained some satisfactory experimental results, in the
traveling salesman problem, routing and scheduling problems, but the standard ant
algorithm is easy to fall into local optimal solution. Considering the cloud com-
puting environment is large in scale, and the quality of service requirements, to
achieve efficient resource scheduling, the shortest path should be found in algo-
rithm and also meet the bandwidth requirements of which cloud services supplied
in the path each node can offer. The proposed adaptive ant algorithm, is based on
running a cloud service according to our country current limited bandwidth, by
setting the appropriate threshold about minimum network capacity to adaptively
adjust searching for the shortest path, can both quickly discover the resource such
as the routing, but also to improve the convergence of the algorithm, and the QOS.
11.2.3 Adaptive Ant Algorithm
To achieve efficient and fast scheduling of cloud services, in a real environment
must fulfill two conditions: (1) the data transmission path must be the shortest path
to reduce the data transmission distance; (2) the shortest path through each node
must have enough bandwidth to run cloud services, it will be congestion in some
nodes. The ant algorithm selects the shortest path by the rapid convergence con-
dition (1) easy to implement. Search the shortest path, all cloud services are run
from the path, is bound to be caused by data traffic increased sharply on the
shortest path. At present, limited the basis of bandwidth in our country that can be
104 Z. Chen
provided by the between of each node network the maximum capacity of is has
been spotty. Once all the business on the shortest path on the transmission, will
make some smaller capacity network node to enter the congestion in advance. If
you cannot adjust in a timely manner will make a follow-up business to continue to
transfer from the shortest path, leading to congestion and data retransmission
exacerbate transmission of cloud services is very unfavorable. The establishment
of a multi-road by the table for the same source knot point to the target knot point.
Search idea is that the network capacity threshold set automatically when preferred
shortest path congestion routing, get a new ‘‘sub-optimal’’ shortest path. And so
on, so as to achieve the normal operation of the entire network in the state of
optimal network capacity. Therefore construct adaptive ant algorithm steps are:
solving the source node to destination node point shortest path process, consider
the capacity constraints and flow changes on the network each path in real time,
i.e., solving the shortest path to the source node to destination node sections
capacity minus the minimum link capacity. When the data was transferred on the
shortest path close to the minimum link capacity, the bottle neck sections of
congestion, the other nodes in which sections of the available network capacity
bottleneck link capacity minus the capacity of the original node, rather than the
capacity of the shortest path remains unchanged.
11.3 Model Designing for Adaptive Ant Algorithm
Firstly, each node in the cloud environment was treated as the point in the map
abstractly. A certain node was set as the start point, will was searched by the ant as
the ‘‘food’’ finally, namely to complete the routing process.
11.3.1 Improvement of Algorithm
Firstly, each node in the cloud environment is abstracted as a connected graph
point to determine a node as a starting point, the final node of the visit as ‘‘food’’
by the ants to search. Routing process was completed when the ants had found the
target in the traditional ant algorithm, the filmon was reserved in line which the
ants searched, the algorithm has been modified in this paper that the filmon was
reserved on each node traversed by the ants. The si represents the amount of
information, H represents the network capacity values the node can carry,
Threshold is set to di, if di [ H, hosted by the node network traffic has exceeded
the network capacity, this means that the congestion would be coming on the node,
then the new business transmission can not be longer allowed in the path. The
choice of the new path must be started in order to bypass the sections in upcoming
congestion, a new sub-optimal path would be created.
11 Adaptive Ant Colony Algorithm 105
With the establishment and the end of the network session, the amount of
information on each node would change when each cycle completed, the Fireman
on each node Adjusted as (11.1, 11.2):
si ðt þ 1Þ ¼ qsi ðtÞ þ Dsi ð11:1Þ
X
mn
Dsi ¼ Dski ð11:2Þ
k¼1
Among them, the k-th ant in the cycle stays in the pixels on the filmon.
11.3.2 Experiment
Figure 11.1 is a simulation example of a network of choice in this article, a total of
24 nodes in the graph, the connection on behalf of the node between the node path,
the connection of a group of figures, respectively, the distance between the node
and the maximum network capacity, it is assumed from the source node 1 to node
23, and the source node to node 24, two paths through the data routing experiment.
Step1: designed to be adaptive ant algorithm to find the shortest path to node 1
? 23;here to take a = 1, b = 2, q = 0.8, Q = 1000, the initial value smin = 60. Get 1
? 23 of the shortest path is : 1 ? 3 ? 6 ? 11 ? 16 ? 20 ? 23, 15 flow
(minimum segment capacity) is 4;
Step2: the capacity of each segment of the shortest path 1 minus the smallest
segment of the shortest path capacity, the results of 16 ? 20 segment of the
capacity is 0, the segment identified as the bottleneck segment, the segment most
prone to congestion;
Step3: According to the capacity change and the connection matrix changes to
look for the source node and destination node 23 the shortest path available to the
Fig. 11.1 Node topology
106 Z. Chen
Fig. 11.2 Cloud node topology
new shortest path `: 1 ? 4 ? 7 ? 12 ? 17 ? 19 ? 20 ? 23, 18 flow rate, and
similarly the capacity of each segment of the shortest path ` subtract the smallest
segment of the shortest path capacity, the result of the capacity of the 4 ? 7
segment 0, that segment is identified as unavailable, repeat the same steps to find
the shortest path ´ change Network: 1 ? 4 ? 9 ? 14 ? 19 ? 20 ? 23, length
19, flow rate of 3; ˆ: 1 ? 2 ? 5 ? 10 ? 15 ? 18 ? 20 ? 23, length 21, flow
rate of 5;
Step4: could not find a feasible route of the destination node of a source node ?
(or less than the number of shortest path to the default limit), the source node or
destination node will become an isolated point, out of the end of the loop;
(Fig. 11.2)
Step5: will find the shortest path sequence as the source node 1 ? purpose node
23 optional routing tables, the same method can be found in the 9 ? 15 optional
routing tables. As the experiment shows, the adaptive ant algorithm constructed in
this paper had got the data listed in Table 11.1. As the experimental results,
dynamic optimal routings from node 1 ? 23 and 9 ? 24, there are four paths of
Optimal routing from node 1 ? 23, and four paths of Optimal routing from 9 ?
24 as the same. According to the provisions of the preferential routing priority, in
order to achieve a dynamic optimal routing because of the network congestion or
partial failure.
11 Adaptive Ant Colony Algorithm 107
Table 11.1 Dynamic optimal routing
Source ? target order selected path Shortest path Flow
1 ? 23 1 ? 23 1 1 ? 3?6 ? 11 ? 16 ? 20 ? 23 15 4
2
1 ? 4?7 ? 12 ? 17 ? 19 ? 20 ? 23 18 4
3 1 ? 4?9 ? 14 ? 19 ? 20 ? 23 19 3
4 1 ? 2?5 ? 10 ? 18 ? 20 ? 23 21 5
9 ? 24 9 ? 24 1 9 ? 14 ? 17 ? 16 ? 15 ? 24 15 4
2 9 ? 14 ? 19 ? 20 ? 18 ? 15 ? 24 18 1
3 9 ? 7?6 ? 8?10 ? 15 ? 24 19 3
4 9 ? 4?3 ? 2?5 ? 10 ? 15 ? 24 21 5
11.4 Discussion
In this paper, the adaptive ant algorithm can effectively avoid congestion in net-
work on the shortest paths, and select the shortest paths automatic by setting the
minimum network capacity threshold for each of the segments in the shortest path
in hadoop cloud platform, provided a choice of an adaptive routing scheme in a
cloud environment.
The experiments about the algorithm are carried out in hadoop cloud platform,
not by simulator, the practicality and adaptability is better. A good self-healing
method was provided for some sections on the shortest path when some sections
were failed, for author/s of more than two affiliations: to change the default, adjust
the template as follows.
The minimum network capacity as the only one factor which was considered in
this method when the paths were selected adaptively on each segment, this is
certainly not enough in the real cloud environment, therefore, the algorithm needs
the further improvement if been applied in the cloud environment.
References
Xue Jing (2010) A brief survey on the security model of cloud computing. In: Proceedings of the
ninth international symposium on distributed computing and applications to business,
engineering and science
Asterisk (2010) Open source communications [CP/OL]
Fetterly D, Haridasan M, Isard M et al (2010) TidyFS: a simple and small distributed file system.
In: Proceedings of the USENIX annual technical conference (USENIX’11), 2011
Guo D, Wu K, Li J et al (2010) Spatial scene similarity assessment on Hadoop. In: Proceedings of
the ACM SIGSPATIAL international workshop on high performance and distributed
geographic information systems, New York, pp 39–42
Hayden C (2009) Announcing the map/reduce toolkit
IP Multimedia Subsystem (IMS) (2009) Stage 2 (Release 9)
108 Z. Chen
Nezamabadi-pour H, Saryazdi S, Rashedi E (2006) Edge detection using ant algorithm. Soft
Comput 10(7):623–628
Han Bing (2010) Research and implementation of future network computer based on cloud
computing (2010). In: Proceedings of 2010 third international symposium on knowledge
acquisition and modeling (KAM 2010)
Satyanarayanan M (2001) Pervasive computing: vision and challenges. IEEE Pers Commun
8(4):10–17
Chapter 12
An Empirical Analysis on Guangdong
Industrial Energy Intensity Based
on Panel Data
Xin-dong Hao
Abstract Energy is of vital importance to the modern industrial economy. Based
on the panel data of 36 industrial sectors in Guangdong, this paper analyzes the
relationship between energy intensity and industrial added value by means of gray
relational analysis. The results show that the energy intensities of industrial sectors
are quite inconsistent with their industrial added values. Guangdong should attach
great importance to the development of ‘‘win–win’’ industrial sectors which
increasing industrial output while reducing energy intensity.
Keywords Energy intensity Grey relational analysis Industrial added value
Panel data
12.1 Introduction
China is facing severe challenges from energy supply and environment protection
due to its heavy dependence on energy. Up to now, Guangdong has made great
progress in industrialization, and its industrial output value has ranked first in
China for many years, and even some of its manufactured goods play a prominent
role on world scene. Industrialization needs large amounts of energy, Guangdong’s
total energy consumption was 152.367 million tones of standard coal in 2011, and
Guangdong try to reduce its per unit energy consumption by 18 % in the 12th
5-Year Plan (2011–2015).
Over the years, Guangdong’s industrial growth rely heavily on high input and
high energy consumption, its industrial energy consumption accounts for about
70 % of the total consumption. At present, Guangdong is facing serious challenges
X. Hao (&)
School of Economics and Management, Wuhan University, Wuhan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 109
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_12,
Ó Springer-Verlag Berlin Heidelberg 2013
110 X. Hao
of resources and environmental, and the energy problem is hindering Guangdong’s
economic growth and industrialization. The energy intensities of industrial sectors
play a dominant role in total energy consumption; many studies have focused on
this problem. A brief literature review as follows.
Sinton and Levine (1994) studied China’s industrial energy intensity of
1980–1990 and found that the reasons for the decline of overall industrial energy
intensity were primarily the improvement of energy efficiency. Garbaccio et al.
(1999) tried to construct an index system on energy intensity and they pointed to
that technological changes being one of main reasons for the decline of energy
intensity. Karen et al. (2004) analyzed the panel data of 1997–1999 on large and
medium-sized industrial enterprises in China, and found that the decline of energy
intensity mainly due to the reduction of coal consumption in industrial sector,
technology development, industrial restructuring.
Lun and Ouyang (2005) analyzed the relationship of Guangdong’s economic
and energy consumption, and they argued that there was a significant positive
correlation between them. He and Zhang (2005) insisted that the increasing of
China GDP energy intensity was mainly due to the fast increase of energy con-
sumption in industry, particularly in heavy and chemical industrial sectors. Fisher-
Vanden et al. (2006) studied China’s energy intensity based on sample data which
coming from a large number of industrial enterprises and they found that the
change of production structure was one of the major factors for the decline of
energy intensity. Hu and Wang (2006) conducted a comprehensive analysis on the
factors of affecting China regional energy efficiency, and they insisted that the
main influencing factors include energy structure, industrial structure, technology
and so on.
Shi and Dong (2007) insisted that the differences of energy efficiency in China
were due to the factors such as economic development level, industrial structure,
energy consumption structure, maketization degree. Yu (2007) thought that the
decrease of Guangdong’s energy intensity was mainly due to the increase of
energy utility efficiency in the second industry. The rising of energy utility effi-
ciency in the key sectors of industry had led to the fall of the whole energy
intensity. Ma and David (2008) analyzed the data of energy intensity between
1980 and 2003 in China and they pointed out that the technological progress was
extremely important for knock-down industrial energy intensity. Zha et al. (2009)
nalyzed the energy consumption of 36 China industrial sub-sectors from 1993 to
2003. The results showed that the industrial structure played a important role in
China’s energy intensity.
Zhu (2010) studied the general characteristics of energy consumption based on
the data of Guangdong, and he designed an energy demand path of low-carbon
development for Guangdong. Cai (2010) insisted that the energy problems had
seriously affected the development of Guangdong’s industry and he suggested that
the reduce of energy intensity would be a direct ways of accelerating the devel-
opment of Guangdong industrialization. Fu and Zhang (2011) analyzed the data of
36 Guangdong industrial sectors from 2001 to 2010, the results showed that the
structural effect and technical effect had a positive role on the reduction of energy
12 An Empirical Analysis on Guangdong Industrial Energy Intensity 111
consumption while the scale effect had a negative role. The development of steel,
nonferrous metal, building materials, chemical and other high energy consumption
industrial sectors was the main reason for the rapid rise of total energy con-
sumption. Zheng (2011) analyzed the general characteristics of energy consump-
tion and energy efficiency based on the data of industrial sectors in Guangdong,
and the result showed that there were many differences in different industrial
sectors.
In this paper, we go through the main study results of previous literature so as to
explore the relationship between industrial energy intensity and industrial added
value in Guangdong basing on gray relational analysis. By this way, we try to find
out which Guangdong industrial sectors should be the priority development ones
by the perspective of energy intensity and industrial added value. Our goal is to
identify the key industrial sectors of ‘‘win–win’’ which increasing industrial output
while reducing energy intensity.
12.2 Methodology of Gray Relational Analysis (GRA)
The gray system theory was first proposed in 1982, and it is a system containing
both insufficient and sufficient information, called ‘‘black’’ and ‘‘white,’’ respec-
tively, is a gray system. A system can be called as a black box if its mathematical
equations or internal characteristics that describe its dynamics are completely
unknown. On the other hand, a system can be named as a white system if the
description of the system is completely known. Similarly, a system that has both
known and unknown information is defined as a gray system. The gray system
theory deals with a system containing insufficient information, the gray relational
analysis can capture the relationship between the main factor and other factors in a
system regardless whether this system has adequate information. The gray rela-
tional analysis can effectively avoid subjective bias and have better performance
than some traditional methods when the study involves economic, environmental
and technical indices problems.
In real life, each system can be considered as a gray system because there are
always some uncertainties in them. Even a simple price system always contains
some gray characteristics because of the various kinds of social and economic
factors. These factors are generally random and make it difficult to obtain an
accurate model. There are many situations in which the difficulty of incomplete or
insufficient information is faced, by gray relational method; we can calculate the
gray relational coefficient and the gray relational grade of a gray system, and to
find out the system’s internal relationships.
A GRA model is a kind of impact measurement model of two series which
named reference series and compare series. During the processes of system
development, the change trends of two series should be consistent, a higher grade
of synchronized change can be considered to have a greater ranking; otherwise, the
grade of relation would be smaller. Thus, the analysis method, which takes the
112 X. Hao
ranking of the relation into account, is established upon the degree of similarity or
difference of the developmental trends of two series to measure the degree of
relation.
12.3 Empirical Analysis
Taking energy intensity (Tons of standard coal/yuan) as reference series, and
taking industrial added value as compare series. The energy intensity come from
the energy consumption per unit industrial added value, the data involving 36
industrial sectors of Guangdong from 2004 to 2010. The compare series are the
industrial added values of above-mentioned 36 industrial sectors. All the data
come from Guangdong Statistical Yearbooks (GSY, 2005–2011). By the meth-
odology of GRA, we summary the gray relational degrees (descending order) in
Table 12.1, we take also Guangdong energy intensity data(ascending order) into
Table 12.1, so as to find out the ‘‘win–win’’ industrial sectors.
The 36 industrial sectors as follows: tobacco manufacturing industry (X1), oil and
gas mining industry (X2), transportation equipment manufacturing industry (X3),
recycling and disposal of waste industry (X4), instrumentation and culture, office
equipment manufacturing industry (X5), electrical machinery and equipment man-
ufacturing industry (X6), pharmaceuticals manufacturing industry (X7), universal
equipment industry (X8), special equipment manufacturing industry (X9), garments,
shoes and caps manufacturing industry (X10), beverage manufacturing industry
Table 12.1 The sorts of energy intensity and grey relational degree (2004–2010)
Sectors Energy intensity Sort Degree Sort Sectors Energy intensity Sort Degree Sort
X1 0.061 1 0.893 6 X19 0.858 19 0.896 3
X2 0.127 2 0.910 2 X20 0.923 20 0.850 18
X3 0.283 3 0.830 29 X21 0.982 21 0.849 19
X4 0.379 4 0.727 34 X22 1.063 22 0.878 9
X5 0.445 5 0.841 25 X23 1.258 23 0.894 5
X6 0.454 6 0.840 27 X24 1.370 24 0.842 24
X7 0.623 7 0.895 4 X25 1.386 25 0.833 28
X8 0.664 8 0.845 22 X26 1.403 26 0.827 30
X9 0.667 9 0.843 23 X27 1.529 27 0.823 31
X10 0.678 10 0.854 16 X28 1.658 28 0.864 12
X11 0.681 11 0.917 1 X29 1.780 29 0.820 32
X12 0.687 12 0.856 13 X30 1.946 30 0.889 8
X13 0.724 13 0.876 10 X31 2.106 31 0.841 26
X14 0.769 14 0.850 17 X32 2.448 32 0.855 15
X15 0.771 15 0.713 35 X33 3.308 33 0.685 36
X16 0.776 16 0.856 14 X34 4.306 34 0.846 20
X17 0.776 17 0.890 7 X35 5.797 35 0.845 21
X18 0.847 18 0.867 11 X36 6.013 36 0.813 33
12 An Empirical Analysis on Guangdong Industrial Energy Intensity 113
(X11), leather, furs, feathers manufacturing industry (X12), food manufacturing
industry (X13), handicrafts and other Manufacturing industry (X14), nonferrous
metals mining and extraction industry (X15), furniture manufacturing (X16), printing
and copying industry (X17), metal products industry (X18), ferrous metals mining and
extraction industry (X19), cultural, educational, sporting articles manufacturing
industry(X20), chemical raw materials and chemical products manufacturing(X21),
agro-food processing industry (X22), nonmetal mineral mining and extraction
industry (X23), plastic products industry (X24), nonferrous metals smelting and
rolling processing industry (X25), wood, bamboo, timber processing industry (X26),
water processing and supplying industry (X27), rubber products industry (X28),
chemical fiber industry (X29), textiles industry (X30), electric power, hot power
producing and supplying industry (X31), papermaking and paper products industry
(X32), gas producing and supplying industry (X33), oil processing, coking and
nuclear fuel processing industry (X34), nonmetal mineral products industry (X35),
ferrous metals smelting and rolling processing industry (X36).
The Table 12.1 shows that the top fifteen industrial sectors from the perspective
of gray relational degree are: beverage manufacturing, oil and gas mining, ferrous
mining, medicine manufacturing, non-metallic Mining, printing recording media
replication, tobacco manufacturing, textiles, agro-food processing, food manu-
facturing, fabricated metal products, rubber, leather, fur, feathers and their prod-
ucts, furniture, paper and paper products industries. Obviously, the top fifteen
sectors from the perspective of energy intensity are different.
12.4 Conclusion
Energy intensity is the comprehensive reflection of a country or region’s economic
structure, mode of growth, level of science and technology, management capacity,
consumption patterns and so on. This study examines Guangdong’s industrial
energy consumption during 2004–2010 by the gray analysis method and it gives us
a new way of thinking (named ‘‘win–win’’) for breaking through Guangdong’s
energy bottlenecks.
The gray relational analysis shows that there are complex relationships between
Guangdong’s industrial energy intensity and industrial added value, i.e. the
industrial added value of the sectors can be high while their energy intensities are
low.
Comparing the rankings of energy intensity and gray relational degree, we find
out the ‘‘win–win’’ industries, i.e. increasing industrial output while lowering
energy intensity industries, include tobacco manufacturing industry, oil and gas
extraction industry, pharmaceutical manufacturing industry, beverage manufac-
turing industry, leather, fur, feather products industry, food manufacturing
industry, they all should be given priority to development. We should also pay
attention to the industrial sectors of transportation equipment manufacturing
industry, recycling and disposal of waste industry, electrical machinery and
114 X. Hao
equipment manufacturing, non-ferrous metal mining industry because their gray
relational degree are low but their ranks of energy intensity are high. Although
such industries cannot play an important role in create industrial added value, but
they consume less energy.
References
Cai MR (2010) The research of energy and industrial development in Guangdong province. In:
Proceedings of information management, innovation management and industrial engineering,
ICIII’10, Hong Kong, VA, pp 333–336
Fisher-Vanden K, Jefferson GH, Jingkui M (2006) Technology development and energy
productivity in China. Energy Econ 28(5):690–705
Fu JY, Zhang HL (2011) An empirical analysis of influential factor of Guangdong’s energy
consumpution in the industrial sector. http://www.paper.edu.cn/index.php/default/releasepaper/
content/201204-424
Garbaccio RF, Ho MS, Jorgenson DW (1999) Why has the energy output ratio fallen in China?
Energy J 20(3):63–91
He JK, Zhang XL (2005) Analysis on the impact of the structural changes in the manufacturing
industry on the rising of intensity of GDP resources and its trend (in Chinese). Environ Prot
12:37–41
Hu JL, Wang SC (2006) Total-factor energy efficiency of regions in China. Energy Policy
17(5):3206–3217
Karen FV, Gary HJ, Liu HM, Tao Q (2004) What is driving China’s decline in energy intensity?
Resour Energy Econ 26(1):77–97
Lun X, Ouyang YG (2005) A study on energy constraint based on Guangdong sustainable
development (in Chinese). Coal Econ Res 28(2):30–32
Ma CB, David IS (2008) China’s changing energy intensity trend: a decomposition analysis.
Energy Econ 30(3):1037–1053
Shi D, Dong L (2007) A study on energy efficiency and energy saving potential and its
influencing factors in China (in Chinese). Nat Gas Technol 1(2):5–8
Sinton JE, Levine MD (1994) Changing energy intensity in chinese industry; the relative
importance of structural shift and intensity change. Energy Policy 22(5):239–255
Yu PG (2007) A factor analysis of the energy intensity in China (in Chinese). Acad Res 2:74–79
Zha Dl Zhou DQ, Ning D (2009) The contribution degree of sub-sectors to structure effect and
intensity effects on industry energy intensity in China from 1993 to 2003. Renew Sustain
Energy Rev 13(4):895–902
Zheng LZ (2011) Study on the energy efficiency of industrial sector in Guangdong province.
http://210.44.1.99/kns50/detail.aspx?dbname=CMFD2011&filename=1011191476
Zhu SX (2010) An analysis of fictitious conditions in Guangdong low coaban growth (in
Chinese). China Opening Herald 148(1):57–61
Chapter 13
An Empirical Research on the Ability
of Sustainable Development for Coal
Resource Exhausted Cities
Bing Zhang
Abstract The sustainable development of coal resource exhausted cities has
become an important part of whole social development. This paper constructs the
sustainable development indicator system of the coal resource exhausted cities
from the four perspectives of the economy, society, resource and environment, we
choose 12 coal resource exhausted prefecture-level cities for examples, analyze the
level of sustainable development of the coal cities from 2006 to 2011 using
principal component analysis method. Finally, it proposes the policy recommen-
dations for sustainable development of coal resource exhausted cities.
Keywords Coal resource exhausted city Empirical research The ability of
sustainable development
13.1 Introduction
Our country has rich mineral resources, it has nearly one hundred large and
medium cities built mainly in mining. Since twenty first century, coal resources are
gradually shrinking and depletion, the resource cities standing due to coal have
difficult to see its former glory, which have seriously lagging behind in economic
development, a structural contradiction of single industry has become increasingly
emergent (Xu and Zhang 2000). With the growing exploitation of resources,
mining geological conditions become more complex, resource extraction enter-
prises are gradually aging and these resource-based cities will face the important
topic of economic transformation (Wang 2003). In the process of continuous
improvement of China’s market economy, how to promote resource exhausted
B. Zhang (&)
College of Management and Economics, Tianjin University, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 115
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_13,
Ó Springer-Verlag Berlin Heidelberg 2013
116 B. Zhang
cities to achieve comprehensive and sustainable development in economic, social,
resources and environment has become an imperative and urgent major strategic
task needing to be addressed (Xia and Shen 1998).
13.2 Overview of Coal Resource Exhausted Cities
Resource-based cities are an important urban-type in our country, they generally
refer to the cities that rely on the natural resources of mineral resources and
forest resources and so on, and these cities have specialized functions in which
resource extraction and initial processing as their pillar industries, the similar
concept is mining city and industrial city and so on (Zhao 2006). The coal
resource exhausted cities are an important part of resource exhausted cities in
our country (Wang et al. 2010). We choose 12 prefecture-level cities for
examples which are the coal resource exhausted cities in the list issued by our
country, including the Zaozhuang city of Shandong Province, the Baishan city of
Jilin Province, the Fushun city of Liaoning Province, the Fuxin city of Liaoning
Province, the Huaibei city of Anhui Province, the Huangshi city of Hubei
Province, the Jiaozuo city of Henan Province, the Liaoyuan city of Jilin Prov-
ince, the Pingxiang city of Jiangxi Province, the Qitaihe city of Heilongjiang
Province, the Shizuishan city of Ningxia Hui Autonomous Region, and the
Tongchuan city of Shanxi Province. In recent years, these cities have already
begun with economic transformation, the level of economic development is
gradually increasing, and the GDP growth rate is slightly higher than the national
average. At the same time, these cities also focus on the coordinated develop-
ment of the society, resources and environment; take various measures to raise
the level of sustainable development.
13.3 Build Index System of Sustainable Development
for Coal Resource Exhausted Cities
The evaluation system of sustainable development for coal resource exhausted
cities should reflect the development status and trend of four systems within the
economy-society-resource-environment complex system, and also reflect the
coordination state of the four systems. In addition to considering the basic ideas
and general principles of sustainable development, it should also reflect the
characteristics of the coal resource exhausted cities. This paper presents a model of
sustainable development indicator systems for coal resource exhausted cities.
CRCSD ¼ f ðK1 ; K2 ; K3 ; K4 ; S; T Þ
13 An Empirical Research 117
In the model, CRCSD is the ability of sustainable development for coal resource
exhausted cities, K1 is the economy system, K2 is the society system, K3 is the
resource system, K4 is the environment system, S is the space variables, it refers to
the different cities, T is the time variables, it refers to the different development
stages.
Each subsystem is further represented by a number of indicators.
Ki ¼ f ðKi1 ; Ki2 ; ; Kin Þ i ¼ 1; 2; 3; 4
Specifically, the index system should highlight the following features. First, it
should reflect the quality and scale of economic development for coal resource
exhausted cities. Second, it should reflect the operating conditions of the social
system; the key is to make a clear assessment of eliminating poverty and
improving the quality of life and so on. Third, it should have a great importance to
the extent of development and utilization of the main resources and the richness of
the existing resources. Four, it should reflect the environment, especially the
capacity of natural environment and the ability of regional sustainable develop-
ment (Li 2005; Li and Lu 2008; Pang and Wang 2012).
According to the function need to be reflected by the above model and index
system, combined with the own characteristics and goals of sustainable develop-
ment for the particular system of coal resource exhausted cities, we establish a
index system for sustainable development and select 28 representative indicators
as shown in Table 13.1 (Li 2005; Li and Lu 2008; Pang and Wang 2012;
Akinbami and Fadare 1997; Marsh 1987).
Table 13.1 The index system of sustainable development for coal resource exhausted cities
Economic indicators Social indicators
GDP per capita Annual average wage of workers
Coal resources industry added value accounts for Disposable income per urban population
GDP Employment rate of urban population
The tertiary industry added value accounts for GDP The poverty rate
Financial revenue accounts for GDP The proportion of people in and above
The growth rate of investment in fixed assets college
Investment in environmental protection accounts for The labor force average years of schooling
GDP Insurance rates for social insurance
The actual utilization of foreign capital Living area per capita
Total amount of import and export
Resource indicators Environment indicators
Existing reserves of coal resource Wastewater treatment rate
The useful life of coal resource Waste gas treatment rate
The utilization rate of coal resource Dustfall concentration
Per capita share of coal resource The comprehensive utilization of solid and waste
The consumption of coal resource Green area per capita in urban
Available arable land per capita Forest coverage rate
Available water resource per capita
118 B. Zhang
13.4 Empirical Study on the Sustainable Development
of Coal Resource Exhausted Cities
According to the above index system of sustainable development for coal resource
exhausted cities, we select some key indicators to measure its capacity for sus-
tainable development. According to the Statistical Yearbook of the coal resource
cities from 2006 to 2011, we select the following variables and data as shown in
Table 13.2 (Statistical yearbook of the various coal resource-exhausted cities from
2006). Then use principal component analysis method to comprehensive measure
the level of sustainable development in these years of coal resource exhausted
cities. Data is processed by the software of EXCEL and SPSS (Gao and Dong
2007).
In Table 13.2, X1 is GDP per capita (yuan), X2 is the tertiary industry accounts
for GDP (percentage), X3 is investment in environmental protection accounts for
GDP (percentage), X4 is annual average wage of workers (yuan), X5 is employ-
ment rate of urban population (percentage), X6 is the proportion of people in and
above college (percentage), X7 is living area per capita (square meters), X8 is per
capita share of coal resource (ton), X9 is wastewater treatment rate (percentage),
X10 is green area per capita in urban (square meters) (Table 13.3).
The correlation coefficient matrix of each variable can be seen, the correlation
coefficient is larger among each variable, they also have a strong correlation. It can
use principal component analysis method to comprehensive reflect the level of
sustainable development through the relatively small number of variable synthetic
indicators.
Table 13.2 The variable data
Years X1 X2 X3 X4 X5
2006 15055 0.3392 0.0071 15473 0.9499
2007 18254 0.3305 0.0078 19096 0.9543
2008 22923 0.3147 0.0101 22361 0.9579
2009 26103 0.3195 0.0106 25026 0.9600
2010 30860 0.3032 0.0126 29391 0.9629
2011 36499 0.3138 0.0167 33301 0.9635
Year X6 X7 X8 X9 X10
2006 0.0624 24.91 275 0.8678 8.93
2007 0.0656 26.35 260 0.8101 9.56
2008 0.0692 27.41 244 0.8613 10.08
2009 0.0724 28.67 231 0.8593 10.99
2010 0.0772 29.53 220 0.8658 12.23
2011 0.0801 30.53 206 0.9034 12.73
Explanation: the 10 indicators are the simple average of the 12 selected coal resource exhausted
cities
13 An Empirical Research 119
Table 13.3 The correlation coefficient matrix of each variable
X1 X2 X3 X4 X5
X1 1.000
X2 -0.823 1.000
X3 0.981 -0.735 1.000
X4 0.998 -0.840 0.972 1.000
X5 0.958 -0.923 0.893 0.969 1.000
X6 0.995 -0.857 0.960 0.998 0.974
X7 0.987 -0.858 0.941 0.991 0.987
X8 -0.992 0.858 -0.953 -0.994 -0.984
X9 0.664 -0.389 0.739 0.621 0.493
X10 0.988 -0.837 0.951 0.992 0.958
X6 X7 X8 X9 X10
X1
X2
X3
X4
X5
X6 1.000
X7 0.992 1.000
X8 -0.994 -0.999 1.000
X9 0.622 0.566 -0.602 1.000
X10 0.996 0.982 -0.982 0.616 1.000
Table 13.4 Total variance explained
Component Initial eigenvalues
Total % of variance Cumulative %
1 8.984 89.837 89.837
2 0.740 7.401 97.238
3 0.215 2.154 99.391
4 0.032 0.316 99.707
5 0.029 0.293 100.000
6 2.84E - 016 2.84E - 015 100.000
7 1.27E - 016 1.27E - 015 100.000
8 -1.6E - 016 -1.60E - 015 100.000
9 -2.6E - 016 -2.59E - 015 100.000
10 -4.0E - 016 -4.01E - 015 100.000
Component Extraction sums of squared loadings
Total % of variance Cumulative %
1 8.984 89.837 89.837
2 0.740 7.401 97.238
120 B. Zhang
Fig. 13.1 Scree plot Scree Plot
Eigenvalue
4
1 2 3 4 5 6 7 8 9 10
Component Number
From the results of principal component analysis in Table 13.4 and Scree plot
in Fig. 13.1 can be seen, the contribution rate of the first principal component is
89.837 %, the contribution rate of the second principal component is 7.401 %, the
cumulative contribution rate of the first two components is 97.238 %. Therefore, it
can extract the first two principal components to measure the ability of the sus-
tainable development for coal resource exhausted cities.
From Table 13.4, the first two principal components eigenvalues respectively
are k1 ¼ 8:984, k2 ¼ 0:740.
The initial solution of the factor loading matrix is extracted by principal
component analysis method is shown in Table 13.5.
pffiffiffiffiffi pffiffiffiffiffi
For each column values in Table 13.5 respectively divided by the k1 , k2 , it
can get the unit eigenvectors corresponding with each eigenvalue e1 , e2 .
Table 13.5 Component Component
matrix
1 2
X1 0.996 0.041
X2 -0.865 0.335
X3 0.966 0.188
X4 0.996 -0.016
X5 0.974 -0.207
X6 0.998 -0.028
X7 0.990 -0.092
X8 -0.995 0.053
X9 0.651 0.732
X10 0.989 -0.020
13 An Empirical Research 121
0 1 0 1
0:996 0:332
B 0:865 C B 0:289 C
B C B C
B 0:966 C B 0:322 C
B C B C
B 0:996 C B 0:332 C
B C B C
1 B 0:974 C B C
C ¼ B 0:325 C
e1 ¼ pffiffiffiffiffiffiffiffiffiffiffi B C
8:984 B 0:998 C B 0:333 C
B B
C
B 0:990 C B 0:330 C
B C B C
B 0:995 C B 0:332 C
B C B C
@ 0:651 A @ 0:217 A
0:989 0:330
So the expression of the first principal component can be drawn as follows.
F1 ¼ 0:332X1 0:289X2 þ 0:322X3 þ 0:332X4
þ 0:325X5 þ 0:333X6 þ 0:330X7 0:332X8
þ 0:217X9 þ 0:330X10
0 1 0 1
0:041 0:048
B C B C
B 0:335 C B 0:389 C
B C B C
B 0:188 C B 0:219 C
B C B C
B C B C
B 0:016 C B 0:019 C
B C B C
1 B C B C
B 0:207 C B 0:241 C
e2 ¼ pffiffiffiffiffiffiffiffiffiffiffi B C¼B C
0:740 B B
0:028 C B 0:033 C
C B C
B 0:092 C B 0:107 C
B C B C
B C B C
B 0:053 C B 0:062 C
B C B C
B 0:732 C B 0:851 C
@ A @ A
0:020 0:023
So the expression of the second principal component can be drawn as follows.
F2 ¼ 0:048X1 þ 0:389X2 þ 0:219X3 0:019X4
0:241X5 0:033X6 0:107X7 þ 0:062X8
þ 0:851X9 0:023X10
According to the values and characteristics of each principal component, it can
calculate the integrated value of the ability of sustainable development from 2006
to 2011 for coal resource exhausted cities, as shown in Table 13.6.
X X
F ¼ k1 = k F1 þ k2 = k F2
¼ 0:8984F1 þ 0:074F2
122 B. Zhang
Table 13.6 The ability of sustainable development for coal resource exhausted cities
Years The first principal component F1 The first principal component F2 Integrated value F
2006 10064.44 444.49 9074.79
2007 12337.07 528.41 11122.73
2008 14979.10 689.14 13508.22
2009 16926.58 790.20 15265.32
2010 19962.27 934.94 18003.29
2011 23140.46 1129.99 20873.01
Explanation: the numerical size of F value has no practical significance, different indicators
system will have different values, it only has the reference value
It can be seen by the values of the ability of sustainable development for coal
resource exhausted cities as shown in Table 13.6, in recent years, the overall
ability of sustainable development of China’s coal resource exhausted cities is
gradually increasing, but relative to the growth rate of GDP per capita at the same
period, the growth rate of the ability of sustainable development is relatively
slower. On the one hand, it thanks to the stable and orderly development of overall
macroeconomic in our country, on the other hand, it also benefits from the strong
support of the development of coal resource exhausted cities in the national policy
(Wang and Geng 2012; Campbell and Roberts 2003; Randall and Ironside 1996).
13.5 Policy Recommendations
According to the index system and empirical analysis, it can take the following
measures to achieve sustainable development for coal resource exhausted cities. Use
industry adjustment assistance policies to broaden the financing channels to increase
the capital investment. Select the appropriate pillar industries to raise the employ-
ment rate. Improve the population quality. Strengthen the building of scientific and
technological contingent. Control pollution, enhance environmental protection.
Industry structure and regional development needs are adapting in coal cities.
References
Akinbami JFK, Fadare SO (1997) Strategies for sustainable urban and transport development in
Nigeria. Transp Policy 4(4):237–245
Campbell GA, Roberts M (2003) Urbanization and mining: a case study of Michigan. Resour
Policy 29(29):1028–1042
Gao XB, Dong HQ (2007) Data analysis and SPSS applications. Tsinghua University Press,
Beijing, pp 342–364
Li J (2005) The study of urban sustainable development index system and evaluation
methods—resource-exhausted cities for examples. Res Fin Econ Issues 259(6):52–56
13 An Empirical Research 123
Li CH, Lu SC (2008) The study of resource-exhausted cities sustainable development of the
evaluation index system. Ind Sci Tribune 7(3):115–116
Marsh B (1987) Continuity and decline in the anthracite towns of Pennsylvania. Ann Assoc Am
Geogr 77(3):337–352
Pang ZQ, Wang BD (2012) Study on evaluation system of resource exhaustion regions’ economic
transformation. Stat Res 29(2):73–79
Randall JE, Ironside RG (1996) Communities on the edge: an economic geography of
resource-dependent communities in Canada. Can Geogr 40(1):17–35
Statistical yearbook of the various coal resource-exhausted cities from 2006 to 2011
Wang QY (2003) The study of resource-based cities with economies in transition. China
Economic Publishing Press, Beijing, pp 59–78
Wang J, Geng J (2012) Empirical study on resources exhausted city, stability of sustainable
development. Econ Probl 1:46–49
Wang Y, Lei GP, Liu ZJ, Yuan ZH (2010) Driving forces study on sustainable use of urban land
of coal resource-exhausted cities. Econ Geogr 30(7):1185–1188
Xia YX, Shen B (1998) The problems and countermeasures of resource development companies
and urban sustainable development in China. Chin Soft Sci 7:115–120
Xu XJ, Zhang J (2000) Resources industry in our country in the 21th century. Min Metall Eng
20(1):5–7
Zhao JH (2006) Overview of studies on urban development of resource-based city in China. Res
Dev Cities 6(3):86–91
Chapter 14
An Extensive Innovation Procedure
to Quality Function Deployment
for Product Design
Chang-tzuoh Wu, Jyh-rong Chou and Chang-shiann Wu
Abstract In this research, a systematic design method based on quality function
deployment and extension method has been developed. The major procedure of
QFD is to identify the customers’ needs for the product and then convert into
appropriate technical measures to fulfill the needs based on the company’s com-
petitive priorities. The priorities of product characteristics can be obtained by
translating important technical measures. According to their characteristics, the
prior engineering parameters will be identified and selected as the key require-
ments to redesign. This article will focus on the integration of QFD and extension
method. With the help of ‘‘matter-element theory and extension method’’, cus-
tomer requirements (CRs) can be transferred into product design attributes more
comprehensively and deeply. An innovative design case, bicycle, successfully
demonstrates that the proposed design process is feasible and efficient.
Keywords Extension method Innovative design Matter-element QFD
14.1 Introduction
A successful new product development should meet various customer demands.
Hence, analyzing customer requirements (the voice of customers) and responding
C. Wu (&)
Department of Industrial Design, National Kaohsiung
Normal University, Kaohsiung, Taiwan
e-mail: [email protected]
J. Chou
Department of Creative Product Design, I-Shou University,
Kaohsiung, Taiwan
C. Wu
Department of Information Management, National Formosa
University, Huwei, Taiwan
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 125
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_14,
Springer-Verlag Berlin Heidelberg 2013
126 C. Wu et al.
to their needs has now become an important and inevitable task for a company’s
product development team. (Liu 2009).
Quality Function Deployment (QFD) is an important quality control theory
proposed by the Japanese quality control masters Yoji Akao and Shigeru Mizuno.
It was Akao who first realized the value of this approach in 1969 and wanted to
utilize its power during the product design stage so that the product design
characteristics could be converted into precise quality control points in the man-
ufacturing quality control chart (Hill 1994); Sullivan (1986a, b) indicated that
Quality function deployment (QFD) provides a means of translating customer
requirements into the appropriate technical requirements for each stage of product
development.
In recent years, the correlative researches of application of QFD have been
developed rapidly and widely. As for the methodology, Eco-QFD, QFD integrated
with Kano Model and QFD integrated with FMEA have been proposed and
applied well.
Ernzer (2003) indicated that methodical support is necessary which integrates
environmental and market issues in a product. He proposed an integrated approach
called EI2QFD, which reduces the effort of carrying out an Eco-QFD. Addition-
ally, Ernzer proposed an effective and efficient application of eco-QFD. Kuo
(2005) introduced a fuzzy theoretic modeling approach to the Eco-QFD. The Eco
design product development problem was formulated as a fuzzy multi-objective
model based on the QFD planning. Chen (2009) proposed fuzzy nonlinear pro-
gramming models based on Kano’s concept to determine the fulfillment levels of
PCs with the aim of achieving the determined contribution levels of DRs in phase
1 for customer satisfaction. Besides, the failure modes and effects analysis
(FMEA) are incorporated into QFD.
Elif Kılıç (2009) indicated that in real world applications, the values of DRs are
often discrete instead of continuous. A new QFD optimization approach com-
bining MILP model and Kano model is suggested to acquire the optimized solution
from a limited number of alternative DRs, the values of which can be discrete.
Cecilia (2010) proposed a tool which utilized a combination of the Quality
Function Deployment (QFD)—Kano model to evaluate service quality. Relevant
information may be obtained about issues that should be improved in order to
increase customer satisfaction by listening to the voice of the customer (VOC).
Most of the associated researches provide valuable and valid methods to
enhance the effect of QFD. Though these studies applied successfully the proposed
methods to solve design problems, they focused on local perspective view only.
This article will concentrate on the creativity way of thinking ‘‘matter-element
theory and extension method’’, with the help of this method to assist to translate
customer requirements (CRs) into engineering characteristics (ECs) more deeply
and widely.
14 An Extensive Innovation Procedure to Quality Function 127
14.2 Innovative Design Method
14.2.1 Quality Function Deployment
QFD, introduced by, Akao 1972 was designed to improve quality in product
development. It has been a successful tool in assisting product designers sys-
tematically incorporate customer requirements (CRs) into product and process
development (Akao 1990). Specifically, QFD systematically brings customer’s
needs to the level of detailed operations. The American Supplier Institute’s (ASI’s)
Four-Phase approach was selected as the framework of this research. As shown in
Fig. 14.1, the Four-Phase approach consists of product planning, part deployment,
process planning, and production planning phases.
QFD is a structured design tool, and is defined as: ‘‘A consumers’ needs ori-
ented tool which establishes the relationship between customer attributes and
design parameters to be quantified through the House of Quality (House of
Quality), and integrated customers’ cognition, finding key design factors in the
product to determine the direction of product development and market position-
ing.’’ HOQ indicates the relationship between customer requirements (what to do)
and engineering characteristics (how to do it). It is the engine that drives the entire
QFD process. In essence, the product planning phase translates qualitative cus-
tomer requirements into measurable engineering characteristics, and identifies
important engineering characteristics. The part deployment phase translates the
output of the product planning into critical part characteristics and explores
the relationship between engineering characteristics and part characteristics. The
process planning phase establishes the relationship between part characteristics
and manufacturing operations related to a part. Critical process parameters are
identified and deployed in operation instructions. The production planning phase
translates the manufacturing operations into production standards or work
instructions, such as the number of parts to be checked, type of tools to be used, the
inspection method to be performed, etc. (Liu 2009).
Fig. 14.1 Four-phase approach of QFD
128 C. Wu et al.
The first phase of QFD, usually called house of quality (HOQ), is of funda-
mental and strategic importance in the QFD system, since it is in this phase that the
customer needs for the product are identified and then, incorporating the producing
company’s competitive priorities, converted into appropriate technical measures to
fulfill the needs. (Chen and Wu 2005) The goal of product planning phase is to
translate customer requirements (CRs) into engineering characteristics (ECs) and
prioritize their importance. Therefore, the CRs must be acquired from market
surveys or customer questionnaires. The acquired information can be used to
calculate the relative importance of CRs, calculate the final importance of CRs,
and identify the Ecs, so that the final importance of ECs will be calculated and the
relationship matrix also be established.
14.2.2 Matter-Element Analysis and Extension Method
In 1983, Cai developed a powerful tool, Matter-element and extension method, to
systematically analyze concrete or intangible products. Extension theory is a
course to study the extensibility, extent rules, performing procedure of matter and
try to employ to resolve contradictive problems. Extension theory and method are
the new science. Their application to the research on conceptual design of the
products and innovative design is a new field. The extension method can help
people resolve problems separately by decomposing and recombining the prob-
lems to search for the feasible solutions.
A matter-element can be combined with the other matter-elements to form a
new one, or be decomposed into a few new matter-elements; new matter-elements
contain qualities that former matter-elements do not have. The extension of matter-
element provides another way of resolving contradiction. Matter-element can be
used to describe every matter in real world. We use an ordered triad.
RðN; c; vÞ ð14:1Þ
The basic element for describing the matter N is called matter-element, where
N represents the matter; c the characteristic name; v is N measure about c. Thus,
v ¼ cðNÞ
The extensibility of matter-element is the basis of dealing with incompatible
problem. It includes divergence, expansibility, conjugate inside the matters and
relativity of matter-element. Divergence is to study the possible routes of outward
extension. It includes the same matter of matter-element, the same characteristics
of matter-element, the measure of matter-element, the same matter and charac-
teristics of matter-element, the same characteristics and measure of matter-ele-
ment, the same matter and measure of matter-element. Extensibility studies
plausibility, integration and separability of matter-element.
Replacement, decomposition, addition/deletion and expansion/contraction are
four basic methods for transformation of matter-element and to be conducive to
14 An Extensive Innovation Procedure to Quality Function 129
Fig. 14.2 General model of extending tree
exchanging or to synthesizing different matter-element. As shown in Fig. 14.2,
general model of extending tree constructed by matter-element symbols expresses
the divergence in simple way.
Study of the internal structure of matter is contributive to solve contradictive
problems by using any element of matters. Systems theory, a description of
structures of matters, is a science to study matters from system components and
internal-external relationships. Through the analysis of a large number of matters,
we found that, the properties of substance, dynamic and opposition can also be the
research topics of the structure of matters in addition to systemic. A matter consists
of the tangible part and intangible part of substance, both of which should also be
considered. As for the dynamic properties, matter has the significant part and
potential part. Such as the development of the disease, it generally includes the
incubation period and the exacerbation period.
The structure of matter can be described more completely and the nature of
variation of the development of matters can also be revealed more deeply while we
explore the matter from the point of view of systemic, substance, dynamic and
opposition. Therefore, we proposed corresponding four pairs of opposing concepts
such as real/imaginary, soft/hard, potential/significant and positive/negative to
describe the composition of matters called conjugate of matters.
14.3 Extensible QFD
Customer needs-orientation is the principal spirit of QFD. The conversion between
functional quality and engineering design parameters is the important creativ-
ity thinking course for the fulfillment of the customer needs-oriented innovative
design.
130 C. Wu et al.
The major procedure of QFD is to identify the customers’ needs for the product
and then convert into appropriate technical measures to fulfill the needs based on
the company’s competitive priorities. The priorities of product characteristics can
be obtained by translating important technical measures. According to their
characteristics, the prior engineering parameters will be identified and selected as
the key requirements to redesign. Throughout the procedure, the creativity activ-
ities with matter-element theory and extension method can be imported or
expanded mainly in two parts,
(1) The course of translating customers’ needs into product design attributes
(technical measures).
(2) The course of identifying corresponding product defect or the parameters
needed to be improved.
The use of extensibility of matter-element and extension method for the
translation of product design attributes and engineering properties can provide
effective assistance for the designer to conceive new products comprehensively
and deeply. In this research, we will discuss the procedure for improving design
attributes by the aids of extensibility of matter-element and extension method.
Matter-element and matter-element with multi-characteristics are defined as
follows,matter-element
R ¼ ðN ðtÞ; c; vðtÞÞ ð14:2Þ
matter-element with multi-characteristics
2 3
N ðtÞ c1 v1 ðtÞ
6 c2 v2 ðtÞ 7
6 7
RðtÞ ¼ 6 .. .. 7 ¼ ðN ðtÞ; C; V ðtÞÞ ð14:3Þ
4 . . 5
cn vn ðtÞ
Based on the divergence of matter-element, matter-element R0 ðN0 ; c0 ; v0 Þ can be
diverged from one or two of N0, c0, v0 to synthesize different matter-elements, and
build an extending tree. Extending tree is a method for matter to extend outwards
to provide multi-orientated, organizational and structural considerations, as oper-
ating in Fig. 14.2. An event is the interaction of matters and described as event-
element. Basic elements for describing an event-element are constructed by verb
(d), name of verb characteristic (b) and u, the corresponding measure about (b).
Event-element
IðtÞ ¼ ðdðtÞ; b; uðtÞÞ ð14:4Þ
Multi-dimensional event-element
IðtÞ ¼ ðdðtÞ; B; UðtÞÞ ð14:5Þ
14 An Extensive Innovation Procedure to Quality Function 131
Relationship-element is form by relationship name s(t), characteristics
a1 ; a2 ; ; an and corresponding measure values w1 ðtÞ; w2 ðtÞ; . . .; wn ðtÞ:
2 3
Sð t Þ a1 w1 ðtÞ
6 a2 w2 ðtÞ 7
6 7
Q ðt Þ ¼ 6 .. .. 7 ¼ ðSðtÞ; A; W ðtÞÞ ð14:6Þ
4 . . 5
an wn ðtÞ
While solving design problems based on Su-field model, diversity and creativity,
the advantages of ‘‘extenics’’ method, will be imported by implementing the
extensibility of matter-element, event-element and relationship-element. Thus, the
solution will not be limited to standard solutions but be inspired.
14.3.1 Procedure for Improving Product Design Attributes
Throughout the procedure of QFD, the extensibility of matter-element and
extension method can be introduced into the course of improving product design
attributes (technical measures) based on product defects found from customers’
voice. Below are the steps of creative activities for matter-element theory and
extension method to be imported or expanded.
Step 1: express the existing products N into an n-dimensional matter-element
Step 2: Obtain the products’ functional defects or the shortcomings lead to
customer dissatisfaction by the customer voice (complain). Define the sub-matter-
elements Ri (i = 1 * m) corresponding to functional defects and rearrange as
2 ð1Þ ð1Þ
3
c1 v1
6 ð1Þ ð1Þ 7
6N c2 v2 7
6 7
6 .. .. 7
6 . . 7
R¼6 ð1Þ ð1Þ 7 ð14:7Þ
6 c v m 7
6 m 7
6 .. .. 7
4 . . 5
ð1Þ ð1Þ
cn vn
Step 3: Transformation on the defects sub-matter-elements Ri (i = 1,2,…,m).
Treat the replacement, addition/deletion, expansion/contraction transformation to
the iso-matter-element but the addition/deletion, expansion/contraction transfor-
mation to the distinct matter-element. That is, either separately or integrated
transform sub-matter-elements for each shortcoming Ri (i = 1,2,…,m)
Ti Ri ¼ Ri ði ¼ 1; 2; . . .; mÞ ð14:8Þ
132 C. Wu et al.
After performing the matter element transformation, a new n-dimensional
matter-element R* which contained the revised sub-matter-elements will be
occurred. The revised sub-matter-elements obtained by transforming the matter-
element with shortcomings to a better one. Corresponding to the elements of R, a
sequence of new products will occur.
14.4 Illustrative Design Case
In this research, design case ‘‘Bicycle’’, the most favorite exercise equipment and
transport is adapt to explain and verify feasibility of the proposed innovative
procedure. The proposed approach, QFD integrated with extension method, has
been used to investigate the users’ needs and relative functional requirements. In
the creative course, we make use of the extension method to improve product
design attributes and to determine the associated engineering design parameters.
The creative solution programs based on customers’ voice will thus be achieved.
14.4.1 Cooperated with Extenics
Two parts are in this stage of research. First, obtain the needs of bicycle by
interviewing the elderly and translate to product design attributes, and then in the
second part, the questionnaire checks for the elderly to pick the important
requirements.
In this study, survey candidates are the cycling sport amateurs, and the cyclists
are included to avoid some problems are ignored. Gathering respondents’ views,
users’ requirements are sorted out as Table 14.1.
Use extending tree method to assist the establishment of HOQ. First, we build
the transformation model by extending tree: A matter refers to multi-characteris-
tics; a characteristic is also mapped by matters. Thus, ‘‘one matter with multi-
characteristics’’, ‘‘one characteristic maps to multi-matters’’, ‘‘one value maps to
multi-matters’’, ‘‘one matter, one value versus multi-characteristics’’, ‘‘one matter,
one characteristic versus multi-values’’ are the extending propositions of matter-
element to resolve contradiction.
Table 14.1 Requirements of bicycle for the sport amateures
Portability operating experience Sturdy and durable
Overall riding comfortable Additional functions
Cost/price Attractive appearance
Flexibility to adjust
14 An Extensive Innovation Procedure to Quality Function 133
2 3
Overall riding comfortable comfortable grip
6 anti-shock 7
6 7
R1 ¼ 6 7
4 comfortable saddle 5
ventilatory saddle
2 3
Portability operating
6 Light weight 7
6 experience 7
6 7
6 Breaking effort 7
6 7
R2 ¼ 6 7
6 Operatederailleur smoothly 7
6 7
6 Chain - link come off the ratchet 7
4 5
and slip gears
Flexible adjustment Easy to adjust seat post
R3 ¼
Quick detach
…etc.
We can translate product functional requirements into product design attributes
as shown in Table 14.2.
We derive and confirm the ‘‘hypostatic characteristics’’ from ‘‘functional
characteristics’’ and then the corresponding ‘‘certain characteristics’’ will also be
derived and confirmed. Then, we use the corresponding four pairs’ concepts such
as replacement, addition/deletion, expansion/contraction transformation to the iso-
matter-element but the addition/deletion, expansion/contraction transformation to
Table 14.2 Product design attributes
Level 1 Level 2 Level 3
Basic function Specifications Simplify parts
Parts design for assembly
Weight
Size
Mechanism Driving mechanism
Breaking mechanism
Derailleur mechanism
Safety Material of frame
Maintenance
Breaking performance
Structural strength
Ergonomic Comfort Heat dissipation of saddle
Suspension performance
Body dimensions
Handling Derailleur maneuverability
Flexible adjustment
Brake system
134 C. Wu et al.
the distinct matter-element to describe the composition of matters and assist the
transformation and extension of matter-elements.
Take the design attribute, driving mechanism, for example. One of the Cus-
tomers’ complaints is that a traditional bicycle uses a chain for transferring the
pedal actuated driving force from the pedal crankshaft to the rear wheel. After
using period of time, the chain-link is prone to come off the ratchet and slip gears.
In addition, the long skirt also has the risk of being sullied or rolled into chain link.
The extension of relationship-element of driving mechanism
2 3
Transmission Front gear set Chain
6 Ratchet and slip gears Chain 7
6 7
6 Drive method Rotation 7
6 7
6 Bad effect Come off rear gear 7
6 7
Q¼6 Structure 7
6 7
6 Safety degree Poor 7
6 7
6 Disadvantages Sully 7
4 5
.. ..
. .
¼ ðs, A, WÞ
We consider making the ‘‘rigid transmission device’’ play the major role of
bicycle devices and then, design problem becomes ‘‘How to transform the flexible
device into a rigid device?’’ This leads to new design problems based on the new
concept. Repeat matter-element extending program. Transform the flexible device
into a new general model by the divergence tree. Therefore, restart the extending
processes by taking a new matter ‘‘rigid transmission device’’ and a new char-
acteristic ‘‘contact type’’ to set up matter-element, event-element, and relationship-
element and extending tree. Many ideas will be obtained. Therefore, a new device,
bevel gear set, can meet the demand. We may consider using ‘‘bevel gear set’’ to
replace ‘‘chain’’. We propose changing the transmission type. Bevel gear mech-
anism will be the new design concept we proposed.
14.5 Conclusions
This study proposes an innovative design and problem-solving process, based on
quality function deployment method with the assistance of extension of matter-
element.
In this research, we assess possibility and advantages to combine quality
function deployment with the extension of matter-elements. The procedures for
improving product design attributes and engineering design parameters transfor-
mation are also proposed. Furthermore, an innovative design case, bicycle, suc-
cessfully demonstrates that the proposed design process is feasible and efficient.
14 An Extensive Innovation Procedure to Quality Function 135
Acknowledgments This work is supported by the National Science Council, Taiwan, and No.
NSC 95-2221-E-017-006-
References
Akao Y (1972) New product development and quality assurance deployment system, (in
Japanese). Stand Qual Control 25(4):243–246
Akao Y (1990) Quality function deployment: integrating customer requirements into product
design. Productivity press, Cambridge
Chan LK, Wu ML (2005) A systematic approach to quality function deployment with a full
illustrative example. Int J Manage Sci Omega 33:119–139
Chen LH, Ko WC (2009) Fuzzy approaches to quality function deployment for new product
design. Fuzzy Sets Syst 160(18):2620–2639
Ernzer M (2003) EI2QFD—an integrated QFD approach or from the results of eco-indicator 99 to
quality function deployment. In: Proceedings on environmentally conscious design and
inverse manufacturing, pp 763–770
Garibay Cecilia, Gutiérrez Humberto, Figueroa Arturo (2010) Evaluation of a digital library by
means of quality function deployment (QFD) and the Kano model. J Acad Librarianship
36(2):125–132
Hauser JR, Clausing D (1988) The house of quality. Harvard Bus Rev 66(3):63–73
Hill A (1994) Quality function deployment. In: Lock D (ed) Gower handbook of quality
management, 2nd edn. Gover, Brookfield, pp 364–386 (Chapter 21)
Kuo TC, Wu HH (2005) Fuzzy eco-design product development by using quality function
deployment. In: Proceedings on environmentally conscious design and inverse manufacturing,
eco design, 4th international symposium, pp 422–429
Lee YC, Sheu LC, Tsou YG (2008) Quality function deployment implementation based on Fuzzy
Kano model: an application in PLM system. Comput Ind Eng 55(1):48–63
Liu HT (2009) The extension of fuzzy QFD: from product planning to part deployment. Expert
Syst Appl 36:11131–11144
Sullivan LP (1986a) The seven stages in company-wide quality control. Quality Prog
19(5):77–83
Sullivan LP (1986b) Quality function deployment. Quality Prog 19(6):39–50
Chapter 15
Analysis on Grey Relation of Labor
Export Mechanism’s Influence Factors
in Poverty-Stricken Areas
Shan-ping Wang and Yi Zhou
Abstract Aiming at how to improve the labor export at the new period of poverty
alleviation work of China, this paper first analyzed the influence factors of labor
export mechanism in our country’s poverty-stricken areas theoretically, then
selected the related influence indexes, chose the poverty alleviation counties as the
research object and used the Grey Correlation Analysis, empirically analyzed each
index’s influence size and order, so as to provide basis of related policy for further
improving the poverty alleviation mechanism of labor export.
Keywords Grey correlation analysis Labor export Poverty alleviation
15.1 Introduction
From the ‘‘The Seven-year Priority Poverty Reduction Program’’, ‘‘Outline for
Poverty Reduction and Development of China’s Rural Areas (2001–2010)’’, to the
issue of the new round of national poverty alleviation tasks and implement of
‘‘Outline for Development-oriented Poverty Reduction for China’s Rural Areas
(2011–2020)’’, all what can see is the government’s spirit of encouraging poverty-
The research is supported by the New Century Training Program Foundation for the Talents of
the Ministry of Education (NET06-0703) and by National Social Science Foundation
(08BJY025).
S. Wang (&) Y. Zhou
School of Business Administration, Hunan University, Changsha, China
e-mail: [email protected]
Y. Zhou
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 137
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_15,
Springer-Verlag Berlin Heidelberg 2013
138 S. Wang and Y. Zhou
stricken areas to export their labor force, so as to help the poor people get
employment in non-agricultural field and then increase their income to finally
overcome poverty. However a poverty policy wants to get implementation and
achieve the expected effects, it needs to meet the basic conditions: the poor people
have the opportunity and ability to benefit from it. But the influence factors of
labor export in poverty-stricken areas are various, just depending on poverty
alleviation strategy or poverty relief funds are not enough to make them realize the
goal successfully and get self-development in the constraint condition. In this case,
the poverty relief investment’s performance to the mechanism is very low, budget
constraints and the requirements of improving efficiency are also not allow the
funds’ increasingly growth. Policy’s limitation highlights the urgency of changing
the focus of poverty strategy. What must be realized is the labor export is not a
simple process of transferring the poverty-stricken areas’ labor force to others; it
needs to have a comprehensive understanding of each affecting factor, and then
choose corresponding policy tools. Only in this way, in a new period of poverty
alleviation and development, the export of labor can be more specific, and its
effects can also be more enduring.
Based on this, this paper collect 2002–2010s related data, and using the Grey
Correlation Analysis method (GCA) to study the influence factors of poverty-
stricken areas’ labor export mechanism, calculating the correlation degree and
sorting, so as to seek for the major factors of this mechanism and make accurate
evaluation.
15.2 Related Literatures Review
Labor export is a poverty alleviation mechanism that points to provide short-term
vocational training to improve laborers’ skills from poverty-stricken areas, after
the training, the labor transfer to town or other developed regions to seek for
higher income employment opportunities, so as to improve the family’s income
level. Existing studies have confirmed, carrying out the export of labor is one of
the main effective poverty alleviation way. Through the flow of labor force, this
mechanism can make reasonable use of labor resources, and it is an important
channel of developing the rural human capital. Furthermore, China’s rural labor
flow’s impact on the farmers’ income increase, labor force optimal allocation and
regional harmonious development is also of great significance. Based on the active
effects from rural labor transfer, the study of the poverty-stricken areas’ labor
export accords to the similar research paradigm. Among these studies, the motive,
barriers of the labor are embodied in the following aspects.
A. Institutional variables, especially the development strategy and the household
registration system’s impact. Cai (2001) research on institutional influence to
rural laborers’ decision-making shows that, the traditional development strat-
egy and household registration system limit potential transfer, and institute
15 Analysis on Grey Relation of Labor Export Mechanism’s Influence Factors 139
reform has eased the restrictions. But the slow process of reform, make labor
flow free is still filled with obstacles and would be a long-term standing
problem. In this case, the explanation of expected income and human capital
for labor transferring is not sufficient. Li (2003) regards the institutional factors
have great influence on the China’s rural labor transfer. Improvement of the
household registration system will promote smooth migrant of the peasant-
worker. Li (2007) establishes rural labor force transformation model which
includes institutional factors, and comes to the similar conclusions.
B. Economic factors’ impacts on the rural labor force transformation. In Chen and
Hu labor transfer analysis framework, the town’s informal employment
department is introduced to explain labor transfer’s causes and obstacles,
observe the theory and practice, and study the transfer strategy and path of
China’s rural surplus labor, thus forming the widely accepted ‘‘Triangle Eco-
nomic Theory’’. Du (1997) observes the China’s rural laborers’ transfer
behavior from the perspective of income or resources, finds that the low
income of farming and shortage of agricultural resource are the main reasons of
rural labor force’s shifting. Bai and Gan’s (2005) paper investigates the Jiangxi
rural labor force transformation’s dynamics and obstacles by establishing the
dynamic game model, and finds that the primary reason of rural labor transfer
is economic income factors.
C. Discuss physical factors (include age, gender, health etc.) and mental factors
(include education degree and non-agricultural working experience), and other
demographic characteristics that effect the rural laborers’ transfer decision-
making. The results of Li (1999) study show that, the household’s education is
proportional to the possibility of labor flow, the higher the household’s education
degree is, the more likely to make the decision of family migration. Zhao con-
siders the influence of formal education on labor transfer decision-making is very
small, as to the labor force’s shift from agriculture to non-agricultural areas, the
influence is significant. Cheng and Shi’s research confirms this conclusion.
In addition, as to the labor from the poverty-stricken areas, they are not only
influenced by these three common factors as the national rural labors are, but also
facing another important influence factor—the government’s poverty alleviation
activities to labor export mechanism. As Wang (2004) study shows that, if the
poverty relief funds of training skills be combined with the labor export, it can put
a more significant effect on the farmers of the poverty-stricken areas and can be
more easily accepted. And he points out that the government should be given more
autonomy and flexibility in choosing specific projects, so that the projects can take
more farmers’ needs into consideration. Through analyzing the changes of China’s
rural poverty properties, Du thinks the poor community which has more migration
probability, the government should provide them the basic resources for migration
to ensure them can make full use of the outside employment opportunities to
improve their own welfare. Li (2007) tries to investigate the way of poverty
alleviation, finally come to the conclusion that the existing method can’t adapt to
140 S. Wang and Y. Zhou
China’s the complexity of rural poverty problems and diversity of peasant
households’ demands, it decreases the accuracy of poverty relief funds. Fang
(2007), based on the theory of ability poverty, points out that the effective poverty
alleviation strategy should develop and enhance the poverty population’s risk
coping ability.
Reviewing the literatures above, we can see the poverty-stricken areas’ influ-
ence factors of labor export mechanism, should be considered from the aspects of
institute, economic, self-own and poverty alleviation activities, in order to examine
whether the existing poverty alleviation way correspond to the factors or not. As to
the research methods about the mechanism, most existing ones deduce the factors
theoretically, but lack of the accurate result of the factors’ contribution degree or
ranking, and empirical evidence is mostly from regression analysis and time series
analysis. However, if using the same empirical method to study the problems of
poverty-stricken areas’ labor export, the problem lies in, the lack of data. Con-
sidering the history of our country’s poor monitoring is still short, so the available
sample data is very poor. But the sample requirement of regression analysis and
time series analysis are at least 30, or the accuracy of analysis cannot be guar-
anteed. In addition, the adjustment of the poverty line, the economic condition’
variation and other external factors, make the sample cannot meet the requirements
of distribution conditions of traditional quantitative analysis. So it is necessary to
adopt new analysis method in exploring poverty-stricken areas’ labor export
mechanism.
15.3 Grey Correlation Analysis of the National Poverty
Alleviation Counties’s Labor Export
As a subsystem of the society, poverty-stricken areas’ poverty problem is caused
by social, economic, nature, history, etc. Combining the analysis above, we can see
that the influence factors related with it is complex, and these factors constitute a
network of relationships, of which there are many feedback loop. It is because of
the relationship between the influence factors are not clear, so the labor export
mechanism has gray system’s characteristics. In this case, adopting the traditional
regression analysis is difficult to achieve the expected effect, but the Grey Cor-
relation Analysis method can compensate for it. The method has no strict
requirement for samples and distribution, and is very convenient; it generally
won’t appear to be the situation that the quantitative result does not accord with the
qualitative analysis. And the method can find out the advantage and disadvantage
factors, or the leading and potential factors for the development of the system, and
judge them. So this paper uses the Grey Correlation Analysis method to empiri-
cally analyze the influence factors’ size and order of the mechanism.
15 Analysis on Grey Relation of Labor Export Mechanism’s Influence Factors 141
15.3.1 Model Specification
This paper’s research object is to study the influence factors’ impact on poverty-
stricken areas’ labor export. The Grey Correlation Analysis is based on grey
correlation degree to analyze the relationship of lord behavior factor and the
relevant behavior factors in the grey system, and then judge the main factors and
secondary factors for the development of the system according to similar degree of
the curves’ geometry. The closer the curves are, the greater correlation between
the corresponding sequences is Deng (1993). If the grey correlation degree
between influence factors and the Labor export index is deeper, then the factor’s
impact on this mechanism is greater. Thus provide policy guiding direction to
promote smoother labor export in the new period of poverty alleviation work.
The basic steps of the grey relation analysis are:
Firstly, confirm the reference sequence that reflects system lord behavior
characteristics and the comparison sequences that influent the system behavior.
Set system behavior sequence and comparison sequences:
i 2 N; N ¼ 0; 1; 2; . . .; ; m = 2; yi ¼ ðyi ð1Þ; yi ð2Þ; . . .; yi ðnÞÞ;
Y ¼ yi
yi ðkÞ 2 yi ; k 2 K; K ¼ ð1; 2; . . .; nÞ; n = 3
Among them: y0(k), (k = 1, 2,…, n) is the reference sequence; yi(k), (i = 1,
2,…, m, k = 1, 2,…, n) are the comparison sequences.
Secondly, the reference sequence y0 and comparison sequences yi are converted
into proper dimensionless indexes in order to uniform dimension, making the
factors comparable. This paper uses the equalization method to eliminate the
dimension:
1X n
y ð kÞ 1X n
y ð kÞ
y0 ¼ y0 ðkÞ; y00 ¼ 0 yi ¼ y0 ðkÞ; y0i ¼ i
n k¼1 0 y0 n k¼1 i yi
(k = 1,2, …, n, i = 1,2,…,m)
Thirdly, calculate the correlation coefficients n0i ðkÞ of the reference sequence
and the comparison sequences to assess the correlation between them:
maxi maxk jy0 ðkÞ yi ðkÞj þ f maxi maxk jy0 ðkÞ yi ðkÞj
n0i ðkÞ ¼
jy0 ðkÞ yi ðkÞj þ f maxi maxk jy0 ðkÞ yi ðkÞj
Among them, f, is the distinguish coefficient, its value range is (0, 1), this paper
take f = 0.5.
Fourthly, calculate the grey correlation degree c0i. Due to the correlation
coefficients are correlation degrees at different times, so we need to prevent from
data too much to compare. It is necessary to average the correlation coefficients.
The formula of grey correlation degree is:
142 S. Wang and Y. Zhou
1X n
c0i ¼ f ðkÞ ðk = 1,2,. . .; n, i = 1,2,. . .; mÞ
n k¼1 i
Fifthly, sort the grey correlation degree. The correlation’s sequence between
factors is mainly used by the grey correlation degree to describe. Sorting the c0i,
can reflect comparison sequences’ impact on reference sequence.
15.3.2 Index Selection and Processing
Through the review of existing literatures, it can be seen that labor export is
restricted by many factors, but considering the analysis method’s feasibility, this
paper, on the basis of existing researches, using the Grey Correlation Analysis
method and combining with the ‘‘Chinese rural poverty monitoring reports’’,
‘‘China Statistical Yearbook’’ and ‘‘Chinese labor statistics yearbook’’, mainly
from the aspects of institute, economic, self-own and poverty alleviation activities
to empirically analyze influence factors of poverty-stricken areas’ labor export.
Since 2002, the State Council determined the 592 national poverty alleviation
work counties (i.e. counties) again according to ‘‘631 index method’’. So con-
sidering the continuity of the statistical data, this paper selects the time series data
of 2002–2010 years for the model analysis.
Y0 is set to represent the counties’ number of export labor who obtains the
employment outside the counties for more than a month. According to the analysis
above, the influence factors are classified into 4 kinds:
• Institutional factor (y1). This paper uses Chen (1999), Jin’s (2001) researches for
reference and selects the following indexes: the private sector employment rate
(y11), namely the non-state sector employment’s proportion of total town
employment; The urbanization rate (y12), that is, the proportion of the urban
population in total population, to measure the impacts of the institutional factors
including economic component change and the household registration system
change on the counties’ labor export respectively. In addition, institutional
changes’ consequence is the proportion of market distributes labor resources
gets bigger. Generally speaking, high degree of marketization of the country, the
proportion of government allocating labor resources is usually low. So this
paper adopts the proportion of GDP distributed by the market to approximate
reflect the Proportion of Market Allocate Resources index (y13), its measure-
ment formula is: (GDP–national finance income)/GDP (national finance income
without debt revenue).
• Economic factor (y2). Because the income gap between urban and rural areas
(y21) reflect the attraction for the rural labor transfer, the proportion of the
secondary and tertiary industry in GDP (y22) reflects the modern industrial and
service sectors’ output, and for the rural surplus labor’s transfer, the secondary
15 Analysis on Grey Relation of Labor Export Mechanism’s Influence Factors 143
and tertiary industry provide them with the huge absorb space, it is an important
way for rural labor to get non-agriculture employment; registered urban
unemployment rate (y23) represents the supply and demand situation of the
urban labor market. So we take these three as the economic factor indexes to
review.
• Self-own factor (y3), The rural human capital stock is the comprehensive
reflection of the rural labor force’s culture quality, so using the Rural Human
Capital Stock (y31) to measure the overall Labor’s culture quality of the
counties, its computation formula is:
X
4
y31 ¼ Qi hi
1
n represents the workforce of the county, Qi is the proportion of rural labor force
with different education degree in total labor force, hi is the Education
conversion coefficients.1 Considering the data’s accessibility, we classify the
rural labor’s education degree into four kinds, namely, illiterate or semiliterate,
elementary school, junior high school, high school and above.
Due to the First Industry Productivity (y32), namely the ration of county’s first
industry output value with number of the first industry employees, and Rural
Industrialization Rate (y33), namely the proportion of the counties’ non-
agricultural labor force in the total labor, the two indexes represent the labor’s
transfer ability from the first industry in the counties, so we choose them to
represent self-own factor indexes.
• Poverty alleviation factor (y4). As for the poverty alleviation factors, this paper
investigates from the following three aspects: poverty relief funds (y41), namely
the amount of poverty relief funds each year; the coverage of Poverty alleviation
projects (y42), namely that the proportion of poor village that participate in the in
poverty alleviation projects; Farmers’ participation (y43) in poverty alleviation
activity, using the proportion of farmers who have been funded by the poverty
alleviation projects to measure.
15.3.3 Empirical Analysis
This paper use the GTMS3.0 software to process the initial data above, then get the
Grey Correlation Degree (GCD) of each single influence index, the result is as
Table 15.1 shows.
1
For the education conversion coefficient, this paper refer Li’s method, assume that illiterate or
semiliterate is 1, the primary school is 1.1, junior high school is 1.2, high school and high school
above is 1.5.
144 S. Wang and Y. Zhou
Table 15.1 Grey correlation degree of the counties’ labor export influence indexes (2002–2010)
Labor export number
Influence indexes GCD/single GCD/mean
Institutional factor (y1) Y11 0.7129(6) 0.7025(2)
Y12 0.7393(4)
Y13 0.6552(12)
Economic factor (y2) Y21 0.6888(7) 0.6830(4)
Y22 0.6813(9)
Y23 0.6788(10)
Self-own factor(y3) Y31 0.6769(11) 0.6957(3)
Y32 0.6828(8)
Y33 0.7274(5)
Poverty alleviation factor (y4) Y41 0.7805(3) 0.8139(1)
Y42 0.8273(2)
Y43 0.8340(1)
According to the Grey Correlation Analysis results, we can see that impacts of
the influence factors on the counties’ labor export are all relatively obvious, the
sequence is: y43 [ y42 [ y41 [ y12 [ y33 [ y11 [ y21 [ y32 [ y22 [ y23 [ y31 [
y13. The result shows grey correlation degree of the poverty alleviation factor is
the biggest, an average of 0.8139, the institutional factor takes the second place,
the economic factors’ is the minimum; As to the single factor, the coverage of
poverty alleviation projects and farmers’ participation in poverty alleviation
activity are of great influence, the grey correlation degree of them are both above
0.8, while the influence of the Proportion of Market Allocate Resources, rural
human capital stock and the town unemployment rate town influence are relatively
small, are 0.6552, 0.6788, 0.6788 respectively. Therefore, we can see:
(1) The poverty relief activities (0.8139) is the most important factor for the
counties’ smooth labor export, especially the labors’ participation in poverty
alleviation activity (0.8273) and projects’ coverage (0.8340). It shows that the
poverty relief projects and related funds to this mechanism are closely correlated.
This can be explained from the existing problems of present poverty alleviation
mode and labor export mechanism.
In the poor counties which with few outworkers, poor information and surplus
labor forces, the peasant lack adequate capital to pay for the initial cost of
transferring, and the majority of them, the transfer is originated from spontaneous.
In this case, only with the government’s involvement, organize and guide the labor
a proper way to transfer, associated with training them corresponding public
knowledge and skills and improving their participation in these activities, then
provide them appropriate initial funds, such as education and training funds or
interest loans, will the ability of labor forces to win non-agriculture employment
get effectively improved, and so will the resistance of the transfer channel
decrease. In addition, considering our country’s present poverty alleviation mode,
it just aims at the poor county, not the real poor people. After the key counties have
15 Analysis on Grey Relation of Labor Export Mechanism’s Influence Factors 145
been determined, resources will be given. But these resources are mainly used in
infrastructure projects, so that it cannot have wide coverage and the labor force
cannot benefit from it directly. Therefore, if the pointing accuracy of individual
increases, and making project cover more people, the poverty relief resources will
be more accurately and timelier transmit to the poor labor force, and make the
labor export mechanism more effective.
(2) The institutional factor (0.7025) is still one of the dominate factors that
affect the counties’ labor transfer. The institutional factor, including urbanization
rate (0.7393) and private sector employment rate (0.7129), raise the threshold of
labor export. The grey relation analysis result also concomitants with the present
situation of our country’s rural labor transfer. because of discriminations from the
census register system and other related institutional factors, the cost of labor
export increases invisibly, making the labor force pay more transaction cost to get
a non-agricultural job; On the other hand, the discriminations also make the labor
forces face the situation that trading places are not fixed and there are no contracts,
they can only seek job in the subprime market, hardly get a long-term stable
employment in the town, result in most of the labor forces wander between towns
and poor areas helplessly.
(3) In the self-own factor (0.6957) and economic factor (0.683), the counties’
proportion of non-agricultural production value (0.7274) and the income gap
between towns and countries (0.6888), are also important to the labor export
mechanism. This shows that the development of the counties’ non-agricultural
industries, the adjustment of economic structure, change of economic growth, and
actively promoting agricultural industrialization and the third industry in the
counties, will release labor force. As to the income gap between urban and rural, it
also can promote the transfer to some extent. What worth noting is, the rural human
capital which has been widely recognized as one of the key factors that affect rural
labor flow, however, in this study, its grey relational grade only gets 0.6769, ranking
No. 11. The reason for this situation may lie in the education pattern in poor counties,
it almost copy the city’ mode, just pay attention to the ordinary education, but ignore
the importance of vocational skills, result in one’s education does not fit him for a
certain job and low effectiveness of the education investment. So the deficiency and
dislocation of poor areas’ human capital make the gray relational grade between
rural human capital stock and labor export is small.
15.4 Conclusion and Suggestions
This paper has empirically researched the influence factors’ impact on the coun-
ties’ labor export base on the Grey Correlation Analysis method, the result shows
that the government’s poverty relief activities have positive impacts on the
counties’ labor export; these positive impacts come into effect mainly through
improving labor’s participation of poverty relief programs, programs’ coverage
and increasing the investment. This means that, increasing poverty alleviation
146 S. Wang and Y. Zhou
investment is necessary but not enough, it is more important to improve the
poverty alleviation projects’ pertinence and coverage, so that more poor labor
force can benefit from the activities, therefore changing their families’ economic
conditions through labor export. In addition, the empirical result also shows that,
for poverty-stricken areas’ labor force, whether they can smoothly export or not,
still related with the institutional and economical factors, such as the household
registration system, economic composition, the income gap and human capital
stock, etc.
In conclusion, it is the national macro institutional arrangements, policies and
regulations, and economic changes that have a profound influence on the poverty-
stricken areas’ labor export. The influence may be direct and clear, or indirect and
potential. Under these constraints, if the poor labor forces want to change their
social economic status through labor export, the government must leave spaces
and opportunities for them to realize it, or promote this change as s driving force of
the outside world. And these measures should by no means just by increasing the
poverty alleviation funds (though its importance is obvious), but should be com-
bined with improvement of poverty relief programs’ coverage and pertinence;
institution reform, especially the household registration system; improvement of
the income gap between urban and rural; human capital investment mode reform
of the poor areas. All in all, only the macro system reform, policy adjustment and
micro poverty activities combined together, implementing more comprehensive
poverty alleviation policies, will China’s poor labor forces indeed benefit from this
labor export mechanism in the new period and eventually get out of poverty.
References
Bai YT, Gan XW (2005) The dynamic model analysis of labor transfer in Jiangxi province.
Enterp Econ (7)
Cai F (2001) The way of China’s population flow. Social Sciences Academic Press, Beijing
Chen ZS (1999) The research of China’s Marketization process. Shanghai People’s Publishing
House, Shanghai
Deng J (1993) Gray control system, 2nd edn. Huazhong University of Science and Technology
Press, Wuhan, pp 310–318
Du Y, Bai N (1997) The transfer of Chinese rural labor and the empirical research. Economic
Science Press, Beijing
Fang LM, Zhang XL (2007) The analysis on the effect of China’s rural poverty reducing policy
based on the theory of poverty-ability inspection. J Finan Econ (2):48–56
Jin YG (2001) The contribution of macro system changes to China’s economic growth in the
transition period. Sci Finan Econ 2:24–28
Li S (1999) China’s rural labor flow & income growth and distribution. Soc Sci China (2):16–33
Li P (2003) Migrant workers: social and economic analysis on chinese migrant workers. Social
Sciences Academic Press, Beijing
Li XY, Tang LX, Zhang XM (2007) The analysis on the input mechanism of poverty alleviation
funds. Probl Agric Econ (10):77–81
Wang S, Li W, Li Y (2004) The analysis on the focus and effect of China’s poverty alleviation
funds. J Agrotech Econ 5:45–49
Chapter 16
Analysis on Trend of Research
and Development Intensity in China
Bin Huang and Lu-cheng Huang
Abstract Research and development (R&D) investment is an important index to
measure the R&D activity scale and evaluate the innovation ability. Based on
review of the theory of R&D intensity from scholars at home and abroad, the paper
analyzes R&D intensity trend in China. GM (1, 1) model is used to simulate and
predict future trend of China R&D intensity.
Keywords China GM (1, 1) model Gray predication R&D intensity
16.1 Introduction
Research and development (R&D) intensity is defined as the ratio between R&D
expenditures and GDP (OECD 2011a). R&D intensity is used as an indicator of an
economy’s relative degree of investment in generating new knowledge. Several
countries have adopted ‘‘targets’’ for this indicator to help focus policy decisions
and public funding (OECD 2011b). The government intends to have R&D
intensity reach 2.2 % by 2015 according to the Twelfth Five-Year Plan for
National Economic and Social Development of the People’s Republic of China
(English section of the central document translation department 2011).
16.2 Research on R&D Intensity at Home and Abroad
The Human Development Report shows that R&D intensity has strong positive
correlation with logarithmic per capita GDP, and R&D intensity of high-income
B. Huang (&) L. Huang
School of Economics and Management, Beijing University of Technology, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 149
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_16,
Springer-Verlag Berlin Heidelberg 2013
150 B. Huang and L. Huang
countries is much higher than that of low-income countries (United Nations
Development Program 2001).
The research induces that approximate ‘‘S’’ curve of R&D intensity, pointing
out that 1.0 and 2.5 % are two points of inflection in the curve of R&D intensity
(Zeng and Tan 2003).
The study finds that the pattern of science and technology (S&T) take-off is
characterized by an abrupt increase from 1 to 2 % in the R&D intensity (Gao and
Jefferson 2007).
According to the UNESCO’s research, the variation in R&D has intimate
relationship with economic development stage. In general, the variation in R&D
intensity is not more than 1.5 % during the initial stage of industrialization; the
variation in R&D intensity is about 1.5–2.5 % during the middle stage of indus-
trialization; the variation in R&D intensity is generally greater than 2.0 % during
the advanced stage of industrialization.
In the initial stage of industrialization, the growth of variation in R&D intensity
is from fast to slow; in the middle stage of industrialization, the variation in R&D
intensity grows rapidly at first, but later it grows more slowly; in the advanced
stage of industrialization, the variation in R&D intensity tends towards stability
(Deng and Ai 2004).
Generally, the country with its variation in R&D intensity below l.0 % is lack
of innovation ability; the country with its variation in R&D intensity between l.0
and 2.0 % makes a difference; the R&D intensity of innovative country is above
2 % (Zhang 2001).
The trend of R&D intensity is showed in Fig. 16.1.
Fig. 16.1 The trend of R&D
R&D/GDP
intensity
(%)
2.5
2.0
1.5
1.0
Growth Slow Take-off Growth Stable
Year
16 Analysis on Trend of Research and Development Intensity 151
16.3 The Trend of R&D Intensity in China
16.3.1 Trend of R&D Intensity
Over the past thirty-four years, China’s economy has moved from being largely
closed to becoming a mayor global player. Gross expenditure on R&D (GERD)
increased consistently from 0.61 % in 1987 to 1.76 % of GDP in 2010. The
government requires above 2.5 % R&D intensity by 2020 according to the
Medium and long-term Science and Technology Strategic Plan (2006–2020) in
china (The Communist Party of China and State Council 2006). China’s R&D
intensity was low between 1987 (0.61 %) and 1998 (0.65 %). R&D intensity fell
from a peak of 0.74 in 1992 to 0.57 % in 1995 or 1996. R&D intensity increased
only marginally between 1997 and 1998 from 0.64 to 0.65 % but increased rap-
idly, particularly after 1998, at 1.07 % of GDP by 2002. China has started to step
into the stage of scientific and technological take-off since 2002. The China’s
R&D intensity trend from 1987 to 2010 is showed in Fig. 16.2.
16.3.2 Reflection of R&D Intensity Trend
China’s R&D intensity is still low. Taken as a whole, China is still in the primary
stage of socialism and remains a developing country. China is in the grouping of
upper-middle-income countries, in the middle stage of industrialization, and its
innovation ability places twenty-ninth in the word according to Global Innovation
Index (GII) 2011.
Fig. 16.2 The China’s R&D intensity trend from 1987 to 2010, Source Statistics of science and
technology of China, http://www.sts.org.cn/index.asp
152 B. Huang and L. Huang
16.4 Gray Prediction of R&D Intensity in China
16.4.1 Gray Prediction
Gray Prediction Theory, based on the Gray System Theory, established by Chinese
Scholar Professor Deng Julong in 1982 (Di 2002), is a new method to solve the
problems that are lack of data and information. GM (1, 1) model construct gray
prediction model to predict the characteristic quantity of a given amount of time,
or the time to reach a certain characteristic quantity by a number of equal interval
observed numbers which reflect the characteristics of the predictable object, such
as production, sales, population, interest rates, etc. (Deng 1982; Liu and Lin 2006;
Deng 1989). In theory, GM (1, 1) model is a continuous function of time,
stretching from the initial value to the future at any time (Deng 1995).
The basic procedure for grey prediction is as follows (Deng 2005):
Step 1: Determine the raw series x(0), class ratio series rð0Þ
The raw series x(0) for modeling are given as follows:
n o
xð0Þ ¼ xð0Þ ð1Þ; xð0Þ ð2Þ; . . .; xð0Þ ðnÞ ð16:1Þ
where n is equal or greater than 4.
Let rð0Þ ðkÞ be class ratio of x(0) at point k,
rð0Þ ðkÞ ¼ xð0Þ ðk 1Þ=xð0Þ ðkÞ ð16:2Þ
Step 2: Determine the semi-finished series xð1Þ ; zð1Þ
xð1Þ is valued as follows:
xð1Þ ¼ AGOxð0Þ , where AGO means Accumulated Generating Operation.
xð1Þ ð1Þ ¼ xð0Þ ð1Þ; where k is equal 1;
xð1Þ ðkÞ ¼ xð0Þ ðk 1Þ þ xð0Þ ðkÞ ð16:3Þ
where k is greater than 1.
zð1Þ is valued as follows:
zð1Þ ¼ MEANxð1Þ , where MEAN means the average of adjoining number of xð1Þ .
xð1Þ ðkÞ ¼ ðxð1Þ ðkÞ þ xð1Þ ðk 1ÞÞ=2 ð16:4Þ
where k is greater than 1.
Step 3: Determine the semi-parameters C, D, E, F
X X
E¼ ðzð1Þ ðkÞ xð0Þ ðkÞÞ; F ¼ ðzð1Þ ðkÞ zð1Þ ðkÞÞ
X X ð16:5Þ
C¼ zð1Þ ðkÞ; D ¼ xð0Þ ðkÞ
where k is greater than 1, and less than n.
16 Analysis on Trend of Research and Development Intensity 153
Step 4: Determine the developing coefficient a and grey input b as follows:
a ¼ ðC D ðn 1Þ FÞ=ððn 1Þ F C CÞ ð16:6Þ
b ¼ ðD F C EÞ=ððn 1Þ F C CÞ ð16:7Þ
Step 5: Determine the Grey differential equation of GM (1, 1) and the white
response of GM (1, 1) as follows:
xð0Þ ðkÞ þ a zð1Þ ðkÞ ¼ b ð16:8Þ
then we can get the white response of GM (1, 1):
x0ð0Þ ðk þ 1Þ ¼ x0ð0Þ ð1Þ b=a eab þ b=a ð16:9Þ
Step 6: Checking residual errors rðiÞ ðkÞ as follows: recuperating value:
xð0Þ ðk þ 1Þ ¼ xð1Þ ðk þ 1Þ xð1Þ ðkÞ ð16:10Þ
residual errors:
h i
dð 1 Þ ð k Þ ¼ xð1Þ ðkÞ x0ð1Þ ðkÞ =ðxð1Þ ðkÞ 100 % ð16:11Þ
k = 0,1,2,… where xð1Þ ðkÞ is the actual value, x0ð1Þ ðkÞ is the modeled value.
Step 7: Check GM (1, 1) model accuracy
According to literatures (Liu et al. 2004), if dð1Þ ðkÞ value is equal or less than
0.01, the model is level 1, if dð1Þ ðkÞ value is between 0.01 and 0.05, the model is
level 2, if dð1Þ ðkÞ value is between 0.05 and 0.10, the model is level 3, if dð1Þ ðkÞ
value between 0.10 and 0.20, the model is level 4.
16.4.2 China’s R&D Intensity Gray Prediction
According to recent R&D intensity data in China, take data on R&D intensity in
China over the past 6 years as original sequence:
xð0Þ ¼ ð1:32; 1:39; 1:40; 1:47; 1:70; 1:76Þ;
ð0Þ
let xn1 as xð0Þ subsequence, also is
ð0Þ ð0Þ ð0Þ ð0Þ
xn1 ¼ ðxðniÞ ; xðniþ1Þ ; . . .; xðnÞ Þ
where n is equal six. Then make models for these sub-semi-groups sequence,
respectively, constantly adjusted to these sequences’ interval boundary value,
repeatedly modeling, thus obtained GM (1, 1) models with different interval
boundary value of the project, also is
154 B. Huang and L. Huang
Table 16.1 Forecast data on Years R&D intensity
China’s R&D intensity
2011 1.89
2012 2.02
2013 2.16
2014 2.31
2015 2.48
2016 2.65
x0ð0Þ ðk þ 1Þ ¼ x0ð0Þ ð1Þ b=a eab þ b=a:
In the end, make residual inspection of GM (1, 1) model with different interval
boundary value, then selected residual qualified model for level or close precision
level as the R&D intensity GM (1, 1) model.
Forecast data on China’s R&D intensity in the next 5 years is showed in
Table 16.1.
16.5 Conclusion
The upward trend of R&D intensity in China is generally in agreement with the
rule of R&D intensity trend. It has entered the second stage Science and Tech-
nology (S&T) take-off which is a crucial period. However the level of R&D
intensity in China is still low. But, this study documents that China has begun a
similar S&T take-off, and the increased ratio of R&D intensity is consistent with
that of GDP. So R&D intensity in China will be able to realize the steady and
quick growth based on the goal of building of innovation-oriented country.
Meanwhile, R&D intensity in China will have reached 2.5 % before 2016.
Acknowledgments The project is supported by the National Natural Science Foundation of
China (70639002).
References
Deng J (1982) Control problem of grey system. Syst Control Lett 1(5):288–294
Deng J (1989) Introduction to Grey System Theory. J Grey Syst 1(1):1–24
Deng J (1995) Extent Information Cover Grey System Theory. J Grey Syst 7(2):131–138
Deng J (2005) The primary methods of grey system theory, 2nd edn. Huazhong University of
Science and Technology Press, Hubei, pp 8–15
Deng X, Ai Q (2004) Analysis on Chinese R&D investment model and relevant policies—
analysis on government leading R&D investment model. Contemp Finance Econ
232(3):23–24
16 Analysis on Trend of Research and Development Intensity 155
Di L (2002) Brief discussion on gray model and its application in physical education. Fujian
Sports Sci 21(1):6–8
English section of the central document translation department (2011) The twelfth five-year plan
for national economic and social development of the People’s Republic of China. Central
Compilation and Translation Press, Beijing, p 16
Gao J, Jefferson GH (2007) Science and technology take-off in China?: sources of rising R&D
intensity. Asia Pacific Business Review, 13(3):357–337
Liu SF, Lin Y (2006) Grey information theory and practical applications. Springer-Verlag,
London, pp 7–12
Liu S, Dang Y, Fang Z (2004) Grey system theory and application, 3rd edn. Science press,
Beijing, pp 163–164
OECD (2011a) Regions at a glance. OECD Press, Paris, p 62
OECD (2011b) Oecd science, technology and industry scoreboard 2011. OECD Press, Paris, p 18
The Communist Party of China and State Council (2006) The medium and long-term science and
technology strategic plan (2006–2020) in China. China Legal Press, Beijing, p 68
UNDP (United Nations Development Program) (2001) Human development report. Oxford
University Press, New York, pp 52–54
Zeng G, Tan W (2003) The development path of states’ R&D and basic research and its
inspirations. Stud Sci 21(3):154–156
Zhang W (2001) A comparative study R&D expenditure between China and U.S. China Soft Sci
12(10):79
Chapter 17
Application of IE Method in Modern
Agro-Ecological Park Planning
Li Yue
Abstract The present paper deals with the comprehensive planning of a new
modern agro-ecological park established by a company. Learning from the prac-
tices of IE method used in the manufacturing and service industries and consid-
ering the characteristics of modern agriculture, we offers some references to the
planning of fungi configuration, functional zone, processing zone and layout of the
park to maximize the overall benefits of the construction of a modern agro-
ecological park.
Keywords IE method Modern agriculture Plan SLP
17.1 Introduction
The IE method has been widely used in the manufacturing and service industries
such as automobile, steel, machinery, home appliance, construction materials, and
information. Japan is the first country to apply IE method in agriculture and this
application is still at its initial stage (Guo 2003; Qi et al. 1999; Yi and Guo 2007).
One company (hereinafter referred to as A company) will establish a modern
agro-ecological park covering 90 mu, which will incorporate fungi cultivation,
personnel training, eco-tourism, entertainment and dining and conference and
leisure. Combining the characteristics of modern agriculture and the practices of
IE method used in manufacturing and service industries, this article considers
solutions to the problems in the planning of a modern agro-ecological park with a
view to maximizing corporate and social benefits (Shi 2002; Shin 2001; Wu et al.
2008).
L. Yue (&)
School of Manufacturing Science
and Engineering, Southwest University of Science and Technology, Mianyang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 157
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_17,
Ó Springer-Verlag Berlin Heidelberg 2013
158 L. Yue
17.2 Research Guideline and Methodology
The construction of modern agro-ecological park shares much in common with that
of modern manufacturing company. This paper employs the IE method in the
research of agro-ecological park planning in the following areas: location and layout
of the park, organizational structure planning, supply chain planning, capacity
planning, industrial chain planning, human resource planning, human factors envi-
ronmental planning and standardized production (Lu and Wang 2004; Wang 2009).
The research guideline and procedures in this article are shown in Fig. 17.1: (1)
using Fuzzy Comprehensive Evaluation method to choose the most suitable fungi
species for cultivation., (2) using SWOT analysis and Historical Analogy Analysis
method to predict the capacity of each functional zone and make personnel and
equipment planning according to the market demand., (3) in accordance with the
results of capacity planning, employing the knowledge of Operations Research,
building a model with the optimal area as its target and using Lindo to find the optimal
solutions to the area of each functional zone., (4) and on the above basis, using SLP
method to make a plan for the layout of the processing zone and the park (Jiang 2001).
17.3 The Species of Edible Fungi Planning
Through the market research and the analysis of cultivation techniques, there are
six selected species of edible fungi to be cultivated: Morchella conica, Lentinus
edodes, Ganoderma lucidum, Coprinus comatus, Pleurotus ostreatus, Flammulina
Fig. 17.1 Research
The application of IE method in the manufacturing and
guideline
service industries
The application of IE method in modern agro-ecological
park planning
The species of edible fungi planning
The capacity planning of each functional zone
Functional zone area Processing zone layout
planning planning
General layout planning
17 Application of IE Method in Modern Agro-Ecological Park Planning 159
Table 17.1 The evaluation Evaluation indicators Weights
of the species of fungi
1. Return on investment 0.25
2. Difficulty of cultivation technique 0.15
3. Sales channels 0.20
4. Competition environment 0.22
5. Managerial expertise 0.10
6. Market response capability 0.08
Morchella conica Lentinus edodes Ganoderma lucidum
Fig. 17.2 Edible fungi
velutipes, but only three of them can be cultivated due to the actual demand of A
company. Therefore, use fuzzy comprehensive evaluation method to make eval-
uations of the selected species. The evaluation indicators and their associated
weights are shown in Table 17.1 (Pan 2003).
According to the Table 17.1, the selected six species are graded, ranking from
high to low, and Morchella conica, Ganoderma lucidum, Lentinus edodes are the
final selections. See Fig. 17.2 (Hang 2003; Zhang 2002).
17.4 Capacity Prediction and Planning
of Each Functional Zone
The agro-ecological park can be divided into 6 zones with some relevantly dis-
tinctive functions: (1) Dining and Training Zone; (2) Morchella conica Cultivation
Zone; (3) Ganoderma lucidum Cultivation Zone; (4) Lentinus edodes Cultivation
Zone; (5) Breeding and Sightseeing Zone; (6) Processing Zone (Ding 2004).
We analyze the geographic conditions and human environment of the park as
the first step, then start from the market analysis and make capacity predictions of
each zone using the historical analogy method and SWOT analysis according to
the various features of each functional zone, and finally make personnel, equip-
ment and construction planning for each zone in conformity with the results of
capacity predictions.
160 L. Yue
17.4.1 Capacity Planning for Catering and Training Zone
Dining and training zone is a major functional zone featuring farmer training,
instructing, dining (mainly for fresh edible fungi), eco-tourism, specialty shops,
DIY zone and leisure and entertainment. It is predicted that the annual number of
visitors will be 27,000–33,000 and about 1,000 edible fungi cultivators will
receive professional training annually.
17.4.2 Capacity Planning for Edible Fungi Cultivation Zone
The market demand for Morchella conica mainly comes from the visitors who are
coming here for relax and leisure. In accordance with the capacity planning for
dinning and training zone, it is projected that the annual demand of Morchella
conica will reach 5,900–7,300 kg.
Ganoderma lucidum has been proven to have medicinal properties, and its
market demand has been increasing year after year. There are two sales channels
for Ganoderma lucidum: (1) fresh Ganoderma lucidum consumed in restaurants;
(2) Ganoderma lucidum spore powder, dried Ganoderma lucidum and powdered
Ganoderma lucidum sold to herbal medicine dealers. Based on the market anal-
ysis, competitor analysis and SWOT analysis, it is forecasted that the annual
demand of Ganoderma lucidum will stand at 149,000–165,000 kg.
Skiitake is rich in vitamin, iron and potassium, treating loss of appetite and
relieving languor. Through the current market analysis and the sales data analysis,
it is calculated that the annual demand of Lentinus edodes will be
455,000–572,000 kg in terms of the historical analogy method.
17.4.3 Capacity Planning for Breeding and Sightseeing Zone
Breeding and sightseeing zone takes breeding and earthless cultivation as its core,
mainly breeding for Morchella conica cultivation zone, Ganoderma lucidum
cultivation zone and Lentinus edodes cultivation zone. Thereupon, the capacity of
breeding zone can be determined to some extent by that of the three cultivation
zones. The capacity planning is as follows: Morchella conica 118–146 bottles,
Ganoderma lucidum 2980–3300 bottles, Lentinus edodes 9100–11440 bottles.
17 Application of IE Method in Modern Agro-Ecological Park Planning 161
Table 17.2 Annual capacity planning of processing zone (ten thousands kg/year)
Species of edible fungi Name Total
Lentinus edodes Powdered Lentinus edodes 319
Dried Lentinus edodes 410
Preserved Lentinus edodes 182
Ganoderma lucidum Powdered Ganoderma lucidum 74
Dried Ganoderma lucidum 124
Preserved Ganoderma lucidum 50
17.4.4 Layout Planning for Processing Zone
The main function of this zone is processing Ganoderma lucidum and shiitake.
Given the development goals of A company, the capacity planning of processing
zone is shown in Table 17.2.
17.5 Optimized Layout of Processing Zone
The processing zone layout has much in common with plant layout, so SLP
method can be applied.
According to the processing techniques of Ganoderma lucidum and Lentinus
edodes, this zone can be divided into 10 operating units: 1 raw material base, 2
cleaning zone, 3 drying room, 4 sun drying zone, 5 sterilization room, 6 cooling
room, 7 pulverizing room, 8 low-temperature packaging room, 9 packaging and
quality monitoring workshop, 10 processing products zone.
First, analyze the processing procedures of the edible fungi and logistics
amount and then establish logistics relationship shown in Fig. 17.3.
Analyze non-logistics relationship in processing zone and establish non-logis-
tics relationship shown in Fig. 17.4.
Then, combine the two above relationships and establish integrated relationship
shown in Fig. 17.5.
Finally, according to the integrated relationship between various operating
units, draw their area diagram (Figs. 17.6 and 17.7).
Take the characteristics of edible fungi processing, relevant regulations and
restraints into account and then determine the overall layout of the processing zone
shown in Fig. 17.8 (Chen et al. 2008; Dong 2005).
162 L. Yue
Fig. 17.3 Logistics relationship diagram
Fig. 17.4 Non-logistics relationship diagram
17 Application of IE Method in Modern Agro-Ecological Park Planning 163
Fig. 17.5 Integrated relationship diagram
Fig. 17.6 The location-
related map of operating units
164 L. Yue
Fig. 17.7 The area-related map of operating units
Fig. 17.8 Overall layout of processing zone
17.6 Optimized Layout of the Park
17.6.1 Preliminary Area of Each Functional Zone
The capacity per unit of each functional zone divided by its overall annual demand
equals the initial area of each zone shown in Table 17.3.
17
Table 17.3 The initial area of each zone
Name Morchella conica Ganoderma lucidum Lentinus edodes Processing Breeding and Catering and
cultivation zone cultivation zone cultivation zone zone sightseeing zone training zone
Area requirement Maximum 1.1 20 44 10 10.8 11
Unit (acre of land) Minimum 0.91 18 35 10 8.6 9
Application of IE Method in Modern Agro-Ecological Park Planning
165
166 L. Yue
Table 17.4 The interest coefficient of yield acre of land in each functional zone
Name Morchella conica Ganoderma lucidum Lentinus edodes Catering and
cultivation zone cultivation zone cultivation zone training zone
Interest 13.4 11.1 7.6 6.5
coefficient
17.6.2 Formula for the Calculation of Optimal Area
of Each Functional Zone
The interest coefficient of each functional zone is shown in Table 17.4.
Set the values of added areas of Morchella conica cultivation zone, Ganoderma
lucidum cultivation zone, Lentinus edodes cultivation zone, breeding zone and
catering and training zone as X1, X2, X3, X4, X5 respectively, and make use of the
knowledge of operations research to build mathematical model with the maximum
benefits of the agro-ecological park as the target.
17.6.2.1 Objective Function
MAX Z ¼ 13:4X1 þ 11:1X2 þ 7:6X3 þ 6:5X5 ð17:1Þ
Constraints:
S.T.
X1 þ X2 þ X3 þ X4 þ X5 71 ð17:2Þ
0:12X1 þ 0:09X2 þ 0:2X3 ¼ X4 ð17:3Þ
0:1X5 þ 1 ¼ X1 ð17:4Þ
0:9 X1 1:1 ð17:5Þ
18 X2 20 ð17:6Þ
35 X3 44 ð17:7Þ
0 X5 2 ð17:8Þ
Use Lindo to find the optimal solutions and calculate the results: X1 ¼ 1; X2 ¼
3; X3 ¼ 40; X5 ¼ 0; X4 ¼ 0:12X1 þ 0:09X2 þ 0:2X3 ¼ 10: Thus the optimal
area of each functional zone is shown in Table 17.5.
17 Application of IE Method in Modern Agro-Ecological Park Planning 167
Table 17.5 The optimal area of each functional zone (mu)
Name Morchella Lingzhi Shiitake Processing Breeding Catering
conica mushroom cultivation zone and and
cultivation cultivation zone sightseeing training
zone zone zone zone
Optimal 1 20 40 10 10 9
area
Table 17.6 Comprehensive relationship between each functional zone
Catering Breeding Morchella Ganoderma Lentinus
Processing and and Conica lucidum edodes
Name
Zone Training Sightseeing Cultivation Cultivation Cultivation
Zone Zone Zone zone Zone
Processing Zone
XX XX X E A
Catering and
Training Zone XX A I E O
Breeding and
Sightseeing Zone XX A I I I
Morchella
Conica
Cultivation Zone X I I E E
Ganoderma lucidum
Cultivation zone E E I E E
Lentinus edodes Cultivation Zone
A O I E E
Close degree
2 8 8 9 14 13
Ranking
6 5 4 3 1 2
17.6.3 Analysis on Comprehensive Relationship Between
Each Functional Zone
The higher ranking the close degree is, the closer to the center of the general layout
the operating unit should be and vice versa.
According to Table 17.6, the rankings are as follows: 1 Ganoderma lucidum
cultivation zone; 2 Lentinus edodes cultivation zone; 3 Morchella Conica Culti-
vation Zone; 4 breeding and sightseeing zone; 5 catering and training zone;
6 processing zone. The location-related map of each functional zone and the
general layout of the park are shown in Figs. 17.9 and 17.10.
168 L. Yue
Fig. 17.9 The location and area-related map of each functional zone
Fig. 17.10 The general layout of the agro-ecological park
17 Application of IE Method in Modern Agro-Ecological Park Planning 169
17.7 Conclusion
This paper examines a new way about modern agricultural park planning theo-
retically and practically. We draw references from the practices of IE method used
in the manufacturing and service industries and make full use of IE knowledge,
making planning for capacity, fungi configuration, park layout and other aspects so
as to maximize the overall benefits of the agro-ecological park.
References
Chen Y, Jia W-q, Huang Y (2008) The design of distributed measuring-control and monitor
system of edible fungi facilities. Hunan Agric Sci 3:149–150
Ding Q (2004) A recycling agriculture model based on edible fungi cultivation. J Anhui Agric Sci
7(5):21–22
Dong H (2005) Facilities planning and logistics analysis. Machinery Industry Press, Beijing
Guo C-q (2003) The status and role of industrial engineering in development of china’s
manufacturing industry. Ind Eng Manag 2:1–3
Hang Y (2003) The problem and countermeasure of edible fungus factory’s facilities cultivation.
Edible Fungi 6:2–4
Jiang W (2001) Development of soilless culture in mainland China. Trans CASE 17(1):10–15
Lu L, Wang J (2004) IE application of IE to agriculture production management. Ind Eng J
7(5):21–22
Pan E-s (2003) Planting planning and control. Shanghai Jiao Tong University press, Shanghai
Qi E-s, Wang Y-l, Lu L (1999) Developing status and tendency of Chinese industrial engineering
subject. Ind Eng J 2(1):1–4
Shi T (2002) Ecological agriculture in China: bridging the gap between rhetoric and practice of
sustainability. Ecol Econ 42(3):359–368
Shin DH (2001) An alternative approach to developing science parks: a case study from Korea.
Pap Reg Sci 80:103–111
Wang J (2009) Analysis of agricultural industry chain construction. North Econ 4:94–95
Wu Y-r, Qiu M-q, Wang C-b (2008) The reflection of modern agriculture demonstration garden’s
construction. Mod Agric Sci Technol 20:329–332
Yi S-P, Guo F (2007) Fundament of industrial engineering. Machinery Industry Press, Beijing
Zhang Y (2002) Probe into regarding the edible mushroom as the central industry in the
development of agricultural circular economy. Sci-Tech Inf Dev Econ 19(20):121–122
Chapter 18
Application of Improved Grey Prediction
Model for Chinese Petroleum
Consumption
Ying Ma and Meng Sun
Abstract An improved Grey-based settlement predictor is promoted. Adopt Grey
prediction as a forecasting means because of its fast calculation with a few of data
inputs needed. However, our preliminary study shows that the general Grey model,
GM (1,1) is inadequate to handle settlement prediction as its only adapt to the data
with exponential law. In this paper, the prediction is improved significantly by
adopting equal dimensionality information fill model aim at enhancing the pre-
diction accuracy. The prediction is made on Chinese petroleum consumption in
future.
Keywords Grey system theory Improved GM (1,1) Petroleum consumption
Prediction
18.1 Introduction
Energy is the life blood of the economy, relations with the development of the
national economy and improving the People’s living conditions. Energy con-
sumption is forecast to the stable and rapid economic development, to speed up the
healthy development of the energy industry, and conducive to the formulation of a
sound energy planning, so energy consumption accurate forecasts that is very
necessary.
Y. Ma (&)
School of Economics and Management, China Petroleum University,
Dongying 266580 Shandong, China
e-mail: [email protected]
M. Sun
School of Management, Shandong University, Jinan 250014 Shandong, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 171
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_18,
Ó Springer-Verlag Berlin Heidelberg 2013
172 Y. Ma and M. Sun
China is the world’s petroleum Producing and consuming nations. In recent
years, along with the rapid economic growth, domestic petroleum consumption
increased substantially. How to better predict petroleum consumption, which has
become an important sustainable development issues on the national economy.
Many researchers developed the settlement prediction methods, ‘Three Point
Method’ (Li et al. 1994), ‘Hyperbola Method’, ‘Grey Theory’ (Zhang et al. 1999),
‘Neural Network’ (Tavenas and Leroueil 1980), ‘Asaokao’ (Asaoka 1978) etc.
Grey theory is one of the most widely used methods. Grey theory, developed
originally by Deng (1982), is a truly multidisciplinary and generic theory that deals
with systems that are characterized by poor information and/or for which infor-
mation is lacking. The fields covered by grey theory include systems analysis, data
processing, modeling, prediction, decision making and control. Grey forecasting
models have been extensively used in many applications (Gao et al. 2003; Liu
2001; Meng et al. 2002; He 2002), and it was improved by many researchers.
18.2 Methodology
18.2.1 GM (1,1) Model
Deng proposed Grey theory to deal with indeterminate and incomplete systems.
Unlike conventional stochastic forecasting theory, Grey theory simply needs few
sample data inputs to construct a Grey model. Since the poor regularity for the
tested settlement data, the AGO technique in Grey forecasting is suitable to reduce
the randomization of the raw data efficiently. Generally, the procedure for GM
(1,1) forecasting is explained as follows:
Step 1. Denote the original data sequence by
n o
ð0Þ
X ¼ X ð0Þ ð1Þ; X ð0Þ ð2Þ; . . .; X ð0Þ ðnÞ ð18:1Þ
where n is the number of data observed.
Step 2. The AGO formation of xð0Þ is defined as:
n o
ð1Þ
X ¼ X ð1Þ ð1Þ; X ð1Þ ð2Þ; . . .; X ð1Þ ðnÞ ð18:2Þ
where
X ð1Þ ð1Þ ¼ X ð0Þ ð1Þ
( )
X k
ð1Þ ð0Þ
and X ðkÞ ¼ X ðiÞ; k ¼ 1; 2; 3; . . .; n ð18:3Þ
i¼1
18 Application of Improved Grey Prediction Model 173
The GM (1,1) model can be constructed by establishing a first order differential
equation for X ð1Þ ðkÞ as:
dxð1Þ ðkÞ=dk þ axð1Þ ðkÞ ¼ b ð18:4Þ
Therefore, the solution of Eq. 18.4 can be obtained by using the least square
method. That is,
X^ ð1Þ ðkÞ ¼ X 0 ð1Þ b eaðk1Þþ b ð18:5Þ
a a
where
1
½a bT ¼ BT B Xn ð18:6Þ
and
2 3
Z ð1Þ ð2Þ 1
6 Z ð1Þ ð3Þ 1 7
B¼6
4
7 ð18:7Þ
5
Z ð1Þ ðnÞ 1
where
Z ð1Þ ðkÞ ¼ axð1Þ ðkÞ þ ½1 axð1Þ ðkÞ
ð18:8Þ
Xn ¼ ½xð0Þ ð1Þ; xð0Þ ð2Þ; . . .; xð0Þ ðnÞT
We obtained ^xð1Þ from Eq. 18.5. Let ^xð0Þ be the fitted and predicted series,
^xð0Þ ¼ ðxð0Þ ð1Þ; xð0Þ ð2Þ; . . .; xð0Þ ðnÞ; . . . ð18:9Þ
where ^xð1Þ ð1Þ ¼ xð0Þ ð1Þ
ð0Þ ð0Þa b aðk1Þ
^x ðkÞ ¼ ð1 e Þ X ð1Þ e ð18:10Þ
a
where ^xð0Þ ð1Þ; ^xð0Þ ð2Þ; . . .^xð0Þ ðnÞ are called the GM (1,1) fitted sequence, while
^xð0Þ ðn þ 1Þ; ^xð0Þ ðn þ 2Þ; . . . are called the GM (1,1) forecast values.
18.2.2 Improved GM (1,1) Model
Grey prediction has lot of advantages, such as less data required, regardless of
distribution, regardless of trends, convenient operation, short-term forecasts of
high precision, easy to test, etc., so it is widely used, and achieves satisfactory
results. But there are some limitations. Many scholars encountered a number of
problems of low prediction accuracy in the application of GM (1,1) model, and
174 Y. Ma and M. Sun
many researchers have proposed methods to improve prediction accuracy. There
are mainly two ways to improve the forecasting model, one is to transform the
original series, and the other is to transform GM (1,1) itself. GM (1,1) improved
methods are generally ^ a parameter correction method, the raw data moving
average processing, and equal dimension information processing, residual treat-
ment etc., here we took equal dimensionality information fill model.
Generally modeling must be based on the latest data as a reference point.
According to the different choice of data, new information model and equal
dimension new information model are commonly used in GM (1,1) model. New
information model will add the new information into the original series when
coming up with new information and modeling after adding the new information
(full-line modeling), also known as metabolic models. It’s mechanism close to the
general theory of forgetting factor modeling ideas.
18.3 Empirical Analysis
In 1993, China became a petroleum importing country. With the rapid develop-
ment of the national economy, petroleum demand was in the rapid growth, the
petroleum supply situation will be even more severe in the future. No matter From
the goal of economic development, or the goal of environmental protection,
adjusting and improving the energy structure, promoting the diversification of
energy use are the only way for China’s sustainable energy strategy. Table 18.1
shows China’s petroleum consumption in recent years.
The raw data xð0Þ ðkÞ are tested by Quasi-smooth conditions. Then Test whether
xð1Þ have the quasi-exponential law. Raw data, for example:
xð0Þ ðkÞ
qðkÞ ¼ ð18:11Þ
xð1Þ ðk 1Þ
Table 18.1 Petroleum consumption of China
Years The proportion of total energy consumption (%) Petroleum (million tons of
standard coal)
Coal Petroleum Nature Hydroelectric, nuclear,
gas wind power
2002 66.3 23.4 2.6 7.7 35520
2003 68.4 22.2 2.6 6.8 38847
2004 68.0 22.3 2.6 7.1 45319
2005 69.1 21.0 2.8 7.1 47414
2006 71.1 19.3 2.9 6.7 49924
2007 71.1 18.8 3.3 6.8 52735
2008 70.3 18.3 3.7 7.7 53334
2009 70.4 17.9 3.9 7.8 54889
2010 68.0 19.0 4.4 8.6 61738
2011 69.7 18.2 4.6 7.4 63278
18 Application of Improved Grey Prediction Model 175
when k [ 3; qðkÞ\0:5, Quasi-smooth conditions are met.
Given
xð1Þ ðkÞ
Given dð1Þ ðkÞ ¼ ð18:12Þ
xð1Þ ðk 1Þ
when k [ 3; dð1Þ ðkÞ 2 ½1; 1:5, then Quasi exponential law are met.
First doing the quasi-smooth test, the results are as follow:
qð3Þ ¼ 0:55; qð4Þ ¼ 0:36\0:5;
qð5Þ ¼ 0:28\0:5; qð6Þ ¼ 0:24\0:5;
qð7Þ ¼ 0:23\0:5; qð8Þ ¼ 0:19\0:5;
qð9Þ ¼ 0:17\0:5; qð10Þ ¼ 0:12\0:5
So when k [ 3, the quasi-smooth conditions is met.
Similarly quasi-exponentially test, when k [ 3, meet the exponential law.
To illustrate the superiority of improved GM (1,1) model in predicting. For the
purpose of fitting the existing data and predicting future demand, we selected
petroleum consumption data and using MATLAB software for data processing.
The result of relative error of calculated values and actual values are as follows in
Table 18.2 by using gray theory method.
Using Gray theory to fit the original data, the model has a higher fitting
accuracy. Because of uncertainty of the future factors, using traditional gray
theoretical model to predict may cause the results are untrue. Sometimes the higher
the reliability of the data closing to the true value and the longer time interval, the
data reliability is easier to change. Therefore using equal dimensionality infor-
mation fill model to predict China’s future petroleum consumption will be better,
we used MATLAB software predicted the results shown in Table 18.3.
Table 18.2 Chinese petroleum consumption fitting value (million tons of standard coal)
Years Petroleum Traditional GM(1,1) model Error test
Predicted value Residual Relative error (%)
2002 35520 35520 0 0.00 C = 0.02
2003 38847 39746 -899 -2.31 P=1
2004 45319 42840 2479 5.47
2005 47414 46776 638 1.35
2006 49924 49770 154 0.31
2007 52735 53645 -910 -1.73
2008 53334 54987 -1653 -3.10
2009 54889 56765 -1876 -3.42
2010 61738 62876 -1138 -1.84
2011 63278 63786 -508 -0.80
176 Y. Ma and M. Sun
Table 18.3 The forecast value china’s petroleum consumption (million tons of standard coal)
Years Petroleum Years Petroleum Years Petroleum
2012 67646 2018 97180 2024 139911
2013 72690 2019 104769 2025 150558
2014 74876 2020 112466 2026 161961
2015 77925 2021 120915 2027 174083
2016 83413 2022 130027 2028 187124
2017 90110 2023 139911 2029 201156
18.4 Conclusion
Through the forecast shows that China’s petroleum consumption in the year of
2029 will increase of 13.351 billion tons standard coal compared to the year of
1998. A big shortfall of petroleum demand will lead a tough situation of the
petroleum security. In the year of 2029, China will basically realize industriali-
zation and modern agriculture, the rapid growth in demand of petroleum will
restrict China’s economic growth, if we do not take some practical countermea-
sures for the security of petroleum.
References
Asaoka A (1978) Observational procedure of settlement prediction. Soils Found 18(4):87–101
Deng JL (1982) Grey system fundamental method. Huazhong University of Science and
Technology, Wuhan
Gao Y, Wang Z, Tao Z et al (2003) Estimation of non-statistical uncertainty in precision
measurement using grey system theory. Int J Adv Manuf Technol 22(2):271–277
He Q-l (2002) Application of grey system theory in the safety evaluation of coal mines. Process
Saf Sci Technol Part B 3:1065–1070
Li Y, Xu H, Zhang Yz (1994) The grey dynamic model and its application in the of grinding
temperature. The first Asia Pacific and second Japan-China international conference on
progress of cutting and grinding, Shanghai, China, 1994, pp 108–115
Liu C-w (2001) Application of grey system theory in mineral economics research. China
University of Mining and Technology, Xuzhou, China, pp 813–816
Meng X-L, Sun L-X, Zhou D (2002) Application and improvement of grey theory for the
assessment of environmental quality. J Harbin Inst Technol 34(5):700–702
Tavenas F, Leroueil S (1980) The behavior of embankments on clay foundations. Can Geotech J
17:236–260
Zhang Y-p, Zhang T-q, Gong X-n (1999) Grey prediction of settlement. Ind Constr 29(3):45–48,
57
Chapter 19
Applying an Integrated SOM Model
on Studying Corporate Governance Data
Jen-Ying Shih
Abstract Corporate governance has gradually become an important consideration
in investment decisions of stock market and credit evaluation of financial market
since several notorious business scandals, such as Enron’s inside information
trading in the US and executive entrenchment of Rebar conglomerate in Taiwan.
However, corporate governance is presented by a large amount of data, which
involves the following facets: managerial ownership, external ownership, board
size, board composition, financial transparency, executive compensation, etc. To
utilize these data, this study applied an integrated self-organizing map (SOM)
model, which has been successfully used for visualizing and clustering high-
dimension data in several applications, to study corporate governance of Taiwan’s
semiconductor companies by recognizing implicit patterns of these data. The
research results demonstrate that eight types of corporate governance have been
identified by the SOM model, and features selected by LabelSOM are used to
interpret the eight styles of companies.
Keywords Corporate governance Data mining Self-organizing maps
Semiconductor companies
19.1 Introduction
In recent decades, corporate governance has gradually become an important
consideration in investment decisions of stock market and credit evaluation of
financial market since several notorious business scandals. For example, Enron
used false accounting information to cheat stock market investors and the U.S.
J.-Y. Shih (&)
Graduate Institute of Global Business and Strategy, National Taiwan
Normal University, Taipei, Taiwan
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 177
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_19,
Springer-Verlag Berlin Heidelberg 2013
178 J.-Y. Shih
Securities and Exchange Commission, which made its executives earn an enor-
mous amount of illegal returns by manipulating the false information and inside
information in securities market. Similar cases also existed in Taiwan, for instance,
Rebar’s executives entrenched assets of the Rebar conglomerate and such behavior
hurt plenty of its stockholders and financial institutions who lend a lot of money to
the Rebar.
With consideration of the above examples, stock market investors prefer to
invest in the promising companies with strong corporate governance to minimize
the loss resulting from weak corporate governance; credit analysts of credit rating
agencies and financial institutions consider corporate governance factors in their
credit granting decisions to avoid the risks associated with bad debt (Standard and
Poor’s 2002; Fitch Ratings 2004; Ashbaugh-Skaifea et al. 2006). The authorities of
financial markets, such as the Financial Supervision Commission, R.O.C. (FSC),
thus required listed companies to file multifaceted corporate governance data,
including ownership structure, information on inside shareholders, institutional
shareholdings, board composition, board size, financial transparency, executive
compensation, etc., to protect stakeholders of these companies.
However, due to that corporate governance is presented by a massive amount of
data, to make a full use of these data, it is important to provide an analytical tool to
interpret the data. Therefore, this study applied an integrated self-organizing map
(SOM) model to study corporate governance of Taiwan’s semiconductor compa-
nies by recognizing implicit patterns of these data. The SOM has been successfully
used for visualizing and clustering high-dimension data (i.e., in this study, each
case company with lots of corporate governance data) in document classification,
construction of knowledge maps, market segmentation, financial statement anal-
ysis, etc. (e.g., (Back et al. 1998; Eklund et al. 2002; Shih et al. 2008; Shih 2011;
Kohonen et al. 2000)).
The remainder of this paper is organized as follows. In Sect. 19.2, we review
corporate governance literature to define features (input variables) used in this
study. Then the dataset of this study is introduced in Sect. 19.3. Subsequently,
Sect. 19.4 describes the methodology applied in this research (i.e., the integrated-
SOM model); Section 19.5 presents results and interpretation of such results.
Finally, the conclusions are provided in Sect. 19.6.
19.2 Corporate Governance
According to (Luo 2005), business corporations are directed and controlled by
their corporate governance system. The system specifies the distribution of rights
and responsibilities among different parties who participate in corporate affairs and
spell out the rules and procedures for making decisions on these affairs. Corporate
governance should assure that all stakeholders receive reliable information about
the value of the firm and it motivates corporate managers to maximize firm value
instead of pursuing personal goals in managerial decision making. Based on this
19 Applying an Integrated SOM Model 179
concept, considerable research investigated the relationship between corporate
governance and firm performance (Kim et al. 2004; Cui and Mak 2002; Short and
Keasey 1999), as well as the relationship between corporate governance and firm’s
credit ratings (or firm’s cost of debt financing) (Ashbaugh-Skaifea et al. 2006;
Sengupta 1998; Bhojraj and Sengupta 2003). Most of the research applied
regression models, such as linear regressions (e.g., (Demetz and Lehn 1985)),
nonlinear regressions (e.g., (Short and Keasey 1999)), piecewise regressions (e.g.,
(Morck et al. 1988)), and two stage least square regressions (e.g., (Kapopoulos and
Lazaretou 2007)), to investigate how corporate governance impact on firm per-
formance or credit ratings. For example, Jensen and Meckling (1976) found that
firm performance was positively affected by managerial ownership and proposed
an alignment effect hypothesis for the relationship between managerial ownership
(one of corporate governance factors) and firm performance.
However, due to limitation of regression method assumptions, previous
research using regression models only investigated the relationships between a few
independent variables (i.e., limited corporate governance variables) and dependent
variables, which only provide partial evidence on specific observation but not on a
whole picture of corporate governance. Instead of using regression models, this
study used a clustering methodology based on self-organizing map (SOM) algo-
rithm, to study this important issue. The major characteristic of SOM is that it
could map high-dimensional input data into a low-dimensional representation
space to facilitate data analysis and data visualization (Shih et al. 2008). This study
adopted a different approach relative to previous studies to explore the corporate
governance issue and its relationship with firm performance.
According to previous research, the factors that influence on corporate gover-
nance cover: (1) managerial ownership or insiders’ shareholdings (Short and Keasey
1999), (2) external ownership or shareholder activism, such as institutional share-
holdings (Smith 1996), (3) board size (Ashbaugh-Skaifea et al. 2006), (4) board
composition (Ashbaugh-Skaifea et al. 2006), (5) financial transparency (Ashbaugh-
Skaifea et al. 2006), and (6) compensation system for board directors, supervisors,
managers, and employees (Cheng and Firth 2005). Therefore, these factors are all
included in the input data used in this study for constructing the corporate gover-
nance maps.
19.3 Dataset
Taiwan’s semiconductor companies have played important roles in the global
semiconductor industry, and they thus have raised huge amounts of capitals from
securities markets and financial institutions for business operation. Therefore, this
study used the semiconductor companies as a test bed because the stakeholders of
these companies are affected seriously by the corporate governance of these
companies.
180 J.-Y. Shih
This study gathered data from 53 semiconductor companies that are listed on
the Taiwan Stock Exchange (TSE). The data set contains a total of 30 corporate
governance variables which are illustrated as follows.
• Managerial ownership or insiders’ shareholdings
The factor is observed by the following variables:
1. Percentage of shares held by board directors
2. Percentage of shares held by supervisors
3. Percentage of shares held by managers
4. Percentage of shares held by family
5. Percentage of shares owned by large shareholders whose holdings exceed 10 %
of outstanding shares
6. Shares pledged by directors to shares owned by directors
7. Shares pledged by supervisors to shares owned by supervisors
• External shareholdings
The factor is observed by the following variables:
8. Percentage of external shareholdings
9. Percentage of shares held by institutional investors
10. Percentage of shares held by foreign institutional investors
11. Percentage of shares held by domestic government
12. Percentage of shares held by domestic financial institutions
13. Percentage of shares held by foreign financial institutions
• Board size
The factor is observed by the following variables:
14. Number of board directors
15. Number of supervisors
• Board composition
The factor is further decomposed into three dimensions, including family
control, independence and division of managerial work, measuring by the fol-
lowing variables.
i. Family control
16. Number of family-controlled board directors
17. Number of family-controlled supervisors
ii. Independence
18. Number of independent board directors
19. Number of independent supervisors
iii. Division of managerial work
20. Number of directors who are also managers in the conglomerate
21. Number of supervisors who are also managers in the conglomerate
19 Applying an Integrated SOM Model 181
22. Whether CEO also serves as the chairman of board
• Compensation system
The factor is observed by the following variables:
23. Compensation of directors and supervisors (in NT$ million)
24. Average compensation of each director or supervisor
25. Compensation of directors and supervisors to earnings before tax
26. Employee bonus to earnings before tax
27. Employees’ cash bonus (in NT$ million)
28. Employees’ stock bonus (in NT$ million)
• Financial transparency
29. Frequency of financial forecast
30. Frequency of financial statement revisions
19.4 Methodology
This study mainly used the SOM algorithm to analyze corporate governance
patterns of Taiwan’s semiconductor companies. Kohonen (1982) proposed the
SOM algorithm, a two-layered fully connected artificial neural network, for
clustering tasks. The SOM model is widely used for generating topology-pre-
serving maps and for data visualization. The most notable characteristic of the
SOM algorithm is that the similarities in input data are mirrored to a very large
extent by their geographical vicinity within the representation space. Similar types
of input data are assigned to neighboring regions on the map (Shih et al. 2008).
The main advantage of the SOM algorithm is that it can map high-dimension data
into a lower dimension representation space. The SOM algorithm has been suc-
cessfully applied in knowledge management (Shih et al. 2008; Schweighofer et al.
2001) and financial statement analysis (Back et al. 1998; Eklund et al. 2002).
In addition, the appropriate map size of SOM is determined by a growing SOM
(GSOM) algorithm (Shih et al. 2008) in this study. The interpretation of SOM maps
is assisted by LabelSOM (Dittenbach et al. 2002) for analyzing the clustering results.
This integrated SOM model is constructed by the following procedures.
Step 1: Initialize all parameters of the integrated SOM model for model
training, including an initial neighborhood range, an initial learning rate of the
SOM, an accepted error measure, an initial map size for SOM training process, a
growing-stopping criterion (s1 ) for growing SOM, a maximum number of labels
and a label threshold for LabelSOM in selecting labels that represent each unit in
the output layer.
182 J.-Y. Shih
Fig. 19.1 Topology of GSOM
Step 2: Start with a virtual layer 0 (see Fig. 19.1) consisting of only one unit
whose weight vector (m0) is initialized as the average of all the input data. Then
calculate the mean quantization error (mqe) by the Euclidean distance between the
weight vector of the unit and all input vectors. The mqe of the unit 0 is calculated
as the following equation.
1 Xn
m0 xj ;
mqe0 ¼
n j¼1
where xj denotes the input vector j, k k represents the Euclidean vector norm, and
n is the total number of the input vectors.
Step 3: Set the initial map size of the output layer in a topology of, for example,
2 9 2 units. This topology is determined by the parameter provided in Step 1.
Step 4: Initialize weight vectors of the output layer.
Step 5: Use input vectors to train the initial SOM by the following algorithm
(Shih et al. 2008), Fig. 19.2.
Step 6: Evaluate the mapping quality by calculating each unit’s mqe in this
map. The mqe of the unit i is calculated as the following equation.
1 X
mi xj ;
mqei ¼
nci x 2C
j i
where mi refers to the weight vector of the unit i, xj denotes the input vector
j mapped to the unit i, Ci denotes the set of all input vectors mapped to the unit i,
and ni is the total number of the input vectors that are mapped to unit i.
A large mqe value means that the input vectors are not clustered well by the
current map topology. Hence, some new units need to be added on this map to
improve the mapping quality (i.e., to decrease the mqe). To determine where to
add the new units, we have to find the error unit (e) which has the largest value of
mqe. The e is determined by the following equation.
19 Applying an Integrated SOM Model 183
Fig. 19.2 SOM algorithm adapted from (Shih et al. 2008)
e ¼ arg maxfmqei g:
i
Then, either a new row or column of units is interpolated between the error unit
and its most dissimilar neighbor (as shown in Fig. 19.3).
The weight vectors of these new units are initialized as the average of their
neighbors. After growing the map, calculate the mean mqe of all units (MQE) in
the current map. A map grows until its MQE is reduced to a predefined fraction
(the growing-stopping criterion, s1 ) of the mqe0 in the virtual layer (as shown in
Eq. 19.1). The lower the value of the quantization error, the better the map has
been trained.
MQE\s1 mqe0 ð19:1Þ
Step 7: After training GSOM, LabelSOM (Dittenbach et al. 2002) is used to
select the features that can represent the input vectors mapped to each unit in the
final map. The selection mechanism is based on the quantization error vector (qik ),
which is determined by the distance between the weight vector mi ¼
ðli1 ; li2 ; . . .; lin Þ 2 Rn of the unit i and all the input vectors xj ¼ ðnj1 ; nj2 ; . . .; njn Þ 2
Rn mapped to the unit i (Eq. 19.2). After sorting the all qik of each unit, according
to the parameters set by step 1 (i.e., the maximum number of labels and the label
threshold), several features are selected as the labels for representing the unit.
184 J.-Y. Shih
Fig. 19.3 Growing mechanism of GSOM adapted from (Dittenbach et al. 2002)
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X ffi
1
qik ¼ ðlik njk Þ2 ; k ¼ 1; . . .; n ð19:2Þ
jCi j xj 2Ci
where Ci is the set of all input vectors mapped to the unit i in the output layer, lik
denotes the kth element of the weight vector of unit i in the output layer, and
njk represents the kth element of the input vector xj mapped to the unit i in the
output layer.
19.5 Results
The training result of the integrated SOM model is shown in Fig. 19.4, which
illustrate that eight types of corporate governance of semiconductor companies in
Taiwan are identified by this study. Based on the LabelSOM, each unit (cluster) is
assigned a set of up to three labels to represent various types of corporate gov-
ernance. The eight types of corporate governance are described as follows.
1. CEO also serves as the chairman of board (unification of directing). The
companies listed on Unit (1, 1)1 and Unit (2, 1) belong to this style; totally, 15
semiconductor companies are clustered in this style.
1
We use the notation (x,y) to refer to the unit in row x and column y, starting with (1,1) in the
upper left corner.
19 Applying an Integrated SOM Model 185
Fig. 19.4 Corporate governance maps of the semiconductor companies and its relationships with
ROA
2. Managers participate in decision making of the board by serving as board
directors. This style is represented by the variable ‘‘number of directors who are
also managers in the conglomerate’’. Those companies listed on Unit (1, 2)
belong to this style; totally, seven companies are clustered in this style. The
major characteristic of this style is that more than two directors also serve as
managers in their business conglomerate.
3. Companies with a large board size and a high percentage of shares held by
institutional investors, especially foreign financial institutions. Only one com-
pany (TSMC) is listed on Unit (1, 3).
4. Companies with a high percentage of shares held by institutional investors,
especially foreign institutions. Three companies are listed on Unit (2, 3). Basi-
cally, this type of companies is similar to the third type of companies (Unit (1, 3))
except that the third type of companies has a large board. Hence, Unit (2, 3) is
near Unit (1, 3).
186 J.-Y. Shih
5. Family-controlled companies with a high percentage of shares held by insti-
tutional investors, family, and board directors. Three companies are listed on
Unit (3, 1).
6. Companies with supervisor internalization and a high percentage of shares held
by domestic financial institutions. Three companies are listed on Unit (3, 2).
7. Companies with a high percentage of shares held by institutional investors, a
large board size, and more independent board directors. Six companies are
clustered in Unit (3, 3).
8. Others. Those companies not belonging to the above styles are clustered in Unit
(2, 2). However, LabelSOM could not provide any label to induce this style
because no feature can be determined to explain this cluster.
19.6 Conclusion
This study explored the corporate governance data of 53 semiconductor companies
listed on TSE in Taiwan by using the integrated-SOM model with 30 corporate
governance features. Eight types of corporate governance have been identified by
the model and some associations between the clustering results and features can be
represented on the maps. The study results show that the SOM could be a feasible
tool for analyzing corporate governance data. It is easy to visualize and interpret
the SOM results and provides a very practical approach to compare the corporate
governance data of different companies. Future research can extend the data set to
the other industries to investigate the feasibility of SOMs for analyzing corporate
governance data.
References
Standard & Poor’s (2002) Standard & Poor’s corporate governance scores: criteria, methodology
and definitions. McGraw-Hill Companies Inc, New York
Fitch Ratings (2004) Evaluating corporate governance: the Bondholders’ perspective. In: Credit
policy special report, New York
Ashbaugh-Skaifea H, Collins DW, LaFond R (2006) The effects of corporate governance on
firms’ credit ratings. J Acc Econ 42:203–243
Back B, Sere K, Vanharanta H (1998) Managing complexity in large data bases using self-
organizing maps. Acc Manag Inf Technol 8:191–210
Eklund T, Back B, Vanharanta H, Visa A (2002) Assessing the feasibility of self-organizing maps
for data mining financial information. In: Proceedings of the 10th European conference on
information systems, Gdańsk, Poland, pp 528–537
Shih JY, Chang YJ, Chen WH (2008) Using GHSOM to construct legal maps for Taiwan’s
securities and futures markets. Expert Syst Appl 34(2):850–858
19 Applying an Integrated SOM Model 187
Shih JY (2011) Using self-organizing maps for analyzing credit rating and financial ratio data. In:
Proceedings of 2011 International conference on asia pacific business innovation and
technology management, Dalian, China, pp 109–112
Kohonen T, Kaski S, Lagus K, Salojävi J, Honkela J, Paatero V, Saarela A (2000) Self-
organization of a massive document collection. IEEE Trans Neural Netw 11(3):574–585
Luo Y (2005) How does globalization affect corporate governance and accountability? A
perspective from MNEs. J Int Manag 11(1):19–41
Kim KA, Kitsabunnarat P, Nofsinger JR (2004) Ownership and operating performance in an
emerging market: evidence from Thai IPO firms. J Corp Finan 10:355–381
Cui H, Mak YT (2002) The relationship between managerial ownership and firm performance in
high R&D firms. J Corp Finan 8:313–336
Short H, Keasey K (1999) Managerial ownership and the performance of firms: evidence from the
UK. J Corp Finan 5:79–101
Sengupta P (1998) Corporate disclosure quality and the cost of debt. Acc Rev 73:459–474
Bhojraj S, Sengupta P (2003) Effect of corporate governance on bond ratings and yields: the role
of institutional investors and the outside directors. J Business 76:455–475
Demetz H, Lehn K (1985) The structure of corporate ownership: Causes and consequences.
J Political Econ 93:1155–1177
Morck R, Shleifer A, Vishny RW (1988) Managerial ownership and market valuation: An
empirical analysis. J Finan Econ 20:292–315
Kapopoulos P, Lazaretou S (2007) Corporate ownership structure and firm performance:
Evidence from Greek firms. Corp Govern Int Rev 15(2):144–158
Jensen MC, Meckling WH (1976) Theory of the firm: Managerial behavior, agency costs and
ownership structure. J Finan Econ 4:305–360
Smith MP (1996) Shareholder activism by institutional investors: Evidence from CalPERS.
J Finan LI(1):227–252
Cheng S, Firth M (2005) Ownership, corporate governance and top management pay in Hong
Kong. Corp Govern Int Rev 13(2):291–302
Kohonen T (1982) Self-organized formation of topologically correct feature maps. Biol Cybern
43:59–69
Schweighofer E, Rauber A, Dittenbach M (2001) Automatic text representation, classification
and labeling in European law. In: Proceedings of ICAIL 2001, ACM Press, Amsterdam
Dittenbach M, Rauber A, Merkl D (2002) Uncovering hierarchical structure in data using the
growing hierarchical self-organizing map. Neurocomputing 48:199–216
Chapter 20
Applying Constraint Satisfaction Methods
in Four-Bar Linkage Design
Gang Chen
Abstract Product design is a very complicated problem. Effective and efficient
computerized methods are hence needed to assist human engineers in original
product design and design modifications. Constraint satisfaction problem (CSP)
provides the insight of problem solving. Many practical algorithms have been
developed along with the exploration of CSP’s theoretical foundation. This paper
applies the general CSP methods into solving product design problems. A crank-
rocker mechanism is used as an example to illustrate the proposed ideas.
Keywords Change propagation Constraint satisfaction Graph theory Product
design
20.1 Introduction
Mechanical products are usually composed of several components or parts. For a
product to realize its designated functions, these parts must be linked in a specified
way. Hence, one of the important issues in product geometric design is finding a
proper configuration and linking relations for all parts. This task is usually not easy
due to the complexity of most mechanical products.
Constraint satisfaction problem (CSP) provides the theoretical foundation and
many practical algorithms for finding solutions to complex constraints among
multiple variables.
This paper applies CSP methods in four-bar linkage mechanism design. Firstly,
theory and basic methods of CSP are briefly introduced; secondly, the four-bar
linkage mechanism design is modeled as a CSP with CSP elements specified.
G. Chen (&)
School of Mechanical Engineering, Tianjin University of Science and Technology,
Tianjin, People’s Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 189
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_20,
Springer-Verlag Berlin Heidelberg 2013
190 G. Chen
How to apply constraint propagation and searching methods to find a solution for
four-bar linkage mechanism is explained. The purpose is to explore automatic and
efficient computerized methods for mechanical or mechatronic product design.
20.2 Constraint Satisfaction Problem
A constraint satisfaction problem is consisted of three elements: a set of variables (V),
a set of possible values for each variable (its domain, D), and a set of constraints
specified among variables (C). Constraints are rules that impose a limitation on the
values that a variable, or a combination of variables, may be assigned simultaneously.
A solution to a CSP is an assignment of a value to each variable from its domain such
that all constraints are satisfied (Dechter 2003).
Figure 20.1 uses an 8-queen problem as an illustration of CSP. The problem is
to place eight queens on an 8 9 8 chessboard satisfying the constraint that no two
queens will be a threat to each other (Tsang 1996). One way to model an 8-queen
problem as a CSP is as follows: each column in the chessboard is treated as a
variable, its domain is all rows in the chessboard, and the constraints are that no
two queens should be placed on the same row, column or diagonal.
Constraint satisfaction problems can be represented as constraint graphs in
which nodes represent variables while arcs represent constraints between two
variables (for binary CSPs). Figure 20.2 is a graph representation of the above-
mentioned 8-queen problem.
Fig. 20.1 A possible
solution to the 8-queen
problem (Tsang 1996)
20 Applying Constraint Satisfaction Methods in Four-Bar Linkage Design 191
Fig. 20.2 Constraint graph of the 8-queen problem
Fig. 20.3 Constraint hypergraph (Dechter 2003)
However, for most practical problems, such as mechanical products design,
since most constraints are specified among multiple variables, simple graphs for
binary constraints are insufficient. Constraint hypergraph is a more general and
suitable representation. Figure 20.3b shows a constraint hypergraph in which each
hyperedge (e.g. the circle around variables 1, 2, 3, 4, 5) represents a constraint
specified on a set of variables. Arcs between hyperedges represent that two related
constraints share one or more variables (labels on arcs).
Two main CSP solving techniques are problem reduction and searching.
(1) Problem reduction: Although problem reduction through propagating and
processing constraints among related variables can not solve a CSP by itself,
however, an original CSP usually can be transformed into an equivalent
problem which is easier to be solved.
192 G. Chen
Fig. 20.4 Constraint propagation
For example, in Fig. 20.4, if the first queen (variable A, which has its value
domain as 1–8) is put at cell (1, A), i.e. assigning variable A value 1, then the
shaded cells in Fig. 20.4a are excluded from domains of the remaining
7 variables. Figure 20.4b shows further constraint propagation after variable B
has been assigned value 3. The problem has become easier to be solved with
domain reduction.
(2) Search: This might be the most fundamental technique for solving CSPs.
Variables are instantiated one by one. With a partial solution, a new variable is
selected and instantiated. If it is impossible to find a feasible value after tra-
versing the new variable’s domain, then the searching process is backtracked to
select a new value for the previous variable. For a CSP with limited domains,
the searching process will carry on until either a solution is found or all the
combinations of possible values have been tried and have failed (Tsang 1996).
Figure 20.5a shows an inconsistent partial instantiation for the first three
variables in a 4-queen problem since no value is feasible for variable D. After
backtracking, i.e. assigning a new value to variable A, a solution is found as
illustrated in Fig. 20.5b.
A lot of improvements have been made to the simple backtracking algorithm,
such as forward checking constraints which involve the most variables or
heuristic-guided backtracking instead of chronicle backtracking.
Each constraint may have a weight that indicates its importance. Weight setting
represents the priority of a constraint during decision-making. Hard constraints
(i.e. constraints that must be satisfied) have higher weights than soft constraints,
Fig. 20.5 Instantiation
through searching (Dechter
2003)
20 Applying Constraint Satisfaction Methods in Four-Bar Linkage Design 193
Fig. 20.6 A prototype system for constraint solving
which are negotiable or can be relaxed if necessary. For practical problems in real
life, a complete solution which satisfies all constraints is usually impossible. It is
more realistic to satisfy as many constraints as possible or the most important
constraints (with higher weights).
The application of CSP modeling and solving techniques in mechanical design
or manufacturing area is rare except for workshop scheduling problem. It might be
due to the following reasons:
(1) Realistic problems in mechanical engineering are usually very complex. For
example, variables may have different types and abstraction levels; constraints
may involve multiple variables and may be represented as complicated
mathematical functions.
(2) Value domains of variables in mechanical engineering are usually unlimited
and continuous. Traditional searching techniques must be modified.
Figure 20.6 shows a prototype constraint solving system developed using
Visual Basic. Interval computing is used to propagate constraints and prune off
value domains. Domain discretization based on the precision requirements can be
used to transform a continuous problem into a discrete problem.
20.3 Four-Bar Linkage Mechanism Design
The planar four-bar linkage mechanism (crank-rocker mechanism as illustrated in
Fig. 20.7) is used here as an example to illustrate how to apply the general CSP
solving techniques to mechanical design problems.
194 G. Chen
Fig. 20.7 Crank-rocker
mechanism (Zhang 2011)
As shown in Fig. 20.7, a planar crank-rocker mechanism has four revolute
joints, A, B, C, and D. Link 4 is fixed as the ground link. Link 1 (crank) and link 3
(rocker) are connected to the ground link. Respectively, they are the input and
output links of the system.
For a four-bar linkage mechanism to work properly, it must obey several
implicitly or explicitly specified rules. For example, let:
S ¼ length of the shortest link
L ¼ length of the longest link
P ¼ length of one remaining link
Q ¼ length of the other remaining link
Then if : S þ LP þ Q
the Grashof condition indicates that at least one link will be capable of making a
full revolution with respect to the ground plane (Norton 1999). Hence, for the
mechanism shown in Fig. 20.7 to be a crank-rocker mechanism, lengths of four
links must follow the Grashof condition. If lengths of four links are modeled as
four variables, then the Grashof condition represented a constraint among variables
that the crank-rocker mechanism must satisfied.
One method to design a crank-rocker mechanism is using the specified rocker
length, rocker oscillating angle, and the travel velocity-ratio coefficient K as input
to synthesize crank length, coupler length, and frame length. The design procedure
is described as follows (see Fig. 20.8):
(1) Calculating limitation location angle h
K1
h ¼ 180 ð20:1Þ
Kþ1
(2) Selecting a position for joint D, using rocker length L3 and rocker oscillating
angle w to determine two extreme rocker positions C1 and C2 (as shown in
Fig. 20.8);
20 Applying Constraint Satisfaction Methods in Four-Bar Linkage Design 195
Fig. 20.8 Design procedure of a crank-rocker mechanism (Zhang 2011)
(3) Determining the position of point F as follows: drawing a line C1F which is
perpendicular to line C1C2; drawing another line C2F with the angle
\C1C2F = 90-h, h is the limitation location angle as calculated in step 1;
(4) Making the circumscribed circle of DC1C2F, another joint A must locate at this
circle;
(5) Using the minimum transmission angle c as the optimization object to finally
determine the position of joint A as well as the frame (AD) length L4;
(6) As shown in Fig. 20.9, the crank length L1 and the coupler length L2 can be
calculated as follows:
Fig. 20.9 Calculating the crank length and the coupler length (McCarthy and Soh 2010)
196 G. Chen
L1 ¼ ðAC2 AC1 Þ=2
ð20:2Þ
L2 ¼ ðAC2 þ AC1 Þ=2
The above-mentioned design procedure can be modeled as a CSP which has the
following variables:
(1) Crank length L1;
(2) Coupler length L2;
(3) Rocker length L3;
(4) Frame length L4;
(5) Travel velocity-ratio coefficient K;
(6) Rocker oscillating angle w;
(7) Transmission angle c;
The lower and upper limits for each variable should be specified. The value
domain for each variable should be discretized based on the precision require-
ments. For example, continuous domain (3–8) for rocker length can be
transformed into discrete domain (3.0, 3.1,…, 7.9, 8.0).
Some intermediate variables, such as limitation location angleh, positions of C1,
C2, F, and A, should also be presented in the constraint graph (see Fig. 20.10).
The Grashof condition, Eqs. (20.1) and (20.2), etc. are modeled as constraints
among variables.
The constructed CSP model is shown in Fig. 20.10. Constraint propagation and
searching techniques can be used to solve the problem and find a feasible solution.
Fig. 20.10 Constraint graph
of crank-rocker mechanism
design
20 Applying Constraint Satisfaction Methods in Four-Bar Linkage Design 197
20.4 Discussion
A lot of issues need to be addressed before the general CSP methods can be
realistically applied into solving mechanical engineering problem (Lottaz et al.
2000; Ribeiro et al. 2008; Ouertani and Gzara 2008; Ermolaeva et al. 2004; Chen
et al. 2006; Zhao et al. 2002; Li and Xiao 2004; Jie and Sun 2007; Xu et al. 2002;
Li and Xiong 2002). These issues are listed as follows:
(1) Design variables usually have different abstraction levels, how to represent
this characteristic using hypergraph;
(2) How to represent complicated mathematical functions as constraints;
(3) How to tackle and manage the complexity of a design problem;
(4) Many disciplines are usually involved in product design, how to represent
them in a constraint network;
(5) Product design usually involve several phases in product life cycle, how to
model the temporal relations among variables in a constraint graph.
20.5 Conclusion
This paper proposes applying CSP methods to solve mechanical design problems
with a crank-rocker mechanism as an example. The purpose is to explore an
effective and efficient computerized method for product design. This work is still
very primitive. A prototype system is currently under development to illustrate the
feasibility of the proposed ideas.
Acknowledgments This research is sponsored by Natural Science Foundation of China
(Grant No.51075300), Research Funding from Tianjin University of Science and Technology
(Grant No.20090407), and Natural Science Foundation of TianJin (Grant No. 10JCYBJC06800).
References
Chen L, Jin G, Deng L (2006) Research on distributed constraint network solving in
multidisciplinary collaborative design. Trans Chinese Soc Agric Mach 37(9):137–140
Dechter R (2003) Constraint processing. Elsevier Science, Amsterdam
Ermolaeva NS, Castro MBG, Kandachar PV (2004) Materials selection for an automotive
structure by integrating structural optimization with environmental impact assessment. Mater
Des 25:689–698
Jie Y, Sun L (2007) Research of the conflict detection of the collaborative design. J Harbin Univ
Sci Technol 12(6):7–9
Li H, Xiao R (2004) Principle of mechanical product design methodology based on constraint
network. Chinese J Mech Eng 40(1):121–126
Li T, Xiong G (2002) Multidisciplinary design based on constraint networks. J Tsinghua Univ
(Sci Technol) 42(9):1257–1260
198 G. Chen
Lottaz C, Smith IFC, Robert-Nicoud Y, Faltings BV (2000) Constraint-based support for
negotiation in collaborative design. Artif Intell Eng 14:261–280
McCarthy JM, Soh GS (2010) Geometric design of linkages. Springer, New York
Norton RL (1999) Design of machinery. McGraw-Hill, New York
Ouertani MZ, Gzara L (2008) Tracking product specification dependencies in collaborative
design for conflict management. Comput Aided Des 40:828–837
Ribeiro I, Pecas P, Silva A, Henriques E (2008) Life cycle engineering methodology applied to
material selection, a fender case study. J Clean Prod 16:1887–1899
Tsang E (1996) Foundations of constraint satisfaction. Academic Press Limited, London
Xu Y, Sun S, Pan Y (2002) Research on constraint-based intelligent collaborated design system
model. China Mech Eng 13(4):333–336
Zhang C (2011) Theory and design of mechanisms and machines. China Machine Press, Beijing
Zhao H, Tian L, Tong B (2002) Constraint-based conflict detection and negotiation in
collaborative design. Comput Integr Manuf Syst 8(11):896–901
Chapter 21
Appraisal of Estate Enterprise’s Social
Responsibility Based on ANP and Fuzzy
Theory
Chun-hua Wu
Abstract The estate industry is one main pole of China’s economy, and it has
done great contribution to economy development as well as improving urban and
rural appearance. But in recent years, the negative reports about building’s quality,
false advertising and so on appear frequently. These have done great damage to
estate enterprises’ social image, so it is necessary to set up one reasonable system
and give out proper method to assess estate enterprises’ social responsibility.
Reasonable appraisal will play an active role in changing estate enterprises’
negative image as well as promoting the realization of harmony society. After
analyzing the existing problems of estate enterprises, this paper sets up one
compressive system and applies ANP and fuzzy theory into the appraisal of estate
enterprise’s social responsibility. At last a case is shown to identify the appraisal
process.
Keywords ANP Estate enterprise Fuzzy theory Social responsibility
21.1 Introduction
Under the big construction and big development background, the estate industry of
China has make great development in recent 10 years, becoming one main pole
of economy and occupying second place in absorbing labor forces only behind of
manufacturing industry (Wu 2011). However, during this process, many problems
appears, especially the absence of social responsibility, such as the false adver-
tising, getting abnormal profit by raising house price deliberately, discard of
ordinary customer’s urgent demand by putting too much resource in developing
C. Wu (&)
School of Architectural and Civil Engineering, An Hui University
of Technology, Ma An Shan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 199
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_21,
Ó Springer-Verlag Berlin Heidelberg 2013
200 C. Wu
high-level residential, and so on (Zheng 2008; Zhang and Qin 2007). It is the time
to do some thing to improve estate enterprises’ social responsibility, not only for
customer’s benefit but also for estate industry’s healthy and sustainable develop-
ment. But relative research in China just start, till now not much papers can be
found, so in order to enrich relative research and do some efforts to improve estate
enterprises’ social responsibility, this paper sets up one comprehensive index
system and gives out one model to appraise estate enterprises’ social responsi-
bility, thus providing some useful references to enterprises and relative govern-
ment department.
21.2 Why Estate Enterprises Should Put Emphasis
on Social Responsibility
21.2.1 The Important Way to Improve Social Image
Because the estate industry of our country founded recently, till now is only about
30 years, relative law and regulation is not perfect, so in recent years, in order to
get windfall profit, some developers sacrifice profit of stakeholders deliberately
(Qu 2007). The negative reports often occur, such as quality problem of house,
bidding up price, peasants’ pay in arrears and so on (Ceng 2007). This phenom-
enon not only despairs customer’s interest, but also destroys the image of estate
industry, thus bringing unfavorable influence to its healthy development. Under
such background, the estate enterprise should take some positive action to improve
its image, so in the process of profit obtaining, the social responsibility should be
put much emphasis. Only by solving the exiting problems felicitously, can estate
industry’s social image will be improved, and then estate industry’s sustainable
and harmonious development can be realized.
21.2.2 The Requirement to Keep in Line with International
Practices
After entering WTO, how to accelerate the speed to take part in international
competition becomes one common problem to domestic enterprises, including
estate enterprises. In recently years, after ISO9000 and ISO14000, now the western
is advocating SA 8000, which is the first worldwide standard pointed to enter-
prise’s social responsibility (Chen and Chen 2006). Its main purpose is to endow
market economy with humanism, so under such background, domestic estate
enterprises must treat social responsibility as important component of develop-
ment, other else, the chance taking apart in international cooperation and com-
petition will be lost.
21 Appraisal of Estate Enterprise’s Social Responsibility 201
21.2.3 Requirement to Cultivate Enterprise’s
Core Competitiveness
The enterprise’s core competitiveness is the integrity of technological core com-
petence, organizational core competence and culture core competence. With the
upgrading of people’s living standard and development of society economy, the
enterprise image and brand consciousness are taking a more and more important role
in enterprise’s core competitiveness. Under such background, the estate enterprises
should aware social responsibility has become a significant part of competitive
advantage in the competition era of brand, so undertaking social responsibility
actively will get public’s approbation, thus improving enterprise’s competitiveness.
21.3 Construction of Appraisal Index System
Since 1970s, researchers began to study enterprise social responsibility, till now
several representative viewpoints have formed. First one is the viewpoint from the
economy, famous scholar Milton Friedman is its main advocator. This viewpoint
holds that enterprises’ only social responsibility is through utilizing resource to
obtain or increase profit under certain game rule. The second one is centered social
economics, to which Howard R. Bowen is the main representative scholar. It
thought increasing profit is not the all of enterprise’s responsibility, besides this,
protecting and upgrading social welfare were also important component (Yan and
Wang 2007). The third one pointed out enterprises should take on 3 aspects social
responsibility: realizing the earnings of finance, society and environment. This is
the famous ‘‘three-dimensional model’’, put forwarded by Archie B. Carroll. Until
early 1990s, more and more scholars and entrepreneurs began to use ‘‘corporate
citizenship’’ to describe enterprise’s social responsibility (Cui et al. 2011). As a
good corporate citizenship, the corporate should be good to all level interest-
related parties, such as customer, employee, district, environment, partnership as
well as stakeholders.
Based on the viewpoints above, this paper sets up one comprehensive index
system from four aspects, including operation and product, employee develop-
ment, social contribution and relation with nature, see Fig. 21.1.
The first aspect mainly concerns enterprise’s operation and product. Estate
enterprise is a business organization, so through smooth and healthy operation and
then bring acceptable profit to stakeholders well as the employees is important,
other else the estate enterprise will lose the most crucial developing power.
Meanwhile, enterprises’ main task is to provide qualified and acceptable product,
this means the house quality is reliable, house type as well as price is acceptable by
ordinary customers.
The second aspect concerns employee’ development. Employee is the most
valuable treasure of enterprise, employee’s benefit should get more attention than
202 C. Wu
return rate of shareholders I11
asset-liability ratio I12
operation and
product I1 ratio of high-grade and ordinary residence I13
complaints times per year I14
career planning I21
Social Employee skill training I22
responsibility development I2 chance of participating decision I23
of safety measure of working environment I24
estate
enterprise situation of paying tax by credibility I31
Social times of public-spirited activity Per year I32
contribution I3
ability of providing employment opportunities I33
utilization rate of construction waste I41
Relation with
contribution rate of technology advance I42
nature I4
enthusiasm of developing energy-saving building I43
Fig. 21.1 The index system of social responsibility
before, because of this, the SA 8000 put employee’s benefit on the most important
status. Employee’s development not only means salary, but also refers upgrading
employee’s skill level, improving working environment and so on.
The third aspect concerns estate enterprise’s social contribution. Estate enter-
prise is one component of society, so it should do proper social contribution. This
means estate enterprises should obey relative laws, pay tax on time, create
employment chance, and be active in social public service.
The fourth aspect mainly concerns estate enterprises’ environment protection
consciousness. The estate enterprises of China doesn’t get rid of extensive growth
pattern, so high resource and energy consuming, low level of productivity and low
contribution of scientific development are still the main characters of estate
enterprises. According to estimation, compared with advanced countries, the estate
industry of China consumes 3–5 times resource, while providing merely 1/5–1/6
productivity of these countries. On the same time, in the process of development,
some estate enterprises ignore the rule of nature, so the lake is filled and farmland
is exploited and so on, leading to ecological environment deterioration (Li 2012).
21.4 Construction of Appraisal Method
Till now the research about estate enterprises’ social responsibility mainly stay on
qualitative analysis level, such as the sense to improve estate enterprise’s social
responsibility, the index study as well as the relation of improving social
responsibility and its healthy development. Although exiting research has play an
21 Appraisal of Estate Enterprise’s Social Responsibility 203
active role in improving estate enterprise’s social image at some degree, but it is
necessary to set up one more comprehensive index system and give out one
quantitative model to measure the level of society more objectively, which can
provide more information to enterprise and government.
The appraisal method put forward by this paper is as follow (Fig. 21.2).
Phrase1: index value fixation. To measure the value of every index scientifically is
first key step of appraisal. As to quantitative index, its value is got directly from its
financial statements or statistical data, based on appraisal subject’s performance.
While to qualitative index, it is difficult to use one objective value, so in order to get
more accurate measurement value, this paper will apply unascertained rational
number to do this task. Unascertained rational number is put forwarded firstly Guang-
Yuan Wang, academician of the Chinese Academy of Engineering. This method can
identify all the possibility and as well as experts’ reliability, thus avoiding the
information distortion or omission. Till now it has been applied in practice widely,
such as production decision, data processing, interval analysis and so on.
Phrase2: index weight fixation. The indexes are not of equal importance, so we
need to fix every index’s weight scientifically. In this paper, the task will be
realized by Analytic Network Process (short of ANP), which firstly put forward by
American Thomas L. Satty at 1996, It developed from AHP, so they have nearly
same decision principle, but their decision structure is different, for ANP network
structure is applied while in AHP is hierarchical structure (Sun et al. 2007). ANP
model can take into the feedback of different levels’ indexes and the inner loop
relation of the same level indexes, while traditional AHP only emphasizes the
domination of upper layer to lower layer and suppose the indexes under same level
is independent, so compared with AHP, ANP is inefficient in tackling complex
problem (Zhang et al. 2012a; Guo and Bai 2011). In a word, ANP can make the
analysis process more close to practical situation, thus making the analysis result
more efficient and more reliable.
Phrase3: social responsibility comprehensive appraisal. On the base of frontal
two phrases, in this phrase fuzzy theory will be used to appraise estate enterprises’
social responsibility. Because in the appraisal process, there is full of unascer-
tained information, so the fuzzy model is adapted and will make the appraisal
qualitative
indexes Phrase 3
Phrase 2
Phrase 1 Unascertained
Analytic Improved
rational Social
network fuzzy
number responsibility
The fixation process model
Fixation of quantitative comprehensive
index weight
index value indexes appraisal
objective value
Fig. 21.2 The main appraisal phrases and its method
204 C. Wu
result more objectively (Ma et al. 2007). But traditional fuzzy model has some
deficiency needing to be overcome, such as the max–min in algorithm of fuzzy set,
the maximum membership degree recognition criteria to ranked evaluation grade
(Liu et al. 2009; Zhang et al. 2012b). So in the application process, some
improvement will be done to fuzzy model from the two aspects.
21.5 The Application
In order to show the appraisal process of the model put forward by this paper as
well as identify its reasonability, this paper applies it to one estate of An Hui
Province.
21.5.1 Fixation of Index Value
To the quantitative index the value is get from this enterprise’s financial statements
or statistical data of relative government department. While to qualitative indexes,
we invite 4 experts to grade from 100 to 0 based on the logic of unascertained
rational number. For example, take index I13 as an example to show the grading
process. Specifically speaking, four experts’ grade intervals to this index are
(75–80), (70–80), (80–85) and (70–75) respectively, and because experts’ expe-
rience and knowledge structure is different, so different experts is allocated dif-
ferent credibility, which is 0.3,0.2, 0.3 and 0.4 respectively. Then based on
unascertained rational number theory, the credibility function is fixed:
8
> 0:3 x 2 ð70; 75Þ
>
>
< 0:4 x 2 ð75; 80Þ
f13 ðxÞ ¼
>
> 0:15 x 2 ð80; 85Þ
>
:
0 other
Through computation, the grade of I13 77.5 can be got. By the same way, other
quantitative indexes’ grade can be got.
Finally, we can get all indexes’ value, as shown in Table 21.1.
Table 21.1 Value of indexes
Index I11 I12 I13 I14 I21 I22 I23
Value 27 % 30 % 77.5 3 78 75 82
Index I24 I31 I32 I33 I41 I42 I43
Value 72 27 % 30 % 77.5 3 78 75
21 Appraisal of Estate Enterprise’s Social Responsibility 205
Fig. 21.3 The network of
indexes social responsibility Target
appraisal layer
Operation and employee
product development
network
Layer
social relation
contribution with nature
21.5.2 Weight Fixation
In order to overcome the shortcoming of AHP, this paper turns to ANP, which can
make the weights more reasonable and scientific. First the network hierarchy is set
up as follow (Fig. 21.3).
Then this paper uses software Super Decision, which is designed specifically to
ANP, to realize the computation process. The weight get from the software are
shown in Table 21.2.
21.5.3 Comprehensive Appraisal Using Improved Fuzzy
Theory
Firstly, based on the experts’ advice and industry standard, this paper sets up
indexes’ rank standard, including quantitative and qualitative indexes, see
Table 21.3. Then the membership function of every index can be set up.
Secondly, based on the logic of fuzzy theory (Qu 2007), substituting index’s
value into corresponding membership function, the membership matrix is got. But
in computation process, some change is done to fuzzy theory’s max–min algorithm
following the way of reference (Ceng 2007).
Finally, the comprehensive appraisal vector V is got,
Table 21.2 Weights fixed by ANP
First level I1 I2 I3 I4
index
Duster 0.29 0.22 0.26 0.23
weights
Second level I11 I12 I13 I14 I21 I22 I23 I24 I31 I32 I33 I41 I42 I43
index
Local 0.26 0.29 0.21 0.24 0.28 0.22 0.31 0.19 0.35 0.38 0.27 0.34 0.35 0.31
weights
206 C. Wu
Table 21.3 Rank of indexes
Rank Low Relativity General Relativity high High
Index
I11 (%) B5 5–10 10–20 20–30 C30
I12 (%) C80 70–80 50–70 40–50 B40
I14 C15 10–15 5–10 5–2 B2
I32 B3 3–6 6–9 9–12 C12
I41 (%) B20 20–40 40–60 60–80 C80
I42 (%) B25 25–40 40–55 55–70 C70
Qualitative indexes B50 50–60 60–70 70–80 C80
2 3
0:578 0:422 0 0 0
6 0:478 0:465 0:057 0 07
6 7
V¼½ 0:29 0:22 0:26 0:23 6 7
4 0 0:14 0:545 0:315 0 5
0 0 0:408 0:592 0
¼ð 0:273 0:261 0:248 0:218 0 Þ
For the index rank is ordered, so the maximum subordinate recognition rule is
not perfect, then this paper turns to incredible recognition rule. Suppose threshold
value k ¼ 0:7, and then we can know this enterprise’s social responsibility level
belongs to rank general. From analysis process, we also can know enterprise’s bad
performance on social contribution and relation with nature is the main reason
lowering its social responsibility level.
21.6 Conclusion
Pointed to the problem of estate enterprise’s social responsibility, this paper set
ups one comprehensive index system, and then applies ANP and fuzzy theory into
responsibility appraisal process. One specific case is given, which identity this
model is feasible and reasonable.
The study of this paper can provide one tool to assess estate enterprise’s
responsibility quantitatively, thus avoiding traditional qualitative methods’ short-
comings. In a word, the study of this paper will do good to estate industry’s healthy
development as well as the realization of harmony society.
References
Sun H-C et al (2007) The appraisal of emergence bridge project design using ANP. Eng Sys
Theor Pract 3:63–70
Ceng R (2007) Study of estate enterprise’ social responsibility under context of harmonious
society. Econ Rev 8:116–118
21 Appraisal of Estate Enterprise’s Social Responsibility 207
Chen C, Chen C (2006) Study of construction estate’s social responsibility (in Chinese). Eng
Constr 20(4):298–399
Cui M, Hua J, Xiao C (2011) Research on evaluation system of construction corporate social
responsibility based on data envelopment analysis method. Sci Technol Eng
11(34):8658–8664
Guo W, Bai D (2011) Selection of manufacturing suppliers with ANP/TOPSIS. J Wuhan Univ
Technol (Information & Management Engineering) 33(1):147–151
Li Y (2012) Core estate cultivation of estate under social responsibility orientation. Knowl Econ
6:109–110
Liu K et al (2009) Fuzzy comprehensive evaluation for conceptual design of mechanic product
based on new membership conversion (in Chinese). J Mech Eng 45(12):162–166
Ma G, Mi W, Liu X (2007) Multi-level fuzzy evaluation method for civil aviation system safety
(in Chinese). J Southwest Jiao Tong Univ 42(1):104–109
Qu Y (2007) Thinking to social responsibility of estate enterprise (in Chinese). Constr Econ
7:94–96
Wu G (2011) The analysis of estate enterprises’ social duty and its realization using game theory.
Constr Econ 8:23–26
Yan W-Z, Wang D (2007) Appraisal of construction estate’s social responsibility based on fuzzy
and AHP. Constr Econ 12:46–48
Zeng R-P (2008) Estate enterprise’s social responsibility study under background of harmony
society. Econ Rev 4:116–118
Zhang L, Qin P (2007) Construction of corporate social responsibility management system in
construction industry based on SA8000 (in Chinese). J Tian Jin Univ (Social Sciences)
9(5):416–419
Zhang S, Hou X, Yang Q (2012a) Performance measurement of contractor based on BSC and
ANP. Stat Decis 1:175–178
Zhang G, Liu J, Wang G (2012b) Fuzzy reliability allocation of CNC machine tools based on
task. Comput Integr Manuf Syst 18(4):765–771
Chapter 22
Bottleneck Detection Method Based
on Production Line Information
for Semiconductor Manufacturing System
Xiao-yu Yu, Fei Qiao and Yu-min Ma
Abstract Semiconductor wafer fabrication system is a typical complex
manufacturing system, since it has large-scale, reentrant, multi-objective, uncer-
tain and other characteristics. It’s too difficult to achieve the capacity balance to
lead the existence of the bottleneck. According to the theory of TOC, the accurate
detection of bottleneck is the key to implement DBR thought. For the character-
istics of semiconductor production line, this paper proposes a bottleneck detection
method based on the starvation and blockage information of the production line.
The method is verified on HP-24 model by simulation. Compared to the relative
load method, the equipment utilization law and the queue length method; the
experimental results show that this method makes performance better than them.
Keywords Bottleneck detection DBR Production line Reentrant
22.1 Introduction
Drum-Buffer-Rope (DBR) theory is put forward by Doctor Goldratt based on the
theory of constraints to solve the production scheduling problem (Rahman 1998).
According to the theory, the whole system’s output is decided by bottleneck’s
output and maximizing the utilization of the bottleneck’s capacity is the key to
improve system productivity and economic benefit. Bottleneck detection is a very
important step for the operation and management of manufacturing enterprise,
because it will not only affect the decision of the feeding strategy, but also
influence the dispatching of the jobs on bottleneck equipment.
X. Yu (&) F. Qiao Y. Ma
The School of Electronics and Information Engineering,
TongJi University, Shanghai, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 209
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_22,
Ó Springer-Verlag Berlin Heidelberg 2013
210 X. Yu et al.
At present, much research effort has been devoted to bottleneck detection and
can be divided into two categories. One is to detect the bottleneck before the start
of production system. Literature (Zhang and Wu 2012) firstly establishes an
optimization model by reducing some traditional constraints of a standard Job-
shop, then calculates the bottleneck’s characteristic value and chooses the excel-
lent by the simulated annealing algorithm. Literature (Zhai et al. 2010) uses the
orthogonal table and different assigned rules to construct a test program, with
production system job target as measurement index to identify bottleneck. The
other method to identify bottleneck is to conduct collection, imitation and simu-
lation analysis of the data from the production system which has been online for a
period of time. As in (Roser et al. 2002, 2003), system bottlenecks are divided into
independent bottlenecks and shift bottlenecks, they are identified by calculating
the maximum active time of the machines, which is similar to the common
equipment utilization method. Literature (Li et al. 2007, 2009; Wang et al. 2008) is
based on the data of the production line, the blockage and starvation information
are made full use of to detect bottlenecks. Literature (Kasemset 2009; Kasemset
and Kachitvichyanukul 2009, 2010) classifies the candidate bottleneck equipment
according to the static and real-time data from simulation and testifies the cor-
rectness of the bottleneck equipment selection via confidence interval level.
Semiconductor wafer fabrication system is recognized as one of the most
complex manufacturing systems, which normally contains as many as three or four
hundreds processing steps. In addition, wafer manufacturing has reentrant char-
acteristics, namely the same product will go through some processing center more
than once, which is different from the traditional Job-shop and Flow-shop system
(Wu et al. 2006). Most methods mentioned above are applicable to Flow-shop;
some can be used for Job-shop as in (Zhai et al. 2010). It will ignore some
stochastic disturbance such as equipment failure or seasonal variation of need
though it is faster and more convenient. Based on literature (Li et al. 2007), this
paper mainly studies the semiconductor wafer fabrication system by changing its
constraints and proposing a concept of relatively blocking rate, which can record
the detailed variation of the count of the jobs in the buffer throughout the whole
production system operation period. The method can make full use of outline and
online information of the production line, and it is not necessary to consider
machine sets, type, product type, processing route and all kinds of factors such as
random fluctuation in the process of identification. Therefore, it is a convenient
and accurate identification method.
22.2 Common Bottleneck Detection Methods
Due to the high complexity of the semiconductor manufacturing system, the
existing manufacturing system bottleneck identification methods are not suitable
as in (Sengupta et al. 2008), which identifies bottlenecks by analysing the
departure time among the equipments. However, it does not exist the logic
22 Bottleneck Detection Method 211
upstream and downstream equipments in semiconductor wafer fabrication system
because of its serious reentrant characteristic. At present, common methods of
bottleneck detection in semiconductor manufacturing system are as follows.
22.2.1 Analyze the Queue Length of the Equipment
In this approach, the queue length or waiting time of the equipment is measured
and the one which has the longest queue length or waiting time is considered as
manufacturing bottleneck, as is shown in formula 22.1
max
Bottleneck ¼ Machinej ¼ max 1 j m 0\t T ðWtj Þ ð22:1Þ
where T stands for simulation period, m is the number of the processing center in a
system or a model, Wtj is the number of jobs in the buffer for equipment j at a
certain time t during the simulation period.
22.2.2 Measure the Utilization Rate of the Equipment
The equipment which has the highest utilization rate is system bottleneck as is
shown in formula (22.2) (Zhou and Rose 2009).
max WTj ðTÞ +OTj ðT)
Bottleneck = Machinej ¼ 1 j m ð22:2Þ
T
where T is the time period considered, m is the number of the processing center in
a system or a model, WTj(T) and OTj(T) are the processing time and off-line time
of equipment j during time period T.
22.2.3 Calculate the Relative Load of the Equipment
On the basis of the order, the load of every processing center is calculated
according to the job’s craft, and the machine which has the maximal relative load
is system bottleneck (Ding et al. 2008), as is shown in formula (22.3) and (22.4).
LB = Max(Lh Þ ð22:3Þ
X
x Xy htij
Lh ¼ qi j¼1 l
ðh ¼ 1; 2; . . .mÞ ð22:4Þ
i¼1
212 X. Yu et al.
where LB is the load of the system bottleneck, work center B is system bottleneck,
i is the type of the job, x is the number of the type of the job, y is the step number
of the job, h is the related coefficient of the equipment (if some job is processed in
the center, then h is 1, otherwise h is 0), tij is the processing time of step j of job i.
The method can identify the system bottleneck through a simple calculation using
some relevant technological parameters.
22.3 Bottleneck Detection Based on the Information
of the Production Line
Using the blockage and starvation information in buffer to detect bottlenecks is a
method based on data, the basic idea is that a bottleneck machine will often cause
the upstream machines to be blocked and downstream machines to be starved,
therefore, the bottleneck machine will often have a lower total blockage plus
starvation time than its adjacent machines. LIN LI validates his theory according
to the thought that the disturbance of the bottleneck equipment has the largest
impact on the system. At last, he applies his idea to the production line including
three and more machines (Li et al. 2007).
Semiconductor production line has a great difference with general industrial
manufacturing line for its complex processing flow and serious reentrant charac-
teristics. In the actual production line, bottleneck equipment is always the one
which owns more reentry times and the jobs in the bottleneck buffer possibly come
from multiple upstream machines, therefore, upstream or downstream machines
can’t be judged by physical location. In addition, the starvation rate and blockage
rate can’t be simply added together because there are lots of parallel and group
equipments. This paper proposes a new method based on the thought of utilizing
historical information in the production line aiming at the characteristics of
semiconductor production line. At first, the capacity of buffer is set to be infinite. If
the capacity is a fixed value, it will lead to buffer overflow when the production
line is crowded and the block degree of different machines can’t be distinguished.
Secondly, the formula of the starvation rate is defined according to its basic
concept. Finally, this method puts forward a concept of relative blocking rate and
the formula is defined. The details are as follows.
(1) Starvation rate The idle time divided by processing period reflects when the
machine is leisure in the production process. It can be expressed as Si ¼ Tidle
T
(i is equipment number, Tidle is leisure time, T is processing period);
(2) Relative blocking rate generally, each buffer will appear the phenomenon of
accumulation, this paper proposes the concept of relative blocking rate to
distinguish the congestion degree among different equipment at different
times, namely the equipment is relative to other equipments in degree of
obstruction. The computing method is: obtain the number of jobs in each
22 Bottleneck Detection Method 213
buffer at regular time called qi (i is equipment number), calculate the total
queue length of jobs of all the equipment buffers at the present moment called
Pn
qi (n is the number of all the equipment), then the relative blocking rate of
i¼1
q
the equipment is Bi ¼ Pn i : Assume that the collection period is Dt, the total
i¼1
qi
P qi
Dt
Pn
i¼1
qi
simulation time is T, then the relative blocking rate is dBi ¼
T
In the real production line, bottleneck machine is the weak link of the whole
system and its processing ability is the weakest. For that reason, the buffer before it
often has a large accumulation of jobs waiting for processing. Compared to the other
machines, blockage occurs more frequently in bottleneck machine and starvation is
on the contrary. Thus, the starvation rate plus the opposite number of the relative
blocking rate will be the least of all the equipments. To avoid confusion, the opposite
number of the relative blocking rate is defined as non-blocking rate which is equal to
1 dBi . According to the above description, the method is expressed as:
Bottleneck = Machinei ¼ min ð1 dBi + Si Þ
0\i n
0 P qi 1
Tidle Dt Pn
B qC
B 1 iC
¼ min B1þB C ð22:5Þ
0\i n@ T C
A
where i is equipment number and n is the total number of the equipments.
Because of the high complexity of semiconductor manufacturing system, it may
exist multiple bottlenecks (Cao et al. 2010). When this method is used to detect
bottleneck, the distribution of starvation rate and non-blocking rate of each
equipment should be analyzed. For the equipment whose result of starvation rate
plus non-blocking rate is similar to the system bottleneck can be considered as the
second bottleneck. For the main aiming of this paper is single bottleneck detection,
we don’t make much analysis on multiple bottlenecks.
22.4 Example Validation
This paper chooses the model of HP-24 semiconductor production line as object of
study and EM-PLANT as simulation platform. HP-24 Model comes from silicon
wafer production technology center lab and most parameters are collected from
real devices. There are 24 equipment groups in the model and most of them are
single machine except the lithography (one group contains two machines, the other
group contains three machines). As one kind of simplified model, HP-24 only
214 X. Yu et al.
Table 22.1 Parameters for machines in Hp-24 model (Murphy and Dedera 1996) and simulation
data
Machine group Count Reentrant times Starvation rate (%) Nonblocking rate (%)
ID Name
1 CLEAN 1 19 14.73 97.72
2 TMGOX 1 5 23.97 99.31
3 TMNOX 1 5 20.89 99.50
4 TMFOX 1 3 58.23 99.90
5 TU11 1 1 83.30 100.00
6 TU43 1 2 56.30 99.98
7 TU72 1 1 81.39 100.00
8 TU73 1 3 59.87 99.93
9 TU74 1 2 71.97 99.98
10 PLMSL 1 3 65.82 99.96
11 PLMSO 1 1 78.44 100.00
12 SPUT 1 2 65.85 99.97
13 PHPPS 2 13 14.22 99.05
14 PHGCA 3 12 5.00 65.33
15 PHHB 1 15 63.23 99.94
16 PHBI 1 11 6.00 64.45
17 PHFI 1 10 53.88 99.93
18 PHJPS 1 4 57.36 99.91
19 PLM6 1 2 22.84 99.80
20 PLM7 1 2 67.12 99.96
21 PLM8 1 4 13.29 99.39
22 PHWET 1 21 37.82 99.44
23 PHPLO 1 23 26.00 99.36
24 IMPP 1 8 9.05 100.00
processes one type of products which has 172 processing steps (Ding et al. 2008).
During the simulation period, the buffer information is collected every half an hour
and the feeding method is subject to uniform distribution. The simulation period is
set to be 1 year and the data collected is shown in Table 22.1. The starvation rate
and non-blocking rate of each equipment are calculated by the formulas introduced
in the third section.
As is shown in Table 22.1 and Fig. 22.1, the starvation rate of machine 14 is
0.05, the non-blocking rate is 0.64, so the starvation rate plus the non-blocking is
0.69, which is the smallest of all machines. Thus, Machine 14 can be considered as
the bottleneck machine according to the thought of the method based on pro-
duction line information. As machine 14 is a parallel processing machine, it can be
regarded as the bottleneck processing center.
Bottleneck machine is the short and fat son of the system, so the system output
depends on the processing speed of bottleneck machine. The following is com-
parisons among different bottleneck detection methods.
(1) To prove the effectiveness of the machine group 14, we feed jobs into the
production line based on the processing speed of machine 14. Through
22 Bottleneck Detection Method 215
Fig. 22.1 Bar graph of starvation rate and nonblocking rate
calculation, the processing speed of machine 14 is L14 ¼ 31:28h, namely
machine 14 needs 31.28 h to process one lot of product.
(2) When we use the relative load method to identify bottlenecks, machine 16
holds the largest load according to the calculation formula introduced in Sect.
22.2. Thus, machine 16 is the bottleneck machine and its processing speed is
L16 ¼ 32:56h.
(3) Under the condition that the feeding speed obeys the uniform distribution, the
equipment utilization of machine 24 is the highest. Thus, machine 24 is the
bottleneck machine and its processing speed is L24 ¼ 30:88h by the method.
(4) By analyzing the queue length of every machine in the whole process,
machine 1 has the longest queue length and it is the system bottleneck with
processing speed L1 ¼ 29:45h.
The feeding tables of all kinds of methods are drawn up according to the
analysis above and they are validated on HP-24 model by simulation. The dis-
patching rule of bottleneck machine and non-bottleneck machines is FIFO (first in
first out). The strengths and weaknesses of different feeding methods and sched-
uling strategies need to be evaluated by performance and the common perfor-
mance indexes including:
(1) Average processing period. In the reentrant manufacturing system, the time
from when a raw job is put into the production line to the time the job leaves
the system is processing period, which can be expressed as CT ¼ Tout Tin
Tout is the time the job leaves the processing system as finished product, Tin is
the time the job enters the system. The scheduling goal is to make the average
processing period minimum.
(2) Productivity. Productivity refers to the number of finished products per unit
Q
time, the formula is PR = : Q is the number of the finished lots during the
T
processing period, T is processing period. Productivity is inverse to processing
216 X. Yu et al.
Table 22.2 Performance of different bottleneck detection methods
Average Processing Productivity Wip Bottleneck
processing period machine
period variance utilization
Production information 60024.49 6961.08 0.655 29.5 0.95
Relative load 63869.54 7097.10 0.643 30.57 0.93
Equipment utilization 78717.38 14856.57 0.634 39.96 0.921
Queue length 83229.23 20667.08 0.623 43.37 0.869
period. The shorter the processing period is, the higher the productivity will
be. Productivity determines the cost of final product, processing period, cus-
tomer satisfaction and so on.
(3) WIP (work in process) is the number of products in process online every day
and the scheduling goal is to make the index minimum.
(4) Utilization rate of bottleneck machine. The formula is UB ¼ TTopen , Twork is the
work
time that the machine is in the state of process, Topen is the uptime of the
machine. An overview of the comparison of different performance of each
method is shown in Table 22.2
From Table 22.2 we can see that when we use the information of the production
line to detect the bottlenecks, for average processing period, there is a 6.0 %
reduction compared to the relative load method, a 23.7 % reduction compared to
the equipment utilization, a 27.9 % reduction compared to the queue length. For
processing period variance, there is a 1.9 % reduction compared to the relative
load method, a 53.1 % reduction compared to the equipment utilization, a 66.3 %
reduction compared to the queue length. For productivity, there is a 1.9 % increase
compared to the relative load method, a 3.3 % increase compared to the equipment
utilization, a 5.1 % increase compared to the queue length. For average day wip,
there is a 3.5 % reduction compared to the relative load method, a 26.2 %
reduction compared to the equipment utilization, a 32.0 % reduction compared to
the queue length. For bottleneck machine utilization, there is a 2.2 % increase
compared to the relative load method, a 3.1 % increase compared to the equipment
utilization, a 9.3 % increase compared to the queue length. The relevant results are
displayed in Table 22.3.
Table 22.3 Performance analysis of the other three bottleneck detection methods compared to
the proposed method
Average Processing Productivity (%) WIP (%) Bottleneck
processing period machine
period (%) variance (%) utilization (%)
Relative load -6.0 -1.9 +1.9 -3.5 +2.2
Equipment utilization -23.7 -53.1 +3.3 -26.2 +3.1
Queue length -27.9 -66.3 +5.1 -32.0 +9.3
22 Bottleneck Detection Method 217
Data analysis presents that the proposed method in this paper has different range of
ascension than other methods and thus proves its practicality and effectiveness.
22.5 Conclusion
Detecting bottleneck accurately is the first step to implement the DBR thought.
Common bottleneck detection method has some limitations in the complex
semiconductor manufacturing system. This paper utilizes the production line
information to detect bottleneck which is based on data mining and obtains good
effect. Historical data underlies the process information of manufacturing system
and it can be regarded as a knowledge base which should be made full use of to
detect bottlenecks. Future work: 1. There are many uncertain factors in a real
production line, a single bottleneck feeding strategy may not be achieved good
effect all the time, it should be combined with the bottleneck scheduling strategy.
2. How to further mine the underlying experience, knowledge and rules of the
historical and online data to optimize the production line needs further study.
Acknowledgments This research was supported by Chinese National Natural Science Foun-
dation (61034004), Science and Technology Commission of Shanghai (10DZ1120100,
11ZR1440400), Program for New Century Excellent Talents in University (NCET-07-0622) and
Shanghai Leading Academic Discipline Project (B004).
References
Cao Z, Peng Y, Wu Q (2010) DBR-based scheduling for re-entrant manufacturing system (in
Chinese). In: 2010 Proceedings of contemporary integrated manufacturing system, vol 2010,
pp 566–572
Ding X, Qiao F, Li L (2008) Research of DBR scheduling method for semiconductor
manufacturing (in Chinese). Integr Mech Electr 2008:29–31
Kasemset C (2009) TOC based Procedure for Job-shop Scheduling. Doctoral Dissertation,
Industrial Engineering and Management, School of Engineering and Technology, Asian
Institute of Technology, Bangkok, Thailand
Kasemset C, Kachitvichyanukul V (2009) Simulation tool for TOC implementation. In:
Proceedings of ASIMMOD 2009, ASIMMOD 2009, Bangkok, Thailand, pp 86–97
Kasemset C, Kachitvichyanukuf V (2010) Effect of confidence interval on bottleneck
identification via simulation. In: Proceedings of the 2010 IEEE IEEM. IEEE, DC, USA,
pp 1592–1595
Li L, Chang Q, Ni J (2007) Bottleneck detection of manufacturing systems using data driven
method. In: Proceedings of IEEE international conference on symposium on assembly and
manufacturing. IEEE, Washington, DC, USA, pp 76–81
Li L, Chang Q, Ni J (2009) Data-driven bottleneck detection of manufacturing systems. Int J Prod
Res 47(18):5019–5036
Murphy RE Jr, Dedera CR (1996) Holistic toc for maximum profitability. In: 1996 IEEE/SEMI
advanced semiconductor manufacturing conference
218 X. Yu et al.
Rahman S (1998) Theory of constraints: a review of the philosophy and its applications. Int J
Oper Prod Manag 18(4):336–355
Roser C, Nakano M, Tanaka M (2002) Shifting bottleneck detection. In: Proceedings of the 34th
conference on winter simulation. IEEE, Washington, DC, USA, pp 1079–1086
Roser C, Nakano M, Tanaka M (2003) Comparison of bottleneck detection methods for AGV
systems. In: Proceedings of the 2003 winter simulation conference. IEEE, Washington, DC,
USA, pp 1192–1198
Sengupta S, Das K, VanTil RP (2008) A new method for bottleneck detection. In: Proceedings of
the 2008 winter simulation conference. IEEE, pp 1741–1745
Wang Z, Chen J, Wu Q (2008) A new method of dynamic bottleneck detection for semiconductor
manufacturing line. In: Proceedings of the 17th world congress the international federation of
automatic control, vol 17(1). Elsevier, Seoul, Korea, pp 14840–14845
Wu Q, Qiao F, Li L (2006) Semiconductor manufacturing system scheduling. Publishing House
of Electronics Industry, Beijing
Zhai Y, Sun S, Wang J, Wang M (2010) Bottleneck detection method based on orthogonal
experiment for job shop (in Chinese). Comput Integr Manuf Syst 16(9):1945–1952
Zhang R, Wu C (2012) Bottleneck machine identification method based on constraint
transformation for job shop scheduling with genetic algorithm. Inf Sci 188:236–252
Zhou Z, Rose O (2009) A bottleneck detection and dynamic dispatching strategy for
semiconductor wafer fabrication facilities. In: Proceedings of the 2009 winter simulation
conference. Institute of Electrical and Electronics EngineersInc, Austin, TX, USA,
pp 1646–1656
Chapter 23
Clinical Decision Support Model of Heart
Disease Diagnosis Based on Bayesian
Networks and Case-Based Reasoning
Man Xu and Jiang Shen
Abstract To boost the accuracy of clinical decision support systems and degrade
their misdiagnosis rates, a hybrid model was proposed with Bayesian networks
(BN) and case-based reasoning (CBR). BN were constructed with the feature
attributes and their casual relationships were learned. The similarities of feature
attributes were measured with the case matching method, as well as the knowledge
of their dependent relationships. Therefore, the accuracy of the diagnosis system
was enriched through the dynamic retrieval method.
Keywords BN CBR Heart disease diagnosis
23.1 Introduction
Reasoning mechanism is one core component of clinical decision support systems
(CDSS). Rule-based reasoning (RBR) is a popular inference technology in most of
the existing expert systems (ES), as well as model-based reasoning (MBR).
However, the relationship between information and knowledge is complex and
fuzzy (Kong et al. 2008) in the area of healthcare, especially in the process of
medical diagnosis. But the medical rules are more difficulty in acquiring and
matching.
Cased-based reasoning (CBR) is the process of solving new problems based on
the solutions of similar past problems. In CBR, processes like retrieval and
matching are typically assumed to be broadly general cognitive processes.
M. Xu J. Shen
TEDA College, Nankai University, Tianjin, China
M. Xu J. Shen (&)
College of Management and Economics, Tianjin University, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 219
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_23,
Ó Springer-Verlag Berlin Heidelberg 2013
220 M. Xu and J. Shen
Therefore, CBR is often used to solve complex problems with incomplete
knowledge or difficulty in acquiring rules. The critical part of CBR is to find a
suitable case retrieval method, which is determined to the efficiency and accuracy
of the solution.
The process of case retrieval relies on the similarity measures for all the case
attributes. The weights and similarities of features are given by physicians during
medical diagnosis. Although the feature weights are taken into account with several
algorithms, including feature evaluation (Shiu et al. 2001), introspective learning
(Zhang and Yang 2001) and Neural Network (Zhang and Yang 1999), they omitted
adaption strategies. Furthermore, Nearest Neighborhood (NN) method is adapted to
case retrieval (Gu et al. 2003; Ling et al. 2006), but the solution cannot regenerate
with database updating. The accuracy of NN decreases rapidly when dealing with
incomplete data of medical problem, as well as its retrieval efficiency.
This paper built a hybrid model integrating CBR with Bayesian Networks (BN)
for clinical decision supports (BN-CBR). Compared with conventional CBR,
BN-CBR contributes to the optimization of the storage and retrieval process for
medical cases, which improves the efficiency of CDSS. Moreover, BN-CBR aims
to boost the accuracy of reasoning and to degrade the rate of misdiagnosis.
23.2 CDSS Based on the BN-CBR Hybrid Model
The diagnosis process of a clinician involves in synthesizing the information of a
patient. This process also involves the circle of observation, diagnosis and treat-
ment. From the perspective of clinic, the tasks of observation mainly consist of
acquiring and mining the information of a patient completely, which reduces the
uncertainty of the patient’ condition to the greatest extent. With the available
information, the clinician administers the treatment to the patient as well as his/her
experience and medical knowledge. This diagnosis process can be regarded as a
reasoning process, and its treatment is a solution for the medical problem. The
treatment relies on the diagnosis and decision-making process, while its accuracy
are determined by the belief and the adequacy of the acquired information during
the period of observation (Ma et al. 2002). The framework of CDSS based on the
BN-CBR model is demonstrated as Fig. 23.1.
As in Fig. 23.1, queries demonstrate the patients’ condition (medical cases).
The processor of CBR obtains the most similar case through the similarity mea-
sure. The conditional probabilities among the attributes of the cases are inferred
through data training on Bayesian Networks (BN), which is an effective data
mining tool. Therefore, the similarity function is formulated to measure the degree
of similarity among cases. Medical knowledge and the information of patients are
stored in the medical database, which is integrated into casebase after data
screening and processing. The casebase as the main knowledge source for rea-
soning can be accumulated, refined and updated during the diagnosis reasoning
episodes.
23 Clinical Decision Support Model 221
Fig. 23.1 The framework of
CDSS based on the BN-CBR Database Data Processing
hybrid model
Bayesian Information
Networks Gain
Casebase
Similarity Optimal
CBR
measure feature set
Most Similar Queries &
Case Feature weights
Clinicians
Yes Add No
New Case Suggestion
23.2.1 Knowledge Representation
Definition 23.1 A medical case is represented as
Ci ¼ ðIndex; X1 ; X2 ; . . .; Xm ; SÞ; Ci 2 CB; Xi 2 X ð23:1Þ
where CB denotes the casebase, which is a finite set; X denotes the set of medical
features; Index denotes the index of medical cases; S denotes the consequence or
suggestion of the diagnosis.
Take the diagnosis of Coronary heart disease (CHD) for example. CB = (#,
Age, Gender, Family history of CHD, chest pain type, blood pressure, cholesterol,
class of CHD), where # is the index of cases. The class of CHD is binary, 0
indicating CHD while 1 indicating other health condition.
Definition 23.2 A medical case database is represented as CB ¼ ½xij mn , where xij
denotes the jth attribute of the ith medical case, m denotes the number of features,
and n denotes the size of casebase. CB consists of the specific knowledge of
previous experiences or historical examples.
Definition 23.3 A medical query is represented as
MD ¼ f ðMPÞ ¼ f ðP; D; M; CB; K; M þ Þ: ð23:2Þ
where MP denotes the medical problem; P denotes the patient; D denotes the
possible states of the patient; M denotes the finite set of clinical symptoms;
K denotes the knowledge set containing all the relationship among M, D and its
causes of P; M þ denotes the vital signs set of the patient, M þ M; DC denotes
222 M. Xu and J. Shen
the diagnose/conclusion for the patient, DC D; f ðÞ denotes the reasoning
function of the clinician’s cogitation.
Clinical symptoms and vital signs are the prerequisite of the inference mech-
anisms in CDSSs. The basic idea is to drive the conclusion from the vital signs set
M þ based on the medical case database CB, the symptoms set M, the knowledge
set K, the possible states D and the reasoning process f .
Definition 23.4 The reasoning process of CDSSs based on CBR is formulated as
fCBR ðC; T; SimÞ ¼ DC, where T denotes the clinical query, and its dimensions
represent the attributes; Sim denotes the similarity measure for two cases, and
Sim : ðCi ; TÞ ! ð0; 1Þ.
23.2.2 Feature Selection
Information gain (IG), also named as mutual information, is an important machine
learning tool for weighting attributes. IG is the average value of information about
the class with or without a feature. When the value of IG is larger, the information
added by the feature is also much larger. Therefore, the values of IG for all the
features are measured during the feature selection epoch. The feature with the
largest value of IG belongs to the optimized feature set.
The idea of feature selection is to calculate the values of IG for all the attributes
in the feature set, and to remove the features whose value of IG are less than a
default threshold. The vital signs set M þ and the symptoms set M is Initialized and
the threshold is set as p. The feature f (f 2 M) is traversed. Given a data sample,
entropy
Xm
IðS1 ; S2 ; . . .; Sm Þ ¼ i¼1
PðCi Þ log2 PðCi Þ ð23:3Þ
where PðCi Þ is the prior probability for samples with Ci , PðCi Þ ¼ Si =S; m denotes
the number of classes, Si the number of samples labeled by Ci , and Si the total
number of samples.
Definition 23.5 Assume that the feature f has v values, ff1 ; f2 ; . . .; fv g. The data
sample is divided into v subsets by f , fS1 ; S2 ; . . .; Sv g, where Sj contains the cases
with feature value fj . The term S1j þ S2j þ þ Smj S is the weight of the jth
subset. Sij is the number of samples with class Ci in Sj , then
Xm
IðS1j þ S2j þ þ Smj Þ ¼ P log2 Pij
i¼1 ij
ð23:4Þ
where Pij is the probability of samples with class Ci in Sj , Pij ¼ Sij Sj .
23 Clinical Decision Support Model 223
Information gain corresponding to f is
Gainðf Þ ¼ IðS1 ; S2 ; . . .; Sm Þ Eðf Þ;
Xv S1j þ þ Smj ð23:5Þ
Eðf Þ ¼ IðS1j þ þ Smj Þ:
j¼1 S
The feature is added into M+ on the condition that Gainðf Þ is larger than the
threshold p.
23.2.3 Parameter Learning of BN
First, the conditional probability is calculated as pðxi ; ri jDl ; hðtÞ Þ for all the
parameters in set D and the attribute xi . Given set D, its likelihood is
X X
lðhjDÞ ¼ ln pðDjhÞ ¼ hðxkj ; rij Þ ln hijk
l ijk
where hðxkj ; rij Þ
denotes the assigned value in the database with xi ¼ k and ri ¼ j.
The maximum of the likelihood h is
hðxkj ; rij Þ
hijk ¼ P ð23:6Þ
hðxkj ; rij Þ
Assuming the initial value hð0Þ , the expectation of the current likelihood hðtÞ is
XX
lðhjhðtÞ Þ ¼ ln pðDl ; Xl jhÞpðXl jDl ; hðtÞ Þ
l
For any h, if Lðhhðtþ1Þ Þ LðhhðtÞ Þ, then
X l
LðhhðtÞ Þ ¼ f akj ; p aj ln hjlk
jlk
The next estimation of the likelihood is determined by finding its maximum
likelihood (MLE),
ðtþ1Þ f ðxkj ; rij Þ
hijk ¼ arg max E½PðDjhÞjD; hðtÞ ; S P k j
f ðxj ; ri Þ
The conditional probability distribution is fixed through the above recursive
algorithm, namely, the relation among the symptom features are formulated.
224 M. Xu and J. Shen
23.2.4 Case Retrieval
As mentioned above, case retrieval and matching is the core phase of CBR, in
which the similarity is measured for the query with the historical cases. In the
literature, the retrieval of CBR is to measure the difference among the features of
cases, and Euclidean distance is an popular tool as the similarity measure function
(Jain and Marling 2001).
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X m
Similarityðx; yÞ ¼ ðxi yi Þ2
i¼1
where xi ; yi denote the observations of the ith feature (fi 2 M þ ) corresponding to
the case x and yrespectively; m is the total number of features.
The problem is that this similarity assessment method omitted the cases during
the updating episode, as a result of losing the relative information. Therefore, a
new similarity measure function is proposed to solve the probability or BN-CBR.
Based on the BN of the features in M þ , the similarity of two medical cases x; y is
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Xm
Similarityðx; yÞ ¼ gðxi ; yi Þ ð23:7Þ
i¼1
8
>
> kxi ; yi k when xi ; yi are numeric;
>
> 0 when xi is equivalent to yi ;
>
>
>
>
>
> and xi ; yi are symbolic;
<
1p when xi ; yi are symbolic,
g(xi ; yi Þ¼
>
> xi ¼ X; the observation of yi is missing;
>
>
>
> 1q when xi ; yi are symbolic,
>
>
>
> yi ¼ Y; the observation of xi is missing;
:
1 Others:
where p ¼ Pðyi ¼ X jy1 ; . . .; yi1 ;yiþ1 ; . . .; ym Þ and q ¼ Pðxi ¼ Y jx1 ; . . .; xi1 ;
xiþ1 ; . . .; xm Þ. The probabilities are obtained with BN to reveal their dependent
relationships.
The best case is initiated as BESTCASE ¼ 0, as well as the best similarity. All
the cases are retrieved with the formula (23.7) for the query y. If Similarityðx0 ; yÞ[
BESTSIM, then BESTSIM= Similarityðx0 ; yÞ and BESTCASE ¼ x0 . The final
BESTCASE becomes the retrieved case.
23.3 Conclusion
This paper proposed the CDSS based on the BN-CBR hybrid model. BN were
established with the feature attributes of heart disease, which improved the
accuracy of diagnosis reasoning through solving the problem of missing data in
23 Clinical Decision Support Model 225
the database. Not only were the similarities of feature attributes measured, but also
the case matching method integrated the knowledge of their dependent
relationships.
Acknowledgments This work was supported by the National Natural Science Foundation of
China (71171143), by Tianjin Research Program of Application Foundation and Advanced
Technology (10JCYBJC07300), and Science and Technology Program of FOXCONN Group
(120024001156).
References
Gu YS, Hua Q, Zhan Y et al (2003) Case-base maintenance based on representative selection for
1-NN algorithm. In: Presented at the international conference on machine learning and
cybernetics, vol 4, pp 2421–2425
Jain AF, Marling CR (2001) Case-based tool for treatment of behavioral problems.
In: Proceedings of the 33rd southeastern symposium on system theory, pp 337–341
Kong G, Xu DL, Yang JB (2008) Clinical decision support systems: a review on knowledge
representation and inference under uncertainties. Int J Comput Intell Syst 1(2):159–167
Ling HF, Guo JY, Yan J (2006) Similarity algorithm of CBR technology applied to fault
diagnosing field. J PLA Univ Sci Technol (Nat Sci Edn) 7(5):480–484 (in Chinese)
Ma XX, Huang XY, Huang M, Ni L (2002) Analysis of information flow based on entropy of
fault diagnosis. J Chongqing Univ (Nat Sci Edn), 25(5):25–28 (in Chinese)
Shiu SCK, Yeung DS, Sun CH et al (2001) Transferring case knowledge to adaptation
knowledge: an approach for case-base maintenance. Comput Intell 17(2):295–314
Zhang Z, Yang Q (1999) Dynamic refinement of feature weights using quantitative introspective
learning. In: Proceeding of the international joint conference in artificial intelligence. Morgan
Kaufmann, San Francisco, pp 228–233
Zhang Z, Yang Q (2001) Feature weight maintenance in case bases using introspective learning.
J Intell Inf Syst 16(2):95–116
Chapter 24
Construction of Project Management
Classification Frame and Classification
Controlling Model Based on Project
Portfolio
Chao Yang and Fa-jie Wei
Abstract Project portfolio management and project classification management are
two important mutual connecting stages (Archer and Ghasemzadeh 1999). The
former is based on project priority appraisal in fact; the latter is project execution
management based on project execution collaboration and level-to-level control-
ling. Both organic combinations will carry out the whole process from the project
portfolio management stage to the enterprise project strategy management stage.
Therefore, It has also solved the problem that enterprise’s strategy execution ability
is insufficient and it is advantageous to enterprise’s strategy implementation.
Keywords Control Project portfolio management Project classification man-
agement frame 6s
24.1 Introduction
The so-called project is completed under certain resource constraints of the human,
financial and other one-time activity of the target. The projects tend to include
more than one task. The realization of each task needs to consume a certain
amount of human, financial resources (Boddy and Macbeth 2000).
The project management is the project organizer operation system, the science
theory and the method. Through carries on the plan, the organization, the coor-
dination, the control to the project, is for the purpose of realizing the project goal
management system.
C. Yang (&) F. Wei
School of Economics and Management, Beihang University, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 227
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_24,
Springer-Verlag Berlin Heidelberg 2013
228 C. Yang and F. Wei
Project portfolio management is project choice process, that is, under the
enterprise resources limited premise, how to chooses the corresponding project, to
dispose appropriate resources, and to realize enterprise’s strategic target better.
The project level-to-level management means that after having chosen the good
project portfolio, according to the project involves the scope, the content, the
complex degree, and project management personnel’s knowledge and ability, it
divides into the project the different rank, and has key carries on the level-to-level
administration.
24.2 Introduce a Question
In the project management company, the daily work is the management, the
implementation of project. Along with enterprise’s development, the enterprise will
face lots of projects, and the manager must consider two most essential questions: (1)
How to guarantee the coherence of the project goal with enterprise’s strategic target;
(2) Under the condition of project multitudinous and limited resources, how to
manage multiple projects on time, according to nature, according to budget man-
agement. The first question is Project portfolio management question, and it has
solved the project priority problem; the second question was that under the condition
of having solved in the project priority situation, it should not take common methods
which was used in the progress of PM, but should consider all projects as a whole to
carry on the management, because enterprise itself was a system strategic whole. In
consider all projects as a whole to carry on the management in the process, because
each project characteristic, the scope, the complex degree are different, which cause
the scope and the content of the project management different. Simultaneously it
needs the project personnel’s different ability and knowledge, therefore it is
important to carry on the level-to-level administration to the project. The classifi-
cation management decides the project implementation management method and
the control pattern. First, the project execution graduation will affect the authorized
management way of the project. Next, the project execution graduation has decided
the project resources disposition way. Finally, the project graduation has decided the
project controlling mode (Cheng et al. 2003).
24.3 Project Classification Standard
According to scope and complex degree, project management scope and content,
as well as personnel’s knowledge and ability involved, may divide into enter-
prise’s in project the enterprise project, the department level project and the
project level project. According to the different rank project situation, the enter-
prise determined the project carries on the starting time, the scope and completes
24 Construction of Project Management Classification Frame 229
the time and so on. Enterprise project graduation standard was shown in
Table 24.1.
According to its the complexity and innovativeness, the enterprise project was
strategic and chrematistic, and it should be carried on the special single row
management by the enterprise project office. This may centralize resources and the
time, and avoid the excessively many business management activity waste in
document (document) and the conference.
The department level project according to its key degree and enterprise’s
strategic sense, suitably enter the project office according to the project group way
to management; the individual project may promote according to the situation as
the company level, and carries on the single row management in the project office
(Dye and Pennypacker 1999).
Table 24.1 Project management standard
Appraisal Range and complexity PM range and content Knowledge and ability
standard of PMP
Classification
Project of The widest range and • Detail plan • Systematic knowledge of
enterprise the most complexity • Comprehensive cost PM, and expert ability
control • With ability to deal with
• Strict progress plan multiple mission
and control system
• Intact document
notes
• Comprehensive risk
management
• Regular meeting
Project of Wide range and more • Moderate detail plan • Mastering knowledge of
organization complexity • Moderate control PM;
plan • With elementary ability ti
• Moderate detail deal with high risk P
programme plan
• Periodic report and
notes
• Moderate risk
consciousness
• Moderate frequency
meeting
Project alone Incomprehensive and • Simple plan and no • Primary knowledge of
no complexity charging program PM;
• Logical budget, no • With elementary ability of
control PM
• No detail period
• Oral communication
• No risk plan
• No meeting
230 C. Yang and F. Wei
The project level project, according to the research and development’ the
technological innovation, the management, may divide into the project group, and
carried on the management by each group.
The project execution graduation is carried out after specific priority, and dif-
ferent projects can be carried out through different program and sequence.
24.4 Integrated Application of PM 6S System
In the project implementation, the 6s system of project management is integrated
application (Froese 1992; Zhang and Zhang 2009). 6S is a important link that
decides the project management success or failure, also is foundation of Project
portfolio management, the project implementation level-to-level administration,
and the construction project level-to-level administration frame. Project manage-
ment 6s system integrated application is shown in Fig. 24.1.
In the project management 6s system, it may divide into following three levels:
(1) project decomposition level, including EPS/WBS; (2) resources safeguard
level, including OBS/CBS/RBS; (3) knowledge support level, including KMS.
24.5 How to Construct Project Management Classification
Frame and Classification Control Model Based
on Project Portfolio
On the basis of project level-to-level administration 6S system, we have con-
structed the project level-to-level administration logical frame which is shown in
graph 3: Enterprise project management platform EPS, the department level
project management platform (Luiten et al. 1998; Van Der Merwe 2002), the
Fig. 24.1 6s-system of project management
24 Construction of Project Management Classification Frame 231
Fig. 24.2 Project classification management frame
project level project management platform. The project of different platform is
different in the resources disposition, the cost management, aspects and so on risk
guard.
In Fig. 24.2, the enterprise project, according to the importance, the strategic
characteristic, should equip with the manpower, the financial resource, and the
material resources sufficiently. Simultaneously it should be carried on the strict
cost, the progress, the quality control. In the project implementation, knowledge
management system (KMS) is necessary to manage the documents material, the
project material of the project entire life cycle, for document accumulation may
provide the reference for the enterprise similar project execution, and may save
certain cost.
According to the department level project characteristic and the important level,
it should dispose the corresponding personnel resources reasonably, and should
have certain risk consciousness (Ward and Chapman 1995). The department level
project management also is a very important level in three level frame based on 6s.
As to unimportant project, because of its multitude, low important level, we
may reduce suitably in resource investment. Of course it should not do harm to the
enterprise project and department project.
Figure 24.2 has given merely project level-to-level administration logical frame
based on the 6s, where as Chart 4 shows the concrete control pattern of different
kinds of project.
It can be seen from Fig. 24.3: classification control pattern emphasizes the
resources coordination assignment in the enterprise level on the basis of many
projects, and solves project implementation control management question of dif-
ferent project ranks. As to enterprise driven by projects, the former in fact is
project portfolio management on the basis of project optimal appraisal, the latter is
the project execution management based on project level to level management.
Both organic syntheses will realize the complete process from the project portfolio
management to the enterprise project strategy management (Yan et al. 2000), thus
232 C. Yang and F. Wei
Fig. 24.3 Project classification control model
has also solved the problem: enterprise strategy executive ability insufficient, and
will be advantageous to enterprise’s strategic implementation.
The author believed that, the level-to-level administration frame and the control
pattern is necessary to solve above questions. Therefore, the enterprise should
adopts the different management and the control pattern to the different rank
project, according to its condition, its cope and the complex degree, the project
management scope and the content, as well as project management essential factor
reasonable.
References
Archer NP, Ghasemzadeh F (1999) An integrated framework for project portfolio selection. Int J
Proj Manag 17(4):207–216
Boddy D, Macbeth D (2000) Prescriptions for managing change: a survey of their effects in
projects to implement collaborative working between organizations. Int J Project Manage
18:297–306
Chapman CB (1997) Project risk analysis and management-PRAM the generic process. Int J Proj
Manag 15(5):273–281
24 Construction of Project Management Classification Frame 233
Cheng M-Y, Su C-W, You H-Y (2003) Optimal project organizational structure for construction
management. J Constr Eng Manag 129(1):70–79
Dye LD, Pennypacker JS (1999) Project portfolio management: selecting and prioritizing projects
for competitive advantage. Center for Business Practices, West Chester
Froese TM (1992) Integrated computer-aided project management through standard object-
oriented models. Doctoral dissertation presented to the Department of Civil Engineering of
Stanford University, at Stanford, California, USA in partial fulfillment of the requirements for
the degree of Doctor of Philosophy
Jennings NR, Wooldridge M (1998) Agent technology: foundations, application and markets.
Springer, Berlin
Lampel J (2001) The core competencies of effective project execution: the challenge of diversity.
Int J Proj Manag 19:471–483
Luiten GT, Tolman FP, Fischer MA (1998) Project-modelling in AEC to integrate design and
construction. Comput Ind 35:13–29
Patterson FD, Neaile K (2002) A risk register database system to aid the management of project
risk. Int J Proj Manag (20):365–374
Platje A, Seidel H, Wadman S (1994) Project and portfolio planning cycle-project based
management for multi-project challenge. Int J Proj Manag 12(2):100–106
Van Der Merwe AP (2002) Project management and business development: integrating strategy,
structure, processes and projects. Int J Proj Manag 20:401–411
Vanegas JA (1987) A model for design/construction integration during the initial phases of
design for building construction projects. Doctoral dissertation presented to the Department of
Civil Engineering of Stanford University, at Stanford, California, USA in partial fulfillment of
the requirements for the degree of Doctor of Philosophy
Ward SC, Chapman CB (1995) Risk-management perspective on the project lifecycle. Int J Proj
Manag 13(3):145–149
Williams TM (1994) Using a risk register to integrate risk management in project definition. Int J
Proj Manag 12(1):17–22
Wooldridge M, Jennings NR (1995) Intelligent agents: theory and practice. Knowl Eng Rev
10(2):115–152
Yan Y, Kuphal T, Bode J (2000) Application of multiagent systems in project management. Int J
Prod Econ 68:185–197
Zhang H, Zhang G (2009) Construction project life cycle management of integrated risk
management. Ind Technol Forum 11(8):235–237
Chapter 25
Product Material Combination Image
Method Based on GEP
Ke Su and Sheng-li Kong
Abstract Materials are the stuff of design. From the very beginning of human
history, materials have been taken from the natural world and shaped, modified,
and adapted for everything from primitive tools to modern electronics. Aiming at
the need for product material combination, this study presents a product material
combination decision-making method based on Gene Expression Programming
(GEP). The semantics differential method is utilized to extract user image, mul-
tidimensional scaling is utilized to select cognitive samples. GEP was utilized to
construct the image and the material models. Finally, implementation of this
proposed method is demonstrated high stability via a seat case.
Keywords GEP Image Material combination Product design
25.1 Introduction
In the new century, the market has highly competition, and is simultaneously
affected in globalization, localization, and individualization. Traditionally black-
boxed designing model can’t match what the consumers require, exactly and
effectively, not mentioning the numerous chooses in the present. Mass custom-
ization relates to the ability to provide customized products or services through
flexible processes in high volumes and at reasonably low costs. The concept has
emerged in the late 1980s and may be viewed as a natural follow up to processes
that have become increasingly flexible and optimized regarding quality and costs.
In addition, mass customization appears as an alternative to differentiate compa-
nies in a highly competitive and segmented market. Therefore, the designers’
responsibilities are for the choices of appropriated use, suitable size, the favorable
K. Su (&) S. Kong
School of Mechanical and Automotive Engineering, Shandong Polytechnic University,
Jinan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 235
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_25,
Springer-Verlag Berlin Heidelberg 2013
236 K. Su and S. Kong
texture, the satisfactory style, and right message transformation in products. Let
products fit in human and let users conduct products designing. Then they are
really product that consumers really need.
Many scholars have done researches on product image research (Hsiao and Tsai
2004; Tsai et al. 2006; Xu et al. 2007; Liu et al. 2009; Kejun et al. 2008, 2009; Chen
2008; Nagamachi 1995; Su and Li 2005). Hsiao used gray theory and back-propa-
gation neural network for the color image evaluation of children’s walker; Tsai used
fuzzy neural network and gray theory for electronic locks color image evaluation;
Xu used genetic algorithms to build images and product modeling optimization
model. The above study focused mainly on the relationship between product image
and modeling and color, but is rarely involved product material. In this paper, based
on the GEP, the relationship between image and semantic are constructed by
quantification I. GEP was utilized to construct the image and the material models.
Finally, implementation of this proposed method is demonstrated high stability.
25.2 An Overview of Gene Expression Programming
The flowchart of a gene expression algorithm (GEA) is shown in Fig. 25.1
(Ferreira 2001). The process begins with the random generation of the chromo-
somes of the initial population. Then the chromosomes are expressed and the
fitness of each individual is evaluated. The individuals are then selected according
to fitness to reproduce with modification, leaving progeny with new traits. The
individuals of this new generation are, in their turn, subjected to the same
developmental process: expression of the genomes, confrontation of the selection
environment, and reproduction with modification. The process is repeated for a
certain number of generations or until a solution has been found.
Note that reproduction includes not only replication but also the action of
genetic operators capable of creating genetic diversity. During replication, the
genome is copied and transmitted to the next generation. Obviously, replication
alone cannot introduce variation: only with the action of the remaining operators is
genetic variation introduced into the population. These operators randomly select
the chromosomes to be modified. Thus, in GEP, a chromosome might be modified
by one or several operators at a time or not be modified at all.
25.3 Material Image Decision-Making Model
Based on GEP
Gene expression programming (GEP) is, like genetic algorithms (GAs) and genetic
programming (GP), a genetic algorithm as it uses populations of individuals,
selects them according to fitness, and introduces genetic variation using one or
25 Product Material Combination Image Method 237
Fig. 25.1 The flowchart of
gene expression algorithm
238 K. Su and S. Kong
Fig. 25.2 Framework of the Customer image
image material based on GEP
complex function modeling
quantification I GEP
Material texture element
decode
Material image select Optimization
more genetic operators. The core technology of GEP is separated the assessment
and the variation process completely (Ferreira 2001). Its variation process uses
fixed-length linear string symbols; assessment process uses ET (Expression Tree),
line with the overall image and the constituent element of the distribution
mechanism, thus suitable for this study. The image of the material based on GEP
research framework shown in Fig. 25.2.
Based on the overall image and the constituent element, function set F = {‘+’,
‘-’, ‘*’, ‘/’, ‘S’, ‘C’, ‘Q’, ‘e’, ‘ln’, ‘pow’, ‘abs’}. S is sine function, in is Natural
logarithm function, pow is Power function, abs is absolute value function.
End set T = {constitute element}, can be expressed as T = {x1, x2, x3,…,
xn}. x1, x2, x3,…, xn is the parameter set.
In the experiment, if population size is N, according to reference documentation
(Ferreira 2001), then fitness function is:
1
ffitness ¼ 1000 ð25:1Þ
Eþ1
The equation above:
1X m
E¼ ðFðxÞ Fj Þ2
m j¼1
is mean square error of the experimental sample, m is the total number of training
set samples, F ð xÞ is mathematical expression based on GEP, Fj is jth training set
output observations. Gene encoding with the relevant parameters which affects the
overall image and construct an initial population, calculate the fitness value of
population by Eq. 25.1, use the principle of survival of the fittest to guide the
evolution of populations.
25.3.1 Complex Functions Modeling
GEP complex functions modeling includes the following steps:
In this study, author selected parameters from F and T randomly, and selected
parameters as a locus. After the completion of this operation we will get a pop-
ulation of individuals. Repeating times of individual coding, the initialization of
population will complete. Figure 25.3 is a gene A with length 13, head length of
25 Product Material Combination Image Method 239
Fig. 25.3 Gene A with 1 2 3 4 5 6 7 8 9 10 11 12 13
length of 13 S + + * / + x 4 x5 x1 x2 x3 x4 x5
gene A is 6, tail length of gene A is 7. Suppose A is the initial population of genes.
Set functions F = {S,+,-,*,/,e}.x1, x2, x3, x4, x5 belong to end set T, to represent
the proportion of metal, plastic, wood, leather, fabric of product respectively.
In accordance with the GEP tree composition rules, gene A expression tree
shown in Fig. 25.4 (13th gene is discarded):
According to the rules of the expression tree (Su and Li 2005), we get
Eq. (25.2), that is FðxÞ:
FðxÞ¼ sinðx1 =x2 þ x3 þ x4 þ x4 x5 Þ ð25:2Þ
Fitness solution: parameters are evaluated by Eq. 25.2, the value obtained is
evaluated by Eq. 25.1, repeated N times, and all the individual fitness values will
appear. The best fitness value individual is the best one, then algorithm terminates.
Otherwise, enter the evolutionary steps: generate a new population.
Generate a new population: generate new populations of genetic manipulation,
including variations, restructuring, insert string, etc. Suppose gene A was chosen
to undergo gene mutation, *function change into +function, obvious, the corre-
sponding expression tree may be changed, read the expression tree to get the
mathematical expression is:
FðxÞ¼ sinðx1 =x2 þ x3 þ x4 þ x4 þ x5 Þ ð25:3Þ
25.3.2 The Statistical Results
In this study, authors use the mean square error (S2 ) and correlation coefficent (R2 )
to test the validity and predictive power of algorithms. S2 calculated as:
Fig. 25.4 Gene A expression
tree
240 K. Su and S. Kong
1Xn
S2 ¼ ðPðijÞ Tj Þ2 ð25:4Þ
n 1
PðijÞ is predictive value of individual j in population i, Tj is the observations of
individual j.
R2 calculated as:
CovðT; PÞ
R2 ¼ ð25:5Þ
rt rp
Above the equation, CovðT; PÞ is covariance, rt and rp are the standard deviation
respectively.
25.4 Experimental Verification
In order to rule out other factors influenced the image experiment, simple form
product should be used as a subjects. In this paper, we select high-speed train seat
to illustrate the effectiveness of the method.
25.4.1 Product Image Vocabulary Collection and Screening
The vocabularies were described by a questionnaire, and we get 30 words. After
that, factor analysis extracted vocabulary according to principle of ‘‘the factor
eigenvalue must be greater than 1 and greater than 85 % of the total contribution
rate’’, the words semantic compose of seven image vocabulary: popular, practical,
simple, simple, refined, elegant, creative (Fig. 25.5).
Fig. 25.5 Product image
sample
1 2 3 4 5
6 7 8 9 10
25 Product Material Combination Image Method 241
25.4.2 Elements of the Material Sample
According to the research of Jian (2002) and Ke (1997), we selected five common
material: wood, leather, metal, plastic, fabric. The 120 subjects (20 designers and
100 consumers) in this study taken part in the research (100 sets of data is training
data, 20 sets of data is test data) viewed the virtual images with 144 material
combinations and evaluated their value on a 100 mm measuring scale. The extreme
left side of the measuring scale represented ‘‘exceedingly disagree’’ and the extreme
right represented ‘‘exceedingly acceptance’’. After the assessment process is
completed, the scores are transferred into quantified values between 0 and 1.
25.4.3 GEP Experimental Data Processing
The sample data corresponding to each word was selected for testing. For example:
Select the data corresponding to ‘‘public’’ participate in this experiment. There are
five data collection parameters. The property name is shown in Table 25.1.
25.4.4 Experimental Data Processing
In the experiment, the extracted parameters are shown in the Table 25.2 GEP:
In the experiment, we used C++, 2.80G Intel Core Duo CPU, 4G RAM, the best
individual mathematical expression is
x5 ðx2 x5 Þ
F ð X Þ ¼ sin sinð9:932985Þ þ
7:256194 x5 þ 2x2
x2 x1
þ cos x8 þ sinðx3 21:864838Þ x5 þ x4 þ
x3
þ x5 x2 þ sinðx1 cosð7:34278Þ þ x5 Þ
x3 sin sinðx1 þ x4 Þ x4
x2 x4
þ
1=2 ðx5 þ x3 þ x2 þ x4 x1 Þ
Table 25.1 Paraments Parameter name Parameter description
property
X1 The proportion of metal
X2 The proportion of plastic
X3 The proportion of wood
X4 The proportion of leather
X5 The proportion of fabric
242 K. Su and S. Kong
Table 25.2 GEP Parameter set modeling
Parameter name Parameter values
Termination of algebra and population size 1000,1000
Set of functions F ¼ fþ; ; ; ; sin; cosg
End set T ¼ fx1 ; x2 ; x3 ; x4 ; x5 g
The number of genes and gene head length 5,15
Additional field Dc length and code range 16,(-10,10)
Genetic connection +
Mutation rate 0.044
Single point, two points recombination rate 0.2
Gene recombination rate 0.05
IS, RIS, Gene transposition rate 0.05
IS, RIS number of translocation element 1, 2, 3
From the experimental results, MSE and CC of F ð X Þ is 0.0385.
25.4.5 Experimental Verification
In order to verify the validity of this method, check the 20 experimental data in the
original model. MSE and CC of F ð X Þ is 0.0517, 0.9431. The mean-square error
and correlation coefficients of the GEP material image model is ideal. In addition,
experiments also showed that GEP algorithm had high efficiency, the total time for
running 100 times is only 3521 s, and 94 times gained sub-optimal solution which
close to the optimal solution.
25.5 Conclusion
Product customization employs virtual images to allow consumers to choose
product material via the Internet/Intranet recently. After choosing different material
combinations, consumers often become confused and are unable to choose material
with more selections. This study presented a design decision-making support
model for customized product material combination. This decision-making support
model composed of the GEP approach and the image compositing technique that can
be used to create a virtual image support system for product customization, allowing
consumers to choose their preferred product style and weight to the image words
according to their own preferences, and then consumers can clarify which product
material and impressions they prefer. Also, the design decision-making support
model of this study can be extended to other products or industries. It can also be
used in any customization problems where a hierarchy of decisions needs to be made
to extract appropriate list of choices.
25 Product Material Combination Image Method 243
Acknowledgments This work is supported by the fundamental research funds for the central
universities project (2010QNA5042); Nation natural science foundation, China (No. 61070075,
61004116, 61002147, 61003147); Zhejiang major science and technology projects
(2011C14018); Wenzhou Science and technology project (H20100042).
References
Chaoming K (1997) Research of material image for vision and touch texture. National Yunlin
university of science and technology, Taiwan (In Chinese)
Chen S-H (2008) Genetic programming: an emerging engineering tool. Int J Knowl Intell Eng
Syst 12(1):1–2
Ferreira C (2001) Algorithm for solving gene expression programming: a new adaptive problems.
Complex Syst 13(2):87–129
Hsiao SW, Tsai HC (2004) Use of gray system theory in product color planning. Color Res Appl
29(3):222–231
Jian L (2002) The research of product material image of sensation and perception. Dissertation,
Private Tunghai university, Taiwan (in Chinese)
Kejun Z, Shouqian S, Hongzong S (2008) Prediction of retention times for a large set of
pesticides based on improved gene expression programming. In: Proceedings of the 10th
annual conference on genetic and evolutionary computation, Atlanta, GA, USA, Association
for Computing Machinery, pp 1725–1726
Kejun Z, Shouqian S, Y Tang, Hongzong S (2009) Gene expression programming for the
prediction of acute toxicity of aldehydes. Chin J Anal Chem 37(3):425–428
Liu G, Cao H, Song X (2009) Products shape design based on genetic algorithm and fuzzy neural
network. Coal Mine Mach 30(11):205–208 (in Chinese)
Nagamachi M (1995) Kansei engineering: a new ergonomics consumer-oriented technology for
product development. Int J Ind Des 15(1):3–11
Su J-N, Li H-Q (2005) Investigation of relationship of form design elements to kansei image by
means of quantification-I theory. J LAN Zhou Univ Technol 31(2):36–38 (in Chinese)
Tsai HC, Hsiao SW, Hung FK (2006) An image evaluation approach for parameter based product
form and color design. Comput Aided Des 38(2):157–271
Xu J, Sun SQ, Zhang KJ (2007) Products shape optimization design based on genetic algorithm.
J Mech Eng 43(4):53–57 (in Chinese)
Chapter 26
D2-Index: A Dynamic Index Method
for Querying XML and Semi-Structured
Data
Yin Zhang, Hua Zhou, Zhi-hong Liang, Jun-hui Liu,
Yun Liao, Peng Duan and Zhen-li He
Abstract To facilitate queries over XML and semi-structured data, various
structural indexes acting as a structural summary have been proposed. Structural
index is derived directly from the data and serve as indices for evaluating path
expressions on XML and semi-structured data. We propose D2-Index: a path-
based dynamic index method for querying XML and semi-structured data. D2-
Index is prefix-based and based on the concept of ORDPATH and BSC. It uses
binary fraction to encode and can be completely capable of static encoding. More
importantly, it can insert nodes into any position efficiently without re-encoding
the existing indexes. According to the experimental results, it shows that the query
plans using D2-Index is feasibility, accuracy and efficiency.
Keywords Prefix-based Semi-structured data Structural index XML
26.1 Introduction
With the growing importance of XML (eXtensible Markup Language) and semi-
structured data in information storage, exchange and query, much research has been
done to provide query mechanisms to extract information from XML and semi-
structured data. XML is an example of semi-structured data which are different
Y. Zhang (&) H. Zhou Z. Liang J. Liu Y. Liao Z. He
School of Software, Yunnan University, Kunming, China
e-mail: [email protected]
Y. Zhang H. Zhou Z. Liang J. Liu Y. Liao Z. He
Key Laboratory of Software Engineering of Yunnan Province,
Yunnan University, Kunming, China
P. Duan
Network and Information Center, Yunnan University, Kunming, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 245
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_26,
Springer-Verlag Berlin Heidelberg 2013
246 Y. Zhang et al.
from traditional data models, such as relational or object-oriented models. Instead,
the underlying data model of semi-structured data is a labeled graph (Chen et al.
2003), and XML is more like a tree (Cohen et al. 2002). In this paper, it discusses
XML as an example of semi-structured data. Generally, XML data is stored in
documents separately, or in relational databases or native XML databases. No
matter how to be stored, a structural index of an XML data can not only provide
efficient insertion but also support extremely high-performance query plans.
Several query languages, including XPath (Clark and De Rose 1999) and
XQuery (Chamberlin 2007), have been proposed for querying semi-structured
XML data. However, these languages based on path expression extract data by
traversing, but their consumptions are too large clearly. We would like to build a
structural index to speed up the evaluation process. Usually, each node of an XML
data is assigned a unique index, which can be used to identify the relation between
any nodes and get the structural information. An efficient XML index method
should have the following two characteristics:
• The index method should support structural queries directly (Wang et al. 2003a).
It can quickly find the XML node by path or other structural expression.
• The index structure should allow dynamic data insertion, deletion, update, etc.
(Wang et al. 2003a). When the XML data is updated dynamically, and the
existing structural index does not need change.
In this paper, it proposes a new index method called D2-Index, which is prefix-
based. This method not only supports dynamic update but also quick and efficient
queries of XML data.
26.2 Related Work
Existing XML index methods can be divided into two categories. One is region-
based, and the other one is prefix-based. The early research of XML index method
is mostly region-based, which has been unable to solve the problem of dynamic
update. Therefore, it has been gradually replaced by prefix-based (Wu et al. 2003).
According to the tree structure of XML data, prefix-based index encodes each
node with two parts, prefix and layer ID (Amagasa et al. 2003; Wang et al. 2003b).
The prefix is the code of parent node, and the layer ID is the unique identifier of
each node which is different from other sibling nodes. DewyIDs (Duong and
Zhang 2005) and ORDPATHs (O’Neil et al. 2004), are prefix-based and used
widely. And both of them reserve code range for insertion to avoid re-encoding
other sibling nodes and their children when inserting a new node. However, when
the nodes inserted are too much, the code range reserved cannot meet demand and
re-encoding is still inevitable. DLN (Bohnle and Rahm 2004) solves this problem
by variable encoding. DLN defines layers with fixed-length binary string, and
creatively defines layer separator, which avoids re-encoding effectively. Yet its
encoding is complex and no flexible. BSC (Wang et al. 2008) presents an encoding
26 D2-Index 247
by using binary fraction, which is indeed effective to avoid re-encoding. However,
the binary representation of BSC has limitations, because it needs to introduce
parameters to encode and decode.
Based on the above methods, this paper proposes a new dynamic index method
for XML data, called D2-Index (Divide by 2-Index). The method completely
avoids re-encoding, and it is able to represent with binary.
26.3 Fundamental D2-Index Concepts
In D2-Index, the binary code of each complete index consists of three parts: a
unique identifier including several layer IDs, the offset of the node in the XML
data, and the length of the node. Figure 26.1 shows the composition of the code for
each node in D2-Index.
The most important component of the binary encoding in Fig. 26.1 is pairs of
variables and their lengths. L defines the length in bits of the succeeding variable V
bitstring, which is familiar with ORDPATH. This encoding has the following
important properties (O’Neil et al. 2004): (1) we know where an L starts and we
can identify where it stops, (2) each L bitstring specifies the length in bits of the
succeeding V bitstring, (3) from the two properties above, we see how to parse the
code bitstrings.
D2-Index uses binary fraction to encode layer IDs, which is different from
ORDPATH. The binary storage of fraction generally references the standard of
IEEE, which makes presentation easily but is not flexible and usually encodes
useless bits because it defines a fraction with a certain length in bits. Thus we
propose a new encoding in this chapter, which can accurately represent a fraction
in relatively short bits.
26.3.1 Motivating Example
In D2-Index, the initial value of each layer ID is 12, and each node’s layer ID is the
previous sibling one’s divided by 2. That is the reason why it is called D2-Index.
For instance, Fig. 26.3 shows the structural index of the XML data in Fig. 26.2
through D2-Index. It illustrates the results with tree structure in order to make it
easier to be understood.
Fig. 26.1 Composition of L off Voff
the code for each node in
D2-Index
Key0 ... Keyk Offset Length
L0 VE0 M0 Llen Vlen
248 Y. Zhang et al.
Fig. 26.2 Sample XML data
26.3.2 Detail Encoding Design
In general, the encoding of index without offset and length (If not described
specially, the following text will be the same) should follow the principles as
follows:
• Uniqueness: Different node has different code.
• Orderly: The order of nodes’ appearance in the XML data should be reflected in
encoding.
1/
1/22
text
name n item item item item
1/2·
1/ 2·1/64
1/64
1/2·
1/2· 1/2
1/2 1/2·
1/2·1/ 4
1/4 1/2·
1/2·1/8
1/8 1/2·
1/2·1/16
1/ 16 1/ 2· 1/32
1/2· 1/32
a a c c c
1/2·
1/2· 1/8·
1/8· 1/2
1/2 1/ 2· 1/16
1/2· 1/16··1/
1/22 1/2·
1/2·1/16
1/16··1/4
1/ 4 1/2·
1/2· 1/32
1/32··1/2
1/2 1/2·
1/2· 1/64
1/64··1/2
1/2
b d
1/2·
1/2· 1/8·
1/8· 1/2·
1/2·1/2
1/ 2 1/2·
1/2·1/
1/64
64·· 1/2·
1/2·1/ 2
1/2
Fig. 26.3 Index tree for XML data of Fig. 26.2
26 D2-Index 249
• Practicality: The process of index encoding and decoding should be practical
and simple.
The most important part of D2-Index is binary fraction presentation for the
layer ID of each node in the XML data. In order to reduce the code length, we
propose a variable-length encoding based on the IEEE standard of fraction storage,
and it meets all the requirements above. In the specification of the IEEE, a fraction
is stored with its binary scientific notation, and it consists of three parts: Sign,
Exponent and Mantissa.
In D2-Index, all the fractions are between 0 and 1, so Sign can be omitted. And
Exponent can be represented by Variable/Length pairs as ORDPATH. The most
difficult part is how to store Mantissa. It cannot be stored in fixed-length bits
because that makes the code length is too large, and contain a lot of useless bits.
Moreover, we cannot represent it with Variable/Length pairs because of the par-
ticularity of Mantissa. For example, there are two binary fractions 0.11 and 0.101.
Obviously, 0.11 is bigger than 0.101. The length of 11 is 2 and 101 is 3, and this
presentation makes 0.101 is bigger than 0.11 because length is the prefix of
variable.
Therefore, we propose a new presentation, which encodes Mantissa scalability
in bit that 10 defines 1 and 01 defines 0. Exponent is presented with Variable/
Length pairs, so we can know where Mantissa starts. And the succeeding bitstring
is an L of another Variable/Length pair. Then if L starts with 00 or 11, it will be
distinguished with Mantissa. In addition, the integer part of binary scientific
notation is always 1 and it has only 1 bit, so we encode it together with Mantissa in
order to reduce the complexity of encoding and decoding.
It is obvious that L must match the Fano condition that no code is prefix of
another code. In Table 26.1, it describes the prefix encoding schemes for L bit-
strings in D2-Index. The L bitstring 110 defines a component Variable/Length
encoding with assigned length L = 3, indicating a 3-bit V bitstring. The following
V bitstrings (000–111) represent V value of the first eight integers (0–7). Thus,
110110 is the bitstring for D2-Index ‘‘6’’. In the next row in Table 26.1, bitstring
10100 identifies an encoding with L = 4, and the 4-bit V bitstrings that follow
represent the range [8,23]; in particular, V = 8 is represented by bitstring 0000, 9
by bitstring 0001, …, up to 23 by bitstring 1111. Similarly, V in the rage [-8,-1]
is associated with the L bitstring 011, with -8 represented by the lowest bitstring
000 and -1 represented by the highest bitstring 111.
Example 26.1 Using L values of Table 26.1, we would generate ‘‘12 34 38’’ in
Table 26.2.
We can directly compare D2-Index values without decoding, and it will be
elaborated in the next section.
250 Y. Zhang et al.
Table 26.1 Lengths L with V ranges represented
Bitstring L Value range
01000010 64 [-18447025552981299544,-281479271747929]
01000011 48 [-281479271747928,-4295037273]
0100010 32 [-4295037272,-69977]
0100011 16 [-69976,-4441]
010010 12 [-4440,-345]
010011 8 [-344,-89]
01010 6 [-88,-25]
01011 4 [-24,-9]
011 3 [-8,-1]
100 3 [0,7]
10100 4 [8,23]
10101 6 [24,87]
101100 8 [88,343]
101101 12 [344,4439]
1011100 16 [4440,69975]
1011101 32 [69976,4295037271]
10111100 48 [4295037272,281479271747927]
10111101 64 [281479271747928,18447025552981299543]
Table 26.2 The result of Example 26.1 after generating
1 3 3
2 4 8
1 1
12 1:1 2 1:1 2 2
001 111 10 001 111 1010 001 110 1010
L0 = 3 VE0 = -1 M0 = 1 L1 = 3 VE1 = -1 M1 = 11 L2 = 3 VE2 = -2 M2 = 11
26.3.3 Comparing D2-Index Values
In the previous section, it mentioned that the order of nodes’ appearance in the
XML data should be reflected in encoding. In other words, we can know the order
and relative position of nodes in the XML data by comparing the encoding values.
In Table 26.3, it lists several encoding values and their binary representation.
From top to bottom it is the order of nodes’ appearance in the XML data, the
document order.
Then we examine whether the results of comparing encoding values in bits
yielding the document order from the aspects as follows:
• Comparing ancestor-offspring nodes (as 1 and 2): It shows that the parent’s code
is the prefix of the child’s, and the child’s bitstring starts with 00. Obviously, we
can know that the order of parent is before their child with their bitstring.
• Comparing sibling nodes (as 1 and 5), non-sibling nodes with the same depth (as
2 and 4), and non-ancestor-offspring node with different depth (as 1 and 4):
When comparing sibling nodes, there are two cases. One case is that the lengths
26 D2-Index 251
Table 26.3 Sample encoding values and their binary representation in document order
Order Key code Bitstring
1 3 001 111 1010
4
2 3 1 001 111 1010 001 111 10
4 2
3 1 001 111 10
2
4 1 1 001 111 10 001 111 10
2 2
5 3 001 110 1010
8
6 5 001 110 100110
16
7 1 001 101 10
8
of their bitstrings are equal, such as node 1 and 5. In this case, the simple
bitstrings or byte by byte comparison yields document order, and the larger one
is former. The other case is that the lengths of their bitstrings are not equal, such
as node 1 and 3. In this case, we compare their bitstring in the shorter length,
and the larger one is former, such as node 3 and 5. But if the comparison result is
equal, the one with longer bitstring will be former, such as node 1 and 3.
In summary, the comparisons in the latter three cases are the same. Further-
more, the nodes with the equal length of bitstring can be compared directly. If not
equal, they will be compared in the shorter length. If the comparison result is not
equal, the larger one will be former. Else we will check the rest of the node with
longer bitstring. If the rest part of its bitstring starts with 00, it will be the other
one’s parent (or ancestor) and it will be former. Otherwise, it is latter than the other
one in document order.
Clearly, in D2-Index method, the index values decrease when the document
order increasing, which complies with the principle of orderly.
26.3.4 Arbitrary D2-Index Insertions
According to the position, D2-Index insertion can be divided into three cases:
leftmost of the siblings, rightmost of the siblings and between two siblings. With
different position, the encoding of insertion is different. It describes the specific
encoding process of inserted layer ID in Algorithm 26.1.
Algorithm 26.1 The Insertion Layer ID Encoding Algorithm
Input The layer ID of the node on the left side of inserted position, left, and the
layer ID of the node on the right side of inserted position, right.
Output The layer ID of inserted node, insertion.
1. If left does not exist (it means the inserted position is at leftmost of the
siblings);
252 Y. Zhang et al.
Table 26.4 The result after Key code Bitstring
insertion in Example 26.2 1 7
28
001 111 10 001 111 101010
1 3 001 111 10 001 111 1010
24
1 5 001 111 10 001 111 100110
28
1 1 001 111 10 001 111 10
22
1 1 001 111 10 001 110 10
24
• Insertion := (right ? 1)/2;
2. Else if right does not exist (it means the inserted position is at rightmost of the
siblings);
• Insertion := left/2;
3. Else (it means the inserted position is between the two siblings);
• Insertion := (left ? right)/2;
Example 26.2 It assumes that ‘‘12 34’’ and ‘‘12 12’’ are two existing sibling nodes,
and there are three nodes need to insert at the leftmost, rightmost of them and
between them. In Table 26.4, it shows the results after insertion.
We can see the indexes after insertion still yield document order as before. In
addition, all the codes are unique and their lengths are variable, so insertion will
not affect the existing node has been encoded.
26.4 D2-Index Query Plans
26.4.1 Secondary Indexes
In the previous chapter, we discuss the primary indexes. In practical we need prove
the query plans with secondary indexes. There are two of the important secondary
indexes as follows (O’Neil et al. 2004):
• Element and Attribute TAG index, supporting fast look up of elements and
attributes by name.
• Element and Attribute VALUE index.
In this chapter we discuss another secondary index called PATH index. Actu-
ally it is familiar with TAG index. The difference between them is that TAG index
is based on the node’s tag, but PATH index is based on the path which consists of
all the nodes’ tags from root to itself. In D2-Index, every different path has a
separate storage space to store all the nodes’ primary index keys, not the whole
index, with such a path.
26 D2-Index 253
26.4.2 Query Plans
XPath (XML Path Language) is a language for selecting and processing parts of an
XML document, and it is published by W3C. In D2-Index, we use XPath to
express the path of the query plans, and get the results from the XML data with the
collection of their primary index keys found by secondary index (He and Yang
2004). The basic steps of query plans are as follows:
1. Enter a query expression.
2. Find the secondary index and get the collection of primary index keys.
3. According to the keys, obtain the offsets and lengths of the nodes.
4. Get the results.
26.5 Experiment
In this section we present results of testing D2-Index, evaluating three aspects:
encoding length, index size and query efficiency. Table 26.5 shows the detailed
information of the main XML data in the experiments, which are from XML Data
Repository (http://www.cs.washington.edu/research/xmldatasets/).
In Table 26.6, it shows the average and maximum length (bit) of D2-Index, the
size of all index files including primary and secondary index, and the expansion
ratio compared with their source files.
Table 26.5 Detailed information of the main XML data in the experiments
File Name File size Elements Attributes Max-depth Avg-depth
reed 277 K 10546 0 4 3.19979
sigmodRecord 467 K 11526 3737 6 5.14107
nasa 23.8 M 476646 56317 8 5.58314
treebank_e 82 M 2437666 1 36 7.87279
SwissProt 109 M 2977031 2189859 5 3.55671
dblp 127 M 24032673 6102230 6 2.90228
Table 26.6 Indexes size and expansion ratio of D2-Index
File name Avg-length Max-length Index size Expansion ratio
reed 36.36 47 200 K 0.72
SigmodRecord 60.53 75 384 K 0.82
nasa 59.41 87 11.6 M 0.49
treebank_e 79.58 305 293 M 3.57
SwissProt 51.08 74 100 M 0.92
dblp 53.9 81 75.5 M 0.59
254 Y. Zhang et al.
Table 26.7 The time XPath Results Time (s)
consumption of the query
plans by D2-Index /dblp/book/@key 845 15.7
/dblp/article/title/sub/sup/i 1 28.5
/dblp/book[300] 1 19.1
We can see that the encoding length and expansion ratio of D2-Index are not
directly related to the number of nodes in XML data, but its depth. As the depth
increases, the encoding length shows a clear upward trend. Moreover, the size of
index files is affected by the depth more than the size of XML data.
In Table 26.7, it shows the time consumption of the query plans with three
XPath expressions to the XML document DBLP by D2-Index, and the query
results are consistent with XQuery (Goldberg 2009).
26.6 Conclusion and Outlook
In this paper, it proposes D2-Index: a path-based dynamic index method for
querying XML and semi-structured data. It uses binary fraction to encode and can
be completely capable of static encoding. More importantly, it can insert nodes
into any position efficiently without re-encoding the existing indexes. According to
the experimental results, it shows that the query plans using D2-Index is feasi-
bility, accuracy and efficiency.
As for the future work, the research will focus on lexical analysis of both
English and Chinese, which can be used to create Elements and Attributes VALUE
index and full-text index.
Acknowledgments This work is funded by the Open Foundation of Key Laboratory of Software
Engineering of Yunnan Province under Grant No. 2011SE13, A Study on XML-based Adaptable-
Event-Driven Integrated Software Framework of Yunnan Science Foundation under Grant No.
2009CD009, and the Postgraduates Science Foundation of Yunnan University under Grant No.
YNUY201131.
References
Amagasa T, Yoshikawa M, Uemura S (2003) QRS: a robust numbering scheme or XML
documents. In: Proceedings of ICDE, pp 705–707
Bohnle T, Rahm E (2004) Supporting efficient streaming and insertion of XML data in RDBMS.
In: Proceedings of the 3rd international workshop Data Integration over the Web (DIWeb),
pp 70–81
Chamberlin Dl (2007) XQuery 1.0: an XML query language. W3C working draft
26 D2-Index 255
Chen Q, Lim A, Ong KW (2003) D(K)-index: an adaptive structural summary for graph-
structured data. In: Proceedings of the ACM SIGMOD international conference on
management of data, San Diego, CA
Clark J, De Rose S (1999) XML Path Language (XPath) Version 1.0. W3C Recommendation
Cohen E, Kaplan H, Milo T (2002) Labeling dynamic XML trees. In: Proceedings of PODS,
pp 271–281
Duong M, Zhang Y (2005) A new labeling scheme for dynamically updating XML data.
In: Proceedings of ADC, pp 185–193
Goldberg KH (2009) XML, 2nd edn. Peachpit Press, Berkeley
He H, Yang J (2004) Multiresolution indexing of XML for frequent queries. In: Proceedings of
the 20th international conference on data engineering, Boston, pp 683–694
O’Neil PE, O’Neil EJ, Pal S (2004) ORDPATHs: insert friendly XML node labels.
In: Proceedings of SIGMOD, pp 903–908
Wang H, Park S, Fan W, Yu PS (2003a) ViST: a dynamic index method for querying XML data
by tree structures. In: Proceedings of the ACM SIGMOD international conference on
management of data, San Diego, CA
Wang H, Perng CS, Fan W, Park S, Yu PS (2003b) Indexing weighted sequences in large
databases. In: ICDE
Wang C, Yuan X, Wang X (2008) An efficient numbering scheme for dynamic XML trees.
In: Proceedings of the 2008 IEEE international conference on computer science and software
engineering (CSSE), pp 704–707
Wu H, Wang Q, Yu JX, Zhou A, Zhou S (2003) UD(k,l)-index: an efficient approximate index for
xml data. In: Proceedings of the 2003 international conference on web-age information
management, pp 68–79
Chapter 27
Data Preprocessing in Web Usage Mining
Xiang-ying Li
Abstract At present, the study on Web Usage Mining mainly focuses on pattern
discovery (including Association Rules, sequence pattern, etc) and pattern analysis.
However, the study on the main data sources, that is to say, the study on web-log
pre-process is relatively rare. Given that high-quality data helps a lot in improving
Pattern mining precision, this paper studies from this aspects, and proposes the high-
effective data preprocessing method.
Keywords Client identification Date cleaning Path completion Web
usage mining
27.1 Introduction
The main data resource of Web Usage Mining is web log, from which we can
know the browsing behaviors of clients. Based on the browsing behaviors of
clients (Shao 2009), we can (1) modify the corresponding web link; (2) get to
know the interested points of clients, and provide personalized pages for them;
(3) subdivide the clients, carry out different promotion strategies for different
customers aiming to improve (ROI) return on investment; (4) find out clients’
clicks on ads, based on which modify the ads setup (Cooley 1997a).
Data preprocessing is the first part in Web Usage Mining. Whether the data
preprocessing is good or bad will directly influence the effect of the following links
(Zhao 2003; Wu 2002), such as Association Rules mining, Sequence Pattern
discovery and the categorical and clustering of clients and so on.
X. Li (&)
College of Information Engineering, Shandong Youth University
of Political Science, Jinan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 257
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_27,
Springer-Verlag Berlin Heidelberg 2013
258 X. Li
Data Client
Session Path Transaction
Cleaning Identification
Identification Completion Identification
Data After The Identificated The Identificated Session Log Transaction
Web Session Log After Data
Client Log
Log
Fig. 27.1 Mining process of web use
In a word, Data Preprocessing can greatly improve the quality of Data Mining
and shorten the time needed in practical Data Mining. Web Usage Mining, whose
object is mainly web log, is even more affected by Data Preprocessing, for web
log, different from the traditional well-structured data base or data from Data
Warehouse, is semi-structured. In addition, the incomplete data in web log lead by
all kinds of reasons, and the purpose of Web Usage Mining (Liu 2007a; Zhang
2006; Liu 2007b), which is unlike that of transaction data mining, require log files
to be preprocessed before mining, converting the log files to format easy to mining
and laying foundation for improving the accuracy and effectiveness of final pattern
mining. As the Fig. 27.1 shows that it is a complete process of Web Usage Mining.
And the paper pays attention to the module of Data Preprocessing, of which
several steps, Data Cleaning, Identify Clients, Session Identification and Path
Completion are included, as shown in Fig. 27.1.
27.2 Data Preprocessing
27.2.1 Data Cleaning
The task of Data Cleaning is to delete data unrelated to mining, such as pictures of
GIF, JPEG and jpg format. These pictures are in WebPages in large numbers,
when clients visit WebPages, pictures and cartoons exist in log files as independent
records. For most mining task, they can be ignored (of course, we have to
reconsider if the websites is special for pictures). Though deleing these records
contributes nothing to improve mining effect (Wang 2000; Liu 2003; Tang 2002;
Xu 2003; Ji 2009), it can decrease the data to be processed afterwards, improve
processing rate and reduce the effect of invalid data upon mining process. The
experimental data adopted is a week’s logs files (2012\3\1–2012\3\7) from http://
my.sdyu.edu.cn/, in all 132 M bytes. Before Data Cleaning, there are 1,274,051
records. After Data Cleaning as the above method, 336,741 records are left. Thus
we can see that this step can greatly reduce the data to be processed later, and
improve the processing rate.
27 Data Preprocessing in Web Usage Mining 259
27.2.2 Clients Identification
Clients’ identification is to identify from logs files which clients visit the website
and each client visits what web pages.
The clients registered are easy to be identified. However, many clients unreg-
istered surf the internet by web proxy server or several clients share one computer,
and the existing of firewall and a client using different browsers, all these add
difficulties to client identification. Certainly we can confirm the visiting clients by
Cookies, but taken personal privacy into consideration, Cookies are forbidden by
many clients’ browsers.
How to distinguish many visiting clients using the same computers? Some ideas
are to detect whether we can directly visit the page from the page we have visited
last time by topology map in websites (Pirolli 1996), if we can’t, it is probably that
many clients use one computer, and then we can identify clients by Client Session
Automatic Generation Technique based on Navigation Patterns (Cooley 1997b).
Considered that the two methods above can only partly solve the problem of
clients’ uncertainty, the paper proposes an algorithm—Client and Session Identi-
fication Algorithm (CSIA). This algorithm takes comprehensive consideration, and
combines Client_IP, topology diagram in websites, the browsers version and
Referer_Page to identify individual clients, with good accuracy and expansibility.
Meanwhile, the algorithm—CSIA also can complete the task of Session Identifi-
cation, the specific content of the algorithm will be given in Session Identification
part, and the paper will first proposes the definition of client as followed:
Definition 1 Clientsi = \Client_ID, Client_IP, Client_URL, Client_Time,
Client_RefPage, Client_Agent [ , 0 \ i \ n, n represents the total numbers of
clients, Client_ID is the identification of clients identification and Client_IP,
Client_URL, Client_Time, Client_RefPage, Client_Agent respectively stands for
the IP address of clients, the pages visiting, the time of visiting pages, the pages
visited and operating system clients used and the version of browsers, from which
the unique client is confirmed.
27.2.3 Session Identification
The time scan of a log file is different, from 1 day, one week–one month etc. In
such a long time, a client can not visit a website only for one time, so how to
distinguish client’s access record left by one—time visit or many times visits? And
this involves the problem of Session Identification. A session is time serials of
URLs in the process that the client visits a webpage. The relative definition is
given as followed:
Definition 2 The definition of Sessionsi: Sessionsi = \ClientID, Sj, [refpagej1,
refpagej2,……refpagejk] [ , 0 \ i \ n, n represents the total numbers of sessions,
ClientID is the identification identified in the process of clients identification,
260 X. Li
Sj refers to the j session of the client, refpagejk is the assemble of Refpages-the
visited pages in session.
The most simple and most commonly used way to distinguish two different
sessions of one client is to set a timeout, usually 30 min. If the request time
interval for two web pages exceeds the threshold preset, then the client is thought
to begin a new session. Taking 30 min as the timeout has been tested by much
practical application, and it is simple and easy to carry out, so the paper adopts this
method.
Definition 3 Cube is the saving mode of clients and sessions identified by the
algorithm—CSIA. The definition is that Cube = \ClientID, Sj, Client_IP,
[refpagej1(tj1), refpagej2(tj2),……, refpagejk(tjk)], among them, ClientID, Sj,
refpagejk have the same definition as that in first one, and different from definition 2,
in definition 3, Client_IP and tjk are added, of which tjk is the page-visiting time.
The thought of algorithm:
A large number of continuous record fragments exist in log files, and each
record fragment is from the same Client_IP. If we inspect the record fragments
item by item through isExistedIP (Cube) and isSameClient (Clientsi), then the
running efficiency is dramatically lowered. Therefore the thought of algorithm in
this paper is as followed: first judge whether the current record has the same
Client_IP as the last one, if they have the same Client_IP, then they are supposed
as the same client; or judge whether the client has been identified, the specific
judgments are shown in the algorithm below.
This refers to a hypothesis that if the several continuously appeared records is
from the same IP address, and then supposes that the continuous records are from
the same client. The hypothesis is based on the following two points:
(1) The acceptable client identification accuracy: In order to check the accuracy of
client identification under the above hypothesis, we exact 200 record frag-
ments as mentioned above from log files (each continuous fragment is from
different IP address). As the hypothesis shows, they are considered as from
200 different clients, however, they are identified as 202 different clients by
isExistedIP (Cube) and isSameClient (Clientsi) respectively. So it is clear that
the above hypothesis has little effect on the accuracy of client identification,
and it is acceptable.
(2) The efficiency of algorithm: Adopting the above hypothesis can avoid
checking large numbers of records item by item, thus greatly improving the
running efficiency of program.
Algorithm: CSIA
Input: log files; TimeSpan
Output: ClientID, Session
(for the specific Algorithm occupies too much space, here proposes the
framework in JAVA language limited to space constrains)
27 Data Preprocessing in Web Usage Mining 261
262 X. Li
27 Data Preprocessing in Web Usage Mining 263
The advantages and disadvantages of algorithm:
(1) High accuracy: Overcome the disadvantage of low identification accuracy
caused by the traditional way adopting only IP address to identify visitors.
(2) High efficiency: Realize the client identification and session identification;
overcome the low efficiency caused by identifying client and session
respectively.
(3) Good data store formats: Store the client and session identified into dynamic
three-dimensional array construed by CSIA, avoiding the waste of store space,
as shown in Fig. 27.2: client-session axis means client and session identified,
among them, n is the total number of clients, k is the k session of some client,
and different sessions are shown by different Client ID, such as 1-1, 2-1, 3-1,
…, n-k, the same client may have several different sessions, such as the client
3 with two sessions 3-1 and 3-2 in Fig. 27.2. IPn is the ip address of the n
client. IP-Urls sequence axis shows the sequence of Urls visited by client in an
individual session, and the sequence is arranged by time, for example, the Urls
of web pages represented by A, B, C, D, E, F… time axis stores the time that
client visit some page, for 2012-3-1 10:21:36 is the time that client 1 visits the
page A, this contributes to the sequence pattern mining after data prepro-
cessing. The Fig. 27.2 represents client 1 visits seven pages in one individual
session. In this way, it is convenient to store the results of client identification
and session identification and related useful information, thus laying founda-
tion for further improving the efficiency and accuracy of pattern mining.
(4) Disadvantage: in judging whether the current client and the client identified is
the same visitor, many factors must be considered, thus causing the decrease
of the operating rate of algorithm. But the analysis of web log file mining is
not real-time, so operating rate is not the top factor considered, and it is worthy
of weighing identification accuracy and operating rate.
264 X. Li
User
1 7 2012-3-1 10:21:36 2012-3-1 10:27:38 2012-3-1 10:32:22
Session
Time
IP—Urls Sequence
Fig. 27.2 A three- dimensional array data storage
27.2.4 Path Completion
Another important step of pre-processing is Path Completion. The number of
URLs those clients browse recorded in web log is less than the actual number those
clients browse, because of the page resident in local cache or agent cache in Web
Browser, the important information transmitted by Post technique unrecorded in
web log and the use of back button in browsing, which also lead to the incomplete
of URLs sequence related to client browsing in log files. Therefore to remedy the
effect of the question upon pattern mining, path completing is needed.
To avoid this question of the incomplete of access path caused by customers’
buffer, we can adopt the HTTP/1.1 agreements that forbid buffering. Though the
current website trends to provide dynamic information service, and buffering can
not be considered for websites adopted dynamic technique, static HTML page is
also applied in website practice, so the treatment of buffering is of great impor-
tance for session identification. Meanwhile, only one method can not solve the
problem of incomplete access path caused by back button, so some strategies need
to take. One strategy is to analyze with the help of topological graph in websites, if
the pages visited by uses include the link pointed to the current page, then suppose
that request should be sent by the present page (Cooley et al. 1999), so the omitted
URLs can be added to the existed path. Though the method can not guarantee the
one hundred percent accuracy, it can yet be regarded as an accessible way for it
can achieve good effect indicated by experiment.
27 Data Preprocessing in Web Usage Mining 265
27.3 Results
Object by a week’s log files (2012\3\1–2012\3\7) from http://my.sdyu.edu.cn/, in
all 132 M bytes, the log files with 336,741 records remaining after data cleaning,
with Athlon (tm) XP 1700+ as test-bed and internal memory 256 M, the test
identifies 5,625 clients and 18,536 sessions by CSIA. If only identifies by IP
address then only 5,270 be indentified, with 5,625-5,270 = 355 clients being
ignored, thus we can see the high client identification accuracy of CSIA. From the
view of marketing and customer service, if the 355 clients are clearly indentified,
and provide personalized service for each of them, then it may be probable that
quite a significant proportion of clients turn to the loyal customers, bring more
interests to company.
27.4 Conclusion
The paper pays attention to the data preprocessing module in web usage mining,
focusing on the specific realization of preprocessing, and proposes the algorithm
CSIA that can better identify client and session simultaneously, with higher
identification accuracy than the common identification algorithm. And the paper
hopes that the work done can help web usage mining researchers.
At present the study on Web Usage Mining at aboard is abundant, and some
universities and research institute at home also have studied on this aspect, but the
influential one is not that much. Considering the business value, wide application
prospect and the large developing space of its related technique of Web Usage
Mining, much more research strength will be put into this area. The focus of
research will have more tendencies on the visualization of pattern analysis and
analysis results, and man–machine interaction aspect based on the continuous
study on pattern discovery.
References
Cooley R (1997a) Web mining: information and pattern discovery on the world wide web. In: 9th
international conference on tools with artificial intelligence (ICTAI’97), New-port Beach,
USA, 1997, pp 558–567
Cooley R (1997b) Grouping web page references into transactions for mining world wide web
browsing patterns. In: Proceeding Of the IEEE knowledge and data engineering exchange
workshop (KDEX-97)
Cooley R, Mobasher B, Srivastava J (1999) Data preparation for mining world wide web
browsing patterns. J Knowl Inf Syst 1:5–32
Ji Y (2009) Application cases of data mining technology. China Machine Press, Beijing
Liu Y (2003) Research on content mining technology based on web. Harbin Institute of
Technology, Harbin
266 X. Li
Liu W (2007a) Design for web usage mining model. Appl Res Comput 24(3):184–186
Liu L (2007b) The Preprocessing of web usage mining. Comput Sci 5:200–204
Pirolli P (1996) Silk from a sow sear: extracting usable structures from the web. In: Proceeding of
1996 conference on human factors in computing systems (CHI-96), Vancouver, British
Columbia, Canada
Shao F (2009) Principle and algorithm of data mining. Science Press, Beijing, pp 379–380
Tang Q (2002) The text mining based on web. Comput Eng Appl 21:198–201
Wang J (2000) Research of web text mining. J Comput Res Dev 37(5):513–520
Wu Q (2002) Client identification in the processing of web log mining. Comput Sci 29(4):64–66
Xu M (2003) Study on text mining on web. Basic Autom (5):44–46
Zhang W (2006) Clustering web client based on interest similarity. Shandong Univ (Nat Sci)
41(6):54–57
Zhao W (2003) Research on data processing technology in web log mining. Comput Appl
23(5):62–64
Chapter 28
Design Knowledge Reduction Approach
Based on Rough Sets of HCI
Qing Xue, Qi-qi Yin, Li-ying Feng and Min-xia Liu
Abstract With the application of high-tech in the battlefield, the battlefield
environment became complicated. Whether the weapon is easy to use or not
depended on its interface, and determines the success or failure of the war. In order
to design weapon display interface and improve the usability of interactive system,
an approach to adaption reasoning based on rough sets is proposed. Condition
attributions of decision tables in the knowledge systems could be reduced, and it
simplified the adaption inference rules and related human-computer interface
design knowledge, which could be applied into the design practices easily. And
concise friendly adaptive human-computer interface could be designed to improve
the efficiency of operations.
Keywords Context Decision table Rough set Weapon display interface
28.1 Introduction
The theory of rough sets was originally proposed by Pawlak as a formal tool for
modeling and processing intelligent systems characterized by insufficient and
incomplete information (PawIak 1991; PawIak 1982). Its main advantage is that
through rough set we can find the connection and characteristics of the data and
extract the implied rules without the apriori knowledge and additional information
(Zhou and Wu 2011).
Q. Xue (&) Q. Yin M. Liu
Institute of Industrial Engineering, Beijing Institute of Technology,
Beijing, China
e-mail: [email protected]
L. Feng
Beijing Institute of Technology Zhuhai, Zhuhai, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 267
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_28,
Springer-Verlag Berlin Heidelberg 2013
268 Q. Xue et al.
A domain U and a equivalence relation R of U is given in the rough set,
8A U. The upper and downer approximate set of A is:
RðAÞ ¼ f xj½ xR \A 6¼ £g
RðAÞ ¼ f xj½ xR Ag
If RðAÞ 6¼ RðAÞ, then said that set A is rough set of equivalence relation R in
domain U (Li and Cercrone 2005; Miao and Li 2008).
Because of the large quantity of targets, complex cooperation relations and
frequent tactical movements in modern wars, it is very difficult to accept an
increasing number and types of data, and process them into information of weapon
system (Zhang et al. 2003; Pawlak 2002). In order to improve the efficiency of
operations and avoid misoperation, concise and friendly interface of multimedia
becomes more and more important. But in the design process, weapon context
information is complex, changeable and indistinguishable. In this paper, the
authors defined the weapon context as follows: context is to decide or influence
weapon display control calculation and the information of man-machine interac-
tive process, which may come from users, equipments and systems, the external
environment, and other entities. If we establish all interfaces corresponding to all
the context knowledge respectively, this will be a very big job (Zhang and Liang
1996; Wang et al. 2006). And there is a lot of information to deal with in the
system, which is a complicated constraint-based reasoning process.
At present, the basic methods of uncertain measurement in the rough set theory
are (Zhang et al. 2001): Precision, rough membership function, attribute depen-
dency attribute significance, inclusion degree and conditional information entropy
etc. Zhang (Cai and Yao 2009) has put forward the concept of inclusion degree as
a measurement of rough set data analysis, indicating that every measurement can
be concluded as inclusion degree. Beauboude et al. (Dai and Li 2002) introduced a
method of measuring the uncertainty of rough set based on information entropy,
which is better to reflect and can measure the uncertainty more precisely.
Design knowledge reduction approach based on rough sets of weapon display
interface is put forward in this paper.
28.2 Design Knowledge Expression System
We get the knowledge acquisition through the knowledge representation system,
and find useful information through analyzing the original data. We should take
the original knowledge expression into a new target expression form (easily for
computer process). Knowledge discovery based on the rough set theory, mainly
use such an effective knowledge expression-information list. The basic elements of
information list are the set of samples. Knowledge of samples described through
designating properties (characteristics) and their property values (characteristic
value) of the samples.
28 Design Knowledge Reduction Approach 269
Table 28.1 Part of the decision table
U C D
Against measure
Con1 Con2 … Conm
Target distance (km) Target speed (m/s) Target type
X1 24 500 Cruise missile D1
X2 18 300 Boomer fighter D2
X3 6 280 Tactical missile D3
… … … …
Xn 35 560 Air-to-ground missile Dt
Through observing the practical operation situation and the use of weapons’
interface to get a record form, it can be considered as a knowledge representation
system, KRS for short. In the KRS table, row presents research objects; and
column presents attribute (Cheng and Sun 2007; Wei et al. 2006). A list can be
regarded as a cluster of equivalence relation of attributes’ definition, and this kind
of list is usually called the decision table.
According to part of the provided data, the decision table has been constructed
as shown in Table 28.1:
In Table 28.1, domain U ¼ fx1 ; x2 ; . . .; xn g means that there are n contexts and
interface mode records; condition attribute set C ¼ fCon1 ; Con2 ; . . .; Conm g mean
that there are m different weapon equipment contexts, such as presenting target
distance, and presenting target speed. Decision attribute set D ¼ fUI1 ; UI2 ; . . .;
UIt gmeans that there are t different interfaces.
28.2.1 Pretreat the Decision Table
In order to obtain valuable knowledge from the decision table, the original data
must be pretreated first. The common pretreatment mainly includes two aspects:
totalize the incomplete data and discrete normalize the attribute value (Wei 2006).
28.2.1.1 Totalize the Incomplete Data
The context information of battlefield environment is often incomplete. For
example, sensor errors may cause target information lost partly. Therefore, before
establishing the decision table, totalize the incomplete data is very necessary. KRS
contains incomplete data, taking Table 28.2 as an example. Totalizing the
incomplete data method is presented as follows.
Table 28.2 is a KRS which contains incomplete data. It records the context
information Con1, Con2 and the corresponding interface decision attributes D.
270 Q. Xue et al.
Table 28.2 KRS which contain incomplete data
U Con1 Con2 D
Target distance (km) Target speed (m/s) Against measure
X1 15 500 D1
X2 25 680 D2
X3 10 400 D3
X4 3.5 D4
X5 6 800 D5
(1) Elimination method
Delete the row record which has attribute values missing, and get a complete
decision table. In Table 28.2, the object X4 should be deleted. Obviously, this
method is very simple, and we use it when in the decision table the incomplete
information object number is far less than complete information data; otherwise,
we cannot use this method.
(2) Compensation method
For the KRS which contain incomplete data, we use the following approaches
to add the missing data and complete the KRS:
(i) According to the actual requirements, take incomplete attribute value as a
kind of special value; (ii) Using statistical principle, assess the missing attribute
values according to the rest of the record object attribute value. If the missing value
is numerical, take the arithmetic mean value of this property in other objects as a
supplement; and if the missing value is no-numerical, take the highest frequency
value of this property in other objects as the supplement. In Table 28.2, the arith-
metic mean value of Con2 in other objects is (500 ? 680 ? 400 ? 800)/4 = 595.
We take 595 as a supplement. In addition, there are some other methods of com-
pensation, such as condition combination compensation method, and compensation
method based on the indiscernibility relation.
28.2.1.2 Discrete Normalize the Attribute Value
The weapon context information with numerous types, some are numerical while
some not, some are continuous while others not, and they have different value
confines and measurement units. It is hard to create a decision table with clear
structure, if we do not discrete normalize the attribute value. Discrete normalize
the attribute value must meet the needs of the two aspects: One hand, the
dimension of the attribute value should be reduced. On the other hand, the
information loss should be avoided to the greatest extent. There are two ways of
discrete normalize the attribute value.
28 Design Knowledge Reduction Approach 271
(1) Partial discretization and normalization
In this method, we consider only one attribute in the decision table. As for a
continuous attribute a, with its range ½amin ; amax , discrete and normalize mean the
generating set of a group, f½d1 ; d2 ; ½d2 ; d3 ; . . .½dn1 ; dn g and d1 ¼ amin ; dn ¼ amax .
When they are in the same range of values, we use ðamax amin Þ=k ¼ ðd2 d1 Þ ¼
ðd3 d2 Þ ¼ ðdn dn1 Þ as a division. If the number of the attributes is m, and
there are k levels, every sample size of value range is m=k.
(2) Global discretization and normalization
In this method we manipulate all the attributes in the decision table, such as the
method based on Boolean logic and rough set theory. Main processes of operation
are as follows: the attribute values can be shown by the set and symbolic which are
used to represent the gap between contact attribute values; disjunction types are
used to represent different decisions; conjunction expressions are used to represent
the disjunction types and could be translated into disjunction types; discretization
results could be obtained from one of the disjunction types.
28.2.2 Reduce the Decision Table
When we design the weapon human-computer interface, there are many data
samples and condition attributes in the decision table. The decision table is very
complex, and it is hard to find the implied knowledge from the data. The reduction
process of decision table is deleting the redundant data, reducing condition attri-
bute dimensionality and reducing the sample size.
In the actual battlefield, the situation of the battlefield is diversity and uncer-
tainty. Sometimes, we got the same sample information, but we made different
decisions; so it is an inconsistent decision table. When we process the inconsistent
decision table, we usually transfer it into a consistent table. The reduction method
is as follows:
(1) Firstly, merge the repeat record and reserve the inconsistent part;
(2) Secondly, simplify the condition attributes. Delete every condition attribute
in turn, if the table changed into inconsistent, the attribute cannot be reduced, or it
can be reduced;
(3) Thirdly, simplify the decision rule. On the basis of above, delete an attribute
for every decision rule. If the table changed into inconsistent, the attribute cannot
be reduced, or it can be reduced. Then get the core value table of decision;
(4) Finally, according to the core value table generated the simplest decision
table. That is to say, if we delete some core value, the decision rule is the same
with others, so it is the simplest decision table.
The core part of above method is, specific algorithm of reducing attribute, and it
is shown as follows (Meng et al. 2008):
272 Q. Xue et al.
Algorithm: The algorithm of reducing attributes
Input: condition set C ¼ fai ji ¼ 1; 2; . . .; ng, and decision set D;
Output: the minimum reduction decision table.
Let B ¼ CORED ðCÞ; A ¼ C CORED ðCÞ.
Step1. [Is |B| = 0?], if |B| = 0, REDD(C) = C and exit; else, calculate and go to
Step4.
Step2. For each ai ^I A, calculate kBE fai g ðDÞ
Step3. For ai ^I A, satisfied kBE fi g ðDÞ ¼ min kBE fi g ðDÞ ; A ¼ A fai g; B ¼
B [ aj
Step4. If (ðkB ðDÞÞ ¼ kC ðDÞ, then REDD(C) = B; else go to step2.
Through reducing the decision table, we get a more simple decision rule. Not
only reducing redundant data, but also improving the decision efficiency obvi-
ously. It can help to establish a simple expression from context space to interface
design space. Using this method can realize the adaptive interfaces easily.
28.3 The Design of a Weapon Defense Interface
The rapid development of modern optoelectronic technology, greatly promotes the
military photoelectron technology to be mature and perfect. In military applications,
the photoelectric precision technology and photoelectric detection technology
develop extensively and rapidly. Currently it became a more perfect equipment
system (Mi and Li 2009). This example is the generation of optoelectronic coun-
termeasure interface in different battlefield situation. The condition attributes are
target distance a1 (km), target type a2, target speed a3 (m/s), and route shortcut a4
(km); and decision attributes d, which are different interfaces with different opto-
electronic countermeasures. In this example the decision attribute d are information
apperceive 1, laser alarm 2, laser countermeasure 3, smoke set 4, and optical
camouflage 5.
According to the provided data, construct the original decision table as shown
in Table 28.3.
28.3.1 Attribute Discredited
Before establishing the final decision table, the original decision table should be
discredited. In this paper, due to the attributes have different nature, we use local
discrete method and adopt experts subjective designed to discrete the attributes.
Discrete standards are provided by application unit shown in Table 28.4.
28 Design Knowledge Reduction Approach 273
Table 28.3 Original decision table
Sample no. a1 a2 a3 a4 d
1 24 Helicopter 380 20 1
2 16 Cruise missile 560 13 2
3 6 Helicopter 360 5 3
4 2 Cruise missile 540 1 2
5 15 Tactical missile 1200 16 2
6 25 Boomer fighter 760 22 1
7 0.06 Air-to-ground missile 460 0.05 4
8 12 Helicopter 340 8 2
9 30 Boomer fighter 660 15 1
10 0.9 Air-to-ground missile 740 0.7 4
11 18 Air-to-ground missile 800 14 2
12 0.05 Cruise missile 700 0.02 5
13 5 Air-to-ground missile 350 18 1
14 3 Helicopter 320 1 4
15 8 Tactical missile 780 7 3
16 26 Boomer fighter 680 15 2
17 0.08 Air-to-ground missile 460 0.06 5
18 7 Tactical missile 1000 5 3
19 5 Tactical missile 750 2 4
20 8 Cruise missile 720 6 3
21 17 Boomer fighter 780 12 2
Table 28.4 Discrete standards
1 2 3 4 5
A1 Above 10 5–10 0–5
A2 Helicopter Boomer fighter Cruise missile Air-to-ground missile Tactical missile
A3 0–250 250–500 500–750 750–1000 Above 1000
A4 Above 15 7–15 0–7
In Table 28.4, A1 represents target distance, A2 represents target type, A3
represents target speed, A4 represents route shortcut. The decision table after
discrete is shown in Table 28.5.
28.3.2 Reduction the Decision Table
Given that the condition set of decision table is C ¼ fa1 ; a2 ; a3 ; a4 g, and decision
attribute set is D ¼ fdg. Let the initial attribute core is COREDðCÞ ¼ . From the
decision table, we get that U=C ¼ffX1 g; fX2 g; fX3 g; fX4 g; fX5 g; fX6 g; fX7 g;
fX8 g:fX9 ; X16 g; fX10 g; fX11 g; fX12 g; fX13 g; fX14 g; fX15 g; fX17 g; fX18 g; fX19 g;
fX20 g; fX21 gg, and U=D¼ffX1 ; X6 ; X9 ; X13 g; fX2 ; X4 ; X5 ; X8 ; X11 ; X16 ; X21 g;
274 Q. Xue et al.
Table 28.5 The decision Sample no. a1 a2 a3 a4 d
table after desecrating
1 5 1 2 1 1
2 4 3 3 2 2
3 3 1 2 3 3
4 2 3 3 3 2
5 4 5 5 1 2
6 5 2 4 1 1
7 2 4 2 3 4
8 4 1 2 2 2
9 5 2 3 2 1
10 2 4 3 2 4
11 4 4 4 2 2
12 1 3 3 3 5
13 5 4 2 1 1
14 2 1 2 3 4
15 3 5 4 2 3
16 5 2 3 2 2
17 1 4 2 3 5
18 3 5 5 3 3
19 2 5 3 2 4
20 3 3 3 3 3
21 4 2 4 2 2
Table 28.6 The decision Sample no. a1 a3 a4 d
table after reducing
1 5 2 1 1
2 4 3 2 2
3 3 2 3 3
4 2 3 3 2
5 4 5 1 2
6 5 4 1 1
7 2 2 3 4
8 4 2 2 2
9 5 3 2 1
10 2 3 2 4
11 4 4 2 2
12 1 3 3 5
13 5 2 1 1
14 2 2 3 4
15 3 4 2 3
16 5 3 2 2
17 1 2 3 5
18 3 5 3 3
19 2 3 2 4
20 3 3 3 3
21 4 4 2 2
28 Design Knowledge Reduction Approach 275
Table 28.7 The minimum reduction decision table
Sample no. a1 a3 a4 d
1,13,6 1
2,5,8,11,21 4 2
3,15,18,20 3 3
4 2 3 2
7,14 2 2 4
9 5 3 2 1
10,19 2 2 4
12,17 1 5
16 5 3 2 2
Fig. 28.1 Adaptive human-computer interface of weapon system
fX3 ; X15 ; X18 ; X20 g; fX7 ; X10 ; X14 ; X19 g; fX12 ; X17 gg. According to the algorithm,
we get, POSðC; DÞ ¼ ffX1 g; fX2 g; fX3 g; fX4 g; fX5 g; fX6 g; fX7 g; fX8 g; fX9 g;
fX10 g; fX11 g; fX12 g; :fX13 g; fX14 g; fX15 g; fX16 g; fX17 g; fX18 g; fX19 g; fX20 g;
fX21 gg; kC ðDÞ ¼ 19=21. So the decision Table 28.5 is inconsistent. And we get
that inconsistent of decision table is because sample X9 and sample X16 have the
same condition attributes but different decision attributes. After remove a1 , get
kCfa1 g ðDÞ ¼ 11=21 6¼ kC ðDÞ, so cannot be remove. That is to say,
a1 I CORED(CÞ. Similarly, we can obtain the dependency of other condition
276 Q. Xue et al.
attributes to the decision attribute D, kCfa2 g ðDÞ ¼ 15=21 6¼ kC ðDÞ,
kCfa3 g ðDÞ ¼ 19=21, kCfa4 g ðDÞ ¼ 19=21. Minimum attribute reduction set is
fa1 ; a3 ; a4 g. Table 28.6 shows the decision table after reducing.
Similar attribute reduction, remove the decision table redundant condition
attribute value, and with the same rules, finally obtained such as the minimum
reduction decision table shown in Table 28.7.
The reduced decision table made the task context and optoelectronic counter-
measure form decision knowledge simply. Then we reasoned and simplified the
design rules. In addition, these rules more simplified than the original records.
28.3.3 Example
Weapons systems require a good real-time and stability, so we choose VxWorks as
the system platform, and use WindML development tools to develop weapon
system interface. We have got the reduction decision table above, and made the
goal situation and photoelectric form decision knowledge simplified. The adaptive
interface for against measures can be obtained according to the sensed actual
context and decision reasoning rules. It is shown in Fig. 28.1.
28.4 Conclusion
In this paper, we used rough sets to reduce conditional attributions of decision
tables’ in the knowledge systems. Also we extracted key attributes from a lot of
original data and established simple expression from context space to interface
design space. The adaptive interfaces realized easily using this method.
References
Cai L, Yao X (2009) Attribute reduction algorithm based on rough set theory. J Sichuan Univ Sci
Eng 22:34–37 (in Chinese)
Cheng S, Sun S (2007) Adaptive human-computer interaction user modeling based on rough sets.
Comp Sci 34:48–52 (in Chinese)
Dai JH, Li YX (2002) Study on discretization based on rough set theory. In: Proceedings of 1st
international conference on machine learning and cybernetics, Piscataway, IEEE Press,
pp 1371–1373
Li J, Cercrone N (2005) Empirical analysis on the geriatric care data set using rough sets theory.
Tech Report, CS-2005-05
Meng J, Liu Y, Mo H (2008) New method of packing missing data based on rough set theory.
Comp Eng Appl 6:175–177 (in Chinese)
Mi R, Li X (2009) The knowledge reduction in incomplete fuzzy decision information system.
J Wuhan Univ Sci Eng 22(3):24–28 (in Chinese)
28 Design Knowledge Reduction Approach 277
Miao D, Li D (2008) Rough sets theory algorithms and application. Tsinghua University Press,
Beijing (in Chinese)
PawIak Z (1982) Rough sets. Int J Comp Inf Sci 15:341–356
PawIak Z (1991) Rough sets-theoretical aspects of reasoning about data. Kluwer Academic
Publishers, Dordrecht
Pawlak Z (2002) Rough sets and intelligent data analysis. Inf Sci 147:1–12
Wang X, Cai N, Yang J, Liu X (2006) Methods of rough set uncertainty measure based on
approximation precision and condition information entropy. Acad J Shanghai Jiao Tong Univ
51(7):1130–1134
Wei D (2006) Research on rough set model and knowledge deductions for incomplete and fuzzy
decision information system, Nanjing University, Nanjing (in Chinese)
Wei D, Zhao Y, Zhou X (2006) A rough set approach to incomplete and fuzzy decision
information system. 2006 IEEE 6th world congress on intelligent control and automation,
June 2006
Zhang W, Liang Y (1996) The theory of uncertainty reasoning. Publishing Houses of Xi’an
Jiaotong University, Xi’an
Zhang W, Wu W, Liang J (2001) Rough set theory and method. Science Press, Beijing,
pp 182–185
Zhang W, Liao X, Wu Z (2003) An incomplete data analysis approach based on rough set theory.
Pattern Recognit Artif Intell 16:158–163 (in Chinese)
Zhou L, Wu W (2011) Characterization of rough set approximations in atanassov intuitionistic
fuzzy set theory. Comp Math Appl 62:282–296
Chapter 29
Diagnostic Testing and Analysis of CAN
Data Bus Based on the Sagitar Power
Transmission System
Wen Fang
Abstract Based on the Sagitar power system of CAN data bus, analysis of a
power transmission system network structure, elaborated the power system of
CAN data bus termination resistor working principle and test method. Diagnostic
testing by the BOSCH FSA740 standard engine analyzer, a power system CAN
data bus circuit breaker, CAN_H and CAN_L data between the wire short circuit,
CAN data bus to the positive or negative electrode short-circuit diagnostic testing
and analysis.
Keywords CAN Data bus Diagnosis Testing Power transmission system
29.1 Introduction
Through the network can satisfy the people to the modern vehicle driving safety,
comfort, emission performance and fuel consumption growing demand, reducing
wire and connector number reduces the required space and the weight of a vehicle.
But the automobile network application for repair and technical personnel raised
taller requirement, is the current automotive electrical network repair of a problem.
Automobile network structure for various manufacturers varies, generally
divided into a power transmission system, comfortable and infotainment systems,
combination of instruments and diagnostic interface system, each system data
exchange through CAN data bus. Wiper motor, light and rain sensor and anti-theft
alarm device assembly using the LIN data bus. Due to the adoption of the central
diagnosis interface or gateway, so the CAN data and the LIN data can be cross-
border data exchange. In this chapter, based on the FAW-Volkswagen power of
W. Fang (&)
Department of Automotive Engineering, Sichuan Vocational and Technical College of
Communications, Chengdu, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 279
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_29,
Ó Springer-Verlag Berlin Heidelberg 2013
280 W. Fang
Fig. 29.1 Sagitar power system network structure
CAN data bus, through a vehicle network diagnostic test for repair, technical staff
to provide a good reference.
29.2 Network Structure of Power Transmission System
Power CAN data bus sub-system includes engine management system control unit,
automatic transmission, with EDS ABS, steering column electronic device, illu-
mination distance adjusting device and airbag. The data transmission rate of
500 kbit/s, each control unit receives information by CAN data bus continuously
from the other control unit, and immediately on power system working condition
changes, for real-time bus. The data bus of short circuit or open circuit causes the
bus off (Fig. 29.1).
29.3 The Principle and Detection of Power System CAN
Data Bus Terminal Resistance
29.3.1 The Principle of Power System CAN Data Bus
Terminal Resistance
For the CAN data bus signal, bus wire end is equivalent to the role of independent
transmitter, so at the end of wire will produce reverse operation of the signal, the
signal is superimposed on the effective signal will cause the distortion, must be in
29 Diagnostic Testing and Analysis 281
the high frequency bus network end termination abort signal, which might
otherwise occur reflection. The process of reflection and crashed into a dock
embankment, then reflection and with follow-up wave superposition is similar,
terminal resistance effect is like sand, if the wave washed onto the beach, the beach
will absorb the energy of waves and causing no wave superposition. So the data
bus to be connected to a terminal resistor termination data transmission, absorption
signal operation to the data wire end when the energy.
High speed bus connects the CAN physical interface typically uses the standard
ISO11898. This standard stipulated the transmission medium for a two bus lines,
two terminal resistances are provided for 120 ohm. But not all manufacturers are
using the standard ISO11898, in the Sagitar power transmission system on the
CAN data bus, the data wire end without installing the standard two 120 ohm
resistor. But the engine management system control unit of resistance of 66 ohm
resistor to assume a central, power transmission system bus remaining on the
control unit has a high resistance, each of the resistance for 2.6 k ohms. Because
the control system of the resistance is connected in parallel, so according to the
following way of calculating the load resistance of the total resistance value:
1 1 1 1 1 1 1 1
Rges ¼ þ þ þ ¼ þ þ þ ¼ 61:35 X
R1 R2 R3 R4 2600 2600 2600 66
29.3.2 The Detection of Power System CAN Data Bus
Terminal Resistance
In the system can use the resistance measuring device for short and open fault
diagnosis. As the power transmission system of CAN data bus through a bus
terminal 15 is switched on and in short-term continue operation after shut down
completely, so it will be in the interval after the time, allowed pull off the control
unit plug, check the CAN_H and CAN_L data between the wire resistance. If the
resistance value is greater than 250 ohms, show that the engine control unit of a
data conductor circuit; if the resistance value is less than 30 ohm, show that the
data wires may exist between short circuit; if the CAN_L or CAN_H data wire and
a grounding between measured resistance value is less than 300 ohm, then the
negative short circuit (Fig. 29.2).
29.4 Power Transmission System CAN Data Bus Diagnosis
Test and Analysis
When the automobile multiplex system for information transmission link (or
communication lines) have a fault occurs, the communications line short circuit,
282 W. Fang
Fig. 29.2 The Sagitar power
system CAN data bus
termination resistor
connection
broken circuit and circuit physical nature of the communication signal attenuation
or distortion, will cause multiple electronic control unit not working or electronic
control system error action. Power transmission system CAN data bus failures that
may occur with the power transmission system of CAN data bus circuit, CAN_H
and CAN_L data between the wire short circuit, CAN data bus to the positive or
negative electrode short circuit. Through the BOSCH FSA740 standard engine
analyzer diagnostic test, observation different of data communication signal and
standard communication data signal.
29.4.1 Power Transmission System CAN Data Bus Normal
Waveform Test
From the oscillogram, CAN_H signal, the CAN_L signal, signals are under the
symmetrical arrangement. Level from dominant to recessive state switching state
without interference. CAN_H and CAN_L signal recessive level is 2.5 V, domi-
nant state CAN_H signal voltage value is about 3.5 V, CAN_L signal voltage
value is about 1.5 V. According to the bus load, these values may be vary in the
hundreds of millivolt range. Voltage signal curve depends largely on the use of
oscilloscope. The general rule is: the higher resolution, signal display more clearly
(Fig. 29.3).
29 Diagnostic Testing and Analysis 283
Fig. 29.3 Normal system oscillogram
29.4.2 Power Transmission System CAN Data Bus Circuit
Breaker Waveform Test
29.4.2.1 CAN_L Wire Breaker
Oscilloscope preferably using two channel and measurement data wire signal
voltage to ground for diagnosis, can be more easily analyzed voltage level and
fault diagnosis. Relative measurement data wire on the voltage signal display the
voltage difference. From the oscilloscope, CAN_H and CAN_L signal level are
2.5 V recessive dominant state CAN_H signal voltage value of about 3.5 V, is
normal waveform; the voltage of the CAN_L signal values of approximately 3 V,
is abnormal waveform. Thus can judge CAN_L wire circuit breaker (Fig. 29.4).
CAN_L fault signal
Fig. 29.4 CAN-L wire breaker oscillogram
284 W. Fang
CAN_L fault signal
Fig. 29.5 CAN-H wire breaker oscillogram
29.4.2.2 CAN_H Wire Breaker
From the oscilloscope, CAN_H and CAN_L signal level are 2.5 V recessive.
Dominant state CAN_H signals voltage value of about 2.3 V is abnormal wave-
form; the voltages of the CAN_L signal values of approximately 1.5 V, is normal
waveform. Thus CAN_H wire circuit breaker can be judged (Fig. 29.5).
29.4.3 Power Transmission System CAN Data Bus: On
the Positive or Negative Electrode Short Circuit
29.4.3.1 CAN_L Short Circuit to Negative or Positive Electrode
CAN_L line voltage is 0 V if the CAN_L data wire short circuit to ground.
CAN_H wire recessive level becomes 0.2 V from about 2.5 V, at the same time
the voltage level becomes the dominant state from the recessive state; if the
CAN_L data wire short circuit to positive electrode, and the engine is shut off and
the connection without contact resistance, CAN_L signal voltage level is increased
to the voltage value 12 V (Fig. 29.6 is about 11.75 V), while in the data wires
cannot be observed changes from the recessive state level to the dominant state
(Fig. 29.7).
29.4.3.2 CAN_H Short Circuit to Negative or Positive Electrode
The oscilloscope must be observed two data wires without voltage signal if
CAN_H data wire on the negative short. CAN_H data wire voltage level becomes
0 V, CAN_L becomes 0.2 V wire level; if the CAN_H data wire short circuit to
positive, CAN_H signal voltage level is increased to the voltage value 12 V when
the engine is shut off and the connection without contact resistance, at the same
29 Diagnostic Testing and Analysis 285
Fig. 29.6 CAN_L short circuit to negative electrode
Fig. 29.7 CAN_L short circuit to positive electrode
Fig. 29.8 CAN_H short circuit to negative electrode
time in the two data wires cannot be observed changes from the recessive state
level to the dominant state (Figs. 29.8 and 29.9).
286 W. Fang
Fig. 29.9 CAN_H short circuit to positive electrode
Fig. 29.10 Between CAN_H and CAN_L data wires short circuit oscillogram
29.4.4 Between CAN_H And CAN_L Data Wires
Short Circuit
Oscillogram shows two voltage signals overlap each other and curves of the same
if data wires are connected to short circuit each other. The two signal voltages are
2.5 V approximately (Fig. 29.10).
29.5 Conclusion
CAN bus (multiplex transmission system) is one of the most promising buses; its
application field is constantly expanding. Multiplex transmission system has
incomparable advantages to the traditional automobile wiring. Because high
technology content, there are still some fault can not accurately find and eliminate
29 Diagnostic Testing and Analysis 287
via the test equipment. Automobile repair and technical personnel must be find
fault by the systematic analysis of multiplex network basic structure and signal
transmission principle, the flexible application of multiplex transmission network
specific diagnostic methods, multi-path transmission network diagnostic test, find
out the multiplex network structure inherent laws.
Chapter 30
Effectiveness to Driving by Interbeds
in Channel Sand of River Delta System
Lun Zhao, Shu-qin Wang, Zi-fei Fan, Lei Zhang
and Ben-wei Li
Abstract Heterogeneity of remaining oil distribution makes it difficult to recover.
This paper analysed the interbed in the channel sand which is the one of the main
factors to form remaining oil, and simulated respectively water and polymer
flooding for channel sand with all kinds of interbeds, and then nailed down the
affect to the distribution of remaining oil by different driving means. Horizontal
interbed which developed mainly in the braided channel and straight distributary
channel would obstruct the fluid during flowing vertically, and the remaining oil
congregates under the horizontal interbed. Polymer driving can accelerate the
moving of most remaining oil, but it is unconspicuous for the one under the
interbed. Lateral interbed which mainly deposit in the meandering and low curving
distributary river would shelter from sideway markedly. The degree of this effect is
based on the relation of directions of interbed and driving.
Keywords River Delta System Channel sand interbed Polymer driving
L. Zhao (&) S. Wang Z. Fan L. Zhang
PetroChina Research Institute of Petroleum Exploration and Development, Beijing, China
e-mail: [email protected]
S. Wang
e-mail: [email protected]
Z. Fan
e-mail: [email protected]
L. Zhang
China University of Geosciences (Beijing), Beijing, China
B. Li
PetroChina Jidong Oilfield Company, Tangshan, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 289
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_30,
Ó Springer-Verlag Berlin Heidelberg 2013
290 L. Zhao et al.
30.1 Introduction
In China, the Mesozoic and Cenozoic basins of continental sediments are wide-
spread with rivers and developed sands.
The river-deltaic sandstone dominates the continental basins. It is ascertained
that in China the majority of oil reserves are located in the river-delta deposition
system. Songliao Basin is China’s largest continental depression-type basin of oil
and gas with the typical dual-structure featured by lower rifting and upper
depression. Before the early Cretaceous Denglouku it is the depression evolution
stage of the basin. The Quantou Period-Neijiang Period marks the evolution stage
of depression (Liu et al. 2010), in which the formation of oil and gas at the middle
and lower part of Daqing Placanticline is a large river-delta depositional system of
depression sediments. The channel sand is the main part of oil and gas reserves. It
is found through the study of depositional mechanism of channel sand and during
the oil field development process that, channel sands are not homogeneous con-
nection (Wang et al. 2009). Changes in various factors like hydropower conditions,
provenance supply and climate lead to the extensive heterogeneity (Sun et al.
2008) within the channel sand; particularly the prevalence of impermeable inter-
bed affects the laws of fluid flow.
The knowledge of interbed of channel sands and genesis and the analysis of
driving effects of various flooding means are particularly important for the
development of oil fields.
In recent years, there is a growing research of interbeds. Jiao Hongbing once
mentioned in the ‘‘Influences of Interbed Distribution of Meandering River on the
Development Effective Driven by Water’’ that: factors like interbed dip angle,
spaces and the ratio of interbed depth to reservoir depth. As the dip angle (3–15°)
is increased, the rate of recovery driven by water is also increased and the 7–15
folds is a moderate growth; the increased interbed space (10–40 m) and the
recovery driven by water exhibits a nearly linear rise; H interbed/H oil layer (0–2/
3) is increased and the recovery driven by water is lowered. After the comprhesive
collection of distribution features of all kinds of channel sands in Daqing Oilfield,
a study of types, genesis of the channel sand interbed and their influences on the
driving effects of various driving means.
30.2 Type and Genesis of Interbeds
Daqing Placanticline Puyi Formation is a well-shaped river-delta deposition,
developed with flood plain, distributary plain, delta-front facies (Fig. 30.1) from
north to south and various fluvial facies sands like braided river, meandering river
and straight distributary (Zhang et al. 2009). The channel interbed consists of two
categories: horizontal interbed and lateral interbed.
30 Effectiveness to Driving by Interbeds in Channel Sand of River Delta System 291
Fig. 30.1 Distribution of
unit channel sand, Daqing
Placanticline PI23 (study on
the ‘‘Detailed Reservoir
Sedimentolog of Songliao
Basin’’, authored by Zhao
Hanqing modified)
30.2.1 Horizontal Interbed
Horizontal interbeds mainly are located in braided rivers of oilfields like Lamadian
and Saertu, low-bend distributary channel of oil fields like Xingshugang and straight
distributary channel. There are two major genetic types: (1) due to changes in the
hydropower condition or provenance supply during the deposition process, interbeds
are formed among sandy laminae reservoir of fluvial facies, including the muddy
intercalation and conglomerate interbed cemented by muds; (2) interbed arising from
physical and chemical effects during the diagenesis of sediments. Such interbed is
292 L. Zhao et al.
mainly calcareous (Yu et al. 2010). In addition, there is a marked difference in the
lithology of sediments, the upper and lower part of reservoir. Because the vertical
accretion of the suspended sediment featured by fine lithology, small depth, exten-
sive distribution and wide scope, are capable of forming an vertically upward
impermeable shelter (Li et al. 2009) during the crest concussion process in the flood
event of braided river.
30.2.2 Lateral Interbeds
Lateral interbeds are mainly developed in the meandering rivers of oils fields like
Shaertu and Xingshugang. Under the action of lateral interbeds, each flood forms a
lateral accretion. An argillaceous layer can be formed during the dry season and
interflood period between two floods (Zhang et al. 2008). The oblique crossing of
lateral accretion with the formation boundary and of overlaying argillaceous layer
with the formation boundary form the argillaceous lateral interbed in the channel.
Generally, the argillaceous lateral interbed is only developed at the upper and middle
part of point bar, thus forming a ‘‘semi-connection’’ featured by the connected
bottom, disconnected upper and middle portions of the point bar of meandering river.
30.3 Impacts of Horizontal Interbeds on Different Means
of Driving Effects
30.3.1 Mechanism of the Horizontal Interbed Affecting
Infiltrating Fluid
The presence of horizontal interbeds affects the flow of reservoir fluids and
changes the distribution of entire field of infiltrating fluid (Wang and Zeng 2008).
The higher the development frequency of interbeds, the more complex distribution
of oil and water movement. The development of a large number of unstable
interbeds reduces the sinking speed and gravity action of injected water along the
moving direction for the sand bed of fining upwards sands, favorable for the
vertical swept volume and and oil driving efficiency; but for coarsening upward
sands, it prevents the the effect of gravity sinkage on the swept volume during the
water driving oil process. The fining upwards sands without interbeds are greater
than the those with interbeds in the flooded intensities. The flooded intervals of the
bottom appears earlier. While the unstable interbeds of the coarse upward sands
reduces the oil scavenging cofficients and oil driving efficiency of gravity in the
coarse upward sands. It is of severe implications on the development effects. The
longer or more interbeds, the worse development effects. It mainly lies in that the
oil driving efficiency near the lower part of interbeds is far less than other regions.
The dead oil area is formed during the latter stage of high water content.
30 Effectiveness to Driving by Interbeds in Channel Sand of River Delta System 293
30.3.2 Impacts of Horizontal Interbeds on the Remaining
Oil During the Water Flooding and Polymer
Flooding Process
According to the reservoir distribution characteristics of major oil reservoirs of
Daqing Placanticline Lamadian and Saertu Puyi Formation, the three-dimensional
geological model of braided river sand of developed interbeds are set up to carry
out the numerical simulation of water flooding and polymer flooding.
During the water flooding process, when the water content accounts for 92.5 %,
the region of oil saturation from 0.4 to 0.6 is narrow, the line of 0.6 and line of 0.7
are basically coincident. If the water content reaches 0.98 with the sustained water
driving, the remaining oil will continue to ‘‘be floated’’. The obstruction effect of
interbeds remains obvious. But generally, the remaining oil is floated on the upper
layer (Fig. 30.2). The horizontal interbeds under specific conditions (within a
small area) can serve to prevent crossflow between layers and expand swept
volume. A portion of water advances along the upper side of interbeds, so that the
recovery of reservoir above the interbed is high (Su 2004). But if the hydropower
is not strong, a lot of crude oil will remain under interbeds, particularly those at the
middle and lower part of reservoirs with most obvious blockage. The more
interbed area, the more remaining oils.
If polymer flooding is used when the water content is 0.925, the profile control
of polymers changes the original fluid direction, the above crude oil will move
downwardly and then upwardly. If the crude oil moves downwardly and meets the
blockage of interbedds, it is favorable for the recovery. If the crude oil moves
downwardly below the interbeds, then rises and is blocked by interbeds, then it is
unfavorable for the enhanced oil recovery. The interbed itself also blocks the
Fig. 30.2 Profile of So. at
different stages of braided
river sands (1)–(5) are water
driving with 92.5 and 98 % of
water content, the polymer
flooding with water content
dropped to 64.6 % and water
driving until reaching the
water content of 98 % after
the end of polymer driving
294 L. Zhao et al.
polymer and exerts negative impacts on the expanded swept volume of the
polymer profile control.
Through the comparison of Fig. 30.2 (3) with (4), it can be seen in the polymer
flooding process, the remaining oil near the interbeds is not significantly improved.
Through the comparison of Fig. 30.2 (1) with (4), the remaining oil is diffused in
the local areas after the polymer flooding. The region with the oil saturation of
between 0.4 and 0.6 is much wider than the that before the polymer flooding. Only
in local aeras like interbed bottom and top, etc., the oil saturation of remaining oil
is as same as that before polymer flooding. It shows the polymer flooding can
develop remaining oil in most regions within the braided channel sands. But due to
the presence of the horizontal interbeds, the remaining oil at the bottom has not
been significantly improved.
30.4 Impacts of Lateral Interbed on Different Means
of Driving Effects
The impacts of lateral interbed on the distribution of the remaining oil are closely
related to the combination of direction of dip and injection direction (Ling et al.
2008). There are three major combination relationship of the injection direction
and interbed inclination, namely the direction, against along the interbeds, or the
extension direction of interbeds fluvial ldirection) (see Fig. 30.3). For the study of
different combinations of direction and the impact of lateral interbeds on the
distribution of remaining oils, three dimensional conceptual model of three
combinational directions are set up for the numerical simulation.
30.4.1 No Interbed Development
It is relatively homogeneous in the channel sand without the interbed development.
In the water flooding process, without the blockage of interbeds, the water flooding
process is mainly controlled by the channel deposition model and is easier based
Fig. 30.3 Schematic view of combinational relationship of lateral interbed and injection
direction
30 Effectiveness to Driving by Interbeds in Channel Sand of River Delta System 295
on sedimentary rhythm. Figure 30.4 is a profile of oil saturation at different stages
of water flooding. In Fig. 30.4 (2) the unused remaining oil distributed in the
triangular area formed by the connecting line of the middle part of oil well and
water well and the reservoir top when the water content is 0.925. If you continue
with the water flooding, the unused remaining oil only stays on the reservoir top
(see Fig. 30.4 (3)). When the water content is 0.925, the recovery rate is 35.51 %;
when the water content by the water flooding reaches 0.98, the recovery rate is
41.28 %.
If polymer driving is used when the water content is 0.925, the phenomenon of
‘‘Oil Wall’’ will appear.
Because the viscosity of polymer solution is greater than that of water, the
pressure on the water well should be increased from 15 to 26 MPa at the time of
water driving during the polymer injection. At the same time the injection grout is
increased by folds. After the polymer solution enters the formation, the pressure
cannot be transferred quickly, the pressure gradient surrounding the water well is
great, the dip angle of water testing surface becomes bigger. Based on the Hubbert
potential formula, the dip angle of oil–water interface dip angle will also become
larger, thus a part of oil will move downwardly. As the pressure advances towards
the oil well, an oil wall is formed in front of polymer aggregation with the V-
shaped profile which appears to move downwardly driven by polymers. When the
oil wells arrive at the oil well, a sudden peak will appear in the oil well yield while
the rear oil wall disappears gradually and the oil well output is decreased
Fig. 30.4 Cross-section of oil saturation with the water content of 0.8, 0.925 and 0.98 with water
driving at the homogeneous sandstone oil reservoir without interbed and the water content of 0.98
through the follow-up water driving
296 L. Zhao et al.
dramatically. The appearance of oil walls or the obvious oil wall is related to the
reservoir non-homogeneity and well spacing. In general, the more severe hetero-
geneity, the greater differences in the positive rhythm, the less likely to appear the
oil wall. The braided channel sand due to sound connectivity of reservoir has less
penetration rate differences, thus the obvious oil wall appears. At the same time the
appearance of oil wall is controlled by well space. Due to the upper lateral interbed
and lower belts of high permeability of meandering river sands, the injected
polymer solution suddenly ingresses along the bottom. In addition, when the
remaining oil in the interbed is displaced, it only flows out of the bottom of lateral
interbed, thus no V-shaped oil wall will be formed.
The polymer flooding is used when the water content is 0.925. Figure 30.4 (4)
and (5), respectively show the profile of oil saturation when stopping the polymer
injection and the follow-up water driving until the water content reaches 0.98, the
remaining oil is still distributed on the top of reservoir near to the oil well. The
ultimate recovery rate for 0.70 PV of injected polymer is 42.43 %. (Table 30.1
shows the equivalent of the relationship between the injection of different volume
of polymers and recovery rate. The ultimate recovery rate after displacement refers
to the corresponding value of ultimate recovery when the follow-up water driving
enables the water content to be 0.98 after the injection of polymer of 0.70 PV).
Compared with the water driving, the ultimate recovery rate only grows by
1.15 %. It can be seen that the effects of water driving is significant. Under this
circumstance, the effects of polymer driving are insignificant and may be
uneconomical.
30.4.2 Driving Along the Interbeds
When injection is along the direction of interbeds, the impact of interbeds on the
development of remaining oils can be seen in Fig. 30.5. After the polymer driving,
more distribution of remaining oil at one side of the water well towards the lateral
interbed and the structural top formed by two interbeds is related to the tilting
direction of oil–water interface and hydrodynamic direction. Due to the obstruc-
tion of interbeds, the recovery rate for the water content of 0.925 by water driving
Table 30.1 Relationship between different volumes of polymers and recovery rate
Driving direction Water content of Water content 0.2 PV 0.4 PV 0.6 PV 0.7 PV
0.925 of 0.98
No interbeds 35.51 41.28 41.24 40.78 41.73 42.43
Injection along interbeds 26.00 30.96 35.05 34.18 36.57 37.26
Injection against 23.38 28.57 34.87 35.78 36.74 36.78
interbeds
Injection along the fluvial 34.94 41.11 40.99 38.73 38.59 41.94
direction
30 Effectiveness to Driving by Interbeds in Channel Sand of River Delta System 297
is 26.00 %. The recovery rate for the water content of 0.98 by complete water
driving is 30.96 %. The recovery rate can reach 37.26 % when the polymer driving
is employed. The three values drop 9.51, 10.12 and 5.17 % in comparison with
those without interbeds. It can be seen that due to the obstruction of interbeds, the
recovery rate decreases around 10 % compared with those without interbeds. After
the polymer driving, the obstructed crude oil has been remarkably improved. The
differential value in the crude oil recovery is reduced to 5 % or so.
30.4.3 Driving Against the Interbeds
Figure 30.6 is a cross-sectional view of oil saturation at all stages of injection
against the interbed. Compared with the injection along the interbed, there are
more remaining oil, mainly due to the more surrounding area of oil–water interface
and the interbed at the time of injection against the interbeds. The included angle
formed by the interbed and the top is a ‘‘dead area’’ which is also the remaining oil
enriched region. Due to the obstruction of interbeds, the recovery ratio for water
content of 0.925 with water driving is 23.38 %. The recovery rate is 28.57 % when
the complete water driving with the water content up to 0.98. The recovery rate
can reach 36.78 if the polymer driving is used. The three values drop 11.13, 12.71
and 5.65 % in comparison with those without interbeds. It can be seen that due to
the obstruction of interbeds, the recovery rate decreases around 10 % compared
Fig. 30.5 Cross-section of oil saturation with the water content of 0.8, 0.925 and 0.98 with water
driving and the water content of 0.98 through the follow-up water driving along the interbeds
298 L. Zhao et al.
Fig. 30.6 Cross-section of oil saturation with the water content of 0.8, 0.925 and 0.98 with water
driving and the water content of 0.98 through the follow-up water driving against the interbeds
with those without interbeds. Through the polymer driving, the differences in the
crude oil recovery is reduced by over five percents.
30.4.4 Driving Along the Extension of Interbeds
The injection along the interbed has the least resistance in theory (maximum
permeability along the fluvial direction without direct obstruction of lateral in-
terbed) and excellent water driving effects. Figure 30.7 shows a profile of oil
saturation at all stages of injection in the direction. The remaining oil is mainly
located on the reservoir top near to the oil well when the water content is 0.98 with
the follow-up water driving. On the profile, the remaining oil between two oil
wells is V-shaped. The unobvious obstruction can be seen near some lateral in-
terbed. The recovery rate is up to 34.94 % when the water content is 0.925. The
ultimate recovery rates for the follow-up water driving and polymer driving are
high, reaching respectively 41.11 and 41.94 %. The water injection along the
interbed has higher recovery rates than those without interbeds. But under this
situation the recovery rate improvement is small, only 0.83 % for the polymer
driving effects. Within less than 7 years of polymer driving, the recovery rate can
even be lowered.
30 Effectiveness to Driving by Interbeds in Channel Sand of River Delta System 299
Fig. 30.7 Cross-section of oil saturation with the water content of 0.8, 0.925 and 0.98 with water
driving and the water content of 0.98 through the follow-up water driving along the fluvial
direction
30.5 Conclusion
The distribution location, scale, injection mean, recovery means and reservoir
rhythm of horizontal interbeds have direct bearing on the distribution law of
remaining oil, capable of preventing upward movement of oil and gas while
obstructing the downward movement of water on the interbeds, thereby expanding
the swept volume. It is an important reason for the formation of remaining oils.
The horizontal interbeds can serve to flood in intervals and prevent the interbed
water breakthrough to a certain extent. During the polymer driving, the horizontal
interbeds can change the fluid direction to prevent oil and gas flow, reducing the
displacement effects.
The lateral interbed in the meandering river is a major factor for the formation
of remaining oil. Although the rise of bottom water can not be obstructed, but at
the upper part of lateral interbed, especially the acute angle formed by the interbed
and top surface is the highest concentration of remaining oils. In most cases, the
closer to oil wells, the more remaining oils in the angle.The lateral interbed will
increase the resistance to oil driving and make the polymer injection difficult. At a
certain time of polymer injection, the recovery rate of crude oil at some stages may
be lower than that at the time of complete water driving. In the sand of developed
lateral interbed, different directions of water injection affect the recovery rate of
crude oils and also make the distribution location of remaining oil in the interbeds
300 L. Zhao et al.
different. In general, the injection directions leading to crude oil recovering rate
from high to low are: driving along the extension of interbeds (river direction),
along the lateral interbed and against the lateral interbed.
References
Li M, Zhao Y, Liu X (2009) Distribution of petroleum enriched areas, Changling Sag, southern
Songliao Basin. Pet Explor Dev 36(4):413–418
Ling Z-f, Wang L-j, HU Y-l (2008) Flood pattern optimization of horizontal well injection. Pet
Explor Dev 35(1):85–91
Liu Z, Sa L, Dong S (2010) Current situation and trend of geophysical technology in CNPC. Pet
Explor Dev 37(1):1–10
Su Q (2004) Distribution mode of remaining oil in the positive rhythm sand and development by
horizontal well. Pet Geol Recovery 11(6):34–38
Sun C, ZHU X-m, ZAN G-j (2008) Sand body types and reservoir properties in Cretaceous
Jiufotang Formation in Luxi Sag. Pet Explor Dev 35(5):569–575
Wang X-z, Zeng L-f (2008) Effect of practical techniques in producing remaining oil in Gudong
Oilfield. Pet Explor Dev 35(4):467–475
Wang H, Liu G, Chen G (2009) Development of mm-level mud-films and significance for
reservoirs in Yingtai area, Daqing Oilfield. Pet Explor Dev 36(4):442–447
Yu J, Yang Y, Du J (2010) Sedimentation during the transgression period in Late Triassic
Yanchang Formation, Ordos Basin. Pet Explor Dev 37(2):181–187
Zhang Q-g, Bao Z-d, Song X-m (2008) Hierarchical division and origin of single sand bodies in
Fuyu oil layer, Fuyu Oilfield. Pet Explor Dev 35(2):157–163
Zhang G, Lan Z, Liu P (2009) Fracturing control method for deep volcanic rock gas reservoirs in
Daqing exploration area. Pet Explor Dev 36(4):529–534
Chapter 31
Effectiveness Valuation of Electronic
Countermeasure on Ground Air Defense
and Anti-missile
Yong-ling Yan, Zhi-feng Zhang, Qing-bo Zhang and Tao Dong
Abstract The effectiveness valuation of electronic countermeasure on ground air
defense and anti-missile has turned into one of the hot subjects of present research.
To the complexity and changefulness of the diverse random events deciding the
competency matrix C and the fact that some criterions lack quantified represen-
tation in ADC method, innovation of the ADC method is proposed in the thesis.
Through combination of qualitative and quantitative process, ADC method, ana-
lytic hierarchical process and Delphi method are used jointly to implement the
effectiveness valuation of electronic countermeasure on ground aerial defense and
anti-missile system. Availability A, dependability D, competency matrix C and the
computational models of their sub models are set up respectively. It is proved that
the model is in validity by example.
Keywords Analytic hierarchy process Delphi method Electronic counter-
measure system Effectiveness valuation Improved ADC method
31.1 Introduction
As the electronic technology in air strikes and air defense against are extensive
used in the field, electronic against becomes an important part of the modern war.
As one of the important forces in ground to air defense, we air-defense unit of the
ground will certainly put up drastic rivalry in electronic against conditions. So,
how to evaluate the electronic against effectiveness is an important issue.
Y. Yan (&) Z. Zhang Q. Zhang T. Dong
Department of Equipment, Air Force Engineering University,
Xi’an, Shaanxi, china
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 301
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_31,
Springer-Verlag Berlin Heidelberg 2013
302 Y. Yan et al.
31.2 Methodology
31.2.1 Improving Evaluation Methods
Currently, there are many methods to evaluate the effectiveness. But in all of the
methods the ADC is more comprehensive, precise and its index is more clear
which can reflects weapon system’s physical advantages. There are also limitations
in this method that every index must have a specific expression (AD A 109549
1981). As the ground to air defense electronic against system is complex and lack
quantitative indexes, it is difficult to analyse its C matrix (Meng et al. 2003; Sang
2008; Li and Wang 2008).
So this article talks about how to ameliorate the ADC model, and its main part
uses the improved ADC method to have a strict process and get an authentic
outcome. In the premise of using as much analytical method as possible, calculate
the weight by APH for some uncountable index and find out the point with Experts
consult method to solve the calculation problem. Combining the quality and
quantity can use the ADC to good advantage and can also make up its disad-
vantages, so that we can evaluate electronic against effectiveness effectively.
31.2.2 Building the Evaluation Index System of Ground
to Air Defense Effect in Electronic
Countermeasure Conditions
Combining with improved ADC model elements, the index system as follows: the
evaluation of ground to air defense effect in electronic against conditions is decided
by A, D, C three matrixes (AD A 109549 2010), A and D are decided by mainte-
nance and reliability, C is decided by anti-jamming matrix C1, the electronic
reconnaissance capability matrix C2, the anti-radiation missile resistance capability
matrix C3, the anti-stealth ability matrix C4, the survival ability matrix C5.
31.2.3 Building the Model of Effectiveness Valuation
of Electronic Countermeasure
31.2.3.1 Analyzing the States of Electronic Countermeasure
Though ground to air defense electronic against systems are different in theory,
function and structure, but the typical electronic against process is: Firstly,
reconnaissance equipment such as satellite, radar and photoelectricity equipment
reconnoitre the radiation source. Secondly, the Data-Processing-Center find out its
31 Effectiveness Valuation of Electronic Countermeasure on Ground 303
Fig. 31.1 Reliability chart of
1 6
electronic countermeasure
system 2 4 5 7
3 8
position and radiation source recognition system identify and pick up features such
as the working frequency. Thirdly, our ground to air defense electronic against
systems takes soft killing or hard killing according to the obtained information.
Soft against methods include radar jamming equipment and photoelectricity
jamming equipment. Hard killing methods are launching missiles to against ARM.
According to its typical process, building elementary model can find out sys-
tem’s reliable frame as Fig. 31.1, and it can also discover system’s original state as
Table 31.1.where:
1 -Equipment of radar reconnaissance;
2 -Equipment of photoelectricity reconnaissance;
3 -Equipment of secondary plane reconnaissance;
4 -Center of data processing;
5 -System of radiation sources identifies;
6 -Equipment of radar disturbing;
7 -Equipment of photoelectricity disturbing;
8 -Equipment of hard destroys.
Explanation: the number in the picture shows the serial number of each part.
31.2.3.2 Building Evaluation Model
The basic model of ADC mean is:
Es ¼ AT ½D½C ð31:1Þ
Table 31.1 Work state of Order State
electronic countermeasure
system 1 All of the parts are normal
2 Part 1 conk out, others are normal
3 Part 2 conk out, others are normal
4 Part 3 conk out, others are normal
5 Part 6 conk out, others are normal
6 Part 7 conk out, others are normal
7 Part 8 conk out, others are normal
8 The system conk out
304 Y. Yan et al.
Es —system’s effectiveness vector;
AT —availability vector;
C —competency matrix
D —dependability matrix
(1) the sub models of availability A
Electronic countermeasure system is made up of eight parts. The availability
level of each part can be got by the formula (Yan et al. 2008):
Ai ¼ MTBFi=MTBFi þ MTTRi ði ¼ 1; 2; . . .; 8Þ
where:
i-number in Fig. 31.1.
Combining eight states in Table 31.1, availability A of electronic counter-
measure system can be got by the calculable models of combined system.
A ¼ ða1 a2 a3 a4 a5 a6 a7 a8 Þ
Y
8 Y
8 Y
8 Y
8
a1 ¼ Ai a2 ¼ ð1 A1 Þ Ai a3 ¼ ð1 A2 Þ Ai a4 ¼ ð1 A3 Þ Ai
i¼1 i¼2 i¼1;6¼2 i¼1;6¼3
Y
8 Y
8 Y
7 X
7
a5 ¼ ð1 A6 Þ Ai a6 ¼ ð1 A7 Þ Ai a7 ¼ ð1 A8 Þ Ai a8 ¼ 1 ai
i¼1;6¼6 i¼1;6¼7 i¼1 i¼1
ð31:2Þ
(2) the sub models of dependability D
The factors of D are decided by dependability level. The dependability level’s
expression of each parts in electronic against system is
Ri ¼ expðki tÞ ði ¼ 1; 2; . . .; 8Þ
where:
ki is the parts’ invalidation possibility, and can be got by: ki = 1/MTBFi
State transfer probability d11–d88 can be got by system’s original state and
every part’s dependability. The d11 means the probability that system runs nor-
mally from beginning to the end. The d12 means the probability that system runs
normally at beginning but radar reconnaissance equipment conk out at last. It can
be got by
Y
8
d12 ¼ ð1 R1 Þ Ri
i¼2
In the same way dependability D can be found out by:
dij ; i j
D ¼ DðtÞ ¼ ði; j ¼ 1; 2; . . .; 8Þ ð31:3Þ
0 i[j
31 Effectiveness Valuation of Electronic Countermeasure on Ground 305
(3) the sub models of competency matrix C
Competency matrix C is determined by anti-jamming matrix C1, the electronic
reconnaissance capability matrix C2, the anti-radiation missile resistance capa-
bility matrix C3, the anti-stealth ability matrix C4, the survival ability matrix C5.
a. anti-jamming Capability C1
Anti-jamming capability can be expressed by the change of radar maximum
detection distance in jamming conditions. If there is no jamming, the maximum
detection distance of radar (Schrick 2008) that it can be denoted by Rmax is
" #14
pt G2t k2 r
Rmax ¼ ð31:4Þ
ð4pÞ3 KT0 Dfr Fn LðS=N Þmin
where:
Pt —radio power;
Gt —antenna gain;
k —wavelength;
r —RCS;
K —Boltzmann constant;
T0 —receiver noise temperature, it can be denotation by 290 K;
Dfr —receiver bandwidth;
L —the loss factor of system;
Fn —noise coefficient;
(S/N)min —minimum SNR
Radar maximum detection distance in jamming
conditions (Yan and Zhang 2009) is
rffiffiffiffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0 1 Rj pPt Gt rKDfj
Rmax ¼ ð31:5Þ
3 h0:5 Kcj Pj Gj Dfr
where:
rj —polarization loss, rj = 0.5;
Pj —jamming power;
Gj —interference machine lord disc plus;
Dfj —jamming signal bandwidth.
C1 can be got by it:
0 .
C1 ¼ 1 C1 ¼ 1 Rmax Rmax Rmax 100% ð31:6Þ
b. The electronic reconnaissance capability matrix C2
C2 is mainly decided by anti-signal intercepted capability N1 and anti-signal
analysis capability N2,
306 Y. Yan et al.
where:
ht Df Rt
N1 ¼ 1 K1 P1 K1 ¼ ð31:7Þ
ht0 DF Rto
K1 —reconnaissance relative cover coefficient;
P1 —intercept probability;
ht —actual cover range in the detectable orientation of system;
ht0 —expected or demanded by campaign mission cover range in the detectable
orientation;
Dft —actual cover range in the detectable frequency orientation of system
(Ribeiro 2006);
DFt —expected or demanded by campaign mission cover range in the detectable
frequency;
Rt —actual reconnaissance distance of system;
Rt0 —expected or demanded by campaign mission reconnaissance distance
(Zhou and Tao 2007)
N2 ¼ P2 q ð31:8Þ
where:
P2 —signal processing probability;
q —recognition confidence
we can get:
C2 ¼ x1 N1 þ x2 N2 ð31:9Þ
where:
x1x2 are decided by experts, x1 = 0.43, x2 = 0.57.
c. The anti-radiation missile resistance capability matrix C3 (Liu 2010), the
anti-stealth ability matrix C4
Because of the complexity and changefulness of the diverse random events
deciding the competency matrix C3, C4, and the fact that some criterions lack
quantified representation in ADC method, innovation of the method is proposed in
the thesis. Through combination of qualitative and quantitative process with ADC
method, hierarchical analytic process and Delphi method are used jointly to
implement the effectiveness valuation of electronic countermeasure. And they can
be broken up to index system in Table 31.2 (Ti 2005a).
The weight of each index can be decided by hierarchical analytic process and
every index’s relative importance can be shown by using ratio build judge matrix
1–9. Taking the anti-stealth ability matrix C4 for an example: assuming that the
sub model’s tactics measure and techniques ability constitute the matrix of remark
collection:
31 Effectiveness Valuation of Electronic Countermeasure on Ground 307
Table 31.2 The standard of grade
Index Subindex The standard of grade (point)
0.75–1 0.5–0.75 0.25–0.5 0–0.25
Anti-radiation Concealment [300 200–300 100–200 \100
missile distance (km)
resistance Warning time [60 s 45–60 s 20–45 s \20 s
capability (Wu (Chrazanowski)
et al. 2010) Jamming and Very good Good Not bad Bad
inveiglement
capability
Attack and against Very good Good Not bad Bad
ability
The anti-stealth Meter wave band Most More Sensitive Not
ability (Ti radar sensitive sensitive sensitive
2005b) Anti-stealth Most More Continual Not
drilling continual continual continual
Radar web Very good Good Not bad Bad
Increase [30 20–30 10–20 s \10
aperture
multiplication
of radar power
Increase the [100 80–100 50–80 \50
number of
impulses dealed
with phasic
parameter
1 3:03
T ¼ ðtmn Þ22 ð31:10Þ
0:33 1
The eigenvector of judge matrix can be calculated by ‘‘addition method’’
according to expression (31.10).
1X 2
tmn
xm ¼
2 n¼1 P
2
tkn
k¼1
And the weight can be got as:
x ¼ ðx1 ; x2 ÞT ¼ ð0:66; 0:34ÞT :
Then ten experts mark basing on Table 31.2, and calculate the average figure.
At last it can be got by linearity addition method:
" #
X2 X
mn
C4 ¼ xm xmk Fmk ð31:11Þ
m¼1 k¼1
308 Y. Yan et al.
where:
mn —the number of sub index;
xm —the weight of the third layer index;
xmk —the weight of sub index’s coefficient;
Fmk —the sub index point given by experts.
d. the survival ability C5
The survival ability can be showed by survival probability. In order to improve
survival ability, we usually assume that there are m information centers which are
redundancies of each other. So the survival probability is
!
Ym Ym YQ
C5 ¼ Pcam ¼ 1 pk ¼ 1 pik ð31:12Þ
k¼1 k¼1 i¼0
where:
K —the number of ruined center;
P0k —the probability of raided on the center;
P1k, P2k,…, PQk —the probability of destroying the center after all effective
against measures are taken.
(4) Calculating system’s effectiveness Es
As analyzed above, the formula can be got:
Y
5
C¼ CK ð31:13Þ
k¼1
and formula (31.1)can be changed to
Y
5
ES ¼ AD CK ð31:14Þ
k¼1
The improved ADC method which can evaluate the effectiveness of electronic
countermeasure on ground air defense and anti-missile can be got by joining
formulae (31.2)–(31.6), (31.9), (31.11), (31.12) to formula (31.14).
31.3 The Example
The effectiveness of two supposed typical ground air defense and anti-missile
systems in electronic against conditions which can be evaluated by the model have
been got. System 2 is partly advanced to system 1 by improving reliability level
and radar’s anti-stealth ability of every part in System 2. The numerical value of
each parameter of 2 systems above can be got from figure Tables 31.3 and 31.4.
31 Effectiveness Valuation of Electronic Countermeasure on Ground 309
Table 31.3 Reliability parameter
MTTR1 MTTR2 MTBF1 MTBF2
Radar 25 13 180 300
Photoelectricity equipment 10 10 200 600
Satellite 50 30 300 1470
Date center 30 20 200 980
Radiosource recognition system 21 20 300 500
Radar jamming equipment 26 20 300 410
Photoelectricity jamming equipment 10 7 400 470
Hard killing methods 40 20 180 650
Elucidation: in Table 31.4, the number which follows every index is its weight;
other numbers in Table 31.4 are points.
From Table 31.3:
A1 ¼ ½ 0:4279 0:0595 0:0214 0:0713 0:0371 0:0107 0:0951 0:277
2 3
0:6392 0:0041 0:0720 0:0168 0:0002 0:0008 0:0001 0:2668
6 0 0:6943 0:0028 0:0513 0:0014 0:0003 0:0002 0:2497 7
6 7
6 0 0 0:7101 0:0082 0:0312 0:0051 0:0001 0:2453 7
6 7
6 0 0 0 0:7452 0:0362 0:0014 0:0006 0:2166 7
D1 ¼ 6
6 0
7
6 0 0 0 0:7642 0 0:0012 0:2346 7
7
6 0 0 0 0 0 0:8013 0:0004 0:1983 7
6 7
4 0 0 0 0 0 0 0:8537 0:1463 5
0 0 0 0 0 0 0 1
A2 ¼ ½ 0:7934 0:0012 0:0132 0:0162 0:0387 0:0118 0:0245 0:101
2 3
0:7421 0:0052 0:0810 0:0170 0:0004 0:0009 0:0002 0:1532
6 0 0:7841 0:0032 0:0812 0:0040 0:0006 0:0001 0:1268 7
6 7
6 0 0 0:8021 0:0044 0:0923 0:0061 0:0004 0:0947 7
6 7
6 0 0 0 0:8428 0:0060 0:0998 0 0:0514 7
D2 ¼ 6
6 0
7
6 0 0 0 0:8690 0:0091 0:0760 0:0459 77
6 0 0 0 0 0 0:9215 0:0410 0:0375 7
6 7
4 0 0 0 0 0 0 0:9705 0:0295 5
0 0 0 0 0 0 0 1
From Table 31.4: C41 = 0.664; C42 = 0.814.
To predigest the problem, these parameters of system1 and system 2 such as
anti-jamming C1, the electronic reconnaissance resistance capability C2, anti-
radiation missile resistance capability C3, the survival ability C5 can be supposed
to equal to 1.
So it can be got that: ES1 = 0.426; ES2 = 0.603.
According to the analysis of result, the conclusion can be made that the
effectiveness of electronic countermeasure can be strengthened obviously with
310 Y. Yan et al.
Table 31.4 Points given by experts to the anti-stealth ability of radar
Sub index The third layer index System System
1 2
Tactics measure Meter wave band radar 0.313 0.62 0.84
0.66 Anti-stealth drilling 0.227 0.87 0.94
Radar web 0.460 0.74 0.89
Techniques ability Increase aperture multiplication of radar power 0.461 0.63 0.88
0.34 Increase the number of impulses dealed with phasic 0.75 0.82
parameter 0.327
radar’s anti-stealth ability and the reliability of system2’s improving. The result is
accordant with practice (Yan et al. 2007; Chin 1998; Packer 2003; Whatmore
2005; Hall and Betts 1994; Rius et al. 1993). And it is fully proved that the
improved ADC method is in validity to evaluate the effectiveness of electronic
countermeasure on ground air defense and anti-missile.
31.4 Conclusion
In conclusion, the improved ADC model is used to evaluate the effectiveness of
electronic countermeasure on ground air defense and anti-missile, and it is proved
by example that the model is in validity (Volakis 1994; Zhang et al. 2000; Levison
and Badler 1994; Badler et al. 2006). Currently, the model has been used to
evaluate C4ISR air defense systems, and a synthetically effectiveness valuation
software has been developed. So the improved ADC model is proved to be
worthful.
References
A methodology to find overall system effectiveness in a multicriterion environment using surface
to air missile weapon systems as an example. AD A109549, (1981)
A methodology to find overall system effectiveness in a multicriterion environment using surface
to air missile weapon systems as an example. AD A109549, (2010)
Badler N, Bindiganavale R, Allbeck J, Schuler W, Zhao L, Palmer M (2006) Parameterized
action representation for virtual human agents. In: Cassell J (ed) Embodied conversational
agents. MIT Press, Cambridge, pp 256–284
Chin HH (1998) Intelligent simulation support system: used in electronic warfare simulator.
American Institute of Aeronautics and Astronautics, Washington (AIAA-92-4811) Sep 1998
Chrazanowski EJ (1990) Radar active countermeasure. Artech House, Norwood, MA
Hall DR, Betts FC (1994) An air-to-air situationally interactive combat model (SICM). In:
NAECON, pp 960–967
Levison L, Badler N (1994) How animated agents perform tasks: connecting planning and
manipulation through object-specific reasoning. Department of Computer and Information
Science, University of Pennsylvania, Philadelphia
31 Effectiveness Valuation of Electronic Countermeasure on Ground 311
Li Z, Wang S (2008) Several ways of effectiveness valuation to weapon. J Univ Navy Force Eng
90(1):97–99 (in Chinese)
Liu Y (2010) Mathematics model of effectiveness for electronic recce equipment. J UEST Chin
24(3):304–306 (in Chinese)
Meng Q, Zhang J, Song B (2003) Effectiveness valuation of fish torpedo. Publishing company of
national defence industry, Beijing (in Chinese)
Packer RJ (2003) Computer modelling of advanced radar techniques: the advanced radar
simulator. In: IEE colloquium on electronic warfare systems
Ribeiro RA (2006) Fuzzy multiple attribute decision making: a review and new preference
elicitation techniques. Fuzzy Sets Syst 78(2):155–18
Rius JM et al (1993) High-frequency RCS of complex radar target in real time. IEEE Antennas
Propag 41(9):1308E–1319E
Sang W (2008) The study on ways of effectiveness valuation of electronic countermeasure. Tech
Electron Countermeasure 24:34–41 (in Chinese)
Schrick G (2008) Interception of LPI Radar Singles. In: IEEE international conference on radar
Ti N (2005a) Formulas for measuring radar ECCM capability. IEE Proc, 132(3):198–200, July
2009
Ti N (2005b) Rader ECCMs new area: anti-stealth and anti-ARM. IEEE Trans Aerosp Rev
31(3):1120–1127
Volakis JL (1994) XPATCH: a high-frequency electromagnetic scattering prediction for complex
three dimensional objects. IEEE Antennas Propag Mag 36(1):65–69
Whatmore L (2005) The evaluation of radar effectiveness using the generic ECM ES model. In:
IEE colloquium on computer modeling and simulation of radar systems
Wu X, Hui Zhang, Wang S (2010) The study on technic measure of aerial defence radar to
confront anti-radiation missile. Electron Countermeasure Spacefl 22(1):10–13 (in Chinese)
Yan Y, Zhang Q (2009) The effectiveness evaluation of the C4ISR system based on the improved
ADC method. J Univ Air Force Eng 34(1):32–35 (in Chinese)
Yan Y, Lei Y, Zhang Q (2007) The design of the C4ISR system effect valuation software. Arith
Tech Aviat 37(1):4–7 (in Chinese)
Yan Y, Lei Y, Zhang Q (2008) Research on reliability of the efficiency evaluation of the C4ISR
system based on ADC method. Firepower Command Control 33(1):21–24 (in Chinese)
Zhang Y, Volz R, Sehgal A (2000) Generating Petri net driven graphical simulation tool for
automated systems. In: Proceedings of the American nuclear association 2000 annual
meeting, pp 140–147
Zhou S, Tao B (2007) Mathematical simulation of using decoying and killing missiles to counter
anti-radiation missiles. (1): ADA320857
Chapter 32
Energy Consumption and Economic
Growth: Cointegration and Granger
Causality Test
Jing-wei Liang and Zhou Liu
Abstract In this paper, we use cointegration technique and VEC model to study
the relationship between energy consumption and economic growth in China from
1953 to 2008 and have the following conclusions: 1. Long-term cointegration
relationship exists between energy consumption and GDP growth. 2. In the long-
term, energy consumption and economic growth have no Granger causality; in the
short term, a bi-direction Granger causality between energy consumption and GDP
exits.
Keywords Cointegration Economic growth Energy consumption Granger
test
32.1 Introduction
With the process of industrialization, energy is becoming the most important
material basis for social development and economic growth. Scholars from various
countries are focusing on the interdependence relationship between energy con-
sumption and economic growth. Although the academic community has done
many researches on the issue since the 1970s, an accordant conclusion is never
agreed by the academics. Due to the different samples from different countries and
different statistical methods, a diverse conclusion is presented on the relationship
between energy consumption and economic growth. Therefore, a study on dif-
ferent countries and regions in different time interval of energy consumption and
J. Liang (&)
School of Management, Tianjin University, Tianjin, China
e-mail: [email protected]
Z. Liu
STS Study Center, Tianjin University, Tianjin, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 313
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_32,
Ó Springer-Verlag Berlin Heidelberg 2013
314 J. Liang and Z. Liu
economic growth has an important theoretical significance. Meanwhile, the
implementation of appropriate energy policies which is based on the empirical
research of the relationship between energy consumption and economic growth
has a crucial impact on a country’s economic development; the issue means a
practical significance. In this paper, cointegration model with vector error cor-
rection model (VECM) is used to portray the long-run equilibrium and long-term
and short-term Granger causality between energy consumption and economic
growth in China, relevant policy recommendations is based on the empirical
results of the study.
The rest part of the paper is organized as follows: the second section gives a
literature review of the empirical research between energy consumption and
economic growth; the third section presents an explanation of cointegration theory
and vector error correction model; the fourth section gives the necessary expla-
nation of variables and data acquisition, preliminary data processing and parameter
estimation; the fifth section presents an economic analysis of the empirical results
and policy recommendations.
32.2 Literature Review
Kraft and Kraf (1978) gave the first study on the relationship between energy
consumption and economic growth. They used GNP and energy consumption data
of U.S. from 1947 to 1974, obtained the one-way causal relationship from GNP to
energy consumption. This means that a conservative policy of the energy will not
cause greater impact on the GNP growth. However, Akarca and Long (1980) used
same method as Kraft, with less sample size, cannot draw the same conclusion.
This means that Kraft’s conclusions sound less robust to the impact of sample size.
Since then, much research has focused on using a different sample, different
modeling methods, different measurement methods to explore the inherent rela-
tionship between energy consumption and economic growth. Masih.et al studied
the case of Asian countries (Masih and Masih 1996). They studied eight countries
and regions in Asian of their income to total energy consumption, the results
showed an inconsistent Granger causality among the eight countries. Cheng and
Lai (1997) studied the situation in Taiwan from 1955 to 1993, the results showed
that one-way causal relationship only GDP to energy consumption is accepted.
Ghali and Elsakka (2004) used the Cointegration technique to explain the Cana-
dian situation. They established a neo-classical production function model
including four variables which is output, capital, labor and energy consumption,
the results showed that a long-term co-integration relationship was existence
among these four variables. Lee and Chang (2007) established linear and nonlinear
models in Taiwan using 1955–2003 data, they concluded the existence of a U-
shaped relationship between energy consumption and economic growth in Taiwan,
and the non-linear model can better fitting the relationship between the two. Gross
(2012) applied ARDL technique to study industrial, commercial and transport
32 Energy Consumption and Economic Growth 315
sectors’ energy consumption and growth, the findings showed that a single causal
relationship between growth and energy consumption of the commercial sector; a
two-way causal relationship of growth and energy consumption in the transport
sector. Cointegration methods and Granger test are also widely used in this topic
(Apergis and Payne 2009; Jumbe 2004; Yoo 2006; Yang 2000; Yu and Choi 1985;
Yu and Jin 1992; Yu et al. 1988; Wolde-Rufael 2005).
32.3 Cointegration Theory and Vector Error Correction
Model
According to the characteristics of the data generation process, time series data can
divided into two types: stationary and non-stationary process. For non-stationary
random process, if after d times difference, the stochastic process becomes a
stationary process, and d-1 times the differential is still a non-stationary process,
we call it an I(d) process. For non-stationary time series, there will be a serious
problem known as the ‘‘spurious regression’’ by using the ordinary least squares
method, and the resulting of the estimated value and the parameter are not valid.
Cointegration theory proposed by Granger and Engle can deal with the two non-
stationary time series data for the long-run equilibrium study. Given two I(1) data
series, if a linear combination of the series is I(0),we call the two series
cointegration.
The vector error correction model (VEC) is a VAR model with constraints. It
applies to non-stationary sequence which is known to have co-integration rela-
tionship. When there is a wide range of short-term dynamic fluctuations, VEC
expression of endogenous variables will limit the long-term behavior in order to
make them converge to the long-term cointegration relationship. Due to a series of
partial short-term adjustment can correct the deviation from the long-run equi-
librium, the cointegration term is also known as error correction. The error cor-
rection term can reflect the dynamic mechanism of the long-run equilibrium short-
term fluctuations deviate from the self-correction. This method can separate the
long-term and short-term Granger cause. Because of the error correction term
contains the long-term cointegration relationship, the long-term Granger cause can
be detected by the significant of the error correction coefficient.
Before using the error correction model, a unit root testing is required. Com-
monly used unit root statistic tests are ADF test, KPSS test, PP test. Then, we must
test the exits of the long-term equilibrium relationship between the non-stationary
random variables, namely the existence of cointegration relationship between them.
Granger pointed out that there must be an expression of error correction model
among cointegration I(1) variable. The usual test methods are EG two-step method
and JJ trace statistic method.
316 J. Liang and Z. Liu
32.4 Empirical Research
32.4.1 Cointegration Study
This paper selects the annual GDP and energy consumption data from 1953 to
2008. The real GDP using 1978 constant prices, energy consumption is calculated
by coal consumption calculation. All data are from the New China 60 years sta-
tistics compilation. All variables are transformed into their natural logarithm.
First, the ADF test is applied to detect the unit root of the two sets of data. We
choose the equation test including the intercept and time trend, lag order is
determined by the Akaike Information Criterion (AIC). Table 32.1 shows the test
results:
As shown above, under the 10 % significant level, the P values of both lngdp
and lnec are greater than 10 %, therefore, both cannot reject the existence of a unit
root hypothesis; the first-order differential data can reject the null hypothesis, so
the differential data is I(0) (Table 32.2).
After the unit root test, we began to study the cointegration relationship
between two variables. We use the E-G two-step method to determine the exis-
tence of a cointegration relationship between two variables. The model is set as
Eq. (32.1):
ln gdp ¼ b0 þ b1 ln ec þ ut ð32:1Þ
Using the OLS method,we can get the following equation:
ln gdp ¼ 4:34 þ 1:19 ln ec
ð0:534Þ ð0:049Þ
Using the ADF test to detect the stationary of the residuals of Eq. (32.1), we can
conclude that, under the 10 % confidence level, the residual is I(0).
The residual is I(0) process means that long-term cointegration relationship
between lngdp and lnec exits. Economic growth and energy consumption in the
long-run equilibrium mechanism can be portrayed by Eq. (32.1).
Energy consumption and GDP in the long-term co-integration relationship
described the long-run equilibrium situation between the two. The error correction
model can describe the short-term fluctuations. Short-term fluctuations can be
described by the following equation:
Table 32.1 ADF test
ADF test Lag P-vaule
lngdp -1.750 3 0.674
Dlngdp -4.660 3 0.010
lnec -3.1179 0 0.1238
Dlnec -4.650 0 0.010
32 Energy Consumption and Economic Growth 317
Table 32.2 The residual ADF Lag P-value
ADF test
-3.4198 3 0.0618
D ln gdp ¼ 0:048 þ 0:396D ln ec þ 0:007ecmt1
This equation describes the short-term fluctuations between GDP and energy
consumption. The error correction coefficient is positive, indicating that the
adjustment mechanism of the reverse does not exist; and the error correction term
coefficient is not significant, indicating that in one period, the system cannot
automatically do the amendment process.
32.4.2 VECM and Granger Causality Test
In this section, we establish a bivariate vector error correction model to explore the
long term Granger causality between energy consumption and economic growth.
The model is as follows:
Xm Xm
D ln gdpt ¼ a2 þ d2 ECMt1 þ b D ln gdpti þ
i¼1 2i
c D ln ecti þ e2t
t¼1 2i
Xm Xm
D ln ect ¼ a1 þ d1 ECMt1 þ i¼1
b1i D ln gdpti þ c D ln
t¼1 1i
ecti þ e1t
ECTt ¼ ln gdpt / ln ect
where, a, b, c, u, stands for estimated coefficient, D means the differential oper-
ator, ECT for error correction term, the lag order is determined by the AIC
criterion
By the AIC criterion, the lag order is 2, and the model estimated coefficient is
shown in the Table 32.3. From Table 32.3, we find that only the GDP in the first
order lag is statistically significant, the other lagged variables affect the dependent
variable are not significant. The coefficient of error correction are positive, the
reverse correction mechanism does not exist, and the error correction coefficient is
not significant, indicating that in the long-run equilibrium, there is no Granger
causality from GDP to energy consumption; at the same time, there also does not
exist a causal relationship from energy consumption to GDP. In other words, the
long term, increase or decrease of the energy consumption of GDP does not
explain the increase or decrease.
After exploring the long-term Granger causality, we study the relationship of
short-term Granger. Short-term Granger relationship in the Table 32.4.
As we can seen from Table 32.4 the two hypothesis under 1 % significant level
is rejected. This shows a bi-direction short-term Granger causality between GDP
and energy consumption. The inconsistency between Long-term Granger causes
318 J. Liang and Z. Liu
Table 32.3 Estimated coefficient
a1 a2 b11 b12 b21 b22 c11 c12 c21 c22 d1 d2 /
Cof. 0.036 0.063 0.033 0.250 0.685 0.250 0.306 0.060 0.219 0.045 0.015 0.019 0.797
s.d. 0.030 0.016 0.048 0.227 0.254 0.227 0.286 0.286 0.105 0.150 0.035 0.019 0.007
Table 32.4 Granger causality test
lag Null hypothesis F Value P Value Conclude (1 %sig. level)
2 Energy consumption does not 7.6032 0.001331 Reject
Granger cause GDP
2 GDP does not Granger cause 5.1206 0.009561 Reject
Energy consumption
and short-term Granger cause may be caused by the lack of control variables. In
the long run, the relationship between energy consumption and GDP could be
impact by other important variables (capital, labor, and import and export, etc).
32.5 Conclusions
In this paper, we use cointegration technique and VEC model to study the rela-
tionship between energy consumption and economic growth in China from 1953 to
2008. We have the following conclusions: 1. Long-term cointegration relationship
exists between energy consumption and GDP growth. 2 In the long-term, energy
consumption and economic growth have no Granger causality; in the short term,
there is a bi-direction Granger causality between energy consumption and GDP.
References
Kraft J, Kraf A (1978) On the relationship between energy and GNP. J Energy Dev 3:401–403
Akarca AT, Long TV (1980) On the relationship between energy and GNP: a reexamination.
J Energy Dev 5:326–331
Masih AMM, Masih R (1996) Energy consumption, real income and temporal causality: results
from a multi-country study based on cointegration and error-correction modeling techniques.
J Energy Econ 18(3):165–183
Cheng SB, Lai TW (1997) An investigation of cointegration and causality between energy
consumption and economic activity in Taiwan province of China. Energy Econ 19:435–444
Ghali KH, El-Sakka MIT (2004) Energy use and output growth in Canada: a multivariate
cointegration analysis. Energy Econ 26(2):225–238
Lee C-C, Chang C-P (2007) The impact of energy consumption on economic growth: evidence
from linear and nonlinear models in Taiwan. Energy 32(12):2282–2294
Gross C (2012) Explaining the (non-) causality between energy and economic growth in the
U.S.—a multivariate sectoral analysis. Energy Econ 34(2):489–499
32 Energy Consumption and Economic Growth 319
Apergis N, Payne JE (2009) Energy consumption and economic growth in central America:
evidence from a panel cointegration and error correction model. Energy Econ 31(2):641–647
Jumbe CBL (2004) Cointegration and causality between electricity consumption and GDP:
empirical evidence from Malawi. Energy Econ 26(1):61–68
Yoo SH (2006) The causal relationship between electricity consumption and economic growth in
the ASEAN countries. Energy Policy 34(18):3573–3582
Yang H (2000) A note on the causal relationship between energy and GDP in Taiwan. Energy
Econ 22(3):309–317
Yu E, Choi J (1985) The causal relationship between energy and GNP, an international
comparison. J Energy 10(12):249–272
Yu E, Jin J (1992) Cointegration tests of energy consumption, income and employment. Res
Energy 14(3):259–266
Yu E, Choi P, Choi J (1988) The relationship between energy and employment: a re-examination.
Energy Syst Policy 11:287–295
Wolde-Rufael Y (2005) Energy demand and economic growth: The African experience. J Policy
Model 27(5):891–903
Chapter 33
Estimation of Lead Time in the
RFID-Enabled Real-Time Shopfloor
Production with a Data Mining Model
Ray Y. Zhong, George Q. Huang, Qing-yun Dai and Tao Zhang
Abstract Lead time estimation (LTE) is difficult to carry out, especially within
the RFID-enabled real-time manufacturing shopfloor environment since large
number of factors may greatly affect its precision. This paper proposes a data
mining approach with four steps each of which is equipped with suitable mathe-
matical models to analysis the LTE from a real-life case and then to quantitatively
examine its key impact factors such as processing routine, batching strategy,
scheduling rules and critical parameters of specification. Experiments are carried
out for this purpose and results imply that batching strategy, scheduling rules and
two specification parameters largely influence the LTE, while, processing routine
has less impact in this case.
Keywords Data mining Lead time Radio frequency identification (RFID)
Real-time Shopfloor production
33.1 Introduction
Lead time estimation (LTE) is significant since it exclusively influences customer
relations and shopfloor management practices (Alexander 1980). The direct out-
come of LTE is due data quoting which indicates the commitment to meet
R. Y. Zhong (&) G. Q. Huang
Department of Industrial and Manufacturing Systems Engineering,
The University of Hong Kong, Hong Kong, China
e-mail: [email protected]
Q. Dai
Faculty of Information Engineering, Guangdong University of Technology,
Guangzhou, China
T. Zhang
Huaiji Dengyun Auto Parts (Holding) Co., Ltd., Zhaoqing, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 321
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_33,
Ó Springer-Verlag Berlin Heidelberg 2013
322 R. Y. Zhong et al.
customer orders on time (Chen and Prorok 1983). Shorting lead time is commonly
used in the manufacturing companies to improve their images and future sales
potentials. However, only its shorting is not adequate because customers always
require the accurate and precise estimation of lead time so as to ensure their
production and delivery due date.
There are several challenges when contemplating to carry out LTE. Firstly, large
number of factors such as the process routine, batching strategy, and scheduling
rules affects the lead time. The studies of their influences on the LTE are limited,
not only in qualitative aspects, but also quantitative facets (Jun et al. 2006).
Secondly, it is commonplace that most portion of lead time is spent upon either
waiting in queues and transit. Actual processing time is difficult to estimate due to
the dynamic manufacturing environment and uncertain disturbances. Thirdly, LTE
heavily relies on the statuses of various manufacturing objects such as machines,
operators and materials, whose real-time information is hardly captured. Therefore,
the contingent situations deteriorate its accuracy and effectiveness.
In order to tackle the above challenges, large number of methods has been
proposed to improve the precision, efficiency and effectiveness of LTE (Ozturk
et al. 2006; Ruben and Mahmoodi 2000; Sudiarso and Putranto 2010; Ward 1998).
However, these methods only concentrate on analytical, experimental and heuristic
aspects. Real-life cases are limitedly reported and studied.
This research is motivated by a real-life automotive manufacturer who has
applied RFID technology to support the real-time production on its manufacturing
shop floors over 5 years. Great myriad of data has been captured, representing the
typical manufacturing objects like machines, operators and materials. Due to the
application of RFID technology, shop floor production has been significantly
improved (Dai et al. 2012). The senior manager contemplated to explore the lead
time from such massive RFID-enabled shopfloor data. Thus, a research team
formed by a group of experts from collaborating universities investigated the entire
company and decided to apply data mining for this purpose.
Several research questions are concerned in this paper. The first question is
what aspects can largely affect the LTE and how we can examine their effects on
the estimation. The second question is how we can work out the practical and
precise lead time under the RFID-enabled real-time production ambience.
In order to answer the above questions, this paper proposes a data mining model
for estimating the lead time from the RFID-enabled shopfloor production data. It
includes four steps: data cleansing, data clustering, pattern mining and data inter-
preting, each of which is equipped with suitable mathematical models. The objec-
tives of this paper are to quantitatively examine the key factors which can influence
the LTE as well as to figure out the precise lead time for various product series.
33.2 Related Work
Related work can be categorized into three domains. They are lead time estimation
(LTE), data mining as well as production planning and scheduling.
33 Estimation of Lead Time 323
33.2.1 Lead Time Estimation
LTE is significant in planning shopfloor operations, thus, many research has been
carried out. Three types of LTE procedures are studied so as to figure out the
effectiveness shopfloor information in bottleneck-constrained production system
(Ruben and Mahmoodi 2000). This work indicates the reduction of bottleneck
shiftiness that is based on the excess capacity at non-bottleneck work centers.
A data mining approach with regression tree was proposed for LTE in make-to-
order manufacturing (Ozturk et al. 2006). This research uses a large set of attri-
butes to work out the prediction and then compares with other methods from the
literature. Another approach using a heuristic algorithm was introduced to estimate
the lead time of a complex product development process (Jun et al. 2006). This
paper adopts computational experiments to show the effectiveness and efficiency
of the reduction of makespan of branch-merge types. Some factors influenced the
LTE were examined in a make-to-order company by a mathematical models
(Sudiarso and Putranto 2010). The experiment results imply the usefulness of the
fuzzy approach for estimating the lead time without simulations.
33.2.2 Data Mining
The basic concepts and techniques of data mining were introduced by Han et al.
(2011). Planning and data mining approach was integrated to create better planners
under the background of unmanned and manned space flight (Frank 2007). This
paper reviews the current work in this area and integrates some technologies for
opening research issues. A RFID-enabled data mining model was demonstrated to
collect the data and analyze the data for exhibition industry (Wang et al. 2009).
This model integrates the exhibitor and relational customers by the data mining
model when executing planning. The nature and implications of data mining
techniques in manufacturing and implementations on two fields—product design
and manufacturing were discussed (Wang 2007). That illustrates the methodology
enabled engineers and managers to understand large amount of data by extracting
useful information. A wide and comprehensive review of data mining used in
manufacturing field was investigated in order to reveal the progressive applications
and existing gaps (Choudhary et al. 2009).
33.2.3 Production Planning and Scheduling
Production planning and scheduling (PPS) addresses decisions on the acquisition,
utilization and allocation of manufacturing resources to meet customer require-
ments through most efficient and effective manner (Graves 1999). In theoretical
324 R. Y. Zhong et al.
facet, a multi-period mixed integer linear programming model so as to deal with
the simultaneous planning and scheduling under the back ground of single-stage
multi-product continuous plants with parallel units (Erdirik-Dogan and Grossmann
2008). A bi-level algorithm is proposed to integrate MILP model and iterative
approach within the upper and lower levels. An entire review on PPS was done in
terms of major models, solution strategies, challenges as well as opportunities
(Maravelias and Sung 2009). That reveals the new area of real-time planning and
scheduling through new technologies like Auto-ID such as RFID technology.
33.3 RFID-Enabled Real-Time Shopfloor Production
Environment
The RFID-enabled real-time shopfloor could be regarded as a hybrid flow-shop.
There are several stages, each of which contains multiple machines with same
functionality. Each stage has a job pool that arranges jobs for each machine. Two
concepts are critical and detailed illustrated in our previous work (Dai et al. 2012;
Huang et al. 2009).
Each machine is equipped by a RFID reader and materials are deployed by
RFID tags. Machine operators carry tags as staff cards. Once deployed by RFID,
they become smart objects (SOs) that can sense, reason and interact with each
other. SOs dynamically generates observations associated with the timestamps and
location changes, carrying significant and useful information.
The typical RFID observations include four categories: worker, machine,
material and job. Worker data contains personal and operation information. Per-
sonal information is stored in a staff card. It indicates which machines and what
types of work can be operated by a worker. Operation information involves a
worker’s critical behavior data such as processed jobs, locations, start and finish
time, quality, and quantity. The information implies who did what at what time in
where. Machine data includes general and behavior information. General infor-
mation contains device type, technical parameters, etc. Behavior information
involves its operators, processed jobs, processing time, and performance. The
information represents the activities a machine performed and is significant to
explore machine utilization. Material data comprises property and trajectory
information. Property information includes specification, texture, etc. Trajectory
information embraces location, shippers, in-time and out-time, as well as hand-
over time. It memorizes the movements of various materials and their changes of
statuses. Job data include instructions, standard operation times, times when a job
enters and leaves a job pool, and priority. Job data can be updated at different
stages through downloading from databases.
This company has accumulated over 5 year’s data. We extracted the data from
03/2008 to 12/2010 with a size of 12.2 GB for LTE. Specifically, 16.9 million
pieces of data are picked up. After data cleansing and clustering, 1.1 million
33 Estimation of Lead Time 325
effective and typical pieces are chosen. Further statistics and sampling approach
has been adopted to select a suitable amount of data for specific analysis.
RFID shopfloor data have several characteristics. Firstly, these data are gen-
erated automatically and instantly, thus the volume is super large due to the daily
production operations. Secondly, while the accuracy of current RFID application
in shopfloor production is improving, there are still some errors such as duplicated,
missing and uncompleted data. Thirdly, RFID data involve large number of
information which reflects the practical situations in a RFID-enabled ubiquitous
manufacturing environment. Physically, each piece of RFID data keeps the sta-
tuses of workers, machines and materials within the entire production cycle.
Logically, a set of RFID data imply the production disciplines such as logistics
trajectory, lead time fluctuation and their key impact factors.
33.4 A Data Mining Model for Estimating Lead Time
In order to suit the above characteristics, this section proposes a four-step data mining
approach, each of which is equipped with a suitable mathematical model. The four
steps are data cleaning, data clustering, pattern mining and data interpreting.
33.4.1 Data Cleaning
A piece of RFID-enabled shopfloor data is formatted and expressed as:
d = \BatchID, TaskID, EPC, ProcessID, StartTime, FinishTime, Quantity[. That
means a batch of materials (BatchID) with certain amount (Quantity) from a
customer order (TaskID) is processed by an operator (EPC) in a machine (Pro-
cessID) from StartTime to FinishTime. The lead time is the sum of all the actual
processing time within all the manufacturing stages.
However, there are some noises such as incomplete, missing and duplicated
data. An improved cleansing model is used in this paper which integrates the
adaptive cleaning method with filter rules from (Rao et al. 2006; Jeffery et al.
2006). The purpose of this model is to detect and remove corrupt or inaccurate
records from raw RFID data. The model is simplified as follows:
D ¼ ½di ð33:1Þ
where
di ¼ ½Bi1 ; Ti2 ; EPCi3 ; Pi4 ; Si5 ; Fi6 ; Qi7
with filter rules formulated as a vector function:
Fun ¼ \F1 aij ; F2 ðdi ; diþ1 Þ; F3 ðai5 ; ai6 Þ; F4 ðai7 Þ [ 1 i n; 1 j 6
326 R. Y. Zhong et al.
33.4.2 Data Clustering
The cleansed RFID data are clustered by using support vector machine (SVM)
manner for several reasons. Firstly, SVM adopts supervised learning methods to
analyze data and recognize patterns (Joachims 1999). Different categories are divided
by a clear gap. That is easy to classify the RFID data which may have tinny differ-
ences. Secondly, multiclass SVM enables the classifications from several elements
like product series, machines etc. Meanwhile, the bounds on error of classification
could be avoided by maximizing the margin and the over-fitting could be minimized
by selecting the maximal margin hyperplane in the feature space (Wang et al. 2004).
This paper uses multiclass SVM model to cluster RFID data. The purposes are
to generate various data aggregations by different standards first, and second to
classify the aggregations by different impact factors. The model is formulated as:
D0 ¼ fðxi ; yi Þjxi 2 #p ; yi 2 hgni¼1 ð33:2Þ
33.4.3 Pattern Mining
Least square (LS) is used for establishing the patterns from a set of RFID-enabled
data. LS minimizes the sum of the squares of the errors in every single equation
(Geladi and Kowalski 1986). Specifically, this paper adopts least square polyno-
mial fitting (LSPF) to work out the best fit in the least-square sense that minimizes
the sum of squared residuals. Then, the models from the best fit are selected for
predicting the processing time for different stages. The sum of processing of
different stages is lead time for a specific product series. In addition, the eigen-
functions of key factors are obtained by LSPF with diversified degree of poly-
nomial to examine the effect on the estimates.
The model from LSPF is
X
n
S ð xÞ ¼ ai ui ð xÞ ð33:3Þ
i¼0
S ð xÞ meets the constrain:
X
m
kdk22 ¼ min ½ Sð x i Þ y i 2
Sð xÞ2u
i¼1
33.4.4 Data Interpreting
The above extracted models must be further interpreted for several reasons. Firstly,
these models must be evaluated in a confidence interval to give the assurance that
33 Estimation of Lead Time 327
the data provided by them are correct. Polynomial interpolation and residual
analysis are used for this purpose. Secondly, the raw RFID data may be normalized
by different and heterogeneous methods within different steps. Data interpreting
ensures that the values from the models could be understood in diversified
applications. Finally, the predicted values may be used by different applications
like ERP, APS with standardized format.
The data interpretation model is formulated as:
Y ¼ U ð S ð x Þ Þ ð33:4Þ
where U is an interpretation function that contains a set of functions in a vector to
interpret the values in different applications. For example, to identify the lead time
of a specific product series, U can be specifically expressed as:
yi ¼ S ðxi Þða1 ; a2 ; . . .; an Þðx1 ; x2 ; . . .; xn ÞT
where ai represents an impact factor with its weight xi .
33.5 Experiments and Discussion
33.5.1 Experiment Design
There are several product series which are divided into categories of diesel, pas-
senger, and racing cars etc. The experiment takes S-series used in passenger
vehicles for example. S-series is determined by three key parameters: total length
(TL), head diameter (HD) and stem diameter (SD). The production of such product
follows batch mode. Each batch contains 180 pieces. The lead time is defined as a
batch passes through all the manufacturing stages before shipping to the customer.
Figure 33.1 demonstrates the S-series products with different processes and total
process time. LTE has been based on the times which have been obtained from
past experiences or time study. It is obtained from the sum of each process as
shown in the bottom of Fig. 33.1 (S00: 330 m, S-01: 340 m and S-02: 420 m, m
means minute). However, they are greatly varied and affected by some key factors
when executing real case.
There are three categories in s-series. They are S-00, S-01 and S-02 with 10, 9
and 12 processes (stages). Their specifications are 100*40*8, 100*50*10 and
100*70*12 respectively. Figure 33.2 presents a statistics analysis on the three
categories. The statistics data comes from the RFID-enabled real-time shop floor
data within 10 months which are the peak season of this company. Each point
represents the mean value of a month. The mean values of three categories are
328.7, 345.6 and 423.6 with the standard deviations are 17.6, 20.0 and 16.1. From
the experiences, this company allows 60 % of the standard deviations. Therefore,
the interval of lead time can be worked out with the maximum values as criteria in
peak season, while, minimum values in off-season. The intervals for S-00, S-01
and S-02 in this case are [339.26, 318.14], [357.6, 333.6] and [433.26, 413.94].
328 R. Y. Zhong et al.
S-series
Unit: mm
S-00 S-01 S-02
(100*40*8) (100*50*10) (100*70*12)
1 2 3 4 5 6 7 8 9 10 11 12
Magnetic Detecting
Coarse Grinding
Seal Inspecting
Rough Lathing
Face Welding
Straightening
Quenching
Tempering
Nitridizing
Washing
Packing
Cutting
Total (m)
40
50
70
35
30
50
25
10
10
10
330
50
60
70
35
45
50
10
10
10
340
60
70
70
50
30
60
15
20
15
10
10
10
420
Fig. 33.1 S-series product processes with standard operation time
Fig. 33.2 Deviation of lead time within S-series
33 Estimation of Lead Time 329
33.5.2 Analysis of Key Impact Factors
Lead time is commonly influenced by several factors, especially in the
RFID-enabled real-time shopfloor production due to the quick data intercommu-
nication among different manufacturing units and just-in-time (JIT) production
fashion. The typical impact factors are process routine, batching strategy, sched-
uling rules and product specification. Figure 33.3 reports on the examinations of
these impact factors and their analysis by using the proposed data mining model on
RFID-enabled shopfloor data.
The process routine means a batch of material moves from the beginning of
operation to the end of operating stage. Within these stages, several processes
could be interchangeable. Ten groups of data from three product series with dif-
ferent processes routines are extracted to examine their impacts on LTE. LSPF
with 2nd degree polynomial is used to figure out the characteristics curves of the
process routine. The curves imply the minor impact on lead time in this case.
The batching strategy is examined by LSPF with 2nd degree polynomial. The
characteristics curves of its impact on LTE are shown on the upper right diagram.
With the increasing of quantity in a batch for three product categories, the lead
time greatly increases with the rate of 0.29, 0.31 and 0.19 respectively. It is
observed that the less quantity in a batch causes more batch amount, while, the
processing time increases. But the increase rate is lower than that caused by the
Fig. 33.3 Analysis of different impact factors
330 R. Y. Zhong et al.
quantity increment in a batch. In order to balance the confliction, this company
compromises a decent quantity—180 pieces in a batch.
Five scheduling rules are examined in this paper to evaluate their impacts on
LTE. They are SPT (Shortest Processing Time), MBR (Material-based Rule), OBR
(Order-based Rule), EDD (Earliest Due Date) and FIFO (First in First out). They
are applied for the S-series product for generating their lead time. It is observed
that they greatly influence the LTE with maximum impact rate of 0.14 and 0.13
respectively. The most suitable rules for this case are MBR and OBR which
sequence jobs (batches) by material property and order information.
Two parameters of product specification are examined. They are HD and SD.
The 3-D diagram in Fig. 33.3 demonstrates their impact on the LTE. The lead time
slightly increases when HD 2 [40, 50] and SD 2 [8, 10]. However, when HD 2
(50, ?) and SD 2 (10, ?), the lead time sharply increases. It implies that these
two factors largely influence the lead time when they reach a certain point (e.g. 50
and 10 mm) in this case. It is significant to instruct the estimation of processing
time or lead time when the customer’s requirements meet the situations.
33.6 Conclusion
This paper introduces a data mining model to analyze the lead time estimation
(LTE) in RFID-enabled real-time production environment. A typical product with
three categories has been examined for estimating the lead time. After that, four
key factors are scrutinized to find their impacts on LTE. It is observed that, in this
case, processing routine almost do not affect LTE, while, batching strategy,
scheduling rules as well as head diameter and stem diameter largely influence
LTE. This paper works out the intervals for predicting the lead time and optimal
quantity in a batch as well as quantitative analysis of the above key factors so as to
guide the LTE.
Future research will be carried out from two aspects. Firstly, case-based rea-
soning (CBR) will be concerned to solve new problems by using the meaningful
RFID shopfloor data. Secondly, the data mining model can be extended to explore
the practical processing time, setup time for real-time production planning and
scheduling in the RFID-enabled real-time environment.
Acknowledgments Authors would like to acknowledge National Natural Science Foundation of
China (61074146), International Collaborative Project of Guangdong (gjhz1005), Modern
Information Service Fund 2009 (GDIID2009IS048), Guangdong Department of Science and
Technology Fund (2010B050100023). Special acknowledgements would be given to Key
Laboratory of Internet of Manufacturing Things Technology and Engineering of Development
and Reform Commission of Guangdong Province.
33 Estimation of Lead Time 331
References
Alexander MS (1980) Manufacturing lead time estimation and the implementation of a material
requirements planning system: a computer simulation study. The University of Western
Ontario, Canada
Chen JS, Prorok PC (1983) Lead time estimation in a controlled screening program. Am J
Epidemiol 118(5):740
Choudhary A, Harding J, Tiwari M (2009) Data mining in manufacturing: a review based on the
kind of knowledge. J Intell Manuf 20(5):501–521
Dai QY, Zhong RY et al (2012) Radio frequency identification-enabled real-time manufacturing
execution system: a case study in an automotive part manufacturer. Int J Comput Integr
Manuf 25(1):51–65
Erdirik-Dogan M, Grossmann IE (2008) Simultaneous planning and scheduling of single-stage
multi-product continuous plants with parallel lines. Comput Chem Eng 32(11):2664–2683
Frank J (2007) Using data mining to enhance automated planning and scheduling. In: Proceedings
of IEEE symposium on computational intelligence and data mining, IEEE, March 1–April 5,
pp 251–260
Geladi P, Kowalski BR (1986) Partial least-squares regression: a tutorial. Anal Chim Acta
185:1–17
Graves SC (1999) Manufacturing planning and control. Handbook of applied optimization,
Massachusetts Institute of Technology, MA, pp 728–746
Han J, Kamber M, Pei J (2011) Data mining: concepts and techniques. Morgan Kaufmann, San
Francisco
Huang GQ, Fang J, Lv HL et al (2009) RFID-enabled real-time mass-customized production
planning and scheduling. In: Proceedings of 19th international conference on flexible
automation and intelligent manufacturing, 6–8 July, Teesside, UK
Jeffery SR, Garofalakis M, Franklin MJ (2006) Adaptive cleaning for RFID data streams. In:
Proceedings of the 32nd international conference on very large databases. VLDB Endowment,
pp 163–174
Joachims T (1999) Making large-scale SVM learning practical. MIT Press, Cambridge
Jun HB, Park JY, Suh HW (2006) Lead time estimation method for complex product
development process. Concurr Eng 14(4):313–328
Maravelias CT, Sung C (2009) Integration of production planning and scheduling: overview,
challenges and opportunities. Comput Chem Eng 33(12):1919–1930
Ozturk A, Kayaligil S, Ozdemirel NE (2006) Manufacturing lead time estimation using data
mining. Eur J Oper Res 173(2):683–700
Rao J, Doraiswamy S, Thakkar H, Colby LS (2006) A deferred cleansing method for RFID data
analytics. In: Proceedings of the 32nd international conference on very large databases,
pp 175–186
Ruben RA, Mahmoodi F (2000) Lead time prediction in unbalanced production systems. Int J
Prod Res 38(7):1711–1729
Sudiarso A, Putranto RA (2010) Lead time estimation of a production system using fuzzy logic
approach for various batch sizes. Proc World Congr Eng 3:1–3
Wang K (2007) Applying data mining to manufacturing: the nature and implications. J Intell
Manuf 18(4):487–495
Wang Y, Wong J, Miner A (2004) Anomaly intrusion detection using one class SVM. In:
Proceedings of the 5th annual IEEE SMC information assurance workshop. IEEE, pp 358–364
Wang W, Chang CP, Huang CT, Wang BS (2009) A RFID-enabled with data mining model for
exhibition industry. In: Proceedings of the 6th international conference on service systems and
service management. IEEE, Xiamen, China, 8–10 June, pp 664–668
Ward MN (1998) Diagnosis and short-lead time prediction of summer rainfall in tropical North
Africa at interannual and multidecadal timescales. J Clim 11(12):3167–3191
Chapter 34
Evaluation of Green Residence Using
Integrated Structural Equation Model
with Fuzzy Theory
Michael H. Hu, M. Y. Ku, C. K. Liao and P. Y. Ding
Abstract The feasibility of constructing green residential houses was investigated
by analyzing questionnaires designed to collect relevant information coped with
indices and design characteristics for green residential houses. The questionnaire
answers were analyzed using the SPSS18.0 software to carry out descriptive sta-
tistical analyses, and the SEM with fuzzy theory to perform model suitability
analyses. The objective of these analyses is to evaluate and identify the factors that
practical affect the construction of green residential houses. Results of analyzing
questionnaire answers reveal that as high as 80 % of the general public agrees to
purchase green residential houses. The results obtained in this research indicate
that these characters must be considered as important factors when developing
green residential houses with qualities meeting demands of the general public so
that a win–win situation can be achieved for both developer and consumers.
Keywords Fuzzy theory Green residential houses Residential houses
Structural equation model
34.1 Introduction
Green environmental protection is currently a topic that is concerned the most in
our society. In 1996, the Administrative Yuan established the ‘‘National Sus-
tainable Development Committee’’ and actively promoted the policy of developing
M. H. Hu C. K. Liao
Department of Industrial Engineering and Management, Yuan Ze University, Taoyuan,
Taiwan, China
M. Y. Ku (&) P. Y. Ding
Department of Industrial Engineering, National Chin-Yi University of Technology,
Taichung, Taiwan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 333
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_34,
Springer-Verlag Berlin Heidelberg 2013
334 M. H. Hu et al.
green buildings as proclaimed in ‘‘Building White Paper’’ published earlier in the
same year. Further, the Administrative Yuan approved ‘‘Project to Promote Green
Buildings’’ on March 8, 2001 and implemented the ‘‘Green Building Logo’’ sys-
tem (National Council for Sustainable Development Network). Although started a
little late, these implementations led to obvious accomplishments; Taiwan became
the fourth nation in the world that officially put certification of green buildings into
practice. Therefore, the factors that affect the practicality of constructing green
residential houses are studied in this research so that the difficulties that may be
encountered to promote green residential houses will be identified. The results will
provide valuable recommendations to be referenced by Government for future
promotion of green residential houses in Taiwan.
34.2 Literature Review
In recent years, a green upsurge in construction is on the rising internationally;
variations nations and relevant organizations actively promote the implementation
of green products in daily applications that enable the popularity of green resi-
dential house internally. Our nation also actively promotes green buildings through
rewarding and subsiding the design and construction of green buildings. However,
the percentage of green residential houses on the real estate market is not as high
as expected. In this research, the practicality of implementing green residential
houses will be studied to identify the future trend for promoting green residential
houses that is accepted by both builders and buyers.
34.2.1 Definition of Green Residential Houses
Housing is one of the basic needs to satisfy the six people’s livelihood of eating,
clothing, housing, transportation, education and entertainment; traditional housing
simply provides sheltering. With social and economical advances, the need for
quality residential houses becomes increasingly demanding. In addition to pro-
viding a safe and comfortable sheltering, green residential houses are becoming a
concerned emphasis. The definition of green residential houses is discussed in this
section.
a. Definition of Residence
’’Residence’’ refers to the house that provides either a permanent or temporary
shelter for the occupant to live and perform daily activities. As defined in Wiki-
pedia (Baidu Baike 2011), a residence is an establishment where it was originally
or currently being used by a host as their main place of dwelling or home. This is
used as the definition of ‘‘residence’’ in this research.
34 Evaluation of Green Residence 335
b. Definition of Residential House
Green buildings that recently emerge in Taiwan cover a wide range of build-
ings. Green residential house is a part of green building. Although most scholars
have proposed the definition of ‘‘green residential houses’’, this term has not been
clearly defined. Based on literature information, the ‘‘green residential house’’ is
defined in this research as ‘‘the residential house that is constructed based on the
concept of low carbon with natural ventilation and lighting to reduce energy
requirement and wastewater discharge’’. A green residential house is a house that
can breathe by itself in some sense.
c. Characteristics of Green Residential Houses
A green residential house has the following integrated characteristics:
1. Ecological Consideration: The broad sense of this consideration was proposed
in 1997 by Kuo and Kuo (1997) based on the viewpoint of ecological envi-
ronment. It is based on the overall consideration to target energy source pro-
tection, climate regulation, and resource implementation for accomplishing the
objective of ecological environment. A more specific definition is applying the
characteristics of cultivated plants and green technology to reinforce or
improve the building and foundation environment. The above statements are
integrated to define the ecological consideration of green residential houses as:
the combination of local environment and relevant climate to make a place for
human living using natural design to reduce impacts on environment, and
achieve ecological balance.
2. Energy Saving Consideration: In 1999, the Architecture and Building Research
Institute, Ministry of the Interior (Taiwan) (Lin 1999), proposed four groups of
green building indices; energy savings is an important index group. If a resi-
dential house meets the requirement of a green residential house, it must have
the characteristics of achieving energy saving. Hence, in this research, the green
residential energy saving features include facilities that lead to savings of water
and energy. These features may include solar water heater, and water-saving
apparatus and pipeline design to achieve conservation of water, energy and
other natural resources.
3. Water Reducing Consideration (Taiwan Architecture and Building Center): For
each square meter of reinforced concrete buildings constructed in Taiwan,
about 1.8-kg dust is produced that seriously threatens human health. A medium-
height residential building generates about 0.14 cubic meters of solid wastes
during the construction period; if demolished and removed later, the building
will generate 1.23 cubic meters of more solid wastes. Thus, the ultimate
treatment and disposal facilities are heavily loaded by the large quantity of
construction wastes (Web-site of Taiwan Architecture and Building Center).
Therefore, green residential houses are constructed with natural materials that
are environmentally friendly and recoverable to reduce the waste quantity so
that the impact on environment can be alleviated.
336 M. H. Hu et al.
4. Health Consideration: Lin (2005) proposed that the green buildings is an
ecological, energy-saving, waste-reducing and healthy building indicating that
promoting the health of occupants is an important characteristics of a green
residential house. The quality of living environment affects to a great extent the
physical and mental health of the occupants. Green residential houses apply the
design of natural ventilation and external solar shed to reduce the energy
consumptions on indoor artificial heating, ventilation and air conditioning as
well as lighting while providing a healthy and comfortable living environment.
5. Sustainable Consideration: Kuo and Kuo (1997) proposed that during its life
cycle, a building should provide the optimal and most comfortable and healthy
living environment with the lowest energy consumption, most efficient man-
agement and resource application, and the least damages to environment.
Hence, symbiotic existence of human, building and environment with mutual
benefits can be achieved to accomplish sustainable development. In this
research, the sustainable characteristics of green residential houses is based on
the concept of sustainable development design to alleviate environmental
damages, and to assure that future environmental development is compatible
with the earth so that sustainable development of our living environment is
assured.
34.2.2 Structural Equation Model
Structural Equation Modeling (SEM) is also called the ‘‘Analyses of Covariance
Structure or Linear Structure Equation’’ that is a statistical method for analyzing
case-and-effect results and examining assumptions (Chen 2007).
a. Application of Structural Equation Model (SEM)
Huang (2006) has pointed out that SEM is the most important emerging sta-
tistical method for quantifying information in modern sociology; it has been
extensively applied in the various fields such as management science, psychology
and economics.
b. Chartered Financial Analyst (CFA)
CFA is used for testing validity to examine the significant and structure of latent
variables.
The above introduction of SEM reveals that SEM is a research and analysis
method of statistical analysis technology for dealing with complicate multivariate
information and data. In this research, the CFA available in SEM is used to
develop factors for evaluating green residential houses so that the results are
reliable and valid.
34 Evaluation of Green Residence 337
Fig. 34.1 The membership uA(x)
functions of a triangular
1
fuzzy number
X
c a b
34.2.3 Fuzzy Theory
The Fuzzy Theory that was proposed by Prof. Zadeh in 1965 emphasizes that the
concept of human thinking, inferring and cognitive ability is somewhat fuzzy for
solving uncertain and fuzzy problems encounter in a real world. Chien (2009)
considered that the fuzzy theory can be used for solving decisions not only for
policy making but also for daily life related matters such as fuzzy control of
variable frequency air conditioning unit. In recent years, the fuzzy theory has been
integrated with many other methods including grey relational analyses, TOPSIS
and AHP, among the many others.
Values of the fuzzy semantic variables are also called semantic values; they can
be expressed by defining the base variable of a fuzzy membership function
(Zimmermann 1987), or the membership function can be considered as quantified
attributes of semantic values. The triangular fuzzy number A in the region of real
number R indicates that any x 2 R is designating a number uA ð xÞ 2 ½0; 1, and that:
8
< ðx cÞ=ða cÞ; c x a
>
uA ð xÞ ¼ ðx bÞ=ða bÞ; a x b ð34:1Þ
>
:
0
The triangular fuzzy number can be expressed by A = (c, a, b) with the
function value between c and b of uA ð xÞ, and the maximum membership of 1. The
region between c and b refers to the possible upper and lower boundaries of the
evaluation data. Regions between c and a, and between a and b reflect the fuzz-
iness of the data as shown in Fig. 34.1.
Based on the above statements, the fuzzy theory is applied in this research for
evaluating the degree of importance attached to the various factors of green res-
idential houses. The importance of these factors is prioritized as the standards for
accessing green residential houses.
34.3 Methodology
The objective of this research is to apply the combined structural equation model
(SEM) and fuzzy theory for investing the possibility of developing green resi-
dential houses in Taiwan. Additionally, questionnaires are distributed to residence
living in Taichung (Taiwan) region for understanding their need and degree of
338 M. H. Hu et al.
Fig. 34.2 Research structure Review of Relevant
Literature on Green Factors for Evaluating
Structural Equation Fuzzy Theory
Analyses of Certification Calculation of Weight
Analyses of Degree of Recognition for Green
recognition of green residential houses, and the feasibility of developing green
residential houses in Taiwan.
34.3.1 The Research Structure
Based on the viewpoint of consumers, this research is carried out for investigating
the factors of green residential houses that satisfy the need of general public.
Review of literature leads to the factors that influence the construction of green
residential houses as included in the following research structure (Fig. 34.2):
34.3.2 Research Model
Based on the review of literature on green residential houses, the factors used in
the model for evaluating green residential houses are grouped into three major
dimensions: ‘‘Indoor Environment’’, ‘‘Energy Saving Facilities’’ and ‘‘Community
Environment’’ as illustrated in the follows:
1. Indoor environment: The occupants of the green residential house enjoy a
comfort and leisure life provided by appropriate residential lighting, ventilation
and insulation.
2. Energy saving facilities: The objective of green residential house is achieved by
providing energy saving features such as electricity saving, water saving and
noise insulation.
3. Community environment: The residence of a community can enjoy the green
residential houses only if the community environment and facilities are well
planned such as wastewater treatment facilities, green zone, storm water
drainage, and facilities to recover garbage.
34 Evaluation of Green Residence 339
34.3.3 Fuzzy Theory Procedures
The fuzzy theory that is used for solving uncertain and fuzzy problems encoun-
tered in the real world is implemented in this research using the following
procedures.
Procedure 1: A fuzzy preference order matrix is established to evaluate the
semantic variables expressed by K professionals and experts based on each cri-
terion. After evaluating the m projects (Ai, i = 1,…, m) based on n criteria
(Cj, j = 1,…,n) for k professionals and experts, geometric mean is used to inte-
grate the evaluation results to yield fuzzy preference order matrix as follows:
2 3
~x11 ~x12 ~x1n
6 ~x21 ~x21 ~x21 7
~ ¼6
D 6 .. .. ..
7
.. 7 8i; j ð34:2Þ
4 . . . . 5
~xm1 ~xm2 ~xmn
where: ~xij ¼ aij ; bij ; cij is a triangular fuzz number that represents the fuzzy value
of ith prospective project in the jth criterion. The following equation is used to
normalize the fuzzy matrixbb R ~ ¼ ~rij :
mn
!
aij bij cij
~rij ¼ þ ; þ ; þ ; j 2 B cj ¼ max cij if j 2 B
cj cj cj i
ð34:3Þ
aj aj aj
~rij ¼ ; ; ; j 2 C cj ¼ max aij if i 2 C
aij bij cij i
The five classes of fuzzy semantic conversion scale proposed by Chen and
Hwang (1992) are used to evaluate the importance attached to green residential
houses to convert the semantic expression into fuzzy numbers as shown in
Table 34.1.
Procedure 2: Weighed coefficients: The weight of each evaluation condition can
be calculated; the weight of evaluation attribute is wi ¼ ðw1 ; w2 ; . . .; wm Þ. In this
research, a simpler gravity method developed by Yager (1980) is used to perform
fuzzy sorting. Results of equation derivation show that the sorting of centroid
fuzzy values for the fuzzy triangular function can be expressed by Eq. (34.4). The
Table 34.1 Correspondence of fuzzy semantic expressions to fuzzy number
Evaluation scale Fuzzy semantic expression Fuzzy number
1 Significantly not important (0,0,0.3)
2 Not important (0,0.25,0.5)
3 Neutral (0.3,0.5,0.7)
4 Important (0.5,0.75,1)
5 Significantly important (0.7,1,1)
Source of Information (Chen and Hwang 1992)
340 M. H. Hu et al.
opinions expressed by experts are integrated using triangular fuzzy numbers
Buckley (1984), and Eq. (34.5) is used to calculate the geometric mean of the
results. The sorting of the centroid fuzzy for the integrated fuzzy triangular
function is expressed as follows:
1
~ ¼ ðlA þ mA þ uA Þ
R A ð34:4Þ
3
!ð1Þ
Y
n n
Wi ¼ ~
R A i ¼ 1; 2; . . .; n ð34:5Þ
i¼1
Procedure 3: Preference Aggregation: The algorithm using the two triangular
fuzzy numbers calculation, ã1 = (a1, a2, a3) and~b ¼ ðb1 ; b2 ; b3 Þ as proposed by
Zadeh.
Procedure 4: Normalized weighted fuzzy matrix is developed:
~ ¼ b~vij cm n ; i ¼ 1; 2; . . .; m; j ¼ 1; 2; . . .; n
V ð34:6Þ
where: ~vij ¼ ~rij W ~ ij with value between [0,1]; W
~ ij is the weighted fuzzy value of
the jth criterion.
Procedure 4: Defuzzification: There are many method of defuzzification. In this
research, the method proposed by Chen and Hsieh (1999) is used by assigning
Ai ¼ ðai ; bi ; ci Þ, i = 1,2,…,m, as m triangular fuzzy numbers. R(Ai) indicates the
representative value of triangular fuzzy number for the ith person being evaluated.
The equation is:
ai þ 4bi þ ci
Rð Ai Þ ¼ ð34:7Þ
6
34.4 Analyses
Through reviewing literature on green residential houses, relevant residential
factors are summed up. After analyzing the information collected using ques-
tionnaire and studying the sample structural analyses, confirmation factor analyses,
and fuzzy theory, the factors that can be implemented and the items that the
general public prefers when purchasing green residential houses can be found. The
questionnaire results indicate that about 80 % of the general public has the willing
to purchase green residential houses but wishes to pay not more than 10 % higher
than a regular house with 5 % higher cost being the most acceptable.
34 Evaluation of Green Residence 341
34.4.1 Data Analyses
The SPSS 18.0 statistical software is used in this research for carrying out sta-
tistical analyses. The statistical methods include descriptive statistical analyses,
reliability analyses and the various examinations. The overall model analyses are
performed using AMOS 17.0, and finally the fuzzy theory is applied for find out
the practicality of green residential houses and the order of importance for factors
considered by the general public for purchasing green residential houses.
34.4.2 Factors Important to the Buyers of Green Residential
Houses
The results obtained by perform the two-stage five-factor examinations are used to
confirm the validity of the three assumptions proposed in this research. The
explanations are provided in the following paragraphs:
1. Constructing green residential houses is an external latent variable that is
reflected by the three internal latent variables, i.e. indoor environment, energy
saving facilities and community environment. All coefficients obtained by
conducting CFA examinations show are greater than 0.7 confirming that green
residential houses can be evaluated using these three dimensions.
2. Indoor environment is an internal latent variable that is reflected by observing
three variables.
3. Energy saving is an internal latent variable that is reflected by observing three
variables.
4. Community environment is an internal latent variable that is reflected by
observing four variables.
34.4.3 Fuzzy Theory
The fussy theory and the centroid method are used to determine relevant weighted
value for investigating the factors important to constructing green residential
houses in the Greater Taichung (Taiwan) region. Procedures for conducting the
fuzzy theory are as follows:
Procedure 1: Standards for evaluation are established. In this research, the
evaluation factor in ‘‘Dimensions for constructing green residential houses’’ are
used to carry out the evaluation of factors affecting green residential houses as
shown in Table 34.2.
Procedure 2: After the original matrix is developed, Eq. (34.5) can be used to
evaluation the weight of attributes by calculating the fuzzy value of the various
evaluation dimensions as shown in Table 34.3. The geometric average is then
342 M. H. Hu et al.
Table 34.2 Factors for evaluating green residential house
Evaluation dimensions Evaluation factors
C1 Indoor environment A1 Residential lighting
A2 Residential ventilation
A3 Residential insulation
C2 Energy-saving facilities B1 Electricity saving facilities
B2 Water saving facilities
B3 Noise insulation
C3 Community environment D1 Wastewater treatment facilities
D2 Green area
D3 Drainage
D4 Facilities for recovering garbage
Table 34.3 Overall fuzzy evaluation results
Evaluation factors Fuzzy number
A1 Residential lighting 0.34 0.57 0.78
A2 Residential ventilation 0.31 0.54 0.75
A3 Residential insulation 0.30 0.53 0.74
B1 Electricity saving facilities 0.34 0.57 0.78
B2 Water saving facilities 0.36 0.59 0.79
B3 Noise insulation 0.38 0.61 0.81
D1 Wastewater treatment facilities 0.36 0.60 0.81
D2 Green area 0.29 0.52 0.73
D3 Drainage 0.32 0.55 0.76
D4 Facilities for recovering garbage 0.34 0.58 0.78
Table 34.4 Weight and ordering of green residential housing
Evaluation factors Weight Ordering
A1 Residential lighting 0.535 5
A2 Residential ventilation 0.500 8
A3 Residential insulation 0.490 9
B1 Electricity saving facilities 0.533 6
B2 Water saving facilities 0.550 3
B3 Noise insulation 0.571 1
D1 Wastewater treatment facilities 0.560 2
D2 Green area 0.482 10
D3 Drainage 0.513 7
D4 Facilities for recovering garbage 0.537 4
calculated to obtain the weight value and order of magnitude for various factors
Table 34.4.
Procedure 3: Eq. (34.7) is used to defuzzify, calculate and sort. Order of factor
important to the construction of green residential houses is shown in Table 34.5.
34 Evaluation of Green Residence 343
Table 34.5 Results of fuzzy analyses on green residential houses
Evaluation factors Fuzzy solutions Ordering
A1 Residential lighting 0.569 1
A2 Residential ventilation 0.537 2
A3 Residential insulation 0.527 3
B1 Electricity saving facilities 0.567 3
B2 Water saving facilities 0.583 2
B3 Noise insulation 0.604 1
D1 Wasterwater treatment facilities 0.593 1
D2 Green area 0.517 4
D3 Drainage 0.548 3
D4 Facilities for recovering garbage 0.572 2
Data in the above table reveal that the need of a green residential house with
energy saving feather by the general public is relatively high. As the indoor
environment is concerned, the residential lighting, which is related to energy
savings, is also emphasized by the general public indicating that the general public
has already acknowledged the importance of energy resources. Hence, further
planning of green residential houses needs to emphasize the energy saving func-
tion in order to satisfy the need of consumers.
34.5 Conclusions
Results of investigation and analyses obtained in this research show that more than
80 % of the general public in Taiwan accepts green residential houses so that
developing green residential houses is feasible. The general public also expresses
that the green residential house costs less than 10 % more than a regular residential
house so that 10 % cost difference is accepted. Additionally, developing green
residential houses needs the cooperative efforts of government, developer and
consumer. The government has to promote programs for dissimilating information
on green residential houses, and encourage developers to construct more green
residential houses through a rewarding system so that the general public is more
willing to purchase green residential houses. Effective promotion of green resi-
dential houses will enable developer to base on specifications of green residential
houses for selecting appropriate materials and methods to construct residential
houses that are healthy and comfortable with low pollution and reduced energy
consumption. The concept and advantages of green residential houses are dis-
similated to the general public so that when purchasing residential houses, con-
sumers consider green residential houses as the primary choice, and are willing to
pay a little more for green residential houses. The idea of green residential houses
can thus be implemented in Taiwan.
344 M. H. Hu et al.
References
Baidu Baike (2011) Green Residential House. http://baike.baidu.com/view/21130.htm/
Buckley JJ (1984) The multiple judge, multiple criteria ranking problem: a fuzzy set approach.
Fuzzy Sets Syst 13(1):25–37
Chen SY (2007) The AMOS operation of structural equation model, 1st edn. San Ming Book
Company Ltd., Taipei
Chen SH, Hsieh CH (1999) Ranking generalized fuzzy number with Graded Mean Integration
Representation. In: Proceeding of eighth international fuzzy systems association word
congress (IFSA’99 Taiwan), vol 2, pp 551–555, Taiwan, Republic of China
Chen SJ, Hwang CL (1992) Fuzzy multiple attribute decision making methods and applications.
Springer, New York
Chien CF (2009) Decision analyses and management. Yeh Yeh Book Gallery, Taipei
Huang FM (2006) Social science statistical methods—structural equation model, 1st edn. Wu
Nan Book Company, Ltd., Taipei
Kuo CY, Kuo GR (1997) Sustainable township and village—The Spring of a Mountain Township
by the Side of Dongshi Creek, Panning of Dongshi Mountain Township. Workshop on the
concept and strategy for sustainable township and village development in Taipei. Architecture
and Building Research Institute, Ministry of the Interior, Taiwan and Chi-Hsin Agricultural
Development Foundation
Lin HT (1999) Four indicators of the green building. Architecture and Building Research
Institute, Ministry of the Interior, Taiwan
Lin HT (2005) Technical manual for design of green buildings. Architecture and Building
Research Institute, Ministry of the Interior, Taiwan
National Council for Sustainable Development Network. http://sta.epa.gov.tw/nsdn/
Taiwan Architecture and Building Center. ww.tabc.org.tw
Yager RR (1980) On a general class of fuzzy connectives. Fuzzy Sets Syst 4:235–242
Zadeh LA (1965) Fuzzy set. Inf Control 8(3):338–353
Zimmermann HJ (1987) Fuzzy sets, decision making, and expert systems. Kluwer Academic
Publishers, Boston
Chapter 35
Evaluation of Population Age Structure
Model Using Grey Clustering Theory
Bao-ping Chen
Abstract Due to the differences in economic development level and fertility
policy enforcement, the population age structures are quite diverse from different
regions in China. Thirty one provinces and Autonomous regions are used as
clustering objects. Population from various age groups and Total dependency ratio
are chosen as clustering indexes and whitening weight functions from corre-
sponding indexes are defined. Meanwhile, clustering models of the population age
structure assessment are provided and corresponding algorithms are compiled.
Thirty one regions are divided into three groups by the age structure: excellent,
normal, and poor. Then, sort again within each group to further analyze the rea-
sons. The conclusion from our research provided a systematic, objective and
accurate evaluation of age structure in various regions, it may has some reference
value for solving China’s population problems.
Keywords Cluster evaluating Evaluation Grey theory Grey clustering
Whitenization weight function Population age structure
35.1 Introduction
After several years of development, the grey system theory has centered on gray
model as the core of the model system, which is built on gray equation, gray
matrix, gray algebra system, etc., relying on the grey relation space analysis (Liu
and Dang 2009). Grey clustering is a branch of gray system theory, which can be
employed in the multivariate statistical analysis. Such method has been used to put
the samples together in a multi-dimensional space and separate these samples with
B. Chen (&)
Department of Computer Information and Management,
NeiMongol University of Finance and Economics, Hohhot, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 345
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_35,
Springer-Verlag Berlin Heidelberg 2013
346 B. Chen
correlated measurement from each other. Edge on dealing with small sample
problems, the grey clustering is one of the most applicable subjects in recent years.
The population problem is one of the key elements of China’s social and
economic development. Demographic statistics and the real world case reminds us
that if the growth of population is out of control, the middle aged generation will
make up the big foundation of pyramid-like age structure; if in the ‘‘only one
child’’ policy, the senior generation will take place of the middle aged people with
a mushroom-like age structure. After 1978, an aging population issue becomes
unavoidable in China. Therefore, the age structure dilemma is not only a society
problem, but a serious economic problem. The situation does not allow of any
delay to deal with the current situation. There are a lot of useful researches of how
age structures influence the growth of economy and CPI, and studies of different
age structure among the regions (Dang et al. 2009). The examples of Liang Zhang,
Elsie adjusting Leslie’s Model of Population age structure (Zhang et al. 2011),
Xiulin, Gen’s The analysis about population age structure based on composition
statistics (Gen 2011), Show ever, in all of these research papers, only few of them
use grey model to demonstrate such situation.
The author apply the grey model of midpoint trigonometrical whitenization
weight function to assess the composition of population age with its bring-up ratios
in 2010 (all original data comes from China Statistical Yearbook 2010) and cat-
egory the age structure into three regions. After the sorting, the author also ana-
lyzes which regions have the overwhelming age structure among the other two
regions. The research shows that the grey model can easily assess the age structure
of the population and also illustrate the aging degree of each area more objectively
and accurately than other models.
35.2 Grey Clustering Theory and Model
Grey systems theory is an important ingredient in the development of information
processing. Grey systems-based techniques are powerful tools in addressing those
systems in which information is partially known and partially unknown. According
to clustering object, Grey clustering can be divided into grey relation clustering
and grey whitenization weight function clustering. Grey relation clustering is
mainly used for the same incorporated factors to make complex system simplified.
Gray whitenization weight function clustering is established through the grey
Numbers whitenization weight function, which gets together all the different
clustering objects to sort each other with same whitenization weight and finally
judge which type those clustering are (Zhang G-h et al. 2011; Wang et al. 2011).
Gray whitenization weight function is a function which uses a quantitative
description method to formulate the degree of grey of the grey cluster and then decide
which type those clustering are. Whitenization weight function can be assured in real
situation. When dealing with the real world cases, it can be either determined by the
involved objects or obtained through the whole picture to determine all the similar
35 Evaluation of Population Age Structure Model 347
sample objects of whitenization weight function. According to the Chinese age data,
the author chose center point trigonometrical whitenization weight function of the
grey evaluation model to make it a reasonable classification.
35.2.1 Center Point Triangle Whitenization Weight Function
of Gray Evaluation Model
Assume there are N objects, m assessment index, s different gray clusters, for
object I of index J sample observation value is Xij, i = 1,2,3,…, n, j = 1,2,3,…,
m, please make diagonal assessment according to the Xij value for the corre-
sponding object I. During sorting different grey cluster, we will determine the
maximum of grey level as ‘‘gray center point’’. The following is the process of
calculation:
Step1: Accordance with the evaluation requirements of grey number s, I define
each grey cluster as 1,2,…, s, each of their center point as k 1, k 2,…k s, in which of
them maybe any kind of the target grey central point (not the center point neces-
sarily). I also define the domain of each index as S correspondingly, which are k 1, k
2,…k s.
Step2: Make connection (kk,1)correspondingly with(kk-1,0)(kk+1,0) and get K
gray triangle whitenization weight function to index J :f kj ðXij Þ ðj ¼ 1; 2; . . .; m;
k ¼ 1; 2; . . .; sÞ. As the measured value X of index J, can be calculated each grey
type (k = 1, 2,…, s) and its grey level f kj ðXij Þ through the following function:
8
>
> 0; x 62 ½kk1 ; kkþ1
>
>
>
< x kk1
; x 2 ðkk1 ; kk Þ
fjk ð xÞ ¼ kk kk1 ð35:1Þ
>
>
>
> kkþ1 x
>
: ; x 2 ðkk ; kkþ1 Þ
kkþ1 kk
Step3: Calculate the comprehensive clustering coefficient rki (i = 1,2,…,m;
k = 1,2,…,s) from the objects I of their grey cluster K.
X
m
rki ¼ f kj Xij gj ð35:2Þ
j¼1
Note: f kj ðXij Þ is the subclass of Whitenization weight function of index J and gj
(j = 1,2,3,……,n) is the weight of Comprehensive clustering of index J.
Step4: Determine the level of the target sample object. If max {rkj } = rkj , and
then we can reckons that objects I belongs to K*, we can get the further results of
the class of each objects and order according to their corresponding comprehensive
clustering coefficient value.
348 B. Chen
35.2.2 Advantages of Center Point Triangle Whitenization
Weight Function
Comparing with other grey clustering methods, the center point triangle white-
nization weight function is based on the most likely point kk as the center point of
the level, so such kind of method can easily calculate center point triangle whit-
enization weight function of each grey clustering through k0,k1,k2,…ks,ks ? 1.
People can be more familiar with how to accurately get the center point of grey
cluster than that of the intervals, therefore, any research conclusion based on such
method might be more reliable and accessible.
35.3 Assessment of Real Case
When UN do population census, they usually begin with the age of 65. Generally,
there are three categories by age: below 14, 14–65, 65 above (65 inclusive) (Zhang
and Lei 2011). There is a fact we cannot ignore that china is a large country, so
there are total different economic development among area, especially the east and
west part of China (Bao 2012; Kang 2009; Ding 2012). Each area, which of the age
structure is changing from middle adult to senior citizen and this gap expends
gradually. The following is that the author applied the grey model of midpoint
trigonometrically whitenization weight function to according to assess the com-
position of population age with its bring-up ratios in 2010 and category the age
structure into three regions, all the data come from China Statistics Yearbook
2010.
35.3.1 Evaluation Model
Considering the different economic development among each area in China and
the research result more reliable, the author will divide the 31 provinces, cities and
autonomous regions into three parts: east, middle and West. East: Beijing, Tianjin,
Hebei, Liaoning, shanghai, Jiangsu, Zhejiang, Fujian, Shandong and Hainan;
Middle: Shanxi, NMG, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei and
Hunan; West: Sichuan, Guizhou, Yunnan, Guangxi, Shanxi, Gansu, Qinghai,
Ningxia and Xinjiang. Age Data collected from all regions is below 14, 15–64 and
65 above (65 inclusive) and the number of population in each area. In addition, the
author also adds young dependency rate and old dependency rate of each part of
China. We can get the weight of age of below 14 years old through dividing the
total amount of population of age below 14 years old by the total sample amount
of population; similarly, we also get each weight of the age 15–64 and 65 above
35 Evaluation of Population Age Structure Model 349
using the same method. Table 35.1 shows each weight of the three different age
group, young dependency ratio and senior dependency ratio of the 31 cities.
The following is the process of evaluation:
Step1: as the sample matrix X:31 areas are the clustering objects (i = 1,2,…, 31),
each age group weight from each region, as well as their dependency ratio corre-
spondingly of each clustering index as J (j = 1,2,3,4,5).
Step2: divided each region into three categories: excellent, normal, poor, which
means grey number S is 3, calculate each gray center point based on its mean,
maximum and minimum value. For example, the age group of below 14 year old,
its level center point of excellent level is 20, medium level is 15, and poor level is
seven.
Table 35.1 The population structure parameters in 31 regions
City The ratio of The ratio of The ratio Young Senior
0–14 15–64 of 65 dependency ratio dependency ratio
Beijing 8.62 82.66 8.72 10.41 10.54
Tianjin 9.81 81.68 8.50 11.99 10.43
Hebei 16.83 74.93 8.24 22.46 10.99
Shanxi 17.11 75.30 7.59 22.70 10.06
Neimenggu 14.08 78.35 7.57 17.99 9.65
Liaoning 11.43 78.26 10.31 14.59 13.17
Jilin 11.98 79.64 8.38 15.06 10.53
Heilongjiang 11.96 79.72 8.33 15.00 10.44
Shanghai 8.64 81.23 10.12 10.62 12.46
Jiangsu 13.01 76.10 10.89 17.09 14.31
Zhejiang 13.21 77.46 9.33 17.05 12.05
Anhui 17.98 71.83 10.18 25.03 14.17
Fujian 15.48 76.63 7.89 20.18 10.30
Jiangxi 21.88 70.52 7.61 31.02 10.77
Shandong 15.73 74.42 9.84 21.15 13.23
Henan 21.01 70.63 8.36 29.73 11.83
Hubei 13.91 77.01 9.08 18.07 11.81
Hunan 17.62 72.61 9.77 24.27 13.47
Guangdong 16.89 76.36 6.75 22.12 8.84
Guangxi 21.70 69.06 9.23 31.44 13.38
Hainan 19.95 72.20 7.84 27.71 10.80
Chongqing 16.98 71.47 11.54 23.76 16.17
Sichuan 16.96 72.08 10.95 23.54 15.19
Guizhou 25.21 66.22 8.58 38.10 12.94
Yunnan 20.73 71.63 7.64 28.93 10.65
Xizang 24.33 70.67 5.00 34.55 7.22
Shanxi 14.71 76.77 8.52 19.16 11.11
Gansu 18.14 73.61 8.25 24.67 11.19
Qinghai 20.96 72.82 6.22 28.75 8.66
Ningxia 21.43 72.22 6.35 29.79 8.89
Xinjiang 20.77 73.04 6.19 28.44 8.47
350 B. Chen
Step3: set up each whitenization weight function respectively in excellent,
normal and poor level
as f kj ðXij Þ (j = 1, 2; ; m, k = 1, 2; ; s) :
f11 ½20; 26; ; 100; f12 ½15; 20; ; 26; f13 ½5; 15; ;
f21 ½78; 83; ; 100; f22 ½73; 78; ; 83; f23 ½50; 73; ; 78
f31 ½; ; 5; 7; f32 ½; ; 7; 9; f33 ½; ; 9; 14
f41 ½25; 39; ; 100; f42 ½21; 25; ; 39; f43 ½6; 21; ; 25
f51 ½; ; 7; 10; f52 ½; ; 10; 13; f53 ½; ; 13; 17
Step4: calculate the comprehensive clustering coefficient rkj from the objects
i(i = 1, 2, …, 31) of their grey cluster k(k = 1, 2, 3). For example: rkj of Bei-
jing = {0.0937, 0.3728, 0.0000, 0.1137, 0.1313}.
Step5: Calculate grey cluster of each region using max {rkj } = rkj .
For example: max {rkj } of Beijing = max{0.0937, 0.3728, 0.0000, 0.1137,
0.1313} = 0.3728
Step6: Based on the value of rkj of each region and get the further sorting of
each grey type, we can drive the conclusions as follows:
35.3.2 Analysis of the Results
From the 1970s, China has implemented the one-child policy, but the enforcement
of this policy differs regionally. Due to the less developed economy and lack of
prospection in central and western parts of the country, this policy is not strictly
carried out, especially in the western parts of China, which results in high birth rate
there because of relatively more tolerant policy environment, This situation result
in the varying degree of aging in different part of China. During the research, the
author found the assessment result was same as the real world case: Age structure
in Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia and
Xinjiang have changed slowly, so these provinces haven’t been in the stage of
aging yet and have relatively more reasonable demographic structure than any
other area. Some metropolis like Beijing, Tianjin and Shanghai, there is a large
number of Talent Introduction under the age of 65, the demographic structure is
still at a rational level in spite of some difference from Sichuan province. But in
Liaoning, Heilongjiang, Shandong, Guangdong Province, and some parts of the
western area, there are larger elderly population cardinal number and more speedy
of aging, therefore, they have most prominent Irrationality in demography
Table 35.2.
35 Evaluation of Population Age Structure Model 351
Table 35.2 The population structure clustering results in 31 regions
Sorting Provinces or cities(from excellent to poor)
results
Excellent Hainan, Chongqing, Sichuan, Guizhou, Yunnan, Xizang, Shanxi, Gansu, Qinghai,
Ningxia, Xinjiang
Normal Beijing, Tianjin, Hebei, Shanxi, Shanghai
Poor Neimemggu, Liaoning, Jilin, Heilongjiang, Jiangsu, Zhejiang, Anhui, Fujian,
Jiangxi, Shandong, Henan, Hubei, Hunan, Guangdong, Guangxi
35.4 Discussion
As an appraisal and decision approach for Uncertainty Problem, grey clustering
method is widely used in different area (Yuan and Liu 2007). We defined the
Whitening weight function of Corresponding indicators based on the data from
China Statistical Yearbook 2010, and build the Grey Clustering Model of the
population age structure. As the result of the evaluation, it combined all the
impacts from each of the sample data, much more close to the original properties
of the subject being evaluated. For a country or a region, structure and changes of
population age are the performance of the quality of the total population. It’s not
only related to the development of the population, but also the development of
society and economy. Demographic theory reveals the age structure of the dif-
ferent types of the population and indicates different population reproduction
scale, speed and development trend, which highlights various social economics
and population problem. Just because of this, people pay much more attention of
the population statistics and the composition of age structure of population. This
paper strives to achieve objectivity and accuracy to evaluate of the age structure of
each area of China and try to overcome any kind of human factors. This paper also
is very helpful to be provided as a scientific reference to any interested scholars
and researchers to solve the population problem.
Acknowledgments Project supported by the Science Research Projects of the Colleges and
Universities in Inner Mongolia (NO.NJZY11106) and the Natural Science Foundation Project in
Inner Mongolia (NO.2010 MS1007).
References
Bao Y-x (2012) An analysis of the effect of population ageing on regional economic development
basing on the model of neoclassical economic growth. Popul Econ (1):1–7
Dang Y, Liu S, Wang Z (2009) Model on gray prediction and decision model. Science Press,
Beijing
Ding D (2012) Continuous dependence on the parameters of self-similar measures. J Hubei Univ
(Nat Sci) 34(1):26–30
Gen X-l (2011) The analyses about population age structure based on composition statistics.
J Appl Stat Manag (1):118–126
352 B. Chen
Kang J-y (2009) Impact of change of population age structure on consumption in China. Popul
Econ (2):60–64
Liu S, Dang YG (2009) Grey system on dealing with the theory and practical applications. Social
Sciences Edition, Beijing
Wang J, Huang K, Wang H, Li Y (2011) A hierarchy cluster method for functional data. J Beijing
Univ Aeronaut Astronaut (Soc Sci edn) (1):86–102
Yuan C, Liu S (2007) A grey degree based on grey whitening weight function. J Jiangnan Univ
(Natl Sci Edn) 6(4):494–496
Zhang G-h, Wu Y, Zhang M (2011) Enterprise emergency management capacity comprehensive
evaluation based on gray cluster analysis. J Quant Econ (3):94–99
Zhang L, Lei L-h (2011) An analysis of the relation of the regional population age structure and
the household consumption in China. Popul Econ (1):16–21
Zhang L, Shi S-l, Dou C-Y (2011) Adjusting Leslie’s model of population structure. Coll Math
(8):99–102
Chapter 36
Evaluation of Recycle Level of Qaidam
Salt Lake Circular Economy
with Intuitionistic Fuzzy Entropy
Measures
Fei Feng, Yu Liu, Jian Zhang, Nan Wang and Xiao-hui Xing
Abstract The purpose of this paper is to build the evaluation index system of
Qaidam salt lake circular economy and sorting the programs according to the
weights with the intuitionistic fuzzy sets conception. Intuitionistic fuzzy sets
conception takes into account the objective and subjective weights comprehensive
to determine weight, and then get the more accurately weight. This paper plays a
referenced role for assessing the circular level of circulation in the Qaidam region.
Keywords Circular economy Evaluation index system Intuitionistic fuzzy
entropy Qaidam salt lake
36.1 Introduction
Circular economy is a completely new kind of ecological economy about human
development raised by western countries. It is the economic model based on the
reducing, reusing and recycling principles (known as the 3R principle) (Huang
2004). The real terms to develop circular economy is minimal resource con-
sumption and the smallest environmental costs, to get maximize the development
benefits, so as to unify economic, environmental and social benefits, and achieve
the goal of sustainable development (Xiao 2007). Germany put forward the recycle
economic Law in October 1996, and have done quite well in this regard.
F. Feng (&) Y. Liu J. Zhang N. Wang X. Xing
School of Economics and Management, Beijing Information Science
and Technology University, Beijing, China
e-mail: [email protected]
Y. Liu
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 353
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_36,
Springer-Verlag Berlin Heidelberg 2013
354 F. Feng et al.
China was first reported circular economy in 1997 (Min 1997), it shows that
Chinese late start in the circular economy.
Zhong et al. (2006) summed up in the present on recycle economic, China’s
mainly research recycle economic is concentrated in the following areas: firstly,
the connotation and principle of recycle economics; Secondly, the role of the
recycling economy and the impact on socio-economic; Thirdly, the discussion of
the pattern of the recycle economic; Fourthly, the supporting conditions to achieve
recycle economic. So, in the circulating levels of evaluation have not done enough,
and some of them based steel. For example, Wenjie and Ma (2007) build a set of
index system about ecological steel industry, include positive benefit and negative
benefits, and use them evaluate the Chinese level of ecological steel industry in
2003; Cui et al. (2008) base on the present index system of circular economy and
the evaluation index system of steel and iron industry clean production, and build
evaluation index system of steel and iron industry based on circular economy. Ma
et al. (2007) build index system and evaluation system of steel and iron industry by
green manufacturing, and use DEA evaluate the green degree of Chinese eighteen
steel and iron companies. But, now there is few people evaluate the level of salt
lake circular economy, and it mainly still locate theory research phase. This paper
analysis present salt lake resource situation and reference to the index system of
steel and iron circular economy and Chinese index and evaluation system, and
combine the feature of salt lake to build index and evaluation system, then use the
intuitionistic fuzzy(IF) sets and empower weight algorithm to evaluation.
36.2 The Theory of Intuitionistic Fuzzy Sets
and Aggregation Operators
Zhang (2006) pointed that there is gap of data in index system of recycle economic,
so some index are fuzzy when statistical investigation. Hong and Choi (2000) and
Li (2005) used to do useful exploration of decision problems based on the intui-
tionistic fuzzy sets. And intuitionistic fuzzy sets is characterized by taking into
account the element of non-empty set degree of membership and non-membership
information, and which makes the ability to express, more flexible, and more
suitable to deal with the reality of the practical problems (Wang and Yang 2010).
36.2.1 Intuitionistic Fuzzy Sets
Intuitionistic fuzzy sets constitute a generalization of the notion of a fuzzy set and
were introduced by Atanassov in 1983 (Atanassov 1986).
Definition Set X is a not empty classic set, and an intuitionistic fuzzy set A in a
universe X is an object of the form A = {\x, lA(x), vA(x) [ |x [ X}, where, for
all x[X, lA(x) [ [0,1] and vA(x) [ [0,1] are called the membership degree and the
36 Evaluation of Recycle Level of Qaidam Salt Lake Circular Economy 355
non-membership degree, respectively, of x in A, and furthermore satisfy
0 B lA(x) ? vA(x) B 1 (Xu 2008). For example, [lA(x),vA(x)] = [0.5, 0.3] in
voting model meaning that 10 people, including five members in favor, three
against and two abstained. The class of intuitionistic fuzzy sets in X is denoted by
IFS[X]. For each A-IFS in, if pA (x) = 1-lA (x)-vA(x), then pA (x) is called the
degree of indeterminacy of x to A.
36.2.2 Intuitionistic Fuzzy Weighted Averaging (IFWA)
Operator
In generally multi-attribute decision-making (MADM) program need be
empowered weight, the weight reflect the relative importance of each attribute.
The method to determine weight usually divide into two types: the one is sub-
jective method, weight is determined by the decision makers’ preferences and
experience, the common methods are expert scoring method and Analytic Hier-
archy Process (AHP), the shortcomings of such methods have greater subjective
arbitrariness in the decision-making. Another is objective method, it basis for
attributive information and using the mathematical theory to empower, so it avoid
subjective and arbitrary, but the disadvantage is to ignore the preferences of
decision makers’. Therefore, this paper used intuitionistic fuzzy intuitionistic
fuzzy sets in dealing with uncertainty information than the traditional fuzzy sets
have stronger entropy to determine the weights, and then based on expert scoring
to correct weights, and to determine the weight integrated.
Set Y = (y1,y2,y3,…,ym) is sets for the evaluation programs, set G =
(g1,g2,g3,…,gn) is set for attribution,Pk = (k1,k2,k3,…,kn) is decision makers’
n
preference for all of attribution, and j¼1 kj ¼ 1. So, the formula of Y about
decision information of G is as follows:
1 Xm
EGj ¼ pij þ hij ð36:1Þ
2m i¼1
The EGj reflect the program sets based on the decision-making information of
attribution Gj ambiguity and uncertainty, and the higher values indicate that the
degree of fuzzy and degree of uncertainty higher,and it meaning that the degree it
rely on attribution Gj less. Let DGj is the degree of deviation of decision-making
information of attribution Gj, where, DGj = 1-EGj , j = 1,2,…,n. Then the for-
mula of subjective weights of attribution Gj is as follows:
DGj
r j ¼ Pn ; j ¼ 1; 2; 3; . . .; n ð36:2Þ
i¼1 DGj
Moreover, taking into account the decision makers’ preferences and experi-
ences, we according to the subjective weights k = (k1,k2,k3,…,kn) for correction,
the formula is as follows:
356 F. Feng et al.
kj r j
wj ¼ Pn ð36:3Þ
i¼1 kj rj
The weighted averaging operator is a common operator in multi-attribute
decision making with integrating data and information.
Definition Let WA: Rn ? R, if
WAx ¼ x1 a1 x2 a2 xn an : ð36:4Þ
Then WA is called a weighted averaging operator, where x = (x1,x2,x3,…,
P
xn)T is the weight vector of aj(j = 1,2,3,…,n), with xj [ [0,1] and nj¼1 xj ¼ 1, R
is the set of all real numbers, and ‘‘’’ is the symbol for the addition of
intuitionistic fuzzy sets (Harsanyi 1955).
Amended weights apply to formula (36.4) can get each attributive information
about sets of programs Yi (i = 1,2,…,m), then use the formula of intuitionistic
fuzzy weighted averaging (IFWA) operator, it is as follows (Qin 2012):
Yn wj Yn
IFWAðYi Þ ¼ 1 j¼1 1 lij ; j¼1 vwj ij ; i ¼ 1; 2; . . .; m ð36:5Þ
Then we can get the value of decision-making programs Yi (i = 1,2,…m) based
on the sets of attribution Gj. Finally, we use formula (36.6) sort based on S (IFWA
(Yi)) and select the maximum score value for the optimal solution. The formula
(36.6) is as follows:
Yn wj Yn
SðIFWAðYi ÞÞ ¼ 1 j¼1 1 lij j¼1 vwj ij ; i ¼ 1; 2; . . .; m ð36:6Þ
36.3 Building Evaluation System of Qaidam Salt Lake
Circular Economy
36.3.1 The Resource of Qaidam Salt Lake Recycle Economic
Development and Situation
There are 33 salt lakes in Qaidam, and the salt lake resources are the unique
advantaged resources, and it play an important position and role in Chinese
national economic construction, moreover, the lithium mine, and chemical fertil-
izer and asbestos are ranking first in China, in Table 36.1.
At present, the main developed salt lake includes the Chaka, Keke Lake,
Chaerhan Lake, and small Qaidam Lake. The main production includes salt,
potassium salt, magnesium salt and boron salt (Yu and Tan 2000). However, the
development of Qinghai salt lake resources, in the past 50 years, always had been
based on potassium resource development as the main product. The associated or
36 Evaluation of Recycle Level of Qaidam Salt Lake Circular Economy 357
Table 36.1 The main miner of the qaidam basin
Minerals Main Maintain Potential Province’s Domestic Mining
ingredients reserves (103 value ratio (%) position area
t) (102 million) number
Lithium mineral LiCl 1388.6 3611.7 100 1 10
Sr deposits Celestite 1589.5 79.6 100 1 3
Salt mineral NaCl 32626.0 122349.7 100 1 24
Magnesium salt MgCl 311866.6 9000.0 100 1 21
MgSO4 167339.7 2174.4
Potassium salt KCl 44299.5 2215.0 100 1 22
Mirabilite K2SO4 668516.4 19246.9 100 1 9
Boron mineral B2O3 1152.5 38.4 100 2 12
a
The material comes from the tenth five-year plan of Qaidam basin mineral resources com-
prehensive development and utilization
symbiotic resources, like lithium, boron, magnesium, rubidium, bromine and other
active ingredients do not use effectively, and most of them as the waste emissions,
and formed a one-way linear process of ‘‘resources—products—waste’’. This
unidirectional extensive irrational development, on the one hand lead to tremen-
dous waste of the salt lake resources, and exacerbated destruction of resources; the
other hand, this mode of production resulting in high production costs and
affecting the economic efficiency of enterprises and the competitiveness of prod-
ucts in the market, and weak ability to withstand market risks (2005). Therefore,
the route of development of circular economy is a priority, and effectively evaluate
the circulatory levels of ability is the most important.
36.3.2 The Principle to Build Qaidam Salt Lake Recycle
Economic Evaluation and Index System
The principle to build salt lake recycle economic evaluation and index system
should be based on the ‘‘reducing, reusing, recycling’’ (3R) as the main criteria and
should take into consideration the levels of corporate, regional and social at the
same time. Wang and Chen (2003) proposed the five principles to build index of the
salt lake resources for evaluating sustainable development: systematic principle,
scientific principle, operability principle, regional principle and dynamic principle.
36.3.3 Determining the Index and Evaluation System
of Qaidam Salt Lake Circular Economy
Salt lake recycle economic evaluation index system is still in the exploratory stage,
and it is still not reached a uniform, accepted standard. Based on the above principles
of constructing, and relying on the instructions on the ‘‘recycle economic evaluation
358 F. Feng et al.
index system’’ in China and the existing statistical system of the National Bureau of
Statistics proposed the output indicator of resources, the consumption indicator of
resources, comprehensive utilization of resources indicator, waste disposal volume
indicator. Besides, Wang and Chen (2003) based on the method of problematic focus
(Yang and Hong 2001) determined index system which contains sustainable utili-
zation of salt lake resources, the impact on the environment of exploiting and using
salt lake resources, and sustainable development of the salt lake industries for
themselves. Wang and Feng (2012) considered the unique characteristics of the
object in the circular economy innovation evaluation, and combined with the phase
characteristics of the economic and social development propose the evaluation index
of the recycle economic innovation situation and effects to evaluate the capability of
innovation. Then, this paper consider these evaluation index comprehensively and
provide the evaluation index system with six first-grade indexes and twenty-two
second-grade indexes, in the Table 36.2.
Table 36.2 The evaluation index system of the salt lake circular economy
Target layer Criterion level layer Index layer
The evaluation index The output indicator of The main mineral resources output rate
system of the salt lake resources The energy output rate
circular economy Water resources output rate
The consumption Water consumption per unit of GDP
indicator of resources Unit GDP energy consumption
Ten thousand production value of ‘‘three
wastes’’ emissions
The comprehensive Industrial waste gas comprehensive
utilization indicator of utilization rate
resources Industrial solid waste comprehensive
utilization rate
Industrial water reuse rate
The proportion of industrial ‘‘three
wastes’’ comprehensive utilization
output value of the production in GDP
The waste disposal Waste recycling rate
volume indicator Industrial solid waste disposal quantity
Industrial waste water emissions
Sulfur dioxide emissions
COD emissions
The impact on the The capability of resource consumption
environment of The capability of environmental bearing
exploiting and using The capability of ecological system
salt lake resources bearing
indicator The capability of resources environmental
protection
Recycle economic The proportion of research funds in GDP
innovation situation The proportion of circular economy
evaluation indicator research funds in scientific and
technological activities funds
The number of patents
36 Evaluation of Recycle Level of Qaidam Salt Lake Circular Economy 359
36.3.4 The Steps and Result Analysis for Evaluation
of Qaidam Salt Lake Circular Economy
Based on the formulas of intuitionistic fuzzy sets, and then use the intuitionistic
fuzzy weighted averaging (IFWA) operator to evaluate salt lake recycle economic
index system. The mainly steps are as follows:
Firstly, based on the formula (36.1) to calculate the every attributive intuition
fuzzy entropy, and we can get six value, that is EG1, EG2, EG3, EG4, EG5, EG6
Secondly, based on the formula (36.2) to calculate the every attributive objective
weight; Thirdly, use the formula (36.3) correct weight with the subject weights,
where, the subject weights could come from experts scoring; Fourthly, use the
formula (36.4) we can get the value of IFWA based on sets of programs; Finally,
based on the formula (36.5) we can get the value of the S (IFWA (Yi)) and select
the maximum score value for the optimal solution.
36.4 Conclusion
With the continuous development of circular economy, the effective index system
is conducive to better monitoring and evaluation for circulating levels, and realize
the minimum consumption to get the maximize development benefits. This paper
try to propose the evaluation index system of the salt lake circular economy, and
use the intuitionistic fuzzy sets and intuitionistic fuzzy weighted averaging
(IFWA) operator algorithms, and take into account the objective and subjective
weights comprehensive to determining weight, then get the more accurately
weight, and evaluation of the program can achieve eventually.
Acknowledgments This paper was supported by ‘The National Key Technology R&D Program’
(Grant No.2012BAH10F01, 2011BAC04B02), ‘Education program for New Century Excellent
talents’ (Grant No.NCET-11-0893), ‘Beijing municipal talents supporting plan for enhancing
education’ (Grant No.PHR20110514), ‘science and technology innovation platform of university
in Beijing’ and ‘Beijing knowledge management research base program’.
References
Atanassov K (1986) Intuitionistic fuzzy sets. Fuzzy Sets Syst 20(1):87–96
Cui S, Wang J, Liu H (2008) The iron and steel industry based on circular economic evaluation
index system construction (in Chinese). Chin Sci Technol 16:300–301
Harsanyi JC (1955) Cardinal welfare, individualistic ethics, and interpersonal comparisons of
utility. J Polit Econ 63:309–321
Hong DH, Choi CH (2000) Multi-criteria fuzzy decision-making problems based on vague set
theory. Fuzzy Sets Syst 114(1):103–113
360 F. Feng et al.
Huang X (2004) Circular economy: industry pattern and policy system. Nanjing university press,
Nanjing, pp 10–10
Li DF (2005) Multi attribute decision making models and methods and using intuitionistic fuzzy
sets. J Comput Syst Sci 70(1):73–75
Li D, Zhang W (2006) The research of evaluation index system of circular economy (in Chinese).
Stat Res 23(9):23–26
Ma S, Qi E, Huo Y, Pan Y (2007) The research of iron and steel industry green manufacturing
evaluation system (in Chinese). Sci Sci Manage Sci Technol 28(9):194–196
Min Y (1997) The German circulation economic law (in Chinese). Environ Her 3:40–40
Qin H (2012) Darong Intuitionistic fuzzy entropy and weighted decision making algorithm (in
Chinese). Math Pract Theory 42(4):255–260
Wang X, Chen J (2003) Study on sustainable development evaluation index of Qinghai salt lakes
resources (in Chinese). J Qinghai Norm Univ 42(4):68–72
Wang M, Feng Z (2012) Research on evaluation index system of circular economy innovation (in
Chinese). Chin Popul Resour Environ 22(4):163–166
Wang K, Yang H (2010) Multiple attribute decision making method based on intuitionistic fuzzy
sets (in Chinese). Fuzzy Syst Math 24(3):114–118
Wenjie L, Ma C (2007) The research of iron and steel industry ecological analysis and evaluation
(in Chinese). Inquiry Econ Issues 8:97–103
Xiao H (2007) Review on the development mode of district circular economy and its evaluating
system research (in Chinese). Ecol Econ 4:52–55
Xu Z (2008) Intuition fuzzy information integration theory and application. Science Press,
Beijing, pp 1–3
Yang C, Hong S (2001) Mineral resources sustainable development index discusses-focus method
(in Chinese). Resour Ind 1:29–31
Yi K (2005) Recycling-based economic production model of exploiting resources of Qinghai salt
lakes (in Chinese). J Salt Lake Res 13(2):20–24
S Yu, Tan H (2000), Exploitation of Chinese salt lake resources and environmental protection (in
Chinese). J Salt Lake Res 8(1)24–29, 2000
Zhong T, Huang X, Li L, Wang C (2006) Assessing regional circular economy development:
approaches and indicator systems: a case study in Jiangsu province (in Chinese). Resour Sci
28(2):154–162
Chapter 37
Evolution Analysis of Standardization
Production Behavior in GI Agricultural
Product Enterprise Cluster
Zhi-fang Li and Tong Chen
Abstract GI agricultural products, as the typical representative of quality
agricultural products, play an important role in promoting the development of rural
economy. However, in recent years, some GI agricultural products enterprises did
not process agricultural products according to the standards, so the accidents
occurred frequently. Based on the problems, with the premise of sharing the
collective revenue, this paper uses the evolutionary game theory to analyze the
behaviors of standardization strategy of GI agricultural product enterprise cluster.
The research shows that it must be properly introduced the supervision of the local
government and the association and the construction of the mechanism of pun-
ishment, which can promote the development of GI agricultural product enter-
prises cluster.
Keywords Enterprise cluster Evolutionary game GI agricultural product
Standardized production
37.1 Introduction
In terms of the definition in ‘‘Trade related intellectual property rights agreement’’
(TRIPS), Geographical indications(GI) refers to the commodities in the origin,
which has specific quality, reputation or other characteristics, are determined by
the natural or cultural factors in the region. According to the definition of geo-
graphical indications, geographical indication of agricultural products is identifi-
cation marking of agricultural products, which is naive to a particular country,
region or in a particular place, the quality characteristics, flavor and reputation of
Z. Li (&) T. Chen
College of Management and Economics, Tianjin University, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 361
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_37,
Ó Springer-Verlag Berlin Heidelberg 2013
362 Z. Li and T. Chen
agricultural product completely or mainly depends on the geographic environment,
natural conditions, human factors and so on in this area (Ren and Huang 2007).
Based on the protection of geographical indications of the agricultural products, it
is conducive to the development of the local leading industry, regional brand; but
also can improve the added value of agricultural products, increase the income of
the farmers, and promote rural economic development (He 2010).
According to ‘‘Geographical indications products protection regulations’’ in
China, in the registration process, there must be detailed plant (production, cul-
tivation) specification, local and national standards, which is to ensure that the
quality characteristics of GI agricultural products, but also to make the sustainable
development (Wang and Du 2011). But in recent years, geographical indications
agricultural products appeared more malignant events, such as 2010, ‘‘consumer
advocate’’ in CCTV reported the massive fraud of Heilongjiang Wuchang rice, the
reason mainly lies in the adulteration of Wuchang rice processing enterprises in
origin. These events seriously disturbed the stability of agricultural product market
and violated consumer rights. The fundamental reason lies in the ‘‘free riding’’
behaviors under sharing collective reputation, processing businesses, which
belongs to small and medium enterprises, do not strictly enforce the standards of
production and processing (Sun 2012).
From production to processing, to enter the market, GI agricultural products
involve multiple bodies. At the same time, processing as the most important link,
multiple small and medium-sized processing enterprises distribute in the origin,
which is consisted of a fully competitive market (Wang and Liu 2008). Therefore,
in view of the current literature on the standardized production of GI agricultural
products being not studied, this paper uses evolutionary game theory and method,
under the premise of the collective revenue sharing, builds behavior evolution
model on using standardized strategy in the processing of GI agricultural products
enterprise clusters in the processing link, analyzes the stability of evolution pro-
cess and identifies the key factors to adopt standardized strategy, in order to
provide decision-making references on promoting the GI agricultural product
standardization production.
37.2 Methodology
Evolutionary game theory takes the group behavior as the research object, from the
bounded rational individual as the starting point, thinks that individual decision-
making behavior is not always optimal, it is often not possible to find the optimal
strategy in the beginning. Individual decision-making is through individual mutual
imitation, learning in dynamic process to achieve, stepwise finds the better strategy
(Sun et al. 2003; Wu et al. 2004).
One of the most important dynamic behavior analysis models in evolutionary
game theory is replicator dynamic model. Replicator dynamics is a dynamic dif-
ferential equation which describes adopted frequency of a particular strategy
37 Evolution Analysis of Standardization Production Behavior 363
(behavior) in a group. It can well describe the change trend of bounded rational
individual in groups and predict individual behavior in groups. Replicator dynamic
model can be expressed as:
h i
dxðtÞ= ¼ xðtÞ ut ðaÞ u ð g Þ ;
dt t
where xðtÞ is the proportion of the involved members in the groups who adopt pure
strategy a during the time t involved in groups; ut ðaÞ is the expected utility of the
involved members who use the pure strategy a in the period t; ut ðgÞ is the group’s
average expected utility (Friedman 1991; Taylor and Jonker 1978; Friedman
1998). Evolutionary game theory provides good analysis methods and tools for
analyzing the mutual game behavior of cluster members in GI agricultural product
cluster enterprises who adopt standardization strategy (Hu 2010).
37.3 Results
37.3.1 Model Building
GI agricultural products processing enterprises are the core bodies who can add the
value of agricultural products, the process of adopting standardized production is
directly related to the GI agricultural product reputation and the economic interest
of relevant interest persons. In the processing link of GI agricultural products, this
paper takes adopting standardized production behavior in GI agricultural products
processing enterprise cluster as an example, analyzes the dynamic behavior evo-
lution (Zhang 2011).
Hypothesis 1 GI agricultural product enterprise cluster has n similar enterprises in
scale. At one time t, each enterprise in clusters is faced with the same set of
strategies Si ¼ fs1 ; s2 g, where is named {adopting the standards, not adopting the
standards}. Thus, it assumes that in the period t, the percentage of adopting
standards in cluster is xðtÞ, so the percentage of not adopting standards is 1 xðtÞ.
Hypothesis 2 The amount of resources which each enterprise has is 1, the amount
of resources which can be put into the standardized production is e.
Hypothesis 3 R is the benefit of standardization production in GI agricultural
product enterprises cluster, which is a regional sharing, for all enterprises sharing.
Here, it assumes that Rðe; n; xðtÞÞ ¼ ðenxðtÞÞ1=a , here 0 e 1, a 0. In the above
formula, the benefit R is decided by the numbers nxt ðtÞ of enterprises who adopt
standardization strategy in cluster and the amount e of resource which is put into
the standardized production, and both size is proportional to R; where a is the
productivity parameters of the standardized production, in this article a is assumed
to be valued 2.
364 Z. Li and T. Chen
Hypothesis 4 The income pc of the enterprises who adopt standardized strategy is
composed of two parts, one part 1 e is devoted by the profits of other agricultural
products, another part is the benefit of the standardized production, Rðe; n; xðtÞÞ.
Therefore,
pc ¼ 1 e þ Rðe; n; xðtÞÞ:
Hypothesis 5 The income pb of the enterprises who do not adopt standardized
strategy is composed of two parts, one part is the original resource 1, another part
is the overflow profit of the enterprises who use standardized production, overflow
value coefficient is k, so pb ¼ 1 þ kRðe; n; xðtÞÞ.
Taking into account all enterprises using the standardization production being
randomly paired into the game, at a time t, it can establish the following 2 9 2
asymmetric game model, as shown in Fig. 37.1 (Shen and Wang 2011).
37.3.2 Model Solution
Assume that in a certain period t, uc is the expect profit of the enterprises which
use standardized strategy, ub is the expect profit of the enterprises which do not use
standardized strategy, u is the average expected returns of GI agricultural product
enterprises cluster. According to Fig. 37.1, it can concluded that
uc ¼ xðtÞ½1 e þ Rðe; n; xðtÞÞ
þ ð1 xðtÞÞ½1 e þ Rðe; n; xðtÞÞ ð37:1Þ
¼ 1 e þ Rðe; n; xðtÞÞ
Enterprise 1
adopting the standards not adopting the standards
adopting the 1 − ε + R(ε , n, x(t)), 1− ε + R(ε , n, x(t)),
standards
1 − ε + R(ε , n, x(t)) 1+ λR(ε , n, x(t))
Enterprise 2
1 + λR(ε , n, x(t)), 1 + λR(ε , n, x(t)),
not adopting
the standards 1 − ε + R(ε , n, x(t)) 1 + λR(ε , n, x(t))
Fig. 37.1 The 2 9 2 asymmetric game model of the standardization production in GI
agricultural products cluster
37 Evolution Analysis of Standardization Production Behavior 365
ub ¼ xðtÞ½1 þ kRðe; n; xðtÞÞ
þ ð1 xðtÞÞ ½1 þ kRðe; n; xðtÞÞ ð37:2Þ
¼ 1 þ kRðe; n; xðtÞÞ
u ¼ xðtÞuc þ ð1 xðtÞÞub
¼ xðtÞ½1 e þ Rðe; n; xðtÞÞ ð37:3Þ
þ ð1 xðtÞÞ½1 þ kRðe; n; xðtÞÞ
Therefore, according to the references (Rosenberg 1983), the changed pro-
portion of choosing standardization strategy in cluster enterprises every period can
be used to describe by the following discrete dynamical system:
xðtÞuc
x ð t þ 1Þ ¼ ð37:4Þ
xðtÞuc þ ð1 xðtÞÞub
The equation shows that if the profits of using standardized strategy is greater
than the average expected return of GI agricultural product enterprise cluster, then
it will increase the probability of using standardized strategy in the next period;
otherwise, the cluster will reduce the proportion of using standardized strategy in
the next period.
Subtracting xðtÞ from the Formula (37.4) in both sides, then, introducing the
Formulae (37.1), (37.2, (37.3) into the Formula (37.4), it can be obtained as fol-
lows a discrete dynamical system:
xðtÞð1 xðtÞÞðuc ub Þ
x ð t þ 1Þ ¼ x ð t Þ þ
xðtÞuc þ ð1 xðtÞÞub
1 ð37:5Þ
xðtÞð1 xðtÞÞ ð1 kÞðnexðtÞÞ2 e
¼ x ðt Þ þ 1 1
1 exðtÞ þ xðtÞðnexðtÞÞ2 þkð1 xðtÞÞðnexðtÞÞ2
Commanding xðt þ 1Þ ¼ xðtÞ, when 0\k\1, there are three equilibrium points
e
in the discrete system, respectively 0,1, ; when k ¼ 1, the Formula (37.5)
nð 1 kÞ 2
xðtÞð1 eÞ
changes into xðt þ 1Þ ¼ , there are two equilibrium points,
1 þ xðtÞð1 eÞ
respectively 0,1.
37.3.3 Stability Analysis
According to the discrete dynamic system theory, it can get the following
conclusion:
Conclusion 1 when 0\k\1, the Formula (37.5) can change into (Shen and
Wang 2011)
366 Z. Li and T. Chen
Fig. 37.2 The behavior y
evolution of the
standardization production in
GI agricultural products
cluster
x (t )
0 ε 1
n (1− λ )
2
dxðtÞ xðt þ 1Þ xðtÞ
y¼ ¼
dt x ðt Þ
ð1 xðtÞÞðuc ub Þ
¼
xðtÞuc þ ð1 xðtÞÞub ð37:6Þ
1
ð1 xðtÞÞ ð1 kÞðnexðtÞÞ2 e
¼ 1 1
1 exðtÞ þ xðtÞðnexðtÞÞ2 þkð1 xðtÞÞðnexðtÞÞ2
e
As shown in Fig. 37.2, when 0 xðtÞ\ , which meets uc \ub , at this
nð1 kÞ2
time, GI agricultural product enterprises are all not using standardized strategy;
e
when xðtÞ 1, which meets uc [ ub , at this time, GI agricultural
nð 1 kÞ 2
e
product enterprises are all using standard strategy; x ðtÞ ¼ , it means that
nð1 kÞ2
e
said the proportion, , of GI agricultural product enterprises in the period
nð 1 kÞ 2
is using standardized strategy, which is a non steady state (Shen 2008).
e
Conclusion 2 When xðtÞ 1, the probability of using standardized
nð 1 kÞ 2
strategy decreases (increases) as the amount of resources required increases
(decreases), as the spillover coefficient of adopting standardized strategy increases
(decreases), as the cluster size increases (decreases).
Conclusion 3 When k ¼ 1, x ðtÞ ¼ 0, no GI agricultural product enterprise in
the period uses standardized strategy, which is a pure strategy equilibrium;
x ðtÞ ¼ 1, all GI agricultural product enterprises in the period have adopted
standardized strategy, which is a pure strategy equilibrium.
37 Evolution Analysis of Standardization Production Behavior 367
37.4 Discussion
According to the above, it can be known that the initial state of the GI agricultural
product enterprise clusters, the amount of resource of the enterprise using stan-
dardized strategy, the spillover coefficient of standardization strategy, the effi-
ciency of standardized strategy and the size of GI agricultural product enterprise
cluster and other factors play important roles on the formation, evolution and
stability of GI agricultural product enterprise cluster.
(1) The amount of resource of the enterprise using standardized strategy e. On the
condition that a certain amount of resources needed by standardized strategy,
the less the enterprises using standardized strategy, the more resources
invested by every enterprise. In addition, the more the resource required to
invest, the higher the ratio threshold of using standardized strategy by GI
agricultural product enterprise clusters is. Therefore, reducing e through the
construction of government subsidies mechanism, improving the agricultural
science and technology extension system and other measures are contribute to
the evolution and stability of adopting standardized strategy of the GI agri-
cultural product enterprise clusters generally.
(2) The spillover coefficient k. When k closes to 1, spillover effect is obvious,
which will lead to more ‘‘free-riding’’ behaviors, so it leads to a decrease in the
quality of agricultural products, and even endanger the collective reputation.
So, only by cluster internal or local government to improve the punishment
system, the ‘‘free-riding’’ acts will be reduced.
(3) The efficiency a. The size of a directly affects the threshold of using stan-
dardized strategy in GI agricultural product enterprise clusters. Thus, in
reality, planning the industrial structure and the coordination of organizational
settings and designing management mechanism can effectively influence the
standardized popularization efficiency.
e
(4) The scale n. From the Fig. 37.2, when the initial area is in nð1k Þ2
xðtÞ 1,
the smaller clusters, the popularization rate of using standardized strategy is
higher, thus more stable. It means that, in reality, it should control the number
of enterprises, ensure the effective supervision and management, being com-
bined with the above measures at the same time.
37.5 Conclusion
This paper constructs the evolutionary game model of using standardized strategy
in GI agricultural product enterprise clusters, and focuses on the analysis of
behavior evolution of cluster members in different conditions. The results show
that using standardized strategy may be influenced by a variety of key factors, so
local government, industry associations and other coordination organization which
368 Z. Li and T. Chen
are used to control these factors should play an important role. Establishing the
supervision and punishment mechanism of standardized strategy implementation,
through the improvement of the initial state, reducing the burden of cluster
members, thereby improving the popularity efficiency of standardization, reducing
the exterior effect of standardized popularization, making standardization to carry
out effectively, safeguarding the characteristics and market competitive advantage
of GI agricultural product.
Acknowledgments This paper is supported by the Humanities and Social Sciences Program of the
Ministry of Education of China (10YAJ630014).
The authors thank the editor and the anonymous referees for their comments and suggestions.
References
Friedman D (1991) Evolutionary games in economics. Econometrica 59(3):637–666
Friedman D (1998) On economic applications of evolutionary game theory. J Evol Econ
8(1):15–43
He X-m (2010) Economics substance to the geographical indications institution of agricultural
products. J Lanzhou Commer Coll 26(3):46–51 (in Chinese)
Hu X (2010) Game analysis on collective action dilemma of medium-sized and small enterprises
cluster. Stat Decis 24:62–64 (in Chinese)
Ren Q-j, Huang Y (2007) Study and economic analysis on the protection of geographical
indication. Areal Res Dev 26(6):57–59 (in Chinese)
Rosenberg RM (1983) Dynamics analysis of the discrete systems. People’s Education Press (in
Chinese)
Shen L (2008) Research on government motivation mechanism with green supply chain
evolutionary game (in Chinese). Technol Econ 27(3):110–113
Shen L, Wang Y (2011) Research on evolutionary game of reverse supply chain. Manag Rev
23(8):46–51 (in Chinese)
Sun Q-z (2012) Geographic logo products and county economic development. J Nanjing Agric
Univ (Soc Sci Ed) 12(1):37–42 (in Chinese)
Sun Q-w, Lu L, Yan G-l, Che H-a (2003) Asymptotic stability of evolutionary equilibrium under
imperfect knowledge. Syst Eng Theory Pract 24(2):11–16 (in Chinese)
Taylor PD, Jonker LB (1978) Evolutionary stable strategies and game dynamics. Math Biosci
40(12):145–156
Wang X-m, Du Z-m (2011) Economic and legal analysis on the standardization of agricultural
products protected by geographical indications. J Northwest A&F Univ (Soc Sci Ed)
11(6):148–154 (in Chinese)
Wang Y-f, Liu J-h (2008) The analysis on mechanism of interest linkage between interest main
bodies in geographical indications protection. World Stand Quality Manag 45(12):50–54
(in Chinese)
Wu H, Yang H, Chen L (2004) An analyzes on complexity and evolutionary stability in
competition games. Syst Eng Theory Pract 23(2):90–94 (in Chinese)
Zhang S (2011) Evolution analysis of standards adoption in the construction industry. China Civil
Eng J 44(5):144–148 (in Chinese)
Chapter 38
Forecasting of Vehicle Capacity Based
on BP Neural Network
Ya-qin An, Ya-jun Wang and Wen-wei Gao
Abstract According to the theory of neural network, a forecasting model of BP
neural network is set up on the basis of studying the influencing factors of vehicle
population. The forecasting accuracy is improved greatly compared with the gray
forecasting model. It is valuable to enact reasonably the resource management
policies and establish expropriation counterplan for freight capacity.
Keywords BP neural network Forecasting Influencing factors Vehicle
population
38.1 Introduction
The civilian vehicle capacity resource is an important component of national
defense reserve forces, it is a basic way to enlarge military traffic transportation
strength and enhance the continued support capacity of military traffic transpor-
tation during wartime. Because of the sudden and uncertainty characteristics of
future war, we must prepare for the vehicle capacity mobilization in advance in
order to meet the sharply increased requirements of military transportation. On the
basis of existing statistical data, therefore, it is important to analyze and forecast
exactly the distribution and trend of vehicle capacity resource of each war area
according to the objective law of vehicle capacity development, which can also
help us enact some policies of capacity resource management and establish the
capacity expropriation counter plans.
We must forecast the vehicle population exactly before confirming civilian
vehicle capacity. There are three ways to forecast the vehicle population at present
Y. An (&) Y. Wang W. Gao
Automobile Transport Command Department, Academy of Military Transportation, Tianjin,
China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 369
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_38,
Ó Springer-Verlag Berlin Heidelberg 2013
370 Y. An et al.
(Cheng and Su 2005; Deng and Hu 2010; Zhu 2011; Zhou et al. 2011; Ren et al.
2011; Zhu et al. 2011; Wen 2000; Hu 2003; Yuan 1999; National Bureau of
Statistics 2010; Wu et al. 2006; Guo et al. 2006; Cheng 1997; Zhang and Xue
2008; Ye 2004). The first one is extrapolation method which forecasts the future
status using the past data, for example, the time series method is used frequently;
The second one is cause and effect method which forecasts the future status by
finding out the relationship between the forecasting variables and correlated
variables according to the past data, for example, regression analysis method; The
last one is judgment analytical method which forecasts the future status depending
on the experience and comprehensively analytical ability of forecast experts.
Because of the correlation among the forecast factors of vehicle population is
anfractuosities, the primary and secondary relation is uncertain and the quantity
relation is difficult to pick up, all above methods have some flaws at different level
simultaneously.
Neural network is an important artificial intelligence technology developed in
recent years (Wen 2000; Hu 2003; Yuan 1999). Because of its characteristics such
as parallel distributed processing, self-organization, self-adaption, self-learning,
associative memory, stronger fault tolerance and so on, it is applied successively in
many areas. Moreover, it is used widely to forecast the multi-factor, uncertain and
nonlinear problems (Zhu et al. 2011). This type of forecasting method similar to a
black box can overcome the shortages of traditional forecasting methods well.
Moreover, it can combine well with some traditional forecasting methods (Deng
and Hu 2010). In this paper, BP neural network is used to forecast the vehicle
population, and a forecast model based on BP neural network is discussed.
38.2 BP Neural Network
The structure of Backpropagation (for short BP) Neural Network model is shown
in Fig. 38.1. The main ideas of BP algorithm is dividing the leaning or training
processes into two stages. The first stage is forward-propagation process which
performs according to an order of inputting the initial input vector, entering into
the input layer, going though the hidden layer and arriving at the output layer by
dealing with every layer and calculating the actual output value of each unit; the
second stage is called counter-propagation process. When the desired output result
is not gotten at the output layer, D-value (namely error) between actual output
result and desired output result is calculated recursively layer by layer. In order to
reduce the error, then, the weights of corresponding joints in different layers are
adjusted according to the D-value. The two processes are applied repeatedly and
the learning process is over until the error signal drops off and meets the rea-
sonable demand.
The neural network trained by learning process extracted the nonlinear mapping
relation contained in the specimens, and stored the values of weights in the net-
work. During the operational stage, when a new example is inputted into the
38 Forecasting of Vehicle Capacity Based on BP Neural Network 371
i j counter -propagation
x
y
input
desired output
vector result
output layer
input layer the first the second
hidden layer hidden layer
Fig. 38.1 Model of BP neural network
established neural network, it can build up the arbitrary nonlinear mappings from
n-dimensional input space to m-dimensional output space. Finally, it can exactly
describe the laws of specimens which are hard to describe using some mathe-
matical equations.
In this paper, the model of BP neural network including three layers, namely,
input layer, one hidden layer and output layer was built and discussed.
38.2.1 Confirming the Network Joint Number of Input Layer
There are many relevant factors influencing the vehicle population. For example,
gross industrial output value is the foundation of national economy, it influences
the vehicle market remarkably; Traffic volume reflects the objective demand to the
transport facilities and can promote the development of vehicle market. So pas-
senger and freight volume is an important factor influencing the vehicle popula-
tion; furthermore, highway total mileage is also a key factor determining the
vehicle population. Thus, the joint number of input layer is identified as three, they
are total product of society, road traffic volume and highway total mileage.
38.2.2 Confirming the Network Joint Number of Hidden
Layer and Output Layer
Because of shortage of strict theory or basis, choosing joint number of hidden layer
is a complex question. Normally it is determined in accordance with the experi-
ence. If the hidden units are less, network may not be trained; however, more
hidden units make the learning time to prolong and the error may increase. In
general, when more hidden units are used, the possibility of forming local mini-
mum decreases; conversely, a platform phenomenon caused by the slowly
descending learning error function value forms easily. So the network joint
372 Y. An et al.
number of hidden layer should be considered synthetically combining accuracy
and generality. In this paper, the joint number of hidden layer is confirmed as five
finally by calculating repeatedly and considering the computational accuracy and
speed synthetically.
The network joint of output layer is object variable, namely vehicle population.
38.3 Forecasting Vehicle Population
38.3.1 Demand Analysis of Civilian Vehicle Capacity
According to the tasks and characteristics of military transportation guarantee in
the future warfare, the necessary vehicle types mainly include passenger service
vehicle, motortruck, birdyback, special purpose motor vehicle and so on. For
example, the common trucks and special vehicles such as dump trucks, tank
trucks, vans trucks, etc. with excellent properties and more than 4 ton payload can
meet the requirement of military transportation. These vehicles can be used for
transportation guarantee such as troop maneuvering, transporting to the frontline of
the warstate organs and important factories, goods and materials transportation
under special circumstances such as emergency and armed conflict, and urgency
transport when grave natural calamities are happened. According to Chinese sta-
tistical yearbook, the heavy motortruck populations from 1989 to 2005 and their
main influence factors are listed in Table 38.1.
Table 38.1 National motortruck population and its main influence factors
Years Motortruck Gross product of society Road freight volume Highway total
population (ten (hundred million Chinese (ten thousand ton/km) mileage (km)
thousand) Yuan)
1989 346.37 16909.2 733781 101.43
1990 368.48 18547.9 724040 102.83
1991 398.62 21617.8 733907 104.11
1992 441.45 26638.1 780941 105.67
1993 501.00 34634.4 840256 108.35
1994 560.33 46759.4 894914 111.78
1995 585.43 58478.1 940387 115.70
1996 575.03 67884.6 983860 118.58
1997 601.23 74462.6 976536 122.64
1998 627.89 78345.2 976004 127.85
1999 676.95 82067.5 990444 135.17
2000 716.32 89468.1 1038813 140.27
2001 765.24 97314.8 1056312 169.80
2002 812.22 105172.3 1116324 176.52
2003 853.51 117251.9 1159957 180.98
2004 893.00 136875.9 1244990 187.07
2005 955.50 182320.6 1341778 193.05
38 Forecasting of Vehicle Capacity Based on BP Neural Network 373
Fig. 38.2 Deviation curve of BP neural network training process
38.3.2 Network Training and Forecasting
According to the BP network established in above section, we choose the national
motortruck population from 1989 to 2005 and main influence factors as an
example. Visual Basic language was used to design an application program for
above BP network; MATLAB program was used to train the specimens and
forecast the result. The data from 1989 to 2002 were used as learning specimens,
and the data from 2003 to 2005 are used as samples for forecasting. Square sum
error of network output is designed as 0.05, learning rate is designed as 0.02, and
the maximum training number is designed as 10,000.
The network meets the requirement after training 368 times, and the network
error curve after training is shown in Fig. 38.2; the forecasting result is shown in
Table 38.2.
In order to avoid data overflow and decrease learning rate, the specimen data must
be normalized before forecasting, then each variable has the same normalization
scale. In this paper, the data of training specimens and forecasting samples are
calculated following (40.1), which makes the data to limit in the scope of [-1, +1].
2ðx xmin Þ
Xn ¼ 1 ð40:1Þ
xmax xmin
Here, x is a variable to be determined; xmin is the minimum of x; xmax is the
maximum of x; Xn is the normalized value of x.
Table 38.2 Forecasting results and data error analysis from 2003 to 2005
Years 2003 2004 2005
Actual value 853.51 893.00 955.50
Forecasting value 818.55 849.93 937.92
Relative error 4.10 % 4.82 % 1.84 %
Mean error 3.59 %
374 Y. An et al.
Table 38.3 Forecasting results of vehicle populations in 2015, 2020 and 2025
Years Motortruck Gross product of society Road freight volume Highway total
population (ten (hundred million Chinese (ten thousand ton/km) mileage (km)
thousand) Yuan)
2015 1968.58 587314.8 3316312 479.80
2020 2601.22 785172.3 4506324 555.52
2025 3523.51 977251.9 5729957 648.98
Cheng and Su (2005) forecasted the data from 1999 to 2001 using gray forecast
method and the average forecast error was 20.94 %. So the application of BP
neural network method makes the mean error to decrease from 20.94 to 3.59 %.
It means that the forecasting accuracy is improved greatly by using the BP neural
network method.
38.3.3 Forecasting Vehicle Population
If the forecasting value is input to the network model again as a new input value,
the vehicle population in arbitrary future year can be forecasted. The influence
factors of vehicle populations in 2015, 2020 and 2025 are forecasted separately
using time series method, which are shown in Table 38.3.
38.4 Summary
BP neural network model sets up in this paper is 3-5-1-type, which fits the raw data
accurately and gets higher forecasting precision. This type of method has an
unparalleled advantage compared with other methods, especially it is a more
important and better method to forecast vehicle population, enact capacity
resource management policies and establish capacity expropriation counterplan.
However, BP Neural Network still has some flaws such as slow training rate,
easily plunging into local minimum. And its forecasting accuracy can be further
improved.
References
Cheng QS (1997) Attribute recognition theoretical model with application. Acta Scientiarum
Naturalium Universitatis Pekinensis 33(1):12–20 (in Chinese)
Cheng CS, Su LL (2005) Forecasting of vehicle population in China based on Hereditary BP
Algorithm. J Changsha Commun Univ 21:79–83
38 Forecasting of Vehicle Capacity Based on BP Neural Network 375
Deng W, Hu SJ (2010) Non-linear combined prediction model of medium and long-term civil
vehicle population of China. J Transp Syst Eng Inf Technol 10:103–108
Guo YS, Yuan W, Fu R (2006) Research on characteristics of assessment indexes for road safety.
J Highway Transp Res Dev 23(5):102–103 (in Chinese)
Hu WS (2003) Neural network theory and engineering application. Transportation Institute of
Southeast University, Nanjing, pp 16–23 (in Chinese)
National Bureau of Statistics (2010) National statistical yearbook. China Statistics Press, Beijing,
pp 56–62 (in Chinese)
Ren YL, Cheng R, Shi LF (2011) Prediction of civil vehicles’ possession based on combined
logistic model. J Ind Technol Econ, 31:90–97
Wen X (2000) MATLAB neural network apply design. Science Press, Beijing, pp 26–29 (in
Chinese)
Wu YH, Liu WJ, Xiao QM (2006) AHP for evaluating highway traffic safety. J Changsha Univ
Sci Technol (Nat Sci) 3(2):7–8 (in Chinese)
Ye QY (2004) The synthetic assessment of occupation craftsmanship appraisal of engineering
surveying. Surv Mapping Sichuan 27(1):43–44 (in Chinese)
Yuan ZR (1999) Manual neural network and apply. Tsinghua University Press, Beijing, pp 35–39
(in Chinese)
Zhang XR, Xue KS (2008) Application of attribute recognition theoretical model to
comprehensive evaluation of eco-industrial park. Environ Sci Manage 33(11):167–171
Zhou JJ, Zhao Y, Long DH (2011) The study of prediction model of taxi quantity of the city.
Technol Econ Areas Commun 85:27–31
Zhu XH (2011) Cars quantities predication based on multivariate linear regression. J Hubei Univ
Technol 26:38–39
Zhu C, Zhou HP, Zhong BQ (2011) Forecasting of vehicle population based on combined model
with mixed neural network. J Changsha Univ Sci Technol (Nat Sci) 8:13–16
Chapter 39
Fuzzy Data Mining with TOPSIS
for Fuzzy Multiple Criteria Decision
Making Problems
Chin-yao Low and Sung-nung Lin
Abstract In this study, we purpose a common fuzzy multiple criteria decision
making model. Different with formerly studies, a new concept—fuzzy data mining
scheme is adopted for considering in the model to establish a fuzzy multiple
criteria decision making with time weight (FMCDMTW) model. In this study, a
real case of fuzzy multiple criteria decision making (FMCDM) problem to be
considering. The problem under investigated is a FMCDM problem with multiple
appraisers, and the data considered is combined with historic data and recent data.
Since the evaluated criteria proposed in the literature cannot be defined precisely
and numerically, fuzzy linguistic terms can be used to aggregate them numerically.
It not only conforms to human cognition but also benefits interpretation. Fur-
thermore, notice that the data considered contains certain amount of historic data.
As a result, fuzzy time weighted technique is adopted to resolve this issue.
Keywords Auction website FMCDM Fuzzy data mining Fuzzy time
weighted TOPSIS
39.1 Introduction
According to the development of the internet network, consumer behavior had
been changed. E-commerce becomes one of the main shopping ways. E-commerce
is a way engaged in the business activity through the internet, and its target and
range include inside/outside enterprises and individual consumer. Its business
activity’s content includes goods, service trade, network auction, financial
C. Low (&) S. Lin
Department of Industrial Engineering and Management, National Yunlin University of
Science and Technology, Yunlin, Taiwan
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 377
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_39,
Ó Springer-Verlag Berlin Heidelberg 2013
378 C. Low and S. Lin
exchange, network advertisement, etc. Among them, the auction website has taken
the majority of e-commerce, and is growing up continuously constantly.
The development of the auction website has not only changed people’s life but
also changed its consumption habit. Consumers don’t need to go out and can stay
at home and surf the Net and buy the goods needed. And utilize the auction
website to have several advantages to buy the goods, price flexible, goods various
in style, rate of exchange easy, 24-h marketing and have no regional constraint etc.
The factors which the auction website could operating continuously, totally
depend on users, whether would like to continue using and trading the platform
and doing a deal. In other words, the service performance of the auction website
has determined whether this website can manage continuously. Therefore, how to
measure the auction website effectively to manage and serve performance, not
only the subject that website’s operator cares about, but also an important refer-
ence indicator while choosing the trade platform of consumer.
The first job of setting up the performance assessing method is to propose the
effective assessment criterion. There is no standard measure criterion for auction
website. Therefore, it is quite difficult to set up the effective assessment criterion
accorded with subject of assessing. Especially when assess and plan executors
have no professional knowledge set up of the criterion; the setting-up of the
criterion must gather the relevant specialty personage’s opinion. However, it is
obtained face to face that the professional personage’s opinion is difficult, or need
to spend a lot of cost and time. For overcome this difficulty, this research propose
one utilize literature review to replace expert opinion probe into method—‘fuzzy
time weighted’ effectively originally.
Assess the criterion in performance about the auction website; quite a lot of
literatures propose different assessment criteria. In order to choose the really
important assessment criterion objectively and effectively, give consideration to
the applicability of gradual progress with time of the criterion at the same time.
This research is summing up the assessment criterion that relevant literature,
arrange in an order and select the assessment criterion that the research institute
need by way of fuzzy time weight (Hoffman and Novak 1996; Hung et al. 2003).
To consider the criterion of auction website service performance, will often be
influenced by the appraiser experience subjectively, course the result indistinct and
uncertainty, in addition, the assessment of auction website service performance are
not assess the criterion singly can be satisfied, must consider a lot of assessment
criteria at the same time. So, after the ones that finish assessing criterion, this
research continues utilizing the criterion of assessing to set up the auction website
to deal in the performance questionnaire. What the questionnaire topic made use of
linguistic variable in the fuzzy method to express appraiser’s linguistic variable is
fuzzy, the ones that assessed the criterion after building and constructing, then
appraised to the behavior in assessing the criterion of website by the user of
auction website. Finally, serve to apply fuzzy theory and combine multiple cri-
terion decision method tool TOPSIS come assessment opinion to combine persons
who assess while being arithmetical performance, thus build and construct the
assessment result that an auction website serves performance. The performance
39 Fuzzy Data Mining with TOPSIS 379
and competitive ability that this method can not merely offer service of
understanding the one’s own one in the auction website, can still offer consumers
to choose the best auction website of a service performance (Troy and Shaw 1997;
Zadeh 1965, 1975).
39.2 Methodology
39.2.1 Fuzzy Theory
The fuzzy theory had been proposed by Professor Zadeh. He viewed its human
cognitive process (mainly in order to think deeply and deduce). Uncertainly, it
come by the way of mathematics expression, and mathematics that is traditional is
from having only ‘True’ and ‘False’s two value logics (Binary logic). Expand to
the continuous n-value (Continuous multi-value) with gray area Logic. And the
fuzzy theory utilizes under the jurisdiction of the function (Membership Function).
Value come, describe one speciality of concept, come, shows from 0 to 1 element
which belong to a certain concept of intensities with 1 number value that asks also,
the ones that call this element to assembling in this value are under the jurisdiction
of degree (Membership grade). Expect that discourse domain
U ¼ x1 ; x2; . . .; xn ;
and the fuzzy set of discourse domain U, A e present as
n o
x1 ; ue ðx1 Þ ; x2 ; ue ðx2 Þ ; . . .; xn ; ue ðxn Þ
A A A
ue ðxi Þ means the degree of membership of xi in set A. e
A
The linguistic variable is in the appointed fuzzy set that is used for describing
the natural language under the land, can turn into narration of natural language
with logic person who infer logic narrate, and the linguistic variable regards word
or sentence in the natural language as the parameter that is worth instead of the
regarding counting as value. For example the persons who assess experience the
appraisal which the network shop service performance, can utilize ‘very unsatis-
fied’, unsatisfied, ‘ordinary’, ‘satisfied’, ‘very satisfied’ the sentences of five kinds
of yardstick express and serve the quality intensity of the performance, and can
turn into number value for language purpose via the expert, change into relevant
fuzzy numbers (Hsu and Chen 1996; Kaufmann and Gupta 1991; Klir and Yuan
1995; Langari and Zadeh 1995; Mendel 1995; Zadeh 1965, 1975).
380 C. Low and S. Lin
39.2.2 Multiple Criterion Decision Methods
In true world, most decision questions have a lot of assessment criteria, it was not
the single indicator that be weighed, but each criterion cannot turn into the same
unit to compare, so utilize many criterion decision technology carry on decision of
assessment method to born because of answering. Common many criteria decision
method as if the analytic approach of the level (AHP), For example: Bi (Chow
et al. 1994) AHP and apply the research that is assessed to supplier’s performance;
In addition SMART analytic approach and TOPSIS. And TOPSIS is by Hwang
and Yoon (Van Heck and Ribbers 1997) Multi Criteria Analysis Model. The
method, this theoretical foundation supposes promptly every assessment indicator
all has a dull characteristic that increases progressively or decreases progressively,
it is solved and so-called ideal solution (Grant and Schlesinger 1995; Hung et al.
2003). It is all criterion optimum value that make up (the greatest one of attribute
value of the interests; the minimum one of attribute value of the cost). Shoulder the
ideal to solve on the contrary (Negative-ideal Solution). It is all criteria difference
make up the most (the minimum one of attribute value of the interests; the largest
one of attribute value of the cost). Evaluate for select scheme calculate with Euclid
Distance, in order to assess indicator comparing to degree of approximation that
ideal solve, the scheme chosen answers and the ideal distance that is solving is
being shortest, it is farthest and shoulder the distance that the ideal is being solved.
TOPSIS has assessed the law and has already been accepted and employed the
order of all kinds of trades to compare and assess extensively in the educational
circles. For example: Wang, Xu (The National Library 2011), employ research of
assessing operation performance of listed company of TOPSIS method; Wang,
Chen (The National Library 2011), assess the listed company of the computer and
manage the performance with the financial indicator.
According to motive and purpose of this research, we carry on collection and
arrangement of literature and relevant materials, added in more than 20 foreign
research quoted in the thesis by 105 theses and dissertation, by fuzzy time weight
method, according to assess criterion occurrence number, relevance and time
weight remit, appear 5 indices, total 25 items auction website serve assessment
criterion of performance mainly exactly, and assess the criterion of these to make
into the questionnaire, regard user of the auction website as testees and carry on
questionnaire investigation. And then assess two big auction websites (Yahoo &
Return) at present service performance. The questionnaire one is mainly divided
into two major parts. First, it is importance degree comparing and assessing the
performance of serving and assessing the criterion. Second and assess the satis-
faction of assessing the criterion; And via the questionnaire retrieving, the rated
value to two major auction websites that utilize the fuzzy theory and evaluate the
weight of the service criterion project of network shop and person who assesses
integration, utilize the decision method TOPSIS method of many criteria finally,
get rank order and good and bad detail that two auction websites serve perfor-
mance (Lambert and Sharma 1990; Parasuraman et al. 1985).
39 Fuzzy Data Mining with TOPSIS 381
39.2.3 Assesses and Service Performance Criterion Appraise
The research of the auction website belongs to quite young subject, but numerous
relevant research issues. This research analyzes and quotes use to extract out the
measure criterion of managing performance of auction website which this research
institute needs to relevant literature. The main facts are involved in 105 theses
(The National Library 2011).
First, in order to overcome the difficulty that the expert opinion has, reduce the
relevant cost which assesses the necessary expenditure of the homework at the
same time. This research uses the way in which the literature probes into to
produce the criterion of assessing, look at every literature the same expert, analyze
and combine assessment performance relevant components of managing the per-
formance in each literature with the auction website, use to extract the important
assessment criterion.
Network auction has been just the trade form coming out in development in
recent ten years; the user is at the stage of break-into this kind of transaction
formally. So, with the progress of relevant science and technology and gradual
progress of time, the user, to the performance that the auction website shows,
concerned angle and proportion are being produced and changed constantly.
Because performance of user’s service for auction website varies with gradual
progress of time greatly; the turnover auctioned adds a new line of consumer’s
consciousness with all increasing on day too. While the literature is put in order,
we find though the time block that relevant literature are issued is not big, but
service performance that consumers mind has some changes every year (Bellman
and Zadeh 1970; Chen and Hwang 1992; Day 1984).
In order to further assess and consider objectively because the time array
produced and assessed the criterion change, this research is during the process of
putting the literature in order, the assessment criterion of putting forward various
years is entrusted to different weights respectively. The assessment criterion after
gathering together wholly is according to issuing number of times and relevance, it
issues times fuzzy weight on taking advantage of, in order to get the real weight of
the criterion. Analyzing the domestic master’s thesis quoted will issue times and
all lie between in 1998 to 2010 because of this research institute, and foreign most
literature definitely also and. So, we build the fuzzy time weight of setting up and
belong to one degree of functions as shown in Fig. 39.1.
This research is set out by the user view, through the collection and gathering
together wholly of the literature, by weight method of fuzzy time, calculate
according to assessing the occurrence number of the criterion and time weight,
score according to criterion weight order elect 5 construct surface, total 25 items
auction website serve assessment criterion of performance mainly (The National
Library 2011; Angehrn 1997; Athanassopoulos 2000; Beam 1999; Bellman and
Zadeh 1970; Cheng 1994; Fornell 1992; O’Connor and O’Keefe 1997; Kim and
Stoel 2004; Nielsen and Tahir 2005; Parasuraman et al. 1988; Paul 1996; Simon
2001).
382 C. Low and S. Lin
Fig. 39.1 Ownership function that assess the weight of criterion
Fig. 39.2 Assess the ownership function of the criterion ‘importance’
39.2.4 Assess and Serve the Criterion Weight
of the Performance
After receiving the criterion of assessing, we entrust the criterion to the weight of
different levels through the fuzzy way to define with justice, make by favorable
follow-up study and questionnaire. Assess the detail as follows:
Step1: Determine to assess the semantic parameter of the criterion importance
and count fuzzily
It is because in design of questionnaire utilize by Likert scale with five grade
scale, can utilize person who assesses each person by different linguistic variable.
To expresses the important measurement value of degree of each assessment
criterion of persons who assess, as shown in Fig. 39.2.
Step2: Calculate every fuzzy weight of assessing the criterion
e j ¼ lj ; mj ; uj ; j ¼ 1; 2; . . .; n
w
ð39:1Þ
lj ¼ Min lij ; i ¼ 1; 2; . . .m
i
39 Fuzzy Data Mining with TOPSIS 383
!1=m
Y
m
mj ¼ mij ; i ¼ 1; 2; . . .; m
i¼1 ð39:2Þ
uj ¼ Max uij ; i ¼ 1; 2; . . .; m
i
Step3: Each assess the solving and melting fuzzily of fuzzy weight of the
criterion
Its main purpose is to change the fuzzy weight of 25 assessment criteria into a
clear single number value (Ofj), Can learn importance degree and priority of each
assessment criterion, and often the fuzzy method to melt of the opinion includes
‘the maxima–minimum set method’, ‘the greatest average law’ and ‘center law’,
among them it is the simple and most easy method to calculate too that it is the
most general that center law is and is adopted, so, this research utilizes center law
to change the fuzzy weight of n assessment criteria, the conversion method is as
follows:
l j þ m j þ uj
ofj ¼ ; j ¼ 1; 2; . . .; n ð39:3Þ
3
39.2.5 Combines the Rated Value to the Auction Website
of Persons Who Assess
Utilize questionnaire that retrieve, serve the measurement of the performance.
Serving the measurement step of the performance about the network shop proves
as follows:
Step1: Determine to assess the semantic parameter of criterion satisfaction and
count fuzzily
Utilize the special five grade yardstick scale too, come, weigh person who
assess to each assessment criterion satisfaction, utilize different and fuzzy lan-
guage purpose parameters (very unsatisfied, is unsatisfied, ordinary, satisfied, very
satisfied) To express the measurement value of each assessment criterion satis-
faction, as shown in Fig. 39.3.
Step2: Combine every indicator of assessing in the assessment value of satis-
faction of auction website
e tj ¼ ltj ; mtj ; utj
X
n o ð39:4Þ
ltj ¼ Min litj ; i ¼ 1; 2; . . .m
i
384 C. Low and S. Lin
Fig. 39.3 Assess the ownership function of the criterion ‘satisfaction’
!1=m
Y
m
mtj ¼ mitj ; i ¼ 1; 2; . . .; m
i¼1 ð39:5Þ
n o
utj ¼ Max uitj ; i ¼ 1; 2; . . .; m
i
39.2.6 Utilizes Service Performance of Every Network Shop
of Commenting Amount of TOPSIS Method
Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) proposed
Hwang and Yoon in 1981 were mainly used for solving the decision problem of
many criteria; its basic idea lies in defining and solving positive ideal solution and
negative ideal solution. First, so-called and ideal to solve, mean, and take the place
of select scheme benefit most heavy or cost minimum criterion value. On the
contrary, shoulder ideal solving minimally for benefit or heaviest criterion value of
cost, that is to say this method purpose lies in looking for a best scheme, and
scheme this distance ‘positive ideal solution’ to close most, from ‘negative ideal
solution’ most far.
According to the concept of TOPSIS, we can be defining out and solving and
shouldering the ideal to solve fuzzily ideally fuzzily, utilize fuzzy decision matrix
to calculate the fuzzy distance that counts of trigonometry, get each auction
website to solving and is being defeated by the distance that the ideal is being
solved ideally, utilize the common performance indicator (OPI) finally Value is
come to arrange in an order, find out and serve the best auction website of the
performance. The operation step is as follows:
Step1: Set up and assess matrix and weight matrix fuzzily
e ¼ xtj
D ; t ¼ 1; 2; . . .; k; j ¼ 1; 2; . . .; n ð39:6Þ
kn
39 Fuzzy Data Mining with TOPSIS 385
e ¼ wj
W ; j ¼ 1; 2; . . .; n ð39:7Þ
1n
e e j are calculate assess fuzzy weight of criterion with assess merger fuzzy
x tj and w
service performance value of criterion each under the auction website in t websites
each in Sects. 39.3.1 and 39.3.2.
Step2: Assess the matrix regularization fuzzily
!
ltj mtj utj
er tj ¼ ; þ; þ ; j2B ð39:8Þ
uþ
j uj uj
uþ
j ¼ max utj if j 2 B
t
Step3: Build and construct the fuzzy decision weight matrix of the
regularization
e ¼ ev tj
V ; t ¼ 1; 2; . . .; k; j ¼ 1; 2; . . .; n ð39:9Þ
kn
where
e
v tj ¼ er tj w
ej ð39:10Þ
Step4: Determine to be shouldering the ideal to solve
vþ
e j ¼ rmin d ð e
v tj ;e
v þ Þ e
v tj ; t ¼ 1; 2; . . .k ; j ¼ 1; 2; . . .n ð39:11Þ
j
v
e j ¼ rmin d ð e
v tj ;e
e
v tj ; t ¼ 1; 2; . . .k ; j ¼ 1; 2; . . .n ð39:12Þ
j Þ
v
v þ
where the absolutely positive solution e j and negative solution ev
j is:
v þ
e j ¼ ð1; 1; 1Þ
v
e j ¼ ð0; 0; 0Þ
Step5: Calculate that is shouldering the ideal and solving the distance
Calculate method as follows:
X
n
dtþ ¼ d ev tj ; vþ
j ; t ¼ 1; 2; . . .k ð39:13Þ
j¼1
X
n
dt ¼ d ev tj ; v
j ; t ¼ 1; 2; . . .k ð39:14Þ
j¼1
Step6: Calculate every auction website service performance indicator (OPI)
with arranging in an order
The movements arranged in an order finally, the good and bad order of
performances of the service that the result arranged in an order offers for every
auction website.
386 C. Low and S. Lin
dt
OPIt ¼ t ¼ 1; 2; . . .; k ð39:15Þ
dtþ þ dt
39.3 Results
39.3.1 Process
This research gathers together and exactly appears 5 and mainly indices via col-
lection and arrangement of literature and relevant materials, total 25 assessment
criteria that service performance of auction website, and assess the criterion of these
to design into a questionnaire, the sample is the website user who often use the
auction website (Yahoo and Return). Testee send out it questionnaire by investi-
gation of carrying on, last questionnaire of 60 copies together this research, retrieve
59 part questionnaires, the effective questionnaire 57 part among them, the effective
questionnaire rate is 95 %. 57 persons who assess evaluate 25 service performances
that two major auction websites include to assess importance degree of the crite-
rion, appraise to the satisfaction of this criterion while using this website actually to
the persons who assess, then we utilize the fuzzy theory and TOPSIS method,
calculate the common performance indicators of two major auction websites, ask
the good and bad level of service performance of two major auction websites.
39.3.2 Results
According to the analysis described above, the good and bad orders of the service
performances of two domestic auction websites are: Yahoo is expressively better
than Return, found finally the buyer presented the obvious difference to two major
domestic service satisfaction of auction website as in Tables 39.1 and 39.2.
Table 39.1 Each assess Auction website diþ di
shouldering the distance that
the ideal is being solved of Yahoo 0 1.4718
the criterion Return 1.4718 0
Table 39.2 Service Auction website OPI Sequence
performance indicator of OPI
value with arranging in an Yahoo 1 1
order Return 0 2
39 Fuzzy Data Mining with TOPSIS 387
39.4 Discussion
Have assessed and contained more criterion and persons who assess more in
performance of auction website, it is fuzzy on possessing with assessing the cri-
terion congenitally at the same time, difficult accurate quantization.
Therefore, it is often difficult to exactly amalgamate and verify its objectivity to
assess the result. Among the subjects especially as assessing when surface of
different literary compositions have either excellent or bad behavior, the whole
performances of websites will become quite difficult to arrange in an order. In
addition, the assessment criterion needing to be established of expert opinion, must
often spend lengthy time and a large amount of cost, on a small scale assessment of
plan have more difficultly implement.
To problem characteristic described above, originally research and propose in
weight way of fuzzy time, extract the important assessment criterion out to replace
and generally remit the whole expert opinion with technology such as Delphi
method from the literature. And then through questionnaire way, received the
performance appraisal of user’s service for comparative theme of website. And
will appraise and melt it fuzzily in order to agree with receiving the fuzzy weight
inborn and fuzzily of language purpose. Then we arranged the website’s perfor-
mance in an order with TOPSIS.
Through the appraisal of 25 assessment performances, we received two com-
parative subjects (Yahoo and Return) in this research the performance is arranged
in an order. We can be found out by the result of this research, the service
performance of Yahoo is far higher than Return.
39.5 Conclusion
Assess the performance of the auction website for the example, this research has
combined several kinds of employing extensively and simple and feasible fuzzy
theory technology of the present stage, aim at proposing a common procedure of
fuzzy much attribute decision, as assess target have many attribute, need many
people assess and assess content have fuzzy difficult while quantizing clearly
characteristic, the way that this research institute puts forward is suitable for the
using of assessment of this kind of problem. Can assess the target to further
arrange in an order clearly while combining TOPSIS technology, especially when
assessing the target and arranging in an order numerously and difficultly. Choose
such as large attribute makes policy regardless of the scheme, if assess the criterion
and persons who assess and can be confirmed, can use the way that this research
institute put forward to assess and arrange in an order the track case performance.
In addition, this research regards literature as experts, put forward the brand-
new literature review concept. Screen the way to assess criterion with the fuzzy
time weight, when it is difficult for the expert to investigate expensively and live,
388 C. Low and S. Lin
can offer the comparatively economic substituting scheme not losing its objec-
tivity. The weight concept of fuzzy time, though the concept is quite simple, we
think we can further probe into the application feasibility in other fields in this kind
of concept.
References
Angehrn A (1997) Designing mature internet business strategies: the ICDT model. Eur Manag J
15(4):361–364
Athanassopoulos AD (2000) Customer satisfaction cures to support market segmentation and
explain switching behavior. J Bus Res 47:191–207
Beam CM (1999) Auctioning and bidding in electronic commerce: the on-line auction. PhD
thesis, University of California, Berkeley
Bellman RE, Zadeh LA (1970) Decision-making in a fuzzy environment. Manag Sci
17(4):141–164
Chen SJ, Hwang CL (1992) Fuzzy multiple attribute decision making: methods and applications.
Springer, Berlin
Cheng SW (1994) Practical implementation of the process capability indices. Qual Eng
7(2):239–259
Chow G, Heaver TD, Henriksson LE (1994) Logistics performance: definition and measurement.
Int J Phys Distrib Logist Manag 24(1):17–28
Day RL (1984) Modeling choices among alternative responses to dissatisfaction. Adv Consumer
Res 11:244–249
Fornell C (1992) A national customer satisfaction barometer: the Swedish experience. J Marketing
1(56):6–21
Grant WH, Schlesinger LA (1995) Realize your customers’ full profit potential. Harv Bus Rev
73(4):59–72
Hoffman DL, Novak TP (1996) Marketing in hypermedia computer-mediated environments:
conceptual foundations. J Marketing 60(3):50–68
Hsu HM, Chen CT (1996) Aggregation of fuzzy opinions, under group decision making. Fuzzy
Sets Syst 79(3):279–285
Hung YH, Huang ML, Chen KS (2003) Service quality evaluation by service quality performance
matrix. Total Qual Manag 14(1):79–89
Kaufmann A, Gupta MM (1991) Introduction to fuzzy arithmetic: theory and applications. Van
Nostrand, New York
Kim S, Stoel L (2004) Dimensional hierarchy of retail website quality. Inf Manag 41(5):620–632
Klir GJ, Yuan B (1995) Fuzzy sets and fuzzy logic theory and application. Prentice-Hall Inc.,
Prentice-Hall, NJ
Lambert DM, Sharma A (1990) A customer-based competitive analysis for logistics decisions. Int
J Phys Distrib Logist Manag 20(1):17–24
Langari JY, Zadeh LA (1995) Industrial applications of fuzzy logic and intelligent systems. IEEE
press, New York
Mendel JM (1995) Fuzzy logic systems for engineering: a tutorial. Proc IEEE 83:345–377
Nielsen J, Tahir M (2005) Homepage usability: 50 websites deconstructed. New Riders,
Indianapolis, pp 130–135
O’Connor GC, O’Keefe B (1997) View the web as a marketplace: the case of small companies.
Decis Support Syst 21:171–183
Parasuraman A, Zeithaml VA, Berry LL (1985) A conceptual model of service quality and its
implications for future research. J Marketing 49:41–50
39 Fuzzy Data Mining with TOPSIS 389
Parasuraman A, Zeithaml VA, Berry LL (1988) SERVQUAL: a multiple-item scale for
measuring consumer perceptions of service quality. J Retail 64(1):12–40
Paul P (1996) Marketing on the internet. J Consumer Marketing 13(4):27–39
Simon SJ (2001) The impact of culture and gender on web sites: a empirical study. Data Base
Adv Inf Syst 32(1):25–33
The National Library (2011) Electronic thesis and dissertation System. http://etds.ncl.edu.tw/
theabs/index.html, Retrieved Dec 2011
Troy JS, Shaw MJ (1997) Characteristics of electronic markets. Decis Support Syst 21:185–198
Van Heck E, Ribbers PM (1997) Experiences with electronic auctions in the Dutch flower
industry. Int J Electron Markets 7(4):29–34
Zadeh LA (1965) Fuzzy sets. Inf control 8:338–353
Zadeh LA (1975) The concept of a linguistic variable and its application to approximate
reasoning I, II, III. Inf Sci 8:199–251, 301–357; 9:43–80
Chapter 40
Knowledge Discovery from Granule
Features Mining
Jian-hong Luo, Xi-yong Zhu and Xiao-jun Wang
Abstract For the learning problem on imbalanced distribution of data sets,
traditional machine learning algorithms tend to produce poor predictive accuracy
over the minority class. In this paper, granule features mining model (GFMM) for
knowledge discovery is proposed to improve classification accuracy on the
minority class. Suitable information granules (IGs) are constructed by ETM-
ART2, and then key features analysis method is proposed to discrete represent the
IGs to mining compact knowledge rules. The final class for new samples inputted
to GFMM can soon be decided by the knowledge rules. Experiments were con-
ducted on three data sets with different skewed level, the results show that GFMM
can lead to significant improvement on classification performance for the minority
class and outperforms individual SVM and C4.5.
Keywords Classification Data mining Information granules Key features
analysis
40.1 Introduction
Data mining is the process that attempts to discover usable information in large
data sets (Fayyad et al. 1996). It is a popular research topic in many fields.
However, the data set sample distribution may be highly imbalanced or skewed. It
means one class in the data set might be represented by a large number of
examples, while the other might be relatively represented by a few. Since tradi-
tional classifiers which focus on seeking an overall accuracy over the full data set
often produce high accuracy on the majority class, while poor predictive accuracy
J. Luo (&) X. Zhu X. Wang
Zhejiang Sci-Tech University, Hangzhou, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 391
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_40,
Springer-Verlag Berlin Heidelberg 2013
392 J. Luo et al.
on the minority class, they are not apposite to deal with skewed data decision tasks
(Batista et al. 2004; Chawla et al. 2004).
There are also some methods proposed to cope with this kind of problem, such
as the methods of sampling (Batista et al. 2004; Tong et al. 2011), moving the
decision thresholds (Cristianini 2000; Jo and Japkowicz 2004) and adjusting the
cost-matrices (Cristianini 2000). Sampling methods tend to reduce data imbalance
level by downsampling-removing instances from the majority class or upsampling-
duplicating the training instances from the minority class or both. Moving the
decision thresholds method tries to adapt the decision thresholds to impose bias on
the minority class. The third kind of methods tries to adjust the cost (weight) for
each class or change the strength of rules to improve the prediction accuracy
(Batista et al. 2004). However, these three kinds of methods lack a rigorous and
systematic treatment on imbalanced data (Huang et al. 2004). For example,
downsampling the data might lose information, while upsampling the data will
introduce noise (Bargiel and Andrze 2002).
How to build an effective and efficient model on this kind of problem is an
important concern of the data mining (Tong et al. 2011; Nie et al. 2011;
Liu et al. 2011). In this study, ‘information granulation’ is conducted to build
knowledge discovery model to deal with this problem. Granular computing rep-
resents information in the form of some aggregates such as clusters, classes, and
subsets of a universe, which is called as information granules (IGs), and then
solves the targeted problem on each information granule (Yao 2001). When a
problem involves incomplete, uncertain, or vague information, human beings tend
to utilize aggregates to ponder the problem instead of numbers. If distinct elements
cannot be differentiated by ordinary methods, one may study information granules
which are collections of entities aggregated together due to their similarity,
functional adjacency and indistinguishability or the like (Huang et al. 2004; Nie
et al. 2011; Liu et al. 2011), and then solves the problem over the information
granules .
The process of constructing information granules is referred to as information
granulation. It was first pointed out by Zadeh (1996) who proposed the term
‘information granulation’. Zadeh also emphasized that a plethora of details may
not necessarily amount to knowledge. Information granulation serves as an
abstraction mechanism to reduce an entire conceptual redundancy. Granular
computing (GrC) which concerns the representation and processing of complex
information entities as information granules is quickly becoming an emerging
paradigm of information processing (Bargiel and Andrze 2002). GrC is a superset
of the theory of fuzzy information granulation, rough set theory and interval
computations, and is a subset of granular mathematics (Zadeh 1997). GrC arise in
the process of data abstraction and derivation of knowledge from information
In this study, we propose granular features mining model for knowledge dis-
covery (GFMM) to improve prediction performance, which should not only pro-
duce good prediction accuracy on the majority class but also produce good
prediction accuracy over the minority class. Firstly, suitable information granules
are constructed by ETM-ART2 (Luo and Chen 2008) which is an improved ART2
40 Knowledge Discovery from Granule Features Mining 393
with good clustering performance, and then information granules based key feature
analysis method is proposed to discrete represent the IGs to mining compact
knowledge rules. According to the knowledge rules, it can soon predict the final
class of new inputted data. Experiments for a glass classification problem show
that our method outperforms individual SVM and C4.5 classifiers.
40.2 Granule Features Mining Model
for Knowledge Discovery
Imbalanced data sets occur in many practical industries, such as business, medical
and fraud detecting data. Imbalanced data sets exhibit skewed class distribution in
which most instances belong to the majority class while far fewer instances belong
to a smaller class. Usually, the majority class has bigger proportion than the
smaller class. Take fraud data for example, there are usually very few instances of
fraud as compared to the large number of honest instances because it should be
honest rather than fraud by nature. Information granulation has a better insight into
its essence of the data sets and can help to comprehend the target problem. In this
section, our proposed granule features mining model for knowledge discovery is
described in detail. Figure 40.1 shows the process.
IGs are constructed by ETM-ART2 according to the similarity of numerical
data. The total number of IGs may be remarkably smaller than the size of
numerical data set. Most of all, the number of IGs in the majority class would also
be remarkably reduced compared to the size of numerical data. And it would
increase the proportion of the minority class and improve the imbalanced data
situation. By construct IGs, we also can get main characteristics of the data
essences by reduce some redundancy from the data sets. Then we propose key
feature analysis method processing IGs to mine compact knowledge rules.
40.2.1 Information Granulation
40.2.1.1 Information Granules Construction by ETM-ART2
Denote a data set D ¼ fdi gNi¼1 , it includes N sample patterns with M classes and
each is p-dimension. There is an attribute to express its class name for each sample
pattern.
IGs are usually grouped according to the similar ‘size’ (that is granularity) in a
single layer. Since IGs can exist at different levels of granularity, the appropriate
levels of granularity can be helpful to deal with the target problem and it should be
discovered as knowledge. There are many approaches to construct IGs, such as rough
sets, Self-Organizing Map (SOM) network, and etc. (Bargiel and Andrze 2002).
394 J. Luo et al.
Fig. 40.1 Granule features
mining model for knowledge
discovery
In this paper, we utilize ETM-ART2 network to construct IGs. Because ETM-ART2
network is an improved ART2 network with enhanced triplex matching mechanism,
which has better clustering analysis performance and can be easier to construct useful
IGs. Furthermore, ETM-ART2 network has a vigilance parameter (q) which can be
adjusted to control the degree of similarity of patterns grouped to the same cluster,
this is viewed as an information granule. So q can be used to adjust the granularity of
IGs. The vigilance parameter is the major difference and superiority between ART
network and other unsupervised neural networks. Compared to some other clustering
methods which do not implement vigilance parameter, ETM-ART2 network don’t
40 Knowledge Discovery from Granule Features Mining 395
automatically group all input patterns into one cluster if they are not similar
sufficiently. Furthermore, ETM-ART2 network shows the same appropriate stability
properties as ART2 network, which is also self-organizing network capable of
dynamic, on-line learning and with higher accuracy of hierarchical clustering. So we
use ETM-ART2 network to construct IGs.
40.2.1.2 Selection of Granularity
The granularity determines how similar patterns should be grouped together to
form one IG. It is not an easy task to select an appropriate granularity to construct
IGs.
Firstly, we propose two basic indexes to depict the similarity of IGs. One is IAR
and the other is PIR. IAR is designed to measure the consistency of the class for the
patterns in one IG, which depicts the similarity of patterns in one IG. And it is
defined as follows:
m
IAR ¼ 100 % ð40:1Þ
n
where n denotes the number of all patterns in one IG and m is the number of
patterns possessing the majority class. The value of IAR is larger, the IG could be
more ‘pure’ since more patterns in one IG possess the same class. If IAR = 1, the
IG is called as a pure IG. Then another index can be conduct to measure the purity
of granularity, which is denoted as PIR and defined as
e
PIR ¼ 100 % ð40:2Þ
H
where H is the number of all IGs and e is the number of pure IGs.
Secondly, we propose granularity index to help choosing appropriate granu-
larity. Let HðvÞ denotes the number of IGs by different granularity v, and IARðvÞ
denotes the average IAR. Then granularity index is defined as follows:
IARðvÞ PIRðvÞ
GIðvÞ ¼ ð40:3Þ
HðvÞ
where IARðvÞ and PIRðvÞ are larger, the IGs could be purer, and at the same time,
HðvÞ is smaller, this granularity is a good solution.
The vigilance parameter of ETM-ART2 can be set to about 1 and then decrease
gradually until find a satisfying vigilance with maximum GIðvÞ. Then the vigilance
which make maximum GIðvÞ by granularity vis used to construct appropriate IGs.
And the suitable level of granularity by ETM-ART2 is selected.
396 J. Luo et al.
40.2.2 IGs Representation
Some IGs are formed through ETM-ART2 network learning according to the
selected granularity. Denote the number of IGs is q, usually q\\N, the number of
IGs for the ith class is qi.
It is difficult to mining knowledge rules directly from the attributes values of all
IGs. So we propose the key feature analysis method for IGs to get key feature that
can differentiate with the other classes significantly.
Firstly, we give some definitions:
For patterns of the ith class, the minimum and maximum value on the jth
dimensional attribute can compose the class’s jth dimensional attribute value
range, which is denoted as CFij .
For patterns in the kth granule of the ith class, the minimum and maximum
value on the jth dimensional attribute can compose the granule’s jth dimensional
ði;kÞ
attribute value range, which is denoted as GFj . Any minimum or maximum
value point of granules is denoted as Cut Point on CFij . The interval between any
two cut points is abbreviated as ICF.
For all patterns, the minimum and maximum value on the jth dimensional
attribute can compose the data set’s jth dimensional attribute value range, which is
denoted as QFj.
Then key features are analyzed as follows:
40.2.2.1 Cover Degree of Granules
Make the jth dimensional attribute of the ith class as an example, let Sij denotes a
ði;kÞ
ICF on CF ij . If GF j \ Sij 6¼ /, we say the kth IGs cover Sij . Suppose the number
of IGs of the ith class that cover Sij is gði; Sij Þ, the number of IGs of non-ith class
that cover Sij is gði; Sij Þ. Then we can compute the cover degree of IGs on Sij :
gði; Sji Þ gði; Sji Þ
Gcdði; Sji Þ ¼ ð40:4Þ
qi
It depicts the differentiate capability for the ith class from the other classes on
Sij . The cover degree is larger, the differentiate capability is better.
For each attribute, compute the cover degree of each ICF for all classes
according to Eq. (40.4).
40.2.2.2 Find Key Features
For the jth dimensional attribute of the ith class, choose the ICF with maximum
cover degree as the key feature range of the jth dimensional attribute for ith class,
40 Knowledge Discovery from Granule Features Mining 397
and denoted as KF ij . If the number of ICF with maximum cover degree is larger
than 1, choose the ICF with minimum gði; Sij Þ value as the key feature. Then the
key features of all attributes for all classes can be found.
40.2.2.3 Getting Sub-Attributes
For the jth dimensional attribute, make two endpoint of key feature range of each
class segment QFj into several sub-ranges which are denoted as sub-attributes. And
the jth dimensional attribute can be segmented into several sub-attributes.
40.2.2.4 IGs Discretely Represented by Sub-Attributes
ði;kÞ
According to the sub-attributes, each IG’s attribute GF j can be discretely rep-
resented by discrete value 1 or 0. Including the class of each IG, a decision table
U can be figure out by each IG’s sub-attributes.
Though the number of total attributes is increasing, the sub-attributes parti-
tioned by key features has significant differentiate capability for class and they are
non-strongly correlated, and is represented by discrete value 1 or 0. So mining
from the decision table compose by sub-attributes is easier to get compact
knowledge rules.
40.2.3 Knowledge Rules Mining and Application
According to the decision table U, we use C4.5 decision tree to mining knowledge
rules since C4.5 decision tree algorithm is fast, easy to implement to get useful
rules.
A mined rule has the following format: if a1 ¼ x1 and …and ap ¼ xp then
c ¼ ct . ap ¼ xp denotes sub-attribute ap value is xp , and ct is the decision class.
We use support and confidence to measures the knowledge rules. The number
of instances that contain all items in the antecedent denoted as A(R), and the
number of instances that contain all items in the consequent parts denoted as Y(R).
Then support and confidence can be defined as SupðRÞ ¼ jAðRÞ \ Y ðRÞj=jU j and
Cer ðRÞ ¼ jAðRÞ \ Y ðRÞj=jAðRÞj.
Then we define instance match degree for rule R as MatðRÞ. It is the ratio of the
number of condition attributes the instance meet in the all condition attributes in
antecedent.
The steps to predict the test instance’s class by the knowledge rules are as
follows:
398 J. Luo et al.
(1) Make the instance’s attributes discrete according to the sub-attribute found
above.
(2) Compute the MatðRÞ for each knowledge rule.
(3) Predict the test instance’s class by the rule with maximum Mat (R).
40.3 Experiments and Evaluation
40.3.1 Data Sets
We use glass data set from UCI machine learning repository (Lake and Merz
1998), on which assemble three data sets to make experiments. They are depicted
in Table 40.1. Glass1 includes two classes: building_windows_float_processed,
others. Glass2 includes two classes: headlamps, others. Glass3 include three
classes:
vehicle_windows_float_processed,
vehicle_windows_non_float_processed, others
40.3.2 Evaluation Methodology
Fivefold cross-validation was used to evaluate the performance of the algorithm,
and implemented our experiments using WEKA. For imbalanced data sets, the
measure of overall classification accuracy is not effective to evaluate the prediction
performance of the proposed approach. So we use the geometric mean (GM) to
evaluate our GFMM model. GM is defined as
vffiffiffiffiffiffiffiffiffiffiffiffiffi
u T
u Y
GM ¼ t Tci
T
ð40:5Þ
i¼1
where Tci is the classification accuracy for the ith class.
Table 40.1 Data sets Data sets Number Attributes No. %Class
Glass1 214 9 32.71 %:67.29 %
Galss2 214 9 13.55 %:86.45 %
Glass3 214 9 40.65 %:35.51 %:23.83 %
40 Knowledge Discovery from Granule Features Mining 399
Table 40.2 Information of Data sets q No. of IGs PIRðqÞ/ %
IARðqÞ/ %
parameters and IGs
Glass1 0.85 63 93.92 81.81
Glass2 0.82 59 98.07 96.57
Glass3 0.82 57 90.05 79.87
Table 40.3 Comparison of GFMM, C4.5 AND SVM
Data sets Method Acc (%) Tc1 (%) Tc2 (%) Tc3 (%) GM (%)
Glass1 GFMM 83.72 80.00 85.51 – 82.68
C4.5 80.84 74.29 84.03 – 79.01
SVM 83.64 62.86 93.75 – 76.77
Glass2 GFMM 95.81 86.67 97.30 – 91.76
C4.5 95.79 82.76 97.84 – 89.98
SVM 95.79 79.31 98.38 – 88.33
Glass3 GFMM 81.78 86.21 77.63 80.39 81.33
C4.5 73.83 81.61 65.79 72.55 73.03
SVM 77.10 85.06 67.11 78.43 76.50
40.3.3 Results
When input the samples to the ETM-ART2 to construct IGs, we set q to 0.9 and
then decrease to 0.5 stepped by 0.05. The granularity v is selected with maximum
GIðvÞ. The experimental data about IGs is list in Table 40.2
Compared to the size of training patterns, the number of IGs is reduced to about
30 %. AverðIARÞ is above 90 % and PIR is above about 80 %. So these IGs are
help to mining useful knowledge rules. Then test patterns were predicted their
classes by the knowledge rules.
To compare the performance of GFMM, we conducted the experiments by two
other approaches: (1) non-information granulations for single C4.5 decision tree
algorithm. (2) SVM (Cristianini 2000; Platt 1999), which is based on Person VII
Kernel Function and use Sequential minimal optimization. The results are listed in
Table 40.3
It can be seen that GFMM has better classification accuracy especially for data
set of minority class with manifest higher GM. C4.5 and SVM show good overall
classification accuracy, while they show poorer accuracy for minority class and
GM is poorer.
40.4 Conclusions
In this paper, granule features mining model for knowledge discovery (GFMM) is
studied. We proposed granularity index and use ETM-ART2 to construct useful
information granules, and then proposed information granules based key feature
400 J. Luo et al.
analysis method to discrete represent the IGs to mining compact knowledge rules.
New patterns can be soon predicted by the knowledge rules.
In many real world data sets, most of the instances belong to a larger class and
far fewer instances belong to a smaller. Usually, the smaller class should be
concerned and interesting. And the cost is usually high when the smaller (positive)
class instances are misclassified. Normal systems which ignore the imbalanced
distribution of class tend to misclassify the minority class instances as majority,
and will lead to a high false rate of the minority class. In this paper, we consider
IGs instead of numerical data, it might increase the distribution proportion of
minority class and improve imbalanced situation of data. We also utilize ETM-
ART2’s property to construct suitable IGs conveniently. Through extensive
experiments with different imbalance ratio datasets, our proposed method is shown
to be effective and superior to several other classification methods such as single
C4.5 and SVM.
Acknowledgments This work is supported by the Research Foundation of Education Bureau of
Zhejiang Province, China (Grant No. Y201122135), Science Foundation of Zhejiang Sci-Tech
University (No. 1005841-Y, No. KY2011005).
References
Bargiel A, Andrze J (2002) Granular computing: an introduction. Kluwer Academic Publishers,
Boston
Batista G, Prati RC, Monard MC (2004) A study of the behavior of several methods for balancing
machine learning training data. SIGKDD Explor 6(1):20–29
Chawla NV, Japkowicz N, Kolcz A (2004) Editorial: special issue on learning from imbalanced
data sets. SIGKDD Explor 6(1):16
Cristianini N (2000) An introduction to support vector machines and other kernel-based learning
methods. Cambridge University Press, Cambridge
Fayyad UM, Piatetsky-Shapiro G, Smyth P, Uthurusamy R (1996) Advances in knowledge
discovery and data mining. American Association for Artificial Intelligence Menlo Park, CA,
USA
Huang K, Yang H, King I, Lyu MR (2004) Learning classifiers from imbalanced data based on
biased minimax probability machine. In: Proceedings of the 2004 IEEE computer society
conference on computer vision and pattern recognition, Washington, DC, USA, pp 558–563
Jo T, Japkowicz N (2004) Class imbalances versus small disjuncts. SIGKDD Explor Newslett
6(1):40–49
Lake CL, Merz CJ (1998) UCI repository of machine learning database. Available: http://
www.ics.uci.edu/*mlearn/mlrepository.html
Liu Y, Yu X, Huang JX, An A (2011) Combining integrated sampling with SVM ensembles for
learning from imbalanced datasets. Inf Process Manage 47(4):617–631
Luo JH, Chen DZ (2008) An enhanced ART2 neural network for clustering analysis. In:
Proceedings of the first international workshop on knowledge discovery and data mining,
Washington, DC, USA, pp 81–85
Nie G, Rowe W, Zhang L, Tian Y, Shi Y (2011) Credit card churn forecasting by logistic
regression and decision tree. Expert Syst Appl 38(12):15273–15285
40 Knowledge Discovery from Granule Features Mining 401
Platt JC (1999) Fast training of support vector machines using sequential minimal optimization.
In: Schoelkopf B, Burges C, Smola A (eds) Advances in kernel methods: support vector
learning. MIT Press, Cambridge, pp 185–208
Tong L-I, Chang Y-C, Lin S-H (2011) Determining the optimal re-sampling strategy for a
classification model with imbalanced data using design of experiments and response surface
methodologies. Expert Syst Appl 38(4):4222–4227
Yao YY (2001) On modeling data mining with granular computing. In: Proceedings of the 25th
international computer software and applications conference on invigorating software
development, Washington, DC, USA, pp 638–643
Zadeh LA (1996) Fuzzy sets and information granularity. In: Klir GJ, Yuan B (eds) Fuzzy sets,
fuzzy logic, and fuzzy systems. World Scientific Publishing, River Edge, pp 433–448
Zadeh LA (1997) Announcement of GrC
Chapter 41
Multivariable Forecasting Model Based
on Trend Grey Correlation Degree and its
Application
Li-ping Fu, Lu-yu Wang and Juan Han
Abstract Among the current forecasting methods, trend grey correlation degree
forecasting method is limited to a single variable time series data, but cannot solve
the problem of multivariable forecasting. While multiple regression forecasting
can only be used for multivariable linear forecasting, and can be easily affected by
random factors. Therefore, this paper combines trend grey correlation degree
forecasting based on optimization method and the multiple regression forecasting,
generates the multivariable forecasting model based on trend grey correlation
analysis, and uses this model to forecast GDP in Henan Province, not only to
overcome the effect of random factors on time series, but also to comprehensively
consider the various factors that affect the development of objects, thus to achieve
the effect of improving accuracy and increasing the reliability of forecasting. And
this paper also provides a new method for the study of multivariable combination
forecasting.
Keywords Grey prediction Multivariate regression forecasting Optimization
model Trend grey correlation degree
41.1 Introduction
The grey system theory was first proposed in 1982 by the Chinese scholar, Professor
Deng Julong. This theory requires starting from the system perspective to study the
relationship between the information, that is, to study how to use known
L. Fu L. Wang
Management and Economic Department, Public Resource
Management Research Center, Tianjin University, Tianjin, China
J. Han (&)
Management and Economic Department, Tianjin University, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 403
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_41,
Ó Springer-Verlag Berlin Heidelberg 2013
404 L. Fu et al.
information to reveal unknown information. The grey system theory needs a small
amount of calculation and does not require the typical distribution regularity, so it
has been widely used in many areas. The earliest and most typical forecasting model
of the grey system theory is GM (1, 1) model, but there has been a case with too
large deviation in specific application (Baozhang 1994). In order to overcome the
shortage of GM (1, 1) in relevant applications, people have proposed various cor-
rection methods of GM (1, 1), such as GM (1, 1) with residual error correction, GM
(1, 1) with K correction (Zhang and Dai 2009), etc. In terms of the multivariable
grey prediction, the first model was GM (1, N) (Deng 1990). After that, people
proposed other improved models, for example, GM (n, h), widening GM (n, h), and
so on. However, these models are exponential model, which is unrealistic in many
practical applications (Xiong et al. 2003). For this reason, Peng and Xiong 2008
proposed trend grey correlation degree forecasting model based on time series
correlation analysis that proposed by Zhao et al. 2004, using orthogonal design
strategy to consider the combination when comparing the size between the absolute
differences. But this method only selected some representative combinations, did
not exhaust all the possible combinations, which is bound to affect the forecast
accuracy of the results. In order to achieve the purpose of exhausting the data
combination containing the value to be predicted in series calculation process,
according to the goal of forecasting method based on trend correlation degree,
Wang 2011 established the objective functions to construct optimization problem,
and then solved the value to be predicted in the global scope. However, this method
is limited to the single variable time series, can not solve the problem of multi-
variable forecasting. Therefore this paper combines trend grey correlation degree
forecasting based on optimization method and the multiple regression forecasting,
not only to overcome the effect of random factors on time series, but also to
comprehensively consider the various factors that affect the development of objects,
thus to achieve the effect of improving the accuracy and reliability of forecasting.
41.2 Model Building and Checking
41.2.1 The Optimization Forecasting Model Based on Trend
Grey Correlation Degree
(1) Trend Grey Correlation Degree Model
Definition 1: Provide series
X0 ¼ ðx0 ð1Þ; x0 ð2Þ; . . .x0 ðnÞ; Þ
X1 ¼ ðx1 ð1Þ; x1 ð2Þ; . . .x1 ðnÞ; Þ
41 Multivariable Forecasting Model 405
Let:
Dx0 ðk þ 1Þ ¼ x0 ðk þ 1Þ x0 ðkÞ
ð41:1Þ
Dx1 ðk þ 1Þ ¼ x1 ðk þ 1Þ x1 ðkÞ
If x0 ðk þ 1ÞDx1 ðk þ 1Þ 0 ðk ¼ 1; 2; . . .; n 1Þ; then said series Dx0 and Dx1
have the same trend.
Definition 2 (Zhao et al. 2004):
X0 ¼ ðx0 ð1Þ; x0 ð2Þ; . . .; x0 ðnÞÞ
; ði ¼ 1; 2; . . .; mÞ
Xi ¼ ðxi ð1Þ; xi ð2Þ; . . .; xi ðnÞÞ
Provide X0, Xi are time interval series which have the same length, and
X
n1
jxi ðk þ 1Þ xi ðkÞj 6¼ 0 ði ¼ 0; 1; . . .; mÞ
k¼1
Let:
1 X n1
Di ¼ jxi ðk þ 1Þ xi ðkÞj ð41:2Þ
n 1 k¼1
1
yi ðkÞ ¼ xi ðkÞ ð41:3Þ
Di
In these formulas, k ¼ 1; 2; . . .; n and i ¼ 0; 1; 2; . . .; m.
Dyi ðk þ 1Þ ¼ yi ðk þ 1Þ yi ðkÞ ð41:4Þ
(k ¼ 1; 2; . . .; n 1; i ¼ 0; 1; 2; . . .; m)
8
>
> 0
>
>
< if Dy0 ðk þ 1ÞDyi ðk þ 1Þ ¼ 0
jDy0 ðk þ 1Þj þ jDyi ðk þ 1Þj ð41:5Þ
>
> sgn Dy0 ðk þ 1Þi ðk þ 1Þ
>
> 2maxðjDy0 ðk þ 1Þj; jDyi ðk þ 1ÞjÞ
:
if Dy0 ðk þ 1ÞDyi ðk þ 1Þ 6¼ 0
1Xn1
cðX0 ; Xi Þ ¼ f ðk þ 1Þ ð41:6Þ
n k¼1 i
cðX0 ; Xi Þ is said to be the trend grey correlation degree, fi ðk þ 1Þ is said to be the
relational coefficient of X0 related to Xi at Kþ1 point.
(2) The Optimization of the Mathematical Model Building (Wang 2011)
Based on trend grey correlation degree model, we are going to predict the
system state xðtÞ at time t of a single series X ¼ ðxð1Þ; xð2Þ; . . .; xðt 2Þ; xðt 1Þ;
xðtÞÞ, from the single series we can construct three time series written as:
406 L. Fu et al.
X0 ¼ ðxð1Þ; xð2Þ; . . .; xðt 2ÞÞ
X1 ¼ ðxð2Þ; . . .; xðt 2Þ; xðt 1ÞÞ ð41:7Þ
X2 ¼ ðxð3Þ; . . .; xðt 1Þ; xðtÞÞ
According to similarity principle, the sampling series of the grey system in the
two sampling periods on the adjacent imply almost the same dynamical infor-
mation. In this sense, the value to be predicted should satisfy that the trend grey
correlation analysis cðX0 ; X1 Þ between X0 and X1 is equal to the trend grey cor-
relation analysis cðX1 ; X2 Þ between X1 and X2.
It can construct optimal objective function:
minf = jcðX0 ; X1 Þ cðX1 ; X2 Þj ð41:8Þ
In this formula, cðX1 ; X2 Þ is the function of the value to be predicted, and its
function form is decided by the definition process of the trend grey correlation
degree. When calculating the relational measure, we must meet the problems of
comparing Dy1 ðk þ 1Þ and the value to be predicted Dy2 ðk þ 1Þ and determining
whether Dy2 ðk þ 1Þ is zero, so its function has the form of the phase function,
every comparison and judgment produce a combination of calculating the formula
of relational measure, and result in the relational measure expressed by a func-
tional form of the value to be predicted. Solving the optimization problems can
obtain the value to be predicted. In particular, using the algorithms with global
search feature can solve the whole optimal solution, in order to avoid possible loss
of the combination when calculating relational measure.
41.2.2 Multivariate Forecasting Model Based on Grey
Correlation Degree Forecasting
For single-factor trend analysis and forecasting, we often use trend extrapolation
method (Xu 1998; Yu and Huang 2003). But when using this method to the trend of
series with random factors, it will engender considerable errors, and make predic-
tions distorted. If using the principle of multiple regressions to establish a multi-
variable grey prediction model, it will be conducive to combine various factors
impacted on the development of objects to do comprehensive prediction, and to
improve the forecasting accuracy (Su et al. 2007; Chen and Wang 2012). Building
multivariate forecasting model based on trend grey correlation degree is as follows:
^yðtÞ ¼ b0 þ b1^x1 ðtÞ þ b2^x2 ðtÞ þ þ bk ^xk ðtÞ ð41:9Þ
In this formula, ^yðtÞ is the predictive value of the object at time t; ^xi ðtÞ is the
predictive value of the variable Xi at time t, which is obtained by using optimi-
zation model to solve the trend grey correlation degree model; bi ði ¼ 1; 2; . . .; kÞ is
the parameter to be estimated.
In the model, bi is decided by the least squares method (Li and Pan 2005):
41 Multivariable Forecasting Model 407
B ¼ ðb0 ; b1 ; . . .; bk Þ0 ¼ ðX 0 XÞ1 X 0 Y ð41:10Þ
where:
2 3
1 x1 ð1Þ xk ð1Þ
6 1 x1 ð2Þ xk ð2Þ 7
6
X¼4 7
5
1 x1 ðnÞ xk ðnÞ
Y ¼ ½ yð1Þ yð2Þ yðnÞ 0
Substitute historical data into formula (41.10), we can obtain B.
Finally, substitute the predictive value of each variable ^xi ðt) into the model ^yðt),
we can get a final predictive value.
41.2.3 Model Test
(1) The Coefficient of Determination R2 (Zhang 2009)
P
n
ðyi ^yi Þ2 =ðn k 1Þ
2 RSS=ðn k 1Þ i¼1
R ¼1 ¼ P n ð41:11Þ
TSS=ðn 1Þ
ðyi yÞ2 =ðn 1Þ
i¼1
In this formula, RSS represents the Residual Sum of Squares, TSS stands for
Total Sum of Squares, k is the factor number, n is the total number of equations.
The closer R2 to 1, the better goodness of fitting the estimated regression
function effect on the sample points, that is, the stronger the role of the inde-
pendent variable explaining on the dependent variable (Zhang 2009).
(2) F-Test
P
n
ð^yi yÞ2 =k
ESS=k i¼1
F¼ ¼ n ð41:12Þ
RSS=ðn k 1Þ P
ðyi ^yi Þ2 =ðn k 1Þ
i¼1
ESS is Regression Sum of Squares. F follows Fðk; n k 1Þ distribution.
Given the significant level a, if F Fa=2 ðk; n k 1Þ, indicates that the linear
regression model is not significant and can not be used to predict; if F [ Fa=2
ðk; n k 1Þ, indicates that the linear regression model is significant, can be used
to forecast (Li and Liu 2010).
(3) T Test
T test is to test whether the coefficient of each explanatory variable is signifi-
cant, the test statistic is:
408 L. Fu et al.
^
bj
t¼ ðj ¼ 1; 2; . . .; kÞ ð41:13Þ
sðbj Þ
In this formula, t follows t(n-k-1) distribution. Given the significant level a,
checking a t test table for critical value ta=2 ðn k 1Þ. If jtj [ ta=2 ðn k 1Þ,
rejecting the null hypothesis, which means that, the impact of the independent
variable on the dependent variable is significant (Li and Liu 2010).
41.3 The Application of the Model
This paper selects the State-owned units’ Investment in Fixed Assets (x1), the Total
Revenue from Leasing of the Use Right of State-owned Land (x2) as the factor
variables, studies the effect of two factors on GDP (y) of Henan province, and
forecasts. The data come from the Statistical Yearbook of Henan Province
(2005–2011) (Statistics Bureau of Henan Province 2005).
First, use formula (41.7) to build three series, they are X0, X1 and X2, according
to the State-owned units’ Investment in Fixed Assets (x1).
X0 = (1127.02, 1367.02, 1608.59, 1715.78, 2127.02, 2586.9)
X1 = (1367.02, 1608.59, 1715.78, 2127.02, 2586.9, 2857.41)
X2 = (1608.59, 1715.78, 2127.02, 2586.9, 2857.41, S1)
Second, use formulas ((41.1)–(41.6)) to calculate fi ðk þ 1Þ and cðX0 ; X1 Þ:
f1 ð2Þ ¼ 0:9903;
f1 ð3Þ ¼ 0:7173;
f1 ð4Þ ¼ 0:6330;
f1 ð5Þ ¼ 0:9565;
f1 ð6Þ ¼ 0:7881;
cðX0 ; X1 Þ ¼ 0:8176
Then, calculate cðX1 ; X2 Þ containing the predictive value ^x1 ðt) of 2011, sub-
stitute into formula (41.8), use the Matlab software, get:
S1 ¼3139:5
Similarly, obtain fi ðk þ 1Þs and cðX0 ; X1 Þ of the Total Revenue from Leasing
of the Use Right of State-owned Land (x2) and GDP in Henan Province (y), and the
predictive values S2 and S0. The results are shown in Table 41.1.
According to the data of y, x1 and x2 from the years of 2000–2010, use the
software SPSS18.0 to obtain the numerical value of the estimated parameters bi
and the prediction model.
41 Multivariable Forecasting Model 409
Table 41.1 Grey correlation degree correlation coefficient and predictive value of each variable
GDP(y) Total revenue from leasing of the use
right of state-owned land (x2)
f1 ð2Þ 0.8814 0.6006
f1 ð3Þ 0.8834 0.5929
f1 ð4Þ 0.9957 0.8988
f1 ð5Þ 0.7125 0.5271
f1 ð6Þ 0.7316 0.5461
cðX0 ; X1 Þ 0.8409 0.6331
^x1 ðt) 20688 789.28
^yðtÞ ¼ 264:102 þ 7:440^x1 ðtÞ þ 4:036^x2 ðtÞ
ð41:14Þ
t ð0:255Þ ð6:133Þ ð2:869Þ
R2 ¼ 0:987; F ¼ 294:976 ðp ¼ 0:000Þ
Upon examination, the t value of and x1 are greater than the critical value.
R2 = 0.987 is close to 1, which means the goodness of fit is better. F = 294.976 is
larger, and the corresponding p equals to zero, indicating the formulation is sig-
nificant and can be used to forecast.
Finally, substitute the calculated the predictive values of variables x1 and x2 into
formula (41.14), and get the final predictive value ^yðtÞ = 26279.298.
41.4 Conclusions
Actually, GDP of Henan Province in 2011 is 2.723204 trillion Yuan (Henan
channel on Xinhuanet). Comparing the predictive value and the result simply
calculated by trend grey correlation analysis prediction based on optimization
method (S0 = 26088):
2723204 S0 ¼ 6544:04
2723204 ^yðtÞ ¼ 952:74
It is obviously that the deviation of ^yðtÞis smaller, which is closer to the actual
situation. And this shows that combining trend grey correlation analysis prediction
based on optimization method and the multiple regression prediction, has not only
overcome the effect of random factors on time series, but also comprehensively
considered the various factors that affect the development of objects, and made the
results of prediction more accurate. This also provides a new method for the study
of multifactor combination forecast.
Acknowledgments Project Subject: Soft Science Research Project in Henan Province: Research
on State-owned Assets Supervision Mode in Henan Province (112400430009).
410 L. Fu et al.
References
Chen X, Wang B (2012) The grey forecasting of multiple factors and its algorithm. Math Pract
Theory 42(1):80–83 (In Chinese)
Henan channel on Xinhuanet http://www.ha.xinhuanet.com/add/hnnews/2012/02/29/content_24
799412.htm
Deng J (1990) The tutorial of grey system theory. Huazhong University of Science and
Technology Press, Wuhan, pp 1–30 (In Chinese)
Li G, Liu D (2010) Experimental course of econometrics. China Economic Press, Beijing, p 37
(In Chinese)
Li Z, Pan W (2005) Econometrics. Higher Education Press, Beijing (In Chinese)
Peng W, Xiong H (2008) The prediction method based on grey correlation analysis. J Changjiang
Univ 5(1):129–131 (In Chinese)
Statistics Bureau of Henan Province (2005–2011) Statistical yearbook of Henan province. China
Statistic Press, Beijing (In Chinese)
Su B, Cao Y, Wang P (2007) Research on grey forecasting model of multi-factor time series.
J Xian Univ Archit Technol 39(2):289–292 (In Chinese)
Wang Y (2011) Application of the optimization method in the forecasting based on the grey
correlation degree of time series data. J Qingdao Univ Sci Technol 32(3):317–319
(In Chinese)
Xiong H, Chen J, Qu T (2003) Grey forecasting method based on analogy relation. J Wuhan
Transp Univ 1–20 (In Chinese)
Xu G (1998) Statistic forecast and decision-making. Shanghai University of Finance &
Economics Press, Shanghai (In Chinese)
Yu G, Huang H (2003) The selection method of time series data model. J GuangXi Norm Univ
21(1):191–194 (In Chinese)
Zhang X (2009) Mathematical economics. China Machine Press, Beijing, p 113 (In Chinese)
Zhang H, Dai W (2009) The improvement of grey GM (1, 1) forecasting model. J Zhejiang Univ
Sci Technol 26(1):142–145 (In Chinese)
Zhao B, Tian J, Zhang H (2004) The grey method on correlation degree analysis of time series
data. J Shandong Norm Univ 19(4):17–19 (In Chinese)
Zhu B (1994) Study and reviews on the essential methods of grey system. Syst Eng Theory Pract
14(4):52–54 (In Chinese)
Chapter 42
Partner Selection About the PPP
Reclaimed Water Project Based
on Extension Evaluation Method
Xu-kuo Gao and Xiao-hu Chen
Abstract The model of Private-Public-Partnership can solve the shortage of
reclaimed water project funds effectively, which in favor of the reclaimed water
popularization and application. Choose a suitable partner is one of the most
important steps of Private-Public-Partnership project. In order to evaluate Potential
Partners effectively, we set up evaluation index system of Private-Public-Part-
nership reclaimed water project, which based on the characteristics of reclaimed
water project that combined the comprehensive characteristics of home and
abroad. This extension index system applied extension evaluation method to build
the evaluation model of matter-element, to evaluate partners comprehensive and
choose a suitable partner.
Keywords Extension Partner selection Private-Public-Partnership Reclaimed
water project
42.1 Introduction
In recent years, at home and abroad, the practice and research about construction
and management of public project has pursued energetically Private-Public-Part-
nership (PPP) (Gao 2007). Its aim is to attract non public sectors involved in
infrastructure projects, alleviate the government financial pressure and improve the
level of public services. Now PPP mode has been widely used in global scope, and
is becoming the government achieves its economic goals and upgrades the level of
public services to the core ideas and measures (Li and Qu 2004). The recycled
water in city after depth handling can satisfy the need of industry, agriculture and
X. Gao X. Chen (&)
School of Management, Xi’an University of Architecture and Technology,
Xi’an City, Shanxi, People’s Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 411
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_42,
Springer-Verlag Berlin Heidelberg 2013
412 X. Gao and X. Chen
view in the city fluid matter requests. As an effective path that alleviated shortage
of water resources in water-deficient area. It is paid close attention. But, mean-
while, a lot of factors restrict the project’s construction and development, while the
funds problem is main factor in these factors.
As a constituent part of public project, the project finance mode of Private-
Public-Partnership can effectively work out the funds problem that the recycled
water item faces and push forward the construction of recycled water project. The
PPP project is based on the cooperation of government and the private section. So
that constructs a rational and an effective relationship determines the success of
PPP project (Li 2009). Therefore, one of the first questions that the public sector
facing in the project finance mode of PPP, is how to choose appropriate private
partners (Meng and Meng 2005). And it’s also a very important link that deter-
mines the success of PPP project. The research about selecting partners mostly
concentrates on the supply chain business enterprise look for partners. But the
research about government selecting private partners in PPP project is very few.
This paper, according to recycled water PPP project’s characteristics and
experience at home and abroad, tries to analysis the index sign of selecting
partners. Then it uses the extension evaluation method to evaluate partners rea-
sonably and effectively, optimized partner selection.
42.2 Build Evaluate Index Sign System
The recycled water project belonging to infrastructure occupies large sum of funds,
and has the characteristics of low-return and long return, long recovery period
characteristics. It also faces many risks, such as market, price and fluid matter etc.
This paper, according to recycled water PPP project’s characteristics and the
experience at home and abroad, builds the selection of PPP recycled water project
partner’s evaluation index sign system (Hou 2011).
(1) The credit status of enterprise. The credit status of enterprise mainly
includes two aspects: commercial credit which gets it according to credit rating
and social credit which get it according to the implementing of social responsi-
bility condition. If lacking commercial credit, it will increase project risk, and
increase trades cost. If lacking social credit, it will be hard to balance business
enterprise benefits and social benefits, and increase public private conflict, damage
to the overall interests of the project goals.
(2) The scale of the enterprise cooperation. The size of the enterprise cooper-
ation including three main conditions: registered capital, production capacity and
staff number. Usually, the larger companies, its comprehensive strength and ability
are stronger to resist risk. Choosing a larger cooperative enterprise can digest part
of the risk independently in the face of the environmental factors changing. And it
leads to the success.
(3) The cooperation enterprise’s financial situation. The cooperation enterprise
financial conditions include the total assets, cash flow, the asset-liability ratio,
42 Partner Selection About the PPP Reclaimed Water Project 413
return on assets and other major economic and financial index. It can understand
the assets structure, capital conditions and operating conditions according to
analysis this financial index.
(4) The technology level of cooperation enterprise. The technology level of
cooperation enterprise includes three aspects: quantity of technical personnel,
equipment, and similar project experience. Reviewing the technique of the
cooperation enterprise helps to learn the advantage of the cooperation enterprise.
That’s what determines the result of the item.
(5) The operating and management level of cooperation enterprise. The oper-
ating and management level includes: profit ability, management quality, man-
agement system management system, organization form, and others. The recycled
water project is a newer industry than many other industries. Public sector lacks
experience of project management, construction and operation. Meanwhile, the
recycled water project has long pay-back period. So the recycled water project
needs partner has a higher management level.
(6) The willingness and ability to bear risks of cooperation enterprise. It mainly
includes the following aspects: risk management ability, risk attitude and risk
sharing level. Reasonable risk sharing is an important factor of success for PPP
project (Smith 2004). And it is the foothold of stable cooperation relations. At
present stage, the recycled water industry is still at the further promotion and
development stage, has lots of complicated risks. If project partners lack of
willingness and ability to share the risk, not only can increase the project of
construction and operation cost, but also can reduce the overall efficiency of the
implementation of the project (Cai 2000).
Synthesizes the above analysis, building PPP reclaimed water project partner
evaluate index sign system.
42.3 The Basic Theory of Extenics
The extenics takes framework for the matter-element theory and extension
mathematics theory. With ordered triple form R = (N, C, V) describing objects’
basic unit, short for matter-element. Matter-element is triples that composed by
object, feature and eigenvalue, recorded as R = (N, C, V). N represents object, C
represents feature and V represents eigenvalue, recorded as V = (a, b). They are
called three factors of matter-element. In the matter-element, V = C(N) reflects
the object relationship of quality and quantity. It composed the name of features
(C) and their eigenvalue (V) (Cai 1999). An object has many characteristic ele-
ments; it can use n dimension matter-element to description:
2 3 2 3
N C1 V1 R1
6 C V 7 6 R 7
6 2 27 6 27
Ri ¼ 6 .. .. 7 ¼ 6 .. 7
4 . . 5 4 . 5
Cn Vn Rn
414 X. Gao and X. Chen
Ri = (N, Ci, Vi) (i = 1, 2, …, n).
Build extension evaluate model of cooperative partner.
Above-mentioned has two levels evaluate index, so that can build cooperative
partner multi-level extension evaluate matter-element model.
42.3.1 Extension Evaluate Matter-Element Model
If the letter P signifies the problem which evaluated cooperative partners
R1, R2, …, Rn (n 1) are potential cooperative partners matter-element; r repre-
sents partners comprehensive evaluation of matter-element conditions. So the
comprehensive evaluation problem is: P ¼ Rx r; Rx 2 fR1 ; R2 ; . . .; Rn g (Huang
2007).
The above partner evaluation system has 6 first grade evaluation index
C1, C2, C3, C4, C5, C6, 18 level-two evaluation indexes C11, C12, …, C62.
According to the classification of evaluation index system, establish the multi-level
extension synthesis assessment method. Conditions matter-element r corresponding
attribute
2 value is [a3i, bi2 ]. Enterprise choose 3 partners conditions matter-element is
N C 1 V1 N C1 ½a1 ; b1
6 C 2 V2 7 6 C2 ½a2 ; b2 7
6 7 6 7
r¼6 .. .. 7 ¼ 6 .. .. 7R’s level-two index can be expressed
4 . . 5 4 . . 5
C 6 V6 C6 ½a6 ; b6
as conditions
2 matter-element:
3 2 3
N Ci1 Vi1 N Ci1 ½ai1 ; bi1
6 Ci2 Vi2 7 6 Ci2 ½ai2 ; bi2 7
6 7 6 7
Ri ¼ 6 .. .. 7 ¼ 6 .. .. 7(i = 1, 2, …, 6). Each sub-
4 . . 5 4 . . 5
Cin0 Vin0 Cin0 ½ain0 ; bin0
goal includes n0 indexes.
42.3.2 Partners Extension Synthesis Assessment Method
(1) The evaluation index classification
The evaluation index has six types
C = {C1, C2, C3, C4, C5, C6};
The sub-goal Ci has n0 indexes, Ci = {Ci1, Ci2 … Cn0}.
(2) Determine the weight of every index
Determine the weight of every evaluation index extension synthesis evaluation
method is one of the key problems. The weight of each index should try to
accord with the actual situation. The method is often used the Delphi method
and analytic hierarchy process (AHP) (Zhao et al. 1996). In this paper, the
42 Partner Selection About the PPP Reclaimed Water Project 415
indexes weights determines by the Delphi method. The sub-goal each weight
is:
!
X
n
ai ði ¼ 1; 2; . . .; nÞ ; ai ¼ 1 :
i¼1
(3) Calculating the acceptable degree and optimal degree.
(1) Dependent function. As each evaluation index dimension is different,
some evaluation index is the bigger the better. And some may be smaller
may be more beautiful. So build different correlation functions (Dai and
Qu 2011).
The bigger the better:
8
< KðxÞ ¼ ðx aÞ=ðb aÞ; x 2 \a; b [
KðxÞ ¼ 1; xb ð42:1Þ
:
bKðxÞ ¼ 0; xa
8
The smaller the better:
< KðxÞ ¼ ðb xÞ=ðb aÞ; x 2 \a; b [
KðxÞ ¼ 0; xa ð42:2Þ
:
bKðxÞ ¼ 0; xb
(2) Evaluate each sub-goal Ci for first class evaluation.
Use dependent function calculated review things N for Ci relationship of
degree.
X
n0
Ki ðNÞ ¼ aik Kðcik Þ ð42:3Þ
k¼1
(3) The relationship of degree of review things N is:
KðNÞ ¼ ½K1 ðNÞ; K2 ðNÞ; . . .; Kn ðNÞ:
The goodness is:
X
n
CðNÞ ¼ ai Ki ðNÞ ð42:4Þ
i¼1
Composite indicators of enterprises to be evaluated can be quantified by
comparing the goodness of potential partners, and a preferred partner can be
selected.
416 X. Gao and X. Chen
42.4 Case Analysis
Supposing that now we need to evaluate the following four potential cooperative
enterprises R1, R2, R3, R4, and then select the suitable companies as project
partners. Each enterprise index values are as in Table 42.1.
Thus, In order to simplify the problem, we define that each sub-index weight is
equal, and use the correlation function and the weight to obtain the qualified
degrees of various indicators of various enterprises. Use formula (42.3) to calcu-
late the correlation degree K (N) of the enterprise sub-goals Ci
(i = 1, 2, 3, 4, 5, 6), and then use level analysis method to judge each weight of
sub-goals of the matrix a = [0.21, 0.14, 0.23, 0.1, 0.16, 0.16], and finally use
formula (42.4) to calculate the goodness C (N1), C (N2), C (N3), C (N4). Good-
ness of each enterprise C (Ni) = (0.48, 1.25, 0.36, 0.58), arrange each enterprise’s
goodness: C (N2) [ C (N4) [ C (N1) [ C (N3), C (N2) is largest, and then the
optimal partner is enterprise R2.
Table 42.1 Enterprise index values
Evaluation Evaluation Classic R1 R2 R3 R4
objectives index domain
Credit Commercial credit C11 (70, 100) 93 85 87 96
conditions (C1)
Potential Social credit C12 (70, 100) 95 87 85 95
partners Scale (C2) Registered capital (0.5, 1) 0.8 0.6 0.5 1
(million) C21
Production Capacity C22 (60, 100) 90 92 95 95
Number of employees C23 300 200 260 400 300
Total assets C24 (1, 2) 1.5 0.8 0.8 1.2
Financial Cash flow C31 (0.05, 1) 0.1 0.08 0.07 0.2
situation (C3) Asset-liability ratio C32 (0, 0.6) 0.35 0.4 0.5 0.4
Return on capital C33 (0.1, 1) 0.2 0.15 0.2 0.25
Technical Number of technical (20, 100) 50 55 40 60
level (C4) personnel C41
Equipment C42 (60, 100) 90 85 80 90
Operating and Similar project (80, 100) 90 88 87 91
management experience C43
level (C5) Management quality C51 (70, 100) 91 89 87 90
Forms of organization C52 (60, 100) 90 80 85 90
Management system C53 (60, 100) 89 85 88 89
Profitability C54 (0.15, 1) 0.3 0.3 0.25 0.35
Risk willingness Risk management (70, 100) 87 84 85 88
to take on (C6) capabilities C61
The level of risk (80, 100) 80 87 87 90
sharing C62
42 Partner Selection About the PPP Reclaimed Water Project 417
42.5 Conclusions
The paper constructs evaluation index system about partners selection of PPP
reclaimed water project, and applies extension evaluation method on a compre-
hensive evaluation of potential partners. Considering From multiple perspectives
and multiple factors, rational use the collected information and establish the
evaluation model of multi-level indicator parameters, use quantitative values to
show assessment results and can completely reflect the comprehensive level of
program to make the evaluation simple, easy to operate and more practical. It
provides a simple method for PPP recycled water projects to select partners.
References
Cai W (1999) Extension theory and its application. Chin Sci Bull 44(17):1538–1548
Cai W (2000) The extenics overview. Syst Eng Theory Pract 2:92–93
Dai B, Qu X (2011) Research on a model for choosing cooperation innovation partners based on
fuzzy comprehensive evaluation. Sci Technol Prog Policy 28:120–122
Gao X (2007) The recycled water recycle project risk analysis and control. Group Economics
Research, pp 222–223
Hou L (2011) Game analysis of government investment public construction projects. J Eng
Manag 27:323–324
Huang Y (2007) Research on evaluation method and decision making of PPP project. Tongji
University, Shanghai
Li C (2009) Research on the evaluation of urban transportation sustainable development based on
extension method. Beijing Jiaotong University, Beijing
Li C, Qu J (2004) Project finance, 2nd edn. Science Press, Beijing, p 198
Meng F, Meng J (2005) Partners extension synthesis assessment method. Computer Integrated
Manufacturing System, pp 869–871
Smith NJ (2004) Engineering project management. Blackwell Science Ltd, Oxford, pp 260–263
Zhao H, Xu S, He J (1996) Analytical hierarchy process. Science Press, Beijing
Chapter 43
Personnel BDAR Ability Assessment
Model Based on Bayesian Stochastic
Assessment Method
Zhi-feng You, Tian-bin Liu, Ning Ding and Kai-xuan Cui
Abstract It is important for us to evaluate each trainee’s ability with the aim of
improving BDAR training efficiency. The typical assessment methods include
fuzzy evaluation, gray correlation evaluation, neural network and so on. All of
these methods are not able to make full use of the historical information.
Determining membership function in first two methods is not easy. And ANN
needs a lot of data sample which is difficult to obtain in BDAR training. So we
can’t use these methods to model the assessment of personnel BDAR ability. Then
we introduce Bayesian Stochastic Assessment Method which can deal well with
the nonlinear and random problem. Each indexes’ standard is given according to
the characteristics of BDAR training. A modified normal distribution which can
make full use of historical information was put forward to determine the prior
probability. And the poster probability is determined by distance method.
Keywords Bayesian Stochastic assessment Bayesian theorem Personnel
BDAR ability Prior probability
43.1 Introduction
Battlefield damage assessment and repair is a series of activities, which can make
equipments’ basic functions recover quickly by emergency diagnosis and rush-
repair technologies. It consists of battlefield damage assessment and damage repair
Z. You (&) T. Liu K. Cui
Department of Management Engineering, Mechanical Engineering College,
Shijiazhuang, China
e-mail: [email protected]
N. Ding
Department of Basic Sciences, Mechanical Engineering College, Shijiazhuang, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 419
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_43,
Springer-Verlag Berlin Heidelberg 2013
420 Z. You et al.
(Li et al. 2000). BDAR is the multiplier of equipments’ functions in wartime.
The quick and effective assessment and repair are keystones of a war. It needs
management, technology and training to form the BDAR ability. Training plays an
important role in mastering the basic skills, acquainting the BDAR procedure and
improving their BDAR ability. Based on the ITS technology (Hyacinth 1999), we
must train trainees in accordance with their aptitude. Hence, it must analyze the
ability’s composing and make a best plan for a certain trainee. It can make best use
of training resources and time, and thus the effectiveness can be greatly improved.
There are many assessment methods, such as Fuzzy comprehensive evaluation
(Wang and Sun 2008), Gray correlation analysis method (Gu et al. 2011), Bayesian
discriminate analysis (Liang 2011), artificial neutral net (Jin and Sun 1996) and so
on. They are applied on many domains and the effect is excellent. However, the
BDAR ability assessment for personnel is on another way. There are more than
one effected factors for BDAR assessment. And the relations between these factors
are nonlinear and random. The class of BDAR ability is fuzzy. The history
information which is not considered in the former assessment methods plays an
important role in it. It must know the weight or white weight function of each
index in the first three methods. The result must be affected by the artificial factors,
which is not objective. Artificial neutral net (ANN) can learn and organize auto-
matically, and it is very suitable to model the non-linear problem. Now, it is
widely used in pattern recognition, fault diagnosis, comprehensive evaluation and
so on. We need lots of suitable data swatches to train the ANN for good genera-
lization. In other words, the ANN doesn’t suit to model this problem. But we
cannot get enough data swatches for personnel BDAR ability. As a consequence, it
is not suitable to use the ANN to model BDAR ability assessment. To solve these
shortcomings, we set up an assessment model based on Bayesian stochastic
assessment method. Its applying result is effective and practical.
This article introduces the Bayesian stochastic assessment method and its
application in terms of personnel BDAR ability assessment. In Sect. 43.1, we
review the assessment methods used in other domains and analyze their disad-
vantages. Then, we introduce Bayesian stochastic assessment method to model
ability assessment in Sect. 43.2. The BDAR ability assessment index system and
their standard are given in this part and the method of determining the prior
probability using modified normal distribution and poster probability using
distance are promoted here. In Sect. 43.3, we give an instance to validate this
method. Lastly, the conclusion is given in Sect. 43.4.
43 Personnel BDAR Ability Assessment Model 421
43.2 Bayesian Theorem Stochastic Assessment Method
43.2.1 Bayesian Stochastic Assessment Principle
Let B1, B2,…,Bn denote the events group. Bi \ Bj ¼ /ði 6¼ jÞ, where / is impos-
S
n
sible event, and Pð Bi Þ ¼ 1, PðBi Þ [ 0. Then based on the Bayesian theorem in
i¼1
Probability Theory (Cheeseman et al. 1988), we can get the next formula:
PðBi ÞPðAjBi Þ
PðBi jAÞ ¼ P
n ð43:1Þ
PðBi ÞPðAjBi Þ
i¼1
where, PðAjBi Þ, PðBi jAÞ are both the conditional probability, PðBi Þ is the prior
probability of every events.
43.2.2 Bayesian Stochastic Assessment Model
The personnel BDAR ability can be classified into four levels: B1, B2, B3, B4. The
meanings of them are: BDAR engineer, BDAR assessment, normal technician and
equipment operator (Cheeseman et al. 1988; Fried et al. 1997). P(Bi) (i = 1, 2, 3,
4) is an initial estimation for each trainee’s ability level without any data sample
information, it is prior probability. We can get the prior probability based on the
correlative indexes information in general grade evaluation. But it is reasoned only
by experiments and judgments without enough data sample. Considering on both
the BDAR characteristic and the requirement of applying trainee’s history infor-
mation enough, we modify the normal distribution to estimate the prior probability
for each grade which each person’s BDAR level belongs to. This method will be
discussed detailedly in next part.
P(A|Bi) is the probability of index A belong to level Bi. P(Bi|A) is the
re-estimation of each ability levels’ probability after knowing the index A. It is
poster probability. Bi is the standard of ith personnel ability level, i = 1, 2, 3, 4;
A is index for BDAR assessment; Akj is the value of ith assessment index for kth
BDAR person; j is the assessment index’s number (j = 1, 2,…, m). Based on the
characteristic of personnel BDAR ability level evaluation, we can know: m = 4;
Then we can apply (43.1) in this problem as following:
PðBi ÞP Akj jBi
P Bi jAkj ¼ P4 ð43:2Þ
i¼1 PðBi ÞP Akj jBi
(1) Personnel BDAR ability assessment index and standard: The personnel
BDAR ability assessment is different as normal evaluations in which real man or
422 Z. You et al.
Fig. 43.1 Index system for
BDAR Ability
assessment
Irrelevance Losting number of
Help number Operate time
operate number inevitable operate
woman operates real equipments and evaluated by real experts. This is because the
circumstances needed by BDAR training cannot be well set in real world. So we
must apply the simulation to train. This is the way that real man or woman
operates the virtual prototypes and it is automatically evaluated by the assessment
software. So the assessment index must be suitable to this characteristic. By
analyzing (Li et al. 2000, 2003), we use the help numbers, the operator time, the
irrelevance operator number and the losing number of inevitable operating to
assessment the personnel BDAR ability (Fig. 43.1).
Where the fist three indexes is a certain value and the last one is a variable
according to certain battle damage. We suppose the allowing max time is T, and
then we can get the standard for each level (Wang et al. 2006). It is shown in
Table 43.1.
(2) Prior probability (Ronald and Myers 1978): The prior probability is the
primary estimation of personnel BDAR ability without any sample information.
The prior probability will be uniform distribution on (Li et al. 2000; Gu et al. 2011)
if the trainee is firstly evaluated without any ability history information in database.
Its value is 1/n. If the trainee is not firstly assessed, we will use a modified normal
distribution which considers the ability history information enough in order to
estimate the prior probability (Cheng et al. 1985). This method is detailed in
following.
Firstly, setting a interzone value for Table 43.3 each ability level: B[ {B1, B2,
B , and B4} = {[100, 80), [80, 60), [60, 30), [30, 0]}, the median of B is
3
B 2 fB1 ; B2 ; B3 ; B4 g ) f90; 70; 45; 15g;
Secondly, applying the median of each level Br as expectation (l), the r is
deviation and it meets the next condition:
max B1 Br ; B4 Br ¼ 3r
Table 43.1 Assessment index standard on each level
Assessment Help Irrelevance Operate Losting number
index BDAR number operator time of inevitable operate
ability level number
B1 enigneer 0–1 0–2 T/3-T/2 0–1
B2 assessmentor 2–3 3–5 T/2-2T/3 2–4
B3 technican 4–6 6–8 2T/3-4T/3 5–6
B4operator 7–8 9–10 4T/3-5T/3 8–10
43 Personnel BDAR Ability Assessment Model 423
1
B2
0.9
B4 B3 B1
0.8
0.7
Prior probability
B2
0.6 B4
B1
0.5 B3
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90 100
History assessment value for trainee
Fig. 43.2 Distribution figure of the trainee on each ability level
Then the prior probability subjects to modified normal distribution Nðl; r2 Þ;
i 2
n B l
^
pðBi Þ ¼ pffiffiffiffiffiffiffiffi exp ð43:3Þ
2pr 2r2
where n is the compensation factor which can avoid the too small value of prior
probability and too craggedness of the distribution curve. Its value is got by Monte
Carlo simulation (Fishman et al. 1996; Patz and Junker 1999). Now we only give
the result: when B1 and B4 n ¼ 55, when B2; n ¼ 40, and when B3; n ¼ 35, and
then we can get the distribution figure for the trainee on each ability level.
Fig. 43.2.
Thirdly, we can get the initial probability of each ability level by inputting the
median (Bi ) of each level into (43.3). Because of the adding of compensation
P
factor, 4i¼1 p B^ i 6¼ 1. Then we normalized the initial probability to get the final
prior probability using the next formula:
p B^i
pðBi Þ ¼ P4 ð43:4Þ
p B^i
i¼1
(3)Calculating of P(Bi/Akj): Here, we apply the distance method to calculate the
probability (Cheng et al. 1985). Firstly, we calculate the distance between the
actual value of each index and the standard value of each ability level L. Secondly,
compute the reciprocal value of L.
Then we can use the next formula to calculate P(Bi/Akj):
.
1 Lji
P Akj jBi ¼ P4 ð43:5Þ
i¼1 1 Lji
424 Z. You et al.
where k = 1,2,…,n; j = 1, 2, 3, 4; i = 1,2,3,4; Lji = |Akj - Bij|, k is the number of
the trainee. Bij is the arithmetic mean value of the upper and lower bounds on jth
index on ith level. And then we can get the single index’s conditional probability:
P(Akj/Bi).
(4) The poster probability of comprehensive assessment: When doing the
comprehensive assessment, we use Weighting method to calculate the single
index’s conditional probability (Cheeseman et al. 1988):
X
4
Pi ¼ P Bi =Akj Wj ð43:6Þ
j¼1
where Wj is the weight of each assessment index. There are many methods to
calculate the index’s weight, such as Analysis Hierarchy Process (AHP), binomial
coefficient weighted sum, Rough set method, Delphi, Entropy method, main factor
analysis method, multiple regression analysis and so on (Park and Kim 1999).
Each method has its own advantages and disadvantages. Based on the BDAR
ability assessment’s characteristic, we apply the modified AHP to determine the
weight for each index as shown in Table 43.2 (Jia 1995).
(5) Final decision of BDAR ability level based on maximum probability:
According to the maximum likelihood classification rule, we can get the trainee’s
BDAR ability level (Ph) by choosing the maximum in all probabilities of each
level:
Ph ¼ max Pi ði ¼ 1; 2; 3; 4; 5Þ ð43:7Þ
Table 43.2 Modified AHP method to determine the weight
Assessment Help Irrelevance Operate Losting Weight
index number operator time number
number of inevitable
operate
Help 1 1 2 1 0.2875
number
Irrelevance 1 1 2 1 0.2875
operator
number
Operate 0.5 0.5 1 0.5 0.1429
time
Losting 1 1 2 1 0.2875
number
of inevitable
operate
43 Personnel BDAR Ability Assessment Model 425
43.3 Applying Instances
43.3.1 Experiment Results
By experimenting on certain scenarios in virtual BDAR training system, we get all
of the three trainees’ statistical results in every index as shown in Table 43.3.
We can get the following results by (43.5) and (43.6):
P1 = 0.1429*0.2875 ? 0.5455*0.1429 ? 0.2338*0.2875 ? 0.0979*0.2875 =
0.2124
P2 = 0.2065*0.2875 ? 0.2727*0.1429 ? 0.5453*0.2875 ? 0.1592*0.2875 =
0.2960
P3 = 0.4647*0.2875 ? 0.1091*0.1429 ? 0.1443*0.2875 ? 0.0425*0.2875 =
0.3129
P4 = 0.1859*0.2875 ? 0.0727*0.1429 ? 0.0767*0.2875 ? 0.3184*0.2875 =
0.1956
Because P3 = max{P1,P2,P3,P4} = 0.3129, then the first trainee’s BDAR
ability level is classified to third level, he is technician.
In similarity, the second trainee and the third trainee are both the fourth level,
they are equipment operators.
43.3.2 Results Validating
To validate this method, we compare it with Fuzzy comprehensive evaluation,
Gray correlation analysis method, BP artificial neutral net and PCNN method (Tan
and Steinbach 2011). There are consistencies like Table 43.4. So the Bayesian
stochastic assessment theory for personnel BDAR ability comprehensive assess-
ment is effective and scientific.
Table 43.3 Statistical results
Assessment standard Help Irrelevance operator Operate Losting number of
trainee number number time inevitable operate
First trainee 7 2 2.6T/6 7
Second trainee 8 5 5T/6 3
Third trainee 2 10 8T/6 0
Table 43.4 Comparing of different method
Method Gray correlation Fuzzy BP-ANN Bayesian
trainee analysis comprehensive stochastic
evaluation assessment
First B3 B3 B3 B3
trainee
Second trainee B4 B4 B4 B4
Third trainee B4 B4 B4 B4
426 Z. You et al.
43.4 Conclusion
The Bayesian stochastic assessment methods can use the history information
effectively and it can deal with the complex nonlinear relationship in assessment.
Its physical meaning is explicit, the calculating steps are very easy and the
effectiveness is very good. By analysis, we regard help numbers; irrelevance
operator number, operator time and Losing number of inevitable operate as the
assessment factors. And then we build a comprehensive model for personnel
BDAR ability based on Bayesian stochastic assessment methods. It is consistent
with other assessment methods mentioned before and proved by the applying
instance. In the future, we will apply this method to develop the BDAR training
assessment system to improve the effectiveness of the BDAR training simulation.
References
Cheeseman P, Kelly J, Self M, Stutz J, Taylor W, Freeman D (1988) Auto class: A Bayesian
classification system. In: Proceedings of the 15th International conference on machine
learning, vol 140, pp 52–65
Cheng P, Chen XR, ChenGJ, Wu CY (1985) Parameters estimation (Chinese) Shanghai science
technology princess, Pudong, pp 20–100
Fishman, Monte Carlo (1996): Concepts, algorithms, and applications, Springer, New York,
pp 85–122
Fried N, Geiger D, Goldszmidt M (1997) Bayesian network classifiers. Mach Learn
2–3(29):131–163
Gu XP, Ai JL, Han H (2011) Direction-determination ability evaluation based on interval number
grey relational analysis(Chinese). Aerosp Electron Warf 27(3):26–29
Hyacinth SN (1999) Intelligent tutoring systems: an overview. Artif Intell Rev 40(4):251–277
Jia H (1995) A modified arithmetic determining weight in AHP (Chinese). WTUSM Bulletin of,
Sci Technol, pp 25–30
Jin KR, Sun CH (1996) Chaotic recurrent neural networks and their application to speech
recognition. Neuron Computing 13(224):281–294
Li JP, Shi Q, Gan MZ (2000) BDAR theory and application (Chinese) Army industry publitions,
pp 10–50
Li M, Sun SY, Lv GZ (2003) Design and implementation of evaluation model of simulative
maintenance training (Chinese). Comput Eng 29(9):186–188
Liang C (2011) Evaluating quality model of concrete construction project using Bayes
disciminant analysis method (Chinese) Concr 259:50–53
Park KS, Kim SH (1999) Tools for interactive multi-attribute decision: Range-based approach.
Eur J Oper Res 118:139–152
Patz Richard J, Junker Brian W (1999) A Straightforward approach to Markov chain Monte Carlo
methods for item response models. J Educational Behav Stat 24(2):146–178
Ronald E, Myers WRH (1978) Probability and statistics for engineers and scientists, 2nd,
MacMillan, New York
Tan PN, Steinbach M (2011) Introduction to data mining Posts &telecom press, Jan, pp 139–155
Wang SM, Sun YC (2008) Evaluation of virtual maintenance training in civil airplane based on
fuzzy comprehensive evaluation (Chinese). Aircr Des 28(6):42–45
Wang RS, Jia XS, Wang RQ (2006) Study of intelligence frame based on case of battlefield
damage assessment (Chinese). Comput Eng 32(7):174–184
Chapter 44
Power Control of Cellular System
in CDMA Technology Integrated with SIC
Pan-dong Zhang and Juan-juan Min
Abstract In terms of the problems in power control in the CDMA system of the
existing SIC, whether SIC is perfect or not should be considered first, i.e., whether
there is error in channel estimation or judgment. Secondly, the multi-cell situation
should also be considered with outer-cell interference no longer be assumed
independent of the transmitted (received) power of users. In the context, the paper
firstly derives the expressions for optimal power allocation in the decoding order;
then puts forwards the optimal decoding order which can minimize the outage
probability of the system for the restriction of transmitted (received) power of
users, and finally deduces the optimal decoding order that can minimize the total
transmitted power. The paper comes to the conclusion of being able to minimize
the total transmitted power on the condition that the Eb/I requirements of users are
met through simulation comparison of total transmitted power in different
decoding orders.
Keywords CDMA Cellular system Decoding order Power control SIC
Many users share the same frequency band in the cellular system based on the
CDMA technology and the signals among users can be distinguished through
spread spectrum code word. However, due to the multi-path effect of wireless
channel transmission and limitation of spread spectrum code word, the user signals
are not orthogonal at the receiver and accordingly, there is mutual interference or
multi-access interference between the received signals of users, which will result
in problems (Paulraj et al. 2004) like near-far effect and corner effect in the system.
The power control technology can to some extent reduce the negative impact of
this series of problems on the system performance. Meanwhile, the adoption of
proper transmitted power will lengthen the service time of terminal battery and
extend the life time. Moreover, the control of users’ transmitted power and
P. Zhang (&) J. Min
School of Information, Jiujiang University, Jiujiang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 427
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_44,
Springer-Verlag Berlin Heidelberg 2013
428 P. Zhang and J. Min
transmission rate is also a key method to guarantee the Quality of Service (QoS)
requirements of different types of users and to improve the system capacity in case
of co-existence of multiple operations.
At present, joint optimization (Ivanek 2008) has been conducted for the power
control and decoding order adjustment in the CDMA system integrated with Serial
Interference Cancellation (SIC). Nevertheless, in this kind of research, the opti-
mization only aims to adjust the decoding order to minimize the total transmitted
power of users under the premise of catering to the users’ requirements on bit-
energy-to-interference-spectral-density ratio, Eb/I. And there are following limi-
tations in these operations respectively: the assumption (Paulraj et al. 2004) of SIC
is perfect, i.e., the previously detected user signal interference has been eliminated
completely in the subsequent detection of user signals; the interference of outer-
cell user signal on the in-cell user signal (outer-cell interference for short) is
assumed independent of the constant of users’ transmitted (received) power or
considered only for the single cell cases (Zahariadis and Doshi 2004); there is no
limitation (Jorguseski 2001) of users’ transmitted (received) power.
However, in practice, the channel estimation error and decoding judgment error
make it impossible for the SIC to be perfect. That is to say the reconstructed
signals of the user whose decoding has been detected will never be equivalent
correctly to the received signals of the user and thus, after the elimination of the
reconstructed signals of the user whose decoding has been detected from the
composite signal, if there is still residual signal component of this kind of users in
the composite signal, it will impact the system performance in decoding detection
of the current users.
44.1 System Model
Suppose that there are K users in the target cell, the received composite signal r0(t)
at the base station is composed of received signals xi(t) (i = 1,2,…,K) of various
users, outer-cell interference signals i(t) (i.e., interference of user signals from
outer-cell on in-cell signals) and background noise n(t). The expression is as
follows:
X
K
r0 ðtÞ ¼ xi ðtÞ þ iðtÞ þ nðtÞ ð44:1Þ
i¼1
where the received signal xi(t) (i = 1,2, …, K) of user i is
pffiffiffiffiffi
xi ðtÞ ¼ gi ðt si ÞÞ ej/iðtsi Þ Ei bi ðt si Þ ai ðt si Þ; i ¼ 1; 2; . . .; k ð44:2Þ
where gi(t) and /i ðtÞ are respectively the growth rate of channel gain and phase
component of user i and Ei, bi(t) and ai(t) are respectively the transmitted power,
44 Power Control of Cellular System 429
data bit and the sequence of spread spectrum code word, and si refers to the
relative time delay.
Figure 44.1 shows the structure (Jantti and Kim 2000) of CDMA base station
receiver integrated with SIC technology. The base station receiver integrated with
SIC detects and decodes the signals of various users in the received composite
signals. In the event of detection and decoding of the user i, the base station
receiver will reconstruct the received signals of the user through channel esti-
mation with the reconstructed signal recorded as si (t). Eliminate the reconstructed
signal from the composite signal before detection of the subsequent users; that is to
say, before the detection and decoding of user i, the composite signal should be
renewed as
ri ðtÞ ¼ ri1 ðtÞ si1 ðtÞ; i ¼ 1; 2; . . .k ð44:3Þ
where s0 (t) = 0. The mth bit is decoded and decided (Saghaei and Neyestanak
2007) by the following means for user i.
Z
^ 1 mTi þsi
bi;m ¼ sgnf riðtÞ ai ðt si Þdtg ð44:4Þ
T ðm1ÞTi þsi
where, Ti indicates the duration of the bit symbol. This process will be repeated in
the base station until the signals received by the K users are detected and decoded.
Accordingly, the signal to interference-plus-noise ratio (SINR) for the noise
received by user i is:
P
SINRi ¼ Pi1 PK PK ; i ¼ 1; 2; . . .k ð44:5Þ
j¼1 hPj þ j¼iþ1 Pj þ f ð j¼1 Pj Þ þ N0 W
It is assumed that all hi share the same value, and represented by h, which is the
peak value of all hi on conservative terms (Gu et al. 2005). In the above expres-
sion, N0W indicates the background noise power, W Pthe band width and N0 the
K
corresponding power spectral density. Moreover, f j¼1 Pj indicates the outer
cell interference power (Liao et al. 2007), and f the ratio of the outer cell
Fig. 44.1 SIC-based CDMA
base station receiver
430 P. Zhang and J. Min
interference power to the in-cell received power, which means the outer cell
interference is no longer independent of users’ transmitted (received) power. This
is due to the fact that different received power requirements of users mean different
transmitted power requirements under different decoding orders; therefore, the
interference is imposed by in-cell users on the outer cell varies. Likewise, the
signal power differs with different decoding orders. That is why it is unjustifiable
to assume the outer-cell interference to be a constant independent of users’
transmitted (received) power in actual multi-cellular system.
44.2 Optimal Power Allocation for Multi-Cell
The expression for optimal power allocation with certain decoding order has been
provided in documentation (Kwok and Lau 2003); however, it is only applicable to
the supposition that the outer-cell interference is a constant independent of users’
transmitted (received) power or aims at single-cell. The expression of optimal
power allocation for multi-cell will be deduced in this article where the outer-cell
interference is no longer assumed to be independent of users’ transmitted
(received) power.
Users in the system have their specific requests for QoS, and only when each
QoS is satisfied can normal communication be ensured:
ðW=Ri Þ SINRi ci ; i ¼ 1; 2; . . .k ð44:6Þ
where, !i indicates the SINR user i requests at the rate of Ri. In practice, if users’
transmitted (received) power allocated by the system can validate the above
expression, then there will be specific power allocated that can validate the fol-
lowing expression (Lee et al. 2002):
ðW=Ri Þ SINRi ¼ ci ; i ¼ 1; 2; . . .; k ð44:7Þ
This is provable in reference to how an individual user examines the receiver
(Gu et al. 2005), but the proof will not be given here. Accordingly, it is the best
way to allocate power with a given decoding order and minimum total transmitted
power. Therefore, QoS requested by users is given by expression (44.7).
It is assumed that !i ¼ W=ðRi !i Þ so as to deduce the received power requested
by users with the Eb/I requirement met. The following equations can be got using
expressions (44.5) and (44.7):
Xi1 XK XK
Yi Pi þ Pi ¼ j¼1
hP i þP i þ j¼iþ1
P i þf j¼1
P j þ N0 W; i ¼ 1; 2; . . .; k
ð44:8Þ
44 Power Control of Cellular System 431
Xi2 Xi2 XK
Yi1 Pi1 þ h Pi1 ¼ h Pi1 þ j¼1
hP þ P þf
j¼1 i j¼1
P i þ N0 Wi
¼ 1; 2; . . .; k
ð44:9Þ
where the proportions at the right side of expressions (44.8) and (44.9) are the
same; therefore it can be deduced that the received powers of two users with
neighboring decoding orders can be expressed as follows:
h þ Yi1
Pi ¼ Pi1 ; i ¼ 2; . . .; K ð44:10Þ
1 þ Yi
Further, suppose that i ¼ 1 and i ¼ K in expression (44.8), along with
expression (44.10), the received powers requested by the user whose decoding is
first examined (user l) and the user whose decoding is last examined (user K) are
shown as follows:
N0 W
P1 ¼ nP Qi o ð44:11Þ
K
ðY1 f Þ ð1 þ f Þ i¼2 j¼2 h þ Yj1 1 þ Yj
N0 W
Pk ¼ nP Qi o ð44:12Þ
K
ðYk f Þ ðh þ f Þ i¼2 j¼2 1 þ Yj1 h þ Yj
The above expressions give the allocation means of users’ received powers
more commonly when compared with the existing documentations. The expres-
sions (44.11) and (44.12) give the same requested received powers as in docu-
mentation (Chatterjee et al. 2007) if there is no f , the ratio of outer-cell
interference to in-cell interference, namely the outer-cell interference is supposed
to be a constant independent of users’ transmitted (received) power or only the
single cell is taken into consideration. The expressions (44.11) and (44.12)
describe the examination of the receiver by individual user (Lee et al. 2002) when
residual power factor h equals 1. The two expressions give the same requested
received power as in documentation (Jorguseski 2001) if there is no f and residual
power factors are somehow different, and the same requested received power as in
documentation (Ivanek 2008) if f does not exit and residual power factor h is 0,
meaning, there is no perfect SIC for residual power after the interference of each
user in a single cell is eliminated.
44.3 Decoding Order for Optimal Outage Performance
Definition 1 In given decoding order, if the constraint Eqs. (44.7) and (44.11) can
not be satisfied, the condition is called the outage of the system in the decoding
order.
432 P. Zhang and J. Min
If there are K users in the system, there may exist K! decoding orders, and the
system has different outage in performance in different orders. In case the trans-
mitted (received) power is limited, the research on optimal decoding order will be
helpful to minimize the outage probability for the design of better system meeting
actual requirement.
We can conclude the following inferences based on the Definition 1.
Inference 1: Among all possible decoding orders, the system can get the
minimum outage probability in ZD (Zs Descending) order.
Demonstration: According to Definition 1, once the system is practicable, the
constraint Eqs. (44.7) and (44.11) can be realized in decoding order ZD, even if the
equations are not practicable in other decoding orders. Thus, we can conclude that
among all possible decoding orders, the system can get the minimum outage
probability in decoding order ZD. Q.E.D.
To comprehensively and intuitively understand the rationality of decoding
order ZD in system minimum outage probability, we can consider the following 2
special cases:
(1) All the users have the same rate and corresponding Eb/I requirements, namely,
all the Y i ¼ W=ðRi ci Þ results are the same. In such case, decoding order ZD is
corresponding to the decoding order confirmed by the descending order Pmax i
The later a user is decoded, the smaller degree other users are interfered, and
the lower demand the received power requires based on Eqs. (44.7) and
(44.11). So the constraint Eqs. (44.7) and (44.11) are most likely to be satisfied
in the decoding order confirmed according to the Pmax i descending order which
is adopted by the K users. Furthermore, if all the users have the same maxi-
mum transmitted power, then the decoding order confirmed according to the
Pmax
i descending order is corresponding to the decoding order confirmed
according to the descending order based on the user channel gain.
(2) All the users have the same maximum received power Pmax i . In such case, the
decoding order ZD is corresponding to the decoding order confirmed by the
descending order based on Y i ¼ W=ðRi ci Þ. In the same way, according to Eqs.
(44.7) and (44.11), the bigger the Y i ¼ W=ðRi ci Þ of a user is, the lower
requirement of the corresponding SINR is. Meantime, the earlier a user is
decoded, the more interference it will get from other users. Thus the Eqs.
(44.7) and (44.11) probably be satisfied in the decoding order of all the users
confirmed by the descending order based on Y i ¼ W=ðRi ci Þ. Further, if all the
users have the same Eb =I requirement, the decoding order confirmed by the
descending order based on Y i ¼ W=ðRi ci Þ is corresponding to the decoding
order confirmed by the ascending order based on Ri.
44 Power Control of Cellular System 433
44.4 Decoding Order Fro Minimum System Gross
Transmitted Power
If there are K users in the system, there will be K! possible decoding orders. And
in different decoding order, the gross transmitted power of the system is different.
Low transmitted power is helpful to prolong the service time of the mobile ter-
minal battery and extend its lifetime, and reduce the electromagnetic radiation
pollution to environment. So it is necessary to study on the optimal decoding order
for the minimum gross transmitted power. Though documentation (Saghaei and
Neyestanak 2007) has considered the combination of the CDMA system in SIC in
minimum gross transmitted power, its objects are mainly single cells and it
assumes that the interference power from outer cells is a constant not dependent on
user transmitted (received) power. In the cellular system, the cells are not isolated,
and the change of the in cell’s transmitted (received) power can influence that of
outer cells. Even though documentation (Jorguseski 2001) has considered the
interference of outer cells, it is only based on perfect SIC.
This paper analyzed the question of how to adjust the decoding order for a
minimum gross transmitted power in multi-cell condition which assumes that the
outer-cell interference is dependent on the user transmitted (received) power and at
the same time the SIC is not in perfect condition.
44.4.1 Theoretical Analysis
Based on the Theorem 1 mentioned above, we can prove and obtain the following
theorem and determine how to adjust decoding order for the total transmitted
power minimization according to the following theorem.
Theorem 2 The decoding order is identified pursuant to the descending sort of
users’ channel gains, and then the transmitted (received) power will be distributed
under the decoding order according to expressions (44.10), (44.11) and (44.12).
Consequently, the total transmitted power can be minimized under the premise of
satisfying the user’s Eb/me requirement.
Demonstration: we suppose that the users A and B are respectively the Lth and
(L ? 1)th with decoding detected and for the user A0 s channel gain hA is lower
than B0 s hB, that is hA \ hB. Therefore, we investigate the changes of total
transmitted power fore and after exchange of A0 s and B0 s decoding orders. It is
supposed that during the exchange, both users’ received powers are respectively Pi
(i = 1, 2, …, K) and P~ (i = 1, 2, …, K) under expressions (44.10), (44.11) and
(44.12). So we consider the following three conditions to facilitate the analysis:
(1) L = 1 means that A is the first user getting decode detected among the
original decoding order. And the rest users’ received powers remain unchanged
except A and B before the exchange of decoding orders of A and B according to
434 P. Zhang and J. Min
Theorem 1, that is, Pi ¼ P ~ i , i C 3. Correspondingly, the difference of total
transmitted power before and after exchange of decoding orders of A and B is as
follows
XK XK
Pi =hi P ~ B hB þ P
~ i hi ¼ ðPA =hA þ PB =hB Þ P ~ A hA
i¼1 i¼1
ð1 þ Y3 Þð1 þ YB Þ ð 1 þ Y3 Þ ð1 þ Y3 Þð1 þ YA Þ ð1 þ Y3 Þ
¼ P3 hA þ hB hB hA
ðh þ YB Þðh þ YA Þ ð h þ YB Þ ðh þ YA Þðh þ YB Þ ð h þ YA Þ
ð1 þ Y3 Þ ð1 hÞ
¼ P3 ð1=hA 1=hB Þ [ 0
ðh þ YA Þ ðh þ YB Þ
ð44:13Þ
That is, the total transmitted power is reduced after the exchange.
(2) L = K-1 means B is the last user getting decode detected among the
original decoding order. And the rest users’ received powers remain unchanged
except A and B before the exchange of decoding orders of A and B according to
Theorem 1, that is, Pi ¼ P ~ i , i B K-2. Correspondingly, the difference of total
transmitted power before and after exchange of decoding orders of A and B is as
follows
XK XK
Pi =hi ~ i hi ¼ ðPA =hA þ PB =hB Þ P
P ~ B hB þ P
~ A hA
i¼1 i¼1
ðh þ Yk2 Þ ðh þ YK2 Þðh þ YA Þ ðh þ Yk2 Þ ðh þ Yk2 Þðh þ YB Þ
¼ Pk2 hA hB hB hA
ð1 þ YA Þ ð1 þ YA Þð1 þ YB Þ ð1 þ YB Þ ð1 þ YB Þð1 þ YA Þ
ðh þ Yk2 Þ ð1 hÞ
¼ Pk2 ð1=hA 1=hB Þ [ 0
ð1 þ YA Þ ðh þ YB Þ
ð44:14Þ
So the total transmitted power can be lowered after the exchange.
(3) If 1 \ L\K-1, A is the first user with decode detected among the original
decoding order while B is not the last user. Therefore, the rest users’ received
powers remain besides A and B before the exchange of decoding orders
according to Theorem 1, that is, Pi ¼ P~i , i B L-1 or i C L+2. Similarly to
above-mentioned (1) and (2), the total transmitted power can be lowered after
the exchange.
From the above three conditions, the requirements of total transmitted power
can be lowered after the exchange of decoding orders for adjacent users A and B.
Repeat the adjacent users’ exchange of decoding orders until the decoding order
is the one identified by descending order of users’ channel gains. Based on the
above-mentioned three conditions, the requirement of total transmitted power can
be continuously lowered during such exchange process of decoding orders of
users. Thus, the Theorem 2 is established. Q.E.D.
In this paper, gains descending (GD) means the decoding order identified
through the descending order of users’ channel gains. The users’ powers will be
distributed under such decoding order according to expressions (44.10), (44.11)
and (44.12). Consequently, the total transmitted power can be minimized under the
premise of satisfying the user’s Eb/I requirement.
44 Power Control of Cellular System 435
In addition, although the above-mentioned conclusions are the same to the
documentation (Ivanek 2008), the author takes account of SIC under non-perfect
conditions. Moreover, the outer-cell interference is no longer supposed to be
independent from the users’ transmitted (received) powers while this paper aims at
multi-cell conditions. So such model is different from the existing model.
44.4.2 Simulation Experiment and Result Analysis
The following simulation experiment verifies the minimization of the system’s
total transmitted power under GD decoding order.
Figure 44.2 shows the change curves of the total transmitted power require-
ments versus the number of users under different decoding orders, among which
the number of users changes from 30 to 60 and the residual power factor h remains
0.1. The x-coordinate means the number of users while the y-coordinate the total
transmitted power in Fig. 44.2. Figure 44.3 shows the change curves of the total
transmitted power requirements versus residual power factor h under different
decoding orders, among which the residual power factor h changes from 0.05 to
0.5 and the number of users remains 36. The x-coordinate means the residual
power factor h while the y-coordinate the total transmitted power in Fig. 44.3. In
this section, gains ascending (GA) is the decoding order determined by ascending
order of users’ channel gains, rates descending (RD) is the decoding order
determined by descending order of users’ rates, and rates ascending (RA) is the
decoding order determined by ascending order of users’ rates.
It can be known from Figs. 44.2 and 44.3 that the total transmitted power
requirement is the lowest under GD—decoding order identified by descending
order of users’ channel gains. This conclusion is consistent with theoretical analysis
Fig. 44.2 Change curve of
total transmitted power
versus number of users
436 P. Zhang and J. Min
Fig. 44.3 Change curve of
total transmitted power
versus residual power factor h
of Sect. 44.1. In addition, it is also highlighted from Figs. 44.2 and 44.3 that the
requirement of total transmitted power grows in direct proportion to the number of
users or of the residual power factor h, which conforms to actual situation.
44.5 Conclusion
This paper takes account of the relevant power control in cellular system with
CDMA technique in combination with SIC. The expression of optimal power
distribution is firstly deduced under the given decoding order; and then, the optimal
decoding order is proposed to minimize the outage probability in the system if there
are limitations to users’ transmitted (received) powers; finally, the optimal
decoding order is deducted to minimize the total transmitted power. The simulation
compares the total transmitted powers of the system under different decoding
orders, and the results support the analysis of this paper.
Acknowledgments The research work in this paper is supported by Research on the Technology
of Load Balancing on Server Cluster under the grant 2012KJ05.
References
Chatterjee M, Lin HT, Das SK (2007) Rate allocation and admission control for differentiated
services in CDMA data networks. IEEE Trans Mob Comput 6(2):179–197
Gu et al. (2005) QoS based outer loop power control for enhanced reverse links of CDMA
systems. IEEE Electr Lett 41(11):659–661
Ivanek F (2008) Convergence and competition on the way toward 4G. IEEE Microw Mag
9(4):6–14
44 Power Control of Cellular System 437
Jantti R, Kim SL (2000) Second-order power control with asymptotically fast convergence. IEEE
J Sel Areas Commun 18(3):447–457
Jorguseski L (2001) Radio resource allocation in third generation mobile communication
systems. IEEE Commun Magaz 39(2):117–123
Kwok YK, Lau VKN (2003) System modeling and performance evaluation of rate allocation
schemes for packet data services in wideband CDMA systems. IEEE Trans Comput
52(6):804–814
Lee JW, Mazumdar RR, Shroff NB (2002) Downlink power allocation for multi-class CDMA
wireless networks. Proceedings of IEEE INFOCOM, vol. 3. pp 1480–1489
Liao CY, Wang LC, Chang CJ (2007) A joint power and rate assignment algorithm for multi-rate
soft handoffs in mixed-size WCDMA cellular systems. IEEE Trans Veh Technol
56(3):1388–1398
Paulraj AJ, Gore DA, Nabar RU (2004) An overview of MIMO communications—a key to
gigabit wireless. Proc IEEE 92(2):198–218
Saghaei H, Neyestanak AAL (2007) Variable step closed-loop power control in cellular wireless
CDMA systems under multi-path Fading. IEEE Pacific Rim conference on communications,
computers and signal processing. PacRim 2007. IEEE Press, pp. 157–160
Zahariadis T, Doshi B (2004) Applications and services for the B3G/4G era. IEEE Wirel
Commun 11(5):3–5
Chapter 45
Research of Embedded Intelligent
Decision Support System
for Natural Disasters
Jiang Shen, Tao Li and Man Xu
Abstract Based on the requirement of real-time, accuracy and stability of natural
disasters emergency interact decision support system, the essay does research on
the structure of natural disasters’ decision-making support system, realization of
information sharing platform and key technology of embedded system, establishes
fuzzy lookup, connective-administration workstation and bus structure and system
support structure, etc. Then the author ties the interfaces of the decision data layer,
the logic layer and the application layer to the whole emergency system interface,
and establishes embedded intelligent decision support system for natural disasters
which achieve the optimum combination of hardware and software and informa-
tion resources sharing and emergency command in natural disaster emergency
system. Finally, the paper discusses the system technology performance index and
taste tests method.
Keywords Natural disasters Embedded system Intelligent decision support
system Information sharing
45.1 Introduction
With the higher level of urbanization, function becomes more complex, the
potential problems of natural disasters is more and more, human beings’ depen-
dence on energy supply and the city configuration is increasing and system
J. Shen T. Li
College of Management and Economics, Tianjin University,
Tianjin, People’s Republic of China
M. Xu (&)
TEDA College, Nankai University, Tianjin 300457, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 439
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_45,
Springer-Verlag Berlin Heidelberg 2013
440 J. Shen et al.
security and reliability of the machines which work in cities is required much more
higher. In 2010, a severe storms triggered landslide in Zhouqu County, Gansu. In
2008, our country suffered a rare snow disaster and the Wenchuan earthquake. All
about these show that our country’ natural disaster integrated emergency response
system have many shortcomings. Based on Bayes Consulting Co., Ltd.’s research,
with the perfection of natural disasters emergency system, share of investment of
Chinese informatization is increasing. For example, some great cities and
municipalities such as Beijing invest an average of 100 million RMB in the natural
disaster integrated emergency response system and follow-up system’s upgrade
and engineering investment will reach 100 million RMB in five years. So the next
five years will be a large-scale construction period of emergency support platform.
As the further development of the natural disaster emergency application market,
the proportion of software and services will sharp increase which produces larger
market development space. Decision support system is the new market demand in
recent years on the basis of comprehensive emergency (UNDP 2004; Apel et al.
2004; Arnell 1989).
The construction of integrated emergency response and intelligent decision
support system for natural disasters will contribute to the comprehensive efficiency
and benefits of existing emergency system, which make the whole system produce
more social benefits and economic benefits and help information sharing function
do much better. The study establishes an embedded decision support integrated
system for natural disasters and realizes information sharing and emergency
command of natural disasters emergency system on the basis of the research on
embedded information sharing system and embedded support system and intelli-
gent decision support platform. The study will promote the process of natural
disaster emergency management research, and improve the level of natural disaster
management research and emergency response (Blaikie et al. 1994). Also, it may
make the daily management countermeasure and emergency plan system and
disaster mitigation plan on natural disasters for related administrative departments,
and provide scientific proof and technical support for disaster preparedness and
response and urban development planning.
45.2 Situation at Home and Abroad and Technology
Development Trend
45.2.1 Embedded Information Sharing System
Embedded system is a kind of computer system in which software links with
hardware tightly. The hardware bases on embedded microprocessor and integrates
storage systems and all kind of input/output devices. The software contains set-up
procedure, driver, embedded operating system and application program. The
organic combination of the software constitutes the integration software, which has
45 Research of Embedded Intelligent Decision Support System 441
many characteristics such as effective combination of hardware and software,
solidified source, hard real-time and high reliability. Dunkels et al. do research on
how to realize embedded network communication protocol stack by combination
of embedded system of limited resources, software-hardware environment and
network application. Geoff et al. propose a lightweight, flexible, system software-
oriented OPENCOM embedded component models. Bruneton et al. propose a
Fractal model which is independent of programming language and any system and
running services. The model can use Fractal component in low software layers and
within operating system. On the existing foundation, embedded system includes
two research directions: one is real-time operating system (32 bit embedded
microprocessor chips take over 4 bit, 8 bit, 16 bit microprocessor chips gradually),
another is Internet-oriented system (it combines embedded technology and internet
technology), which is developing to be a portable, cuttable, high real-time, highly
available network platform and standard.
Field bus structure and Ethernet structure has commonly used to automatic
control field in the aspect of embedded information sharing structure. We can use
different LAN protocols according to different application such as TCP/IP,
NETBIOS, NOVELL, IPX/SPX, Appletalk, SNA, Banyanvines and so on which
can achieve to link LAN/LAN with LAN/Internet well. The fieldbus is commu-
nication network which is the connection between field equipment and control
device in bi-directional, serial, digital, multi-node way. This open control network
adapts to the decentralized, networked, high-speed direction. The international
fashion fieldbus mainly contains: FF of fieldbus foundation, Profibus of German,
Lon works of Echelon, CANBUS of Bosch and Dupline of Sweden and so on.
Fieldbus Control System is taking over closed DCS System gradually to become
the major in process control. However, it should further to construct the common
interface and structure mode to adapt to two heterogeneous platforms to improve
the system compatibility and the reconstruction in key technology.
45.2.2 Intelligent Decision Support Platform
The following are the foreign exemplary decision support system: Portfolio
Management System, which supports for investors to manage customers’ securi-
ties, has the following functions: stock analysis, securities processing and classi-
fication; Brand aid, a hybrid market model which is used to product promotion,
product pricing and advertising decisions, can help the manager and management
analyses strategies and make decisions by connecting goods sales, profits with the
manager’s action; Projector helps the manager to construct and explore the
methods to solve problems, which supports enterprise’s short-term planning.
Geodata Analysis and Display System, developed by IBM Research, is used to
aided design of police patrol routes, urban design and school jurisdiction scope
arrangement. Ford did research on a decision support system for the flood disaster
warning decisions; William et al. constructed spatial decision support system to
442 J. Shen et al.
decide dangerous chemicals transport routes; Ezio tried to utilize real-time deci-
sion support system to manage natural disaster risk. The following are domestic
application and research: The government macro-economic management and
public management issues; The water resources allocation and early warning
system for flood control; Industry planning and management and all kinds of the
development and utilization of resources; Decision-making of ecological and
environmental control system; Natural disaster management.
Zhongtuo Wang described the decision support system by functionality. During
his analysis, the experience of the decision makers penetrated the mathematical
model of the decision support system; Sprague et al. proposed decision support
system three parts structure, namely dialogue component, data component (data-
base and database management system) and model component(model base and
model base management system); Bonczak et al. proposed DSS three system
structure, namely Language system, problem processing system and knowledge
system. Now, Aided decision-making ability of decision support system develops
from single decision model to more comprehensive decision-making model; By
combining decision support system and expert system, a decision support system
combining intelligent decision support system (Yong et al. 2007), qualitative aid
decision making and quantitative aid decision making is developed.
45.3 The General Framework of Integrated System
This research studies the structure of the decision support system for natural
disasters, the realization of information sharing platform and the key technology of
embedded systems such as interface issues (microprocessors, scalable RAM and
ROM, the data latch and bus coprocessor) to build a fuzzy lookup table,to connect
the management station and the bus architecture, system software and the
knowledge of the natural disasters, information resource database integration, and
system support structure. In this study, we use embedded technology to embed
information fusion technology integration and information-based decision support
system into the entire emergency management system based on interface tech-
nology, which achieves the optimal combination of hardware and software and the
sharing of information resources of natural disaster response system and emer-
gency command. Through the embedded technology, we connect data layer, logic
layer and application layer of decision-making platform with interface of the entire
emergency response system to enhance the efficiency and effectiveness of the
entire complex system that is to optimize the design, faster system responsiveness,
reduce the consumption of system resources and reduce hardware costs, while
enabling a variety of resources for the entire emergency response system to share
information (Yong et al.2007; Sun et al. 2007).
The overall framework of natural disasters embedded decision support inte-
grated system constructed in this study includes the following five levels.
45 Research of Embedded Intelligent Decision Support System 443
• The presentation layer. To provide a classification of the client as the user
interface part, for the different object needs.
• Computing layer. Composed by multiple servers, the object may demand
extended to N-tier structure. The server is responsible for receiving the specific
data of the main online information source, the monitor image signal (Fig. 45.1).
• Data layer. Architecture DB-ODS-DW, and the establishment of the DBMS and
the DLA data model.
• Control layer. Use RS-232 standard, and instead of hard-wired to wireless serial
communication.
• Embedded layer. Including multiple terminals embedded chip and external
devices, such as information collection, sensors, etc.
Higher ......
Internet
Database Database
Firewall server server
Controlling
signal
...... Data signal
Natural disaster
Supervisory emergency
control Decision-making Server Server Embedded decision
desk plan support platform
Controlling Data
signal signal
Ethernet
Data
Controlling signal
signal
Linkage unit ......
emergency system
emergency system of Professional Monitoring and Emergency Other system and
the county emergency warning system support system emergency resources
government / committee System of
Weiban Board
Controlling Data signal
signal
CPU 412 -2DP Extension rack
PROFIBUS Bus
Remote IO Remote IO AI
DI module module 431 -1KF10
ET200 M ET200 M Transducer
421 -1BL00
Button, Discrete input
input Controlling Controlling BP
operation signal 1 signal 1 ECG
box
AO
DO Controlling Controlling
signal N signal N 432 -1KF10 Signal
422 -1BL00 Discrete conditioning
output
Instructions, Distributed control system 1 Distributed control system N
relay
AI/AO System
DI/DO System
Fig. 45.1 Emergency interact embedded decision support integration system framework for
natural disasters
444 J. Shen et al.
45.3.1 The Main Hardware
The CPU uses Siemens CPU416-2DP. Switch DI template uses the 421-1BL00 for
field devices such as switches, operating box I/O points of access. DO template
uses the 422-1BL00. Both are 24VDC*32CH-type template. The use of interme-
diate relays to control equipment used to indicate the unit to direct output. Analog
template AI uses the 431-1KF10, which is use to access for instrumentation
monitoring signals, such as pressure sensors. The AO templates use the 432-
1KH10, in which channel type can be any configuration for controlling the reg-
ulator valve 9+. Ethernet card with the CP443-1 communicates over twisted pair.
Image there are two racks: One is central rack and other one is expansion rack. We
use template Interface IM460 to connect central rack and expansion rack and
utilize the CPU built-in PROFIBUS-DP interface to achieve communication with
the inverter. Then we use expansion unit—M communication card to connect to a
remote station ET200 M. The system’s communication networks uses Siemens
high speed industrial Ethernet the Culan connection for communication between
the PLC and the PLC, in which the communication speed reaches 10 Mb/s. We use
PROFIBUS-DP, one of the international fieldbus standards, and inverter, used to
communication and control and belongs to the device bus, for the complex field
devices and distributed I/O. The physical structure is the RS-485, the transmission
rate is up to 12 Mb/s. As we use the same protocol standards and transmission
media, and utilize ET200 M as remote I/O station, we should set ET200 M in a
location where there exist many I/O such as operating floor. In this way we can
reduce cable volume effectively. Twisted-pairline media and TCP/IP open pro-
tocol are used for communication between monitoring station and the PLC.
45.3.2 Embedded Support System
The hardware of the embedded support system takes the embedded microprocessor
as the core and integrates storage systems, and various input/output devices;
software contains the system startup programs, drivers, embedded operating sys-
tem, applications, etc.
(1) Embedded chip
Include embedded microprocessors, embedded microcontrollers, embedded
DSP processor and embedded on-chip system.
(2) Interface circuit of embedded system
We design or select the hardware interface according to the receipt signal that
need to deal with or the control signals that should be send to by the integrated
storage systems and a variety of input/output devices. These interfaces
include: embedded microprocessors, scalable RAM and ROM, the data latch
and bus coprocessor selection, design and field testing.
45 Research of Embedded Intelligent Decision Support System 445
(3) Management station and the bus structure
Realize Interconnection of simplicity, flexibility and versatility. Research and
test in the way of Bus communication provide more features and flexibility to
support multiple transport protocols and different types of services, develop
API functions and class libraries, and reduce on-site hardware and software
requirements.
(4) Embedded software system
Build embedded software which can be divided into embedded operating
systems and embedded applications.
(5) Generic component model
To meet embedded systems’ non-functional constraints, such as real-time
features, reliability, and stability. We combine the characteristics of the gen-
eral component technology and embedded systems and use the embedded
component technology to build the embedded set up model. The general
component model includes the CORBA Component Model defined by the
OMG and ezCOM of KETAI Century Company.
45.4 Technical Performance Indicators
(1) Embedded system interface circuit
Microprocessor: 80C51 chips, by calling the data within the RAM and ROM
process to complete the encoding and decoding of data. 80C51 power con-
sumption: 120 mW; working environment -40 –+85.
Scalable RAM and ROM: in accordance with the procedures and fuzzy lookup
table status, 2716, 2764, 6116 or 6264. Extended memory is directly con-
nected by bus with a microprocessor, directly addressable to the MCU.
Data latch and bus coprocessor: the management of the address and data
signals on bus multiplexed bus clock cycle, the successful completion of the
data read-write operations. Chip 74LS138 or 74LS373.
(2) The establishment of the fuzzy lookup table
Build up a fuzzy lookup table whose addressing speed of up to 2–3 times the
ordinary INDEX addressing according to structure-based data storage of
Binary tree, so that the system response time reduced by an average of
40–50 ms.
(3) Management workstation and the connection of the bus structure
Using Sockets to achieve the connection of the Fieldbus Fieldbus and man-
agement workstation MS.
Transmission is completed through the Fieldbus module; the data type of the
third layer of the fieldbus specification sub-layer FMS package of different
devices to the user application layer UA provides a unified data type interface
in order to achieve interchangeability and interoperability of equipment.
Socket Transmission only transmits two parts of the data of physical Zhen
446 J. Shen et al.
(one is FMSPCI and other one is UED) to improve the transmission efficiency
of the line. The following table is the form of data Zhen of physical layer.
PRE SD DLL PCI FAS PCI FMS PCI UED FCS ED
PRE: synchronization sequence
SD:Starting defined operator
DLL PCI:DLL protocol Information Mining
FAS PCI:FAS protocol control information
FMS PCI:FMS protocol Information Mining
UED:The user coding data
FCS:frame check sequence
ED:Final value defined operator
(4) Embedded application software based on network
(1) Enhanced TCP protocol, propagation delay time: \20 ms
(2) Improved interface RS232 of external device, propagation delay
time: \5 ms
(3) Maximum number of concurrent TCP session: [50
(4) Correct rate for data transmission: [98 %
(5) Embedded module
(6) Operating system: Embedded operating system based on WINDOWS
45.5 Taste Tests Method
According to the characteristics of the embedded system, a system test should be
done after the completion of the preliminary system development. We choose
dynamic test method including control flow testing, data flow testing, test, function
test points domain and random testing. The following are the related test content:
(1) Current test
(2) Compatibility test
Hardware compatibility, BIOS compatibility, operating system compatibility,
the corresponding software with PC compatibility should also be taken into
account.
(3) Reliability evaluation
(1) Maturity measure
False Discovery Rate DPP (Defect Detection Percentage)
` Test coverage measure
45 Research of Embedded Intelligent Decision Support System 447
(2) Fault tolerance evaluation
(3) Control the fault tolerance measure
The control ability of concurrent processing
` Error’s quality of revision and the ability to deal with
(4) Data fault tolerance measure
Tolerance of the illegal input data
` Tolerance of the requirements of the conflict and the illegal combinations
´ Tolerance of the rationality of the output data
(5) Tolerance of hardware’s recovery measure
(6) Tolerance of the ability to recovery
(4) Recovery measure
(1) Recovery of space intensity
(2) Recovery of time intensity
(3) Recovery of data intensity
(4) Recovery of abnormal communications
(5) Recovery of data breach
(6) Recovery of battery limit
45.6 Conclusion
This article uses embedded technology to connect the data level, logical level and
application level of the decision-making system with the interface of the whole
stand-by system and meanwhile form an integrated decision-making support
system of natural disasters. This can improve the efficiency and outcome of the
whole complicated system, that is, it can optimize the design, raise its responding
speed, reduce its resource consumption and its costs of hardware. In addition, it
can also realize the information sharing of its various resources in the whole stand-
by system.
This article will contribute to a raise in both the response to natural disasters
and the anti-risk abilities. It will also provide better services, that is, the stand-by
system, embedded supporting system and information-sharing platform, to insti-
tutions including government offices, legislation, tax, police, administration, flood
prevention and so on.
Acknowledgments This work was supported by the National Natural Science Foundation of
China (71171143), by Tianjin Research Program of Application Foundation and Advanced
Technology (10JCYBJC07300), and Science and Technology Program of FOXCONN Group
(120024001156).
448 J. Shen et al.
References
Apel H, Thieken AH, Merz B et al (2004) Flood risk assessment and associated uncertainties. Nat
Hazard Earth Syst Sci 4(2):295–308
Arnell NW (1989) Expected annual damage and uncertainties in flood frequency estimation.
Water Res PI Manag 115:34–107
Blaikie P, Cannon T, Davis I et al (1994) People’s vulnerability and disasters. Routledge,
(Natural Hazards) London pp 189–190
Sun XD, Xu XF, Wang G et al (2007) Multi-objective optimization of process based on resource
capability. J Harbin Inst Technol 14(4):450–453
UNDP. Reducing disaster risk: a challenge for development. John S. Swift Co., USA. (2004)
www.undp.org/bcpr
Yong H, Jia Xing H, Ju Hua C (2007). Software project risk management modeling with neural
network and support vector machine approaches. In: Third international conference on natural
computation, Hainan, pp 358–362
Chapter 46
Research on Classifying Interim Products
in Ship Manufacturing Based
on Clustering Algorithm
Lin Gong, Yan Song, Hao Lin and Zi-xu Chen
Abstract This paper first reviews some methods of classifying interim products
during shop manufacturing. Then, the general cluster analysis is present. The
progress of classifying interim products based on cluster analysis is given, and
seven features of interim products are present to analysis. Finally, a case is cal-
culated to show the effective of the method, which provides a useful way for
effective classification of the intermediate products in the shipbuilding process.
Keywords Classifying Clustering algorithm Interim products Ship
manufacturing
46.1 Introduction
The shipbuilding industry is a strategic industry which relates to national security
and national economic construction. In 2011, shipbuilding capacity in China was
76.65 million Deadweight tons (DWTs), orders for new ships reached 36.22 million
DWTs, and by the end of December, handheld shipbuilding orders was 149.91
million DWTs. Chinese shipbuilding capacity, new orders received and handheld
orders accounted for 45.1, 52.2 and 43.3 % of world market share (http://
zbs.miit.gov.cn/n11293472/n11295142/n11299228/14428463.html, http://wenku.
baidu.com/view/0ec320befd0a79563c1e7203.html.) At present, ship production
model for assembly, as the representative of modern shipbuilding mode, is
L. Gong (&) Y. Song H. Lin Z. Chen
School of Mechanical Engineering, Beijing Institute
of Technology, Beijing, China
e-mail: [email protected]
Y. Song
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 449
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_46,
Springer-Verlag Berlin Heidelberg 2013
450 L. Gong et al.
vigorously in our shipyard. Ship production model for assembly includes segmental
construction and pre-outfitting on the basis of the group technology. Using interim
products oriented, it combines the hull construction, outfitting and painting work,
which makes modern shipbuilding mode of hull, outfitting and painting integration.
In this process, the traditional decomposition mode which is in system-oriented
must be changed to the mode which is dispersed specialized production of interim
products-oriented. This mode can allocate resources rationally, which means every
production risk can be allocated to the corresponding production resources. The
key issue of it is the classification of interim products by product characteristics
and production characteristics.
This paper first reviews some methods of classifying interim products during
shop manufacturing. Then, the general cluster analysis is presented. The progress
of classifying interim products based on cluster analysis is given, and seven fea-
tures of interim products are present to analysis. Finally, an instance is calculated
to show the effective of the method, which provides a useful way for effective
classification of the intermediate products in the shipbuilding process.
46.2 Research Background
Classification of the intermediate products rely on the experience of qualitative
cannot meet the requirements of the modern ship production technology. It must
use mathematical methods to do quantitative classification work (Wu et al. 2008
Liu and Liang 2011). At present, there was some classification study of the interim
products of the shipbuilding system. Zhong Yuguang has proposed a classifier of
ART-2 artificial neural network on the basis of the Adaptive Resonance Theory. It
can automatically classify and identify the input data by analyzing the character-
istics of interim products in shipbuilding (Zhong et al. 2008). Zhong Hongcai
presented an interim products characteristics measurement method and used cluster
analysis method to classify sub-assemblies (Zhong et al. 2003a, 2003b). In addition,
Wang Shili designed a hull piece coding system based on modern shipbuilding
mode, which focuses on project number, block number, assembly coding, piece
kind, piece number and piece bending kind configure, etc. (Wang et al. 2006).
The cluster analysis method is simple and effective. Using the cluster analysis
method, the interim products can be classified in quantitative division on the basis
of the digital characteristics of interim products. The research on the interim
products classified using of cluster analysis methods is given in this paper.
46.3 General Cluster Analysis Methods
Cluster analysis is a technology which focuses on finding the logical or physical
relationship between data. The data set is divided into several classes of similar data
points in nature by certain rules. The results of cluster analysis not only can reveal
46 Research on Classifying Interim Products in Ship 451
the intrinsic link and differences between the data, but also provides an important
basis for further data analysis and knowledge discovery (Jain et al. 1999).
46.3.1 Cluster Analysis Methods
Cluster analysis methods include the following:
AHP: AHP is through the decomposition of a given set of data objects to
create a hierarchy. Two types of hierarchical decomposition of the formation of
hierarchical methods can be divided further into cohesion (Agglomerative) and
division (Divisive).
`Divisions method: Given a data collection contained n objects, Divisions
method is to build data k divided, each division represents one cluster, and k B n.
Moreover, these groups meet the following conditions: each group contains at least
one object; each object must belong to and only belongs to one group (the
requirements may be relaxed in the fuzzy partition techniques). For a given k, this
method first creates an initial partition. Then the iterative method to change the
grouping makes the group every improvement program than before. Good division
of the criteria is: the distance in the same grouping as close as possible to different
groups, the greater the distance, the better.
´Density-based method: As long as the density of objects or data points in a
region is more than a certain threshold, then continue to clustering. Each data point
in a given class in the area of a given range must contain at least a certain number
of points.
ˆGrid-based method: The grid-based method divide the data space into a grid
structure of a finite number of units, all the processing is a single unit. The main
advantage of this method is the processing speed; the processing time is only
depend on the number of units quantify of the space in each dimension, no rela-
tionship to the event the number of data objects.
˜Model-based method: Model-based method is to assume a model for each
class to find the best fit of the given model. Such a model may be the probability
density function or other functions of the data points in space. Its underlying
assumption is: the target data set is determined by a series of probability distri-
bution. A model-based algorithm may locate the cluster by constructing a density
function reflected the data points in the spatial distribution. Representative model-
based methods include statistical method and neural network method.
In addition to the above five categories, there are a large number of clustering
methods, such as dealing with high dimensional data clustering method (Strehl and
Ghosh 2003), to deal with dynamic data clustering method, based on genetic
algorithm clustering method, as well as basic clustering method a variety of new
technology combined with the clustering method.
452 L. Gong et al.
46.3.2 The Related Concepts of Cluster Analysis
m samples to be the clustering can be described by n indicators and characteristic
values. This constitutes an m n data set or data matrix Xmn , the matrix elements
is xij , that is value of j characteristic index of sample i (i = 1,2,…,m;
j = 1,2,…,n).
2 3
x11 x12 x1j x1n
6 x21 x22 x2j x2n 7
6 7
6 .. .. .. .. .. .. 7
6 . . . . . . 7
X¼6 6 xi1 xi2 xij xin 7
7
6 7
6 . .. .. .. .. .. 7
4 .. . . . . . 5
xm1 xm2 xmj xmn
Sample i can be expressed by vector ! xi , ! xi ¼ ðxi1 ; xi2 ; ; xin Þ, This vector is
called the pattern vector of the sample i.
The sample is the classification of objects of a single individual. In classifying
subsections, each small segment is a sample.
The class is a collection of similar samples. The class not only can contain one
or more samples, but also can contain all the samples.
The similarity coefficient or distance is to describe the similarity degrees of
statistic between two samples or two classes, or between one sample and one class.
Distance is used to stress difference between classes or samples, but the similarity
coefficient is used to stress the similarity.
The distance can also be used to describe the degree of similarity, which
includes Euclidean distance, Minkowski distance, Mahalanobis distance, and the
distance of the mixed data set.
Two dimensional Euclidean distance of expression is:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u 2
uX 2
Dkl ¼ t xkj xlj ð46:1Þ
j¼1
k and l are two samples, xkj, xlj are value of sample k and l on dimension j, the
distance between them is Dkl .
N-dimensional expression is:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX
u n n
Dkl ¼ t xkj xlj ð46:2Þ
j¼1
Because dimensions are different in the mixed data set, the data should be
regulated before the distance calculation. Generally, the following methods for
data regulation are use.
46 Research on Classifying Interim Products in Ship 453
There are two samples k and l and the index j, the coefficient a j ðk; lÞ is defined
in the following:
If j is a binary variable:
j 0ðK and 1 values is sameÞ
a ðk; lÞ ¼ ð46:3Þ
1ðK and 1 values is not sameÞ
If j is a nominal index:
j 0ðK and 1is belong to the same classÞ
a ðk; lÞ ¼ ð46:4Þ
1ðK and 1is belong to the different classÞ
If j is an indicator of a sequence, set the value of j is desirable 1, 2, …, t:
x x
kj lj
a j ðk; lÞ ¼ ð46:5Þ
t
If j is an interval indicator or arithmetic index:
xkj min xfj xlj min xfj
f f xkj xlj
a j ðk; lÞ ¼ ¼ ð46:6Þ
max xfj min xfj max xfj min xfj max xfj min xfj
f f f f f f
Distance:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u X
u1 n
d ðk; lÞ ¼ t ½a j ðk; lÞ2 ð46:7Þ
n j¼1
46.4 Cluster Analysis of the Ship Manufacturing
Interim Products
46.4.1 The Segment Characteristics of Ship Manufacturing
Interim Products
In the ship manufacturing process, a variety of characteristics of interim products
(segmentation) can be chose, to describe before cluster analysis. According to the
actual production, the seven characteristics to describe the segmentation are: type,
welding length, pipe length, weight, projected area, height, and whether to turn
over. Of course, in the production practice, the shipyards can constantly add
characteristics for their own production style and need. In the above seven char-
acteristics, type and whether to turn over belong to the qualitative features, weight,
length of welding, pipe length, projected area, height are quantitative character-
istics (Song et al. 2010).
454 L. Gong et al.
(1) Type
Type is to describe the main types of segmentation from the shape, which can
be divided into flat segmentation, special flat segmentation, curved segmentation,
special curved segmentation and superstructure. Process route of different types of
sub-manufacturing are not the same.
(2) Assembly length
Assembly length means the connection length in part assembling,assembly
assembling, and segmentation assembling. The assembly length reflects the size of
the assembly workload to a certain extent.
(3) Welding length
Welding length is the welding length of puzzle, part, assembly, and segmen-
tation. In order to reflect the welding workload and degree of difficulty, it is
necessary to transfer welding length into the conversion length. This conversion is
derived by multiplying the coefficient. The coefficient of this effect sometimes are
multiple, such as the size and thickness of steel plates and profiles, weld type,
welding edge forms, working conditions and steel factors, etc.
(4) Weight
The segment weight determines the segmentation of the demand for production
equipment. Generally based on the lifting equipment and transport equipment
capacity, the weight is divided into several levels: less than 10 T, 10–30 T,
30–60 T, 60–100 T, etc.
(5) Projected area
The size of segmentation projected area reflects the size of the site area of the
segmented manufacturing production.
(6) Height
Segment height is the height of subsections and the base surface of the con-
nected construction, which decides whether need to install the framework in
welding connection components.
(7) Turn Over (Zhong et al. 2003)
Whether to turn over is an important production characteristic in manufactur-
ing. By turning over, some welding work in of the original overhead position
becomes down welding, which is lower in work difficulties. However, turn over
needs lifting equipment, at the same time, some assistive devices should be welded
for turning over. The process route including turn over has a huge impact in
manufacturing process.
46.4.2 Cluster Analysis Process
The cluster analysis process of interim products is shown in Fig. 46.1
The first step is to establish the characteristics index matrix. m interim products
are analysis by cluster analysis with n characteristic indexes. The following matrix
is created.
46 Research on Classifying Interim Products in Ship 455
Fig. 46.1 Cluster analysis
Start
flow chart of interim products
Input characteristic index X of interim
products, the number of groups N
Calculate the distance matrix of
interim products: D
Search the minimum in D: d min
Merge d min to mapping of Calculate the distance matrix D
interim productts between the groups
No
Determine whether the number
of groups is N
Yes
End
2 3
x11 x12 x1j x1n
6 x21 x22 x2j x2n 7
6 7
6 .. .. .. .. .. .. 7
6 . . . . . . 7
X¼6
6 xi1
7
6 xi2 xij xin 77
6 . .. .. .. .. .. 7
4 .. . . . . . 5
xm1 xm2 xmj xmn
The original distance matrix is calculated by calculate the distance between the
interim products:
2 3
0
6 d21 0 7
6 7
Dð0Þ ¼ 6 .. .. .. 7
4 . . . 5
dm1 dm2 0 mm
Find the minimum value dmin in D(0), the corresponding interim products are
merged into one class. Then, the distance between the class and other interim
products are calculated to get the new distance matrix D(1). The distance between
the new class which combines p and q, with the class r can be calculated by
weighted pair group method of the average distance.
d ðu; v þ wÞ ¼ ½dðu; vÞ þ d ðu; wÞ=2 ð46:8Þ
456 L. Gong et al.
Continue to find the minimum value of D(1), the corresponding interim
products are grouped. Repeat this process until the last interim products are
merged.
46.5 Instance Validation
Table 46.1 shows some ship segmentations. According to the actual production
needs, these six segmentations should be grouped into four classes for production.
Its original distance matrix D (0) is as follows:
2 3
0
6 0:89 0 7
6 7
6 0:63 0:56 0 7
6 7
6 0:30 0:78 0:59 0 7
6 7
4 0:57 0:81 0:51 0:69 0 5
0:63 0:62 0:28 0:67 0:52 0
Minimum value dmin = 0.28, so Section C, F segmentation are combined into
one class.
2 3
0
6 0:89 0 7
6 7
6 0:30 0:78 0 7
6 7
4 0:57 0:81 0:69 0 5
0:63 0:59 0:63 0:52 0
The minimum value dmin = 0.30, so Section A, D segmentation are combined
into one class.
At this point, these segments are divided into four groups: C and F, A and D, B, E.
Table 46.1 Output of four projects
Segment Segmentation characteristics
name
Type Weight(t) Assembly Welding Projected Height Turn
length (m) length (m) area (m2) (m) over
A Superstructure 72 320 671 73 9.7 No
B Special curved 45 173 347 90 3 Yes
segmentation
C Curved 63 265 513 77 5.5 Yes
segmentation
D Superstructure 60 247 499 65 8.6 Yes
E Special flat 78 352 721 79 7 Yes
segmentation
F Curved 54 303 589 89 8.3 Yes
segmentation
46 Research on Classifying Interim Products in Ship 457
46.6 Conclusion
This paper mentions a method to classify interim products in ship manufacturing
based on clustering algorithm. It is a useful way for effective classification of the
intermediate products, which means every production risk can be allocated to the
corresponding production resources. The further work is to refine the index system
and study other models of classifying to get a more accurate result.
References
Wu X, Kumar V, Ross Quinlan J, Ghosh J, Yang Q, Motoda H, McLachlan G, Ng A, Liu B, Yu P,
Zhou Z, Steinbach M, Hand D, Steinberg D (2008) Top 10 algorithms in data mining. Knowl
Inf Syst 14(1):1–37
Liu L, Liang Q (2011) A high-performing comprehensive learning algorithm for text
classification without pre-labeled training set. Knowl Inf Syst 29(3):727–738
Zhong Y, Qiu C, Xue K (2008) Research on classification of interim product families in
shipbuilding based on ART-2 network. Ship Build China 47(2):108–113 (in Chinese)
Zhong H, Jiang R, Tan J (2003a) Research on formation of interim product family in
shipbuilding. Ship Eng 25(3):57–61 (in Chinese)
Zhong H, Jiang R, Tan J (2003b) Interim products clustering method in shipbuilding. J Shanghai
Jiaotong Univ 37(8):1234–1236 (in Chinese)
Wang S, Liu J, Jin Y, Sun J (2006) Study and application of hull piece coding based on modern
shipbuilding mode. Shanghai Shipbuild 4:44–47 (in Chinese)
Jain AK, Murty MN, Flynn PJ (1999) Data clustering: a review. ACM computing surveys
(CSUR) 31:264–323
Strehl A, Ghosh J (2003) Relationship-based clustering and visualization for high-dimensional
data mining. INFORMS J Comput 15:208–230
Song X, Li B, Gong L (2010) Research on the resources management platform for the modern
shipbuilding enterprise. Ship Sci Technol 7:12–15
Chapter 47
Research on Financial Accounting
Information Disclosure of China’s Social
Security Fund Based on Game Analysis
Sha-sha Dai, Ke Pan and Yuan-yuan Dai
Abstract Due to the non-consistency of the interests and objectives, confrontation
and conflict will naturally occur between the accounting information providers of
social security fund and regulators of the social security funds, and consequently
the game will be generated. Applying the basic principles of game theory, the
author firstly analyzes the mixed-strategy game between the social insurance fund
management institutions and regulators; then conducts improvement and analysis
on game model by regulatory measures like joining the civil regulatory and
reporting; and finally concludes the relevant policy recommendations.
Keywords Basic social security fund Financial and accounting information
Information disclosed Game analysis
47.1 Introduction
The financial and accounting information disclosure of social security funds takes the
social security funds as the accounting entity, and it provides the information formed
through processing and sorting the financial accounting information of this entity to
users of financial statements, and makes it to the public, which is an important part of
the social security fund finance. As the purpose of social security funds operation is to
pursue self-balance, to ensure the normal balance of payments, and to achieve the
maintenance and appreciation of funds’ value; the primary purpose of the financial
information disclosure is to manage responsibilities so as to ensure that the owners of
the social security funds—the state, pay units and individuals are clear about the
S. Dai (&) Y. Dai
Business school, Hunan University, Changsha, China
e-mail: [email protected]
K. Pan
Bank of Communications Hunan Provincial Branch, Changsha, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 459
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_47,
Ó Springer-Verlag Berlin Heidelberg 2013
460 S. Dai et al.
operational status of the various funds. The posterior purpose is to provide decision-
relevant information to the users of financial statements. However, due to the con-
fusion of the decentralized management system of social security funds and the lack
of social security funds’ supervision, the funds are often diverted, misappropriated,
wasted and special funds are difficult to be used only for special purpose, which
reduces the efficiency of social insurance, and seriously impacts on security of social
security funds and continuous operation of the system. For example: Officials from
the Ministry of Labor and Social Security claimed that since 1998, the Ministry of
Labor and Social Security have carried out 5 inspections with other ministries, in
which they found that more than 170 billion Yuan was misappropriated in 2000 and
discovered 16 provinces are diverted and misappropriated in 2004. Especially the
Shanghai social security fund case and the Guangzhou social security fund misap-
propriation case in recent years have triggered common concern on the operational
and regulatory issues of social security funds. In the meantime, serious issues such as
the loose daily supervision of social insurance funds, the backward regulatory legal
system of the social insurance funds, over-dependence on the system of adminis-
trative supervision, the absence of social regulation, and the lack of medium-term
and the long-term early warning regulatory are exposed.
47.2 Literature Review
The publish of the classic paper Uncertainty and Welfare Economics written by
Kenneth Arrow in 1963 marks the establishment of Health Economics. This paper
discusses risk aversion, moral hazard, asymmetric information, externalities of
charitable actions and a lot of other issues that occupy important positions in the
later health economics research (Arrow 1963). Cho and Kreps (1987) study that a
rational person will choose a strictly dominant strategy in game. In 1991, Claire
et al. applied measurement techniques to measure the level of information
disclosure in financial reports and the effects have achieved by use of this mea-
surement technique till now (1991). Gong and Zhou (2006) believed that the
financial information disclosed have three basic theoretical models. Zhou (2001)
thinks each of social security funds is an accounting entity. In 2001, Wang
proposed that in-depth research must be done to improve the pay off vectors of
investors and listed companies in the game matrix and Nash equilibrium; while in
the game of government and listed companies, the supervision level should be
improved so as to control the can false accounting information disclosed by the
listed companies (Wang 2001). In 2002, Gou believed that the objective of
the social security accounting was to provide information users with accounting
information based on financial information and with certain qualitative charac-
teristics (Gou 2002). In 2003, Joseph and Terry examined the relationship between
the independence of the regulatory members and the information disclosure quality
of the company with financial difficulties, and proved that there was a positive
correlation between them (Joseph and Terry 2003). Wei et al. (2003) utilized game
47 Research on Financial Accounting Information Disclosure 461
theory to systematically analyze the attitudes and behaviors of the stakeholders
like operators, accountants, investors, certified public accountants, government
regulators in the process of disclosing the accounting information, pointed out the
various reasons for accounting information fraud, and raised some suggestions for
governance (Deng 2003). Deng (2003) put forward countermeasures and sugges-
tions on standardizing the surplus behavior of the listed companies in China from
the perspective of game theory, hoping to provide reference for the relevant
departments to establish and perfect the sound accounting system, accounting
standards and policies, laws and regulations of governance structure (Wei et al.
2003) In 2004, Du took the corporate governance as the background, accounting
development history as the evidence, and game theory as the analysis tools, he
focused on elaborating the formation process of institutional structure of the
accounting information disclosure regulation during the evolution process of
corporate governance (Du 2004). In 2008, Zhao put forwards some relevant
suggestions on strengthening the legal construction for supervising the social
security funds, the establishment of the supervision and budgeting systems for the
social security fund and improving the information disclosure system of fund
regulation (Zhao 2008). In 2010, based on the operations and regulation theories of
current social insurance fund, Li Wen-bo and Duan Hong-bo conducted in-depth
analysis and evaluation on the regulatory situation of Chinese social insurance
fund, and studied to establish social insurance fund monitoring system with the
characteristics of Chinese society from many aspects and based on China’s
national conditions (Li et al. 2010). In 2011, Tong and Liu established several
different game models, used game analysis method to study the social security
fund managers and the behaviors of the social security funds’ regulators, and
concluded the necessity of social security funds’ regulators and their impact on the
social security funds (Tong and Liu 2011).
In recent years, a lot of the articles apply game theory to explain the information
disclosure of the listed companies emerge, and those studies mainly conduct
research from the topics like information disclosure behavior, information disclo-
sure and supervision as well as information distortion and accounting irregularities.
However, few scholars have used game theory to explain the accounting information
disclosure of social security funds.
47.3 Theoretical Analysis
47.3.1 Analysis of Supply and Demand of Financial
Accounting Information of Social Insurance Funds
Financial
Accounting information as a public good also has supply and demand like other
products. Though the reason for the generation of demand and supply of
accounting information is that the information learned by the supply and demand
462 S. Dai et al.
parties is asymmetric, and asymmetric information leads to the imbalance of both
power and interests. Therefore, the demand party produces the need for accounting
information with its purpose of maintaining the balance of power and interests of
both supply and demand parties; while the supply party certainly wants to keep the
status of its information advantage to obtain additional benefits. Because of this,
both supply and demand parties will ultimately form the disclosure of financial
accounting information through continuous game. In the study of accounting
information disclosure, the demand party generally refers to the main users of
accounting information. However, from our analysis of the accounting goals, no
matter it is ‘‘accountability theory’’ or ‘‘decision availability’’, they all take the
separation of ownership from management as the premise. The agency relationship
between management authorities and property owners is the major contradiction in
the study of accounting information disclosure. Therefore, the demanders for
accounting information in this paper are researched, which mainly means to
provide accounting information to external information demanders in the form of
financial information reports.
According to our analysis, the providers of accounting information refer to the
party with the information superiority position in the game of supply and demand.
In general, it refers to the business owners as well as other internal managers.
Accounting information providers disclose the accounting information for two
motives; one is voluntary disclosure, the other is non-voluntary disclosure. The
supply party in the status of information superiority will not easily disclose
accounting information unfavorable to itself. In order to safeguard the healthy and
good economic order, the State will force accounting information providers to
provide the required information through administrative means such as developing
the accounting system and criteria by the government and relevant organizations.
47.3.2 Game Theory
Economist Nash’s solution to the classic game case of—the ‘‘prisoner’s dilemma’’
indicates that the premise of the agreement to be complied with is that the benefits
of complying with the agreement outweigh the benefits of breaking the agreement,
or the loss of breaking the agreement is greater than the loss of complying with the
agreement. Otherwise, the parties will not have the interest to comply with the
agreements. It is believed that game theory has great reference significance on
analyzing the selection of the parties between the financial and accounting
information disclosers on the social security funds and the regulators. Meanwhile,
the both game parties will induce the parties to produce the motive of breaking the
agreement for the results of their own interests pursuit. Accounting principles,
accounting standards and other related economic laws and regulations, are a kind
of agreement. How to ensure that this agreement is implemented, and effectively
restrains the concerned parties, especially the behavior of regulators and the
financial and accounting information disclosers? According to the basic idea of
47 Research on Financial Accounting Information Disclosure 463
the ‘‘Nash equilibrium’’, the key premises to achieve the ‘‘Nash equilibrium’’ are
the scientific and reasonable system design, strictly working in accordance with
the system, complying the system, and everyone is equal before in the system
(Zhang 1996).
47.4 Game Analysis
47.4.1 Game Analysis on the Accounting Information
Providers of Basic Social Security Fund
and Government Regulators
Assumption: 1. Game participants in the game model are rational, and they will
make the optimized rational decision under some constrained conditions.
2. Assumption that there are two kinds of actions for the government regulators to
choose: check or not check; two kinds of actions for financial and accounting
information discloser to choose: financial accounting fraud or financial accounting
credit. (Here in order to simplified model, fraud include actions that cause dis-
tortion of accounting information disclosure such as profit forecast inaccurate,
reporting manipulation and the disclosure of financial information still exists in
time) 3. Assumption that understanding of each participant about action selection
of other participant is not necessarily accurate, and the information of participants
in this model is not complete. According to the above assumptions, complete
information static state game model between financial and accounting information
discloser and government regulators can be built (see Table 47.1). This game is a
static state game between government regulators and the financial accounting
information discloser of social insurance fund.
Table 47.1, the first number in the table shows earnings of the government
regulators; the second number shows earnings of financial and accounting infor-
mation discloser. Strategies taken by game participants and condition of earnings are
common knowledge and common information of both sides, whose game are
developed by these. When financial accounting credit of financial accounting
information discloser of social insurance fund and government regulators don’t
check and both sides have no extra income and loss, ‘‘extra income’’ of both sides at
this moment can be seen as 0. When government regulators don’t check and financial
accounting of the social security fund managers makes a fraud, the obtained extra
income is E. when financial accounting of the social security fund managers makes a
fraud and is investigated by the government regulators, fines should be paid besides
extra income E is confiscated and its income is-F(fine). Fines in real life usually are
several times the illegal extra income. However, the government regulators should
pay the cost C, such as dispatch of labor power and material resources, time to check
and so on. The performance reward of government regulators obtained by investi-
gating fraud of financial accounting information is B. Here, E, R, F, C, B are positive.
464 S. Dai et al.
Table 47.1 Static state game model between financial and accounting information discloser and
government regulators
Government regulatorsc Accounting information discloser
Accounting Accounting
fraud (c) credit (1 - c)
Irregularity (h) E ? B ? F – C, -F -C, 0
Non-inspection (1 - h) 0, E 0, 0
According to this payoff matrix, we can calculate expected profit function of the
financial accounting information discloser of social insurance fund:
S ¼ c hðF Þ þ c ð1 hÞ E ð47:1Þ
Expected profit function investigated by the government regulators are:
L ¼ c hðE þ F þ B CÞ þ ð1 cÞhðCÞ ð47:2Þ
In this formula, expected profit function of the financial accounting information
discloser of social insurance fund is S. According to fraud probability, we can get
its reaction function by derivation:
oS
¼ E h ðE þ F Þ ð47:3Þ
oc
Similarly, for expected profit function investigated by the government regula-
tors L, we can get its reaction function by derivation according to regulatory
probability:
oL
¼ c ð E þ F þ BÞ C ð47:4Þ
oh
Combination of (47.1) and (47.2) can get the optimal regulation probability of
the government regulators and the optimal fraud probability of the financial
accounting information discloser of social insurance fund, which are as follows:
E C
h ¼ c ¼
EþF EþFþB
It is synthesizing the above mathematical reasoning process; we can get Nash
equilibrium of mixed-strategy game.
Because the additional income E obtained from the fake information disclosure
is independent of this model, it will be regarded as constants out of human control
in the analysis. It can be seen that in the Nash equilibrium solution in the model, is
directly proportional to C and inversely proportional to F ? B.
The fraud in financial and accounting information disclosure of the social
security fund depends on the government regulators: if the smaller the law
enforcement of the government regulators C is, the bigger the awards B obtained
47 Research on Financial Accounting Information Disclosure 465
by the government regulators due to investigating the frauds in the accounting
information disclosure are and the penalties income F is and the smaller the
probability of the frauds in the accounting information disclosure is. The proba-
bility of the regulation and supervision of the government regulators mainly
depends on the extra benefits E obtained from false accounting information and
degree of punishment on the accounting information disclosure fraud F. The
higher the fraud income E is, the lower the punishment degree F is, and the higher
the probability of supervision and regulation of the government regulators is for
their own work, performance and reputation.
47.4.2 Improved Game Model After Adding Public
Supervision and Report Measures in the Above Game
Model
On the basis of the government regulation is not in place due to lack of financial
and material resources, public supervision institutions and the system for reporting
the social security fund disclosure fraud are considered to be established, or the
report can be done through the network or telephone.
As mentioned in Sect. 47.4.1, assume that the social security fund management
institutions are found to choose to disclose the fake accounting information and
punished seriously, which can make the Nash equilibrium point close to the Nash
equilibrium with best total social utility Pareto. It is not quite as simple as that,
though. First, it is uncertainty whether the government regulators have enough
energy to verify all the social security fund management institutions. Second, even
if verification is conducted, the verification shall not be guaranteed to be fully
successful, that is, the concealment of some false accounting information is too
intense to be found by the existing technology. Next, it shall be dependent on the
degree of the punishment on those agencies which disclose the false accounting
information. If the punishment is less than the benefits obtained from the disclo-
sure of false information, this phenomenon can’t be governed well. In the end, the
regulatory effect will depend on the benefits and costs of the regulators. All of
these show that social security agencies and the government regulators optimize
their action strategies in the game constantly.
In Table 47.2, the first figure indicates the benefits of the regulators and the
second one indicates the benefits of accounting information disclosers.
Suppose P is the probability that the government regulators receive the reports
without inspecting the social security fund management institutions. Therefore, the
probability that the government regulators don’t receive the reports without
inspection is 1-P. The probability that the regulators succeed in the verifying the
social securities fund disclosers is K while the probability of failed verification is
1-K. The reputation losses of the government regulators due to the fraud in the
466 S. Dai et al.
Table 47.2 The benefits of the regulators and accounting information disclosers
Accounting Government regulators
information discloser
Irregularity (h) Non-inspection (1 - h)
Confirmed (K) Not confirmed No result Report and verify the
(1 - K) (1 - P) irregularities (P)
Accounting fraud (c) E?B?F– C, E 0, E E ? F - R, -F
C, - F
Accounting integrity -C, 0 -C, 0 0, 0 0, 0
(1 – c)
financial and accounting information disclosure of the social securities fund are
R (Reputation).
1. Where the situation of the probability of the fraud in financial and accounting
information disclosure of the social security fund is given, the expected benefits
obtained in the event of inspection and non-inspection conducted by the regulators
is respectively:
p1 ¼ ½ðE þ B þ F CÞ K þ ðCÞð1 KÞc
þ ½ C K C ðL KÞð1 KÞ ð47:5Þ
p2 ¼ ðE þ F RÞc P þ 0 ð47:6Þ
C
p1 ¼ p2 ; c ¼
ð E þ F Þ ð K PÞ þ K B þ R P
2. Where the probability that the regulators conduct the inspection is given, the
expected benefits obtained by social securities fund from fake information dis-
closure and true information disclosure are separately:
p3 ¼ ½Eð1 K Þh FKh
ð47:7Þ
þ ½Eð1 PÞ EPð1 hÞ
p4 ¼ 0 ð47:8Þ
ðE þ F ÞP E
p3 ¼ p4 ; Get h ¼
ðE þ F ÞðP K Þ
Therefore, the improved mixed-strategy Nash Equilibrium is
C
c ¼
ðE þ F ÞðK PÞ þ K B þ RP0
ðE þ FÞP E
h ¼
ðE þ FÞðP KÞ
The improved mixed strategy Nash equilibrium indicates: if the social security
fund managers choose to undertake the false accounting information disclosure in
47 Research on Financial Accounting Information Disclosure 467
the form of c [ c , then the best choice for the regulators is to inspect. Otherwise,
the regulators will not inspect. If c ¼ c to undertake the false information
disclosure, the best choice for the regulators is not to inspect. Similarly regulators
conduct the inspection in the form of h [ h , the best choice for the social
securities fund management institutions is to disclose the true accounting infor-
mation. Otherwise, the reverse. Because the additional income E obtained from the
fake information disclosure and irregularities report and investigation P are
independent of this model, they will be regarded as constants out of human control
in the analysis. It can be seen that the probability of the fraud in financial and
accounting information disclosure of the social security fund depends on the
government regulators:
(1) when K–P is more than 0: if the smaller the law enforcement of the
government regulators C is, the bigger the awards B obtained by the government
regulators due to investigating the frauds in the accounting information disclosure
are and the penalties income F is. Besides, the bigger R and the probability for
regulators to succeed in the verification K is, the smaller the probability of the
frauds in the accounting information disclosure is.
At this time, the probability of the regulation and supervision of the government
regulators is mainly dependent on the degree of punishment F. The bigger F is, the
greater the probability of the supervision is.
(2) When K–P is less than 0: if the smaller the law enforcement of the government
regulators C is, the bigger the awards B and R obtained by the government regulators
due to investigating the frauds in the accounting information disclosure and the
probability for regulators to succeed in the verification K are. Besides, the less the
penalties income F is, the smaller the probability of the frauds in the accounting
information disclosure is.
At this time, the probability of the regulation and supervision of the government
regulators is mainly dependent on the degree of punishment F. The smaller F is,
the greater the probability of the supervision is.
Usually, K–P is greater than zero. If K–P is less than 0, it indicates that the
regulators don’t have any supervision effects. This situation is not allowed to
appear. As a result, the probability of the frauds in the accounting information
disclosure is mainly inversely proportional to B, F, R, K and directly proportional
to C. The measures shall be taken to contain B, F, R, K and C together, which will
minimize the probability of the frauds in the accounting information disclosure.
47.5 Conclusion
A. Social Securities Fund Accounting System and Social Securities Fund Financial
System issued in 1999 should be improved and updated. From the point of view of
game theory, the system arrangement is the rule of games. At the same time, the
system arrangement is the result of the game. Unreasonable or imperfect system
468 S. Dai et al.
arrangement will update and improve in the continuous dynamic game to reach a
new equilibrium at last. That is, the system designer and system executor conduct
a dynamic game. A good system arrangement should be perfected continuously. In
a certain sense, the ‘‘system’’ is more important than the game itself.
B. The government shall establish not only the efficient, honest, relatively inde-
pendent social security fund supervisory institutions, but also the non-governmental
supervisory institutions independent of the government regulators and provide the
public phones, web sites, mailboxes and other report measures. At the same time, this
can make the various regulators supervise each other.
C. On the basis of the application of the advanced computer technology, the
construction of the social security fund information management platform, the
supervision of revenues and expenditures of the social security fund and
management in advance, and the network management of the whole progress of fund
usage, the off-site surveillance and dynamic supervision will be realized gradually.
D. To establish an account checking system between the social security
administration department and the social security fund administrative agencies to
prevent the deviations of the their accounts. At the same time, actuarial reports and
submissions of the enrolled actuaries and the audit reports and submissions of the
certificated public accountants shall be provided in the financial accounting reports
and management reports of the social security fund.
References
Arrow KJ (1963) Uncertainty and the welfare economics of medical care. Am Econ Rev
53:941–973
Cho I-K, Kreps D (1987) Signaling games and stable equilibria. Q J Econ 102:179–221
Claire L, Marston Philip, Shrives J (1991) The use of disclosure indices in accounting research: A
review article. British Account Rev 23:195–210
Deng C (2003) A study on the problem of earnings management based on game theory. Acc Res
5:37–42 (in Chinese)
Du X (2004) Corporate governance development and regulation of accounting information
disclosure-game analysis and historical evidences. J Finance Econ 9:74–84 (in Chinese)
Gong L, Zhou H (2006) The key points for the auditing of social insurance fund. J Guizhou Univ
(Social Sciences) 25 (in Chinese)
Gou Y (2002) Current situation and reform idea of china’s social security accounting. Contemp
Finance Econ 2:73–76 (in Chinese)
Joseph VC, Terry LN (2003) Audit committee independence and disclosure: choice for
financially distressed firms. Corp Governance Int Rev 11(4):289–299
Li W, Duan H, Kong X, Ma X (2010) Study on the regulation mode of Chinese social security
fund. Int J Bus Manag 5(9):124–126
Tong G, Liu C (2011) Game analysis of fund supervision. information technology. In:
International conference on computer engineering and management sciences (ICM), 2011,
pp 197–199
Wang X (2001) A game analysis on the accounting information publication of listed companies.
Econ Surv 28(6) (in Chinese)
Wei S, Xue H, Lu T (2003) Analysis on game among parties in information disclosure. Economic
Problems 10:51–53 (in Chinese)
47 Research on Financial Accounting Information Disclosure 469
Zhao L (2008) The perfection of China’s social security fund supervision system. Shanghai
Finance 11:55–57 (in Chinese)
Zhang W (1996) Game theory and information economics. Shanghai people’s publishing house,
Shanghai
Zhou Y (2001) Social security fund of accounting research. Dongbei University of finance
&economics press, Dalian (in Chinese)
Chapter 48
Research on the Multi-Target Tracking
Algorithm Based on Double FPGA
Xu-dong Liu
Abstract The accuracy and real-time of the multi-target tracking system have been
the main study problems in the targets tracking field. In this research, two slices of
FPGA are applied as the main processing chips. Kalman Filter and Particle Filter, the
two commonly used filtering algorithms are dynamically integrated to make use of
their characteristics in multi-target tracking processing. The real-time of multi-target
tracking is realized through the complementation of double FPGA chips.
Keywords FPGA Kalman filter Multi-target tracking Particle filter
48.1 Introduction
With the increasing complexity of the targets tracking background and the rapid
development of digital technology, the requirements for radar multi-target tracking
ability in the interference background become much higher than ever. The multi-
target tracking technology has been widely used in the fields of military, commu-
nications, satellite navigation, and remote sensing etc. The core of the targets
tracking study is the filter algorithm. Therefore, it is the hot spot and difficult point of
the research field to find a better performance filtering algorithm to deal with the
linear and nonlinear problems in the actual system. In order to conduct more
effective real-time multi-target tracking in complex environments, the paper intro-
duces a new filtering algorithm by applying double FPGA as the processor core, and
combining Kalman Filter Algorithm and the Particle Filter Algorithm together.
X. Liu
Changchun University of Science and Technology, Changchun, China
X. Liu (&)
Economic Management Cadre College of Jilin Province, Changchun, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 471
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_48,
Ó Springer-Verlag Berlin Heidelberg 2013
472 X. Liu
48.2 The Filter: A Algorithm Realized by FPGA
The multi-target tracking is the process of identifying and tracking of the
multi-target motion parameters, such as position, velocity, acceleration and targets
classification feature, decomposing and estimating the measurement data from the
receiver, and finally forming each target trajectory (Liu 2003). In the whole sys-
tem, filtering is of vital importance. The effect of filtering exerts a direct impact on
the result of the whole tracking system. The so-called filtering refers to the optimal
estimation of the current value from the observation values got currently and in the
past. The filter theory is the system state or parameters estimation theory and
method, which is in accordance with certain filter criterion and based on mea-
surements of the system observed signals.
Since the filter process needs to access and calculate large state data, the main
processing chips of the tracking system require high operation speed and large
storage space. Meanwhile, the main processing chips have data communication
with the peripheral circuits. Therefore, the main processing chips should have
strong interface communication ability to ensure the real-time of the Multi-target
tracking. Based on SP3, Spartan-3A is an I/O optimization FPGA launched by
Xilinx for users’ more I/O demands. The advantage of SP3A lies in its highest I/O
density in the same FPGA series, which can support more I/O standards. It is an
ideal select as the main processing chips in the Multi-target tracking. The filter
algorithm in this research is realized by adopting FPGA Spartan -3A.
For the multi-target tracking system, there are various kinds of filter algorithms,
such as Bayes Filter, Optimal Estimation Filter, Suboptimal Estimate Filter, etc. At
present, Kalman Filter (a type of Optimal Estimation Filter) and Particle Filter are
more mature, and commonly used. The two kinds of filter methods are dynamically
integrated and studied in this paper.
48.2.1 Kalman Filter
Kalman Filter describes system by state space method, and is composed of state
equations and measurement equations. Kalman Filter estimates the state current
value with the previous state estimated value and some recent observation data, and
describes the current value in the form of state variable estimated value (Ding 2003).
Suppose the system state variable is Xk at the moment k, the state equation and
the measurement equation are as follows.
Xkþ1 ¼ Ak Xk þ xk ð48:1Þ
yk ¼ Ck Xk þ vk ð48:2Þ
In the equations, ‘‘k’’ means time, and here it refers to the ‘‘k’’ iteration value of
the corresponding signal. The input signal ‘‘xk’’ is white noise. The observation
48 Research on the Multi-target Tracking Algorithm 473
Fig. 48.1 Kalman Filter
Signal Model
noise of the output signal ‘‘Vk’’ is White Noise, too. The branch gain of the input
signal to the state variable equals to 1. ‘‘A’’ is the gain matrix among the state
variables, which can change with the time. ‘‘C’’ means the gain matrix between the
state variables and output signals. The signal model is shown in Fig. 48.1. If time
variable ‘‘k – 1’’ replaces ‘‘k’’ in the state equation, the equations are as follows.
Xk ¼ Ak1 Xk1 þ xk1 ð48:3Þ
y k ¼ C k Xk þ v k ð48:4Þ
Here ‘‘Xk’’ is a state variable. ‘‘xk-1’’ means the input signal, which is white
noise. ‘‘Vk’’ is observation noise. ‘‘Yk’’ is observation dada.
Kalman Filter Algorithm by FPGA is shown in Fig. 48.2. The calculation of the
covariance and the calculation of filter gain are realized by one FPGA. While the
posterior state estimate and the posterior error covariance estimate are conducted
by another FPGA. By this way, double FPGA can work simultaneously on the
algorithm computation and storage, and improve the performance of the real-time.
For the linear Gaussian random system, Kalman Filter can get an accurate
analytical solution of posterior probability density function. But in some cases, it is
hard to get an accurate posterior probability density function, so the approximate
suboptimal estimates are required (Xu 2008; Abdel-Hakim and Farag 2006).
Besides, because of the poor system’s observability and the low-degree linearity of
the state space model, Kalman Filter Algorithm can not meet the requirements in
the aspects of the convergence precision and the convergence time (Deng 2008). In
order to solve the problem, Gordon introduced re-sampling algorithm into
sequential importance sampling, and put forward the importance sequential
importance re-sampling algorithm. Thus the algorithm becomes more perfect. The
Sequential Importance Re-sampling Algorithm is also the base of the Particle
Filter Algorithm. Compared with conventional Kalman Filter Algorithm and its
improving algorithm (Zawar and Malaney 2006), the Particle Filter can perform
better in non-linear & non-Gaussian environment (Wu et al. 2009). When the
particle quantity is enough, its estimate can approximate the optimal solution.
474 X. Liu
Fig. 48.2 Kalman Filter
FPGA Algorithm Process
48.2.2 Particle Filter
The Particle Filter is a filter method for non-linear & non-Gaussian system, based
on Monte Carlo Methods (MCM) and the Recursive Bayes Estimation. As the
fundamental idea of Particle Filter, firstly, in terms of the experience conditional
distribution of the system state vector (Czyz et al. 2007), a group of random
samples set is generated by state space sampling. The sample set is called ‘‘Par-
ticle’’. And then the particles weight and the sample position is adjusted cease-
lessly in accordance with the observation data (Wang 2009). Finally, the correction
initial experience conditional distribution and the estimation system states and
parameters can be achieved through the adjustments of the particles information.
Algorithm shows in Fig. 48.3.
Normally, the descriptions of nonlinear & non-Gaussian system are as follow.
Xk ¼ fk ðXk1 ; vk Þ ð48:5Þ
Yk ¼ hk ðXk ; wk Þ ð48:6Þ
In the equations, ‘‘Xk [ Rn’’ means system state of ‘‘K’’ time. ‘‘Yk [ Rn’’ means
measurement. ‘‘Vk [ Rn’’ and ‘‘Wk [ Rn’’ represent the independently and
simultaneously distributed system noise and observation noise sequence respec-
tively. State space model describes dynamic system model. The implied time is
white variable. In the study of multi-target tracking, state vector contains goal
48 Research on the Multi-target Tracking Algorithm 475
Fig. 48 3 Signal Model of
Particle Filter
dynamics characteristic parameters, for instance, the specific location in the
coordinate system and the speed value of a definite direction. This algorithm is a
kind of recursive filtering algorithm, and in principle, it can estimate the state and
parameters of an arbitrary nonlinear & non-Gaussian stochastic system (Zhang
2010). Thus it overcomes the shortcomings of Kalman Filter Algorithm.
In the process of Particle Filter, particle degradation exists (Feng 2008). Re-
sampling method can be adopted to solve the problem preliminarily (Vo and Ma
2006). But this will bring forth more computation amount, and add more burden to
the main processor, and influence the real-time of the system (Qian 2011). The
Particle Filter FPGA algorithm process is shown in Fig. 48.4.
The application of double FPGA processing form can solve the problem
thoroughly. The two slices of FPGA chips adopt parallel operations model, one is
used mainly to handle Kalman Filtering process, Particle Filter re-sampling pro-
cess and the communication between the double chips, the other completes the
main operation process of Particle Filter. Thus, the filter method can switch from
one to the other according to the specific signal condition, and two kinds of filter
methods can run synchronously through making full use of resources to cope with
more complex non-linear& non-Gaussian system (Wang 2004). In some maneu-
vering multi-target tracking system, the system needs to switch at any time in any
sorts of models (Hu 2010). The mutual cooperation of the two FPGA can provides
guarantee that the system can perform real-time tracking to the multiple targets
(Hu 2008).
48.3 Simulation Contrast
The paper only takes first 100 frame images in the simulation process as reference.
For simple realization, suppose both the system noise and the measurement noise
are zero mean Gaussian white noise.
The simulation results shows in Fig. 48.5. The track constituted by solid circles
is the actual multi-target motion trajectory, and hollow rings form the algorithm
estimate trajectory. From the graph, for each target, the difference value between
the filtering algorithm estimate trajectory and the actual target motion trajectory is
476 X. Liu
Fig. 48.4 FPGA Algorithm Process of Particle Filter
Fig. 48.5 Kalman Filter and
Particle Filter Simulation
Results
small. In the process of the target movement, even if there exists a high acceler-
ation, the filtering effect can still be achieved, which completely meet the
requirement of the multi-target tracking.
48 Research on the Multi-target Tracking Algorithm 477
48.4 Conclusion
This paper designed and realized the application of multi-target tracking system
algorithm in complex environment by using FPGA as its main processing chips,
which can satisfy the requirement of the subsequent processing for multi-target
detection and tracking. The simulation results show that the expected objectives
can be achieved. If there is a further demand to improve the system signal real-
time processing speed, a piece of DSP (TMS320C6416) can be applied for signal
preprocessing before the two pieces of FPGA perform signal processing specifi-
cally. By this way, better signal characteristics can be distinguished. Besides, it
will have a bright market prospect to add a USB form output interface in the
system which uses this kind of algorithm.
References
Abdel-Hakim AE, Farag AA (2006) A sift descriptor with color invariant characteristics. IEEE
Comput Soc Conf Comput Vis Pattern Recognit 67(10):1978–1983
Czyz J, Ristic B, Macq B (2007) A particle filter for joint detection and tracking of color objects.
Image Vis Comput 21(9):1271–1281
Deng WT (2008) Research of particle filter algorithm based on FPGA implementation. Mater
dissertation, Beijing Jiaotong University, Beijing, China
Ding YM (2003) Digital signal processing. Xidian University Press, Xian, pp 48–61
Feng Y (2008) The research on data association in multiple target tracking. Mater dissertation,
Xidian University, Xi’an, China
Hu P (2010) Research on video object tracking with Kalman filter. Mater dissertation, College of
Computer Science of Chongqing University, Chongqing, China
Hu S (2008) Visual target tracking based on Particle filter. Ph.D. dissertation, Nanjing University
of Science and Technology, Nanjing, China
Liu G (2003) The research and realization of multiple target tracking algorithm. Ph.D.
dissertation, Northwestern Polytechnical University, Xi’an, China
Qian Z (2011) A multi-target tracking problem of occlusion. Mater dissertation, Guangdong
University of Technology, Guangzhou, China
Vo B, Ma W (2006) The Gaussian mixture probability hypothesis density filter. IEEE Trans
Signal Process 49(10):4091–4104
Wang TT (2009) Particle filter algorithm application and research on target tracking. Mater
dissertation, North University of China, China
Wang Y (2004) Research on occlusion in object tracking. Mater dissertation, Huazhong
University of Science and Technology, Wuhan, China, 2004
Wu C, Yang D, Hao ZC (2009) Color image tracking algorithm based on particle filter. Optics
Precis Eng 17(10):2542–2546
Xu LZ (2008) Visual tracking based on particle filter and kalman filter under complex
environments. Mater dissertation, Zhejiang University, Hangzhou, China
Zawar S, Malaney RA (2006) Particle filters and position tracking in Wi-Fi networks. In:
Proceedings of the 63rd IEEE vehicular technology conference. Melbourne, Australia, 7–10
May 2006, pp 613–617
Zhang C (2010) Research on object detection algorithm of partially occluded. Mater dissertation,
Xi’an Technological University, Xi’an, China
Chapter 49
A Study of the Effect of Xi’an Subway
on the Rent of Office Building
Ying Wang and Wen-jia Wei
Abstract On the basis of comprehensive consideration of internal and external
factors affecting the rent of office building, this article aims at the rent of the office
buildings along the Xi’an Metro Line 2. It applies BP neural network model to
quantitatively analyze the degree of influence of Xi’an subway on the office rent.
The analysis results demonstrate that Xi’an Subway Line 2 has a significant effect
on office rent. It ranked the fourth in 11 factors, effect weight of which is 10.5 %. It
shows that Office enterprises especially prefer on the convenience of the subway
traffic, which indicates this subway line plays an important role in enhancing the
value of surrounding office properties.
Keywords Office rents Xi’an Subway Line 2 BP neural network model
49.1 Introduction
Whether investors will develop a real estate, or firms choose office location, office
space location is a critical factor. Existing urban rail transit construction experi-
ence has shown that urban rail transit through the ‘‘value capture’’ phenomenon in
improving the quality and attractiveness of urban areas can play a significant role
(Du and Mulley 2007). It will not only improve the traffic convenience along the
region, and can attract a variety of life, business, education, entertainment and
other facilities along the metro aggregation, enhance and improve the life envi-
ronment of the office along the subway.
Therefore, it has strong stimulation to the development of surrounding property,
thus contributes to the high-density development of the site and enhancement of
Y. Wang W. Wei (&)
School of Management, Xi’an University of Architecture and Technology, Xi’an, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 479
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_49,
Ó Springer-Verlag Berlin Heidelberg 2013
480 Y. Wang and W. Wei
the real estate value along the subway. In general, the closer the office building is
to the Metro arrivals, the greater the transit has impact on office rental. However,
some studies suggest no impact on the rent of office along the subway.
The impact of the subway on the rent of office building reflects not only the
strength of the external economic role of the subway construction, but the driving
agglomerating force on the surrounding economy. In essence, it reflects the input
and output value of the subway construction, rail transport space resource values
and the economic sustainability of the future development of urban rail transit.
Therefore, this research can provide valuable suggestions, which is about more
rational and scientific development of urban rail transit planning and use-planning
of land along the subway in order to fully realize the optimal allocation of urban
space resources, and also provide strong support for the decision-making of the
investors on the siting.
Subway Line 2 is currently the only subway line in Xi’an, which has been built
and put into operation. It is a total length of 26.4 km, runs through north and south
of the city. The installment is from the Xi’an North Passenger Transport Station to
south, through the Bell Tower, along the South Hamlet and Xiao-zhai to the
Qu-jiang International Convention and Exhibition Centre. It officially has put into
operation in September 2011. This paper studies the rental office along Xi’an
Subway Line 2 and conduct an empirical analysis of the impact of subway con-
struction on office rentals by employing BP neural network model.
49.2 Domestic and International Research Review
Since the 1970s, there has been extensive and in-depth research about interests
influence on urban rail transit project development in foreign countries. Therein,
the impact research of rail transit construction on the surrounding property values
is an important research topic in this field. Mejia-Dorantes et al. (2011) studied
Madrid metro and by using spatial statistical technique come to the conclusion that
the subway plays a very important role in improving transport efficiency and urban
planning. Pagliara and Papa (2011) studied the value impact of the city region
along the subway, and research shows as time goes, the value of surrounding land
will gradually improve.
The subway from the Beijing Railway Station to Western Suburb Apple Orchards
is the China’s first metro line, and started trial operation in 1971. In recent years, with
the rapid development of domestic metro, the relationship of rail transportation and
real estate values along it is the concern of many scholars. Some studies focused on
the residential real estate, such as the studies of Yang and Shao (2008) and Liu and
Shang (2009). Other studies aimed at the office market such as Jiao et al. (2009). It
carried out the empirical results on the office rentals change along Beijing Metro 10
line and show that there is no significant effect of urban rail transit on the office space
rental changes. The main reason was that the center effect of the city center and the
aggregation effect of shopping district weakened the impact of urban rail transit. The
49 A Study of the Effect of Xi’an Subway 481
tool of empirical analysis of these studies mainly has transportation cost model, the
hedonic price model or neural network analysis. Transportation cost model is the
model based on the relationship between transport costs and real estate value without
considering the value impact of real estate factors. The hedonic price model com-
monly uses the form of linear equations functions mainly to solve the static problem.
But the neural network model with particular autonomous learning is able to accu-
rately simulate and solve non-linear relationship. It has the higher accuracy and
reliability (Yang and Shao 2008).
On the basis of existing research, the paper uses the BP neural network model to
analyses the influence strength of Xi’an Metro Line 2 on the office rental con-
sidering various internal and external factors affecting office rentals.
49.3 Empirical Analysis of the Impact of Office Rentals
Alone Xi’an Subway
49.3.1 Sample and Variable Selection
In order to get more qualified samples, the samples are limited to the following: (1)
samples are taken from rental offices surrounding some great station, such as
Chang’an Road Station, South gate Door Stations, taking into account of the big
gap in commercial atmosphere and traffic conditions along the north–south of the
Metro Line 2; (2) samples are only the rental offices 800 m depart from the
subway, due to no effects beyond a certain range; (3) excluding the exotic sample
of which the rent is significantly low or high. In accordance with above the limit,
this paper collected 80 samples out of the 20 leased office apartments along Xi’an
Subway Line 2. Based on existing research, the paper selects 11 impact factors
variables of office rentals as following: distance from the subway exit, office type,
commercial agglomeration degree, public facilities, traffic conditions, image of the
appearance, decoration level, internal support facilities, where the floor, property
costs and property level. All variables are defined as follows in Table 49.1 (Wang
and Chen 2009).
Among them, the descriptive variables in the sample are quantified, referring to
the real estate valuation method and expert scoring method. First a standard
sample is selected for each independent variable, Scoring 100, and then the other
samples are scored one by one comparing to the standard sample rate. The
quantified samples are normalized, the formula of normalization is:
! !
X n X n X n
Yi ¼ Xi Min Xi Max Xi Min Xi
i¼1 i¼1 i¼1
Xi and Yi are the Values before and after the conversion (Fang and Chen 2004) .
482 Y. Wang and W. Wei
Table 49.1 The definition of the model variables
Variable name Definition
Distance from the subway The distance to the subway station (m, quantitative data)
exit
Office type The grade level of the office (given class of data)
Commercial agglomeration The flow of people around station (descriptive data)
degree
Public facilities Hotels, shopping malls, hospitals, schools and other facilities
(descriptive data)
Traffic conditions Traffic convenience (descriptive data)
Image of the appearance Building height, shape, wall decoration and building condition
(descriptive data)
Decoration level Interior grade (given class of data)
Internal support facilities Power, communications, transportation, electronics and other
equipment and facilities (descriptive data)
Where the floor Location of floor (quantitative data)
Property costs Property service fees (Yuan\m2\mouth, quantitative data)
Property level The level of property services (given class of data)
49.3.2 BP Neural Network Model Analysis and Results
This article selects BP neural network model with three-layer hierarchy network,
nodes of input layer, node of output layer and nodes of hidden layer. There are a
number of hidden nodes on the adjacent; neurons of adjacent on the lower are fully
connected. That is, each lower neuron with each upper neuron to achieve the right
connections, while each floor no connection between neurons. First, the input
signal propagates forward to the hidden layer nodes, after the role of function, then
the hidden output signal transmitted to the output node, and finally output is
obtained. The role of sigmoid-type function of node activation function is selected
the non-linear physical. Network training is selected TRAINGDM optimization
algorithms to improve the network generalization ability and convergence speed,
and the largest training sated 500, the target error value sated to 0.01, the fre-
quency showing sated 50. Eleven main factors are as the input of the BP neural
network model, rental of office space rent is the output. The paper selects 60
groups as the training samples and 20 groups as the simulation data. The results of
the model Training is shown in Figs. 49.1 and 49.2.
The result of model after repeated experiments is that hidden layer nodes is 30.
The best results Fig. 49.1 shows that the error of the desired output of training
sample and calculating output is less than 1 % after 168 times simulation training
by the network. When the network reaches a predetermined target error, model
training will get success. Figure 49.2 shows a sample model fitting, R = 0.83912,
the model fitting of the sample data close relatively high to 84 %.
From Table 49.2, using the trained model to test, the maximum error between the
output value and the ‘‘expectation’’ is 10.52 %, with an average error of 5.2 %.
Comprehensive evaluation result is consistent with the expected data. This shows
49 A Study of the Effect of Xi’an Subway 483
Fig. 49.1 Training results
Fig. 49.2 Fitting figure
that using comprehensive evaluation of the three-layer BP neural network method
to simulate the data is valid. Training office influencing factors B, by the 3-layer
neural network model can be explained the effects of office influence factors on rent.
Table 49.3 is the impact proportion of office rental influencing factors of the
model output, the greater effect proportion factors include: commercial aggrega-
tion, public facilities, and internal support facilities, distance from the subway exit
and traffic conditions. Because office rentals form from the office market demand
and supply conditions, the impact of various factors on office rentals reflects the
demand of the enterprises settled in offices for office space preferences. Business
concentration’s impact proportion of office rentals is 15.7 %, indicating that the
impact of the overall atmosphere of the shopping district and the maturity on office
rentals is the most important. It is also the most important for Business enterprises.
The effect proportion of public facilities is 13.30 % ranking the second. Besides,
484 Y. Wang and W. Wei
Table 49.2 The samples tested in the input and model estimates the value of output
Tested samples (Yi) 1 2 3 4 5 6 7
Real value 0.1607 0.2047 0.0689 0.2216 0.063 0.1198 0.1524
Model estimates 0.1022 0.1803 0.1642 0.1682 0.1065 0.028 0.1501
Relative error 0.0585 0.0244 -0.095 0.0533 -0.0435 0.0918 0.0023
Tested samples (Yi) 8 9 10 11 13 14 15
Real value 0.0398 0.2392 0.4625 0.5199 0.1069 0.1606 0.157
Model estimates 0.145 0.2021 0.4589 0.4839 0.1744 0.2084 0.2084
Relative error -0.105 0.0372 0.0036 0.036 -0.0675 -0.048 -0.0515
Tested samples (Yi) 16 17 18 19 20
Real value 0.0398 0.0188 0.2502 0.3473 0.045
Model estimates 0.1046 0.0781 0.2622 0.2631 0.0536
Relative error -0.065 -0.059 -0.012 0.0842 -0.0086
the shopping malls, hotels, apartments, hospitals and other public facilities can
greatly enhance the value of office space. Internal supporting facilities have a great
impact on office rentals, namely 11.7 %, due to the intelligent management of the
office bringing great convenience to the business office.
The impact of distance from the subway exit on office rentals is 10.5 %, located
the fourth in the selected factors, which indicates that the impact of the Xi’an
Subway Line 2 on the rents of office buildings along it is more significant. In
recent years, following Xi’an’s rapid economic development, a sharp increase in
foreign population and the rapid growth of private car, the problem of urban traffic
congestion is increasingly serious. Xi’an Subway Line 2 is located in the city
north–south traffic artery and it is opened to a greater extent to mitigate this traffic
load in the trunk on the ground, and raise the transportation convenience of office
properties, but also be lower transportation costs of enterprises settled, thus the
greater value of office properties.
However, ‘‘the city of Xi’an, rapid rail transit construction plan’’ shows that
Xi’an will build a total of six subway lines and a total length of 251.8 km, forming
the network structure of the board radial layout. Now, in Xi’an, only Metro Line 2
is completed, the length is only one tenth of the total length of all metro Planning,
and it can only solve the convenient traffic problems on a trunk road. Since metro
Table 49.3 Significance of Factors Affecting
Affect Commercial Public Internal support Distance from Traffic
agglomeration degree facilities facilities subway exit conditions
Importance 15.70 13.30 11.70 10.50 9.60
(%)
Affect (%) Property level Where Image of Decoration level Property
the appearance costs
floor
Importance 7.40 7.10 6.80 6.70 6.20
(%)
49 A Study of the Effect of Xi’an Subway 485
transport network has yet to be not formed, the synergy between the subway lines
also cannot play, and powerful subway externalities are not adequately reflected.
Competition between the subway lines will also have a negative effect for the
rental of office premises, so the combined effect of these two aspects of com-
parative study has yet to be further after the completion of the subway network
(Chen and Wu 2005).
49.4 Conclusion
Urban rail transit not only is able to respond to rapid economic development of
urban traffic congestion, but also can improve the quality and attractiveness of
urban areas. Based on BP neural network model for quantitative, the paper anal-
yses the influence degree of the Xi’an Subway Line 2 on the rent of office
buildings. Studies show that Xi’an Line 2 subway has a significant impact on the
office buildings rent along, the importance weight is of 10.5 % and places the
fourth out 11 office influencing factors. The construction of the Xi’an Subway Line
2 not only enhances the convenient transportation and the convenience of the
people’s work and life, also enhances the value of office buildings along it.
However, as the Metro Line 2 is the only one metro line in Xi’an, and metro
transport network has not yet formed, the value enhancement of the property
brought by synergy effect among the subway lines cannot be fully manifested, On
the other hand, the subway line of this article will lose the current the monopoly
competitive strength due to the formation of the future metro network. Thus
weakening the role of the subway line for the office rent promotion, Hence, the
future research is very significant after the metro network completion (Ma and
Yang 2010).
Acknowledgments Supported by the key discipline of Shaanxi Province and Supported by the
featured discipline of philosophy and general social sciences of Shaanxi higher education
institutions.
References
Chen F, Wu Q-B (2005) Quantitative research on the increase of real estate price alongside the
urban rail transit (Chinese). Urban Rail Transp Res 10:12–17 (Chinese)
Du H, Mulley C (2007) The short-term land value impacts of urban rail transit: quantitative
evidence from Sunderland, UK. Land Use Policy 24:223–233
Fang X-Y, Chen Z-N (2004) Review of impacts of urban rail transit on property values in foreign
countries. Trop Geogr 09:270–274 (Chinese)
Jiao X, Gu L-P, Hu C-M (2009) Prices analysis of rail transit impact on real estate based BP
neural network. J Chongqing Inst Technol Nat Sci 10:135–139 (Chinese)
Liu X-J, Shang P (2009) Analysis of the impact of real estate prices along the Xi’an Subway Line
2. Railway Transp Econ 02:41–43 (Chinese)
486 Y. Wang and W. Wei
Ma C-Q, Yang F-S (2010) Impact of rail transit on price increase of residential real estate.
J Traffic Transp Eng 4:91–96 (Chinese)
Mejia-Dorantes L, Paez A, Vassallo JM (2011) Transportation infrastructure impacts on firm
location: the effect of a new metro line in the suburbs of Madrid Original Research Article.
J Trans Geogr 10:91–97
Pagliara F, Papa E (2011) Urban rail systems investments: an analysis of the impacts on property
values and residents’ location Original Research Article. J Trans Geogr 19(2):200–211
Wang C-C, Chen Y-H (2009) Urban rail transit impact on the surrounding office rentals—
empirical analysis based on the Beijing Subway Line 10 (Chinese). Construction Economy,
2009, 4-6–4-9 (6 Month suppl)
Yang L-Y, Shao C-F (2008) Integrated forecasting model for real estate price along urban rail
transit based on BP neural network and Markov chain. J Jilin Univ Eng Technol Ed
3:514–519
Chapter 50
Urban–Rural Income Gap in China:
Evolutionary Trend and Influencing
Factors
Cun-gui Li
Abstract This paper is concerned with the evolutionary trend of Chinese urban–
rural residents’ income gap and find out its main influencing factors. Firstly, the
current situation and historical evolutionary trend of urban–rural residents’ income
gap in China were analyzed. The results demonstrated that since the reform and
opening up in 1978, the Chinese urban–rural income gap shows the phase change
characteristics: reduced—expanded—again reduced—again expanded—flattened.
Then, by using the data from 1978 to 2010, multiple linear regression models were
established to identify the correlation between urban–rural income gap and its
influencing factors. The study proves that the urban–rural dual structure,
employment structure and urbanization have positive correlation with urban–rural
income gap, and there is a negative correlation between rural financial develop-
ment level and urban–rural income gap.
Keywords Evolutionary trend Influencing factors Multiple regression analysis
Urban–rural income gap
50.1 Introduction
For a long time, the development strategy of ‘‘industrial priority and urban bias’’
was implemented in China; with respect to industrial policy, stressed the priority
of urban industrial; with respect to capital flows, the agricultural surplus was
transferred from countryside to city through taxation, price scissors between
industrial goods and agricultural products, financial institutions and so on; with
respect to population migration, there were many obstacles and constraints for
labors to migrate from countryside to city. Those urban–rural dual policies result
C. Li (&)
School of Economics, Henan University of Science and Technology, Luoyang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 487
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_50,
Ó Springer-Verlag Berlin Heidelberg 2013
488 C. Li
in the urban–rural income gap is increasingly widening. In 2011, the per capita
annual disposable income of Chinese urban households is 21 810 Yuan, and the
Per Capita Annual Net Income of Rural Households is 6 977 Yuan. The ratio of
urban income to rural income reaches to 3.13:1. Moreover, a lot of hidden incomes
are not included in the income of urban households, such as housing subsidies
medicate, medical insurance, unemployment insurance, old-age pension, guarantee
a minimum income and other social welfare. If we took them into account, the
urban–rural income gap would become much bigger (Li and Luo 2007).
Currently China has become one of the countries of largest urban–rural income
gap all over the world. The enlarging urban–rural income gap will inevitably affect
China’s economic sustainable development. Levin (2001) holds that like many
other developing countries, the urban-biased policies are the main reason of
enlarging urban–rural income gap (Levin 2001). A study by Cai (2003) has shown
that the urban–rural relationship in many developing countries is mandatory,
namely, urban biased. The urban–rural income gap caused by urban biased policies
is the common phenomenon in developing countries (Cai 2003). Hertel (2006)
found that the household registration system, rural land contract system, and non-
agricultural labor migration restrictions have significant impact on urban–rural
income gap (Hertela and Zhai 2006). Lei and Cai (2012) draws a conclusion that
the distortion of primary income distribution and urban-biased fiscal expenditure
policy enlarge the urban–rural inequality significantly (Lei and Cai 2012). Chen
and Peng (2012) thinks that the opportunity inequality caused by the household
registration system is an important reason for the urban–rural income gap (Chen
and Peng 2012). Zuo draws a conclusion that the main measures to narrow the
urban–rural income gap is to speed up the process of urbanization (Zuo 2012).
Based on these literatures, using the time series from 1978 to 2010, this paper will
built multiple regression models to test the relationship between urban–rural income
gap and its influencing factors, such as urban–rural dual economic structure,
urbanization, growth rate of GDP, employment structure, rural industrialization
level, and rural financial development level. The purpose is to find out the key factors
of influencing the urban–rural income gap and provide basis for decision making of
reducing urban–rural dual economic system and forming a new pattern that inte-
grates economic and social development in urban–rural areas. The arrangements of
following parts of this paper are as follows: the second part will analyze the current
status and historical evolutionary trend of urban–rural residents’ income gap in
China; the third part will make an empirical analysis on the influencing factors of
urban–rural income gap; in the fourth part, the useful conclusions will be given.
50.2 The Evolutionary Trend of Urban–Rural Income Gap
According to Fig. 50.1, since 1978, although the growth rate of absolute urban–
rural income, the ratio of urban residents’ income to rural residents’ income
becomes bigger. From 1978 to 2011, the per capita disposable income of urban
50 Urban–Rural Income Gap in China 489
Fig. 50.1 The changing trend of urban–rural income growth rate and gap from 1978 to 2011 in
China
households and the per capita net income of rural households have increased from
343.40 Yuan and 133.60 Yuan to 21 810 Yuan and 6 977 Yuan. The annual
average growth rate is 13.40 and 12.73 % respectively. Because the urban resi-
dents’ income grows faster than rural, the absolute urban–rural income gap is
becoming wider.
As shown in Fig. 50.1, the urban–rural income gap shows obvious phase
characteristics since the reform and opening up in China.
(1) The first phase (1978–1983), the urban–rural income gap shows the trend of
fast reduction. The central task of this phase was rural economic system
reform, and the rural household contract responsibility system was approved
and promoted. The enthusiasm of the peasants for production was greatly
motivated, the rural productive force was liberated and the agricultural labor
productivity was raised by a big margin. On the contrary, the urban economic
system reform was relatively backward. As a result, the growth rate of rural
residents’ income was higher than that of urban residents. From 1978 to 1983,
the per-capita net income of rural households rose by 18.32 %, the per capita
disposable income of urban residents by 10.46 %. The relative urban–rural
income gap dropped from 2.57 in 1978 to 1.82 in 1983.
(2) The second phase (1984–1994), the urban–rural income gap shows the trend of
fast expansion in general. At this phase, the focus of economic reform changed
from countryside to city, result in a rapid growth of the urban economy. Over
the same period, the incentive effects of the rural household contract
responsibility system on the agriculture was diminishing, the agricultural labor
force increased gradually, and the price of agricultural means of production
increased substantially, therefore, the rural economic growth slowed down.
From 1984 to 1994, the average annual growth rate of the urban per capita
disposable income and rural per-capita net income was 18.30 and 13.14 %
respectively. The relative urban–rural income gap increased from 1.83 in 1984
to 2.86 in 1994.
490 C. Li
(3) The third phase (1995–1997), the urban–rural income gap shows a diminishing
trend once again. At this phase, the purchasing price of agricultural and
sideline products was raised considerably; the rural township enterprises
developed rapidly, so the income from wages and salaries of rural households
increased by a large margin. During the same period, the government
implemented tight macro-control policies, leading to a slow down of urban
economic development and a soft landing for the economy. From 1995 to
1997, the average annual growth rate of the urban per capita disposable
income was 13.86 % and the average annual growth of rural per-capita net
income was 19.62 %. The growth rate of income of urban residents urban was
much lower than the growth rate of rural residents’ income. The relative
urban–rural income gap dropped from 2.71 in 1995 to 2.47 in 1997.
(4) The fourth phase (1998–2003), the urban–rural income gap shows the trend of
expansion once again. At this phase, there commonly appeared the following
situations, such as it was difficult for peasants to sell food, the output
increased, but income did not increase. The pace of development of rural
township enterprises slowed down, its ability to absorb rural surplus labor
force weakened gradually and its contribution to increase farmers’ income
decreased obviously. In addition, the urban-oriented policies and institutional
arrangements with the household registration system at its core have great
constraints for the rural surplus labor to migrate to non-agricultural industries
of rural areas or urban areas. Household registration system and urban
employment system have become the primary restricting bottleneck of
increasing income of peasants. From 1998 to 2003, the average annual growth
rate of the urban per capita disposable income was 9.32 % and the average
annual growth of rural per capita net income was only 3.94 %, less than half of
urban. The relative urban–rural income gap increased from 2.51 in 1998 to
3.23 in 2003.
(5) The fifth phase (2004-present), the urban–rural income gap tend to become
flattened. From 2004 to 2012, nine No. 1 central documents were issued by
central committee of the CPC to solve the three dimensional rural issues. A
series of policy measures to support and benefit agriculture, rural areas and
farmers were implemented, such as agricultural tax, livestock tax and taxes on
special agricultural products were rescinded throughout the countryside;
increase financial support for agriculture; carry out the subsidy policies of
direct subsidies to grain peasants, fine breed allowance and Farm machinery
purchase subsidy; implement the strategy of balancing urban and rural eco-
nomic and social development. Both the urban and rural income has a rela-
tively high growth. From 2004 to 2011, the average annual growth rate of the
urban per capita disposable income was 12.74 % and the average annual
growth of rural per-capita net income was 13.16 %. The relative urban–rural
income gap is reduced from 3.21 in 1998 to 3.13 in 2011.
50 Urban–Rural Income Gap in China 491
50.3 Empirical Analysis
50.3.1 Selection of Variables
This paper takes the urban–rural income gap (Y) as the explained variable
(dependent variable), which is measured by the ratio of the urban residents’
incomes to rural residents’ incomes. Urban–rural income gap = disposable
income of urban residents/per capita net income of rural residents. The bigger it is,
the larger urban–rural income gap will be.
It is unnecessary to consider all factors effected urban–rural income gap.
According to the significance of the influence and availability of the data, this
paper mainly considers the following independent variables.
(1) The coefficient of urban–rural dual structure (X1). Lewis (1954) proposed the
essay ‘‘economic development model with unlimited supplies of labor’’. This
essay assumed that an unlimited supply of labor was available in traditional
agricultural sector and the marginal labor productivity is negligible, zero, or
even negative, much lower than urban labor productivity generally. Accom-
panied by the transfer of labor, the dual economic structure will eliminate
(Lewis 1954). The coefficient of urban–rural dual structure (X1) = (the sec-
ondary and tertiary industries output/the secondary and tertiary industries
labor force)/(the primary industry output/the primary industrial labor force).
(2) The urbanization level (X2). In theory, with the improvement of the urbani-
zation level, the urban–rural income gap will decrease. On the one hand, the
aggregation effect of cities can absorb a lot of rural labors, which will result in
the improvement of agricultural productivity and increase of farmers’ income
(Lu and Chen 2004). On the other hand, the cities have a strong radiation and
driving function to rural economy. The city can promote rural economic
development by means of technology transfer, industries transfer, capital
output, information dissemination and so on. China’s current urbanization is
incomplete, which lead to agricultural scale management cannot realize and
rural migrant workers in cities cannot live in cities permanently (Ma 2008).
Meanwhile, among rural residents, only those who possess higher human
capital or have certain skills can easily migrate to cities; those who stay in
countryside are usually low education, lack of necessary skills, elder and
female persons. These people are also known as ‘‘386199 troops, 38 refer to
women, and 61 refer to the child, 99 refer to the elderly. The massive spillover
of human capital in rural areas of China may account for widening urban–rural
income gap.
(3) Growth rate of per capita GDP (X3). Kuznets (1995) proposed the ‘‘inverted
U’’ hypothesis to explain the income distribution gap changes with the process
of economic development in industrial countries. Kuznets inversed U curve
indicates that at the beginning of the process of economic development,
especially in the national per capita income increased from the lowest to the
492 C. Li
middle level, the income gap will increase along with the growth of per capita
GDP; when the economic development reaches to a certain stage, the income
gap will gradually reduce with further growth of per capita GDP (Kuznets
1995).
(4) Employment structure (X4). In theory, with the labors migrating from the
primary industry to the secondary and tertiary industries, the urban–rural
income gap will decrease. On the one hand, the transfer of the primary
industrial labor is helpful to realize agricultural large-scale management and
improve the agricultural labor productivity. On the other hand, non-agricul-
tural employment will improve rural residents’ wage income, which would
reduce the urban–rural income gap. But, since the reform and opening up in
1978, the restrictions on rural labor mobility were loosen gradually, but the
urban-biased welfare systems still exist. When rural labors find a job in cities,
they will subject to discriminatory treatment in the aspect of industry access,
wages, rights and interests maintenance, and so on. The employment structure
(X4) = non-agricultural industries employed persons/total employed persons.
(5) Rural industrialization level (X5). The development of rural industrialization
not only can absorb a lot of rural labors, but also can promote the process of
rural modernization. When more and more rural persons left land, the farmers
will possess more resources. Thus, the appropriate large-scale management of
agriculture will achieve (Hou 2005). Rural industrialization level (X5) = rural
non-agricultural industries employed persons/rural total employed persons.
(6) Rural financial development level (X6). The production theory of economics
shows that the increase of capital will result in the growth of output. The
development of rural financial can increase rural capital stock, and then lead to
the growth of agricultural output and improvement of farmers’ income. The
Rural financial development level (X6) = rural loans/the out of primary
industry (Liu and Hu 2010).
50.3.2 Model Building
Multiple linear regression models are often used as empirical models when more
than one independent variable is involved, which can approximate the true
unknown functional relationship between the dependent variable and independent
variables (Pao 2008).
Based on the statistical data from 1978 to 2010 (Table 50.1), the multiple
regression model is built to test the relationship between urban–rural income gap
and its influencing factors.
X
6
Y ¼ b0 þ bj Xj þ e ð50:1Þ
j¼1
50 Urban–Rural Income Gap in China 493
Table 50.1 The statistical data of urban–rural income gap and its influencing factors in china
(1978–2010)
Years Y X1 X2 (%) X3 (%) X4 (%) X5 (%) X6 (%)
1978 2.57 6.10 17.92 10.19 29.48 7.57 15.17
1979 2.42 5.08 18.96 6.15 30.20 7.71 14.14
1980 2.50 5.09 19.39 6.50 31.25 8.52 18.20
1981 2.20 4.56 20.16 3.90 31.90 8.86 18.02
1982 1.95 4.26 21.13 7.46 31.87 8.88 18.17
1983 1.82 4.10 21.62 9.26 32.92 10.20 19.14
1984 1.83 3.76 23.01 13.67 35.95 14.18 29.85
1985 1.86 4.18 23.71 11.93 37.58 16.01 30.53
1986 2.12 4.19 24.52 7.24 39.05 17.73 39.30
1987 2.17 4.09 25.32 9.81 40.01 18.81 43.91
1988 2.17 4.22 25.81 9.50 40.65 19.51 43.67
1989 2.29 4.48 26.21 2.48 39.95 18.84 45.83
1990 2.20 4.05 26.41 2.33 39.90 18.43 47.66
1991 2.40 4.56 26.94 7.70 40.30 18.59 55.71
1992 2.58 5.06 27.46 12.85 41.50 19.86 65.94
1993 2.80 5.27 27.99 12.66 43.60 22.38 69.49
1994 2.86 4.79 28.51 11.81 45.70 24.95 48.52
1995 2.71 4.38 29.04 9.73 47.80 27.53 24.88
1996 2.51 4.16 30.48 8.86 49.50 28.98 50.82
1997 2.47 4.45 31.91 8.18 50.10 28.95 57.82
1998 2.51 4.66 33.35 6.80 50.20 28.24 67.65
1999 2.65 5.09 34.78 6.69 49.90 26.98 74.16
2000 2.79 5.64 36.22 7.58 50.00 26.34 73.28
2001 2.90 5.95 37.66 7.52 50.00 25.22 76.83
2002 3.11 6.28 39.09 8.35 50.00 23.86 82.83
2003 3.23 6.57 40.53 9.34 50.90 23.79 92.47
2004 3.21 5.71 41.76 9.43 53.10 25.85 83.65
2005 3.22 5.88 42.99 10.66 55.20 27.71 86.67
2006 3.28 5.94 44.34 12.05 57.40 29.57 80.82
2007 3.33 5.71 45.89 13.57 59.20 30.74 78.74
2008 3.31 5.45 46.99 9.07 60.40 31.15 74.43
2009 3.33 5.34 48.34 8.67 61.90 32.03 87.01
2010 3.23 5.16 49.95 9.91 63.30 32.56 85.74
Data resources: China statistical yearbook, almanac of China’s finance and banking, China
township enterprise yearbook and China compendium of statistics 1949–2004
In above formula (50.1), e is random error term, b0 is a constant term, bj ðj ¼
1; 2; . . .; 6Þ is the partial regression coefficient, which represents the variation of
the predicted score when the corresponding independent variable Xi varies 1 unit.
_
They are estimated by bj ðj ¼ 1; 2; . . .; 6Þ separately.
494 C. Li
50.3.3 Multicollinearity Diagnosis
Variance inflation factor (VIF) is estimated for each independent variable to
identify causes of multicollinearity. When VIFj is more than ten, it reveals that the
j-th independent variable is highly correlated with other independent variables of
the model.
Using SPSS (19.0) and selecting the enter method in regression analysis, the
result is shown in Table 50.2. The result of equation fitting showed that the sample
multiple correlation coefficient R = 0.973, its value gives the proportion of the
variance of the dependent variable that can be predicted from the independent
variables. The determination coefficient R2 = 0.947, F-statistics = 76.825, sig-
nificant P & 0.000, indicating that the strength of the straight-line relationship is
high. But, as is seen from Table 50.2, some regression coefficients cannot pass the
t-Test of significance, such as X3 and X6. Meanwhile, the variance inflation factors
(VIF) of X2, X4 and X5 are much larger than 10, showing that there is multicol-
linearity in the model.
50.3.4 Result of Stepwise Regression Analysis
In econometrics, there are many methods for overcoming multicollinearity prob-
lem, such as to expand sample size, to transform model form, stepwise regression,
ridge regression, principal component regression and so on. In this paper, the
stepwise regression is adopted to overcome multicollinearity problem. According
to Table 50.3, the following two multiple-linear regression models were obtained
for the relationship between the urban–rural income gap and the above mentioned
influencing factors using the statistical data of the Table 50.1.
Model (1):
_
Y ¼ 0:194 þ 0:300X1 þ 0:030X2 ð50:2Þ
Table 50.2 Regression analysis result using with the method of enter-coefficients a
Model Unstandardized Standardized coefficients t Sig. Collinearity
coefficients statistics
B Std. Error Beta Tolerance VIF
(Constant) -2.849 0.910 -3.132 0.004
(X1) 0.413 0.045 0.654 9.212 0.000 0.407 2.457
(X2) -0.102 0.042 -2.007 -2.398 0.024 0.003 341.301
(X3) -0.006 0.008 -0.033 -0.675 0.506 0.872 1.147
(X4) 0.189 0.065 3.832 2.918 0.007 0.001 839.643
(X5) -0.089 0.040 -1.444 -2.225 0.035 0.005 205.142
(X6) 0.001 0.003 0.041 0.244 0.809 0.072 13.893
a
Dependent variable: Urban–rural Income gap (Y)
50 Urban–Rural Income Gap in China 495
Table 50.3 Regression analysis result using with the method of stepwise-coefficients a
Model Unstandardized Standardized coefficients t Sig. Collinearity
coefficients statistics
B Std. Error Beta Tolerance VIF
1 (Constant) 0.194 0.185 1.046 0.304
(X2) 0.030 0.004 0.587 8.181 0.000 0.652 1.533
(X1) 0.300 0.045 0.475 6.618 0.000 0.652 1.533
2 (Constant) -0.756 0.233 -3.252 0.003
(X1) 0.392 0.039 0.620 10.025 0.000 0.588 1.700
(X4) 0.037 0.005 0.759 7.350 0.000 0.211 4.735
(X6) -0.005 0.002 -0.254 -2.183 0.037 0.167 5.989
a
Dependent variable: Urban–rural Income gap (Y)
R2 = 0.899, R = 0.948, F-statistics = 133.824, significant P & 0.000, the
model can pass the F-test, indicating that there were significant correlations
between dependent variable and all independent variables.
Model (2) :
_
Y ¼ 0:756 þ 0:392X1 þ 0:037X4 0:005X6 ð50:3Þ
2
R = 0.935, R = 0.967, F-statistics = 138.327, significant P & 0.000, the
model can pass the F-test, indicating that there were significant correlations
between dependent variable and all independent variables.
According to Table 50.3, all the regression coefficients of model (1) and model
(2) can pass the t-Test of significance and the estimated VIF coefficients of all
variables are less than 10, indicating that the multicollinearity problem have been
overcoming and the above two models are effective, which can be used to
approximate the true functional relationship between the urban–rural income gap
and its influencing factors.
50.4 Conclusion
Since the reform and opening up in 1978, the urban–rural income gap in China
shows the phase change characteristics of reduced—expanded— again reduced—
again expanded—flattened, and each change has a great relationship with the
affection of policy changes.
According to empirical analysis, the relationship between urban to rural income
gap and its key influencing factors is as follows.
(1) There is a positive correlation between the urban–rural dual structure and
urban–rural income gap. According to the model (1) and model (2), the partial
regression coefficients of X1 are both positive and maximum, proving that the
urban–rural dual structure is the most key factor that leads to the expansion of
496 C. Li
urban–rural income gap. Therefore, changing the dual economic structure
should be the first choice to narrow the urban–rural income gap.
(2) There is a positive correlation between urbanization and urban–rural income
gap. As shown in the model (1), the partial regression coefficient of X2 is
0.030, indicating that when the urbanization level increases by one unit, the
urban–rural income gap will increase by 0.030 units averagely, assuming that
all other factors remain unchanged. China’s urbanization process has not
reduced the urban–rural income gap, but led to the urban–rural income gap
expand unceasingly. The fundamental reason is that the patterns of urbani-
zation is inappropriate and the process of urbanization is slow, rather than
urbanization itself. In the process of urbanization restricted by kinds of dis-
criminatory policy constraints, it maybe leads to the outflow of human capital
and physical capital in rural areas, and then the urban–rural income gap would
expand.
(3) There is a positive correlation between employment structure and urban–rural
income gap. As shown in the model (2), the partial regression coefficient of X4
is 0.037, indicating that the urban–rural income gap will increase by 0.037 unit
in average when the ratio of non-agricultural industries employed persons to
total employed persons increases by one unit, assuming that all other factors
remain unchanged. It was suggested that under the influence of institutional
barriers that represented by the household registration system, the mobility of
labor hasn’t the effect of narrowing the urban–rural income gap (Murata 2002).
(4) There is a negative correlation between rural financial development level and
urban–rural income gap. According to the model (2), the partial regression
coefficient of X6 is -0.005, indicating that the urban–rural income gap will
decrease by 0.005 unit in average, when the rural financial development level
increases by one unit, assuming that all other factors remain unchanged. The
development of rural financial plays an important role to reduce the urban–
rural income disparity.
Acknowledgments The work was supported by the tender subject of Henan government
decision-making research (No. 2012B239) and the investigation subject of Henan federation of
humanities and social sciences and Henan federation of economic organizations (No. SKL-2012-
3318).
References
Chen W-T, Peng X-M (2012) Household registration system, employment opportunity and the
income gap between urban and rural residents in China (in Chinese). Econ Surv
29(2):100–104
Cai F (2003) Rural urban income gap and critical point of institutional change (in Chinese) Soc
Sci China 40(5):16–25
Hertela T, Zhai F (2006) Labor market distortions, rural–urban inequality and the opening of
China’s economy (in Chinese). Econ Model 23(1):76–109
50 Urban–Rural Income Gap in China 497
Hou Y-j (2005) The development of rural industrialization and urban–rural dual industrialization
(in Chinese). J Xingtai Univ 20(2): 33–35
Kuznets S (1995) Economic growth and income inequality. Am Econ Rev 45(1):1–28
Lei G-q, Cai X (2012) The distortion of primary income distribution, urban–biased fiscal
expenditure policy and urban-rural inequality (in Chinese). J Quant Tech Econ 29(3):76–89
Levin J (2001) China’s divisive development. Harv Int Rev Fall 23:40–42
Lewis WA (1954) Economic development with unlimited supplies of labor. Manchester Sch Econ
Soc Stud 22(2):139–191
Li S, Luo C-L (2007) Re-estimating the income gap between urban and rural households in China
(in Chinese). J Peking Univ 44(2):111–119
Liu Y-w, Hu Z-y (2010) Empirical study on rural financial development on the urban–rural
income gap (in Chinese). J Shanxi Finan Econ Univ 32(2):45–52
Lu M, Chen Z (2004) Urbanization, urban-biased economic policies and urban–rural inequality
(in Chinese). Econ Res J 50(6):50–58
Ma X-S (2008) The negative effect and countermeasures of incomplete urbanization (in Chinese).
Jiangxi Soc Sci 35(1):176–185
Murata Y (2002) Rural-urban interdependence and industrialization. J Dev Econ 68(68):1–34
Pao H-T (2008) A comparison of neural network and multiple regression analysis in modeling
capital structure. Expert Syst Appl 35:720–727
Zuo Y-H (2012) The analysis of urban–rural income gap in status and income source contribution
(in Chinese). Econ Probl 36(1):27–31
Chapter 51
Sensitivity Analysis on Optimized
Sampling for Sealing Performance
of GVTP
Long Yu, Shu-rong Yu and Wen-li Yang
Abstract According to the method of sensitivity analysis, responding factors of
sealing performance on gate valve of thermal power (GVTP) are simplified. The
relationship of sensitivity and displacement of sealing surface is expressed by
mathematical equation. From 41 design parameters, it chooses 10 major global
parameters to compare the degree of sensitivity and get the highest sensitivity
parameters. The quantity of counting is reduced and the times of iteration are
shortened obviously. Using the Monte-Carlo method, sampling points are reduced
by 35 % and the maximum total deformation is reduced by 21 %. By using of
optimized, the sealing performance of the GVTP is improved, the leaking on the
sealing surface is restrained completely. The result of sensitivity analysis can
provided reference in the optimized process in power system with high parameter.
Keywords Displacement Sensitivity analysis Sampling Optimization
Monte-Carlo method
51.1 Introduction
Gate valve of thermal power (GVTP) is used to nuclear power and thermal power
station which work in high pressure and temperature. The technological parame-
ters still increase continuously for improving thermal efficiency. According to the
L. Yu S. Yu W. Yang (&)
Lanzhou University of Technology, Lanzhou, China
e-mail: [email protected]
L. Yu
e-mail: [email protected]
S. Yu
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 499
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_51,
Springer-Verlag Berlin Heidelberg 2013
500 L. Yu et al.
forecast provided by British Ministry of Industry and Trade (BMIT),The initial
parameters of steam of ultra-supercritical power unit have higher temperature and
pressure which are 40 MPa and 700 C, and the thermal efficiency will reach to
50–55 % in the future 5 years. So the performance of the GVTP must be enhanced
to fit the higher parameters of ultra-supercritical power station.
The plants of nuclear power and thermal power are composed by four major
units witch are constituted by boiler, turbine, pump and heat exchanger. Steam
generator provides superheated steam to driving turbine and outputting kinetic
energy. The turbine drives generator to producing power and making kinetic
energy turn to electric energy. The GVTP can lead the steam to heat exchanger for
protecting the turbine. The mechanics performance of the bypass valve is
importance for the safety in total power system. The GVTP has the role of con-
trolling the steam in front of turbine, because the pressure may be abrupt changed
and the temperature is lower than necessary. The sealing performance of the
GVTP is importance to the safety in total power system.
There have some results on the related valves, Nagaki Alegre et al. (2007),
Chen and Li (2007), studied on low cyclic fatigue failure for the steam valve in
high pressure. Zhu and Lei (2011), Wang et al. (2010), presented that the effect of
lower parameters valve be optimized relationship in its internal structure and flow
characteristics. Lu (2007), pointed out the views that the ultra-supercritical valve
has the relationships between mathematical models with operating parameters.
Zhu and Chuan (2010), Xiang et al. (2006), Tao and Cai (2009), presented the
method of three dimensional numerical simulation on flow field for ultra-
supercritical valve. Yu and Yu (2007), studied on the sensitivity analysis for
structure parameters on valve in normal temperature. But there are rarely related
documents discussed the method of optimizing sealing performance for the GVTP.
The paper discussed the optimized method of enhancing sealing performance
for the GVTP. It chooses design variables from global parameters of the GVTP,
and appoints target function which is displacement distribution on sealing surface
counted by the FEM. It gets the optimal solutions through the sampling method of
Monte Carlo. The relationship of sensitivity response between global parameters
and displacement is counted. The sealing properties of the GVTP are improved
obviously by revising higher sensitivity parameters.
51.2 Mathematical Model
The global parameters of sealing performance of the GVTP are design variables
and the quantity is so numerous that is hard to optimize through general counting.
The sensitivity analysis is the method that judges the responsible degree of the
target function following with the disturbance of design variable. The optimized
target function can be chosen from the result of the FEM. For enhance the sealing
performance, the total deformation on the sealing surface of the GVTP is regarded
as target function to optimize, and convergence direction is approaching minimum
51 Sensitivity Analysis on Optimized Sampling 501
displacement. The Monte-Carlo method can decrease the quantity of sampling
obviously and enhance the accuracy of counting, and it plays the importance role
in the optimized process.
(1) Sensitivity analysis: Sensitivity analysis is a method which optimized the
responsive relationships between design variations and objective functions.
Sensitivity analysis is used to optimize design for searching the relationships
and the impact degrees of input parameters on different simulation outputs. It
can be applied in the study of sealing performance for the valve working in
thermal power station.
The testing model of sensitivity analysis imposes the variance-based technique.
The model can be shown in the form of Y = f(x1,x2,…,xn), where x1,x2,…,xn are
imported design variations and Y is the exported parameter as objective function.
The total variance is presented as follows (Saltelli et al. 2009; Chan et al. 2000):
X
k X
V ðY Þ ¼ Vi þ Vij þ . . .þV1;2;;n ð51:1Þ
i¼1 1 i\j n
where V(Y) is the total variance of exported variable Y. Vi is the main response of
xi, and the other parameters are analyzed the response degree of interaction.
The sensitivity indices can be computed using a Monte Carlo method (Homma
and Saltelli 1996). The principle is to generate randomly samples of parameters
within their permissible ranges and to estimate V(Y), Vi and V-i as follows
(Fesanghary 2009):
(A) Assumed the base sample dimension N.
(B) Point dimension N 9 k to two sampling matrices M1 and M2 and N.
(C) Define a matrix Ni formed by all columns of M2, except the ith column which
is taken from M1, and a matrix NTi complementary to Ni, formed with the ith
column of M1 and with all the remaining columns of M2.
(D) Compute the model output for all the input values in the sample matrices M1,
and, obtaining three column vectors of model outputs of dimension N-1:
(E) The sensitivity indices are computed based on scalar products of the above
defined vectors of model outputs.
1X N
Vi ¼ Y ðjÞ Y 0ð jÞ f02 ð51:2Þ
N j¼1
1X N
0 ð jÞ
Vi ¼ Y ðjÞ YT f02 ð51:3Þ
N j¼1
502 L. Yu et al.
And finally
Vi
Si ¼ ð51:4Þ
V ðY Þ
Vi
Sn ¼ 1 ð51:5Þ
V ðY Þ
where Si is the first-order sensitivity of the ith parameter. Vi is the sum of all the
variance parameter except i. f is the function of Y as variable. N is the times of
iteration. Using the displacement parameters S as the object function, and the
sensitivity response can be obtained to optimize the global parameters of the valve
for improving sealing performance.
(2) Sampling of the Monte-Carlo method: The basic theory of the Monte-Carlo
method is loose the iteration times by finite samplings. While the probability
distribution of the state variable xi (i = 1,2,…,n) is known and independent, it
generates the random count for xi fitting its probability distribution according
to the limit condition Z = g(x1,x2, …,xn). The series xi(i) is put into the state
function Z = g(xi) and can be present as independent series Z(i) by iteration.
Assuming the quantity M of state function as safety coefficient if Z(i) \ 1, and
security reserve if Z(i) \ 0, while the times of simulation are enough, the
frequency M/N is close to probability by the law of Theorem, and the damage
probability can be presented as Pf = P{Z-g(X1,X2,…,Xn)}. The Monte-Carlo
method can accurate fitting the distribute function of probability G(Z) for Z,
and counts the mean value uz and normalized value rz.
The program structure of the Monte-Carlo method is simple and bears no
relation to the computational efficiency and dimensionality of the problems to be
studied, lots of samples are needed to get precise results (Zhang et al. 2010). This
method can speed up the process for searching the design point, which is then used
as the sampling center and importance sampling is conducted. Monte-Carlo
method is further employed to obtain the failure probability (Su et al. 2009).
Using the importance sampling Monte-Carlo method can choose the core of the
sampling and reduce the sampling quantity of the non-core. The accuracy of
counting is improved and the time of simulation is deceased by the method. The
Monte-Carlo method has the function as follows:
100
N R ð51:6Þ
X I ½Gð xÞ hfXXððxxÞÞ hX ð xÞdx
1 ; Gð xÞ\0
I ½Gð xÞ ¼ ð51:7Þ
0 ; G ð xÞ [ 0
51 Sensitivity Analysis on Optimized Sampling 503
where X is the define region of X,I[G(x)] is the indicative function, x is the basic
random vector, and X = [X1,X2,…,Xn], fX(x) is the probability density function,
hX(x) is the importance sampling density function.
51.3 Simulation and Result
The material of the GVTP is forgeable and creep resistant, and the material is
ASTM A335 F92 according to Case 2179-6. There have not perfect materials to be
applied to thermal cycle system more than 600 C in internet so far, and the
material using in the temperature is on studying and testing period. ASTM A335
F92, P112 and E911 are used to valve of ultra-supercritical power unit widely and
F92 has better forging property and stability in higher temperature than former
materials, and it have less thermal coefficient of expansion and more thermal
conductivity coefficient which can reduce thermal stress on inside and outside
wall. These materials have better creep strength in high temperature fitting thermal
stress requirement caused by thermal expansion than other alloy steels, and the
creep strength working high temperature and 105 h can attain to 90–100 MPa. The
applied range of ASTM A335 F92 is wider than P112 and E911 in internet and it is
applied materials of the GVTP in this paper.
Sensitive analysis can get the solution directly by the FEM (Kahraman 2009;
Lin 1999), and use the Monte-Carlo method to reduce the quantity of sampling
points (Melchers and Ahammed 2004; Ahammed and Melcher 2006). The per-
formance of F92 is attained to physical parameters for counting of the FEM. The
operating station is 33.1 MPa and 560 C, the medium is superheat steam flowing
12.3 m/s. The initial conditions are defined to the FEM for getting the distribution
of total deformation on the sealing surface of the GVTP. The performance of A335
F92 can be seen in Table 51.1.
The methods of structural optimization are control deformation mostly, and the
processes are making minimization of deformation with increasing the quality of
the structure (Wang et al. 2009; Pettit and Wang 2000). The displacement on
sealing surface of the GVTP can be seen as the deformation and defined as
objective function. Constraint condition is certain control domains for solving the
Table 51.1 Performance of ASTM A335 F92
Parameters Value
Temperature T [C] 610
Allowable stress [r] [MPa] 68.0
Tensile strength rb [MPa] 620
Yield strength rs [MPa] 440
Coefficient of thermal conductivity k[W/(mK)] 45.5
Specific heat capacity cp[J/(kgK)] 460
Density q[kg/m3] 7820
504 L. Yu et al.
Table 51.2 Performance of design variables
Lab Range [mm] Distribution Approximation direction
P1 142 / 160 ? 175 N I
P2 130 / 150 ? 176 R D
P3 247 / 260 ? 278 N I
P4 328 / 340 ? 369 A R
P5 187 / 250 ? 240 N D
P6 260 / 270 ? 280 R D
P7 20 / 25 ? 30 A I
P8 30 / 60 ? 90 N R
P9 75 / 90 ? 115 A I
P10 95 / 115 ? 130 N R
extreme value of target function. The improper constraint condition can leads to
lengthy solving process if the range is oversize, or makes more errors if the range
is undersize. The constraint condition of design variable is listed in Table 51.2.
There are 10 parameters been distributes as Pi (i = 1,2,…,10), and value range is
marked in the bracket. The distribution of sampling has three patterns, N means
normal distribution, A means average distribution and R means random distribu-
tion. The approximation direction of iteration has three patterns, I means
increment direction, D means decrement direction and R means random direction.
51.4 Simulation and Result
The minimum total deformation is defines as target function for optimized. By the
Monte Carlo method, the sampling points choosing in global parameters decrease
from 240 to 156 and iterating times is reduced from 1032 to 697. The efficiency
and accuracy of counting is improved obviously. The sensitive relationship of
structure parameters can be expressed by bar chart directly, respond degree of
sensitivity can be contrasted by the length of structure parameters, the larger length
mean it have more affecting degree to objective function and the shorter length
mean the sensitive degree is lower than others. Positive value of sensitive response
means sensitive degree of objective function increases follow with the value
increasing of structure parameter, and negative value means sensitive degree
increases follow with the value reducing. The sensitive relationship of structure
parameters can be seen from Fig. 51.1.
In Fig. 51.1, there have 10 series global parameters be chosen to sensitivity
analysis. Absolute values of sensitive distribute from 0.73 to 0.04, and composed
by 3 negative values and 7 positive values. The responsive relationship of sensitive
analysis can be seen that P1 is the most sensitive parameter for the displacement
on sealing surface and its value of sensitivity is -0.87, its means that the total
deformation should increases with reduces of P1. The sensitivities of P5 and P3 are
51 Sensitivity Analysis on Optimized Sampling 505
Fig. 51.1 Sensitive P1 P2 P3 P4 P5 P6 P7 P8 P9 P10
1.0
relationship of structure
parameters 0.8
0.6
Sensitive Response
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
Global Parameters
0.73 and 0.48 that are less than P1. The sensitivity of P4 is 0.04 that is the
minimum value than the others and it means the P4 can be neglected in optimized
process. To modify more sensitivity parameters in the process of design and total
deformation can be improved obviously, and this project should be used at design
firstly.
In Fig. 51.2, the responsive figure is three-dimensional space range that is
constituted by P1, P5 and maximum total deformation on sealing surface of the
GVTP. It is shown that maximum total deformation is greater following with the
increase of P1 and P5. The maximum equivalent stress is minimum value while
P1 is 164.5 mm and P5 is 211.4 mm, and its value is 7.7 9 10-2mm less than
10 9 10-2 mm that is the maximum deformation of allowable sealing, and
maximum total deformation is reduced by 21 %, so the series parameters should
be taken into account in design process. The design parameters can interfere each
other and lead to counting error, so the change arrange of design parameters should
Fig. 51.2 Optimized result
0.150
of total deformation and
Total Deformation Maximum u[mm]
parameters
0.135
0.120
0.105
0.090
0.075
174
192 168
200
208 162
P5[
mm
216 156 m]
224 [m
] 150 P1
506 L. Yu et al.
no more than 15 %. It has not extreme values in the optimizing process at some
times and the choosing method of noninferior solution on multiobjective optimi-
zation is important.
51.5 Conclusion
The sealing performance of the GVTP is hard to amend for concerning with a large
number of global and operation parameters. By the sensitivity analysis, the number
of the parameters is decreased to 10, and the displacement on the surface of the
GVTP is affected directly by these major parameters for deciding with sensitivity
analysis. It is shown that P1,P3 and P5 are the higher sensitivity than others, and
correcting the parameters can enhance the sealing performance. The number of
sampling is decreased by 35 % and maximum total deformation is reduced by
21 %.
As usually, the way of improving sealing performance of the GVTP is
strengthen the local place out of leaking point. By using the sensitivity analysis,
there has higher pertinence in the process of optimized, the quantity of design
variable is reduced obviously. Improving the global parameters of higher sensi-
tivity which leads to leaking, the sealing performance of the GVTP is optimized
and the safe degree of power system is ensured.
References
Ahammed M, Melcher RE (2006) Gradient and parameter sensitivity estimation for systems
evaluated using Monte-Carlo analysis. Reliab Eng Syst Saf 91(10):594–601
Alegre JM, Preciado M, Ferren D (2007) Study of the fatigue failure of an anti-return valve of a
high pressure machine. Eng Fail Anal 14(2):408–413
Chan K, Tarantola S, Saltelli A, Sobol IM (2000) Variance-based methods. In: Saltelli A, Chan
K, Scott EM (eds) Sensitivity analysis. Wiley, New York, pp 167–197
Chen ZB, Li GQ (2007) Low cyclic fatigue life study for the intermediate pressure main steam
valve housing in the power station. Heat Treat Met 32(12):122–127
Fesanghary M, Damangir E, Soleimani I (2009) Design optimization of shell and tube heat
exchangers using global sensitivity analysis and harmony search algorithm. Appl Therm Eng
29(8):1026–1031
Homma T, Saltelli A (1996) Importance measures in global sensitivity analysis of nonlinear
models. Reliab Eng Syst Saf 52:1–17
Kahraman A (2009) Natural model of planetary gear trains. J Sound Vib 173:125–130
Lin J, Parker RG (1999) Analitical characterization of the unique properties of planetary gear free
vibration. J Vib Acoust 228(1):100–128
LU HC (2007) The development of super-critical and ultra super-critical thermal power plant
valve. Valve 36(3):1–4
Melchers RE, Ahammed MA (2004) Fast approximate method for parameter sensitivity
estimation in Monte-Carlo reliability. Comput Struct 82(6):55–61
51 Sensitivity Analysis on Optimized Sampling 507
Pettit RG, Wang JJ, Toh C (2000) Validated feasibility study of integrally stiffened metallic
fuselage panels for reducing manufacturing costs. Langley Research Center, Hampton
Saltelli A, Tarantola S, Campolongo F (2009) Sensitivity analysis as an ingredient of modeling.
Stat Sci 15(4):377–395
Su C, Li PF, Han DJ (2009) Importance sampling monte-carlo method based on neumann
expansion response surface techniques. Eng Mech 26(12):1–11
Tao ZL, Cai DS, Yan CL (2009) Three dimensional numerical simulation and experimental study
on fluid field of control valve in power station. J Eng Thermophys 24(12):63–69
Wang XM, Liu ZY, Guo DM (2009) Sensitivity analysis of topology optimization micro flexible
mechanical structure by homogenization method. Mech Eng China 10(11):1264–1267
Wang AL, Wu XF et al (2010) Multidisciplinary optimization of a hydraulic slide valve based on
CFD. J Shanghai Jiaotong Univ 44(6):1767–1772
Xiang XW, Mao JR, Sun B (2006) Numerical investigation of flow characteristic of control valve
of steam turbine in the entire range of operating mode. J Xi’an Jiaotong Univ 40(7):290–296
Yu L, Yu SR (2007) Sensitivity analysis on performance of sealing pressure in triple offset
butterfly valves. Fluid Mach 32(3):163–165
Zhang Y, Zhu LS, Zhang YM (2010) Reliability based sensitivity analysis of mechanical strength
via Quasi-Monte Carlo Method. J Northeast Univ 31(11):1594–1598
Zhu QG, Chuan GG (2010) Numerical simulation on unsteady flow field in the main stop and
control valve system of a 1000 mw ultra-supercritical steam turbine. J Chin Soc Power Eng
30(8):743–749
Zhu GY, Lei LJ (2011) Water cone valve pilot valve port cfd f low field analysis and structure
optimization. Sci Technol Eng 11(2):591–598
Chapter 52
Study on Application of Logistic Curve
Fitting and Forecast from Inbound
Tourist Market
Wei-qi Tan
Abstract The rising of logistic curve basically coincides with the development and
growth of the tourist market. It has great advantages in improving forecast precision
to use the logistic growth curve to fit the developing trend of the tourist market and
to make forecast. The calculation method of K value is proposed in the paper
according to the characteristics of curve symmetry. Based on this, a and r parameter
values can be calculated and the logistic curve model of China’s inbound market
can be fitted by using the method of general curve regression. By the test of v2 ,
fitting logistic curve regression meets the requirements. So logistic curve regression
can be used as a general method for forecasting the tourist market.
Keywords Advantage Characteristics Fitting Logistic curve Parameter
52.1 Introduction
In the recent 30 years, tourist industry is treated as pillar industry of national
economy development in our country. We hope to increase the economy and
promote the proportion of the third industry with the development of tourism so as
to optimize the industrial structure. However, tourism economy has significant
volatility with the influences of natural factors, worldwide social economic factors,
and so on. The volatility of the growth in tourism economy makes tourism
investment a high risk one. So it’s an important means of reducing the risk of
tourism investment to improve the prediction quality of the tourist market. The
quantitative forecast method mainly uses historical data to build a quantitative
model and then use the model to forecast the development and clinic, psychology
W. Tan (&)
Anqing Vocational and Technical College, Anqing, Anhui, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 509
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_52,
Ó Springer-Verlag Berlin Heidelberg 2013
510 W. Tan
Fig. 52.1 A logistic vurve
and marketing (Huang et al. 2009). With the influences and restrictions from the
growth rate, environmental restrictions and time, the logistic growth curve begins
to slow down. Once it reaches the critical point, it grows rapidly but never
indefinitely. When it gets saturated, the growth speed slows down because of
intense competition for resources. The growth curve is slightly elongated as an
‘‘S’’ which is shown in Fig. 52.1: changes in future. Based on time series, there are
various of quantitative methods that can be used to predict. According to math-
ematical principles, there are two kinds: autoregressive conditional heteroske-
dasticity model and stochastic volatility model. But the two methods are naturally
limited in predicting tourist market with greater volatility. This paper is going to
introduce the fitting and forecast of the logistic curve to tourist market in order to
improve the quality of the forecast (Wand et al. 2009).
52.2 Research Background
52.2.1 The Application and Research of Logistic Curve
Equations
Logistic curves are derived from the study process of population growth. They
have been widely used in the field of natural science and social science. The curve
equations can be shown as follows:
k
y¼ k [ 0; a 2 R; r [ 0 ð52:1Þ
1 þ aert
Among the letters, k represents the environment or resource limitation, a is an
undetermined constant, r for growth and t for time. It has revealed the law of changes
in growth rate, environmental restrictions and time. The similar relationship
52 Study on Application of Logistic Curve Fitting and Forecast 511
between them widely exists in natural science and social science. The curve has been
widely used in the field of natural science and social science. Nowadays, the logistic
regression is a commonly used analysis and forecast method in social science.
52.2.2 The Application of Logistic Curve Equations
for Tourism Industry
The logistic curve model is applied to fit and forecast the population growth,
economic development and life cycle of tourism areas, which has significant
effects (see Hu 2011; Huang et al. 2011; Li 1993; Yang 2008, 2009).
Based on the study of logistic curve equations, a Canadian scholar Butler
combined them with changes of tourist destinations and put forward the theory of
life cycle of destination in 1980. He divided the tourist destinations into 5 different
growth stages: the exploration stage, the involvement stage, the development
stage, the consolidation stage and the stagnation stage. From then on, many
scholars did researches on the theory of life cycle of destination. Although there
are different views on it, scholars all agreed that the law of life cycle exists in
developing tourism products and tourist areas. This law is of important theoretical
and practical significance in forecasting the tourist market. Due to our growth
tourist market, the trend of tourist market development increases sharply when we
use regression analysis method to forecast, which cannot reflect the medium-term
and long-term trend of the development in the tourist market. Logistic curve
equations are established on the basis of the trend in growth forecast, so they can
overcome the defects and are of great significance to the prediction of the tourist
market (Zhang and Xue 2009).
52.3 Characteristics and Advantages of Logistic curves
There are exponential function curves, logarithm function curves, power function
curves, hyperbolic function curves, sigmoid curves, and so on. They can be used to
fit curves, analyze and predict according to the scatter plots. Scholars have con-
ducted extensive research on the logistic curve fitting. Several methods are used to
determine the parameters : the search method, the Gauss–Newton method, the
improved Gauss–Newton–Marquit method, and so on. Since the coupling problem
between the parameters cannot be solved in the process of solving the problems
which need alternate iterations, the selection of initial iteration value is always the
key point of being successful during the solving process (Huang et al. 2011). The
determination of parameters of the logistic curve equation is closely connected
with the characteristics of the curve. We need to conduct research on the char-
acteristics of logistic curve equations to find more scientific and reasonable
512 W. Tan
methods for parameter estimation. Besides the logistic curve shown in Eq. (52.1),
there is another curve equation (52.2):
k
y¼ k [ 0; a 2 R; r [ 0 ð52:2Þ
1 þ eart
Among the letters, k represents saturated level and r for growth factor. The ea in
formula (52.2) can be regard as a in formula (52.1). In this way, (52.1) and (52.2)
are coincident in essence. This article uses the curve shown in formula (52.1).
52.3.1 Characteristics of the Logistic Curve
1. Asymptotic lines
k k
lim rt
¼ 0; lim ¼ k;
t!1 1 þ ae t!þ1 1 þ aert
We can see from (Jin 2010) that the logistic curve has two asymptotic lines: y = 0
and y = k
2. Monotonicity
The speed function from the growth process of the logistic curve can be formed
by working out the first-order derivative of the logistic curve.
karert
Since y0 ¼ ð1þae rt Þ2
[ 0, the logistic curve is monotonically increasing.
3. Three key points
If we want to work out the second-order derivative of the logistic curve, we can
rt
ðarert rÞ
have y00 ¼ kareð1þae rt Þ3
. Make it equal to 0, we have t ¼ lnr a. When
t\ lnr a ; y00 [ 0, the curve is concave; when t [ lnr a ; y00 \0, the curve bulges.
If we want to work out the third-order derivative of thelogistic curve, we can
3 rt
ð14arert þa2 e2rt Þ
have y000 ¼ kar e
ð1þaert Þ4
. Make it equal to 0, we have
1 4arert þ a e2 2rt
¼ 0. Solve the equation we have t1 ¼ ln a1:317 r ;
ln aþ1:317
t2 ¼ r . In this way, there are three key points in the logistic curve:
t1 ¼ ln a1:317
r ; t2 ¼ lnr a ; t2 ¼ ln aþ1:317
r .
52.3.2 Advantages of the Logistic Curve
The curve can be divided into four growth stages.
1. Formative period: 0 t1 ¼ ln a1:317
r , the market has just formed and grows
slowly.
52 Study on Application of Logistic Curve Fitting and Forecast 513
2. Growth period: t1 ¼ ln a1:317
r t2 ¼ lnr a, the market has developed and grows
rapidly.
3. Mature period: t2 ¼ lnr a t2 ¼ ln aþ1:317
r , the market continues to grow but the
growth speed has decreased.
4. Degenerating stage: t2 ¼ ln aþ1:317
r þ 1, the growth speed continues to fall
and eventually saturates. We can see from the above analysis that the logistic
curve and the process of tourism market development are basically consistent.
In this way, using the logistic growth curve to fit and forecast the development
of tourist market has great advantages (Yang 2009).
52.4 The Parameter Estimation in Logistic Curves
52.4.1 General Methods for the Fitting Curve
The curve fitting refers to the process of conducting curve regression analysis with
two variable data to get a significant curve equation. Usually there are three steps:
(1) Choose appropriate curve type according to the exact relationship between
variable X and Y. (2) As for the selected curve type, configurate linear regression
equation with least square method after linearization and test significantly. (3)
Convert linear regression equation into the corresponding curvilinear regression
equation and make inference on the related statistical parameters. The logistic
curve has k, r, a the three selected parameters, so the general curve fitting can’t
determine the parameters (Yang 2008)
52.4.2 The Determination of Parameters in the Logistic
Curve
52.4.2.1 The Determination of Parameter
Choose three time points willfully, make it meet t2 ¼ t1 þt
2 , we have:
3
8 rt1
< y1 ¼ k=ð1 þ ae Þ
>
y2 ¼ k=ð1 þ aert2 Þ
>
:
y3 ¼ k=ð1 þ aert3 Þ
Then,
y22 ðy1 þ y3 Þ 2y1 y2 y3
k¼ ð52:3Þ
y22 y1 y3
514 W. Tan
Once k is determined, the value of a and r can be determined according to the
general methods for the curve fitting (Wu and Huang 2004).
52.4.2.2 The Estimate of Parameters
Take the logarithm from both sides in Eq. (52.1), then
ky
In ¼ In a rt ð52:4Þ
y
If y0 ¼ ln ky
y ; then we get the linear regression equation:
by 0 ¼ In a rt
Parameters a and r can be determined by Eq. (52.5), of which
9
r ¼ SPy0 t =SSt =
In a ¼ y0 þ rt ð52:5Þ
;
a ¼ eIn a
X 1X 2
X
SPy0 t ¼ ðti tÞ2 ¼ ti2
ti ;
n
X X 1 X X 0 1X 0 1X 0
SSt ¼ ðti tÞðy0i y0 Þ ¼ tiy0i ti yi t ¼ tiyi ¼ yi
n n n
52.4.3 Parametric Test
Since the logistic regression equation curve still contains parameter k besides
regression parameters a and r, we need to discuss the testing method for goodness-
of-fit (Lie and Zeng 2002). The related coefficients are as follows:
P 2
ðy0i ^
y 0Þ
i¼1 SSR 1
R2 ¼ 1 ¼ ¼ :
Pn
SST 1 þ SSE
ðy0i y0 Þ2 SSR
i¼1
The SST is for total sum of square, the SSE for residual squares and the SSR for
sum square within groups. There are two types of distribution: SSE 2
r2 v ð1Þ and
SSR 2 2
r2 v ðn 1Þ. We use distribution v ðn 1Þ for goodness-of-fit test. Look up
the data in chart v2 , if we take a ¼ 0:05 and have v2 \v20:05 ðn 1Þ, the equation
52
Table 52.1 China’s inbound tourist numbers from 1978 to 2009 (unit: one hundred million)
The particular year 1978 1979 1980 1981 1982 1983 1984 1985
t 1 2 3 4 5 6 7 8
Inbound tourist numbers 0.18092 0.42039 0.57025 0.77671 0.79243 0.9477 1.28522 1.78331
The particular year 1986 1987 1988 1989 1990 1991 1992 1993
t 9 10 11 12 13 14 15 16
Inbound tourist numbers 2.28195 2.69023 3.16948 2.45014 2.74618 3.33498 3.81149 4.15269
The particular year 1994 1995 1996 1997 1998 1999 2000 2001
t 17 18 19 20 21 22 23 24
Inbound tourist numbers 4.36845 4.63865 5.11275 5.75879 6.34784 7.27956 8.34439 8.90129
The particular year 2002 2003 2004 2005 2006 2007 2008 2009
t 25 26 27 28 29 30 31 32
Inbound tourist numbers 9.79083 9.16621 10.90382 12.02923 12.49421 13.18733 13.00274 12.64759
The particular year 2010
Study on Application of Logistic Curve Fitting and Forecast
t 33
Inbound tourist numbers 13.37622
Source National Bureau of Statistics of China & National Tourism Administration of the People’s Republic of China
515
516 W. Tan
performs well in fitting, which means the expected value is mainly fit with the
actual value. The statistics of v2 can be described in the equation below:
X ðyi ^yi Þ2
v2 ¼ ð52:6Þ
^yi
52.5 The Fitting Logistic Curve of China’s Inbound
Tourist Market
52.5.1 Source of the Data
The data in Table 52.1 were collected from the original data provided online from
National Bureau of Statistics of China and National Tourism Administration of the
P.R.C.
52.5.2 The Determination of Parameters
Substitute the data of t = 16, 17, 18 and calculate K according to formula (52.3),
then k = 14.257.
By using dfittool matlab, we get r = 0.189, lna = 3.822. Put SSE: 3.183, R-
square: 0.9711, Adjusted R-square: 0.9701 in goodness-of-fit test and the results
showed that the line fit well. Calculate lna = 3.822 and we have a = 45.6955.
And we can get the logistic regression curve equation of China’s inbound tourism:
14:257
^y ¼ ð52:7Þ
1 þ 45:6955e0:189t
52.6 Conclusion
The rising of logistic curve basically coincides with the development and growth
of the tourist market. It has great advantages in improving forecast precision to use
the logistic growth curve to fit the developing trend of the tourist market and to
make forecast. The growth patterns of China’s inbound tourist market follow the
logistic curve equations which can be used to fit and forecast China’s inbound
tourist market. There are many kinds of tourist markets in China. We can consider
using logistic curve equations for regression forecast provided that their growth
scatter plots are similar to the sigmoid curves.
52 Study on Application of Logistic Curve Fitting and Forecast 517
References
Hu X-h (2011) Parameters estimation for logistic curve and application. Math Theory Appl
31(04):32–36 (in Chinese)
Huang Y, He P, Li Z (2009) Empirical study on product life cycle based on logistic model.
Jiangsu Univ Sci Technol (Soc Sci Edn) 9(04):51–55 (in Chinese)
Huang H, Ma F, Ma Y (2011) Logistic curve model for regional economy medium-term and
long-term forecast. J Wuhan Univ Technol (Inf Manage Eng) 13(01):94–97 (in Chinese)
Li Z-w (1993) Foundation and application of gray logic curve model. Math Pract Theory
23(01):49–52 (in Chinese)
Lie Q-j, Zeng Q (2002) Logistic regression model and its research progress. J Prev Med Inf
18(05):417–420 (in Chinese)
Q-s Jin (2010) The foundation of higher vocational mathematics. Jilin University Press, p 79 (in
Chinese)
Wand J-w, Han Y-q, Wand H-z (2009) Application of ecological mathematical models to
population predication—taking Daqing for example. Saf Environ Eng 16(07):30–35 (in
Chinese)
Wu J, Huang Z-f (2004) Study on the application of logistic curve simulating tourism destination
lifecycle. Geogr Geo-Inf Sci 20(05):91–95
Yang L-F (2008) A study on the law of Chinese internet-user growth based on logistic curve.
J Xiamen Univ Technol 17(04):86–89,108 (Chinese)
Yang C-F (2009) Construction of tourism destination state forecasting model and case study. Res
Sci 6:1015–1021 (in Chinese)
Zhang X-M, Xue D (2009) A comparative study on two mathematical models of tourism area life
cycle. Tour Sci 29(04):6–13 (in Chinese)
Chapter 53
Study on Decision Mechanism Choosing
by Cost Model for Projectized
Organization
Hua-ming Zhang
Abstract The cost of making decision in enterprise differs when they have dif-
ferent decision mechanism. So decision mechanism choosing is very important. A
model for calculating the cost of projectized organization is designed in this article
with the parameters of the organization information and the relationship among
projects. The cost of projectized organization is analyzed under different decision
mechanism, on the base of which some tactic is advanced for projectized orga-
nization to choose decision mechanism.
Keywords Decision mechanism Organization structure Projectized organization
53.1 Introduction
Organization is a collaboration system in which we consciously coordinate two or
more persons’ activity or capacities (Barnard 1948). The objective of organization
is to organize the activities of members inside the system, confirm the best purpose
of operation and to rule the assignment of the members and their relationship
(Brown and Duguid 1991). By building organization structure, we can easily
manage system destination so that it can be realized (Joseph et al. 1999). The goal
of organization structure contains division of labor, group of labor and activity
collaborate (Stephen and Timothy 1997). When establish organization structure,
we should think about activity specialization, activity departmentalization, com-
mand chain, span of management, centralization and separation of powers and
normalization (Gang and Sun 2006).
H. Zhang (&)
School of Economics and Management, Nanjing Forestry University, Nanjing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 519
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_53,
Springer-Verlag Berlin Heidelberg 2013
520 H. Zhang
In projectized organization, project departments are independent to a certain
degree (Stephen and Timothy 1997; Project Management Institute 2000). Each
project department has its own technician and administration framework. It keeps
weak touch with matrix enterprise by phased progress report. Some enterprises
give detailed discipline about finance, human resource and administration. But
some enterprises leave all the duty and power to project departments (Harold 2002;
Xing 2003).
According to the distinction of centralization degree of the enterprise and the
competitive relationship and coordinate relationship between the project depart-
ments, the decision mechanism can be sorted as follow (Masahiko et al. 2001).
Classical mechanism is the simplest decision mechanism. The enterprise
decides the working capacity of the departments on the enterprise information and
the department information. The functional departments carried out the work.
In separation mechanism functional departments decide the work capacity on
its predicted department information but enterprise information.
In information assimilation mechanism, functional departments decide the work
capacity on the same predicted enterprise information but their own information.
They undertake the enterprise responsibility together.
In horizontal mechanism, functional departments decide the work capacity on
its predicted department information and the same predicted enterprise
information.
In information decentralization mechanism, functional departments decide their
work capacity by its predicted enterprise information and predicted department
information.
Information dissimilation mechanism is similar to assimilation mechanism.
Functional department decide their work capacity on its predicted enterprise
information. The difference between the two mechanisms is that in the former the
enterprise information is different for departments but in the latter the enterprise
information is same.
53.2 Modelling
When building the model, we hypothesize that the enterprise decides its work
capacity to minimize expected cost. The objective of building the model is to
analyze how the decision mechanism affect expected cost and which mechanism is
suitable to projectized organization. So that advice can be drawn for enterprise to
choose decision mechanism.
For the convenient of studying, the number of project departments in the
enterprise is represented by parameter n. The projectized organization can be
described in Fig. 53.1, (Lian and Fu 2007). In this kind of organization there may
be functional departments in the enterprise, but they can not affect project. And
executive of the enterprise directly control project manager.
53 Study on Decision Mechanism Choosing by Cost Model 521
Fig. 53.1 Projectized Enterprise
organization
Project manager p1 Project manager p2 Project manager pn
Project member Project member Project member
Project member Project member Project member
Project member Project member Project member
... ... ...
To build the cost model of projectized organization, we suppose parameter a as
the factors that affect both project departments such as project alteration or break
off. We call them system oscillation. The parameter ci(i = 1, 2) means the
information changing which occurs in project departments such as machine
stoppage or tool damage and can only affect one project department. We call them
individual oscillation. Both the average of a and c are 0, and variance are fi and fi .
They are random variable.
With the increase of the number of projects, it is inevitable that the relationship
of competition and coordination emerge between project departments. They
compete for capital and human resource in enterprise. The parameter B represents
the competition level between departments. The parameter D represents coordi-
nation level. They are treated as constant. The relationship between parameter B
and D represents the relationship of competition and coordination between func-
tional departments.
The work capacities of project departments is represented by variable yi(i = 1,
…, n), yi 3 0.
The operating cost of the enterprise can be conducted as follow.
X
n
1X n X n 2
c ¼c ða þ ai Þyi þ B yi þ yj
1
2 i¼1 j¼iþ1
ð53:1Þ
1X n X n 2
þ D yi yj
2 i¼1 j¼iþ1
Project department decides the value of yi on the information wi (i = 1, …, 2)
given by enterprise, the constitution of which is decided by on the decision
mechanism of the enterprise. The expected operating cost is as follow.
2 X 3
n
1X n X n 2
6 c ð a þ a Þy
i i þ B y i þ y j 7
6 1
2 i¼1 j¼iþ1 7
6
E6 7 ð53:2Þ
X n X n 2 7
4 1 5
þ D yi yj
2 i¼1 j¼iþ1
522 H. Zhang
53.3 Decision Mechanism Analyzing
In classical mechanism, all decision-making power concentrates upon enterprise
level. So the decision criterion is that choose suitable yi so that expected operating
cost is minimum. When partial differentials of yi are all equal to zero, expected
cost is minimum.
By solving these equations can get a conclusion that when y1 ¼ y2 ¼ ¼
yn ¼ 0 minimum c is c.
Under separation decision mechanism, variable wi DH ¼ ai þ ei refers to the
information project departments decided on.
The variable ei is the prediction error. It fit normal distribution with average
zero and variance re 2 . The prediction error of department fit the same distribution.
And they are mutual independent. We also hypothesize that the distribution of
system oscillation individual oscillation and prediction error are mutual indepen-
dent. AS cost function is quadratic and the random variables related to it fit to
normal distribution, decision criteria is linear equation about wi . yi DH Can be
defined as
yi DH ¼ k þ twi ð53:3Þ
Substituting (53.3) into cost function (53.2) can get minimum value of cost
function.
Decision criteria of separation mechanism is
QDH wi DH 1
yi DH ¼ ð53:4Þ
BþD n1
The minimum of expected cost is
QDH ra0 2 n
C DH ¼ C ð53:5Þ
B þ D 2ðn 1Þ
2
The parameter QDH is defined as QDH ¼ r 0r2aþr
0
e
2.
a
The parameter QDH is the proportion of necessary information ai in the infor-
mation wi DH of i department which has prediction error. It reflects the accuracy
departments processing information. Bigger value indicates information accurately
processed and smaller minimum value of expected enterprise cost. It is easy to get
the conclusion that the cost of separation mechanism is smaller than classical
mechanism.
In information assimilation mechanism all the project departments ignore
individual information oscillation. They consider the same system information
which has prediction error e0 . The variance e0 is independent to a. It obeys normal
distribution with average zero and variance re 2 . The system information can be
defined as: wAI ¼ a þ e0 :
53 Study on Decision Mechanism Choosing by Cost Model 523
Decision mechanism is
QAI wAI 1
yAI ¼
2B n1
Minimum value of expected cost function is
QAI ra 2 n
C¼C ð53:6Þ
2B 2ðn 1Þ
The parameter QAI in the equation is defined as follow.
ra 2
QAI ¼
ra þ re 2
2
Similar to QDH in separation mechanism, QAI reflects the accuracy enterprise
processing information. Bigger value indicates information accurately processed
and smaller minimum value of expected enterprise cost. At the same time the cost
is relative to system oscillation ra 2 , parameter B and the number of projects. The
cost will increase with the decrease of ra 2 and an increase of parameter B and n. If
system oscillation is very big assimilation mechanism is efficient.
In horizontal mechanism project departments consider the same system infor-
mation a þ e0 and its own predicted individual oscillation ci þ ei . The sum of
system information and individual oscillation can be defined as wi HH .
wi HH ¼ a þ e0 þ ai þ ei
Minimum value of expected cost function is
QHH ðra 2 þ ra0 2 Þ n
C¼C r 2 þr2
ð53:7Þ
ðB þ DÞ þ ðB DÞQHH raa2 þr e0 2 2ð n 1Þ
a
Decision mechanism is
QHH wi HH 1
yHH ¼ 2 þr 2
ðB þ DÞ þ ðB DÞQHH rraa2 þr e
2
n1
a0
The parameter QHH in the equation is defined as follow.
ra 2 þ ra0 2
QHH ¼
ra þ ra0 2 þ re 2
2
The parameter QHH reflects the accuracy processing information. Bigger value
indicates lower minimum value of expected enterprise cost. At the same time the
cost is relative to the sum of system oscillation and individual oscillation. It is also
relative to parameter B and parameter D and n. The cost will increase with the
decrease of ra 2 þ rc 2 and the increase of n, B ? D and B - D. Horizontal
mechanism would be suitable when system oscillation and individual oscillation is
524 H. Zhang
very big. When individual oscillation is very small its cost is approximate to that
of assimilation mechanism.
In decentralization mechanism project departments consider different system
information and different predicted individual oscillation. The sum of system
information and individual oscillation can be defined as wi DI .
wi DI ¼ a þ ai þ ei
Minimum value of expected cost function is
QDI ðra 2 þ ra0 2 Þ n
C¼C 2 ð53:8Þ
ðB þ DÞ þ ðB DÞQDI ra 2rþr
a
02
2 ð n 1Þ
a
Decision mechanism is
QDI wi DI 1
yDI ¼ ra 2
ðB þ DÞ þ ðB DÞQDI n1
ra 2 þra0 2
The parameter QDI in the equation is defined as follow.
ra 2 þ ra0 2
QDI ¼
ra þ ra0 2 þ re 2
2
The parameter QDI reflects the accuracy processing information. Bigger value
indicates lower minimum value of expected enterprise cost. At the same time the
cost is relative to the sum of system oscillation and individual oscillation. It is also
relative to parameter n, parameter B and parameter D. The cost will increase with
the decrease of ra 2 þ rc 2 and the increase of n, B ? D and B - D.
In dissimilation mechanism what departments consider is wi D which is defined
as wi D ¼ a þ ei :
Minimum value of expected cost function is
QD ra 2 n
C¼C ð53:9Þ
ðB þ DÞ þ ðB DÞQD 2ðn 1Þ
Decision mechanism is
QD wD 1
yD ¼ i
ðB þ DÞ þ ðB DÞQD n 1
The parameter QD in the equation is defined as follow.
ra 2
QD ¼
ra 2 þ re 2
The parameter QD reflects the accuracy processing information. Bigger value
indicates lower minimum value of expected enterprise cost. At the same time the
53 Study on Decision Mechanism Choosing by Cost Model 525
cost is relative to the sum of system oscillation and individual oscillation. It is also
relative to parameter B and parameter D. The cost will increase with the decrease
of ra 2 and the increase of B ? D and B - D. So dissimilation mechanism is
suitable when system oscillation is very big.
53.4 Comparative Analysis of Decision Mechanism
From the result conducted by the model, conclusion can be easily made as follow.
When the number of projects in the enterprise is not large the appropriate
decision mechanism is in connection to the proportion of individual oscillation or
system oscillation in all information.
In the fields that have rapid changing environment the system oscillation will be
large. If the individual oscillation in the project departments is very small and can
be neglected the efficiency of the four decision mechanisms in the front would be
equal. But the information processed in horizontal mechanism and decentralization
mechanism is more than that of assimilation mechanism and dissimilation
mechanism. In consideration of the cost for processing information it is proper to
choose assimilation mechanism and dissimilation mechanism.
In the fields that have mature technology and stable environment system
oscillation is relatively small and individual information is relatively large. The
cost of assimilation mechanism, dissimilation mechanism, horizontal mechanism
and decentralization mechanism is equal. The cost of separation is the lowest. So it
is more appropriate than the other mechanisms.
If the system oscillation and individual oscillation are all very large and cannot
be neglected horizontal mechanism and decentralization mechanism process more
information than other mechanism. So they are more efficient. But they are also
different. If the competitive factor in among projects is larger than coordination
factor decentralization mechanism will be more efficient. And horizontal mecha-
nism will be more efficient on the contrary.
With the increase of the number of projects in enterprise the competition among
projects would increase rapidly and the coordination among them would decrease
rapidly. If the number increase to a certain degree competition factor would be far
larger than coordination factor and the coordination can be neglected. The cost
function of the enterprise would increase with the increase of the number of
projects in the enterprise. If the number is large the expected cost would be large.
The extreme condition is that the number is infinite. The minimum of expected
cost will be c. At this time projectized organization would not be efficient. The
reason is that when the number of projects in enterprise is small the competition
among projects is small and can be solved by project managers. But with the
increase of the number in enterprise project managers can no longer solve
the competition by themselves. It is necessary to add new department to manage
the resource that all projects compete for. They are functional departments.
526 H. Zhang
53.5 Conclusion
In this paper cost model established with parameter defined as competition rela-
tionship and coordination relationship and the number of projects in the enterprise.
The cost of projectized organization is analyzed under different decision mecha-
nism. Selection strategy is proposed on choosing decision mechanism for pro-
jectized organization. So it is feasible to analyze the cost of projectized
organization quantitatively. But it is not enough to analyze organization only from
the perspective of cost. Operating efficiency and performance should also be
analyzed. These work remains to be further studied.
Acknowledgments This work is supported by Jiangsu Education Department Grant #2010SJB
630038 to Hua-ming Zhang.
References
Barnard C (1948) Organization and management. Harvard University Press, Cambridge,
pp 35–36
Brown JS, Duguid P (1991) Organizational learning and communities-of-practice: toward a
unified view of working, learning, and innovation. Organ Sci 2(1):40–57
Gang Z, Sun M (2006) Summarize of research on organizational innovation. Reach Dev Manag
18(4):24–29
Harold K (2002) A systems approach to planning, scheduling, and controlling. China Electronic
Industry Press, Beijing, pp 2–20, 74–103
Joseph MP, Heinz W, Harold K (1999) Essentials of management (An Asian Perspective). China
Machine Press, Beijing, pp 154–189
Lian M-K, Fu J (2007) Simulation modeling of organizational structure (in Chinese). J Shanghai
Jiaotong Univ 33(10):1314–1315
Masahiko A, Avner G, Paul M (2001) Toward a comparative institutional analysis. MIT Press,
Massachusetts pp 95–140
Project Management Institute (2000) A guide to the project management body of knowledge.
Project Management Institute, Newtown Square, pp 12–150
Stephen PR, Timothy AJ (1997) Organizational behavior. China Renmin University Press,
Beijing, pp 419–445
Xing Y-Q (2003) Discussion on process oriented organization (in Chinese). Soft Sci 17(4):42–45
Chapter 54
Study on Impact of Investor’s Herd
Behavior on Corporation
Investment Behavior
Hai-ming Wang and De-ming Zeng
Abstract From the perspective of investor irrationality, this paper empirically
analyzes the effect of investors’ herd behavior on corporate investment behavior.
The empirical results show that the buyer’s herd behavior has positive effect on the
level of corporate investment, and the seller’s herd behavior has inhibitory effect
on the level of corporate investment. The degree of influence on corporate
investment varies between the buyer’s herd behavior and that of the seller’s.
Keywords Corporate investment behavior Equity financing Herd behavior
Investor irrationality
54.1 Introduction
Corporate investment behavior refers to corporations invest financial resources to
expect for future returns. Traditional finance theory implied a hypothesis that
investors are rational. However, with the appearance of some abnormal financial
phenomenon which could not be explained by traditional financial theory, scholars
began to pay attention to the influences of investors’ irrational behavior on cor-
porate investment behavior and tried to research corporate investment behavior in
the consideration of the influence of capital market (Baker et al. 2003; Baker
2009). However, it is very difficult to measure the investors’ irrational degree
directly in fact. A lot of researchers indirectly measured the irrational degree by
measuring the impacts of investors’ irrational behavior on stock price. As there are
different types of investors irrationality such as herd behavior, over-confidence and
so on, current researches still could not explain what specific influences would
H. Wang (&) D. Zeng
College of Business Administration, Hunan University,
Changsha, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 527
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_54,
Ó Springer-Verlag Berlin Heidelberg 2013
528 H. Wang and D. Zeng
various investors’ irrational behaviors have on corporate investment. Therefore, it
is valuable and meaningful to study the influences of investors’ specific irrational
behavior on corporate investment behavior.
Since the 1980s, as one aspect of irrational manifestations of investors,
investors’ herd behavior has attracted more and more attention from scholars. Herd
behavior is the behavior that investors follow each other into and out of the same
securities or follow themselves into and out of the same securities within some
period ignoring the value of private information. Keynes (1936) mentioned this
concept first, and used the beauty contest game to illustrate such a phenomenon.
Later, scholars made a large number of empirical studies. The literature about herd
behavior of investors could be divided into two branches. First, Wermers (1999)
and Sias (2004) studied the buy and sell decisions of investors based on the results
of Lakonishok et al. (1992). Second, Christie and Huang (1995), and Chang et al.
(2009) derived a time series based on the distribution of returns around the market
to measure herd behavior of investors. These researches showed the existence of
herd behavior in securities market. Li et al. (2010), Chun et al. (2011), Li and Li
(2011) found that investors’ herd behavior exists generally in Chinese securities
market, and there were some differences between individual investors’ herd
behavior and institutional investors’. After then, scholars further researched the
effect of investors’ herd behavior on stock price. Most scholars argued that herd
behavior would aggravate stock price fluctuation and destroy market stability
(Chiang and Zheng 2010; Lao and Singh 2011; Maug and Naik 2011); while other
scholars believed that herd behavior would reduce stock price fluctuation and
stabilize the stock market (Wu et al. 2004).
To sum up, existing studies generally agreed on that investors’ herding behavior
exists in the securities market and investors’ herd behavior would affect the stock
price and its volatility, but few studied whether investors’ herd behavior impact
corporate investment and what the effects are. Considering the existence of herd
behavior universally in Chinese securities market, this paper tries to research the
influence of investors’ herd behavior on corporate investment by relaxing the
assumptions of traditional financial theory.
54.2 Hypothesis Development
As the subject of the economic behavior in securities market, investors will make
investment decisions in accordance with the information they obtained. What they
pursue is the maximum returns with fixed risks or the minimum risk with the fixed
returns. However, due to information asymmetry in reality, most investors will
follow others’ to make investment decisions leading to herd behavior happen.
Buying-in or selling-out the same stock of most investors at the same time will
greatly aggravate the fluctuation of the stock price, then cause the stock price to be
overestimated or underestimated, and finally affect corporate investment behavior.
This kind of influence can be divided into two situations as follows.
54 Study on Impact of Investor’s Herd Behavior 529
On one hand, in order to maximize the basic value and pursue the goal of doing
better and bigger, the corporate operator needs to increase investment and
implement projects with positive net present value. The buyer’s herd behavior
means that most investors buy a large number of the same stock during the same
period, and indicates that most investors are optimistic about the development of
corporate. This will result into the rise of stock price, which deviates from its inner
value. And then it will ease financing constraint and financing costs accordingly,
providing an important guarantee for the smooth implementation of projects with
positive net present value in terms of financing. Therefore, the corporate operator
will invest more projects with positive net present value, and then expand the scale
of the corporate. On the contrary, the seller’s herd behavior means most investors
sell the same stock during the same period, which indicates that most investors are
not optimistic about the development of corporate. This will cause the stock price
to fall, which deviates from its inner value. And then it will increase financing
constraint and financing costs accordingly, influencing the implementation of
projects with the positive net present value. Therefore, the corporate operator will
give up some projects with the positive net present value to avoid getting into
financial dilemma when they are making investment decisions.
On the other hand, the corporate operator will take the corresponding measure
to maximize the stock price according to the stock price fluctuation. The buyer’s
herd behavior gives rise to the overestimated stock price, which is consistent with
the goal of the corporate operator. In order to maintain the high stock price, the
corporate operator will improve the level of investment to cater for the majority of
investors’ expectations. The seller’s herd behavior leads to the underestimated
stock price, which goes against the purpose of operators. In order to go through the
worse situation, the corporate operator will consider share repurchase at the
appropriate time. Share repurchase requires a large amount of cash, which imposes
restrictions on the corporate investment.
Therefore, the paper proposes the following hypotheses based on the above
analysis:
Hypothesis 1: the higher the degree of buyer’s herd behavior is, the higher the
level of corporate investment is.
Hypothesis 2: the higher the degree of seller’s herd behavior is, the lower the
level of corporate investment is.
54.3 Research Design
54.3.1 Sample Selection and Data Sources
Considering Chinese securities market has undergone tremendous changes since
the split-share reform, we select all A-share firms listing on the Shanghai Stock
Exchange (SHSE) and the Shenzhen Stock Exchange (SZSE) over the period from
2007 to 2011 as the research samples.
530 H. Wang and D. Zeng
Considering the availability of the data, we use the open-end funds as the
representative of investors. As the main force of the investors, the open-end funds’
herd behavior can be used to measure the herd behavior of investors to achieve the
empirical requirement of this paper. The original data of the investors’ herd
behavior is the open-end funds’ shareholding details of the annual report data. We
compare the number of shareholding stock i of the open-end funds in the period
t annual report with the number of shareholding the same stock in the period t-1
annual report. If the number of shareholding increased compared with the last
period, we call the open-end funds of stock i in the t period as net buying. On the
contrary, we call it as net selling. Further, we can get the number of investors who
are the net buyers or sellers of stock i in period t by data statistics. In order to
guarantee the validity, we eliminate abnormal sample data which the sum of the
net buying and net selling is less than five and the number of net buying or net
selling is zero. Then, we exclude the ST group, PT group, and companies whose
net assets are negative and financial data is missing. At last, we get 3049
observations.
All the data comes from the CSMAR database.
54.3.2 Variables Definition
There exists phenomenon of manipulating profits by asset impairment in Chinese
capital market, but it is more difficult to manipulate the data in the cash flow
statement. Therefore, this paper chooses the cash which is used to pay for con-
struction of fixed assets, intangible assets and other assets divided the final total
assets to measure the level of corporate investment.
We amend the LSV model constructed by Lakonishok et al. (1992) to measure
the herd behavior degree of investors. The LSV model can measure the degree of
deviation between the stock trading in the same direction transaction and the stock
trading in the independent transaction during the same period. The LSV model is
as follows:
HMi;t ¼ Pi;t EðPi;t Þ AFi;t ; ð54:1Þ
where HMt represents the herd behavior degree of investors, Pi,t is the proportion
of number of investors who are the net buyers of stock i to the number of investors
who are all trading of stock i; E(Pi,t) is the expected value of Pi,t; AFi,t is an
adjustment factor.
The LSV model calculation steps are as follows:
Firstly, the calculation formula of Pi,t is as follows:
54 Study on Impact of Investor’s Herd Behavior 531
Bi;t
Pi;t ¼ ; ð54:2Þ
Bi;t þ Si;t
where Bi,t is the number of investors who are the net buyers of stock i during
period t, Si,t is the number of investors who are the net sellers of stock i during
period t.
Secondly, according to the study of Lakonishok et al. (1992), we use the pro-
portion of all stock trades which are purchased by investors during period t. The
specific calculation formula is as follows:
Pi¼Ni;t
Bi;t
Pt ¼ Pi¼Ni;t i¼1 Pi¼Ni;t ; ð54:3Þ
i¼1 Bi;t þ i¼1 Si;t
Thirdly, the calculation formula of AFi,t is as follows:
AFi;t ¼ EPi;t EðPi;t Þ ¼ EPi;t Pt ; ð54:4Þ
where E|Pi,t-Pt| is calculated under the null hypothesis Bi,t following a binomial
distribution with parameter Pt. Thus,
P Bi;t ¼ k ¼ CNk i;t Pkt ð1 Pt ÞNi;t k ; ð54:5Þ
we can figure out AFi,t by substituting Eq. (54.5) into Eq. (54.4).
Fourthly, we calculate the herd behavior degree of buyers and sellers. Given
that the degree of herd behavior is different when buying and selling stocks, its
impaction on corporate investment will also be different. Therefore, according to
‘‘buy herding measure’’ and ‘‘sell herding measure’’ proposed by Wermers (1999),
in this paper HMt is defined as follows: if HMt [ 0, then HMt represents the
buyer’s herd behavior; if HMt \ 0, then HMt represents the seller’s herd behavior.
Values of HMt significantly different from zero are interpreted as evidence of herd
behavior. The greater the absolute value of HMt is, the more serious the investors’
herd behavior is.
On the basis of previous studies, in this paper we select some control variables.
Specific definitions are in Table 54.1.
Table 54.1 The control variable definition
Variable Definition
SIZEt The natural logarithm of asset
CASHt-1 The cash flow from operations in period t-1
CASHt The cash flow from operations in period t
SHAREt The holding ratio of largest shareholder against second-largest shareholder
LEVt-1 The ratio of total liabilities to total assets
GROWTHt The revenue growth rate
RETt The stock market-adjust returns
YEAR Year dummy variables
532 H. Wang and D. Zeng
54.3.3 Model Design
In order to verify the above hypothesis, we construct the model as follows:
It ¼ c þ b1 HMt þ b2 SIZEt þ b3 CASHt1 þ b4 CAHSt þ b5 SHAREt
X ; ð54:6Þ
þ b6 LEVt1 þ b7 GROWTHt þ b8 RETt þ b9 YEAR þ e1
where It is the level of corporate investment, HMt is the herd behavior of investors,
others are control variables.
54.4 Empirical Results
The results of regression results of model (54.6) are shown in Table 54.2.
The coefficient of investors’ herd behavior degree is significantly positive at the
95 % confidence level, suggesting that the higher the degree of buyer’s herd
behavior is, the higher the level of corporate investment is; and the higher the
degree of seller’s herd behavior is, the lower the level of corporate investment is.
Hypothesis 1 and hypothesis 2 are verified.
As the regression results of control variables, the larger the corporate scale is,
the higher the level of corporate investment is. This may be due to the domestic
listed corporation, especially large corporate tend to invest more to achieve the
goal of doing better and bigger. The greater the cash flow of previous period in
corporate is, the higher the level of corporate current investment is. The initial cash
Table 54.2 The total sample regression results
Variable Coefficient t-value p-value
(Constant) -0.1423*** -6.8546 0.0000
HMt 0.0205** 2.2188 0.0266
SIZEt 0.0086*** 9.0456 0.0000
CASHt-1 0.1216*** 9.3180 0.0000
CASHt 0.0985*** 7.3498 0.0000
SHAREt -0.0001** -2.5652 0.0104
LEVt-1 -0.0069 -1.0828 0.2790
GROWTHt -0.0001 -1.1787 0.2386
RETt -0.0011 -0.8638 0.3878
YEAR Control
Adjusted R2 0.1041
F-statistics 36.3991
(0.0000)
Durbin-Watson 1.9702
Note The superscript ** indicates obvious herd behavior at the level of 5 %, and the superscript
***
indicates obvious herd behavior at the level of 1 %
54 Study on Impact of Investor’s Herd Behavior 533
Table 54.3 Two subsamples regression results
Variable The buyer’s herd behavior The seller’s herd behavior
Coefficient Coefficient
(Constant) -0.1557*** -0.1101***
HMt 0.0277** 0.0211*
SIZEt 0.0092*** 0.0074***
CASHt-1 0.1347*** 0.1022***
CASHt 0.0865*** 0.1163***
SHAREt -0.0001** -0.0001*
LEVt-1 -0.0043 -0.0074
GROWTHt 0.0000 -0.0024***
RETt -0.0030* -0.0020
YEAR Control Control
Adjusted R2 0.1094 0.1000
F-statistics 20.6365 21.1208
(0.0000) (0.0000)
Durbin-Watson 2.0457 1.7991
Note The superscript * of HMt indicates obvious herd behavior at the level of 10 %, the super-
script ** indicates obvious herd behavior at the level of 5 %, and the superscript *** indicates
obvious herd behavior at the level of 1 %
flow is a kind of expression of the level of previous investment. Since corporate
investment has some continuity, the level of previous investment has certain
effects on current corporate investment. Consequently the initial cash flow has
positive effects on the level of corporate investment. The greater the current cash
flow is, the higher the level of current investment is. As corporate ownership is
more concentrated, the corporate operator will be more cautious for investing, so
the level of investment will be lower. In general, control variables are basically
consistent with the actual situation.
In order to eliminate the interaction of the data and explore the similarities and
differences of the influence which the buyer’s herd behavior and the seller’s have
on the level of corporate investment, regarding the data of the buyer’s herd
behavior and the seller’s as two subsamples, we make regression analyses
respectively. The results are shown in Table 54.3.
After controlling the scale of corporate, the cash flow, equity structure, lever-
age, revenue growth, stock return and annual factor, the regression coefficient of
buyer’s herd behavior is 0.0277, p value is 0.0415, which means that the coeffi-
cients is significantly positive at the 95 % confidence level, so the hypothesis 1 is
confirmed. While the regression coefficient of seller’s herd behavior is 0.0211,
p value is 0.0984, which means that the coefficient is significantly positive at the
90 % confidence level, so the hypothesis 2 is confirmed.
Further, comparing the regression coefficients of buyer’s herd behavior with the
seller’s, we can conclude that the buyer’s herd behavior has a greater impact on
corporate investment. The possible reason is that the buyer’s herd behavior is more
serious than the seller’s in securities market.
534 H. Wang and D. Zeng
54.5 Conclusion
Research results show that, the buyer’s herd behavior can significantly improve the
level of corporate investment; and the seller’s herd behavior has significant
inhibiting effect on the level of corporate investment. What’s more, the impact of
the buyer’s herd behavior on corporate investment is greater than the sellers’.
Acknowledgments This work was supported by the National Social Science Foundation of
China (07AJL005), Program for Changjiang Scholars and Innovative Research Team in Uni-
versity (IRT0916), and Science Fund for Innovative Groups of Natural Science Foundation of
Hunan Province of China (09JJ7002).
References
Baker M (2009) Capital market-driven corporate finance. Annu Rev Financ Econ 1(1):181–205
Baker M, Stein JC, Wurgler J (2003) When does the market matter? Stock prices and investment
of equity-dependent firms. Q J Econ 118(3):969–1005
Chang EC, Cheng JW, Khorana A (2009) An examination of herd behavior in equity markets: an
international perspective. J Bank Finance 24(10):1651–1679
Chiang TC, Zheng D (2010) An empirical analysis of herd behavior in global stock markets.
J Bank Finance 34(8):1911–1921
Christie WG, Huang RD (1995) Following the pied piper: do individual returns herd around the
market? Financ Anal J 51(4):31–37
Chun WD, Li XY, Li M (2011) Study on herd behavior in Chinese stock market based on
dispersion model. Forecasting 30(5):25–30 (in Chinese)
Keynes JM (1936) The general theory of employment, interest and money. Macmillan, London,
pp 133–134
Lakonishok J, Shleifer A, Vishny RW (1992) The impact of institutional trading on stock prices.
J Financ Econ 32(1):23–43
Lao P, Singh H (2011) Herding behaviour in the Chinese and Indian stock markets. J Asian Econ
22(6):495–506
Li XF, Li JM (2011) Investors’ individual herd behavior: distribution and degree—a matrix
model based on partitioning clustering algorithm. Stud Int Finance 28(4):77–86 (in Chinese)
Li CW, Yu PK, Yang J (2010) The herding behavior difference between institutional investor and
individual investor. J Financ Res 54(11):77–89 (in Chinese)
Maug E, Naik N (2011) Herding and delegated portfolio management: the impact of relative
performance evaluation on asset allocation. Q J Finance 1(2):265–292
Sias RW (2004) Institutional herding. Rev Financ Stud 17(1):165–206 (Spring)
Wermers R (1999) Mutual fund herding and the impact on stock prices. J Finance 54(2):581–622
Wu FL, Zeng Y, Tang XW (2004) Further analysis of Chinese investment funds’ herding
behavior. Chin J Manag Sci 21(4):7–12 (in Chinese)
Chapter 55
Study on the Evaluation Framework
of Enterprise Informationization
Qing-wen Yuan, Shu-wei Yu, Yuan-yuan Huo and Dan Li
Abstract Evaluation of enterprise informationization (EI) is an important part of EI
construction and applications, which has direct impact on the effect of the acqui-
sition of information technology, investors confidence and the healthy development
of following steps. This paper attempts to establish a systematic evaluation frame-
work of EI, analyze the process of that evaluation, and especially build the corre-
sponsive applied model, offering relative references for the work of EI evaluation.
Keywords Application stage Benefit evaluation Enterprise informationization
(EI) Evaluation model
55.1 Introduction
Enterprise informationization is a evolutionary process in which an enterprise, by
focusing on two objectives, the economic efficiency and the competitiveness of
enterprises, and by using the modern information technology, develops and uses
Q. Yuan (&)
Department of Finance and Economics, Shandong University of Science and Technology,
Jinan, China
e-mail: [email protected]
S. Yu
College of Economics and Management, Shandong University of Science and Technology,
Qingdao, China
e-mail: [email protected]
Y. Huo
College of Economics and Management, Shandong Women’s University, Jinan, China
e-mail: [email protected]
D. Li
Shandong Chenlong Energy Sources Group Co., Ltd, Tengzhou, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 535
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_55,
Ó Springer-Verlag Berlin Heidelberg 2013
536 Q. Yuan et al.
corporate information resources effectively and improves its own operation and
management and the abilities of research and development, meanwhile strength-
ening its core competitiveness (Ke and Li 2007).
In the process, the evaluation of EI is an important part of the information-
ization construction and application. The quality of the evaluation directly influ-
ences the effects of the EI application as well as the development in following step.
However, the current work on matter on the theory or on application has not
formed a mature and practical system, which is an obstruction on the healthy
development of informationization work. To sort out the evaluation of EI, forming
a certain constructive ideas to provide a reference for the preparation or being
implemented information technology companies is the purpose of this article.
Evaluation of EI is not only a guide to build enterprise informationization, but
also scales to measure the level of enterprise informationization. The evaluation
not only can guide enterprises to understand accurately the connotation of infor-
mation technology and to clear the purpose of information technology, but also
guide enterprises to correctly formulate the information technology strategy and to
protect the implementation of the informationization project. EI evaluation can not
only improve the overall quality, the sustainable development ability, and the
international competitiveness of enterprises, but also promote local economy and
social development. It is obvious that for improving the efficiency and the prof-
itability of EI process the evaluation work come into spotlight.
55.2 Evaluation and Significance of the EI
The evaluation of EI is the important part of the construction and application of
informationization. It is a comprehensive measurement and metrics to the EI
infrastructure, information resources, human capital, technology, transaction costs,
resource consumption, competitiveness and innovation capability and other
aspects, finding out the relationship between the enterprise information and
business performance and corporate value, and establishing the evaluation index
system so as to determine the level of corporate information technology and its
benefits. Measuring the EI, should not only reflect the level, but also reflects the
effectiveness and efficiency, should not only reflect the direct benefits, but also to
reflect the development potential and value.
Evaluation of EI is not only a guide to build enterprise informationization, but
also scales to measure the level of enterprise informationization. The evaluation
not only can guide enterprises to understand accurately the connotation of infor-
mation technology and to clear the purpose of information technology, but also
guide enterprises to correctly formulate the information technology strategy and to
protect the implementation of the informationization project. EI evaluation can not
only improve the overall quality, the sustainable development ability, and the
international competitiveness of enterprises, but also promote local economy and
55 Study on the Evaluation Framework of Enterprise Informationization 537
social development. It is obvious that for improving the efficiency and the prof-
itability of EI process the evaluation work come into spotlight.
55.3 The Content of Evaluation of EI
Evaluation of EI basically includes two aspects: the level evaluation of EI and
benefit evaluation of EI.
The evaluation of the level of EI is to measure from the stage of the EI and
basic indicators of information measure to the representation of the physical and
chemical form of indicators, including the infrastructure in terms of information
diffusion, information resources development and utilization of information sys-
tem applications, and from the organizational structure, quality of personnel as
well as corporate IT investment accounted for the proportion of fixed investment.
Construction of enterprise information has both economic benefits and social
benefits, meaning that it can not only product direct benefits, but also generate
indirect benefits. Its effectiveness and overall business performance are insepara-
ble. Therefore, to judge and evaluate the effectiveness of information technology,
it cannot simply treat informationization as a common project investment which is
viewed in isolation, but should start from the improvement of the overall corporate
performance, according to their differing stages, to comprehensively evaluate. In
the evaluation of enterprise information, because of the diversity of its evaluation
level and main body, the product of benefits has invisibility, stealth, lag, long-term,
accelerated increments and other features which make great difficulties to the set
of informationization evaluation index, determination of index weight, and the
choice of evaluation methods. This is one of the reasons for the formation of the
easy operation of the accepted system of scientific evaluation system.
55.4 Analysis on the Evaluation of EI
55.4.1 Framework of EI Evaluation
The evaluation of EI is a systematic work to analyze the factors ofi nter-
nal resources and capabilities, external environment, in the case of considering the
matching of the enterprise strategy and information strategy to clear the applica-
tion stage and establish two main evaluation systems which is the level of eval-
uation and effectiveness evaluation of the EI (Wang and Liu 2007). EI evaluation
framework diagram is shown in Fig. 55.1.
To carry out the evaluation of EI, it firstly should be in conjunction with the
division of the enterprise information application stage to define the enterprise
information. The application stage of division is to facilitate the level of
538 Q. Yuan et al.
The external environment Internal resources and capabilities
Corporate strategy
Other business strategies Information strategy
The level of information (stage)
Other business systems
Information system Benefit Evaluation System
Support
Subsystems Evaluation criteria
Evaluators
Benefit Evaluation Evaluation system
Fig. 55.1 EI evaluation framework diagram
evaluation. For benefit evaluation as a precondition to fully consider the level of
evaluation and effectiveness evaluation of convenience, applicability, systemic and
operational evaluation of characteristics, we must consider the purpose of the
application of information technology, the characteristics and the purpose of the
evaluation of information technology, rather than just consider the evaluation
division stage.
Secondly, it should be combined with the purpose of the evaluation to make the
indicator selection and settings, the choice of evaluation methods, and the level of
evaluation and effectiveness evaluation in a phased manner. This process is a
dynamic process, indicating that in the different stages of the evaluation, with the
different purposes, the selected indicators, weights and evaluation methods should
change accordingly.
55.4.2 The Problems of EI Evaluation
55.4.2.1 Moderate Informationization, Emphasizing the Strategic
Matching
EI strategy is an important supporting part of the overall business strategy, on the
basis of the analysis on the internal and external environment of EI, to make the
scientific assessment for the ability of EI it is the planning and decision which is
made the strategy and implementation steps for EI.
The formulation and implementation of EI strategy not only need to match with
the corporate strategy, emphasizing the unity of purpose to meet the needs of
enterprise development, but also to standardize and restrict the direction and
content of EI construction and the application. It needs the modest investment and
emphasizes on building applications effectiveness to achieve the maximum returns
on investment, avoiding the aimlessness and disorder of informationization
construction.
55 Study on the Evaluation Framework of Enterprise Informationization 539
The matching of moderate informationization strategy needs to consider the
influencing factors of external environment, competitive strategy, competitiveness,
industry characteristics, EI demand and other factors, and to make the overall
consideration and analysis of these factors (Wen 2010).
55.4.2.2 Do Well the Phases of Work on the EI with the Evaluation
The phasing of EI is the early work on the evaluation of EI. For the evaluation of
EI work (Liu et al. 2004), the appropriate information phasing can play a multi-
plier effect on the smooth development of information technology evaluation.
In order to carry out the work of stage division, we need to choose the phasing
basis suitable for EI evaluation under the principle of phasing and to master the
main features of each stage.
Under the normal circumstances, the phasing of EI should follow three prin-
ciples: (a) between each level it should have continuity which can reflect the
gradual process of development and trend. (b) Each phase should have significant
and easily described characteristics. (c) The phase characteristics identified should
reflect the change of the enterprise as soon as possible.
Under the three principles of framework above, the writer conducts the system
analysis about the stage research of local and foreign enterprise information
application, and makes sure the main stage division by mainly using the following
information both at home and abroad: information system input and effective
(Nolan); the role of information (Synnott); information technology application
(Mische); IT application mode (Boar); outside, the efficiency of benefit, dispersion,
comprehensive (Edwords). These serve as the reference of classifying the infor-
mationization evaluation stage (Nolan 1975; Churchill et al. 1969; Synnott 1987;
Mische 1995; Edwards et al. 1991).
55.5 Rour-Stage Application Model Adapted
for the Evaluation of EI
The evaluation of EI is a multi-stage dynamic process, and before the benefit
evaluation of the informationization, it’s important to confirm the application stage
of the enterprise and to confirm that the application stage should integrate the level
of EI. The evaluation level of EI and the benefit evaluation of EI is continuous
process on the same level.
Refer to the research result of Mische & Boar, we divide EI into three levels
four stage according to information technology application and function level.
Three levels refer to construction phase, function application level, strategic
management level. Among them, infrastructure level corresponding fundamental
construction stage, function application levels corresponding internal integration
540 Q. Yuan et al.
stage, strategic management level corresponding strategic management stage.
Strategic management can be divided into process innovation and strategic inno-
vation two stages.
The basic construction phase is the early step of the beginning and technical
application, and it mainly consists informational network, data-base construction
and local application with an emphasis on the complete function of information.
The informationization assessment is carried out as a technical project which aims
at the revolution of working format and the improvement of efficiency, thus, the
main point is not about making strategies and gaining investment income.
After the completion of the basic construction of the informationization, the
requirement of the enterprisers on application of information will come to a
strategic level. So far, the internal integration phase has been playing an important
role for most of the time on the application phase, and this phase is mainly about
the application of informational techniques, improving the business.
The assessment to the internal integration phase is mainly about the incomes, as
well as a little invisible benefit which occupies a small portion and can be decided
according to specific situations of the enterpriser.
The phase of strategic management is gradually developed from the functional
application phase. It is a phase when the functional application becomes mature
and the informational strategy shows its support. and with the help of the infor-
mational support, the enterprise transits from redesigning the internal process to
looking for better cooperation and combination with the relative enterprises on
supply chains and redesigns the enterprise boundary.
By enhancing the competitiveness of enterprises as well as competitive
advantages of products, the informationization strategy becomes a vital part of the
enterprise’s overall strategy. An informational management department can be
founded in which CIO plays an important role on information support in the
enterprise’s decision making.
The assessment to the strategic management phase is focused on the invisible
benefits and strategic benefits while both incomes and invisible benefits are
assessed, and it is mainly about the changes of index and the designing of weights.
The four phases of informationization application is suitable for the assessment
of informationization. The classification of the four phases is suitable for differ-
entiating and properly setting the indices on both quality and quantity by making
the indices more available. Also, it reduces the difficulty of assessing and increases
the practicability. What’s more, it is suitable for adjusting the indices and weights
in different phases which ensures that the assessment is subject and accurate.
55 Study on the Evaluation Framework of Enterprise Informationization 541
55.6 Conclusion
Evaluation of EI is a multi-level and stages dynamic evaluation process. In the
evaluation process, the evaluation of EI should combine the application stage to
choose the appropriate evaluation index and evaluation method. How to construct
the evaluation of EI system depends on the specific industry and enterprise.
Acknowledgments Soft Science Project of Shandong Province. 2011RKGA7066.
References
Churchill NC, Kempster JH, Uretsky M (1969) Computer based information systems for
management: a survey. National Association of Accountants , New York
Edwards C, Ward J, Bytheway AJ (1991) The essence of information systems. Prentice Hall
International (UK) Ltd, Prentice Hall, pp 28–31
Ke J, Li C (2007) The investigation of enterprise informationization performance appraisal.
Intelligence Magazine 10:30–35 (Chinese)
Liu Y, Hao W, Wei L (2004) The development patterns of the enterprise information stage and
the stage characteristic analysis. Sci Technol Manag Res 2:101–103 (Chinese)
Mische MA (1995) Transnational architecture a reengineering approach. Inf Manag 12(1):98–100
Nolan RL (1975) Thoughts about the fifth stage. Database 7(2):4–10
Synnott WRL (1987) The information weapon: winning customers and market with technology.
Wiley, New York, pp 88–93
Wang K, Liu B (2007) Stage evolution and index setting of enterprise informationization benefit
evaluation stage evolution. Intelligence Magazine 11:29–31 (Chinese)
Wen C (2010) Informationization impact mechanism to enterprise competitive capability and the
strategic choice. Jilin University 4 (Chinese). Doctorial thesis
Chapter 56
Supplier Selection Model Based
on the Interval Grey Number
Zhi-yong Zhang and Sheng Wu
Abstract Supplier selection plays an important role in the management of
purchasing. Decision support tools and methodologies can help managers make
more effective decisions. Many tools have been developed with a variety of formal
modeling techniques, but these techniques are limited for the reason that they are
not available when the criteria value are interval grey numbers. There are many
reasons for the criteria values should be expressed by the interval grey numbers.
This paper proposes a new approach for the supplier selection which is prepared to
deal with the interval grey criteria based on the definition of grey circle/radius and
distance degree. The interval criteria value and its whitenization weight function
(WWF in shorted) make the model more realistic and accurate. Through a
numerical case, on both the theoretical and the practical level, it has been proved
that this model is correct and feasible.
Keywords Grey system theory Grey circle/radius Interval grey number
Supplier selection
Z. Zhang (&) S. Wu
Beijing Modern Logistics Research Center, Beijing Wuzi University,
Beijing, People’s Republic of China
e-mail: [email protected]
S. Wu
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 543
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_56,
Ó Springer-Verlag Berlin Heidelberg 2013
544 Z. Zhang and S. Wu
56.1 Introduction
Supplier selection is an important part of purchasing management. Usually, the
purchase cost could account for the more than 70 % of the total cost of a product.
Supplier selection is one of the most important functions of purchasing manage-
ment (Ghodsypour and O’Brien 1998).
The criteria of supplier selection can be classified as quantitative and qualitative
in general. In many cases, the quantitative criteria values of suppliers should be
expressed by interval grey numbers. Such as the price, the suppliers always need to
define the range of the price interval for bargaining. The range of the price interval
is an interval grey number. The profit goal of Ford Motor Company is 15–20 %, so
the price is an interval grey number based on target-return pricing. The quality
requirement of white carbon black is that PH should between 5 and 8. The delivery
performance can be represented by delivery period and quantity. According to
international trade sales contract terms, in terms of delivery period, delivery period
is always interval grey number. The actual delivery quantity is also interval grey
number, because of the human error and unavoidable risk such as natural disasters
and damage of cargo etc. So we need a method to select supplier when there are
criteria values of interval grey numbers.
The remainder of this paper is organized as follows. Section 56.2 is a literature
review of supplier selection methods. The basic concepts and the supplier selection
model based on interval grey number are proposed in Sect. 56.3. Then, in
Sect. 56.4, an illustrative example presents applying the proposed approach to the
supplier selection problem, and then we discuss and show how the method based
on the interval grey number is effective. Finally, conclusions are presented in
Sect. 56.5.
56.2 Methods of Supplier Selection
Four evaluation models for supplier selection feature prominently in the literature:
LW models, the total cost models, the mathematical programming methods and
the grey correlation models. Each of these is introduced below.
(1) Linear-weighting models: LW models evaluate potential suppliers using
several equally weighted factors, and then allow the decision-maker to choose the
supplier with the highest total score (Timmerman 1986). Although this method is
simple, it depends heavily on subjective judgment. In addition, these models
weight the criteria equally, which rarely happens in practice (Min 1994;
Ghodsypour and O’Brien 1998).
In contrast to the equal weighting utilized in LW models, AHP is an effective
method for providing a structured determination of the weights of criteria by using
pair wise comparison to select the best suppliers. Several researchers have used
AHP to deal with the supplier selection issues. These include Nydick and Hill
56 Supplier Selection Model Based on the Interval Grey Number 545
(1992), Barabarosoglu and Yazgac (1997), Tam and Tummala (2001), Bhutta and
Huq (2002) and Handfield et al. (2002).
(2) Total cost of ownership models: TCO models attempt to include the
quantifiable costs that are incurred throughout the purchased item life cycle into
the supplier selection model. Monczka and Trecha (1988), Smytka and Clemens
(1993), Roodhooft and Konings (1996), Chen and Yang (2003) attempted to
integrate the total cost into their evaluation models.
(3) Mathematical programming methods: Mathematical programming methods
can be used to formulate the supplier selection problem in terms of an objective
function to be maximized or minimized by varying the values of the variables in an
objective function. Several papers have used single objective techniques to solve
the supplier selection issues, these include linear programming (Pan 1989;
Ghodsypour and O’Brien 1998), goal programming (Buffa and Jackson 1983;
Karpark et al. 1999) or mixed integer programming (Chaudhry et al. 1993;
Rosenthal et al. 1995).
(4) Grey correlation models: Grey system theory can be used to solve uncer-
tainty problems in cases with discrete data and incomplete information (Deng
2002). It is, therefore, a theory and methodology that deals with poor, incomplete,
or uncertain systematic problems. Several papers have used Grey correlation
models to select suppliers. These include Tsai et al. (2003), Jadidi et al. (2008) and
Yang and Chen (2006).
All the existing methods are effective when the criteria values are real numbers,
but it cannot be used for the interval grey number situations. So vendor selection
urgently needs a solution for the interval grey criteria. This paper proposes a new
model to solve this problem.
56.3 Supplier Selection Based on Interval Grey Numbers
There are many patterns of supplier selections. The pattern in this paper is
described as follows: The managers of purchasing department have the criteria
value of the ideal referential supplier, the criteria value of ideal referential supplier
is called optimum criteria value; we should compare the optimum criteria value
with the criteria value of supplier candidates and chose the closest one. We will
give some basic concepts and methods for the convenience of the reader to
understand.
56.3.1 Basic Concepts
In our research, the value of criteria is an interval grey number, so the criteria
value sequences are interval grey sequences.
Definition 1 Liu et al. (2008) if the value ranges is known and exact value is
unknown, then this number is called grey number, noted as . If grey number has
546 Z. Zhang and S. Wu
both upper bound b and lower bound a, then this number is called interval grey
number, noted as 2 ½a; b.
Definition 2 Zeng et al. (2010) if a sequence is consist of interval grey number,
then this sequence is called an interval grey number sequence, noted as
X ðÞ ¼ ððt1 Þ; ðt2 Þ; . . .; ðtn ÞÞ ð56:1Þ
where
ðtk Þ 2 ½ak ; bk ðk ¼ 1; 2; . . .; nÞ ð56:2Þ
The criteria value with different dimension cannot form comparable criteria
value sequence, so criteria value should be transformed to no-dimensional
sequence. Standardization method is a mostly-used dimensionless method, and the
mean method is better than the former (Guo et al. 2011). But for Interval grey
criteria sequence, there are no existing methods that can transform the sequence to
no-dimensional sequence.
Now, we give a method as following: we can just transform the midpoint
sequence of interval grey criteria value, and the lengths of interval grey number are
invariant. The midpoint can be calculated according to (56.3) and be transformed
to no-dimensional sequence according to (56.4).
aik þ bik
mik ¼ aik þ ð56:3Þ
2
mik
m0ik ¼ ð56:4Þ
k
m
Where aik is the lower bound of criteria k and
bik is the upper bound of criteria
k is the mean value of the midpoints’ value of criteria k.
k. m
Definition 3 Liu et al. (2008) if the starting point and endpoint of the function are
determined and the graph of the function is left up, right down, this function is
called typical whitenization weight function (TWWF in short), see Fig. 56.1.
Fig. 56.1 Typical
whitenization weight function
56 Supplier Selection Model Based on the Interval Grey Number 547
Fig. 56.2 Grey figure/center and grey circle/radius
Where xki ð1Þ, xki ð2Þ, xki ð3Þ, xki ð4Þ are called the turning points of fik ð xÞ.
TWWF can be used to describe the probability that the grey interval criteria
take values in its range. TWWF of interval grey criteria value can be determined
by experience of the purchasing managers.
In order to describe the geometrical characteristic of interval grey criteria value
and its TWWF, we proposed the definition of grey figure and grey point as follow:
Definition 4 The figure combined by interval grey line and WWF is called grey
figure, noted as 4, Grey figure’s gravity center is called grey center, noted as G,
see Fig. 56.2. The circle which has the equal area with grey figure is called grey
circle, noted as O, it’s radius is called grey radius, noted as r, see Fig. 56.2`.
56.3.2 Fundamental Idea
The fundamental idea of this paper is that we calculate a grey closeness degree
between the compared supplier alternatives set and the ideal referential supplier
alternative to determine the ranking order of all alternatives of supplier and to
select the ideal supplier based on interval grey numbers.
Firstly, the quantitative criteria should be converted into the proper dimen-
sionless indexes, and regard the dimensionless criteria value of supplier as an
interval grey number sequence. Then, map the interval number sequence and its
WWF on the rectangular coordinates system, see Fig. 56.3.
We assume that <i is the grey figure sequence of CSi and <0 is grey figure
sequence of the ideal referential supplier. The closeness degree between <i and <0
is judged based on the similarity level of grey figure and the distance between
them. The similarity level of grey figure is represented by the difference between
grey radius and the distance between grey figures is represented by the grey point’s
distance.
548 Z. Zhang and S. Wu
Fig. 56.3 Geometric patterns of interval criteria value
56.3.3 Specific Methodology
The specific methodology of this model is mainly includes the following steps:
Step 1 Calculate the Grey points’ horizontal ordinate sequence and vertical
ordinate sequence according to (56.5) and (56.6).
2xki ð3Þ þ xki ð4Þ 2xki ð2Þ xki ð1Þ
xi ðtk Þ ¼ k ð56:5Þ
3½xki ð3Þ þ xki ð4Þ xki ð2Þ xki ð1Þ
2 2 2 2
xki ð4Þ þxki ð3Þ xki ð4Þ þ xki ð3Þ xki ð2Þ xki ð1Þ xki ð1Þ xki ð2Þ
yi ðtk Þ ¼
3½xki ð3Þ þ xki ð4Þ xki ð2Þ xki ð1Þ
ð56:6Þ
Where xki ð1Þ is the turning point of TWWF.
Step 2 Calculate the distance between grey points according to (56.7).
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
d0i ðtk Þ ¼ ½xi ðtk Þ x0 ð tk Þ2 þ½yi ðtk Þ y0 ð tk Þ2 ð56:7Þ
Where xi ðtk Þ, yi ð tk Þ are the grey points’ horizontal ordinate and vertical
ordinate of supplier candidate i, x0 ð tk Þ, y0 ðtk Þ are the grey points’ horizontal
ordinate and vertical ordinate of ideal referential supplier.
The longer of the grey points distance the smaller the similarity of different
sequence. So the distance degree (noted as Di0) is calculated according to (56.8).
" #
m X
X n X
n
D0i ðtk Þ ¼ di0 ðtk Þ di0 ðtk Þ di0 ðtk Þ ð56:8Þ
i¼1 k¼1 k¼1
Step 3 Calculate the weight that reflects the difference of the circle’s radius
which has the equal area with grey figure according to (56.9) and (56.10).
56 Supplier Selection Model Based on the Interval Grey Number 549
P
n
1 r00i ðkÞ r00i ðkÞ
k¼1
x00i ðkÞ ¼ ð56:9Þ
n1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
xk ð3Þ þ xk ð4Þ xk ð1Þ xk ð2Þ xk0 ð3Þ þ xk0 ð4Þ xk0 ð1Þ xk0 ð2Þ
r00i ðkÞ ¼ i i i i
2p 2p
ð56:10Þ
Step 4 Compute the weighted sums of grey point’s distance by use of the weight
in step 3.
56.4 Case study
In this section, we will concentrate on how to apply the new supplier selection
method based on interval grey number to evaluate vendors. It is not our intention
to develop an evaluation model with considering all criteria as mentioned above.
Therefore, this paper will only provide a feasibility study on the new model for
vendor evaluations.
56.4.1 Selection of Criteria
In a review of earlier research, Dickson (1966) identified 23 different criteria as
having been used in the supplier selection process—including quality, delivery,
price, technical capability, and financial position. A review of 74 supplier selection
articles by Weber et al. (1991) founds that cost, product quality, delivery perfor-
mance, and supply capacity were most frequently used as selection criteria. The
five criteria select by Amir et al. (2010) were product quality, delivery perfor-
mance, price, technical capability and tenacity. Four criteria are accepted by most
researchers, there are price, product quality, delivery performance and technical
capability.
56.4.2 The Application of Supplier Selection Method
Based on Interval Grey Number
Before using the new method, quantitative criteria should be converted into the
proper dimensionless indexes. The processed information of five candidate sup-
pliers is present in Table 56.1. The turning point of WWF is present in Table 56.2
550 Z. Zhang and S. Wu
Table 56.1 Processed information of candidate suppliers
CS Criteria
Price Quality Delivery performance Technical capability
1 [75,81] [88,93] [28,32] [96,98]
2 [81,85] [92,96] [25,31] [92,96]
3 [78,80] [79,81] [40,46] [89,92]
4 [90,94] [91,95] [33,40] [82,88]
5 [85,88] [84,88] [25,27] [78,83]
Table 56.2 Turning point of processed information of candidate suppliers
Turning point of CS Criteria
Price Quality Delivery performance Technical capability
1 (78,81) (89,92) (29,32) (97,98)
2 (81,83) (94,95) (25,30) (94,94)
3 (79,79) (79,80) (42,44) (90,91)
4 (92,94) (93,93) (35,38) (84,86)
5 (86,87) (86,88) (26,26) (80,83)
Step 1 Purchasing department (PD in short) determines the optimum interval
grey value of criteria, and the processed data is presented in Table 56.3. The
turning points of the optimum interval grey criteria value’s WWF function are
present in Table 56.4.
Step 2 Calculate the distance between grey points.
Firstly, calculate the grey point’s horizontal ordinates and vertical coordinates,
see Tables 56.5 and 56.6.
Then, calculate the distance and distance degree between grey points of the
optimum criteria value and the suppliers’ criteria value respectively, see
Tables 56.7 and 56.8.
Table 56.3 Processed optimum criteria
Criteria
Price Quality Delivery performance Technical capability
[76,80] [89,93] [30,33] [96,98]
Table 56.4 Turning point of whitenization weight function
Criteria
Price Quality Delivery performance Technical capability
Turning point (78,80) (90,92) (32,32) (96,97)
56 Supplier Selection Model Based on the Interval Grey Number 551
Table 56.5 Grey points horizontal ordinates
CS Horizontal ordinate
Price Quality Delivery performance Technical capability
CS1 0.556 1.542 2.542 3.556
CS2 0.533 1.6 2.515 3.667
CS3 0.667 1.556 2.583 3.583
CS4 0.556 1.667 2.567 3.583
CS5 0.583 1.556 2.667 3.542
PD 0.556 1.556 2.667 3.556
Table 56.6 Grey points horizontal ordinates
CS vertical ordinate
Price Quality Delivery performance Technical capability
CS1 78.667 90.5 30.238 97.222
CS2 82.556 94.2 27.758 94
CS3 79 79.778 43 90.5
CS4 92.444 93 36.5 85
CS5 86.5 86.444 26 80.958
PD 78.444 91 32.667 96.778
Table 56.7 Distance between grey points
CS Distance di0(tk)
Price Quality Delivery performance Technical capability
CS1 0.223 0.500 2.432 2.049
CS2 4.112 3.200 4.911 3.359
CS3 0.567 11.222 10.333 6.581
CS4 14 2.003 8.834 11.942
CS5 8.056 4.556 6.667 15.948
Step 3 calculates the weight that reflects the difference between the area of the
optimum grey figures and compared figures, which are noted as xPDCSi ðkÞ, see
Table 56.9.
Step 4 Compute the weighted sums of grey point’s distance by the use of
weights in step 3.
SC1 ¼ 109:834 SC2 ¼ 96:666 SC3 ¼ 81:762
SC4 ¼ 78:022 SC5 ¼ 73:193
So, candidate supplier 1 is selected
552 Z. Zhang and S. Wu
Table 56.8 Distance degree between grey points
Distance degree Di0(tk)
Price Quality Delivery performance Technical capability
CS1 111.33 111.053 108.553 109.505
CS2 96.411 97.767 95.725 97.608
CS3 88.584 78.151 79.151 82.546
CS4 71.865 83.862 82.365 73.909
CS5 73.707 77.207 74.761 68.816
Table 56.9 Weights that reflects the grey figure difference
xPDCSi ðkÞ
Price Quality Delivery performance Technical capability
CS1 0.290 0.107 0.269 0.333
CS2 0.333 0.22 0.255 0.191
CS3 0.203 0.243 0.253 0.3
CS4 0.333 0.252 0.279 0.135
CS5 0.292 0.333 0.143 0.232
56.5 Conclusion
The ultimate goal of supplier selection is to select appropriate suppliers that can
provide faster delivery, lower cost and better quality, and further to increase the
corporate competitiveness. But the information of delivery, prize, quality and
technical capability is usually uncertain, and the criteria values might be interval
grey numbers. This problem could not be solved by the existing methods. So we
put forward a new method based on the interval grey number.
The method, in this paper, could deal with the supplier selection problem in an
uncertain environment. Though it was demonstrated that the new method, which is
used to select the best supplier, it is a good means of evaluation. The proposed
model has two disadvantages.
(1) The interval grey value of criteria is not easy to obtain in practice. Because
the suppliers’ information which existing evaluation methods needed is real
number, there is little interval grey information in hand.
(2) There is no existing scientific method to determine the WWF of the interval
grey criteria value. Most WWFs could only be determined by the experience of the
managerial and/or the technical staffs.
Our suggestions are, Firstly, establish an interval grey information system for
the supplier selection of your department. Secondly, study the appropriate method
for determining the interval grey criteria value’s WWF.
After all, the new model should be widely applied for supplier selection in
practices. In comparison with other models, this model is more applicable and
effective.
56 Supplier Selection Model Based on the Interval Grey Number 553
Acknowledgments This research is supported by Beijing Municipal Education Commission (SZ
201010037015) and partly supported by the Researching Platform of Logistics management and
Engineering of Beijing Wuzi University.
References
Amir S, Farid S, Yazdankhah A (2010) Group decision making process for supplier selection with
VKOR under fuzzy environment. Expert Syst Appl 37:24–30
Barabarosoglu G, Yazgac T (1997) An application of the analytical hierarchy process to the
supplier selection problem. Prod Inventory Manag J 1st q 38(1):14–21
Bhutta KS, Huq F (2002) Supplier selection problem: a comparison of the total cost of ownership
and analytical hierarchy process approaches. Int J Supply Chain Manag 7(3):126–135
Buffa FP, Jackson WM (1983) A goal programming model for purchase planning. J Purch Mater
Manag, Fall, pp 27–34
Chaudhry SS, Forst FG, Zydiak JL (1993) Vendor selection with price breaks. Eur J Oper Res
70:52–66
Chen CC, Yang CC (2003) Total-costs based evaluation system of supplier quality performance.
Total Qual Manag 14(3):325–339
Deng JL (2002) The basis of grey theory. Huazhong University of Science and Technology press,
Wuhan, pp 8–16, 122–202
Dickson GW (1966) An analysis of vendor selection systems and decisions. J Purch 2(1):28–41
Ghodsypour SH, O’Brien C (1998) A decision support system for supplier selection using an
integrated analytic hierarchy process and linear programming. Int J Prod Econ 56/57:199–212
Guo YJ, Ma FM, Dong QX (2011) Analysis of influence of dimensionless methods on deviation
maximization method. J Manag Sci China 4(5):19–28
Handfield R, Walton S, Sroufe R, Melnyk S (2002) Appling environmental criteria to supplier
assessment: a study in the application of the AHP. Eur J Oper Res 141:70–87
Jadidi O, Yusuff RM, Firouzi F, Hong TS (2008) Improvement of a grey based method for
supplier selection problem. J Achiev Mater Manuf Eng 31(2):770–777
Karpark B, Kumcu E, Kasuganti R (1999) An application of visual interactive goal programming:
a case in vendor selection decisions. J Multi-Criteria Decis Anal 8:93–105
Liu SF, Dang YG, Fang ZG (2008) Grey system theory and its application, 4th edn. Science
Press, Beijing, pp 3–4, 50–55
Min H (1994) International supplier selection: a multi-attribute utility approach. Int J Phys Distrib
Logist Manag 24(5):24–33
Monczka RM, Trecha SJ (1988) Cost-based supplier performance evaluation. J Purch Mater
Manag 24(2):2–7
Nydick RL, Hill RP (1992) Using the analytical hierarchy process to structure the supplier
selection procedure. Int J Purch Mater Manag 28(2):31–36
Pan AC (1989) Allocation of order quantity among suppliers. J Purch Mater Manag, Fall,
pp 36–39
Roodhooft F, Konings J (1996) Vendor selection and evaluation: an activity based costing
approach. Eur J Oper Res 96:97–102
Rosenthal EC, Zydiak JL, Chaudhry SS (1995) Vendor selection with bundling. Decis Sci
26(1):35–48
Smytka DL, Clemens MW (1993) Total cost supplier selection model: a case study. J Purch Mate
Manag 29(1):42–49
Tam MCY, Tummala VMR (2001) An application of the AHP in vendor selection of a
telecommunications system. Omega 29:171–182
Timmerman E (1986) An approach to vendor performance evaluation. J Purch Mater Manag
26(4):2–8
554 Z. Zhang and S. Wu
Tsai C-H, Chang C-L, Chen L (2003) Applying grey relational analysis to the vendor evaluation
model. Int J Comput 11(3):45–53
Weber C, CurenT J, Benton W (1991) Vendor selection criteria and methods. Eur J Oper Res
50:2–18
Yang C-C, Chen B-S (2006) Supplier selection using combined analytical hierarchy process and
grey relational analysis. J Manuf Technol Manag 17(7):926–941
Zeng B, Liu SF, Meng W, Zhang J (2010) Incidence degree model of interval grey number based
on space mapping. Syst Eng 28(8):122–126
Chapter 57
The Accuracy Improvement of Numerical
Conformal Mapping Using the Modified
Gram-Schmidt Method
Yi-bin Lu, De-an Wu, Ying-zi Wang and Sha-sha Zheng
Abstract In this paper, a method to improve the accuracy of numerical conformal
mapping is considered. This method calculates new charge points using a gen-
eralized eigenvalue problem, the matrix of which is calculated using the modified
Gram-Schmidt method. We illustrate the efficiency of the proposed method by
some numerical results.
Keywords Generalized eigenvalue problem Modified gram-schmidt method
Numerical conformal mapping Vandermonde matrix
57.1 Introduction
Conformal mapping theory has been widely applied in the fluid dynamics and
many other fields with a strong vitality. Computing conformal mapping is a very
difficulty question. Amano et al. have achieved many numerical results on the
studies of the charge simulation method for conformal mappings (Amano 1987,
1988a, b, 1994) in which the charge points is often gained by experience. On the
other hand, Sakurai and Sugiura proposed a numerical method for conformal
mappings by using Padé approximation (Sakurai and Sugiura 2002), allows the
Y. Lu (&) S. Zheng
Faculty of Science, Kunming University of Science and Technology
of China, Kunming, People’s Republic of China
e-mail: [email protected]
D. Wu
School of Mathematical Sciences, University of Electronic Science
and Technology of China, Chengdu, People’s Republic of China
Y. Wang
Computing Center, Kunming University
of Science and Technology of China, Kunming, People’s Republic of China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 555
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_57,
Ó Springer-Verlag Berlin Heidelberg 2013
556 Y. Lu et al.
charge points to be computed automatically. High accuracy of numerical con-
formal mappings can be gained through a few charge points by using Padé
approximations, but numerical accuracy is degraded when exceeding a certain
number of charge points.
A numerical method is proposed for conformal mapping in this paper that
resolves this accuracy issue. First we introduce how to compute charge points
using the charge simulation method and the Padé approximations. We then con-
struct the generalized eigenvalue problem using two diagonal matrices and a
vandermonde matrix for obtaining new charge points (Niu and Sakurai 2003a, b).
A highly accurate orthonormal matrix is constructed for solving the generalized
eigenvalue problem using the modified Gram-Schmidt method (Golub and Van
Loan 1996; Saad 2003). The numerical accuracy of conformal mapping is
improved. Some numerical results are presented to demonstrate the efficiency of
the proposed method.
57.2 Computing Charge Points
There is a conformal mapping from a domain D exterior to a closed Jordan curve C
in the z plane onto the exterior of the unit disk jwj [ 1 in the w plane. Without lose
generality, we assume that the origin z ¼ 0 lies in D and f ð0Þ ¼ 0. Then the
mapping functions
z
f ðzÞ ¼ expðgðzÞ þ ihðzÞÞ
c
are uniquely determined by the normalization conditions f ð1Þ ¼ 1 and
f 0 ð1Þ [ 0 (Amano 1988a), and gðzÞ is the solution of the dirichlet problem of
exterior
8 2
< r gðzÞ ¼ 0;
> z2D
gðzÞ ¼ log c logjzj; z 2 C
>
:
gð1Þ ¼ 0
where c is the radius of mapping and hðzÞ is conjugate harmonic function of gðzÞ in
D. The charges q1 ; . . .; qN can be calculated at the charge points f1 ; . . .; fN
arranged inside the boundary C and the collocation points z1 ; . . .; zN arranged on
the boundary C by
8
>
> XN
>
> q logjzi fj j ¼ logjzi j logC ði ¼ 1; 2; . . .; NÞ
>
< j¼1 j
>
> X
N
>
> qj ¼ 0
>
:
j¼1
57 The Accuracy Improvement of Numerical Conformal Mapping 557
in the charge simulation method, where C is an approximation to c (Amano
1988a).
Let EðzÞ be the rational expression
X
N
qj
EðzÞ :¼ ~1 z1 þ l
¼l ~2 z2 þ ;
j¼1
z fj
the residue theorem implies that
X
N
~k ¼
l qj fk1
j ; k ¼ 1; 2; : ð57:1Þ
j¼1
Application of the Padé approximations yields EðzÞ AðzÞ=BðzÞ, where AðzÞ ¼
P Pn1
a0 þ n1 j
j¼1 aj z and BðzÞ ¼
j n
j¼0 bj z þ z .
~ n be the n n Hankel matrix
Let H
H ljþkþ1 n1
~ n :¼ ½~ j;k¼0 ; ð57:2Þ
then the coefficient vector x ¼ ½b0 ; b1 ; . . .; bn1 T of BðzÞ can be obtained by
solving H~ n x ¼ m in (Stewart 1973), where m ¼ ½~ lnþ1 ; l ~2n T .
~nþ2 ; . . .; l
The zeros of BðzÞ can be regarded as new charge points calculated by using
Padé approximations when a high accuracy result of the charge simulation method
is gained. Thus the charge points are computed automatically. This numerical
method for conformal mappings using the charge simulation method and the Padé
approximations was originally proposed by Sakurai and Sugiura (2002). When
increasing the number of charges n gained by Padé approximations, the solution x
is not accurate because H~ n is ill conditioned (Tyrtyshnikov 1994). Then the zeros
of BðzÞ can not be calculated accurately.
57.3 Improving the Accuracy
In this section, the problem of computing new charge points is transformed to
solving the generalized eigenvalue problem, in which an orthonormal matrix Q is
calculating by using the modified Gram-Schmidt method.
Let H~\
n be the n n shifted Hankel matrix
~\ ljþkþ2 n1
n :¼ ½~ ð57:3Þ
H j;k¼0 :
\
Note, in the formula above, that the elements of H ~ n are defined via (57.1). Thus
from (57.2) and (57.3), new charge points can be gained by calculating the gen-
eralized eigenvalue problem H ~\ ~
n x ¼ k H n x, instead of calculating the roots of
BðzÞ (Kravanja et al. 1999, 2003; Sakurai et al. 2003).
558 Y. Lu et al.
Let V be the N n Vandermonde matrix
2 3
1 f1 f21 fn1
1
6 1 f2 f2 fn1 7
6 2 2 7
6 2
fn1 7
V :¼ 6 1 f3 f3 3 7;
6. . .. . . .. 7
4 .. .. . . . 5
1 fN f2N n1
fN
DN :¼ diagðq1 ; q2 ; . . .; qN Þ; ZN :¼ diagðf1 ; f2 ; . . .; fN Þ: ð57:4Þ
Then, from (57.1) it can be verified that
H ~\
~ n ¼ V T DN V; H T
n ¼ V DN ZN V: ð57:5Þ
Here, V can be expressed as
~ ~ 0 R
V ¼ QR ¼ ½Q; Q ¼ QR ð57:6Þ
O
~ 2 CNN is an unitary
for QR decomposition, where Q 2 CNn ; Q0 2 CNðNnÞ : Q
~ 2 CNn is expressed as
matrix, and R
~¼ R :
R
O
In the matrix above, R 2 Cnn is an upper triangular matrix and O 2 CðNnÞn is
a zero matrix.
Note that V is the Vandermonde matrix constructed by the mutually distinct
charge points f1 ; . . .; fN , thus the numbers of linearly independent rows and col-
umns of V are N and n, respectively. Therefore we have
rankðVÞ ¼ n: ð57:7Þ
By comparing the rank of various matrices via (57.6), it follows that
rankðVÞ ¼ rankðQRÞ rankðRÞ: ð57:8Þ
Since R is the n n matrix, then
rankðRÞ n: ð57:9Þ
From (57.8) and (57.9), we can verify that
rankðRÞ ¼ n: ð57:10Þ
Therefore the upper triangular matrix R is a regular matrix. Then the eigen-
values can be described as the following theorem.
~\
Theorem The eigenvalues of H ~
n k H n are equal to the eigenvalues of
T T
Q DN ZN Q kQ DN Q:
57 The Accuracy Improvement of Numerical Conformal Mapping 559
Proof From (57.5), it can be obtained that
~\
detðH ~ T T
n k H n Þ ¼ detðV DN ZN V kV DN VÞ:
Then it follows from (57.6) that
~\
detðH ~ T T T T
n k H n Þ ¼ detðR Q DN ZN QR kR Q DN QRÞ
¼ detðRT ðQT DN ZN Q kQT DN QÞRÞ
¼ detðRT ÞdetðRÞdetðQT DN ZN Q kQT DN QÞ:
Since R is the regular matrix via (57.10), then
detðRT Þ 6¼ 0; detðRÞ 6¼ 0:
Therefore, it can be verified that
\
~ n k H
detðH ~ nÞ ¼ 0
and
detðQT DN ZN Q kQT DN QÞ ¼ 0 ð57:11Þ
have identical eigenvalues. Thus the theorem is proved.
The theorem implies that the eigenvalues k1 ; . . .; kn can be gained by (57.11).
We obtain the following algorithm for numerical conformal mapping from the
discussions in this section. The approximate charge points of Padé approximations
are obtained by solving (57.11). Then the mapping functions can be calculated
through the use of k1 ; . . .; kn by calculating the charge simulation method.
Algorithm
1. Give the charge number n of the Padé approximations and the charge number N
of the charge simulation method, the collocation points z1 ; . . .; zN , and the
charge points f1 ; . . .; fN .
2. Calculate the charges q1 ; . . .; qN by using the charge simulation method.
3. Calculate diagonal matrices DN and ZN by (57.4).
4. Construct the matrix Q.
5. Calculate the eigenvalues k1 ; . . .; kn of the pencil QT DN ZN Q kQT DN Q .
6. Calculate conformal mapping by the charge simulation method regarding
k1 ; . . .; kn as charge points.
It is necessary to obtain Q in step 4 to calculate the generalized eigenvalue
problem in the algorithm. The modified Gram-Schmidt method is considered for
construction of Q (Table 57.I). Here V is constructed as
V ¼ ½u; ZN u; ZN2 u; . . .; ZNn1 u;
where u ¼ ½1; 1; . . .; 1T . In the Krylov subspace
560 Y. Lu et al.
Table 57.1 Construction of Q
r11 ¼ jjujj; q 1 ¼ u=r11 ; p ¼ u
for k ¼ 2; 3; . . .; n
p ¼ ZN p
for j ¼ 1; 2; . . .; k 1
rjk ¼ ðp ; qj Þ; p ¼ p rjk qj
end
rkk ¼ jjp jj; q k ¼ p =rkk
end
kn ðZN ; uÞ ¼ spanðu; ZN u; ZN2 u; . . .; ZNn1 uÞ
is expressed using column vectors of V with the construction of
Q ¼ ½q1 ; q2 ; . . .; qn .
57.4 Numerical Example
In this section, we provide the numerical example of the charge simulation method
and the proposed method. We draw a comparison between the proposed
method and the charge simulation method. The algorithm of the charge simulation
method is denoted by M0. In contrast, the proposed algorithm by using the
modified Gram-Schmidt method is referred to as M1. On a Microsoft Windows
operating system, the calculations were performed by using Matlab. The numerical
error is defined as the maximal distance from the point which is the map of the
point on the boundary C in the z plane, onto the circumference of the unit circle in
the direction of radius in the w plane. Here the eigenvalues of QT DN ZN Q
kQT DN Q were calculated by using the eig command in Matlab.
Example Exterior of trochoid: the boundary is given by
x ¼ 0:9 cos t þ 0:1 cos 3t
:
y ¼ 0:9 sin t 0:1 sin 3t
Collocation points and charge points in the charge simulation method are placed
by (Watanabe 1984).
In Fig. 57.1, the error curves of the conformal mappings by various numerical
calculation methods are shown. The results of charge simulation with N ¼ 200
were used for the calculation of (57.1). The accuracy of M1 is superior to M0. M1
gained the best accuracy at n ¼ 29. Figure 57.2 showed the locations of the charge
points for M1 at n ¼ 29. From the results of example, we see that the accuracy of
57 The Accuracy Improvement of Numerical Conformal Mapping 561
Fig. 57.1 Errors of
conformal mappings
Fig. 57.2 Locations of
charge points (N = 200,
n = 29)
M1 is superior to M0. Figure 57.3 shows the exterior of the trochoid. The con-
formal mapping of Fig. 57.3 computed by M1 at n ¼ 29 is shown in Fig. 57.4.
From the result of M1 in Fig. 57.4, we see that the boundary of the trochoid is well
mapped onto the unit circle.
562 Y. Lu et al.
Fig. 57.3 Exterior of
trochoid
Fig. 57.4 Conformal
mapping of fig. 57.3 (M1)
57.5 Conclusions
In this paper, a numerical method using the modified Gram-Schmidt method has
been proposed for improving the accuracy of conformal mapping. The applica-
bility of our method has been demonstrated with numerical results. The accuracy
of conformal mapping by the proposed method is better than achievable by the
charge simulation method. The error analysis for the proposed method will be
investigated in the future.
57 The Accuracy Improvement of Numerical Conformal Mapping 563
Acknowledgement This work is supported by Yunnan Provincial Natural Science Foundation
of China under Grant No. 2011FZ025, and Fundamental Research Funds for the Central Uni-
versities of China, No. ZYGX2010J111
References
Amano K (1987) Numerical conformal mapping based on the charge simulation method. Trans
Inform Process Soc Japan 28:697–704 (in Japanese)
Amano K (1988a) Numerical conformal mapping of exerior domains based on the charge
simulation method. Trans Inform Process Soc Japan 29:62–72 (in Japanese)
Amano K (1988b) Numerical conformal mapping of doubly-connected domain based on the
charge simulation method. Trans Inform Process Soc Japan 29:914–92 (in Japanese)
Amano K (1994) A charge simulation method for the numerical conformal mapping of interior,
exterior and doubly-connected domains. J Comput Appl Math 53:357–370
Golub GH, Van Loan CF (1996) Matrix computations, 3rd edn. The Johns Hopkins University
Press, Baltimore
Kravanja P, Sakurai T, Van Barel M (1999) On locating clusters of zeros of analytic functions.
BIT 39:646–682
Kravanja P, Sakurai T, Sugiura H, Van Barel M (2003) A perturbation result for generalized
eigenvalue problems and its application to error estimation in a quadrature method for
computing zeros of analytic functions. J Comput Appl Math 161:339–347
Niu X, Sakurai T (2003a) An eigenvalue method for finding the multiple zeros of a polynomial.
Trans Japan Soc Ind Appl Math 13:447–460 (in Japanese)
Niu X, Sakurai T (2003b) A method for finding the zeros of polynomials using a companion
matrix. Japan J Indust Appl Math 20:239–256
Saad Y (2003) Iterative methods for sparse linear systems, 2nd edn. SIAM, Philadelphia
Sakurai T, Sugiura H (2002) A method for numerical conformal mapping by using Padé
approximations. Trans Inform Process Soc Japan 43:2959–2962 (in Japanese)
Sakurai T, Kravanja P, Sugiura H, Van Barel M (2003) An error analysis of two related
quadrature methods for computing zeros of analytic functions. J Comput Appl Math
152:467–480
Stewart GW (1973) Introduction to matrix computations. Academic Press, New York
Tyrtyshnikov EE (1994) How bad are Hankel matrices? Numer Math 67:261–269
Watanabe N (1984) A collection of figures for conformal mappings. Cangye Bookstore, Japan (in
Japanese)
Chapter 58
The Application of AHP in Biotechnology
Industry with ERP KSF Implementation
Ming-Lang Wang, H. F. Lin and K. W. Wang
Abstract This research focused on the production of Phalaenopsis, particularly in
its implementation of enterprise resources planning, dimensions and the adoption
of AHP (Analytical Hierarchy Process). The importance of assessment index and
its attributes were reviewed. The results from this research have shown that factors
that led to the incorporation of ERP Key Success Factor in the Biotechnology
Industry (KSF) were employees’ training, the full support from executives in ERP
system integration, communication with company, assistance in training and
technology transfer, real-time and system accuracy and efficiency and flexibility in
resources allocation. The results from this study will benefit the Biotechnology
industry in understanding the important factors that contributed to the success of
ERP, and thus is instrumental for the development of products and marketing
strategies. It serves as a reference for breaking into the renovation market.
Keywords Enterprise resource planning (ERP) Key success factor (KSF)
Analytical hierarchy process (AHP)
58.1 Introduction
The application of ERP system is very complex and relies on background and
motivation. Performance management and analysis of the KSF are required to
ensure successful application of the ERP system in the biotechnology industry
(Wong and Keng 2008). To cope with the competitive orchid market in Taiwan in
M.-L. Wang (&)
Department of Industrial Management, Chung Hua University, Hsinchu, Taiwan
e-mail: [email protected]
H. F. Lin K. W. Wang
Department of Technology Management, Chung University, Hsinchu, Taiwan
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 565
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_58,
Ó Springer-Verlag Berlin Heidelberg 2013
566 M.-L. Wang et al.
the future, it is necessary to reduce capital, increase profit and enhance competi-
tiveness. The objective of this study is to explore enterprise resource planning in
the KSF-Biotechnology industry for effective integration of ERP system (Li et al.
2007). A clear picture of ERP implementation will promote biotechnology
industry’s market.
Based on the above mentioned research backgrounds and motives, this sturdy is
intended for realizing the following purposes: (1) To consolidate and summarize
the literature of KSF of ERP SYSTEM implementation; (2) To understand the
KSF of ERP SYSTEM implementation; (3) To understand the difficulties and
obstacles encountered at each stage of ERP system implementation and solution to
overcome.
58.2 Literature Review
58.2.1 Enterprise Resource Planning System
ERP is short for Enterprise Resource Planning, conceptualized by the Gartner
Group in the early 1990s. Davenport (1998) had viewed ERP as a technology for
enterprise information integration with a simple database as its core. The database
pools and processes various commercial activities within the enterprise according
to the functions, department and regions. Through the internet it creates a network
for data sharing and supports application modules to comply with policies, orga-
nizational characteristics and corporate culture. The integration of ERP system can
maximize efficiency in planning, management, control and utilization of corporate
resources.
58.2.2 Key Success Factor
Daniel (1961) was one of the first to propose the concept of Key Success Factor
(KSF) or critical success factor. It was highlighted that the success of most
industry is determined by three to six factors, such are known as the KSF. KSF
after which economist Commons (1974) referred to as the ‘‘limiting factor’’ and
had the concept applied in economy management and negotiation. Thereafter
Barnard (1976) applied the concept in management decision making theory. He
considered the analysis required for decision making was essentially looking at
‘‘strategic factors’’. In addition, Tillett (1989) applied the concept of strategic
factors to dynamic system theory. He viewed that the ample resources of an
organization was the key factor. Policies were established to maintain and ensure
maximum utilization of resources. In addition, they were important in resources
forecast. KSF is the top priority in industrial analysis. It is important in the
management of control variables, as well as the source of competitive advantage.
58 The Application of AHP in Biotechnology Industry with ERP KSF Implementation 567
In recent years, policy management has overtaken ‘‘information system manage-
ment’’ which was the main focus in earlier research. KSF has thus being applied in
areas beyond information system management.
58.2.3 The Blue Gold of Biotechnology Industry:
Phalaenopsis
Taiwan is known as the kingdom of orchids. The suitable climate, environment
and along with government policies and effort in related measures have placed
Taiwan’s orchid industry in a pivotal position in the international arena. The sales
of Phalaenopsis, Oncidium and Paphiopedium (from Xinxing) rank first in the
world. The production process of Phalaenopsis is divided into three stages. The
upstream development involves development and species identification, breeding,
propagation; the middle stream carries out domestication, small seedlings, medium
seedling, large seedling, growing and harvesting; while logistics, marketing and
branding are steps in the downstream pathway. The application of key technology
and its importance varies among the three production stages. At upstream, the
value-added of breeding and seedling is higher than development, identification
and seedlings. In the middle stages, the value-added of the flowering and seedling
stages is higher than growing, harvesting and domestication. While in the down-
stream stage, the value-added of logistics, channels, brand and marketing is higher
than cash flow and information flow (Huang et al. 2008).
With government guidance, the production of Phalaenopsis has moved toward
mass production. The stringent computerized monitoring on production and cul-
tivation management, in addition a large demand from the domestic and interna-
tional market has led to the expansion of the production lines across cities and
nation-wide. Apart from ‘‘Taiwan Orchid bio Park’’, the national orchid industry
has also established the Taiwan Orchid Growers Association (TOFA), which aims
to promote production and sale of orchids, to develop domestic and foreign
markets, to improve marketing strategy and to assist Taiwanese government in
policy making for the flora industry.
58.3 Methodology
58.3.1 Brief Description and Purpose of AHP
The purpose of AHP analysis is to simplify complex problems into elementary
hierarchy system. It gathers scholars, experts and decision makers at all levels for
comparison between pairs of elements, also known as Pair wise Comparison. Upon
quantization, comparative matrix pairs (Pair wise Comparison Matrix) is
568 M.-L. Wang et al.
established according to the matrix of eigenvectors (Eigenvector). Thereafter the
elements of the vector are employed to establish a hierarchy of priorities and thus
finding the largest eigenvalue. This value provides a reference point for policy
makers in decision making by assessment of the relative strength between the Pair
wise Comparison Matrix consistency indexes. The consistency index is made up of
at least two levels, while AHP links all levels in deriving the AHP hierarchy
among the various factors of relative priority and strength. This is followed by
AHP connecting all the Consistency Index and Consistency Ratio before the final
evaluation of the high and low levels consistency of the hierarchy.
58.3.2 Hypothesis of AHP
In ‘‘The features and application of hierarchy analysis method’’ by Deng
Zhengyuan, Zeng Guoxiong (1989), it was mentioned that the basic assumptions
of AHP analysis included 9 of the following:
(1) A system can be broken down into a number of classes or components to form
a network-level structure.
(2) Each level is assumed independent within the hierarchy.
(3) The factors within each level make use some of or all of the factors in the level
above for assessment and evaluation.
(4) Absolute values can be converted to a Ratio Scale in comparative assessment.
(5) Regular matrix can be employed after Pair wise Comparison.
(6) Preferential relations satisfy transitivity.
(7) Consistency level needs to be measured in case of transitivity.
(8) The degree of advantage of factors is evaluated by the Weighting Principle.
(9) For elements that appear in the class structure, they are considered and
assessed as a whole structure (regardless of their strength), rather than of
independent of review class structure.
58.3.3 Design of a Questionnaire
To study the KSF and their role in ERP implementation, the questionnaire was
divided into two parts:
(1) First Stage Questionnaire: (1) Target audience: the junior and middle man-
agement level of companies involved in ERP system; (2) The company’s key
success factors of ERP.
(2) Second Stage Questionnaire: (1) Target audience: the junior and middle
management level of companies involved in ERP system; (2) The part was
based on the first part and had the factors that contributed to the success of
ERP analyzed.
58 The Application of AHP in Biotechnology Industry with ERP KSF Implementation 569
(3) The aim was to seek the relative importance of KSF: (1) the motive behind
companies in ERP implementation; (2) The second level as the measurement
index: the four dimensions: internal factors, ERP system features, ERP soft-
ware support, results followed by ERP implementation; (3) Third level gauged
the KSF of second level indexes.
58.3.4 Research Structure
See Fig. 58.1.
58.4 Results
58.4.1 Analysis and Results of First Stage Questionnaire
This section covered the first stage of the analysis. Its target audience was the
junior and middle management level of the companies that was involved in suc-
cessful ERP system integration. Out of the total of 20 questionnaires issued, 14
were received. Of those received 13 were valid questionnaires after omitting one
that was incomplete (70 % completed). Likert’s five-point level score system was
used for data mining. The reliability of this study was summarized by Cronbach’ s
a coefficients where the reliability a of the four dimensions in ERP implementation
fell within the range of 0.50 \ a \ 0.90. It is thus a good indicator of the
research’s reliability and value (Wang 2004).
58.4.2 Analysis and Results of Second Stage Questionnaire
(1) Subjective measurement of AHP analysis
Consistency ratio (C.R.) was subjected to ensure questionnaire answers lie
within the valid range in a consistent manner. The measurement of the consistency
ratio in each level was found to be less than 0.1 in all the levels, whether as a
whole factor (first level), internal factor (second level), ERP system features (third
level) or the ERP software support. The C.R values were listed in Table 58.1
Results has shown the factors within the hierarchy were closely related and con-
sistent. Since there were only two factors for comparison in the second level, it
was found to be consistent (A [ B or A \ B) and hence no calculation was
required. In conclusion, C.R was not applicable in the results of ERP
implementation.
570 M.-L. Wang et al.
1. The determination of
executives in
implementation
Internal 2. A highly effective ERP
factors implementation team
across department
3. ERP project team
allowed full authorization
4. ERP implementation
progress
5. Communication between
project team and
departments
6. Staff training
7. Department acceptance in
system implementation
1. Cost of system set-up and
KSF implementation time
2. System integration
capability
ERP 3. Flexibility in modification
system 4. System for modular
design
5. Interface that provides
ease of use
6. Accuracy and real-time
1. Real-time response
service
2. Assists companies in staff
training, and technology
ERP transfer
software 3. Expertise demonstrated
support by vendor
4. Equipment provided by
vendor
5. Understanding the needs
of user
6. Communication with
company
1. Reduced operating costs
2. Increased flexibility and
Results from efficiency in resources
ERP allocation
implementation 3. Increased resources
availability in real-time
4. Smooth purchase process
Fig. 58.1 Research structure
58 The Application of AHP in Biotechnology Industry with ERP KSF Implementation 571
Table 58.1 The consistency Dimension Consistency ratio (C.R.)
ratio in second stage of
questionnaire Whole factor 0.0413
Internal factor of the company 0.9459
ERP system features 0.0400
ERP software support 0.0337
Results from ERP implementation Consistency not required
(2) Compound weight analysis
In ‘‘KSF research on the chain of cafés’’ (Qin 2002), it was mentioned that
hierarchy weighing is also known as local priority which refers to relative com-
parison of weight between each level. The overall weight is known as Global
Priority, which is the weight of the level above (second level) multiplied by the
factors in the current level (third level). This is to display the impact of the factors
in the current level (third level) has on the entire evaluation. Therefore based on
the results from the four dimensions, the compound weights were listed in
Table 58.2 and ranked in order of importance. For better clarity, the values were
multiplied by 100. For example: compound weight (c) = second level weight (a) *
third level weight (b) *100 %.
58.5 Conclusion
The main focus of this study was to identify the Key Success Factors and its level
of importance in ERP implementation. This was carried out successfully through
AHP analysis as well as the survey from questionnaire which was designed based
on interviews and literature review. The factors of ERP implementation in the
company were consolidation in Table 58.3.
In recent years a number of 4 * 6 KSF was usually considered by most
researchers. Therefore, in this case study, the focuses were placed on the first 6
KSF (Yang et al. 2007).
(1) Staff training and education: As the company in this case study is a tradi-
tional industry, the employees were not as highly educated and most did not
possess computer skills. Therefore the company had invested a considerable
amount of time in conducting training courses. Their staff training and education
were divided into two main stages; (1) The team of E- training. They were mainly
responsible for system maintenance, program modifications, and as training
instructors. In addition to basic knowledge, the E- team was required to work with
the ERP software vendors in training courses; (2) The company was responsible
for part of their training courses. The courses were planned in conjunction with the
ERP implementation. A 30-hour course was planned according to the work sys-
tem. The E- team members served as lecturers and all system users were required
to undergo training in a classroom setting during non-working hours. The course
572 M.-L. Wang et al.
Table 58.2 Compound weight analysis of KSF From ERP implementation
(Second level) (Third level) Compound Importance
weight (%)
Level/dimension Weight Question (Index) Weight
(c) = (a)*
(a) (b)
(b)*100 %
Internal factor 0.381 Determination of executives 0.340 12.954 2
in ERP implementation
A highly effective ERP 0.070 2.667 11
implementation team
across department
Progress of implementation 0.119 4.534 7
team
Staff training and education 0.392 14.935 1
ERP system 0.139 Cost of system set-up and 0.150 2.085 12
features implementation time
System integration capability 0.279 3.878 9
Accuracy and real-time 0.372 5.171 5
ERP software 0.249 Real-time response service 0.163 4.059 8
supply Assist company in staff 0.286 7.121 4
training and technology
transfer
Communication with 0.342 8.516 3
company
Results from ERP 0.097 Reduced 0.320 3.104 10
implementation operating cost
Increased flexibility and 0.498 4.831 6
efficiency in resources
allocation
aimed to resolve issues encountered and provide solutions. It also provided an
opportunity to understand and improve the ERP system. At present, the company
continues to conduct staff training and assessment. The assessment would serve as
an encouragement for the staff and accelerate their learning curve. The company’s
executives estimated an additional year is required for staff training and full ERP
system integration. It remains one of the company’s future goals to have everyone
familiarized with the operating system.
(2) Determination of executives in implementation. The company first imple-
mented the ERP system in 2000. The decision to implement was proposed by the
executives through meetings. However, due to lack of support from the senior
management and hence lack of funding by the company, the project was shelved.
In 2001, Taiwan officially became a member of WTO in the 144th meeting. The
influx of large scale manufacturing companies and consulting firms in Taiwan had
led to a perspective change. To enhance competitiveness, the senior management
of the company had decided to reincorporate the ERP system with the full support
of budget.
58 The Application of AHP in Biotechnology Industry with ERP KSF Implementation 573
Table 58.3 Factors of ERP implementation
KSF Level/dimension
(A6) Staff training and education Internal factor
(A2) Determination of executives in ERP implementation Internal factor
(C6) Communication with company ERP software support
(C2) Assist company with staff training and technology transfer ERP software support
(B6) Accuracy and real-time ERP system features
(D2) Increased flexibility and efficiency in resources allocation Results from ERP
implementation
(A4) Progress of implementation team Internal factor
(C1) Real-time response service ERP software support
(B2) System integration capability ERP system features
(D1) Reduced operating cost Results from ERP
implementation
(A2) A highly effective ERP implementation team across Internal factor
department
(B1) Cost of system set-up and implementation time ERP system features
(3) Under the impetus of the executives, staffs were trained to work with the
new system. Staffs that refused change had to be let go. With full budget support
and authorization given by the top management, E-team was able to focus on all
steps and methods of implementation and had complex issues resolved. Therefore
the support from the executives was the main crucial success factor.
(4) Communication between the ERP software vendor and company. The
company in this case study differs from other manufacturing industries. It had to
rely on customized software. The company spent almost a year working with the
software vendor in building a customized ERP system. Meetings were held
between the parties of the E-team members, representatives from every department
and the ERP software vendor. Every one to two weeks pre and during the
implementation period, meetings were held. The meetings were changed to once a
month upon implementation and finally to only emergency situations. ERP soft-
ware vendors to assist companies in training and technology transfer. From the
factor of ERP software vendor features, the company in this case study has placed
less priority on the training and technology transfer. The services provided by the
ERP software vendor were as follows: (1) To provide professionals in staff training
and to arrange training in system conversion. Tailor training courses to meet
customers’ needs; (2) To carry out an assessment on the old system before
determining the means of data transfer (The system requires Windows 2000 Server
operating systems, server software: Tomcat, programming languages: Java, data-
base: Oracle 9i). The E-team was responsible for data transfer operations which
was followed by data integration data into the ERP system by the software
vendors.
(5) Accuracy and real-time system. The progress of the ERP system imple-
mentation could be monitored through network and video systems (cameras).
574 M.-L. Wang et al.
Through real-time monitoring, effective quality control and work progress could
be ensured. Officials could also depend on the system database to identify prob-
lems should they occur during construction.
(6) Flexible and efficient allocation of resource. One of the biggest results upon
ERP implementation was the increased flexibility and allocation of resources,
hence more efficient business operations.
58.5.1 Limitations
(1) The company in this case study is a single case company. The key factors for
implementation mentioned here may not apply in the non-construction
industry. Therefore, there may be limitation on the scope of the findings. The
similar research method is applicable to another industry (e.g., semiconductor
industry) to find any same or distinct conclusions.
(2) This study was not able to widen its scope of survey from a larger pool of
employees due to time, manpower, and financial constrains. This might have
an impact on the results and analysis.
58.5.2 Suggestions and Directions for Follow-up Research
The analysis of this research was based primarily on the csompany in this case
study. It is recommended that subsequent research to be carried out on companies
of difference portfolios in order to compare results and discussion. This will be a
good source of reference for companies and organization considering ERP
implementation.
References
Huang ZQ, Huang B Wang H Lin R (2008) A study on the critical factors and strategy of
Phalaenopsis industry development. Taiwan Agric Assoc Rep 9(6):50–58
Li M, Liu GM, Ding S, Lin Y (2007) A study of KSF of China’s biotechnology industry.
Soochow J Econ Bus Stud 56(2):27–51
Qin JW (2002) The study of KSF of the coffee chain KSF. Tamkang University’s master of
science in management science research papers
Wang HN (2004) Development of the key performance indicator (KPI) system. http://
www.mamage.org.cn
Wong YF, Keng LB (2008) AHP analysis of the KSF of marine museum outsourcing business
model by KSF. Eng Sci Educ J 5(20):200–222
Yang CC, Yang CC, Peng CF (2007) Analyses of employees’ behavior models by introduction of
ERP—an example of the notebook computer industry. Chin-Yu J 25:39–57
Chapter 59
Application of Gray Correlation Method
in the Employment Analysis of Private
Enterprises
Bao-ping Chen
Abstract In recent years, the private enterprises have made outstanding
contributions to China’s employment issue and become the main channel to
assimilate labors. However, the private enterprises in various regions of China are
developing unevenly, resulting in the very big difference in the number of
employment. With the gray correlation method, this paper takes the number of
employment of the private sector in 7 industries in 21 provinces, municipalities
and autonomous regions as the evaluation index to calculate the corresponding
relative correlation, absolute correlation and comprehensive correlation, look for
the major industries affecting the employment of the private sectors, and analyze
the effective way to expand the employment space in the backward areas. The
results show that this method can comprehensively considerate the various factors
of the evaluation problem, which not only avoids the subjectivity of the single
factor but also makes the analysis process more reasonable and objective. Also the
analysis results can accurately reflect the differences between various factors.
Keywords Gray correlative analysis Compositive correlative degree Private
enterprises Employment
59.1 Introduction
At present the reform and opening up, China’s private enterprises have been
developing vigorously from small to large and from weak to strong and become an
important economic growth point in China’s national economic development (Liu
2005). Since the mid-1990s, the employment of the private sector has been
B. Chen (&)
Department of Computer Information and Management,
NeiMongol University of Finance and Economics, Hohhot, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 575
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_59,
Ó Springer-Verlag Berlin Heidelberg 2013
576 B. Chen
developing substantially in scale and growth speed and gradually become the
absolute subject to solve the employment issue in our society, which provides an
important guarantee for China’s social stability. At the same time, it can be seen
that the private enterprises in all regions are developing unevenly in employment
and the number of employment varies greatly (Feng et al. 2010). Then, the private
enterprises in which industries play a leading role in employment? Many scholars
have studied the development of the private sector, such as Feng Tainli’s An
Empirical Study on Political Capital and the Accession to Loan from State-owned
Banks of Chinese Private Enterprises (Song and Dong 2011), Song Qicheng’s On
the Relationship between Development of Private Enterprises and Employment
(Zhang 2004), etc. However, there has been relatively little research conducted on
the degree of impact of the private enterprises in various industries on
employment.
Gray correlation analysis method is one method to quantitatively describe as
well as compare the trend of the development and change of a system (Dang et al.
2009).. Through the determination of the similarity degree of the geometrical
shapes between the reference data column and several comparative data columns,
the closeness of the correlation is estimated and the correlation degree of curves
also is reflected. In the development process of one dynamic system, the major
influential factors can be analyzed through the sequencing of the correlation
degrees. Among them, a low correlation degree means there is no or less influence
from this factor while a high correlation degree means this factor is the major one
influencing the development of the system.
According to the number of employment of the private enterprises in 7
industries in 21 regions of China given by China Statistical Yearbook (2010), this
paper applies the grey correlation analysis to reveal the major sectors impacting
employment and analyze its causes. The empirical studies have shown that: the use
of this method can evaluate the degree of impact of the private enterprises in
various industries on employment more systematically, objectively and accurately,
which has a certain reference value to solve the employment problem of China.
59.2 Procedures of Grey Relation Analysis
Grey correlation method has a low requirement on the sample size and its regu-
larity. It can be applied to the evaluation study with few statistical data, large data
grey, great data fluctuation or non-typical distribution regularity. Grey correlation
method based on the gray system theory is a multi-factor analysis technique which
uses grey correlation to describe the strength, degree and sequence of the corre-
lation between the factors through the calculation of the grey correlation degree.
The specific procedures are as follows.
Step 1: determine the reference sequence and comparative sequence. On the
basis of the qualitative analysis, determine one dependent variable and multiple
independent variables. For m indexes which have n evaluation objects, according
59 Application of Gray Correlation Method 577
to the historical statistics, the reference sequence X0 which reflects the corre-
sponding condition of the things and the comparative sequence which describes
the corresponding situation of m factors are given. Among them:
Reference sequence:
X0 ¼ ðx0 ð1Þ; x0 ð2Þ; . . .; x0 ðmÞÞ ð59:1Þ
Comparative sequence:
Xi ¼ ðxi ð1Þ; xi ð2Þ; . . .; xi ðmÞÞði ¼ 1; 2; . . .; nÞ ð59:2Þ
Step 2: calculate the absolute correlation degree. Set the length of Xi is same as
that of Xj and Xi0 and Xj0 are their respective initial point zero images.
Xi0 ¼ ðxi ð1Þ xi ð1Þ; xi ð2Þ xi ð1Þ; . . .; xi ðmÞ xi ð1ÞÞÞði ¼ 0; 1; 2; . . .; nÞ ð59:3Þ
Calculate the grey absolute correlation degrees of Xi and Xj with formula:
1 þ jsi j þ sj
eij ¼ ð59:4Þ
1 þ jsi j þ sj þ si sj
Among them:
Xn1
1
j si j ¼ x0i ðkÞ þ x0i ðnÞ ð59:5Þ
k¼2 2
Step 3: calculate the relative correlation degree. Set the length of Xi is same as
that of Xj and their initial values are not equal to zero. Xi’ and Xj’ are respectively
the initial value images of Xi and Xj. Take eij 0 , the absolute correlation degree of
Xi’ and Xj’ as the grey relative correlation degree of Xi and Xj. Note as rij. Among
them:
Xi0 ¼ Xi =xi ð1Þ; ð59:6Þ
Xj0 ¼ Xj xj ð1Þ; ð59:7Þ
1 þ s0i þ s0j
rij ¼ ð59:8Þ
1 þ js0i j þ s0j þ s0i s0j
Step 4: solve the comprehensive correlation degree. Set the length of Xi is same
as that of Xj and their initial values are not equal to zero. There is no positive
correlation between eij (the absolute correlation degree of Xi and Xj) and rij (the
relative correlation degree of Xi and Xj). Comprehensive correlation degrees take
both the absolute change and relative change of the data sequences into consid-
eration and at the same time satisfy 4 Axiom of grey correlation degree. Note qij as
the grey comprehensive correlation degrees of Xi and Xj. Among them:
578 B. Chen
qij ¼ heij þ ð1 hÞrij ð59:9Þ
The value of h represents the emphasis on the absolute correlation degree eij and
relative correlation degree rij. Generally, the value of h is 0.5. When h is set, grey
comprehensive correlation degree is unique. However, this kind of conditional
uniqueness does not affect the analysis on the problem.
Step 5: utilize the calculated comprehensive correlation degree analysis qij to
analyze the correlation sequence.
59.3 Analyze the Number of Employment of the Private
Enterprises with Grey Correlation Method
Since the reform and opening up, the private enterprises has mushroomed in the
rapid speed and become the basic force to promote the national economic
development as well as an important guarantee to realize the interests of the
people. The development of the private enterprises plays an important role in
optimizing our resources, increasing the employment rate and promoting the
national economic growth (Wang et al. 2011; Fang et al. 2011; Zhang et al. 2011).
However, the private enterprises develop unevenly in different provinces and cities
of China with a great difference. This paper tries to explore the main reasons
causing the difference with the gray correlation method and find ways to solve the
problem so as to achieve common development and avoid the decline of the
overall level affected by individual backward areas.
Divide the provinces, cities and districts in the country into three parts: the
eastern area includes Beijing, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang,
Shandong, Guangdong, etc.; the central area includes Shanxi, Heilongjiang,
Henan, Hubei, etc. the western area includes Guangxi, Xinjiang, etc. According to
the number of employment of the private enterprises in 21 regions in 2010 given
by China Statistical Yearbook (2010), this paper selects the total employment as
the reference sequence and the number of employment in seven industries, namely,
manufacturing, construction, transportation, wholesale and retail, accommodation
and catering, leasing and business services and others as the main assessment
index. Table 59.1 is the relevant data of 21 cities in 2010.
From Table 59.1, the reference sequence can be obtained:
X0 = (323.2, 123.3, 294.1, 170.7, 216.0, 511.4, 247.3, 293.9, 343.1, 1297.3,
758.8, 397.4, 278.4, 247.5, 636.6, 375.2, 451.8, 372.7, 1233.1, 241.5, 79.8, 43.2,
48.6, 132.4)
Comparative sequence is:
Xi ¼ ðxi ð1Þ; xi ð2Þ; xi ð3Þ; . . .; xi ð21ÞÞði ¼ 1; 2; . . .; 21Þ ð59:10Þ
59
Table 59.1 The number of employment of the private enterprises in 7 regions in 2010 [million people]
Cities Total Manufacturing Construction Transportation Wholesale Accommodation Leasing Others
and retail and catering and services
Beijing 323.2 11.5 8.7 6.6 88.7 21.0 46.8 15.4
Tianjin 123.3 33.4 5.4 5.5 39.8 4.6 10.1 4.7
Hebei 294.1 59.4 9.6 10.2 136.2 20.2 9.2 19.6
Shanxi 170.7 16.1 4.0 3.3 94.1 15.6 5.9 17.1
Neimenggu 216.0 24.4 7.6 10.8 94.9 23.1 9.1 19.2
Liaoning 511.4 87.5 27.0 46.5 193.9 31.0 24.2 32.0
Jilin 247.3 31.5 27.3 10.3 102.5 21.4 7.7 20.0
Heilongjiang 293.9 34.2 8.8 11.3 114.0 28.4 12.6 52.6
Shanghai 343.1 53.5 25.9 13.6 121.5 16.4 45.1 12.0
Application of Gray Correlation Method
Jiangsu 1297.3 475.3 109.5 29.3 408.4 52.5 57.9 57.6
Zhejiang 758.8 259.7 34.0 15.6 269.8 36.1 44.8 38.4
Anhui 397.4 57.5 11.0 6.7 190.7 37.0 13.5 34.4
Fujian 278.4 65.4 8.6 5.1 115.1 16.1 19.7 16.6
Jiangxi 247.5 53.3 4.0 10.2 108.8 22.1 6.6 19.2
Shandong 636.6 144.0 30.3 19.9 279.3 36.0 39.1 35.7
Henan 375.2 61.9 12.5 6.1 186.9 28.8 17.8 27.7
Hubei 451.8 71.5 19.4 15.1 207.6 32.6 20.1 35.2
Hunan 372.7 39.4 12.5 10.1 162.9 20.9 56.2 23.0
Guangdong 1233.1 293.3 30.3 24.4 523.9 71.3 83.9 67.5
Guangxi 241.5 34.7 5.2 11.3 120.5 18.5 12.2 14.7
Hainan 132.4 16.3 4.8 5.4 56.7 14.6 8.2 11.4
579
580 B. Chen
Step 1: Calculate the absolute correlation degree. Take manufacturing calls for
example, through the initialization operation (settled as 1-time interval sequence of
equal length), we can obtain:
X1 = (11.5, 33.4, 59.4, 16.1, 24.4, 87.5, 31.5, 34.2, 53.5, 475.3, 259.7, 57.5,
65.4, 53.3, 144.0, 61.9, 71.5, 39.4, 293.3, 34.7, 4.4, 5.2, 5.2, 16.3)
Through the operation of initial point zero images on X0 sequence and X1
sequence, we can obtain following sequences:
X0 = (0.00, -199.91, -29.19, -152.58, -107.24, 188.19, -75.98, -29.30,
19.89, 974.05, 435.53, 74.10, -44.88, 5.71, 313.34, 51.96, 128.57, 49.41, 909.88,
-81.79, -190.89)
X1 = (0.0, 21.85, 47.91, 4.55, 12.85, 75.98, 19.97, 22.66, 41.97, 463.82,
248.16, 45.99, 53.91, 41.84, 132.47, 50.43, 60.02, 27.84, 281.76, 23.20, 4.75)
Calculate the values of js0 j, js1 j and js1 s0 j, Among them:
js0 j ¼ 2252:91; js1 j ¼ 1679:64; js1 s0 j ¼ 573:27;
Thus, according formula (59.4), the absolute correlation degree of manufac-
turing calls can be calculated and its value is 0.5676. Similarly, the absolute
correlation degrees of all the factors can be calculated. Namely:
e01 ¼ 0:8728; e02 ¼ 0:5501; e03 ¼ 0:5310;
e04 ¼ 0:8926; e05 ¼ 0:5290; e06 ¼ 0:5917;
e07 ¼ 0:5560
Step 2: Calculate the relative correlation degree. Take the manufacturing calls
for example. After the initialization operation, calculate the initial value images of
X0 sequence and X1 sequence. Namely:
X0 = (1.0, 0.38, 0.91, 0.52, 0.66, 1.58, 0.76, 0.90, 1.06, 4.01, 2.3, 1.22, 0.86,
0.76, 1.96, 1.16, 1.39, 1.15, 3.81, 0.74, 0.4095)
X1 = (1.0, 2.89, 5.16, 1.39, 2.11, 7.60, 2.73, 2.96, 4.64, 41.30, 22.56, 4.99,
5.68, 4.63, 12.51, 5.38, 6.21, 3.41, 25.48, 3.01, 1.41)
Calculate the initial point zero images of X0’ and X1’. Namely:
X0’ = (0.0, -0.61, -0.09, -0.47, -0.33, 0.58, -0.23, -0.09, 0.06, 3.01, 1.34,
0.22, -0.13, -0.23, 0.96, 0.16, 0.39, 0.15, 2.81, -0.25, -0.59)
X1’ = (0.0, 1.89, 4.16, 0.39, 1.11, 6.60, 1.73, 1.96, 3.64, 40.30, 21.56, 3.99,
4.68, 3.63, 11.51, 4.38, 5.21, 2.41, 24.48, 2.01, 0.4)
The values of js0 j, js1 j and js1 s0 j can be obtained.
js0 j ¼ 6:96; js1 j ¼ 145:96; js1 s0 j ¼ 138:99
Thus, according formula (59.8), the relative correlation degree of manufac-
turing calls can be calculated and its value is 0.5255. Similarly, the relative cor-
relation degrees of all the factors can be calculated. Namely:
r01 = 0.5255, r02 = 0.6418, r03 = 0.6735, r04 = 0.6828, r05 = 0.9475,
r06 = 0.9007, r07 = 0.7218
Step 3: Calculate the comprehensive correlation degree. Utilize the above
absolute correlation degree and relative correlation degree and formula (59.9) and
59 Application of Gray Correlation Method 581
at the same time set h ¼ 0:5, the comprehensive correlation degrees of all the
factors can be calculated. Namely:
q01 ¼ 0:73; q01 ¼ 0:66; q01 ¼ 0:91; q01 ¼ 0:56;
q01 ¼ 0:76; q01 ¼ 0:90; q01 ¼ 0:82
Step 4: Result analysis.
The result is:
q03 \q06 \q07 \q05 \q01 \q02 \q04 ; Namely:
X3 \ X6 \ X7 \ X5 \ X1 \ X2 \ X4.
It can be seen from the above analysis that X5 is the optimal factor. In other
words, the main factors that affect employment are wholesale and retail, followed
by business services, accommodation and catering, manufacturing, others and
transportation; the construction industry has the minimal impact, which is con-
sistent with the reality. From the point of view of economic development, the
economic growth will largely promote the employment of the private sector.
Meanwhile, the gradient difference in economic development is consistent with the
gradient level of employment. China has a vast territory and the levels of economic
development in the East, center and west are different, which promotes the gradual
expansion of the difference in the number of employment in the private enterprises
in various regions. It can be seen from Table 59.1 that provinces and cities with
the high employment rate such as Shanghai, Beijing, Liaoning, Zhejiang, Jiangsu
and Guangdong are mainly concentrated in the east area; the employment rate in
the central and western areas is relatively low. The employment rates of Shanxi,
Guangxi, Gansu, Inner Mongolia and other provinces rank bottom. If provinces
and cities with the low employment rate want to improve the employment rate of
the private sector, the key is to accelerate the economic development, because the
income gap of residents is a key factor to affect the employment of the private
sector in China. Meanwhile, the local governments take more proactive policies
within the scope of authority to regulate the charge order, provide the employment,
such as: wholesale and retail, business services, accommodation and catering and
manufacturing, appropriately reduce and exempt certain taxes and provide certain
tax grace period. They should also actively address the financial problem of the
private enterprises during the development, give more support and create the
development environment encouraging the private entrepreneurship, which will
have a great pull on the employment of the private enterprises.
59.4 Conclusions
Grey comprehensive evaluation method is a comprehensive evaluation method
which combines the qualitative analysis and quantitative analysis (Liu et al. 2009).
This method can not only solve the problems of evaluation indexes well that the
evaluation indexes are difficult to quantify and accurately statistic, but also exclude
582 B. Chen
the effects of personal factors. All these make the evaluation results more accurate.
Gray correlation analysis method adopts the correlation degree to quantitatively
describe the strength of the influences between things. The calculated value of the
correlation degree falls on the interval. The larger the value is, the stronger the
influence between things is. The geometric significance of the correlation degree is
the difference degree of the geometrical shapes between curves which represent
different things or factors (Sun 2011). If the correlation degree of certain index is
high, it means this index is one major factor affecting things. On the contrary, if the
correlation degree of certain index is low, it means this index has a low influence.
Applying grey correlation method into the principal component analysis to seek
the major influential factors can take several factors into consideration compre-
hensively, which avoids the subjectivity of the single factor. In this way, the
analysis process can be more reasonable and objective and the analysis results can
accurately reflect the differences between various factors. The above case shows
that gray correlation analysis method has a low requirement on the regularity of
the original data and definite objectivity and scientificity. Besides, it is simple for
use, not time consuming and easy to understand.
Acknowledgments Project supported by the Science Research Projects of the Colleges and
Universities in Inner Mongolia (NO.NJZY11106) and the Natural Science Foundation Project in
Inner Mongolia (NO.2010 MS1007).
References
Dang Y, Liu S, Wang Z (2009) Model on gray prediction and decision model. Science Press,
Beijing, China (in Chinese)
Fang F, Tang W, Cheng G (2011) Performance evaluation of beijing innovative enterprises based
on principal component analysis. J Beijing Inf Sci Technol Univ 2011(8):89–94
Feng T, Jing R, Wang G (2010) An empirical study on political capital and the accession to loan
from state-owned banks of Chinese private enterprises. Forecasting 29(3):26–31
Liu Y (2005) Analysis of the features of the boosting trend by private enterprises. J Hebei Univ
Econ Trade 26(2):25–30
Liu S, Dang Y et al (2009) Grey system on dealing with the theory and pratical applications.
Social Sciences Edition, Beijing
Song Q, Dong C (2011) On the relationship between development of private, J Chongqing Univ
Technol (Social Science) 25(7):26–31
Sun L (2011) Comparison between performance of principal component analysis and fuzzy
analysis in water quality evaluation. Environ Sci Manag 8:178–181
Wang H, Li Y, Guan R (2011) A comparison study of two methods for principal component
analysis of interval data. J Beijing Univ Aeronaut Astronaut (Social Sciences Edition)
2011(7):86–89
Zhang F (2004) On the employment potentiality of the private enterprises in Liaoning.
J Shenyang Norm Univ (Social Science Edition) 28(3):19–21
Zhang J, Hu X, Lin X (2011) Research on the financial revenue of Hainan province based on the
principal component analysis. J Hainan Norm Univ (Natural Science) 2011(9):260–264
Chapter 60
The Design of Three-Dimensional Model
for the Economic Evaluation of the Coal
Enterprise Informationization
Qing-wen Yuan and Shu-wei Yu
Abstract According to the coal enterprise characteristics and the current coal
enterprise informationization construction circumstance, the author designed a
three-dimensional model for the economic evaluation of the coal enterprise in-
formationization with a low demand for data and great practicability, so as to
direct the construction of the coal enterprise informationization, and to evaluate
the economic returns.
Keywords Coal enterprise Economic evaluation Informationization Three-
dimensional model
60.1 Introduction
In China, the enterprise informationization is defined as the utilization of infor-
mation technology, the application of information system, and the development
and utilization of the information resources in all aspects, all levels and various
areas in production, operation, and management, which aims at constantly improve
the efficiency and level of production, operation, management, decision-making
and services, so as to improve the enterprise economic benefits and its
competitiveness.
Q. Yuan (&)
Department of Finance and Economics, Shandong University of Science
and Technology, Jinan, China
e-mail: [email protected]
S. Yu
College of Economics and Management, Shandong University of Science
and Technology, Qingdao, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 583
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_60,
Ó Springer-Verlag Berlin Heidelberg 2013
584 Q. Yuan and S. Yu
Coal enterprise informationization is a complex systematic project, which needs
a huge investment, a long cycle, and a high demand for technology. This causes a
dramatic difference between the actual results and the expected results on the
informationization application. Thus, there is an urgent demand for the objective
and fair approach to evaluate the economic returns of the informationization.
60.2 The Construction of a Three-Dimensional Model
for the Economic Evaluation of the Coal Enterprise
Informationization
60.2.1 The Description of a Three-Dimensional Model
for the Economic Evaluation of the Coal Enterprise
Informationization
The three-dimensional model for the economic evaluation of the coal enterprise
informationization (hereafter referred to as the three-dimensional model) is shown
below as Fig. 60.1. The three-dimensional coordinates are delineated in Fig. 60.1.
The volume of the cube that is surrounded by the dimensions in the three-
dimensional model can be calculated by the coordinates quantification and undi-
mensionalization, and it can be used as a reference for the economic evaluation of
the coal enterprise informationization.
60.2.2 The Definition of the Dimension Coordinate Axses
in the Three-Dimensional Model
(1) X-axis: Supporting Capacity of the Current Coal Enterprise Informationization
Condition
Fig. 60.1 The three-
dimensional model for the
economic evaluation of the
coal enterprise
informationization
60 The Design of Three-Dimensional Model for the Economic Evaluation 585
Supporting capacity is the capability of current coal enterprise technology and
management for ensuring the developing information system can be on normal
operation and achieve expected results. The axis can be divided into tech-
nology and management in two directions, and formulates the technology and
management bi-domain model (Zhong 2004). The management domain
defines the current internal organizational management status of the enterprise
and the technology domain defines current informationization of the coal
enterprise.
(2) Y-axis: The Ratio of Tangible Assets to Implementation Cost of the Coal
Enterprise Informationization
The Y-axis is defined as the ratio of tangible assets to implementation cost.
Here is the formula:
RðYÞ ¼ Tangible AssetðTAÞ=Implementation CostðCÞ ð60:1Þ
The cost of informationization investment consists of two aspects. One is the
investment in the system hardware and software at the beginning of the coal
enterprise, and the other one is the cost on the operation, maintenance, and
personal expenditures of the informationization implementation. The tangible
assets are the economic returns that can be measured in currency, and can be
calculated by profit and cost.
(3) Z-axis: The Internal and External Invisible Earnings of the Coal Enterprise
Informationization
The invisible earnings refer mainly to the improvements of the efficiency of
the coal enterprise, and they are difficult to measure in currency.
Based on economist Galbraith’s theory, and the opinions of Edivinsson and
Malone, the coal enterprise informationization is a complex systematic pro-
ject. It not only affects the internal organization and staff in the coal enterprise,
but also plays a role of great importance on the upstream and downstream
firms (Edvinsson and Malone 1997).
In this paper, the author defines the invisible earnings in the coal enterprise as
intellectual capital, and the intellectual capital can be divided into organiza-
tional capital, market capital and human capital.
Considering that the invisible earnings caused by informationization in the
coal enterprise can be regarded as the increase of intangible assets, the author
sets the Z-axis origin as the center, and defines the left interval as the increase
of the internal invisible earnings, which consists of organizational capital and
human capital, and the right interval is defined as the increase of the external
invisible earnings, which is formed by the market capital (KaPlan and Norton
1996; Stewart 1994; Bharadwaj 2000).
586 Q. Yuan and S. Yu
60.2.3 The Dimension Coordinate Axses Quantification in the
Three-Dimensional Model
(1) X-axis: The Technology Domain Quantification and Management Domain
Quantification of the Informationization Supporting Capacity
(a) The Technology Domain Quantification—The Supporting Capacity of the
Coal Enterprise Informationization
The enterprise informationization supporting capacity of the technology
domain of the X-axis in the bidomain model can be evaluated by measuring
current information system supporting capacity of the developing information
system of the enterprise.
In recent years, there have been a lot of information system vcs in the coal
enterprise, and they are available in production, operations, and management.
Network technology plays an important role in the coal enterprise information-
ization application by using advanced coal mining methods, techniques, and
equipment. This results in the change from site control to process control. The
control and management data can be shared by the integration of the device layer,
control layer and information layer of the production field. Also, the application
and implementation of the ERP management informationization system software
can integrate the management informationization system and production automatic
system, together with the intensive integration of the detect ability and control
ability of the system.
The author deems the information systems to have the following relations after
intensive study of the technology foundation, data source, system compatibility,
and function component of each system.
Suppose A, B, and C are different information systems, and C is a lower system
than A and B, then the relations between A and B are shown as below:
(1) If A [ B, then A is a subsystem of B, and the supporting capacity of A to B is
the maximum value
(2) If C [ A, C [ B, and C/A [ C/B, then C is a lower system than A and B, and
the supporting capacity of A to B should be less than the supporting capacity of
B to A.
(3) If C [ A, C [ B, and C/A \ C/B, then C is a lower system than A and B, and
the supporting capacity of A to B should be greater than the supporting
capacity of B to A.
(4) If A and B are not related, then the supporting capacity of A to B is 0.
(b) The Management Domain Quantification—the Supporting Capacity of
Management
The coal enterprise management informationization is the integration of man-
agement and the advanced digital management technology. In this paper, the X-
axis management domain quantification has differences on the system classifica-
tion, organization classification.
60 The Design of Three-Dimensional Model for the Economic Evaluation 587
Table 60.1 Coal enterprise informationization cost reduction
Cost reduction Calculation method
Coal development cost Geological exploration cost reduction ? the layout of mining area
reduction cost reduction ? mine construction cost reduction
Coal production cost Mainly contains the cost reduction of material, purchased fuels, staff
reduction salaries, depreciation, upkeep, etc
Coal safety cost The safety of information, safety accidents reduction (information
reduction disclosure, gas accidents, water leakage accidents, etc.)
Inventory cost reduction The decrease of inventory capital 9 capital funds
Marketing cost reduction Advertising cost reduction ? the decrease of marketing
staff 9 salaries
Transaction cost Paper documents reduction ? the capital increase 9 capital funds
reduction
(2) Y-axis: The Ratio of Tangible Assets to Implementation Cost
(a) The calculation of information system implementation cost.
The information system implementation cost (C) can be divided into devel-
opment cost (C1) and maintenance cost (C2). And both can be subdivided by the
objects of cost. The cycle used in the calculation depends on the information
system, and the cost can be amortized over from 1 to 3 years. The formula of the
cost is as below:
C ¼ C1 þ C2 ð60:2Þ
(b) The calculation of tangible assets.
The profit of the coal enterprise informationization can be achieved by
improving management efficiency, and using labor and materials sparingly. The
tangible assets mainly come from the cost reduction (TA). The whole cost
reduction equals the sum of the development cost reduction, production cost
reduction, safety cost reduction, inventory cost reduction, and transaction cost
reduction of the informationization application.
All the formulas are shown in the Table 60.1 (Li 2005, 2009, 2010).
From the above information we can figure out:
P
RðyÞ ¼ ð60:3Þ
C1 þ C2
(3) Z-axis: The Quantification of the Internal and External Invisible Earnings.
The invisible earnings that are achieved by informationization can be measured by
the definition and classification of the coal enterprise. The coal enterprise can create a
self-evaluation table based on human capital, organizational capital, and market
capital. The result of calculation is the quantification of the domains of the Z-axis.
588 Q. Yuan and S. Yu
60.2.4 The Dimensions Coordinates Undimensionalization
of the Three-Dimensional Model
The quantification of coordinate axses are delineated above, but can not be cal-
culated directly because of the differences among dimensions. In this paper, the
author uses mathematical methods for the undimensionalization.
The catastrophe points are not available in the three-dimensional model, neither
are the sample data; therefore, the mathematical undimensionalization method is
simple and accurate.
The higher value of the coordinates in the three-dimensional model means the
better performance of the informationization. The valuation of X-axis and Z-axis
are the results of formula 60.4. The valuation of the Y-axis is already dimen-
sionless and does not need to be calculated.
The undimensionalizaiton formula is shown as below:
i imin
RðiÞ ¼ imin i imax ð60:4Þ
imax imin
60.2.5 The Calculation of Economic Evaluation
in the Three-Dimensional Model
The coordinates of the dimensions can be figured out through the definition,
quantification and undimensionalization.
As is shown on Fig. 60.1, the valuation of X1 is the result of the undimen-
sionalization of enterprise management domain. The valuation of X2 is the result
of the undimensionalization of technology domain. The valuation of Y1 is the ratio
of tangible assets to informationization cost. The valuation of Z1 is the internal
invisible earnings of the coal enterprise informationization, and Z2 is the external
invisible earnings.
The volume that is surrounded by the three dimensions is the economic eval-
uation of the coal enterprise informationization, and the formula is shown as
below.
B¼XY Z ð60:5Þ
X, Y, Z are the undimensionalization results, and B is the undimensionalization
of the coal enterprise economic benefits.
60 The Design of Three-Dimensional Model for the Economic Evaluation 589
60.3 Conclusions
The three-dimensional model has great practicability, a low demand for data and is
easier to evaluate the economic returns compared with the index synthetic
evaluation.
The three-dimensional model maintains the supporting capacity of the coal
enterprise informationization, and emphasizes the relativity and conditionality of
the enterprise informationization economic returns.
This model can not only be used for evaluating the economic returns of the
developing information system, but also can be used as a reference for choosing an
information system, and plays a role of great importance in the coal enterprise
informationization construction.
Acknowledgments Soft Science Project of Shandong Province. 2011RKGA7066.
References
Bharadwaj AS (2000) Resource-based perspective on information technology capability and firm
performance: an empirical investigation. MIS Q 24(l):169–196
Edvinsson L, Malone MS (1997) Intellectual capital. New York Inc., New York
KaPlan R, Norton DP (1996) The balanced scorecard translating strategy into action. Harvard
Business School Press, Boston
Li S-S (2005) Exploration of coal enterprise informationization. Jilin People’s Publishing House,
Beijing
Li J (2009) Study on the informationization of coal enterprises in China and its level evaluation.
Anhui University of Technology, Anhui
Li S (2010) Study on the evaluation index system of informationization level in coal enterprises.
Coal Eng 4(2):34–38
Stewart TA (1994) Your company’s most valuable asset: intellectual capital. Fortune
130(7):68–74
Zhong P-Z (2004) Research on value Chain of the manufacturing based on informationization.
Shandong University Press, Shandong, pp 64–66
Chapter 61
The Effects Decomposition of Carbon
Emission Changes and Countermeasures
in Shandong Province: Based on De
Bruyn’s Factors Decomposition Model
Guo-hua Wu, Jin-sheng Hou and Lin Wu
Abstract This article calculates the total carbon emission of Shandong Province
and the variations in carbon emission of all industries during 1995–2010 term by
term. Based on the synchronous data of population and economy, decomposition
analysis is applied to decompose the carbon emission of Shandong Province into
scale effect, structure effect and technical effect. Then the article analyzes the
relations between the average GDP’s carbon emission, the total amount of carbon
emission, the per capita of carbon emission and these effects by use of de Bruyn’s
model. Results show that: scale effect and structure effect led to the growth of
carbon emission; technical effect played opposite role to make the amount of
carbon emission lower. The article gives some countermeasures and suggestion
according to the results.
Keywords Carbon emission Countermeasures Decomposition analysis
Factors decomposition model
61.1 Introduction
In recent years, factors decomposition model has been extensively used in the area
of environment research to measure the relative importance of each effects to
environmental pollution changes. De Bruyn (1997) believes that economic scale
expansion, industrial structure change and population strength change lead the
G. Wu (&) J. Hou
Research Center of Resources Economy and Strategy, Shandong University of Finance and
Economics, Jinan 250014, People’s Republic of China
e-mail: [email protected]
L. Wu
China Minsheng Banking Corporation Ltd., Jinan 250002, People’s Republic of China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 591
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_61,
Ó Springer-Verlag Berlin Heidelberg 2013
592 G. Wu et al.
outcome. He called them as follows: scale effect, structure effect and technical
effect (Chen et al. 2004). The method of decomposition analysis becomes more
and more important because it can effectively separate the effects that lead to the
change of population from any possible effects (Stern 2002).
According to the data during 1995–2010, the article will research the effects on
the carbon emission of Shandong Province that caused by economic scale,
industries structure and technical improvement using de Bruyn’s factors decom-
position model and Ang’s PDM (parametric Divisia methods) (Ang 1994), and
then put forward corresponding measures.
61.2 Factors Decomposition Model and Estimation
on the Amount of Carbon Emission
61.2.1 Factors Decomposition Model
In this model, Ct is the amount of carbon emission in the t year, Yt is GDP in the t
year, It is the strength of carbon emission in the t year. Sit ¼ Yit =Yt is the GDP’s
proportion of added value for industry i, Iit ¼ Cit =Yit is the strength of carbon
emission for industry i in the t year. So the change of carbon emission’s amount
can be attributed to the interaction of economic scale, industrial structure and
technical improvement. The formula is:
X
Ct ¼ Yt It ¼ Yt Sit Iit ð61:1Þ
i
To attribute the change of carbon emission completely to the effects, decom-
position of allowances is needed (Sun 1998). The methods used widely now
include fixed-weight method, AWD and allowance balance method (Ang and
Zhang 2000). The article uses Ang’s PDM (parametric Divisia methods) to
decompose the carbon emission into three parts as follow:
DCsca ¼ ðYt Yt1 Þ½It1 þ bðIt It1 Þ ð61:2Þ
X
DCstr ¼ ½Yt1 þ bðYt Yt1 Þ½Iit1 þ bðIit Iit1 ÞðSit Sit1 Þ ð61:3Þ
i
X
DCtec ¼ ½Yt1 þ bðYt Yt1 Þ½Sit1 þ bðSit Sit1 ÞðIit Iit1 Þ ð61:4Þ
i
In the equations of (61.2–61.4), DCsca , DCstr , DCtec represent the improvement
of carbon emission amount caused by economic scale, industrial structure and
technical improvement in turn. b is an adjustment coefficient. If b = 0.5, so we
can get the contribution rate equations of scale effect, structure effect and technical
effect. As follow:
61 The Effects Decomposition of Carbon Emission Changes and Countermeasures 593
ðYt Yt1 ÞðIt1 þ It Þ
Rsca ¼ DCsca =Ct1 ¼ ð61:5Þ
2Yt1 It1
P
ðYt þ Yt1 ÞðIit þ Iit1 ÞðSit Sit1 Þ
i
Rstr ¼ DCstr =Ct1 ¼ ð61:6Þ
4Yt1 It1
P
ðYt þ Yt1 ÞðSit þ Sit1 ÞðIit Iit1 Þ
i
Rtec ¼ DCtec =Ct1 ¼ ð61:7Þ
4Yt1 It1
In the equations of (61.5–61.7), Rsca , Rstr , Rtec represent the contribution rate of
scale effect, structure effect and general technical effect. Ri 0 means that actor i
improve the amount of carbon emission. On the contrary, factor i reduce the
amount of carbon emission (Hu et al. 2008).
61.2.2 Carbon Emission’s Amount Estimation
The article uses equation of (61.8) to estimate the amount of carbon emission in
Shandong Province.
! !
X X X
Ct ¼ p Eti þ Lt ¼ p Ftij Mj þ Ltj Mj ð61:8Þ
i ij j
In the equation of (61.8), Ct is the amount of carbon emission (tc); p is the unit
energy’s carbon emission coefficient (tC/tce). Development and Reform Com-
mission of China believes p = 0.67 tC/tce (Wu 2009). Eti is the amount of energy
consumption of industry i (tce). Lt is the amount of energy consumption of resident
life (tce). Ftij is the amount of energy j’s consumption of industry i in the t year (t).
Ltj is the amount of energy j’s consumption of resident life in the t year (t). Mj is
the conversion factor between energy j and standard coal.
Based on the data of energy consumption caused by industries and resident life,
the article gets the data of total amount of carbon emission, and the carbon
emission of the three industries and resident life during 1995–2010. At the same
time, the article gets the pertinent data of GDP, the added value of the three
industries and the population. The data in Table 61.1 is the base for analysis.
The energy data in this article comes from the part of Shandong Province
Energy Balance Sheet (physical quantity) in ‘‘China Energy Statistics Yearbook’’
(2000–2005) and Integrated Energy Balance Sheet in ‘‘China Statistical Year-
book’’ (2011). The data of economic and population comes from ‘‘Shandong
Statistical Yearbook’’ (1996–2011) (China’s National Bureau of Statistics 2002,
2005; Statistics Bureau of Shandong province 1996).
594
Table 61.1 Basic data for analysis of carbon emission of Shandong
Years Amount of carbon emission(104 tons) Population (104 GDP(108 yuan)
persons)
Total The first The second The third Household The first The second The third
industry industry industry consumption industry industry industry
1995 5853 216 4780 542 315 8701 1263.78 2782.73 1942.10
1996 6116 257 5105 450 304 8747 1347.18 3167.87 2184.46
1997 6111 244 5145 461 261 8810 1353.46 3570.49 2500.11
1998 6024 314 4976 429 305 8872 1429.89 4001.46 2780.14
1999 6062 375 4977 388 322 8922 1497.03 4484.03 3038.67
2000 5485 296 4462 406 321 8975 1553.94 5020.78 3355.93
2001 6674 341 5323 544 466 9024 1553.94 5572.56 3732.79
2002 7380 357 6153 469 401 9069 1658.95 6406.20 4139.66
2003 8739 131 7444 718 446 9108 1750.94 7482.45 4609.95
2004 10607 152 9176 809 470 9163 1872.68 8922.83 5177.39
2005 15601 427 12707 1604 863 9212 1963.51 10478.62 5924.74
2006 17928 295 13942 2353 1338 9282 2064.57 12219.93 6785.73
2007 19548 324 15153 2597 1474 9346 2147.04 14145.95 7773.41
2008 20481 344 15781 2772 1584 9392 2255.53 15847.09 8856.99
2009 21721 368 16620 3035 1698 9449 2349.15 18043.87 9849.62
2010 23009 260 18554 2861 1334 9536 2434.41 20346.93 11182.09
Notes the equation of carbon emission amount for industry i: Ci = pEi, It’s the product of the unit energy’s carbon emission coefficient and the amount of
energy consumption of industry i, the price of GDP is according to the GDP of 2005 and the population is household register number
G. Wu et al.
61 The Effects Decomposition of Carbon Emission Changes and Countermeasures 595
61.3 The Analysis of Carbon Emission Effects
in Shandong Province
According to the equations of (61.5–61.7), the article calculates the contribution
rate of scale effect, structure effect and technical effect. Table 61.2 and Fig. 61.1
reflect the outcome.
Table 61.2 Contribution of different effects to carbon emission in Shandong
Years Scale effect (Rsca) Structure effect (Rstr) Technical effect (Rtec)
1995–1996 11.6454 1.1701 -7.5401
1996–1997 10.5321 1.0257 -10.4468
1997–1998 10.1588 0.8003 -12.6205
1998–1999 9.5920 1.1928 -9.9526
1999–2000 9.3573 0.8643 -19.2023
2000–2001 10.5714 0.5397 8.6138
2001–2002 11.6695 1.9356 -1.3590
2002–2003 13.7051 2.3206 2.5217
2003–2004 15.7992 2.7512 3.2888
2004–2005 17.0911 1.8421 25.3113
2005–2006 14.7321 1.1802 -3.0838
2006–2007 13.8973 0.9678 -5.5482
2007–2008 11.6328 0.1263 -6.6613
2008–2009 11.8451 0.9289 -6.3562
2009–2010 11.9538 0.3628 -3.8932
Average 12.2789 1.2006 -3.1286
30
20
effect contribution (%)
10
0
Average
1995-1996
1996-1997
1997-1998
1998-1999
1999-2000
2000-2001
2001-2002
2002-2003
2003-2004
2004-2005
2005-2006
2006-2007
2007-2008
2008-2009
2009-2010
-10
-20
-30
Scale Effect ( Rsca ) Structure Effect ( Rstr ) Technical Effect ( Rtec )
Fig. 61.1 The effects of carbon emission for 1995–2009 in Shandong
596 G. Wu et al.
61.3.1 Scale Effect
In the research time, the contribution rate of carbon emission scale effect is always
above 0 (9–18 %), the average rate is 12.3 %. It means that the enlargement of
economic scale improves carbon emission amount. The change path likes an
inverted ‘‘N’’ along with the growth of economic (Fig. 61.1). It’s basically the
same with the economic development of Shandong Province. 1995–2000, and the
improvement speed of economic was drop obviously in the background of the
Asian financial crisis, the speed dropped from 12.02 % in 1996 to 10.28 % in 2000
and the influence of scale effect became weaker. The improvement speed of
economic was straight up in the following 5 years, the influence of scale effect
became stronger in this period and it reached the maximum as 15 % in 2005. In the
period of The Eleventh-Five, China government adopted the policy of energy-
saving and emission reduction, the efficiency of energy consumption was
improved clearly in Shandong Province, and the strength of energy consumption
dropped 22.1 % than last 5 years. Because of the Global financial crisis in 2008,
the speed of economic improvement dropped slowly again and dropped to 12.30 %
of 2010, the scale effect became much weaker. The economic improvement path
obviously proved the change of scale effect. The by-product (carbon emission) will
grow large with the development of economic scale. So, obviously, the press from
scale effect of carbon emission will keep long time in the developing Shandong
Province even all over China which is a developing country.
61.3.2 Structure Effect
As the same with scale effect, structure effect also improves the amount of carbon
emission in the research period, with the numbers between 0–3 %, the average
number 1.2 %, and the influence much weaker. It suggested that in the passed
15 years the change of economic structure did not reduce the amount of carbon
emission but improved it, although the number is only 10 % of scale effect. In the
last 15 years, the structure of industries in Shandong Province is 20.4:47.6:32.0 in
1995 moves to 9.2:54.2:36.6 in 2010. The ratio of added value of the first industry
dropped 11.2 % while the second industry and the third industry improved 6.6 and
4.6 % in turn. Although the structure of in industries has improved, the second
industry which emits 5–6 times carbon than the first industry and 3–4 times than
the third industry became larger in the structure. It directly leads the number of
structure effect above 0, becoming the draw power of carbon emission. This sit-
uation means the change of structure is important and urgent to carbon emission
reduction.
61 The Effects Decomposition of Carbon Emission Changes and Countermeasures 597
61.3.3 Technical Effect
In the research period, the fluctuation of technical effect was strong, with the
contribution numbers between -20 and 26 %. In most of the years the numbers
were below 0 except the period from 2000 to 2005, and the average number is -
3.1 %. It means that the contribution of technical effect is not quite obviously and
has a certain amount of randomness. But in average, the technical effect reduced
the amount of carbon emission totally. In fact, the promoted contribution of
technical effect comes from the technical improvement of industry units. In the
period of Eleven-Five, the policy of energy-saving and emission reduction was
executed powerful in Shandong Province. Some specific policies got notable
achievements like ‘‘eliminating backward production capacity’’, ‘‘developing
technical of energy-saving’’. The efficient of energy-using improved large. These
actions led the number of technical effect contribution dropped to -6.7 % in 2008
form 25.3 % in 2005 and retain below 0. The technical effect inhibits the growth of
carbon emission efficiently, although the space is still large. So it’s an important
task to insist for a long time that using energy-saving technology to strength the
technical effect, such as ‘‘cogeneration of heating and electric power’’, ‘‘waste heat
and pressure generate electric power’’, ‘‘new types of motors’’ and so on.
61.4 Countermeasures and Suggestions
The analysis above shows that among the three mechanisms that effect the changes
in carbon emission, scale effect and structure effect enlarge the amount of carbon
emission, technical effect inhibits the amount of carbon emission. According to the
analysis, the article gives several countermeasures to promote the development of
low carbon economic.
61.4.1 Speed up the Adjustment of Industrial Structure,
to Change Structure Effect from Positive to Negative
The proportion of the second industry in the industrial structure of Shandong
Province is too large, that’s why structure effect enlarged the amount of carbon
emission. So, the government should take efficient actions to change the industrial
structure. First, the government should strengthen the review of energy con-
sumption of all investment projects, and control the development of the industries
which cost too much energy. Second, the government should accelerate the
upgrade of traditional industries and encourage the development of new strategic
industries and new types of service industry. These measures will enable structure
effect to change from positive to negative, from enlarging to inhibiting the carbon
emission.
598 G. Wu et al.
61.4.2 Perfect the Policies about Carbon Emission
Reduction, Establish and Sound the Market
Mechanism
First, the government should establish and sound the finance and taxation policies
to encourage energy-saving and emission reduction, enhance financial support to
make energy-saving technology industrialization, to develop new energy and
environmental industry, and to eliminate and modify high energy-consuming
devices. Meanwhile, tax breaks on energy conservation should be implemented to
control high energy-using and emission products and consumption. Second, the
government should speed up the service industries that are energy-saving, practice
marketing methods like ‘‘contract manage the energy-using’’ and ‘‘certificate
energy-saving products’’, cultivate a cluster of energy-saving equipment industry,
strength the competitiveness of energy-saving products, and build long time
mechanism of energy-saving and carbon emission reduction.
61.4.3 Accelerate the Development of Energy-Saving
Technology, to Achieve that Technology Dominates
Emissions Cuts
The special campaigns about energy-saving and emissions cuts should be
continued. By analysis the traits of enterprises and industries with high energy
consumption, the government should choose and provide supports for the devel-
opment, import and absorption of relevant energy-saving technology. Our gov-
ernment should build energy-saving industrial centers and support the
industrialization of major energy-saving technology and equipment, which will
make a mutual promotion between the development of industry and technology,
and enhance inhibiting effect of technological progress on carbon emission.
61.4.4 Determine Energy Consumption Indexes
Scientifically, Control the Amount
of Energy Emission
The situation of energy supplies should be changed. At the basis of local level of
economy, industrial structure, energy structure and the level of energy consump-
tion, the local government ought to determine its target for energy consumption,
and control the total amount of energy cost. According to the discipline that
economy growing should keep a balance with society, energy consumption
indexes system that combines total amount index with strength index of energy
61 The Effects Decomposition of Carbon Emission Changes and Countermeasures 599
consumption should be established, as well as energy efficiency examining
mechanism remained by total amount control, and supplemented by strength
control.
61.4.5 Strengthen Advocacy and Education, Launch
Universal Low Carbon Campaign
Our government should publicize the knowledge about resources environment and
climate changes, such as energy-saving and low carbon economy. Base on NGOs
and grass-roots communities, launching actions focused on energy efficiency and
low carbon widely can help to build low-carbon conception, strength low-carbon
awareness, popularize the pathway how to product and consume in a low-carbon
way, and launch universal low carbon campaign.
Acknowledgments Supported by Chinese Ministry of Education humanities the social sciences
research fund item (NO. 10YJA790199) and Shandong Province Natural Science Foundation
Project (NO. ZR2009HM02).
References
De Bruyn SM (1997) Explaining the environmental Kuznets curve: structural change and
international agreements in reducing sulphur emissions. Environ Dev Econ 2(4):485
Chen L, Wang D, Fang F (2004) Main factors of pollution in China—decomposition model and
empirical analysis. J Beijing Normal Univ (Natural Science Edition) 40(8):561–568
Stern DI (2002) Explaining changes in global sulfur emissions: an econometric decomposition
approach. Ecol Econ 42(2):201
Ang BW (1994) Decomposition of industrial energy consumption: the energy intensity approach.
Energy Econ 18:163–174
Sun JW (1998) Changes in energy consumption and energy intensity: a complete decomposition
model. Energy Econ 20(1):85
Ang BW, Zhang FQ (2000) A survey of index decomposition analysis in energy and
environmental studies. Energy 25(12):1149
Hu C, Huang X, Zhong T, Tan D (2008) Character of carbon emission in china and its dynamic
development analysis. China Popul Resour Environ 3:38–42
Wu G (2009) Research on energy-saving emission reduction strategies in China. Economic and
Science Press, Beijing, pp 15–28
China’s National Bureau of Statistics (2002–2005) Energy statistics yearbook. China Statistical
Publishing House, Beijing
China’s National Bureau of Statistics (2005) China statistical yearbook. China Statistical
Publishing House, Beijing
Statistics Bureau of Shandong province (1996–2010) Shandong statistical yearbook. China
Statistical Publishing House, Beijing
Chapter 62
The Empirical Research of the Causality
Relationship Between CO2 Emissions
Intensity, Energy Consumption Structure,
Energy Intensity and Industrial Structure
in China
Tao Zhao and Xiao-song Ren
Abstract This paper adopts econometrics methodology to explore the causality
relationship among CO2 emissions intensity, energy consumption structure, energy
intensity and industrial structure in China, during the time period 1980–2009.
Multivariate cointegration and vecm model are applied to explore the long-term
equilibrium and short-term fluctuation. Base on the granger causality test results, it
concludes that there exist four group unidirectional causality relationships from CO2
emissions intensity to energy consumption structure, CO2 emissions intensity to
industrial structure, industrial structure to energy consumption structure and energy
intensity to energy consumption structure. In addition, a bidirectional causality
relationship running from CO2 emissions intensity to energy intensity is detected.
Keywords Carbon dioxide emissions intensity Energy consumption structure
Energy intensity Industrial structure Multivariate cointegration Vecm model
62.1 Introduction
The empirical research of the causality relationship between energy consumption,
economic output, carbon dioxide emission and some other variables is a hot topic
in the research field of low carbon economy, since the pioneering research of Kraft
and Kraft (1978) who explored the causality relationship between economic
growth and energy consumption in the United States during the period 1947–1974
(Kraft and Kraft 1978). However, due to diverse countries focused, specific time
span selected, various variables used and different econometric methodologies
adopted, the conclusions of the causality relationship study are uncertain, and even
controversial (Ilhan 2010).
T. Zhao X. Ren (&)
School of Management and Economic, Tianjin University, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 601
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_62,
Ó Springer-Verlag Berlin Heidelberg 2013
602 T. Zhao and X. Ren
Ang used multivariate cointegration and vecm model to explore the dynamic
causality relationships between carbon dioxide emissions, energy consumption,
and economic growth in France for the period 1960–2000 (James 2007). Zhang
and Cheng (2009) adopted the TY-VAR method to discuss the existence and
direction of the causality relationship between carbon emissions, economic
growth, and energy consumption in China from 1960 to 2007 (Zhang and Cheng
2009). Feng et al. (2009) employed cointegration analysis and Granger causality
test to explore the long-run equilibrium relationships, short term dynamic rela-
tionships and causality relationships between energy intensity, energy consump-
tion structure and economic structure in China over the time span from 1980 to
2006 (Feng et al. 2009). Halicioglu (2009) used bounds test and Granger causality
analysis to explore dynamic causality relationships between economic output,
carbon emissions, energy consumption and foreign trade in Turkey for the period
1960–2005 (Halicioglu 2009). Soytas, Sari adopted the TY approach to explore the
long run Granger causality relationship between carbon dioxide emissions, eco-
nomic growth and energy consumption in Turkey (Ugur and Ramazan 2009).
Chang detected the causality relationship between carbon dioxide emissions,
energy consumption and economic growth in China from 1981 to 2006, applying
the vecm model and Granger causality test. Lotfalipour used Toda-Yamamoto
method to examine the causality relationships between carbon emission, energy
consumption, and economic growth in Iran over the time from 1967 to 2007
(Mohammad et al. 2010). Menyah and Wolde-Rufael adopted the bound test and
Granger causality test to analyze the causality relationship between carbon dioxide
emissions, economic growth, and energy consumption in South Africa over the
time from 1965–2006 (Kojo and Yemane 2010). Chang and Carballo used vecm
and var model to explore the cointegration and causality relationships between
carbon dioxide emissions, energy consumption and economic growth for twenty
countries belong to Latin America and the Caribbean region during the period
1971–2005 (Chang and Claudia 2011). Hatzigeorgiou et al. employed the vecm
model to explore the dynamic causality relationship between GDP, CO2 emissions
and energy intensity for Greece during the period 1977–2007 (Emmanouil et al.
2011). Bloch et al. used Johansen multivariate cointegration and vecm model to
investigate the causality relationships among carbon dioxide emission, coal con-
sumption and economic output for China (Harry et al. 2012). Jayanthakumaran
et al. used the bounds cointegration analysis and the ARDL model to test the long
run and short term relationships among carbondioxide emission, economic growth,
trade and energy consumption during the time 1971–2007, comparing China with
India (Kankesu et al. 2012). Chen used a multinomial logit model to examine the
key elements affecting the causality relationships between energy consumption
and economic output for 174 samples (Cheng et al. 2012).
Above all, the conclusions derived from these empirical researches are various
and even conflicted. It is because that the different data sets are collected, different
time spans are selected, different econometric model are applied and different
countries are focused. In this paper, we apply multivariate cointegration, vecm
model and granger causality test to examine the causality relationship among CO2
62 The Empirical Research of the Causality Relationship Between CO2 Emissions 603
emissions intensity, energy consumption structure, energy intensity and industrial
structure for China, over the time period 1980–2009.
The rest parts of this paper are arranged as follows. Section 62.2 presents data
source and process. Section 62.3 describes econometric methodologies and models
applied in this paper. Section 62.4 reveals the results of the empirical research
analysis. Section 62.5 finally puts forward some conclusions and policy
suggestions.
62.2 Data Source and Process
We collect and calculate the annual data of CO2 emissions intensity, energy
consumption structure, energy intensity and industrial structure in China during the
time period 1980–2009 as research samplings. CO2 emissions intensity denotes
CO2 emissions per unit of GDP, which is defined as CI. The coal proportion
accounting for the primary energy consumption is used to describe energy con-
sumption structure, which is called ECS. Energy intensity denotes primary energy
consumption per unit of GDP, which is named EI. The proportion of tertiary
industry is used to represent industrial structure, which is denoted as IS. The
annual data of GDP is collected from China statistical yearbook of 2010, the
primary energy consumption data is from China’s energy statistics yearbook of
2010, and the CO2 emissions data come from the databank of world development
indicators.
62.3 Methodology
62.3.1 Unit Root Test
Cointegration analysis claims that the level time series or the same difference time
series of variables must be stationary. So it is necessary to test the stationarity of
variables using unit root test. ADF test, introduced by Dickey and Fuller, is the
most broadly applied root test method. The model is represented as follows.
X
k
DCIt ¼ qCIt1 þ ki DCIti þ et ð62:1Þ
i¼1
X
k
DECSt ¼ qECSt1 þ ki DECSti þ et ð62:2Þ
i¼1
604 T. Zhao and X. Ren
X
k
DEIt ¼ qEIt1 þ ki DEIti þ et ð62:3Þ
i¼1
X
k
DISt ¼ qISt1 þ ki DISti þ et ð62:4Þ
i¼1
62.3.2 Multivariate Cointegration Test
If the variables pass the unit root test, we can use cointegration tests to explore the
long run equilibrium relationships among the variables, which introduced by
Johansen and Juselius. Base on the maximum likelihood procedure, the Johansen
co integration test are used to confirm the existence of long run equilibrium
relationship among CO2 emission intensity, energy consumption structure, energy
intensity, and industrial structure during the time period 1980–2006. The trace
statistic value can be used to ascertain the existence of cointegration. The model is
as follow.
0 1 0 1 0 1
CI CI CI
B ECS C B C X
p1
B ECS C
DB C ¼ ab0 B ECS C þ Ci DB C
@ EI A @ EI A @ EI A þet ð62:5Þ
i¼1
IS IS t1 IS ti
62.3.3 Vector Error Correction Model
VECM is the cointegrated constraint vector auto regression model, introduced by
Sargan, developed by Engle and Granger, and is applied to deal with the cointe-
gration relationship between the non-stationary time series modeling. VECM can
comprehensively reflect the long-term equilibrium and short-term fluctuations
between carbon dioxide emissions, energy consumption structure, energy intensity
and industrial structure. The VECM model can be specified as follow.
0 1 0 1
CI CI
B ECS C X
p1
B ECS C
DB C
@ EI A ¼ aECMt1 þ Ci DB C
@ EI A þet ð62:6Þ
i¼1
IS IS ti
62 The Empirical Research of the Causality Relationship Between CO2 Emissions 605
Table 62.1 Results of ADF unit root tests
Variables ADF Critical value Critical value Critical value Stationary/
test (1 % level) (5 % level) (10 % level) nonstationary
CI -1.443 -3.689 -2.972 -2.625 Nonstationary
ECS -0.922 -3.679 -2.968 -2.623 Nonstationary
EI -0.944 -3.689 -2.971 -2.625 Nonstationary
IS -1.237 -3.679 -2.968 -2.623 Nonstationary
DCI -3.199 -3.689 -2.972 -2.625 Stationary**
DECS -4.190 -3.689 -2.972 -2.625 Stationary***
DEI -3.379 -3.689 -2.972 -2.625 Stationary**
DIS -3.837 -3.689 -2.972 -2.625 Stationary***
62.4 Empirical Analysis
62.4.1 Stationarity Test Results
Considering the popularity and practical applicability, Augment Dickey–Fuller
(ADF) test is adopted in this paper. The equations of the ADF model all include
the intercept terms, but no trend terms. Table 62.1 represents the results of the
ADF unit root test in level and first difference time series of these four variables.
From the Table 62.1, we can see that the t statistics value of four variables (CI,
ECS, EI, IS) are all larger than the critical values at the 5 % level in the ADF unit
root test. It denotes that the null unit-root hypothesis can be accepted at the 5 %
level, that is to say, these four variables are all non-stationary in their level time
series. DCI, DECS, DEI and DIS are the first differenced variables of CI, ECS, EI
and IS. The ADF unit test statistics of the first differenced variables are smaller
than the critical values at the 5 % significant level; especially the ADF test sta-
tistics of DCI and DEI are smaller than the critical values at the 1 % significant
level. After fist difference, these four variables become stationary. It denotes that
these four variables are integrated of order one, which implies that there may exist
long term cointegration relationships among these four variables.
62.4.2 Multivariate Cointegration Test Results
As reflected in Table 62.2, over the time from 1980 to 2009, the trace test dem-
onstrates that at the 5 % significance level, just one cointegration relationship
exists among CO2 emission intensity, energy consumption structure, energy
intensity, and industrial structure. The cointegration equation is represented as
follow.
CI ¼ 0:201 ECS þ 0:017 EI 0:281 IS ð62:7Þ
606 T. Zhao and X. Ren
Table 62.2 Results of Johansen cointegration tests
Hypothesized Eigenvalue Trace 0.05 Probability **
No. of CE(s) Statistic Critical value
None * 0.621 43.376 40.175 0.023
At most 1 0.311 16.225 24.276 0.364
At most 2 0.181 5.777 12.321 0.464
At most 3 0.007 0.195 4.130 0.714
The cointegration equation expresses that, in the long term, both energy con-
sumption structure and energy intensity have positive role to CO2 emission
intensity, while industrial structure plays an obvious negative part in CO2 emission
intensity. Supposing that other variables remain unchanged, if the energy con-
sumption structure is improved by 1 %, CO2 emission intensity will lessened
accordingly by 0.201 %. With energy intensity dropping by one percentage, CO2
emission intensity is cut down by 0.017 %. If the proportion of tertiary industry of
the industrial structure goes up in 1 %, CO2 emission intensity may decrease by
0.281 %.
62.4.3 VECM Results
CI, ECS, EI and IS have one cointegration relationship, so a corresponding error
correction model can be constructed to represent the long-term equilibrium and
short-term fluctuations in the cointegrating relationship. Carbon dioxide emissions
intensity as explained variables, the error correction model equations is repre-
sented as follows:
DCIt ¼ 0:843DCIt1 þ 0:3394DECSt1 0:005DEIt1 0:049DISt1
ð62:8Þ
0:39ECMt1
ECMt1 ¼ CIt1 0:201ECSt1 0:017EIt1 þ 0:281ISt1 ð62:9Þ
The equations of the VECM demonstrate the long run equilibrium and short
term fluctuation among CI, ECS, EI and IS. In the short term, compared with ECS
and EI, IS has the biggest influence on the volatility of CI, except for the lag term
of CI itself. The Eq. (62.8) shows that the coefficient of the error correction term is
negative, which is in accord with reverse revision mechanism. When carbon
dioxide emissions intensity appears short-term volatility off long-term equilibrium,
the system will be adjusted to the stable equilibrium at the 39 % adjustment speed.
62 The Empirical Research of the Causality Relationship Between CO2 Emissions 607
Table 62.3 Results of Granger causality tests
Null hypothesis F-statistic Prob. Accept/
Reject
ECS does not Granger Cause CI 0.070 0.794 Accept
CI does not Granger Cause ECS 10.985 0.003a Reject
EI does not Granger Cause CI 2.275 0.144 Accept
CI does not Granger Cause EI 3.910 0.059a Reject
IS does not Granger Cause CI 0.288 0.596 Accept
CI does not Granger Cause IS 3.910 0.059a Reject
EI does not Granger Cause ECS 8.445 0.007a Reject
ECS does not Granger Cause EI 0.316 0.579 Accept
IS does not Granger Cause ECS 4.992 0.034a Reject
ECS does not Granger Cause IS 0.252 0.620 Accept
IS does not Granger Cause EI 0.024 0.877 Accept
EI does not Granger Cause IS 3.631 0.068a Reject
a
Denotes rejection of the hypothesis (X does not Granger cause Y) at the 10 % level
62.4.4 Granger Causality Tests
The result of cointegration test and VECM model denote that there exist some
causal relationships among CI, ECS, EI and IS. However, it cannot ascertain the
direction and the number of the causal relationships, and then the Granger cau-
sality test is used. The result shows as Table 62.3.
If the null Hypothesis is accepted at the 10 % significant level, it denotes that
there is not the causality relationship. Table 62.3 reveals that there exist four group
unidirectional causality relationships running from CI to ECS, CI to IS, EI to ECS,
and IS to ECS. Meanwhile, the bidirectional causality relationship between CI and
EI is appeared. The decrease of CO2 intensity will improve energy consumption
structure, which leads to the decrease of coal consumption amount. The decrease
of CI also can promote industrial structure improvement, and stimulate the pro-
gress of the tertiary industry. Meanwhile, if energy intensity drops, the third
industry in the country industrial structure in the total increase will promote the
energy structure to be improved.
62.5 Conclusions and suggestions
62.5.1 Conclusion
Through the above analysis this paper gives conclusions as follows.
(1) Four variables (CI, ECS, EI, IS) are stationary in first differences and have one
cointegration relationship. In the short term, although China’s carbondioxide
emissions intensity, energy consumption structure, energy intensity and
608 T. Zhao and X. Ren
industrial structure have the fluctuation relationship, a stable equilibrium
relationship exists among them in the long run. In this long term equilibrium
relationship, energy consumption structure and energy intensity have a posi-
tive impact on boosting the growth of carbondioxide emissions intensity, while
the industrial structure plays a negative inhibition effect on carbon dioxide
emissions intensity. In comparison, the industrial structure has the strongest
impact on the growth of carbondioxide emissions, followed by energy con-
sumption structure, and energy intensity has the least role among them.
(2) In the short term, compare with energy consumption structure, energy inten-
sity, industrial structure has the biggest impact on the volatility of CO2
emission intensity, except for the lag term of CO2 emission intensity itself.
The coefficient of the error correction term in vecm equation is negative,
which is in accord with reverse revision mechanism. When carbon dioxide
emissions intensity appears short-term volatility off long-term equilibrium, the
system will be adjusted to the stable equilibrium at the 39 % adjustment speed.
(3) There exist four group unidirectional causality relationships from CO2 emis-
sion intensity to energy consumption structure, CO2 emission intensity to
industrial structure, energy intensity to energy consumption structure, and
industrial structure to energy consumption structure. In addition, a bidirec-
tional causality relationship running from CO2 emissions intensity to energy
intensity is detected.
62.5.2 Suggestions
(1) To speed up the adjustment of industrial structure of our country, increase the
proportion of the third industry. With high economic added value and low
energy consumption characteristics, the proportion of third industry has
become an important symbol of development of a country’s low carbon
economy. The industrial structure has the biggest impact on carbondioxide
emissions intensity. At present, the proportion of industries with low added
value, high energy consumption and high pollution, is too big in national
economy of China, this is the main cause carbon of the high dioxide emissions
intensity. So our government has to implement strategic adjustment to the
national economic structure, eliminate energy-intensive and low production
value industries, emphasize on the development of high added value, low
energy consumption and high technology industries, try to develop the new
service industry, and the new energy industry.
(2) Improve the coal-dominated energy consumption structure. The annual aver-
age proportion of coal amounting for the primary energy consumption is 76 %
during period from 1980 to 2009. It also determines that China may not
change the current energy consumption structure in the short term, but could
increase oil and gas consumption proportion, and develop water power, wind,
62 The Empirical Research of the Causality Relationship Between CO2 Emissions 609
nuclear energy, solar energy and other clean energy to gradually improve
China’s energy consumption structure.
(3) Strengthen the low carbon technology innovation and improve energy effi-
ciency. Since the reform and open policy, our country industrialization process
gets the rapid development, along with the phenomenon of excessive energy
consumption and serious waste. Low carbon technology innovation is an
important way to realize national resource conservation and environment
friendly development. In order to support low carbon technology innovation,
the government can levy resource tax, energy tax and environmental taxes to
make subsidies for low carbon technology innovation activities.
References
Chang CC, Claudia CSF (2011) Energy conservation and sustainable economic growth: the case
of Latin America and the Caribbean. Energy Policy 39(7):4215–4221
Cheng P-Y, Chen S-T, Chen C-C (2012) Energy consumption and economic growth—new
evidence from meta-analysis. Energy Policy 44(4):245–255
Ching-Chih C (2010) A multivariate causality test of carbon dioxide emissions, energy
consumption and economic growth in China. Appl Energy 87(11):3533–3537
Emmanouil H, Heracles P, Dias H (2011) CO2 emissions, GDP and energy intensity: A
multivariate cointegration and causality analysis for Greece, 1977–2007. Appl Energy
88(4):1377–1385
Feng T, Sun L, Zhang Y (2009) The relationship between energy consumption structure,
economic structure and energy intensity in China. Energy Policy 37(12):5475–5483
Halicioglu F (2009) An econometric study of CO2 emissions, energy consumption, income and
foreign trade in Turkey. Energy Policy 37(3):1156–1164
Harry B, Shuddhasattwa R, Ruhul S (2012) Coal consumption, CO2 emission and economic
growth in China: empirical evidence and policy responses. Energy Econ 34(2):518–528
Ilhan O (2010) A literature survey on energy-growth nexus. Energy Policy 38(1):340–349
James AB (2007) CO2 emissions, energy consumption, and output in France. Energy Policy
35(10):4772–4778
Kankesu J, Reetu V, Ying L (2012) CO2 emissions, energy consumption, trade and income: a
comparative analysis of China and India. Energy Policy 42(3):450–460
Kojo M, Yemane W (2010) Energy consumption, pollutant emissions and economic growth in
South Africa. Energy Econ 32(6):1374–1382
Kraft J, Kraft A (1978) On the relationship between energy and GNP. J Energy Dev 3:401–403
Mohammad LR, Mohammad FA, Malihe A (2010) Economic growth, CO2 emissions, and fossil
fuels consumption in Iran. Energy 35(12):5115–5120
Ugur S, Ramazan S (2009) Energy consumption, economic growth, and carbon emissions:
challenges faced by an EU candidate member. Ecol Econ 68(6):1667–1675
Zhang X-P, Cheng X-M (2009) Energy consumption, carbon emissions, and economic growth in
China. Ecol Econ 68(10):2706–2712
Chapter 63
The Evaluation of China Construction
Industry Sustainable Development
on DEA Model
Peng-Yang Liu, Jian-Ping Yang and Fan-Fang Wen
Abstract Using the DEA model and based on the panel data of China construction
industry between the year of 1995 and 2008, this paper chooses the number of
employees, fixed assets and the total power of mechanical equipment as input index,
gross output and value added as output index, to conduct an overall evaluation of
China construction industry Sustainable Development. Based on the input–output
index, the study data shows that China construction industry has been in low level of
development, and experienced from dropping to rising and begun to develop well
since 2002. Furthermore, according to the results of the element adjustment, the
sustainable development of China construction industry can be achieved by opti-
mizing employees’ structure and proportion of investment assets, improving the
operational efficiency of mechanical equipment and expanding the market actively.
Keywords Construction industry DEA model Evaluation Sustainable
development
63.1 Introduction
Driven by rapid economic growth, China construction industry is developing
rapidly for more than 30 years of reform and opening up. Therefore, it is partic-
ularly important to evaluate the sustainability of China construction industry. We
P.-Y. Liu (&) J.-P. Yang F.-F. Wen
Department of Management, Xi’an University of Architecture & Technology, Xi’an, Shan
Xi, China
e-mail: [email protected]
J.-P. Yang
e-mail: [email protected]
F.-F. Wen
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 611
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_63,
Ó Springer-Verlag Berlin Heidelberg 2013
612 P.-Y. Liu et al.
can understand the level of development in the construction industry with evalu-
ation material, and provide a scientific basis for the industrial development policy
with the aid of the evaluation results. After collecting and analyzing literatures
above, we can find that there are some problems in determining the index weight in
the use of evaluation methods (Zhang 2010). Therefore, in this paper, data
envelopment analysis (DEA) is applied to evaluate which of the sustainable
development of China construction industry, don’t need to determine the weight of
the index, to avoid subjective factors with the objective validity, which is more
objective and effective to the evaluation results of the construction industry’s
sustainable development. Furthermore, the results are adjusted (Xiao 1994).
63.2 Basic Idea of DEA and Models
63.2.1 Basic Idea of DEA
Data Envelopment Analysis (DEA) is one of the non-parametric methods to
evaluate the efficiency of multiple decision-making units and their relative effec-
tiveness, which was first proposed by A. Chames, W. W. Cooper and E. Rhodes.
The idea of DEA is based on the concept of relative efficiency, DEA regards each
target of evaluation as a decision making unit (DMU) and forms a evaluated group
by all the DMUs, according to the DMU data, using the mathematical program-
ming we can analyzes the input–output data comprehensively, calculate the
indicators of relative efficiency of each DMU to determine whether each DMU is
efficient or not, and find the reason of non DEA efficiency, the direction and extent
of improvement by using the projective method. DEA method does not need pre-
estimate parameters, so it has the advantage in avoiding subjective factors, sim-
plifying calculation and reducing errors (Lu and Cai 2006).
63.2.2 The Model of C 2 R
This paper uses C2 R model of DEA to analyze. The envelopment form of the C2 R
model for evaluating DMUj is as follows:
63 The Evaluation of China Construction Industry 613
8 " !#
>
> X t X t
>
> Min h e þ
sr þ
sr
>
>
>
> r¼1 i¼1
>
>
>
> Xn
>
>
>
> s:t: kj xij þ si h xij0 ¼ 0
>
>
< j¼1
X n ð63:1Þ
> kj yij sþ
>
>
> r ¼ yij0
>
> j¼1
>
>
>
> kj 0; j ¼ 1; 2; . . .; n
>
>
>
>
> s
>
> i 0
>
:
sþ
r 0
where h is the efficiency score which is unconstrained, k ¼ ðk1 ; . . .; kn ÞT is a
‘‘structural’’ variable vector. Suppose we have a set of n peer DMUs consisting of
DMUj ðj ¼ 1; 2; . . .; nÞ, and each DMUj consumes m inputs xij ði ¼ 1; 2; . . .; nÞ and
generates s outputs yrj ðr ¼ 1; 2; . . .; sÞ, s þ
i (slack of input) and sr (slack of output)
represent input excess and output shortfall respectively; and e is a non-Archime-
dean infinitesimal which is defined to be smaller than any positive number; kj , sþ r ,
si and h represent the Parameters to be estimated (Li 2007).
In Eq. (63.1), kj is a combination proportion reconstructed an effective DMU
combination of the j decision unit of DMU combination in relate to DMUi . n is the
number of DMU decision unit; m, t represents respectively the quantity of input
index and output index; xij is the j decision unit for the i type of inputs; yrj is the j
decision unit for the r type of output; sþ
r (slack of input) and si (slack of output)
represent input excess and output shortfall respectively; e is a non-Archimedean
infinitesimal which is generally taken e ¼ 106 ; h is valid values for the DMU
decision unit, that is relative efficiency for the first a decision unit for the input in
relate to output. kj , sþ
r , si and h represent the Parameters to be estimated (Gong
and Zhang 2004).
63.2.3 The Economic Meaning of C2R
(1) When considering the factor of ‘‘time’’ in C 2 R, making a time i in random, if
any hðt iÞ\hðtÞ, the evaluated system is in the good developing track; if
hðt iÞ ¼ hðtÞ existing, the evaluated system is weak less than developing
track; if hðt iÞ [ hðtÞ existing, the evaluated system is less than developing
track. P
(2) For non-effective DMU unit, if ki ¼ 1, DMU is efficient
P to technical effi-
ciency, otherwise, it is valid; supposing K ¼ 1=h kj , when K ¼ 1, the
returns-to-scale of DMU is prevail, when K\1, the return-to-scale of DMU is
increasing prevail, otherwise is decreasing prevail (Deng et al. 2008).
614 P.-Y. Liu et al.
(3) If and only if the efficiency score h is equal to one and all optimal slacks are
zero, DMU is called efficient unit, the formation of the efficient frontier for
constant returns to scale, and the DMU is technical efficiency and scale effi-
ciency. When h is equal to one but sþ
r 6¼ 0 or si 6¼ 0, DMU is called weak
efficient unit; When h \ 1, DMU is inefficient, maybe in technical inefficiency
or scale inefficiency (Sheng 1996).
63.3 Empirical Analysis
63.3.1 Data and Selection of Inputs and Outputs
Using the C 2 R model, this paper takes the Chinese construction 1995–2008 year
for decision unit, selects the calendar year panel data of China construction
industry as the analysis data of input and output indicators. The date originates
calendar year ‘‘Chinese Statistic Almanac’’ from 1999 to 2009.
From the input–output point of view, the sustainable development of the con-
struction industry includes input subsystem and output subsystem, the key elements
of each subsystem are selected as the input and output indicators.
As human, fund, and material are the three types of input elements in the
construction industry, input indicators selected in this paper for the number of
employees of China construction industry, the fixed assets and total power of its
own machinery and equipment at the end of the year, which represent three types
of elements as the labor input, capital inputs, machinery and equipment input;
output indicators selected for the construction industry output value and added
value (Porteous 2002). The construction industry output value can reflect the
overall production of the construction industry efficiency better relative to revenue,
gross profit, and other output indicators, and value added of construction can
reflect the value added capability and space of industry efficiently (Reinhardt
1999).
Therefore, in the DEA model, the input index is determined in this paper for:
the number of employees, fixed assets and total power of its own machinery and
equipment at the end of the year; output index for: the construction industry output
value and added value. So DMU can be defined as 11, it represents the years of
1995–2008; the number of input indexes is 3; the number of output indexes is 2.
The data of input and output are included in Table 63.1.
63.3.2 The Calculation of the C 2 R Model
Take the input and output data from Table 63.1 into the C2 R model, we can get the
solution (Tables 63.2 and 63.3) using Maxdea5.0 software.
63 The Evaluation of China Construction Industry 615
Table 63.1 The data of input and output indexes of China construction 1995–2008
Year Input index Output index
x1 (10,000 x2 (100 millions x3 y1 y2
persons) RMB) (10,000 kWh) (100 millions (100 millions
RMB) RMB)
1995 1497.9 1850.76 7056.5 5795.73 1668.64
1996 2121.9 2685.89 9804.8 8282.25 2405.62
1997 2101.5 3083.81 8668.5 9126.48 2540.54
1998 2029.99 3380.89 8656.52 10061.99 2783.79
1999 2020.1 3752.66 9077.77 11152.86 3022.26
2000 1994.3 4204.71 9228.11 12497.60 3341.09
2001 2110.7 4951.31 10251.72 15361.56 4023.57
2002 2245.2 6183.80 11022.52 18527.18 3822.42
2003 2414.3 6548.74 11712.38 23083.87 4654.71
2004 2500.28 7148.85 14584.05 29021.45 5615.75
2005 2699.9 7621.45 13765.56 34552.10 6899.71
2006 2878.2 8395.68 14156.29 41557.16 8116.39
2007 3133.7 9175.82 15579.39 51043.71 9944.35
2008 3314.9 10258 18195.37 62036.81 11911.65
Data source Calendar year ‘‘Chinese Statistic Almanac’’ from 1999 to 2009
63.3.3 The Analysis of the Solution
Using DEA to evaluate the ability of sustainable development of China con-
struction industry, and evaluation results can be known in Table 63.3. From the
evaluating results in Table 63.3, the ability of sustainable development of China’s
construction industry of the DEA appears valid only in 2008, that is, input and
Table 63.2 The C 2 R model solution of China construction 1995–2008
P
Year h ki s
1 s
2 s
3 sþ
4 sþ
5
1995 0.776 0.140 698.620 0 2929.998 2894.678 0
1996 0.771 0.202 967.150 0 3887.901 4246.407 0
1997 0.709 0.213 783.923 0 2269.228 4104.851 0
1998 0.709 0.234 664.714 0 1885.864 4436.206 0
1999 0.694 0.254 560.000 0 1679.384 4587.307 0
2000 0.684 0.280 434.880 0 1211.137 4903.058 0
2001 0.7 0.338 357.329 0 1028.178 5593.508 0
2002 0.532 0.321 131.403 0 28.670 1380.283 0
2003 0.612 0.391 182.404 0 59.005 1158.224 0
2004 0.676 0.471 128.584 0 1287.775 225.816 0
2005 0.78 0.579 184.763 0 192.436 1382.13 0
2006 0.876 0.681 261.929 363.272 0 713.635 0
2007 0.975 0.835 287.979 382.824 0 747.244 0
2008 1 1 0 0 0 0 0
616 P.-Y. Liu et al.
Table 63.3 Results of trend of China construction industry development capacity 1995–2008
P
Year h ki K Technology efficiency Operating track
1995 0.776 0.140 0.180 Invalid –
1996 0.771 0.202 0.262 Invalid Inferior to track
1997 0.709 0.213 0.300 Invalid Inferior to track
1998 0.709 0.234 0.330 Invalid Inferior to track weakly
1999 0.694 0.254 0.366 Invalid Inferior to track
2000 0.684 0.280 0.409 Invalid Inferior to track
2001 0.7 0.338 0.483 Invalid Benign to track
2002 0.532 0.321 0.603 Invalid Inferior to track
2003 0.612 0.391 0.639 Invalid Benign to track
2004 0.676 0.471 0.697 Invalid Benign to track
2005 0.78 0.579 0.742 Invalid Benign to track
2006 0.876 0.681 0.777 Invalid Benign to track
2007 0.975 0.835 0.856 Invalid Benign to track
2008 1 1 1 Valid Benign to track
Mean 0.750 0.424 0.546 – –
output results achieve the optimal state, and both scale efficiency and technical
efficiency are effective. While, the evaluating results of the DEA is invalid in the
years from 1995 to 2007, the relative values of the efficiency fluctuate largely, and
change overall from downward to upward with a V-type curve, as shown in
Table 63.1. Before 2005, the construction industry develop slowly, and the overall
development trend appears declining. In 2005, the efficiency index is lowest, and
h = 0.528, which indicates that if the returns to scale is in a constant state, 46.8 %
of the input resources are wasted. In the 14 years, there are 7 % when the effi-
ciency of DEA is effective, there are 5 years when the comprehensive efficiency
index is less than 0.7, 6 years when the comprehensive efficiency index is between
0.7 and 0.8, 3 years when the comprehensive efficiency index is more than 0.8, the
average efficiency is 0.88 between 1995 and 2007, So we think that development
of China’s construction industry is lower (Requate 2005).
According to the results of Table 63.2, during the years when the DEA is valid
in 2008, the slack variables s þ
i and sr are both zero, which indicates that China
construction industry system runs without sink and deficit, and the input–output
has reached its optimal state; Otherwise, during the invalid years of DEA, the input
and output present with sink or deficit, which indicates that China construction
industry system runs in the invalid state.
In input, there are some years appears oversupplied, 13 years to employees and
10 years to its own machinery and equipment at the end of the year, which
indicates that the technology quality of employees and mechanization level of
China construction industry is low, the industry is still labor-intensive and rough
(Tieterberg 1994). In output, the construction industry output value appear defi-
cient, while the slack variables of added value are zero, which indicates that trends
of the added value of China construction industry products are good.
63 The Evaluation of China Construction Industry 617
Analyzing from the P technology efficiency, when the operating efficiency
appears valid in 2008, Pki \1, which indicates the technology efficiency is
effective; in other years ki ¼ 1, which indicates the technology efficiency is
invalid, and the combination of input–output elements does not reach the optimal
state and needs further improvement. The good news is, in the years of non DEA
efficiency, technical effectiveness index value shows a rising trend, rising from
0.140 in 1995 to 0.835 in 2007, up to 1 in 2008, which show that the technical
level of China construction industry is in the rising phase (World Commission of
Environment and Development 1987).
Judging from the scale efficiency, the scale efficiency index K \ 1, which
indicates the scale efficiency is invalid.
Before 2004, K \ 0.5, scale returns grow slowly, the industry’s scale efficiency
is low. After 2005, scale efficiency appears to increase, and reaches best in 2008.
On the whole, with the advancement of technological level and expansion of the
market, China construction industry returns to scale in the rising stage, even in
non- DEA efficient year, the scale efficiency appears increasing (Yu et al. 1996).
In addition, from the dynamic operation of China construction industry system
(Table 63.3), we can find that the operation does not work on the track before
2002, while, with the improvement of Construction personnel quality, technical
level and the prosperity of construction market, China construction industry works
on the track after 2002.
After projection and adjustment in the effective plane of non DEA efficiency
data in Table 63.3, adjustment results are as shown in Table 63.4. The conclusion
is as follows, the development of China construction industry is not only achieved
by investing a lot of labor resources and capital, but also consuming a lot of
energy, which can be seen from the savings of employees, fixed assets and total
power of its own machinery and equipment at the end of the year in the devel-
opment process (Koopmans 1951).
63.4 Conclusion
According to the analysis results of overall sustainable development of China
construction industry above, the paper concluded as follows:
(1) China construction industry has been in low level of development, the effi-
ciency average is only 0.75, which the DEA efficient year accounted for only
7 %, and experienced from dropping to rising and begun to develop well since
2002.
(2) In the process of the development of the construction industry, the technical
efficiency is low in most of the year, which is in the invalid state, but the
indexes of technology efficiency are always in the increasing trend. The
industrial scale is inappropriate, but what the results of the analysis found are
the scales are always in the increasing trend.
618 P.-Y. Liu et al.
Table 63.4 DEA projection and adjustment results of China construction industry development
Year The number of employees Fixed assets savings Its own machinery and equipment
savings (10,000 persons) (100 millions RMB) at the end of the year savings
1995 1033.496 413.771 4507.607
1996 1452.398 614.233 6130.150
1997 1394.491 895.964 4787.756
1998 1255.275 983.563 4404.205
1999 1179.051 1149.969 4461.185
2000 1064.492 1327.451 4124.504
2001 990.921 1486.317 4105.608
2002 1181.431 2892.032 5183.669
2003 1118.888 2540.226 4602.183
2004 937.446 2312.714 6005.839
2005 779.766 1679.601 3226.065
2006 619.411 1406.058 1758.283
2007 366.249 612.007 389.124
(3) Adjust DEA results show that investments of China construction industry in
the number of employees, the fixed assets and machinery and equipment
operating efficiency, are all in excess, leading to gross domestic product output
deficit. China construction industry is still labor-intensive and rough.
Based on the conclusion of analysis above, from the point of view of enhancing
the ability of sustainable development of China construction industry, this article
suggests that people should be focusing on optimizing employee’s structure and
quantity, improving their technical level, allocating the proportion of investment
assets rationally, increasing the operating efficiency of the machinery and equip-
ment gradually, reducing invalid investment of productive elements and resources,
and expanding the market space actively, in particular, broadening foreign con-
struction market positively.
References
Deng XH et al (2008) Management information system curriculum design and teaching reform
based on project management technology. Manage Inf 59(9):80–83
Gong ZQ, Zhang ZH (2004) Life cycle assessment and management towards sustainable
construction industry. Qinghai Univ J 47(2):24–29
Koopmans TC (1951) An analysis of production as an efficient combination of activities.
Activitiy analysis of production and allocation, vol 13. Cowles Commission for Research in
Economics, pp 39–44
Li J (2007) Research on financing barrier of retrofit for energy saving of existing buildings and
corresponding countermeasures. Archit Econ 56(12):15–17
Lu N, Cai AY (2006) Synthetic evaluation of sustainable development of construction industry in
China—data of 31 provinces (autonomous regions and municipalities) from 1999 to 2003.
Chongqing Archit Univ J 46(4):94–97
63 The Evaluation of China Construction Industry 619
Porteous C (2002) The new eco-architecture: alternatives from the modern movement. Spon
Press, London, p 121
Reinhardt FL (1999) Bringing the environment down to Earth. Harvard Bus Rev 77(4):149–157
Requate T (2005) Dynamic incentives by environmental policy instrument—a survey. Ecol Econ
54(175):195
Sheng ZH (1996) The theory, method and application of DEA (In Chinese), vol 11. Beijing
Science Press, Beijing, pp 36–47
Tieterberg T (1994) Environmental economics and policy. Harper Collines, New York
World Commission of Environment and Development (1987) Our common future. Oxford
University Press, Oxford
Xiao WP (1994) Target analysis and evaluation models on science-technic progress at the
construction enterprise. Chongqing Constr Univ J 16(1):1–8
Yu G, Wei QL, Brokett PA (1996) Generalized data envelopment analysis model. Ann Oper Res
66:47–89
Zhang YW (2010) The research on China architecture development evaluative method. Econ Res
Guide 48(7):54–56
Chapter 64
The Prediction of Customer Retention
Costs Based on Time Series Technique
Fan Yu, Ji-fang Yang and Ai-wu Cheng
Abstract Customer expenditure is one of the vital factors that impact customer
asset, the measurement and prediction of customer expenditure means a lot to the
measurement of customer asset (Chen 2006). From the perspective of customer
asset, we’d like to study the measurement of customer retention costs—the major
component part of customer expenditure. Firstly, we define the components of
customer expenditure and explain the connotation of customer retention costs;
secondly, using time series technique, we build a prediction model of retention
costs, then we predict the customer future costs on the basis of this model. Last,
this prediction model is used to the case and the results prove that this model is
effective. Besides, this model has reference value to develop the study of the
measurement of customer asset.
Keywords Customer retention costs Expenditure prediction Time series
technique
64.1 Introduction
Customer expenditure, a factor impacting customer value, has been paid attention
after the fact that customer value-oriented becomes a generally recognized mar-
keting concept. To achieve the purpose of measuring customer value accurately
and objectively, we should focus on the measurement of customer expenditure.
However, we lack the theoretical and empirical research about it currently, the
reason for which is that it’s not an easy task to predict the future costs, considering
the uncertainty of future. As a part of customer expenditure, the measurement of
future costs is very vital; it will not only impact the measurement of customer
F. Yu J. Yang A. Cheng (&)
School of Management, Xi’an Polytechnic University, Xi’an, Shanxi, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 621
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_64,
Springer-Verlag Berlin Heidelberg 2013
622 F. Yu et al.
expenditure, but also have an effect on evaluating customer value accurately, then
impacting the expenditure management and marketing decision (Ness et al. 2001).
Therefore, the purpose of this paper is to study the measurement of future costs.
Retention costs are its major expression. We will predict the customer retention
costs by time series technique and historical data, aiming to provide a suitable way
for solving problems of measurement and prediction of customer expenditure.
64.2 The Constitution of Customer Expenditure
Customer expenditure means that enterprises spend some money on attracting and
retaining customers (Liu 2003). There are two ways to classify customer expen-
diture: according to the costs spent in customer expenditure life cycle, it can be
divided into costs incurred and future costs; according to the overall process of
obtaining and retaining customers, it can be divided into customer retention costs
and acquisition costs (Zhao and He 2009). This paper integrates these two ways
under the thought of customer life cycle, and believes that retention costs and
acquisition costs constitute historical and current costs, retention costs are future
costs’ major expression. The sum of historical and current costs and future costs is
lifelong costs (Yang 2011). So it’s very easy to see that we must know acquisition
costs and retention costs if we want to measure customer expenditure. Acquisition
costs are costs incurred, which are available because we have actual data, while
retention costs are future costs, predicted by suitable way, chosen by characters
and rules. Obviously, retention costs are more difficult than acquisition costs.
64.2.1 Customer Acquisition Costs
Acquisition costs refers to operating expenditure spent on attracting customers,
including visit costs, promotional costs, cost of sales (Fei 2007).
Visiting is the basic way to promote productions in order to attract customers.
There are some virtues about this way: flexible, seasonable, bidirectional feedback.
Besides, when developing new customers, this method is very useful in products
introduction and purchasing encouragement. Repeated visiting is very necessary to
build trading relationship when competing for target customers, who have bigger
purchasing potential. Visiting costs include expenses of travel, product samples,
brochures, technical services.
Promotional costs are the customer expenditure spent on promotional activities,
trying to intrigue customers and achieving purchasing goals. When enterprises
develop new customers, it’s necessary to propagate enterprises and productions for
deepening their understandings of us. These promotions should be repeated for the
purpose of encouraging them to purchase. Therefore, these costs account for large
proportion in acquisition costs.
64 The Prediction of Customer Retention Costs 623
Costs of sales which happen in the selling process include policy costs and
service costs. Costs caused by the differences of sales policy include sales com-
mission costs, costs of push money, and costs of cash discounts and so on. Service
costs in completing purchasing include order processing costs, packaging costs,
cargo handling costs, etc.
64.2.2 Customer Retention Costs
Customer retention costs are the spending to retain existing customers. They
include customer sales service costs and customer management costs (Jiang 2006).
Customer after-sales service costs are comprised of several parts: Sales tech-
nical support, training costs, maintenance service costs and product upgrades costs.
After-Sales Service is one of the key marketing policies to improve customer
satisfaction, especially for complex products of high technicality. Whether com-
panies provide after-sales service has become a basic requirement to retain cus-
tomers. Favorable after-sales service can develop corporate image and strengthen
willingness to repeat their purchasing behaviors. But, we have to admit the fact
that it is not a small spending, which is related directly to production environment,
technology and the quality of personnel.
Customer management costs include customer account management costs and
relationship maintenance costs. Customer account management costs are accounts
receivable management costs caused by credit business. They include wages
caused by managing accounts receivable, travel expense caused by collecting
accounts receivable, bad debt losses and bad debt processing fee. This fee has been
on increase because of the sincere competition and the rise of credit business.
Relationship maintenance costs are emotional investment costs caused by
retaining purchasing relationship with existing customers; including regularly
visiting costs, presenting gifts costs, support costs of customers and long-term
relationship redeem costs, etc.
From above we can see acquisition costs are costs incurred, we have current
data, so it’s not a tough task. Retention costs are future costs, characterized by
uncertainty. They are predicted by suitable way chosen by characters and rules.
Therefore, we must know customer retention costs in order to obtain lifelong costs.
That’s why we write this paper.
64.3 The Choice of the Way to Predict Retention Costs
From above, we know many parts of retention costs and factors which impact it are
relevant with customers themselves. These factors’ varieties make customer
expenditure uncertain. So, we must consider their effects on retention costs, if we
want to establish prediction model. But, there are too many factors, it’s impossible
624 F. Yu et al.
to analyze every factor and build models. Besides, their effects have been shown in
the time series of past and current time. In a word, this paper will use time series
technique to build prediction model of customer retention costs.
The basic idea of time series technique is that we should find regularity that
appearance changes over time from historical data. Besides, we assume that this
regularity will continue to the future. We can predict the future data according to
this regularity. We use single-element time series in this paper. The purpose of
building the model is to use historical data and random errors to predict the change
of variable. Generally, we assume that random errors et at different times are
statistically independent and normally distributed random variables. There are
three steps when using time series technique: pretreatment of time series, the
establishment of model, short-term forecasts of customer costs (Wang 2008).
There are two parts of the pretreatment of time series: the judgment of sta-
tionary nature and pure randomness testing. At first, we should judge the stationary
nature of time series by drawing timing diagram. If the time series is not of
stationary nature, we should do zero mean and difference stationary processing of
series. If they are of stationary nature, we should do pure randomness testing. Pure
randomness series are also called white noise sequences; they are the series that
the past behaviors will not impact the future development. The series can be
divided into different types. Every type has a method to analyze (Guo et al. 2006).
There are three phases of the model building: order determination, parameter
estimation and adaptive test. There are three ways to do order determination: ACF
order determination, PACF order determination, residual variogram order deter-
mination method, the best criterion function order determination method. What
should be noticed is that if we adopt the first method, the order we judge is not the
certain result, the exact order can be obtained by other ways.
When we finish this step, the next is parameter estimation: build ARMA (p, q)
model by a set of sample data sequence, judge the order (p, q) and parameter.
There are three ways to estimate parameter: moment estimation, maximum like-
lihood estimation, least squares estimation. Least squares estimation is more
accurate because it uses information at utmost (Cheng and Li 2007).
After the establishment of the model, we will do adaptive test to ensure whether
we get enough information. The null hypothesis is: residuals series are white noise
sequences. If we refuse null hypothesis, which means residuals has relevant infor-
mation, fitting model is not significant. If not, that means the model is significant.
After two steps, we can apply the model to predict customer retention costs.
This paper will use SAS to build ARMA (p, q) model.
64.4 Case Study
A company is a high-tech enterprise, which produces power monitoring and
measurement system, power energy-saving technology and other commercial
electronic trading products. We select company’s cost of five customers from first
64 The Prediction of Customer Retention Costs 625
Table 64.1 The results of white noise testing
To Log v2 DF Pr [ v2 ………Autocorrelations……
6 39.50 6 \0.0001 0.702 0.373 0.294 0.362 0.265
quarter of 2001 to second quarter of 2010. We take customer A for example and
show its data processing. Other companies have the same data processing.
In the judgment of stationary nature, the results show that retention costs of this
customer are sequential; fluctuations are in smooth and have a certain periodicity.
We judge it the stationary time series.
Next, we do pure randomness testing.
The results of white noise testing shown in Table 64.1.
The results are: P values of LB test statistic are very small (\0.0001). There-
fore, we are sure that (confidence level[99.999 %) customer retention costs series
are smooth non-white noise sequences.
Thirdly, we experience order determination. We adopt ACF order determina-
tion, PACF order determination. We can see that all autocorrelation coefficients
decay to two times of the standard deviation range of fluctuations. It shows that
sequences obviously are short-term related. But the sequence of significant non-
zero correlation coefficient of attenuation for the little finger volatility process is
quite continuous and slow, the autocorrelation coefficient can be regarded as not
cut tail. In addition, it also shows that delayed two-order partial autocorrelation
coefficient are significantly greater than two times the standard deviation, the other
partial autocorrelation fluctuate in two times the standard deviation range of the
random of small value. The process of attenuating from non-zero correlation
coefficient to small value fluctuations is very prominent coefficient. Partial auto-
correlation coefficient can be regarded as two order censored. Therefore, we
et
consider using model AR (2) xt ¼ 1/ B/ B2 ; et Nð0;r2 Þ to Fit the values of the
1 2 e
observed sequences.
Through further model fitting and optimization, in an ARMA (p, q) model
where moving average delay of the order is less than or equal to five, BIC relative
minimum amount of information is an ARMA (2, 0), so we are certain that the
model is model AR (2). Then we use least squares method to do parameter esti-
mation of model AR (2). Results show that mean MU and other parameters are
significant, (t test statistic P values are less than 0.0001).
So Sample data fitted model is xt ¼ 1:16163xt1 0:66487xt2 þ et .
Next step is the adaptive test of the model. Results show that P values of LB
statistic, delayed orders are significantly greater than a. The fitted model is sig-
nificantly available.
Last, we use the model to predict the customer expenditure. Sample data is
extended to the second quarter of 2011. We can get the retention costs of third,
Fourth quarter of 2010 and first, second quarter of 2011. The retention costs of
future four quarters are 246015. Using the same way, we get the results of the
retention costs in a year: 203178, 313482, 235837 and 173109.
626 F. Yu et al.
64.5 Conclusion
As a part of customer expenditure, retention costs are of great importance. In this
paper, we use time series technique to predict the customer retention costs. We get
these conclusions: firstly, time series technique is a good way to predict future
costs; secondly, this method makes the dynamic econometrics Forecast of reten-
tion costs possible, which lays the foundation for further application.
Although, this paper makes some breakthroughs in measuring customer
expenditure, it’s still a long way to deepen the study of this field, if we want to
develop the measurement of customer asset. For example, we can consider the
prediction of retention costs from the constitution. It’s the fact that there are many
differences between every part, so it’s necessary to analyze every part, and build
prediction models. In the future, we will continue this research and make efforts to
make other breakthroughs and make contribution to the measurement of customer
asset.
Acknowledgments Foundation item: The National Natural Science Fund (70672116).
This research was supported by the National Natural Science Foundation of China under Grant
70672116.
References
Chen Y (2006) Analysis of customer value from the perspective of the cost of services. Master
thesis, NanJing Normal University, NanJing
Cheng Z-y, Li X (2007) (6) The research of customer costs in customer profitability analysis.
Coast Enterp Technol 3rd episode:25–26
Fei Y (2007) Customer profitability analysis based on ABC—the use of chemical enterprises.
Master thesis, Xi’an Polytechnic University, Xi’an
Guo C-m, Shen Y-a, Wang X-r, Gui L-j (2006) Research on the model of parametric cost
estimation based on ABC. Syst Eng Pract 26(2):55–61
Jiang Y (2006) The measurement and management of customer asset, Master dissertation, Hunan
University, Hunan
Liu X (2003) Prediction model of the value of customer asset and its use in marketing decision.
Quant Tech Econ Res 5th episode:158–161
Ness JA Schrobeck Mj, Letendre RA, Douglas WJ (2001) The role of ABM in measuring
customer value. Strateg Finance 82(9):32–37(10); 44–49
Wang Y (2008) The use of Time sequence analysis. Renmin University of China Press, Beijing
Yang J-f (2011) The measurement of customer expenditure based on the customer asset. Master
thesis, Xi’an Polytechnic University, Xi’an
Zhao Q-k, He C-c (2009) The costs of customers. Manag Aspect 6:116–117
Chapter 65
The Research on the Location Selection
of the Bank Outlets Based on Triangular
Fuzzy Analytic Hierarchy Process
Yang Han and Fa-shan Dai
Abstract This paper analyzes the influencing factors of the location of bank
outlets under the use of the triangular fuzzy analytic hierarchy process, then counts
and gets the weights of each factor. Furthermore, it shows the feasibility of this
model approach in the process of the site, in order to provide a reference for the
decision-makers.
Keywords Bank outlets Fuzzy analytic hierarchy process (FAHP) Location
Triangular fuzzy numbers
65.1 Introduction
Physical outlets play a key part in bank marketing as the most important places
where banks manage various business activities. They are the operational platform
and the extension of information antennae of banks. However, commercial bank
outlets in China which are often installed in the administrative level, lack scientific
basis. Repeatability exists in the same business circle. Similarity in products and
scales leads to low operation efficiency (Guo 2010). Nearly each bank carries out
the research on the planning and transformation of the bank outlets as the result of
the fierce horizontal competition. In this case, how to get scientific sites is full of
practical significance.
Y. Han (&) F. Dai
College of Economics and Management, China Three Gorges University,
Yichang, China
e-mail: [email protected]
F. Dai
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 627
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_65,
Ó Springer-Verlag Berlin Heidelberg 2013
628 Y. Han and F. Dai
65.2 Analysis
65.2.1 Classifications in Influence Factors of Bank Outlets
The same as the location of other establishment, various factors should be taken
into account. This paper will state it in the following five aspects.
(1) Geographical site factor: Location belongs to the geographical category. More
convenient the geographical conditions are, bank outlets are more likely to
gather together (Xu 2008). The indexes it involves are: road access, parking
numbers, bus stop numbers, numbers of public places, communities, malls and
enterprises in this area. There is feasibility to set up outlets as the business
circle is more dense.
(2) Competitors factor: Try to have a understanding of your competitor numbers
and your market share in this area. Distinction exists between service products
in different banks. Compared to hardware facilities conditions of rivals, you’re
allowed to have an explicit self-positioning, learn your own advantages and
disadvantages. If you want to gain more market and customers, you’d better
improve the core competitive power and turn to personalized services.
(3) Marketing factor: Banks of different position have different location. For
example, Bank of China position business priority in high-end customers, so
their outlets mostly dot in urban district. The Agricultural Bank focus on
agricultural support programs, so that they tend to place in suburbs. Sup-
porting costs and rents must be different in each outlet. Higher cost will reduce
the efficiency, and then influence the performance of banks. Marketing anal-
ysis mainly contains: bank positioning, own orientation, supporting costs and
rents, floor space, outdoor visual effect.
(4) Population economic factor: Potential target customers depend on total pop-
ulation and economy, so do the bank size. Meanwhile, per capita income
levels and persons flow rates in unit time also have an impact on the location.
(5) Urban development planning: As a typical factor, urban planning is an overall
arrangement which municipal government is to deploy the urban land use,
spatial distribution and various construction, according to the development
goal in a period. It contains urban reconstruction of old areas and program-
ming of new districts. In addition, urban traffic planning is also included.
65.2.2 Description of FHAP
Analytic hierarchy process (AHP) is a multi-objective decision analysis method
combined with qualitative analysis and quantitative analysis, which is put forward
by American operations research expert Satty in the 1970s (Xu 1988). Its feature is
that dividing various factors of complexed problems into interconnected orderly
hierarchy. Then experts can get judgment matrix by mutual estimating the
65 The Research on the Location Selection of the Bank Outlets 629
significance of each factor, ensure its relative importance in the hierarchical
structure. At last we can get the total order of each factor from the result of
comparative judgment. In the multi-objective decision, increase of the index mean
making up more judgment matrix with deeper levels and larger numbers
(Cao 2009). In general we use AHP’s pairwise comparison to explain its relative
importance by number one to nine. But the more the index is, the judgment of the
importance about each two indexes is more difficult (Li and Huang 2009). It even
has an effect on consistency of the orders, failing to pass the consistency check.
Because of the complexity of the objects or the one-sided understanding of the
things, it may be not necessarily reasonable to find out eigenvectors. The main idea
of AHP is to make up judgment matrix by using integer 1–9 and its reciprocal, but
it neglects people’s fuzzy judgments. It’s difficult to obtain an accurate numerical
to express comparison of the importance. When one expert regards index Ci is
important to Cj , the numerical may approximate for 2, or between 1 and 2, this is
fuzzy judgment rather than an accurate numerical. Based on it, taking fuzzy
mathematics into AHP can make the structured judgment matrix more reasonable.
Location of bank outlets is a decision-making process full of factors. Generally we
make a choice in the given area after analyzing each factor. It can reduce the
subjectivity of expert scoring by using triangular fuzzy numbers to a certain extent
(Jiang and Liu 2010). We can ascertain the impact of each factor scientifically and
systematize the location to get the best scheme according to the actual situation.
3 Specific application of FHAP in location of bank outlets
(1) Triangular fuzzy number
Definition 1 Fuzzy Set: In the domain U, for any x 2 U, x often belongs to A to a
certain extent l(l [ [0,1]), but not x 2 A or x 62 A, all the fuzzy sets are signified
with F(U).
Definition 2 Subordinate Function: Set domain U, if present lA ð xÞ: U ? [0,1],
lA ð xÞ is known as x 2 A membership, thus lA ð xÞ is generally called A’s subor-
dinate function (Tao and Zhang 2012).
Definition 3 Triangular fuzzy number: Set fuzzy number M in the domain U, if
M’s subordinate function lM :U ? [0,1] is expressed as (Yan and Zhu 2009)
8
1 l
< mx x ml x 2 ½l; m
lM ð xÞ ¼ mu 1 u
x mu x 2 ½m; u
:
0 x 2 ð1; l [ ½u; þ1Þ
In the formula l O m O u, l and u are said as M’s upper bound and lower bound
m is the mid-value of M in the membership of 1. General triangular fuzzy number
M is expressed as (l, m, u) (Chen 2004).
Operation method of two triangular fuzzy numbers M1 and M2 is:
630 Y. Han and F. Dai
M1 ¼ ðl1 ; m1 ; u1 Þ; M2 ¼ ðl2 ; m2 ; u2 Þ
M1 þ M2 ¼ ðl1 þ l2 ; m1 þ m2 ; u1 þ u2 Þ
M1 M2 ¼ ðl1 l2 ; m1 m2 ; u1 u2 Þ
1 1 1 1
; ;
M u m l
(2) Steps of FHAP
(1) Setting up the index system of the hierarchical structure model. Generally
it’s ascertained as target layer, rule layer, scheme layer. According to
actual situation, sub-rule layer can be set up under the rule layer.
(2) On the basis of expert scoring, making up fuzzy judgment matrix (Xuan
and Hua 2008). Expert gives a mark of pairwise comparison in the form of
triangular fuzzy number, uses aij showing the importance degree of index i
and j, including aij ¼ lij ; mij ; uij .
(3) Determining the initial fuzzy weights of the index Dki is said as the ele-
ment is comprehensive fuzziness in K layer, including,
!
X n Xn X n
k k k
Di ¼ aij aij ; i ¼ 1; 2; . . .; n ð65:1Þ
j¼1 i¼1 j¼1
(4) Calculating the normalized weights of the index Setting M1 ðl1 ; m1 ; u1 Þ and
M2 ðl2 ; m2 ; u2 Þ are triangular fuzzy numbers. The possibility that M1 is
more equal to M2 can be defined in triangular fuzzy number as:
vðM1 M2 Þ ¼ supx y ½minðuM1 ð xÞ; uM2 ð yÞÞ
8
>
< 1 m1 m2
ð65:2Þ
vðM1 M2 Þ ¼ lðdÞ ¼ ðm1u1l2u1Þðm2l2Þ m1 m2; u1 l2
>
:
0 otherwise
The possibility that one fuzzy number is more equal to other K fuzzy numbers
can be defined as:
d0 ðCi Þ ¼ vðM M1 ; M2 ; . . .; Mk Þ ¼ min vðM Mi Þ; i ¼ 1; 2; . . .; k ð65:3Þ
Then we can get weight vectors of all rules:
T
W 0 ¼ ½d0 ðC1 Þ; d0 ðC2 Þ; . . .; d0 ðCn Þ ð65:4Þ
After the normalized processing, we can get normalized weights of each rule:
W ¼ ½d ðC1 Þ; dðC2 Þ; . . .; d ðCn ÞT ð65:5Þ
65 The Research on the Location Selection of the Bank Outlets 631
Table 65.1 Factors hierarchy chart of locatiom
Target Rule Sub-rule
Location of bank outlets Urban development planning Urban traffic planning C18
A B5 Urban reconstruction of old areas and
programming of new districts C17
Population economic factor Persons flow rates in unit time C16
B4 Per capita income levels C15
Economy C14
Population C13
Marketing factor B3 Outdoor visual effect C12
Floor space of outlets C11
Supporting costs and rents C10
Bank positioning and own orientation
C9
Competitors factor B2 Hardware facilities condition C8
Type of service products C7
Competitor numbers C6
Geographical site factor B1 Numbers of malls and enterprises C5
Bus stop numbers C4
Outlets towards the entrance C3
Parking numbers C2
Road access C1
(3) Application of the model in the location
(1) Taking the location of bank outlets as the target, we classify influential
factors into different layers, then subdivide it to establish corresponding
level structure model (as shown in Table 65.1).
(2) Designing the questionnaires combined with all factors and distributing
them to experts who are from Construction Bank, Postal savings Bank and
Rural commercial Bank for scoring. We can get three sets of data from
several experts’ scores by doing arithmetic average. Thus we could gain
Fuzzy judgment matrix data about target layer and rule layer, also rule
layer and sub-rule layer included. The following data is triangular fuzzy
number, using nine scale method of AHP (Orlovsky 1986). The middle
numerical is the most likely value. The former and later numbers are
respectively ceiling and floor. This paper takes target layer A and rule
layer B as an example to state and calculate (such as in Table 65.2). The B
layer and C layer are not listed but only given the final score results.
(3) Finding out comprehensive fuzzy triangle matrix of each layer. According
to Table 65.1, doing arithmetic average operations again on the basis of
the former experts’ scores. We can get comprehensive fuzzy value of each
factor in rule layer, then comprehensive fuzzy triangle matrix is offered
(Ji et al. 2007) (such as in Table 65.3).
632 Y. Han and F. Dai
Table 65.2 A–B fuzzy evaluation matrix
A–B B1 B2 B3 B4 B5
B1 (1 1 1) (1/3 1/2 3/4) (3/2 2 5/2) (3/4 1 3/2) (4/3 2 9/4)
(1 1 1) (1/5 1/3 1/2) (1 3/2 2) (1 2 5/2) (1 2 5/2)
(1 1 1) (2/3 1 3/2) (2/3 1 3/2) (1/4 2/7 1/3) (4/5 1 3/2)
B2 (4/3 2 3) (1 1 1) (3 4 9/2) (3/2 2 3) (3 7/2 4)
(2 3 5) (1 1 1) (2 3 4) (1 3/2 2) (2 3 4)
(2/3 1 3/2) (1 1 1) (1/2 1 7/5) (2/9 1/4 1/2) (1 3/2 2)
B3 (2/5 1/2 2/3) (2/9 1/4 1/3) (1 1 1) (1/4 1/2 2/3) (4/5 1 4/3)
(1/2 2/3 1) (1/5 1/4 2/7) (1 1 1) (5/2 3 7/2) (1 3/2 2)
(2/3 1 3/2) (5/7 1 2) (1 1 1) (2/9 1/4 1/2) (1 7/4 2)
B4 (2/3 1 4/3) (1/3 1/2 2/3) (3/2 2 4) (1 1 1) (8/5 2 3)
(2/5 1/2 1) (1/4 1/3 1/2) (2/7 1/3 2/5) (1 1 1) (1/2 1 6/5)
(3 7/2 4) (2 4 9/2) (2 4 9/2) (1 1 1) (1 3/2 2)
B5 (4/9 1/2 3/4) (1/4 2/7 1/3) (3/4 1 5/4) (1/3 1/2 5/8) (1 1 1)
(2/5 1/2 1) (1/4 1/3 1/2) (1/2 2/3 1) (5/6 1 2) (1 1 1)
(2/3 1 5/4) (1/2 2/3 1) (1/2 4/7 1) (1/2 2/3 1) (1 1 1)
Table 65.3 A–B Comprehensive fuzzy triangle matrix
A–B B1 B2 B3 B4 B5
B1 (1 1 1) (0.4 0.611 (1.056 1.5 2) (0.667 1.095 (3 4.333
0.917) 1.167) 5.667)
B2 (1.333 2 3.167) (1 1 1) (2.333 3 3.967) (1.241 1.75 (2.833 3.667
2.444) 5)
B3 (0.522 0.722 (0.368 1.5 (1 1 1) (0.991 1.25 (1.533 2.333
1.056) 0.873) 1.556) 3)
B4 (1.356 1.667 (1.194 1.611 (1.262 2.111 (1 1 1) (1.533 2.333
2.111) 1.889) 2.967) 3)
B5 (0.187 0.244 (0.2 0.278 (0.444 0.579 1) (0.361 0.484 (1 1 1)
0.361) 0.373) 0.708)
X
5 X
5
aij ¼ð1 1 1Þ þ ð0:4 0:611 0:917Þ þ þ ð0:556 0:722 1:208Þ
i¼1 j¼1
þ ð1 1 1Þ ¼ ð23:553 31:799 41:276Þ
X
5
aij ¼ð1 1 1Þ þ ð0:4 0:611 0:917Þ þ ð1:056 1:5 2Þ
j¼1
þ ð0:667 1:095 1:444Þ þ ð1:044 1:667 2:083Þ
¼ð4:167 5:873 7:444Þ
From the above formula (65.1)
65 The Research on the Location Selection of the Bank Outlets 633
!1
X
5 5 X
X 5
DB1 ¼ aij aij ¼ ð0:1010 0:1847 0:3161Þ
j¼1 i¼1 j¼1
Similarly we can get:
!1
X
5 5 X
X 5
DB2 ¼ aij aij ¼ ð0:1714 0:3014 0:5364Þ
j¼1 i¼1 j¼1
!1
X
5 5 X
X 5
DB3 ¼ aij aij ¼ ð0:0927 0:1537 0:2659Þ
j¼1 i¼1 j¼1
!1
X
5 5 X
X 5
DB4 ¼ aij aij ¼ ð0:1335 0:2481 0:4260Þ
j¼1 i¼1 j¼1
!1
X
5 5 X
X 5
DB5 ¼ aij aij ¼ ð0:0721 0:1121 0:2081Þ
j¼1 i¼1 j¼1
Then to blur, according to the above formula 65.2–65.4, we can get:
0:1714 0:3161
vðDB1 DB2 Þ ¼ ¼ 0:5536
ð0:1874 0:3161Þ ð0:3014 0:1714Þ
vðDB1 DB3 Þ ¼ 1
0:1335 0:3161
vðDB1 DB4 Þ ¼ ¼ 0:7423
ð0:1874 0:3161Þ ð0:2481 0:1335Þ
vðDB1 DB5 Þ ¼ 1
d0 ðB1Þ ¼ min vðDB1 DB2 ; DB3 ; DB4 ; DB5 Þ
¼ minð0:5536 1 0:7423 1Þ ¼ 0:5536
Similarly we can get:
d0 ðB2 Þ ¼ 1 d0 ðB3 Þ ¼ 0:3902
d0 ðB4 Þ ¼ 0:8269 d0 ðB5 Þ ¼ 0:1611
By calculation, we can get:
W 0 ¼ ð0:5536 1 0:3902 0:8269 0:1611Þ
After the normalized, we can get:
W ¼ ð0:1888 0:3411 0:1331 0:2820 0:0549Þ
634 Y. Han and F. Dai
Table 65.4 Index total weight
Index of Weight of the index Index of Weight of the index Weight of the total
layer B of layer B layer C of layer C ranking of layer C
B1 0.19 C1 0.30 0.057
C2 0.14 0.030
C3 0.17 0.032
C4 0.19 0.036
C5 0.20 0.038
B2 0.34 C6 0.42 0.143
C7 0.33 0.112
C8 0.25 0.085
B3 0.13 C9 0.34 0.044
C10 0.26 0.034
C11 0.23 0.030
C12 0.17 0.022
B4 0.28 C13 0.29 0.081
C14 0.27 0.076
C15 0.22 0.062
C16 0.22 0.062
B5 0.06 C17 0.46 0.028
C18 0.54 0.032
The same method can be obtained the weights that the first class index relative
to the second class index, just like Table 65.4 shows, based on this we can gain the
total order that the sub-rule layer C relative to the target layer A.
From Table 65.4, we know that competitors factor play an important part in the
location, while the influence of urban development planning is relatively minimal.
So numbers of competitor in this area should be given priority when a new site is
set up (Li et al. 2005). Try to learn the diversity of services and products between
rivals from market research. If you want more market share, innovation is indis-
pensable. At the same time, population and economy should be taken into con-
sideration. It’s necessary to expand service radium and improve the service level.
65.3 Summary
In the hierarchical analysis of general problem, structured judgment matrix
neglects people’s fuzzy judgment, only including two possible extreme conditions:
to choose an index in a membership of 1, meanwhile, to negate other scale value
with the same membership (or other, to choose an index in a membership of 0)
(Liu and Fan 2005). Sometimes experts may give some fuzzy quantity when to be
consulted. It seems necessary to bring in fuzzy numbers to improve analysis.
FHAP is the combination of AHP and the fuzzy comprehensive evaluation
method. Evaluation based on it contributes to scientific and effective decisions.
65 The Research on the Location Selection of the Bank Outlets 635
Due to the limitation of scoring, the method is general applied in particular area
(Fan et al. 2005). It cannot be denied that it has reference value by eliminating
personal subjective judgment. Also, it has good evaluation in a given scheme to
choose the best.
References
Cao Y (2009) Location decision analysis of chain stores based on the fuzzy analytic hierarchy
process. Bus Perspect 594:48–49
Chen X (2004) Application of fuzzy hierarchy process for optimum selection in decision-making.
Comput Eng Des 10:1847–1849
Fan LF, Jiang HB, Chen KS (2005) The application of fuzzy analytic hierarchy process in the
location selection of distribution centers. Mod Logist 11:15–17
Guo XP (2010) The research on influential factors about layout of commercial bank outlets.
Anhui University, Hefei
Ji D, Songg B, Yu TX (2007) The method of decision-making based on fuzzy analytic hierarchy
process and its application. Fire Control Command Control 11:38–41
Jiang Y, Liu D (2010) Evaluation of industrial cluster comprehensive performance based on
fuzzy analytic hierarchy process. Stat Decis 02:31–33
Li B, Huang S (2009) Application of method for evaluating in quality assurance system of MMS.
Tech Econ 28:50–53
Li Y, Hu XH, Qiao J. (2005) An improved fuzzy analytic hierarchy process method. J Northwest
Univ 01:11–16
Liu LJ, Fan RG (2005) The application to supplier partner selection of the fuzzy analytic
hierarchy process based on the triangular fuzzy numbers comparative theory. Logist Sci
Technol 127:117–121
Orlovsky SA (1986) Decision-making with a fuzzy preference relation. Fuzzy Sets Syst
18:105–120
Tao C, Zhang H (2012) Risk assessment for the third-party damage to pipeline based on fuzzy
hierarchy process. Gas storage Transp 31:99–102
Xu S (1988) Principle of analytic hierarchy process-practical decision-making and methods.
Tianjin University Press, Tianjin
Xu F (2008) Empirical research on the location of bank outlets. Zhejiang University of
Technology, Hangzhou
Xuan Z, Hua L (2008) The evaluation of the port site selection based on fuzzy hierarchy process.
Chin Water Transp 12:68–70
Yan T, Zhu R (2009) The research of universities’ financial risk identification based on fuzzy
analytic hierarchy process. Stud Finan Account Educ 03:26–30
Chapter 66
The Study of Sino-Russian Trade
Forecasting Based on the Improved Grey
Prediction Model
Zhen-zhong Zhang, Shuang Liu and Li-xia Tian
Abstract In this paper, we improved the traditional GM (1,1) model with the
other-dimensional gray-scale by-ways, which has a higher accuracy, and predicted
the Sino-Russian future trade. First of all, we introduced the theory of GM (1,1)
grey and GM (1,1) grey equidimensional filling vacancies. Secondly, we estab-
lished GM (1,1) grey forecasting model of equidimensional filling vacancies by
using the trade volume between China and Russia from 2000 to 2011. Then, we
forecasted the Sino-Russian trade in 2012. At the end of the paper, we analyzed the
forecast results, and we found that Sino Russian trade still has very large devel-
opment space.
Keywords GM (1, 1) Grey forecasting model of equidimensional filling
vacancies Grey theory Sino-Russian trade forecasting
66.1 Introduction
Along with the deepening of world economy integration, the economic connection
between each country is increasingly close. China is the most populous country in
the world. Russia, China’s largest neighbors, is the world’s largest country. They
have a more than 4300 km common boundary line and a long trading history. At
present, both of China and Russia are permanent members of the UN Security
Z. Zhang (&) S. Liu L. Tian
Department of Economy and Management, North China Electric
Power University, Baoding, China
e-mail: [email protected]
S. Liu
Department of Quality Technology Supervision, Hebei University,
Baoding, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 637
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_66,
Ó Springer-Verlag Berlin Heidelberg 2013
638 Z. Zhang et al.
Fig. 66.1 Volume of trade between China and Russia from 1994 to 2010
Council, members of WTO, and play a significant role in international politics and
economic affairs.
From Fig. 66.1, we can see, Sino-Russian trade history is a long and tortuous
development process. For example, from 1999 to 2008, because of the reforms of
Russia’s foreign trade policy, the rapid growth of Chinese economy and other
reasons, Volume of trade between China and Russia grew stably and rapidly, while
the Sino-Russian trade sharp declined under the influence of the global financial
crisis in 2008 (Du 2011; Ren and Wang 2011; Zhang and Liang 2011; Ma et al.
2006; Li et al. 2012; Chen 2009; Zhao 2010).
From the above situation, we can see, the foreign trade volume growth presents
a certain degree of volatility and uncertainty, because of the impact of trade policy,
international market demand, emergencies and many uncertainties, and so on,
which has increased difficulty to the accurate prediction of trade volume. However,
accurate prediction of volume of foreign trade has important significance in pro-
moting the national economy stable and sustained growth, and in developing
reasonable and effective foreign trade policy (Deng 2005; Li 2008; Chu 2011;
Wang et al. 2008).
Because the volume of foreign trade will be influenced by many uncertain
factors, some scholars use the traditional GM (1,1) model for foreign trade volume
prediction. Because of its small sample size, the higher prediction accuracy, the
gray prediction technology has been widely used. While, GM (1,1) models has its
limitations as other forecasting methods. For example, when the data with a
greater degree of dispersion, that is to say, the greater the gray scale data it is, the
prediction accuracy worse; and it is not suitable for predicting long-term forecast
for several years after the push (Zhou and Jiang 2004; Zhou and Zhou 2011; Wang
et al. 2009; Zhang et al. 2009; Niu et al. 2006). In this paper, we improved the
traditional GM (1,1) model with the other-dimensional gray-scale by-ways, and
predicted the Sino-Russian future trade.
66 The Study of Sino-Russian Trade Forecasting 639
66.2 Introduction of Grey Forecasting Model
of Equidimensional Filling Vacancies GM (1, 1)
66.2.1 GM (1, 1) Model
In 1982, Professor Deng Ju-long, a famous scholar in China, proposed and
developed Grey system. When modeling the gray model of Sino-Russian trade
forecasting, we usually use historical data to establish differential equations, which
is used as the Sino-Russian trade forecasting models. Because there are many
uncertain factors affect Sino-Russian trade, which is called gray system, the vol-
ume of trade between China and Russia is no rule. With the help of Grey theory,
these seemingly irregular historical data, after generated by the cumulative,
showed a clear exponential growth law compared with the original value. At the
same time, differential equations have the same form of exponential form. That is
to say, the prediction process can be divided into three steps. Firstly, fit expo-
nential growth to generate regular data column by using differential equations.
Secondly, conducting the Sino-Russian trade forecast. Thirdly, generate the actual
Sino-Russian trade forecasts with the help of regressive reduction method. GM
(1,1) model is one of the most simple models in gray system theory. The modeling
process is as follows:
Recorded the Sino-Russian trade raw data in previous years is:
xð0Þ ¼ fxð0Þ ð1Þ; xð0Þ ð2Þ; . . .; xð0Þ ðnÞg ð66:1Þ
The result of 1-AGO is:
xð1Þ ¼ fxð1Þ ð1Þ; xð1Þ ð2Þ; . . .; xð1Þ ðnÞg ð66:2Þ
Among of them,
X
k
xð1Þ ðkÞ ¼ xð1Þ ðiÞ ðk ¼ 1; 2; . . .; nÞ ð66:3Þ
i¼1
The sequence xð1Þ ðkÞ has the law of exponential growth, so we generally believe
the sequence xð1Þ ðkÞ meets the exponential growth in the form of general solutions
of first-order linear differential equation:
dxð1Þ
þ dxð1Þ ¼ u ð66:4Þ
dt
Under normal circumstances, the Sino-Russian trade data we get is discrete
data, while the index should be a continuous type of equation. At this time, the
general approach is: using xð0Þ ðk þ 1Þ to represent the differential term of the
discrete form. Here, xð1Þ take the average load of k and k ? 1, namely:
640 Z. Zhang et al.
1 h ð1Þ i
xð1Þ ¼ x ðkÞ þ xð1Þ ðk þ 1Þ ð66:5Þ
2
Therefore, the equation is transformed into:
1 h i
xð0Þ ðk þ 1Þ þ a xð1Þ ðkÞ þ xð1Þ ðk þ 1Þ ¼ u ð66:6Þ
2
The results will be written in matrix form as follows:
2 ð0Þ 3 2 3
x ð2Þ 12 ½xð1Þ ð1Þ þ xð1Þ ð2Þ 1
6 xð0Þ ð3Þ 7 6 1 ½xð1Þ ð2Þ þ xð1Þ ð3Þ 1 7
6 7 6 7 a
4 ... 5 ¼ 4 ð66:7Þ
2
... ...5 u
xð0Þ ðnÞ 12 ½xð1Þ ðn 1Þ þ xð1Þ ðnÞ 1
2 ð0Þ 3
x ð2Þ
6 xð0Þ ð3Þ 7 a ,
Here, we order Yn ¼ 4 6 7 ,A¼
... 5 u
xð0Þ ðnÞ
2 3
12 xð1Þ ð1Þ þ xð1Þ ð2Þ 1
6 1 xð1Þ ð2Þ þ xð1Þ ð3Þ 1 7
B¼6
4
2 7, then: Yn ¼ BA.
... ...5
12 xð1Þ ðn 1Þ þ xð1Þ ðnÞ 1
" #
^
By solving the equation, we know A ¼ ðBT BÞ B Yn ¼ a^ , then put
1 T
u
parameters back to the original equation, we know:
" ^
# ^
ð1Þ ð1Þ u ^
aðkþ1Þ u
b
x ðk þ 1Þ ¼ x ð1Þ ^ e þ ^ ðk ¼ 0; 1; 2; . . .Þ ð66:8Þ
a a
After 1-IAGO, we can get the discrete form:
bx ð0Þ ðk þ 1Þ ¼bx ð1Þ ðk þ 1Þ bx ð1Þ ðkÞ
" ^
#
^
a ð0Þ u a^ k ð66:9Þ
¼ð1 e Þ b x ð1Þ ^ e ðk ¼ 0; 1; 2; . . .Þ
a
Equations (66.8) and (66.9) are the time response function model of the GM
(1,1) model. Among of them, when k ¼ 0; 1; . . .; n 1, bx ð0Þ ðk þ 1Þ is the fitted
x ð0Þ ðkÞ ðk ¼ 1; 2; . . .; nÞ. While, b
values of the original data sequence b x ð0Þ ðk þ 1Þ
ð0Þ
is the predictive value of the original data sequence b x ðkÞ ðk ¼ 1; 2; . . .; nÞ,
when k n.
66 The Study of Sino-Russian Trade Forecasting 641
66.2.2 Grey Forecasting Model of Equidimensional Filling
Vacancies GM(1, 1)
Because of the first-order differential equation, which is used in Gray GM (1, 1)
model, is exponential, GM (1,1) model is applied to a strong exponential Sino-
Russian trade forecasting. But it requires the data is equidistant, adjacent, no
jumping, and requires using the latest data as a reference point, the earliest data is
dispensable, but the latest data have to be added. For these reasons, the application
of Gray GM (1, 1) model is limited. So in this paper, the author would improve it.
The essence of Grey forecasting model of equidimensional filling vacancies
GM(1,1)is to get each new forecast data into the original data, while removing one
of the earliest data, thereby maintaining the same number of data. Then, use the
sample sequence with the times to rebuild gray GM (1,1) model to predict the next
value. Repeat the above process, forecast one by one, elected one by one. When
using it to predict, it can replenish the use of new information and increase degree
of gray plane albino. Finally, Sino-Russian trade forecasting accuracy will be
improved significantly. The modeling process is as follows:
First of all, handling the data respectively by 1-AGO and 1-IAGO as the
traditional GM (1,1) model, then you can get the corresponding time series
xð0Þ ðk þ 1Þ þ 12 a[xð1Þ ðk) þ xð1Þ ðk þ 1Þ ¼ u of GM (1,1) model:
x ð0Þ ðk þ 1Þ ¼ b
b x ð1Þ ðk þ 1Þ b x ð1Þ ðkÞ
" ^
#
^
a ð0Þ u ^a k
¼ ð1 e Þ b x ð1Þ ^ e
a
ðk ¼ 0; 1; 2; . . .n 1Þ
The dynamic process of equidimensional filling vacancies on the original
sequence: Removexð0Þ ð1Þ, add b
x ð1Þ ðn þ 1Þ. Thus, the original data sequence turns
into:
n o
xð0Þ ¼ xð0Þ ð2Þ; xð0Þ ð3Þ; . . .; xð0Þ ðnÞ; b
x ð1Þ ðn þ 1Þ
Based on this adjusted the original data sequence, re-use the traditional GM
(1, 1) model to predict the next value.
Finally, repeat the above steps until get the final demand forecast results.
66.2.3 Sino-Russian Trade Forecasting
Sino-Russian trade has experienced a long and tortuous development history. In
the modeling process, we selected Sino-Russian trade, from 2000 to 2010, as the
raw data, and the 2011 Sino-Russian trade data as the year of testing the merits of
642 Z. Zhang et al.
Table 66.1 The Sino-Russian trade from 2000 to 2011 (billion dollars)
Years 2000 2001 2002 2003 2004 2005
Volume of trade 80.03 106.71 119.27 157.58 212.26 291.01
Year 2006 2007 2008 2009 2010 2011
Volume of trade 333.87 481.55 569.09 387.52 555.33 835
the standard model. Then, we forecast the Sino-Russian trade in 2012. The Sino-
Russian trade from 2000 to 2011 as follows (Table 66.1):
After using Matlab to get the original series and after series of accumulated
generating trends, the exponential growth trend can be found. That is to say, we
can use GM (1, 1) to predict (Fig. 66.2).
With the help of Matlab, the result could be got, as follows:
" #
^
a 0:1539
A¼ ^ ¼
u 1296409:5357
Therefore:
x ð1Þ ðk þ 1Þ ¼ 9224037:68e0:1539k 8423713:68
b
From 1-IAGO to get tired by the gray prediction model is:
x ð0Þ ðk þ 1Þ ¼ 9224037:68 1 e0:1539 e0:1539k
b
ðk ¼ 0; 1; 2. . .Þ
With the help of traditional GM (1,1) model and improved gray GM (1,1)
model, we forecast the Sino-Russian trade in 2012, the results are as follows
(Table 66.2):
Fig. 66.2 The original series and additive series trends
66 The Study of Sino-Russian Trade Forecasting 643
Table 66.2 Compare the predictions between traditional GM (1, 1) model and improved model
Traditional grey Improved grey model
Years Actual value Predictive value Residuals Predictive value Residuals
2011 835 833.78 1.22 834.50 0.5
Table 66.3 Accuracy Accuracy class p c
assessment model
First grade:Good [0.95 \0.35
Second grade:Qualified [0.80 \0.5
Third grade:Reluctantly [0.70 \0.65
Fourth grade:Failure \= 0.70 [ = 0.65
Table 66.4 The posterior p c
margin of the two models
Traditional GM (1,1) model 1.61 0.067
Improved GM (1,1) model 2.28 0.039
There are two indicators about difference test after test: poor ratio of posterior
‘‘c’’ and small error probability ‘‘p’’. ‘‘c’’ is smaller, the model is better; ‘‘p’’ is
greater, the model is better (Tables 66.3, 66.4).
It can be seen from the result, the accuracy of the model no matter improved or
not is good. However, the accuracy of the improved model was better than before
the improvements. In other words, improved gray model has a higher extrapolation
in Sino-Russian trade forecasting.
66.3 Conclusion
Because the foreign trade volume between the two countries is affected by the
country’s economic conditions, trade policy, international market demand, unex-
pected events and many uncertain factors, which belongs to the gray areas of the
system. Because of foreign trade volume prediction in gray scale is too large, the
traditional GM (1, 1) model to forecast precision is reduced which results in not
applicable. In this paper, we improved the traditional GM (1, 1) model, and used the
new model predicted Sino-Russian trade. At the end of the paper, we found that the
improved model has much higher accuracy than the traditional one, through
comparing the predict results which were got by using the traditional one and
the improved one. However, the greater the need to predict the amount, the greater
the computation it is. For this, the model still needs to improve again.
644 Z. Zhang et al.
References
Chen D (2009) On the influences of the international financial crisis on china’s foreign trade and
countermeasure. J Hubei Polytech 5(4):67–72
Chu X (2011) A forecast of foreign trade of Beijing based on the gray system theory. China Bus
Market 2011(5):54–58
Deng J (2005) Basic methods of gray system. Hua-Zhong University of Science and Technology
Press, Wu Han, pp 60–70
Du Y (2011) China’s foreign trade development status and countermeasures. China Bus Trade
2011(5):201–202
Li S (2008) A trend forecast of China import and export trade total volume based on the gray
system model. Commer Res 2008(3):113–115
Li L, Dunford M, Yeung G (2012) International trade and industrial dynamics: geographical and
structural dimensions of Chinese and Sino-EU merchandise trade. 32(1):130–142
Ma T, Li B, Fang C, Zhao B, Luo Y, Chen J (2006) Analysis of physical flows in primary
commodity trade: a case study in China. Resour Conserv Recycl 47(1):73–81
Niu D, Zhang B, Chen L, Zhang T (2006) Application of intelligent optimization grey model in
middle-term electricity demand forecasting. East China Electr Power 1(1):8–11
Ren T, Wang Y (2011) Present situation and prospect of Sino-Russian economic and trade
relations. China Econ Trade Herald 3:64–65
Wang Y, Sun L, Xu C (2008) The dynamic analysis of export structure versus china trade
competitiveness index. Oper Res Manag Sci 17(2):115–120
Wang Z, Dang Y, Liu S, Lian Z (2009) Solution of GM (1, 1) power model and its properties.
Syst Eng Electron 10(10):2380–2383
Zhang J, Liang S (2011) China’s foreign trade development trends and policy measures in post-
crisis era. J Yunnan Univ Finance Econ 6:43–48
Zhang L, Ji P, Du A, He Q (2009) Comparison and application of several grey-forecasting models
to mid-long term power load forecasting. J China Three Gorges Univ (Natural Sciences)
3(6):41–45
Zhao M (2010) China-Russia relations enter a new period of historical development. Russian
Cent Asian East Eur Stud 1:62–67, 96
Zhou J, Jiang Z (2004) China’s exports forecast based on gray system model of GM (1.1). J Int
Trade 2004(2):27–29, 39
Zhou Z, Zhou F (2011) The application of grey model of equidimensional filling vacancies in
forecasting GDP. J Huanggang Norm Univ 31(6):26–28
Chapter 67
The Role of Preference and Emotion
in Environmental Risk Perception
Charlene Xie, Yang Liu, Shengxiang She and Dixi Song
Abstract Environmental risks are becoming increasingly frequent and severe
across the world, especially in China. An in-depth understanding of how public
perceives risk is of crucial importance to effective environmental risk communi-
cation and management. Risk preference and emotion are two critical factors in
environmental risk perception. This paper summarizes existing researches in risk
preference and emotion, and reflects upon their roles in environmental risk per-
ception. Based upon existing literatures, it argues that delay affects environmental
risk perception more than time preference and proposes a research on disentan-
gling time preference and risk preference, i.e. risk preference at different time
periods. Under the guidance of Appraisal Theory, this paper explores the roles of
emotion in environmental risk perception and attempts to combine it into the
future examining of delay effect on risk preference.
Keywords Environmental risk perception Risk preference Time dealy effect
Emotion Apprial theory
67.1 Introduction
The recent years has seen the sharp increase in environmental risk frequency and
severity, causing casualty and financial loss. The petroleum leakage accident in the
Bohai Sea and the cadmium pollution in Guangxi are typical examples of recent
C. Xie Y. Liu D. Song
Shenzhen Graduate School, Harbin Institute of Technology,
Shenzhen, China
S. She (&)
School of Management, Guilin University of Technology,
Guilin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 645
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_67,
Ó Springer-Verlag Berlin Heidelberg 2013
646 C. Xie et al.
environmental risk events in China. Environmental risk management has become
an important task for central and local governments in China.
When reviewing upon public reactions in these environmental risk events, we
find that common people typically lack objective judgment and usually overreact
in panic. However, people are apparently ignoring environmental risks like air
pollution, climate change, and soil contamination in daily life. Such an obvious
contrast cannot be fully explained by existent theories in decision research, psy-
chology and risk management. According to the results of latest researches in this
field, risk preference and emotion are two important underlying factors in such an
obvious contrast (Scherer 2009; Keller et al. 2012).
Traditional rationality paradigm in decision theory attributes people’s responses
to risks to the probability and severity. For long-term risks like environmental risks
which have noticeable time delays, scholars combine time preference into the
decision process. As for the relationship between time preference and risk pref-
erence (manifested by subjective probability), most researches have treated them
as independent from each other. Some scholars even simply equal subjective
probability to statistical probability, absolutely ignoring the non-linearity of peo-
ple’s real risk preference.
People have long felt that their instant emotion holds a great power on their
perception of uncertainty, environmental risks included. Researches into emotion
and risk perception in the past few decades have shown that the role of emotion in
risk perception is more important than previously thought. Advancement in
appraisal theory has brought many new and more convincing insights into how
emotion is elicited, combined with appraisal, and why people’s emotional
responses vary so much for the same environmental risk.
To research further into the relationship between time and risk preference and
combine emotion into it is of great importance to better understanding of envi-
ronmental risks. This paper will review researches in risk preference and emotion
in environmental risk perception. It will proceed as follows. Section 67.2 will
reflect on research development in the entangling relationship between risk pref-
erence and time preference, emphasizing its influence on environmental risk
perception. Section 67.3 will summarize theories and findings concerning emo-
tion, its effect upon environmental risk perception in particular. Section 67.4 will
discuss our research plan to explore the time delay effect on risk preference,
especially its influence on environmental risk perception, and its combination with
emotion in specific environmental risks like nuclear leakage, and air pollution.
Section 67.5 will discuss our expected results, and the problems we meet in design
our research plan. Also directions for future researches are shown in this section.
67.2 Risk Preference in Environmental Risk Perception
Environmental risks are typically intertemporal risks, with delayed consequences
and high uncertainty (Gattig and Hendrickx 2007). Perception of environmental
67 The Role of Preference and Emotion 647
risks involves both objective factors and subjective factors. Risk preference is one
of the two main subjective factors in environmental risk perception, together with
time preference. Although it is generally accepted that people can be risk-averse,
risk-neutral, or risk-seeking, it is more practical to assume all human beings are
averse towards environmental risks, since they cannot get any benefit from envi-
ronmental risks. Also, it is reasonable to assume that all people require compen-
sation for delayed benefit, i.e. people are averse to time delay.
Risk preference manifests people’s attitudes towards risks. It is usually mea-
sured via the comparison between expected value and subjective certainty
equivalent. When the subjective certainty equivalent of a risky option is equal to
its expected value, the decision maker is risk-neural; when lower than expected
value, he is risk-averse; and when greater than expected value, he is risk-seeking.
Extensive explorations have been conducted in this field. The first and perhaps
most influential result is the expected utility theory. This model received wide-
spread support and became the basis for decision under uncertainty in several
decades, due to clarity and simplicity in both logic and form. While in financial
field Markowitz proposed his risk–return model to explain the St. Petersburg
Paradox. However, the findings of more and more anomalies, like framing effect,
challenged the two theories. It was under this situation that prospect theory was
proposed.
Time preference shows people’s discounting of future gain or loss. Extensive
researches have been conducted in time preference, bringing forth multiple the-
ories and findings. The Discounted Utility (DU) model is the first widely accepted
model, in which future utility is manifested by utility discounting factor. Later
researches found many anomalies, greatly undermining the validity of the DU
model. To better explain people’s real activities under time delay researchers
proposed models like hyperbolic discounting (Pender 1996), and hyperboloid
discounting (Green et al. 1997).
For intertemporal risks, discounted expected utility has long been the dominant
model to explain people’s attitudes towards both time delay and uncertainty at the
same time. However, its underlying assumption that people are risk neutral
towards gains and losses is under criticize.
Although both risk preference and time preference in intertemporal risks have
been researched in detail, few have focused on the relationship between them,
simply treating them as orthogonal dimensions. Recently, some studies have
noticed that time delay can affect subjective probability judgment (Epper et al.
2009). Some researchers treat time delay as an implicit risk, arguing that time
delay increases aversion towards risk (Baucells and Heukamp 2010). Some other
researchers use impatience to argue that time delay makes people more risk-
averse. While Construal Level theory (Liberman and Trope 2008) suggests that
time delay reduces risk aversion. Latest behavioral experiments show that time
delay makes people more risk tolerant. Anyhow, researches in neuroscience
(Loewenstein 2001) and biology (Boyer 2008) strongly support that time has an
impact on risk preference.
648 C. Xie et al.
Risk preference greatly affects people’s perception of environmental risks. Risk
preference changes people’s subjective probability of a risk event. For example,
higher risk aversion increases people’s subjective probability of environmental
risk. Studies have found that people with different risk preference levels exhibit
greatly different attitudes towards environmental risks and relevant policies and
measures (Rundmo and Moen 2006). In a cross-cultural research as for environ-
mental risk perception, Duan and Fortner found that Chinese people are more risk
averse than Americans, and consequently more concerned with environmental
risks (Hongxia and Rosanne 2010).
Combing the fact that time delay affects both time preference and risk pref-
erence, we can logically deduce that time delay affects people’s perception of
long-term risks like environmental ones more than time preference does. Our
finding can improve our understanding of environmental risk perception in a more
precise way. This new finding can better explain and predict people’s environ-
mental risk related behavior.
67.3 Emotion in Environmental Risk Perception
In the traditional rationality paradigm, emotion was treated as an external inter-
ference in risk perception, like many other subjective factors. With the psycho-
logical researches in environmental risk field, it has been found and widely
accepted that environmental risk perception is closely related to individual traits,
besides that objective condition of environmental risks (Vastfjall et al. 2008). In
fact, many studies have shown that emotion is very important in environmental
risk perception, more than only being external interference. The newly proposed
Appraisal Theory has attempted to combine both cognition and affection into
environmental risk perception, in favor of a dual-process model.
Appraisal theory has found that feelings involved in environmental risks are
more than simply bad or good affect (Peters et al. 2004). Anger, fear and other
specific emotions have been found very important in perception of environmental
risks. The role of emotion in environmental risk perception can be analyzed from
three kinds of different models—relational, process, and structural models. Rela-
tional models attempt to explain why individual’s emotional responses to the same
environmental risk are so different (Scherer 2009). It has been found that values,
abilities, goals and needs are valid reasons for the difference in environmental risk
related emotions (Lerner and Keltner 2000). Process models aim at finding how
people’s emotion is elicited when faced with risks. However, no specific research
in environmental risk perception has been conducted under this type of model.
Structural models explore the implicit relationship between emotion and envi-
ronmental risk appraisal. These three kinds of models have the potential to be
integrated as tried by Peters and his colleagues (Smith and Kirby 2009). Studies
have confirmed that emotion is closely related to environmental risk appraisals
(Watson and Spence 2007). Causes of emotion in environmental risks can be
67 The Role of Preference and Emotion 649
classified as the agency, coping potential, fairness, certainty, and outcome desir-
ability (Watson and Spence 2007).
Emotion is one of the causes of behavioral tendency (Scherer 2009). Different
emotion in risk perception can cause different environmental risk actions, implying
the important role of emotion in environmental risk management. However,
emotion is far from being the sufficient condition of behavior (Gattig and Hend-
rickx 2007; Pender 1996; Green et al. 1997). Environmental risk types and cultural
factors coordinate the relationships between emotions and behaviors. In addition,
different people have different ways to express their emotions. It requires further
researches to find what emotions are relevant in specific environmental risks, and
what are the effects of appraisals on emotional reactions in specific environmental
risks. In brief, future research can focus on the particular factors in emotion in
specific environmental risks, like nuclear leakage, water pollution, and soil
contamination.
67.4 Future Research
The intertwinement of time preference and risk preference in environmental risk
perception requires the effort to disentangle them. The research into the effect of
time delay on risk preference is the key, since it has been found that time pref-
erence is stable under different uncertainties (Gattig and Hendrickx 2007).
The following simple question can show the existence of time delay effect upon
risk preference. If a man treats $1,000 which is to be paid one year later as
equivalent to $800 today, and treats $800 as the certainty equivalent to a $2,000 or
$0 risky option with probability of 50–50, will he be indifferent to the same risky
option to be resolved and paid one year later and a certain amount of $640 ($800/
($1,000/$800))? Some recent studies have found people will prefer risky option to
certain amount of $640 (Noussair and Wu 2006).
We will extend the study in this topic, and expects to get reasonable and
convincing result on time delay effect upon risk preference. The further research
plan would consist of two types of research, laboratory behavioral experiments and
questionnaire survey. For laboratory behavioral experiments, we will design a
bunch of lotteries with different probabilities and results under both gain and loss
situations. A possible design for loss situation is shown in Table 67.1.
As shown in the Table 67.1, we will use 10 different lotteries with different
combination of results and probability. Each lottery will be exercise under four
Table 67.1 Behavioral experiment lottery design
1 2 3 4 5 6 7 8 9 10
X -1,200 -1,200 -600 -1,200 -600 -1,000 -1,200 -1,200 -1,200 -1,200
P 1/6 2/6 2/6 2/6 2/6 2/6 2/6 3/6 4/6 5/6
y 0 0 0 -600 -400 -900 0 0 0 0
650 C. Xie et al.
different time delays, 0 day, 6 week, 12 week, and any uncertain time between
0 day and 12 weeks. Therefore, participants in the experiment will face 40 dif-
ferent choices.
Different from Noussair and Wu 2006, we will assume no specific models on
expected values, especially the linear relationship between probability and prob-
ability weighting function, i.e. subjective probability. We will ask subjects to
report certainty equivalent on exercise date for each lottery rather the present
value. Data will be collected and analyzed, comparing the certainty equivalents for
the same lottery under different time delays. By imposing specific form on
probability, we will be able to analyze further into the structure of time delay effect
on risk preference.
Given the influence of emotion on risk perception, we will also test how dif-
ferent emotions affect people’s risk preference under delay. We will try to elicit
different emotions among subjects with different contents and ask them to answer
questions which are designed to test their emotions. By comparing the certainty
equivalents of the same lottery with the same time delay under different emotions,
it is able to catch insight into how specific emotions affect risk preference.
The questionnaire survey will feature specific environmental risks, like nuclear
leakage, water pollution, soil contamination or air pollution. With properly
designed questionnaire, we will gain data reflecting both people’s risk preference
in a specific environmental risk, and how their instant emotion affects time delay
effect on risk preference.
The subjects will include students, teachers, and white collar workers. Those
who join in laboratory behavioral experiments will be well guided before the
experiments to well understand the effect of their choice and get some amount of
money as reward according to their performance.
67.5 Conclusion
The roles of risk preference and time preference in environmental risks are crucial.
The finding that risk preference can be affected by time delay is of value in
understanding people’s environmental risk perceptions. We expect to find that
time delay makes people more risk tolerant. Based upon this, we will gain a new
approach to explain people overreaction towards temporal environmental risks,
and their neglect for long-term environmental risks like global warming. As for
emotion, we anticipate to find that overwhelming emotions can greatly change
people’ risk preference and time delay effect on risk preference.
This research will be the first study to explore the effect of time delay on
environmental risk perception. It is also supposed to be the first study that com-
bines risk preference and emotion in environmental risk perception. Also, it will
provide evidence for identifying the difference between people’s risk preference
under gain situations and that under loss situations. We hope to catch insight into
the role of risk preference and emotion in environmental risk perception,
67 The Role of Preference and Emotion 651
especially the effect of time delay on risk preference with application in envi-
ronmental issues. Our research will provide deeper insights to people’s attitudes
towards environmental risks, and thus offer better guidance for relevant environ-
mental risk communication and management.
Acknowledgments This work was supported by the National Science and Technology Support
Program (2009BAK53B06), the National Natural Science Foundation of China (71101035) and
the Humanities and Social Sciences of Education Ministry (12YJA880130).
References
Scherer KR (2009) The dynamic architecture of emotion: evidence for the component process
model. Cognit Emot 23(7):1307–1351
Keller C, Bostrom A, Kuttschreuter M, Savadori L, Spence A, White M (2012) Bringing
appraisal theory to environmental risk perception: areview of conceptual approaches of the
past 40 years and suggestions for future research. J Risk Res 15(1):237–256
Gattig A, Hendrickx L (2007) Judgmental discounting and environmental risk perception:
dimensional similarities, domain differences and implications for sustainability. J Soc Issues
63(1):21–39
Pender JL (1996) Discount rates and credit markets: theory and evidence from rural India. J Dev
Econ 50(2):257–296
Green L, Myerson J, McFadden E (1997) Rate of temporal discounting decreases with amount of
reward. Mem Cognit 25(5):715–723
Epper T, Fehr-Duda H, Bruhin A (2009) Uncertainty breeds decreasing impatience: the role of
risk preferences in time discounting. In: Working paper in institute for empirical research in
economics, University of Zurich, Zurich
Baucells M, Heukamp FH (2010) Common ratio using delay. Theory Decis 68(12):149–158
Liberman N, Trope Y (2008) The psychology of transcending the here and now. Science
11(11):1201–1205
Loewenstein GF, Weber EU, Hsee CK, Welch N (2001) Risk as feelings. Psychol Bullet
127(2):267–286
Boyer P (2008) Evolutionary economics of mental time travel? Trends Cognit Sci 12(6):219–224
Rundmo T, Moen BE (2006) Risk perception and demand for risk mitigation among experts,
politicians and lay people in Norway. J Risk Res 9(6):623–640
Hongxia D, Rosanne F (2010) A cross-cultural study on environmental risk perception and
educational strategies: implications for environmental education in China. Int Electron J
Environ Educ 1(1):1–19
Vastfjall D, Peters E, Slovic P (2008) Affect, risk perception and future optimism after the
tsunami disaster. Judgm Decis Mak J 3(1):64–72
Peters EM, Burraston B, Mertz CK (2004) An emotion-based model of risk perceptionand stigma
susceptibility: cognitive appraisals of emotion, affective reactivity, worldviews, and risk
perceptions in the generation of technological stigma. Risk Anal 24(5):1349–1367
Lerner JS, Keltner D (2000) Beyond valence: toward a model of emotion-specificinfluences on
judgment and choice. Cognit Emot 14(4):473–493
Smith CA, Kirby LD (2009) Putting appraisal in context: toward a relational model of appraisal
and emotion. Cognit Emot (23)7:1352–1372
Watson L, Spence MT (2007) Causes and consequences of emotions on consumer behavior—a
review and integrative cognitive appraisal theory. Eur J Mark 41(3):487–511
Noussair C, Wu P (2006) Risk tolerance in the present and the future: an experimental study.
Manag Decis Econ 27(6):401–412
Chapter 68
The Model Research on Risk Control
Qing-hai Zhang
Abstract With the development of the society and the growth of technical
complexity, the risks of many problems are increasing, which promotes a pressing
need to conduct a research on the technology and methods of risk control. Having
identified and assessed all the possible risks, the paper divides them into four
types, and designs the risk control model for each type aiming at minimizing the
risk probability and harm degree. Furthermore, the paper also extends the model
concerning the economic costs to the one concerning social benefits and other
factors.
Keywords Risk Risk control Model
68.1 Introduction
Risk is the phenomenon that widely exists in people’s work and life. Risk have the
following characteristics. Firstly, risk is the cause of disaster and accident, or the
economic loss and casualties of disaster and accident. Secondly, the occurrence of
risk is uncertain. Thirdly, the loss degree of risk is uncertain, and there’s difference
between the probable result and anticipated outcome (Doherty 2000a).
The risk management was born for the existence of risk phenomenon, which is
new management science about occurrence rule of risk and risk control technology
deriving from the USA in 1950s. The three kernel stages in risk management are
risk identification, risky appraisal and risk control. Risk identification is to identify
the present and potential factors which may cause loss in the management course,
analyze if there is uncertainty in the factors and determine if the uncertainty exists
objectively. Risk identification is the first step and stage of risk management and
the basement of the whole risk management. Risk appraise is also important work
Q. Zhang (&)
Basic Courses Department, Military Economy Academy, Wuhan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 653
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_68,
Ó Springer-Verlag Berlin Heidelberg 2013
654 Q. Zhang
that tests, weighs and estimate the risk, size up the probability and predicate the
serious degree.
The final aim of risk identify and assess is to avoid and control risk, achieving
optimum political, economic, and social benefits with the minimum cost, reduce
the probability of risk accidents and the scope and effect of loss by the greatest
extent (Jarrow and Turnbull 1998; Dohety 2000; Arther Williams and Heins RM Jr
1997). Risk control, the most popular risk management technique is fit to plan and
implementation stage, which is vividly described as ‘‘admit risk, try to reduce the
occurrence and the effect’’.
Specially, the risk control includes prior control, press control and subsequent
control. For the prior control, we need to compare risk of different schemes and
choose the best plan one giving consideration to every aspect. For the press control,
there are two conditions. In the first condition, the risk is unacceptable. In the
second condition, however, the risk is acceptable. If the risk exceeds the maximum
acceptable level of risk, we have to cancel the present scheme and choose the
alternate one, or rescue the present scheme by reducing the assess indicator and
adjust the tactical and technical data requirements. If the risk is acceptable, we
should continue monitoring the risk. For the subsequent risk, we need to summarize
and popularize advanced experience and take warning from the failure.
68.2 Risk Assortment
By risk recognize and assess, we can find out the main risk and fix its probability
and harm degree. There are four types of risk.
Risk 1, low probability, low harm. This risk is secondary and acceptable.
Risk 2, high probability, low harm. This risk should be well controlled to reduce
its probability. Though the dangerous level of individual risk is not high, it’s
necessary to guard against accumulator risk.
Risk 3, low probability, high harm. This risk is seldom, but it will be subversive
dangerous once it happens. So precautionary measures should be taken and new
type of risk should be kept a weather eye on.
Risk 4, high probability, high harm. This risk is essential, and precautionary,
shifting, diminishing measures should be taken to reduce the influence of this risk
and prevent the subversive risk (LiuJun 2008).
68.3 P–C MODEL
68.3.1 The Design Thinking of the Model
The purpose of risk control is to reduce the probability of risk accident by the
greatest extent and reduce the scope of the loss. Apparently, this is a optimization
68 The Model Research on Risk Control 655
problem. So we need to create risk control model with the method of mathematical
programming.
This model is designed for high probability and low harm risk that is risk 2.
This risk can be divided into two types. Firstly, the harm is acceptable when the
risk happens. Secondly, the harm is small but unacceptable. In the first case, the
risk can be defined as acceptable risk. In the second case, as the probability of risk
2 is high, control measures are needed to reduce the probability. Besides, the
control measures demand for cost that is called control cost. So we need to fully
consider probability P and cost C. The model is based on the integration, so it’s
called P C model (Vincent and Jeryl 1985). P C model aims at increasing
economic efficiency, which is suitable to apply to normal risk control.
68.3.2 The Creation of the P–C Model
The First Model: Single factor model
Take s1 ; s2 ; ; sn as the n measures, x1 ; x2 ; ; xn are the n control measures,
x1 ; x2 ; ; xn are the corresponding variables. For measure si ,there are two choices
which are taking measure si and not taking measure si . If measure si is taken, then
xi ¼ 1; if not, then xi ¼ 0. Apparently, xi should be 0-1 variable:
0 ; take si
xi ¼ :
1 ; nottake si
Design n one dimension functions Pk ¼ Pk ðxi Þ ; ðk ¼ 1; 2; ; nÞ, for exam-
ple, P1 ð0Þ is the risk probability when si isn’t taken; P1 ð1Þis the risk probability
when si isn’t taken. Take C1 ; C2 ; ; Cn as the corresponding control cost of
s1 ; s2 ; ; sn .
Take control measure si as an example. Whether to choose si , we should
consider Pi ð0Þ, Pi ð1Þ and Ci overall. Generally, there are three conditions:
1) If Pi ð1Þ\\Pi ð0Þ, then choose si .
2) If Pi ð1Þ\Pi ð0Þ, we should first define
control revenueRi : control revenue is revenue resulting from si which reduces the
risk probability. Mathematic expectation method is a popular method of control
profit measuring. Suppose the loss resulting from the risk is A, then control profit is
Ri ¼ ½ Pi ð0Þ Pi ð1Þ A.
There are two conditions. If Ri Ci , then don’t take measure si ; if Ri [ Ci , then
take measure si , but if Rj Cj [ Ri Ci , then take measure sj , that’s to calculate
maxfRi Ci g ; i ¼ 1 ; 2 ; ; n to determine which control measure to take.
Common ground is told, ‘‘whether the control cost is worthy’’.
3) Pi ð1Þ Pi ð0Þ is unrealistic.
In the same way, we can analyze other control measures.
The Second Model: Several factor model
656 Q. Zhang
In single factor model, there are two hypothesizes about si which are taking
measure si and not taking measure si . xi is 0-1 variable:
0 ; take si
xi ¼ :
1 ; nottake si
But in practice, we can choose several measures to control the risk, and
determine the extent of each measure. So create several factor model based on
mathematic programming knowledge (Benink 1995).
Create n dimension function P ¼ Pðx1 ; x2 ; ; xn Þ, in which xi is the extent of
control measure si , according to different extent, give xi different numerical value. P
is the corresponding risk probability when taking different extent of control mea-
sures such as s1 ; s2 ; ; sn . Suppose xi ði ¼ 1; 2; ; nÞ is continuous variable, so
function P ¼ Pðx1 ; x2 ; ; xn Þ is a n dimension continuous function of
x1 ; x2 ; ; xn . Our target is to reduce the probability of this risk, so take the function
as the target function of the programming problem, and calculate the minimum :
min Pðx1 ; x2 ; ; xn Þ:
Take C1 ; C2 ; ; Cn standing for the control cost of control measures
s1 ; s2 ; ; sn . Apparently, different extent of control measure si brings different
control cost, which means Ci is a function of xi (Jiang 2002):
Ci ¼ Ci ðxi Þ ; ði ¼ 1; 2; ; nÞ:
Then we get the restrain condition:
C1 ðx1 Þ þ C2 ðx2 Þ þ þ Cn ðxn Þ C;
C is the total acceptable cost.
Then we get the programming problem:
min Pðx1 ; x2 ; ; xn Þ
(
C1 ðx1 Þ þ C2 ðx2 Þ þ þ Cn ðxn Þ C
s:t: :
Ci ðxi Þ 0 ; i ¼ 1; 2; ; n
By calculating the programming model, we get the numerical value of
x1 ; x2 ; ; xn corresponding to the minimum of Pðx1 ; x2 ; ; xn Þ. Accordingly, we
can decide the type and extent of control measures to control risk 2 and provide the
risk control experts with accurate evidence.
68.3.3 The Extension of the P–C Model
The above model focuses on economic efficiency. But in actual life, except eco-
nomic efficiency, social efficiency is more important in some cases. Then, S which
68 The Model Research on Risk Control 657
stands for social efficiency is instead of C which stands for control cost. So the new
model can be named as P S model whose create method and steps can refer to
P C model. In the same way, the model can extend and changes into other forms
according to various needs.
68.4 r-C MODEL
This model aims at risk 3 with low probability and high harm. The risk can be
divided into two parts. The probability of the first risk is acceptable. The proba-
bility of the second risk is low, but the risk is unacceptable. In the first case, the
risk is considered to be acceptable risk (Jorion 1997; Arrow 1971; Smith 1998;
Delianedis and Geske 1998). In the second case, as the harm is serious when risk 3
happens, so it’s necessary to take control measures to reduce the loss.rsymbolizes
the loss, which is the difference between the final result and the intended goal. The
cost results from the control measures is called control cost. So we need to fully
consider loss scope r and cost C. The model is based on the integration, so it’s
called r Cmodel.
The create method, steps and extension of the r C model are similar to the
P C model.
68.5 P-r MODEL
This model aims at risk 4 with high probability and serious harm that is the key
point. Both reducing the probability P and lightening the loss scope are necessary
(Ward 1999). As a result, the model has two target functions, making use of double
target programming model:
min Pðx1 ; x2 ; ; xn Þ
min rðx1 ; x2 ; ; xn Þ
(
C1 ðx1 Þ þ C2 ðx2 Þ þ þ Cn ½xn C
s:t: :
Ci ðxi Þ 0 ; i ¼ 1; 2; ; n
This model will consider the different focus levels and demands of P and r
according to the practice. Operation researches always change double target
programming into single target programming(Editorial Board of Operational
Research 2002). That’s to make a weighted analysis of probability and loss scope
to form a target: min k1 P þ k2 r.
658 Q. Zhang
In this formula, k1 is the weight of P, k2 is the weight of r. k1 and k2 fit the
relationship:
k1 þ k2 ¼ 1; 0 k1 ; k2 1:
68.6 Conclusion
Above all, the risk2, risk3 and risk4 have been talked out. So for risk1 with low
probability and little serious, we generally consider it as secondary and acceptable
risk. If the probability or the harm is unacceptable, we need to reduce the prob-
ability or the harm when the above models could be used.
In the particular problems, there are many methods of risk control. But theory
and practice of risk management prove that either method has its applicability and
limitation. So we need to choose different control method according to different
problem and characteristic of risk. The three models in this paper reduce the
probability of the risk and lighten the loss scope of four types of risk, using the
programming knowledge together with economic cost. It provides risk control
experts with theoretical basis from the view of methodology (Doherty 2000b;
White 2004). But in actual risk control, the risk control experts ought to overall
consider different factors to decide which method to take with their own
experience.
References
Arrow KJ (1971) Essays in the theory of risk bearing. NorthHolland, New York, pp 86–93
Arther Williams C,.Heins RM Jr (1997) Risk management and insurance. McGraw-Hill Higher
Education, Boston, pp. 17-31, 41–47
Benink HA (1995) Coping with financial fragility and systemic risk boston. Klumer Academic
Publishers, London, pp 43–47
Delianedis G, Geske R (1998) Credit risk and risk-neutral default probabilities: information about
rating migrations and defaults. Paper presented at the Bank of England conference on credit
risk modeling and regulatory implications, London, 21–22 Sept 1998
Doherty NA (2000) Integrated risk management techniques and strategies for managing corporate
risk. McGraw-Hill, New York, pp 65–67
Doherty NA (2000) Integrated risk management techniques and strategies for managing corporate
risk. McGraw-Hill, New York, pp 134–167
Neil A. Dohety (2000) Integrated risk management. McGraw-Hill Companies, New York,
pp 37–40
Editorial Board of Operational Research (2002) A brief introduction to operational research. Qing
hua university press, Beijing, pp 75–79
Jarrow RA, Turnbull SM (1998) The intersection of market and credit risk. Paper presented at the
Bank of England Conference on Credit Risk Modeling and Regulatory Implications, London,
21–22 Sept 1998
Jiang Q (2002) Mathematic model. Qing hua university press,Beijing, pp 79–84
68 The Model Research on Risk Control 659
Jorion P (1997) Value at risk: the new benchmark for controlling market risk. The McGraw-Hill
Companies, Inc., New York, pp 122–126
LiuJun (2008) An introduction to risk management. Qing hua university press, Beijing
Smith ML (1998) Risk management and insurance. McGraw-Hill Inc , New York, pp 106–118
Vincent TC, Jeryl M (1985) Risk analysis and risk management: an historical perspective. Risk
Anal 5(2):103–120
Ward SC (1999) Assessing and managing important risks. Int J Proj Manag 17:331–336
White L (2004) Management accountants and enterprise risk management. Strateg Financ
43:10–14
Chapter 69
TOPSIS Based Power-Saving Plans
Choice for Manufacturing Enterprise
Dong-sheng Wang and Kuan-ming Zheng
Abstract To accommodate the demand of low-carbon economy and lowering
enterprise’s production costs, the issues of one electronic manufacturing enterprise
on power-saving are analyzed, and the plans for its power-saving are devised. With
TOPSIS, these plans are appraised and optimized, and the effects on cutting down
costs and improving efficiency are summed up.
Keywords Low-carbon economy Manufacturing enterprise Power-saving
TOPSIS method
69.1 Introduction
Since the twenty-first century, low-carbon economy has been attached more and
more importance to. For manufacturing enterprises, the low-carbon developing
model means that diminishing their energy consuming, improving the utilizing
efficiency, and curtailing waste discharging are the foremost issues to be resolved.
Improving production efficiency and lowering equipment cost are the basic
conditions for the optimization of any enterprise’s power-saving plans. The opti-
mization methods include Fuzzy Evaluation Model, Analytical Hierarchy Process,
Gray Comprehensive Evaluation, TOPSIS, etc., (Yue 2003) among which
TOPSIS, as a simple statistical method, is of high reliability and of low error (Guo
and Jin 2010). This paper, taking one electronic manufacturing enterprise as the
case, puts forward three power-saving plans, optimizes these plans with TOPSIS,
and analyzes the effects of the optimal choice.
D. Wang (&) K. Zheng
School of Management, Shaanxi University of Technology, Hanzhong, Shaanxi, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 661
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_69,
Ó Springer-Verlag Berlin Heidelberg 2013
662 D. Wang and K. Zheng
69.2 Methodology
69.2.1 Basic Ideas of TOPSIS Method
Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is an
important multi-attributes decision model, which finds the positive ideal plan and
negative ideal plan with normalized initial matrix, to calculate the relative
closeness between one plan and two ideal plans (Xu 2010). Based on the relative
closeness, the sequence of the evaluation results is achieved. The positive ideal
plan is the optimal one, which is a virtual plan, and its attributes are the best. And
the negative ideal plan is the worst one, which is also virtual plan, and its attributes
are the worst (Feng and Liu 2005).
69.2.2 Steps of the TOPSIS Method
Step 1: Set up normalized decision matrix (Verma and Pullman 1998)
Xij
Yij ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðj ¼ 1; 2; 3. . .nÞ ð69:1Þ
Pm
X 2 ij
i¼1
Step 2: Set up weighted normalized decision matrix (Lin et al. 2008)
Vij ¼ Wi Yij ði ¼ 1; 2. . .n j ¼ 1; 2. . .nÞ ð69:2Þ
Step 3: Calculate the positive solution V+ and negative ideal V- (Liao and
Rittscher 2007)
V þ ¼ maxvij ; j J1 ; minvij ; j J2 ; i ¼ 1; 2; 3; . . .; m
ð69:3Þ
V ¼ minvij ; j J1 ; maxvij ; j J2 ; i ¼ 1; 2; 3; . . .; m
Step 4: Calculate Euclidean distance (Barbarosoglu 2000)
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Xm 2
Dþ ¼ Wi Vij Viþ
i¼1
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð69:4Þ
X m 2
D ¼ Wi Vij Vi
i¼1
Step 5: Calculate the relative adjacent degree (Pawlak 1982)
D
i
Ci ¼ ð69:5Þ
Dþ
i þ D
i
Step 6: Obtain the optimal plan from Ci (Bin and Li-jie 2006)
69 TOPSIS Based Power-Saving Plans Choice 663
69.3 Application of TOPSIS Method to Enterprise’s
Power-Saving Plans
69.3.1 Troubles in the Power-Saving of the Case Enterprise
The case enterprise is an electronic manufacturing factory. There are too many
testing processes with high energy consumption. Before packaging, there are for
steps of test for the products. Its testing workshops use LCD equipment to display
the testing information. The followings are the existed troubles for the enterprise:
Large space is occupied by LCD equipment. The information displayed by
LCD equipments is very simple and much equipment is used for the tests, which
lead to large space occupation. The whole produce lines are crowded and the
layout of lines is in chaos, which caused the operating personnel to walk
frequently, and to walk in large range. These factors speed up personnel’s fatigue,
and the efficiency is low.
The testing equipments are of high cost. Through statistics, there are 373 testing
stations in the whole workshop. The cost of LCD equipment is 1200 Yuan per
station, and the total cost is up to 447,600 Yuan.
The energy consumption is high. Through statistics, the power fee from one
workshop is high up to 96,300 Yuan per year.
69.3.2 TOPSIS Based Power-Saving Plans Choice
Through brain-storm, three plans are addressed. The first is that replace the CMC
for LCD equipment which is a kind of apparatus for displaying scanning and
testing information. The advantage of this plan is of comprehensive information
and of good effects; and the weakness is of low visual angles and of high cost. The
second plan is that substitute LED equipment for LCD. The advantage is of good
visualization, of efficient information and of low equipment cost; and the weakness
is of too simple in displaying information. The third plan is that change LCD
equipment for LED indicator lamps with different colors. The advantage is of
direct observation, of efficient information and of low price; and the weakness is of
little information and of chaos in visualization.
Through analysis and research, lowering equipment cost, diminishing space
occupation, saving power, and improving the efficiency are the indexes for plan
choice. The steps to use TOPSIS for the choice of the power-saving plans are the
followings:
Step 1: Set up decision indexes set C
664 D. Wang and K. Zheng
2 3
y1 ¼ lowering cost
6 y ¼ improving efficiency 7
6 2 7
C¼6 7
4 y3 ¼ occupying space 5
y4 ¼ saving power
Step 2: Set up plans set X
2 3
x1 ¼ plan1
6 7
X ¼ 4 x2 ¼ plan2 5
x3 ¼ plan3
Step 3: Decide the weight of indexes and set up the weight set W
2 3
W1 ðlowering costÞ ¼ 0:25
6 W ðimproving efficiencyÞ ¼ 0:15 7
6 2 7
W¼6 7
4 W3 ðoccupying spaceÞ ¼ 0:35 5
W4 ðsaving power Þ ¼ 0:25
Step 4: Decide the actual value of indexes
Through market survey and field measurement, the price of CMC is 1500 Yuan
per station, and the space occupation is 0.00179 m3; the price of LED monitor is
200 Yuan per station, and the space occupation is 0.00145 m3; the price of LED
indicator is 150 Yuan each, and the space occupation is 0.00062 m3. According to
the above data, the score of the three plans’ efficiency-improving and power-
saving can be obtained as Table 69.1:
Step 5: Ascertain normalized decision matrix Y
The dimensions of the above values are different. Thus it is necessary to
transform the value to normalized ones with Eq. 69.1.
Y11 ¼ 0:986394; Y12 ¼ 0:131519; Y13 ¼ 0:098639; Y21 ¼ 1:053182;
Y22 ¼ 1:180127; Y23 ¼ 0:52718; Y31 ¼ 0:32014; Y32 ¼ 0:260452;
Y33 ¼ 0:11146; Y41 ¼ 0:840125; Y42 ¼ 0:74425; Y43 ¼ 1:04825
From the above values, the decision matrix Y can be obtained:
2 3
0:986 1:053 0:320 0:840
Y ¼ 4 0:132 1:180 0:260 0:744 5
0:100 0:527 0:111 1:048
Step 6: Calculate weighted decision matrix V with Eq. 69.2.
2 3
0:247 0:158 0:112 0:210
Y ¼ 4 0:032 1:180 0:091 0:186 5
0:025 0:527 0:039 0:262
69 TOPSIS Based Power-Saving Plans Choice 665
Table 69.1 Initial value of Lowering Improving Occupying Saving
decision indexes Index
Plan cost (Yuan) efficiency space (m3) power
Plan 1 1500 8 0.00179 9
Plan 2 200 9 0.00145 8
Plan 3 150 4 0.00062 9
Table 69.2 Euclidean Plan 1 Plan 2 Plan3
distance of the three plans Plan
Item
D+ 0.240 0.092 0.099
D- 0.082 0.236 0.246
Step 7: Calculate positive and negative ideal solution with Eq. 69.3.
V þ ¼ ð 0:025 0:177 0:039 0:262 Þ
V ¼ ð 0:247 0:079 0:039 0:262 Þ
Step 8: Calculate Euclidean distance D with Eq. 69.4.
Dþ þ
1 ¼ 0:240; D1 ¼ 0:082; D2 ¼ 0:092;
D2 ¼ 0:236; D3 ¼ 0:099; D
þ
3 ¼ 0:246
According to the above value, Table 69.2 can be ascertained.
Step 9: Calculate the relative adjacent degree C with Eq. 69.5.
C1 ¼ 0:255; C2 ¼ 0:720; C3 ¼ 0:713
C2 [ C3 [ C1
Among which, the second plan is the most optimal as the final choice.
69.4 Performance Analysis on the Plan Choice
Until present, the case enterprise has introduced plan2 to display testing infor-
mation, and the performance is fairly well. Before improvement, the total cost of
equipment is up to 447,600 Yuan and the power fee from one workshop is high up
to 96,300 Yuan per year. After the enforcement of plan 2, the total cost is
74,600 Yuan and the power fee is 12,000 Yuan per year. The cost saving is
373,000 Yuan, and the power fee is lowered by 84,300.
At the same time, the substitute of LED monitor for LCD equipment makes the
production line in order, improves personnel’s morale, and allows workers to see
the information clearly without much motion. And through speech set on the
666 D. Wang and K. Zheng
equipment, the information of code type can not only be identified by visualization
but also by sound, which facilitates workers’ operation. And the replaced LCD
equipment can be made use by other department and production line. The goal of
lowering cost and saving power is attained.
69.5 Discussion
Since the early twenty-first century, we have focused on low carbon, which
emphasizes protecting our environment. For the public, it means leading a simple
and saving life. And for the enterprises, it means eliminating the redundant
emission, lowering power using, comprehensively reusing the material called
waste before, and recycling the materials that can be reused such as packages,
bottles, and so on.
Unfortunately, in the early stage of the low-carbon economy, there are a
number of requirements on the publics at present. Those on enterprises seem to be
neglected. Someone will say that environment protection has been awakened by
many governments in the world, and the relative institutions have been initiated
since the 1960s. But an obvious fact must be laid on the desk that compared with
the public the amount of power-using by enterprises is fairly huge. Thus, enter-
prises should be the priority to low-carbon economy.
For enterprises, to meet the requirement of reducing emission, reusing and
recycling, they should enforce their consciousness and self-discipline besides
abiding by the outer institutions, which means they should adjusting their opera-
tion strategies to cover reducing, reusing and recycling., and some scientific
methods should be adopted.
This paper takes a specific case to explore the way to lower its power-using
with TOPSIS. But faced with the pressure of profit earning and growing, quite a
few enterprises will not consider too much low-carbon. Seemingly, it is reason-
able. But through deep exploring, the fact that one enterprise which complies to
the demand of low-carbon usually affords high costs, which will be a burden for its
development. And those enterprises that consider these factors of low-carbon will
not obtain anything in return. Although some stimulus measures have been
enforced in low-carbon by some governments; these measures are usually treated
as temporary ones. And form the long period perspective, these measures will
cause damage to the enterprises’ operation. The reasons are that some enterprises
will rely highly on the allowances form government, and some will use these
allowances to compete with others, which will cause unfairness, even international
trade disputes such as anti-dumping and anti-subsidies.
So, for enterprises, the low-carbon economy needs more innovations including
technology, operation and management. The technology innovation means some
techniques that are efficient in low-carbon will be invented and adopted by
enterprises. The operation innovation means the enterprises must change their
visions. From the present research and practice, low-carbon supply chain seems as
69 TOPSIS Based Power-Saving Plans Choice 667
one effective strategy. Its application will equalize the costs of reusing, reducing
and recycling on the supply chain. This strategy needs some supportive means
such as low-carbon contracts for all members on the chain, interest-collaboration
between different members, and so on. And on the basis of operation innovation,
the management must innovate on some aspects such as information management,
outsourcing, supplier management, channel management, etc.
In a word, this paper probes the way to low-carbon development of one
enterprise’s power-saving. To comprehensively realize low-carbon economy, the
field must extend from the public to all industries. And enterprises should play
relatively important role in the process. The cost of the realization of low-carbon
economy requires a full range of innovations.
References
Barbarosoglu G (2000) A decision support model fore customer value assessment and supply
quota allocation. Prod Plan Control 11(6):608–616
Bin S, Li-jie W (2006) Study of method for determining weight based on rough set theory.
Comput Eng Appl 29:216–217
Feng K, Liu H (2005) A new Fuzzy TOPSIS Algorithm for MADM based on decision maker’s
subjective preference. In: Proceedings 24th Chinese control conference, Guangzhou, P.R.
China, pp 1697–1701
Guo X, Jin L (2010) Grey incidence TOPSIS for multiple attribute decision making (in Chinese).
Sci Technol Manage CHN 12(5):49–51
Liao Z, Rittscher J (2007) A multi-objective supplier selection model under stochastic demand
conditions. Int J Prod Econ 105(1):150–159
Lin M-C, Wang C-C, Chen M-S, Chang CA (2008) Using AHP and TOPSIS approaches in
customer-driven product design process. Comput Ind 59(1):17–31
Pawlak Z (1982) Rough sets, theoretical aspects of reasoning about data. Int J Comput Inform Sci
11:314–356
Verma R, Pullman ME (1998) An analysis of the supplier selection process, vol. 26, no. 6.
Omega, pp 739–750
Xu K (2010) TOPSIS method based on pairwise comparisons (in Chinese). Math Pract Theory
CHN 40(5):110–114
Yue C (2003) Theory and methods for decision (in Chinese). Science Press, Beijing, pp 133–140.
Chapter 6
Chapter 70
Research on Central Control DDSS
System for Fund Portfolio Management
Cheng Hu and Er-shi Qi
Abstract In order to satisfy the demand of fund portfolio management and based
on the feature of balancing in the central control and distribution decision, a
central control DDSS is schemed based on the systemization of decision-making
theory. The scheme provides a balance point of a dynamic management, and
makes the fund investment more controllable and flexible.
Keywords Portfolio management Systemization of decision-making Distrib-
uted decision support system (DDSS)
70.1 Introduction
The basic principle of portfolio management is making investment scattered in
many different kinds of assets by a certain percentage under the standard of
investor risk appetite, to achieve the goal of utility maximization (Elton et al.
1996). In fact, investment funds reflect this concept of a portfolio management.
Investment fund is a portfolio of investments, including various securities and
assets; investor who purchases a fund is equivalent to purchase a group of com-
pany’s stocks or a combination of different asset classes.
In practice, investment funds sometimes are decentralized managed by several
different managers, rather than managed by a single manager. Large fund man-
agement companies will choose several managers to form a combination of
investment managers, or consider the use of three to five funds to compose a Fund
group. This is actually a portfolio management concept that spread the risks of
investment objectives to different managers. Individuals have different risk
C. Hu (&) E. Qi
Management School of Tianjin University, 300072 Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 669
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_70,
Ó Springer-Verlag Berlin Heidelberg 2013
670 C. Hu and E. Qi
preferences and profitability goals in different environments, a particular
combination of the investment managers will wash away the non-rational devia-
tion through their own independent operations to achieve both portfolio risk and
profitability objectives.
Distributed Decision Support System (DDSS) is a tool that is compatible with
this multi-team portfolio management style. DDSS is the combination of distrib-
uted decision-making method, distributed databases, distributed operating system
and distributed supporting (Swanson 1990). The supporting environment for
DDSS is a computer network which is constituted by various physically separated
processing information node, each node of the network contains at least one
decision-making supportive system or a number of decision supporting functions
(Kirn and Schlageter 1992). Comparing with the centralized DSS, this distributed
decision supporting system is closer to the actual situation of large-scale organi-
zational decision-making activities, especially related to fund management system
which is an organization that has a team of decision-makers with different
responsibilities in a multi-level decision-making system.
But the idea of the classic distributed decision supporting system is to consider
each individual decision-makers or decision-making organization as an indepen-
dent, physically separated processing information node and the system provides
individual, group and organizational support for processing and decision-making
in these nodes (Gao 2005). Thus, the fully independent decision-making of each
node and multi-team management of investment fund is different in reality. When
multi-team of traders is operating the fund, the total amount of funds is always
limited, the funds between the various teams are allocated, and the funds for each
team in the selection of investment targets and resources are also limited.
Therefore, central control has become necessary in DDSS system.
70.2 Theory
Systemization of decision-making theory is originally introduced by Herbert
Simon (Yue 2003). He introduced the concept of systemization of decision-making
after he studied a large number of highly organized linear programming applica-
tions, and tried to implement the method to solving company problems (Janis and
Mann 1977). The decision comes from the detailed analysis and discussion in
advance, gets the understanding of participants to input defined data, and generates
decision in the form consistent with the expected results (Roy 1977). So we can
determine in advance a type of decision that is wanted, run a series of fixed
calculation, and get the expected result which is in line with organizational
objectives (Zeleny 1977).
If the decision-making process is systemized, it can be prepared in routinization,
so that you can delegate the task to first-line manager with independent decision-
making (Chung 1993). Given controlled constraints and values within the expected
range, they can make independent decisions. In this case, all decisions are
70 Research on Central Control DDSS System 671
Fig. 70.1 A systemized decision-making model
constrained under the systemization of decision-making process. Decision-making
process starts with restriction when developing decision. Another major restriction
in decision-making process is that the available resources. These resources include
the personnel and tools for decision-making that can be arranged. It is the
restrictions that constitute a central control of distributed decision making.
The systemization of decision-making model can be expressed as shown in
Fig. 70.1.
All decision-making runs in a systemized decision-making model under the
constraints of a linear program. The top management will not make many decisions
by following one routine; they are only responsible for setting the rules of first-line
managers in decision-making. The systemization of decision-making system is
developed analysis discussed previously with the understanding of the participants.
By inputting defined data, the result produced by the system will meet our expec-
tation. They can pre-determine the action of subordinates or the computer analysis.
Subordinates or computer will follow the series of fixed calculation to achieve the
expected results. This hierarchical decision-making process provides the basic
concept of central control for the distributed decision support system in reality.
70.3 Model
In the DDSS classic model, each subsystem has its own database, model base and
method base system. Data resources and decisions made are exchanged on each
node through network information system; the result of processing all nodes would
give us the final decision. This design emphasizes: the exchanges between the
nodes, independent decision-making of the node, as well as the combined result of
all decision made by all the independent nodes.
However, the structure of such systemized decision-making model lacks the
control over decision-making in the system. As mentioned earlier, in order to
672 C. Hu and E. Qi
systemize a DDSS, the DDSS model should also include a common database, a
common model base, a common method base, and a common knowledge base.
These four libraries are the constraints for all the subsystem when it comes to
decision making. Comparing portfolio management in practice with DDSS, the
common database stores data regarding fund managers team goals and the allo-
cation of the resources; method base stores data that represents portfolio managers
investment combinations and risk control policies; model base holds models to
processing asset valuation and risk assessment formulas; knowledge base is the
storage for the various subsystems decision-making results, and it also make
adjustments to the other three database’s constraint and algorithm.
A DDSS model with the four libraries as mentioned above is shown in
Fig. 70.2:
Fig. 70.2 Central control DDSS
70 Research on Central Control DDSS System 673
Fig. 70.3 DDSS model with multi-management and central control
In the above DDSS model, the four public database access the DDSS through
the network information system, all independent decision-making subsystems use
data from all four common database and provides result under the given con-
straints. In general, the four common databases require constraints inputted
through the man–machine interface. In some special cases, these data can come
from an intelligent decision support system with specialized control indicators.
674 C. Hu and E. Qi
70.4 Application
Based on the above structure of DDSS model, a DDSS model with multi-
management and central control is developed, as shown in Fig. 70.3:
In this framework, the portfolio management committee inputs the investment
funds, investment objectives, portfolio principles and other fund associated data to
DDSS. Decision-making support system decomposes all managers’ constraints
into the four libraries, and they will become indicators and constraints in man-
ager’s decision-making process. At the same time, the knowledge base, model
base and database provides the supporting environment for the subsystem on
decision making by providing target parameters, constraints, and some general
data, which determine the overall system control of the portfolio. Fund Manager of
each team has its own independent decision support system, and they are the
subsystem of a DDSS. This subsystem can be structure to be used by a single
decision maker’s DDS. Depending on the fund manager’s knowledge of intelligent
tools, the subsystem can be structured for a single decision-maker with a smart
DDS. Fund managers make decisions under the above constraints and environment
with their own judgments.
70.5 Conclusion
In this paper, in order to satisfy the demand of fund portfolio management and
based on the feature of balancing between central control and distribution decision,
a systemized central control DDSS scheme for portfolio management is proposed.
By introducing the systemization of decision-making theory and model, and
adding overall resource constraints and operational constraints of a central data-
base onto the classic DDSS model, the DDSS model is more controllable and
flexible for a dynamic portfolio management in terms of balancing between the
central control and distribution decision of DDSS model.
References
Chung HM (1993) Distributed decision support systems: characterization and design choices. In:
Proceedings of the 26th annual Hawaii international conference on system sciences,
pp 660–667
Elton EJ, Gruber MJ, Brown SJ, Goetzmann WN (1996) Modern portfolio theory and investment
analysis, 6th edn. Wiley Publishers, New York
Gao H (2005) Decision support systems (DSS) theory, case. Tsinghua University Press, Beijing
Janis IL, Mann L (1977) Decision making: a psychological analysis of conflict, choice, and
commitment. Free Press, New York
Kirn S, Schlageter G (1992) Distributed decision support in federative knowledge based systems.
In: 2nd ISDSS conference, Ulm, Germany
70 Research on Central Control DDSS System 675
Roy B (1977) A conceptual framework for a prescriptive theory of decision aid, multiple criteria
decision making, TIMS studies in the management sciences, vol 6. North-Holland Publishing,
Amsterdam, pp 179–210
Swanson EB (1990) Distributed decision support systems: a perspective. In: Proceedings of the
23rd annual Hawaii international conference on system sciences, pp 129–136
Yue C (2003) Decision theory and methods. Science Press, Beijing
Zeleny M (1977) Adaptive displacement of preferences in decision making, multiple criteria
decision making, TIMS studies in the management sciences, vol 6. North-Holland Publishing,
Amsterdam, pp 147–159
Author Biographies
HU Cheng (1954) Male, Chairman of the Board for Hong Kong Licheng Capital Group, is engaged in
investment banking industry. No. 43 Queen’s Road East, Hong Kong, Room 1607, telephone (852)
68762311
QI Er-shi (1954) Male, Professor, Doctor Tutor, Tianjin University, is engaged in the research in
various fields of industrial engineering. Management School, Tianjin University, Tianjin 300072,
China. Tel (022) 27405100
Chapter 71
The Evaluation and Application
of Residential Structure System Based
on Matter-Element Model
Sen Yu and Xiang-ju Liu
Abstract According to the progress of building industry, residential construction
model suits to Chinese population, resource and environment development is
required to be established as soon as possible. Residential industrialization is the
inevitable course of development with premise of new technology, new material
and appropriate structure system. The paper comprehensively evaluates the current
residential structure by Matter-element theory to chose the most efficient resi-
dential structure system and supply theoretical basis for investment decisions.
Keywords The matter-element model Structure system Comprehensive
evaluation Application
71.1 Introduction
The current residential structure system contains Cast-in situ concrete structure,
steel structure, the assemble type concrete structure with both advantages and
disadvantages (Lei and Chen 2010). For example, Cast-in situ concrete structure is
good at safety and durableness, but has a complex process and a high energy
consumption; steel structure supplies a larger space (Bi 2008), needs a shorter
construction period with higher cost, poor refractory and corrosive resistance and
the assemble type concrete structure charactered a short construction period,
S. Yu (&)
School of Management,
Xi’an University of Architecture and Technology, Xi’an 710055, China
e-mail: [email protected]
X. Liu
Department of Technology, The Engineering Co. Ltd of China Construction,
Hefei 230000, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 677
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_71,
Ó Springer-Verlag Berlin Heidelberg 2013
678 S. Yu and X. Liu
energy efficiency, good quality but single structure form and poor seismic per-
formance (Jia et al. 2010). Consider the above, it is extremely important to make
an overall assessment on residential structure system in technological and eco-
nomical aspects (Zhang 2010a).
Face the different structure of the residential system, how to carry out the
effective evaluation become a key issues (Zhang 2010b). Until now, our evaluation
system has been stayed at the primary stage, only used construction cost and
energy consumption as main indexes without any analysis for influence on envi-
ronment and society nor different degree of residential technical and economic
performance between different social groups (Mi et al. 2010). To solve the
question, a comprehensive evaluation system is called for.
Related Matter-element theory and some qualitative and quantitative research
methods are used in this paper to evaluate the current residential structure system
scientifically, synthetically and reasonably. Hoping to give a guidance to the
government and real estate company, in order to promote popularize of the most
effective residential structure and supply scientific basis for development of
Residential industrialization (Huang and Zhu 2009).
71.2 Selection of the Case
Take a six layer residential building for example, its total length is 65 m, total
width is 12 m, total height is 18 m, every floor area is 780 m2, the construction of
the residential area is 97.5 m2, seismic intensity is 7°, the site category is 3rd.
Three types of residential structures are chosen, the supporting dates are recorded
in Table 71.1.
71.3 The Establishment of Performance Evaluating
Indicator System
The establishment of performance evaluating indicator system should follow
principles like Scientific principle, systematic principle, operability principle and
others (Porter 1990). And residence life cycle is selected as research object, and
Table 71.1 The cost date
Structure Construction cost Energy consumption
(yuan/m2) (yuan/m2)
Cast-in situ concrete structure 1430 65
Steel structure 1980 72
The assemble type concrete structure 1760 58
71 The Evaluation and Application of Residential Structure System 679
Table 71.2 Each index Cij relative to the index system of T weights
First grade indexes Weigh Second grade indexes Weigh
Applicability C1 0.18 Reconstruction C11 0.08
Flat surface layout C12 0.06
Indoor and outdoor traffic condition C13 0.04
Economy C2 0.26 Construction cost C21 0.16
Use-cost C22 0.10
Safety durability C3 0.36 Durability of construction C31 0.16
Durability of decoration C32 0.12
Building Fire Protection C33 0.06
Control of indoor pollution C34 0.02
Sustainable development C4 0.2 Building energy saving C41 0.11
Green building materials C42 0.04
Rationality of the water resources utilization C43 0.05
indexes of applicability c1, economy c2, safety durability c3 and Sustainable
development c4 are taken into consideration (Porter 1985). Each level index
weighs are calculated by AHP-LSDM, and the results are recorded in Table 71.2.
71.4 The Comprehensive Evaluation Based
on the Matter-Element
71.4.1 The Establishment of the Matter-Element
Evaluation Model
1. Determining the matter-element matrix of joint region, classics region and the
matrix to be evaluated.
The system of matter-element is an unit to describe the object, which consists of
the name of object N, characteristic c and the value v of the object’s characteristic
c. N, c, v is called the three key elements of the matter-element (Baldwin and Kim
1997).
The normalized form of matter-element is described as follows: R = [N, c, v].
Usually, an object has more than one characteristic. Given n characteristics of
an object c1 ; c2 ; . . .cn; and relevant values v1 ; v2 ; . . .; vn; , which we can use to
describe the object called n-dimension matter-element, recorded as:
2 3 2 3
N c1 v1 R1
6 c v 7 6 R2 7
6 2 2 7 6 7
R¼6 .. .. 7 ¼ 6 .. 7
4 . . 5 4 . 5
cn vn Rn
680 S. Yu and X. Liu
(a) Determining the matter-element matrix of partial unit
2 3
P; c1 hap1 ; bp1 i
6 c2 hap2 ; bp2 i 7
6 7
Rp ¼ ½P; ci ; xpi ¼ 6 .. .. 7
4 . . 5
cn hapn ; bpn i
In the murderer, P shows the whole effect level, xpi ¼ hapi ; bpi i shows the value
range of domain object on the characteristic ci (Baldwin and Kim 2000).
(b) Determining the matter-element matrix of classics region
2 3
Nj; c1 haj1 ; bj1 i
6 c2 haj2 ; bj2 i 7
6 7
Rj ¼ ½Nj ; ci ; xji ¼ 6 .. .. 7
4 . . 5
cn hajn :bjn i
In the murderer, Nj shows j effect levels divided to evaluate ðj ¼ 1; 2; . . .; mÞ,
ci shows the characteristic of effect level Nj ði ¼ 1; 2; . . .; nÞ, xji ¼ haji ; bji i
shows the value range of Nj specified by ci , that is the numerical scope of effect
level on the characteristics (Liu 2010).
(c) Determining the matter-element matrix to be evaluated
2 3
P0 ; c 1 x 1
6 c2 x2 7
6 7
R0 ¼ 6 .. .. 7
4 . . 5
cn xn
In the murderer, R0 shows matter-element to be evaluated, P0 shows subject
matter, xi shows the value of P0 specified by ci .
2. Determining the correlation function and the order of evaluation
Correlation function is:
8
> qðx ;x Þ
< i ji xi 2 xji
jxji j
K j ðxi Þ ¼ qðxi ;xji Þ ð71:1Þ
>
: xi 62 xji
qðxi ;xpi Þqðxi ;xji Þ
1 1
q xi ; xji ¼ xi aji þ bji bji aji ð71:2Þ
2 2
1 1
q xi ; xpi ¼ xi api þ bpi bpi api ð71:3Þ
2 2
71 The Evaluation and Application of Residential Structure System 681
3. Determining the comprehensive correlation and the order of evaluation
X
n
Kj ðP0 Þ ¼ wij Kj ðxi Þ; ð71:4Þ
i¼1
In the murderer wij For each characteristics of cij is weight coefficient. Eval-
uation of the level for the subject matter:
kj ¼ max kj ðp0 Þ ðj ¼ 1; 2; . . .; mÞ:
71.4.2 Application in Residential Structure
Evaluation System
1. Divising the standard of evaluation result
According the research interview of different projects and the specialistic, we
make the evaluation result into four levels in broad outline: excellent, good,
ordinary, bad (Li 2008) (see Table 71.3).
2. Determining the matter-element matrix of joint region, classics region and
the matrix to be evaluated
(a) Determining the matter-element matrix of partial unit
Rp ¼ P; ci; xpi
excellent---bad reconstruction C11 h0; 1i
flat surface layout C12 h0; 1i
¼ .. ..
. .
ationality of the water resources utilization C43 h0; 1i
(b) Determining the matter-element matrix of classical domain
R1 ¼ N1 ; ci; x1i
excellent; reconstruction C11 h0:75; 1i
flat surface layout C12 h0:75; 1i
¼ .. ..
. .
ationality of the water resources utilization C43 h0:75; 1i
Table 71.3 Index level standard
Evaluation index Excellent Good Ordinary Bad
C11 reconstruction h0:75; 1i h0:5; 0:75i h0:25; 0:5i h0; 0:25i
C12 flat surface layout h0:75; 1i h0:5; 0:75i h0:25; 0:5i h0; 0:25i
.. .. .. .. ..
. . . . .
C43 rationality of the water resources utilization h0:75; 1i h0:5; 0:75i h0:25; 0:5i h0; 0:25i
682 S. Yu and X. Liu
R2 ¼ N2 ; ci; x2i
good; reconstruction C11 h0:5; 0:75i
flat surface layout C12 h0:5; 0:75i
¼ .. ..
. .
ationality of the water resources utilization C43 h0:5; 0:75i
R3 ¼ N3 ; ci; x3i
ordinary; reconstruction C11 h0:25; 0:5i
flat surface layout C12 h0:25; 0:5i
¼ .. ..
. .
ationality of the water resources utilization C43 h0:25; 0:5i
R4 ¼ N4 ; ci; x4i
bad; reconstruction C11 h0; 0:25i
flat surface layout C12 h0; 0:25i
¼ .. ..
. .
ationality of the water resources utilization C43 h0; 0:25i
(c) Determining the matter-element matrix of evaluated
Cast-in situ concrete structure, steel structure and the assemble type concrete
structure are selected as research object, through investigation and analysis, can
get 12 of evaluation index of the normalized data, the matter-element matrix as
follows:
3. Determining the correlation function and the order of evaluation
(Table 71.4).
Take Cast-in situ concrete structure for example, by Eqs. (71.1), (71.2) and
(71.3) can draw C11—C44 about the level of correlation degree index and use
Eq. (71.4) to review that the subject matter P0 about each level of correlation
degree, see Table 71.5.
After the same steps, we can get the result of three structures about each level of
comprehensive correlation degree, see Table 71.6.
For the Cast-in situ concrete structure, k2 ð pÞ ¼ max kj ðpÞ; j 2 ð1; 2; 3; 4Þ, that
means that it belongs to level good.
Table 71.4 Normalized date
Structure C11 C12 C13 C21 C22 C31 C32 C33 C34 C41 C42 C43
Cast-in situ concrete 0.91 0.71 0.65 0.47 0.56 0.74 0.65 0.46 0.32 0.43 0.32 0.65
structure
Steel structure 0.82 0.68 0.64 0.28 0.24 0.65 0.52 0.42 0.55 0.56 0.46 0.42
He assemble type 0.65 0.76 0.85 0.79 0.81 0.68 0.71 0.79 0.81 0.68 0.55 0.78
concrete structure
71 The Evaluation and Application of Residential Structure System 683
Table 71.5 Correlation of the index
Evaluation index Weigh Excellent Good Ordinary Bad
C11 0.08 -0.2222 0.4000 -0.3000 -0.5333
C12 0.06 0.0435 -0.0400 -0.5200 -0.6800
.. .. .. .. .. ..
. . . . . .
C43 0.05 0.1579 -0.1200 -0.5600 -0.7067
Kj ðp0 Þ 0.1784 -0.0142 -0.4998 -0.6664
Table 71.6 Three structures about each level of correlation degree
Structure Comprehensive correlation degree
Cast-in situ concrete structure -0.4362 0.0384 -0.0834 -0.571
Steel structure 0.4845 -0.0837 -0.0756 -0.8862
Assemble type concrete structure 0.1784 -0.0142 -0.4998 -0.6664
For the steel structure, k1 ð pÞ ¼ max kj ð pÞ; j 2 ð1; 2; 3; 4Þ, that means that it
belongs to level excellent.
For the assemble type concrete structure, k1 ðpÞ ¼ max kj ðpÞ; j 2 ð1; 2; 3; 4Þ,
that means that it belongs to level excellent.
In order to compare the two structures further, after we get each of the result
normalized, we evaluate the results secondly. If kj0 ð pÞ ¼ max kj ðpÞ; j
2 ð1; 2; . . .; mÞ, that means that p belongs to level j0 (Li et al. 2007) if
Kj ðpÞ minKj ðpÞ
Kj ðpÞ ¼
maxKj ðpÞ minKj ðpÞ
X
s X
s
j ¼ j Kj ðpÞ= Kj ðpÞ ð71:5Þ
j¼1 j¼1
That means that j is Eigenvalue of the variable level of p.
For example, if j0 ¼ 1; j ¼ 1:9, that means it belongs to the first lever tend to
the second lever, or rather, it is closer to the second lever (Guo 2006). So we can
see the extent it biased in favor of the other. Now we make j1 j2 as Eigenvalue of
the variable level of steel structure and assemble type concrete structure.
By Eq. (71.5) can draw j1 ¼ 1:7128, j2 ¼ 1:5923. Because the Eigenvalue of
the variable level of assemble type concrete structure is closer to 1, so it biased in
favor of the lever excellent higher. Therefore, We can draw the conclusion that the
assemble type concrete structure is superior to others.
684 S. Yu and X. Liu
71.5 Conclusions and Implications
The paper first calculates the weigh of indexes by AHP-LSDM, then make a
comprehensive assessment on three types of structure systems by Matter-element
model, and finally finds that the assemble type concrete structure is the better
choice for its advantages and promotion of residential industrialization. The
research is helpful to reduce the blindness of invest and promote the research and
development of residential construction system.
References
Baldwin CY, Kim BC (1997) Managing in an age of modularity. Harvard Bus Rev 75(5):84–93
Baldwin CY, Kim BC (2000) Design rules: the power of modularity. Cambridge MIT Press,
Cambridge
Bi Jl (2008) Research on residential industrialization in China. Chongqing University, Chongqing
Guo F (2006) Application of the CS theory in real estate industry. Sci Technol Manag
2006:15–17
Huang Y, Zhu J (2009) Based on the matter-element model of the urban transportation evaluation
and empirical research. Syst Eng 27(2):79–84
Jia H, Wu X, Li H (2010) Simple discuss project management information. Proj Manag Technol
2010(8):86–89
Lei Zy, Chen W (2010) Applications research in performance evaluation of project management
information based on matter-element. Proj Manag Technol 2010(8):86–89
Li Y (2008) Customer satisfaction strategy in the application of residential. Bus Res
2008(9):70–723
Li P, Zhang X, Zhang J (2007) Empirical study on the driving factors of [14] real estate industry
customer satisfaction. J Hunan Univ 21(6):50–54
Liu S (2010) Grey system theory and application. Science Press, Beijing
Mi S, Jia H, Wu X, Li H (2010) Simple discuss project management information. Proj Manag
Technol 2010(8):86–89
Porter ME (1985) Competitive advantages: creating and sustaining superior performance. The
Free Press, New York, pp 33–61
Porter M (1990) The competitive advantage of nations. Harvard Bus Rev 68(2):74
Zhang F (2010a) Project management information construction problems and countermeasures.
Theor Invest 13(5):231–238
Zhang Z (2010b) Project management information to the development trend of the research.
China Build Inf 16(14):48–51
Chapter 72
Logistic Financial Crisis Early-Warning
Model Subjoining Nonfinancial Indexes
for Listed Companies
Shao-fang Ding, Ying-chao Hou and Pei-pei Hou
Abstract The occurrence of financial crisis is related with financial factors. Many
nonfinancial ones also contain important information relevant to the occurrence of
financial crisis. If merely financial factors are taken into consideration, much
useful information will be lost. Thus, the early-warning capacity of the model will
be reduced. What’s more, we will fail to learn the cause for the occurrence of
financial crisis at a more profound level. It is imperative to draw nonfinancial
index into the study of financial crisis early-warning and build a more effective and
more complete financial crisis early-warning model. The paper introduces not only
financial index, but also nonfinancial index including enterprise ownership struc-
ture, corporate governance, and major item, etc, while it takes a preliminary
identification and screening about the study sample, paired sample and early-
warning indicators. Then we set up enterprise’s financial crisis early-warning
model to complete the warning index system.
Keywords Financial crisis early-warning Nonfinancial indexes Logistic
regression Factor analysis
S. Ding (&)
Beijing Polytechnic, Beijing, China
e-mail: [email protected]
Y. Hou
Department of Economics and Management, NCUT, Beijing, China
e-mail: [email protected]
P. Hou
Department of Foreign Languages, Xi’an Jiaotong University, Xi’an, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 685
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_72,
Ó Springer-Verlag Berlin Heidelberg 2013
686 S. Ding et al.
72.1 Introduction
In previous papers, in the application of nonfinancial index, scholars study more
about relationship between nonfinancial index and company performance. Even in
financial crisis early-warning model including nonfinancial index, they apply
mainly the financial indexes without a comprehensive nonfinancial index system,
while they introduce just a few nonfinancial indexes, as there is only a single
ownership structure or corporate governance. However, this paper introduces
ownership structure, corporate governance situation, and major item into the early-
warning model, according to different samples, data characteristics, to achieve
better warning effect.
72.2 The Selection of Financial Crisis Early-Warning
Sample and Index
72.2.1 The Preliminary Determination of Samples and Data
This paper will take listed companies who have received ST for operation in China
securities market A-shares as samples. 87 companies in total in Shanghai Stock
Exchange and Shenzhen Stock Exchange are chosen: 25 ones who were the first to
receive ST in 2007, 34 ones who received ST in 2008, and 28 ones who received
ST in 2009. The financial and nonfinancial index information of those listed
companies in the three years before ST is used to forecast whether they are
financial crisis companies.
In order to find out the early-warning index which has an impact on ST
companies by comparing ST companies with non-ST companies, this paper also
chooses 87 non-ST companies by the ratio of 1:1 as paired samples. To guarantee
the consistency and comparability with the original sample data, the paired
samples are in the same or similar industry and in the similar asset size with the
original ones while the same last three years’ information is used as study object.
Sample data come from Wind, CSMAR and RESSET Databases.
This paper classifies early-warning indexes into financial ones and nonfinancial
ones. On the basis of previous research, 31 indexes are chosen, according to the
principle of sensitivity, accuracy, representativeness and comprehensiveness.
Among those indexes, there are 16 financial indexes, selected in accordance with
debt-paying ability, operating capacity, earning power and development capacity.
The other 15 indexes are nonfinancial ones, selected by shareholding structure,
corporate governance, significant matters and other factors. See Table 72.1.
Table 72.1 Early-warning index
72
Index Evaluation items Index Computational formula
classification
Financial Debt-paying ability Current ratio X1 Current assets/current liabilities
Index Quick ratio X2 (Current assets—inventory)/current liabilities
Debt asset ratio X3 Total indebtedness/total assets
Working capital total assets ratio Working capital/average total assets
X4
Operating capacity Receivables turnover ratio X5 Main business income/average receivables
Inventory turnover ratio X6 Cost of goods sold/average inventory
Current assets turnover ratio X7 Main business income/average current assets
Total assets turnover ratio X8 Main business income/average total assets
Earning power Main business Profit rate X9 Main business profit/main business income
Net profit rate to total assets X10 Net profit/average total assets
Net profit margin on sales X11 Net profit/main business Income
Profit margin on net assets X12 Net profit/average net assets
Logistic Financial Crisis Early-Warning Model
Development capacity main business’s increasing rate of (Main business income of this year—main business income of last
income X13 year)/main business income of last year
Rate of capital accumulation X14 Growth of owner’s equities this year/owner’s equities at the
beginning of the year
Increasing rate of net assets X15 (Net assets of this period—net assets of last period)/net assets of
last period)
Increasing rate of total assets X16 (Total assets of this period—total assets of last period)/total assets
of last period
(continued)
687
Table 72.1 (continued)
688
Index Evaluation items Index Computational formula
classification
Nonfinancial Shareholding Shareholding proportion of theThe ratio of the shares of the controlling shareholder to the total
index structure (Tan and Zhang 2005) controlling shareholder Y1 shares of the company
Herdindhal_5 Index Y2 Sum of the squares of the first five substantial shareholders in the
company’s shareholding proportion
Z IndexY3 Shareholding proportion of the first substantial shareholder/
shareholding proportion of the second substantial shareholder
CR_5 Index Y4 Sum of the shareholding proportion of the first five substantial
shareholders in the company
Corporate governance (Deng and Ratio of independent director Y5 Independent director/all director
Wang 2006; Wang and Ji 2006) Ratio of state shares Y6 The amount of state shares/capitalization
Ratio of upper management shares (Shareholding of board of directors +shareholding of management
Y7 layer) /capitalization
Position set of chairman of the If the chairman of the board and the general manager is the same
board and general manager Y8 person, score 1; otherwise, score 0.
Significant matters (Lv 2006) Whether involving related party If yes, score 1; otherwise, score 0.
Transaction Y9
Whether having violation record If yes, score 1; otherwise, score 0.
Y10
Whether involved in Lawsuit or If yes, score 1; otherwise, score 0.
Arbitration Y11
Whether involving external If yes, score 1; otherwise, score 0.
guarantees Y12
Other Factors (Wan and Wang Whether having changed If yes, score 1; otherwise, score 0.
2007; Yang 2008) accounting firms Y13
Whether having altered If yes, score 1; otherwise, score 0.
abbreviation Y14
Type of audit opinion Y15 If the auditor presents standard clean opinions, score 1; otherwise,
score 0.
S. Ding et al.
72 Logistic Financial Crisis Early-Warning Model 689
72.2.2 The Preliminary Selection of Early-Warning Indexes
72.2.2.1 Normal Distribution Inspection
First, we take normal distribution test of these primary early-warning indexes. By
means of the K–S test method in SPSS statistical software, we test the 31 primary
indexes selected from the samples of two group.
K statistic ¼ maxðjSðXi Þ FðXi ÞjÞ
In the formula, S(Xi) is the actual cumulative probability value of each different
observing samples, of which F(Xi) is the theoretical value. Under significant level
of a = 0.05, the bilateral progressive probability’s P value of X3, X13, X16, Y1, Y2,
Y5, Y12 is greater than 0.05, which means the 7 of them pass the inspection and
overall accord with the normal distribution. By two independent samples test
method of significant test, the rest 24 indexes do not accord with normal distri-
bution, so we use a nonparametric test, Mann–Whitney test, to test their
significance.
72.2.2.2 T test of Two Independent Samples
The equation of T (Zhang and Cheng 2004) statistic:
1 X
X 2 ðl1 l2 Þ
T¼ qffiffiffiffiffiffiffi
r212
The results are as follows:
Under significant level of a = 0.05, X3, X13, Y1, Y2 pass the T test, which
means that these 4 indexes have significant differences, while X16, Y5, Y12 does
not, which means they have no significant differences.
72.2.2.3 U-Test of Two Independent Samples
This paper selected the most effective alternative method of parameter test, Mann–
Whitney test.
U-test equation (Liao et al. 2008) is as followed:
mðm þ 1Þ Xm
Uxy ¼ mn þ R;
i¼1 i
2
nðn þ 1Þ X n
Uyx ¼ mn þ R;
j¼1 j
2
690 S. Ding et al.
The test results are as followed: the indexes’ values of X1, X2, X4, X8, X9, X10,
X11, X12, X14, X15, Y4, Y8, Y10, Y11, Y13, Y14, Y15, 17 in all, are smaller than the
significant level, while the other 7 indexes do not pass the significant test.
In general, a total of 21 indexes pass the significant test.
72.2.3 A Further Integration of Early-Warning Index
The tests of significance above identify 21 early-warning indexes, including 12
financial indexes and 9 nonfinancial indexes. Both non- and financial indexes
reflect a company’s financial performance. Accounting for that these early-warn-
ing indexes may be relevant between each other, the paper convert the multiple
observable variables into a few uncorrelated integrated indexes, by the method of
factor analysis, to best simplified the high dimension data. Because some of the
nonfinancial early-warning indexes are virtual variables, of which the data is not
continuous, they can not be integrated.
72.2.3.1 KMO Test (Yang 2007)
We take a KMO test before factor analysis to determine whether the financial
ratios involved are suitable for it (Table 72.2).
By KMO test, the results show 0.729 of KMO test coefficient, indicating a high
relevant between the indexes, so they are suitable for factor analysis. The 744,202
of Bartlett Chi square value and 0.000 \ 0.05 of P value show the 12 financial
indexes are not independent and there is a certain relationship between them.
72.2.3.2 Factor Analysis (Gui and Wu 2007)
We screen out 12 financial indexes by significance test above: X1, X2, X3, X4, X8,
X9, X10, X11, X12, X14, X15. Then we take factor analysis on these 12 indexes, and
find that, the characteristic values of first 4 common factor are greater than 1, and
their accumulated contribution rates reach 84.819 %, recorded as F1, F2, F3, F4.
To explain them reasonable, we need to get the correlation coefficients between the
4 common factor and the 12 initial financial indexes. So the paper uses orthogonal
rotation maximum variance method to do the conversion, and gets the factor
loading matrix as followed:
Table 72.2 KMO test
KMO and Bartlett test
Enough sampling Kaiser–Meyer–Olkin test 0.729
Bartlett sphericity test Chi square 744.202
df 66
Sig. 0.000
72 Logistic Financial Crisis Early-Warning Model 691
Table 72.3 Factor loading matrix
Rotated factor matrix
Factor
F1 F2 F3 F4
Liquidity ratioX1 0.043 0.943 0.057 -0.049
Quick ratioX2 0.064 0.910 0.065 -0.062
Asset liability ratioX3 -0.292 -0.672 0.396 0.072
Working capital to total asset ratioX4 -0.044 0.733 -0.096 0.026
Total assets turnover ratioX8 0.789 0.232 0.273 0.107
Main business profit rateX9 0.890 0.161 -0.196 0.029
Total net asset profit rateX10 0.859 0.091 0.298 -0.061
Sales net profit rateX11 0.942 0.107 0.033 -0.033
Net assets income rateX12 0.231 0.187 0.594 0.090
The growth rate of main businessX13 0.256 -0.098 0.112 0.645
Capital accumulation rateX14 0.075 -0.017 0.006 0.962
Net asset growth rateX15 0.383 0.212 0.733 0.166
Extraction method: principal component analysis method
rotation method: Kaiser standardized orthogonal rotation method
From the factor loading matrix after rotation above, we can see that the 4 factor
variances respectively yield high load capacity in different index variables.
According to the factors’ load distribution, we can make a further analysis as
followed (Table 72.3):
(1) Index factor load capacity of F1 on X8, X9, X10, X11, is far greater than that of
other indexes. It shows the company’s operating profit level and the ability.
(2) Index factor load capacity of F2 on X1, X2, X3, X4, is far greater than that of
other indexes. It shows the company’s solvency.
(3) Index factor load capacity of F3 on X12, X15 is far greater than that of other
indexes. It shows the company’s profitability and growth ability.
(4) Index factor load capacity of F4 on X13, X14, is far greater than that of other
indexes. It shows the company’s ability to grow.
By calculating the coefficients in the linear combination of common factors, as
dependent factors, and initial index variables, as the independent factors, we get
the initial linear expression as followed:
F1 ¼ 0:048x1 0:037x2 0:038x3 0:074x4 0:228x8 þ 0:271x9 þ 0:293x10 þ 0:307x11 þ 0:078x12 þ 0:176x13 0:057x14 þ 0:068x15
F2 ¼ 0:314x1 þ0:301x2 0:194x3 þ 0:250x4 þ 0:183x8 0:013x9 0:028x10 0:033x11 þ 0:063x12 0:079x13 þ 0:029x14 þ 0:035x15
F3 ¼ 0:065x1 þ0:070x2 þ 0:291x3 0:061x4 þ 0:236x8 0:116x9 þ 0:269x10 þ 0:064x11 þ 0:487x12 þ0:051x13 þ 0:048x14 þ 0:554x15
F4 ¼ 0:090x1 0:025x2 þ 0:084x3 þ 0:056x4 þ 0:130x8 0:050x9 0:118x10 0:108x11 þ 0:101x12 þ0:626x13 þ 0:930x14 þ 0:106x15
692 S. Ding et al.
72.3 The Construction of Logistic Financial Crisis
Early-Warning Model
72.3.1 The Construction of Logistic Model Based
on Financial Indexes Alone
In the construction of Logistic financial crisis early-warning model based on
financial indexes alone, the previous three years’ data of the 44 ST companies and
44 non-ST companies are taken as original data and F1, F2, F3, F4 as dependent
variables. Multiple Logistic regression is employed to do the analysis. The
regression results are presented in Table 72.4.
The above chart illustrates that the coefficient of every explanatory variable is
significant when it is a ¼ 0:1, which implies that the model fits well. Hence, the
company’s Logistic financial crisis early-warning model based on financial
indexes alone in the year T is:
1
P¼
0:730 4:261F1 0:748F2 0:400F3
1 þ exp
0:687F4
72.3.2 The Construction of Logistic Model Injecting
Nonfinancial Indexes (Chen 1999; Wu 2001)
Conduct regression analysis with the four common factors F1, F2, F3, F4 obtained
by factor analysis and the nine nonfinancial index variables Y1, Y2, Y4, Y8, Y10,
Y11, Y13, Y14, Y15, which have been through parameter T test and non-
parameter U test. Through forward gradual selection variables method, the
synthetical early-warning model based on both financial and nonfinancial indexes
is constructed. The regression results are presented in Table 72.5.
The above chart illustrates that the coefficient of every explanatory variable is
significant when it is a = 0.05, which implies that the model fits well. Through the
Table 72.4 The logistic regression results based on financial indexes alone
Variables in equation
B S.E. Wald df Sig. Exp (B)
Step 1a F1 -4.261 1.069 15.884 1 0.000 0.014
F2 -0.748 0.343 4.749 1 0.029 0.473
F3 -0.400 0.369 1.172 1 0.079 0.670
F4 -0.687 0.624 1.212 1 0.071 0.503
Constant -0.730 0.363 4.042 1 0.044 0.482
a
inputting Variables F1, F2, F3, F4 in step 1
72
Table 72.5 The regression results of the Logistic synthetical model injecting nonfinancial indexes
Variables in equation
B S.E. Wald df Sig. Exp (B)
Step 3a F1 -3.219 1.094 7.390 1 0.003 0.039
F2 -2.114 1.168 3.908 1 0.023 0.121
F3 -2.103 1.612 6.367 1 0.014 0.122
F4 -1.601 1.701 9.948 1 0.005 0.202
Shareholding Proportion of the Controlling Shareholder Y1 3.437 1.806 6.312 1 0.006 31.094
Logistic Financial Crisis Early-Warning Model
CR_5 Index Y4 -2.108 1.236 7.836 1 0.018 0.121
Whether having Violation Record Y10 3.262 1.155 7.975 1 0.005 26.102
Whether involved in Lawsuit or Arbitration Y11 3.285 1.019 10.404 1 0.001 26.709
Whether having altered abbreviation Y14 3.923 1.201 10.678 1 0.001 50.552
Type of Audit Opinion Y15 -2.888 1.098 6.915 1 0.009 0.056
Constant 1.130 1.024 1.217 1 0.007 3.095
693
694 S. Ding et al.
coefficients of the variables in the chart above, the Logistic financial crisis
synthetical early-warning model injecting nonfinancial indexes is obtained:
1
P¼ 2 0 13
1:130 3:219F1 2:114F2 2:103F3
6 B 1:601F4 þ 3:437Y1 2:108Y8 C7
1 þ exp6 B
4@
C7
A5
þ3:262Y10 þ 3:285Y11
þ3:923Y14 2:888Y15
From the above synthetical early-warning model, one can see that it is positive
correlation between the nonfinancial index variable Shareholding Proportion of the
Controlling Shareholder Y1 and the occurrence of financial crisis probability P,
which implies that the higher the shareholding proportion of the controlling
shareholder, the greater the probability of financial crisis. It is negative correlation
between the nonfinancial index variable CR_5 index Y4 and P, which indicates
that the higher the shareholding proportion of the first five substantial shareholders
and the ownership concentration, the less the probability of financial crisis.
Meanwhile, if the company is involved in violation record, lawsuit or attribution
and abbreviation alteration, the probability of financial crisis will be further
greater.
72.4 The Test of the Early-Warning Model
72.4.1 The Test of Logistic Early-Warning Model
Based on Financial Indexes Alone
Since the ratio between the original samples and the paired samples is 1:1, hence 1
is to represent companies with financial crisis while 0 is to represent companies
without financial crisis. P = 0.5 is taken as discriminating section ratio. If P [ 0.5,
it is marked as company with financial crisis; if P \ 0.5, it is marked as company
with normal financial condition.
Input the index variable data of the 86 companies in the testing samples,
consisting of 43 ST listed companies and 43 non-ST listed companies, into the
early-warning model based on financial indexes alone to test the model’s veracity.
Testing results are illustrated in Table 72.6.
The above chart shows that the constructed early-warning model based only on
financial indexes is able to discriminate accurately 32 ST companies and 35 non-
ST companies, taking P = 0.5 as predicted discriminating point and the actual 43
ST listed companies and 43 non-ST ones as testing samples. In other words, the
accuracy rates of the early-warning model based only on financial indexes to the
prediction for the ST companies and non-ST ones respectively are 74.42 and
81.39 %. the average percentage is 77.91 %.
72 Logistic Financial Crisis Early-Warning Model 695
Table 72.6 Testing results of the Logistic model based on financial indexes alone
Classification tablea
Observed value Predicted value
Group Accuracy rate Misjudgment rate
(%) (%)
ST Non-ST
company company
Group ST company 32 11 74.42 25.58
Non-ST 8 35 81.39 18.61
company
Total percentage 77.91 22.09
a
Discriminant Piont.500
72.4.2 The Test of the Logistic Synthetical Early-Warning
Model Injecting Nonfinancial Indexes
Input the index variable data of the 86 companies in the testing samples, consisting
of 43 ST listed companies and 43 non-ST listed companies, into the Logistic
synthetical early-warning model based both on financial and nonfinancial indexes
to test the model’s veracity and compare the testing results of the two models.
Testing results are illustrated in Table 72.7.
From the above chart one can see that the constructed Logistic synthetical
early-warning model injecting nonfinancial indexes is able to discriminate accu-
rately 35 ST companies and 37 non-ST companies, taking P = 0.5 as predicted
discriminating point and the actual 43 ST listed companies and 43 non-ST ones as
testing samples. Thus, the accuracy rates of the Logistic synthetical early-warning
model injecting nonfinancial indexes to the prediction for the ST companies and
non-ST ones respectively are 81.39 and 86.05 %. The average predicting
percentage is 83.72 %.
Table 72.7 Testing results of the Logistic synthetical early-warning model injecting nonfinan-
cial indexes
Classification tablea
Observed value Predicted value
Group Accuracy rate Misjudgment rate
(%) (%)
ST Non-ST
company company
Group ST company 35 8 81.39 18.61
Non-ST 6 37 86.05 13.95
company
Total percentage 83.72 16.28
a
Discriminant Piont.500
696 S. Ding et al.
By comparing the testing results of the two models, one can see that after
drawing nonfinancial index variables in, model’s accuracy rate increases by
5.81 %, which manifests that it enhances effectively the predicting accuracy rate of
the model to draw nonfinancial index into the study of financial crisis early-
warning.
References
Chen J (1999) Empirical analysis of listed company financial deterioration prediction. Acc Res
6:31–38
Deng X, Wang Z (2006) Financial distress prediction from the nonfinancial perspective. Manag
Sci (3):71–80
Gui M, Wu S (2007) Financial distress model study of nonfinancial perspective. Financ Econ
(22):132–133
Liao Y, Zhang L, Liu L (2008) Empirical study of financial early warning based on nonfinancial
and financial information. Mod Manag Sci 4:57–59
Lv J (2006) An empirical study of financial distress and symptom analysis based on nonfinancial
indicators—from manufacturing listed companies. J Grad Sch Chin Acad Soc Sci 2:52–58
Tan Y, Zhang L (2005) Research of bankruptcy prediction method subjoining nonfinancial
variables. Sci Technol Ind 5(10):31–34
Wan X, Wang Y (2007) Fuzzy warning model research for financial crisis of enterprise based on
nonfinancial index. J Manag 4(2):195–200
Wang K, Ji M (2006) Company in deficit finance early warning study based on the financial and
nonfinancial index. J Financ Econ 32(7):63–72
Wu S (2001) Financial distress prediction model research of our listed companies. Econ Res
6:46–55
Yang H (2007) Nonfinancial index application research in financial crisis early warning model.
Acc Commun (Compr Ed) 5:31–32
Yang Y (2008) Review and evaluation of the selection of nonfinancial index in early warning
research. Acc Commun 6:100–101
Zhang M, Cheng T (2004) Audit opinion’s information content in early warning. Acc Commun
12:47–48
Chapter 73
Evaluation Research on Logistics
Development of the Yangtze River Port
Based on the Principal Component
Analysis
Gao Fei
Abstract This article analyzes the significance of port logistics as well as factors
influencing the development of Yangtze River port logistics. On this basis, a
scientific evaluation system of the Yangtze River port logistics development and a
principal component analysis model of the port logistics development level
evaluation are established. Taking the ports group along the Yangtz river in Anhui
province as an example, this article justifies the validity of the river port logistics
development level evaluation system.
Keywords Ports along the Yangtze River Port logistics Evaluation system
Principal component analysis
At present, related research on port logistics evaluation has become one of the
focuses of the theory. Many scholars have done a lot of work in this field, such as
Cao Weidong, Cao Wave, Wang Ling, Wei Ran etc. Some use a specific object for
the evaluation and analysis of the port logistics system. However, most researches
focus on the application of modeling methods while paying little attention to the
evaluation index system. In addition, inaccurate understanding of the port logis-
tics’concept leads to a one-sided evaluation index system, which to some extent
affects the evaluation result. Combined with previous research, this paper attempts
to discuss the connotation of the port logistics, build a relatively reasonable river
port logistics evaluation index system on the basis of analyzing influencing factors
of port logistics’ developmental level, and conduct a case study of ports along the
Yangtze river through evaluation model by applying the principal component
method (Xu 2004).
G. Fei (&)
Anqing Vocational and Technical College, Anqing 246003 Anhui, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 697
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_73,
Springer-Verlag Berlin Heidelberg 2013
698 G. Fei
73.1 The Meaning of Port Logistics
Port logistics refers to that the center port cities make use of its own port’s
advantages, relying on advanced hardware and software environment, to
strengthen its radiation ability in logistics activities around the port, highlight the
port set of goods, inventory, distribution features, with harbor industry as the basis,
with information technology as the support, aim at integration of port resources
and develop a comprehensive port service system covering the features of all links
in the industry chain of logistics. Port logistics is a special form of the integrated
logistics system and also is an irreplaceable and important node, which completes
the basic logistics service and the value added service to supply the whole chain
logistics system (Play 1995).
73.2 The Construction of Evaluation Index System
73.2.1 Factors Influencing the Port Logistics’
Developmental Level
Port logistics’ developmental ability along the Yangtze River: logistics’ devel-
opmental ability reflects the existing development of port logistics’ capability,
based on their own advantages and competitive resource, its outcome and the
status of the past and present logistics market’s development. The port logistics
development ability can be reflected from the logistics’ infrastructure equipment,
haven dimensions, informatization level, the standardization of logistics and port’s
developmental level.
The river port logistics developmental environment: port logistics develop-
mental environment is an extrinsic factor for measuring the port’s logistical
development, and is the guarantee of the present developmental ability and the
cultivation basis of potential development. Port’s overall environment has very
important influence on the development of logistics. For instance, logistics ser-
vices and hinterland economic development level will have direct impact on
logistics service demand and growth potential. Port logistics developmental
environment usually depends on the economic environment, policy environment,
human resource environment and so on (Han and Wang 2001).
The port logistics’ capability of sustainable development: the sustainable
development capacity of port logistics is a measure of port logistics’ subsequent
development ability. Logistics sustainable development must be in accordance
with the carrying capacity of nature. Only by guaranteeing the sustainability of
resource and ecology can we make the sustainable development of logistics pos-
sible. This requires that in the pursuit of logistics development, we must pay
73 Evaluation Research on Logistics Development of the Yangtze River Port 699
attention to protection of environment and resources to ensure sustainable use of
resources. The sustainable utilization of resources and good ecological environ-
ment are important symbols of the logistics’ sustainable development (Han and
Wang 2001).
73.2.2 The Index System of Port Logistics Evaluation
Based on the analysis of the port logistics’ connotation and its influencing factors
above, this paper divides port logistics’ evaluation index system into three levels:
the first level is the target level (Mao 1996), namely the evaluation of river port
logistics development level; the second level is an first-class indicator. Based on
the analysis of factors influencing the port logistics, it establishes two first-class
indicators, which are logistics developmental ability and environment and influ-
ence of logistics development respectively; the third level is the second-class
indicator. This is the core part of the index system as well as the operable indi-
cator’s component. This article identifies 14 two-level index according to the three
factors influencing the port logistics system while considering the theoretical and
practical possibility (Han and Micheline 2001) (Table 73.1):
Table 73.1 14 two-level index according to the three factors influencing the port logistics
system
Target layer Level indicator Two level index
The port logistics Logistics development V1 Waterfront line length (KM)
development V2 Berth number
level of V3 Cargo throughput
V4 The port number of employees
V5 The level of public information
platform
V6 Logistics standardization level
V7 Profit ability
V8 The level of logistics services
V9 Investment in fixed assets (Million
yuan)
Logistics development V10 Hinterland economy GDP (Billion
environment and yuan)
influence V11 Total retail sales of consumer goods in
the hinterland of (Billion yuan)
V12 Hinterland trade (Billion yuan)
V13 Policy environment
V14 College school student number
700 G. Fei
73.3 The Case Study of the Port Logistics Development
Level Evaluation
The principal component analysis is used to calculate the comprehensive strength
index of the port logistics. Its basic principle and steps are: assumption on the N
port logistics development in M indexes of comprehensive evaluation, the index
set of matrix is Xij (I = 1, 2,…, N; J = 1, 2,…, m) (Tian 2000). In order to
eliminate the effects of different classes of magnitude, it is necessary to standardize
the original matrix, form a new index set matrix Yij, calculate Yij correlation
coefficient matrix R, obtain eigenvalue ki through inner product vector of R, get
the corresponding feature vector Iij by solving inverse compact transform, the
characteristic value of the cumulative percentage determination of P components
Zi, and for each principal component contribution rate Pj; calculation of the I port
logistics of each principal component scores of Zij; finally, calculated for each port
logistics development comprehensive strength index Qi (Nevem Working Group
1989).
X
p
Qi ¼ pj Zij ði ¼ 1; 2; . . .; nÞ ð73:1Þ
j¼1
The value of Qi is in proportion to the port logistics’ development level of i
years, and vice versa. By that analogy, we can calculate the comprehensive
strength index of various port logistics development (Lu Avenue 1988).
Anhui port group includes 5 major ports like Ma’anshan, Wuhu, Tongling,
Chizhou and Anqing. The port system has unique geographical advantages and
important strategic position. With the promotion of Anhui along-river area’s
openness and development as well as the demand of accepting industrial transfer,
various regions implement the strategy of ‘‘port prospers city’’ one after another
and increase haven infrastructure construction. Some areas rely on their regional
advantages of port to develop port logistics in order to realize port city’s sustained
and rapid economic development, along with the specific characteristics of the port
(Xu 1998).
73.3.1 Computational Results Analysis
Choosing four years, 2000, 2003, 2007, 2010 from 2000 to 2010, according to the
‘‘China City Statistical Yearbook’’ (2001, 2004, 2008 and 2011), ‘‘statistical
yearbook of Anhui province’’ (2001, 2004, 2008 and 2011) and the Anhui Yangtze
River 5 ports’ statistics report (Chun 2001), through the establishment of 5 Port
Logistics comprehensive strength evaluation index database, using SPSS13.0
software, to analyze and process the data. According to standard extracting main
factor with factor eigenvalue greater than 1 and the cumulative contribution rate
73 Evaluation Research on Logistics Development of the Yangtze River Port 701
Table 73.2 Anhui River Port Logistics comprehensive strength index along the Yangtze (Qi)
2001 2004 2007 2010
Ma’anshan 6.82711 19.57205 5.32404 12.74210
Wuhu, 69.977337 87.99985 67.10115 91.96178
Tongling -18.49716 -25.48272 -45.71171 -55.06373
Chizhou -61.99780 -55.49272 -30.70364 -5.56749
Anqing 40.69597 50.40360 18.99016 61.92710
more than 85 %, and according to the formula (73.1), comprehensive strength
index of the port logistics (Table 73.2) for different years are calculated (Xiao and
Han 2001). Qi represents Port Logistics comprehensive strength index.
73.3.2 Analysis of Port Logistics Development Level
From Table 73.2 we can see that Port Logistics comprehensive strength index of
Wuhu harbor, Ma’anshan port and Anqing harbor is always positive, indicating that
logistics development level in the area of the port logistics development is above the
average (Foster 1992); Chizhou Port Logistics comprehensive strength index has
always been negative, which indicates that the port logistics development level
has always been below the average; the development of Tongling port logistics has
obvious ups and downs, with comprehensive strength index turning from
-18.49716 in 2001 into 55.06373 in 2010, and it continues to be negative, sug-
gesting that the port logistics development has been below the average (Helen 1992).
73.3.3 Analysis of Port Logistics Development
and Performance
According to the comprehensive strength index of the port logistics development
trend, Anhui port logistics development is divided into the following types: (1)
rising type, mainly represented by Chizhou port logistics. Chizhou Port Logistics
comprehensive strength index increases from -61.99780 in 2001 into -5.56749 in
2010, a markable rise; (2) declining type, mainly represented by Tongling port
logistics (Thompson 2002). Tongling Port Logistics comprehensive strength index
drops from 18.49716 in 2001 into -55.06373 in 2010, big dropping range; (3)
fluctuations, including Ma’anshan, Wuhu and Anqing port logistics. Ma’anshan
Port Logistics comprehensive strength index rise from 6.82711 in 2001 into
19.57205 in 2004, reaching a peak, and then falling to 12.74210 in 2010; 2001
Wuhu Port Logistics comprehensive strength index is 69.977337, reaching the
peak of 91.96178 in 2001, while the development of port logistics level will
always be in the first place; Anqing Port Logistics comprehensive strength index
702 G. Fei
rank second in 2001, and its value is 40.69597, with its index 18.990162010 falling
to its low ebb. In 2010, its port logistics comprehensive strength index rise to
61.92710, ranking second.
On the port logistics development level evaluation system, we should first pay
attention to the research of evaluation index system (Chames et al. 1978). Only
with an in-depth analysis of the influencing factors of port logistics system
established on the basis of scientific and reasonable index system can we conduct
further evaluation. At the same time, we should also take the development level of
hinterland economy as the important evaluation index (Saul and Adam 1999).
References
Cao Play (1995) Preliminary study of the port system along the Yangtze river in Anhui province.
Geogr Sci 15(2):154–162
Chames A, Cooper W, Rhods E (1978) Measuring the efficiency of decision making units.
European J Opt Res (6):429–444
Cloud Chun (2001) The Development of the port and into the transformation of the logistics
center. Port Handl 4:23–25
Eiichi Taniguchirusellg Thompson (2002) Modeling city logistics. J Transp Res Board (1):45–51
Thomas A Foster (1992) Logistics benchmarking: searching for the best. Distribution (3):31–36
Han JW, Micheline K (2001) The data mining: concepts and techniques. In: Fan X (ed.) Meng
translation. Mechanical Industry Press, Beijing, pp 76–77
Han ZL, Wang G (2001) Port logistics characteristics and influencing factors. Chinese Ports
(8):38–40
Han ZL, Wang G (2001) Port logistics characteristics and influencing factors. Dalian Port Ocean
Dev Manag (4):39–42
Helen R (1992) Improve quality through benehmarking. Transp Distribut (10):l2–20
Lu Avenue (1988) Location theory and regional research methods. Science Press, Beijing
Mao HY (1996) Shandong province sustainable development indicator system. Geography
15(4):16–23
Nevem Working Group (1989) Performance indicators in logisticsf. IFS Publication, Bedford,
pp 36–39
Saul E, Adam R (1999) Enterprise performance and ownership: the case of Ukraine. European
Econ Rev (4–6):1125–1136
Tian Yu (2000) Logistics efficiency evaluation method. Logist Technol 2:34–36
Xiao P, Han ZL (2001) Coming of age of integrated logistics and port function of the evolution.
Trop Geogr (3):41–43
Xu JW (2004) Port logistics development. World Ocean 27(2):31–32
Xu Shubo (1998) Analytic hierarchy principle. Tianjin University Press, Tianjin, China
Chapter 74
A Game Analysis of New Technical
Equipment Procurement
Ai-hua Zhang, Zhi-wei Yang, Qiong Zhang
and Hong-wei Fu
Abstract With the rapid development of modern science and technology,
equipment construction plays an increasingly active role in Chinese military
modernization construction. Given the limited funds input in the military equip-
ment and the pressing need to strengthen the funds management, the paper ana-
lyzes the game process of the new technical equipment procurement and put
forwards some scientific suggestions, which helps to win a priority, improve the
benefits in the procurement game, and finally realize the optimization of the source
distribution.
Keywords New technology Equipment Procurement Game
74.1 Introduction
Equipment construction plays an increasingly active role in Chinese military
modernization construction With the rapid development of modern science and
technology. Equipment procurement, the critical link in the entire life cycle of
weapon equipment, is restrained by equipment funds input and benefits (Zhao et al.
2011). Serious contradiction between military expenditure and demand leads to
limited funds pouring into equipment construction. Given this condition,
strengthening the management of the funds and analyzing the Procurement Process
of the new technical equipment help to win a priority and improve the benefits in
the procurement game (Xie et al. 2011; Xiang and Xin 1997).
Equipment purchase expenditure is the money the purchaser pays the supplier
for the equipment (International Society of Parametric Analysts 2007). For the
A. Zhang (&) Z. Yang Q. Zhang H. Fu
Basic Courses Department, Military Economics Academy, Wuhan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 703
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_74,
Ó Springer-Verlag Berlin Heidelberg 2013
704 A. Zhang et al.
convenience of expression, Party A and Party B may be applied somewhere in the
following text to represent the purchaser and the supplier respectively. Party A
expects to purchase the equipment they need as cheaply as possible, while Party B
hopes to sell it at the best possible price. Then the contradiction between the two
Parties appears, namely, each Party becomes one of the main roles in the game of
business or bargaining (Aghion and Bolton 1992; Weitzman 1980).
The following three circumstances usually occur during the process of equipment
procurement: mature technology of producing equipment and a single supplier; `
mature technology of producing equipment and more than one suppliers; ´ imma-
ture technology of producing equipment and no available supplier. Under the first
two circumstances, the essence of the concern both parties focus on is price (Hou and
Wang 2001). This article highlights the third circumstance, under which, not only
price but also risk control should be seriously considered. Lacking mature technol-
ogy, the supplier need to pour an amount of money into research. There is some
uncertainty about how much money should be poured into research, whether the new
equipment can be developed, and if so, whether the function of the newly-developed
equipment meet the requirement of the purchaser. So given this condition, The
procuring behavior of the purchaser can also considered as the venture investment.
74.2 The Model of New Technical Equipment Procurement
The major problem of new technical equipment procurement lies in the uncertainty
of its development and manufacture (Hartley 2007). And for military purchasers
and the new equipment suppliers, this uncertainty is characterized by the difficulty
of conquering new technology. Different difficulty leads to different developing
costs, which can not be confirmed before the procurement contract is signed, but
its probabilistic distributions can be estimated (Aliprantis and Chakrabarti 2000).
The concrete costs the new technical equipment developer (Party B) pay out is
their privacy. So for the good of the company, they always claim their high
technical difficulty leads to high costs even with low technical difficulty and low
costs. Then it is so difficult for the purchaser(Party A) to judge whether the cost
information form Party B is true. But Party A can choose different quantity of the
purchase, also their own privacy, to avoid the moral risk of Party B.
Suppose both parties have to reach an agreement in the purchase and sale contract
that the price under high technical difficulty is Ph , and under low technical difficulty
the price is Pl . Then under high technical difficulty, the marginal cost of Party B is
Ch , and purchase quantity of Party A is Qh ; under low technical difficulty the
marginal cost of Party B is Cl , and purchase quantity of Party A is Ql . And,
Ph [ Ch ; Pl [ Cl ; Ql [ Qh
Here, Ch and Cl represent the privacy of Party B, which are the fixed values,
while Qh and Ql are variables in the free charge of A, which has to become fixed
74 A Game Analysis of New Technical Equipment Procurement 705
Table 74.1 Gain matrix of new equipment game based on high technical difficulty I
Party B
Ph Pl
Party A Qh -Ph, Qh (Ph-Ch) -Pl, Qh (Pl-Ch)
Ql -Ph, Ql (Ph-Ch) -Pl, Ql (Pl-Ch)
after the contract is signed to facilitate the development process of Part B. So
before signing the contract, Party A is expected to fix the values of Qh and Ql
under the condition of both the high and low technical difficulties.
There exists the moral risk that Party B may present the low technical difficulty
as high one and vice versa. And at the same time, Party A may present the high
quantity as low one and vice versa (Wang 2011).
Party A, who has the last word on quantity of the purchase, is aiming at
purchasing the equipment they need as cheaply as possible, while Part B, who has
the last word on the price, is hoping to sell the newly developed technical
equipment at the best possible price. That is to say, Party A pursues minimizing
the average cost of the equipment purchase, while Party B pursues maximizing the
selling profits. And the gain matrix game is shown in the following table
(Table 74.1)
In this table, there is no difference to Party A whichever quotation Party B
chooses. However, Party A prefers low quantity given the high quotation, and
prefers high quantity given low quotation. In order to identify the difference in this
preference, the matrix in the table should be modified as follows.
According to the Table 74.2, expecting to get the low price, all Party A could
do is to force Party B to take the initiative in choice of low quotation. The
quotation Party B chooses depends on the their possible profits, which is clearly
shown as follows:
Qh ðPh Ch Þ [ Qh ðPl Ch Þ
Ql ðPh Ch Þ [ Ql ðPl Ch Þ
So, whatever choice Party A makes, Party B will inevitably choose to quote
high price, and there is nothing party A can do to force Party B into preferring low
quotation. That is to say, game equilibrium will be reached under the condition
that Party A requires low quantity and Party B presents high quotation.
Under the circumstances of low technical difficulty, gain matrix game is shown
in the following table.
Table 74.2 Gain matrix of new equipment game based on high technical difficulty II
Party B
Ph PI
Party A Qh -Ph ? D, Qh (Ph-Ch) –Pl, Qh (Pl-Ch)
Ql -Ph, Ql (Ph-Ch) –Pl ? D, Ql (Pl-Ch)
706 A. Zhang et al.
Table 74.3 Gain matrix of new equipment game based on low technical difficulty
Party B
Ph PI
Party A Qh -Ph ? D, Qh (Ph-Ch) -Pl, Qh (Pl-Cl)
Ql –Ph, Ql (Ph-Cl) -Pl ? D, Ql (Pl-Cl)
Like the game of new equipment based on high technical difficulty, this game is
also equilibrated under the condition that Party A requires low quantity and Party
B presents high quotation (Table 74.3).
The main reason why both games are equilibrated based on the same condition
is that whatever choice Party A makes, Party B inevitably chooses high quotation
which is beneficial. Namely, high quotation is the dominant strategy of Party B and
low quotation is his strict dominated strategy.
Under the condition of high technical difficulty, if Pl [ Ch ; Ql ðPl
Ch Þ [ Qh ðPh Ch Þ; the disequilibrium game point that Party A requires high
quantity and Party B presents low quotation: ðPl þ D ; Ql ðPl Ch ÞÞ, is strictly
superior to the equilibrium game point that Party A requires low quantity and Party
B presents high quotation: ðPh þ D ; Qh ðPh Ch ÞÞ. Then, the game between
both parties get stuck in the ‘‘prisoner’s dilemma’’, which stems from the fact that
Part B always pursue the optimal profit which is however considered the worst to
Party A, who has to choose the suboptimal point to improve the unfavorable
condition, so does Party B, and subsequently both sides are bound to reach a
suboptimal equilibrium rather than optimal equilibrium which is unstable.
The same analysis is also applicable to the purchase of new technical equipment
under the condition of low technical difficulty.
To avoid ‘‘prisoner’s dilemma’’ and arrive in the optimal condition, both parties
may reach an agreement beforehand and may sign the following two contracts:
The quotation of Party B is Ph while the purchase quantity of Party A is Qh ; ` The
quotation of Party B is Pl while the purchase quantity of Party A is Ql . And Party
B is allowed to choose either of two, at the same time, Party A may make a
promise during the game that Party A does choose low quantity if Party B prefers
high quotation and vice versa. This promise is made by Party A without risk, just
conveying the message to Party B that those who reap profits at the expense of
others will end up ruining themselves.
74.3 Conclusions
Whether the procurement price is accurate and rational is closely related to the
improvement of military equipment and the benefits of military expenditure on
equipment purchase (Li et al. 2011). Therefore, new technical equipment pro-
curement seems especially important that it is urgent to win a priority and improve
74 A Game Analysis of New Technical Equipment Procurement 707
the benefits in the procurement game, though the optimal equilibrium is not stable,
which promotes a pressing need to take specific measures as follows.
74.3.1 Measures from a Macroscopic Point of View
Firstly, Equipment procurement rules should be gradually improved based on the
administrative rules of military equipment procurement expenditure and some
related administrative regulations on military product price, to form a compre-
hensive system of laws and regulations on equipment procurement management,
ensuring that there are laws to go by and rules to obey (Zhang and Zhang 2007;
Hang and Tan 2011).
Secondly, the building of procurement team, the scientific group, should be
strengthened to make every talent of it be good at price review work, under-
standing both techniques and finance (Hao and Jiang 2010).
74.3.2 Measures from a Microscopic Point of View
On the one hand, given many related links and departments involved in the
equipment procurement, purchasers should build up the sense of responsibility to
deal well with and strengthen all kinds of relationships (Zhang et al. 2009).
On the other hand, purchasers should make a good job of price review work,
which requires them to actively focus on or participate in scientific research,
gaining an adequate understanding of the details (quality, performance, design,
material, manufacture, etc.) of the equipment to accumulate some related infor-
mation, and also pushes them to get acquainted with the critical information
(business concepts, pricing strategy, foaming quotes, rational price, etc.) of the
suppliers to make a good preparation for the subsequent work (Yuan and Hu 2008;
Wang et al. 2007). All the measures mentioned above, if taken completely, can not
only effectively prevent the suppliers exaggerating the equipment cost but also
give firm guarantee for a rational quotation and an effective procurement contract.
References
Aghion P, Bolton P (1992) An incomplete contracts approach to financial contracting. Rev Econ
Stud 1992(6):473–494
Aliprantis CD, Chakrabarti SK (2000) Games and decision making. Oxford University Press,
New York, Oxford
Hang H, Tan G (2011) Equipment purchase power supervision based on reason. J Liaoning Tech
Univ (Nat Sci Ed) 2011(A01):211–216
708 A. Zhang et al.
Hao S-c, Jiang Y-n (2010) Research on purchasing corruption reasons and countermeasure based
on game theory model. Storage Transp Preserv Commod 2010(3):95–97+75
Hartley K (2007) The Arms industry, procurement and industrial polices. In: Sandler T, Hartley K
(eds) Handbook of defense economics, Vol 2. Elsevier, Amsterdam
Hou D-P, Wang Z-J (2001) Theoretical discuss and applications of nonlinear assessment. China
University of Science and Technology Publishing House, Hefei
International Society of Parametric Analysts (2007) Parametric estimating handbook, Fourth
edn., ISPA/SCEA Joint Office, Vienna, VA, pp 77–78
Li J, Gan M, Wang F (2011) A game approach to collusion in purchasing and pricing of military
reserves. Logistics Technol 2011(7):214–216
Sun Z-b, Jin C-h, Peng l (2011) Game analysis of weapon and military equipment procurement.
Mil Econ Res 2011(6):26–28
Wang H-m, Qu W, Bai H-w (2007) Study on model and strategy based-on asymmetric of
equipment procurement information’s game. J Acad Equip Command Technol 2007(5):31–34
Wang J-k (2011) Gambling analysis on anticorrosion and supervision to the Government
procurement. Value Eng 2011(17):135–136
Weitzman M (1980) The ratchet principle and performance incentives. Bell J Econ 1980:302–308
Xiang F-x, Xin W-f (1997) On enhancing three awarenesses and deepening administrative reform
of equipment procurement expenditure. Mil Econ Res 18:52–56
Xie X-h, Wang J-w, Yang M-j (2011) Some key issues in improvement of competitive system of
Chinese military equipment procurement. J Mil Econ Acad 2011(5):155–157
Yuan Y-q, Hu L (2008) Game analysis on military materials procurement under the lowest bid
price. Logistics Technol, 2008(10):259–261
Zhang H-y, Zhang W-j (2007) Analysis on anti-collusion based on game theory in tendering
procurement. Logistics Technol 2007(4):22–24+39
Zhang T, Cao M-y, Ou Y (2009) Incentive pricing model for equipment acquisition based on
game theory. J Armored Force Eng Inst, 2009(6):20–22+39
Chapter 75
Constructing Performance Measurement
Indicators in the Government’
Information Unit in Taiwan: Using
Balanced Scorecard and Fuzzy Analytic
Hierarchy Process
Yi-Hui Liang
Abstract The purpose of the study is to establish balanced scorecard (BSC) in
performance measurement of Government’ MIS Department. We take a broader
definition of Government’ MIS Department as “an assembly which brings forth
some specific functional activities to fulfill the task of MIS.” BSC used as a
measurement tool to assess study subjects, according to its strategy and goal
formed by its assignment property, can be divided into four dimensions: internal
process, customer, business value, and future readiness, which can provide us with
a timely, efficient, flexible, simple, accurate, and highly overall reliable mea-
surement tool. In order to extract the knowledge and experience from related
experts to pick out important evaluation criteria and opinion, this study combines
fuzzy theory and the analytical hierarchy process (AHP) to calculate the weights.
After completing weighted calculation of every dimension and indicator, the BSC
model is thus established. The findings of this study show that the indicator
weightings between and among all the levels are not the same, rather there exists
certain amount of differences. The degrees of attention drawing in order of
importance among all dimensions are internal process, customer, business value,
and future readiness. After comprehensively analyzing indicators of performance
measurement included in every level, the highly valued top three indictors are,
when conducting dimension performance measurement in Government’ MIS
Department, “Control cost,” “Satisfy end user demand,” “Operate and maintain
information technologies efficiently”. From these studies we will be able to
develop the indicators and the calculated weights of the four dimensions and the
indicators mentioned above. This model can be utilized by the information units of
Y.-H. Liang (&)
Department of Information Management, I-SHOU University,
Kaohsiung, Taiwan, Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 709
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_75,
Ó Springer-Verlag Berlin Heidelberg 2013
710 Y.-H. Liang
the governments for constructing the strategies and blueprints for self evaluation.
Further, these can also provide important information for effective resource
investment in Government’ MIS Department.
Keywords Government MIS Performance measurement Balanced scorecard
75.1 Introduction
Performance appraisal system is the most effective tool used for government
reengineering. Performance appraisal aims to help people achieve their strategies,
missions, visions and goals.
Wu (2000) supposed that Good performance appraisal systems can enable
government departments to allocate reasonable resources, prioritize resource
investment, further improve departmental effectiveness and efficiency, and orga-
nizational members adopt identical methods to pursue their goals, encourage their
morale, and cause them to focus on organizational vision.
Traditional government departments usually developed their information sys-
tems according to their individual requirements, and hence did not communicate
with each other, leading people to develop bad impressions and stereotypes
regarding government performance owing to inefficient government operations.
Balanced Scorecard (BSC), which was developed by Kaplan and Norton
(1992), is a useful and popular method of identifying business performance using
lagging and leading indicators based on the foundation of visions and strategies.
Balanced Scorecard implies that organizational performance is evaluated not only
utilizing financial indicators, but also simultaneously non-financial indicators.
Balanced Scorecard built a framework to transform organizational vision and
strategies into a series of consistent performance indicators, and thus execute and
control organizational administration, allow organizational members to more
concretely learn the vision and strategies of organization, and also l and also help
managers track the outcomes of implemented strategies.
Since Executive Yuan, Republic of China implemented performance reward
and performance management plan in 2003, this plan followed the BSC spirit.
However, Executive Yuan then consider the business properties, organizational
culture, and management and check, so as to authorize each government depart-
ment to set up its own performance evaluation process and evaluation indicators
(Directorate-General of Personal Administration 2005). Until now, Executive
Yuan does not force government departments to set up their own performance
evaluation process and evaluation indicator (Chu and Cheng 2007).
The analytical hierarchy process (AHP) (Kaplan and Norton 1992), which is the
multi-criteria technique, is considered appropriate for solving complex decision
problems (Directorate-General of Personal Administration 2005). The AHP is
based on theory, and offers information on the relative weight of the BSC
75 Constructing Performance Measurement Indicators 711
performance indicator (Chu and Cheng 2007; Searcy 2004). Otherwise, (Zadeh
1965) (Liedtka 2005) developed fuzzy theory to handle uncertain problems
involving fuzziness and vagueness. Lee et al. (Martinsons et al. 1999) posited that
traditional BSC failed to consolidate diverse performance indicators. Lee et al.
(Martinsons et al. 1999) also suggested the fuzzy AHP method as as an answer for
this problem.
BSC can help managers of government organizations holistically evaluate
information technology (IT) investments, as well as the performance of informa-
tion system (IS) departments. This study builds a Framework for evaluating
government MIS departments based on BSC. The study summarizes how to
combine the BSC and fuzzy AHP to serve as a decision tool for government
organization. The tool can be used not only to assess the contribution of a specific
government MIS department, but also analyze the performance and direct the
activities of government MIS departments.
75.2 Methodology
75.2.1 Research Structure
This study builds a Framework for evaluating government MIS departments based
on BSC and fuzzy AHP.
75.2.2 Select Research Variables
This study adopted the dimensions and indicators which developed by Martinsons
et al. (1999), Liang et al. (2008), and related government MIS experts to develop
my proposed the dimensions and indicators. The research variables are showed as
Table 75.1.
75.2.3 Fuzzy AHP
Step 1: Construct hierarchical framework of the BSC performance evaluation
criteria
From the four BSC perspectives, the hierarchical framework of the BSC per-
formance evaluation criteria is constructed.
Step 2: Using AHP method to calculate the weight
If get the weightmatrix W in pairwise comparison matrix A, standardize geo-
metrical mean of row vectors, multiply element in every row, get geometrical
mean and normalize it.
712 Y.-H. Liang
Table 75.1 Research variables
Dimension Indicator
Customer 1. Build and maintain the good image and reputation with end users. A1
2. Have the opportunity to develop IT. A2
3. Maintain a good relation with user communities. A3
4. Satisfy end user requirement A4
5. Perceived the preferred IS products and services provider by end users. A5
Business 1. Manage the good image and reputation. B1
value 2. Make sure IS projects to offer business value. B2
3. Control cost. B3
4. Be onerous to offer the suitable IS products and services to the third party. B4
Inner process 1. Expect and affect the demands from end users and managers. C1
2. Plan and develop IT efficiently. C2
3. Operate and maintain IT applications efficiently. C3
4. Obtain and test new hardware and software. C4
5. Offer to satisfy the end user trainings with effective cost. C5
6. Manage the IS problems effectively. C6
Future 1. Expect and prepare the IS problems. D1
readiness 2. Train and develop regularly to improve IS skills. D2
3. Promote regularly IS applications mix. D3
4. Increase regularly IS hardware and software. D4
5. Implement cost-effective and new technological researches which are D5
suitable for organizations.
Step 3: Construct Positive Reciprocal Matrix
Every evaluation member use fuzzy AHP evaluation scale to express relative
weight between each dimensions and criteria, and construct fuzzy Positive Reci-
procal Matrix.
Step 4: Consistency Check
The check methods are as follows:
4.1 Consistency Index (C.I.)
According to Consistency Index (C.I.), C.I = 0 indicate that evaluation has
perfect consistency; C.I [ 0 indicate that evaluation has consistency; C.I. \ 0.1
indicate that evaluation has evaluation has tolerant bias.
4.2 Consistency Rate (C.R.)
Saaty (1980) (Kaplan and Norton 1992) supposed that Consistency Rate (C.R.)
to evaluate the consistency of pairwise comparisons in a matrix among criterions.
Under the condition of different rank of matrix, it produce different random index
(R.I.). Under the condition of the same rank of matrix, the ratio of C.I. to R.I. is
called C.R.. When C.R ≤ 0.1, the consistency level is acceptable.
Step5: Calculate fuzzy weight value
Utilize the Lambda-Max method which Csutora and Buckley (2001) proposed,
calculate the fuzzy weight of evaluation criterions. The steps of calculation are as
follows:
75 Constructing Performance Measurement Indicators 713
5.1 When α = 1, use α-cut to get median Positive Reciprocal Matrix. Then,
calculate the weight use AHP method to get the weight matrix.
5.2 When α = 0, use α-cut to get minimum positive reciprocal matrix and
maximum positive reciprocal matrix. Then, calculate the weight use AHP
method to obtain the weight matrix.
5.3 In order to make sure that calculated weight value is fuzzy number,
therefore, adjusted the coefficient.
5.4 After obtained adjusted coefficient, calculate minimum positive reciprocal
weight matrix and maximum positive reciprocal weight matrix of every mea-
surement dimension.
5.5 Combing adjusted minimum, maximum and median values to get the fuzzy
weight in kth evaluation member and kth measurement dimension.
5.6 Utilize average method to integrate the fuzzy weight of evaluation members
and measurement dimensions.
75.3 Results
75.3.1 Survey Candidates
Based on previous studies on applying the BSC approach to information systems,
this study used the 21 indicators as performance evaluation indicators to construct
the research model and develop the questionnaire items based the model. The 20
indicators are showed in Table 75.2.
Next, take the central engineering government department as the example, and
calculate the weights of all dimensions and indicators of the model using Fuzzy
AHP method. The questionnaire was distributed among Director and Vice Director
of the direct department, 7 Director of first-class independent unit, 3 Section
Manager of the direct department, and Director of Information Technology, and a
total of 13 valid questionnaires were returned and censored 2 invalid question-
naires (refusing answer, incomplete answer, or don’t passing the consistency
check). The result of this study is showed in Table 75.2.
The results demonstrated that the importance weights of all dimensions were
ordered as follows: internal process, customer, business value, and future readi-
ness. Additionally, the top three importance weights of performance evaluation
indicators were top three indicators are “Control cost,” “Satisfy end user demand,”
“Operate and maintain information technologies efficiently”.
714 Y.-H. Liang
75.3.2 Results of Fuzzy AHP Method
Table 75.2 Results
Dimension Weight Indicator Weight Rank
Customer 0.277 1. Build and maintain the good image 0.046 11
and reputation with end users.
2. Have the opportunity to develop IT. 0.037 15
3. Maintain a good relation with user communities. 0.053 8
4. Satisfy end user requirement 0.098 2
5. Perceived the preferred IS products and services 0.043 12
provider by end users.
Business 0.239 1. Manage the good image and reputation. 0.050 10
value 2. Make sure IS projects to offer business value. 0.050 9
3. Control cost. 0.100 1
4. Be onerous to offer the suitable IS products and 0.040 13
services to the third party.
Inner 0.328 1. Expect and affect the demands from end users and 0.055 7
process managers.
2. Plan and develop IT efficiently. 0.060 5
3. Operate and maintain IT applications efficiently. 0.064 3
4. Obtain and test new hardware and software. 0.032 16
5. Offer to satisfy the end user trainings with effective 0.062 4
cost.
6. Manage the IS problems effectively. 0.055 6
Future 0.156 1. Expect and prepare the IS problems. 0.029 18
readiness 2. Train and develop regularly to improve IS skills. 0.030 17
3. Promote regularly IS applications mix. 0.029 19
4. Increase regularly IS hardware and software. 0.028 20
5. Implement cost-effective and new technological 0.040 14
researches which are suitable for organizations.
75.4 Conclusion
Performance appraisal systems for profit organizations have traditionally measured
performance financially. For non-profit organizations a different approach is used,
since for profit is not a main objective for such organizations, and possibly is even
a constraint. Financial performance represents a subjective measure of how well a
firm can use assets from its primary mode of business to generate revenues.
Notwithstanding, for government organizations and other non-profit organizations,
financial performance primarily represents a measure of how efficiently a gov-
ernment organization can use its budget. Using financial performance to measure
is not sufficient to measure the government organization performance.
This study develops BSC framework and calculate the weights of the four
perspectives and the indicators mentioned above. These performance evaluation
indicators will then be utilized by the information units of governments for
75 Constructing Performance Measurement Indicators 715
constructing the strategies and blueprints for self evaluation. Further, these can
also provide other related departments for effective resource investment in infor-
mation units of governments.
Compared to Miller and Doyle (1987) 和 (Saunders and Jones 1992), the
proposed IS evaluation dimensions and indicators more focus non-profit organi-
zations characteristics.
References
Chu HL, Cheng CJ (2007) An empirical investigation of the governmental performance
management system: using the balanced scorecard and the analytic hierarchy process.
J Manag 24(6):689–705
Csutora R, Buckley JJ (2001) Fuzzy hierarchical analysis: the lambda-max method. Fuzzy Sets
Syst 120(2):181–195
Directorate-General of Personal Administration (2005) Executive Yuan, performance reward and
performance management plan
Kaplan RS, Norton DP (1992) The balanced scorecard-measures: that drive performance. Harv
Bus Rev 70(1):72–79
Liang YH, Hsieh HC, Wang CY (2008) A study of implementing balanced scorecard of
government mis department-a case study of engineering department of government. Electron
Commer Stud 64(4):425–445
Liedtka SL (2005) Analytic hierarchy process and multi criteria performance management
systems. Cost Manag 19(6):30–38
Martinsons M, Davison R, Tse D (1999) The balanced scorecard: a foundation for the strategic
management of information systems. Decis Support Syst 25(1):71–88
Miller J, Doyle BA (1987) Measuring the effectiveness of computer-based information systems in
the financial service sector. MIS Q 11(1):107–124
Saaty TL (1980) The analytic hierarchy process. McGraw-Hill, New York
Saunders FA, Jones JW (1992) Measuring performance of the information systems function.
J Manag Inf Syst 8(4):63–82
Searcy DL (2004) Alignment the balanced scorecard a firm’s strategy using the analytic hierarchy
process. Manag Acc Q 5(4):1–10
Wu A (2000) The new direction for performance evaluation. The Chin BAS Mon 530:43–52
Zadeh LA (1965) Fuzzy sets. Inf Control 8:338–353
Chapter 76
A Decision Model and Its Algorithm
for Vehicle Routing Problem to
Meet Emergencies
Xing Liu, Min Zhong and Ya-hong Ma
Abstract A kind of vehicle routing problem is discussed to meet emergencies, in
which some routes may be destroyed or blocked uncertainly. The value of a route
in an uncertain situation is analyzed and a two-stage integer program model is
constructed. To simplify the computation of the model a method is put forward.
And in this method, only the maximum value and the minimum value are
calculated for the object value. In the end, a two-stage taboo search algorithm is
designed and an example is given.
Keywords Logistics Taboo search algorithm Uncertain decision Vehicle
routing problem
76.1 Introduction
People had and have to face so many natural disasters, for example, the Kobe
earthquake (1995) in Japan, the hurricane ‘‘Rita’’ (2005) in the United States, the
Wenchuan earthquake (2008) in China, etc. These natural disasters had destroyed a
lot of traffic facilities and made emergency rescue very difficult. For example, some
key bridges, tunnels and line hubs of the traffic lines may be damaged or destroyed. If
vehicles can pass through the key bridges (tunnels), transportation mileage will be
shortened. But if these bridges (tunnels) were destroyed, the vehicles may have to
make a detour to transport, or even backtrack (Li and Guo 2001; Gan et al. 1990;
X. Liu (&) Y. Ma
Department of Military Transportation, University of Military Transportation,
Tianjin, China
e-mail: [email protected]
M. Zhong
Department of Basic Science, University of Military Transportation,
Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 717
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_76,
Ó Springer-Verlag Berlin Heidelberg 2013
718 X. Liu et al.
Liu and Jiao 2000; Wu and Du 2001; Zhang et al. 2002; Renaud and Boctor 2002). It
may delay the mission. So decision-making of multi-vehicle path decision problem
in the emergency environment became more and more important.
This article discusses the multi-vehicle path decision problem when some
critical sections (bridges, tunnels, etc.) of the road network may be destroyed. The
mathematical model is constructed and the taboo heuristic algorithm is given to
solve the problem.
76.2 Description of the Problem and Complexity Analysis
76.2.1 Description of the Problem
The problem can be described as follows:
Let G ¼ ðV; AÞ be the transportation network graph. Let V ¼ fv0 ; v1 ; v2 ; ; vn g
be the set of vertices. And let v0 be the freight yard, which is the point that the vehicles
will start from and will get back when delivery finished. The other vertices
v ði ¼ 1; 2; . . .; nÞ are the locations which are in need of services. Let A ¼
i
vi ; vj i 6¼ j; vi ; vj 2 V be the set of edges. Each edge represents a road between
two vertices. Vector set C ¼ cij is defined on the edge set A to present the distance,
travel costs or travel time between the two vertices.
Suppose ½b1 ; b2 2 A is an edge to represent a road which may be destroyed at
any time. And the destruction time and the probability of uncertainty could not be
known. Graph G ^ would be the new network graph after ½b1 ; b2 2 A was destroyed.
^
The graph G will be almost the same as graph G except the edge ½b1 ; b2 . Customer
demand at vertices vi will be noted as qi ði ¼ 1; 2; . . .; nÞ. All the demand will be
met by M vehicles whose maximum cargo capacity is Q. Now problem is how to
find the smallest expenses (the shortest time, the smallest mileage, or other
smallest comprehensive index) transportation routes.
76.2.2 Complexity Analysis
The vehicle routing problem is recognized NP-hard. If there are n demand vertices,
there will be n! kinds of optional routes. The number of optional routes which
include ðb1 ; b2 Þ or ðb2 ; b1 Þ is 2ðn 1Þ!. Those optional routes which include
ðb1 ; b2 Þ will be at least ðn kÞ! when kðk nÞ demand vertices have been fulfilled
and the route ahead of the vehicles has been destroyed. Only consider k ¼ 1, the
number of optional routes may be
n! 2ðn 1Þ! þ 2ðn 1Þ!ðn 1Þ! n! þ 2ðn 1Þ!ðn 1Þ!
76 A Decision Model and Its Algorithm for Vehicle Routing Problem 719
If n ¼ 10, the number of optional routes is about 2:6 1011 . When the eval-
uation of each scheme should be calculated, it will take too long to finish. If
consider k [ 1 and the whole network be taken into account, the calculation time
will be longer more.
76.3 Mathematical Model
For any pre-planning transport route scheme Ri ri1 ; ri2 . . .rij . . .; riM (rij represents
the route of the vehicle j). dj is used to represent the set of demand vertices that the
vehicle j should service. Then the set of all demand vertices can be noted as
D ¼ [dj . If rij doesn’t contain the road ðb1 ; b2 Þ or ðb2 ; b1 Þ which may be damaged
or destroyed, the scheme does not need to adjust in transit. If rij contains the road
ðb1 ; b2 Þ or ðb2 ; b1 Þ, hj is used to represent the current location of the vehicle j when
the vehicle j gets the information that the road ðb1 ; b2 Þ or ðb2 ; b1 Þ was destroyed.
The set of all vehicles’ current location is H ¼ hj j j M . dj0 is used to represent
the rest set of demand vertices that the vehicle j S hasn’t serviced. Then the set of all
the rest demand vertices can be noted as D0 ¼ dj0 .
The results of re-optimizing the line of the vehicle j should be related to the
current location of all vehicles, the remaining demand vertices, and the trans-
portation network graph G. ^ Since the time when the road will be destroyed is
uncertain, the factors above are uncertain too.
So the problem can be represented by a two-stage uncertain planning problem.
If all the vehicles are collaborative the path global collaboration optimization
model can be expressed as follows:
X
n X
n X
M X
M X
e X
n X
n X
M
min z ¼ cij xijm þ switchð ðxb1 b2 m þ xb2 b1 m ÞÞ phb ½min z0 crs xrsb
i¼0 j¼0 m¼1 m¼1 hb ¼1 r¼0 s¼1 m¼1
r; s 2 ðd d0 Þ [ f0g ð76:1Þ
M X
X n X
n
min z0 ¼ cij x0ijm ð76:2Þ
m¼1 i¼0 j¼0
X
n X
n
qj xijm Qm m ¼ 1; . . .M ð76:3Þ
i¼1 j¼1
X
M X
n
xijm ¼ 1 j2d ð76:4Þ
m¼1 i¼1
720 X. Liu et al.
X
M X
n
xijm ¼ 1 i2d ð76:5Þ
m¼1 j¼1
M X
X n M X
X n
x0jm ¼ xi0m ¼m ð76:6Þ
m¼1 j¼1 m¼1 i¼1
X
M X
n
x0hi jm ¼ 1 hi 2 H ð76:7Þ
m¼1 j¼1
M X
X n
x0i0m ¼ M ð76:8Þ
m¼1 i¼1
X
M X
n
x0ijm ¼1 i 2 d0 ð76:9Þ
m¼1 j¼1
M X
X n
x0ijm ¼1 j 2 d0 ð76:10Þ
m¼1 i¼1
x0b1 b2 m ¼ x0b2 b1 m ¼ 0 ð76:11Þ
xijm ¼ 0; 1 x0ijm ¼ 0; 1 ð76:12Þ
Expression (76.1) is the first-stage objective function of the two-stage planning.
xijm is a 0–1 variable that it will be 1 when the vehicle m passes through the road
ði; jÞ, otherwise it will be 0. The first term on the left of expression (76.1) means
the value of the first planning. And the second term means re-optimizing value
when the line was destroyed and a new scheme was planed. switchðÞ is a switch
function. It will be 1 when the line contains sections may be destroyed and will be
0 otherwise. The serial number of the first vertex in destroyed line rib is noted as e.
phb is the probability that the line might be re-optimized at the location hb . minz0 is
Pn Pn PM
the re-optimized value of the line and r¼1 s¼0 m¼1 crs xrsm is the milage
planed in first stage but won’t be achieved because of re-optimizing. Expres-
sion (76.2) is the second-stage objective function of the two-stage planning.
Similarly, x0ijm is a 0–1 variable in the second-stage that it will be 1 when the
vehicle m passes through the road ði; jÞ, otherwise it will be 0.
Expressions (76.3)–(76.6) are constraints of the first-stage planning. (76.3)
means that the transport needs of each vehicle must be less than the gross vehicle
weight. (76.4) and (76.5) mean that each demand vertex can only access one time.
(76.6) means the vehicle issued from the warehouse must be the same of the
returned and equal to the number of all transport vehicles. (76.7)–(76.12) are
constraints of the second-stage planning. (76.7) means that the vehicle involved in
the collaboration will start from the current location. (76.8) means that the vehicles
76 A Decision Model and Its Algorithm for Vehicle Routing Problem 721
involved in the collaboration will return to the warehouse. (76.9) and (76.10) mean
that each remaining demand vertex can only access one time. (76.11) means the
vehicle can not pass by the destroyed road.
76.4 Evaluation of the Schemes
If each value of phb in the expressions (76.1) is taken into account, the algorithm
will need too long time to calculate. And the law of the destruction is hard to find
in emergency environment. In fact, the decision makers always concerned about
worst-case situation or best-case situation. So the evaluation of the schemes can be
simplified to the maximum evaluation value in worst-case situation, or the mini-
mum evaluation value in best-case situation.
The minimum milage of the scheme Ri is noted as bestðRi Þ and the maximum
milage of the scheme Ri is noted as worstðRi Þ. The utility of the scheme Ri can be
noted as
uðRi Þ ¼ uðbestðRi Þ; worstðRi ÞÞ:
If ðb1 ; b2 Þ 62 Ri , namely if there is no road being destroyed then
uðRi Þ ¼ f ðRi Þ ¼ bestðRi Þ ¼ worstðRi Þ:
If ðb1 ; b2 Þ 2 Ri , when decision-maker gets the information that the road was
destroyed, the scheme should be renewed.
Therefore, the worst-case situation is that the information of damaged road is
known when the vehicle has been arrived at the vertex of the damaged road and the
vehicle may have to make a long detour to transport. If ðb1 ; b2 Þ 2 Ri , sk ðRi Þ is
the serial number of demand vertex k in rib . And ff ðRi ði; jÞÞ is the milage of all the
vehicles
when the vehicle b runs from demand vertex i to demand vertex j.
ff R0i ðsk Þ is the milage of all the vehicles after the scheme has been renewed at sk .
Theorem If ðb1 ; b2 Þ 2 Ri , the maximum milage for the transport process is
f1 ðse ; Ri Þ, and f1 ðse ; Ri Þ f1 ðse ; Ri Þ, in which se ¼ minðsb1 ; sb2 Þ and sk se .
Proof According to the path of the scheme Ri
f1 ðsk ; Ri Þ ¼ ff ðRi ð0; sk ÞÞ þ ff R0i ðsk Þ ð76:13Þ
f1 ðse ; Ri Þ ¼ ff ðRi ð0; se ÞÞ þ ff R0i ðse Þ ð76:14Þ
Let (76.13)–(76.14), get
f1 ðse ; Ri Þ f1 ðsk ; Ri Þ ¼ ff ðRi ðsk ; se ÞÞ þ ff R0i ðse Þ ff R0i ðsk Þ ð76:15Þ
Because ff ðRi ðsk ; se ÞÞ is the milage of all the vehicles when the vehicle b runs
from demand vertex sk to demand vertex se , then
722 X. Liu et al.
ff ðRi ðsk ; se ÞÞ 0:
Because ff ðR0i ðsk ÞÞ is the milage of all the vehicles after the scheme has been
renewed at sk and ff ðR0i ðse ÞÞ is the milage of all the vehicles after the scheme has
been renewed at se and sk se ,
ff ðR0i ðse ÞÞ ff ðR0i ðsk ÞÞ:
So
f1 ðse ; Ri Þ f1 ðsk ; Ri Þ ¼ ff ðRi ðsk ; se ÞÞ þ ff ðR0i ðse ÞÞ ff ðR0i ðsk ÞÞ 0
Hence
f1 ðse ; Ri Þ f1 ðsk ; Ri Þ:
76.5 Taboo Search Algorithm
The main process of the taboo search algorithm is as follows:
(1) Generate a path sequence randomly for the initial solution. Code n demand
vertices and note warehouse as 0. Path solution is a random arrangement of the
number 0 to n. The head and tail of the path solution are 0. There are M 1
0 s in the middle of the path solution. And they are situated among 1 to n
randomly. The numbers between two 0 s represent the service path of one
vehicle.
For example, to 6 demand vertices and two cars, a path solution is 0-1-2-3-0-4-
5-6-0.
(2) Generate neighborhood of the solution. Exchange the situation of two demand
vertices or insert a new vertex into the service path to gain new solutions.
(3) Process the constraints of the vehicle capacity. If the total demands of demand
vertices exceed cargo capacity of the vehicle, the scheme should be
eliminated.
(4) Taboo objects are two adjacent demand vertices. The length of the taboo list
increases with the increase of evolution generation. If current value is better
than the optimal value in history, the taboo is lifted.
(5) Evaluate the value of the scheme. Calculate the minimum mileage and
maximum mileage of each scheme. When calculate the maximum mileage of
the scheme which contains the destroyed sections of the path, use an inner
taboo search algorithm that is similar to the external algorithm. The final
evaluation value depends on the decision-making criteria, such as optimistic
criteria, pessimistic criteria, compromise guidelines and expectations criteria,
etc. (Qu et al. 2004; Golden et al. 1998; Lia et al. 2005; Renaud et al. 1996;
Stenger et al. 2012; Li and Li et al. 2011; Qian 2011; Liu et al. 2005).
76 A Decision Model and Its Algorithm for Vehicle Routing Problem 723
(6) Consider the direction of the path. On optimistic criterion, the changed value
of the destruction has not been calculated to evaluate the value. In order to
compensate for this deficiency, a number of sub-optimal evaluation values will
be picked up, and the worst case results will be calculated for decision-making
reference. The schemes that have the same sequence but different direction are
considered to be different. Such as 0-1-2-3-4-0 and 0-4-3-2-1-0 are different
schemes.
76.6 Example
There are 1 warehouse vertex and 10 demand vertices in the Table 76.1. And data
in the table is the distance between two vertices.
For each demand vertex, the demand is 5. There are two vehicles being used to
transport. For each vehicle, the maximum cargo capacity is 35 (Table 76.2).
Table 76.1 The distance between the vertices (I)
Vertices 0 1 2 3 4 5
0 0 1000 38 37 45 25
1 1000 0 1000 14 1000 1000
2 28 1000 0 1000 25 28
3 27 14 1000 0 40 1000
4 45 1000 25 40 0 1000
5 25 1000 28 1000 1000 0
6 1000 39 36 1000 1000 10
7 36 36 1000 1000 1000 25
8 41 1000 1000 1000 1000 13
9 1000 17 1000 15 1000 1000
10 30 1000 1000 34 1000 1000
Table 76.2 The distance between the vertices (II)
Vertices 6 7 8 9 10
0 1000 26 41 52 40
1 39 36 1000 37 1000
2 36 1000 1000 1000 1000
3 1000 1000 1000 15 44
4 1000 1000 1000 1000 1000
5 10 25 13 1000 1000
6 0 25 19 1000 1000
7 25 0 12 1000 14
8 29 12 0 1000 31
9 1000 1000 1000 0 30
10 1000 14 41 40 0
724 X. Liu et al.
Table 76.3 The optimal solutions when 2–4 may be destroyed
Decision Optimism Compromise Pessimistic criteria
making criteria criteria
criteria
a b a b
301 301 312 324 323
The best situation 7 7 7 4 7
10 10 10 3 8
9 9 9 9 5
1 1 1 1 6
3 3 3 6 2
0 0 0 2 0
4 4 4 0 4
2 2 2 5 3
6 6 6 8 1
8 8 8 7 3
5 5 5 10 9
10
323 382 323 324 323
The worst situation 7 7 7 4 7
8 10 8 3 8
5 9 5 9 5
6 1 6 1 6
2 3 2 6 2
0 0 0 2 0
4 4 4 0 4
3 3 3 5 3
1 1 1 8 1
3 7 3 7 3
9 8 9 10 9
10 5 10 10
6
2
Evaluation value 301 301 312 324 323
Between vertices 2 and 4, there is so important a bridge that if it was damaged it
could not be repaired. The optimal solutions are calculated out using the tabu
search algorithm under the different criteria (Table 76.3).
76.7 Conclusion
Decision-making of multi-vehicle path decision problem in the emergency envi-
ronment became more and more important. This article discusses the multi-vehicle
path decision problem when some critical sections (bridges, tunnels, etc.) of the
traffic network may be destroyed. The mathematical model is constructed and the
taboo heuristic algorithm is given to solve the problem. Research results can be
used for uncertain environment logistics distribution decision.
76 A Decision Model and Its Algorithm for Vehicle Routing Problem 725
References
Gan Y, Tian F et al. (1990) Operations research. Tsinghua University Press, Beijing
Golden BL, Wasil EA, Kelly JP, Chao I-M (1998) The impact of metaheuristics on solving the
vehicle routing problem: algorithms, problem sets, and computational results. In: Crainic T,
Laporte G (eds) Fleet management and logistics. Kluwer, Boston, pp 33–56
Lia F, Goldenb B, Wasilc E (2005) Very large-scale vehicle routing: new test problems,
algorithms and results. Comput Oper Res 35:1165–1179
Li J, Guo Y (2001) Vehicles scheduling theory and method. China supplies press, Beijing
Liu C, Jiao S (2000) Urban post-earthquake relief system relief decision. J Nat Disasters 3
Li Q, Li Q (2011) Based on spatio-temporal crowding emergency evacuation route optimization
method. J Mapp 55(4):517–523
Liu X, He G, Gao W (2005) The multiple vehicles coordinated stochastic vehicle routing model
and algorithm. Syst Eng 23(4):105–109
Qu Z, Cai L, Li C (2004) The frame for vehicle routing problem of large-scale logistics systems.
J Tsinghua Univ (Nat Sci Ed) 44(5):43–44
Qian W (2011) Tabu in combination with genetic results on the distribution routing optimization
research and application. Comput Appl Softw 2011:53–57
Renaud J, Boctor FF (2002) A sweep-based algorithm for the mix vehicle routing problem. Eur J
Oper Res 140:618–628
Renaud J, Laporte G, Boctor FF (1996) Tabu search heuristic for the multi-depot vehicle routing
problem. Comp Ops Res 21(3):229–235
Stenger A, Enz S, Schwind M (2012) An adaptive variable neighborhood search algorithm for a
vehicle routing problem arising in small package shipping. Transp Sci 47(1):64
Wu Y, Du G (2001) Management science foundation. Tianjin university press, Tianjin
Zhang F, Wu X, Guo B et al. (2002) Logistics network usability research. Syst Eng Theory
Method 12(1):77–80
Chapter 77
A Solution to Optimize Enterprise
Business and Operation Process
Xue-wu Chang, Xiao-yuan Ji and Jian-xin Zhou
Abstract Process continuous optimization plays an important role in enterprise
informatization; however, there are few solutions with good operation and prac-
ticability for enterprises to implement. In this paper, a solution was proposed,
which included a decomposition method of business process named ‘‘Timeline-
Place-Roles (TPR)’’ and an ‘‘8 steps’’ value-added method of process analysis and
continuous optimization based on the value chain. Then this paper took the
‘‘product-prepared’’ business process of a Chinese continuous manufacture
enterprise as a case, and made the case analysis and optimization by using the
solution. The results proved that the solution could optimize enterprise business
process and improve the effect continuously and deepen informatization applica-
tion to a certain extent.
Keywords Business process Depth application Enterprise informatization
Operation process Process optimization Solution
77.1 Introduction
On July 8, 2011, as an informatization web portal, e-works sponsored ‘‘2011
Chinese Manufacture Management Informatization Conference’’ in Zhengzhou,
China. The depth application of Enterprise Resource Planning (ERP) was an
X. Chang X. Ji (&) J. Zhou
State Key Laboratory of Material Processing and Die and Mould Technology,
Huazhong University of Science and Technology, Wuhan, China
e-mail: [email protected]
X. Chang
e-mail: [email protected]
J. Zhou
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 727
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_77,
Ó Springer-Verlag Berlin Heidelberg 2013
728 X. Chang et al.
important topic for discussion at the conference, and the enterprise business pro-
cess optimization was considered as a key method. The Chinese manufacture
enterprises have begun to enter the stage of process optimization, so it is mean-
ingful to research and propose a solution of process optimization.
In a corporation, enterprise management consists of many business processes,
and a business process is composed of many operation processes executed by
different roles at different places. The aim of enterprise informatization is to
improve enterprise management, and the improvement of enterprise management
could be realized by optimizing the business process, and the optimization of
business process could be supported by optimizing the operation process, the
relationship among them is shown in Fig. 77.1. So the level of enterprise infor-
matization could be advanced by optimizing the operation process continuously.
The aims of optimizing enterprise business process should include 2 points:
(1) To decrease the total time consumed of the business process;
(2) To strengthen the value-added capability of the business process and reduce
the non-value-added link.
There were many researches about business process optimization. Some people
use the petri net technology to construct a model for a business process, and then
use software to analyze and optimize it (Wang et al. 2008; Wang 2007; Aalst 1998;
Ling and Schmidt 2000; Pan et al. 2005; Li and Fan 2004; Pang et al. 2008). The
method could find the key route and figure out the shortest route, but it does not
tell us how to describe, decompose and sort out the business process. Someone
presented a process log method of business process digging by using the users’
operation log in some information system, such as business process management
(BPM) system, ERP system and so on (Zhang 2010; Feng 2006; Gaaloul et al.
2005, 2009). This process log method could represent the invisible business pro-
cess, but there are few such information systems which could support it. To
optimize business process, somebody used cost analysis method (Hu et al. 2003;
Cooper 1990; Spoede et al. 1994), or a Business Process Reengineer (BPR)
method based on the value chain (Baxendale et al. 2005), but the methods lacks
maneuverable steps; Prof. Lan proposed a new method, which established a
general equilibrium relationship of the enterprise value chain based on the dual
theory of linear program (Lan et al. 2011), but the operation of calculating the
value by using the cost was also a little complex.
Fig. 77.1 The relationship
between process optimization Enterprise
Enterprise management
with enterprise informatization improvement
informatization
Operation Business
process process
optimization optimization
77 A Solution to Optimize Enterprise Business 729
In fact, nowadays, the design and optimization of business process is completed
by hand in many Chinese companies generally. So the solution to improve busi-
ness process should be easy, and maneuverable. Business process optimization
requires modeling conveniently on one hand, and needs to add value on the other
hand. Thus, this paper proposed a whole set of solution with a decomposition
method of business process named ‘‘Timeline-Place-Roles (TPR)’’ and an ‘‘8
steps’’ value-added method of process analysis and continuous optimization based
on the value chain, which the chain value was estimated by time-consumed.
77.2 Methodology
The solution consists of two parts. The first part is a method to decompose a
business process into a series of continuous operation processes, which is named
Timeline-Place-Roles (TPR) method, as shown in Fig. 77.2. Based on the timeline,
decompose a business process into many activities according to activity places,
and make sure the places of each two adjacent activities are different. Then
decompose every activity into several operation processes according to activity
roles to assure that the roles of each two adjacent operation processes are not the
same. All the operation processes construct the business process. Record the place,
roles, and time consumed of every operation process in a table after the operation
process has been decomposed.
The second part is an ‘‘8 steps’’ value-added method to optimize every oper-
ation process of the business process, which contains 8 steps from simplification to
appreciation in order to make it more valuable, as shown in Fig. 77.3.
The optimization steps are as follows:
(1) Judge whether it could be deleted. If it is meaningless or repeating, then delete it.
(2) Judge whether it could be executed at the same time with other operation.
(3) Judge whether it could be carried out by the computer.
(4) Judge whether it could be simplified.
(5) Judge whether it could be standardized.
(6) Judge whether it could be extended to contain more information.
(7) Judge whether it could add another new operation process to make it more
valuable.
(8) Judge whether it could strengthen the value-added part.
Fig. 77.2 The timeline-
place-roles (TPR) method to A Place Roles Many
Business Many Operation
decompose a business process Activities
Process Processes
Timeline
Every operation: Place, Roles,Time consumed
730 X. Chang et al.
Fig. 77.3 The method to
optimize an operation process Operation
Delete? Value added?
Parallel? Once a Add?
week or
month
Automate? Extend?
Simplify? Standardize?
Meanwhile, the business process could be optimized once a week or once a
month by using these 8steps.
77.3 Case
The casting factory A is a continuous manufacture enterprise. Its enterprise
management was disorder and inefficient before its business process was optimized
to find an informatization platform by using the HZERP system (Zhou et al. 2008).
For example, it was a long time from when the plan was made to when workers
began to product in the workshop, which seriously impacted on the production
schedule. The business process included in the period could be called ‘‘product-
prepared’’ business process, as shown in Fig. 77.4.
In this paper, the ‘‘product-prepared’’ business process was taken as a case, and
then the solution was used to decompose and optimize the business process.
77.4 Results and Discussions
According to the first step of the Timeline-Place-Roles method, the previous
‘‘product-prepared’’ business process in the foundry enterprise A was decomposed
into 7 activity places, and each place had some activity roles, as show in Fig. 77.5.
Plan Draw material Production
Fig. 77.4 The previous ‘‘product-prepared’’ business process in the company A
77 A Solution to Optimize Enterprise Business 731
Place Roles
product dept. planner, production director
technical dept. planner, technical director
workshop planner, workshop director
TimeLine
manager room planner, production manager
workshop planner, workshop director, workshop
statistician, workshop material member
warehouse warehouse director, workshop material
member, warehouse director
workshop workshop material member
Fig. 77.5 The method to optimize an operation process
This Figure shows that the ‘‘planner’’ and ‘‘workshop material member’’ are the
main roles.
According to the second step of the Timeline-Place-Roles method, the activities
were decomposed into 14 operation processes, as shown in Table 77.1. The Table
records process Serial Number (SN), place, and roles of every operation process.
Then work out the average Time Consumed (TC, minutes) of every role in every
operation process. For instance, it would cost the planner 5 min (and 5 min back)
to send the plan to the workshop and give it to the work director who would spend
1 min to scan and check the plan. Then workshop statistician would summarize
into a material requisition according to the plan and the Bill of Material (BOM) of
the related product, and it cost would him 30 min.
According to the 14 operation processes in the Table 77.1 and the ‘‘8 steps’’
optimization method of the solution, the previous ‘‘product-prepared’’ business
process was optimized into 8 operation processes combining the information base
which was formed by the HZERP system in the company A. The optimized
‘‘product-prepared’’ business process was shown in Table 77.2. The three opera-
tions (Audit by production director, Audit by technical director and Audit by
workshop director) could be paralleled and simplified by the HZERP system; the
three roles could audit the plan and sign their names in no order or even at the
same time.
From Table 77.2, the company A could optimize its business process according
to the information system. Three operations were deleted or replaced by the system
automatically. Especially, the operation of summarizing into a material requisition
by the workshop statistician was carried out by the MRP subsystem of the HZERP
system, which saved time and made the result accurate.
732 X. Chang et al.
Table 77.1 The operation processes of the previous ‘‘product-prepared’’ business
SN Operation Place Roles TC(m)
1 Draw up production schedule Product dept. Planner /
2 Audit by production director Product dept. Planer, production director 2, 1
3 Audit by technical director Technical dept. Planer, technical director 2, 1
4 Audit by workshop director Workshop Planer, workshop director 10, 1
5 Audit by production manager Manager room Planer, production manager 5, 1
6 Send to workshop Product dept. Planer, workshop director 10, 1
7 Summarize into a material Workshop Workshop director, workshop 0, 30
requisition statistician
8 Audit by workshop director Workshop Workshop statistician, workshop 0, 1
director
9 Audit by warehouse director Warehouse Warehouse director, workshop 0, 3, 1
material member, warehouse
director
10 Prepare material by warehouse Warehouse Warehouse director, warehouse 0, 20
keeper keeper
11 Sign warehouse keeper’s name Warehouse Warehouse keeper 1
12 Carry material to vehicle and sign Warehouse Workshop material member 3
13 Retain the inventory on account Warehouse Warehouse keeper 10
14 Send material to workshop by Workshop Workshop material member 5
vehicle
Table 77.2 The optimization of operation processes of the previous ‘‘product-prepared’’
business
SN Operation Optimization method Optimization detail
1 Draw up production schedule / /
2 Audit by production director Parallel, Simplify Check the plan and sign the name in
2 Audit by technical director Parallel, Simplify the HZERP system
2 Audit by workshop director Parallel, Simplify
0 Audit by production manager Delete An extra step
0 Send to workshop Automate Send the plan to the workshop
automatically by the HZERP
system
3 Summarize into a material Automate, Standardize Summarize into a material
requisition requisition automatically by the
MRP subsystem
4 Audit by workshop director Parallel, Simplify Check the plan and sign the name in
4 Audit by warehouse director Parallel, Simplify the HZERP system
5 Prepare material by warehouse / /
keeper
6 Sign warehouse keeper / Sign the name in the HZERP system
7 Carry materials to vehicle and sign / Sign the name in the HZERP system
0 Retain the inventory on account Automate Account inventory automatically
after signing warehouse keeper’s
name
8 Send materials to workshop by / /
vehicle
77 A Solution to Optimize Enterprise Business 733
Table 77.3 The time consumed (including work and wait) before and after process optimization
Role TC/Before (m) TC/After (m) Proportion
of value
added (%)
Planner 2 ? 2 ? 10 ? 5 ? 10 = 29 0 100
Production director 1 1 0
Technical director 1 1 0
Workshop director 1?1?0?1?0=3 2 0
Production manager 1 0 100
Workshop statistician 30 ? 0 = 30 0 100
Workshop material 3 ? 20 ? 1 ? 3 ? 10 ? 5 = 42 3 ? 3 ? 5 = 11 73.8
member
Warehouse director 1 1 0
Warehouse keeper 20 ? 1 ? 3 ? 10 = 34 20 ? 1 = 21 38.2
Total 112 37 70
Total business duration 2 ? 2 ? 10 ? 5 ? 10 ? 30 1 ? 1 ? 1 ? 20 ? 3 66.7
? 1 ? 3 ? 20 ? 1 ? 3 + 3 ? 5 = 34
? 10 ? 5 = 102
Finally, analysis and contrast with the time consumed (including work time and
wait time) before and after the process optimization, the result was shown in
Table 77.3. TCB means the time consumed before, and TCA means the time
consumed after. The proportion of value added was characterized by the propor-
tion of the time saving, and was equal to (TCB - TCA)/TCB.
Table 77.3 shows that the total TCA was 112 min, while the total TCB was
only 37 min with 70 % value added. The total business duration before optimizing
process cost 102 min, while the one after optimizing process cost 34 min with
66.7 % value added. The time consumed by the three key roles (planner, workshop
statistician and workshop material member) was decreased largely, and the work
of the workshop statistician was omitted.
The results proved that the solution could optimize enterprise business process
and improve the effect continuously and deepen informatization application to a
certain extent, and the effect would be much better combining the usage of the
information system. Enterprises could improve the effect continuously and deepen
informatization application by using the solution proposed in this paper.
77.5 Conclusion
(1) The relationship among the process optimizing and enterprise informatization
was discussed.
(2) A Timeline-Place-Role (TPR) method of decomposing business process into
operation processes was proposed.
734 X. Chang et al.
(3) The ‘‘8 steps’’ value-added method of operation process analysis and contin-
uous optimization based on the value chain was proposed.
(4) The TPR method and ‘‘8 steps’’ method made up a complete set of solution.
(5) Take a business process of a continuous manufacture company as a case, this
paper made the case analysis and optimization by using the proposed solution
and compared with the time consumed before and after process optimization.
(6) The results proved that the solution could optimize enterprise business pro-
cess. And the effect would be much better by combining the usage of the
information system, which is also a new trend to promote management
technological upgrading.
Acknowledgments This study was financially supported by HUST self-determined and inno-
vative research funds for national defense (2011), Fundamental Research Funds for the Central
Universities (2011TS039), and Program for New Century Excellent Talents in University (No.
NCET-09-0396).
References
Aalst Wvd (1998) The application of petri nets to workflow management. J Circuits Syst Comput
8(1):21–66
Baxendale S, Gupta M, Raju PS (2005) Profit enhancement using an ABC model. Manag
Account Quart 6(2):11–21
Cooper (1990) Cost classification in unit based and activity-based manufacturing cost systems.
J Cost Manag (3):4–14
Feng C (2006) Workflow mining algorithm on time-based log. Computer software and theory.
Master dissertation, Shandong University, Shandong
Gaaloul W, Alaoui S, Baina K, Godart C (2005) Mining workflow patterns through event-data
analysis. In: Proceedings of 2005 symposium on applications and the internet workshops,
SAINT2005, Trento, pp 226–229
Gaaloul W, Gaaloul K, Bhiri S, Haller A, Hauswirth M (2009) Log-based transactional workflow
mining. Distributed Parallel Databases 25(3):193–240
Hu YG, Wang TM, Qiao LH (2003) BPR method based on enterprise value chain analysis (in
Chinese). Aeronaut Manuf Technol (8):55–59
Lan BX, Wang YM, Wang W (2011) Enterprise resource optimization and value chain analysis
(in Chinese). Chinese J Manag Sci (1):69–76
Li HF, Fan YS (2004) Workflow model analysis based on time constraint petri nets. Ruan Jian
Xue Bao/J Softw 15(1):17–26
Ling S, Schmidt H (2000) Time petri nets for workflow modelling and analysis. In: Proceedings
of the 2000 IEEE international conference on systems, man and cybernetics, Nashville,
pp 3039–3044
Pan Y, Tang Y, Tang N, Yu Y, Dao W (2005) A workflow model based on fuzzy-timing petri
nets. In: Proceedings of the 9th international conference on computer supported cooperative
work in design, Coventry, pp 541–546
Pang H, Fang ZD, Zhao Y (2008) Simplification analysis and schedulability verification of timing
constraint workflow model. Comput Integr Manuf Syst 14(11): 2217–2223 ? 2230
Spoede C, Henke EO, Unmble M (1994) Using activity analysis to locate profitability drivers.
Manag Account (3):43–48
77 A Solution to Optimize Enterprise Business 735
Wang ZZ (2007) Research of multimodal transportation process optimization based on petri net.
Vehicle operation engineering. Ph.D. dissertation, Jilin University, Jilin
Wang YP, Li SX, Wang ZZ, Li SW, Dong SW, Cui LX (2008) Optimization of production
logistics process of automobile manufacture enterprise based on petri net (in Chinese). J Jilin
Univ (Engineering and Technology Edition) 38(2):59–62
Zhang LQ (2010) Research on block-structured process mining techonology for business process
modeling. Computer software and theory, Ph.D. dissertation, Shandong University, Shandong
Zhou JX, Liu F, Chen LL, Liu RX (2008) Application status and prospects of ERP system to
China’s foundry enterprises (in Chinese). Foundry 57(9):885–891
Chapter 78
An Approach with Nested Partition
for Resource-Constrained Project
Scheduling Problem
Zhen-yuan Liu and Wen-min Yu
Abstract A time-based nested partition (NP) approach is proposed to solve
resource-constrained project scheduling problem (RCPSP) in this paper. In itera-
tion, one activity is selected as the base point of which the finish time interval
calculated by CPM is divided into two parts to form two subregions on the basis of
the promising region of the last iteration. Then sampling is taken in both subre-
gions and the surrounding region to determine the promising region and aggregate
the other as the surrounding region of this iteration so that whether the back-
tracking or the moving operation being performed is determined. Double justifi-
cation is also performed in iteration to improve the results. The results of
numerical tests on PSPLIB show the effectiveness and time-efficient of the pro-
posed NP method.
Keywords Double justification Nested partition RCPSP Sampling
78.1 Introduction
Since the nineties of the last century, resource-constrained project scheduling
problem (RCPSP) has become a standard problem generated by standard project
generator ProGen, which forms a resource-constrained project scheduling problem
Z. Liu (&) W. Yu
Institute of Systems Engineering, Huazhong University of Science
and Technology, Wuhan, China
e-mail: [email protected]
W. Yu
e-mail: [email protected]
Z. Liu W. Yu
Key Laboratory of Education Ministry for Image Processing
and Intelligent Control, Wuhan, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 737
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_78,
Springer-Verlag Berlin Heidelberg 2013
738 Z. Liu and W. Yu
library (Kolisch and Sprecher 1997). Many algorithms have been developed to
solve the problem and compare and analyze the results with experiment using the
standard problem in the library.
The essence of the resource-constrained project scheduling is to arrange the
execute time of each activity in the network under the constraints of resources and
precedence relations. There are three kinds of solution method which are opti-
mization (Bianco and Caramia 2012), heuristics (Kolisch 1996) and intelligent
algorithm (Kolisch and Hartmann 2006).
Nested Partition is to partition the feasible solution space so that more search
efforts can be expanded in the subregions that most likely contain the best solution.
One of its important features is its flexibility that it can incorporate many efficient
heuristics into its search procedure in order to get better solution. Another one is
that parallel computing capacities can be taken advantage of and searching in
subregions can be done independently and in parallel with only a little coordi-
nation overhead. Therefore, it’s usually used to solve large-scale problems
(Shi and ölafsson 2000).
In this paper, a time-based Nested Partition Framework is proposed to solve
RCPSP, where the operations in NP will be discussed respectively. And the whole
framework will be tested on PSPLIB.
78.2 Definition of Problem
The classic RCPSPs can be stated as the following. It’s assumed that a single
project consists of j = 1,…, J activities with a non-preemptive duration of dj
periods, respectively. Due to technological requirements, precedence relations
between some of the activities enforce that an activity j = 2,…, J may not be
started before all its immediate predecessors i 2 Pj (Pj is the set of immediate
predecessors of activity j) have been finished. Without loss of generalization, we
can assume that activity 1 is the only start activity and activity J is the only finish
activity. K types of renewable resources supplied by the partners will be consumed
during the project. It is assumed that the project needs rjk units of resource k to
process activity j during every period of its duration. Let At be the set of activities
being executed in period t.
The capacity of resource k supplied is noted by Rk . The due date of the project
is D. With a given D, we can get the earliest finish time ej and the latest finish time
lj of activity j by using Critical PathMethod (CPM). The time parameters in the
problem are all integer valued.
We use a set of continuous decision variables xj 2 ðej ; lj Þj ¼ 1; ; j to be the
finish time of activity j. The decision variable can be stated as X ¼ x1 ; x2 ; . . .; xj
jxj 2 ej ; lj ; j ¼ 1; . . .; jg.
78 An Approach with Nested Partition 739
The model of RCPSP can be presented as follows:
min xj ð78:1Þ
s:t: xi xj dj ; 8i 2 Pj ; j ¼ 1; . . .; j ð78:2Þ
X
rjk Rk ; k ¼ 1; . . .; D ð78:3Þ
j2At
xj 2 Zþ1 ; j ¼ 1; . . .; J ð78:4Þ
(78.1) is the objective that minimize the makespan of the problem, (78.2) dem-
onstrates the precedence relation constraints among activities, (78.3) demonstrates
the resource constraints, (78.4) demonstrates the natural constraints of each
activity’s finish time.
78.3 The Generic Nested Partition Framework
The Nested Partition method is partitioning and sampling based strategy. In iter-
ation of the algorithm, the entire solution space is viewed as a union that comprises
a promising region and a surrounding region. The four operations of Nested
Partition method are as follows.
(1) Partitioning: This step is to partition the current most promising region into
several subregions and aggregate the remaining regions into the surrounding
region. With an appropriate partitioning scheme, most of the good solutions would
be clustered together in a few subregions after the partitioning.
(2) Random Sampling: Samples are taken from the sub regions and the sur-
rounding region according to some sampling procedure. The procedure should
guarantee a positive probability for each solution in a given region to be selected.
As we would like to obtain high quality samples, it is often beneficial to utilize
problem structure in the sampling procedure.
(3) Calculation of the Promise Index: For each region, we calculate the promise
index to determine the most promising region. The promise index should be
represented as the performance of the objective.
(4) Moving: The new most promising region is either a child of the current most
promising region or the surrounding region. If more than one region is equally
promising, ties are broken arbitrarily. When the new most promising region is the
surrounding region, backtracking is performed. The algorithm can be devised to
backtrack to either the root node or any other node along the path leading to the
current promising region.
740 Z. Liu and W. Yu
78.4 Partition Operation
78.4.1 Definition of Feasible Space and Partition
We know the solution space is
X ¼ x1 ; x2 ; . . .; xj jxj 2 ej ; lj ; j ¼ 1; . . .; j;
xi xj dj ; 8i 2 Pj ; j1; . . .; J
X
j2At rjk Rk ; t ¼ 1; . . .; D; k ¼ 1; . . .; K:
By critical path method (CPM), the earliest finish time ej and latest finish time lj
of each activity j can be calculated as variable xj 2 ej ; lj with the relaxation of
resource constraints.
The whole searching space of the problem without resource-constraints and
precedence relations in the space can be described as:
X
: ½e1 ; l1 ½e2 ; l2 . . . ½eJ ; lJ :
78.4.2 Partition in the Searching Space
In order to partition the J-dimensional space, an activity is selected as the base
point of which feasible time interval is divided.
Different mapping mechanism of selecting the activities and different way to
decide the size of subregions as well as the number of the activities being selected
once would diverse the partition operation.
Since the time interval change of the activity would influence the time interval
of the activities that have precedence relationship with it, activity with more
precedence constraints in the network is preferred to make the partition more
effective. The end time by initial schedule of the selected activity is chosen as the
boundary point to get more precisely definition of the searching space as well as
efficiency and concentration of the searching. And only one activity is chosen
once. Here such mechanism of partition operation is named as Biased Partition.
Fig. 78.1 The example
2 5
activity network
1 3 8
4
7
78 An Approach with Nested Partition 741
Table 78.1 The parameters of activities in the project shown in Fig. 78.1
j 1 2 3 4 5 6 7 8
dj 0 4 5 1 8 1 10 0
rj 0 1 1 3 1 1 1 0
ej 0 4 5 1 12 2 11 12
lj 12 16 24 14 24 24 24 24
78.4.3 Subregion Redefinition
By CPM, change of the activity’s earliest finish time (EF) would influence the EF
of the activities that has direct or indirect successor relationship with it; change of
the activity’s latest finish time (LF) would influence the LF of the activities that
has direct or indirect precedence relationship with it.
Such influence is considered in order to narrow the searching space as possible.
Here complete partition operation is demonstrated below. Figure 78.1 shows the
network of a project.
In Table 78.1, j represents activity sequence, dj represents the time of activity
j being executed rj represents the resource consumed units of activity j per period
during the activity j being executed ej and lj represents the EF and LF of activity
j by CPM setting deadline D = 24 respectively. Initial searching space of the
project is denoted as follow:
[0, 12] 9 [4, 16] 9 [5, 24] 9 [1, 14] 9 [12, 24] 9 [2, 24] 9 [11, 24] 9 [12, 24]
Suggest that activity 2 is selected as the base point and the finish time interval
of activity 2 is divided into two parts as [4, 10] and [11, 16]. For the part [4, 10],
the LF is changed, which results in the modification of the activities that are
predecessors or transitive predecessors of activity 2. So the finish time interval of
activity 1 and activity 6 are changed into [0, 6] and [11, 16] respectively.
For the part [11, 16], the EF is changed, which results in the modification of the
activities that are successors or transitive successors of activity 2. So the finish
time interval of activity 5 and activity 8 are changed into [11, 16] and [19, 24].
78.5 Sampling Method
Sampling is to generate schedules. Here two basic schedules generate schemas
(SGS) is demonstrated as below.
78.5.1 Series Scheduling Schema
In every stage, an activity is selected to expand a partial schedule, setting its start
time and finish time without violating resource constraints and precedence con-
straints among activities. The procedure is as below.
742 Z. Liu and W. Yu
Definition:
X
pRkt :¼ Rk r ; 8k; t;
j2At jk
En :¼ jjj 62 Cn ; Pj Cn ;
Initialization:
n ¼ 1; Cn ¼ £;
while jCn j\JDostagen
define En ; pRkt ; t ¼ 1; . . .; T; k ¼ 1; . . .; K;
j ¼ min jjtð jÞ ¼ inf tðiÞ ;
j2En i2En
ej ¼ max FTj jj 2 Pj þ dj ;
( )
tjej t lj ;
FTj ¼ min 0
;
rj k pRkt0 ; t ¼ t dj þ 1; . . .; t; 8k
Cn ¼ Cn [ fj g;
n ¼ n þ 1;
78.5.2 Parallel Scheduling Schema
In every stage, there is a schedule time to be determined in order to release some
resource capacity, and then some activities are selected to do scheduling, setting
their start time and finish time without violating constraints. The procedure is as
below.
Definition:
X
pRk :¼ Rk r ; 8k;
j2A jk n
( )
jjj 62 fCn [ An g; Pj Cn ;
En :¼ ;
rjk pRk ; 8k; tn þ d 2 ej ; lj
( )
jjj 62 fCn [ An g; Pj Cn ;
En0 :¼ ;
rjk pRk ; 8k; tn þ d [ lj
( )
jjj 62 fCn [ An g; Pj Cn ;
En00 :¼ ;
rjk pRk ; 8k; tn þ d\ej
78 An Approach with Nested Partition 743
Initialization:
n ¼ 1; tn ¼ 0;
En ¼ f1g; An ¼ Cn ¼ £;
pRk :¼ Rk ; 8k;
whilejCn [ An j\JDostagen
ð1Þtn ¼ min FTj jj 2 An1 ;
An ¼ An1 n jjj 2 An1 ; FTj ¼ tn ;
Cn ¼ Cn1 [ jjj 2 An1 ; FTj ¼ tn ;
definepRk ; En ; EN0 ; En00 ;
if En0 6¼ £;
Exit without feasible solution;
if En 6¼ £; gotoð2Þ;
if En ¼ £ and if An 6¼ £; gotoð1Þ;
else if En00 6¼ £;
tn ¼ min ej dj jj 2 En00 ;
Re define En ; gotoð2Þ
ð2Þ j ¼ min jjtðjÞ ¼ inf tðjÞ ;
j2En i2En
STj ¼ tn ; FTj ¼ tn þ dj ;
An ¼ An [ fj g;
Re define pRk ; En ;
if En 6¼ £; gotoð2Þ;
elsen ¼ n þ 1;
78.5.3 Sampling Schema
Activities need to be selected in both serial schedule and parallel schedule generate
schema. To select appropriate activities and selecting different activities in
744 Z. Liu and W. Yu
different schedules can guide to better results and avoid repeated optimal solution.
The follow are several sampling schemas.
(1) Random sampling: Activities are selected randomly.
(2) Biased sampling: Activities are selected according to some mapping
mechanism that with a priority rule to get corresponding priority rule uð jÞ of
activity j, the feasible activity is selected by a probabilityUð jÞ ¼ P uj2Eð jÞuð jÞ
n
(3) Regret-based biased sampling: It’s similar to biased sampling; the differ-
ence is to set a set of regret value uð jÞ which compares the priority value of
activity j with the worst sequence in feasible activity set as qð jÞ :¼ maxi2En uð jÞ
uð jÞ where a ‘‘minimal’’ priority rule is employed. Then, the probability for j to be
selected is Uð jÞ. In these sampling schemas, the priority rules can be the same as
those in
ðqð jÞ þ 1Þa
Uð jÞ ¼ P a
i2En ðqðiÞ þ 1Þ
78.6 Optimization of the Solution
78.6.1 Promising Region Predefinition
We can get initial schedule from the last iteration that can be taken advantage in
this iteration for iteration depth deeper than 1. Therefore, the promising region can
be estimated with the finish time in initial schedule of the selected activity where
double searching effort would be expanded.
78.6.2 Double Justification
Double Justification is testified to be an efficient method to improve a feasible
solution which may be found by taking most kinds of heuristics (Vallsa and Ballest
2005). It consists of a backward pass and a forward pass.
(1) Backward pass: in a decreasing order, it is to completing a partial activity
sequence by scheduling each activity as late as possible based on a schedule
generated by a basic SGS.
(2) Forward pass is similar to backward pass in an ascending order.
The double justification process is presented in pseudo code as below.
Definition:
Si ¼ fjjFTi þ 1 [ ¼ STj g
Pi ¼ fjjSTi [ ¼ FTj þ 1g
78 An Approach with Nested Partition 745
X
pRkt ¼ Rk rjk ; 8k; t
j2OLn
OLn ¼ /
BackwardPass:
ComputeFTj ; j 2 ½1; J
descending order activity sequence by FTj
to createUL0 ¼ ðp0 ; p1 ::pJ1 Þ
n ¼ 0; i ¼ 1
WhileULn 6¼ /
Begin
NewLatestFT ¼ minSTSi 1
NewFTi ¼ maxftjFTi t NewLatestFT;
0
rik prkt0 ; t ¼ t di þ 1; . . .; t; 8kg
ULnþ1 ¼ ULn npi
OLnþ1 ¼ OLn [ pi
i¼iþ1
End
ForwardPass:
ComputeSTj ; j 2 ½1; n
ascending order activity sequence by STj
to createUL0 ¼ ðq1 ; q2 ::qn Þ
n ¼ 0; i ¼ 1
WhileULn 6¼ /
Begin
NewEaliestST ¼ maxFTPi þ 1
NewFTi ¼ maxftjFTi t NewLatestFT;
0
rik prkt0 ; t ¼ t di þ 1; . . .; t; 8kg
746 Z. Liu and W. Yu
ULnþ1 ¼ ULn nqi
OLnþ1 ¼ OLn [ qi
i¼iþ1
End
Pi is the set of immediate predecessors of activity i, Si is the set of immediate
successors of activity i. OLn represents union of the activity sequence that has been
rearranged at the step n. ULn represents union of the activity sequence that has not
been rearranged at the step n.
78.7 Moving
For each region dj , the promising index as the best performance value within the
region is calculated.
I dj ¼ minxj ; j ¼ 1; 2. . . M þ 1
Promising region is selected by
dj ¼ arg minxj
If promising region is one of the sub regions then we set n ¼ n þ 1 as well as
dn ¼ dj , and dn is to be partitioned in the next iteration.
Otherwise, backtrack operation is performed then we set n ¼ n 1 as well as
dn ¼ dn1 , and the promising region in the previous iteration is to be partitioned.
Here the iteration is ended when the schedule number reaches some specific
amount.
78.8 Computational Experiment
Based on some basic experiments, we paid more attention to the following con-
figuration of parameters in this NP framework: Biased Sizing, Regret-based Biased
Sampling with priority rule LFT. In addition, we employ Double Justification to
improve feasible solution. In the iteration we schedules 30 times in both promising
region and surrounding region in experiment of 1000 schedules and 120 times in
experiment of 5000 schedules.
We employ test sets J120 instance in PSPLIB and make a comparative study as
below (Table 78.2)
We can read from the table that NP method performs excellent among algo-
rithms that are not intelligent and good among all the algorithms.
78 An Approach with Nested Partition 747
Table 78.2 Average deviation from critical path lower bound-J120
Method Author 1000 5000
Frog-leaping Chen and Ling (2012) 34.83 33.2
ACCOS Chen et al. (2010) 35.19 32.48
Random GA Mendes (2003) 35.87 33.03
NP This study 36.14 35.47
ANGEL Tseng et al. (2006) 36.39 34.49
Improved GA Alcaraz et al. (2004) 36.53 33.91
Self-adapting GA Hartmann (2002) 37.19 35.39
BA-FBI Ziarati et al. (2011) 37.72 36.76
Activity list GA Hartmann (1998) 39.37 36.74
LFT sampling Kolisch (1996) 39.60 38.75
Adaptive sampling Schirmer and Riesenberg (1998) 39.85 38.70
Table 78.3 Computational time of different schedule method of J120
NP-SGS2 SGS2 NP-SGS1 SGS1 NP-SGS3 SGS3
1080 3.159 3.26 3.929 20.12 3.296 12.09
5400 14.39 14.79 18.63 – 11.65 –
The specification of Lenovo computer used in the experiment is as follow:
Intel(R) Core(TM) 2 Duo CPU [email protected] GHz, 2.93 GHz, 1.96 GB. SGS1
represents serial schedule. SGS2 represents parallel schedule. SGS3 represents
hybrid schedule. NP-SGS1 represents serial sampling in NP framework.
As Table 78.3 shows, NP method is time efficient especially in serial schedule
since it sharply decreases the searching space. For sampling with parallel schedule
in NP, though searching space is also decreased, the computing time spent on the
other three operations in NP especially the partition operation offsets it.
78.9 Conclusion
The paper shows how the NP framework is applied in the RCPSP.
As a global, open, fast convergence algorithm, NP method can be well applied
to the solution of the RCPSP problem. It performs well in getting the minimized
makespan of the project especially in greatly reducing the searching time.
Acknowledgments This work has been supported by the Chinese National Science Fund under
the Grant 71071062, the Hubei Science Fund under the Grant 2009 CDB242 and the Scientific
Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry of
China and Open Foundation of Key Laboratory of Education Ministry for Image Processing and
Intelligent Control.
748 Z. Liu and W. Yu
References
Alcaraz J, Maroto C, Ruiz R (2004) Improving the performance of genetic algorithms for the
RCPS problem. In: Proceedings of the 9th international workshop on project management and
scheduling, pp 40–43
Bianco L, Caramia M (2012) An exact algorithm to minimize the makespan in project scheduling
with scarce resources and generalized precedence relations. Eur J Oper Res 219(1):73–85
Chen F, Ling W (2012) An effective shuffled frog-leaping algorithm for resource-constrained
project scheduling problem. Comput Oper Res 39(5):890–901
Chen W, Shi YJ, Teng HF, Lan XP, Hu LC (2010) An efficient hybrid algorithm for resource-
constrained project scheduling. Inf Sci 180(6):1031–1039
Hartmann S (1998) A competitive genetic algorithm for resource-constrained project scheduling.
Naval Res Logist 45(7):279–302
Hartmann S (2002) A self-adapting genetic algorithm for project scheduling under resource
constraints. Naval Res Logist 49(5):433–448
Kolisch R (1996) Efficient priority rules for the resource-constrained project scheduling problem.
J Oper Manag 14(3):179–192
Kolisch R, Hartmann S (2006) Experimental investigation of heuristics for resource-constrained
project scheduling: an update. Eur J Oper Res 174(1):23–37
Kolisch R, Sprecher A (1997) PSPLIB-A project scheduling problem library. Eur J Oper Res
96(1):205–216
Mendes JJM (2003) A random key based genetic algorithm for the resource constrained project
scheduling problem workshop on computer science and information technologies. Ufa, Russia
Schirmer A, Riesenberg S (1998) Case-based reasoning and parameterized random sampling for
project scheduling. Technical report, University of Kiel, Germany
Shi L, ölafsson S (2000) Nested partitions method for global optimization. Oper Res
48(1):390–407
Tseng LY, Chen SC (2006) A hybrid metaheuristic for the resource-constrained project
scheduling problem. Eur J Oper Res 175(2):707–721
Vallsa V, Ballest F (2005) Justification and RCPSP: A technique that pays. Eur J Oper Res
165(2):375–386
Ziarati K, Akbaria R, Zeighamib V (2011) On the performance of bee algorithms for resource-
constrained project scheduling problem. Appl Soft Comput 11(4):3720–3733
Chapter 79
An Approximate Dynamic Programming
Approach for Computing Base Stock
Levels
Hong-zhi He
Abstract This paper studies the classical model for stochastic inventory control,
i.e. a finite horizon periodic review model without setup costs. A base-stock policy
is well known to be optimal for such systems. The author gives a new heuristic
computation procedure for calculating the base-stock levels. The idea is based on
approximate dynamic programming. A numerical example is provided to serve as
an illustration.
Keywords Approximate dynamic programming Base-stock policy Stochastic
inventory control
79.1 Introduction
This paper is an attempt to get an approximate solution in stochastic inventory
control problems by utilizing approximate dynamic programming. Although what
we study in the current paper is the simplest stochastic inventory system where
single location and single item are assumed, and with negligible setup costs, the
approach we use is of particular significance for solving large-scale problems
where multi-item and multi-location are assumed.
The idea is based on approximate dynamic programming. First the next-period
optimal value function is approximated by a quadratic polynomial. Then the base
stock level from this approximation is computed. After that, three points are taken
for the approximate optimal value to be computed, and then the Lagrange method
is utilized to get the quadratic polynomial.
H. He (&)
Department of Engineering Management, Luoyang Institute of Science and Technology,
Luoyang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 749
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_79,
Ó Springer-Verlag Berlin Heidelberg 2013
750 H. He
Using a quadratic polynomial function is reasonable. This is because the
optimal value function for the simplest inventory system possesses convexity and
hence exhibits a ‘U’ shape.
Base-stock policy is widely used in inventory systems where setup cost is
negligible. Such a policy is well known to be optimal for simplest single-item
stochastic inventory models (see, for example, Gaver 1959; Zipkin 2000).
Although the base-stock levels can be computed recursively by exact dynamic
programming, the classical computing procedure is nevertheless arduous. This
paper provides a new algorithm for computing such base stock levels. By this
algorithm, the computational effort can be greatly cut down, and extension to
multi-item, multi-location models is of much possibility.
Literatures in the realm of computation of base-stock levels mainly consider
infinite horizon models. Karlin (1960) studied the infinite horizon problem of the
most basic type. Zipkin (1989) used an alternative approach to compute the base-
stock levels. Roundy and Muckstadt (2000) studied an infinite horizon problem
with capacity-limited facility and gave a heuristic computation procedure. Zipkin
(2000) gave the solution for a continuous review, infinite horizon problem. Chen
and Song (2001) studied an infinite horizon, multiechelon inventory system.
Iglehard and Karlin (1962), Srinagesh and Sridhar (2001) and Levi et al. (2007)
studied the exact and approximation algorithms for stochastic inventory model
with correlated and non-stationary demand. Song and Zipkin (1993) provided both
a value iteration and an exact algorithm for computing base stock levels in a
fluctuating demand environment. Huh et al. (2009) provided an adaptive algorithm
to minimize the long run average cost. There are a few papers addressing finite
horizon models, and they mainly use myopic policies as heuristics. See, for
example, Lovejoy (1992) and Morton and Pentico (1995). The only exception is
Iida and Zipkin (2006), who used a piecewise linear function to approximate the
cost function. Their idea is similar to ours.
Surprisingly, literatures for computing finite horizon base-stock levels with
quadratic function approximations are nonexistent, although the idea is straight-
forward. This paper fills into this gap.
79.2 Formulation
This paper considers the simplest inventory system where setup cost is negligible.
The inventory policy is implemented on a periodic-review basis. The finite time
horizon is assumed to be T periods.
Let Jt ðyÞ represents the optimal value function at time t, x represents the order-
up-to level, Dt represents the random customers’ demand during time t and t ? 1,
ft ðDt Þ represents the density function of the demand distribution during period t,
Ft ðDt Þ represents the cumulative distribution function of customers’ demand
during period t, p represents penalty cost for one unit of backlogged product during
79 An Approximate Dynamic Programming Approach 751
one period, h represents the unit holding cost, c represents the unit purchasing cost,
the stochastic inventory control model can be represented as:
Jt ðyÞ ¼ min Efhðx Dt Þþ þ pðDt xÞþ
xy
þ cðx yÞ þ Jtþ1 ðx Dt Þg
Zx Z1
¼ min h ðx Dt Þf t ðDt ÞdDt þ p ðDt xÞf t ðDt ÞdDt ð79:1Þ
xy
0 x
Z1
þ cðx yÞ þ Jtþ1 ðx Dt Þf t ðDt ÞdDt :
0
It’s well-known that this value function is convex, and the optimal policy is a
base stock policy. Computing the base stock level s becomes the key question in
this paper. In the end of the last period T, the salvage value function is assumed to
be JTþ1 ðxÞ, which is a convex, decreasing and differentiable function in x.
79.3 The Computation Procedure
The computing procedure is described in this section. In the beginning of period T,
solve the base stock level. In (79.1), taking derivative over x, one has
Zx Z1 Z1
h f T ðDT ÞdT p f T ðDT ÞdDT þ cðx yÞ þ JTþ1 ðx T Þf T ðDT ÞdDT ¼ 0;
0 x 0
ð79:2Þ
i.e.
Z1
hFT ðxÞ pð1 FT ðxÞÞ þ c þ J0Tþ1 ðx DT Þf T ðDT ÞdDT ¼ 0: ð79:3Þ
0
It deduces
Z1
ðh þ pÞFTðxÞ ¼ p J0Tþ1 ðx DT Þf T ðDT ÞT c: ð79:4Þ
0
This equation determines the last-period base stock level sT , i.e. sT should
satisfy
752 H. He
R1
p J0Tþ1 ðx DT Þf T ðDT ÞdDT c
0
FT ðsT Þ ¼ ð79:5Þ
hþp
Now let’s interpolate the value function for period T by a quadratic function.
Take three points: y ¼ s2T ; y ¼ 3s2T ; y ¼ 2sT . For y ¼ s2T ; from base stock
policy we have x ¼ sT at minimum. Value function
hs i ZsT Z1
T
JT ¼h ðsT DT ÞfT ðDT ÞdDT þ p ðDT sT ÞfT ðDT ÞdDT
2
0 sT
ð79:6Þ
hs i Z1
T
þ c sT þ JTþ1 ðsT DT Þf T ðDT ÞdDT :
2
0
3s 3s
For y ¼ 2
T
; from base stock policy we have x ¼ 2
T
at minimum. Value
function
3sT
Z½ 2
Z1
3sT 3sT 3sT
JT ¼h DT fT ðDT ÞdDT þ p DT fT ðDT ÞdDT
2 2 2
0 3sT
½2
Z1
3sT
þ JTþ1 DT f T ðDT ÞdDT :
2
0
ð79:7Þ
For y ¼ 2sT ; from base stock policy we have x ¼ 2sT at minimum. Value function
Z2sT
JT ð2sT Þ ¼ h ð2sT DT ÞfT ðDT ÞdDT
0
Z1 Z1
þp ðDT 2sT ÞfT ðDT ÞdDT þ JTþ1 ð2sT DT Þf T ðDT ÞdDT :
2sT 0
ð79:8Þ
Let a quadratic function ~ Jtþ1 ðyÞ ¼ atþ1 y2 þ btþ1 y þ ctþ1 be an approximate
value function. Put the above three pair of points into the approximate value
function, one gets atþ1 ; btþ1 ; ctþ1 .
Put ~
Jtþ1 into the dynamic programming equation,
79 An Approximate Dynamic Programming Approach 753
Jt ðyÞ ¼ min Efhðx Dt Þþ þ pðDt xÞþ þ cðx yÞ þ ~Jtþ1 ðx Dt Þg
x;x y
Zx Z1
¼ min h ðx Dt Þf t ðDt ÞdDt þ p ðDt xÞf t ðDt ÞdDt þ cðx yÞ
xy
0 x
Z1h i
þ atþ1 ðx Dt Þ2 þbtþ1 ðx Dt Þ þ ctþ1 f t ðDt ÞdDt :
0
ð79:9Þ
Calculate the function values at three points, then use the Lagrange method to
get an interpolation polynomial, and use this polynomial to approximate the value
function in (79.1). Solve st by
Z1
hFt ðxÞ pð1 Ft ðÞÞ þ c þ ½2atþ1 ðx Dt Þ þ btþ1 f t ðDt ÞdDt ¼ 0; ð79:10Þ
0
i.e.
ðh þ pÞFt ðxÞ þ 2atþ1 x ¼ p þ 2kt atþ1 btþ1 c: ð79:11Þ
s 3s
One solves st by the above equation. Then let y ¼ 2t ; 2 t ; 2st , and compute
Jt s2t ,
Jt 32st and Jt ð2st Þ. For y ¼ s2t , x ¼ sT ,
Zst Z1
st st
Jt ¼h ðst Dt Þft ðDt ÞdDt þ p ðDt st Þft ðDt ÞdDt þ c st
2 2
0 st
Z1
þ ~Jtþ1 ðst Dt Þf t ðDt ÞdDt
0
Zst Z1
st
¼h ðst Dt Þf t ðDt ÞdDt þ p ðDt st Þf t ðDt ÞdDt þ c st
2
0 x
Z1
þ ½atþ1 ðst Dt Þ2 þ btþ1 ðst Dt Þ þ ctþ1 f t ðDt ÞdDt :
0
ð79:12Þ
3s 3s
For y ¼ 2
t
;x ¼ 2
t
,
754 H. He
3st
Z½ 2 Z1
3st 3st 3st
Jt ¼h DT fT ðDT ÞdDT þ p DT fT ðDT ÞdDT
2 2 2
0 ½32st
Z1
~ 3st
þ Jtþ1 Dt f t ðDt ÞdDt
2
0
3st
Z½
2 Z1
3st 3st
¼h Dt f t ðDt ÞdDt þ p Dt f t ðDt ÞdDt
2 2
0 3
s
½ 2t
Z1 " #
2
3st 3st
þ atþ1 Dt þbtþ1 Dt þ ctþ1 f t ðDt ÞdDt :
2 2
0
ð79:13Þ
For y ¼ 2st ; x ¼ 2st ,
3st
Z½ 2 Z1
3st 3st
Jt ð2st Þ ¼ h DT fT ðDT ÞdDT þ p DT fT ðDT ÞdDT
2 2
0 3st
½2
Z1
3st
þ ~Jtþ1 Dt f t ðDt ÞdDt
2
0
3st
Z½ 2
Z1
3st 3st
¼h Dt f t ðDt ÞdDt þ p Dt f t ðDÞdDt
2 2
0 ½32st
" #
2
3st 3st
þ atþ1 Dt þbtþ1 Dt þ ctþ1 f t ðDt ÞdDt :
2 2
ð79:14Þ
Fitting the above three (y, x) pairs into quadratic function ~Jt ðyÞ ¼ at y2 þ bt y þ ct ,
one gets at ; bt and ct .
One loop is finished. Repeating the above procedure, one gets fst ; Jt g
for t ¼ T 1; T 2; . . .; 1:
79 An Approximate Dynamic Programming Approach 755
79.4 A Numerical Example
Let p = 10, c = 3, h = 1, T = 12 and Dt follow a Poisson process with a mean
rate k = 4. Let J13 ðxÞ ¼ cx be the last period salvage value function. From
(79.4), in the beginning of the last period: s12 satisfies
ð1 þ 10ÞFðs12 Þ ¼ 10: ð79:15Þ
s 3s
Thus it deduces that s12 ¼ 7. Therefore 212 ¼ 3, 212 ¼ 10, 2s12 ¼ 14. Inputting
y = 3, y = 10, y = 14 into (79.1) one gets J12 ð3Þ ¼ 6:9302, J12 ð10Þ ¼ 5:0698
and J12 ð14Þ ¼ 11:9560.
By using the Lagrange interpolation method one gets
~
J12 ðyÞ ¼ 0:1833y2 4:8330y þ 19:7794:
Inputting the above expression into (79.8), and then taking derivatives over x
one gets:
Z1
hF ð xÞ pð1 F ð xÞÞ þ c þ 2a12 x þ b12 2a12 D11 f ðD11 ÞdD11 ¼ 0; ð79:16Þ
0
i.e. ðh þ pÞF ð xÞ þ 2a12 x ¼ p þ 8a12 b12 c, which deduces 11F (x) ? 2
* 0.1833x = 10 ? 8 * 0.1833 ? 4.8330 - 3. Thus s11 is solved: s11 ¼ 7:
Inputting y = 3, y = 10, y = 14 into (79.8), one gets: y = 3, x = 7,
X
7
ð7 kÞ4k X
1
ðk 7Þ4k
~
J 11 ð3Þ ¼ 1 e4 þ 10 e4 þ 3 4
0
k! 7
k!
1 h
X i 4k
þ a12 ð7 kÞ2 þ b12 ð7 kÞ þ c12 e4 :
0
k!
By using Matlab one gets ~J11 ð3Þ ¼ 41:8583.
y = 10, x = 10,
X
10
ð10 kÞ4k X
1
ðk 10Þ4k
~
J 11 ð10Þ ¼ 1 e4 þ 10 e4
0
k! 10
k!
1 h
X i 4k
þ a12 ð10 kÞ2 þ b12 ð10 kÞ þ c12 e4 :
0
k!
By using Matlab one gets ~J11 ð10Þ ¼ 27:8862:
y = 14, x = 14,
756 H. He
X
14
ð14 kÞ4k X
1
ðk 14Þ4k
~
J11 ð14Þ ¼ 1 e4 þ 10 e4
0
k! 14
k!
1 h
X i 4k
þ a12 ð14 kÞ2 þ b12 ð14 kÞ þ c12 e4 :
0
k!
By using Matlab one gets ~J11 ð14Þ ¼ 36:0792:
By using the Lagrange interpolation method one gets
~
J11 ð yÞ ¼ 0:3677y2 6:7756y þ 58:8762:
The data for the following periods are listed in Table 79.1.
Finally, we solved the base stock level for period 1 is 14. It departs largely from
the expected mean of one-period demand (k = 4). This is because of the accu-
mulative effect of the relatively high penalty cost. The base stock levels are
roughly decreasing over time periods, which validates the intuition that the base
stock levels grow as the remaining time horizon grows. The base stock level for
period 1 is rather high compared with the mean demand in that period. This is
because the holding cost is low compared with the penalty cost.
Table 79.1 Data for periods
t st y x ~
Jt ð yÞ
11 7 3 7 0:3677y2 6:7756y þ 58:8762
10 10
14 14
10 8 4 8 0:3875y2 7:4077y þ 63:6730
12 12
16 16
9 9 5 9 0:3516y2 8:3598y + 88:0152
14 14
18 18
8 10 5 10 0:2886y2 7:9682y + 105:5914
15 15
20 20
7 11 5 11 0:2354y2 7:2567y + 120:5217
16 16
22 22
6 11 5 11 0:1914y2 6:4681y + 134:7710
16 16
22 22
5 11 5 11 0:1654y2 5:8599y + 149:9170
16 16
22 22
4 10 5 10 0:1374y2 5:3099y + 165:3747
15 15
20 20
3 10 5 10 0:3122y2 9:6663y + 200:5834
15 15
20 20
(continued)
79 An Approximate Dynamic Programming Approach 757
Table 79.1 (continued)
t st y x ~
Jt ð yÞ
2 14 7 14 0:2739y2 9:2883y + 218:9859
21 21
28 28
1 14
79.5 Conclusion
In this paper, a single item/single location inventory system is studied. A compu-
tation procedure is devised, and a numerical example is provided. The computation
procedure proves to work effectively. Because the value function in multi-item and
multi-location systems can be approximated by a quadratic polynomial, the method
in this paper will possibly solve stochastic inventory management problems in such
systems approximately. This will be a future research direction for the authors.
References
Chen F, Song JS (2001) Optimal policies for multiechelon inventory problems with Markov-
modulated demand. Oper Res 49:226–234
Gaver DP (1959) On base-stock level inventory control. Oper Res 7:689–703
Huh WT, Janakiraman G, Muckstadt JA, Rusmevichientong P (2009) An adaptive algorithm for
finding the optimal base-stock policy in lost sales inventory systems with censored demand.
Math Oper Res 34(2):397–416
Iglehard D, Karlin S (1962) Optimal policy for dynamic inventory process with non-stationary
stochastic demands. In: Arrow KJ, Karlin S, Scarf H (eds) Studies in applied probability and
management science. Stanford University Press, Stanford, pp 127–147
Iida T, Zipkin PH (2006) Approximate solutions of a dynamic forecast-inventory model. Manuf
Serv Oper Manag 8:407–425
Karlin S (1960) Dynamic inventory policy with varying stochastic demands. Manag Sci
6:231–258
Levi R, Pál M, Roundy RO, Shmoys DB (2007) Approximation algorithms for stochastic
inventory control models. Math Oper Res 32:284–302
Lovejoy W (1992) Stopped myopic policies in some inventory models with generalized demand
processes. Manag Sci 38:688–707
Morton TE, Pentico DW (1995) The finite horizon nonstationary stochastic inventory problem:
near-myopic bounds, heuristics, testing. Manag Sci 41:334–343
Roundy RO, Muckstadt JA (2000) Heuristic computation of periodic-review base stock inventory
policies. Manag Sci 46:104–109
Song J, Zipkin P (1993) Inventory control in a fluctuating demand environment. Oper Res
41(2):351–370
Srinagesh G, Sridhar T (2001) An efficient procedure for non-stationary inventory control. IIE
Trans 33:83–89
Zipkin P (1989) Critical number policies for inventory models with periodic data. Manag Sci
35:71–80
Zipkin P (2000) Foundations of inventory management. The McGraw-Hill Companies, Boston,
pp 375–378
Chapter 80
An Inventory Model Considering Credit
Cost and Demand Rate Varying
with Credit Value
Zi-quan Long and Ran Gao
Abstract The credit of suppliers is influenced by factors such as quality of
products, promptness of delivering, service level, etc. Reduction in credit value
can lead to suppliers’ loss of profit, and this profit loss is named credit cost. Putting
forward the concept of credit cost for the first time, this paper, starting from
delivering situation, studied how backorder affects credit cost, established two
models considering respectively the existence and absence of communication
between suppliers and customers, and under the condition that credit loss affects
demand rate, verified the existence of credit cost utilizing numerical calculation.
Keywords Credit cost Credit value Demand rate Inventory model
80.1 Introduction
Costs produced by an inventory system in the operational process, which include
ordering cost, holding cost, and backorder cost, are a major criterion for the
evaluation of inventory controlling policy (Jie et al. 2011; Zhao and Huang 2008).
The above three types of cost vary with respect to change in credit of an inventory
company: for instance, price discounts and transaction costs differ due to different
credits, and demand rate drops while credit declines (Gupta and Vrat 1986; Henery
1990; Hariga 1994; Pando and Garcia-Laguna 2011). Credit cost, the cost brought
about by creditability factors, is the hidden cost incorporated in the three types of
cost (Ritchie 1980).
A variety of factors, such as price and quality of goods, transportation cost,
delivering situation and service level, can affect the major credit of inventory
Z. Long (&) R. Gao
School of Economic and Management, Wuhan University, Wuhan 430072, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 759
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_80,
Ó Springer-Verlag Berlin Heidelberg 2013
760 Z. Long and R. Gao
(Paul et al. 1996; Pando 2011; Zhou and Wang 2009). Assume the ideal condition
is that when a supplier provides a certain product to a fixed group of customers, the
supplier will not take the initiative to increase its credit (for example by adver-
tising), and consequently the system will be constantly involved in a vicious circle
where backorder leads to the drop of credit which subsequently cause the reduction
in demand rate. Under such premise, the supplier needs to formulate its optimal
inventory policy to achieve maximum operating profit. This paper will analyze
how backorder affects credit cost, starting from delivering situations.
80.2 Model Assumptions
The following assumptions are made to simplify the researching model: (1) Initial
demand rate is a constant (assume it is 1), and loss of accumulative credit caused
by backorder directly affects demand rate; (2) Replenishment is made instanta-
neously and replenishing ability is infinite, i.e. any amount of replenishment can be
achieved promptly; (3) There is no lead time for ordering, i.e. goods are replen-
ished instantaneously when orders are made; (4) backorder is allowed; (5)
Ordering time of goods is finite, but the operational process begins and ends
without backorder; (6) Price discount is not considered.
Definitions of each notation included in the model are as follows:
H = time span of ordering;
c1 = inventory holding cost per unit per time span;
c2 = backorder cost per unit per time span;
c3 = cost per order;
ti = time point of backorder;
Ti = time point of ordering;
li = time span during which we have inventory in hand in the ith cycle,
li ¼ ti Ti1 ;
Li = total time span of the ith cycle, Li ¼ Ti Ti1 ;
a = credit losing coefficient, i.e. the proportion of people who lose faith in the
company when backorder occurs, a 2 ½0; 1;
Pi = the accumulated credit after the ith cycle, and
l1 l2 l3 li
Pi ¼ 1 a i þ þ þ þ ;
L1 L2 L3 Li
ki = demand rate in the ith cycle, ki ¼ Pi1 k, where k is the initial demand rate
which is a constant;
Qi = amount of order at Ti1 ;
s = selling price per unit;
k = purchase price per unit.
80 An Inventory Model Considering Credit Cost and Demand 761
Fig. 80.1 Change of
inventory level in Model A
-
Amount of Inventory
O t1 T1 t 2 T2 Tn-1 Tn
Time
80.3 Mathematical Model
80.3.1 Model A
Assume that there is no exchange of information among customers within a certain
cycle of supply, and the customers receive the complete backorder information at
the end of the cycle. In this case, demand rate changes in segments with respect to
credit. Change of inventory level with respect to time is shown in Fig. 80.1.
The ith order occurs at t ¼ Ti1 . The storing cost during Ti1 t Ti is:
1 l1 l2 li1
1 ði 1Þa þ a þ þ þ kc1 l2i ;
2 L1 L2 Li1
the backorder cost is:
1 l1 l2 li1
1 ði 1Þa þ a þ þ þ kc2 ðLi li Þ2 ;
2 L1 L2 Li1
and the cost per order is c3 .
Then the total cost during Ti1 t Ti is:
1 l1 l2 li1
CðLi Þ ¼ 1 ði 1Þa þ a þ þ þ kc1 l2i
2 L1 L2 Li1
1 l1 l2 li1
þ 1 ði 1Þa þ a þ þ þ kc2 ðLi li Þ2 þc3
2 L1 L2 Li1
ð80:1Þ
At the end of the ith cycle, the accumulated credit of the supplier is:
l1 l2 l3 li
Pi ¼ 1 a i þ þ þ þ
L1 L2 L3 Li
Total cost within the time span H is:
762 Z. Long and R. Gao
TC ¼ C ðL1 Þ þ C ðL2 Þ þ þ C ðLn Þ;
and total revenue is:
X
n
TI ¼ ðs kÞ Qi ;
i¼1
thus total profit is:
TP ¼ TI TC:
Since TP is a function of ti , Ti and n, the above problem can be written as the
following nonlinear programming problem:
max TPðn; li ; Li Þ
s:t: li Li
ln ¼ Ln ð80:2Þ
L1 þ L2 þ þ Ln ¼ H
Li 0; li 0
80.3.2 Model B
Assume that there is exchange of information among the customers within a
certain cycle of supply, and each customer possesses complete information. In this
case, demand rate changes continuously with respect to credit. Change of inven-
tory level with respect to time is shown in Fig. 80.2.
The ith order occurs at t ¼ Ti1 . The storing cost during Ti1 t Ti is:
Fig. 80.2 Change of
inventory level in Model B -
Amount of Inventory
O t1 T1 t2 T2 Tn-1 Tn
⎛ t − t1 ⎞
⎜1 − α ⎟λ
⎝ t ⎠
Time
80 An Inventory Model Considering Credit Cost and Demand 763
1 l1 l2 li1
1 ði 1Þa þ a þ þ þ kc1 l2i ;
2 L1 L2 Li1
the backorder cost is:
ZTi
l1 l2 li
kc2 ðTi tÞ 1 ia þ a þ a þ þ a dt
L1 L2 t Ti1
ti
and the cost per order is c3 .
Then the total cost during Ti1 t Ti is:
1 l1 l2 li1
C ðLi Þ ¼ 1 ði 1Þa þ a þ þ þ kc1 l2i þ
2 L1 L2 Li1
ZTi ð80:3Þ
l1 l2 li
kc2 ðTi tÞ 1 ia þ a þ a þ þ a dt þ c3
L1 L2 t Ti1
ti
At the end of the ith cycle, the accumulated credit of the supplier is:
l1 l2 l3 li
Pi ¼ 1 a i þ þ þ þ :
L1 L2 L3 Li
Total cost within the time span H is:
TC ¼ C ðL1 Þ þ CðL2 Þ þ þ C ðLn Þ
P
and total revenue is TI ¼ ðs kÞ ni¼1 Qi ; thus the total profit is TP ¼ TI TC.
Same as Model A, it is a nonlinear programming problem.
80.4 Example of Calculation
Here is a specific example of calculation. The numerical values of the systems
factors are as follows: c1 = 5; c2 = 2; c3 = 2000; H = 30; k ¼ 200; s ¼ 130; and
k ¼ 100.
First, let us consider the situation when n = 2, i.e., there are two orders within a
fixed time period. LINGO is used to solve the decision making problem with the
value of a varying, and the results are demonstrated in the Table 80.1.
We can observe from Table 80.1 that: 1. The difference between L1 and l1
decreases as a increases, which indicates that the more stringent are people’s
requirement to the company, the less backorder will occur; for instance, backorder
can lead to considerable loss of customers when a company is facing numerous
competitors, hence the company will reduce backorder to the greatest extent; and
vice versa. 2. TP drops as a rises. When a ¼ 0, backorder will not cause loss of
credit, and this is indeed the inventory model in which credit cost is not considered.
764 Z. Long and R. Gao
Table 80.1 Sensitivity analysis about a
a l*1 L*1 TP*
Model A Model B Model A Model B Model A Model B
1 6.9735 8.6109 22.5300 21.4670 63701.16 40458.38
0.9 6.9666 8.4700 22.6835 21.8280 64879.24 43443.88
0.8 6.9501 8.3071 22.8092 22.1378 66074.18 46587.89
0.7 6.9267 8.1275 22.9139 22.4010 67282.84 49874.96
0.6 6.8982 7.9355 23.0023 22.6225 68503.01 53291.80
0.5 6.8657 7.7342 23.0779 22.8073 69733.10 56827.15
0.4 6.8300 7.5265 23.1432 22.9602 70971.95 60471.61
0.3 6.7919 7.3144 23.2001 23.0857 72218.67 64217.29
0.2 6.7517 7.0994 23.2500 23.1876 73472.58 68057.56
0.1 6.7099 6.8832 23.2941 23.2692 74733.17 71986.77
0.0 6.6667 6.6667 23.3333 23.3333 76000.00 76000.00
As a increases from 0 to 1, TP declines, and the amount of its reduction is in fact the
cost defrayed by the company owing to reduced credit, i.e., the credit cost. 3. When
the value of a is fixed, the credit cost in model B is generally larger than that in model
A, and this is due to the loss brought about by dissemination of information.
Next, let us consider the situation when n 3, i.e., there are more than two
orders within a fixed time period. LINGO is used to solve the problem with the
value of n or a varying, and the results are showed in the Table 80.2.
We can see from Table 80.2 that: 1. When a is fixed and n ascends from 2 to 4,
total profit rises. 2. For a fixed value of a, relative credit cost does not increase
regularly with the rise of n, but depends on the value of a. When a exceeds the
critical point, the opposite situation will occur.
Table 80.2 TP* and relative credit cost when n or a varies
n 2 3 4
a *
TP Relative credit TP* Relative credit TP* Relative credit
cost cost cost
1 63701.16 0.1618 79855.07 0.3218 91820.71 0.3089
0.9 64879.24 0.1463 81041.56 0.3117 92732.61 0.3021
0.8 66074.18 0.1306 82241.15 0.3016 93713.33 0.2947
0.7 67282.84 0.1147 83572.40 0.2903 94744.21 0.2869
0.6 68503.01 0.0986 87004.28 0.2611 95183.13 0.2836
0.5 69733.10 0.0825 91059.03 0.2267 98334.33 0.2599
0.4 70971.95 0.0662 95762.20 0.1867 102209.2 0.2308
0.3 72218.67 0.0498 100823.7 0.1437 103807.6 0.2187
0.2 73472.58 0.0333 106182.1 0.0982 109464.3 0.1762
0.1 74733.17 0.0167 111822.3 0.0503 121840.6 0.0830
0.0 76000.00 – 117750.0 – 132869.6 –
Note Credit cost is TP* -76000, and relative credit cost is the ratio of credit cost to 76000
80 An Inventory Model Considering Credit Cost and Demand 765
80.5 Conclusion
This paper proposed for the first time the concept of credit cost, analyzed how
backorder affects credit cost starting with delivering conditions, considered the
influence of credit loss to demand rate, and derived an appropriate model which
verified the existence of credit cost with specific numerical calculation. The result
indicates that suppliers will reduce backorder while the credit losing coefficient a
increases; less profit will be achieved in models which take credit cost into con-
sideration compared with those do not, and the reduction in profit is in fact credit
cost; rate of change in credit cost descend gradually as a increases; besides, when a
is the same, the credit cost in model B is generally larger than that in model A
owing to the fact that concealing backorder information will reduce credit cost.
References
Gupta R, Vrat P (1986) Inventory models for stock dependent consumption rate. Oper Res
23(01):19–24
Hariga M (1994) Inventory lot-sizing problem with continuous time-varying demand and
shortages. J Oper Res Soc 45(07):827–837
Henery RJ (1990) Inventory replenishment policy for cyclic demand. J Oper Res Soc
41(07):639–643
Jie M, Zhou Y, Liu Y, Ou J (2011) Supply chain inventory model with two levels of trade credit
and time varying demand. Syst Eng Theory Pract 31(2):262–269
Pando V, Garcia-Laguna J (2011) Maximizing profits in an inventory model with both demand
rate and holding cost per unit time dependent on the stock level. Comput Ind Eng
62(2):599–608
Paul K, Datte TK, Chaudhuri KS, Pal AK (1996) Inventory model with two component demand
rate and shortages. J Oper Res Soc 47(08):1029–1036
Ritchie E (1980) Practical inventory replenishment policies for a linear trend in demand followed
by a period of steady demand. J Oper Res Soc 31(07):605–613
Zhao X, Huang S (2008) Inventory management. Tsinghua University Press, Beijing
Zhou Y, Wang S (2009) Theory and method for inventory control. Science Press, Beijing
Chapter 81
An M/M/1 Queue System with Single
Working Vacation and Impatient
Customers
Xiao-ming Yu, De-an Wu, Lei Wu, Yi-bing Lu and Jiang-yan Peng
Abstract In this paper, an M/M/1 queueing system with a single working vacation
and impatient customers is considered. In this system, the server has a slow rate to
serve during a working vacation and customers become impatient due to a slow
service rate. The server waits dormant to the first arrival in case that the server
comes back to an empty system from a vacation, thereafter, opens a busy period.
Otherwise, the server starts a busy period directly if the queue system has cus-
tomers. The customers’ impatient time follows independently exponential distri-
bution. If the customer’s service has not been completed before the customer
becomes impatient, the customer abandons the queue and doesn’t return. The
model is analyzed and various performance measures are derived. Finally, several
numerical examples are presented.
Keywords Impatient customers M/M/1 queue Probability generating functions
Single working vacation
81.1 Introduction
Queueing systems with customer impatience such as hospital emergency rooms
handling critical patients, and inventory systems that store perishable goods, is
very popular. Due to their potential application, many authors are interest in
studying the queueing systems with impatient customers and treat the impatience
X. Yu D. Wu (&) L. Wu J. Peng
School of Mathematical Sciences, University of Electronic Science
and Technology of China, Chengdu, People’s Republic of China
e-mail: [email protected]
D. Wu Y. Lu
Faculty of Science, Kunming University of Science and Technology,
Kunming, Yunnan, People’s Republic of China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 767
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_81,
Springer-Verlag Berlin Heidelberg 2013
768 X. Yu et al.
phenomenon by various assumptions. The first to analyze queueing systems with
impatient customers seems to be Palm’s pioneering work (1953) by considering
the M=M=c queue, where the waiting times is assumed to be independent expo-
nential distribution. Daley (1965) studied the GI/G/1 queue with impatient cus-
tomers, in which customers may abandons the system before starting or
completing their service when they have to wait too long. And these results are
extended by several authors in many different directions. A customer abandons the
queue when he has a long wait already experienced, or a long wait anticipated
upon arrival, for example Takacs (1974), Baccelli et al. (1984), Boxma and de
Waal (1994), Van Houdt et al. (2003), and Yue and Yue (2009). Recently, the
queueing models with server vacation and impatient customers are analized by
Altman and Yechiali (2006) .
This work is supported by the Fundamental Research Funds for the Central
Universities, No. ZYGX2010J111, No. ZYGX2011J102, and Yunnan Provincial
Natural Science Foundation of China under Grant No. 2011FZ025.
In 2002, A class of semi-vacation policies had been introduced by Servi and
Finn (2002). In this system the server don’t completely stop service but works at a
lower rate during a vacation. This system is called a working vacation (WV)
system. The vacation time is assumed to be exponentially distributed. The service
times during a regular service period and during a working vacation follow
exponential distribution but with different rates. Zhao et al. (2008) ultilize the
quasi birth and death process and the matrix-geometric solution method to study an
M/M/1 queueing system with a single working vacation and obtain various per-
formance measures. An M/G/1 queue with multiple working vacations has been
studied by Wu and Takagi (2006). In addition, many authors extended these results
such as Baba (2005), Li et al. (2009).
However, there are few literatures which take customers’ impatient during a
working vacation into consideration. An M/M/1 queueing system with working
vacations and impatient customers is considered by Yue and Yue (2011). We
extend the research in Yue and Yue 2011 to an M/M/1 queueing system with a
single working vacation and impatient customers. For practical application, the
single working vacation policy has been widely used in managing science. When
the system has relatively few the number of customers, it becomes important to
economize operation cost and energy consumption, so the system needs to work at
a lower rate operation state. The model (Yue and Yue 2011) is one case of our
analysis. We analyze the queueing system where the server has a slow rate to serve
during a working vacation and customers become impatient due to a slow service
rate in this paper. The server waits dormant to the first arrival in case that the
server comes back to an empty system from a vacation, thereafter, opens a busy
period. Otherwise, the server starts a busy period directly if the queue system has
customers. The customers’ impatient time follows independently exponential
distribution. If the customer’s service has not been completed before the customer
becomes impatient, the customer abandons the queue and doesn’t return.
We organize the rest of this paper as follows: the model description is given in
Sect. 81.2. Then we derive the balance equations for the system. A differential
81 An M/M/1 Queue System with Single Working Vacation and Impatient Customers 769
equation for G0 ðzÞ is obtained, where G0 ðzÞ is the generating function of the queue
size when the server is on vacation. It is easy to calculate the fractions of time the
server being in working vacation or busy period. Various performance measures
including the mean system size, the mean sojourn time of a customer served is
obtained. Section 81.3 gives are some numerical results.
81.2 Analysis of the Model
81.2.1 Description of the Model
We study an M/M/1 queue with a single working vacation and impatient cus-
tomers. We assume that arrival process follows Poisson proccess with parameter k,
and the rule of serving is on a first-come first-served (FCFS) basis. If the server
returns from a working vacation to find the system empty, he waits dormant to the
first arrival thereafter and open a busy period. The working vacation follows
exponential distribution with parameter c, and the service times during a service
period and a working vacation follow exponential distribution with parameters lb
and lv , respectively, where lb [ lv . During the working vacation, customers
become impatient and the impatient times follows exponential distribution with
parameter n. If the customer’s service has not been completed before the customer
becomes impatient, the customer abandons the queue and doesn’t return.
Remark 1. If lv ¼ 0, the current model becomes the M/M/1 queueing model
which has a single vacation and impatient customers, and was studied by Altman
and Yechiali (2006). If n ¼ 0, the current model has a single working vacation, it
was studied by Zhao et al. (2008).
81.2.2 Balance Equations of the Model
The total number of customers in the system and the number of working servers
are denoted by L and J, respectively. In other words, when J ¼ 0. The server is in a
working vacation, and when J ¼ 1, the server is in a service period. Then, the pair
fJ; Lg constructs a continuous-time Markov process with transition-rate diagram
shown in Fig. 81.1. Here, the steady-state transition probabilities are defined by
Pjn ¼ PfJ ¼ j; L ¼ ng. j ¼ 0; 1; n ¼ 0; 1; 2; . . ..
Then, we can get the set of balance equations as follows:
770 X. Yu et al.
Fig. 81.1 State transition
rate diagram
8
>
>
>
>
> ðc þ kÞP00 ¼ ðn þ lv ÞP01 þ
<n ¼ 0
>
j¼0 lb P11 ð81:1Þ
>
>
>
> ðk þ c þ lv þ nnÞP0n ¼
>
>
:n1
kP0;n1 þ ½lv þ ðn þ 1ÞnP0;nþ1 ;
8
<n ¼ 0 kP10 ¼ cP00
j¼1 ðk þ lb ÞP1n ¼ kP1;n1 þ ð81:2Þ
:n1
cP0n þ lb P1;nþ1 :
The probability generating functions (PGFs) is defined by
X
1
Gj ðzÞ ¼ Pjn zn ; j ¼ 0; 1:
n¼0
Then, multiplying (81.1) by zn , summing over n, and rearranging terms, we get
2 0
kz ðk þ c þ lv Þz þ lv G0 ðzÞ þ nzð1 zÞG0 ðzÞ ¼ lv P00 ð1 zÞ lb P11 z ;
ð81:3Þ
0 dG0 ðzÞ
where G0 ðzÞ ¼ dz :
We also use the same way and obtain from (81.2)
2
kz ðk þ lb Þz þ lb G1 ðzÞ þ czG0 ðzÞ ¼ lb P10 ð1 zÞ þ lb P11 z: ð81:4Þ
Remark 2.
(1) Letting lv ¼ 0 in (81.3), we get
0 0
ðk þ cÞG0 ðzÞ þ nzG0 ðzÞ ¼ kzG0 ðzÞ þ nG0 ðzÞ þ lb P11 : ð81:5Þ
This agrees with Altman and Yechiali (2006) (see (5.3), p. 274).
(2) Letting n ¼ 0 in (81.4), we get the M=M=1 with a single working vacation.
This agrees with Zhao et al. (2008).
81 An M/M/1 Queue System with Single Working Vacation and Impatient Customers 771
81.2.3 Solution of the Differential Equation
For z 6¼ 0 and z 6¼ 1, (81.3) can be written as follows:
0 k c l l P00 l P11
G0 ðzÞ þ þ v G 0 ðzÞ ¼ v b : ð81:6Þ
n nð1 zÞ nz nz nð1 zÞ
Multiplying both sides of (81.6) by
k c lv
e nz ð 1 z Þ n z n ;
we get
d h knz c lv
i 1 l P
lb P11 knz c lv
v 00
e ð1 zÞ z G 0 ðzÞ ¼
n n e ð1 zÞn z n : ð81:7Þ
dz n z 1z
Integrating from 0 to z we have
lb P11 k1 ðzÞ þ lv P00 k2 ðzÞ
G0 ðzÞ ¼ k c lv ð81:8Þ
nenz ð1 zÞn z n
where
Z z
k c lv
k 1 ðzÞ ¼ enx ð1 xÞn1 x n dx;
0
Z z
k c lv
k 2 ðzÞ ¼ enx ð1 xÞn x n 1 dx:
0
Equation (81.8) expresses G0 ðzÞ in terms of P00 and P11 . Also, from (81.4),
G1 ðzÞ is a function of G0 ðzÞ, P00 , P11 . Thus, once P00 and P11 are calculated, G0 ðzÞ
and G1 ðzÞ are completely determined. We derive the probabilities P00 , P11 and the
mean system sizes in the next subsection.
81.2.4 Derivation of Various Performance Measures
P P1
Let Pj: ¼ Gj ð1Þ ¼ 1
0
n¼0 Pjn and E Lj . ¼ Gj ð1Þ ¼ n¼0 nPjn j ¼ 0; 1; . . .. Then
from (81.3) we get
lb P11 ¼ cG0 ð1Þ ¼ cP0: : ð81:9Þ
From (81.8), we have
k
en c
G 0 ð 1Þ ¼ ½lb P11 k1 ð1Þ þ lv P00 k2 ð1Þ limz!1 ð1 zÞn ; ð81:10Þ
n
772 X. Yu et al.
P
since G0 ð1Þ ¼ P0: ¼ 1 n¼0 P0n [ 0,
nc
and limz!1 ð1 zÞ ¼ 1, we must have
lb P11 k1 ð1Þ þ lv P00 k2 ð1Þ ¼ 0
implying that
lb P11 k1 ð1Þ cG0 ð1Þk1 ð1Þ
P00 ¼ ¼ : ð81:11Þ
lv k2 ð1Þ l v k 2 ð 1Þ
And (81.4) can be written as
czG0 ðzÞ þ lb P10 z lb P11 z lb P10
G1 ðzÞ ¼ : ð81:12Þ
ðkz lb Þð1 zÞ
By using L’Hopital rule, we derive
0
lb P10 þ cG0 ð1Þ
G1 ð1Þ ¼ ; ð81:13Þ
lb k
implying that
0 lb k P00
E½L0 ¼ G0 ð1Þ ¼ P1 l: ð81:14Þ
c k b
On the other hand, from (81.3),
0
0 ðlv kÞG0 ð1Þ þ cG0 ð1Þ lv P00
E½L0 ¼ lim G0 ðzÞ ¼ ; ð81:15Þ
z!1 n
implying that
0 ðk lv ÞG0 ð1Þ þ lv P00
G 0 ð 1Þ ¼ : ð81:16Þ
cþn
Equating the two expressions (81.14) and (81.16) for E½L0 , and using
1 ¼ P0 þ P1 , we get
knlb þ kclb nk2 ck2 lv k2 ð1Þ
G0 ð1Þ ¼ ; ð81:17Þ
g1 þ g2
ck1 ð1Þ
P00 ¼ G0 ð1Þ; ð81:18Þ
lv k 2 ð 1 Þ
where
g1 ¼ knlb þ kclb kclv nk2 lv k2 ð1Þ
g2 ¼ lb c2 þ cnlb þ kclv ck1 ð1Þ
81 An M/M/1 Queue System with Single Working Vacation and Impatient Customers 773
From (81.12) we have
g3
E½L1 ¼ ; ð81:19Þ
2 ð lb k Þ 2
where
00 0
g3 ¼ ðlb kÞcG0 ð1Þ þ 2clb G0 ð1Þ þ 2clb P00 :
From (81.3) we get
0
00 2kG0 ð1Þ þ 2ðk c n lv ÞG0 ð1Þ
G 0 ð 1Þ ¼ : ð81:20Þ
c þ 2n
Thus, the mean system size
E½L ¼ E½L0 þ E½L1 : ð81:21Þ
From (81.14), the necessary and sufficient condition for stability k\lb , for
otherwise E½L0 ¼ P00 \0.
81.2.5 Sojourn Times
Denote S the total sojourn time of a customer in the system, which is measured
from the moment of arrival to departure, including completion of service and a
result of abandonment. By using Little’s law,
E½L E½L0 þ E½L1
E ½ S ¼ ¼ ð81:22Þ
k k
Denote Sserved the total sojourn time of a customer who completes his service.
We note thatSserved is more important porfermance measure. Denote Sjn the con-
ditional sojourn time of a customer who does not leave the system, and ðj; nÞ is the
state on his arrival. Evidently, E½S1n ¼ ðn þ 1Þ=l, but this is for n ¼ 0; 1; 2; . . .
rather than for n 1 in (81.15).
Now, we derive E½S0n by using the method used by Altman and Yechiali
(2006), when J ¼ 0, for n 1,
c 1 k 1
E½S0n ¼ þ E½S1n þ þ E½S0n
hnþ1 hnþ1 hnþ1 hnþ1
nn þ lv 1
þ þ E½S0;n1 ; ð81:23Þ
hnþ1 hnþ1
where
hn ¼ c þ k þ lv þ nn
774 X. Yu et al.
for n ¼ 0; 1; 2;…. The second term above is derived from the fact that a new
arriving customer does not influence the sojourn time of a customer present in the
system. The probability n=ðn þ 1Þin the third term comes from the fact that when
one of the n þ 1 waiting customers abandons, our customer is not the one to leave.
Then,
hn cðn þ 1Þ
½c þ ðn þ 1Þn þ lv E½S0n ¼ þ þ ðnn þ lv ÞE½S0;n1 :
hnþ1 lb ð81:24Þ
We also have
k 1 1 k 1 l
E½S00 ¼ þ þ þ E½S00 þ 2v ; ð81:25Þ
h1 h1 lb h1 h1 h1
implying that
1 h0 c
E½S00 ¼ þ : ð81:26Þ
c þ n þ l v h 1 lb
Iterating (81.23) we obtain, for n 1,
1 hn ðn þ 1Þc
E½S0n ¼ þ
c þ ðn þ 1Þn þ lv hnþ1 l )
X n b Y n
1 hi ði þ 1Þc jn þ lv
þ þ :
i¼1
c þ ði þ 1Þn þ lv hiþ1 lb j¼i
c þ ðj þ 1Þn þ lv
ð81:27Þ
Finally, we use the expression for E½S1n and drive
X
1 X
1
E½Sserved ¼ P1n E½S1n þ P0n E½S0n : ð81:28Þ
n¼0 n¼0
implying that
E½L1 þ P1: X 1
E½Sserved ¼ þ P0n E½S0n : ð81:29Þ
lb n¼0
But the first sum in (81.28) starts from n ¼ 0, it’s different from [15].
81.3 Numerical Results
In this section, we show the numerical examples for the results obtained in Sect.
81.2.4. The effectes of lv and n on E½L0 and E½L in (81.14) and (81.21) are shown
in Fig. 81.2. Evidently, with lv increaseing, namely, the server works faster and
81 An M/M/1 Queue System with Single Working Vacation and Impatient Customers 775
Fig. 81.2 The relation of
E½L0 and E½L with n and lv
faster, the mean system size of working vacation E½L0 and the mean system size
E½L decreases when n is fixed. We also find that if n is smaller, E½L0 and E½L are
bigger. From the numerical analysis, the influence of parameters on the perfor-
mance measures in the system is well demonstrated. The results are suitable to
practical situations.
81.4 Conclusions
In this paper, we have studied M/M/1 queueing systems with a single working
vacation and impatient customers. In this system, the server has a slow rate to
serve during a working vacation and customers become impatient due to a slow
service rate. The server waits dormant to the first arrival in case that the server
776 X. Yu et al.
comes back to an empty system from a vacation, thereafter, opens a busy period.
Otherwise, the server starts a busy period directly if the queue system has cus-
tomers. The customers’ impatient time follows independently exponential distri-
bution. If the customer’s service has not been completed before the customer
becomes impatient, the customer abandons the queue and doesn’t return. The
probability generating functions of the number of customers in the system is
derived and the corresponding mean system sizes are obtained when the server was
both in a service period and in a working vacation. Also, we have obtained closed-
form expressions for other performance measures, such as the mean sojourn time
when a customer is served.
References
Altman E, Yechiali U (2006) Analysis of customers impatience in queues with server vacations.
Queueing Theory 52:261–279
Baba Y (2005) Analysis of a GI/M/1 queue with mul-tiple working vacations. Oper Res Lett
33:201–209
Baccelli F, Boyer P, Hebuterne G (1984) Single-server queues with impatient customers. Adv
Appl Probability 16:887–905
Boxma OJ, de Waal PR (1994) Multiserver queues with impatient customers. ITC 14:743–756
Dalay DJ (1965) General customer impatience in the quence GI/G/1. J Appl Probab 2:186–205
Li J, Tian N, Zhang Z (2009) Analysis of the M/G/1 queue with exponentially working vacations
a matrix analytic approach. Queueing Syst 61:139–166
Palm C (1953) Methods of judging the annoyance caused by congestion. Tele 4:189–208
Servi LD, Finn SG (2002) M/M/1 queues with working vacations (M/M/1/WV). Perform Eval
50:41–52
Takacs L (1974) A single-server queue with limited virtual waiting time. J Appl Probab
11:612–617
Van Houdt B, Lenin RB, Blonia C (2003) Delay distribution of (im)patient customers in a
discrete time D-MAP/PH/1 queue with age-dependent service times. Queueing Syst 45:59–73
Wu D, Takagi H (2006) M/G/1 queue with multiple working vacations. Perform Eval 63:654–681
Yue D, Yue W (2009) Analysis of M/M/c/N queueing system with balking, reneging and
synchronous vacations. Advanced in queueing theory and network applicantions. Springer,
New York, pp 165–180
Yue D, Yue W (2011) Analysis of a queueing system with impatient customers and working
vacations. QTNA’11 Proceedings of the 6th international conference on queueing theory and
network applications. ACM, NY, USA, pp 208–212
Zhao X, Tian N, Wang K (2008) The M/M/1 Queue with Single Working Vacation. Int J Inf
Manage Sci 19:621–634
Chapter 82
Analysis of Bank Queueing Based
on Operations Research
Liang Li, Jia-lin Wu and Jin-xiang Ding
Abstract This paper considers the bank queueing problem based on M/M/n
model. Through the analysis of the special situation of input distribution function
and loss of customer, this paper proposes segmented k input model and the losing
customer queueing model. With the actual data collected from the bank, the
calculation of the probability of losing customer provides a suggestion for the bank
to further improve.
Keywords Changeable customer arrival rate Changeable input rate Impatient
customer Queueing model Stable distribution
82.1 Introduction
With the development of economy, modern finance has become an essential part of
the society. Bank, the main body of financial industry, has also become one of the
most important service systems. Queueing is a common phenomenon in our daily
life, the queueing problems in the bank is also a hot issue that is very much
concerned. The queueing makes much contribution to the loss of customers which
followed the loss of money. So how to solve the queueing problem efficiently and
economically becomes more and more vital (Fan and Yuan 2005).
L. Li (&) J. Wu J. Ding
School of Mechanical Engineering, Beijing Institute of Technology, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 777
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_82,
Springer-Verlag Berlin Heidelberg 2013
778 L. Li et al.
82.2 Queueing Model of the Bank
The queueing theory is a mathematical study of waiting lines, or queues. The
classical theory enables mathematical analysis of several related processes,
including the arrival process, the queue’s discipline, and servers. This paper
adopted the data collected form a network of major commercial bank and con-
structed our system based on M/M/n (Jiang and Yang 2009) model.
82.2.1 The Arrival Process
82.2.1.1 The Probability Distribution of Customer Arrivals
Some standard notations for distributions are Poisson distribution, Erlang distri-
bution with j phases, degenerate (or deterministic) distribution, general distribu-
tion (arbitrary), and a phase-type distribution. One of the most common adopted
notations is Poisson distribution (Xiang 2012).
A discrete stochastic variable X is said to have a Poisson distribution with
parameter k [ 0, if for k = 0, 1, 2… the probability mass function of X is given by
ek kk
PðX ¼ kÞ ¼ ð82:1Þ
k!
where k is the mean arrival rate.
Poisson distribution required satisfying the following characteristic (Qin 2008):
(1) Stable. During any period of length t of the time interval, the probability of an
event happening is only relevant to the value of t other than the position of it.
(2) Memoryless (or Markov processes). In two mutually disjoint time interval of
T1 and T2, the numbers of events are independent.
(3) Generalized. The probability of more than one events (or more than one
customer arrived at the bank) happened at the same time can be ignored.
All the above-mentioned characterized in the queueing system for:
(1) The event of single customer reaches the bank is separate and random. Two or
more than two customer reaches the bank at the same time is considered to be
small probability event which could be ignored.
(2) The interarrival time distribution is random.
(3) The amount of customer is infinite.
Considering only the input process of the formation of steady state condition.
Since the opening time of the bank is from 9:00 am to 5:00 pm and some of the
customer arrived at the bank before the network started operating, we selected the
82 Analysis of Bank Queueing Based on Operations Research 779
data from 8:30 pm to 5:00 am. In the meantime, we picked the statistical unit of
the data as 15 min per person.
According to the collected data from Fig. 82.1, more than one peak appeared as
the customer arrival reaches a relatively high value. The first peak happened
between 9:00 and 10:00 am every day, for the banks opened at 9:00 am, and many
customers are in a rush to do business. The second peak is between 1:00 and
2:00 pm as many office workers go to the bank after the lunch during lunch break.
The last peak will be between 4:00 and 5:00 pm, as customers want to catch the
last chance before the bank closed for the day. During the time other than these
three peaks, the amount of customer will be relatively low.
82.2.1.2 The Chi Squared Test of Poisson Distribution
This paper segments the sampling time, and collects, statistics of customer num-
bers of arrival in different time period as X, and the frequency distribution of
customer numbers as n (Liu and Zhang 2008).
The null hypothesis of the test is as followed:
H0 : X PðkÞ ð82:2Þ
The unknown k uses the maximum likelihood estimation:
1 X
1
^
k¼
x¼ ni xi ð82:3Þ
n i¼1
The Pearson’s Chi squared test is used to test the null hypothesis. Karl Pearson
calculated the frequency as pi ¼ ni =n (Yao and Liu 2011), and then the value of
the test-statistic.
X
n
ðni npi Þ2
v2 ¼ ð82:4Þ
i¼1
npi
If the distribution function of X in null hypothesis H0 contains unknown value,
this paper calculated the maximum likelihood estimation value using the sample
data assuming that null hypothesis H0 is true, and calculated the estimation of pi as
Fig. 82.1 Customer arrival
frequency
780 L. Li et al.
^ ðX 2 Ai Þ using the maximum likelihood estimation value as parameter.
pi ¼ P
^
X
n
pi Þ 2
ðni n^
v2 ¼ ð82:5Þ
i¼1
n^
pi
R. A. Fisher’s exact test proved that if the null hypothesis failed to be rejected
and the value of n is large enough, the test-statistic generally follows a Chi squared
distribution as the degree of freedom being n-r-1 (Wang 2010). The critical
region of a significance level a is.
W = v2 [ v2a ðk r 1Þ ð82:6Þ
where k is the number of groups, r is the number of unknown values used to
calculate the sample data estimation of test-statistics.
Table 82.1 shows the statistics of customer arrival within the length of 15 min
as the data had been segmented into different groups.
After calculation and examination, the overall customer arrivals obey the k ¼
12:1 Poisson distribution. Considering the customer arrival frequency, the effect of
difference between days can be neglected based on the analysis of long-term data.
As for different hours within a day, the changes of parameter cannot be ignored.
It will be more practical and accordant with the status of actual customer arrival
of the bank to adopt specific parameter for different hours within a single day.
Since the peaks and the troughs are generally within an hour, the value of k is
considered to be a constant in the 1 h (Yang 2008). To eliminate the influence of
the peaks and troughs, we selected an hour as a unit to further analyzes the Poisson
flow of parameter variation.
82.2.2 The Servers
Following Kendall’s notation it indicates a system conclude two notations
involving with servers, which are the service time distribution and the number of
servers. In most banks, the number of server will not be limited as one. The service
Table 82.1 Distribution of number and frequency
Number Frequency Number Frequency Number Frequency
1 0 9 3 17 1
2 0 10 5 18 0
3 0 11 6 19 0
4 0 12 0 20 0
5 1 13 3 21 1
6 0 14 4 22 0
7 0 15 4 23 1
8 4 16 1
82 Analysis of Bank Queueing Based on Operations Research 781
time is relatively different and independent according to request of the customer.
The service time could be considered to be random variables which are mutually
independent and followed the same distribution. The analysis of the data shows
that the service time of the bank server is exponentially distributed with the
parameter l as follows:
F ðtÞ ¼ 1 elt ; t 0 ð82:7Þ
Let l be a constant and l [ 0 representing the mean service rate in a unit of
time, and 1=l being the mean service time.
82.2.3 The Queue’s Discipline
When the customer arrive at the system, the respond are of three different types.
The first type of customer will leave the system if all the servers are occupied. The
second type of customer will wait even if the traffic intensity levels exceed the
number of servers. The last type of customer will leave the system if the number of
waiting customers reached its capacity. In the second and third type, the queueing
discipline defines the way the customer will be served, the order in which they are
served could be First in first out, Last in first out, Processor sharing or Priority. The
queue can also be defined as single queue or multiple queue system.
The queueing discipline in the bank is quite typical second type as mentioned
above, and the capacity of queue can be considered to be infinite with loss of
customers. With the help of the queueing machine, the queue became single and
organized. The customer with priority will be separate from the ordinary customer
and be served without much delay in VIP room which cannot serve the ordinary
customer. The paper will be focusing on the ordinary customer queue based on
First come first serve discipline of single queue.
82.3 Parameters and Evaluations of Queueing System
M/M/c model indicates a system where arrivals are a Poisson process, service time
is exponentially distributed with c servers.
The main appropriate parameters are as follows (Deng 2010):
Ls: the number of customer, including the customer being served and in the queue;
Lq: the number of customer waiting;
Tq: the average waiting time of the customer in the system;
k: the average rate of arrival;
l: the average rate of service;
q: the service intensity, value being the ratio of average arrival rate k and average
service rate l, q ¼ k=l.
782 L. Li et al.
82.4 The Model of Losing Customer in the Bank
Queueing System
In the real-life situation, customers continuously arrive at the system, but not all of
them enter the system to wait and be served. Some customers arrive at the bank
and leave once they think there are too many waiting customers. Some other
customers leave the system for being in the queue for too long before they receive
service. It has become a serious problem that waiting time and the expected
waiting time is too long to cause the loss of customer of the bank. The problem of
the loss of customer in the bank’s queueing system is not being fully researched
which is exactly what the paper wants to study and analyze.
82.4.1 The Queueing Model of a Changeable Input Rate
The first thing a customer will do when he or she arrives at the bank is to decide
whether to join the system or not. The decision wasn’t made upon the calculation
based on previous waiting time before getting service but the expected waiting
time calculated based on the number of servers, the average service time and the
number of waiting customer. Generally, with the growth of the queue’s length, the
probability of customer joining the.system will diminish. Also, with the growth of
the number of servers, the probability will go up.
Let ak be the probability of the customer joining the system when the length of
queue is k and the number of server is n.
In previous papers, the situation of customer joining the single queue single
server system with the probability of ak ¼ 1=ðk þ 1Þ (Lu 2009) and ak ¼
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffi
k þ 1 k (Ren 2010) has been discussed.
This paper will introduce the parameter bðb [ 0Þ, and generalize the proba-
bility distribution into multiple servers, and the function of customer joining the
system as follows:
n
ak ¼ ðk 2 N; b as normalÞ ð82:8Þ
bk þ n
The introduction of the server number n and the parameter b improves the
relationship between the probability and the customer number k by making it more
realistic.
kn
kk ¼ kak ¼ ð82:9Þ
bk þ n
The status flow diagram (Thomas 2000) can be drawn as Fig. 82.2.
According to the status flow diagram, k’s equation (Tai and Gao 2009) can be
applied under equilibrium conditions as follows,
82 Analysis of Bank Queueing Based on Operations Research 783
Fig. 82.2 The status flow diagram of the queueing model of a changeable input rate
For status 0,
k
kp0 ¼ lp1 ) p1 ¼ p0 ¼ qp0 ; ð82:10Þ
l
For status 1,
kn kn q2 n
p1 ¼ 2lp2 ) p2 ¼ p1 ¼ p0 ; ð82:11Þ
bþn 2lðb þ nÞ 2ð b þ nÞ
For status 2,
kn kn q3 n2
p2 ¼ 3lp3 ) p3 ¼ p2 ¼ p0 ; ð82:12Þ
2b þ n 3lð2b þ nÞ 3!ð2b þ nÞðb þ nÞ
For status n-1,
kn qn nn1
pn1 ¼ nlpn ) pn ¼ p0 ;
ðn 1Þb þ n n!ðb þ nÞð2b þ nÞ ½ðn 1Þb þ n
ð82:13Þ
For status n,
kn qnþ1 nn
pn ¼ nlpnþ1 ) pnþ1 ¼ p0 ; ð82:14Þ
nb þ n nn!ðb þ nÞð2b þ nÞ ½nb þ n
For status n ? r-1,
qnþr nnþr1
pnþr ¼ p0 ; ð82:15Þ
nr n!ðb þ nÞð2b þ nÞ ½ðn þ r 1Þb þ n
For all status,
8 9
>
> q k nk >
>
< p0; k ¼ 1; 2; . . .; n >
>
=
k!nðb þ nÞð2b þ nÞ ½ðk 1Þb þ n
pk ¼ ð82:16Þ
>
> qk nk >
>
>
: knþ1 p0; k ¼ n þ 1; . . . ; >
n n!ðb þ nÞð2b þ nÞ ½ðk 1Þb þ n
From the Regularity condition.
X1
k¼0
pk ¼ 1 ð82:17Þ
784 L. Li et al.
We can get:
Xn qk nk
p0 ¼ 1 þ k¼1 k!nðb þ nÞð2b þ nÞ ½ðk 1Þb þ n
1 ð82:18Þ
X1 qk nk
þ knþ1
k¼nþ1 n n!ðb þ nÞð2b þ nÞ ½ðk 1Þb þ n
It is noteworthy that when b ¼ 1; n ¼ 1
1
p0 ¼ 1 þ q þ q2 2! þ q3 3! þ ¼ eq ð82:19Þ
qk q
pk ¼ e ; ðk ¼ 0; 1; 2; . . .Þ ð82:20Þ
k!
which means the stable distribution is Poisson Distribution with parameter q.
The value of b can be calculated through the analysis of a sample survey of the
customer arrival.
The evaluation can be calculated as follows,
(1) the average rate of customer arrival
X
1 X
n
kqk nkþ1
k¼ kk p k ¼ p0
k¼0 k¼0
k!nðb þ nÞð2b þ nÞ ½kb þ n
ð82:21Þ
X
1
kqk nkþ1
þ p0
k¼nþ1
nknþ1 n!ðb þ nÞð2b þ nÞ ½kb þ n
(2) the average service intensity of the system
k X
n
qkþ1 nkþ1
¼
q ¼ p0
l k¼0 k!nðb þ nÞð2b þ nÞ ½kb þ n
ð82:22Þ
X
1
qkþ1 nkþ1
þ p0
k¼nþ1
nknþ1 n!ðb þ nÞð2b þ nÞ ½kb þ n
(3) the average length of the waiting queue
X
1 X
1
kqkþn nkþn
Lq ¼ kpkþn ¼ p0 ð82:23Þ
k¼1 k¼1
nkþ1 n!ðb þ nÞð2b þ nÞ ½ðk þ n 1Þb þ n
(4) the average length of the system
(
k; k n
Ls ¼ ð82:24Þ
Lq þ n; k [ n
(5) when the customer k ? 1 arrives at the bank noticing k customers are in the
system, the probability of joining the system is, and the probability of loss is
82 Analysis of Bank Queueing Based on Operations Research 785
X
1 X
1 X
1
Ploss ¼ P ð Ls ¼ k Þ ð 1 a k Þ ¼ pk ak pk ¼ 1 k k ð82:25Þ
k¼0 k¼0 k¼0
82.4.2 The Queueing Model with Impatient Customer
Let there be a system with servers of n, infinite capacity, customer arrival being
Poisson Distribution, average arrival rate of k. As a customer arrives at the bank
with all servers occupied, he or she will wait in the queue. The customer will be
impatient and even leave the system when the length of the queue is too long or the
average service time is too long while the queue is not too long (He et al 2009).
A research shows that when the waiting time exceeds 10 min, the customer
started to feel impatient. When it exceeds 20 min, the customer become irritable,
and when it exceeds 30 min, they might leave because of angry (Sun 2010). The
research indicates that a precaution to reduce the average waiting time is of
importance.
The intensity of customer leaving being at is relevant to the waiting time k.
Take the simplest model when at ¼ dtðd [ 0Þ into consideration. Collect the data
as the customer leave the bank and analyze the pattern in statistics angle. As the
establishment of model based on time measurement is too vague, the number of
customer may be k ¼ nt=l when the leaving customer enter the system consid-
ering the average service time is 1=l. The customer waits for time of t and leaves
with at ¼ dtðd [ 0Þ can be transformed into leaving the system with ak ¼ kd=n ¼
hkðd [ 0Þ before entering the system. The status flow diagram can be drawn as
Fig. 82.3
Analyze the flow diagram based on k’s equation and we can get:
8
>
>
> ðnqÞk
< p0 ; 0 k n
k!
pk ¼ ð82:26Þ
>
> nn q k
>
: p0 ; k [ n
n!ð1 þ bÞð1 þ 2bÞ ½1 þ ðk nÞb
where q ¼ q1 =n; b ¼ h=ln:
( )1
Xn
ðnqÞk X1
nn q k
p0 ¼ þ ð82:27Þ
k¼0
k! k¼nþ1
n!ð1 þ bÞð1 þ 2bÞ ½1 þ ðk nÞb
Fig. 82.3 The status flow diagram of the queueing model with impatient customer
786 L. Li et al.
And the probability of losing customer is
X
1 Xn
ðnqÞk
Ploss ¼ 1 p k ak ¼ 1 hp0
k¼0 k¼0
ðk 1Þ!
ð82:28Þ
X
1
nn qk hk
p0
k¼nþ1
n!ð1 þ bÞð1 þ 2bÞ ½1 þ ðk nÞb
82.5 Conclusion
The core concept and basic function of the bank is to satisfy the need of customer.
Thus the bank should be devoting to improve efficiency or add the number of
servers to avoid the situation of customer leaving the system for the expected
waiting time being too long. This paper takes the variable parameters into con-
sideration and builds a more real-life concerned model. The customer-losing rate
can be used as an evaluation to fulfill the basic function of the bank. The bank can
compare the expense of opening a new server with the loss of losing customer to
get a customer loss rate in the balance as P. When the actual losing rate is larger
than this value, a bank should open a new server to reduce the loss.
References
Deng C (2010) M/M/n queuing model of changeable input rates. J Mianyang Norm Univ
29(8):1–3
Fan W, Yuan H (2005) The study about bank customer services system based on queue theory.
Value Eng 24(12):126–128
He J, Li D, Li H (2009) Optimization of the service windows deployed. Mod Electron Tech
32(6):134–136
Jiang C, Yang L (2009) Queue model in personal banking and sensitivity analysis. J Guangdong
Coll Finance Econ 8(3):69–73
Liu R, Zhang Z (2008) Hypothesis test in the judgment on a binomial distribution and a Poisson
distribution. J Qinghai Univ (Nat Sci) 26(1):44–47
Lu C (2009) Queue theory. Beijing University of Posts and Telecommunications Press, Beijing,
pp 31–99
Qin Z (2008) Mathematical model and solution of the problem based on the queue at the bank.
Pop Sci Tech 10(3):24–26
Ren X (2010) Study of an M/M/1 queuing system with a variable input rate. J Neijiang Norm
Univ 25(10):35–36
Sun Z (2010) The application of queueing theory in the bank window and teller staffing. West
China Finan 31(6):20–21
Tai W, Gao S (2009) A M/M/1 queuing model with variable input rate. J Chongqing Norm Univ
(Nat Sci) 26(1):69–77
Thomas G R (2000) Computer networks and systems: queueing theory and performance
evaluation, 3rd edn. Springer, New York, pp 22–30
82 Analysis of Bank Queueing Based on Operations Research 787
Wang L (2010) Probability theory and mathematical statistics. Dalian University of Technology
Press, Dalian, pp 245–255
Xiang H (2012) Discussion on poisson distribution and its application. Learn Wkly 6(10):205
Yao X, Liu R (2011) Probability theory and mathematical statistics. Peking University Press,
Beijing, pp 135–142
Yang M (2008) Bank queuing system data analysis and counter setting optimization research.
J Wuhan Univ Technol (Inform Manag Eng) 30(4):624–627
Chapter 83
Application of DEA-Malmquist Index
in Analyzing Chinese Banking’s Efficiency
Man Ding, Chao-qun Ma, Zhong-bao Zhou and Wen-bin Liu
Abstract The efficiency of commercial banks is important for Chinese banking to
prevent risk and improve the competitiveness, and also the core for China to
deepen financial reform. By using the DEA-Malmquist indices approach, the
efficiency and the efficiency changes of 14 China’s commercial banks during the
period of 2007–2010 were analyzed in this paper. Results showed that the average
efficiency of State-owned commercial banks is generally lower than that of joint-
stock commercial banks and the urban commercial banks, the average efficiency of
urban commercial banks is the highest; the overall efficiency of 14 banks is
improved in 4 year due to efficiency progress and technical advance.
Keywords China’s listed commercial banks DEA Efficiency Malmquist
index
83.1 Introduction
Commercial banks are the main body of China’s financial industry; they play an
important role for China’s economy development and the improvement of people’s
living standard. However, after the 1990s, the control of the commercial banks is
becoming more relaxed with the globalization of economy and financial freedom
in China, especially after China’s entry into the WTO which allows foreign banks
run the business in China, makes the competition between commercial banks
M. Ding (&) C. Ma Z. Zhou
School of Business Administration, Hunan University,
Changsha 410082, China
e-mail: [email protected]
W. Liu
Kent Business School, University of Kent, Canterbury,
CT2 7PE, UK
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 789
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_83,
Ó Springer-Verlag Berlin Heidelberg 2013
790 M. Ding et al.
intensified. Since the U.S subprime mortgage crisis, the financial tsunami makes
we know that the commercial bank efficiency is the key for them to have a place in
the competition. How to improve the efficiency of Chinese commercial banks is
the vital problem to be solved for the bank authorities and the bank decision
makers. Therefore, evaluating the efficiency situation of China’s commercial
banks clearly and correctly, exploring the measures and preferences to improve the
efficiency of China’s commercial banks is priority.
At present, the widespread use of the bank efficiency evaluation method is Data
envelopment analysis (DEA) in academia. DEA which is raised by Charnes et al.
(1978) is an approach for measuring the relative efficiency of peer decision making
units (DMUs) that have multiple inputs and outputs. The essence of DEA is using
the ‘‘frontier analysis’’, according to a certain standard to construct a production
frontier and the gap between the evaluated bank and the frontier is its efficiency.
The advantages of the DEA method are that it need not give the weight of each
index by people, also need not given the production function form of frontier in
advance, and it can deal the project with multiple outputs and multiple inputs
(Chen and Zhu 2004). So this method used in the same industry to analyze the
efficiency has its own unique advantages. Penny (2004) investigates X-efficiency
and productivity change in Australian banking between 1995 and 1999 using DEA
and Malmquist productivity indexes, and finds that regional banks are less efficient
than other bank types. Total factor productivity in the banking sector was found to
have increased between 1995 and 1999 due to technological advance shifting out
the frontier. Zhu et al. (2004) measured the efficiency of China’s largest 14
commercial banks over the period 2000–2001 using the super-efficient DEA model
and Tobit regression method. The results show that the overall efficiency of the
four state-owned commercial banks is far less than 10 joint-stock commercial
banks, and the excessive number of employees is a major bottleneck restricting the
state-owned commercial banks efficiency. Chen et al. (2005) examine the cost,
technical and allocative efficiency of 43 Chinese banks over the period 1993–2000.
Results show that technical efficiency consistently dominates the allocative effi-
ciency of Chinese banks, and the large state-owned banks and smaller banks are
more efficient than medium sized Chinese banks.
But from most related literature of evaluating the efficiency of commercial
banks, we may find that they mainly adopt DEA model to describe the commercial
banking efficiency status, but the description is basically a static comparison, or
even though the dynamic description, the description is incomplete (Mette and
Joseph 2004; Ariff and Can 2008; Laurenceson and Yong 2008). Therefore, to
evaluate bank efficiency by DEA based on Malmquist index began to be widely
used. Zhang and Wu (2005) analyzed the efficiency change of China’s commercial
banks during the period of 1999–2003 using all Input-Oriented Malmquist Index
approach. Gao (Gao et al. 2009) studies the panel data of primary commercial
banks over the period of 1997–2006, and it calculates the total factor productivity
and its decomposition indexes based on the DEA-based Malmquist productivity
index. Maria et al. (2010) develop an index and an indicator of productivity change
that can be used with negative data and use RDM efficiency measures to arrive at a
83 Application of DEA-Malmquist 791
Malmquist-type index, which can reflect productivity change, and use RDM
inefficiency measures to arrive at a Luenberger productivity indicator.
This paper has introduced the theory of related DEA model and Malmquist
index and analyzed the inputs and outputs selection of 14 China’s listed com-
mercial banks, and measured the efficiency and the dynamic changes of the effi-
ciency of 14 China’s listed commercial banks during the period of 2007–2010.
83.2 Methodology
83.2.1 DEA Model
Consider we have n DMUs, and that each DMUj ðj ¼ 1; 2; . . .; nÞ has m inputs and
s outputs. Suppose Xj , Yj are the input and output of DMUj , and Xj ¼ ðx1j ; x2j ;
. . .; xmj Þ, Yj ¼ ðy1j ; y2j ; . . .; ysj Þ, then we can define the DEA model as follows:
min h
Xn
s:t: kj Xj hXj0
j¼0
ð83:1Þ
X
n
kj Yj Yj0
j¼0
kj 0; j ¼ 0; 1; 2; . . .; n
where kj are the weights of input/output indexes, h is the efficiency score. And if
h\1, the DMU is inefficient; If h ¼ 1, the DMU is efficient.
83.2.2 Malmquist Index Model
Malmquist index model was brought out by Malmquist in 1953 in the process of
analyzing consumption. Nishinizu and Page firstly used this index to measure the
change of productivity, since then the Malmquist index model was combined with
DEA theory and has a wide use in measuring the efficiency of production
(Nishimizu and Page 1982). The Malmquist index is defined as:
12
t;tþ1 Dtþ1 ðxtþ1 ; ytþ1 Þ Dt ðxtþ1 ; ytþ1 Þ
M ¼ tfp ¼ ð83:2Þ
Dtþ1 ðxt ; yt Þ Dt ðxt ; yt Þ
From model (2) we cam see that Malmquist index is an efficiency index, which
represents the efficiency changes from t to t þ 1. If Malmquist [ 1, then the
792 M. Ding et al.
efficiency of DMU is improved; If Malmquist ¼ 1, then the efficiency of DMU is
unchanged; If Malmquist\1, then the efficiency of DMU is declined.
Further, Malmquist Index can be decomposed into two components, which is
efficiency progress index (effch) and technical change index (tech):
Dtþ1 ðxtþ1 ; ytþ1 Þ
effch ¼
Dt ðxt ; yt Þ
t tþ1 tþ1 1 ð83:3Þ
D ðx ; y Þ Dt ðxt ; yt Þ 2
tech ¼
Dtþ1 ðxtþ1 ; ytþ1 Þ Dtþ1 ðxt ; yt Þ
Effch is defined as the efficiency improvement part, represented the manage-
ment level change in two periods—‘‘catching-up effect’’ which measures the
commercial bank management performance is more close to the current production
frontier. Tech is defined as part of the technology progress, representing the shift
of the production frontier in two periods—‘‘frontier-shift effect’’. What’s more, the
effch can also be decomposed into pure technical efficiency (pech) and scale
efficiency (sech):
effch ¼ pech sec h ð83:4Þ
83.3 Empirical Analysis
83.3.1 Data Sources
According to the principle of representative, accessibility and integrity, this paper
selects 14 listed commercial banks as research objects during 2007–2010. They are
Bank of China (BC), China Construction Bank (CCB), Industrial and Commercial
Bank (ICBC), Bank of Construction (BCM), China Merchants Bank (CMB),
Industrial Bank (CIB), Citic Bank (CITIC), Shanghai Pudong Development Bank
(SPDB), Minsheng Bank (CMBC), Hiaxia Bank (HXB), Shenzhen Development
Bank (SDB), Beijing Bank (BOB), Nanjing Bank (NOB) and Ningbo Bank (NBB).
The input–output data of the banks is from each listed commercial bank’s
‘‘Published Financial Statements in 2007–2010’’ and ‘‘Chinese Financial Statistics
Yearbook in 2007–2010’’.
83.3.2 Indexes Selection
Reasonable definition of inputs and outputs of banks is the key problem using
DEA model to measure efficiency of the commercial banks. Recently, the gen-
erally accepted method in division of inputs and outputs of the bank in
83 Application of DEA-Malmquist 793
international financial academics are three principal schools: production approach,
the intermediation approach and asset approach (Zelenyuk 2006; Feng and Serletis
2010; Giokas 2008).
The production method regards banks as producers which use labor force and
their own capital to generate deposits and loans. The number of loan and deposit
account is usually seen as outputs, while the number of employees and the capital
are seen as outputs in this method.
The intermediary approach consider banks as financial intermediaries where
deposits are converted into loans, and getting the profits from the income of loans and
investment. Therefore, intermediary method takes fixed capital, labor, and various
interest cost as the inputs, and takes all kinds of loans and investments as the outputs.
The Asset approach also regards banks as financial intermediaries. Usually the
liabilities on the balance sheet are regarded as inputs, and the loans and invest-
ments are regarded as outputs.
According to the China’s listed commercial banks’ characteristics and the
requirements of DEA model, the inputs of the China’s listed commercial banks in
this paper are the number of employees, fixed assets, operating expenses and
deposits; the outputs are loans and revenue which can seen in Fig. 83.1.
83.3.3 The Empirical Results
Based on DEA and Malmquist indices, we calculate the efficiency of China’s listed
commercial banks from 2007 to 2010 which can be seen from Table 83.1 and
Efficiency change of China’s listed commercial banks’ Malmquist index during
2007–2010 which can be seen from Table 83.2 by matlab.
According to the empirical results in Table 83.1, we conduct the following
analysis:
First, from the aspect of time window, we can find that the average efficiency
scores of 14 China’s listed commercial banks are all less than 1, this shows that
each listed commercial bank is DEA inefficient. And the overall efficiency of
China’s banks shows a declined trend, from 0.982 in 2007 down to 0.979 in 2010.
Fig. 83.1 The inputs and number of
outputs of the China’s listed employees
commercial banks loans
operating
expenses
fixed assets Bank
revenue
deposits
794 M. Ding et al.
Table 83.1 The efficiency score of China’s commercial banks during 2007–2010
Banks 2007 2008 2009 2010 Average
BC 0.970 0.938 0.949 0.980 0.959
CCB 0.971 0.901 0.981 1.000 0.963
ICBC 0.930 0.858 0.943 0.900 0.908
BCM 1.000 0.988 1.000 1.000 0.997
CMB 1.000 0.980 0.936 1.000 0.979
CIB 1.000 1.000 1.000 1.000 1.000
CITIC 1.000 0.990 1.000 1.000 0.998
SPDB 0.993 1.000 1.000 1.000 0.998
CMBC 1.000 1.000 1.000 1.000 1.000
HXB 0.935 0.936 0.923 0.962 0.939
SDB 1.000 1.000 1.000 0.992 0.998
BOB 1.000 1.000 1.000 1.000 1.000
NOB 1.000 1.000 1.000 1.000 1.000
NBB 0.947 0.909 1.000 1.000 0.964
Average score of state-owned banks 0.968 0.921 0.968 0.970 0.957
Average score of joint-stock commercial banks 0.990 0.987 0.980 0.994 0.987
Average score of urban commercial banks 0.982 0.970 1.000 1.000 0.988
Average 0.982 0.964 0.981 0.988 0.979
Table 83.2 The Malmquist Index score of China’s commercial banks during 2007–2010
Banks effch tech pech sech tfp
BC 1.003 1.026 1.000 1.003 1.029
CCB 1.010 1.063 1.000 1.010 1.073
ICBC 0.989 1.039 1.000 0.989 1.028
BCM 1.000 1.027 1.000 1.000 1.027
CMB 1.000 0.995 1.000 1.000 0.995
CIB 1.000 1.034 1.000 1.000 1.034
CITIC 1.000 0.982 1.000 1.000 0.982
SPDB 1.002 0.999 1.000 1.002 1.001
CMBC 1.000 0.832 1.000 1.000 0.832
HXB 1.010 0.991 1.015 0.995 1.001
SDB 0.997 1.042 1.000 0.997 1.048
BOB 1.000 1.015 1.000 1.000 1.015
NOB 1.000 1.023 1.000 1.000 1.023
NBB 1.018 0.989 1.000 1.018 1.007
Average score of state-owned commercial banks 1.001 1.039 1.000 1.001 1.039
Average score of joint-stock commercial banks 1.001 0.982 1.002 0.999 0.985
Average score of urban commercial banks 1.006 1.009 1.000 1.006 1.015
2007–2008 0.981 1.039 1.001 0.981 1.019
2008–2009 1.018 0.936 0.999 1.019 0.953
2009–2010 1.008 1.037 1.004 1.004 1.045
2007–2010 1.002 1.003 1.001 1.001 1.005
83 Application of DEA-Malmquist 795
At the same time, the efficiency score in 2008 is the lowest which indicates that the
financial crisis had more adverse impact on Chinese banking industry and the risk
defense ability of Chinese banking industry is insufficient.
Second, from the bank ownership form, we can find that average efficiency
score of urban commercial banks is the highest in the 4 years, reach up to 0.988;
the average efficiency score of joint-stock commercial banks is the second; The
lowest efficiency score is state-owned commercial banks, only 0.957. In addition,
we may find that the average efficiency scores of the urban commercial banks all
achieve the DEA efficient respectively on 2008 and 2009 which shows the
Operation efficiency of urban commercial banks are Overall good.
Third, from the average efficiency score of each commercial bank in 4 years,
we can find that all state-owned commercial banks are DEA inefficient. For the
joint-stock commercial banks, CIB and CMBC are all DEA efficient in 4 years,
BOB and NOB are also DEA efficient. Moreover, ranking in the last three places
respectively is BC, HXB and ICBC.
We evaluate the efficiency of 14 China’s listed commercial banks above using
DEA method, but this method calculates the efficiency of the commercial banks
from static goniometer, namely horizontal comparison of the efficiency in the same
period of different commercial banks which is not suitable for the longitudinal
description of the dynamic changes of the efficiency in a period. So we measure
the dynamic changes of the efficiency of the commercial banks in China using the
Malmquist index to make the evaluation of the efficiency of Chinese commercial
banks more detailed and more comprehensive.
According to the empirical results in Table 83.2, we conduct the following
results:
First, the average Malmquist index of the 14 China’s listed commercial banks
from 2007 to 2010 is 1.005, greater than 1, which means the overall efficiency of
China’s banks is rising. The overall efficiency during the period of 2007–2008 and
2009–2010 are all rising but the efficiency in 2008–2009 is only 0.953, less than 1
the reason is the negative influence of financial crisis. In addition, the average
Malmquist index of joint-stock commercial banks is less than 1, is declined in four
yeas. The largest increased efficiency is state-owned commercial banks. The
efficiency of CMB, CITIC and CMBC is declined in 4 years, and the efficiency of
SPDB and HXB is nearly unchanged, and the others are rising.
Second, the overall improved efficiency of 14 China’s listed commercial banks
is due to the increase of the efficiency progress (effch) and the technical change
(tech). The overall declined efficiency of the joint-stock commercial banks is due
to the decrease of the technical change (tech). And the increase of the state-owned
commercial banks is mainly due to the decrease of the technical change (tech).
Moreover, for the increase of efficiency progress (effch) of the urban commercial
banks is mainly due to the increase of scale efficiency (sech).
796 M. Ding et al.
83.4 Conclusion
The paper has introduced the theory of related DEA model and Malmquist index
and analyzed the inputs and outputs selection of 14 China’s listed commercial
banks, measured the efficiency and the dynamic changes of the efficiency of 14
China’s listed commercial banks during the period of 2007–2010. Through the
analysis, the results show that the average efficiency scores of 14 China’s listed
commercial banks are all DEA inefficient. The average efficiency score of urban
commercial banks is the highest in the 4 years, the average efficiency score of
joint-stock commercial banks is the second and the lowest efficiency score is state-
owned commercial banks. The average Malmquist index of the 14 China’s listed
commercial banks is greater than 1, and the overall improved efficiency of 14
China’s listed commercial banks is due to the increase of effch and tech.
Acknowledgments This paper is supported by National Science Fund for Distinguished Young
Scholar (70825006); The Funds for Innovation Research Changjiang Scholar (IRT0916) in China
and Hunan Provincial Natural Science Foundation of China(09JJ7002).
References
Ariff M, Can L (2008) Cost and profit efficiency of Chinese banks: a non-parametric analysis.
China Econ Rev 21:260–273
Charnes A, Cooper WW, Rhodes E (1978) Measuring the efficiency of decision making units. Eur
J Oper Res 2(6):429–444
Chen XG, Michael S, Brown K (2005) Banking efficiency in China: application of DEA to pre-
and post-deregulation eras: 1993–2000. China Econ Rev 16:229–245
Chen Y, Zhu J (2004) Measuring information technology’s indirect impact on firm performance.
Inf Technol Manage J 5:9–22
Feng G, Serletis A (2010) Efficiency, technical change, and returns to scale in large US banks:
panel data evidence from an output distance function satisfying theoretical regularity. J Bank
Finance 34(1):127–138
Gao M, Yang SY, Xie BC (2009) Research on changing tendency of commercial banks
productivity efficiency in China. J Xidian Univ Soc Sci Ed 19(5):51–55
Giokas DI (2008) Assessing the efficiency in operations of a large Greek bank branch network
adopting different economic behaviors. Econ Model 25:559–574
Laurenceson J, Yong Z (2008) Efficiency amongst China’s banks: a DEA analysis five years after
WTO entry. China Econ Rev 1(3):275–285
Maria CA, Portela S, Thanassoulis E (2010) Malmquist-type indices in the presence of negative
data: an application to bank branches. J Bank Finance 34:1472–1483
Mette A, Joseph CP (2004) Combining DEA window analysis with the Malmquist Index
Approach in a study of the Canadian banking industry. J Prod Anal 21(1):67–89
Nishimizu M, Page JM (1982) Total factor productivity growth, technical efficiency change:
dimensions of productivity change in Yugoslavia in 1965–1978. Econ J 92:929–936
Penny N (2004) X-efficiency and productivity change in Australian banking. Aust Econ Pap
43(2):174–191
Zelenyuk V (2006) Aggregation of Malmquist productivity indexes. Eur J Oper Res
174:1076–1086
83 Application of DEA-Malmquist 797
Zhu N, Zhuo X, Deng Y (2004) The empirical analysis of the efficiency and reform strategy of the
state-owned commercial Banks in China. Manage World 2:18–26
Zhang J, Wu H (2005) The empirical analysis of efficiency of Commercial Bank of China based
on Malmquist Index Approach. J Hebei Univ Technol 34(5):37–41
Chapter 84
The Improvement on R. G. Bland’s
Method
Yu-bo Liao
Abstract Cycling may occur when we use the simplex method to solve linear
programming problem and meet degeneration. Such cycling problem can be
avoided by the Bland method. In this paper, we will present an improved Bland
method with more iterative efficiency than the Bland method.
Keywords Bland method Linear programming Linear optimization Simplex
method
84.1 Introduction
In plain English one can say that a linear optimization (LO) problem consists of
optimizing, i.e., minimizing or maximizing, a linear function over a certain
domain. The domain is given by a set of linear constraints. The constraints can be
either equalities or inequalities.
The simplex method for linear programming problems was first proposed by
Dantzig in 1947 (Dantzig 1948), which can be described as follow:
Supposing that the given standard linear programming problem is
mins ¼ cx
Ax ¼ b
x0
0 1 0 1 0 1
a11 a1n x1 b1
B . C B . C
where A ¼ @ A, x ¼ @ .. A, b ¼ @ .. A,
am1 amn xn bm
Y. Liao (&)
School of Basic Science, East China Jiaotong University, Nanchang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 799
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_84,
Springer-Verlag Berlin Heidelberg 2013
800 Y. Liao
c ¼ ð k1 kn Þ
The rank of A ¼ ðaij Þmn is m, n m 1. The steps of the simplex method can
be summarized as follow:
• The first step: B ¼ ðpj1 ; pj2 ; . . .; pjm Þ is the known feasible basis, and the
ð0Þ
canonical form and the basic feasible solution xB ¼ B1 b ¼ ð b10 bm0 ÞT o
• The second step: Check the testing number. If all testing numbers satisfy
kj 0; ðj ¼ 1; 2; nÞ, the corresponding basic feasible solution xð0Þ is the
optimal solution. All the process is ended, otherwise go to next step;
• The third step: If some testing number kr [ 0 and B1 pr ¼ ðb1r ; b2r ;
; bmr ÞT 0, there is no optimal solution for this problem. All the process is
ended, otherwise go to next step;
• The forth step: If some testing number kr [ 0 and there is a positive number in
ðb1r ; b2r ; ; bmr ÞT , make xr be the entering-basis variable (if there are a few of
positive testing numbers, choose the largest one in order to improve the iterative
efficiency. This method n is namedoas the largest testing number method), and the
minimum ratio is min bbi0ir bir [ 0 ¼ bbs0sr Hence the leaving-basis variable xjs can
be determined (if there are a few same minimum ratios, choose the minimum-
subscript variable as the leaving- basis variable). Substitute pr for pjs , obtain the
and then go to next step;
new basis B,
ð1Þ 1 b,
• The fifth step: Obtain the canonical form and the basic feasible xB ¼ B
corresponding to new basis B (which can be realized directly by elementary row
transformation of the corresponding simplex tableau in manual calculation).
Afterwards, substitute B for B, substitute xð1Þ ð0Þ
for x , and then return to the
B
second step.
For the non-degenerate linear programming problems, using the largest testing
number simplex method in iteration, after finite iterative steps, the optimal solution
must be obtained or not existed. But for degenerate linear programming problems,
this method may not be valid because basis cycling may appear. In 1951,
A. J. Hoffman first designed one example where appears cycling in iterations. In
1955, E. M. L. Beale designed a simpler example to show the possible cycling
problem (Beale 1955; Tang and Qin 2004; Zhang and Xu 1990).
To avoid infinite cycling, R. G. Bland proposed a new method in 1976 (Bland
1977). In the Bland method the cycling can be avoided in calculation if abiding by
two rules which are shown as following (Andersen et al. 1996; Nelder and Mead
1965; Lagarias et al. 1998; Bixby 1994; Herrera et al. 1993; Wright 1996; Han
et al. 1994; Hapke and lowinski 1996; Zhang 1999; Terlaky 1985; Terlaky 2000;
Terlaky and Zhang 1993; Wagner 1958; Ward and Wendell 1990; Wolfe 1963;
Wright 1998; Elsner et al. 1991; Han 2000):
• Rule 1: Once there are a few positive testing numbers, choose the corresponding
minimum-subscript basic variable as the entering-basis variable;
84 The Improvement on R. G. Bland’s Method 801
• Rule 2: Once a few ratios bbi0ir , in different rows reach the minimum at the same
time, choose the corresponding minimum-subscript basic variable as the leav-
ing-basis variable.
Rule 2 determines the leaving-basis variable, and it is same as the forth step of
the simplex method. However, the entering-basis variable is determined by Rule 1,
but the largest testing number method. The advantage of the Bland method is
simple. However because it only considers the minimum subscript, but the
decreasing speed of the target function, its iteration times are often much more
than those of the largest testing number method. In this paper, we will first prove a
theorem, and then use this theorem to propose an improved Bland method with
much more computation efficiency.
84.2 The Improvement of Bland’s Method
Theorem 1 If the linear programming problem has an optimal solution, there
appears degenerate basic feasible solution in some iterative step with the simplex
method, but it is not optimal, and only one basic variable is zero in the degenerate
basic feasible solution, the degenerate basic feasible solution will not appear again
after this iterative step (even if the entering-basis variable is determined by largest
testing number method).
Proof First suppose that the corresponding basis is B ¼ ðpj1 ; pj2 ; . . .; pjm Þ, the
corresponding basic feasible solution is xð0Þ , the corresponding simplex tableau is
c B1 b cB B1 A c
TðBÞ ¼ B 1 o in this iterative step. The corresponding
B b B1 A
canonical form is
X
min s ¼ sð0Þ kj x j
j6¼j1 ;j2 jm
8 P
< xji þ bij xj ¼ bi0 ði ¼ 1; 2; mÞ
j6¼j1 ;j2 jm
:x 0 ðj ¼ 1; 2; mÞ
j
There is only one zero in bi0 ði ¼ 1; 2; mÞ, and now assume that bs0 ¼ 0 and
bi0 [ 0. After this iterative step, according to the hypothesis, because only one
basic variable is zero, only if the row in which leaving-basis variable locates is not
s row, the value of target function will decrease and xð0Þ will be transferred;
Moreover, because the target value will not increase in iteration, xð0Þ will not
appear again. Therefore, if the conclusion is not valid, there is only one case: In the
iteration afterwards, The row in which the leaving-basis variable locates is s row,
802 Y. Liao
and hence the entering-basis variable will be the leaving-basis
variable
in each
iteration. This kind of variable is only in the set xj j ¼ j1 ; j2 jm [ xjs .
Because the number of the set is finite, if there appears cycling, there must be
some variable xq which leaves the basis and then enters again. Supposing that the
corresponding simplex tableau is TðBt Þ when xq is the leaving-basis variable and
ðtÞ ðtÞ
the entering-basis variable is xr in this tableau, bsq ¼ 1, kðtÞ ðtÞ
q ¼ 0, bsr [ 0, kr [ 0
Supposing that the corresponding simplex tableau is TðBtþk Þ. When xq is the
entering-basis variable, kðtþkÞ
q [ 0 (because it’s still not optimal). TðBt Þ becomes
TðBtþk Þ after iteration, and then
ðtÞ
bsq
bðtþ1Þ
sq ¼ ðtÞ
[ 0; kðtþ1Þ
q ¼ kðtÞ ðtÞ ðtþ1Þ
q kr bsq \kðtÞ
q ¼0
bsr
ðtþkÞ
The rest may be deduced by analogy, bsq [ 0, kðtþkÞ
q \0, which contradicts
kðtþkÞ
q [ 0, So the conclusion is valid. The proof is ended.
When there appears degenerate case, from Theorem 1 we can obtain: If only
one basic variable is zero in the degenerate basic feasible solution, we can still use
the largest testing number method and there will not appear cycling. Therefore, we
can modify Rule 1 of the Bland method in order to improve efficiency of iteration:
Improved rule 1: when there are a few of positive testing numbers, if only one
basic variable is zero in the corresponding basic feasible solution at most, the
entering-basis variable can be determined by the largest testing number; If more
than one basic variable is zero in the corresponding basic feasible solution, the
entering-basis variable can be determined by Rule 1 of the Bland method.
84.3 Conclusion
In summary, the large testing number method has high iteration efficiency, but it
has the cycling problem; the Bland method can avoid the cycling problem, but
results in low iteration efficiency. In order to eliminate those two disadvantages,
we proposed an improved method which can prevent the cycling theoretically with
higher computation efficiency.
Acknowledgments I would like to thank the support provided by East China Jiaotong Uni-
versity Research fund and Jiangxi province Research fund.
84 The Improvement on R. G. Bland’s Method 803
References
Andersen ED, Gondzio J, Meszaros Cs, Xu X (1996) Implementation of interior point methods
for large scale linear programming. In: Terlaky T (ed) Interior point methods of mathematical
programming. Kluwer Academic Publishers, Dordrecht, pp 189–252
Beale EM (1955) Cycling in the dual simplex algorithm. Nav Res Logist Quart 2:269–276e
Bixby RE (1994) Progress in linear programming. ORSA J Comput 6(1):15–22
Bland RG (1977) New finite pivoting rules of simplex method. Math Oper Res 2:103–107
Dantzig GB (1948) Programming in a linear structure. Comptroller USAF, Washington, DC
Elsner L, Neumann M, Vemmer B (1991) The effect of the number of processors on the
convergence of the parallel block Jacobi method. Linear Algebra Appl 154–156:311–330
Han L (2000) Algorithms for unconstrained optimization. Ph.D. Thesis, University of Connecticut
Han S, Ishii H, Fuji S (1994) One machine scheduling problem with fuzzy duedates. Eur J Oper
Res 79:1–12
Hapke M, lowinski RS (1996) Fuzzy scheduling under resource constraints. Proceedings on
European workshop on fuzzy decision analysis for management, planning and optimization,
pp 121–126
Herrera F, Verdegay JL, Zimmermann H-J (1993) Boolean programming problems with fuzzy
constraints. Fuzzy Sets Syst 55:285–293
Lagarias JC, Reeds JA, Wright MH, Wright PE (1998) Convergence properties of the
Nelder–Mead simplex algorithm in low dimensions. SIAM J Optim 9:112–147
Nelder JA, Mead R (1965) A simplex method for function minimization. Comput J 7:308–313
Tang HW, Qin XZ (2004) Applied optimal method, Dalian Science and Technology University
Press, Dalian
Terlaky T (1985) A convergent criss-cross method, Mathematics of Operationsforschung und
Statistics. Ser. Optimization 16:683–690
Terlaky T (2000) An easy way to teach interior point methods. European Journal of Operations
Research 130(1):1–9
Terlaky T, Zhang S (1993) Pivot rules for linear programming: A survey on recent theoretical
developments. Ann Oper Res 46:203–233
Wagner HM (1958) The dual simplex algorithm for bounded variables. Nav Res Logist Quart
5:257–261
Ward JE, Wendell RE (1990) Approaches to sensitivity analysis in linear programming. Ann
Oper Res 27:3–38
Wolfe P (1963) A technique for resolving degeneracy in linear programming. J SIAM
11:205–211
Wright MH (1996) Direct search method: once scorned now respectable. In: Griffiths DF, Watson
GA (eds) Numerical analysis 1995: proceedings of the 1995 dundee biennial conference in
numerical analysis. Addison-Wesley, Harlow, pp 191–208
Wright MH (1998) The interior-point revolution in constrained optimization, Numerical analysis
manuscript 98–4-09. AT & T BellLab’s, Murray Hill
Zhang S (1999) A new variant of criss-cross pivot algorithm for linear programming. Eur J Oper
Res 116(3):607–614
Zhang JZ, Xu SJ (1990) Linear programming, Science Press, Beijing
Chapter 85
Influence Mechanism of Lean Production
to Manufacturing Enterprises’
Competitiveness
Hong-liang Zhang and Zhan-wen Niu
Abstract The success of Toyota as well as other enterprises of Japan has proved
that lean production can improve manufacturing enterprises’ competitiveness
greatly. However, lean production’s application in other countries is not ideal. One
of the reasons is that lean production is treated as a tool set not as system engi-
neering, so under such background, this paper studies the influence mechanism of
lean production to manufacturing enterprises’ competitiveness upgrading from
systematic perspective. In this paper, lean production is not merely confined to
improvement tools, but is treated as a system, including improvement tools, lean
culture and staff factor. The direct and indirect effect of the three aspects to
manufacturing enterprises’ competitiveness is analyzed by SEM using SMOS17.0.
Analysis result demonstrates the influence mechanism of LP to competitiveness
clearly. The study of this paper has practical sense to lean implementation in China
and meanwhile it enriches lean production theory.
Keywords Competitiveness upgrading Influence mechanism Lean imple-
mentation SEM
85.1 Introduction
Lean production (short of LP) is from Toyota Production System, whose superi-
ority has been proved by success of Toyota Motor Corporation as well as other
Japanese manufacturing corporations. Because it integrates the characters of Ford
H. Zhang (&)
Management Science and Engineering School, An Hui
University of Technology, Ma An Shan, China
e-mail: [email protected]
Z. Niu
Management and Economy Department, Tianjin University,
Tianjin, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 805
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_85,
Ó Springer-Verlag Berlin Heidelberg 2013
806 H. Zhang and Z. Niu
production Mode and handicraft production mode—low cost with high quality,
and can satisfy the diversified need of customer-focused marketing, so it is
regarded as the third production mode. After 1990s, especially after the publication
of book named the machine that changed the world, more and more enterprises
outside of Japan began to learn and apply LP. From theoretical point of view, LP
can upgrade manufacturing enterprises’ competitiveness greatly, but its 20 years’
application process is not smooth, not a few enterprises claim their lean imple-
mentation is failure or didn’t gain desired outcome. Atkinson, Hines et al., Sim and
Rodgers dictated that less than 10 % of UK organizations have accomplished a
successful lean implementation (Bhasin 2012). Famous IE expert of China Er-shi
Qi also pointed out lack of lean environment, enterprises of China encountered
high failure rate in lean implementation process. The reason leading to this phe-
nomenon may be complicated, but the fact that treating LP as merely a tool set
may be one of the key factors.
Under such background, this paper will regard LP as an engineering system and
aims to study the influence mechanism of lean implementation to competitiveness
of manufacturing enterprises, finding out direct and indirect effect of LP’s different
dimensions to manufacturing enterprises’ competitiveness upgrading.
85.2 Related Theory and Hypothesis
LP as one new production mode is not just a set of improvement tools or
technology; in essence, it is complicated system engineering. Many researchers
have aware that besides improvement tools, lean implementation should include
lean culture and staff factor, and some of them have researched single dimen-
sion’s part to manufacturing enterprises’ competitiveness, but few has studied the
relation of the three dimensions as well as their synthetic effect to enterprises’
competitiveness.
85.2.1 Improvement Tools
The viewpoint that improvement tool is one main component of lean implemen-
tation is accepted by many researchers and lean practitioners. Because improve-
ment action must be implemented by some means of tools and lean thought needs
improvement tools to identify, so many researchers paid attention to it. Monden
(2008) pointed out that LP is the compound of JIT production, including field
management, resource management, TQM and information system management
(Monden 2008). Shah and Ward (2007) pointed out LP comprises three aspects
tools, including tool set about supplier management, tool set about customer
management and tool set of inner operation management (Shah and Ward 2007).
Fullerton and McWatters (2002) did appraisal to LP using 10 tools, they are
85 Influence Mechanism of Lean Production 807
focused factory, group technology, Single Minute Exchange of Die, TPM, multi-
skills operator, level operation, purchase on time and TQM (Fullerton and
McWatters 2002). Kojima and Kaplinsky (2004) thought LP system mainly con-
tain three aspects technology, flexibility, quality and persistence (Kojima and
Kaplinsky 2004). Based on the introduction above, this paper gets the following
hypothesis.
H1: application of improvement tools has active influence to manufacturing
enterprise’s competitiveness.
85.2.2 Staff Factor
Famous management expert Peter F. Drucker once said to staff is the only resource
of enterprise, thus management’s crucial purpose to mine staff’s potential. To lean
implementation, staff also plays an irreplaceable role, because staff is the executor
of improvement tool and the carrier of lean culture. As to its importance, FujioCho
once said a sentence ‘‘before making car must first made man’’. Many researchers
also support this viewpoint. In Toyota mode, the internal training material of
Toyota Corporation, respecting for people and continuous improvement are treated
as two pillars of TPS (Ohno 2001). Lander (2007) also pointed out staff is the most
valuable resource of Toyota, so training education and career development is every
important to enterprises (Lander 2007). Monden (2008) thought in order to satisfy
the need of change, the flexibility of staff is very important (Fullerton and
McWatters 2002). Besides direct influence, staff also has indirect influence to
upgrading of competitiveness. As the carrier of lean tool, staff will develop and
adjust lean technology, making it suitable to demand of specific environment and
requirement. So based on the extant research, the following hypothesis is put
forward.
H2: lean staff has positive effect to manufacturing enterprises’ competitiveness.
H3: lean staff has active influence to improvement tool’s development.
85.2.3 Lean Culture
Lean culture cultivation is one important factor to propel lean implementation.
Without dense lean culture, LP can’t be implemented completely for lack of
abiding impetus. Cho (2003), former president of Toyota Motor Corporation, said
Toyota’s strong lies in its shared culture, which means staff of Toyota own the
same values and consciousness (Cho 2003). Koole (2005) also pointed out that
although LP’s outward manifestation is improvement tool, its core is organiza-
tion’s learning ability, so lean implementation effect will be damaged greatly if too
much emphasis is put on tools while lean culture is ignored. Liker (2008) said to
merely applying lean tools or methods is far from enough, only through setting up
808 H. Zhang and Z. Niu
Fig. 85.1 Concept model of
study Improvement tool
Staff factor Enterprise s
competitiveness
Lean culture
talent cultivation system and fostering lean culture can enterprises’ competitive-
ness will be improved everlastingly (Liker 2008). Besides this aspect, dense lean
culture will make staff more actively taking part in improvement and provide
strong dynamic to ensure the improvement is unremitting. On the base of above
discussion, this paper put forward one hypothesis:
H4: lean culture cultivation has positive direct effect to manufacturing enter-
prises’ competitiveness.
H5: lean culture will has active influence to lean staff.
Based on the analysis above, the concept model of this paper is got, see
Fig. 85.1.
85.3 Methodology
85.3.1 Method
This paper will apply structural equation modeling (short of SEM) to verify above
hypothesis. Through seeking variables’ inner structure relation, it can verify
whether the model assumption is reasonable and if theoretical model has fault, it
can point out how to revise. SEM is a group of equations reflecting relation of latent
and observed variables, through measuring observed variables it can infer latent
variables’ relation and verify model’s correctness (Gong et al. 2011). Observed
variables can be measured directly, which is signified by box in path chart, while
due to things’ complexity and abstraction, it is difficult to measure latent variables
directly and in path chart it is signified by elliptic. SEM can substitute for multiple
regression, path analysis, factor analysis as well as covariance analysis and so on
(Zhang and Gao 2012), its application began at late of twentieth century in society,
psychology, education, management, economy as well as other fields.
In studying relation between LP and manufacturing enterprise’s competitive-
ness, traditional quantitative methods are not applicable, because they can not
analysis the relation between multiple latent variables and multiple observed
variables as well as the relevance among latent variables, so SEM is used in this
paper.
85 Influence Mechanism of Lean Production 809
85.3.2 Construction of Variables
Based on extant research and considering the characters of LP and manufacturing
enterprises, this paper designs the observed variables. In the process of designing
questionnaire, this paper first constructs preliminary questionnaire, and then
invests 5 experts of LP and 4 practitioners of LP to give out amendment sug-
gestion, after proper revision, the final questionnaire is got, which adopts 5-rank
Likert scale. Based on the acceptance degree, all the items will provide 5 different
rank answers, 5-completely agree, 4-basically agree, 3-diffcult to determine, 2-
don’t quite agree, 1-completely disagree.
As to improvement tools, by referring extant research, this paper deigns 4
indexes to investigate it, including tools in product design, tools of production
process, tools of field management and tools about supplier management. Under
lean culture, this paper designs the following indexes, supporting of organization,
reward system of improvement, improvement atmosphere and sharing of value. To
staff factor, the investigation indexes include enthusiasm of participating
improvement, career development plan and team work.
As to competitiveness of manufacturing enterprises, not a few researchers got
the conclusion that LP can improve enterprises’ operation efficiency. Liao (2005)
pointed out that lean implementation can make enterprises have many aspects of
advantages, such as lowering WIP, upgrading production flexibility, strengthening
quality control ability and so on (Liao 2005). Besides operation efficiency, LP also
has an active impact on financial performance. Fullerton et al. (2003) got the
conclusion that lean implementation can brought enterprise high profitability,
including return on assets, return on sales and cash flow margin (Fullerton et al.
2003). The third aspect is non-financial performance. Although this aspect was
often ignored by researcher but it relates to enterprise’s long term development.
M. Barad and D. Even Spair aware that Toyota Corporation had more stable
relation with supplier than enterprises of western countries (Barad and Even Spair
2003). Liker also pointed out that putting much attention to business partner and
gave them help as could as possible is one main principle of LP. The study of Gary
Andrew O’ Dell (2003) showed that Japanese manufacturing enterprises imple-
menting LP performed much better on indexes such as pollutant emission, gen-
eration of pollutant and other environmental indexes. So In this paper,
manufacturing enterprise’s competitiveness will be studied from 3 aspects,
including operation efficiency, financial aspect and non-financial aspect.
85.3.3 Data Collection
In data collection, three main ways were used. Firstly, the MBA of Tian Jin
University, who engaged in production management, are investigated in written
form. Secondly, the questionnaire is emailed to potential respondents, located in
810 H. Zhang and Z. Niu
Tian Jin, He Bei, Shan Dong, An Hui and Jiang Su province. Thirdly, field survey.
In this manner chief of production management and employee engaged in lean
improvement are invited to fill the questionnaire. 500 questionnaires are given out,
and 245 effective questionnaires are collected, the recovery rate is 49 %.
85.4 Empirical Study
85.4.1 Data Reliability and Validity
Based on application procedure of SEM, we need to check the reliability and
validity of data. In this paper software SPSS18.0 is used to check data’s reliability
and validity. Firstly, factor analysis is realized by SPSS18.0 and the common
factors are got by principal component analysis. During this process, KMO and
Bartlett’s test are chose to analyze every index. The calculation result shows that
KMO of data sample is 0.826 and Bartlett’s test’s F value reaches significance
level at 0.001, which means the data is suitable to factor analysis. Meanwhile, all
observed variables’ loading coefficient is above 0.69 and common factors’ reli-
ability coefficient belong to interval (0.709, 0.815), the contribution rate of
accumulative total of variance of the 4 common factors is 76.782 %, these means
that the data validity and reliability is favorable. Besides, the analysis result shows
the coefficient of Cronbach a to every index is above 0.8, which means the internal
consistency of data is good. Analysis result is showed in Table 85.1
Table 85.1 Reliability coefficient and factor analysis result
Measurement items Factors loadings Loading Reliability Rate of
coefficient coefficient accumulative
total of variance
Improvement tools Tools in product design 0.723 0.775 76.782%
Tools of production process 0.832
Tools of field management 0.785
Tools of supplier management 0.759
Lean culture Supporting of organization 0.826 0.709
Reward system of 0.867
improvement
Improvement atmosphere 0.766
Sharing of value 0.797
Staff factor Participating improvement 0.782 0.815
Career development plan 0.771
Team work 0.756
Enterprises Operation efficiency 0.692 0.759
competitiveness Financial performance 0.757
Non-financial performance 0.802
85 Influence Mechanism of Lean Production 811
Table 85.2 Test of model fitting goodness
Fit index v2/df RMSEA AGFI IFI NFI CFI
Fitted value 2.21 0.042 0.906 0.931 0.917 0.925
Adaptation standard \3 \0.05 [0.90 [0.90 [0.90 [0.90
Fig. 85.2 Path loading
coefficient Improvement tool
0. 51 **
0. 40*
0. 68** Enterprise’s
Staff factor
competitiveness
0. 45**
0. 57***
Lean culture
85.4.2 Model-Fitting Degree Analysis
After satisfying measurement requirements, this paper does statistical test to
concept model showed in Fig. 85.1, using software AMOS 17.0. Analysis result is
listed in Table 85.2, which shows that the fitting degree of concept and the data is
favorable.
85.4.3 Path Coefficient Analysis
Figure 85.2 shows that except the path coefficient (045) of staff factor to
improvement tool merely arrives a = 0.05 significance level, all other coefficients
reach a = 0.01 significance level, especially the coefficient of lean culture to
enterprise’s competitiveness arrives a = 0.001 significance level. So the five
hypotheses put forward in this paper are all supported. The Fig. 85.2 shows both
the direct and indirect influence of lean dimension to manufacturing enterprise
competitiveness. Concretely, to improvement tool, its direct influence is 0.51 and
indirect influence is 0, so its comprehensive influence to competitive is 0.51; to
staff factor, its direct influence is 0.68 and indirect influence is 0.40 9 0.51 =
0.204, so its comprehensive is 0.884 and to lean culture, its direct influence is 0.57,
indirect influence is 0.45 9 0.68 = 0.306, so its comprehensive influence is 0.876.
812 H. Zhang and Z. Niu
85.5 Conclusion
This paper studies the relation of LP and manufacturing enterprise’s competi-
tiveness from systematic angle, which no longer confine LP to improvement tool.
The analysis result tells us that all the three dimensions have positive effect to
manufacturing enterprises’ competitiveness. So in implementing LP, the enter-
prises should not ignore whichever aspect. As to staff factor, it has both direct and
indirect effect to manufacturing enterprises’ competitiveness, and its comprehen-
sive influence coefficient is maximal in the three aspects, so the enterprises should
put much emphasis on this aspect, taking measures to encourage employee to take
part in improvement, perfecting their career development plan and encouraging
employee to participate all kinds of improvement team. In order to cultivate lean
culture, the organization should actively develop dense atmosphere and adjust
traditional award system to adapt to lean implementation. Improvement tool,
which has been gave much attention by researcher and enterprises, has direct effect
to upgrading enterprise competitiveness, but in lean implementation, the imple-
menter should not limit it to production link merely, much more emphasis should
be put to improvement tool about facility layout, supplier and customer
management.
In a word, the study of this paper will rich the theory of LP and has an active
part for successfully implementing LP in China. But confined to ability and time,
the study of this paper is not deep enough, related study can be done further.
Firstly, the investigation sample mainly distribute in Tian Jin, He Bei, An Hui,
Shan Dong and Jiang Su, so if the range of investigation is enlarged the result may
be different. Secondly, the study of this paper is done under comprehensive
dimension, so if the study refines to specific index under every dimension, the
result will be much richer.
Acknowledgments This work is financially supported by national natural science foundation of
China (71071107) as well as education department of An Hui province of China (SK2012B085).
References
Barad M, Even Spair D (2003) Flexibility in logistic system-modeling and performance
evaluation. Int J Prod Econ 85(2):155–170
Bhasin S (2012) An appropriate change strategy for lean success. Manag Decis 50(3):439–458
Cho F (2003) Toyota—Kingdom on the evolution. Japan Econ 2:23
Fullerton RR, McWatters CS (2002) The role of performance measures and incentive system in
relation to the degree of JIT implementation. Acc Organ Soc 27(27):711–735
Fullerton RR, McWatters CS, Fawson C (2003) An examination of the relationships between JIT
and financial performance. J Oper Manag 21(4):383–404
Gong W, Wang J, Chen J, Ge C (2011) Empirical study of influencing factors on cooperation
performance in reverse supply chain. Ind Eng Manag 16(1):6–11
85 Influence Mechanism of Lean Production 813
Kojima S, Kaplinsky R (2004) The use of a lean production index in explaining the transition to
global competitiveness: the auto components sector in South Africa. Technovation
24(3):199–206
Koole SE (2005) Removing borders: the influence of the Toyota Production System on the
American office furniture manufacturer. Ph. D. Grand Valley State University, America
Lander E (2007) Implementing Toyota-style systems in high variability environments. Ph.D.
Dissertation, Michigan University, America
Liao I-H (2005) Designing a lean manufacturing system: a case study. Ph.D. Dissertation,
Binghamton University, America
Liker JK (2008) Toyota culture- the heart and soul of the Toyota way. China Machine Press
Monden Y (2008) Toyota production system. He Bei University Press, China
Ohno T (2001) The Toyota way 2001. Toyota Motor Corporation, Toyota
Shah R, Ward PT (2007) Defining and developing measuring of lean production. J Oper Manag
25(1):785–805
Zhang W, Gao X (2012) Structural equation model analysis of foreign investment, innovation
ability and environmental efficiency. China Soft Sci Mag 1(3):170–180
Chapter 86
Mobile Device User Research in Different
Usage Situation
Wei Liu and Jiu-zhou Li
Abstract In this paper, we report the difference of users’ cognition and operating
efficiency in three typical situations, such as noisy, dark, walking condition. The
single-factor experiment’s result data suggest noisy effects mobile user’s cognition
significantly, and walking situation will affect users’ performance to some extent,
but user experience does not relevant to different situation.
Keywords Mobile device Operating efficiency User cognition User
experience
86.1 Introduction
With the rapid development of mobile devices and mobile Internet, user has
entered the ‘‘experience economy era’’ (Luo 2010). The great successes of the
Apple’s range of products proved user-centered-design and great attention paid on
user experience are very important for a company. Compared to product in other
area, mobile devices are been used in a more complex environment. And user’s
cognition, operating efficiency and subjective experience are not the same in
different situation. Mobile device screen space is also an important factor distin-
guished from other products in the field. How to improve the user experience is a
very challenging task.
In this paper, we study mobile users’ cognition, operating efficiency and user
experience in there different typical situations, aiming to provide basis to enhance
mobile device user experience.
W. Liu J. Li (&)
Beijing University of Posts and Telecommunications, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 815
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_86,
Ó Springer-Verlag Berlin Heidelberg 2013
816 W. Liu and J. Li
86.1.1 User Experience of Mobile Device
An abstract definition of the user experience: users’ all aspects of perception when
they interact with products and services (UPA 2006). Garrett thinks that the user
experience includes brand characteristics, information availability, and function-
ality, content and other aspects of experience. Mobile user experience involves a
wide range, in addition to hardware, now more concern is paid on the operating
system, applications, and interface design experience.
86.1.2 User Context Cognitive Psychology
Situated cognition theory thinks the cognitive process is constructed by the situ-
ation, guidance and support, and individuals’ psychology usually active in the
context (Du 2007). When people use mobile devices, their brain is composed by
attention, comprehension and retention. Attention can be divided into centralized,
decentralized, and transfer. While there is a clear demand or potential interests,
user tends to be concentrated. But when there is interference information and time
is uncontrollable, they will be distracted. User will transfer their attention to
explore information. Fitts’s Law suggests that reducing the distance between the
starting position to the target distance and increasing the size of target can
accelerate the speed user find the target (Luo 2010). User’s operation habits and
interaction expectation own its unique mental model when they using mobile
device.
86.1.3 Operating Ergonomics of Mobile Users
Efficiency is the indicator to measure the relationship between correctness of user
task and the degree of completion with the amount of resources used to complete
the task. In the field of mobile device interaction, efficiency can be considered as:
high efficiency means users complete the task goals with fewer operations and less
time (Ingwersen 2000). The methods to evaluating performance on mobile device
include focus groups, cognitive path method, and heuristic evaluation. User per-
formance testing is to observe and analyze the performance of the user’s actions
under the experimental conditions (Du 2007).
86 Mobile Device User Research 817
86.1.4 Experiment Purposes
This paper studies in different situation mobile device (In this paper, refers to the
cell phone) users’ cognitive, operating ergonomics and user experience differ-
ences, aiming to provide a theoretical basis for the mobile design. This study
belongs to the scope of psychology and ergonomics, and we analysis experimental
quantitative data and supplemented by qualitative methods in the experiment.
86.2 Methodology
86.2.1 Experiment and Participants
24 participants (11 males and 13 females), who are 22–25 years old, and familiar
with mobile phones, with a certain touch-screen mobile phone operating experi-
ence, but never use HTC Desire HD, phones with Android 2.3 operating system
and 365 curriculum applications. Their vision is normal and corrected visual acuity
above 1.0, right-handed. After a brief study, all of them can cooperate with host to
complete the test and questioners independently. They never participated in similar
experiments.
86.2.2 Experiment Apparatus
In this paper, we use the HTC Desire HD phone which screen size is 4.3 inches,
resolution is 480 * 800 px, and operation system is Android 2.3.2. All the
experiment is been hold in a lab where participants can walk small-scale.
86.2.3 Experiment Material
Experiment materials include four Icon list picture, 365 curriculums Version1.1
(Android), the availability of subjective evaluation questionnaire and user expe-
rience evaluation questionnaire (Figs. 86.1 and 86.2).
86.2.4 Experiment Design
The experiment use single-factor methods to divide participants into three
experimental groups and control group by different usage scenarios. Experimental
818 W. Liu and J. Li
Fig. 86.1 HTC Desire HD and 365 curriculum applications
Fig. 86.2 Pictures of icon table
group variables are noisy, dark and waking situation (Yamabe 2007). Control
group participants have test in light-sufficient, place-fixed, and quiet indoor. The
users just participate in one group. After the experiment, all participants complete
subjective assessment questionnaire (Fig. 86.3).
86.2.5 Experiment Procedure
Each participant is tested separately. After completing basic personal information
form, host tell participant about the experiment procedure and precautions.
First part of the experiment is icon list testing. Before the start, participant
learns icons’ meanings. In the formal experiment, participant browse four picture
of icon list successively on the phone at a stipulated time. Then they need to
86 Mobile Device User Research 819
Fig. 86.3 Experiment in four different situations
answer some questions about the pictures. Those questions are used to measure
participants’ cognition.
Experiment second part is testing participants’ performance of operating 365
curriculum application. Before the experiment, participants have some time to use
the application indecently. In the formal experiment, there are three tasks for
participants to complete. The first is to find the timetables of the day, at the
meantime host records the time participants used to complete the task. Then
participants need to answer two questions about the operation. The second task is
to remove the day of a lesson, also host records the time to complete the task. The
third task is to set the personal information, and host records the time. After
completing those tasks, participants have to fill in the form of testing subjective
experience. The form is 7-point scale and has 10 questions.
86.2.6 Data Processing
In this paper, we give task performance and experience evaluation different
weights according to participants’ interviews (Li 2009). As to operating perfor-
mance evaluation experiment, task 1, task 2 and task 3 are set to 0.4, 0.2 and
0.4.And Assessment of user experience five measure, icon test experience degrees,
test experience degrees, mobile phone experience degrees, mental state, experi-
mental performance of the self-assessment are set to 0.25, 0.35, 0.20, 0.15, 0.05
according to experts suggests. We use SPSS 17 for data management and analysis.
820 W. Liu and J. Li
Fig. 86.4 User cognition
comparison among four
different situations
86.3 Results
86.3.1 The Effect of Situation to Cognition
After standardizing the data of each group answers to objective questions, the
result shows in Fig. 86.4. It suggests participants’ cognition have certain differ-
ences between the four groups, and the best one is control group which are without
any interferences, on the opposite side, the noisy group’s cognition is lowest.
Table 86.1 shows the result of experiment data to two-tailed T-test. The
bilateral Sig = 0.049 \ 0.05 of noisy group data to the control group data suggests
in the 95 % significance level, the noisy group’s cognition are significantly dif-
ferent to that of control group. It means noisy satiation impact on user cognition
greatly.
86.3.2 The Effect of Situation to Operating Performance
The results of homogeneity of variance, multiple comparisons to the three
experiment groups with control group show in Table 86.2. Comparing with three
experiment groups’ operating time with that of control group, the significances are
larger than 0.1. It means noisy, dark and walking environment doesn’t impact
operating performance significantly.
Figure 86.5 shows the three variables of the experiment don’t impact users’
operating performance significantly. But walking user need more time to complete
the task than other groups.
86 Mobile Device User Research 821
Table 86.1 User cognition T test in four different situations
N AV SD Sig
Normal 6 0.755 0.238 None
Noisy 6 0.500 0.121 0.049
Dark 6 0.640 0.209 0.393
Table 86.2 User performance multiple test in four different situation
N AV SD Sig
Normal 6 10.053 1.676 None
Noisy 6 9.907 2.816 0.049
Dark 6 11.360 3.055 0.393
Walking 6 12.440 3.670 0.246
Fig. 86.5 User performance
comparison among four
different situations
86.3.3 The Effect of Situation to User Experience
In this paper, we use several aspects to assess the degree of the overall user
experience by setting weights to aspects.
Figure 86.6 suggests that the overall experience evaluation of the four groups in
the experiment is not much difference. Probably the user experience of the eval-
uation itself is a subjective evaluation, so the evaluation of the product due to
different user groups will be very different.
86.4 Discussion
According to the analysis to the experiment data, noisy, dark and walking situation
have a certain on mobile users’ cognition, operating performance and user
experience.
822 W. Liu and J. Li
Fig. 86.6 User experience
comparison among four
different situations
Table 86.3 Comparison between noisy and normal situation
Cognition Performance User experience
Noisy 0.500 9.907 0.720
Normal 0.755 10.053 0.732
Sig 0.049 0.931 0.849
Table 86.4 Comparison between dark and normal situation
Cognition Performance User experience
Noisy 0.640 11.360 0.750
Normal 0.755 10.053 0.732
Sig 0.393 0.444 0.772
In a noisy environment, users’ cognition is lower than that of users in normal
situation, but there are little difference between operating performance and level of
user experience. So designers need to consider users’ the lower awareness in noisy
situation, and arrange interface information reasonably, such as reducing the
information in mobile interface (Table 86.3).
In dark environment, mobile users’ cognition, operating performance and the
degree of user experience is not much different with those in normal situation, so
designer need a more in-depth understanding of the user needs, such as adjusting
the screen brightness interface style (Table 86.4).
Table 86.5 shows that in the walking situation operating performance and user
experience difference is not significant, but compared to the other two, in walking
situation the user’s operating performance have significant difference. So designers
should focus on users’ depth needs in the walking situation, adjusting the structure
of the interface, and simplify the task of the interactive process, which can enhance
the user experience on the equipment.
86 Mobile Device User Research 823
Table 86.5 Comparison between walking and normal situation
Cognition Performance User experience
Noisy 0.590 12.440 0.702
Normal 0.755 10.053 0.732
Sig 0.246 0.169 0.748
86.5 Conclusion
The experiment studies in different situation, the difference of mobile device (This
article refers to the cell phone) users’ cognitive, operating performance and user
experience. The experiment uses methods of control groups and the single factor
control to analysis the experiment data. And the result of experiment shows the
impact of noise on the cognitive situation of mobile device users significantly;
while the walking environment has a certain impact on the user’s operating per-
formance. The impact of different situations to user experience is not significant.
Actual design should analysis the characteristics of the product and the users’ own
characteristics deeply.
References
Du Y (2007) Context-aware learning for intelligent mobile multimodal users interfaces, pp 1–5
Ingwersen P (2000) Cognitive information retrieval annual review of information science and
technology
Li J (2009) Comprehensive evaluation of the cognitive load of human-computer interaction
process
Luo S (2010) Context-based user experience design in the mobile interface
UPA Usability Body of Knowledge, 2006.9. http://www.usabilitybok.org/glossary
Yamabe T (2007) Experiments in mobile user interface adaptation for walking users
Chapter 87
Optimal Enterprise Cash Management
Under Uncertainty
Xiao-yun Wei and Li-yan Han
Abstract We present a dynamic model for enterprise cash management under
uncertainty. The numerical method was used to obtain optimal level of cash
holding. The results show that higher yield volatility of financial assets, liquidation
cost of financial assets and coefficient of risk aversion will raise the demand for
cash. It also shows that the optimal choice of inter-temporal model is different
from that of single-period model. The former makes the manager choose to hold
more cash. The reason is that long-horizon managers have an intrinsically larger
need for cash to quell possible transaction and precautionary demand.
Keywords Cash management Financial assets Uncertainty
87.1 Introduction
Optimization models for cash management can be divided into two main groups
based on objection function. The first deals with demand by cost-benefit or loss-
benefit analysis, pioneered by Baumol–Tobin model (Baumol 1952; Tobin 1956),
and extended, among others, by Frenkel and Jovanovic (1980, 1981), Bar-Ilan
(1990), Dixit (1991), Ben-Bassat and Gottlieb (1992), Chang (1999) and Perry and
Stadje (2000). In this approach the optimal demand for cash is decided by the
trade-off between opportunity cost and benefits of cash holding. The second cat-
egory of models concerns the demand by drift control theory, pioneered by Miller
and Orr (1966), and extend by Bar-Ilan et al. (2004), Bar-Ilan and Lederman
(2007).
X. Wei (&) L. Han
School of Economics and Management, Beihang University, Beijing 100191, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 825
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_87,
Springer-Verlag Berlin Heidelberg 2013
826 X. Wei and L. Han
However, the authors above mainly consider cash only and they consider either
single period or infinite horizons only. In this paper, we present a model to obtain
the optimal allocation ration between cash and financial assets based on utility
maximization for different horizons. The model departs from portfolio choice
theory (Barberis 2000), Aizenman and Lee (2007) and multi-period newsboy
model (Matsuyam 2006), and instead emphasizes the importance of cash in pro-
viding insurance for bankruptcy.
The rest of paper is organized as follows. In Sect. 87.2 we introduce the
framework of cash management. A numerical example is presented particularly to
calibrate our model in Sect. 87.3. Section 87.4 offers some concluding remarks.
87.2 The Model
We assume the manager, as a centralized decision maker in the enterprise,
determines the split of the given level of assets Wt between cash and financial
assets, where Rt and St denote two assets in the end of t period respectively. The
enterprise has to meet the demand for payments such as transaction and debt
repayment thus to reduce the probability of a possible bankruptcy. Then enterprise
generally put its cash in commercial bank or purchase government bonds in order
to seek greater security and liquidity but lower risk-free return. In contrast, the
manager will typically pursue higher return by investing the remaining assets in
risky long-term assets such as longer-term government bonds, stocks, corporate
bonds, oil, minerals and real estate. The objective of the manager is to earn more
profit on the basis of enterprise stability. The question, which is concerned, is to
maximize the enterprise’s utility function. In order to formulate this problem, the
dynamics of the total assets Wt of enterprise will be introduced.
Consider we are at initial time 0 and want to write down the allocation problem
for a manager with a horizon of t periods. We suppose the real annually interest
rate on cash is rf ;t . The return on financial assets rs;t is assumed to follow an
independently identical distribution, and cov rs;s ; rs;t ¼ 0; 8s 6¼ t; s; t ¼ 1; 2; 3. . ..
Except the fundamental function of cash for transaction and debt payments,
another function is to enhance confidence of investors, which is not considered by
Bar-Ilan et al. (2004). This is the paradox of cash management—the more cash
holdings, the lower may be the demand for them. Then the real demand for cash in
period t, Yt is given by
Yt ¼ gðXt ; Rt1 ; St1 Þ ð87:1Þ
where Xt , with a corresponding density function f ðXt Þ, is the value of demands
when the enterprise has no cash and financial assets. Then we have
gðXt ; 0; 0Þ ¼ Xt . And Yt is a strictly increasing function of Xt and decreasing
function of Rt1 and St1 . The joint density function for ðX1 ; X2 ; X3 ; . . .; Xt Þ is
given by
87 Optimal Enterprise Cash Management 827
Y
t
f ðX1 ; X2 ; X3 ; . . .; Xt Þ ¼ f ð Xn Þ ð87:2Þ
n¼1
In other words, the distributions of Xn and Xm ðn 6¼ mÞ are independent each
other. We assume that Lt and Ct are lower and upper bound of demand Xt , and
lt ¼ Lt =Wt1 , ct ¼ Ct =Wt1 .
If cash holding Rt1 at the end of period t 1 or at beginning of period t is
higher than the demand Yt in period t, the remaining cash Rt1 Yt will earn rf ;t
rate of return and financial assets St1 earn rs;t rate of return. When Rt1 is lower
than Yt , the liquidation takes place and the level of cash reduces to zero. ht is the
liquidation cost which must be paid for a unit of cash when the demand can’t be
complied with. Then ðYt Rt1 Þð1 þ ht Þ units of financial assets must be liqui-
dated to obtain Yt Rt1 units of cash. We denote xt ¼ Rt1 =Wt1 is the allo-
cation ratio to the cash at the beginning of period t. yt ¼ Yt =Wt1 is the proportion
of demand for cash to total assets Wt1 at the end of period t 1. Thus if t is larger
than 1, xt is decided passively. Then Wt is given by the following expressions
(87.1)–(87.3), where the first subscript of Wt and xt denotes the period and the
second subscript denotes the scenario.
(1) t = 1
y1 x1 ) W1;1 ¼ W0 ðx1 y1 Þ 1 þ rf ;1 þ ð1 x1 Þ 1 þ rs;1 þ y1 ;
y1 [ x1 ) W1;2 ¼ W0 ½ð1 x1 Þ ðy1 x1 Þð1 þ h1 Þ 1 þ rs;1 þ y1 ;
(2) t = 2
y1 \x1 ; y2 \x2;1 )
W0 ðx1 y1 Þ 1 þ rf ;1
x2;1 ¼ ;
W1;1
W0 ð1 x1 Þ 1 þ rs;1
1 x2;1 ¼ ;
W1;1
W2;2 ¼ W1;1 1 x2;1 y2 x2;1 ð1 þ h2 Þ 1 þ rs;2 þ y2 ;
828 X. Wei and L. Han
y1 [ x1 ; y2 \x2;2 )
x2;2 ¼ 0;
" #
ð1 x1 Þ
W0 1 þ rs;1
ð y 1 x 1 Þ ð 1 þ h1 Þ
1 x2;2 ¼ ;
W1;2
" #
x2;2 y2 1 þ rf ;2
W2;3 ¼ W1;2 ;
þ 1 x2;2 1 þ rs;2 þ y2
y1 [ x1 ; y2 [ x2;1 )
x2;2 ¼ 0;
" #
ð1 x1 Þ
W0 1 þ rs;1
ðy1 x1 Þð1 þ h1 Þ
1 x2;2 ¼ ;
W1;2
(" # )
1 x2;2
W2;4 ¼ W1;2 1 þ rs;2 þ y2 ;
y2 x2;2 ð1 þ h2 Þ
y1 [ x1 ; y2 [ x2;1 )
W0 ðx1 y1 Þ 1 þ rf ;1
x2;1 ¼ ;
W1;1
W0 ð1 x1 Þ 1 þ rs;1
1 x2;1 ¼ ;
W1;1
W2;1 ¼ W1;1 ðx21 y2 Þ 1 þ rf ;2 þ 1 x2;1 1 þ rs;2 þ y2 ;
(3) t
Let a set Rþ be defined by
Rþ ¼ faja 0; a 2 Rg
Where R denotes a set of all real numbers. Moreover, At and Bt , t ¼ 1; 2; 3. . .,
are defined by
At ¼ fyt jyt xt ; yt 2 Rg; Bt ¼ Rþ At
87 Optimal Enterprise Cash Management 829
Then we denotes
X 1 ¼ A1 A2 At ;
X2 ¼ A1 A2 At1 Bt ;
X2t 1 ¼ B1 B2 Bt1 At ;
X2t ¼ B1 B2 Bt :
y ¼ ðy1 ; y2 ; . . .; yt ÞT
y 2 X1 )
Wt2;1 xt1;1 yt1 1 þ rf ;t1
xt;1 ¼ ;
Wt1;1
Wt2;1 1 xt1;1 1 þ rs;t1
1 xt;1 ¼ ;
Wt1;1
" #
xt;1 yt 1 þ rf ;t
Wt;1 ¼ Wt1;1 ;
þ 1 xt;1 1 þ rs;t þ yt
y 2 X2 )
Wt2;1 xt1;1 yt1 1 þ rf ;t1
xt;1 ¼ ;
Wt1;1
Wt2;1 1 xt1;1 1 þ rs;t1
1 xt;1 ¼ ;
Wt1;1
(" # )
1 xt;1
Wt;2 ¼ Wt1;1 1 þ rs;t þ yt ;
yt xt;1 ð1 þ ht Þ
……
y 2 Xt1 )
xt;2t1 ¼ 0;
" #
1 xt1;2t1
Wt2;2t2 1 þ rs;t1
yt1 xt1;2t1 ð1 þ ht1 Þ
1 xt;2t1 ¼ ;
Wt1;2t1
" #
xt;2t1 yt 1 þ rf ;t
Wt;2t 1 ¼ Wt1;2t1 ;
þ 1 xt;2t1 1 þ rs;t þ yt
830 X. Wei and L. Han
y 2 Xt )
xt;2t1 ¼ 0;
" #
1 xt1;2t1
Wt2;2t2 1 þ rs;t1
yt1 xt1;2t1 ð1 þ ht1 Þ
1 xt;2t1 ¼ ¼ 1;
Wt1;2t1
(" # )
1 xt;2t1
Wt;2t ¼ Wt1;2t1 1 þ rs;t þ yt ;
yt xt;2t1 ð1 þ ht Þ
The manager’s preferences over terminal wealth are described by constant
relative risk-aversion utility functions of the form
Wt1A
uð W t Þ ¼ ð87:3Þ
1A
The manager’s problem is to solve equation
V ðWt Þ ¼ max E E0 uðWt Þrs;1 ; rs;2 ; . . .; rs;t
x1
82 Z 3 9
> >
>
>6 u W t;1 h ð y Þdy >
>
>
> 7 >
>
>
> 6 X 1 7 >
>
>
<66 Z 7 >
= ð87:4Þ
7
¼ max E 6 þ6 7
u Wt;2 hð yÞdy. . . 7jrs;1 ; rs;2 ; . . .; rs;t
x1 >
> 6 X2 7 >
>
>
>6 Z 7 >
>
>
> 4 5 >
>
>
> >
>
: þ u Wt;2t hð yÞdy ;
X 2t
Where max denotes the problem is solving the optimal x1 , and denotes the fact
x1
that the manager calculates the expected return from the beginning of period 1 on.
E is the expectation operator of rs;t . hð yÞ is the joint density function for
ðy1 ; y2 ; y3 ; . . .; yt Þ.
When budget constraint binds, that is, ð1 xt Þ ðyt xt Þð1 þ ht Þ\0, the
final wealth in period t is 0. Now even though the manager liquidates all financial
assets, they could not satisfy the payments, the bankruptcy is likely to occur.
87.3 Numerical Example
The numerical solution is solved as follows. We assume
PðXt ¼ iÞ ¼ 0:25; i ¼ 1; 2; 3; 4 ;
P rs;t ¼ 0:03 ¼ P rs;t ¼ 0:08 ¼ 0:5; Lt ¼ 1; Ht ¼ 4 ;
Yt ¼ Xt 0:005Rt1 0:002St1 and W0 ¼ 12 ;
rf ;t ¼ 0:04; ht ¼ 0:5; A ¼ 5; t ¼ 1; 2; 3; 4; 5; 6:
87 Optimal Enterprise Cash Management 831
Fig. 87.1 Optimal allocation
to cash for different horizons
The simulation results are reported in Fig. 87.1. The optimal x1 maximizes the
function (87.4).
The results show that the optimal choice of inter-temporal model is different
from that of single-period model. The former makes the manager choose to hold
more cash. The reason is long-horizon managers has an intrinsically larger need
for cash to quell possible transaction and precautionary demand. And we also
conclude that higher yield volatility of financial assets, liquidation cost of financial
and coefficient of risk aversion will raise the demand for cash. For saving space,
we omitted the figure in the paper.
87.4 Concluding Remarks
This paper presents a dynamic model for enterprise cash management under
uncertainty. The numerical method was used to obtain optimal level of cash
holdings. The results show that higher yield volatility of financial assets, liqui-
dation cost of financial assets and coefficient of risk aversion will raise the demand
for cash. It also shows that the optimal choice of inter-temporal model is different
from that of single-period model. The former makes the manager choose to hold
more cash. The reason is long-horizon managers has an intrinsically larger need
for cash to quell possible transaction and precautionary demand.
References
Aizenman J, Lee J (2007) International reserves: precautionary versus mercantilist views, theory
and evidence. Open Econ Rev 18(2):191–214
Barberis N (2000) Investing for the long run when returns are predictable. J Finance
LV(1):225–264
Bar-Ilan A (1990) Overdraft and the demand for money. Am Econ Rev 80:1201–1216
Bar-Ilan A, Lederman D (2007) International reserves and monetary policy. Econ Lett
97:170–178
832 X. Wei and L. Han
Bar-Ilan A, Perry D, Stadje W (2004) A generalized impulse control model of cash management.
J Econ Dyn Control 28:1013–1033
Baumol W (1952) The transaction demand for cash—an inventory theoretic approach. Q J Econ
66:545–556
Ben-Bassat A, Gottlieb D (1992) Optimal international reserves and sovereign risk. J Int Econ
33:345–362
Chang F (1999) Homogeneity and the transactions demand for money. J Money Credit Bank
31:720–730
Dixit A (1991) A simplified exposition of the theory of optimal control of Brownian motion.
J Econ Dyn Control 15:657–673
Frenkel JA, Jovanovic B (1980) On transactions and precautionary demand for money. Q J Econ
94:24–43
Frenkel J, Jovanovic B (1981) Optimal international reserves: a stochastic framework. Econ J
91:507–514
Matsuyam K (2006) The multi-period newsboy problem. Eur J Oper Res 171:170–188
Miller M, Orr D (1966) A model of the demand for money by firms. Q J Econ 81:413–435
Perry D, Stadje W (2000) Martingale analysis of a stochastic cash fund model. Insur Math Econ
26:25–36
Tobin J (1956) The interest elasticity of the transaction demand for cash. Rev Econ Stat
38:241–247
Chapter 88
Problem Analysis and Optimizing
of Setting Service Desks in Supermarket
Based on M/M/C Queuing System
Chun-feng Chai
Abstract Queuing problem is an important factor that affects operation level and
efficiency of the supermarket. To solve the queuing issue properly through
effective measures has become a top priority for supermarket. This paper provides
reference for decision at the issue of optimizing the quantity of service cashier
desks, improving service efficiency and decreasing operating costs. It analyses
Supermarket cashier system queuing issues through establishing M/M/C queuing
model on basis of operation research.
Keywords M/M/C queuing model Operation research Optimization of
queuing system Queuing theory
88.1 Introduction
In general we do not like to wait. But reduction of the waiting time usually
requires extra investments. To decide whether or not to invest, it is important to
know the effect of the investment on the waiting time. So we need models and
techniques to analyze such situations (Adan and Resing 2001). Going shopping in
the supermarket during spare time has become a kind of our life habit today
popular with supermarket. We enjoy the life of shopping, also are worried about
the problems of supermarket cashier service system. Less opening number of
cashier desks will lead to more customers waiting for too long service, which will
cause the customer dissatisfaction and running away. If the supermarket opens too
many service desks, it can reduce customer’s waiting time and length of line, but
this will increase the supermarket operating costs. The supermarket operator must
C. Chai (&)
School of Economy and Management, Taiyuan University of Science
and Technology, Taiyuan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 833
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_88,
Ó Springer-Verlag Berlin Heidelberg 2013
834 C. Chai
consider how to balance the two factors. As the deal terminal between supermarket
and consumer, the service desk has a direct influence on the image of the service
quality and efficiency. Also it can affect the operation level and efficiency of the
whole supermarket. Therefore, how to dynamically and reasonably arrange the
number of service desks according to the customer flow and the time needed and
how to balance the customer satisfaction and operational cost, are the problems
which should be solved by the enterprise.
88.2 Methodology
88.2.1 M/M/C Queuing Models
Queuing theory, as a branch of the operational research, is a mathematical theory
and method which research system of gather phenomenon and random service
system work process, also called the stochastic service system theory. M/M/C
queuing model (see Fig. 88.1) assume service system having the following fea-
tures (Mandelbaum and Hlynka 2009).
The interval of customer reaching the service desks stochastically obeys
parameter k ([0) index distribution. That means that a single customer reached
service desks independently at random according to Poisson Distribution, by
Poisson distribution definition, there are k average arriving customers in unit
interval. When customer approach to the desk, if the desk is spare, customers
immediate get the service, and if the service desk are busy, the customer wait in
line until a desk gets leisure. Queuing rules obey first-come- first- service, and
customers leave after getting service. Assuming service system capacity is infinite,
and there are C desks working independently for each other, and service time
obeys negative exponential distribution with parameters l ([0). Generally
speaking, evaluation index are average length of line and the average waiting time
(Hlynka 2010; Takacs 1962).
Fig. 88.1 M/M/C queuing Queuing system
model
1
Service cashier desk
Customer Source
2 Leaving
Poisson
3
Process
Queue
C
88 Problem Analysis and Optimizing of Setting Service Desks in Supermarket 835
88.2.2 Model Assumption
Any queuing system is consist of three parts. They are respectively input process,
queuing discipline and service agency. From these three parts, a hypothesis for
Supermarket cashier service system can be made (Yan 2012).
88.2.2.1 Input Process
The hypothesis of input process is mainly for the customer getting to the super-
market cashier service system (Zheng and Gu 2005). First of all, customer source
is infinite. Secondly, in the description of the features, customer reaching cashier
desk is random and independent. Besides, the following features can be learned:
the arrival of the customer numbers any time has nothing to do with arriving
moments, but only with the time interval; the probability of the arrival of two
customers at the same time is almost zero. Through the above analysis, it can
assume that the input process of supermarket cashier service system is a Poisson
process and that the number of arrival customers in unit interval meets Poisson
distribution with parameters k.
88.2.2.2 Queuing Discipline
The customer to the supermarket cashier service system is random. If there is free
service, then into service; if not, then the customers wait in line. If Supermarket
layout rationally, the space is enough without jams, the customers will generally
choose the shortest lines to wait for service (Wang and Miao 2012). When the
other lines get shorter while waiting, the customers will change the line imme-
diately, so the queuing discipline of cashier service system in supermarket is the
rules of first- come- first- service in waiting rules (Huang and Xiao 2009).
88.2.2.3 Service Agency
There are C check stands in the supermarket and work independently among each
other. Based on customer order of line up service customer, one customer is served
once. Because the goods kinds and number customers brought are different, the
service time of check stand is random. The service time of check stand of
supermarket can be assumed to meet the negative exponential for the l parameter
(Liu and Liu 2009).
836 C. Chai
88.2.3 Establishing the Model
According to the above analysis and assumptions established, we know that in the
system there are C working check stands, arrival of the customer meet the k
Poisson distribution. The service time of every customer is independent and meets
the negative exponential for the l parameter. System capacity infinite. If the
service window is busy when customer arrives, then wait. The Supermarket
cashier queuing system is a M/M/C queuing system (Deng 2000; Liu et al. 2011).
According the Little formula to set parameters q service intensity, k as cus-
tomer arrival rate, l as average service time, c is the number of cashier open.
When q\1, the system can achieve a steady state, and has a smooth distribution
(Li et al. 2000; Zhou 2011).
" c #1
c1 k
X 1 k 1 1 k
P0 ¼ þ ð88:1Þ
k¼0
k! l c! 1 q l
ck k
k! qc P0 0 k\c
Pk ¼ c k ð88:2Þ
c! q P0 kc
Through analysis of the system, the following corresponding target parameters
can be drawn (Sun 2007).
Average waiting length of team:
ðcqÞc q
Lq ¼ P0 ð88:3Þ
c!ð1 qÞ2
The average of system length of team (or the average number of customers
waiting in the system)
k ðcqÞc q k
L ¼ Lq þ ¼ P þ
2 0
ð88:4Þ
l c!ð1 qÞ l
Average waiting time for the customer in the system:
Lq
Wq ¼ ð88:5Þ
k
Average staying time for the customer in the system
L 1
W¼ ¼ Wq þ ð88:6Þ
k l
When the system is stable, if less than the longest average waiting time and
team length customer can endure, the supermarket opened the least service desks,
which not only can get customer satisfaction, and also reduce the operation cost of
the supermarket. Here firstly supposing the longest waiting time is T1, the longest
88 Problem Analysis and Optimizing of Setting Service Desks in Supermarket 837
team length customer can endure is L1, from the above analysis, the conclusion is
as follows (Miller 1981; Zhang et al. 1997).
The system can run normally, that is, service intensity q\1; the waiting time is
less than the longest waiting time, that is Wq T1 ; the wait queue length is smaller
than the longest, that is Lq L1 .
Only the minimum service unit who meet the three requirements will be the
best. Among them, C is unknown and k, l, T1, L1 is known, so P0 is known. The
model of checkout counter can be described by the following model is:
( )
c c
k ðcqÞ q ðcqÞ q
c ¼ min cq ¼ \1; Wq ¼ P0 T1 ; Lq ¼ P 0 L1
cl kc!ð1 qÞ2 c!ð1 qÞ2
ð88:7Þ
The constraint condition is:
8
> k
>
> q ¼ cl \1
>
>
>
>
>
> ðcqÞc q
>
< Wq ¼ P 0 T1
kc!ð1 qÞ2 ð88:8Þ
>
>
>
> ðcqÞc q
>
> L ¼ P L1
>
>
q
c!ð1 qÞ 2 0
>
>
:
k; c; l; T1 ; L1 0
Among them, k, l, T1, L1 are all known and c is unknown. c* is the best
number of service units. After optimization, the average wait time and average
wait queue length in the system respectively are:
ðc qÞc q
Wq ¼ P0 ð88:9Þ
kc !ð1 qÞ2
ðc qÞc q
Lq ¼ P0 ð88:10Þ
c !ð1 qÞ2
88.3 Data Collect and Analysis
88.3.1 Data Collect
According to the survey of customer flow and service time in the supermarket, the
supermarket whose business hour is from 8 to 22, for a total of 14 h, had 50
checkout counters. The research time was various parts of the day in weekend and
weekdays. In each time slot, we conducted random surveys of 200 unit time (each
838 C. Chai
Table 88.1 Arrival rate of customers in each time slot and the number of available checkout
counters
Time interval Arrival rate of customers(people/h) Number of service desks
Weekday Weekend Weekday Weekend
8:00 * 9:00 756 840 15 16
9:00 * 10:00 984 1116 15 16
10:00 * 11:00 1332 1536 22 24
11:00 * 12:00 1524 1560 24 24
12:00 * 13:00 1296 1620 22 25
13:00 * 14:00 1380 1656 22 25
14:00 * 15:00 1524 1884 24 27
15:00 * 16:00 1680 2052 26 30
16:00 * 17:00 1716 1668 26 25
17:00 * 18:00 1668 1908 26 28
18:00 * 19:00 1680 1992 26 30
19:00 * 20:00 1776 1800 27 27
20:00 * 21:00 1980 1524 27 25
21:00 * 22:00 1416 852 22 16
unit time is 5 min), the Table 88.1 below is the statistics of the customer arrival
situation.
88.3.2 Data Analysis
According to the survey, the service time of cashier desk obeys negative expo-
nential distribution with the parameters l (l = 58.32). Discussing the working
time interval from 9:00 to10:00 as an example, according to the data in Table 88.1,
the average service strength of the system for this moment is:
k 984
q¼ ¼ ¼ 1:125 [ 1 ð88:11Þ
nl 15 58:32
This shows that the system has been very crowded at this time, and the cus-
tomers have to wait for a long time to get service. Now the customers must be not
satisfied with the system very much. The reality of the situation is the same, which
should be improved.
88.4 Designing of the Prioritization Scheme
In the system, only the service strength q\1 the system will reach the balance.
88 Problem Analysis and Optimizing of Setting Service Desks in Supermarket 839
Table 88.2 Service index
Number of service q P0 Lq L W Wq
desks
17 0.992 3.8*10-9 119.5 136.4 0.139 0.121
18 0.937 2.35*10-8 10.6 27.47 0.028 0.011
19 0.888 3.41*10-8 4.11 20.98 0.021 0.004
20 0.844 4*10-8 2.01 18.88 0.019 0.004
k 984
When q ¼ nl \1, with 58:32n ¼ 16:87
n \1, got n C 17, this means that at least 17
service desks should be open to achieve the system balance state from 9:00 to
10:00.
Taking 9:00–10:00 as an example, according to the state equation of service
desks negative exponential distribution queuing system, it can work out the
probability of system free, the team length of average waiting and the relationship
expression of service desk number C. And put C = 17 k = 984 l = 58.32, q ¼
0:992 into formula (88.1–88.6), the final solution:
P0 ¼ 3:8 109 ; Lq ¼ 119:5; L ¼ 136:4; Wq ¼ 0:121; W ¼ 0:139:
Bring the parameters into formula (88.7–88.10):
1
c ¼ min cjq\1; Wq ; Lq 4
12
Then got the system service model in the time interval from 9:00 to 10:00.
Compared with different C value, the optimal number of cashier dest c* can be
obtained, and the detailed analysis is as follows:
Then got the following various system service indexes (see Table 88.2) with
different number of service desks from 9:00 to 10:00.
According to the above the data we can clearly see that with the increasing
opening number of cashier desks, the service strength of whole system declined
gradually and the team length also reduce gradually, so that the customer needs the
less time to wait. when the cashier desk opening number reaching 20, a customer
almost don’t have to wait for getting service. In the actual investigation, the
longest waiting time of customers T1 equals 5 min, and put the longest team
length of customers waiting L1(L1 = 4) into formula, the final solution: C = 18.
That means that it is reasonable to open 18 cashier desks, as the service strength
is 0.937, the customer waiting time is 0.011 h, and the average waiting customers
is less than two. It not only makes the customer get fast service, but also saves the
operation cost as much as possible and the cashier service strength is moderate.
Similarly, by the above optimization methods and standards in each time
intervals for optimized design in our workdays and weekend, the optimal service
cashier numbers in each time slot can be obtained, as follows (see Table 88.3).
840 C. Chai
Table 88.3 Optimization number of service desks
Time interval Customer arrival rate(number/h) Number of opening cashier desks
Workday Weekend Workday Weekend
8:00 * 9:00 756 840 14 15
9:00 * 10:00 984 1116 18 20
10:00 * 11:00 1332 1536 24 27
11:00 * 12:00 1524 1560 28 28
12:00 * 13:00 1296 1620 23 29
13:00 * 14:00 1380 1656 24 29
14:00 * 15:00 1524 1884 28 33
15:00 * 16:00 1680 2052 30 36
16:00 * 17:00 1716 1668 31 29
17:00 * 18:00 1668 1908 30 34
18:00 * 19:00 1680 1992 30 35
19:00 * 20:00 1776 1800 32 32
20:00 * 21:00 1980 1524 36 27
21:00 * 22:00 1416 852 25 15
After optimization, the Numbers of service desks in various time interval reach
the optimal, as the customer do not need long queue for leave, and customer
service satisfaction for supermarkets will increase, the loyal customers of super-
market will be more and more customers, so that the income will increase as well.
88.5 Conclusion
This paper utilizes the classic queuing theory to solve the queuing problem and
optimize service strategy. According to different passenger flow volume in dif-
ferent time, the opening number of cashier number should be set in a flexible way,
so as to shorten the waiting time of customers, improve customer satisfaction,
reduce the cost and improve the competitive power of enterprises.
References
Adan I, Resing J (2001) Queuing theory. Department of Mathematics and Computing Science
Eindhoven University of Technology, Eindhoven, p 7
Deng X (2000) The optimal product line desing based on the queuing theory. Oper Res Manage
Sci 9(3):64–69
Hlynka M (2010) An introduction to queuing theory modeling and analysis in applications.
Techno metrics 52:138–139
Huang Z, Xiao YJ (2009) M/M/C/? queuing system model and its application example analysis.
Technol Dev Enterp 11:92–93
Li W, Li M, Hu Y (2000) Operational research. Tsinghua University press, Beijing, pp 310–348
88 Problem Analysis and Optimizing of Setting Service Desks in Supermarket 841
Liu W, Liu Z (2009) Applying queuing theory to quantitative analysis on outpatient billing
service in hospital. Chin Med Equip J 10:87–89
Liu L, Xu J, Zhang T (2011) Statistical analysis of bank service window scheduling strategy
model. Stat Res 28:75–79
Mandelbaum M, Hlynka M (2009) History of queuing theory in Canada prior to 1980. INFOR
11:335–353
Miller DR (1981) Computation of steady-state probability of M/M/1 priority queues. Oper Res
29(5):945–948
Sun B (2007) The model of bank queue and its appliance in improving the bank quality of
services. Hefei University of Technology, Hefei
Takacs L (1962) Introduction to the theory of queues. Oxford University Press, New York, p 161
Wang R, Miao W (2012) Based on the M/M/n the queue theory and large travel scenic spot
internal queuing phenomenon. Reform Econ Syst 3:177–178
Yan W (2012) Problem analysis and system optimization strategy research about bank queuing.
Finance Econ 6:63–65
Zhang ZG, Vickson RG, van Eenige MJA (1997) Optimal two-threshold polices in an M/G/1
queue with two vacation types. Perform Eval 29:63–80
Zheng H, Gu F (2005) Optimization of the queuing system in large supermarket. Chin J Manage
2(2):171–173
Zhou W (2011) Application research of queuing theory model in medical service system.
HuaZhong University of Science and Technology, Wuhan
Chapter 89
Proposed a Novel Group Scheduling
Problem in a Cellular Manufacturing
System
Y. Gholipour-Kanani, N. Aghajani, R. Tavakkoli-Moghaddam
and S. Sadinejad
Abstract This paper presents a new integrated mathematical model for a cellular
manufacturing system and production planning. The aim of this model is to
minimize machine purchasing, intra-cell material handling, cell reconfiguration
and setup costs. The presented model forms the manufacturing cells and deter-
mines the quantity of machines and movements at each period of time that min-
imizes the aforementioned costs. It is so difficult to find an optimal solution in a
reasonable time. Thus, we design and develop a meta-heuristic algorithm based on
a genetic algorithm (GA). This proposed algorithm is evaluated, and the related
results confirm the efficiency and effectiveness of our proposed GA to provide
good solutions, especially for medium and large-sized problems.
Keywords Cellular manufacturing system Genetic algorithm Intra-cell
material handling
Y. Gholipour-Kanani (&)
Department of Management, Islamic Azad University—Qaemshahr Branch,
Qaemshahr, Iran
e-mail: [email protected]
N. Aghajani
Department of Industrial Engineering, Islamic Azad University—Qazvin
Branch, Qazvin, Iran
R. Tavakkoli-Moghaddam
Department of Industrial Engineering, College of Engineering,
University of Tehran, Tehran, Iran
S. Sadinejad
Department of Industrial Engineering, Islamic Azad
University—Research and Science Branch, Tehran, Iran
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 843
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_89,
Springer-Verlag Berlin Heidelberg 2013
844 Y. Gholipour-Kanani et al.
89.1 Introduction
Most of production environments involve the changes in the input parameters,
such as demands over time. In such a case, managing production resources and
balancing them between successive time periods with the aim of minimizing
production costs are known as production planning. Cellular manufacturing sys-
tems (CMSs) are one of the well-known and efficient alternatives for production
environments with high variety and high volume of products. The main goal of
CMSs is to minimize the material handling costs in the shop floor.
Various manufacturing production planning and inventory control problems
have been studied extensively by many production management researchers.
Different models and methods developed to solve these problems can be found in
widely used textbooks of production engineering or manufacturing systems (Riggs
1981; Singh 1996). Inventory control models from the simple EOQ to more
complicated MRP, kanban and CONWIP models (Monden 1983) have been
developed and widely used in today’s manufacturing industries. Some of them are
very successful in practical applications. Prominent manufacturing and features,
such as production flexibility and manufacturing cell formation, are not usually
considered in developing cellular manufacturing production (Chen 1998).
A comprehensive review of the DCMS’s literature can be found in Safaei et al.
(2008). Production planning in the CMSs was discussed in Chen (2001) and Ol-
orunniwo (1996). Schaller et al. (1998) proposed a two-stage approach named CF/
PP for integrating the cell formation and production planning in a cellular
manufacturing system. Chen and Cao (2004) proposed an integrated model for
production planning in a CMS minimizing the inter-cell material handling cost,
fixed charge cost of setting up manufacturing cells, cost of holding the finished
items over the planning horizon, cost of setting up the system to process different
parts in different time periods, and machine operating cost. Defersha and Chen
(2006a) proposed a mathematical model for the design of cellular manufacturing
systems. The model incorporates a dynamic cell configuration, alternative rout-
ings, lot splitting, sequence of operations, multiple units of identical machines,
machine capacity, workload balancing among cells, operation cost, and cost of
subcontracting part processing, tool consumption cost, cell size limits, and
machine adjacency constraints (Defersha and Chen 2006b).
The main contributions of this paper are as follows:
1. The intra cell handling is done in the form of batching and the intra cell
handling cost for each batch of the kind of parts is determinate separately.
2. The set up cost was added to other costs related to cellular manufacturing and
this shows the integration of production planning and cellular manufacturing
system.
3. The intra cell handling constraint is also shown in this model.
89 Proposed a Novel Group Scheduling Problem 845
89.2 Problem Formulation
This section presents a new integrated pure integer linear programming model of
the CMS and PP under following assumptions.
89.2.1 Assumptions
1. The processing time for all operations of a part type is known and deterministic.
2. The capabilities and time-capacity of each machine type are known and con-
stant over the planning horizon.
3. Parts are moved in a batch within cells. Intra cell batch handling cost is known
and constant. It independent on distance.
4. The number of cells is known and constant over the planning horizon.
5. The upper and lower bounds of cell sizes are known and constant.
6. Relocation cost of each machine type from one cell to another between periods
is known. All machine types can be moved to any cell. Relocation cost is the
sum of uninstalling and installing costs. Note that, if a new machine is added to
system, we have only the installation cost. On the other hand, if a machine is
removed from the system, we have only the uninstallation cost.
7. The set up cost for all parts is known.
8. The batch sizes for all parts and in each period are constant.
9. The independent demand of parts is deferent from period to another period.
89.2.2 Descriptions and Symbols
k Time period index
i Part type index
j Index of operations
f Machine index
c Cell index
89.2.3 Parameters
dik Known demand of part type i for time period k
kji Time processing of part i
si Set up cost to produce part type i
Pf Purchase cost of machine f
Df Available capacity of machine f
846 Y. Gholipour-Kanani et al.
rfþ Install cost of machine f
rf Remove cost of machine f
Vi Unit cost to move part type i in batches
LBc Minimum number of machines in cell c
UBc Maximum number of machines in cell c
M Large positive number Batch: batch size
89.2.4 Decision variables
nfck Number of machine type f in cell c during period k
yþ
fck Installed number of machine type f in cell c during period k
y
fck Removal number of machine type f in cell c during period k
1; if one unit of machine type f is placed in cell c, at periodk
rfck ¼
0; otherwise
1; if part type i isprocessed during period k
zik ¼
0; otherwise
1; if operation j of part i to be processed is done in cell cduring period k
Xjick ¼
0; otherwise
1; ifoperation j of part ito be intra cell handled is done in cell cduring period k
bjick ¼
0; otherwise
89.2.5 Mathematical Model
Consider a manufacturing system consisting of a number of machines to process
different part types. Each part type may require some or all of the machines for
processing. In addition, consider the manufacturing system in a number of time
periods k, where k = 1,…,T, with T [ 1. One time period could be a day, a week,
or a month. Demands for different part types are assumed to be known from work
orders or from forecast.
X X X X X X X X
þ þ
Minz ¼ k i
s i z ik þ k c f
ðr f y fck þ r f y fck Þ þ Pn
f f fck
X X X X k c
dik
þ k i j
Vb
c i jick batch
ð89:1Þ
89 Proposed a Novel Group Scheduling Problem 847
X X
nfck nfc;k1 0; 8ðf ; kÞ ð89:2Þ
c c
X
c
Xjick ¼ zik ; 8ðj; i; kÞ ð89:3Þ
X X
i j
dik kji Xjick Df nfck ; 8ðc; kÞ ð89:4Þ
X
LBc f
nfck UBc ; 8ðc; kÞ ð89:5Þ
nfck ¼ nfc;k1 þ yþ
fck yfck ; 8ðf ; c; kÞ ð89:6Þ
Xjick þ Xjþ1;ick bjick 1; 8ðj; i; c; kÞ ð89:7Þ
1; if dik [ 0
zik ¼ ð89:8Þ
0; if dik ¼ 0
rfck nfck ; 8ðf ; c; kÞ ð89:9Þ
nfck M:rfck ; 8ðf ; c; kÞ ð89:10Þ
Xjick ; bjick ; rfck ; zik 2 f0; 1g; 8ðj; i; f ; c; kÞ ð89:11Þ
nfck ; yþ
fck ; yfck 2 f0; 1; 2; . . .g; 8ðf ; c; kÞ ð89:12Þ
A mathematical programming model is developed to solve this cellular man-
ufacturing production planning problem. Owing to the above problem features.
The mathematical programming model becomes a pure integer programming
model. The objective function of this model is to minimize machine purchasing,
intra cell material handling, cell reconfiguration, and set up costs.
Model objective function: The objective function given in Eq. (89.1) comprises
several cost terms. The firs term of the objective function is the machine purchase
cost. The second term of the objective function is the intra cell material handling
cost. The third term is cell reconfiguration cost and the last term in the function is
the set up cost. The minimization of this cost function is subject to certain
conditions.
Capacity limitations of the machines are expressed in Eq. (89.2). Equation
(89.4) implies that the number of type k machines used in any time period is
greater than or equal to that in the previous period. This means that the model is
not going to remove extra machines of any type if that type of machines happens to
be in excess in a certain time period. The presence of extra machines in the system
increases system flexibility and reliability by providing alternative routes during
machine breakdown.
One constraint (89.3) is to ensure that, if operating j of part type i will be
processed in one of the cells in time period k, then the corresponding binary
variable for system set up must be 1. Normally there is an upper limit to the
848 Y. Gholipour-Kanani et al.
number of machines in each cell due to the limit of the physical space. In addi-
tional, there should be at least one machine in each cell; otherwise the cells
disappear. Equation (89.5) specifies the lower and upper bounds of cell sizes.
Equation (89.6) states that the number of type k machines in the current period in a
particular cell is equal to the number of machines in the previous period, adding
the number of machines being moved in, and subtracting the number of machines
being moved out of the cell. Equation (89.7) specifies the intra cell material
handling. Equation (89.8) specifies the corresponding binary variable for system
set up. Equations (89.9) and (89.10) set the value of equal to 1 if at least one unit of
type k machine is placed in cell l during period t or 0 otherwise. Equations (89.11)
and (89.12) are integrality constraint.
89.3 Genetic Algorithm Implementation
The genetic algorithm (GA) is a population-based algorithm that uses analogies to
natural, biological, and genetic concepts including chromosome, mutation,
crossover, and natural selection. Basically, it consists of making a population of
solutions evolve by mutation and reproduction processes. The best fitted solutions
of the population survive while the worse fitted are replaced. After a large number
of generations, it is expected that the final population will be composed of highly
adaptable individuals, or in an optimization application, high-quality solutions of
the problem at hand. The basic steps of a canonical GA are as follows (Tavakkoli-
Moghaddam et al. 2008).
Step 1. Initialize the population and enter Step 2.
Step 2. Select individuals for recombination and enter Step 3.
Step 3. Recombine individuals generating new ones and enter Step 4.
Step 4. Mutate the new individuals and enter Step 5.
Step 5. If the stopping criterion is satisfied, STOP; otherwise, replace old
individuals with the new ones restructure the population tree and return to Step 2.
89.3.1 Solution Representation
The first step in the proposed GA is to consider a chromosome representation or
solution structure. We use a presented structure in Fig. 89.1 to represent the
solution of the extended model. The chromosome representation in this study
represents each job in the schedule as a gene in a chromosome; in which each
chromosome consists of ((k ? c) 9 f) genes. An example to depict this definition
is provided in Fig. 89.1 (for 2 period, 3 cells and 8 machines).
89 Proposed a Novel Group Scheduling Problem 849
Fig. 89.1 Chromosome
Cell 1 1 0 0 7 8
encoding
Period 1 Cell 2 0 2 3 0 0
Cell 3 6 5 4 0 0
Cell 1 0 3 4 8 0
Period 2 Cell 2 5 0 1 7 0
Cell 3 0 6 0 0 2
89.3.2 Create Population
This procedure creates the initial population (Pop), which must be a wide set
consisting of disperse and good solutions. Several strategies can be applied to get a
population with these properties. The solutions to be included in the population
can be created, for instance, by using a random procedure to achieve a certain level
of diversity. In this study, an initial population of the desired size is generated
randomly. For example, when there are five parts, the algorithm generates 10
solutions randomly, depending on the problem size.
89.3.3 Fitness
Each solution has a fitness function value, which is related to the objective
function value (OFV) of the solution. However, the population can have feasible
and infeasible solutions. An option to manage the infeasibility is to use both cost
and feasibility. This can be written as fitness cost feasibility; where s is the
solution, cost of the objective function value. Feasibility is equals to 1 if the
solution is feasible; otherwise it is zero. Therefore, the fitness is not one value;
however it is two, namely the cost and the feasibility of the solution.
89.3.4 Parent Selection Strategy
The parent selection is important in regulating the bias in the reproduction process.
The parent selection strategy means how to choose chromosomes in the current
population that will create offspring for the next generation. Generally, it is better
that the best solutions in the current generation have more chance to be selected as
parents in order to create offspring. The most common method for the selection
mechanism is the ‘‘roulette wheel’’ sampling, in which each chromosome is
assigned a slice of a circular roulette wheel and the size of the slice is proportional
to the chromosome’s fitness.
850 Y. Gholipour-Kanani et al.
Fig. 89.2 Create new solution
89.3.5 Crossover Operator
The ‘‘Improve Solutions’’ method is applied every solution, S, generated by the
combination method. This method aims at restoring the feasibility of solutions
from the diversification method and enhancing these solutions and those solutions
obtained from the combination method. The solutions to be combine and the
crossing points are to be selected randomly. In this study, we use single point
crossover. For example consider a problem with 10 parts and 3 cells. To create a
new solution, we exchange element positioning at the right hand of the cut point in
a solution. Figure 89.2, shows a typical example to create a new solution.
89.3.6 Mutation Operators
The main task of the mutation operator is to maintain the diversity of the popu-
lation in the successive generations and to exploit the solution space. In this paper,
a mutation operator, called Swap Mutation, consist of swapping any two randomly
chosen genes in a chromosome (Torabi et al. 2006). At first, we define ‘‘mutation
strength’’, demonstrator of the maximum number of swap moves performed. If the
strength of the mutation is chosen to be one, then it performs a single swap move,
provided a given probability P(M). So the strength of the mutation shows the
number of consecutive swaps on the individual chromosome.
89 Proposed a Novel Group Scheduling Problem 851
89.4 Computational Results
Delphi7 program was used for designing algorithm Genetic. The executer pro-
cessor of algorithm Genetic and Lingo is a computer with characteristic of
1.8 GHz and 768 MB. The calculation of optimal value especially in large
dimensions is difficult because CMS planning model solution is complicated. So,
the answers of Lingo8 software are an answer near to optimal. The answers are
shown in Table 89.1. We compare the obtained objective value from genetic
algorithm and Lingo software methods in small dimensions problems. We deter-
mine the difference percentage from Lingo answer and study memory size and
CPU time. We can see the results of the problem solution by genetic algorithm and
lingo software are the same in small dimension problems. So, this thing shows the
efficiency of algorithm. Lingo software does not be able to solve the large prob-
lems in acceptable time that genetic algorithm make optimised answer or near it in
most proper time. The result of some test problem is shown in Table 89.1 and also
Growth of solution time for genetic algorithm and lingo software are compared in
Fig 89.3.
Table 89.1 Comparison of LINGO and GA algorithm solutions
Problem Integer Constraints Solution OFV CPU time Memory Gap
variables no. no. method (s) size (%)
1 176 220 GA 530695 197 201 0
LINGO 530695 1 35
2 176 220 GA 775355 203 201 0
LINGO 775355 1 36
3 264 280 GA 939142.5 199 201 0
LINGO 939142.5 4 37
4 936 1468 GA 1913705 220 201 2
LINGO 1876115 22 44
5 1872 2487 GA 3550545 231 201 5
LINGO 3038832 362 53
6 1856 2601 GA 2396190 236 201 7
LINGO 2229278 297 59
7 2088 3623 GA 2895435 250 201 2
LINGO 2836348 66 56
8 2784 4025 GA 4135362.5 266 201 10
LINGO8 3769162 504 60
9 2784 4833 GA 3654797.5 251 201 10
LINGO8 323348 2989 60
10 3328 4462 GA 4284182.5 253 201 6
LINGO8 4045608 15151 69
852 Y. Gholipour-Kanani et al.
89.5 Conclusion and Future Extensions
The model considered in this article is the model for minimizing machine pur-
chasing, intra cell material handling, cell reconfiguration, and set up costs.
According to the researches done, this problem is the type of NP-hardness that is it
is solution with the optimisation software will be impossible if dimension of the
problem increases. The approaches such as branch & bound and dynamic planning
have computational time limitation and saving limitation in company. So, using of
heuristic algorithm would be effective. The results obtained are as follows:
• With the extension of problems, the computational time will be increase by
Lingo while this increase would be little in comparison with Lingo in case of
using heuristic algorithm.
• Variation of the productions increases and industry moves toward using of
cellular manufacturing for using of its benefits. So using of usual methods in
planning of variation of the productions increases and industry moves toward
using of cellular manufacturing for using of its benefits. So using of usual
methods in planning of cellular manufacturing systems doesn’t have good
performance and it should be pay attention to the heuristic methods.
Followings are some suggestion for the future research.
• Some of the parameters of this problem can be considered in the fuzzy and
converted to the fuzzy cellular manufacturing systems.
• Multiple routes do not be considered in this problem considering multiple routes
can make closer the problem to the real condition. So that investigation can be
valuable.
• Inventory cost does not be considered in the article. That can be considered in
the future article.
Fig. 89.3 Diagram of 16000
computational time respect to Genetic Algorithm
14000
the problem
Lingo
12000
10000
time
8000
6000
4000
2000
0
1 2 3 4 5 6 7 8 9 10
example
89 Proposed a Novel Group Scheduling Problem 853
References
Chen M (1998) A mathematical programming model for system reconfiguration in a dynamic
cellular manufacturing environment. Ann Oper Res 77:109–128
Chen M (2001) A model for integrated production planning in cellular manufacturing systems.
Integr Manuf Syst 12:275–284
Chen M, Cao D (2004) Coordinating production planning in cellular manufacturing environment
using Tabu search. Comput Ind Eng 46:571–588
Defersha F-M, Chen M (2006a) Machine cell formation using a mathematical model and a
genetic-algorithm-based heuristic. Int J Prod Res 44(12):2421–2444
Defersha FM, Chen M (2006b) A comprehensive mathematical model for the design of cellular
manufacturing systems. Int J Prod Econ 103:767–783
Monden Y (1983) Toyota production systems. Industrial engineering and management press,
Norcross
Olorunniwo F-O (1996) Changes in production planning and control systems with implemen-
tation of cellular manufacturing. Prod Inventory Manag 37:65–70
Riggs J (1981) Production system: Planning analysis and control. Wiley, New York
Safaei N, Saidi-Mehrabad M, Jabal-Ameli M-S (2008) A hybrid simulated annealing for solving
an extended model of dynamic cellular manufacturing system. Eur J Oper Res
185(20):563–592
Schaller J-E, Erengüç SS, Vakharia A-J (1998) A methodology for integrating cell formation and
production planning in cellular manufacturing. Ann Oper Res 77:1–21
Singh N (1996) System approach to computer integrated design and manufacturing. Wiley, New
York
Tavakkoli-Moghaddam R, Gholipour-Kanani Y, Cheraghalizadeh R (2008) A genetic and
memetic algorithm approach to sequencing and scheduling of cellular manufacturing systems.
Int J Manag Sci Eng Manag 3(2):119–130
Torabi SA, Fatemi Ghomi SMT, Karimi B (2006) A hybrid genetic algorithm for the finite
horizon economic lot and delivery scheduling in supply chains. Eur J Oper Res 173:173–189
Chapter 90
Regional Eco-Environment Optimization
Based on Multiple Parallel Sub-Systems’
Efficiency
Li Wang and Ning Li
Abstract With the rapid growth of economy, China evidence sharp increase on its
GDP. But at the same time it feels more and more pressure from industry wastes in its
environment. To relieve the pressure on environment while still maintaining the
sustainable development in China, decision makers starts to focus more on mea-
suring the efficiency of waste treatment process. Because the industry wastes are
divided into three classes (i.e. waste water, waste gas and solid wastes), we should
apply different treatments to deal with them. In this paper, we propose a multiple
parallel DEA methodology and apply it to calculate the efficiency for the treatment
of these three kinds of wastes. By formulating the three types of treatments as three
parallel sub-systems in ecological environment optimization, the efficiency of
overall, as well as of individual wastes’ treatment can be calculated. The statistic
data from 30 individual provinces of China in 2010 are used to demonstrate the
effectiveness of our approach. Suggestions on optimizing the ecological environ-
ment in different regions based on our measurement are given at the end of the paper.
Keywords Ecological regions Overall efficiency Parallel DEA Sub-system
efficiency Treatments of wastes
90.1 Introduction
In the past, China relies mainly on the former Soviet mode to fuel its own economy
development, which focuses on increasing the inputs, especially labor and capital
investment. In that process, we scarified our environment and limited resources for
L. Wang (&)
School of Management, Changchun Institute of Technology, Changchun, China
e-mail: [email protected]
N. Li
School of Economics and Management, China University of Petroleum, Qingdao, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 855
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_90,
Springer-Verlag Berlin Heidelberg 2013
856 L. Wang and N. Li
the economy growth. With the influence of global greenhouse effect and serious
pollution, decision makers begin to shift from traditional production mode and lay
more emphasis on the treatments of wastes. Generally, we can classify the wastes
into three types, i.e. waste gas, waste water and solid wastes. The treatments of
three types of wastes are pivotal measures to build environmental-friendly regions.
Clarke et al. (1991) discussed water quality management issues in Oregon, USA
and proposed constructive measures to enhance the capability of waste water’s
treatment. At the same time, other two types of wastes, viz. waste gas and solid
wastes also play important roles in ecological environment. Guan et al. (2011)
proposed a coordination of Energy-Economy-Environment System to express the
close relationship between energy, economy and environment.
The evaluation of wastes’ treatment should be applied to identify the develop-
ment level of ecological optimization (Wu et al. 2005). Murtaugh (1996) proposed a
statistical methodology with ecological indicators. However, the treatment process
of waste gas, waste water and solid wastes can be modeled as parallel system with
almost no interaction among them. At the same time, the three processes cover all
aspects of wastes’ treatment. In this paper, we applied parallel DEA model to
calculate the efficiency of each individual treatment and the overall efficiency for
whole region.
The rest of this paper is organized as follows. In Sect. 90.2, the parallel DEA
models are introduced. In Sect. 90.3, we identified the indicators for efficiency
calculation and illustrate the collection of the corresponding data. The calculation
results are presented in Sect. 90.4.
90.2 Parallel DEA Methodology
DEA model CCR was proposed by (Charnes et al. 1978), which applied an optimal
linear programming formula to calculate efficiency of DMUs. Suppose we have n
DMUs, and that kth DMUk (k = 1, 2, …, n) has m inputs, denoted as xik (i = 1, 2,
…, m), and s outputs, denoted as yrk (r = 1, 2, …, s). The traditional CCR DEA
model can be expressed by the following formula (90.1).
X
s
Ek ¼ max ur yrk
r¼1
8X m
>
> vi xik ¼ 1
>
>
>
>
>
>
i¼1
>
> ð90:1Þ
<X s Xm
s:t: u r y rj vi xij 0; j ¼ 1; . . .; n
>
> r¼1 i¼1
>
>
>
>
>
> ur ; v i e
>
>
:
r ¼ 1; . . .; s; i ¼ 1; . . .; m
90 Regional Eco-Environment Optimization 857
Fig. 90.1 Parallel structure
for DMU
Sub-system 1
Sub-system 2
Sub-system q
By calculating with DEA models, the optimal weights can be allocated for each
DMU, denoted as vi ¼ ðv1j ; v2j ; . . .; vmj Þ, ur ¼ ðu1j ; u2j ; . . .; usj Þ, which guarantee
the kth DMU with the maximum efficiency value. If the objection of model (90.1)
equals to 1, then the DMU is denoted as DEA efficient DMU. If the objection of
model (90.1) is less than 1, then the DMU is denoted as DEA inefficient DMU.
DEA models have obvious advantages in measure the performance of multiple
inputs and outputs system. However, traditional DEA model take system as a black
box and ignores the internal structure of system.
In general, the inside of DMU can be classified into different structures and the
internal structure can affect the overall efficiency of whole system. For each of
sub-systems, its efficiency has strong impact on the system’s overall efficiency. In
this paper, we will use the DEA model to deal with parallel sub-system structures.
To overcome the shortcomings of traditional DEA models, Kao (2009) pro-
posed parallel DEA model for measure the relationship between sub-systems and
DUM. Figure 90.1 shows the diagram for ‘‘Parallel structure’’ DEA model for a
DMU.
For the kth DMU, there are q sub-systems and each of sub-system has the same
number and types of inputs and outputs. The q sub-systems are denoted as sub-
system 1, sub-system 2,…, sub-system q. we use Xikp and Yrkp to express the ith
input and rth output, respectively, of the pth sub-system. The relative inefficiency
of a set of n DMUs, each has q parallel sub-systems, can be calculated by
following formula:
858 L. Wang and N. Li
X
q
min spk
p¼1
8 m
> X
>
> vi Xik ¼ 1
>
>
>
>
>
>
i¼1
>
>
>
> Xs X m
>
> p
ur Yrk vi Xikp þ spk ¼ 0
>
>
>
> r¼1 i¼1 ð90:2Þ
>
<
s:t: X X
s m
p
>
> u Y
r rj vi Xijp 0
>
>
>
> r¼1 i¼1
>
>
>
> ur ; vi e; p ¼ 1; 2; . . .; q;
>
>
>
>
>
>
> j ¼ 1; . . .; n; j 6¼ k
>
>
>
: r ¼ 1; . . .; s; i ¼ 1; . . .; m
The model (90.2) above should be calculated for n times to obtained the
inefficiency slacks of systems as well as their sub-systems.P However, the ineffi-
ciency slacks is not equal to inefficiency scores because m v X
i¼1 i ik
w
is not equal to
1 for kth DMU with wth sub-systems.
Pm Therefore, the inefficiency score calculated
by swk should be divided by w
i¼1 vi Xik . Hence, the final efficiency score is
w
1 sk Pm vi X w .
i¼1 ik
90.3 Structure Analysis of Eco-Environment Optimization
90.3.1 Development of Eco-Environment Optimization
in China
Within the framework of ‘‘Society-Nature-Environment’’, these three factors are
closely related and influence each other. During 1978–2000, China’s economy
development is largely resources based, in which the development relies on
enlarging both inputs and outputs. Although GDP of China increased rapidly, we
also got the penalty from eco-environment. The pollutions of waste water, waste
gas and solid wastes obviously affect eco-environment in China.
Since 2000s, Chinese government realized the problem and began to shift from
previous resource based development mode into environmental friendly devel-
opment mode. The government increased its investment in treating environmental
pollution (i.e. waste water, waste gas, solid wastes). Figure 90.2 shows the trend
for government’s investment on protecting eco-environment (Liu et al. 2005).
As Fig. 90.2 shows, the total investment amount of in treatment of environmental
pollution and the investment in urban environment infrastructure facilities from
90 Regional Eco-Environment Optimization 859
Fig. 90.2 2000–2010 the investment in eco-environment
Chinese government is increasing year by year (Guo et al. 2007). The investment in
the environment infrastructure not only enhances the capability to treat environment
pollutions for now but also for future. It is clear that Chinese government is putting in
more and more resources and effort on optimize eco-environment. In order to
effectively optimize the eco-environment, we need to know what the efficiency of
treatment pollution in China is. As we are evaluating the efficiency of treatment in
eco-environment, we need analyze the structure of wastes’ treatment.
90.3.2 Structure and Indexes of Wastes’ Treatments
In general, the wastes are divided into three types: (a) waste gas, waste water and
solid wastes. To optimize the eco-environment, we should also apply corre-
sponding treatment measures for these three types of wastes. In our model, we
divided the optimization of eco-environment into three parallel processes, i.e.
waste gas treatment, waste water treatment and solid wastes’ treatment. If we
represent each waste treatment as a sub-system, there are multiple indexes, which
can be listed to measure the efficiency of each process of waste treatments in the
view of multiple inputs and outputs. The indexes are shown in Table 90.1
(Bao et al. 2006).
For waste gas treatment, we use 1-in-4-out indexes to interpret the sub-system’s
efficiency. For waste water treatment, we design 2-in-2-out to explain the effi-
ciency of sub-system. For solid wastes’ treatment, we apply 1-in-3-out to measure
the sub-system’s efficiency.
860 L. Wang and N. Li
Table 90.1 Indexes for each of waste treatments
Treatment Indexes
Inputs Outputs
Waste gas Total volume of industrial Number of facilities for treatment
treatment waste gas emission of waste gas (set)
(100 million cu.m) Volume of industry Sulphur
Dioxide removed (10,000 tons)
Volume of industrial Soot
removed (10,000 tons)
Volume of industrial dust
removed (10,000 tons)
Waste water Total volume of waste water Number of facilities for treatment
treatment discharge in industry of waste water (set)
(10,000 tons)
Consumption waste water Industrial waste water meeting
discharge (10,000 tons) discharge standards (10,000 tons)
Solid wastes Volume of industrial Volume of industrial solid wastes
treatment solid wastes produced utilized (10,000 tons)
(10,000 tons) Volume of industrial solid wastes
in stocks (10,000 tons)
Volume of industrial solid
wastes treated (10,000 tons)
90.4 Calculation and Results
Based on those indexes listed on Table 90.1, we collect 30 provinces’ corre-
sponding statistic data from ‘‘China Statistic Year Book 2011’’. With 30 provinces
as DMUs and three treatments as sub-systems, the CCR DEA efficiency and
overall efficiency for each system are calculated. The results are shown in
Table 90.2. The third column in Table 90.2 shows the CCR efficiency scores for
all 30 provinces using traditional CCR DEA model. Since the CCR efficiency
score for all regions are equal to 1, we can’t differentiate the performance of eco-
environment optimization in each of regions. CCR model is not effective for
evaluating those regions eco-environment optimization level. The first and second
columns in Table 90.2 show the inefficiency and efficiency score using our 3-way
parallel structure DEA model. Our model is able to give different efficiency/
inefficiency scores for different regions. Therefore, our model is demonstrated as
an effective way to measure the performance of eco-environment development
level for each province.
Fourteen regions, 47 % of 30, reach 1 for the overall efficiency value. Those
regions perform well in the eco-environmental optimization. Among these efficient
regions, Beijing, Tianjin and Zhejiang are advanced developed regions with large
amount of inputs in the treatments process (Xu and Tang 2005), i.e. emission of
waste gas, discharged waste water and produced solid wastes. The main reason for
their high score is the capability of treatments for all three types of wastes.
90 Regional Eco-Environment Optimization 861
Table 90.2 Efficiencies of 30 Regions in Wastes’ Treatments
Regions Inefficiency score Efficiency score CCR efficiency
Beijing 0 1 1
Tianjin 0 1 1
Hebei 0.0029 0.9971 1
Shanxi 0.0389 0.9611 1
Inner Mongolia 0.0888 0.9112 1
Liaoning 0.0704 0.9296 1
Jilin 0 1 1
Heilongjiang 0 1 1
Shanghai 0.0107 0.9893 1
Jiangsu 0.0079 0.9921 1
Zhejiang 0 1 1
Anhui 0.0130 0.987 1
Fujian 0 1 1
Jiangxi 0 1 1
Shandong 0.0038 0.9962 1
Henan 0.0057 0.9943 1
Hubei 0.0270 0.973 1
Hunan 0.0350 0.965 1
Guangdong 0.0501 0.9499 1
Guangxi 0.0110 0.989 1
Hainan 0 1 1
Chongqing 0.0473 0.9527 1
Sichuan 0.0257 0.9743 1
Guizhou 0 1 1
Yunnan 0.0706 0.9294 1
Shaanxi 0 1 1
Gansu 0 1 1
Qinghai 0 1 1
Ningxia 0 1 1
Xinjiang 0 1 1
Therefore, the three regions have the characteristics of large inputs and lager
outputs.
For Jilin, Heilongjiang, Fujian and Hainan, they are efficient regions too
(efficiency score = 1). Those regions are middle developed regions. Heilongjiang
and Jilin locate in the northeast part of China. Although these regions are industry
basement in 1980s, the center of industry development has transferred into coastal
regions. Therefore, the transformation relieved the pressure of eco-environment in
those regions. Fujian and Hainan are coastal provinces, who are not industry
centers or basements. Therefore, the pollution in Fujian and Hainan are relatively
less than other coastal regions.
The other 7 efficient regions are Jiangxi, Guizhou, Shaanxi, Gansu, Qinghai,
Ningxia and Xinjiang, which are located on the west part of China. Those regions’
development of industry lagged behind other eastern regions.
862 L. Wang and N. Li
Fig. 90.3 Eco-environmental optimization efficient regions’ map
Figure 90.3 shows the efficient regions on the map of China. The green
provinces in the map are the efficient regions with high performance on eco-
environmental optimization.
In Table 90.2, there are 16 regions whose efficiency scores are less than 1.
To optimize eco-environment and keep sustainable development mode in China,
we should empower the treatment capability in the next few years. At the same
time, we notice that the average efficient values are more than 0.9. The means the
gaps between different regions in eco-environmental optimization are small.
Therefore, it is quite feasible to optimize the overall eco-environment in China.
90.5 Conclusions
Over the past 30 years, China enjoy economy booming at the expense of envi-
ronment pollution. How to enhance the capability of deal with those wastes should
be important measures to make our environment friendly. Now, Chinese gov-
ernment recognized the importance of protecting eco-environment and invests
heavily on improving environment. To quantify the results of eco-environmental
optimization, comprehensive evaluation method should be applied to measure the
90 Regional Eco-Environment Optimization 863
efficiency accurately. In this work, we propose the parallel DEA model and apply
it to analyze the eco-environment efficiency for 30 individual provinces of China.
Our results demonstrate that with our model, the government can get accurate eco-
environmental optimization levels for those 30 regions and make corresponding
measures to enhance optimization capability of eco-environment in China.
Acknowledgments The main work of this paper is supported and sponsored by Young Foun-
dation of Ministry of education, humanities and social science research projects (11YJC630100),
project of Shandong Economic and Information Technology Committee (No. 2012EI107) and the
Fundamental Research Funds for the Central Universities (11CX04031B).
References
Bao C, Fang C, Chen F (2006) Mutual optimization of water utilization structure and industrial
structure in arid inland river basins of Northwest China. J Geog Sci 16(1):87–98
Clarke SE, White D, Schaedel AL (1991) Oregon, USA, ecological regions and subregions for
water quality management. Environ Manage 15(6):847–856
Charnes A, Cooper WW, Rhodes E (1978) Measuring the efficiency of decision making units. Eur
J Oper Res 2:429–444
Guan H, Zou S, Zhou X, Huang Z (2011) Research on coordinated evaluation of regional energy-
economy-environment system. Commun Comput Inf Sci 225(2):593–599
Guo R, Miao C, Li X, Chen D (2007) Eco-spatial structure of urban agglomeration. Chin Geogr
Sci 17(1):28–33
Kao C (2009) Efficiency measurement for parallel production systems. Eur J Oper Res
196:1107–1112
Liu Y, Li R, Li C (2005) Scenarios simulation of coupling system between urbanization and eco-
environment in Jiangsu province based on system dynamics model. Chin Geogr Sci
15(3):219–226
Murtaugh PA (1996) The statistical evaluation of ecological indicators. Ecol Appl 6(1):132–139
Wu K, Hu S, Sun S (2005) Application of fuzzy optimization model in ecological security pre-
warning. Chin Geogr Sci 15(1):29–33
Xu Y, Tang Q (2005) Land use optimization at small waterhed scale on the Loess Plateau. J Geog
Sci 19(5):577–586
Chapter 91
Research on Brand Strategy to Small
and Medium-Sized Enterprises
Xin-zhu Li
Abstract Based on the analysis of brand attributes, from the perspective of value
chain theory, this paper presents that brand strategy is a significant strategy for the
small and medium-sized enterprises (SMEs) to realize higher additional value of
products and gain competitive advantage in market. Formation of scientific brand
development strategy planning, clear definition to core brand value, cultivation of
self-owned brand, occupation of competitive advantage by correct brand posi-
tioning, selection of correct brand appeal, and adoption of innovative brand
operational model are the important approaches and means for SMEs to realize
brand strategy, extricate from operational predicament, and promote additional
value of products.
Keywords Brand strategy Small and medium-sized enterprise (SME) Smiling
curve Value chain
91.1 Introduction
Brand, as the symbol and identification to the enterprise and its product and
service, delivers specific information to consumers. As an important link in the
value chain, brand plays a decisive role in promotion of the whole value. Favor-
able brand innovation strategy is a powerful force to increase the additional value
of product and service, and to contribute to enhance competitive advantage and
cultivate core competitive competence.
X. Li (&)
Economics and Management School, Wuhan University, Hubei, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 865
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_91,
Ó Springer-Verlag Berlin Heidelberg 2013
866 X. Li
91.2 Relevant Theories Review on Brand and Value Chain
Philip Kotler, an expert in modern marketing, defines brand as a name, mark,
design or their combination, which is used to distinguish the products and even the
enterprise from other competitors. For the enterprise, brand stands for potential
competitiveness and profitability. For consumers, it is a warranty of quality and
credit, which can reduce purchasing cost and risk, and eliminate information
asymmetry. Brand is an important intangible asset for the enterprise and has direct
relation with its competitive power.
In the ‘‘value chain’’ theory proposed by Michael Porter, a professor of Harvard
University, it is supposed that value creation can be realized in all operating
activities, which include basic activities and supportive activities. The former one
mainly refers to the production and sale of products, with five elements involved,
i.e. internal logistics, production and operation, external logistics, marketing and
after-sale service; and the latter one refers to those supportive basic activities
inside the enterprise with four elements involved, i.e. enterprise infrastructure,
human resource management, technological development and purchasing. All
links of value chain are interactive and interrelated, and they integrally work to
create corporate value (Porter 2005) (Fig. 91.1).
Extending the value chain of enterprise to the whole industrial chain, Stan Shih,
with his ‘‘smiling curve’’ theory proposed in Rebuilding of Acer in 1992, regards
that, in the smiling curve, the upward ends indicate higher additional value at both
ends in the industrial chain, i.e., design and sales, and while the lowest point in the
middle part indicates that the lowest additional value happened at the intermediate
link, i.e., manufacture. In the future, the industry shall focus the development at
both ends to strengthen research, development and design on the left side to gain
Enterprise infrastructure (finance, plan etc)
Margin
Human resource management
Supportive
Activities Research and development
Purchase
Internal Production External Sales After-sales
logistics and Logistics
Margin
Operation
Basic activities
Fig. 91.1 Model of value chain
91 Research on Brand Strategy to Small and Medium-Sized Enterprises 867
Fig. 91.2 Model of smiling
curve
Additional value
Intellectual Brand/service
property
Research,
development Manufacture Marketing
and design
Direction of value chain
competitive advantage by intellectual property and to strengthen customer-
oriented brand and service to dominate the market with sales (Shih 2005)
(Fig. 91.2).
91.3 Attributes of Brand
The attributes of brand can be divided into material attributes and social attributes.
The material attribute embodies use value of commodity, which belongs to its
essential attributes and exists before purchased, reflecting the relationship between
human being and commodity. For example, Jetta and Benz both represent vehicle
in respect to their material attribute.
The social attribute embodies symbol value of commodity, which belongs to
social derived attributes and is not shown until purchased and used, reflecting the
relationship between human beings and the commodity. For example, Jetta is only
regarded as a convenient and fast vehicle, while Benz is the symbol of nobleness,
success and social status.
91.4 Functions of Brand
91.4.1 Embodiment of Marketing Signal
The essential function of brand is to distinguish commodity and eliminate pur-
chasing information asymmetry as a marketing signal. It embodies the quality of
commodity and creates discrepant value for customers. According to ‘‘Lemon
Market’’ theory, a commodity without brand locates in the middle or lower end of
consumers’ expected price, and premium only occurs at commodity with a
favorable brand.
Brand strategy is not necessary for all enterprises or commodities, which is the
case for commodities with premium inferior to promotion cost, commodities
inferior to that at consumers’ expected price, commodities with unnoticeable
868 X. Li
quality difference (such as river sand used for building), and commodities inde-
pendently priced for monopoly (such as water, power, and coal gas) (Aker 1990).
91.4.2 Tool for Independence in Market
Brand not only refers to a name or signal, but also represents many-sided com-
mitment an enterprise make and a major communication channel between the
enterprise and consumers. The brand with good image reflects consumers’ trust to
the enterprise.
For enterprises without brands, they lose an opportunity to earn consumers’
trust as well as to demonstrate their strength. Owning to lack of trust, many SME’s
may become scapegoat for dominant large enterprises in the value chain and be
confronted with enormous market risks. Without the support of brand, it is difficult
for SMEs to directly display their competitive advantage, losing many commu-
nication opportunities (Bhat and Reddy 2001). As a result, SMEs can only attach
themselves to the lower end of the value chain and hardly obtain independence in
market only with small margin.
91.4.3 Warranty for Long-Term Development
For SMEs that are satisfied with a small margin in the manufacture value chain and
even without trademark, as their competitive edge relying on low-cost labor and
resource is shrinking, the original extensive operation model cannot guarantee
long-term development. Therefore, to increase additional value, transform oper-
ation model and pursue long-term development, brand strategy is a practical
choice for enterprises to expand the market, get rid of price competition and
enhance competitiveness.
91.5 Implementation Strategy of Brand
91.5.1 Formation of Scientific Brand Development Plan
In the process of SMEs transition from low end to high end of value chain, brand
building shall be recognized as a systematic strategy project featured by integrity,
constancy and total involvement, and shall be regarded as the core component of
SMEs development strategy. All operation activities shall be designed, launched,
maintained, managed, guided, and coordinated by centering on the brand, to
enhance brand equity through long-term and dedicated work.
91 Research on Brand Strategy to Small and Medium-Sized Enterprises 869
91.5.2 Clear Definition of Core Brand Value
Brand embodies the relationship between the enterprise and consumers. By offering
unique values demanded by consumers, brand is applied to establish firm relation-
ship with consumers, rather than to endow a good name or earn popularity in a short
time at a high promotion cost. Core brand value shall be defined based on demands of
target consumers, in addition to correct perception of the brand (Smith 2001).
Many enterprises, for lack of correct cognition of the brand, equate brand
building and advertising, and believe that well-known brand and even strong brand
can be built in a short period through advertisement. Many such enterprises as Qin
Chi, Sanzhu and Aidor, once pursued for popularity by overspreading advertise-
ment; however, these brands piled up by high advertising costs have vanished in the
market for a long time. The lessons they leave to later generations are very profound.
91.5.3 Clear Brand Positioning
As important idea of market competition, the concept of positioning proposed by
Al Ries and Jack Trout has been recognized by the market (Ries and Trout 2004).
Brand is the image to the enterprise just like the image to a person. Correct brand
image is just the image the enterprise reveals in front of consumers, so clear brand
positioning is required based on the consumers’ requirements. By brand posi-
tioning the core value and competitive advantage of the brand shall be demon-
strated, to help consumers to seize intention of enterprise brand and aware the
special interest the brand offers.
For SMEs with limited resources, brand positioning contributes to pursuit for
competitive edge and focus on meeting specific consumption demand, so as to
improve operation efficiency of the brand. For ‘‘concentration, precision and
uniqueness’’ are the main sources of competitive advantage for SMEs, brand
positioning not only serves for definition of own brand target, but also works to
accurately convey their competitive edge to target consumers (Shao 2005). For
example, ‘‘Fotile’’ has always been devoted to providing high-quality cooking
utensil to ‘‘make better feeling of home’’, and improving product design and brand
building, and thus acquires consumers’ acceptance.
91.5.4 Correct Choice of Brand Appeal
For products of industrial equipment which face specific users, they are the raw
materials, accessories or production means for users, so the material attributes are
emphasized in brand promotion with focuses laid on safety, quality, practical and
other use value, and the brand appeal is usually to create value for users.
870 X. Li
To ordinary consumer goods, the social attributes embody the symbolic value
which is more appealing to the consumers and reflects the relationship among
human beings. For example, the brand promotion for food and beverage emphasize
on cheerfulness, exercises, and vitality, that for high-end automobiles emphasize
on dignity and elegance, and that for telecommunications and home appliances
emphasize on harmony, family love and convenience (Christensen 2010).
Due to the restrictions of various objective situations, there are many practical
difficulties for many enterprises to maintain long-term leading edge in technology
and quality, thus homogenization of product function is basically inevitable. In
order to differentiate their products from competitors’, provide differentiated
product value and obtain consumer’s sustained favor, the emotional communica-
tion with consumers appears very important (David 1991).
However, many enterprises are often restricted to attract consumers by func-
tional benefits of the brand while ignoring the expression of emotional benefits.
Purely functional benefits appeal is likely to make the brand in dilemma of
homogenization competition. In order to avoid the price competition brought about
thereby, the consumers’ satisfaction and loyalty to the band can be promoted
through emotional communication.
91.5.5 Unification of Brand Image
Professor Don Schultz deems that the investment philosophy of brand building
shall transfer from media-oriented model to one focusing on brand connection or
brand contact points (Schultz and Schultz 2005). While the problem many
enterprises encounter during the brand building is the confusion of brand images
and lack of unified brand image in the mind of consumers, which seriously affects
consumers’ cognition of the brand.
In fact, brand building is a systematic project. The enterprises shall start from
the research on consumer behavior to find out the contact point of the brand and
deliver a consistent brand message and create a unified brand image through
effective management of brand contact point. By studying the media contact habits
of the target customers, the enterprises can choose the specific approach for brand
communication and improve communication efficiency by precise work. Espe-
cially for SMEs lacking of funds, effective management of brand contact points to
deliver a unified brand image is an important means to reduce the brand promotion
cost and improve efficiency.
91.5.6 Innovative Brand Operational Model
Innovation is an important guarantee of the sustainable development of the brand.
Featured by flexible organization mechanism and market adaptability, SMEs shall
91 Research on Brand Strategy to Small and Medium-Sized Enterprises 871
adjust measures to local conditions and actively take innovative operation mode
and sales mode of the brand in the process of brand building to earn their own
competitive advantages (Porter 2002).
Many former small enterprises stand out and grow by taking distinctive brand
operational models. Some examples can be taken here, online shopping, TV
shopping and other non-traditional store-free direct selling is an innovation for
marketing channel model and is gradually nibbling the traditional retail market;
Canon replaces Xerox to be the leader of copier market, which is an innovation of
redefining the customer market; the ‘‘straight-thorough processing’’ of Dell is an
innovation of computer customization (Kreinsen 2008). The rapid development of
social economy and the complexity of consumer demand request the innovation of
brand operational model. Adapting to this changing trend, it is possible to create a
miracle within the industry by brand management and operational mode
innovation.
91.6 Optimization Effect of Value Chain Theory
on Brand Strategy
Profit is the ultimate goal of enterprises in the value chain theory. In the
increasingly competitive market and growing product homogenization, brand is an
important tool to provide differentiated value to consumers and plays an
increasingly significant role in market competition. Enterprises may optimize their
value chain to achieve long-term development of the brands (Pavitt 1984). The
optimization effects of value chain on multi-brand strategy are reflected in the
following aspects:
First, the value chain analysis can be applied to enhance the brand value.
Through analysis of value chain to identify the elements which can enhance
product functions and features and factors that may affect brand image, the pro-
duction costs can be reduced and the optimal resource allocation can be realized
(Xu 2009).
Second, the systematic management of brand based on detailed elements and
links of the value chain can improve the value of enterprise image. The brand
value is reflected just due to the asymmetry of consumers’ understand of product
information. Meanwhile the enterprises shall concentrate on exploring the brand
culture, creating product differentiation and forming their own characteristics to
meet the customers’ emotional demands and create personalized brand image.
Third, the differentiation of products can be employed from the perspective of
value chain to define enterprise strategy. Each link in the value chain is inde-
pendent and also interact each other. Through the analysis of various value chains,
the enterprise can recognize whether these chains are separated or coordinated
each other to achieve synergy effect and realize product and enterprise brand
optimization (Chen and Zheng 2009).
872 X. Li
91.7 Conclusions
To change the position from low end of the value chain, extend from the bottom of
the ‘‘smiling curve’’ to both ends, improve the additional value of the product and
enhance the competitiveness, brand construction is one important link for the small
and medium-sized enterprises. Brand strategy is a request for adaptation to eco-
nomic restructuring and also a strategic issue for sustainable development.
Brand is the image of the enterprise as well as its products, just like the image
of a person. To promote overall value creational capability, SMEs shall apply the
tool of brand strategy through formulation of appropriate brand strategy, estab-
lishment of scientific brand development plan, definition of clear brand position,
choice of correct brand appeal, and adoption of innovative brand operation model
in accordance with their own operational situations.
References
Aker D (1990) Consumer evaluations of brand extensions. J Mark 54(1):53–55
Bhat S, Reddy SK (2001) The impact of parent brand attribute association and affect on brand
extension evaluation. J Bus Res 53(3):111–112
Chen J, Zheng G (2009) Innovation management: to gain a sustainable competitive advantage
(in Chinese). Peking University Press, Beijing, p 111
Christensen CM (2010) Innovator’s dilemma (trans: Hu J). CITIC Publishing House, Beijing,
p 55
David B (1991) A process-tracing study of brand extension evaluations. J Mark Res 28(1):16–28
Kreinsen H (2008) Low technology: a forgotten sector in innovation policy. J Technol Manage
Innov 3(3):33
Pavitt K (1984) Sectoral patterns of technical change: towards a taxonomy and a theory. Res
Policy 13(6):13
Porter M (2002) National competitive advantage (trans: Li M, Qiu R). Huaxia Publishing House,
Beijing, p 71
Porter M (2005) Competitive strategy (trans: Chen X). Huaxia Publishing House, Beijing, p 105
Ries A, Trout J (2004) 22 laws of brand. Shanghai People’s Publishing House, Shanghai, p 121
Schultz D, Schultz H (2005) Don Schultz’s views on brand. People’s Posts and Telecommu-
nications Press, Beijing, p 135
Shao L (2005) What are the critical success factors for luxury goods companies looking to expand
in China? (in Chinese). Financial Times (12):95
Shih S (2005) Rebuilding of Acer. CITIC Publishing House, Beijing, p 206
Smith KR (2001) Treats to the external validity of brand extension research. J Mark Res
38(3):326–331
Xu Q (2009) Research development and technical innovation management (in Chinese). Higher
Education Press, Beijing, p 410
Chapter 92
Research on Information Mining About
Priority Weight of Linguistic
Judgment Matrix
Cai-feng Li
Abstract The thesis makes mining information of decision maker as the
breakthrough point, puts forward a ranking method involving parameter based on
the linguistic judgment matrix, whose different value corresponds to different
priority weight. It is necessary to add parameter in ranking method based on the
linguistic judgment matrix, and there are three methods to select parameter, whose
practicality and efficiency can be demonstrated by numerical examples.
Keywords Linguistic judgment matrix Method to select parameter Parameter
Preference
92.1 Introduction
In multi-attribution decision making, due to complexity and uncertainty of
objective thing, and fuzziness of human thinking pattern, even for specialists, it is
difficult to evaluate the attribute of any project, so it is convenient and reliable to
make decision on some attributions with linguistic phrase. In accordance with
existed research, there are two kinds of methods to make decision on linguistic
information, one is ranking method based on the consistency of linguistic judg-
ment matrix, and the other is applying operators to assemble decision-making
information and ranking projects. Based on consistent linguistic judgment matrix
or satisfactory linguistic judgment matrix, through shift formula, the first method
which can rank transforms linguistic judgment matrix to real-value matrix. Lit-
erature (Chen and Hwang 1992) puts forward the shift scale method, and literature
(Chen and Fan 2004) provides the method of transforming linguistic matrix to
C. Li (&)
Department of Hechi, University Guangxi, Yizhou,
People’s Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 873
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_92,
Ó Springer-Verlag Berlin Heidelberg 2013
874 C. Li
positive reciprocal matrix. The second method involves the induced ordered
weighted averaging (IOWA) operator put forward by literature (Yager 2003),
linguistic ordered weighted averaging (LOWA) operator in literature Herrera et al.
(1996), linguistic weighted arithmetic averaging (LWAA) and extensive ordered
weighted averaging (EOWA) operator in literature (Xu 1999),and other operators
in literatures (Herrera et al. 1995, 1996, 2000; Herrera and Herrera-Viedma 2000;
Umano et al. 1998; Wang and Fan 2002, 2003).
In ranking methods based on the consistency of linguistic judgment matrix, few
literatures add parameters into ranking methods, although there is parameter in
shift formula of literature (Chen and Fan 2004), it contains no active substance.
The paper puts forward a ranking method of linguistic judgment matrix involving
parameters, called parameter ranking method based on linguistic judgment matrix.
The information of decision maker would be mined in methods, and then decision
maker gets the better priority weights of linguistic judgment matrix.
92.2 Linguistic Judgment Matrix and its Consistence
Literature (Chen and Fan 2004; Fan and Jiang 2004) describes linguistic judgment
matrix and its consistency. Assuming there is linguistic phrase set S ¼
fSa ja ¼ t; . . .; 1; 0; 1; . . .; tg; and decision-making problem is limited to finite
set A ¼ fa1 ; a2 ; . . .; an g; where ai denote the project i. Decision maker uses a
matrix P ¼ ðpij Þnn to describe the information of the project set A, where pij
evaluate project ai and project aj , when pij ¼ fS1 ; S2 ; . . .; St g, project ai is better
than project aj , the more pij is, the greater project ai is superior to project aj , in
contrast, pij 2 fSt ; ; S1 g, project aj is better than project ai , the smaller pij is,
the greater project ai is inferior to project aj , while pij ¼ S0 , project ai is as good as
project aj , matrix P is called as linguistic judgment matrix.
Definition 1 (Herrera et al. 1995) Let S ¼ fSa ja ¼ t; . . .; 1; 0; 1; . . .; tg denote
natural language set, where Si is the i-th natural language, the subscript i and the
corresponding natural language can be obtained from following function I and I 1 :
I : S ! N; IðSi Þ ¼ i; Si 2 S
I 1 : N ! S; I 1 ðiÞ ¼ Si
Definition 2 (Chen and Fan 2004) With respect to P ¼ ðPij Þnn , 8i; j; k 2 J, its
elements satisfy with the following equation:
Iðpij Þ þ Iðpjk Þ þ Iðpki Þ ¼ 0 ð92:1Þ
Then the linguistic judgment matrix is consistent.
I : S ! N; IðSi Þ ¼ i; Si 2 S
92 Research on Information Mining About Priority Weight 875
I 1 : N ! S; I 1 ðiÞ ¼ Si
92.3 Parameter Ranking Method Based on Linguistic
Judgment Matrix
The logistic relation between linguistic judgment matrix and priority weight is put
forward in this chapter, called as the parameter ranking method based on linguistic
judgment matrix in the paper.
Theorem 1 A sufficient and necessary condition of the consistent linguistic
judgment matrix P ¼ ðpij Þnn is that there exist a positive normalized vector x ¼
ðx1 ; x2 ; . . .; xn ÞT and h, which satisfy the following formula:
xi
Iðpij Þ ¼ logh ; i; j 2 J; where h [ 1: ð92:2Þ
xj
Proof: Necessary condition Assume h [ 1; let
P
n P
n
1
n IðpikÞ X
n 1
n IðpikÞ
xi ¼ h k¼1 = h k¼1 ; i2J ð92:3Þ
i¼1
P
then 8i 2 J; xi [ 0, it is obvious to exist ni¼1 xi ¼ 1.
Because pij is related to xi and xj , it is reasonable to assume
Iðpij Þ ¼ gðxi Þ gðxj Þ, where gðxi Þ ¼ ði 2 J Þis monotonously increasing func-
tion. If the linguistic judgment matrix P ¼ ðpij Þnn is consistent, 8i; j 2 J,
Iðpij Þ ¼ gðxi Þ gðxj Þ, Iðpij Þ ¼ Iðpji Þ from Eq. (92.1), Iðpij Þ ¼ Iðpik Þ Iðpjk Þ,
Pn 1
Pn
1 Iðp Þ
Therefore, xi =xj ¼ hn k¼1 Iðpik Þ =hn j¼1 jk ¼ hIðpij Þ , also Iðpij Þ ¼ logh x xj ; i; j 2 J.
i
Sufficient condition
If pij of linguistic judgment matrix satisfies with Iðpij Þ ¼ logh x xj ; i; j 2 J, where
i
Pn
h [ 1, xi [ 0; xi [ 0, and i¼1 xi ¼ 1, it is easy to draw the following con-
clusion: I ðpIJ Þ þ Iðpjk Þ þ Iðpki Þ ¼ 0, logh x xi xk xi xi xk
xj þ logh xk þ logh xi ¼ logh xj : xk : xi
i
¼ logh 1 ¼ 0. So, the linguistic judgment matrix P ¼ ðpij Þnn is consistent. From
formula (92.2) and h [1, it is not difficult to draw the following conclusion:
pij 2 fs1 ; s2 ; . . .; st g , I pij [ 0 , x
xj [ 1 , xi [ xj , the more pij is, the more
i
xi
xj is, in other word, the project ai is prior to the project aj to greater extent;
pij 2 fst ; ; s1 g , I pij \0 , x xj \1 , xi \xj ; the smaller pij is, the
i
xi
smaller xj is, in other word, the project ai is inferior to the project aj to greater
876 C. Li
extent; pij ¼ s0 , I pij ¼ 0 , x
xj ¼ 1 , xi ¼ xj which demonstrate that it is
i
reasonable to regard the positive normalized vector x ¼ ðx1 ; x2 ; . . .; xn ÞT as the
priority weighty of evaluation the project. From theorem 1 and formula (92.3), it is
easy to draw the following conclusion: under the condition of the same linguistic
scale and consistency, the priority vectors are a family of ranking vectors
involving parameter, the priority should changed with parameter, which provides
some suggestions how to establish the priority vector of project from linguistic
judgment matrix, meanwhile, which puts forward a new method of defining the
priority vector of project.
92.4 The Necessity of Selecting Parameter
The following example 1 demonstrates that different parameter may induce dif-
ferent ranking project.
Example 1 There are two selectable projects with two attributes u1 ; u2 . After a
decision maker grades every attribute from 0 to 100, the decision-making matrix B
can be obtained, whose normalized matrix is R. The decision maker constructs the
linguistic judgment matrix H through pairwise comparison in accordance
with
S0 S 4
linguistic scale S ¼ fSa ja ¼ 5; . . .; 1; 0; 1; . . .; 5g, H ¼ , it is obvi-
S4 S0
ous that the judgment H is consistent.
Next, we consider the following two situations. First, assuming h ¼ 1:4953,
from formula (92.3), x ¼ ð0:8333; 0:1667Þ can be obtained, through utilizing
simply weighing method, the evaluation of above two projects is Z = (0.5056,
0.4944), so a1 a2 (Tables 92.1, 92.2).
Secondly, assuming h ¼ 1:6818, from formula (92.3), x ¼ ð0:8889; 0:1111Þ
can be obtained, through utilizing simply weighing method, the evaluation of
above two projects is Z = (0.4926, 0.5074), so a2 a1 . Example 1 demonstrates
that different parameter may induce different ranking result under multiple stan-
dards, so it is necessary to select reasonable parameter and to put forward ranking
method by introducing parameter in the decision based on judgment matrix.
Table 92.1 Decision- u1 u2
making matrix B
a1 70 70
a2 80 30
92 Research on Information Mining About Priority Weight 877
Table 92.2 Normalized u1 u2
matrix of R from B
a1 0.4667 0.7
a2 0.5333 0.3
92.5 The Method to Select Parameter
In accordance with above analysis, in order to have reasonable weight, it is
essential to obtain suitable parameter, the paper puts forward following three
methods.
92.5.1 The First Comprehensive Weight Method Based
on Linguistic Judgment Matrix
When there are less than 5 selectable projects, it is considerable to apply the first
comprehensive weight method, whose stages are as follows. First, the decision
maker selects two projects from projects A1 ; A2 ; . . .; An , such as A1 ; A2 . Secondly,
the decision
P maker
gives the two projects real-valued weight
0 0 2 0 0 0
x1 ; x2 i¼1 xi ¼ 1 . Thirdly, insert x1 ; x2 into the formula (92.2), and obtain
the following formula:
x01
Iðp12 Þ ¼ logh ð92:4Þ
x02
Fourthly, from formula (92.4), it is not difficult to solve the parameter h which
embodies the preference of the decision maker. Finally, it is important to insert the
value of h into the formula (92.3) to solve the priority weight which embodies the
preference of decision maker to greater extent.
Example 2 If decision maker gives following linguistic judgment matrix A and
real-value matrix A0 induced from A.
S0 S1 S0 S0 1 h1 1 1
S S0 S1 S1 h 1 h h
A ¼ 1 ; A 0
¼
1 h1 1 1
S0 S1 S0 S0
S0 S1 S0 S0 1 h 1 1
The decision maker gives the projects A1 ; A2 the weight ðx01 ; x02 Þ ¼ ð0:4; 0:6Þ, it
is obvious that A22 is consistent, principle submatrices of A1 ; A2 also are
consistent.
x0
From formula (92.4), Iðp12 Þ ¼ logh x10 , 1 ¼ logh 0:4 , obtain parameter h ¼ 1:5,
2
1
Pn 0:6 P 1
Pn
then insert h ¼ 1:5 into formula(92.3): xi ¼ hn k¼1 Iðpik Þ = ni¼1 hn k¼1 Iðpik Þ , so
878 C. Li
h1=4 1:51=4
x1 ¼ ¼
h1=4 þ h3=4 þ h1=4 þ h1=4 1:51=4 þ 1:53=4 þ 1:51=4 þ 1:51=4
h3=4 1:53=4
x2 ¼ ¼
h1=4 þ h3=4 þ h1=4 þ h1=4 1:51=4 þ 1:53=4 þ 1:51=4 þ 1:51=4
h1=4 1:51=4
x3 ¼ 1=4 1=4 1=4
¼
h þh 3=4
þh þh 1:51=4 þ 1:53=4 þ 1:51=4 þ 1:51=4
h1=4 1:51=4
x4 ¼ ¼
h1=4 þ h3=4 þ h1=4 þ h1=4 1:51=4 þ 1:53=4 þ 1:51=4 þ 1:51=4
The priority vector of the project is
x ¼ ð0:2222; 0:3333; 0:2222; 0:2222Þ:
92.5.2 The Second Comprehensive Weight Method Based
on Linguistic Judgment Matrix
If there are more selectable projects whose number is between 5 and 9, considering
the complexity and diversity of decision making and human thinking, it is possible
to have deviation, so the paper puts forward the second comprehensive weight
method based on the consideration that the weight obtained from formula (92.3)
and the subjective weight of decision maker should be smaller. The stages of the
above method are as follows: First, every decision maker gives the subjective
weight to arbitrary three projects in order to obtain more preference information.
Secondly, the optimization model should satisfy with the following equation:
0 P 3 P
3 12
X
3 1
3 Iðpik Þ X
3 1
3 Iðpik Þ
minf ðhÞ ¼ @h k¼1 x0 h k¼1 A ;
i
i¼1 i¼1
s:t h 1
P
where x01 ; x02 ; x03 is the subjective weight, and 3i¼1 x0i ¼ 1
Thirdly, the parameter can be obtained from above model, and is inserted in the
formula (92.3), the priority weight of project can be found.
P
Let di ¼ 13 3k¼1 Iðpik Þ; i ¼ 1; 2; 3, then the above model can be simplified.
!2
X
3 X
3
minf ðhÞ ¼ h di
x0i h di
i¼1 i¼1
s:t h 1
92 Research on Information Mining About Priority Weight 879
Example 3 If decision maker gives following linguistic judgment matrix A and
real-value matrix A0 induced from A.
S0 S1 S0 S0 S3 1 h1 1 1 h3
S1 S0 S1 S1 S2 h 1 h h h2
A ¼ S0 S1 S0 S0 S3 ; A0 ¼ 1 h1 1 1 h3
S0 S1 S0 S0 S3 1 h 1 1 h3
S3 S2 S3 S3 S 0 h3 h2 h3 h3 1
If one specialist gives the projects A1 ; A2 ; A3 the subjective weight
ðw01 ; w02 ; x03 ÞT¼ ð0:25; 0:5; 0:25Þ, it is reasonable to minimize the difference
between the weight obtained from formula (92.3) and subjective weight
x01 ; x2 ;0 x03 , and to construct mathematical model:
!2
X 3 X3
di0 0
minf ðhÞ ¼ h x0i d
hi
i¼1 i¼1
s:t h 1
P3
where di0 ¼ 3 1
Iðpik Þ; i ¼ 1; 2; 3;
k¼1
h i2
min f ðhÞ ¼ 2 h1=3 0:25ðh1=3 þ h2=3 þ h1=32 Þ
h i2
þ h2=3 0:5ðh1=3 þ h2=3 þ h1=32 Þ ; s:t h 1
h ¼ 1:833, insert h ¼ 1:833 into formula (92.3), and obtain the priorityvector
x ¼ ð0:124; 0:3047; 0:124; 0:124; 0:3882Þ.
92.5.3 The Third Comprehensive Weight Method Based
on Linguistic Judgment Matrix
If there are more selectable projects whose number is between 5 and 9, considering
the deviation of decision maker understanding the scale, the paper puts forward the
third comprehensive weight method based on the linguistic judgment matrix,
whose stages are as follows. First, every decision maker gives the subjective
weight to arbitrary three projects. Secondly, insert the weight into formula (92.3)
to obtain three equations, and solve the unknown parameters h1 ; h2 ; h3 . Thirdly,
compute the average h of h1 ; h2 ; h3 ; h ¼ 13 ðh1 þ h2 þ h3 Þ. Finally, after inserting h
into the formula (92.3), the priority weight of project can be found. If the decision
maker gives the subjective weight w01 ; w02 ; x03 to the projects A1 ; A2 ; A3 , the fol-
lowing three equations can be obtained from formula (92.3).
880 C. Li
P
3 P
3
1
3 Iðp1k Þ X
3 1
3 Iðpik Þ
x1 ¼ h k¼1 = h k¼1 ð92:5Þ
i¼1
P
3 P
3
1
3 Iðp2k Þ X
3 1
3 Iðpik Þ
x2 ¼ h k¼1 = h k¼1 ð92:6Þ
i¼1
P
3 P
3
1
3 Iðp3k Þ X
3 1
3 Iðpik Þ
x3 ¼ h k¼1 = h k¼1 ð92:7Þ
i¼1
Through applying the third method, the example 3 also can be solved.
92.6 Conclusion
The paper discusses the problem about parameter of priority of linguistic judgment
matrix, demonstrates the necessity of adding parameter in the sorting method
based on linguistic judgment matrices, puts forward the conclusion of obtain the
parameter value through mining information of decision maker, and some methods
of selecting parameter through making full use of the preference information
which can be reflected by the subjective weight.Project supported by the Scientific
Research Foundation of the Higher Education Institutions of Guangxi Zhuang
Autonomous Region (Grant No. 201204LX394).
References
Chen Y, Fan Z (2004) Study on consistency and the related problems for judgment. Syst Eng
Theory Pract 24:136–141 (in Chinese)
Chen S, Hwang CL (1992) Fuzzy multiple attribute decision-making. Springer-Verlag, Berlin
Fan Z, Jiang Y (2004) A judgment method for the satisfying consistency of linguistic judgment
matrix. Control Decis 19(8):903–906 (in Chinese)
Herrera F, Herrera-Viedma E (2000) Linguistic decision analysis: steps for solving decision
problems under linguistic information. Fuzzy Sets Syst 115(10):67–82
Herrera F, Herrera-Viedma E, Verdegay JL (1995) A sequential selection process in group
decision-making with linguistic assessments. Inf Sci 85(4):223–229
Herrera F, Herrera-Viedma E, Verdegay JL (1996) Direct approach processes in croup decision
making using linguistic OWA operators. Fuzzy Sets Syst 79:175–190
Herrera F, Herrera-Viedma E, Verdegay JL (1996) Direct approach processes in group decision
making using linguistic OWA operators. Fuzzy Sets Syst 78(2):73–87
Herrera F, Herrera-Viedma E, Martinez L (2000) A fusion approach for managing
multi-granularity linguistic term sets in decision making. Fuzzy Sets Syst 114(9):43–58
Umano M, Hatono I, Tamura H (1998) Linguistic labels for expressing fuzzy preference relations
in fuzzy group decision making. IEEE Trans Syst Man Cybern Part B Cybern 28(2):205–218
92 Research on Information Mining About Priority Weight 881
Wang XR, Fan ZP (2002) A topsis method with linguistic information for group decision making.
Chin J Manag Sci 10(6):84–87 (in Chinese)
Wang XR, Fan ZP (2003) An approach to multiple attribute group decision making with
linguistic assessment information. J Syst Eng 18(2):173–176 (in Chinese)
Xu Z (1999) Uncertain attribute decision making: methods and application. Tsinghua University
press, Beijing (in Chinese)
Yager RR (2003) Induced aggregation operators. Fuzzy Sets Syst 137:59–69
Chapter 93
Research on the Project Hierarchy
Scheduling for Domestic Automobile
Industry
Peng Jia, Qi Gao, Zai-ming Jia, Hui Hou and Yun Wang
Abstract To improve the accuracy and performability of the vehicle R&D project
scheduling of domestic automobile enterprises, a ‘‘4 ? 1’’ hierarchy process
system of domestic automobile enterprises is analyzed and summarized. A
corresponding four levels scheduling management mode of the vehicle R&D
project is presented based on the hierarchy process system, and a planning
approach of three-month rolling schedule for the fourth level is proposed. Sche-
dule minor adjustment and modification are given to solve the different extent
change of rolling schedule.
Keywords Hierarchy process
Hierarchy scheduling Rolling schedule
Schedule minor adjustment Schedule modification
93.1 Introduction
The research and development (R&D) of automobile products is complicated
system engineering. The automobile R&D has a long cycle, and involves wide
range of knowledge. At present, the R&D of a new model on a new platform
P. Jia Q. Gao (&)
School of Mechanical Engineering, Shandong University, Jinan, China
e-mail: [email protected]
P. Jia Q. Gao
Key Laboratory of High Efficiency and Clean Mechanical Manufacture, Shandong
University, Ministry of Education, Jinan, China
Z. Jia
Department of Navigation, Qingdao Ocean Shipping Mariners College, Qingdao, China
H. Hou Y. Wang
Technical Division, Shan Dong Shan Kuang Machinery Co., Ltd, Jining, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 883
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_93,
Ó Springer-Verlag Berlin Heidelberg 2013
884 P. Jia et al.
generally requires 36–48 months, even on an existing platform it still requires
18–32 months (Wang 2009). The R&D process involves lots of knowledge and
technology, such as mechanics, aerodynamics, structural mechanics, aesthetics,
electrotechnics, electronics, cybernetics, computer science, etc. (Lin 2008).
Therefore, professionals from different functional departments and different sub-
ject areas need to collaborate with each other to complete the R&D work.
For such a complicated huge system engineering, how to develop automobile
products which meet the market and customer demands in limited time? This puts
forward higher request to R&D process and R&D project management of new
products. In recent years, Chinese automobile enterprises continue to learn from
international leading enterprises, and many management tools and management
methods such as advanced product quality planning (APQP) (Chrysler Corpora-
tion, Ford Motor Company, and General Motors Corporation 1995; Chen 2008)
and project management (Liu 2009; Ju 2008; Zhang 2008; Sun 2004) are intro-
duced to local automobile R&D process. They build their own R&D process based
on learning from the standard R&D process of foreign enterprises.
However, due to the shortage of time, data and experience, immaturity of
management and many other reasons, the R&D process cannot be properly
implemented in domestic automobile enterprises, and the advanced management
theory of project management cannot be well applied in the automobile R&D
process management. Therefore, this paper analyzes and summarizes the current
product R&D process of domestic automobile enterprises, presents corresponding
management mode of the vehicle R&D project scheduling, and gives the solution
to schedule decomposition problem caused by the long cycle and wide range of
products R&D.
93.2 The Hierarchy Process of Vehicle R&D
Most domestic automobile enterprises have adopted hierarchy process to manage
the R&D process. The vehicle R&D flow is divided layer by layer following a
coarse-to-fine sequence in order to facilitate the process management. According
to the R&D process, the vehicle R&D flow is divided into four levels: company
quality gates (Q-Gates) level (the first level flow), cross majors/fields level (the
second level flow), cross departments level (the third level flow), department level
(the fourth level flow). The enterprises usually define a level with foundational
fixed flow to implement of the four levels of R&D flow smoothly. So the flow
system of automobile enterprises is the four levels of R&D flow plus a level of
foundational fixed flow.
(1) Company Q-Gates level. Based on the quality requirements of APQP and the
management theory of Stage-Gate, the vehicle R&D process is divided into
several stages, and the R&D quality of products is ensured through setting up a
93 Research on the Project Hierarchy Scheduling 885
Q-Gate between two stages. These R&D stages and Q-Gates constitute the
flow of company Q-Gates level.
(2) Cross majors/fields level. On the basis of division of stages and Q-Gates, all
the flow nodes of company Q-Gates level are subdivided according to majors
or fields which are involved by automobile products, and the second level
subflow of cross majors or fields is defined.
(3) Cross departments level. The flow nodes of cross majors/fields level are
subdivided to departments which are involved by all majors or fields, and the
third level subflow of cross departments is defined.
(4) Department level. The R&D flow is defined within the departments according
to their business scope in the vehicle R&D process. The flow is also the
subdivision of cross departments flow nodes, so it is the fourth level flow in the
vehicle R&D process system.
(5) Foundational fixed level. In order to improve the R&D efficiency, the enter-
prises establish many fixed flows for certain R&D activities, as the basic
supporting of the vehicle R&D process system. The flows can realize the
automatic transfer among different steps of flow activities, avoid repeated hand
labor and reduce manual workload.
Figure 93.1 Shows the ‘‘4 ? 1’’ process system of Q automobile Co., Ltd.
which is established based on their new product R&D manual (http://
doc.mbalib.com/view/d9fd9a8d5538f64af4cfb3; http://www.docin.com/p-
220283401.html; http://wenku.baidu.com/view/dd7f9633a32d7375a4178066.html
). In the first level, the vehicle R&D process is divided into eleven stages from P0
to P10, and eleven Q-Gates are set up, such as new project research instruction,
project R&D instruction, engineering start instruction, digital prototype and so on.
In the second, the stages are subdivided according to the involved majors or fields.
Taking P2 sculpt design stage as an example, it’s divided into many tasks belong
to mechanical design, manufacturing process, marketing, and other majors. In the
third level, the flow of cross majors or fields is subdivided to departments. Taking
the first round structural design as an example, the flow is divided to platform
technology department, battery system department, CAX design simulation
department and others. In the fourth level, the R&D flow in every department is
defined. Take the first round assembly design as an example, the flow defines four
flow steps that are the definition of system function and performance, the design of
system parts and components, the definition of parts and components function and
performance and the summary of system data. Document approval flow is a
foundational fixed flow, and the typical procedures include compile-proofread-
audit-approve.
886 P. Jia et al.
The vehicle R&D flow of Q automobile Co., Ltd
P0 P1 P2
P0 Stage New project P1 Stage Project P2 Stage Engineering P3 Stage
Project pre-research Research Project approval R&D Sculpt design start Detailed
instruction engineering design
instruction instruction
P3 P4 P5
P4 Stage P5 Stage Validation P6 Stage P6 P7 Stage
Digital Product design Verification Product design Small PVS Zero
prototype prototype prototype
verification validation batch trial freeze production
P9
P8 Project P10 Stage P10
P7 P8 Stage P9 Stage Project continuous
0S freeze Mass production SOP Project summary summary Project
The First freeze report improvement acceptance
Level Flow
Edit the second edition Quality analysis
of product historical models of Select suppliers
configuration table the same platform
Marketing Department Quality department Purchasing Department
The second round CAE analysis
The first round general layout vehicle & parts and Issue conceptual
structural design data (M0)
design components
Design
departments Design departments Design departments Design departments
Analysis process
The Second
Level Flow Manufacturing
process department
Edit CAE Checklist
Platform technology
department CAE analysis and
Battery system optimization The first round some
department parts and components
Electric drive CAX design simulation structural design
technology department department Platform technology
System control department
technology department Battery system
department
Make CAS data Check the vehicle The first round Issue CAS data Electric drive
general layout assembly design technology department
Outsourcing design Platform technology Platform technology Outsourcing design System control
company department department company technology department
Make clay Experiment clay
Design details modeling modeling
The Third
Level Flow Outsourcing design Outsourcing design Experiment
company company technology department
Define parts and
Define system function Design system parts components function Summarize
and performance and components and performance system data
The Fourth
Level Flow Design engineer Design engineer Design engineer Design engineer
Fig. 93.1 ‘‘4 ? 1’’ process system of Q automobile co., ltd
93.3 The Scheduling of Vehicle R&D Project
93.3.1 Hierarchy Schedule
In automobile enterprises, the thought of project management is introduced to
manage the vehicle R&D process. The hierarchy schedule is developed for the
vehicle R&D project corresponding to the hierarchy R&D flow model of auto-
mobile products to help enterprises implement the automobile R&D process more
convenient. Based on the four-level R&D flow, the vehicle development project
schedule can be decomposed into four levels: the first level schedule (big sche-
dule), the second level schedule (cross majors/fields schedule), the third level
93 Research on the Project Hierarchy Scheduling 887
Company Q-Gates Level The First Level Flow Big Schedule The First Level Schedule Project Manager
Cross Majors/ Fields Level The Second Level Flow Cross Majors/ Fields Schedule The Second Level Schedule Project Manager
Cross Departments Level The Third Level Flow Cross Departments Schedule The Third Level Schedule The Head of Majors/Fields
Department Level The Fourth Level Flow Department Operation Schedule The Fourth Level Schedule The Head of Department
Foundational Fixed Flow
Fig. 93.2 Corresponding relationship between hierarchy R&D process and hierarchy schedule of
automobile
schedule (cross departments schedule) and the fourth level schedule (department
operation schedule). It is corresponding layer by layer between the four-level R&D
flow and the four-level schedule, as shown in Fig. 93.2.
(1) The first level schedule, also known as big schedule, is planned by project
manager based on the vehicle R&D Q-Gates flow. The vehicle R&D stages
correspondingly constitute the stage summary tasks in the first level schedule
The Q-Gates correspondingly constitute the milestones.
(2) The second level schedule is cross majors/fields schedule. It is planned by
project manager based on the vehicle R&D cross majors/fields flow. The flow
nodes correspondingly constitute the tasks in the second level schedule, and
are put under the phase tasks which correspond to the high level flow nodes of
the second level flow nodes, so this level schedule is also the decomposition of
the first level stage schedule. In addition, the second level schedule tasks will
be assigned to the appropriate majors or fields.
(3) The third level schedule is cross departments schedule in the project. These
tasks of this level schedule are the decomposition of the majors or fields tasks
by the head of majors or fields based on the cross departments flow, and the
tasks will be assigned to departments. The third level flow nodes corre-
spondingly constitute the tasks in the third level schedule.
(4) The fourth level schedule is operation schedule within the departments. The
head of departments decomposes the work of the third level schedule tasks
based on the department flow and the functions and responsibilities of the
department. The fourth level flow nodes correspondingly constitute the fourth
level schedule tasks, and the tasks will be assigned to the project members.
There is no direct relationship between the foundational fixed flow and the
decomposition of the project schedule, but the flow can support the implementa-
tion of the schedule tasks.
Figure 93.3 shows the four-level R&D project schedule of Q automobile Co.,
Ltd. which is planned based on the company’s four-level vehicle R&D flow.
888 P. Jia et al.
The second level schedule
The third level schedule
The first level schedule
The fourth level
schedule
Fig. 93.3 Vehicle R&D project schedule of Q automobile co., ltd
93.3.2 Rolling Scheduling
The vehicle R&D project has long project cycle, involves wide majors and fields
range, needs many coordinated interaction among departments, exists more
uncertainty factors. Therefore, in the stage of project approval, the schedule cannot
be decomposed exhaustively, only the first, the second and the third level schedule
can be initially decomposed based on the standard R&D stages, the involved
departments and overall R&D requirements. The detailed fourth level specific
operation schedule within department is difficult to accurately plan.
The method of rolling scheduling (http://baike.baidu.com/view/1359753.htm)
can effectively solve the above problem as it can regularly revise future schedule.
The schedule is planned based on the principle of detailed recent and coarse
forward. It means to plan detailed specific recent schedule and the coarse forward
schedule at first, and then regularly make the necessary adjustments and revision to
the schedule according the situation of implementation and the technical problems.
The method combines recent scheduling and forward scheduling. On the one hand,
it can plan the next R&D tasks in advance. On the other hand, it can solve the
contradiction between the relative stability of schedule and the uncertainty of
93 Research on the Project Hierarchy Scheduling 889
actual situation better, and effectively improve the accuracy and performability of
schedule.
The R&D cycle of domestic automobile products is usually 3–5 years, so the
three-month rolling period for domestic automobile enterprises is reasonable and
easy to manage and achieve. Therefore, it is needed to plan three-month rolling
schedule in the period of vehicle R&D project. It means that the fourth level
schedule of the next 3 months is planed in every month.
In the process of planning three-month rolling schedule, the schedule will be
adjusted and revised according to the actual situation. It may lead to different
extent change of the schedule. Two ways can be used to deal with different extent
change.
(1) Schedule minor adjustment. The project managers can adjust the project
schedule for small change of schedule which does not affect the milestone
tasks and the key tasks on the critical path.
(2) Schedule modification. When the change affects the milestone tasks and the key
tasks on the critical path, the project managers must modify the project sche-
dule. The schedule modification will be achieved through implementing the
change flow of project schedule, and then increase the version of the schedule.
93.4 Conclusion
Based on the management status of the vehicle R&D of domestic automobile
enterprises, the paper analyzes and summarizes the ‘‘4 ? 1’’ flow system of
domestic automobile enterprises, which includes company quality gates(Q-Gates)
flow, cross majors/fields flow, cross departments flow, department flow, and a level
of basic foundational fixed flow. According to the four levels R&D process of
vehicle, the paper presents corresponding four levels schedule management mode
of the vehicle R&D project, which includes big schedule, cross majors/fields
schedule, cross departments schedule, department operation schedule. The paper
proposes a planning way of three-month rolling schedule as the fourth level
schedule is difficult to detailly and accurately plan. Finally, the paper gives two
ways, which are schedule minor adjustment modification to deal with different
extent change of rolling schedule.
Acknowledgments The project is supported by the National High Technology Research and
Development Program of China (through grant No. 2012AA040910) and the Natural Science
Foundation of Shandong Province (through grant No. ZR2012GM015).
890 P. Jia et al.
References
Chen G (2008) FA corporation’s R&D process reengineering Based on APQP and project
management (in Chinese), Xiamen University
Chrysler Corporation, Ford Motor Company, and General Motors Corporation (1995) Advanced
product quality planning (APQP) and control plan reference manual, in press
Information on http://www.docin.com/p-220283401.html, (in Chinese)
Information on http://baike.baidu.com/view/1359753.htm, (in Chinese)
Information on http://wenku.baidu.com/view/dd7f9633a32d7375a4178066.html, (in Chinese)
Information on http://doc.mbalib.com/view/d9fd9a8d5538f64af4cfb3 b7593da32b.html, (in
Chinese)
Ju Y (2008) Research on application of the project management in R&D of CA 305 automobile
self-determination-oriented products (in Chinese), Tianjin University
Lin J (2008) Research on methods and application of project management to the new vehicle
research & development Project (in Chinese), Tianjin University
Liu S (2009) The application of project management on entire vehicle design (in Chinese),
Shanghai Jiao Tong University
Sun W (2004) Application of project management in automobile product development, (in
Chinese). Automob Sci Technol 4:44–46
Wang W (2009) Research of project management based on vehicle products R&D (in Chinese),
HeFei University of Technology
Zhang Q (2008) Application research of the project management based on lifecycle in automobile
development (in Chinese). Shanghai Auto 8:24–27
Chapter 94
SVM-Based Multi-Sensor Information
Fusion Technology Research in the Diesel
Engine Fault Diagnosis
Jian-xin Lv, Jia Jia and Chun-ming Zhang
Abstract According to engine’s characteristics of running mechanism and prone
to failure, using integration based on the sub-module decision-making output
multi-sensor information fusion model, this paper discusses the use of SVM-based
multi-sensor information fusion technology on the diesel engine fault diagnosis.
As the real data of the fault vehicles experiment shows, compared to the traditional
diagnostic methods, SVM-based multi-sensor information fusion technology is
more effective on identifying the agricultural diesel failure type.
Keywords Diesel engine Fault diagnosis Multi-sensor information fusion
Support vector machine (SVM)
94.1 Introduction
The vigorous development of the automotive market has led to the improvement of
diesel engine fault diagnosis technology has become the mainstream of diesel
engine fault diagnosis, diagnostic techniques based on sensor signals. Tradition-
ally, the relative maturity of the spectrum-based signal analysis algorithms, but
such methods due to lack of time local analysis function, and is not suitable to
analyze non-stationary signals. The diesel engine vibration signal contains a large
number of high-frequency, low frequency and its harmonic components. By
Vapnik’s support vector machine (SVM) (Vapnik 1995) is a new learning machine
J. Lv J. Jia (&) C. Zhang
Department of Transportation Engineering, Engineering College
of Armed Police Force, Xi’an, Shaanxi, China
e-mail: [email protected]
J. Lv
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 891
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_94,
Springer-Verlag Berlin Heidelberg 2013
892 J. Lv et al.
based on statistical learning theory. Compared to the neural network, which will
use heuristic learning methods in the implementation with a lot of experience in
composition. SVM avoid the local minimum problem, and not overly dependent
on the quality and quantity of the sample, greatly improving its generalization
ability. Multi-sensor information fusion technology can improve the integration
and integration of information between the different sensors, information redun-
dancy, complementarity, timeliness and accuracy. The theory of SVM is intro-
duced into the multi-sensor information fusion technology, and applied in the
agricultural diesel engine fault diagnosis, and achieved good results.
94.2 Information Fusion Based on SVM
94.2.1 The Concept of Support Vector Machines SVM
For a linearly separable sample set N, N is ðxi ; yi Þ,…, ðxk ; yk Þ, xi 2 Rn ,
yi 2 ðk; kÞ; i ¼ 1; 2; . . .; k. Seeking an optimal. hyper-plane x x þ b ¼ 0 will be
2
two types of separation and spacing of the kxk2 largest. The relaxation factor
P
ni 0 ði ¼ 1; 2; . . .; kÞ; ki¼1 ni that allows the sample to the degree of misclassi-
P
fication, and its minimum. That is, solving: min 12 kxk2 þC ki¼1 ni ; C penalty
coefficient, to correct the misclassification sample caused by deviation in accor-
dance with the degree of importance. When the sample linear non-time-sharing,
decision function Kðx; xiP Þ can be divided into the data of high dimension space by
kernel function, gðxÞ ¼ SV ai yi Kðx; xi Þ þ b 0 ai C; ai is the Lagrange factor.
The decision output is that: dð xÞ ¼ sgn½gðxÞ.
Through 1-on-1 to promote the SVM to multi-class classification: N Construction
on the number of separator N ðN 1Þ=2. Discrimination, the new test sample x can
be obtained N ðN 1Þ=2 discrimination results and vote. x belong to the highest
classification. Token the category subscript class if appeared flat votes.
94.2.2 Fusion Method Based on the Output Sub-Module
Decision-Making
Training of large-scale data is not only time-consuming and memory demanding
on the hardware, there would be insufficient memory space training. Were solved
using this small module based on SVM information fusion technology will be the
number of large-scale data decomposition, the ultimate fusion. Assume that
decomposition of the overall problem into K sub-module, each module containing
the N type of data, including information fusion method based on SVM
(Hu et al. 2005; Platt 1999; Hsu and Lin 2002): (1) decision-making output of the
94 SVM-Based Multi-Sensor Information Fusion Technology Research 893
sub-module integration; (2) sub-module in N the decision function value on the
class integration; integration of the weighted value of the decision function (3)
sub-module in the class N; (4) sub-module the number of votes in the class N
fusion.
This article taken the first category, the decision-making output of the sub-
module integration:
X k
dðxÞ ¼ arg maxfV1 ; V2 ; . . .Vk g; Vj ¼ dij; dij;
i¼1
1; di ðxÞ ¼ j
¼ ; ði ¼ 1; 2; . . .K; j ¼ 1; 2; . . .N Þ
0; di ðxÞ 6¼ j
such as the number
Pp of votes are p, discrimination in accordance with
dðxÞ ¼ arg max fij ðxÞ .
i¼1
94.3 Diesel Engine Sensor Fault Diagnosis Application
According to characteristics of diesel engines, commonly used cylinder head
vibration acceleration sensor, the instantaneous speed sensor, cylinder pressure
sensor three types of sensors to collect the required information, the amount of
feature extraction, and in accordance with the integration of decision-making
output of the above sub-module fusion. Prone to the actual running of each cyl-
inder power imbalance in the type of fault, for example. The failure of the cylinder
power imbalance is a common fault. Cylinder head vibration sensors, in theory, the
cylinder pressure sensor, instantaneous speed sensor fusion of the three diagnoses
can get the best results. But in the actual diagnostic process, the high temperature
and high pressure cylinder environmental damage the performance and life of the
pressure sensor, greatly increasing the difficulty of the measurement of cylinder
pressure. And by calculating the conversion speed is obtained by the instantaneous
cylinder pressure, the conversion formula (Kennedy and Eberhart 1995; Coello
and Lechuga 2002):
( ) ( )
XN
2 €
XN
J þ m1 R 2
½f ðh / Þ h þ m1 R
k
2
½f ðh / Þgðh / Þ h_
k k
k¼1 k¼1
X
N
¼ Ap R ½fpðkÞ ðhÞf ðh /k Þ T
k¼1
The test measured in normal and pipeline oil spill two states under the cylinder
head vibration 5 signal for each, is calculated to extract diagnostic indicators such
as Table 94.1.
Recourse to Table 94.1 data to establish a diagnostic model. Assume that the
indicators in the state vector X = [X1, X2, X3] T in accordance with and other
894 J. Lv et al.
Table 94.1 Sample data
Sample number Detonation pressure of normal state Detonation pressure of oil spill state
8.367 MPa 7.608 MPa
P1-1 P1-2 P1-3 P1-1 P1-2 P1-3
1 63.1677 8.2508 12.4569 51.3794 7.5608 10.9037
2 60.1160 8.9492 13.0616 50.6096 7.3252 10.6371
3 61.5950 7.8080 11.9880 50.6898 7.6953 11.1545
4 64.2261 8.8686 12.9638 52.4091 7.9606 11.7211
5 62.6066 9.9479 15.1454 52.5400 7.9170 11.5639
P
covariance matrix normal distribution, denoted as X * NðlðhÞ; Þ. Sample mean
to estimate the l (h):
2 3 2 3
62:342 51:5256
lðh ¼ 8:365Þ ¼ 4 8:7649 5 lðh ¼ 7:607Þ ¼ 4 7:6918 5
13:1231 11:1961
P
To estimate the sample covariance matrices in the :
2 3
X Xk Xni 1:8874
1 i Þ ¼ 4 0:1707 0:3926
l ÞðX1l X 5
¼ ðX l X
n k 1 l¼1 i¼1 1
0:1595 0:5986 0:9834
94.4 Comparison with Other Traditional Diesel Engine
Fault Diagnosis Results
The current methods commonly used in diesel engine fault diagnosis, including
wavelet analysis, artificial neural network diagnosis, extended rough set theory,
and so on. Each method has the characteristics for diesel engine operation of the
law and prone to failure characteristics, compare the pros and cons of various
methods in dealing with the diesel engine fault diagnosis is the key to promote the
further development of diesel engine fault diagnosis technology.
Fault data processing capabilities of several methods for comparing the above, in
the experiments from the actual testing of the diesel engine, select the total number of
features for 1820, the normal signal, the total number of features for the 714’s
imbalance signal, the total number of features for the 1148 collision friction signal.
From randomly selected 70 % of the characteristics of data for network training, the
remaining 30 % for network testing. Therefore, training in normal working condition
the signal characteristics for 1274, the imbalance in the number of signal charac-
teristics for 497, collision friction signal characteristics 812. Signal characteristics of
normal conditions in the test were 546, the number of features of the unbalanced
signal 217, and collision characteristics of friction signal for 336. The experimental
results of the training set and test set, respectively, as shown in Table 94.2.
94 SVM-Based Multi-Sensor Information Fusion Technology Research 895
Table 94.2 Test set classification comparison
Test set classification comparison Normal Unbalanced Friction and
collision
The total characteristic numbers 546 217 336
Artificial neural networks Correct 525 207 279
classification
number
Correct 96.15 % 95.39 % 83.04 %
classification
rate
Generalized rough sets theory Correct 539 217 294
classification
number
Correct 98.72 % 100 % 87.5 %
classification
rate
Wavelet analysis Correct 532 210 273
classification
number
Correct 97.44 % 96.77 % 81.25 %
classification
rate
SVM-based multi-sensor Correct 536 215 304
information fusion technology classification
number
Correct 98.17 % 99.08 % 90.48 %
classification
rate
From the experimental results can be seen, for diesel engine fault diagnosis,
artificial neural network methods require a large amount of data is not dominant.
The wavelet analysis method in the training set for the high recognition rate of the
normal signal, but performance degradation is more obvious in the test set. And for
the failure of the diesel engine, we put more weight on the test set under diesel
imbalance signal and friction collision signal to identify the correct rate. The
imbalance signal recognition, SVM-based multi-sensor information fusion tech-
nology and generalized rough set theory is almost equal, there are certain
advantages in the identification of friction collision signal.
94.5 Conclusion
In this paper, we use the SVM-based multi-sensor information fusion technology
for diagnosis of diesel engine failure. And with the example of the multi-cylinder
power imbalance failure, acquisition failure diesel real vehicle data, using a variety
of diagnostic methods for the comparison test. The results show that SVM-based
896 J. Lv et al.
multi-sensor fusion technology can effectively identify diesel engine operating
status and faults category. And more focused subsequent posterior distribution
compared to single sensor, compared with methods such as artificial neural net-
work, the same confidence level confidence interval is smaller, higher accuracy.
References
Coello CAC, Lechuga MS (2002) MOPSP: a proposal for multiple objective particle swarm
optimization. In: Proceedings of the IEEE congress on evolutionary computation, Honolulu,
Hawaii, USA
Hsu CW, Lin CJ (2002) A comparison of methods from multi-class support vector machines.
IEEE Trans Neural Netw 46(13):415–425
Hu Z-h, Cai Y-z, Li Y-g et al (2005) Data fusion for fault diagnosis using multi-class support
vector machines. J Zhejiang Univ Sci 6A(10):1030–1039
Kennedy J, Eberhart RC (1995) Particle swarm optimization. In: Proceedings of IEEE
international conference on neural network. Perth, Australia: 1942–1948
Platt JC (1999) Fast training of support vector machines using sequential minimal optimization.
In: Proceedings of advances in kernel methods support vector learning. MIT Press,
Cambridge, pp 185–208
Vapnik V (1995) The nature of statistical learning theory. Springer-Verlag, New York
Chapter 95
The Purchase House Choice Research
Based on the Analytic Hierarchy Process
(AHP)
Zhi-hong Sun, Lu Pan, Yan-yan Wang and Da-hu Zhang
Abstract Analytic Hierarchy Process (AHP) is a powerful tool to analyze
multiobjective and multicriteria complex system, which is a systematic and hier-
archical analytical method with the combination of qualitative and quantitative
analysis. By using AHP method to quantitatively analyze a variety of consider-
ation factors in house purchasing decision-making, and then using the results to
help people to make a scientific and rational decision-making in purchasing a
house process.
Keywords Ahp Multiobjective decision-making The consistency test The
indicators of purchasing house Weight
95.1 Introduction
Buying a satisfied house is the dream of many people, however, today, with the
fluctuant development of real estate industry, to realize this dream is not so simple.
In the actual purchasing process, buyers’ requirements for houses are not limited to
residential and other simple functions but require more humane, more comfortable.
Therefore, the respects that buyers concern about are increasingly broad, and the
requirements tend to fine, including real estate lots, product price, design style,
Z. Sun (&) Y. Wang
Department of Mathematics, Department of Foundation,
Air Force Logistics College, Xuzhou, Jiangsu, China
e-mail: [email protected]
L. Pan
Department of Computer, Department of Foundation,
Air Force Logistics College, Xuzhou, Jiangsu, China
D. Zhang
Department of Aviation Oil Supplies, Air Force Logistics College,
Xuzhou, Jiangsu, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 897
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_95,
Springer-Verlag Berlin Heidelberg 2013
898 Z. Sun et al.
landscape supporting, property services, district supporting, the quality of housing,
the developers’ credibility and traffic conditions, etc. Although what buyers con-
cerned about are more and fine, but every respect to achieve the wish of home
buyers is unlikely. How to buy a most satisfactory house relatively from many
houses? When the indicators conflict, what should be as home buyers’ primary
standards? These are the problems that buyers pay close attention to very much.
95.2 Analytic Hierarchy Process
Analytic Hierarchy Process (AHP) was proposed by Operational Research Expert
Professor Saaty in the 1970s, which is a systematic and hierarchical analytical
method with the combination of qualitative and quantitative analysis (Hu and Guo
2007). It is an effective method to transform the semi-qualitative and semi-
quantitative problems into quantitative problems, and is also a powerful tool of
analyzing multiobjective and multicriteria complex system. Its core is to decom-
pose a complex problem into a number of different factors, then in accordance with
the relationship between the factors to establish a hierarchical structure model,
with forming target layer, rule layer and scheme layer, and then with pairwise
comparison method to construct judgment matrix. AHP has been widely used in
various fields, such as economic planning and management, energy distribution,
the military command, etc. (Chen et al. 2007, 2011; Zhao 2007; Jin et al. 2011; Qi
2008; Li et al. 2012). The main steps of AHP are as follows:
(1) To establish a hierarchical structure model.
(2) To construct the judgment matrix.
(3) To calculate the relative weights and consistency test.
(4) To calculate the total ranking of levels and consistency test.
The formula of judgment matrix consistency index CI is:
kmax n
CI ¼ ð95:1Þ
n1
where, n is the order of Matrix, kmax is the largest eigenvalue of judgment matrix.
The formula of judgment matrix random consistency ratio CR is:
CI
CR ¼ ð95:2Þ
RI
where, the value of RI refers to the experimental results of Saaty (Hu and Guo
2007), seen in Table 95.1.
Table 95.1 Values of the random consistency index RI
n 3 4 5 6 7 8 9 10 11
RI 0.58 0.9 1.12 1.24 1.32 1.41 1.45 1.49 1.51
95 The Purchase House Choice Research 899
According to Saaty’s rule of thumb, when CR \ 0.1 that the judgment matrix
has a satisfactory consistency, now, the normalized eigenvector, corresponding to
the largest eigenvalue kmax , as the weight vector of the judgment matrix.
95.3 Applying AHP to Make a Purchasing House Choice
95.3.1 Establishing a Hierarchical Structure Model
Here we build a three-tier hierarchy structure, as follows:
The first layer is the target layer (O), namely to select the appropriate region to
buy a house. The second layer is the criteria layer (C), that is, the indicators of
purchasing house, including nine aspects, such as real estate lots, product price,
design style, landscape supporting, property services, district supporting, the
quality of housing, the developers’ credibility and traffic conditions, each aspects
in turn to be recorded as Ck ðk ¼ 1; 2; . . .; 9Þ. And the third layer is the program
layer (P), here taking four houses of one city as the program layer, from which a
buyer wants to buy a house, and be denoted by Pn ðn ¼ 1; 2; . . .; 4Þ. As a result, the
structure chart of hierarchical model is builded in Fig. 95.1 (Yuan 2012; Mo 2007;
Huang et al. 2006; Yang et al. 2004; Pan et al. 2010).
95.3.2 Determining the Weights W1 of the Criteria Layer (C)
to the Target Layer (O)
A large survey network has done a large-scale survey to the people who want to
buy houses, and according to the large amounts of data, it has obtained the houses
buyers’ concern degree to all aspects in the purchasing house process, shown in
Table 95.2.
Based on Table 95.2, we can obtain the degrees of influence of the purchasing
indicators when people buy a house. Then according to Saaty’s comparison
The target layer O To select a appropriate region to buy a house
Real estate Product Design Landscape Property District Quality Developers’ Traffic
The criteria layer C
lots price style supporting services supporting of house credibility conditions
The program layer P Central time zone China shuiyun stream Purple pavilion dongjun Riverfront Maple city
Fig. 95.1 The structure chart of the hierarchical mode
900 Z. Sun et al.
Table 95.2 The concern Indicators of purchasing house Concern degrees (%)
degrees to indicators of
purchasing house Real estate lots C1 48.08
Product price C2 34.35
Design style C3 49.13
Landscape supporting C4 23.43
Property services C5 32.17
District supporting C6 27.62
Quality of housing C7 47.23
Developers’ credibility C8 27.43
Traffic conditions C9 45.97
criterion (Hu and Guo 2007), through the pairwise comparisons of the 9 aspects in
the criteria layer, we can establish the judgment matrix A, as follows:
0 1
1 4 1=2 6 4 5 2 5 3
B 1=4 1 1=4 4 2 3 1=4 3 1=4 C
B C
B 2 4 1 6 4 5 3 5 3 C
B C
B 1=6 1=4 1=6 1 1=4 1=3 1=5 1=2 1=5 C
B C
A¼B B 1=4 1=2 1=4 4 1 3 1=4 3 1=4 C
C
B 1=5 1=3 1=5 3 1=2 1 1=5 2 1=5 C
B C
B 1=2 4 1=3 5 4 5 1 5 2 C
B C
@ 1=5 1=3 1=5 2 1=3 1=2 1=5 1 1=5 A
1=3 4 1=3 5 4 5 1=2 5 1
By calculating, we can get the largest eigenvalue of the judgment matrix A, that
is kmax ¼ 9:7920, and the corresponding normalized eigenvector is
w1 ¼ ð0:2126; 0:0694; 0:2653; 0:0229; 0:0599; 0:0385; 0:1637; 0:0301; 0:1375ÞT
Using (95.1), CI1 = 0.0990, and the corresponding random consistency index is
RI9 = 1.45 (n = 9), so using (95.2), we can get
CI 1 0:0990
CR1 ¼ ¼ ¼ 0:0683\0:1:
RI9 1:45
The above result indicates that it passed the consistency test. So w1 is the
weight vector of the criterion layer C to the target layer O.
95.3.3 Determining the Weights W2 of the Program Layer
(P) to the Criteria Layer (C)
According to the assessment of experts on the various indicators of the four
regions, we establish the judgment matrices of Ck-P, and conduct consistency test,
with the results in Table 95.3.
95 The Purchase House Choice Research 901
Table 95.3 the judgement matrices of Ck-P and the results of consistency test
Layer P Layer C
C1 C2 C3 C4 C5 C6 C7 C8 C9
w21 w22 w23 w24 w25 w26 w27 w28 w29
P1 0.4675 0.4554 0.0955 0.0919 0.4675 0.4733 0.4718 0.4675 0.4718
P2 0.2771 0.2628 0.16 0.3016 0.2771 0.2842 0.1643 0.2771 0.1643
P3 0.16 0.1409 0.2771 0.1537 0.16 0.1696 0.2562 0.16 0.2562
P4 0.0955 0.1409 0.4675 0.4528 0.0955 0.0729 0.1078 0.0955 0.1078
kj 4.031 4.0104 4.031 4.1658 4.031 4.0511 4.0458 4.031 4.0458
Cl2j 0.0103 0.0035 0.0103 0.0553 0.0103 0.017 0.0153 0.0103 0.0153
CR2j 0.0114 0.0039 0.0114 0.0614 0.0114 0.0189 0.017 0.0114 0.017
From Table 95.3, we can see that the consistency ratio CR of the various
indicators are all less than 0.1, that is, all passed the consistency test. Then the
weight of layer P to layer C is:
w2 ¼ ðw21 ; w22 ; ; w29 Þ49
95.3.4 Determining the Combined Weights W
of the Program Layer (P) to the Target Layer (O)
According to the C–O weights w1 and the P–C weights w2, we can obtain the P–O
weights:
w ¼ w2 w1 ¼ ðw21 ; w22 ; . . .; w29 Þ w1
¼ ð0:3608; 0:2119; 0:2189; 0:2083ÞT ;
and the consistency ratio of the combination is:
P9 2
j¼1 aj CIj
CR ¼ P9 ¼ 0:014\0:1
j¼1 aj RI4
where aj (j = 1, 2,…, 9) corresponds to the various weights of w1. Therefore, the
combined weights w can be the basis for objective decision-making.
95.3.5 Comprehensive Ranking
According to the combined weights w, we can finally obtain the ranking of the four
intentive regions, that is, Central time zone region is better than Purple pavilion
902 Z. Sun et al.
dongjun region, Purple pavilion dongjun region is better than China shuiyun
stream region, and China shuiyun stream region is better than Riverfront Maple
city region. Therefore, funds permitting, buying a house in Central time zone
region could better meet the demand on all aspects.
95.4 Conclusion
AHP method is practical, and its calculation is simple and easy to be operated.
Using AHP method to analyze a variety of consideration factors in house pur-
chasing decision-making, it can guide consumers to buy houses scientifically and
rationally. This method is also applied in the purchasing decision-making of other
consumer goods, such as car buying, insurance buying, etc. (Song and Wang 2012;
Zhang and Lin 2012; Kang and Zhu 2012). In summary, AHP method has certain
guiding significance in solving similar multiobjective problems.
References
Chen D, Li D, Wang S (2007) Mathematical modeling (in Chinese). Science Press, Beijing, ch. 8,
pp 195–201
Chen J, Cheng Z, Liu Y (2011) Analytic hierarchy comprehensive evaluation of the power
transmission project (in Chinese). J Electr Power 26(5):408–412
Hu Y, Guo Y (2007) Operational research tutoria (in Chinese). Tsinghua University Press,
Beijing, ch. 13, pp 422–425
Huang D, Wu Z, Cai S, Jiang Z (2006) Emergency adaption of urban emergency shelter: analytic
hierarchy process-based assessment method (in Chinese). J Nat Disasters 15(1):52–58
Jin Q, Wang Y, Jia Z (2011) AHP and its application in recruitment (in Chinese). Comput Mod
27(8):190–192
Kang J, Zhu Q (2012) Application of AHP in project investment decision (in Chinese). Ningxia
Eng Technol 11(1):25–28
Li Y, Liu S, Zhang S (2012) Risk assessment of military logistics outsourcing based on matter-
element analysis (in Chinese). J Mil Jiaotong Univ 14(3):71–75
Mo S (2007) AHP in Decision-making in the public purchase (in Chinese). Econ Forum
21(19):49–51
Pan S, Duan X, Feng Y (2010) The application of AHP in the selection of the investment program
of military engineering (in Chinese). Mil Econ Res 31(8):40–41
Qi Y (2008) Research of performance appraisal system based on AHP (in Chinese). J Xi’an
Polytech Univ 22(3):125–128
Song Y, Wang Z (2012) Research of agricultural product logistics park location basis on AHP (in
Chinese). Storage Transp Preserv Commod 34(3):90–92
Yang R, Zhang Z, Wu Y, Lei L, Liu S (2004) An application of AHP to selection for designs and
lots of lottery tickets (in Chinese). J Chengdu Univ Inf Technol 19(3):451–457
Yuan N (2012) Evaluation of tourism resources based on AHP method in ancient villages—a case
of world heritage site of Xidi and Hongcun (in Chinese). Resour Dev Market 28(2):179–181
Zhang Y, Lin J (2012) Powder distribution scheduling based on AHP (in Chinese). J Jiamusi Univ
(Nat Sci Edn) 30(1):86–90
Zhao S (2007) The application research of AHP on comprehensive evaluation of physical fitness
(in Chinese). J Beijing Univ Phys Educ 30(7):938–940
Chapter 96
The Relations Tracking Method
of Establishing Reachable Matrix in ISM
Xue-feng He and Yue-wu Jing
Abstract Interpretative Structural Modeling (ISM) is a common Structure
modeling technology. For this technology, the establishment of reachable matrix is
a quite important step. By analyzing the relations among many factors in a system,
the relations tracking method for reachable matrix is presented in this paper.
Through a comparative analysis of several methods, the relations tracking method
is proved to be rapid and effective for establishment of reachable matrix mean-
while avoiding complex matrix operations, and can be used to enhance the
applicability of ISM.
Keywords Adjacency matrix Directed graph Interpretative structural modeling
Reachable matrix
96.1 Introduction
The Interpretative Structural Modeling (ISM) is a common structure modeling
technology and widely used in many fields (Li 2011; Thakkar et al. 2004, 2007;
Singh 2007; Kanungo and Bhatnagar 2002; Bolaños et al. 2005). In terms of the
workflow of ISM, when directed graph is obtained, the establishment of reachable
matrix is a quite important step (Wang 1998). Usually, the main methods for
establishing reachable matrix are formula method (Wang 2000), experience
X. He (&)
School of Economics and Management, Southwest University
of Science and Technology, Mianyang, China
e-mail: [email protected]
Y. Jing
Institute of Technology, China Academy of Engineering Physics,
Mianyang, China
e-mail: jywxfh163.com
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 903
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_96,
Springer-Verlag Berlin Heidelberg 2013
904 X. He and Y. Jing
dialogue method (Zhenkui 1998) and Warshall algorithm (Lipschutz and Lipson
2002; Wang and Ge 1996) and so on. In this paper, the relations tracking method
for reachable matrix is presented, compared with the other methods, this method
can avoid complex matrix operations.
96.2 Relations Tracking Method
96.2.1 Definition
Relations tracking method, just as its name implies, the relationships among many
factors in a system should be tracked firstly and then the reachable matrix may be
established. In this paper we use Fig. 96.1 as an example to illustrate.
96.2.2 Steps
The first step is to find out the direct reachable set of each node.
Direct reachable set is the element set that a node can reach directly not
including the node itself, expressed with D\i[. For example, for Fig. 96.1, the
direct reachable set of node S2 is D\2[ = {3, 4}. Similarly, D\4[ = Ø. All the
direct reachable sets are shown in Table 96.1.
The second step is to find out the tracking reachable set of each node.
Tracking reachable set is the element set that each node can reach, whether
directly or indirectly, which includes the node itself, expressed with R\i[. For
reachable matrix, this step is quite important.
The core idea of relations tracking method is as follows. Each node can be
viewed as a source node and the direct reachable set of the node will be obtained,
then each node in the direct reachable set can be viewed as a branch node which
can be used as the next level branch node, thus the tree branch of each node can be
obtained and all the nodes of tree branch constitute the tracking reachable set of
Fig. 96.1 Directed
connection diagram S1
S6
S2
S3 S5
S4
96 The Relations Tracking Method of Establishing Reachable Matrix in ISM 905
Table 96.1 The direct reachable sets
Si S1 S2 S3 S4 S5 S6
D \ i[ 2 3,4 1, 4, 5 Ø 3 1,3
each node. If a direct reachable set is an empty set, then the tracking reachable set
is the node itself. For example, in Table 96.1, the direct reachable set of S4 is
empty, then R\4[ = {4}.
The important principles: in the process of branching, if a node is repeated, then
the node should be omitted, that is the node is no longer continue to branch.
For example, for the tracking reachable set of S1, the tree branch obtained is
shown in Fig. 96.2.
Thus, R \ 1[ = {1, 2, 3, 4, 5}, similarly, R \ 2[ = {1, 2, 3, 4, 5},
R \ 3[ = {1, 2, 3, 4, 5}, R \ 5[ = {1, 2, 3, 4, 5}, R \ 6[ = {1, 2, 3, 4, 5, 6},
the tree branches of S2, S3, S5 and S6 are shown in Fig. 96.3.
The third step is to write out the reachable matrix.
The reachable matrix M is shown below.
2 3
1 1 1 1 1 0
61 1 1 1 1 07
6 7
61 1 1 1 1 07
M¼6 6 7
7
60 0 0 1 0 07
41 1 1 1 1 05
1 1 1 1 1 1
This method can avoid fundamentally complex matrix operations, and it just
needs to track the relations among nodes.
96.2.3 The Explanation of Relations Tracking Method
The relations tracking method reflects the essence of reachable matrix. In terms of
this method, the tracking reachable set can be obtained directly based on the
directed graph, so the reachable matrix obtained from relations tracking method is
just requested. From reachable matrix M it can be known that the elements in row
1, 2, 3 and 5 are same, which suggest that S1, S2, S3, S5 may form loops.
Fig. 96.2 The tree branch
1 S1 is the source node
of S1
2 D<1>={2}
3 4 D<2>={3 4}
D<3>={1 4 5},D<4>= Ø
5
906 X. He and Y. Jing
Fig. 96.3 The tree branches
of S2, S3, S5, S6 2 3
1 4 5
3 4
1 5 2
5
6
3
1 3
1 4
2 4 5
2
The relations tracking method can avoid the repeated searching because a
repeated node will be no longer continue to branch. Breadth First Search is a
method looking for the shortest path between two nodes in a directed graph (Wang
et al. 1994; Lu and Feng 2006; Yuan and Wang 2011), although the repeated
searching can be avoided, it requires that a clear hierarchical relation be estab-
lished firstly. However, in the ISM process, the hierarchical relationship is just
obtained after reachable matrix.
96.3 Comparative Analysis of Several Methods
96.3.1 Formula Method
With formula method, the reachable matrix is obtained in terms of
(A ? I)r = (A ? I)r-1 = M. A is adjacency matrix, I is unit matrix, M is
reachable matrix. A can be obtained based on directed graph.
Usually, in terms of the formula, after the sequential operation, Ai can be
obtained, that is A1=A2=…=Ar-1 = Ar = Ar+1, Ar = (A ? I)r, r B n-1, n is
order number. Because M = (A ? I)r, then M = Ar = Ar-1. Also taken
Fig. 96.1 as an example, the operation process is shown below.
96 The Relations Tracking Method of Establishing Reachable Matrix in ISM 907
2 3
0 1 0 0 0 0
60 0 1 1 0 07
6 7
61 0 0 1 1 07
A¼6
60
7
6 0 0 0 0 077
40 0 1 0 0 05
1 0 1 0 1 0
A1 ¼ A þ I
2 3
1 1 0 0 0 0
60 1 1 1 0 07
6 7
6 7
61 0 1 1 1 07
¼660
7
6 0 0 1 0 077
6 7
40 0 1 0 1 05
1 0 1 0 1 1
A2 ¼ðA + IÞ2 = A21
2 3
1 1 1 1 0 0
61 1 1 1 1 07
6 7
6 7
61 1 1 1 1 07
¼660
7
6 0 0 1 0 07
7
6 7
41 0 1 1 1 05
1 1 1 1 1 1
A3 ¼ðA + IÞ3 = A2 :A1
2 3
1 1 1 1 1 0
61 1 1 1 1 07
6 7
6 7
61 1 1 1 1 07
¼660
7
6 0 0 1 0 077
6 7
41 1 1 1 1 05
1 1 1 1 1 1
A 4 ¼ ðA + IÞ4 = A3 :A1 = A3 = M
2 3
1 1 1 1 1 0
61 1 1 1 1 07
6 7
6 7
61 1 1 1 1 07
¼6 60
7
6 0 0 1 0 07 7
6 7
41 1 1 1 1 05
1 1 1 1 1 1
908 X. He and Y. Jing
Formula method is a traditional method and used widely, but the biggest
drawback of this method is the complicated matrixes calculation which is only
acceptable when elements are quite few. Actually a system is often large with
many elements, and the relationships among elements are quite complicated.
Therefore complicated matrix calculation decreases the practicality of this method
(Tian and Wang 2003).
96.3.2 Warshall Algorithm
The steps of this algorithm are shown below. P / A. `k / 1. ´i / 1. ˆif
pik = 1, then pij / pij_pkj, j = 1,2,…,n. ˜i / i+1. If i B n, then turn to step ˆ.
Þk / k+1. If k B n, then turn to step ´, otherwise stop. A is adjacency matrix, P
is reachable matrix.
Among these steps, the step ˆ is crucial. For Fig. 96.1, the operation process is
shown as follows.
96 The Relations Tracking Method of Establishing Reachable Matrix in ISM 909
910 X. He and Y. Jing
Because the calculation process is quite tedious, so the partial steps (k = 4 and 5)
are omitted.
When k = 6, the comparing result is as follows.
k 6 i 1 p16 ¼ 0
i 2 p26 ¼ 0
i 3 p36 ¼ 0
i 4 p46 ¼ 0
i 5 p56 ¼ 0
i 6 p66 ¼ 0
96 The Relations Tracking Method of Establishing Reachable Matrix in ISM 911
The essence of Warshall algorithm is matrix comparing and matrix updating
based on the comparisons, at last the reachable matrix can be obtained. The
comparing scope of formula method is between matrix and matrix, while the
comparing scope of Warshall algorithm is limited in a matrix. However, Warshall
algorithm also involves a large number of repeated comparing. From the opera-
tions above it can be known that when k = 3, i = 5, matrix P do not change, but
according to this algorithm, the comparing of remainder must go on.
Another problem of Warshall algorithm is, reachable matrix obtained from
Warshall algorithm cannot reflect the circumstance that a node gets to itself.
However, the reachable matrix obtained from formula method can reflect the
circumstance that each node gets to itself. The reason is that the operation of
formula method is made after the adjacency matrix and unit matrix added together.
This paper argues that when reachable matrix obtained from Warshall algorithm
plus unit matrix, the problem can be solved effectively.
Compared with formula method and Warshall algorithm, relations tracking
method take the form of tree branch to track the reachable set, thus the compli-
cated operations can be avoided. Of course, with computer programming the
establishment of reachable matrix may be more efficient.
96.4 Conclusion
When establishing ISM, the calculation of reachable matrix is always quite crucial
and tedious, in order to solve this problem, the relations tracking method for
reachable matrix is presented, compared with the other methods, this method can
avoid complex matrix operations and consequently enhance the practical opera-
bility of ISM.
Acknowledgments This research is supported by the Sichuan Province Cyclic Economy
Research Center (XHJJ-1222).
References
Bolaños R, Fontela E, Nenclares A, Pastor P (2005) Using interpretive structural modeling in
strategic. Manage Decis 43(6):77–895
Kanungo S, Bhatnagar VV (2002) Beyond generic models for information system quality: the use
of interpretive structural modeling (ISM). Syst Res Behav Sci 19(6):531–549
Li C (2011) System optimize of urban rail transit based on interpretation structure model.
J Beijing Jiaotong Univ 35(5):58–63
Lipschutz S, Lipson M (2002) Discrete mathematics (in America). Science Press, Beijing
Lu Q, Feng X (2006) Search path building algorithm based on breadth-first search. Mod Comput
12:87–89
912 X. He and Y. Jing
Singh MD (2007) Knowledge management barriers: an interpretive structural modeling approach.
In: presented at the 2007 IEEE International Conference on Industrial Engineering and
Engineering Management, Singapore, 2007
Thakkar J, Kanda A, Deshmukh SG (2004) Evaluation of buyer-supplier relationships using an
integrated mathematical approach of interpretive structural modeling (ISM) and graph
theoretic matrix: the case study of Indian automotive SMEs. J Manuf Technol Manage
19(1):92–124
Thakkar J, Deshmukh SG, Gupta AD, Shankar R (2007) Development of a balanced scorecard:
an integrated approach of interpretive structural modeling (ISM) and analytic network process
(ANP). Int J Prod Perform Manage 56(1):25–29
Tian Z, Wang H (2003) The application of interpretative structural modeling in the design of
blueprint. Ind Eng Manage 4:46–50
Wang Y (1998) System engineering theory methodology and application. Higher Education
Press, Beijing, pp 35–38
Wang Z (2000) System engineering. Shaanxi Science and Technology Press, Xi’an, pp 22–24
Wang Q, Ge L (1996) The Warshal1 algorithm for computing reaehability matrix. J Xi’an Univ
Technol 12(1):80–82
Wang S, Wei S, Wensheng J (1994) Comparison of the shortest path algorithms. Syst Eng
Electron 1:43–49
Yuan H, Wang D (2011) The reachable matrix approach for the shortest path. Inform Control
40(2):202–210
Zhenkui W (1998) Notes about the methods of ISM, AHP and fuzzy appraisal. J Tianjin Univ
Comm 5:45–46
Chapter 97
The Selection of the Regional Center City
Under the Policy of Expanding Domestic
Demand
Jian-wei Cheng and Juan Shang
Abstract From the perspective of region economics, this paper makes the com-
parison on regional advantages and the selection of the regional center city
between sixteen provincial capital cities in the Central, Northwest and Southwest
Region in China, by exploiting AHP and constructing the index system with
twelve secondary indexes on the five factors of geography location, traffic facili-
ties, economics, population and human capital. The following research conclusions
are drawn. First, in Central Region, Wuhan, having the highest composite score,
should be selected as the regional center city. Second, although it is not located the
geographical center in the Northwest Region, Xi’an has the highest composite
score and each secondary index is highest, which should be selected as the regional
center city. Third, in Southwest Region, Chongqing should be selected as the
regional center city.
Keywords City Region center Expand domestic demand AHP
97.1 Introduction
With the development of the Chinese economy, Chinese government has been
transforming the mode of economic growth and expanding domestic demand. In
the twelfth five-year guideline for national economic and social development of
the People’s Republic of China, it is stated clearly to improve the pattern of
regional development and expand inland development. For example, ‘‘Build long-
term mechanism to expand domestic demand, rely on consumption, investment
and exports, and promote economic growth.’’ ‘‘Take the expansion of consumer
J. Cheng (&) J. Shang
Department of Economy and Management, Xidian University, Xi’an, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 913
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_97,
Ó Springer-Verlag Berlin Heidelberg 2013
914 J. Cheng and J. Shang
demand as a strategic focus to expand domestic demand, further release the
consumption potential of urban and rural residents, and gradually realize the
domestic market scale to be one of the largest sizes of in the world’’ (The twelfth
five-year guideline for national economic and social development of the People’s
Republic of China). Under this macroscopic environment, inland region having the
resources, labor and other advantages will actively undertake international and
coastal industrial transfer. In recent years, the domestic consumption level has
risen greatly. Per capita consumption expenditure of urban residents has increased
from 4998 yuan in 2000 to 13471 yuan in 2010, while that of rural residents has
increased from 1670 yuan in 2000 to 4382 yuan in 2010. Under the policy of
expanding domestic demand, consumer demand in China will be greatly improved.
The inland cities have more advantages of geography location over coastal cities
such as shorter transportation distance and lower logistics cost so that they can
better meet the increasing consumer demand after expanding domestic demand.
Moreover, Chinese government has accelerated the development of Central and
Western regions, which put forward the strategy of ‘‘The Rise of Central China’’
and ‘‘China Western Development’’. The regions of central and western will have
more policy support. Therefore, in the new economic situation, the cities in the
regions of central and western will have great development.
97.2 Research Situations and Problems
Many scholars have made a lot of researches on the development of the cities and
regions. Linneker and Spence (1996) concluded the positive relationship of
transport infrastructure with regional economic development. Lawson (1999)
researched the competence theory of the region. Siegfried (2002) presented that it
has closer economic associations in adjacent regions.
The comparative researches between the cities are made mainly from the
perspective of urban competitiveness. Hao and Ni (1998) researched empirically
seven cities’ competitiveness of Beijing, Tianjin, Shanghai, Dalian, Guangzhou,
Xiamen, and Shenzhen from the 21 subdivision indexes by using principal com-
ponents analysis method. Ning and Tang (2001) designed a city competitiveness
model, based Michael E. Porter and IMD national competitiveness model. Li and
Yu (2005) presented that city competitiveness is the sustainable development
capacity of a city to attract, acquire, control and convert resources, and then create
value and wealth and improve the living standard of the presidents. Wei-zhong Su,
Lei Ding, Peng-peng Jiang and Qi-yan Wang made the empirical study for the
tourism competitiveness between different cities (Ding et al. 2006; Su et al. 2003;
Jiang and Wang 2008; Wang and Wang 2009). Cheng-lin Qin and Jun-cheng made
the researches on the polycentric urban-regional structure (Qin and Li 2012; Zhu
et al. 2012). However, the studies are mostly made from the city’s current situa-
tion. There are few studies from the perspective of Regional Economics, from the
basic potential factors, such as geographical location, transport and radiation.
97 The Selection of the Regional Center City 915
According to the ‘‘Growth pole theory’’ by Francois Perroux, regional eco-
nomic development depends mainly on the minority regions or industries with
better conditions, which should be fostered to the growth pole of region economic.
The surrounding areas or related industries will be affected and promoted through
the polarization and diffusion effect of the growth pole (Luan 2008). Therefore,
under the present circumstances, to obtain the rapid development of the central and
western region, China should put limited funds into the cities with the regional
advantages in the central and western region to realize polarization and diffusion
effects so as to promote the development of the whole region. However, the central
and western regions have vast territory, in which there are so many different cities.
It is an urgent problem which city has more geography location advantages so as to
have faster development, and has the role of regional centers to radiate sur-
rounding areas. This paper tries to solve this problem. From the perspective of
region economics, this paper makes the selection research on the regional center
city for three regions, based on the empirical data of the central, northwest and
southwest region in china.
In Central, Northwest and Southwest Region, their provincial capital cities are
selected as the research object in this paper since they are the largest city of their
provinces with the greatest policy advantage and economic advantage. According
to the national statistical standards, the provincial capital cities in Central Region
are Taiyuan (Shanxi), Zhengzhou (Henan), Wuhan (Hubei), Changsha (Hunan),
Hefei (Anhui), Nanchang (Jiangxi). The provincial capital cities in Northwest
Region are Xi’an (Shaanxi), Lanzhou (Ganshu), Xining (Qinghai), Yinchuan
(Ningxia), Urumqi (Xinjiang). The provincial capital cities in Southwest Region
are Chongqing, Chengdu (Sichuan), Guiyang (Guizhou), Kunming (Yunnan),
Lhasa (Tibet).
97.3 Index selection
According to the regional economics theory, such as ‘‘agricultural location theory’’
by J.H.Thünen, ‘‘industrial location theory’’ by Alfred Weber, ‘‘transport location
theory’’ by Edgar M. Hoover, Market Location theory by August Losch, ‘‘Central
Place Theory’’ by Walter Christaller and ‘‘growth pole theory’’ by Francois Per-
roux, a regional center must consider the geographical location, population, cap-
ital, labor cost, transportation, marketing and other factors. Through
comprehensive analysis, this paper presents that a regional central city should have
the following factors. First, it is located in the geographical center. Second, it has
convenient transportation and logistics facilities, which can form great polarization
and diffusion effect. Third, it has good economic base, which not only reflects a
region economic strength, competitiveness and consumption ability, but also a
regional market environment and commercial atmosphere. Fourth, it has certain
population. The place with numerous populations can provide a large number of
labor forces as well as potential consumers. Fifth, it has the advantage of human
916 J. Cheng and J. Shang
Table 97.1 Influence index of the selection of regional center city
First index Secondary index Symbols
Geography Distance sum between the city and other cities X1
location
Traffic facilities Railway mileage per ten thousands square kilometers in the province X2
Highway mileage per ten thousands square kilometers in the X3
province
Economics City GDP X4
Total retail sales of consumer goods X5
Per capita disposable income of urban residents X6
Per capita net income of rural residents X7
Population Total population number of the province X8
Employment number of the province X9
Human capital Total number of college student in the province X10
Average number of people having the higher education per ten X11
thousands persons
Average number of the college student per ten thousands persons X12
capital. In the knowledge economy era, human capital is the main factor reflecting
competition ability. In this paper, considering the above five factors and the data
collection, twelve secondary indexes are selected, shown in Table 97.1.
As shown in Table 97.1, Geography location is compared by calculating the
distance sum between the city and other cities. The city with the smallest distance
sum is located relatively in the center, which can bring the polarization and dif-
fusion effects. The city area is small.
Railroad and highway is distributed in a mesh structure in the surrounding area.
The railway and highway conditions within the city can not reflect a city traffic
convenience. Moreover, the population and human capital have mobility. The
quantity within city can not reflect the city’s regional competitive advantage.
Therefore, the traffic facilities, the population and the human capital select the
province data as the index. In order to eliminate the influence of provincial area,
the railway mileage and the highway mileage per ten thousands square kilometers
are selected as the index reflecting a city traffic convenience. The total population
number and the employment number of the province are selected as the population
index. Relatively, the total number of college student in the province is selected to
reflect the total situation of human capital. The average number of people having
the higher education per ten thousands persons is selected to reflect the average
situation of human capital, as well as the average number of the college student per
ten thousands persons. The GDP, total retail sales of consumer goods, per capita
disposable income of urban residents and per capita net income of rural residents
of that city are selected as the index of economics.
97 The Selection of the Regional Center City 917
97.4 Model Selection and Data Comprehensive Processing
97.4.1 Index Weight
Weight is the coefficient of index importance. Calculating the weights of the
indexes commonly uses these methods, such as the subjective weighting method,
the objective weighting method and the Analytic Hierarchy Process. Analytic
Hierarchy Process (AHP) is proposed in the mid 1970s by TL Saaty, a professor of
the University of Pittsburgh. It decomposes the complex question into several
component factors, which is divided further into the target layer, rule layer and
index layer so as to form a model with multi-objective and multi-level and ordered
hierarchical structure. The comparison of each factor is made to determine the
relative importance of the factors. The steps are the establishment of the hierar-
chical structure, constructing judgment matrix, hierarchical ranking and consis-
tency check (Peng et al. 2004).
AHP is used to determine the index weigh in the paper. The index comparison
values of judgment matrix are determined after comprehensive consideration of
the theoretical analysis, statistics data and expert score. Target layer A, rule layer
B and index layer P are made as shown in Table 97.2. In order to facilitate
comparison and reduce the occupied space, the analysis result after calculating is
attached to the last two columns in Table 97.2.
According to AHP, the largest eigenvalue kmax and the corresponding eigen-
vector x can be calculated by constructing the judgment matrix. For the judgment
matrix A-B, shown in Table 97.3, we can conclude that kmax ¼5:0304; x ¼
ð0:375; 0:215; 0:215; 0:121; 0:074ÞT ; CI ¼ kmax n
n1 ¼ 0:0076. And the judgment
coefficient CR ¼ CI/RI ¼ 0:0068\0:10, which means that the consistence of
judgment matrix is satisfied. Therefore, we get that the weight value of B1, B2, B3,
B4, and B5 for A are 0.375, 0.215, 0.215, 0.121, 0.074, in which B1, geographical
location index, is the greatest. Similarly, we can construct the judgment matrix B3-
P (Table 97.4) in which the judgment coefficient CR ¼ 0:0045\0:10 and the
consistence of judgment matrix is satisfied. The weight value of B31 , B32 , B33 and
B34 for B3 are 0.423, 0.227, 0.227, 0.123. The judgment matrix B5 P (Table 97.5)
can be constructed, in which the judgment coefficient CR ¼ 0:0083\0:10 and the
consistence of judgment matrix is satisfied. The weight value of B51 , B52 and B53
for B5 are 0.539, 0.297 and 0.164.
B2 and B4 have only two secondary indexes, which are given that the weight
values are both 0.5. So CI ¼ RI ¼ 0 .We make a consistence test for the total
hierarchy, which is the weightP value of the
P lowest index for the overall target. The
judgment coefficient CR ¼ ni¼1 ai CIi = ni¼1 ai RIi ¼ 0:0052\0:10, in which ai is
the weight value of Bi for A and CIi is the consistence coefficient of Bik for Bi, so
the consistence of judgment matrix is satisfied. The final result is shown as
Table 97.2, of which the last column is the weight value of each index for A.
918
Table 97.2 Index weight value by AHP
Target layer A Rule layer B Weight value of B Index layer P Weight value of P Weight value of P
for A for B for A
Composite Geography 0.375 B11for distance sum X1 1.000 0.375
score A location B1
Traffic facilities 0.215 B21 for railway mileage X2 0.500 0.108
B2
B22 for highway mileage X3 0.500 0.107
Economics B3 0.215 B31 for city GDP X4 0.423 0.091
B32 for total retail sales X5 0.227 0.049
B33 for disposable income X6 0.227 0.049
B34 for net income X7 0.123 0.026
Population B4 0.121 B41 for total population X8 0.500 0.061
B42 for employment number X9 0.500 0.061
Human capital B5 0.074 B51 for total number of college student X10 0.539 0.040
B52 for average number of people having the higher 0.297 0.022
education X11
B53 for average number of the college student X12 0.164 0.012
J. Cheng and J. Shang
97 The Selection of the Regional Center City 919
Table 97.3 Judgment matrix A-B
A B1 B2 B3 B4 B5 x
B1 1 2 2 3 4 0.375
B2 1/2 1 1 2 3 0.215
B3 1/2 1 1 2 3 0.215
B4 1/3 1/2 1 2 0.121
B5 1/4 1/3 1/3 1/2 1 0.074
kmax ¼ 5:0304; CI ¼ 0:0076; RI ¼ 1:12; CR ¼ 0:0068\0:10
Table 97.4 Judgment matrix B3-P
B3 B31 B32 B33 B34 x
B31 1 2 2 3 0.423
B32 1/2 1 1 2 0.227
B33 1/2 1 1 2 0.227
B33 1/3 1/2 1 0.123
kmax ¼ 4:0123; CI ¼ 0:0041; RI ¼ 0:9; CR ¼ 0:0045\0:10
Table 97.5 Judgment matrix B5-P
B5 B51 B52 B53 x
B51 1 2 3 0.539
B52 1/2 1 2 0.297
B53 1/3 1/2 1 0.164
kmax ¼ 3:0096; CI ¼ 0:0048; RI ¼ 0:58; CR ¼ 0:0083\0:10
97.4.2 Data Dimensionless Processing
Because each index has different units and dimensions, dimensionless processing
for these indexes must be done in order to compare and summarize. The normal
methods of index standardization are range transformation method, linear pro-
portional method, normalized method, standard sample transformation method,
vector normalization method and taking reciprocal. In the paper, linear propor-
tional method is used on the standardization
and summation of these indexes.
In the decision matrix X ¼ xij mn ;
For the positive index, given xj ¼ max xij 6¼ 0, then
1im
xij
yij ¼ ; ð1 i m; 1 j nÞ ð97:1Þ
xj
For the reverse index, given xj ¼ min xij 6¼ 0, then
1im
xj
yij ¼ ; ð1 i m; 1 j nÞ ð97:2Þ
xij
920 J. Cheng and J. Shang
Y ¼ yij mn is called as the linear proportional standard matrix.
In the twelve indexes, the distance sum is the reverse index, which means the
higher the distance sum is, the lower the transform score is. So the index of
distance sum should be standardized by the Eq. (97.2). The other eleven indexes
are the positive index, which should be standardized by the Eq. (97.1). It must be
noted that the maximum and minimum values are the respective region value
within central, southwest and northwest regions, not the total value.
97.4.3 Evaluation Model
Linear weighting method is used as the evaluation model of.
P
12
Yj ¼ bi Xij ði ¼ 1; 2; . . .12; j ¼ 1; 2; . . .16Þ
i¼1
In which Yj is ‘‘j’’ city’s composite score, bi is the weight value of ‘‘i’’ index,
Xij is the dimensionless value of ‘‘i’’ index of ‘‘j’’ city. According to the weight
value in the Table 97.2, the final formula is:
Y ¼0:375X1 þ 0:108X2 þ 0:107X3 þ 0:091X4 þ 0:049X5 þ 0:049X6
þ 0:026X7 þ 0:061X8 þ 0:061X9 þ 0:040X10 þ 0:022X11 þ 0:012X12
97.4.4 Calculation Results and Description
Considering the data consistency, all the data are in 2010. The data of X1, X2, X7,
X8, X9, X10 and X12 are from the ‘‘2011 China Statistical Yearbook’’. The data of
X3, X4, X5, and X6 are from Statistical Communiqué on the 2010 National Eco-
nomic and Social Development on these 16 cities. The original data are not listed
for saving the space. By calculating, the final result is shown on the Table 97.6, in
which the Y column is the final composite score.
As shown in Table 97.6, by considering and calculating comprehensively the
twelve secondary indexes, it is concluded that the composite score rank of the six
cities in Central Region are Wuhan (87.52), Zhengzhou (85.75), Changsha (74.83),
Hefei (73), Nanchang (64.03) and Taiyuan (54.38), which means Wuhan, the
highest score city, has the greatest advantages to become the regional center city in
Central Region. But the gap between Zhengzhou and Wuhan is very small, only
1.77. After analysis of various secondary indexes, Both of Wuhan and Zhengzhou
have five indexes being the highest score (100). By comparing the secondary
respectively, Wuhan has the advantages of better geography location and eco-
nomics, while Zhengzhou has the advantages of better traffic and more population
and labor force. Zhengzhou is predominant on the total number of human capital.
97
Table 97.6 The composite scores of sixteen cities
Region City X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 Y
Central Taiyuan 52.3 93.4 57.2 32.2 32.7 75.6 71.5 38.0 27.6 38.6 91.5 73.4 54.38
region Zhengzhou 79.1 100.0 100.0 72.5 66.5 82.8 86.7 100.0 100.0 100.0 67.1 63.3 85.75
Wuhan 100.0 70.5 75.6 100.0 100.0 91.2 78.0 60.9 51.6 89.0 100.0 100.0 87.52
Changsha 72.9 68.0 73.3 82.4 71.8 100.0 100.0 69.9 66.3 71.9 79.7 70.6 74.83
Hefei 85.5 79.6 72.9 49.0 33.3 83.5 66.9 63.3 63.7 64.5 70.3 63.4 73.00
Nanchang 80.0 66.2 57.4 40.0 30.3 80.1 67.6 47.4 38.2 56.0 71.8 74.4 64.03
Northwest Xi’an 69.3 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 99.3 100.0 88.58
The Selection of the Regional Center City
region Lanzhou 100.0 28.9 39.0 33.9 33.8 63.2 59.2 68.5 73.4 41.1 70.7 58.7 66.74
Xining 95.9 13.0 12.0 19.4 14.4 63.3 71.2 15.1 15.1 4.8 81.0 34.9 50.33
Yinchuan 89.6 94.9 47.3 23.5 14.0 76.8 79.5 16.9 16.7 8.6 86.1 58.2 62.54
Urumqi 39.0 12.8 12.8 40.4 35.0 64.7 96.3 58.5 43.7 27.1 100.0 45.7 38.50
Southwest Chongqing 99.4 100.0 100.0 100.0 100.0 91.1 64.3 35.9 38.3 48.1 100.0 100.0 88.77
region Chengdu 100.0 43.2 38.7 70.3 84.0 100.0 100.0 100.0 100.0 100.0 77.2 74.2 83.11
Guiyang 95.6 67.1 60.6 14.2 16.8 79.7 72.8 43.2 48.1 29.8 61.2 46.0 66.17
Kunming 90.5 37.1 37.4 26.9 36.8 90.6 70.8 57.2 56.3 40.4 66.9 57.7 63.18
Lhasa 48.6 2.6 3.6 2.3 3.1 79.6 61.0 3.7 3.5 2.9 63.7 56.9 27.35
921
922 J. Cheng and J. Shang
Wuhan is predominant on the average situation of human capital. Compared
comprehensively the six cities in Central Region, it can be concluded that the
scores was hierarchy distribution and the overall development of the Central
Region is relatively balanced.
The composite score rank of the five cities in Northwest Region are Xi’an
(88.58), Lanzhou (66.74), Yinchuan (62.54), Xining (50.33), Urumqi (38.50).
Xi’an is not located on the center of the Northwest Region, but its composite score
is far higher than the second city, Lanzhou, and it has ten indexes to be the highest
score (100). Xi’an has the greatest advantages to become the regional center city in
Northwest Region. Similarly, compared comprehensively the five cities in
Northwest Region, it can be concluded that the scores was of great difference. The
development gap between the cities of the Central Region was very large.
The composite score rank of the five cities in Southwest Region are Chongqing
(88.77), Chengdu (83.11), Guizhou (66.17), Kunming (63.18), Lhasa (27.35),
which means Chongqing has the greatest advantages to become the regional center
city in Southwest Region. But the gap between Chengdu and Chongqing is small,
only 5.66. From each secondary index, Chongqing and Chengdu have both six
indexed being the highest score. Chongqing has the great advantage of traffic
facilities. Correspondingly, Chengdu has the advantage of population and labor
force. Similarly, compared comprehensively the five cities in Southwest Region, it
can be concluded that the scores was also of great difference. The overall devel-
opment of the Southwest Region is obvious uneven.
97.5 Conclusion and Suggestion
Under this macroscopic environment of expanding domestic demand, the cities in
Central and West Region are facing tremendous development opportunity. From
the perspective of region economics, this paper makes the comparison on regional
advantages and selection of the regional center city between sixteen provincial
capital cities in the Central, Northwest and Southwest Region in china, by con-
structing the index system with twelve secondary indexes on the five factors of
geographical location, traffic facilities, economics, population and human capital
and exploiting AHP. The following conclusions are drawn.
Wuhan, Xi’an, Chongqing have the more regional advantages to meet
increasing consumer demand. Chinese government should position Wuhan as the
regional center city in Central Region, Xi’an as the regional center city in
Northwest Region, and Chongqing as the regional center city in Southwest Region,
giving the three cities more policy and funds support. The three cities should be
built to be the regional growth pole so as to enlarge the polarization and diffusion
effect and drive the three regional developments.
At the same time, Wuhan should take full advantages of its economic and
human capital, and concentrate on capital industry and science and technology
industry. Xi’an has long distance with the other cities in the Northwest Region.
97 The Selection of the Regional Center City 923
The traffic of the Northwest Region and the regional gap is worst in the three
regions, which means the polarization and diffusion effect of Xi’an for Northwest
Region is weaker. Xi’an should strengthen the connection so as to build the
Northwest growth pole in Asia-Europe continental bridge and drive the develop-
ment of Northwest Region. The composite scores of Chongqing and Chengdu are
both on the top in the southwestern region, much greater than other provincial
capital cities in the Southwest Region. Furthermore, the two cities are much close
on the geographical location, Chongqing and Chengdu should reinforce the
cooperation and become the regional growth dual pole so as to drive the devel-
opment of Southwest Region.
Acknowledgments Based on empirical data of the Central, Northwest and Southwest Region in
China
References
Ding L, Wu X, Wu Y, Ding J (2006) A system of evaluation indicator for the urban tourism
competitiveness. Econ Geogr 26(5):511–515 (Chinese)
Hao S, Ni P (1998) The study on the China city competitiveness: a case of several cities. Econ Sci
20(3):50–56 (Chinese)
Jiang P, Wang X (2008) Research on competitiveness of coastal tourism cities in China: an
empirical study of Dalian, Qingdao, Xianmen and Sanya. Tour Sci 28(10):12–18 (Chinese)
Lawson C (1999) Towards a competence theory of the region. Camb J Econ 23(2):151–166
Li N, Yu T (2005) On urban competitiveness and the methods, process of evaluation. Hum Geogr
20(3):44–48 (Chinese)
Linneker B, Spence N (1996) Road transport infrastructure and regional economic development:
The regional development effects of the M25 London orbital motorway. J Transp Geogr
4(2):77–92
Luan G (2008) Regional economics. Tsinghua University Press, Beijing, pp 18–39 (Chinese)
Ning Y, Tang L (2001) The concept and indicator system of urban competitive capacity. Urban
Res 16(3):19–22 (Chinese)
Peng G, Li S, Sheng M (2004) AHP in evaluating government performance: determining
indicator weight. China Soft Sci 19(6):136–139
Qin C, Li H (2012) Progress of studies on polycentric in western countries. Hum Geogr
27(1):6–10 (Chinese)
Siegfried J John (2002) The economics of regional economics associations. Q Rev Econ Finance
42(1):1–17 (Chinese)
Su W, Yang Y, Gu C (2003) A study on the evaluation of competitive power of urban tourism.
Tour Tribune 15(3):39–42 (Chinese)
Wang Q, Wang D (2009) Construction and application of Chinese city tourism competence
evaluation system. Stat Res 26(7):49–54 (Chinese)
Zhu J, Zhang M, Son C, Tang J (2012) Polycentric urban-regional structure and its coordinal
symbiosis in Wuhan Urban Circle. Urban Stud 19(3):7–14 (Chinese)
Chapter 98
A Research on Mine Pressure Monitoring
Data Analysis and Forecast Expert System
of Fully Mechanized Coal Face
Hong-bing Qiao, Hai-long Xu, Ao-shuang Pang, Chang-dong Zhou
and Yi-lun Wang
Abstract Based on the research of mine pressure monitoring system of fully
mechanized coal face, we have designed a set of software, whose perfect function
is mine pressure monitoring data analysis. The data from mine pressure monitoring
system could be collected and analyzed scientifically and effectively. We have
accomplished the establishment of four databases combined with safety production
condition of working face and the result of data analysis. Besides, we have
designed a mine pressure forecast model and forecast expert system with the four
established databases. It is of practical significance for safety production of
working face.
Keywords Data analysis Database Forecast expert system Mine pressure
monitoring
98.1 Introduction
In coal industry, the roof accident has been the major security hidden danger for
coal mine workers for many years (Qian and Shi 2003). According to statistics, the
coal roof accidents make up 42 % of various accidents. The roof accidents are
threatening the life of coal mine workers seriously. One of the main reasons, which
cause all kinds of mine disasters, is mine pressure appearance (Cen 1998).
Therefore, it is particularly important to carry out pressure monitoring for
hydraulic support which is the roof support equipment. In recent years, with the
development of science and technology, the constant improvement of mining
technology and strengthening the requirement of safety production in coal mine,
H. Qiao H. Xu (&) A. Pang C. Zhou Y. Wang
School of Mechanical Electrical and Information Engineering,
China University of Mining and Technology Beijing, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 925
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_98,
Springer-Verlag Berlin Heidelberg 2013
926 H. Qiao et al.
mine pressure monitoring has been carried out extensively to some extent (Sun
et al. 2006). We have designed a set of software, which owns perfect function, for
mine pressure monitoring data analysis of fully mechanized coal face. It plays a
positive role in the research on appearance regularity of mine pressure and the
prevention of coal mine roof accidents. Based on date analysis we have also
designed mine (Zhang 2004).
98.2 Selection of Data Analysis Software Development
Environment on Mine Pressure Monitoring
98.2.1 Design of the Software Interface
Software interface is designed by using Configuration King which owns the fea-
tures of strong adaptability, extensive application, easy extension, low cost and
short development cycle. Besides, it owns rich graphics library and all kinds of
communication interface. It is compatible with other programming languages, and
it could be extended using Visual Basic 6.0 and Visual C++. It has the alarm
window. Meanwhile, it could generate all kinds of reports and real-time trend
curve expediently.
98.2.2 Design of Data Processing and Analysis
Data processing and analysis are designed based on Matlab which is a business
mathematics software developed by MathWorks company of American. Matlab is
the senior calculation language used for algorithm development, data analysis and
numerical calculation. Besides, this software, with kind working platform
and programming environment, has enhanced functionality of data analysis and
graphics processing.
98.2.3 Design of Databases
Database is established using SQL database tool which is an operation command
set specially designed for the establishment of database. SQL is a kind of database
language with all ready function. When using it, users only need to send out the
command of ‘‘what to do’’, without considering ‘‘how to do’’. It has become the
basis of the database operation with features of powerful function, easy learning
and convenient using.
98 A Research on Mine Pressure Monitoring Data Analysis 927
Fig. 98.1 Main function
User login module
modules of analysis software
Data management module
Function
module
design Calculation and analysis module
Of the
software
Inquiring and display module
Monitoring and daily reporting module
98.3 Function Module Design of Data Analysis Software
on Mine Pressure Monitoring
Main function modules of data analysis software on mine pressure monitoring are
shown in Fig. 98.1: Main function modules of data analysis software on mine
pressure monitoring are shown in Fig. 98.1 (Gong and Wang 2011):
98.3.1 User Login Module Design
Different users have different operating authority, which is also the key to guar-
antee the normal operation of the system. User login module is the necessary
channel through which users enter the main application program. Here users need
to complete the information authentication. We need to understand the process of
user information. The flow diagram is shown in Fig. 98.2.
98.3.2 Data Management Module Design
See Fig. 98.3
98.3.3 Calculation and Analysis Module Design
According to the observation, we complete the conventional calculation work,
such as pillar load, under remove and the closer quantity between top and bottom
floor.
928 H. Qiao et al.
Login screen display
The user inputuser name and password
To judge the user name and password
Y
Y
N
Is the user name and password correct?
whether try again when error prompted
Login success and main interface display
N
N
Quit program
Fig. 98.2 Data management module design
Fig. 98.3 Data management
module design Data superaddition
Record deletion
Record insertion
Data management Record amendment
module
Unconditional output
Conditional output
Data storage
98 A Research on Mine Pressure Monitoring Data Analysis 929
Histogram
Statistical analysis
Frequency analysis
Calculation and analysis
Conventional calculation
module
Unitary regression
Regression analysis
Multiple regression
Fig. 98.4 Calculation and analysis module design
According to the process, every interval percent can be calculated when the
extreme value and the number of intervals given. On the basis of it, the probability
distribution examination and the characteristic value calculation could be carried
out.
Unitary linear or nonlinear and multivariate linear or nonlinear regression
analysis could be taken on the basis of shared database. It also adds the multi-
variate stepwise regression. The program can automatically screen factors to
ensure the effectiveness of the regression equation. Besides, the program could
carry on weight analysis of influence factors (Mu et al. 2012).
Calculation and analysis method of module is shown in Fig. 98.4.
98.3.4 Inquiring and Display Module Design
Any database could be queried in this module. In a single query condition, people
can find out the maximum, the minimum and the average. The record number and
percentage which meet certain condition could be found out. Search results can
then be displayed and printed according to user requirements. This module con-
tains content as shown in Fig. 98.5.
98.3.5 Monitoring and Daily Reporting Module Design
It needs strict scientificalness and clear pertinence in monitoring and daily
reporting module. People could make bold discussion about newfound problems
and expound their own and academic points of view. The so-called scientificalness
means processing all the observation data using theory of probability and
930 H. Qiao et al.
Inquiring record number
Single condition inquiring
Complex conditions inquiring
Inquiring and display
module
Inquiring results display
Printing inquiring results
Switching database
Fig. 98.5 Inquiring and display module design
Support quality evaluation
Supporting effect analysis
Monitoring and
Monitoring index
daily reporting
confirmation
output
Mine pressure forecast
Super pressure alarm
display
Fig. 98.6 Monitoring and daily reporting module design
mathematical statistics method, but not choosing optionally or analyzing using the
wrong method. The so-called pertinence means summarizing the achievements
practically and realistically according to concrete contents, purpose and the used
instruments of this observation, then obtaining the mine pressure characteristics,
the roof control method and the improvement way of this working face or tunnel
(Cao 2011). This module contains several contents as shown in Fig. 98.6.
98 A Research on Mine Pressure Monitoring Data Analysis 931
98.4 The Establishment of Database
The database establishment not only includes all of the observed contents and the
relevant information of working face production, but also meets convenient data
processing (Li et al. 2002). Four original databases are established (Zhao et al.
2011; Tan 2000).
98.4.1 Observation Line Database
We record every observation line which is set every ten hydraulic supports.
Observation contents include the support working state and roof bolting effects.
98.4.2 Daily Observation Database
The record, which is made for every day or every shift, includes the position of
winning machine, the setting load of stanchion and the end resistance.
98.4.3 Stanchion Resistance Database
This database, taking mining face cycle for a record, could carry out various
calculations of the stanchion resistance.
98.4.4 Basic Parameters Database of Working Face
This section contains the basic parameters, such as the length of the working face,
the slope angle, the coal thickness, the coal hardness and the surrounding of
working face.
98.5 The Establishment of Forecast Expert System
It is an indispensable security technology to take mining pressure prediction.
Realizing automation and intellectualization is the urgent task which would turn
the technology into high-tech products (Zhou et al. 2011).
932 H. Qiao et al.
According to the results of data analysis and coal mine safety production
experience for many years, this system explores with mine pressure prediction
mechanism deeply. Mine pressure prediction expert system is on the basis of mine
pressure data analysis, prediction model, reasoning strategy, monitoring method
and the data processing method. The core of the system is the knowledge base and
reasoning machine. The knowledge base is the set of mine domain knowledge that
mine pressure prediction needs, including basic facts, rules and other relevant
information (Huang et al. 2008). The knowledge representation is various,
including framework, rules, semantic networks and so on. The knowledge,
deriving from field experts, is the key to expert system capacity, namely the
quality standard of the expert system depends on the quality and quantity of the
knowledge. As the knowledge base and the expert system are independent of each
other, user can change and perfect knowledge contents to improve the performance
of the system. While reasoning machine, which is actually proceeded to explain
the knowledge, is the executive core part of problem solving. According to the
semanteme of knowledge, it conducts interpretive execution for the knowledge
found through certain strategy and records the results in the appropriate space of
the dynamic base.
Mine pressure forecast expert system model is shown in Fig. 98.7.
The establishment of the system should include the following three factors:
• It should possess the knowledge of experts who are in the mine pressure forecast
field.
• It can simulate specialistic thought.
• It also possesses expert-level problem-solving.
The process of establishing mine pressure forecast expert system could be
called ‘‘the knowledge engineering’’. In other words, the software engineering
thought is applied to design system based on the knowledge. It includes several
aspects below (Jiang et al. 1995):
• Knowledge acquisition.
• The selection of appropriate knowledge representation.
• Software design.
• The engineering accomplishment with the right computer programming
language.
The system, which sets up a bridge between automatic forecast and forecast
mechanism, is able to predict mine pressure rule and display overpressure alarm. It
plays a significant role in safety production, roof management, reasonable selec-
tion of hydraulic support and optimized support design.
98 A Research on Mine Pressure Monitoring Data Analysis 933
Predicting regular
Guiding roof Selection of Guiding equipment
pattern of mine
management hydraulic support operation
pressure
User interface with big screen Man-machine dialogue system
Model collection Consultation and Knowledge
Data collector
system interpretation system acquisition system
Model Automatic reasoning Knowledge Data
management machine management management
Establishing an Knowledge
Model base effective decision Database
base
model
Information Comprehensive Prediction expert system
from sensor data analysis based on four databases
Fig. 98.7 Mine pressure forecast experts system diagram
98.6 Conclusion
a. User Login Module Design (1) Through the mine pressure monitoring data
analysis of the fully mechanized mining face, people could grasp mine pressure
distribution, working resistance of hydraulic support, pressure cycle, caving
span, first pressure span and so on. It has theoretical significance in safety
production.
b. The result of data analysis could provide important basis for the correct
selection of hydraulic support and occupy a significant role in giving full play
to the performance of mining equipment.
934 H. Qiao et al.
c. Mine pressure forecast expert system could predict roof accident effectively.
People would move support before the peak value comes and find hidden
danger of the support, such as tilt, roof exposure, sealing performance and so
on. The application of mine pressure monitoring system is an important mea-
sure to avoid blindness and empiricism of roof management. And it could also
provide reliable basis for working out mining regulations of similar coal seam.
References
Cao J (2011) Mine pressure monitoring and data analysis of roadway in working face with soft
coal seam and great height, (in Chinese). Coal Mine Support 04:19–24
Cen C (1998) Stope roof control and monitoring technology, (in Chinese). Press of China
University of Mining and Technology, Xuzhou
Gong L, Wang Q (2011) Research and development of data analysis system of mine pressure
monitoring, (in Chinese). West-China Explor Eng 05:167–169
Huang Y, Yao Q, Ding X, Zhang L, Wang Y, Li L (2008) Mining pressure prediction of upper
roof in condition of hard roof, (in Chinese) 12:56–57
Jiang F, Song Z, Song Y, Yang Y, Zhao W, Qian M (1995) Basic study of expert system for
predicting weighting in coal face, (in Chinese). J China Coal Soc 20(3):225–228
Li H, Long Y, Zhou D (2002) The application of database in data processing, (in Chinese). Meas
Tech 01:42–44
Mu H, Wang F, Mu Y (2012) Mine pressure monitoring and data analysis of 1401E roadway in
Zhaizhen coal mine, (in Chinese). Shandong Coal Sci Technol 01:189–190
Qian M, Shi P (2003) Mine pressure and stratum control, (in Chinese). Press of China University
of Mining and Technology, Xuzhou
Sun Y, Wen Z, Zhang H, Liu Z (2006) Dynamic forecasting and roof support quality monitoring
of rigid roof located, (in Chinese). Coal Technol 25(9):64–65
Tan H (2000) C language programming, (in Chinese). Tsinghua University Press, Beijing
Zhang K (2004) The application of pressure monitoring system of fully mechanized support in
coal mine roof management, (in Chinese). Mining Industry Institute of Shandong University
of Science and Technology
Zhao L, Zhang B, Xiao K, Lu X (2011) Research of C language programming method with zero
defect, (in Chinese). Softw Eng 01:50–52
Zhou X, Bai C, Lin D, Wang Z (2011) Research on mechanisms of roof pressure prediction, (in
Chinese). J China Coal Soc 36(S2):299–303
Chapter 99
The U-Shaped Relationship Between
Corporate Social Disclosure
and Corporate Performance: Evidence
from Taiwan’s Electronics Industry
Chin-Shien Lin, Ruei-Yuan Chang and Van Thac Dang
Abstract This study investigates the corporate social disclosure (CSD) of the
electronics industry in Taiwan and examines the relationship between corporate
social responsibility disclosure and corporate economic performance. The annual
reports of 600 out of 929 companies on the Taiwan Market Observation Post
System and in the Taiwan Economic Journal database in 2009 were hand-col-
lected. The results reveal the practice of corporate social responsibility disclosure
in Taiwan’s electronics industry. More specifically, this paper finds that the
relationship between corporate social responsibility disclosure and corporate
economic value-added is best illustrated by the U-shaped curve. On the one hand,
the findings of this study help to build knowledge of corporate social responsibility
in Taiwan’s business companies. On the other hand, the results of this study
somewhat explain the inconsistent findings of the relationship between corporate
social responsibility disclosure and corporate performance in the previous litera-
ture. This study provides important implications for both academics and
practitioners.
Keywords Corporate social disclosure Economic value-added U-shaped
Content analysis
C.-S. Lin (&)
Department of Business Administration, National Chung
Hsing University, Taiwan, Republic of China
e-mail: [email protected]
R.-Y. Chang
Department of Finance, Providence University, Taiwan,
Republic of China
e-mail: [email protected]
V. T. Dang
Department of Business Administration, National Chung
Hsing University, Taiwan, Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 935
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_99,
Ó Springer-Verlag Berlin Heidelberg 2013
936 C.-S. Lin et al.
99.1 Introduction
In a modern society, business firms have been viewed as open systems that interact
with and are integral parts of their environment. Business firms not only obtain
material, financial, and human resources from the outside environment, but also
gain support and legitimacy from their stakeholders and the whole society (Ott
et al. 2010). Business firms are now facing increasing pressure from stakeholders,
regulators and society as the latter demands more comprehensive and transparent
information regarding the former’s financial soundness, employee policies, envi-
ronmental policies, social responsibility involvement, etc. As a result, an
increasing number of business firms are now disclosing their corporate social
responsibility activities (Bebbington et al. 2007). Although corporate social
responsibility disclosures (CSD) has been the subject of substantial academic
research for the last two decades, CSD has so far remained mainly a phenomenon
of developed countries in Western Europe, the USA, and Australia (Bebbington
et al. 2007; Patten 2002). In fact, very few papers have discussed this issue in the
context of developing countries, and the few ones that have focused on Hong
Kong, Korea, Malaysia, Singapore, and some South African countries (Bebbington
et al. 2007; Choi 1998; Tsang 1998). In Taiwan, to our knowledge, only some
limited empirical studies have concentrated on this issue. Because of this gap in
the empirical studies, it is very difficult to know the practices of CSD in Taiwan.
For this reason, our first purpose is to examine the CSD practices of companies in
Taiwan.
In addition, the relationship between CSD and firm performance remains
inconsistent in the previous literature (Garay and Gonzalez 2010). One view is that
increased disclosure will produce costs of equity capital and negatively impact
firm performance (Dhaliwal et al. 2009). A contradictory view holds that by
increasing their social and environmental disclosures, firms will enhance their
reputations (Armitage and Marston 2008), which in turn helps them to gain support
and legitimacy from stakeholders and society. Therefore, the undefined relation-
ship between CSD and firm performance in previous research is one of the most
important issues for future research (Richardson and Welker 2001).
Recently, a study by Wang et al. (2008) find that the relationship between
corporate philanthropy and firm financial performance follows an inverted U-
shape. Their findings provide an important reference for our arguments in this
study. Additionally, based on the private costs theory and the agency theory, when
firms disclose more social and environmental information to the public, they will
incur substantial direct and indirect costs (Barnett and Salomon 2006). These costs
would suggest a negative relationship between CSD and firm performance.
However, according to the stakeholder theory, the increased disclosure of firms
will reduce information asymmetry, thus lowering the estimation risk of the dis-
tribution of returns, which consequently enables firms to gain the trust of investors
and stakeholders. As a result, the firms will obtain resources controlled by these
stakeholders, such as human capital, financial capital, social capital, etc.
99 The U-Shaped Relationship 937
(Donaldson and Preston 1995). As a result of these benefits, we can expect a
positive relationship between CSD and firm performance. The costs and benefits of
both of these effects result in a rough, non-linear relationship between CSD and
firm performance. This paper contributes to the literature by testing empirically the
existence of a U-shaped relationship between CSD and firm performance. To
address the aforementioned gap in previous empirical studies, a sample of annual
reports of 600 out of 929 electronics companies was hand-collected on the Taiwan
Market Observation Post System and in the Taiwan Economic Journal database in
2009 so as to examine the practice of CSD in Taiwan and to test the relationship
between CSD and firm performance. The findings of this study provide a better
understanding of the practices of CSD in Taiwan. More specifically, the finding of
a U-shaped relationship between CSD and economic value-added somewhat helps
to explain the inconsistent results in the previous literature.
99.2 Literature Review and Hypothesis Development
Over the past two decades, CSD has been one of the most commonly discussed
issues in many developed economies (Bebbington et al. 2007). However, previous
findings on the relationship between CSD and firm performance have been
inconsistent (Garay and Gonzalez 2010). According to Gray et al. (1993), it is
unnecessary to report social and environmental disclosures because of the absence
of any demand for such information and the absence of any legal requirement for
CSD. If firms disclose their CSR in such circumstances, the costs would outweigh
the benefits (Solomon and Lewis 2002). It would also be irrational for firms to
disclose any information harmful to themselves. Consequently, a burden cost will
have a negative impact on firm performance (Dhaliwal et al. 2009).
However, a number of scholars have suggested that by making social and
environmental disclosures in their annual reports, firms enjoy multiple benefits
(Armitage and Marston 2008; Godfrey 2005), such as enhanced firm reputation
(Armitage and Marston 2008), effective response to pressure and prediction of
future environmental regulations (Blair 2000), reduction in information asymmetry
and boost in investor interest (Gray et al. 1995), as well as establishment and
maintenance of good stakeholder relationships, which are conducive to gaining
support and legitimacy from stakeholders and society (Milne and Patten 2002).
Yet still, some other researchers claim that there is no significant relationship
between CSD and firm performance (Freedman and Wasley 1990). According to
their findings, it is likely that the relationship between CSD and firm performance
is more complex, and not simply a direct one as proposed by previous studies
(Wang et al. 2008). For this reason, it is necessary to take a further step to clarify
the relationship between CSD and firm performance.
To understand the relationship between these two variables, it is essential to
consider simultaneously the costs and benefits of CSD activities. Based on the
private costs theory, CSD is a costly endeavor, leading to direct expenses that have
938 C.-S. Lin et al.
a negative impact on corporate performance (Haley 1991). Similarly, from the
viewpoint of agency theory, managers tend to engage in opportunistic behavior to
further their own gain. When managers pursue any purpose other than corporate
performance through CSD activities, CSD incurs burden costs. Nevertheless, the
benefits of CSD must be taken into account simultaneously. CSD is often viewed
as a communication tool between firms and their stakeholders. By making CSD,
firms disseminate information about operating conditions and other dimensions of
the business, such as financial situations, environmental policies, employee poli-
cies, customer relationship management, community relations, and so on. Such
information will give stakeholders a better understanding of the firms and build a
strong relationship and trust between firms and their stakeholders. In this case,
CSD activities invariably generate various benefits. For example, stakeholders and
investors will be more willing to invest money and resources (Donaldson and
Preston 1995), regulators and communities will appreciate firms with a high
reputation and lend more support to them (Milne and Patten 2002), and the firms
will respond effectively to pressure and predict future environmental regulations
(Blair 2000). However, the impact of CSD on firm performance will depend on the
trade-off between these costs and benefits of CSD. If the level of CSD is too low, it
is not sufficient for stakeholders and investors to know the firm well. In this case,
information asymmetry and uncertainty will make stakeholders and investors
hesitant when making investment strategies. At the same time, due to non-trans-
parent information, managers may engage in opportunistic behavior through CSD
for their own advantage. Hence, for a low level of CSD, the costs outweigh the
benefits, leading to a negative impact on firm performance. On the other hand, if
adequate CSD is made above a given threshold and meet the expectations of the
stakeholders, firms can obtain many benefits from those stakeholders and investors
as well as society, such as tangible resources (money and resources invested by
stakeholders and investors) and intangible resources (public images, trust, legiti-
macy, human resource capital, etc.), and reduced opportunistic behavior due to
information transparency. As a result, the benefits generated will surpass costs,
leading to a positive effect of CSD on firm performance. Therefore, by considering
these costs and benefits of CSD simultaneously, a U-shaped relationship emerges
between CSD and firm performance. With this, the following hypothesis is
posited:
Corporate social responsibility disclosure and corporate performance have a U-
shaped curvilinear relationship.
99.3 Methodology
(1) Sample and data collection
The electronics industry is one of the core industries with great contribution to
the development of Taiwan’s economy. According to Taiwan’s Ministry of Eco-
nomic Affairs, electronic product exports account for nearly 30 % of the total
99 The U-Shaped Relationship 939
exports in 2009. In order to maintain sustainable competitiveness of the Taiwanese
electronics industry, many companies are emphasizing their concern with envi-
ronmental and green products. Because the electronics industry contributes greatly
in both quantity and quality to the development of Taiwan’s economy, this study
uses it as the subject for investigation. A sample of annual reports of 929 com-
panies was collected on the Taiwan Market Observation Post System and in the
Taiwan Economic Journal database in 2009. After discarding incomplete data and
missing variables, the final complete sample consists of 600 corporate annual
reports.
(2) Corporate social responsibility disclosure
Consistent with previous research, in this study, CSD is measured according to
the method of content analysis (Gray et al. 1995; Naser et al. 2006). Content
analysis is a methodology used to measure objectively, systematically, and qual-
itatively the content of communication (Naser et al. 2006). Also, it is a method of
collecting data that forms codifying quantitative information and categories to
obtain scales of different levels of complexity (Gray et al. 1995; Naser et al. 2006).
Additionally, using the number of sentences is more easily distinguishable and
prevents the problems of allocations and standardizing the number of words
(Haniffa and Cooke 2005). For these reasons, the number of CSD dimensions and
the length of CSD items expressed in terms of number of words and sentences
were used to capture CSD in this study. Drawing from previous studies (Patten
2002; Haniffa and Cooke 2005; Newson and Deegan 2002), this study integrates
seven dimensions of CSD, including environment, energy, human resources,
community involvement, fair business practices, products and services, and
customers.
(3) Firm performance: Economic value-added
The measurement of firm performance in previous studies can be divided into
non-financial and financial measurements. Non-financial measurements such as
customer satisfaction, process efficiency, etc. lack consistency in standardized
quantity of indicators and easily cause errors (Shane and Spicer 1983). Financial
measurements such as Tobin’s Q, return on assets (ROA), and return on equity
(ROE) are widely used in previous research (Garay and González 2010). Then
there are scholars who use economic value-added (EVA) to measure firm per-
formance (Palliam 2006). The advantage of using economic value-added is that it
can precisely evaluate a firm’s true value, capture simultaneously the firm’s
internal financial situation and market dynamics, and consider the firm’s long-term
development (Brewer et al. 1999). Because EVA integrates the advantages of
accounting-based measurements (ROA and ROE) and market-based measurements
(Tobin’s Q), this study uses EVA as a proxy for firm performance.
(4) Control variables
To control other variables that may have a significant impact on firm perfor-
mance, this study selects systematic risk, capital expenditure, ratio of net value to
market value, and sales growth rate as control variables (O’Byrne 1996).
940 C.-S. Lin et al.
99.4 Empirical Results
(1) Descriptive statistics
Table 99.1 shows the means, standard deviations, and correlation coefficients
for variables in this study. It is shown that CSD in terms of words and CSD in
terms of sentences are not significantly related to firm economic value-added.
However, the square of CSD in terms of words is significantly positively related to
economic value-added (r = 0.13, p \ 0.01), and the square of CSD in terms of
sentences is also significantly positively related to economic value-added
(r = 0.11, p \ 0.01). These results indicate that CSD may not linearly relate to
economic value-added.
(2) Results of the curve regression analysis
The results of the curve regression analysis are presented in Table 99.2. In
order to determine the relationship between CSD and economic value-added, this
study examines simultaneously the CSD in terms of words and sentences. As
shown in Table 99.2, for CSD in terms of words, the adjusted R square is 0.04, the
F value is statistically significant (F = 10.89, p \ 0.01), and the relationship
between CSD and economic value-added follows a U-shaped curve (CSD in terms
of words: b = -0.14, p \ 0.05; square of CSD in terms of words: b = 0.21,
p \ 0.01). Similarly, for CSD in terms of sentences, the adjusted R square is 0.04,
Table 99.1 Means, standard deviations, and correlation coefficients
1 2 3 4 5 6 7 8 9
1. EVA 1.00
2. b -0.16** 1.00
3. N/M_Ratio 0.07 -0.16** 1.00
4. Sale_GR -0.07 0.05 -0.19** 1.00
5. Cap_Ex 0.12** -0.05 -0.14** 0.05 1.00
6. CSD_Words 0.05 0.09* -0.07 -0.04 0.05 1.00
7. CSD_ 0.00 0.16** -0.09* -0.04 -0.04 0.80** 1.00
Sentences
8. CSD_Words 0.13** 0.02 -0.03 -0.01 0.16** 0.82** 0.47** 1.00
square
9. CSD_ 0.11** 0.02 -0.01 -0.00 0.08* 0.74** 0.71** 0.73** 1.00
Sentences
square
Mean 0.01 -0.00 0.00 -0.00 0.01 -0.04 -0.01 0.47 0.93
Standard 0.98 1.00 1.00 1.00 0.98 0.68 0.97 3.16 2.77
deviation
a
n = 600 *** \0.001 ** \ 0.01 * \ 0.05
Notes EVA is economic value-added, b is beta value represents systematic risk, N/M_Ratio is the
ratio of net value to market value, Sale_GR is sales growth rate, Cap_Ex is capital expenditure,
CSD_Words is corporate social responsibility disclosure in terms of number of words, CSD_
Sentences is corporate social responsibility disclosure in terms of number of sentences,
CSD_Words square is the square of CSD_Words, and CSD_Lines square is the square of
CSD_Lines
99 The U-Shaped Relationship 941
Table 99.2 Results of regression analysis on the relationship between CSD and economic value-
addeda
CSD_ Sentences CSD_Words
Model 1 Model 2
b VIF b VIF
Control variables
b -0.13** 1.05 -0.12** 1.07
N/M_Ratio 0.05 1.10 0.05 1.11
Sale_GR -0.05 1.04 -0.06 1.04
Cap_Ex 0.00 1.09 0.01 1.07
Independent varialbes
CSD_ Sentences/Words -0.16* 3.61 -0.14* 2.52
CSD_ Sentences/Words square 0.26** 3.67 0.21** 2.47
R2 0.05 0.05
Adjusted R2 0.04 0.04
F 11.65** 10.89**
a
n = 600 *** \ 0.001 ** \ 0.01 * \ 0.05
the F value is statistically significant (F = 11.65, p \ 0.01), and the relationship
between CSD and economic value-added also follows a U-shaped curve (CSD in
terms of sentences: b = -0.16, p \ 0.05; square of CSD in terms of sentences:
b = 0.26, p \ 0.01). The results of CSD in terms of number of words and sen-
tences offer evidence to support the U-shaped relationship between CSD and
economic value-added, thus providing support for the hypothesis of this study. In
addition, among the four control variables, only systematic risk is significantly
negatively related to economic value-added; other control variables, including the
ratio of net value to market value, sales growth rate, and capital expenditure, are
not significantly related to economic value-added.
99.5 Discussion and Conclusion
Social and environmental disclosures are widely discussed in recent years; how-
ever, the discussions are limited to developed countries in Western Europe. Due to
cultural and national differences among developed and developing countries, the
results of previous studies on these developed countries cannot be generalized to
other developing countries (Bebbington et al. 2007; Choi 1998; Tsang 1998). This
study examines the relationship between CSD and firm performance. For the
measurement of CSD, this study adopts content analysis of CSD in terms of
number of words and sentences, which is consistent with the methods employed by
previous research. In terms of firm performance measurement, this study uses
economic value-added to capture each firm’s internal financial situation and
market dynamics. The finding of a U-shaped relationship between CSD and
942 C.-S. Lin et al.
economic value-added bears important implications for both researchers and
business firms. From a theoretical point of view, the results of this study help to
explain the inconsistent findings of the relationship between CSD and firm per-
formance in previous literature, because previous studies consider costs and
benefits as two separate parts of CSD and only examine the linear relationship
between CSD and performance. This study analyzes simultaneously the trade-off
between the costs and benefits of CSD and discovers a U-shaped curve between
CSD and economic value-added, which may be a reasonable explanation of the
relationship between CSD and firm performance. On a practical level, the findings
of this study imply that firms should use CSD as a tool for obtaining their goals.
Firms making social and environmental disclosures are likely to incur burden costs
at the beginning because the lower level of CSD is not sufficient to gain recog-
nition from stakeholders and investors. In this case, reporting CSD is an unprof-
itable activity. However, as the level of CSD exceeds a certain threshold, reporting
CSD will help firms reduce information asymmetry between them and their
stakeholders, obtain recognition from investors and communities, and thus build
good reputation and attain social capital, relational capital, and resources con-
trolled by those stakeholders and investors. Subsequently, the more CSD firms
report, the less costs they will incur and the more benefits they will obtain.
References
Armitage S, Marston C (2008) Corporate disclosure, cost of capital and reputation: evidence from
finance directors. Biristish Account Rev 40:314–336
Barnett ML, Salomon RM (2006) Beyond dichotomy: the curvilinear relationship between social
responsibility and financial performance. Strateg Manage J 27:1101–1122
Bebbington J, Larrinaga C, Moneva JM (2007) Corporate social reporting and reputation risk
management. Account Auditing Account J 21(3):337–361
Blair A (2000) Richer and Greener, Prime Minister’s speech to the Confederation of British
Industry/Green Alliance conference on the environment
Brewer C, Chandra G, Hock CA (1999) Economics value added (EVA): its uses and limitations.
SAM Adv Manage J 64(2):4–12
Choi JS (1998) An evaluation of the voluntary corporate environmental disclosures a Korean
evidence. Soc Environ Account 18(1):2–7
Dhaliwal D, Li OZ, Tsang A, Yang YG (2009) Voluntary non-financial disclosure and the cost of
equity capital: the case of corporate social responsibility reporting. Chin Univ Hong Kong,
Hong Kong
Donaldson T, Preston LE (1995) The stakeholder theory of the corporation: concepts, evidence,
and implications. Acad Manage Rev 20(1):65–91
Freedman M, Wasley C (1990) The association between environmental performance and
environmental disclosure in annual reports and 10Ks. Advances in public interest accounting,
pp 183–193
Garay A, Gonzalez M (2010) Internet-based corporate disclosure and market value: evidence
from Latin America. Paper presented at the annual meeting of the BALAS annual conference,
ESADE, Barcelona, Spain
Godfrey C (2005) The relationship between corporate philanthropy and shareholder wealth: a risk
management perspective. Acad Manag Rev 30(4):777–798
99 The U-Shaped Relationship 943
Gray H, Bebbington KJ, Walters D (1993) Accounting for the environment: the greening of
accountancy part 2. Paul Chapman, London
Gray H, Kouhy R, Lavers S (1995) Corporate social and environmental reporting: a review of the
literature and a longitudinal study of UK disclosure. Account Auditing Account J 8(2):47–77
Haley CV (1991) Corporate contributions as managerial masques: reframing corporate
contributions as strategies to influence society. J Manage Stud 28(5):485–509
Haniffa M, Cooke TE (2005) The impact of culture and governance on corporate social reporting.
J Account Public Policy 24(5):391–430
Milne J, Patten DM (2002) Securing organizational legitimacy: an experimental decision case
examining the impact of environmental disclosures. Account Auditing Account J
15(3):372–405
Naser K, Al-Hussaini A, Al-Kwari D, Nuseibeh R (2006) Determinants of corporate social
disclosure in developing countries: the case of Quatar. Adv Int Account 19:1–23
Newson M, Deegan C (2002) Global expectations and their association with corporate social
disclosure practices in Australia, Singapore and South Korea. Int J Account 37:183–213
Ò
O’Byrne SF (1996) EVA and market value. J Appl Corp Finance 9(1):116–126
Ott JS, Shafritz JM, Jang YS (2010) Classical readings in organization theory. Wadsworth
Cengage Learn, Canada
Palliam R (2006) Further evidence on the information content of economic value added. Rev
Account Finance 5(3):204–215
Patten M (2002) The relationship between environmental performance and environmental
disclosure: a research note. Acc Organ Soc 27:763–773
Richardson J, Welker M (2001) Social disclosure, financial disclosure and the cost of equity
capital. Account Organ Soc 26:597–616
Shane B, Spicer BF (1983) Market response to environmental information produced outside the
firm. Account Rev 58(3):521–538
Solomon A, Lewis L (2002) Incentives and disincentives for corporate environmental disclosure.
Bus Strategy Environ 11:154–169
Tsang WK (1998) A longitudinal study of corporate social reporting in Singapore: the case of the
banking, food and beverages and hotel industries. Account Auditing Account J 11(5):624–635
Wang H, Choi J, Li J (2008) Too little or too much? Untangling the relationship between
corporate philanthropy and firm financial performance. Organ Sci 19(1):143–159
Chapter 100
Empirical Study on the Five-Dimensional
Influencing Factors of Entrepreneurial
Performance
Xin Lan
Abstract Entrepreneurial is the current hot issues, but the entrepreneurial success
rates are not high. Entrepreneurial performance relates to survival or extinction of
the enterprises, researching the factors that influence entrepreneurial performance
is the key points to capture success. In this study, we make use of theoretical
analysis and empirical research to explore five-dimensional factors that affect
entrepreneurial performance, which consists of capital dimension, innovative
dimension, team dimension, market dimension and environmental dimension.
Then through regression analysis, we sorted out the stepwise regression model
coefficients and test values, to establish the regression equation. We analyzed the
influence of five-dimensional factors to entrepreneurial performance. The study is
conducive to entrepreneurial activities to overcome difficulties and to entrepre-
neurial success.
Keywords Entrepreneurial Entrepreneurial performance Five-dimensional factors
Empirical research
100.1 Introduction
Entrepreneurship can not only cultivate the national innovative capability, but also to
improve national productivity and employment rates and speed up the construction
pace of knowledge-based economy strategy. It is a favorable way to alleviate the
current difficult of employment. In this context, our many business support policies
were introduced, in particular, to encourage and support college students entrepre-
neurial. However, the participation and success rates of college student start remain
X. Lan (&)
Business Institute, China West Normal University, Nanchong, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 945
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_100,
Ó Springer-Verlag Berlin Heidelberg 2013
946 X. Lan
low. As a result, to improve the business performance of our college business should
be a subject worthy of attention in entrepreneurial management discipline study. The
starting point of this study is to explore the real key dimensions of factors impact on
business performance, but the ultimate goal is to improve business performance and
the rate of business success.
So, what are the key factors currently affecting business performance, how to
make business out of the woods effectively. In this study, we have searched much
literatures, we found many literatures only remained in the relatively study on
macro education and the management level to put forward some frameworks and
models, few articles from the entrepreneur’s perspective to analysis the reality of
current entrepreneurial difficulties, use quantitative analysis to reflect the impact
dimensions of the entrepreneurial performance, and make use of the relevant
strategies to promote entrepreneurship success (Lan and Yang 2010).
100.2 Theoretical Basis and Research Hypothesis
Definition of entrepreneurship through a different interpretation, Shane illustrates
how entrepreneurs identify and exploit entrepreneurial opportunities in new ven-
tures based on their prior, and then produce a variety of results (Shane and
Venkataraman 2000). Yu Yi-hong found that entrepreneurship is a complex
process that discovering and capturing opportunities and creating new products
and services, and realize their potential value process. Bemadin believed that
performance should be defined as the results of work, which reflected the
achievement and results gained by people in their work (Schumpeter 1982).
Nowadays, operating profit created by young entrepreneurs is mainly reflected in
their profit or loss. The study on the achievement of these companies was based on
their performance in recent 2 years.
100.2.1 Capital Dimension and Entrepreneurship
Performance
The three elements of Timmons Model tell us that, business opportunity, entre-
preneurial team and entrepreneurial resource can promote the development of
entrepreneurship during their consistently matching and balancing with each other.
These elements drive each other in different development stages of the company
while changing their relation from imbalance to balance. The capital factor in
entrepreneurship is the most important dimension in resource elements of Timmons
Model; it appears that difficulties faced in entrepreneurship are often the problem
of capital. Capital deficiency will make it very difficult to transform innovation into
real productivity or to carry out business operation (Yang and Lan 2011b).
100 Empirical Study on the Five-Dimensional Influencing Factors 947
Some youths really have very good entrepreneurial plans, but they have not starting
money, as a result, their plans cannot be put into practice; or even they start their
entrepreneurial activities, the capital problem will also impact operating profit.
At present, all levels of governments are widely introducing various kinds of
supporting policies, such as patent application grant, innovation fund for medium
or small-sized enterprises, industry-specific subsidies, business subsidies to college
graduates, entrepreneurship competition prize. Therefore, we proposed the
following assumptions:
H1: Lacking of starting money is the first difficulty in developing entrepreneurial
activities; the more starting money is the better entrepreneurship performance will
be.
H2: In the operation of the venture company, good capital chain will promote
the company to develop in a healthy manner; the circulating fund has a significant
impact on entrepreneurship performance.
H3: Smooth financing channel will speed up the development of the venture
company, and promote the company to achieve better entrepreneurship
performance.
H4: Government’s support toward entrepreneurial activities through offering
innovation fund, business subsidies, etc. will help improve entrepreneurship per-
formance, thus government support has a positive impact on entrepreneurship
performance.
100.2.2 Innovation Dimension and Entrepreneurship
Performance
Schumpeter pointed out in his Theory of Economic Development, the economic
development was obtained from innovation, while the subject of innovation is
entrepreneurs (Schumpeter 1991). Drucker thought that innovation was a special
tool featuring entrepreneurship, which gave resources a new capacity to make
fortune. Drucker even thought that one person could not be called an entrepreneur
if he did not have a venture company (Drucker 2005). From the above, we draw a
conclusion that there is a close relation between innovation and entrepreneurship.
Samuelson found that in the annual growth rate per capita of 1.8 %, about
1.5 % was from technology innovation after studying on the economic growth in
the USA from 1990 to 1996 (Samuelson and William 1992). What is worthy to be
mentioned is: innovation not only refers to invention, it also means real devel-
opment of invention which is called entrepreneurship. Compared with entrepre-
neurship, invention is relatively simple and easy to do, and difficulties often appear
in real development of invention. Technology innovation will influence the
research and development and production of new products, while only good
quality products have market potential, which may further influence the entre-
preneurship performance. Therefore, we put forward the following assumptions:
948 X. Lan
H5: Technology innovation can promote the research and development and
production of new products, which will further significantly influence entrepre-
neurship performance.
Cheng (2010) thought that, development difficulties of venture companies in
our country were closely related to the insufficient of innovative motivation,
unreasonable internal and external incentive systems. Zhou (2009) pointed out that
the key to solve the difficult situation in transformation from technology imitation
to independent innovation was to strengthen the research and development and
supply of generic technology. The abovementioned studies were done from the
point of view of how to improve innovation, and all of them placed much emphasis
on independent innovation, but, nowadays, college graduates or young entrepre-
neurs are facing the difficulty of transforming technology into real productivity,
although they have some technology innovations, they fail to put these technology
innovations into real operation, which as a result leads to the low pioneering
success rate. According to the above analysis, we proposed several assumptions
for testing:
H6: Successful transformation of innovation into real operation has a positive
impact on entrepreneurship performance.
Business model is a substantial factor including positioning and channels for a
company to provide its products and services to the customers and the company’s
operation structure which enable the company to achieve its business goals. The
first innovation of a company is the innovation of its business model, which is the
foundation of development and profit. Therefore, I put forward the following
assumption that:
H7: The innovative business model has a significant impact on entrepreneurship
performance.
100.2.3 Cooperation Team Member Dimension
and Entrepreneurship Performance
Factors such as the quality, experience and expertise of the entrepreneur and his
team members will influence the success of entrepreneurship. According to our
investigation and survey, entrepreneurial team members’ experience in business
management as well as their relevant management knowledge impact greatly on
entrepreneurship performance. Therefore, in this paper, we proposed the following
assumptions:
H8: Practical experience in business operation has a positive impact on
entrepreneurship performance.
H9: Expertise has a positive impact on entrepreneurship performance.
H10: Cooperation of team members has a positive impact on entrepreneurship
performance.
100 Empirical Study on the Five-Dimensional Influencing Factors 949
100.2.4 Market Dimension and Entrepreneurship
Performance
The supply-and-demand difficulty existing in the entrepreneur market seriously
restricts the combination between entrepreneurs and technology innovation capital,
and weakens the function of entrepreneurs’ innovation and innovative vigor of their
companies. Market structure and competition situation the entrepreneurs faced with
are important factors which impact the success of entrepreneurship. Entrepreneurs
must adopt proper competition strategies in such market structure, if fail to conduct
market positioning correctly or adopt the wrong marketing mix strategies, their
companies may stop developing or even become unable to exist. We proposed
assumptions for testing as follows:
H11: Wrong or unclear target market strategy will be a disadvantage to the
success of entrepreneurship, correct target market strategy has a positive impact on
entrepreneurship performance.
H12: Products/services meeting market demands significantly have a positive
impact on entrepreneurship performance.
H13: Competitive price has a positive impact on entrepreneurship performance.
H14: Reasonable marketing strategies have a positive impact on entrepreneurship
performance.
100.2.5 Environment Dimension and Entrepreneurship
Performance
Entrepreneurial environment refers to the aggregate of a series of concepts which
significantly promote the success of entrepreneurship. Gu et al. (2008) conducted a
comprehensive analysis on the current situation of Chinese supporting policies on
entrepreneurship and their weaknesses based on the investigation to the theory
framework of the relative policies in other countries, and proposed a system to
improve such polices in China from five aspects respectively entrepreneurship
financing, entrepreneurship services, entrepreneurship cluster, entrepreneurship
education and entrepreneurship culture. Here we proposed two assumptions for
testing as follows:
H15: The effective support from the government and social organization will
have a significantly positive impact on entrepreneurship performance.
H16: Support from colleges and families have a significant impact on entre-
preneurship performance.
950 X. Lan
100.3 Research Devising
100.3.1 Definition and Measurement of Variable
The scope of this analysis is mainly those entrepreneurial activities of entrepre-
neurs who developed their entrepreneurial activities at their first time indepen-
dently, such kind of entrepreneurial activities do not have their initial capital
accumulation, and their capital resources are mainly from self-raised funds or from
financing activities. The explained variable in this analysis is the entrepreneurship
performance, which is mainly reflected in the profit or loss situation of venture
companies, the situation of companies we investigated and surveyed in this
research all reflects their performance of the last 2 years.
In this research, we used Likert Scales in the measurement of the influencing
factors of entrepreneurship performance viewing from the angle of entrepreneurs.
We designed a 7-grade scale (1 = completely don’t agree, 4 = remain neutral,
7 = fully agree), and the entrepreneurs investigated should score carefully
according to their degree of acceptance to the questions.
100.3.2 Sample Selection and Data Collection of the Formal
Questionnaire
After retrieving, consulting and studying related literature, in order to make the
results of the study more general, we took various entrepreneurial projects into
consideration in our questionnaire to make it suited to the Pioneer Park for Chinese
college students (Chengdu) and the Liaison student entrepreneurial base of the
institutions including Youth (College Student) Pioneer Park of Chengdu Hi-Tech
Zone, Technology Park of University of Electronic Science and Technology of
China, Sichuan Normal University Chengdu College, Chengdu University
of Information Technology, Sichuan University Jincheng College, University of
Electronic Science and Technology of China Chengdu College, Chengdu Voca-
tional & Technical College, Sichuan Top Vocational Institute of Information
Technology College, and the 376 enterprises and project teams in the Chengdu
Hi-Tech Zone Innovation Center. The targets of this investigation are all business
or project leaders who are developing entrepreneurial activities, and all of them are
able to answer the questions in the questionnaire. In this investigation, we
distributed 900 questionnaires in total, among which 756 were taken back, that is,
the recovery rate of it is 84 %; and 698 were valid questionnaires, the validity rate
is 92 %. The questionnaires were distributed reasonably in those companies, and
the industries selected are also representative industries, which including 5 ones,
for example, information technology, where college graduates are more willing to
do business in these fields.
100 Empirical Study on the Five-Dimensional Influencing Factors 951
100.4 Result of the Study
100.4.1 Reliability Test of Samples
In this study, we made descriptive statistics, and reliability and validity tests of
the collected sample data. And the result of reliability analysis shows that the
‘‘Cro-banch a’’ value of capital dimension, market dimension, innovation
dimension, environment dimension and cooperation team member dimension are
0.931, 0.969, 0.943, 0.952 and 0.674, respectively. According to the suggestion
of Churchilli, the coefficient of the items is more reliable when the value of
‘‘Cro-banch a’’ is greater than 0.7.
100.4.2 Exploratory Factor Analysis
In this study, in order to test the validity of the assumption of the five influencing
factors of the achievement of entrepreneurial activities, we made an exploratory
factor analysis on the 16 questions in the questionnaire. In the factor analysis, we
adopted a principle component analysis method, and varimax in rotation. We also
used SPSS17.0 in Bartlett sphericity test and KMO measurement, the observation
of the statistic product of Bartlett test of sphericity is 1633.350, and the corre-
sponding probability is close to 0. When significance level ‘‘a’’ is 0.05, due to the
probability ‘‘p’’ is less than the significance level ‘‘a’’, as a result, the null
hypothesis of Bartlett test of sphericity is rejected, and it can be considered that
there is a significant difference between correlation matrix and unit matrix. Then
let us observe the KMO value, if it is greater than 0.7, it can be said that this
project has passed factor analysis and it conforms to the standard of KMO mea-
surement which is often used in factor analysis as proposed by Kaiser, therefore,
we can explore the key influencing factors of entrepreneurship success by using
factor analysis method.
Questions in the questionnaire are systematically subordinate to several factors
at the same time, the factor analysis indicates that capital dimension questions are
four-factor structure, and the factor loading of them are between 0.735 and 0.915;
the innovation dimension questions are three-factor structure, the factor loading of
them are between 0.816 and 0.917; the cooperation team member dimension
questions are three-factor structure, the market dimension questions are four-factor
structure, the factor loading of them are between 0.661 and 0.891; the environment
dimension questions are two-factor structure, the factor loading of them are
between 0.722 and 0.746, the factor loading of them are between 0.729 and 0.877.
As there is no cross loading, see from the loading coefficient of each factor in the
corresponding dimension, all the loading are very large, which shows that
the factors have a good convergence, and the questions are reasonable, and the
dimensions established based on assumption also conform to the study.
952 X. Lan
100.4.3 Multiple Regression Analysis
The study also tested the abovementioned assumptions by using regression analysis
model, made an analysis on the impact of capital dimension, innovation dimension,
cooperation team member dimension, market dimension and environment dimen-
sion on the entrepreneurship performance, and then formed Table 100.1 as follows
of Model 1, Model 2, Model 3, Model 4 and Model 5.
In Model 1, the accessibility of initial capital (b = 0.489, P \ 0.01), financing
channel (b = 0.943, P \ 0.01), circulating fund (b = 0.840, P \ 0.05) have a
significant impact on entrepreneurship performance, which conforms to assump-
tions H1–H3. However, assumption H4 of whether received government funding
has no significant correlation with entrepreneurship performance and it is not
verified, which shows that the direct funding of government did not help to
improve the entrepreneurship performance.
In Model 2, the assumption H6 of conversion of entrepreneurship achievement
(b = 6.099, P \ 0.01) will impact the entrepreneurship performance, assumption
H7 of having innovative business model (b = -1.518, P \ 0.01) will also impact
the profitability of the companies, and it is verified. But assumption H5 of tech-
nology innovation and product research and development has no significant cor-
relation with entrepreneurship performance, which shows the pressure of
technology research and development of these companies are very big.
In Model 3, having experience in business operation (b = 3.058, P \ 0.01) has
a significant impact on entrepreneurship performance, then the assumption H8 is
verified. But surprisingly, the relation between expertise and entrepreneurship
performance is relatively weak, and the cooperation of team members has no
significant correlation with entrepreneurship performance, that is, the assumptions
H9 and H10 are not verified.
In Model 4, the degree of clearance of target market position (b = -0.096,
P \ 0.01) promotes the venture companies to develop in a healthy manner, if they
don’t have a clear idea about the target market they may fall into difficulties, thus
the assumption H11 is verified. The degree of the products/services meeting the
market demand has a significant impact on entrepreneurship performance, which
shows that the assumption H12 is verified. Competitive price and efficient sales
promotion strategies have a significant impact on entrepreneurship performance;
therefore, assumptions H13 and H14 are verified (Yang and Lan 2011a).
In Model 5, if receiving efficient support from the government and social
organization (b = -0.341, P \ 0.01), then the entrepreneurship performance of
the company will improve significantly, thus the assumption H15 is verified. The
efficient support from colleges and families (b = -1.187, P \ 0.01) has a sig-
nificant impact on entrepreneurship performance, and the assumption H16 is
verified.
100 Empirical Study on the Five-Dimensional Influencing Factors 953
Table 100.1 Verification results on the five-dimension supporting system model of entrepre-
neurship performance
Variables Model 1 Model 2 Model 3 Model 4 Model 5
H1: Enough capital has 0.489*** 0.989*** 1.405***
been raised
H2: Smooth financing 0.943*** 0.701*** 0.237***
channel
H3: Sufficient 0.840** 4.475*** 5.543***
circulating fund
H4: Received support -0.420 -3.979*** -4.873
from the government
funding
H5: Technology -.183 -.653
innovation and
product research and
development
H6: Conversion of 6.099*** 4.476***
entrepreneurship
achievement
H7: Innovative business -1.518*** -1.792***
model
H8: Have experience in 3.058***
business operation
H9: Have expertise 0.444
H10: Cooperation of 0.102
team members
H11: Degree of 0.096***
clearance of target
market position
H12: Degree of the 1.086***
products meeting the
market demand
H13: Competitive price 0.026***
H14: Degree of using 0.014**
sales promotion
strategies
H15: Degree of support 1.187***
from the government
and social
organizations
H16: Degree of support 0.341***
from colleges and
families
F 81.196 142.009 119.599 63379.427 246.445
R2 0.391 0.661 0.702 0.998 0.489
Note * means P \ 0.1, ** means P \ 0.05, *** means P \ 0.01 (two-tailed)
954 X. Lan
Fig. 100.1 The key factors
from the five-dimension
model of entrepreneurship Model 2: Verified Model 1: Verified
assumptions H6 and H7 of assumptions H1-H3 of
performance innovation dimension capital dimension
Named as: Named as: Financing
Achievement channel support
transformation support
Model 3: Verified Model 4 Verified
assumptions H8 of
cooperation team assumptions H11-H14 of
member dimension market dimension
Named as: Enterprise Named as: Marketing
practice support support
Model 5: Verified assumptions H15-H16 of
environment dimension
Named as: External environment support
100.5 Conclusion
This study is made from the angle of entrepreneurs, starting from the influencing
factors of entrepreneurship performance, and based on the features of real diffi-
culties appear in the entrepreneurial activities, with combination with results of
extensive research literature. It conducted an empirical research on the structure of
the five dimensions which influence the entrepreneurship performance, through
five-dimension supporting system model building, exploratory factor analysis and
regression analysis, we concluded that five-dimension influencing factors con-
sisting of capital dimension, innovation dimension, cooperation team member
dimension, market dimension and environment dimension. All the abovemen-
tioned five dimensions have factors successfully verified, in this analysis; we
named all the verified factors in each dimension and obtained a supporting system
model of entrepreneurship performance and its key factors, as shown in Fig. 100.1.
In a word, five-dimension supporting system model of entrepreneurship
performance from perspective of entrepreneurs is a mirror, by using such model to
analyze the influencing factors of entrepreneurship performance during certain
period and in a certain place, we can conclude the structure of government support
system during that period and in that place. The key factors costing much effort
play an instructive role on enhancing pioneering success rate.
Acknowledgments Project: It is a financial support project of China West Normal University
‘‘empirical research on marketing plight and strategies of the new ventures’’ (No. 11A016).
100 Empirical Study on the Five-Dimensional Influencing Factors 955
References
Cheng C (2010) On barriers to growth of venture companies in China-quality defect of
entrepreneurs and system dimension. J Southwest Unive Natly Humanit Soc Sci 4:171–175
Drucker P (2005) Innovation and entrepreneurship (translated by Zhang W.) Shanghai People’s
Publishing House, Shanghai, p 12
Gu S, Xiao D, Hong Q (2008) Research on strategies of improving policy system for
entrepreneurship. Chin J Popul Sci 1:10–18
Lan X, Yang A (2010) An empirical study on the entrepreneurial behavior of college students in
campus. Stat Educ 127(4):27–31
Samuelson PA, William D (1992) Nordhaus: economics, vol 2. China Development Press,
Beijing, pp 13–28
Schumpeter JA (1982) The theory of economic development. Transaction Publishers, London,
pp 172–174
Schumpeter JA (1991) Theory of economic development (translated and proofread by Wei H
et al.) The Commercial Press, Beijing, p 59
Shane S, Venkataraman S (2000) The promise of entrepreneurship as a field of research. Acad
Manage Rev 25(1):217–222
Yang A, Lan X (2011a) Innovation and entrepreneurship financing policies and strategies of the
Prisoner’s Dilemma. Chin Bus Trade 6:241–242
Yang A, Lan X (2011b) Study on the prisoner’s dilemma and the strategies of the supporting
policies for innovative entrepreneurship. Chin Bus Trade 2:241–242
Zhou M (2009) Structural problems of innovation and solutions in China. Reform 5:98–102
Chapter 101
A Multistart Local Search Heuristic
for Knapsack Problem
Geng Lin
Abstract Knapsack problem is one of classical combinatorial optimization
problems, and has a lot of applications. It is known to be NP-hard. In this paper we
propose a multistart local search heuristic for solving the knapsack problem.
Firstly, knapsack problem is converted into an unconstrained integer programming
by penalty method. Then an iterative local search method is presented to solve the
resulting unconstrained integer programming. The computational results on three
benchmarks show that the proposed algorithm can find high quality solutions in an
effective manner.
Keywords Knapsack problem Local search Heuristic
101.1 Introduction
Given n items to pack in a knapsack of capacity c. Each item i is associated with a
weight wi and a profit pi . The objective of the knapsack problem is to maximize
the profit sum without having the weight sum to exceed c. The problem can be
mathematically formulated as follows (Martello et al. 2000; Pisinger 1995):
8
> P n
>
> max f ðxÞ ¼ pi x i
>
< i¼1
ðKPÞ Pn
>
> s:t: wi xi c
>
>
: i¼1
x 2 S;
G. Lin (&)
Department of Mathematics, Minjiang University, Fuzhou, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 957
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_101,
Ó Springer-Verlag Berlin Heidelberg 2013
958 G. Lin
where S ¼ f0; 1gn and xi takes a value of 1 if item i is to be included in the
knapsack, and 0 otherwise. Without loss of generality, we assume that wi c, for
iP¼ 1; . . .; n to ensure that each item considered fits into the knapsack, and that
n
i¼1 wi [ c to avoid trivial solutions.
KP is one of the classical optimization problems in combinatorial optimization,
and has a lot of applications (Gorman and Ahire 2006) in production, logistics,
material cutting and financial problems. In solving large combinatorial optimiza-
tion problems, KP is also regarded as a sub-problem to deal with. It has been
widely studied in the last few decades due to its theoretical interest and its wide
applicability, see Kellerer et al. (2004) and references therein. KP is known to be
NP-hard, so the exact algorithms with polynomial complexity can only exist in the
case P ¼ NP. It can be solved in pseudo polynomial time by dynamic program-
ming (Papadimitriou 1981). A lot of heuristic algorithms have been considered for
approximately solving the knapsack problem, such as tabu search (Hanafi and
Freville 1998), genetic algorithm (Zhao et al. 2011; Tian and Chao 2011; Shan and
Wu 2010), artificial fish school algorithm (Li et al. 2009), ant colony algorithm
(Liao et al. 2011).
Local search algorithms are widely applied to numerous hard optimization
problems, including problems from mathematics, operations research, and engi-
neering. A local search algorithm starts from an initial solution and iteratively
moves to a neighbor solution. Every solution has more than one neighbor solu-
tions, the choice of which one to move to is taken using only information about the
neighborhood of the current one. When no improving configurations are present in
the neighborhood, local search is trap in a local optima.
In this paper, a new local search method is proposed for knapsack problem.
When local search is stuck at a locally optimal solution, we restart the local search
procedure from a new initial solution.
The remainder of the paper is arranged as follows: In Sect. 101.2, some defi-
nitions and local search methods which have been used in the literature are
introduced. Section 101.3 presents a new multistart local search method for the
knapsack problem. Experiments were done on some benchmarks, computational
results and comparisons are presented in Sect. 101.4, and concluding remarks are
put in Sect. 101.5.
101.2 Methodology
Local search is a well-known approach for solving a lot of combinatorial opti-
mization problems. When using a local search procedure to a given instance of an
optimization problem, we need to define a ‘‘neighborhood’’, which is a subset of
the solution set, for each solution. Local search algorithm begins with an initial
solution and searches its neighborhood, then moves from solution to solution in the
neighborhood by applying local changes, until the current solution is better than its
neighbors.
101 A Multistart Local Search Heuristic for Knapsack Problem 959
There are two neighborhood structures have been considered for the knapsack
problem: the 1-flip and 1-flip-exchange neighborhoods. If two solutions differ
exactly on one assignment, they are 1-flip neighbor.
Definition 1 For any x 2 S, the 1-flip neighborhood Nf ðxÞ of x is defined by
Nf ðxÞ ¼ fy 2 Sj kx yk1 1g. The 1-flip neighborhood
Nf ðxÞ can be reached by
adding or removing one item from x. Hence, Nf ðxÞ ¼ n þ 1.
If two solutions are 1-flip-exchange neighbors if one can be obtained from the
other by exchanging two items. It is an extension of 1-flip neighborhood.
A lot of algorithms for knapsack problem used above two neighborhood
structures. They start from an initial solution and iteratively move to the best
solution of the neighbor, until the current solution is better than its neighbors.
These local search methods are belonging to greedy algorithm, and trap in a local
optima easily.
101.3 The Proposed Local Search Method
In this section, firstly, we convert knapsack problem equivalently into uncon-
strained integer programming. Then a new local search method for the resulting
unconstrained integer programming is proposed.
101.3.1 Equivalent Unconstrained Integer Programming
We use penalty method to transform knapsack problem to unconstrained integer
programming. Constructing the following unconstrained integer programming:
8
< Pn
max gðxÞ ¼ pi xi khðxÞ
ðNKPÞ i¼1
:
s:t: x2S
P
where k [ 0 is a penalty parameter, and hðxÞ ¼ maxf ni¼1 wi xi c; 0g.
Lemma 1 If k [ pmax , where pmax ¼ maxfp1 ; . . .; pn g, problems KP and NKP
have the same optimal solution and optimal value.
101.3.2 Local Search Method
A lot of local search methods used in the existing algorithms for knapsack problem
based on greedy method. It trap into a local optima easily. We present an iterative
local search method for knapsack problem. The main idea of the algorithm is to
960 G. Lin
flip a bit at a time in an attempt to maximize the profit sum without having the
weight sum to exceed c. Define the gainði; xÞ of item i as the objective value of the
problem (NKP) would increase if the i bit is flipped, which is as follows:
gainði; xÞ ¼ gðx1 ; . . .; xi1 ; 1 xi ; xiþ1 ; . . .; xn Þ gðxÞ:
Note that an item’s gain may be negative. For each item i, the local search
algorithm computes the gainði; xÞ. It starts with a randomly solution in the solution
space S and changes the solution by a sequence of 1-flip operations, which are
organized as passes. At the beginning of a pass, each item is free, meaning that it is
free to be flipped; after a bit is flipped, it become unfree, i.e., the bit is not allowed
to be flipped again during that pass. The algorithm iterative selects a free item to
flip. When a item is flipped, it becomes unfree and the gain of free items are
updated. After each flip operation, the algorithm records the objective value of
(NKP) achieved at this point. When there are no more free item, a pass of the
algorithm stops. Then it checks the recorded objective values, and selects the point
where the maximum objective value was achieved. All items that were flipped
after that point are flipped. Another pass is then executed using this solution as its
starting solution. The local search algorithm terminates when a pass fails to find a
solution with better value of the objective value of (NKP).
When the local search algorithm traps in a local optima, we restarts the local
search algorithm from a randomly solution.
Let V be a set of items which are free to flip in a pass. The multistart local
search algorithm can be stated as follows:
Step 0. Choose a positive number max iter as the tolerance parameter for termi-
nating the algorithm. Set N ¼ 0, xglobal ¼ 0.
Step 1. Generate a solution x ¼ fx1 ; . . .; xn g randomly.
Step 2. Set V ¼ f1; . . .; ng, t ¼ 1, x0 ¼ x. Calculate gainði; xÞ, for i 2 V.
Step 3. Let gainðj; xÞ ¼ maxfgainði; xÞ; i 2 Vg. Set xt ¼ ðx1 ; . . .; 1 xj ; . . .; xn Þ,
and V ¼ Vnfjg, x ¼ xt , t ¼ t þ 1.
Step 4. If V 6¼ [, calculate gainði; xÞ for i 2 V, go to Step 3. Else go to Step 5.
Step 5. Let xmax ¼ maxfxt ; t ¼ 1; . . .; ng. If gðxmax Þ [ gðx0 Þ, set x ¼ xmax , go to
Step 2. Else, if gðxglobal Þ [ gðxmax Þ, let xglobal ¼ xmax . Go to Step 6.
Step 6. If N\max iter, let N ¼ N þ 1, go to Step 1. Else output xglobal .
101.4 Numerical Experiment
In this section, we test the proposed multistart local search algorithm. The
experiments were performed on a personal computer with a 2.11 GHz processor
and 1.0 GB of RAM. For our experiments we employ the following three
benchmark instances, which are also used to test the genetic algorithm for knap-
sack problem in (Shan and Wu 2010).
101 A Multistart Local Search Heuristic for Knapsack Problem 961
Problem 1. ðw1 ; . . .; w20 Þ = (92, 4, 43, 83, 84, 68, 92, 82, 6, 44, 32, 18, 56, 83,
25, 96, 70, 48, 14, 58), ðp1 ; . . .; p20 Þ = (44, 46, 90, 72, 91, 40, 75, 35, 8, 54, 78, 40,
77, 15, 61, 17, 75, 29, 75, 63), c ¼ 878.
Problem 2. ðw1 ; . . .; w50 Þ = (220, 208, 198, 192, 180, 180, 165, 162, 160, 158,
155, 130, 125, 122, 120, 118, 115, 110, 105, 101, 100, 100, 98, 96, 95, 90, 88, 82,
80, 77, 75, 73, 70, 69, 66, 65, 63, 60, 58, 56, 50, 30, 20, 15, 10, 8, 5, 3, 1, 1),
ðp1 ; . . .; p50 Þ = (80, 82, 85, 70, 72, 70, 66, 50, 55, 25, 50, 55, 40, 48, 50, 32, 22,
60, 30, 32, 40, 38, 35, 32, 25, 28, 30, 22, 50, 30, 45, 30, 60, 50, 20, 65, 20, 25, 30,
10, 20, 25, 15, 10, 10, 10, 4, 4, 2, 1), c ¼ 1000.
Problem 3. ðw1 ; . . .; w100 Þ = (54, 183, 106, 82, 30, 58, 71, 166, 117, 190, 90,
191, 205, 128, 110, 89, 63, 6, 140, 86, 30, 91, 156, 31, 70, 199, 142, 98, 178, 16,
140, 31, 24, 197, 101, 73, 169, 73, 92, 159, 71, 102, 144, 151, 27, 131, 209, 164,
177, 177, 129, 146, 17, 53, 164, 146, 43, 170, 180, 171, 130, 183, 5, 113, 207, 57,
13, 163, 20, 63, 12, 24, 9, 42, 6, 109, 170, 108, 46, 69, 43, 175, 81, 5, 34, 146, 148,
114, 160, 174, 156, 82, 47, 126, 102, 83, 58, 34, 21, 14), ðp1 ; . . .; p100 Þ = (597, 596,
593, 586, 581, 568, 567, 560, 549, 548, 547, 529, 529, 527, 520, 491, 482, 478,
475, 475, 466, 462, 459, 458, 454, 451, 449, 443, 442, 421, 410, 409, 395, 394,
390, 377, 375, 366, 361, 347, 334, 322, 315, 313, 311, 309, 296, 295, 294, 289,
285, 279, 277, 276, 272, 248, 246, 245, 238, 237, 232, 231, 230, 225, 192, 184,
183, 176, 174, 171, 169, 165, 165, 154, 153, 150, 149, 147, 143, 140, 138, 134,
132, 127, 124, 123, 114, 111, 104, 89, 74, 63, 62, 58, 55, 48, 27, 22, 12,6),
c ¼ 6718.
The proposed algorithm uses a parameter maxiter as a termination parameter.
In the experiment, we set maxiter ¼ 30. We run the proposed algorithm 10 times
to above three benchmarks. The test results are given in Table 101.1. In order to
compare with genetic algorithm proposed in (Shan and Wu 2010), the results of
greedy algorithm, basic genetic algorithm, hybrid genetic algorithm (Shan and Wu
2010) are also listed in Table 101.1, and the results quote from (Shan and Wu
2010) directly. Table 101.1 gives the best solutions found by greedy algorithm,
basic genetic algorithm, hybrid genetic algorithm. P and W denotes the sum of the
profit, and the sum of weight, respectively. g means algorithm found the best
solution within g generations.
The following observations can be made based on the experimental results in
Table 101.1.
(1) The proposed algorithm found the solution better than those of greedy algo-
rithm and basic genetic algorithm found.
(2) The proposed algorithm and hybrid genetic algorithm found the same best
objective value.
(3) Note that our proposed used only 30 initial solutions. It shows that the pro-
posed can reduce the chance that local search process becomes trapped at local
optima.
962 G. Lin
Table 101.1 Experiment results
Problem Greedy algorithm Basic genetic algorithm Hybrid genetic The proposed
(Shan and Wu 2010) (Shan and Wu 2010) algorithm (Shan algorithm
and Wu 2010)
P=W P=W=g P=W=g P=W
1 1023/825 1024/878/29 1024/878/12 1024/878
2 3095/996 3077/1000/192 3103/1000/50 3103/1000
3 26380/6591 25848/6716/319 26559/6717/147 26559/6717
101.5 Conclusion
A multistart local search algorithm is proposed to find approximate solutions for
knapsack problems. Penalty method is used to transform knapsack problem to
unconstrained integer programming. An iterative local search method is presented
to solve the resulting unconstrained integer programming. It can reduce the chance
of trapping at local optima. Experiments were done on three benchmarks from
literature. Compare with some existing algorithms, it shows the proposed algo-
rithm is effective.
Acknowledgments This research is supported by the Science and Technology Project of the
Education Bureau of Fujian, China, under Grant JA11201.
References
Gorman MF, Ahire S (2006) A major appliance manufacturer rethinks its inventory policies for
service vehicles. Interfaces 36:407–419
Hanafi S, Freville A (1998) An efficient tabu search approach for the 0–1 multidimensional
knapsack problem. Eur J Oper Res 106:663–679
Kellerer H, Pferschy U, Pisinger D (2004) Knapsack problems. Springer, Berlin
Li KS, Jia YZ, Zhang WS (2009) Genetic algorithm with schema replaced for solving 0–1
knapsack problem. Appl Res Comput 26:470–471
Liao CX, Li XS, Zhang P, Zhang Y (2011) Improved ant colony algorithm base on normal
distribution for knapsack problem. J Syst Simul 23:1156–1160
Martello S, Pisinger D, Toth D (2000) New trends in exact algorithms for the 0–1 knapsack
problem. Eur J Oper Res 123:325–332
Papadimitriou HC (1981) On the complexity of integer programming. J ACM 28:765–768
Pisinger D (1995) An expanding-core algorithm for the exact 0–1 knapsack problem. Eur J Oper
Res 87:175–187
Shan XJ, Wu SP (2010) Solving 0–1 knapsack problems with genetic algorithm based on greedy
strategy. Comput Appl Softw 27:238–239
Tian JL, Chao XP (2011) Novel chaos genetic algorithm for solving 0–1 knapsack problem. Appl
Res Comput 28:2838–2839
Zhao XC, Han Y, Ai WB (2011) Improved genetic algorithm for knapsack problem. Comput Eng
Appl 47:34–36
Chapter 102
Heterogeneity of Institutional Investors
and Investment Effects: Empirical
Evidence from Chinese Securities Market
Ying Jin
Abstract With social security funds and securities investment funds as research
objects, this paper makes an empirical study on cross-sectional data in the period
2008–2010 of listed companies of which the stocks are heavily held by institu-
tional investors. Using property rights theory and agency theory, this paper verifies
the following hypothesis: securities investment funds and social security funds
face different political and social pressure, and have different payment mecha-
nisms for managers, thus the fund owners may have conflict or convergence of
interests with companies’ administration, which may affect contrarily the invest-
ment value of companies. This paper contributes by demonstrating the influence
on companies’ investment effects of heterogeneity of Chinese institutional
investors, which provides new evidence for judging, in the era of diversified
institutional investors, the different roles of different institutional investors in
corporate governance and performance, and offers supporting evidence for China
to formulate development strategy for institutional investors.
Keywords Corporate governance Investment value Securities investment
fund Social security fund
102.1 Introduction
In mature capital markets, the supervision of institutional investors tends to exert
important influence on the corporate governance, and it is a reliable mechanism for
addressing corporate governance issues. As super-conventional development of
Chinese institutional investors leads to a diversified structure, people are
Y. Jin (&)
School of Business, Jinling Institute of Technology,
Nanjing, Jiangsu, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 963
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_102,
Ó Springer-Verlag Berlin Heidelberg 2013
964 Y. Jin
concerned whether it is possible to effectively supervise the company’s adminis-
tration and alleviate the problem of internal control to accrue the value of the
company, and whether different institutional investors have different effects on
corporate governance as well as investment value. Diversified institutional
investors mean that except securities investment funds, institutional investors have
held a certain amount of stocks, and may have significant impact on the capital
market. This article will study difference between securities investment funds and
social security funds in aspects such as the political and social pressure they face
and the incentives, and make an empirical research on the different effects caused
when these two kinds of funds hold stocks.
102.2 Literature Review
At present, a lot of domestic and foreign research has been made on institutional
investors’ participation in corporate governance. On whether institutional investors
are involved in corporate governance, there are different types of opinions among
foreign scholars. Scholars who believe in ‘‘shareholder activism’’ think that
institutional investors have favorable conditions for supervision, for example, they
are professional investors and hold a large amount of stocks, and therefore they
obtain information superiority. In addition, heavily held stocks make them sus-
ceptible to liquidity losses when they withdraw from the market, and they have to
bear a strict fiduciary responsibility. This means investors can benefit from
supervision. The above factors indicate that ‘‘free rider’’ problem can be avoided
(David and Kochhar 1996; Grossman and Hart 1980; Smith 1996). Scholars who
believe in ‘‘shareholder passivism’’ presume that because of reasons like legal
restrictions, difficulty of supervision, high cost and liquidity, institutional investors
are not proactively involved in supervision of the companies; their stock holding
has no significant impact on the value of the companies (Agrawal and Knoeber
1996; Bhide 1994; Coffee 1991). Those who hold the eclectic opinion believe that
due to different funding sources, amount of shareholdings, and whether the
institutional investors have conflict of interest with the companies, investors’ roles
in corporate governance differ (Cornett et al. 2007).
Previously, Chinese researchers think that due to their own conditions and
external environmental constraints, institutional investors (mainly those of secu-
rities investment funds) have very limited roles in corporate governance. For
instance, Huang (2006) considered that Chinese institutional investors were highly
dependent on the government, and that the government intervened in operations of
institutional investors with resource control or political control. However, huge
disparity existed between government targets and business goals, therefore insti-
tutional investors were not qualified to supervise the businesses. Over time, a
growing number of scholars believe that Chinese institutional investors do resort
their oversight capacity to govern. For example, Wang et al. (2008) presumed that
as transformation of our government’s functions and reform of split share structure
102 Heterogeneity of Institutional Investors and Investment Effects 965
went on, the government was being phased out as a manager and a supervisor.
Since the types of institutional investors augmented and their share proportions
increased, it was possible for them to become qualified company oversight bodies.
According to associated preliminary findings in China, these studies mainly
focused on securities investment funds, or they regarded all institutional investors
as homogeneous funds. The findings simply presumed that institutional share-
holders had either no significant impact or positive effect on firm value. However,
they ignored that differences between institutional managers’ business objectives
might affect adversely the firm value.
102.3 Research Hypotheses
Many foreign scholars believe that unlike banks and insurance companies which
have business ties with the invested companies, public pension funds are relatively
independent institutional investors (Cornett et al. 2007) as well as long-term funds,
they are suitable to be the overseers of enterprises. Woidtke (2002) divided pen-
sion funds into two types: public pension funds and private pension funds. By
comparing effects on industry’s adjustment Tobin’s Q of companies when these
two types of funds hold stocks in companies, Woidtke found that shareholdings of
public pension funds are negatively related with industry’s adjustment Tobin’s Q,
while the shareholdings of private pension funds are positively related with
industry’s adjustment Tobin’s Q. She suggested that the remarkable differences
were resulted from the fact that public pension funds faced greater political
pressure than private pension funds and their incentives were decoupled from
performance. Domestic scholars, such as Zhang and Sun (2006), believed that the
social security funds had long-term goals, and were suitable to be institutional
investors who could participate in corporate governance and stabilize market.
However, they ignored the inconsistency between objectives of social security
fund managers and those of corporate. Wang (2008) proposed that one particular
problem of agency during management of public pension reserve funds was the
intervention of political factors in the funds’ operations. Based on the above
analysis, we propose Hypothesis 1.
H1: The shareholding proportions of social security funds are negatively cor-
related with investment value of companies.
The minority shareholders have the rights to transfer fund shares, which is the
rights provided by the redemption mechanism (for open-end funds), or the rights of
transfer in capital markets (for closed-end funds), and can pose strong constraints
on fund managers. In addition, an essential part of incomes of the fund companies
is the management fees charged in accordance with the size of the trust fund. The
rights to transfer of minority shareholders provide incentives for fund managers to
monitor management of the companies; this can maximize the interests of minority
shareholders, and can also make the goals of the fund managers less susceptible to
administrative intervention. From the perspective of incentives of the funds, at the
966 Y. Jin
beginning of each year, fund managers and fund companies sign a performance
contract, in which both parties agree on certain performance targets. The targets
are usually about how high the annual cumulative rate of return of the fund
administered by a fund manager must rank among the same type of funds. The
ranking is directly linked with the performance bonus that fund manager can
obtain. This performance-sensitive payment system urges fund managers to strive
to safeguard the interests of minority shareholders, and strengthen supervision as
the shareholdings of funds expand. Such supervision can ease the company’s
problem of agency, reduce agency costs, and increase the value of the companies
as well as that of the funds. Moreover, Chinese fund companies have strict trust
and agency relationship with minority shareholders. This means fund managers are
under dual supervision of the trustees and the general assembly of fund holders,
and they are responsible, on behalf of the minority shareholders, to oversee the
companies and to protect and increase the interests of minority shareholders.
Therefore, fund managers and the firms have the same goal: maximization of
investment value. Based on the above analysis, we propose Hypothesis 2.
H2: The shareholding proportions of securities investment funds are positively
correlated with the investment value.
Because securities investment funds account for a majority of institutional
investors, this paper proposes Hypothesis 3.
H3: The shareholding proportions of institutional investors as a whole
(including securities investment funds, social security funds and insurance com-
panies) are positively correlated with the investment value.
102.4 Study Design
102.4.1 Sample Source and Selection
We have chosen the period 2008–2010 as the sample interval. Since value indi-
cators lag by 1 year, we verified how the shareholding proportions of securities
investment funds, social security funds and institutions as a whole in the period
2008–2009 affected corporate investment value of 2009–2010. We selected 817
samples of 2008, and 1178 samples of 2009. The data used in this paper comes
from the CSMAR database.
102.4.2 Variable Setting and Model Design
As for indicators of investment value, in addition to earnings per share and net
assets yield that represent companies’ accounting performance, Tobin’s Q used to
study relationship between corporate governance and the value was also chosen.
102 Heterogeneity of Institutional Investors and Investment Effects 967
Tobin’s Q, which is a market indicator, equals to the ratio of the company’s market
value to replacement value of the company’s assets, and can reflect the company’s
future development potential. What’s more, Tobin’s Q can reflect not only the
public shareholder activism, but also the value effects of nonpublic shareholder
activism, for example private negotiations (Woidtke 2002; Sun and Huang 1999).
To fully reflect the investment effect, capital expenditure, which is the com-
pany’s most important investment decision, was also considered. Many scholars
believe that capital expenditure is likely to become an important tool for the
controlling shareholders or administrators of the company to secure personal
interests and damage the interests of minority shareholders (Hu et al. 2006). Under
the institutional context in which companies are controlled by the largest share-
holders, investigating the impact of active shareholder behavior of institutional
investors, which is an emerging governing mechanism, on investor protection from
the perspective of capital expenditure will help understand the effect of the
supervision of the institutional investors, and provide backing evidence for Chi-
nese authorities’ vigorous decisions on supporting institutional investors.
We selected capital expenditure as the proxy indicator of investor protection,
and used ‘‘cash for building the fixed assets, intangible assets and other long-term
assets’’ on the cash flow statement as the proxy variable of total capital expen-
diture (Zhang 2010). Referring to the article of Hua and Liu (2009), we used the
following variables as control variables. First, we used GROW to represent the
company’s growth. Capital expenditures differ as companies’ growth differs. More
developed companies have more potential investment opportunities and thus will
spend more capital. Operating revenue growth rate is frequently used as indicators
for measuring growth. Second, we used CASH to represent net cash flow generated
from operations. The above mentioned net cash flow is an important factor
affecting the company’s capital expenditure level.
In terms of indicators for institutional shareholding (INS), this paper used
shareholding proportions of securities investment funds, shareholding proportions
of the social security funds and those of institutional investors as a whole. In terms
of control variables, we used the ownership structure variables to represent internal
mechanism of corporate governance (Bai and Liu 2005). We selected the share-
holding proportions of the largest shareholders (TOP1) and those of the second to
the tenth largest shareholders (TOP2–10). TOP1 reflects corporate holding structure
with Chinese characteristics; TOP2–10 reflects the roles of the second to the tenth
largest shareholders in balancing internal control of the largest shareholders.
Company size and financial leverage (asset-liability ratio) were used to represent
the other control variables that affect the corporate investment value.
We took into consideration that institutional investors might expand their
shareholding proportions the same time when investment value increased. That is
to say, institutional investors may invest in the company due to recent growth in
investment value, rather than company stocks in order to improve companies’
increase investment value. For example, institutional investors supervise com-
pany’s administration after holding more their shareholdings of the company after
finding that corporate performance is better. At this point, institutional investors’
968 Y. Jin
ownership and investment value are also positively correlated. Therefore, we
measured indicator variables of investment value in the following year, to ensure
investment improvement derived from the influence of institutional investors’
shareholdings increase on the company’s administration decisions. Due to the
hysteretic nature of accounting statements, the ownership structural variables,
indicators of financial leverage and firm size have also been brought forward a
year. The regression equation is as follows:
X
TBQ ¼ a þ b INS þ bi Controli þ e ð102:1Þ
X
EPS ¼ a þ b INS þ bi Controli þ e ð102:2Þ
X
ROE ¼ a þ b INS þ bi Controli þ e ð102:3Þ
X
CAP ¼ a þ b INS þ bi Controli þ e ð102:4Þ
The variables used in this paper are showed in Table 102.1.
102.5 Empirical Results and Analysis
In this paper, we used cross-sectional data to make least-square linear regression of
the above variables, and analyzed the impact of year by year change in the
macroeconomic environment on the institutional investors’ shareholdings with
annual dummy variable YEAR (for data of shareholdings in 2008, YEAR = 0; for
data of shareholdings in 2009, YEAR = 1).
From statistical characteristics of variables, we can learn that in the period
2009–2010, average Tobin’ Q of stocks heavily held by institutional investors is
2.84, average earnings per share is 0.45, average return of net assets is 12.04,
average rate of capital expenditures is 6.24 %. In the period 2008–2009, average
shareholding proportion of securities investment funds is 4.54 %, average share-
holding proportion of social security funds is 0.29 % and that of institutional
investors as a whole is 12.04 %. In listed companies of which the stocks are
heavily held by institutional investors, the average shareholding proportion of the
largest shareholders is 38.55 %, while that of the second to the tenth largest
shareholders is 19.69, the average asset-liability ratio is 49.7 %.
According to the results of multiple linear regressions(results table omitted), we
can see that VIF values of the regression equations are much less than 10, indi-
cating that the regression model is not affected by the multicollinearity. The
Durbin-Watson stat is also close to 2, which demonstrates that the auto-correlation
between the variables is low. The regression results show that the negative cor-
relation between shareholding proportions of social security funds and TBQ, EPS,
ROE, CAP is significant, which validates H1. The remuneration of managers in
social security funds is not linked to performance. Under political social pressure,
102 Heterogeneity of Institutional Investors and Investment Effects 969
Table 102.1 Definition of variables
Classification of variables Names of Descriptions of variables
variables
Dependent Indicators of market value TBQ Tobin’s Q of the company
variables Indicators of earning EPS Earnings per share
capacity ROE Return on net assets
Protection indicators of CAP Capital expenditure rate = Cash paid to
interests of minority build fixed assets, intangible assets
shareholders and other long-term assets/total assets
Independent Institutional ownership INSSE Shareholding proportions of securities
variables variables (of the top investment funds
ten shareholders) INSSO Shareholding proportions of social
security funds
INS Shareholding proportions of institutional
investors as a whole
Control Characteristics of TOP1 Shareholding proportions of the largest
variables ownership structure shareholders
TOP2–10 Shareholding proportions of the second to
the tenth largest shareholders
Other variables LOGSIZE Log of total assets (company size)
LEV Financial leverage (debt ratio)
CASH Net cash flow generated from operations/
total assets
GROW Operating revenue growth rate
social security funds have different goals with the listed companies, and they will
exercise a negative impact on the companies’ investment value. The positive
correlation between shareholding proportions of securities investment funds,
shareholding proportions of institutional investors as a whole and TBQ, EPS,
ROE, CAP is significant, which validates H2 and H3. With expanding of share-
holding proportions of securities investment funds, securities investment funds and
institutional investors as a whole can overcome ‘‘free rider’’ problem of minority
shareholders. They are motivated and capable to oversee the company’s admin-
istration, and can play imperative roles in promoting the company’s investment
value. In addition, we can conclude the following from the empirical results: (1)
the ownership structural variables have no significant impact on the investment
value, indicating that the largest shareholders used their advantages of control to
violate company assets and undermine the interests of outside investors. Other
large shareholders, because they have different targets, do not manage to form
effective balance with the largest shareholders. This also demonstrates that insti-
tutional investors as a whole can inhibit large shareholders from infringing the
interests of minority shareholders, protect their interests, and mitigate the problem
of agency; (2) the negative correlation between investment value of companies and
company size as well as asset-liability ratio is significant. The investment value of
companies is negatively correlated with company size; this conforms to the fact
that investment value of larger companies is prone to be underestimated while the
970 Y. Jin
investment value of smaller companies is prone to be overestimated. High debt
ratio will increase creditors’ constraints on the companies, and diminish the
necessity of oversight and institutional investors’ interest in investment, thereby
reducing investment value of companies; (3) the positive correlation between the
net cash flow generated from operations and capital expenditure rate is significant,
indicating that investment spending is influenced by the scale of internal financing.
However, the negative correlation between the company’s growth and capital
expenditure level is insignificant and does not pass statistical test. That is to say,
companies’ investment spending decreases with the improvement of investment
opportunities, therefore reflecting a possible shortage of investment in Chinese
listed companies.
102.5.1 Conclusions and Recommendations
The results of empirical tests show that although the social security funds are
relatively long-term funds and have conditions for supervising administration of
listed companies, they, under the political and social pressure, have different
operating objectives with listed companies and will pose a negative impact on their
market value. The securities investment funds’ incentives are highly related with
performance, making them less vulnerable to the political and social pressure. To
increase in funds’ shareholdings will urge the funds to supervise more closely
listed companies, thereby accruing the investment value. Moreover, it is verified
that the overall shareholdings of institutional investors have a positive impact on
the investment value of listed companies.
This paper demonstrates that heterogeneity exists in institutional investors, due to
differences between incentives and conflicts of interests, different institutions have
different impacts on corporate governance and value of listed companies, which
provides new evidence for judging roles of different institutions in corporate gov-
ernance in the era of diversified Chinese institutional investors. Given the ineffective
supervision of Board of Directors and roles of institutional investors as a whole in
promoting the value of companies, supervision of institutional investors has become
a reliable mechanism for overseeing listed companies. Chinese authorities should
continue to vigorously support institutional investors. But given that social security
funds have negative influence on the investment value of listed companies, Chinese
government, when supporting diversified institutional investors, should reduce the
political and social pressure on institutional investors and set up payment systems
that are closely linked to performance, in order to enable the funds to be independent
market participants and create a harmonious governance structure.
Acknowledgments Fund: This paper explains the preliminary results of the philosophy and
social science project (2010SJD630048) of Education Department of Jiangsu Province.
102 Heterogeneity of Institutional Investors and Investment Effects 971
References
Agrawal A, Knoeber CR (1996) Firm performance and mechanisms to control agency problems
between managers and shareholders. J Financ Quant Anal 31(3):377–397
Bai C, Liu Q (2005) Empirical research on Chinese listed companies governance structure. Econ
Res 51(2):81–91
Bhide A (1994) Efficient markets, deficient governance: U. S. securities regulations protect
investors and enhance market liquidity, but do they alienate managers and shareholders?’’
Harv Bus Rev 72(5):128–140
Coffee J (1991) Liquidity versus control: the institutional investor as corporate monitor.
Columbia Law Rev 91(2):1277–1368
Cornett MM, Marcus AJ, Saunders A, Tehranian H (2007) The impact of institutional ownership
on corporate operating performance. J Bank Finance 31(6):1771–1794
David P, Kochhar R (1996) Barriers to effective corporate governance by institutional investors:
implication for theory and practice. Eur Manag J 14(5):457–466
Grossman S, Hart O (1980) Takeover bids, the free rider problem, and the theory of the
corporation. Bell J Econ 18(11):42–64
Hu G, Huang J, Qiu Y (2006) Ownership structure and capital expenditure decisions: theoretical
and empirical analysis. Manag World 22(1):137–144
Hua G, Liu Z (2009) Governmental control, stock holding of institutional investors and protection
of investors’ rights. J Finance Econ 35(4):119–130
Huang X (2006) On the premise of the cause of effective supervising subject. J Huazhong Univ
Sci Technol (Soc Sci Edn) 27(1):92–96
Smith M (1996) Shareholder activism by institutional investors: evidence from CaPERS.
J Finance 51(4):227–252
Sun Y, Huang Z (1999) Equity structure and performance of listed companies. Econ Res
45(12):23–31
Wang W (2008) Administration structure of public pension fund. Justice China 27(3):94–96
Wang Z, Hua F, Yang Z (2008) The influence of institutional investors’ role in corporate
governance. J Huazhong Univ Sci Technol (Soc Sci Edn) 22(4):112–116
Woidtke T (2002) Agents watching agents? Evidence from ownership and firm value. J Financ
Econ 63(3):99–131
Zhang W (2010) Control of large shareholders, institutional equity and investors protection. Econ
Probl 32(8):94–97
Zhang W, Sun Z (2006) Research on involvement in corporate governance of social security
funds and other institutional investors. Econ Rev 23(4):31–33
Chapter 103
Integer Linear Programming Model
and Greedy Algorithm for Camping
Along the Big Long River Problem
Zhen-ping Li and Xiao-dong Huang
Abstract In this paper, we investigate the problem of camping along the Big
Long River: How to schedule the X trips in a rafting season of the Big Long River
so that the total meets of any boats are minimal? By introducing the proper
variables, the problem is formulated into an integer linear programming model.
For small size problem, this integer linear programming can be solved by Lingo
software; for large size problem, we design a greedy algorithm to arrange the
schedule of the given X boats. Finally, we do some simulations of the above model
and algorithm and obtain the optimal solution.
Keywords Camping along the river Integer linear programming model Greedy
algorithm Simulation The optimal solution
103.1 Introduction
Visitors to the Big Long River (225 miles) can enjoy scenic views and exciting
white water rapids. The river is inaccessible to hikers, so the only way to enjoy it is
to take a river trip that requires several days of camping. River trips all start at First
Launch and exit the river at Final Exit, 225 miles downstream. Passengers take
either oar-powered rubber rafts, which travel on average 4 mph or motorized
boats, which travel on average 8 mph. The trips range from 6 to 18 nights of
camping along the river. Currently, X trips travel down the Big Long River each
year during a six month period (the rest of the year is too cold for river trips).
There are Y camp sites on the Big Long River, distributed fairly uniformly
Z. Li (&)
School of Information, Beijing Wuzi University, Beijing, China
e-mail: [email protected]
X. Huang
Department of Postgraduate, Beijing Wuzi University, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 973
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_103,
Springer-Verlag Berlin Heidelberg 2013
974 Z. Li and X. Huang
throughout the river corridor. In order to make sure the passengers enjoy a wil-
derness experience and also for the sake of their safety (http://wenku.baidu.com/
view/19fab121192e45361066f5e4.html), we should try to avoid the meet of two
groups of boats on the river. Besides, due to the capacity constraints, no two sets of
campers can occupy the same site at the same time.
Every year, before the rafting season, the park managers must arrange the
schedule of these X trips that will raft along the Big Long River in the rafting
season. The key problem is how to schedule these X trips so that the total meets
among all boats in the river is minimum. In this paper, we will solve this problem.
This paper is organized as follows: in Sect. 103.2, we make some assumptions (Fu
2008; Xiang and Xu 2011) and introduce several variables (Gan et al. 2005), and
then we formulate the problem into an integer linear model. Then we will design a
greedy algorithm in Sects. 103.3 and 103.4 are the simulation results. The con-
clusion is given in Sect. 103.5.
103.2 The Integer Linear Programming Model
103.2.1 Assumptions and Variables
103.2.1.1 Assumptions
• Once people choose one of the propulsion (oar- powered rubber rafts or
motorized boats) at first, they can not change again on the way;
• The duration of each trip ranges from 6 to 18 nights on the river;
• There are X trips, each trip has a given duration;
• There are Y camps distributed fairly uniformly throughout the river corridor;
• There is so enough fuel and power for each boat that no breakdown might occur
on the whole river trip;
• Each boat, controlled by the specialized staff, will run exactly on schedule;
• There are 180 days in the Big Long River’s rafting season;
• There are 8 h open for the river trip at daytime;
• Each rafting boat must stay in one camping site at night.
103.2.1.2 Variables
X: the total number of available trip boats;
Y: the total number of camping sites;
rik: the time when boat i arrives at camping site k;
dik: the time when boat i leaves from camping site k;
103 Integer Linear Programming Model and Greedy Algorithm 975
1; boat i occupys camping site k
xik ¼
0; otherwise
1; boat i and j meet at river between camping site k and k þ 1
ckij ¼
0; otherwise
k1 0; if dik [ djk and rikþ1 [ rjkþ1
cij ¼
1; otherwise
0; if dik \djk and rikþ1 \rjkþ1
ck2
ij ¼
1; otherwise
Ti: the total trip duration of boat i (measured in nights on the river);
Pi: from which day boat i start off (Pi is an integer) at the First Launch;
vi, min: the minimal speed of boat i;
vi, max: the maximum speed of boat i.
103.2.2 The Integer Linear Programming Model
The problem of camping along the Big Long River can be formulated into an
Integer Linear Programming Model (Liu et al. 2009; Wang 2010; Li and Wang
2009), in which the schedule time table of all given X boats (the total number of
available trip boats) can be obtained with the total meets among all boats in the
river be minimal.
103.2.2.1 One Essential Definition
Firstly, we define the open time for river trip. According to the given information,
we could define that the river trip is only allowed at daytime from 08:00 to 16:00
clock; for the other time, passengers have to stay at the camping site.
103.2.2.2 The Integer Linear Programming Model
As analyzed above, the objective function is constructed to minimize the total
meets between any pair of boats on the river. The constraints including: (1) the
river trip can only be allowed at daytime from 08:00 to 16:00; (2) the river trip
duration of each boat is an integer, ranging from 6 to18 nights; (3) no two sets of
campers can occupy the same site at the same time; (4) the season for river trip
only lasts 6 months (180 days). Based on the descriptions above, we formulate the
integer linear programming mode as follows:
976 Z. Li and X. Huang
X X
X X X
Y
minnbsp; z ¼ Cijk ð103:1Þ
i j k
s.t
0 rik dik ; i ¼ 1; 2; . . .; X; k ¼ 1; 2; . . .; Y ð103:2Þ
w w 225
rik diðk1Þ ; w¼ ð103:3Þ
vi max vi min Y þ1
24Pi þ 8 dio 24Pi þ 16 ð103:4Þ
! !
X
k Xk
24 Pi þ xis þ 8 dik 24 Pi þ xis þ 16 ð103:5Þ
s¼1 s¼1
! !
X
k X
k1
24 Pi þ xis þ 8 rik 24 Pi þ xis þ8 ð103:6Þ
s¼1 s¼1
X
Y
xis ¼ Ti ð103:7Þ
s¼1
xik dik rik Mxik ð103:8Þ
X
Y
6 Pi þ xis 180 ð103:9Þ
s¼1
8
>
> dik djk Mck1 ij
>
> k1
>
< riðkþ1Þ rjðkþ1Þ Mcij
k2
dik djk þ Mcij ð103:10Þ
>
>
>
> riðkþ1Þ rjðkþ1Þ þ Mck2
>
: k
ij
cij ck1 k2
ij þ cij 1
jrik rjk j 16 Mð2 xik xjk Þ ð103:11Þ
8
>
> rik 0
>
>
>
> x ik ¼ 0; 1
>
>
< ckij ¼ 0; 1
Pi 0 ðintegerÞ ð103:12Þ
>
>
> Ti 0
>
>
>
>
> i; j ¼ 1; 2; . . .; X
:
k ¼ 1; 2; . . .; Y
The objective function (103.1) is to minimum the total meets between any pair
of boats on the river.
103 Integer Linear Programming Model and Greedy Algorithm 977
Constraint (103.2) means the time when boat i leaves site k is later than that
when boat i arrives at site k.
Constraint (103.3) guarantees the time for boat i to travel from site k-1 to site
k is between the lower and upper bound.
Constraint (103.4) means that boat i will begin its trip in the Pith day at the open
time, where Pi ¼ 1; 2; . . .; 174.
Constraint (103.5) guarantees that boat i leaves camping site k at the open time.
Constraint (103.6) guarantees that boat i arrives at camping site k at the open
time.
Constraint (103.7) means the duration of boat i is Ti, where Ti is an integer
ranging from 6 to 18.
Constraint (103.8) describes the condition whether boat i occupies the camping
site k.
Constraint (103.9) guarantees all the boats will finish their river trips in six
months.
Constraint (103.10) describes the condition whether boat i and boat j meets on
the river.
Constraint (103.11) guarantees no two sets of campers can occupy the same
camping site at the same time.
Constraint (103.12) describes the value range of variables.
103.3 Greedy Algorithm
As a matter of fact, we could use Lingo software to solve this problem; however,
the problem scale is so large that the time consuming will be too long. So it’s not
wise enough to use Lingo software in this situation. Here we design a greedy
algorithm (Chen et al. 2008; Chen and Xu 2011; Su and Zhang 2011; Liang et al.
2005; Wang and Li 2008) to solve this problem.
103.3.1 Several Essential Supposes
According to our model, we will write a procedural by using the MATLAB
software, but before this we give some essential supposes again:
• To avoid the meet between any pair of boats on the river, the boats have to start
off by some gap time, and we define the gap time as follows: Gap time = 225/
(Y ? 1)/v, where v is the speed of boat.
• The boat will run at the same speed during the whole trip.
• The actual time of river trip for each boat every day is not more than 5 h.
• Passengers who want to have a river trip need to make a booking in advance,
and then we will divide the six months’ rafting season into several cycles
978 Z. Li and X. Huang
according to the passengers’ booking. During one cycle time, we will arrange
the boats by their duration like this: boat of 6 nights first, and then it will be 7,
8,… in turn, boats of 18 nights will be the last to be arranged.
• The number of boats (represented with Q) arranged every day will depend on its
duration and could be calculated by Q = [Y/Tdur]. For example, for boats whose
duration is 6, the maximum number of this type boats we can arrange every day
is [Y/6]; for boats whose duration is 7, the maximum number we can arrange
every day is [Y/7]. Why? We can explain this by the following graph (see
Fig. 103 1).
Suppose Y = 24, then for boats whose duration is 6, the maximum number of
boats we can arrange every day is 4:
In Fig. 103.1, the bold horizontal line denotes the riverbank while the thin
vertical line represents the camping site and the arrow symbolizes the boat. At the
first day we arrange 4 boats. Then we can see these four boats as a whole and its
whole trip process could be described vividly in the above graph. By this method
we can guarantee to utilize the camping sites in the best possible way. As to other
boats of different duration, we can draw the similar graph like the above.
103.3.2 The Greedy Algorithm
The greedy algorithm can be described as follows.
BEGIN
Fig. 103.1 Trip process of boats arranged in the first and the second days
103 Integer Linear Programming Model and Greedy Algorithm 979
Classes all X boats into 13 groups according to their duration. Denote the
number of boats whose duration is i by X ðiÞ; i ¼ 6; 7; . . .; 18.
Provided that the maximum river trip time every day is no more than 5 h,
calculate the maximum number of boats whose duration is i we can arrange every
day, denoted it by m(i), where mðiÞ ¼ ½Y=i.
for i ¼ 6 : 18
Arrange the schedule of all boats with duration i.
For all X(i) boats with duration i, arrange them to start off from the First Launch
in XðiÞ=mðiÞ consecutive days, with the time gap between two successive boats is
225/(Y ? 1)/v daytime.
After all X(i) boats with duration i arranged, we can arrange the following
X(i ? 1) boats with duration i ? 1 in the following Xði þ 1Þ=mði þ 1Þ days…, till
all X boats are arranged.
end
END
103.4 Simulation Results
In this section, we will do some simulations of the model and algorithm described
above.
Supposing Xi denotes the number of boats whose duration is iði ¼ 6; 7;. . .; 18Þ,
where
X6 ¼5; X7 ¼ 6; X8 ¼ 11; X9 ¼ 12; X10 ¼ 10; X11 ¼ 8; X12 ¼ 6; X13 ¼ 15;
X14 ¼ 7; X15 ¼ 12; X16 ¼ 12; X17 ¼ 11; X18 ¼ 9:
Given Y = 53.
We run our procedural coded by MATLAB software, the simulation result are
as follows:
D1: 5 boats whose duration is 6 nights and 3 boats whose duration is 7 nights
will start off in the first day;
D2: 3 boats whose duration is 7 nights and 4 boats whose duration is 8 nights
will start off in the second day;
D3: 6 boats whose duration is 8 nights will start off in the third day;
D4: 1 boat whose duration is 8 nights and 5 boats whose duration is 9 nights
will start off in the fourth day;
D5: 5 boats whose duration is 9 nights will start off in the fifth day;
…
D31: 2 boats whose duration is 18 nights will start off in the 31st day;
D32: 2 boats whose duration is 18 nights will start off in the 32nd day;
D33: 1 boat whose duration is 18 nights will start off in the 33rd day.
The detail rafting schedule of all boats can be described in Fig. 103.2.
980 Z. Li and X. Huang
Fig. 103.2 The detail schedule of all boats obtained by greedy algorithm
103 Integer Linear Programming Model and Greedy Algorithm 981
According to the simulation results, we find that we can arrange the 124 boats
in about 50 days. This inspired us that we can divided the rafting season
(180 days) into several (for example 3) periods. Arrange X/3 boats in every period
respectively according to the greedy algorithm. This can avoid all boats with same
duration be arranged in several concentrate days.
Remarks: By using the greedy algorithm, we can give a solution to the problem;
however this solution might not be the optimal one. But based on this solution, we
can take some measures to improve. By continuous adjusting, we can finally find a
satisfied solution.
103.5 Conclusion
The problem of Camping along the Big Long River is very complex and the
solution should be of great openness. In this paper, we formulate this problem into
an integer linear programming model and design a greedy algorithm to arrange the
schedule of boats. Then by doing some simulations with this algorithm, we give a
solution to the problem. The results show that this method can obtain the optimal
solution by continuous improving. Furthermore; we can estimate the capacity of
the river by this greedy algorithm.
Although river trip is quite interesting and exciting, it is also very risky and
need some spirit of adventure. Any accident may happen during this process, such
as bad weather, passenger’s injuries and so on. These potential factors might have
a great impact on the supervisor’s decision and management. We don’t take these
factors into account. In the future, we will consider these factors in the model and
algorithm.
Acknowledgments This work is supported by National Natural Science Foundation of China
under Grant No.11131009 and the Funding Project for Academic Human Resources Develop-
ment in Institutions of Higher Learning under the Jurisdiction of Beijing Municipality
(No.PHR201006217).
The author gratefully acknowledges the valuable discussions and comments that Huang qiuai,
Chen Qing, Li Tingting and Wang Shan provide for this paper.
References
Chen H, Xu L (2011) Greedy algorithm computing minkowski reduced lattice bases with
quadratic bit complexity of input vectors. Chin Ann Math 32B(6):857–862
Chen D, You S, Han B (2008) Algorithm to create and rate sudoku puzzles. MCM PRO BLEM B
Fu Y (2008) The study of whitewater drifting tourism product development based on tourists
experience at Hongkou scenic spots of Dujiangyan city. Master dissertation of Southwest
Jiaotong University, pp 33–36
Gan Y, Tian F, Li W, Li M, Chen B, Hu Y (2005) Operations research, 3rd edn, vol 6. Tsinghua
University Press, Beijing, pp 122–126
982 Z. Li and X. Huang
Li Z, Wang H (2009) A feasible mathematical model for the marshalling and dispatching problem
of railway. Internet Fortune 11:92–93 (in Chinese)
Liang L, Chen Y, Xu M (2005) Schedule arrangement algorithm based on greedy method.
J Yunnan Normal University (Nat Sci Edn) 25(3):9–16 (in Chinese)
Liu D, Zhao J, Han D, Chen Z (2009) Model and algorithm for the marshalling and dispatching
problem of railway freight train. Math Pract Theory 39(16):162–172 (in Chinese)
Su F, Zhang J (2011) Research on greedy algorithm to solve the activity arrangement. Softw
Guide 10(12):43–44 (in Chinese)
Wang P (2010) The study on train operation simulation: real-time scheduling model and
algorithm. Master dissertation of Beijing Jiaotong University, pp 9–11
Wang B, Li Z (2008) Research and implementation of automatic course system based on greedy
algorithm. Comput Eng Design 29(18):4843–4846
Xiang W, Xu C (2011) Analysis of the factors influencing whitewater rafting experience. J Guilin
Inst Tourism 3(6):56–60
Chapter 104
Research on End Distribution Path
Problem of Dairy Cold Chain
Zhen-ping Li and Shan Wang
Abstract The vehicle routing problem of dairy cold chain end distribution with
random demand and time window is investigated in this paper. Considering the
characteristics of dairy cold chain end distribution, the chance constrained theory
and the penalty function is introduced to establish a mathematical model of this
problem. A scanning-insert algorithm to solve the model is proposed. The algo-
rithm can be described as: firstly, according to the capacity of the vehicle and time
window restrictions, the customers are divide into several groups by scan algo-
rithm; then find a feasible routing line for each group of customers; finally, using
the idea of recent insertion method to adjust the vehicle route and find the final
optimal distribution vehicle route.
Keywords Dairy cold chain Random demand Mathematical model Scan-
ning-insert algorithm
104.1 Introduction
Vehicle routing problem with time windows refers to the transportation problem in
general under the premise of customer‘s requirements of time window. Solomon
and Desrosiers etc. (Solomon 1987; Solomon and Desrosiers 1988) consider joined
time window constraint to the general vehicle routing problem in 1987. Desorchers
et al. (1988) used to concise summary and summarized various kinds of method
solving vehicle routing problem with time windows further in 1988. Sexton and
Z. Li (&)
School of Information, Beijing Wuzi University, Beijing 101149, China
e-mail: [email protected]
S. Wang
Department of Postgraduate, Beijing Wuzi University, Beijing 101149, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 983
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_104,
Ó Springer-Verlag Berlin Heidelberg 2013
984 Z. Li and S. Wang
Choi (1986) used the Decomposition method proposed by Bender to solve the
single vehicle pick-up and delivery problem with time window restriction.
Chance constrained mechanism use the default value constraint error to return
to probability in the essence of vehicles service process, and additional cost caused
by service failure is not within planning (Chen 2009). Stewart (Stewart and Golden
1983) and Laporte (Laporte et al. 1989) used respectively chance constrained
program change SVRP into equivalent deterministic VRP under some assump-
tions. Dror (Dror and Trudeau 1986) used Clark-Wright algorithm to solve vehicle
routing optimization problem.
This paper’s main consideration is regular route for distribution mode under the
target of minimizing the cost. It means that the customer or the number of nodes
and their position are fixed in every day visit, but each customer’s demand is
different, and their demands meet Normal Distribution.
104.2 Analysis of the Cost in Cold Chain Logistics
Distribution
104.2.1 Fixed Costs
Distribution center has to pay for the fixed costs for the use of each vehicle. These
costs include the driver’s wages, insurance, lease rental of the vehicle.
X
m
c1 ¼ fk
k¼1
104.2.2 The Transportation Cost
The transportation cost of a vehicle is the relevant expenses caused by travel,
which includes fuel consumption, maintenance, maintenance fee.
n X
X n
c2 ¼ ckij dij xkij
i¼1 j
104.2.3 The Cost of Damage
In the cold chain, the main factors causing fresh products damaged are storage
temperature, food of microbes in water activity, PH value, oxygen content (Wang
2008). Assume the damage rate is k, the unit value of the products is P, and
capacity of vehicle k is Qk.
104 Research on End Distribution Path Problem 985
c3 ¼ PkQk
104.2.4 The Cost of Energy Consumption
The heat load vehicle refrigeration equipment is mainly due to difference heat
transfer between the vehicle body inside and outside. Suppose the temperature
difference between inside and outside of the vehicle is fixed in a certain period,
then the cost of energy consumption can be expressed as:
X
m
c4 ¼ A ðek sk Þ
k¼1
104.2.5 Penalty Cost
Soft time window can allow the distribution vehicle to arrive outside the time
window, but outside the appoint time must be punished. Delivery time can be
divided into three categories: service in advance, service by time window, service
delay (Zhan 2006; Thangiah et al. 1991), which is shown in Fig. 104.1.
(1) Service in advance is that the distribution vehicles arrive in time window [a, g).
Immediate delivery may cause customers’ inconvenience and complaint, but it
can reduce the energy consumption.
(2) Service by time window means that the distribution vehicle arrives in the time
window [g, h]. Immediate delivery and the energy costs relate to time is a
constant.
(3) Service delay means that the distribution vehicle arrive in time window (h, b].
Immediate delivery and the energy and relevant penalty costs will increase.
Fig. 104.1 Time window M
Time t
a g h b
986 Z. Li and S. Wang
In conclusion, the penalty cost function is:
8
>
> M; tik \tai or tik [ tbi
< k k
ðtgi ti Þh; tai tik \tgi
uðtik Þ ¼
>
> 0; tgi tik thi
: k
ðti thi Þg; thi \tik tbi
104.3 Mathematical Model
104.3.1 Related Hypothesis
(1) The model only considers the pure delivery problem.
(2) There are enough delivery vehicles in the distribution center, and each vehi-
cle’s capacity is limited.
(3) The stock in the distribution center is enough for all the customers and all
customers’ time windows are known.
(4) All vehicles start off from the distribution center, and return to the distribution
center again after completion.
(5) The position of each customer is given, but quantities demand Di of each
customer i is random, it satisfies a normal distribution Di Nðli ; r2i Þ, and they
are mutual independent.
(6) The route of each vehicle is determined and will not change in the deliver
road.
(7) Products in the transportation process can stay in a fixed transport temperature,
and the vehicle’s energy consumption is only related to their travel time.
104.3.2 Symbols and Mathematical Model
fk : Fixed cost of vehicle k;
ckij : The unit transportation cost of vehicle k in the travel road from customer i
to customer j;
dij : Distance from customer i to customer j;
A : Unit cost of energy consumption;
ek : The time when vehicle k return to the distribution center;
sk : The time when vehicle k start off from the distribution center;
tk : The time when vehicle k arrives at customer i;
Di : Demand of customer i;
1; If vehicle k come to customer j from customer i
xkij ¼
0; otherwise
104 Research on End Distribution Path Problem 987
1; If vehicle k service for customer i
yki ¼
0; otherwise
The mathematical model can be formulated as follows:
X
m X m X
n X
n m X
X n
min z ¼ ½f k þ PkQk þ A ek sk þ ckij dij xkij þ uðtik Þ
k¼1 k¼1 i¼1 j¼1 k¼1 i¼1
ðs:t:Þ
X
m
m; i¼0
yki ¼ ð104:1Þ
k¼1
1; i ¼ 1; 2; . . .; n
X
n
ykj ¼ xkij ; i 6¼ j; j ¼ 1; 2; . . .; n ð104:2Þ
i¼1
X
n X
n
xkip xkpj ¼ 0 p ¼ 1; 2; . . .; n ð104:3Þ
i¼0 j¼0
dij
tjk tik þ ð1 xkij ÞM; j ¼ 1; 2; . . .; n; k ¼ 1; 2; . . .; m ð104:4Þ
v
doi
tik sk þ ð1 xkoi ÞM; i ¼ 1; 2; . . .; n; k ¼ 1; 2; . . .; m ð104:5Þ
v
djo
ek tjk þ ð1 xkjo ÞM; j ¼ 1; 2; . . .; n; k ¼ 1; 2; . . .; m ð104:6Þ
v
tai tik tbi ; i ¼ 1; 2; . . .; n; k ¼ 1; 2; . . .; m ð104:7Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Xn Xn
k 1
li yi þ U ðbÞ r2i yki Qk ð104:8Þ
i¼1 i¼1
yki ¼ 0; 1 i ¼ 1; 2; . . .; n; k ¼ 1; 2; . . .; m;
xkij ¼ 0; 1 i; j ¼ 1; 2; . . .; n; i 6¼ j; k ¼ 1; 2; . . .; m;
The objective function minimum the total cost.
Constraint (104.1) means that each customer will be serviced by one vehicle,
and each vehicle’s route start from and ended at the distribution center.
Constraint (104.2) means that if vehicle k arrives at customer j, then it must
service for customer j.
Constraint (104.3) means that if vehicle k arrives at customer p, then it must
leave from customer p after finishing service.
988 Z. Li and S. Wang
Constraints (104.4–104.5) are the conditions that the arriving time for vehicle k
come to customers i and j must satisfy.
Constraints (104.6–104.7) are the time window restrictions.
Constraint (104.8) means that the probability of each vehicle’s capacity is no
less than the total demands of all the customers it serviced is great than b.
104.4 Algorithm
The algorithm can be described as follows:
A. Set up the polar coordinates system:
B. Partition the customers into several groups.
(1) Starting from zero Angle and rotating along counterclockwise direction,
pick up the customer into a group one by one until the total demands of all
customers in this group exceed a vehicle’s capacity limit.
(2) For each group of customers, ordered them into sequence according to the
demand time window and form initial solution route. Then determine
whether the solution route satisfies the time window constraints.
(3) If the solution route satisfies the time window constraints, then a new
group is created. Go to (1). Continue to rotate along counterclockwise
direction, and the rest of the customers will be added one by one into a new
group. Otherwise, if the solution route does not satisfy the time window
constraints, we can adjust the order of customers and find another feasible
solution route sequence. If no feasible solution route satisfying time
window constraints exists, delete a customer who does not satisfy time
window constraints from the group, and add it into the next new group as a
necessary customer. Go to (1).
C. Repeat step B until all customers are partitioned into groups
D. Optimize the vehicle’s route by the recent insertion method in each group
(1) Select the earliest time requirements customer to form a sub-route with
distribution center 0.
(2) Insert customer point vk as the next demand point according to time
window series. Find an arc ðvi ; vj Þ in the sub-route, insert customer node vk
between customer nodes vi and vj to form a new sub- route such that the
new sub-route satisfies time window and the cost increment is minimum.
(3) Repeat step (2), until all customer nodes are added into a route.
E. Repeat step D, until all groups are optimized.
104 Research on End Distribution Path Problem 989
104.5 Simulation Results
There are 30 customers needed to be serviced. Suppose all vehicles are of the same
type. The capacity of each vehicle is 48; fixed costs is 100; the vehicle speed is
30 km/h; Unit of energy consumption cost $0.5 per minute; unit distance trans-
portation cost is $5 per kilometer; punishment coefficient h is 0.4 and g is 0.5; b is
95 %, k is 0.01, P is 100. The experimental data is random generation through the
computer under experimental hypothesis.
A. Set up the polar coordinates system
B. Partition the customers into several groups
(1) Starting from zero Angle and rotating along counterclockwise direction,
we can find the first group customers are 2, 3, 5, 6, 7, and 9. The detail
information is listed in Table 104.1.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Q1 ¼ 3 þ 4 þ 9 þ 5 þ 7 þ 11 þ 1:65 1 þ 2 þ 3 þ 3 þ 4 þ 5 ¼ 45:4 \ 48
(2) Find an initial solution sequence 0-2-3-7-5-6-9-0, The initial route is
shown in Fig. 104.2.
(3) Continue to rotate counterclockwise to build new group. Repeat the pro-
cess until all customers are picked into a group.
C. Optimize the initial route of each group by the recent Insertion method
(1) Select customer 2 whose requirement time window is the earliest to form a
sub-route with distribution center 0. Insert customer 3 as the next customer
point according to its requirement time window. Then customer 3 will be
inserted between distribution center 0 and customer 2, forming a new sub-
route 0-2-3-0 satisfying time window and with minimal cost increment.
See Table 104.2.
(2) Insert customers 7, 5, 6, 9 into the sub-route one by one. We can find the
optimal route 0-2-7-9-3-6-5-0. The objective function value is 796.6. The
optimal route is shown in Fig. 104.3 and Table 104.3.
Similarly, we can use the same method to find the optimal routes in the other
groups. The results are shown in Fig. 104.4.
Table 104.1 Basic information table
Customer X Y Requirements T Accept time Demand quantity
2 22 10 5:00–5:30 4:00–6:00 Q * N(3,1)
3 12 20 5:50–6:30 5:00–7:00 Q * N(4,2)
7 25 30 6:10–6:40 5:10–7:20 Q * N(9,3)
5 10 2 6:30–7:20 6:00–7:50 Q * N(5,3)
6 8 15 7:10–7:40 6:20–8:00 Q * N(7,4)
9 13 35 7:20–7:50 6:20–8:40 Q * N(11,5)
990 Z. Li and S. Wang
40
35 9
30 7
25
20 3
15 6
10 2
5
5
00 5 10 15 20 25 30
Fig. 104.2 Initial route
Table 104.2 The time table of sub-route
dij 24 14 23
Customer 0 – 2 – 3 – 0
Time 4:30 5:18 5:46 6:32
Dt 0 -4 0
40
35 9
30 7
25
20 3
15 6
10 2
5
5
0
0 5 10 15 20 25 30
Fig. 104.3 The optimal route
Table 104.3 Time table of the optimal route
dij 24 20 13 15 6 13 10
Customer 0 – 2 – 7 – 9 – 3 – 6 – 5 – 0
Time 4:30 5:18 5:58 6:24 6:54 7:06 7:32 7:52
Dt 0 -12 -56 24 0 12
104 Research on End Distribution Path Problem 991
Fig. 104.4 The optimal 11 9
routes of all groups 14
4
7
12
8 3
13 6
10 2
1
15 5
17
26 28
18 20 25
22 29
23 27 30
19
16 21 24
104.6 Conclusion
The vehicle routing problem of dairy cold chain end distribution with random
demand and time window is investigated in this paper. A mathematical model is
constructed, and an algorithm is proposed.
Vehicle routing problem with time windows is a real problem the enterprises
facing with at the end of city distribution. It is obvious that to pursuit minimum
cost may cause to drop the quality of service and eventually lead to the loss of
customers. To establish a suitable mode of long-term sustainable development, the
enterprise should find a balance between service quality and cost. As a result, the
enterprises could meet customer requirements with the highest level of service and
minimum cost. In addition, this paper did not consider the asymmetry of road
network and the time handling factors. In the future, we will investigate the
problem with these factors.
Acknowledgments This work is supported by National Natural Science Foundation of China
under Grant No.11131009 and the Funding Project for Academic Human Resources Develop-
ment in Institutions of Higher Learning under the Jurisdiction of Beijing Municipality
(No.PHR201006217).
References
Chen B (2009) Ant colony optimization algorithm for the vehicle routing problem in the research
on the application (in Chinese). Harbin Industrial University, Harbin
Desorchers M, Lenstra J, Savelsbergh M, Esoumis (1988) Vehicle routing with time windows
optimization and apporximation, vehicle routing: methods and studies. North-Holland,
Amsterdam, pp 64–84
992 Z. Li and S. Wang
Dror M, Trudeau P (1986) Stochastic vehicle routing with modified savings algorithm. Eur J Oper
Res 23:228–235
Laporte G, Louveaux F, Mercure H (1989) Models and exact solutions for a class of stochastic
location-routing problems. Eur J Oper Res 39:71–78
Sexton TR, Choi YM (1986) Pickup and delivery of partial loads withSoft time windows. Am J
Math Manage Sci 6(4):369–398
Solomon MM (1987) Algorithm for the vehicle routing and scheduling problems with time-
windows constraints. Oper Res 35(2):254–265
Solomon MM, Desrosiers J (1988) Time windows constrained routing and scheduling problems.
Transp Sci 22(2):1–13
Stewart WR, Golden BL (1983) Stochastic vehicle routing: a comprehensive approach. Eur J
Oper Res 14:371–385
Thangiah S, Nygard K, Juell PG (1991) Agenetic algorithms system for vehicle routing with time
windows. In: Miami proceedings of the seventh conference on artificial intelligence
applications, Florida, pp 322–325
Wang Y (2008) Cold-chain logistics distribution center mode study (in Chinese). Changsha
University of Science and Technology, Changsha
Zhan S (2006) Parallel algorithm with the soft time Windows in the vehicle routing problem of
the application (in Chinese). Wuhan University of Science and Technology, Wuhan
Chapter 105
Improved Evolutionary Strategy Genetic
Algorithm for Nonlinear Programming
Problems
Hui-xia Zhu, Fu-lin Wang, Wen-tao Zhang and Qian-ting Li
Abstract Genetic algorithms have unique advantages in dealing with optimization
problems. In this paper the main focus is on the improvement of a genetic algorithm
and its application in nonlinear programming problems. In the evolutionary strategy
algorithm, the optimal group preserving method was used and individuals with low
fitness values were mutated. The crossover operator uses the crossover method
according to the segmented mode of decision variables. This strategy ensured that
each decision variable had the opportunity to produce offspring by crossover, thus,
speeding up evolution. In optimizing the nonlinear programming problem with
constraints, the correction operator method was introduced to improve the feasible
degree of infeasible individuals. MATLAB simulation results confirmed the
validity of the proposed method. The method can effectively solve nonlinear pro-
gramming problems with greatly improved solution quality and convergence speed,
making it an effective, reliable and convenient method.
Keywords Nonlinear programming Genetic algorithm Improved evolutionary
strategy Correction operator method
105.1 Introduction
Nonlinear programming problem (NPP) had become an important branch of
operations research, and it was the mathematical programming with the objective
function or constraints being nonlinear functions. There were a variety of tradi-
tional methods to solve nonlinear programming problems such as center method,
gradient projection method, the penalty function method, feasible direction
H. Zhu (&) F. Wang W. Zhang Q. Li
School of Engineering, Northeast Agriculture University, Harbin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 993
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_105,
Springer-Verlag Berlin Heidelberg 2013
994 H. Zhu et al.
method, the multiplier method. But these methods had their specific scope and
limitations, the objective function and constraint conditions generally had con-
tinuous and differentiable request. The traditional optimization methods were
difficult to adopt as the optimized object being more complicated. Genetic algo-
rithm overcame the shortcomings of traditional algorithm, it only required the
optimization problem could be calculated, eliminating the limitations of optimi-
zation problems having continuous and differentiable request, which was beyond
the traditional method. It used the forms of organization search, with parallel
global search capability, high robustness and strong adaptability, and it could
obtain higher efficiency of optimization. The basic idea was first made by Pro-
fessor John Holland. The genetic algorithm had been widely used in combinatorial
optimization, controller’s structural parameters optimization etc. fields, and had
become one of the primary methods of solving nonlinear planning problems
(Operations research editorial group 2005; Bazarra and Shetty 1979; Bi et al. 2000;
Liang et al. 2009; Holland 1975; Hansen 2004; Saleh and Chelouah 2004; Uidette
and Youlal 2000; Lyer et al. 2004).
In this paper, the evolution strategy was improved after analyzing the process of
the genetic algorithm and the improved algorithm took full advantages of genetic
algorithm to solve unconstrained and constrained nonlinear programming prob-
lems. In the MATLAB environment, the numerical example showed that the
proposed improved genetic algorithm for solving unconstrained and constrained
nonlinear programming was effective, and the experiment proved it was a kind of
algorithm with calculation stability and better performance.
105.2 Nonlinear Programming Problems
The nonlinear programming problems could be divided into unconstrained prob-
lems and constrained problems (Operations research editorial group 2005; Sui and
Jia 2010). We presented the mathematical model here in its general form.
The unconstrained nonlinear programming model:
min f ð X Þ X 2 En ð105:1Þ
T
where, the independent variable X = (x1, x2, …, xn) was an n dimensional vector
(point) in Euclidean space. It was the unconstrained minimization problem that
was for the minimum point of the objective function f(X) in En.
The constrained nonlinear programming model:
min f ð X Þ X 2 En
s:t: hi ð X Þ ¼ 0; i ¼ 1; 2; . . .; m ð105:2Þ
gj ð X Þ 0; j ¼ 1; 2; . . .; l
where, ‘‘min’’ stood for ‘‘minimizing’’ and symbol ‘‘s.t’’ stood for ‘‘subject to’’.
It was the unconstrained minimization problem that was for the minimum point of
105 Improved Evolutionary Strategy Genetic Algorithm 995
the objective function f(X) in En. Here, hi(X) = 0 and gj(X) C 0 were the con-
strained conditions.
For max f(X) = -min[-f(X)], only the minimization problem of objective
function was needed to take into consideration without loss of generality.
If some constrained conditions were ‘‘B’’ inequality, they were needed to be
multiplied at both ends of the constraints by ‘‘-1’’. So we could only consider the
constraint in the form of ‘‘C’’.
105.3 Analysis and Description of the Improved
Genetic Algorithm
Based on the simple genetic algorithm, the following gave the analysis design and
description of the algorithm which improved the genetic evolution strategy of
genetic algorithm.
105.3.1 Encoding and Decoding
We used the binary encoding and multi-parameter cascade encoding. It meant that
we made each parameter encoded by means of the binary method and then con-
nected the binary encoded parameters with each other in a certain order to con-
stitute the final code which represented an individual including all parameters. The
bit string length depended on the solving precision of specific problems, the higher
precision we required, the longer the bit string.
If the interval of someone parameter was [A, B] and the precision was c digits
after decimal point, then the calculation formula for bit string length was:
ðB AÞ 10c 2L ð105:3Þ
Here, L took the smallest integer which made the above equation valid.
If the interval of someone parameter was [A, B], the corresponding substring in
the individual code was bL bL1 bL2 b2 b1 , then its corresponding decoding
formula was:
X L
BA
X ¼Aþð bi 2i1 Þ L ð105:4Þ
i¼1
2 1
105.3.2 Production of the Initial Population
There were two conditions when producing initial population. One was to solve
the unconstrained problem; the other was to solve the constrained problem.
996 H. Zhu et al.
Suppose the number of decision variables was n, the population scale was m, ai
and bi were lower limit and upper limit of a decision variable respectively. For the
unconstrained problem, binary encoding was adopted to randomly produce initial
individuals of the population.
For constrained problems, the initial population could be selected at random
under certain constraint conditions. It also could be produced in the following
manner:
First, a known initial feasible individual X(0)
1 was given artificially. It met the
following conditions:
ð0Þ ð0Þ ð0Þ ð0Þ ð0Þ
gj ðX1 Þ ¼ gj ðX11 ; X12 ; X13 ; . . .; X1n Þ [ 0
The other individuals were produced in the following way (Wang et al. 2006):
ð0Þ
X2 ¼ A þ r2 ðB AÞ ð105:5Þ
Here, A = (a1, a2, a3,…,an)T, B = (b1, b2, b3,…,bn)T, r2 = (r21, r22, r23,…,
r2n)T, random number rij 2 Uð0; 1Þ.
ð0Þ
Then checking whether X2 satisfied the constraints or not. If the constraints
ð0Þ
were satisfied, another individual would produce as X2 . If the constraints were
ð0Þ
not satisfied, X2 would be corrected by correction operator.
105.3.3 Correction Operator Method
When genetic algorithm was applied to deal with constrained nonlinear pro-
gramming problems, the core problem was how to treat constraint conditions.
Solving it as unconstrained problems at first, checking whether there were con-
straint violations in the search process. If there were no violations, it indicated that
it was a feasible solution; if not, it meant it was not a feasible solution. The
traditional method of dealing with infeasible solutions was to punish those
infeasible chromosomes or to discard infeasible chromosome. Its essence was to
eliminate infeasible solution to reduce the search space in the evolutionary process
(Gao 2010; Wang et al. 2003; Ge et al. 2008; He et al. 2006; Tang et al. 2000;
Wang and Cao 2002). The improved evolution strategy genetic algorithm used the
correction operator method, which selected certain strategy to fix the infeasible
solution. Different from the method of penalty function, the correction operator
method only used the transform of objective function as the measure of the
adaptability with no additional items, and it always returned feasible solution. It
had broken the traditional idea, avoided the problem of low searching efficiency
because of refusing infeasible solutions and avoided early convergence due to the
introduction of punishment factor, and also avoided some problems such as the
result considerably deviating constraint area after mutation operation.
If there were r linear equations of constraints, and the linear equations’ rank
was r \ n, all decision variables could be expressed by n - r decision variables.
105 Improved Evolutionary Strategy Genetic Algorithm 997
Taking them into inequality group and the objective function, the original
n decision variables problem became n - r decision variables problem with only
inequality group constraints. So we could only consider problems with only
inequality group constraints. The production of initial individuals, offspring pro-
duced by crossover operation and individuals after mutation, all were needed to be
judged whether they met the constraints, if not, fixed them in time. Such design of
genetic operation made solution vectors always bounded in the feasible region.
The concrete realization way of correction operator was:
Each individual was tested whether it satisfied the constraints. If so, continued
the genetic operation; if not, let it approach the former feasible individual
ð0Þ
(assumed X1 and the former feasible individual should be an inner point). The
approaching was an iterated process according to the following formula:
ð0Þ ð0Þ ð0Þ ð0Þ
X2 ¼ X1 þ aðX2 X1 Þ ð105:6Þ
where, a was step length factor. If it still did not satisfy the constraint, then the
accelerated contraction step length was used, that was a = (1/2)n, here, n was
search times.
Big step length factor could affect the constraint satisfaction and reduce the
repairing effect and even affect the search efficiency and speed, whereas, too small
step length factor could not play the role of proper correction. So the method of
gradually reducing the step length factor could both protect the previous correction
result and give full play to correction strategy.
ð0Þ
Thus X2 was made to feasible individual after some times of iteration, then
ð0Þ ð0Þ
X3 was produced as X2 and become feasible individual. In the same way, all the
needed feasible individuals were produced. For binary genetic algorithm, these
feasible individuals were phenotype form of binary genetic algorithm. Real coding
individuals were converted into binary string according to the mapping relation-
ship between genotype and phenotype. Then the feasible individuals of binary
genetic algorithm were obtained.
This kind of linear search way of infeasible individual moving to the direction
of feasible individual had the advantage of improving infeasible individual, ini-
tiative guiding infeasible individuals to extreme point of population, making the
algorithm realize optimization in global space. This paper introduced the correc-
tion operator to improve the feasibility of infeasible individuals. This method was
simple and feasible. And the treatment on infeasible individuals was also one
novelty of improving evolution strategy of genetic algorithm.
105.3.4 Fitness Functions
If the objective function was for minimal optimization, the following transfor-
mation was applied (Wang et al. 2007):
998 H. Zhu et al.
cmax f ðxÞ f ðxÞ\cmax
Fitðf ðxÞÞ ¼ ð105:7Þ
0 other
Here, cmax was an estimated value which was enough large for the problem.
If the objective function was for maximal optimization, the following trans-
formation was applied:
f ðxÞ cmin f ðxÞ [ cmin
Fitðf ðxÞÞ ¼ ð105:8Þ
0 other
Here, cmin was an estimated value which was enough small for the problem.
105.3.5 Selection Operator
Selection operator used the roulette selection method. The selection probability of
individual i:
X
m
ps ¼ f i = fi ð105:9Þ
i¼1
105.3.6 Crossover Operator
The number of decision variables might be more in practical problems. Because
binary encoding and multiparameter cascade encoding were adopted, the one point
crossover would make only one decision variable cross in a certain position in this
encoding mode, leaving no crossover for other variables. So the segmented
crossover mode of decision variables was used, giving each decision variable the
probability pc of single point crossover. Each decision variable had a cross
opportunity to produce offspring. This improvement was also another novelty of
evolutionary strategy of genetic algorithm.
105.3.7 Mutation Operator
Alleles of some genes were randomly reversed according to mutation probability
pm. Parent population individuals and child population individuals after crossover
were sorted together according to their fitness values before mutation and only
individuals with low fitness values were mutated. Thus not only good schema
could avoid being destroyed, but also mutation probability could be appropriately
increased, so generating more new individuals. It was good to increase the pop-
ulation’s diversity, to traverse all of the state, and to jump out local optimum.
105 Improved Evolutionary Strategy Genetic Algorithm 999
105.3.8 Population Evolution
In the process of population evolution, parent population individuals and child
population individuals after crossover were put together to form a new temporary
population, and the fitness value of each individual in the new temporary popu-
lation was calculated, m individuals with high fitness values were preserved, then
m individuals with low fitness values were mutated and the mutated m individuals
and the m previous preserved individuals were put together to form a new tem-
porary population. Thereafter individuals in the new temporary population were
sorted according their fitness values and m individuals with high fitness value were
selected as the next generation to accomplish the population evolution.
The evolution method was based on the traditional elite preserving method,
realizing preserving optimal group. The advantage of this method was to reduce
possibilities of optimal solution being destroyed by crossover or mutation in the
process of evolution. Moreover, premature convergence was avoided which might
be present in traditional elite preserving method because all individuals approa-
ched one or two individuals with high fitness values quickly. This was another
novel place of improving the evolution strategy of genetic algorithm.
105.3.9 Algorithm Stopping Criteria
Two criteria were adopted to terminate algorithm:
(1) The number of generations was more than a preset value;
(2) The difference of fitness value between two successive evolutions was less
than or equal to a given precision, namely to meet the condition:
jFitmax Fitmin j e ð105:10Þ
Here, Fitmax was the individual’s maximum fitness value of a population; Fitmin
was the individual’s minimum fitness value of a population.
105.4 Experimental Data and Results
105.4.1 Experimental Data and Parameters
In the experiment, simulations of two examples were used to validate the cor-
rectness of the algorithm and to test the performance of the algorithm. The
hardware environment in the experiment were Intel Pentium Dual-Core
[email protected] GHz, 2 GB RAM. The operating system was Microsoft Windows
XP, compile environment was MATLAB 7.11.0 (R2010b).
1000 H. Zhu et al.
105.4.2 Experimental Results and Analysis
In the below table, the interval lower bound was a, the interval upper bound was b,
the precision was c digits after decimal point, the population size was m, the
maximum evolution generation was T.
Example 1:
X
n
min f1 ð xÞ ¼ x2i ð105:11Þ
i¼1
f1(x) was a continuous, convex, single peak function. It was an unconstrained
optimization problem. Only one global minimum in the 0, the minimum was 0. We
selected n = 2, n = 5, n = 10 in the simulation experiment to verify the cor-
rectness of Improved Evolutionary Strategy Genetic Algorithm (IESGA). 100
times were executed for f1(x) with crossover probability 0.75, mutation probability
0.05, the end precision 0. All runs converged to the optimal solution. Parameter
settings and calculation results were shown in Table 105.1:
Observing the optimization results in Table 105.1, Improved Evolutionary Strat-
egy Genetic Algorithm had faster computing speed and higher accuracy, and could
robustly convergence to global optimal solution. With the increment of the number of
decision variables, the number of generation to obtain the optimal solution for the first
time also increased. This accorded with the objective law, and was also correct.
Example 2:
max f2 ð xÞ ¼ 2x21 þ 2x1 x2 2x22 þ 4x1 þ 6x2
s:t: 2x21 2x2 0
ð105:12Þ
x1 þ 5x2 5
x1; x2 0
The objective function f2(x) was a quadratic, polynomial function. Under the
conditions of inequality, linear and nonlinear constraints, the theoretical optimal
value f2(0.658, 0.868) = 6.613. In the simulation experiment, the crossover
probability was 0.75, mutation probability was 0.05, the end of precision was 0,
the maximum number of evolution generation was 70, 100 times were executed for
f2(x), and all converged to the optimal solution. The comparison of simulation
results which used methods of Feasible Direction (FD), Penalty Function (PF)
(Tang and Wang 1997) and Improved Evolutionary Strategy Genetic Algorithm
(IESGA) was shown in Table 105.2.
Observing the optimization results in Table 105.2, the result of using Improved
Evolutionary Strategy Genetic Algorithms was better than the two others, and the
optimal solution got the theoretical value. This showed that using Improved
Evolutionary Strategy Genetic Algorithms to optimize constrained nonlinear
programming was correct and effective and it was a reliable and efficient global
optimization algorithm.
Table 105.1 Parameter settings and calculation results
f1(x) a b c m T Number of generation to Variable values Optimal
obtain the optimal solution
solution for the first time
n = 2 -5.12 5.12 6 80 100 45 (0, 0) 0
n = 5 -5.12 5.12 6 80 200 145 (0, 0, 0, 0, 0) 0
n = 10 -5.12 5.12 6 80 500 350 (-0.000023, 0, -0.000030, 0, 0, 0, 0
-0.000396, -0.000396, 0,
105 Improved Evolutionary Strategy Genetic Algorithm
0.000010)
1001
1002 H. Zhu et al.
Table 105.2 The comparison of simulation results Of FD, PF and IESGA
f2(x) x1 x2 Optimal solution
FD 0.630 0.874 6.544
PF 0.645 0.869 6.566
IESGA 0.658872 0.868225 6.613083
105.5 Conclusions
(1) The Improved Evolutionary Strategy Genetic Algorithm preserves the optimal
groups based on the traditional elite preservation method. The advantage of
this method is that it reduces the possibility of optimal solutions being
destroyed by crossovers or mutations in the process of evolution. Premature
convergence, which may be present in the traditional elite preservation
method, is avoided because all individuals quickly converge to one or two
individuals with high fitness values.
(2) The correction operator breaks the traditional idea and avoids some problems
such as low searching efficiency by refusing infeasible solutions, early con-
vergence by introducing a punishment factor and deviation from the constraint
area considerably after mutation operation.
(3) The combination of the improved evolutionary strategy and the method of
correction operator can effectively solve many nonlinear programming prob-
lems, greatly improve solution quality and convergence speed, realize the
linear search method of moving infeasible individuals towards feasible indi-
viduals, and effectively guide infeasible individuals.
The disposal of infeasible individuals by the correction operator is simple and
effective. It is proved to be an effective, reliable, and convenient method.
Acknowledgments Natural Science Foundation of China (31071331)
References
Bazarra MS, Shetty LM (1979) Nonlinear programming theory and algorithms. John Wiley
&Sons, New York, pp 124–159, 373–378
Bi Y, Li J, Li G, Liu X (2000) Design and realization of genetic algorithm for solving nonlinear
programming problem (in Chinese). Syst Eng Electron 22(2):82–89
Gao J (2010) Genetic algorithm and its application in nonlinear programming. Master
dissertation, Xi’an University of Architecture and Technology, Xi’an, China
Ge Y, Wang J, Yan S (2008) A differentiable and ‘almost’ exact penalty function method for
nonlinear programming (in Chinese). J Nanjing Normal Univ Nat Sci Ed 31(1):38–41
Hansen JV (2004) Genetic search methods in air traffic control. Comput Oper Res 31(3):445–459
He D, Wang F, Mao Z (2006) Improved genetic algorithm in discrete variable non-linear
programming problems (in Chinese). Control and Decision 21(4):396–399
Holland JH (1975) Adaptation in natural and artificial systems. University of Michigan, USA
105 Improved Evolutionary Strategy Genetic Algorithm 1003
Liang X, Zhu C, Yan D (2009) Novel genetic algorithm based on species selection for solving
constrained non-linear programming problems (in Chinese). J Central South Univ Sci Technol
40(1):185–189
Lyer SK, Saxena B et al (2004) Improved genetic algorithm for the permutation flowshop
scheduling problem. Comput Oper Res 31(4):593–606
Operations research editorial group (2005) Operations research (3rd edn) (in Chinese). Tsinghua
University Press, Beijing, pp 133–190
Saleh HA, Chelouah R (2004) The design of the global navigation satellite system surveying
networks using genetic algorithms. Eng Appl Artif Intell 17(1):111–122
Sui Y, Jia Z (2010) A continuous approach to 0–1 linear problem and its solution with genetic
algorithm. Math Pract Theor 40(6):119–127
Tang J, Wang D (1997) Improved genetic algorithm for nonlinear programming problems (in
Chinese). J Northeast Univ Nat Sci 18(5):490–493
Tang J, Wang D, Gao Z, Wang J (2000) Hybrid genetic algorithm for solving non-linear
programming problem (in Chinese). Acta Automat Sin 26(3):401–404
Uidette H, Youlal H (2000) Fuzzy dynamic path planning using genetic algorithms. Electron Lett
36(4):374–376
Wang D, Liu Y, Li S (2003) Hybrid genetic algorithm for solving a class of nonlinear
programming problems (in Chinese). J Shanghai Jiaotong Univ 37(12):1953–1956
Wang D, Wang J, Wang H, Zhang R, Guo Z (2007) Intelligent optimization methods (in
Chinese). Higher Education Press, Beijing, pp 20–80
Wang F, Wang J, Wu C, Wu Q (2006) The improved research on actual number genetic
algorithms (in Chinese). J Biomathematics 21(1):0153–0158
Wang X, Cao L (2002) Genetic algorithm—theories, applications and software realization (in
Chinese). Xi’an Jiaotong University Press, Xi’an, pp 1–210
Chapter 106
Simulation and Optimization of a Kind
of Manufacturing and Packing Processes
Chun-you Li
Abstract There are many factors that influence each other in production-
packaging processes. Resources, objects, processes and their properties and
behaviors can be simulated to construct a computer simulation model across the
whole production-packing process. Usually, the minimized cost, maximized profit
or reasonable utilization was targeted as the decision objective, and concerned
parameters was configured as conditions in the simulation model. With enough
repeated runs, the optimization module can seek the best equipment combination
and the best production schedule.
Keywords Manufacturing and packing Simulation Optimization
106.1 Introduction
In some industries such as food and tobacco industry, the terminal product is
generally made from the production line and packaged into small boxes or small
bags. The basic process is to produce these products through one or more pro-
duction lines, and delivery or transfer products to packaging. Finally, the small-
packaging products are filled into a larger container through one or more pack-
aging equipments continuously or partially and gotten out the lines. Figure 106.1
is the schematic diagram of such production and packaging process, two manu-
facturing lines produce the same kind of products, then transfer them to three
C. Li (&)
College of Transportation & Logistics, CSUFT, Changsha, Hunan, China
e-mail: [email protected]
C. Li
Accounting School, GXUFE, Nanning, Guangxi, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1005
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_106,
Ó Springer-Verlag Berlin Heidelberg 2013
1006 C. Li
coordinate packing lines through a series of buffer vessels, finally package them
into finished goods.
In cases such as designing and recreating a new packaging process or managing
an existing process, we often have to face the following questions as: how to
reduce process failure? How to reduce the processing cycle time? How much is the
reasonable buffer capacity and the buffer stock? How to deal with the change of
production scale? And is it necessary to add more and higher ability production
lines, packaging lines and container?
To simplify the analysis, we can do analysis and make decision with single
process and single factor. For example, the expansion of the production and
logistics ability can be determined with the output of production or both speeds.
But the whole process is complex and the relationship between processes is
uncertain. Additionally, there are various factors influencing mutually in processes
and these factors always interact dynamically along with time and events. Once a
suggestion was put forward or a measure was imposed, it may be hard to predict
the ultimate effect brought by the changes, so it is very difficult to determine the
exact priority order of the measures. For example, in order to maintain the reli-
ability and the inventory balance across processes, designers can use greater
inventory to deal with the less reliability of equipments. On the contrary, by
improving the reliability of the upstream through equipment, he can reduce the
storage in the processes. Both measures can ensure to meet the needs of the
downstream material and ensure that production runs smoothly. Even if the rela-
tionship among production, package or buffer process is certain, there are still
many factors interact each other. For example it is difficult to evaluate the influ-
ence degree of the mutual isolated factors on the production scheduling, as well as
the control of the operation sequence and rhythm, the product quantity, structure of
the production, or the characteristics of products and so on. These factors may
affect production speed, lead to the failure or production conversion. In addition,
due to manufacturing and packing may be arranged in different positions apart
from each other, and the difference of scheduling method and the enterprise culture
material
Buffer 1
material
Package 1
Machine 1 Buffer 2
material
Package 2 Final goods
Buffer 3
material
Machine 2
Buffer 4 Package 3
Fig. 106.1 The schema of production and packaging process
106 Simulation and Optimization 1007
require different rhythms of production, the complexity to solve the problems will
also be increased.
There is a variety of decision tools and the experimental methods to deal with in
this kind of problems that schedule manufacturing and packing with multifactor.
This paper presents a simulation method that simulates the process and interacting
factors. It analyses and evaluates the problems and forecast the effect of the
decision that have been designed or improved. It provides a tool to test design idea
or improvement effect by developing a manufacturing-packing simulation model.
Just as simulation driving-cabin can help the driver to learn driving in the best way
and to build a good habit, a manufacture and packaging simulation model can be
used to test and optimize the manufacture and package (Wang et al. 2002).
106.2 Modeling and Simulating Generally
The tool that is used to simulate and solve the problem is the simulation model, or
called simulator. Specific simulation model is based on the research goal con-
cerning with the problem to solve. This paper involves a factory to build a new
manufacture and packing system. Preliminary design assume there were two
production lines in the factory and each production line can produce any kind of
the basic specifications. There were three packaging lines to pack the products into
kinds of size and the shape of the containers with different labels. There were
many parallel buffer tanks between production lines and packing lines which can
receive products from any production line, then put the products to any packing
line. The production lines and the buffer tanks should be cleaned before transform
the line for new products.
The simulation model is developed to solve the following problems: Can the
new equipments match the new production combination and schedule? What kind
of scheduling strategy will make best operation of new equipment? How many
buffer tanks will be need and what is the reasonable specification of buffer tanks?
How is the effect of improvement of manufacture and packaging reliability? What
influence will the production cycle time brings? Are more packaging lines or
production lines needed?
The model is implemented in the ExtendSim that is a simulation platform
developed and published by Imagine That. It is a set of software that contains
simulation libraries and tools. It is used to simulate discrete event, continuous
process and discrete process based on rate. The continuous flow stands for large-
amount or high-speed flow, this software includes controls and schedule parts used
for modeling process, and layered structure templates used to represent a higher
level (Krahl 2010, 2011).
Figure 106.2 is the general simulation model that shows the two production lines,
four buffer tanks and three packing lines. The actual parts of the model are included in a
level module. You can double-click any region of higher level modules image to open
lower level modules. The timetable, equipment performance, fault characteristics, the
1008 C. Li
Fig. 106.2 The simulation model of manufacture and packaging
conversion rate and other data are included in a built-in a database table of the model
and could be visited through a logical scheduling structure.
In this case, the model manufactures and packs products by running manu-
facturing machines and packing equipments under the control of a logic scheduler.
The scheduler controls the simulation by an order table that lists products and the
amount. The utilization ratio of equipment was set by the logical scheduler and we
can use the logical scheduler to instruct equipment for conversion when it is
necessary. The report submitted by the operation model is like the actual business
report. The researcher can check out these subject reports, and points out the
existing problems.
The simulator is widely used for helping the project team of the factory, to
make sure the number and configuration of the new manufacturing lines, pack-
aging lines and buffer tanks. A few test schedules are developed to represent
production requirements in the typical situation and in the extreme cases. These
models are based on the existing factory model and a series of packing line designs
that are recommended.
106.3 Simulating Objects, Attributes and Activities
106.3.1 Creation and Transformation of Objects
When the simulation model operates, the simulation clock keep recording the run-
ning time increasing with simulation steps. Productions and packaging steps are not
predefined with a table, but caused by events. In this model, items as objects are
produced by a Create module, and the production rate of the item is decided by
‘‘interval time of two items’’. The interval can be represented by species and
parameters of designated random distribution that depicts the item production
condition of the making line. For example, the interval in this case is described with
an exponential distribution. The mean value for one conduction line is 0.2 and
106 Simulation and Optimization 1009
another is 0.4. Both location values are all zero. Production characteristic of the two
production lines is depicted with different values of the parameter (Hu and Xu 2009).
The manufacture or the packaging process could change the measurement unit.
For example, five small items of product form a larger one after packing. It can be
simulated with a merger module named Batch, it allows multiple sources of
objects merged into one thing. It will be of great help in coordinating different
machines to assembling or fusing different parts. In the module dialog box, we can
set the number of each input objects to produce an output objects, as well as
objects from other input are not allowed to put into this module if some input
objects are not arrive or quantity is not enough.
106.3.2 Production and Packaging
We can simply use a process with time parameters to simulate the manufacturing
or packaging. The most important activity module of ExtendSim is Activity. Its
basic parameter is Delay that is the processing time of the activity module. In
addition, it can also define and deal with several items at the same time. In the
module dialog box, the processing time can be designated as a fixed value, or input
from the D(Demand) port of the module, or fromattribute value of another module,
as well as from an inquiring form. The last three ways can realize more plenty and
slightly modeling for processes. In this case, the initial processing time of two host
machines was set constants, more detailed time table can be set up in the sub-
sequent according to the specific situation of the machine.
In order to coordinate the input and manufacture or packaging, a Queue module
is needed between input and the activity as a buffer. Queue module will store items
and wait to release them to next module based on the rules predetermined. In the
module dialog box we can select or set queuing rules such as resource pool
queuing, according to attribute value, FIFO(first in first out), LIFO(last in first
out), priority, etc. With rule Resource pool queuing, the resource will be caught
from the resource pool module where the resource number is limited. A queue
based on the attribute value will sort items by an optional attribute. FIFO queue is
the most common queuing way; LIFO is a kind of reverse queuing, also known as
stack, which means the latest item into the stack will leave first; under Priority
queuing way, the module uses the Priority attributes to determine the releasing
order of item (Fig. 106.3).
Fig. 106.3 Production
process model
1010 C. Li
106.3.3 Inventory and Buffer Tank
There exist differences of space distance and rate between the Processing and
packing process. In practice, it often use the way of setting up stock in factory or
production inventory to coordinate the contradiction. The specific physical form of
this inventory may be a universal warehouse, some cargo space or buffer tanks exist
between the processing and packing line. In this case we assume the way of buffer
tank. A buffer tank store the same kind of products only, but receive the products
from different production lines, it can also release to any packing line for packaging.
We still use Queue module to realize the simulation of buffer tank. But since
there are multiple buffer tanks, the model needs to choose the reasonable buffer tank
to store the products from the end of production lines. And because there are more
than one packaging lines, the model should also do a reasonable choice when the
buffer tanks release products so as to put the products into the free packing line. We
can use two kinds of module like Select Item In and Select Item Out to realize the
simulation of the product routing. A Select Item In module receives items from
more input branches and release items out through its only output port. A Select
Item out module receives items from its only input port and release items by
choosing one of export branches. The selections in the dialog box include based on
priority, random selection, sequential selection or based on Select port selection. In
this case, between production line and buffer tank, and firstly establish a Select Item
In with two inputs branches, output randomly to Select Item out with four branches,
completing the routing of bulk products from the buffer tank to the packing line.
Between the buffer tanks and packing lines, establish a Select Item In with four
input branches, randomly output to a Select Item out with three output branches to
realize the route from the buffer tank to the packaging line (Fig. 106.4).
106.3.4 About the Schedule
The simulation model is driven to operate and interact by a schedule, just like an
opera showing step by step with a script written by the editor. The model can be
Fig. 106.4 Routing of production lines to buffers
106 Simulation and Optimization 1011
used to evaluate potential of a specific schedule in real-world, or compare one
schedule against another, on the basis of how well schedules perform in the model.
We can discover which schedule is the best by testing with the different schedules,
as well as what kind of scheduling generating rules is the best. Since for each
schedule, the model will show real-world operation results such as production,
utilization, and down time (Pinedo 2002).
A simulation model of using schedule should be able to introduce a sequence of
operations that was created by embedded data table or database, or use an external
table or external database to manipulate the schedule. There is a built-in module in
some simulation model, for setting scheduling method, rapidly and flexibly to
generate the typical schedule or the experimental schedule. It can be based on
demand conditions, or on the actual demand schedule. These scheduling modules
can calculate some certain evaluation measures, also can put the execution of the
schedule to the simulation model to get a complete operation results by capturing
the dynamic situation of real world. That is generating any sure schedule for the
whole simulation scope, or producing internal schedule obey to the model com-
mands and the set time or conditions in advance in the entire operation (Pinedo and
Chao 1999). It can be showed in the model what is the demand degree, as well as
the state of all products in the whole time range.
106.4 To Evaluate and Optimize Processes
106.4.1 Evaluation Indexes
The simulation model is used to analysis the manufacturing and packaging system
and the problems are to be solved. Some models involve manufacturing operation
only or involve packaging operation only, and the other involves both. Normally, it
is based on problems which will be solved to determine the corresponding eval-
uation index such as the equipment utilization, the processing cost and the queue
length (Jiang et al. 2009).
Utilization. Utilization is the ratio of the working time and all running time of
the equipment. Low utilization means that the resources have not been made fully
use, but high utilization rate is sometimes not a good thing because it means ability
tension. Once the equipment runs failure, inevitably leads to the production halt
and extend the production cycle, which leads to the failure of the production
scheduling (Dessouky et al. 1994).
Pn
ti
Ult ¼ i¼1 ð106:1Þ
T
Ult equipment utilization
Ti the processing time of product number i
1012 C. Li
i ID of products
T running time of the equipment.
Cost. Any manufacturing and packaging process must spend some resource and
its cost is a key management tool. An activity module has cost parameters in the
dialog box that can simulate the cost of the process. We can set the cost information
in the cost page in the dialog box. According to the character of cost, we can set two
kinds of cost, fixed cost and time cost. Fixed cost is the cost that happens when deal
with every product, its value is a constant, unrelated with the delay of products. The
time cost is relevant with the processing time, it equals to the multiplication of the
cost per unit time and the operation time (Harrell 1993). The module will auto-
matically calculate its cumulative cost and display on the plot.
X
n
Total Cost ¼ ti Cpertimeunit þ qi Cperitem ð106:2Þ
i¼1
Total Cost total processing cost
Cpertimeunit unit time processing cost
Cperitem fixed cost of every processing a product
qi total products exited from the module.
106.4.2 Optimization
The simulated optimization, also named goal seeking, is seeking the optimal
answer to the question automatically, or the optimal value of parameters. In
parameter range given in the model, we can run the model repeatedly to search the
solution space and find out the best parameter-values that satisfy the conditions as
well as reach the decision target. In the optimization model including an Optimizer
module, the issue is usually presented as a target function or a cost-profit equation.
In order to realize the cost minimization or the profit maximization, the ExtendSim
simulation models help researchers not only find the best solution automatically
but also put out of the long boring process that repeated trying different parameter-
values (Wang et al. 2009).
The running conditions can be changed in the optimization model. For example,
we can set the value range, value method and constraint conditions of parameters
by limiting value scope of decision-making or defining constraints equations. We
can also affect the solving precision by the setting run parameters, such as deciding
the total sample cases, the search times of each case, when to check convergence,
and optimal number of member cases in the convergence (Zhang and Liu 2010).
Optimizer does not have the function of refusing faults, so any Optimizer maybe
106 Simulation and Optimization 1013
converges to the second best solution not the first best solution, especially when its
running time is not long enough. So we should consider to run more times and get
enough operation results, and ensure getting the same convergence to close to the
optimal solution before using the best solution for actual application.
106.5 Conclusion
The simulation is a good tool to study the problems of manufacture and packaging
operation. It is because of multiple factors interact through a variety of means.
A simulation model can provide more real ways to improve decision effect than
other research models. In fact, a simulation model is just like a virtual factory to
test the new design idea, or assessment recommended projects. The simulation
model discussed in this paper is designed in the ExtendSim condition. It can be
used to optimize the process of manufacture, package and logistics in the factory.
References
Dessouky Y, Maggioli G, Szeflin D (1994) A simulation approach to capacity expansion for the
Pistachio Hulling Process. In: Proceedings of winter simulation conference, IEEE, New
Jersey, pp 1248–1252
Harrell CR (1993) Modeling beverage processing using discrete event simulation. In:
Proceedings of the winter simulation conference, IEEE, New Jersey, 1993, pp 845–850
Hu S, Xu LW (2009) Simulation and optimization for Noshery Service System. Paper presented
at Information Engineering and Electronic Commerce at the international symposium,
pp 721–723
Jiang LF, Sun GT, Zhang N (2009) Layout research of campus traffic system based on system
simulation. In: Technology and innovation conference 2009 (ITIC 2009), pp 1–5
Krahl D (2010) ExtendSim advanced technology: Integrated simulation database. In: Proceedings
in winter simulation conference, 2010, pp 32–39
Krahl D (2011) ExtendSim technology: scenario management. In: Proceedings in winter
simulation conference, 2011, pp 12–19
Pinedo M (2002) Scheduling: theory, algorithms, and systems, 2nd edn. Prentice Hall,
Englewood Cliffs, pp 124–126
Pinedo M, Chao X (1999) Operations scheduling with applications in manufacturing and services.
McGraw Hill, New York, pp 68–74
Wang Y, Perkins JR, Khurana A (2002) Optimal resource allocation in new product development
projects: a control-theoretic approach. IEEE Trans Autom Control 47(8):1267–1276
Wang R, Li Q-M, Zhu H-B, Peng Y-W (2009) A simulation model for optimizing support
resources in requirement of warship mission reliability. In: International conference computer
technology and development, 2009, pp 144–148
Zhang Z-C, Liu J-H (2010) Extend-based research in positioning and optimization the bottleneck
process of MTO enterprises. In: International conference on computer application and system
modeling, 2010, pp 479–481
Chapter 107
Equilibrium and Optimization
to the Unequal Game of Capital-Labor
Interest
Ming-liang Wang and Yu Lu
Abstract Imbalance of labor-capitals’ interests is the inevitable result of both
employers and employees unequal game. Labor for the respect of the existing dis-
tribution system just because they have no better choice. The key to reverse the
imbalance in the distribution of labor interests is to strengthen labor game ability.
Formation of autonomous trade unions will increase the labor collective game
capabilities, and specific training will help to expand the labor game resources to
enhance the ability of the individual game; a sound social security system will help
to ease the worries of the failure of labor negotiations, thus improve the labor’s
tolerance to the time costs for labor-capital consultation, strengthen the credibility of
labor ‘‘threat’’. Meanwhile, the sound signal-transformation mechanism is helpful to
build the reputation of incentives to stimulate corporate social responsibility.
Keywords Imbalance of labor-capitals’ interests Unequal game Signal-
transformation mechanism Multi-dimensional network coordination mechanism
107.1 A Labor-Capital Relation is a Unequal Game
with ‘‘Strong Capital and Weak Labor’’
At this stage, the distribution of labor-capital benefits is severely uneven.
According to preliminary estimates, China’s Gini coefficient has been over 0.5.
How to adjust the distribution to prevent polarization and share the results of the
reform has become a hot topic of the current social. Generally speaking, the
domestic academic interpretation of reason to the imbalance are mainly the fol-
lowing: First, that China’s current economic and population structure led to the
imbalance of the interests, which was exacerbated by the economic globalization;
M. Wang (&) Y. Lu
Guangdong University of Technology, Guangzhou 510520, Guangdong, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1015
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_107,
Ó Springer-Verlag Berlin Heidelberg 2013
1016 M. Wang and Y. Lu
second is that China’s trade unions and other legal deficiencies in the system
resulted in imbalance; the third is that it mainly due to government, whose over-
emphasis on the investment environment caused damage to the workers’ interest;
the last one is that the imbalance stems from the differences between employers
and employees capacity to safeguard the rights (Qi 2008).
According to game theory, in market economy conditions, the distribution to
the labor-capitals’ benefits depends on both employers and employees’ game
power, and Imbalance of labor-capitals’ interests is the inevitable result of the
unequal game. In theory, under the existing conditions, the decisive factors to the
game power include the resources they have, the credibility of the threat, risk
aversion and time preference (Jack 2009). ‘‘Strong Capital and weak labor’’ is an
objective fact, which can be said is a worldwide phenomenon. This is determined
by the system of market economy in the modern times. As the distribution subject,
the corporate system, which dominated by the principle of ‘‘absolute ownership’’,
is still dominated by the ‘‘shareholders center’’ Wang 2008). At the same time,
from employers and employees’ market attributes, the incompetence of the labor
market and capital flows around the world further exacerbates the worldwide
imbalance of labor game capacity (Zaheer 2003).
Based on the principle of freedom of contract in market economy, labor contract
is seen as the one which employers and employees freely negotiate and deal with
their respective interests, the labor interests’ distribution is also seen as the outcome
of the game, which can be a one-time or repeated several times. Macroscopically
speaking, existence of market economy depends on the minimum cooperation of
employers and employees and thus, the labor relations in the market economy is a
long-term competition and cooperation, a bargaining relationship on the distribu-
tion of benefits; but from the micro level, the specific employment relationship may
be a non-cooperative relationship at random. Reflected in the labor contract, short-
term labor contract (or one-time labor contract) means that the one-off game, or
non-cooperative relationship between employers and employees; long-term con-
tract (or non-fixed term labor contract) means repeated game, or competition and
cooperation between the two parties. Employers and employees concern about the
overall benefits in the Long-term labor contracts, and thus there is the possibility to
adjust their game strategy, and it can inhibit the ‘‘short-sighted behavior’’ of the
game subject to a certain extent. When the interests of labor-management coop-
eration and win–win cooperation is over the inputs (such as special human capital
investment), labor-management cooperation will become the norm.
In addition to the above factors, the general industrial labor supply far exceeds
the market demand in our country, which determines the congenital weakness of
labor in the game. Meanwhile, low labor skills, absence of the autonomic trade
union and the deletion of methods such as ‘‘strike’’ and ‘‘threat’’, as well as
defection of unemployment insurance, all compress the game strategy space of the
labor, and reduce labor tolerance to time cost required for labor-capital consul-
tation and the credibility of the means. Thus, relative to other developed countries,
labor game of power even more unbalanced in China. Determined by the stage of
economic development, homogenous competition among cooperation is intense
107 Equilibrium and Optimization to the Unequal Game of Capital-Labor Interest 1017
(e.g., the price competition), enterprises lack the power and capacity to improve
labor rights, and labor rights lack protection. Therefore, our labor respect a serious
injustice distribution system, not because they endorse, nor is a Pareto improve-
ment, but simply because they have no better choice. It is inevitable to adjust the
pattern of labor game and strengthen the ability of labor game in order to pursue
better social justice, optimize the allocation pattern, avoid the deterioration of
labor relations and achieve the stable development of the market economy.
It must be clear that emphasizing the balance of labor interests is not said to
fundamentally reinvent the wheel, but is the proper adjustment of distribution of
labor benefits under the socialist market economy. In essence, it is appropriate
optimization to the game framework with ‘‘strong capital and weak labor’’ to
achieve sustainable economic development and create a harmonious social
environment.
107.2 Balance of the Capital-Labor Interests
in the Unequal Game
The unfair distribution of labor interests is from unequal game power. It is sig-
nificant to explore the theory to realize balance in the unequal game. Assume that
Player A and Player B are on behalf of employers and labor, the game strategy
combination as the following table, then, we can make the following analysis on
strategy equilibrium of the game.
Player A (capital\employer) Player B (labor\employee)
L R
L DA, DB x, x ? eB
R x ? e A, x DA, DB
In this model, if DA, B \ x, then, there will be two equilibrium results, namely
(R, L) and (L, R). D value indicates the payment which the perpetrator will get
when he fails to reach an equilibrium outcome, or failure value. eA, B [ 0, e is said
as the advantages of distribution of behavior.
If DA = DB, failure values are equal, that means it is a peer-to-peer game. If
DA [ DB, or DB [ DA, that means an unequal conditions. In such repeated games
with incomplete information
A ‘s probability (employer) to select R is: ðx þ eB DB Þ=ð2x þ eB 2DB Þ;
probability to select L is: ðx DB Þ=ð2x þ eB 2DB Þ.
Accordingly, B (Labour) probability:
If, p ¼ ðx þ eB DB Þ=ð2x þ eB 2DB Þ, then B (Labour) have no preference
between L and R.
1018 M. Wang and Y. Lu
If, p [ ðx þ eB DB Þ=ð2x þ eB 2DB Þ, then B (Labour) will be preferred to
select L, and receive less income x. p is the subjective probability of protocol
between the weak B (Labour) and the strong A (employers).
Further it can be deduced:
op=oDB ¼ eB =ð2x þ eB 2DB Þ2
The significance of the equation is that starting point value also increases with
the increase of the failure value. This means that with the decline of the failure
value of perpetrator, proportion of strong man whose preference for strategy L also
falls. This suggests that the weak labor prefer to adopt the strategy of L while
strong employers are more likely to adopt strategies R in the game with strong
capital and weak labor.
Tolerance to the time costs and the credibility of a threat to stop negotiation
may also affect the game subject’s choice. To measure time preference by d, if d is
greater than x/(x ? eB), the weaker B (labor) will choose L in the bargaining
model. The second is the ‘‘threat’’. Credibility of the strong A (employers) will be
greater than that of the weak B (Labour) In the unequal bargaining. If the strong A
(employers) impose penalties C on the weak B (Labour), which will change the
income of the balance of (L, R) (DB [ x+eB-c). In order to avoid further reduction
in revenue, L will become the weak B (Labor)’s main strategy. As retaliation (such
as resignation) is expensive, if there is no external remedies (such as unemploy-
ment insurance), it will be difficult for weak B (Labor)to implement retaliation
strategy, and it will not be able to influence the choice of the strong A (employers)
for the R .
The above analysis shows that even if the labor is not sure about the employer’s
strategy, when the labor knows he is the weak, he is more likely to accept the other
conditions. The attitude of labor depends on the judgment in the strong position of
capital, the higher probability he makes, the greater chance to yield. The weaker
labor force is, the higher effective the employer’s strategy is. A employer with
more game recourses will be less dependent on labor consultation results, and
more patient to bear the time cost which translates into consultations advantage for
employer to influence the outcome of the game. In the game, the ‘‘threat’’ of
employers is more credible than that of labor according to the different game
resources, influencing weak labor choice and prompting labor more vulnerably to
accept the conditions mentioned by the employer.
107.3 Allocation Mechanism of Labor Interests Based
on Collective Bargaining
There is no doubt that labor relations in the market economy exists congenital
defect of ‘‘strong capital weak labor’’. How to take appropriate measures to reverse
unfavorable situation of the labor and coordinate capital-labor interests is the
common problem faced by all market economy countries. The academic
107 Equilibrium and Optimization to the Unequal Game of Capital-Labor Interest 1019
community discussed much and put forward a variety of theoretical models and
policy recommendations, which can be summarized as the following:
The first is the new classical school. It abandons ethical factors in classical
economics, emphasizing that the natural order of the market and economic
exchange is the main way to resolve labor conflicts. It advocates mobilizing labor
enthusiasm for production through the adoption of wages, bonuses and other
incentives to realize capital-labor win–win situation. The second is the manage-
ment school. It sees the limitations of ‘‘pure market’’ regulation and labor conflicts
triggered and emphasizes the common development of both employers and
employees on the basis of the same interests. The third is the new system school. It
believes that capital-labor conflict of interest can be solved by constructing
common interests and advocates establishment of a diversified economy and
political system to ensure the bargaining right between employers and employees
and independent trade unions to ensure the labor’s interest and eliminate labor
conflicts through institutionalized ways. The fourth is freedom of reform school. It
advocates the establishment of strong trade unions, develop a strict legal to reg-
ulate labor relations. It also believes that the government should implement
positive economic and social policies to restrict and change the recurrent negative
impact on the market economy. The fifth is the new Marx school. It maintains that
a system should be established so that the labor becomes the owner and manager to
participate in corporate decision-making and profit-sharing (Zhao 2009). All the
above make certain sense and to some extent, reflect the basic requirements of the
labor relations adjustment under market economy conditions, and are adopted in
some countries.
From the practice in the world, the adjustment models of labor relation are
various and different. As a result of congenital defect, ‘‘strong capital and labor
weak ‘‘, of the market economy, all countries have the same objective which is to
strengthen the labor game ability, let the government and other social subjects play
appropriate ole in coordination and prevent excessive imbalance. Among them,
allocation mechanism of labor interests based on collective bargaining has become
the mainstream to solve imbalance.
The unequal bargaining model shows that the hinge to decide the labor force are
the game resources, time preference and the credibility of the ‘‘threat’’. Therefore,
the key to optimize labor allocation pattern is to strengthen the game ability of
labor and optimize labor game structure. Achieve above goals, the following steps
should be taken. One is to enhance individual labor game ability. Theory and
Western experience have shown that the level and specialization of labor skills
connects closely with the game capacity.’’Asset specificity’’ decides comparative
advantage of the parties (Oliver 2002). Therefore, strengthening school education,
vocational training and work skills training is critical to enhance labor game
capabilities. The second is the formation of the labor collective game force.
Reform the existing trade union structure, strengthen the union representative, and
actively build autonomous trade unions in the enterprises, progressively develop
the industrial trade union to strengthen the capacity of the labor collective game, at
the same time, improve labor ‘‘threat’’ power. Draw on Western experience and
1020 M. Wang and Y. Lu
moderate trade unions power of ‘‘limited monopoly’’ and ‘‘orderly lay-off’’,
improve the unemployment insurance and other social security measures to alle-
viate the worries of labor, and strengthen the credibility of the labor policy of the
‘‘threat’’ to enhance the labor tolerance to the time for the negotiation process. The
third is to establish Dynamic Labor Standards, develop appropriate laws and
regulations, timely adjust game expectation, form a new ‘‘focal point’’ of labor
game to promote the labor interests of the coordination to a higher level.
With the rapid development of knowledge-based economy and globalization,
‘‘mass customization’’ has increasingly become multinational prevailing produc-
tion. It strengthened the advantages of technological innovation of the developed
countries and promoted the process of industrialization in developing countries; on
the other hand, it formed the imbalanced value in the international industry
division, reconstructed the international labor relations. For the reputation, the
multinational companies launched the ‘‘Production Code’’ to constraint the
foundry company’s employment practices. The strategy which compete to the
bottom among developing countries weakened the government’s ability to coor-
dinate labor interests and gave birth to various types of labor rights protection
organizations (NGOs). At the same time, the relocation of production processes
also led to domestic structural unemployment in developed countries. In order to
ease domestic political pressure, the developed countries began to strengthen the
global labor standards. All of these mean that labor relation is undergoing a
fundamental change in the world (Wang 2011). Therefore, we must actively adjust
the concept of social management and give full play to the community, smooth
information disclosure channels, improve the delivery mechanism of the ‘‘signal’’,
let concerning subjects play an active role, and effectively connect informal
mechanisms such as labor game with legal and other formal mechanisms, to build
multi-dimensional network coordination mechanism based on the labor collective
bargaining (Aoki 2001).
In short, the imbalance of labor interests is an objective reality, and it is nec-
essary to optimize t distribution of benefits. The importance and complexity of
labor relations determines that the adjustment of labor interests needs all kinds of
formal and informal mechanisms of coordination and cooperation. Strengthen the
labor ability through various means and build the labor collective bargaining
system and improve the labor game platform. Meanwhile, play government’s role,
make up market failures, mobilize all parties to build a multi-dimensional network
coordination mechanism based on a labor collective bargaining and achieve cap-
ital-labor long-term cooperation and win–win.
Acknowledgments Fund Project: The article is the achievement of the humanities and social
science planning project of the Ministry of education (project number:10YJAZH079), Natural
Science Fund Project of Guangdong Province (project number: 10151009001000003), Guangz-
hou society and’’ eleven five’’ planning fund project (project number:10Y73).
107 Equilibrium and Optimization to the Unequal Game of Capital-Labor Interest 1021
References
Aoki CY (2001) Analysis on comparative institution, Leann. Shanghai Far East Press, Shanghai,
pp 385–392
Jack N (2009) Institutions and social conflict, Zhou Weilin. Shanghai People’s Publishing House,
Shanghai, pp 130–141
Oliver WE (2002) The economic institutions of capitalism, Duan Yicai, ed. The Commercial
Press, Beijing, pp 78–84
Qi X (2008) Research on labor relation imbalance. J Jiangxi Adm Coll 10(4):47–50
Wang M (2008) Legal mechanism to generate corporate social responsibility. Theory Guide
30(4):101–104
Wang M (2011) On the coordination of labor relation of ‘‘three mechanisms’’ to implement the
social foundation and limitation. Theory Guide 33(1):34–37
Zaheer DA (2003) Breaking the deadlock: ‘‘why and how developing countries should accept
labour standards in the WTO’’. Stanf J Law Bus Finance 9:69–104
Zhao X (2009) Research on Chinese labor relation adjustment mechanism during the transition
period. Economic Science Press, Beijing, pp 34–35
Chapter 108
Innovative and Entrepreneurship
Education in Underdeveloped Western
Regions of China
Chang-jiang Lu, Yan Feng and De-wen Chen
Abstract This paper makes an analysis on the major problems for innovative and
entrepreneurship education (IEE) in underdeveloped western regions of China, and
outlines a set of implications for local governments and universities. The authors
suggest that a more practical and flexible cultivation system rooted in regional
contexts should be established for bringing a radical change to the backward IEE
in western China. It is important to implement the ‘‘4C’’ concepts in IEE, namely
cross-culture, cross-region, cross-discipline and cross-specialty through strength-
ening international cooperation and mutual regional support, integrating the IEE
into the university curriculum, and building a four-dimensional nexus via part-
nerships between universities, industries, governments and families. While the
paper is written mainly from the perspective of underdeveloped western regions of
China, the discussion allows for generalization, and thus should be applicable to
the development of IEE in other nations facing similar problems.
Keywords Western China Innovative and entrepreneurship education
108.1 Introduction
Along with the occupied population boom, China is entering a new economic
transformation phase. The two words ‘‘innovation’’ and ‘‘entrepreneurship’’ (IE)
are more closely combined than ever before and have become an important
C. Lu (&) Y. Feng D. Chen
Sino-Canadian International College, Guangxi University,
Nanning, China
e-mail: [email protected]
Y. Feng
e-mail: [email protected]
D. Chen
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1023
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_108,
Ó Springer-Verlag Berlin Heidelberg 2013
1024 C. Lu et al.
internal force for China’s economic growth. Social development is in urgent need
of innovative and entrepreneurial (IE) talents. With innovation and entrepre-
neurship education (IEE) in universities as its focus, cultivation of IE talents
forged ahead by Chinese governments at various levels is in full swing across the
country. Provinces throughout China have been making great efforts in developing
IE talents cultivation modes fitted in regional contexts. The number of start-up
businesses sees a continuous increase, yet most of the enterprises are not estab-
lished on the basis of innovative concepts, knowledge, skills or innovations.
Therefore, it has become a great concern for Chinese local governments, educators
and researchers on how to produce more quality IE talents through integration of
innovation and entrepreneurship.
108.2 What is Innovative and Entrepreneurship Education
Innovation, according to the 2004 Report ‘‘Innovate America’’ by the United
States Council on Competitiveness, refers to turning ideas and technologies into
new products, new processes and methods as well as new services that can create
new market values drive economic growth and improve people’s living standards.
Entrepreneurship means one who undertakes innovations, finance and business
acumen in an effort to transform innovations into economic goods. This may result
in new organizations or may be part of revitalizing mature organizations in
response to a perceived information, resource or opportunity. There is an obvious
strong relationship of the two terms: first, they are not always with the same
understanding, but in essence there is a considerable overlap between them in such
aspects as denotations, functions, objectives and processes; second, innovation
forms a significant part of the foundation for entrepreneurship, and entrepre-
neurship is the carrier and manifestation of innovation (Xie 2009). IEE, developed
by western scholars in the 1980s, can be simply defined as an educational system
aimed at cultivating IE talents through fostering their IE awareness, spirit,
knowledge, competencies and skills.
108.3 Review on Innovative and Entrepreneurship
Education in China and Abroad
IE Research and practice spreads across the world, though in varying degrees. In
developing countries like the United States where entrepreneurship receives
general recognition and concern, entrepreneurial enterprises contribute to 40 % of
the value created by all enterprises and have created 75 % of the new job
opportunities in that country.
108 Innovative and Entrepreneurship 1025
Entrepreneurship in China, which started from the 1970s when China adopted
the open-up policy, has gone through six stages in its development marked with
four climax periods. Chinese governments at various levels have begun to build
entrepreneurial cities since 2009, followed by national wide popularization of IEE.
Internationally, researches on entrepreneurship education (EE) began from the
1940s and witnesses fruitful results. In the past decade, western scholars carried
out studies on EE centered on thirteen hotspots, such as EE adjustment and cultural
interpenetration, business and management education, entrepreneurship manage-
ment, business models and EE courses (Li 2007).
In China, there has been a remarkable improvement in EE research and practice
since 2006. Some successful EE modes were formed to solve problems like
ambiguous orientation, unsuccessful localization of western IEE concepts and
modes, ineffective teaching and practice (Lu 2011). Chinese scholars from various
disciplines began to show interest in IEE from the very beginning of the 21st
century and expressed their views on IEE from different perspectives. In respect of
performance evaluation, Professor Xie Zhiyuan employed an analytic hierarchy
process in the qualitative research on the performance evaluation system for
China’s IEE (Xie and Liu 2010); Vesper’s Seven Elevation Factors was introduced
into the comprehensive evaluation on IEE in Chinese universities. These resear-
ches are important for us to learn about the development of IEE in China. How-
ever, it appears that most studies are introductions of western IEE concepts and
experiences or generalized suggestions, and there is an obvious lack of studies on
IEE in regional contexts, especially empirical studies, of more practical value.
108.4 Major Problems for Innovative
and Entrepreneurship Education in Western China
It has always been a serious common problem for western China in developing
IEE that very little progress could be made, though a large amount of manpower,
materials, money and time have been devoted into it, and IEE research and
practice are still on a superficial level where innovation has not been fully sub-
stantialized. To be specific, IEE in western China is confronted with big imped-
iments in respect of resources, concepts as well as educational and supporting
systems.
First, most of the western regions in China are undeveloped with relatively
limited resources and educational funds. Local governments, as the policy maker
and allocator of various social resources, are still clinging to conservative
administrative concepts and beliefs, while subordinate departments, affected by
such work style and attitude of their higher authorities will usually hold a wait-
and-see attitude toward the polices for IEE. This has led to the current situation of
IEE that policy making far outweighs implementation.
1026 C. Lu et al.
Second, economic structure, development level and entrepreneurship environment
vary greatly in different regions. Conservative concepts from the society and families
confront university graduates, who are the mainstay of potential entrepreneurs, with
making a choice between getting employed and starting up a business. And the fact is
that most graduates who choose to start up a business are ‘‘necessity entrepreneurs’’
rather than ‘‘entrepreneurial’’ ones.
Third, IEE in western China is still at its initial stage with many pressing
problems crying out for solutions. Inside the university, there are not enough
specialized teachers, especially teachers with entrepreneurial experiences; students,
because of little access to IEE, lack in entrepreneurial awareness, knowledge and
experiences, which will lead to low entrepreneurial competencies. Outside the
university, ineffective encouragement and supporting systems fail to promote a
robust entrepreneurial culture and strong motivation for IEE development.
Therefore, not many successful startup-businesses are seen so far in western
China, though local governments have adopted many encouragement and prefer-
ential policies to promote entrepreneurship, and some governments even set up
special funds for financial support.
108.5 Implications for Developing Innovative
and Entrepreneurship Education in Western China
Local governments and universities play a very important role in making a radical
change to the backward IEE in western China. Local governments, as the policy
maker and allocator of resources, should be more IEE-supportive and promote
effective utilization of intellectual, manpower, financial and material resources.
Universities, as the main implementer of IEE, is responsible for achieving sub-
stantial progress in IEE by integrating mass education with elite education on the
basis of introducing advanced international IEE concepts. On one hand, they
should popularize basic IE knowledge among all students, guide students to
internalize IE concepts and develop IE competencies. On the other hand, intensive
education should be accessible to students with entrepreneurship mindset, aptitude
and potentials. We hereby would like to outline some preliminary policy and
educational implications for governments and academics on establishing IEE
cultivation mode rooted into the regional contexts in western China.
108.5.1 Develop Students-Oriented IEE Concepts
Concept determines how we act, so it is a must that universities should break away
from the conservative concept which prevents people from risk-taking, which is
one of the main causes for the slow development of IEE in western regions. IEE, in
108 Innovative and Entrepreneurship 1027
essence, is quality-oriented education which targets at the all-round development
of people. Thus, great efforts should be made by local universities in the following
aspects for promoting sound development of IEE in Western China.
First, enhance the exchanges and communication between universities, teachers
and students for achieving a common understanding that the students are wrongly
positioned in the education system and the current situation must be changed from
‘‘university-and-teachers-orientated’’ to ‘‘students-oriented’’, and think about what
they can do for students with entrepreneurship intensions.
Second, integrate IEE into the university curriculum and promote innovative
campus culture. A more open learning environment needs to be created which
entails a flexible IEE practice system and a corresponding diversified evaluation
system. Innovative concepts should be deeply integrated into each step of the
educational process of each discipline. In this way, IE will become an internal
need for students themselves.
Third, based on the intervention-process education concept, create a non-
boundary ‘‘macro-learning’’ environment that integrates the outside society and
jobs with classroom teaching and learning, and provide students with more
opportunities for IE practice to facilitate the internalization and reinforcement of
their IE awareness, knowledge and competencies, and thus, vigorous engagement
in IE practice.
Forth, encouragement and support from families and education by the univer-
sity are two of the most influential factors for students’ entrepreneurship intentions
(Wu and Zhang 2008), according to the findings of empirical studies. The
implication is that universities should also strengthen their connection and com-
munication with the students’ families to achieve their recognition on IE and
support for the students’ entrepreneurship intentions and actions.
108.5.2 Integrate and Optimize Various Resources to Open
New Channels for Cultivating IE Talents
To realize the effective integration and optimization of limited resources is of
crucial importance for IEE development in western regions. Solving this problem
needs joints efforts from governments and universities.
On one hand, universities should take advantage of opportunities brought by
internationalization of education and preferential policies made by governments to
achieve the following goals. First, based on the cross-disciplinary attributes of
IEE, employ the 4C cultivation concepts, namely cross-culture, cross-region,
cross-discipline and cross-specialty throughout the IEE process. Second, enhance
international cooperation and regional mutual support. Third, improve the culti-
vation mode via promoting integration of disciplines and specialties to improve
students’ knowledge structure, upgrade their concepts and beliefs and develop
their competencies.
1028 C. Lu et al.
On the other hand, a supporting system of government-led four-dimensional
nexus which involves partnerships between universities, industries, governments
and students’ families need to be established for universities to adopt a more
flexible credit system for IE courses and practice which allow teachers and stu-
dents to work or study in governments or enterprises. And government officials
and entrepreneurs should be absorbed into the university teaching faculties, which
is a key to the problem of lacking in specialized teachers with entrepreneurial
experiences. Only in this way can we realized the integration and optimization of
limited resources from various regions, industries and fields.
108.5.3 Integrate IEE into the Whole Education System
to Benefit People from all Walks of Life
with Entrepreneurship Intensions
A research team from the Experimental Zone for the Reform Pilot Project to
Cultivate Interdisciplinary Entrepreneurial Talents in China-ASEAN Free Trade
Zone conducted survey on the current situation of IEE in underdeveloped western
regions of China in 2010. One of the focuses of the survey is to learn about
university students’ self-evaluation on IE. According to its empirical analysis,
students think themselves of medium level in terms of the first-level entrepre-
neurship indicators which consist of awareness, psychological qualities, knowl-
edge and competencies. For most of the students, entrepreneurial knowledge is
lower than the other three indicators and is considered of greater need for uni-
versity students with entrepreneurship intentions. Among the 28 second-level
indicators, professional abilities, innovative abilities, learning abilities and foreign
language communicative abilities are considered most important. Result of the
survey is recognition on the necessity and feasibility of popularizing IEE in uni-
versities in western China. Yet, IEE is a continuous and dynamic life-long process
and should be integrated into the whole education system. Universities, while
popularizing IEE among students throughout their university study on a basis of
the process-oriented education concept, which will help students to lay a solid
foundation for future business-startup, should also provide opportunities for
graduates and people from all walks of life with entrepreneurship intensions to get
access to IEE via continuing education or in more flexible ways, such as distance
training programs, lectures, and the like. This during-and-after-university mode of
IEE sees better IE prospects. So we maintain that university students do not need
to make a choice between getting employed and starting up a business upon
graduation. The choice should be made when everything is ready.
108 Innovative and Entrepreneurship 1029
References
Li G (2007) Hotspots of international entrepreneurship education. High Educ Dev Eval China
27(4):70–76 (Chinese)
Lu B (2011) Establishment of education mode for cultivating innovative and entrepreneurial
talents. Hei Longjiang High Educ Res China 207(7):140–141 (Chinese)
Wu Q, Zhang H (2008) An empirical study on the influence of environment for innovative and
entrepreneurship on students’ entrepreneurship intensions. Hei Longjiang High Educ Res
China 175(11):129–131 (Chinese)
Xie Z (2009) Localization of entrepreneurship education at undergraduate level. Explor Educ
Dev China. 30(4):81–832 (Chinese)
Xie Z, Liu W (2010) Evaluation system for innovative and entrepreneurship education in
universities. Innovative Entrepreneurship Educ China 1(6):3–8 (Chinese)
Chapter 109
Network-Based Optimal Design
for International Shipping System
Er-shi Qi, Lian-yu Zhu and Meng-wei Yu
Abstract Lean concept and lean thinking are means of expressions for industrial
engineering reflect in different countries, enterprises and environment. Cost
management in international shipping system is an application of system optimi-
zation used lean management theory and method. After optimization, lean cost
management can be realized.
Keywords Cost Lean Network Optimization Shipping
109.1 Introduction
Based on analysis of plenty of literatures and summaries, so far related research
usually adopt one time, single phased, static decision making using minimized
vessel amount and transportation cost, major in transporting a certain kind of cargo,
specific quantity, constant port of call, fixed freight and a clearly transportation
period. There are rare strictly route integrated optimization model for tramp
transportation, same as balance in multi phased transportation and scheduling
suitable tonnage in existing research. The inevitable trend of the future research is
to establish a series of decision supporting systems including route optimization,
marketing, ship management, safety management, cost management, performance
appraisal, business pattern and fleet development (Lu 2008; Liu 2010).
E. Qi L. Zhu
Management and Economics Department, Tianjin University, Tianjin, China
L. Zhu M. Yu (&)
CCCC International Shipping Corp, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1031
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_109,
Ó Springer-Verlag Berlin Heidelberg 2013
1032 E. Qi et al.
Shipping Lane
Shortest route Climatological route Meteorological route Optimal route
Chart 109.1 Shipping lane-2 (Zhang 2011)
Chart 109.2 Optimization r 10
P2 P5
of shipping lane
r 14
r1 r9
r4
r 11
r2 P3 r8 P6 P8
P1
r 15
r5 r 12
r7
r3
r 13
P4 P7
r6
At present, China is continuously exploring in lean management theory, related
theory and method still imperfect. Improve work efficiency, reduce management
cost and waste by introducing and exploring advanced lean theory will dramati-
cally affect survival and development of the industry (Yang 2010; Liang 2010).
In international shipping system, it usually involves complicated network
system, and lean cost in management of international shipping is the goal which
enterprise pursue. An optimal sailing route is the key in shipping system. Optimal
sailing route is usually defined as, with respect to the navigation safety, the route
which has shortest transit time and best economic benefit. Optimal sailing route
infinitely close to the ideal condition will bring higher economic benefit that has an
important practical significance especially for long distance ocean transportation
(Guo 2008; Hu and Wang 2011; Zhao and Xiao 2009).
109.2 Methodology
Marine cargo transportation is very important in international shipping. The profit
point lies in the lean cost management. There are many factors which affect the
cost, for example ship’s tonnage, stowage, port of call and order, optimal shipping
lane etc. Shipping lane is most important one. There are many factors which affect
the shipping lane for instance meteorological condition, sea condition, barrier,
location, ship’s condition and practice course. Meteorological condition is rela-
tively obvious among the above including monsoon, depression, sea mist, drift ice
109 Network-Based Optimal Design for International Shipping System 1033
Table 109.1 Shipping lane-1 (Huang 2007)
Shipping lane Philosophy of measurement
The rhumb line When sailing in low latitudes, near north and south or in short voyage, the
rhumb line is very close to the shortest route.
Great circle sailing Great circle sailing line is an aggregation of plenty of rhumb line.
line
Parallel sailing Define as sailing follow a same latitude circles, it is an exception of the
line rhumb line.
Composite sailing The shortest route under limited latitudes, it is a combination of great circle
line sailing line and parallel sailing line.
and iceberg. Sea condition such as ocean circulation, swell. Ship’s conditions
comprise vessel age, draft, speed, tonnage, stowage and ship’s crew.
Voyage cost, usually refers to cost of unit time in navigation (Pa) multiplied by
the transit time (Ta) plus the sum of cost of unit time in port when loading/
discharging (Pp) multiplied by the time in port (Tp).
Computational formula:
X
P¼ ðPa Ta Pp TpÞ ð109:1Þ
The semi-submersible vessel operated by Group K is a typical representative of
ocean transportation, this cross ocean transportation we call it shipping lane. Far
off coast, sailing route is regularly long and easily affected by meteorological
condition is the outstanding features of shipping lane. During transit, it is hard to
avoid severe sea condition. Especially during transit in unfamiliar regions, ship’s
crew can only rely on the bridge documentation. In particular, region vessel needs
to apply to sail with convoy in order to avoid being exposed under the risk of
piracy attack and transit time and route will be restricted.
The Table 109.1 lists four basic shipping lanes (Liu 1992).
Considering hydrological and meteorological conditions.
109.3 Results
Based on lean theory, assume workflow and continuity of the workflow as object,
research on how to ensure the continuity of the workflow, reduce waste in
workflow in uncertain period condition synthetically applied Cycle Operation
Network (CYCLONE), Genetic Algorithm (GA), 4D-CAD, Line of balance
(LOB), Theory of Constraints (TOC), Extensible Markup Language (XML), and
established workflow integrated management method, and realize project work-
flow lean management. Workflow integrated management consists of following
three modules: simulation module, optimization module, visual module.
There are plenty of optimization models in shipping system based on different
view points. For example queuing system targeting in maximizing the benefit of
1034 E. Qi et al.
the vessel or make it most economical in certain index requirement. In general
condition, improving service level both in quantity and quality will reduce the wait
expenses of shipper and increase the cost of service supplier, hence the optimal
condition is to minimize the total amount of two expenses, take both shipper’s and
ship owner’s interest into consideration, optimize fleet size, and minimize costs
and expenses. In steady state circumstance, all kinds of cost and expenses are
calculated on time basis, Z represents the sum costs in unit time, which is the
optimization of the accumulated cost.
Shortest route is a sticking point when choosing sailing route. Moore-Dijkstra
algorithm is widely used in shortest route design and optimization. It is a kind of
tagging method, the basic ideas starts from origin, explores path of shortest length
and marked every vertex in the process, gets the fixed tag from the previous vertex
or the temporary tag from the origin (Du 1995; Wu 2010; Yang 2010; Xu 2011;
Liu 2006).
First make L(P1) = 0, L(P) =+?(P = P1),
T1 ¼ P1 ; S0 ¼ fPn g ð109:2Þ
L Pj ¼ min L Pj ; SðPn Þ þ Knj ð109:3Þ
Make L Pjn ¼ min L Pj Pj 2 Sn ð109:4Þ
we can get four kinds of shipping lane.
When j = 0, L(P2) = r1 minimum
When j = 1, L(P3) = r2 minimum
When j = 2, L(P5) = r1 ? r10 minimum
When j = 3, L(P6) = r2+r8 minimum
When j = 4, L(P4) = r3 minimum
When j = 5, L(P7) = min{(r3 ? r6), (r2 ? r7)} minimum
When j = 6, L(P8) = min{(r2 ? r8 ? r15), (r1 ? r10 ? r14)} minimum
109.4 Conclusion
Refer to the different quality, quantities, forms, value, units of measurement of
various goods, the demands for transportation presents the hierarchical charac-
teristics, different in freight, time, safety level and frequency. In the meantime,
with the rapid development of social economy and constantly improved of regional
comprehensive transport network, multimodal transportation has become the
direction and goal of modern transportation and logistics development. In this
context, it is with great significance in considering complex cargo, choosing the
optimal dispatch, type of transportation and route, highlighting the effect of
characteristic to optimize comprehensive transportation system.
109 Network-Based Optimal Design for International Shipping System 1035
International marine logistic network is defined as an aggregation of connected
organization and facilities in goods flow between different countries and regions. It
serves to prompt international trade, transnational operation and accordance of
marine system in difference countries. Regional marine logistic network is defined
as an aggregation of connected organization and facilities in special cargo flow
within different regions. It has realized the optimal allocation of resources and
prompted logistic system accordance in different regions, countries and ports.
Various shipping companies, ship owners and charterers are very concerned
about a shorter, economic and safety route, it is also a significant issue in shipping
development. So far there isn’t a complete, comprehensive and practical route
optimization theory in worldwide, and existing methods are combination and
iteration based on various kinds of algorithm that meet the different needs of user.
References
Du Y (1995) A practical optimal method of scheduling, system theory and practice. vol 2, Hubei,
Wuhan, China
Guo R (2008) Transportation systems engineering. National Industry Publishing Company,
Beijing, pp 126–127
Hu M, Wang Y (2011) Marine transportation business. China Communications Press, Beijing,
China, pp 238–250
Huang Y (2007) The optimization of Yangtze freshwater liner shipping routes. Shanghai Jiao
Tong University, Shanghai, China
Liang J (2010) Study on resource allocation decision of railway container terminal. Southwest
Jiaotong University, Chengdu, Sichuan, China, March 2010
Liu F (1992) A new calculating method for minimum-time route. J Dalian Marine Coll
18(3):231–235
Liu J (2006) An optimization study for the tramp shipping system, Shanghai Maritime University,
Shanghai, China
Liu T (2010) Optimization of multi-transportation organization mode for complicated freight
flow. Wuhan University of Technology, Hubei, Wuhan, China, May 2010
Lu W (2008) Research on work flow lean management for construction projects. Harbin Institute
of Technology, Harbin, Heilongjiang, China, June 2008
Wu L (2010) Study on optimization of bulk shipping process in domestic of a company. Lan
Zhou University, Lanzhou, Gansu, China
Xu J (2011) Graph theory and applications. China science and technology university press,
Beijing, pp 100–106
Yang G (2010) Study on the Evolution mechanism and optimization of regional logistics network
structure. Central South University, Changsha, May 2010
Yang Q (2010) Study on optimization of middle-East line of company A. Dalian Maritime
University, Dalian
Zhang Y (2011) Research on china coastal container transportation market and the optimal
shipping route. Logist Eng Manage 33(11):71–75
Zhao R, Xiao Y (2009) Maritime navigation. China Communications Press, Beijing, China,
pp 382–390
Chapter 110
A Dynamic Analytic Approach to Study
on the Interaction Between Product
Innovation and Process Innovation
of the Equipment Manufacturing
Enterprises
Ting Wang, Ying Wang, Jing Liu and Yang Gao
Abstract The coordinated development of product innovation and process
innovation is an important factor for enterprise’ technology innovation success. In
this work, a systematic dynamic model is proposed to explore the correlation
between product innovation and process innovation. The model is based on system
dynamics theory and methodology. In the case study, experimental analysis of JY
Kinetics Co Ltd. has been carried out with the proposed model. The results show
that the model is able to analyze the correlation between product innovation and
process innovation. It is useful for enterprises to develop technology innovation
strategies and promote its implementations.
Keywords Equipment manufacturing Interactive relationship Product inno-
vation Process innovation System dynamics (SD)
110.1 Introduction
From the industrial point of view, the technology progress has a deep impact on
advanced and rationalized industrial structure of the equipment manufacturing
(Feng 2008). As technological innovation is the main source of the technological
progress in equipment manufacturing, the coordinated development of product
innovation and process innovation is an important factor for enterprise’ technology
innovation success (Kim and Choi 2009). Guizhou Province is a traditional
manufacturing province which has good foundations and development opportu-
nities. However, the rise of the emerging manufacturing provinces through the
T. Wang Y. Wang (&)
Department of Management, Guizhou University, Guiyang, China
e-mail: [email protected]
J. Liu Y. Gao
Department of Mechanical Engineering, Guizhou University, Guiyang, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1037
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_110,
Ó Springer-Verlag Berlin Heidelberg 2013
1038 T. Wang et al.
country is a great challenge to Guizhou equipment manufacturing in recent years.
At present, the equipment manufacturing in Guizhou Province is in urgent need for
technical innovation in order to improve the competitiveness and survivability, so
this research needs to be strengthened. As the relationship between product
innovation and process innovation has the characteristics of complexity, nonlin-
earity, delay, dynamic.ect (Sun 2007), this paper takes the Guizhou JY Kinetics Co
Ltd. for example, through building the system dynamics model to explore the
interaction between product innovation and process innovation and provide policy
analysis tools to help decision-making for equipment manufacturing enterprises,
and finally lay a foundation for the following study of the technological innovation
of the Guizhou regional equipment manufacturing.
110.2 Literature Review
In the 1970s, William J. Abernathy and James M. Utterback put forward the AU
model of the innovation type and innovation degree changing with technology life-
cycle (Utterback and Abernathy 1975), which creates a precedent for collaborative
research of product innovation and process innovation (Bi et al. 2007). After the AU
model was built, Hayes and Wheelwright promoted the formalized relation model of
product innovation and process innovation, namely product—process matrix con-
ceptual model. This model provides a quantitative basis for enterprise production and
market diversification decision-making choices (Hayes and Wheelwright 1979).
Peter M.Milling and Joachim Stumpfe innovatively used the system dynamics (SD)
method, starting from the complexity of product and process which makes changes
with the innovation, so the research of the interaction between them becomes more
systematic (Milling and Stumpfe 2000). Some domestic scholars put forward the
interactive model of product innovation and process innovation corresponding to
Chinese national conditions on the basis of oversea studies. Bi Kexin built the SD
model of the interaction between product innovation and process innovation to have
a simulation study on a particular manufacturing enterprise (Bi et al. 2008). Now,
domestic studies rarely use the SD method to study the interaction between product
innovation and process innovation of the equipment manufacturing. Therefore, it has
an important practical significance to have a study on SD simulation for an equipment
manufacturing enterprise.
110.3 The Establishment of SD Model
110.3.1 Model Assumption
According to assumptions of the resource scarcity in economy and the actual
situation of the majority manufacturing enterprises, this paper assumes that the
110 A Dynamic Analytic Approach 1039
manufacturing enterprise innovation resources are limited. Increasing innovation
investment to a subsystem means reducing another subsystem innovation invest-
ment (Jackson 2005).
Considering that there are many ways of investment of resources in technology
development, but some input methods are difficult to quantify (Zhu 2009), such as
technological innovation incentive policies, science and technology information,
science and technology personnel, etc. But these factors are also affected by the
total investment of technological innovation funds, so this article defines the
investment of research and development (R&D) resources as the capital
investment.
This article studies product innovation and process innovation in R&D and
technology transfer process which is transferred with new products and technology
based on patented technology (Chen et al. 2009).
110.3.2 The Overall Structural Analysis of the Model
From the process of product innovation and process innovation, the interaction
between product innovation subsystems and process innovation subsystem occurs
mainly in decision-making process, R&D process and the manufacturing process.
In the decision-making stage, decision-maker should allocate resources for product
innovation and process innovation to determine the proportion of technology
development inputs and product process innovation (Labeaga and Ros 2003). In
the early R&D stage, product development and design department should
exchange information more frequently with the R&D department (Guo 1999). The
main purpose is to set the framework for the development of process and product.
In the manufacturing stage, there will be more exchange of technical information
between the various departments (Eswaran and Gallini 1996). The overall structure
which displays the interaction between product innovation subsystems and process
innovation subsystem is shown in Fig. 110.1.
Decision-making process
Product innovation subsystem Process innovation subsystem
R & D process Manufacturing process
Fig. 110.1 The overall structure of the model
1040 T. Wang et al.
Number of Potential
Product
Product Innovations
+ innovation rate
+ of Implemented
Number
Product Innovations
+ +
Desired Product Product
+ Innovation - Maturity of Product Attractivenes
+ Life Cycle
Product Features with Resources for
+ Product R&D
Respect to Its Complexity
+ + Number of
Flexibility of
Manufacturing Process Production
Correlation Complexity and Variety
+ - between product of the Product Line + Technology
and process
Transfer Factor
Yield Factor
Process Features Referring to +
-
the Level of Systemization and Desired Process + Process
+
Interconnection Innovation Innovation Rate
- Number of Potential
Process Innovations1
Maturity of Process +
Life Cycle + Number of Implemented Resources for
Process Innovations Process R&D
Fig. 110.2 Product and process causal loop diagram
110.3.3 Causal Loop Diagram
The main effect variable in the subsystem of product innovation and process
innovation is only selected to structure the causal loop (see Fig. 110.2) according
to the actual situation of equipment manufacturing and the point of view of
availability and maneuverability.
110.3.4 Research Model
The variables are quantified and structured simple SD model (see Fig. 110.3)
according to the causal loop diagram. In this model, number of implemented
Number of Potential Delay of Product Number of
Product Implemented Product
Product Innovations Innovation Innovations
Resources for Innovation Rate
Product R&D Fixed Time
Delay(Product Product
Desired Product Maturity of Product Attractivenes
Innovation) Innovation Life Cycle
Quantity of Initial
Number of the
Production
Patents Registered Product Features Maximum Point of
with Respect to Its Product S-curve Quantity of
Complexity Production
Investment Resources Correlation
Proportion R&D between Product
Technology and Process Flexibility of Quantities of
Transfer Factor Manufacturing
Process Yield Factor Production
Complexity and
Variety of the
R&D Product Line Process Features with
Technology Respect to Its
Resources for and Equipment Fixed Time Interconnection
Process R&D Fixed Net Delay(Process
Desired Process Maturity of Process
Innovation) Life Cycle
Innovation
Maximum Point of
Number of Potential Delay of Process Process S-curve
Number of
Process Innovations Innovation Process Implemented Process
Innovation Rate Innovations
Fig. 110.3 System dynamic simulation of product innovation and process innovation
110 A Dynamic Analytic Approach 1041
product innovations and number of implemented process innovations are two level
variables, while product innovation rate and process innovation rate are two rate
variables, and other variables are auxiliary or constant variables.
110.4 Simulation Analysis Model
The simulation target of this paper is a famous equipment manufacturing enter-
prise, the JY Kinetics Co Ltd. in Guizhou Province. JY Kinetics Co Ltd. has more
than 30 years research and development history, which is very popular in Guizhou.
This company is a typical example, which has ability to innovate, and the new data
of product innovation and process innovation in this company have been obtained
from interviews and questionnaires.
The SD software, vensim_ple 5.11a is used to do the simulative analysis. The
operation time of the model is from 2006 to 2015, and time step is one year.
110.4.1 Calibration of the Model
As far as possible, the model was calibrated to make the consistency of model’s
behavior and system behavior. The model passed the unit consistency test and
model test firstly. Then according to the practical data of JY Kinetics Co Ltd., the
model is repeatedly calibrated, and at last the behavior of the model becomes very
close to the reality. In the end, the relative error is less than 10 %.
110.4.2 Policy Tests
From the simulation of product innovation and process innovation, Fig. 110.4
shows the tendency of JY Kinetics Co Ltd.’s product-process interaction at
present. In this graph, product innovation rate is gradually diminishing, at the time,
process innovation rate is surpassed by product innovation rate in 2012.
Because this paper focuses on the political simulation and effect prediction in
order to aid decision making. Thus we choose the three variables as below which
can be regulated and controlled by the managers.
Change investment proportion. Under the circumstance which other variables
are not changed, adjust investment proportion = 2 (current strategy) to investment
proportion 1 = 0.5, investment proportion 2 = 8. Then we got Fig. 110.5.
Compared with Figs. 110.4 and 110.5, we can figure out that when investment
proportion reduces, the resources for product R&D reduce and then product
innovation rate reduces a lot. At the same time, the resources for process R&D
increase and process innovation rate increases in certain extent. However, along
1042 T. Wang et al.
Fig. 110.4 The result of SD Interaction between Product Innovation and
simulation based on basic Process Innovation
strategy 1.5
1.125
Item/Year
0.75
0.375
0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Time (Year)
Product Innovation Rate : current
Process Innovation Rate : current
with the passing time, process innovation rate will gradually reduce. When
investment proportion increases, the resources for product R&D increase too while
product innovation rate is higher than ever before based on basic strategy, but it
still tends to go down as time goes by. Process innovation rate increases, which
causes a small-scope fluctuation.
Change resources R&D. Under the circumstance which other variables are not
changed, adjust the resources R&D = 15 Million Yuan (current strategy) to
resources R&D1 = 7.5 Million Yuan, resources R&D2 = 30 Million Yuan. From
the result of the simulation shown in Fig. 110.6, we can figure out that when
resources R&D reduces, both product innovation rate and process innovation rate
reduce by a large margin. When resources R&D increases, both product innovation
rate and process innovation rate are higher than ever before, but the tendency of
product innovation rate is down and the amplification of process innovation rate
last long as time goes by.
Change correlation between product and process. Under the circumstance
which other variables are not changed, adjust the correlation between product and
process = 0.6 (current strategy) to correlation between product and process
1 = 0.3, correlation between product and process 2 = 0.9. The result of SD
Interaction between Product Innovation and Process Innovation Interaction between Product Innovation and Process Innovation
1.5 1.5
1.125 1.125
Item/Year
Item/Year
0.75 0.75
0.375 0.375
0 0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Time (Year) Time (Year)
Product Innovation Rate : Investment Proportion1 Product Innovation Rate : Investment Proportion2
Process Innovation Rate : Investment Proportion1 Process Innovation Rate : Investment Proportion2
Fig. 110.5 The result of SD simulation by adjusting investment proportion
110 A Dynamic Analytic Approach 1043
Fig. 110.6 The result of SD Interaction between Product Innovation and
simulation by adjusting the Process Innovation
resources R&D 1.5
1.125
Item/Year
0.75
0.375
0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Time (Year)
Product Innovation Rate : Resources R&D1
Process Innovation Rate : Resources R&D1
Interaction between Product Innovation and
Process Innovation
1.5
1.125
Item/Year
0.75
0.375
0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Time (Year)
Product Innovation Rate : Resources R&D2
Process Innovation Rate : Resources R&D2
simulation by adjusting the correlation between product and process is illustrated
in Fig. 110.7. When the correlation between product and process reduces, process
innovation rate don’t increase very much until 2014, which is affected by yield
factor. When the correlation between product and process increases, product
innovation rate and process innovation rate gradually converge as time goes by.
110.5 Conclusions and Policy Implications
By means of computer simulation, it can help to learn about dynamic complexity
and the sources of policy resistance, as well as provide more effective policy
suggestions (Lee and von Tunzelmann 2005). It’s important for equipment man-
ufacturing enterprises to develop technology innovation strategies and promote its
implementations. These SD simulation results show that the development of JY
Kinetics Co Ltd’s product innovation and process innovation is relatively slow.
Thus, the policy suggestions of this paper are as follows. First, the scale value
1044 T. Wang et al.
Fig. 110.7 The result of SD Interaction between Product Innovation and
simulation by adjusting the Process Innovation
1.5
correlation between product
and process
1.125
Item/Year
0.75
0.375
0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Time (Year)
Product Innovation Rate : Correlation between Product
and Process1
Process Innovation Rate : Correlation between Product
and Process1
Interaction between Product Innovation and
Process Innovation
1.5
1.125
Item/Year
0.75
0.375
0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Time (Year)
Product Innovation Rate : Correlation between Product
and Process2
Process Innovation Rate : Correlation between Product
and Process2
between the investment of product and process innovation should be kept in a
certain scope, and the advisable investment proportion is among 0.5–2 for JY
Kinetics Co Ltd. Second, equipment manufacturing enterprises should paid more
attention to the process innovation. The simulation results show that increasing the
resources for process R&D makes very positive effects on the process innovation
and these effects will last for a long time. Compared to this, the resources for
product R&D which maintains the increase can only make sure the stable devel-
opment of product innovation. Third, the higher the correlation between product
and process is, the more attention the company should paid to the management of
product innovation and process innovation. Moreover, from the long-term devel-
opment trend of the equipment manufacturing enterprises, they should focus on the
coordinated growth of product innovation and process innovation as well as pay
more attention to the coordinated management of product innovation and process
innovation. With the help of system dynamic method to analyze the development
110 A Dynamic Analytic Approach 1045
trend of product innovation and process innovation, we try to make out the optimal
strategy about how to make these two factors which can develop steadily. Fourth,
enterprises should have more people who have strong ability on conversion
achievement, and only by this way can the transfer ability of company’s research
be enhanced. It’s very important to communicate with stakeholders, and then it’s
possible for us to make a positive and active innovative environment to attract
innovative personnel and promote technological innovation.
Acknowledgments This project is supported by Humanities and Social Sciences Foundation of
Department of Education in Guizhou Province and soft science Foundation of Guiyang Science
and Technology Department under the grant number 2010-2-27.
References
Bi K-x, Ai M-y, Li B-z (2007) The classification study on analysis models and approaches of
synergy development between product innovation and process innovation (in China). Chin J
Manage Sci 15(4):138–149
Bi K-x, Sun D-h, Li B-z (2008) Product innovation and process innovation—a system dynamic-
based simulation of the interaction in manufacturing enterprises (in China). Sci Sci Manage of
Sci Technol 12:75–80
Chen S, Bi K-x, Gao W (2009) Systematic analysis on associated factors between product
innovation and process innovation based on manufacturing industry (in China). Chin J Mech
Eng 20(6):709–713
Eswaran M, Gallini N (1996) Patent policy and the direction of technological change. RAND J
Econ 27(4):722–746
Feng M (2008) Study on technological progress of equipment manufacturing industry in China:
1996–2006 (in China). World Econ Polit Forum 2:67–69
Guo B (1999) Study on modes and the interaction between product innovation and process
innovation (in China). Sci Manage Res 6:51–55
Hayes RH, Wheelwright SC (1979) The dynamic of process-product life cycles. Harvard Bus Rev
57(2):127–136
Jackson MC (2005) Systems thinking—creative holism for managers, ch. 5. China Renmin
University Press, Beijing, pp 35–43
Kim SW, Choi K (2009) A dynamic analysis of technological innovation using system dynamics.
In : Proceedings of the POMS 20th annual conference, Orlando, 1–4 May 2009
Labeaga JM, Ros EM (2003) Persistence and ability in the innovation decision. Business
Economics, Series 1, 2003
Lee T-L, von Tunzelmann N (2005) A dynamic analytic approach to national innovation systems:
the IC industry in Taiwan. Sci Direct Res Policy 34:425–440
Milling PM, Stumpfe J (2000) Product and process innovation-A system dynamics-based analysis
of the interdependencies. In: Proceedings of the 18th international conference of the system
dynamics society
Sun D-h (2007) Study on the interaction between product innovation and process innovation in
manufacturing enterprises (in China). Harbin University of Science and Technology, Harbin
Utterback JM, Abernathy WJ (1975) A dynamic model of product and process innovation.
Omega 3(6):639–656
Zhu T-b (2009) Research on the interaction between influencing factors of regional equipment
manufacturing technical innovation (in China). Harbin University of Technology, Harbin
Chapter 111
A Multi-agent Simulation System
Considering Psychological Stress
for Fire Evacuation
Fan-xing Meng, Qin-lin Cai and Wei Zhang
Abstract Multi-agent simulation is an important method to study fire evacuation.
The present study developed a multi-agent simulation system for fire evacuation.
This system adopted a new-proposed EID (External stimuli-Internal status-
Decision making) behavior model, emphasizing the influence of psychological
stress to people’s escape behavior. Some simulation trials were conducted to
compare the difference between simulation without and with psychological stress.
It was shown that when the influence of psychological stress was not considered,
the simulation results tended to be more ‘‘optimistic’’, in terms of number of
escaped people and average escape time.
Keywords Fire evacuation Multi-agent Psychological stress Simulation
system
111.1 Introduction
Modeling and simulation tools for analyzing fire evacuation are useful in public
place design for enhancing passenger safety (Sharma et al. 2008), and different
tools have been developed to study fire safety (Owen et al. 1996; Galea and
Galparsoro 1994). Among these tools, multi-agent simulation system is used in a
F. Meng (&) Q. Cai W. Zhang
Department of Industrial Engineering, State Key Laboratory of Automobile
Safety and Energy, Tsinghua University, Beijing, China
e-mail: [email protected]
Q. Cai
e-mail: [email protected]
W. Zhang
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1047
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_111,
Ó Springer-Verlag Berlin Heidelberg 2013
1048 F. Meng et al.
growing number of areas (Drogoul et al. 2003; Zhang et al. 2009; Gonzalez 2010).
This tool is based on a computational methodology that allows building an arti-
ficial environment populated with autonomous agents. Each agent is equipped with
sensors, decision-making rules and actuators, which is capable of interacting with
environment (Pan et al. 2007). In the area of fire evacuation study, for moral and
legal reasons, we are not permitted to deliberately expose normal experimental
participants to real fire condition, which will pose a life-threatening degree of risk
(Hancock and Weaver 2005). Multi-agent simulation technique can potentially
help in achieving a better understanding of fire evacuation process without safety
threatens to real people.
Unfortunately, there is a lack of multi-agent simulation frameworks to allow
human factors, such as psychological stress, to be taken into account (Sharma et al.
2008). Fires are perceived as very stressful and a person, who has to decide how to
get out of a building, and away from an uncontrolled fire, is under extremely
psychological stress (Benthorn and Frantzich 1999). According to theories of
information processing, how people interpret information depends on the degree of
stress they are experiencing (Janis and Mann 1977; Miller 1960). Thus, psycho-
logical stress in fire will affect people’s perception of various environmental
factors and thereby influence their actions in the process of fire evacuation (Ozel
2001; Nilsson et al. 2009). For example, a person’s interpretation of emergency
information and other people’s actions, as well as his or her subsequent behavior,
i.e., decision to evacuate, choice of exit and pre-movement time, is partly related
to his or her psychological stress level (Nilsson et al. 2009).
Psychological stress is an important factor affecting people’s evacuation
behavior, which must be considered in multi-agent simulation system. This paper
proposed a new model to describe individual evacuation behavior and developed a
new multi-agent simulation system for fire evacuation, where people may behave
differently depending on their psychological stress level. This simulation system
considered the uncertainty in behavior under psychological stress, which could
obtain more realistic results about fire evacuation.
111.2 System Architecture and EDI Behavior Model
111.2.1 System Architecture
System architecture is shown in Fig. 111.1. In this multi-agent simulation system,
the simulation environment is defined in the ‘‘Simulation Environment Module’’,
such as environment size, layout, and number of exits. The crowd population in the
simulation environment is first initialized in ‘‘Crowd Initialization Module’’ by
setting different types of agents based on their ages, genders and relations with
other agents. In simulation, each agent behaves based on the pre-defined behavior
model, to interact with other agents, as well as the simulation environment. ‘‘Data
111 A Multi-agent Simulation System Considering Psychological Stress 1049
Simulation
Environment
Module Data Collection
Module
Agent’s
Input Crowd Initialization Output
Module Behavior Model
Visualization
Interventions Module
Generation
Module
Fig. 111.1 System architecture of the multi-agent simulation
Collection Module’’ is used to collect simulation data, including changes of
internal attributes of agents, evacuation time of each agent, etc. By the ‘‘Visual-
ization Module’’, the 2D and 3D scenes of evacuation are displayed, which are
valuable to investigate the behavior of each agent and to get comprehensive
understand of fire evacuation process. Furthermore, some intervention methods can
be added to simulation to study their effects on fire evacuation by ‘‘Intervention
Generation Module’’. For example, the effect of leaders in a crowd, the effect of
different warning signals, etc.
111.2.2 EDI Behavior Model
In multi-agent simulation system, the most important part is the behavior model of
agent, which determines the validity of simulation results. The behavior model in
this study adopts the process of ‘‘External stimuli-Internal status-Decision mak-
ing’’ (see Fig. 111.2), which is called EID model in the present study. The basic
idea in EID model is that external stimuli have direct effects on a person’s internal
status, which in turn influence the process of decision making (Luo et al. 2008).
In EID model, external stimuli are divided into four categories by people’s
sensory system: visual stimuli, auditory stimuli, thermal stimuli and olfactory
stimuli. Visual stimuli include the burning fire and smoke, which are quantified by
fire size and smoke density, respectively. The scale of fire and smoke will increase
with simulation time goes on. Auditory stimuli include the sound of fire alarm and
the sound of burning fire. They are quantified by their sound intensity. Fire alarm is
activated when the fire breaks out, and its sound intensity keeps constant during
fire. The sound intensity of burning fire varies with the scale of fire. Thermal
stimuli represent the heat caused by burning fire, which is quantified by envi-
ronment temperature. The olfactory stimuli represent the smell of smoke, which is
quantified by smoke density. All these quantified parameters have five levels from
0 to 5. ‘‘0’’ means there is no stimulus and ‘‘5’’ means the stimulus has reached its
maximum. All these external stimuli collectively impact a person’s psychological
stress level. Psychological stress influences people’s decision making process,
mainly including route choice and travelling speed.
1050 F. Meng et al.
Visual Stimuli Auditory Stimuli Thermal Stimuli Olfactory Stimuli
Intensity
Fire
Burning Smoke Fire Alarm Burning Heat Caused Smell of
Fire Sound by Fire Smoke
Sound
Sound Sound Intensity Environment
Fire Size Smoke Density Intensity Smoke Density
of Fire Alarm of Burning Fire Temperature
Stress Level
Travelling
Route Choice
Speed
Fig. 111.2 EID behavior model in the multi-agent simulation system
Emergency signs are set up at each intersection, which direct agents to the exit
by the optimal route. The decision of route choice is a two-stage process. In the
first stage, agents recognize the directions given by emergency signs at intersec-
tions. Previous studies showed that most participants in fire evacuation do not pay
enough attention to emergency signs (Tang et al. 2009), thus whether an agent can
find signs at intersection is a probability event, whose probability is influenced by
psychological stress level of this agent. It is believed that agents tend to neglect
emergency signs under high psychological stress, thus the probability of finding
emergency signs is negative correlated with agents’ psychological stress level. In
the second stage, agents evaluate the hazard level of each direction by scale of fire
and smoke. If an agent doesn’t find signs in the first stage, it will choose the
direction with least hazard level from all optional directions; if it recognizes
the directions given by signs, it will continue to judge whether the hazard level of
the given direction exceeds the pre-defined threshold. If no, it will travel to the
direction given by the sign; if yes, it will choose the direction with least hazard
level from other directions. It is believed that people will increase walking speed
under stress, and even start to run when surfing extreme psychological stress. In
the behavior model, agents’ travelling speed is positively correlated with their
psychological stress level (Fig. 111.3).
111.3 System Development and Simulation Results
111.3.1 System Development
Based on the system architecture described above, a multi-agent simulation system
is developed. The test scenario is a public place where all the passengers are
111 A Multi-agent Simulation System Considering Psychological Stress 1051
Reach an
intersection
Find the sign No
Yes
Evaluate the hazard Evaluate the hazard
level of chosen route level of all routes
Choose the route given Higher than the
No Choose the route with
by sign threshold
least hazard level
Yes
Choose the route with
least hazard level
from other routes
Fig. 111.3 The decision process of route choice
unfamiliar with the layout of environment. Human behaviors in normal and
emergency conditions are simulated.
The simulation environment is set up in the ‘‘Simulation Environment Mod-
ule’’. It is created as a virtual public place where seven horizontal and vertical
corridors are crossed (see Fig. 111.4), forming an area of 75 9 75 m. The width of
each corridor is 3 m.
The crowd is initialized in ‘‘Crowd Initialization Module’’. 100 agents are
uniformly dispersed along these corridors. Under normal condition, their travelling
speed is set as 0.8 m/s, which will be increased according to agents’ psychological
level in fire emergency. The maximum value of travelling speed is 5 m/s.
At the beginning of simulation, all agents travel normally in the environment. A
fire in the middle of the environment can be activated by a pre-defined button and
fire alarm is also sounded. Fire and smoke spread along the corridor and agents
begin to escape based on their behavior model. In ‘‘Visualization Module’’, 2D and
3D scene of evacuation are displayed (see Fig. 111.5). 2D scene provides us
macroscopic view of the simulation process, which can show the travel charac-
teristic of agent flow, for example, how fire and smoke influence the route choice
of different agents. 3D scene can give more detailed view of agents’ escape
behavior and the interaction between different agents can be observed clearly.
1052 F. Meng et al.
Fig. 111.4 The top views of test environment in normal condition (left) and in fire emergency
(right)
Fig. 111.5 2D (left) and 3D (right) scenes of simulation environment
‘‘Data Collection Module’’ is used to collect real-time data of each agent during
simulation. The data include: travel speed, psychological stress level, probability
of recognizing emergency signs and escape time. These data are used to analyze
the efficiency of fire evacuation.
111.3.2 Simulation Results
The most important part of the present simulation system is to consider the
influence of psychological stress. To demonstrate the effect of psychological
stress, two different kinds of simulation are conducted, and their results are
compared. In the first kind of simulation, agents’ psychological stress level is not
taken into consideration, that is, agents’ behavior are not affected by external
stimuli. In the other simulation, the influence of psychological stress is added, just
111 A Multi-agent Simulation System Considering Psychological Stress 1053
Table 111.1 Summary of simulation results
Number of escaped agents Average escape time/s
mean s.d. mean s.d.
Without stress 94.9 2.5 37.5 3.0
With stress 83.8 3.7 43.9 4.3
p value \0.001 0.002
as what is described above. For each kind of simulation, ten trials are conducted. In
each trial, the number of escaped agents, which represents how many agents have
escaped before the fire spreads all around the environment, and the average escape
time of each agent, which indicates how much time an agent spends to reach the
exit averagely, are recorded to make comparison.
Summary of simulation results are displayed in Table 111.1. The t-test is
conducted to compare the difference between the two simulations. The number of
escaped agents in the simulation without stress (mean = 94.9, s.d. = 2.5) is sig-
nificantly higher than that in the simulation with stress (mean = 83.8, s.d. = 3.7),
with a p value of less than 0.001. The average escape time is 35.5 s (s.d. = 3.0 s)
and 43.9 s (s.d. = 4.3 s) in the simulation without stress and with stress, respec-
tively, and their difference is significant (p value = 0.002).
111.4 Discussion and Conclusion
In the present study, a multi-agent simulation system is developed, where the
influence of psychological stress to escape behavior is considered. An EID behavior
model is proposed to describe how external stimuli affect the internal status
(psychological stress) of people and further change their decision making process.
Based on this model, agents in simulation system can behave more realistically and
can be more similar with a real person. To demonstrate the influence of psychological
stress to people’s evacuation behavior, two different kinds of simulation, without and
with psychological stress, were compared. Simulation results also showed that there
was significant difference in whether psychological stress was taken into account or
not. Under high psychological stress, agents exhibited poor performance, and the
evacuation efficiency was also low.
However, there are still some limitations in the present study. First, the group
behaviors are not considered. The social behaviors, such as queuing and herding,
were believed to have some influence on people’s evacuation behaviors (Yuan and
Tan 2011), which will be focused on in the future studies. Second, other indicators
of internal status are not included. For example, visibility range can affect the
human behavior significantly (Yuan and Tan 2011). More human factors should be
considered to promote the validity of multi-agent system.
1054 F. Meng et al.
Acknowledgments The authors would like to acknowledge the support of research program of
Foxconn Technology Group.
References
Benthorn L, Frantzich H (1999) Fire alarm in a public building: how do people evaluate
information and choose an evacuation exit? Fire Mater 23(6):311–315
Drogoul A, Vanbergue D, Meurisse T (2003) Multi-agent based simulation: where are the agents?
Multi-agent-based simulation II:43–49
Galea E, Galparsoro J (1994) EXODUS an evacuation model for mass transport vehicles. Fire Saf
J 22:341–366
Gonzalez RA (2010) Developing a multi-agent system of a crisis response organization. Bus
Process Manage J 16(5):847–870
Hancock P, Weaver J (2005) On time distortion under stress. Theoret Issues Ergonomics Sci
6(2):193–211
Janis IL, Mann L (1977) Decision making: a psychological analysis of conflict, choice, and
commitment. Free Press, New York
Luo L, Zhou S, Cai W et al (2008) Agent-based human behavior modeling for crowd simulation.
Comput Anim Virtual Worlds 19(3):271–281
Miller JG (1960) Information input overload and psychopathology. Am J Psychiatry
116(2):695–704
Nilsson D, Johansson M, Frantzich H (2009) Evacuation experiment in a road tunnel: A study of
human behaviour and technical installations. Fire Saf J 44(4):458–468
Owen M, Galea ER, Lawrence PJ (1996) The EXODUS evacuation model applied to building
evacuation scenarios. J Fire Prot Eng 8(2):65–84
Ozel F (2001) Time pressure and stress as a factor during emergency egress. Saf Sci
38(2):95–107
Pan X, Han CS, Dauber K et al (2007) A multi-agent based framework for the simulation of
human and social behaviors during emergency evacuations. AI Soc 22(2):113–132
Sharma S, Singh H, Prakash A (2008) Multi-agent modeling and simulation of human behavior in
aircraft evacuations. IEEE Trans Aerosp Electron Syst 44(4):1477–1488
Tang CH, Wu WT, Lin CY (2009) Using virtual reality to determine how emergency signs
facilitate way-finding. Appl Ergonomics 40(4):722–730
Yuan W, Tan KH (2011) A model for simulation of crowd behaviour in the evacuation from a
smoke-filled compartment. Physica A 390:4210–4218
Zhang Q, Zhao G, Liu J (2009) Performance-based design for large crowd venue control using a
multi-agent model. Tsinghua Sci Technol 14(3):352–359
Chapter 112
A Multi-Granularity Model for Energy
Consumption Simulation and Control
of Discrete Manufacturing System
Jun-feng Wang, Shi-qi Li and Ji-hong Liu
Abstract The sustainable manufacturing makes the discrete industry considering
the energy efficiency of the production process. Energy consumption becomes a
very important indicator of energy efficient manufacturing. Discrete event simu-
lation plays a vital role in evaluating the performance of the production plan.
Energy related decisions making of the production plan by simulation need a
formal energy consumption model to evaluate the manufacturing process. In this
paper, a multi-granularity state chart model is proposed to simulate and control the
energy consumption process of the production. A general energy consumption
profile is defined and some key states in a working cycle of a CNC machine are
clarified for energy audit and saving control purpose. A CNC machine with five
energy consumption states is used as an example to illustrate the use of the model.
Some performance indicators are collected from the simulation and compared to
show the effective of the model.
Keywords Discrete manufacturing system Energy consumption Simulation
State chart model
112.1 Introduction
With the climate change of the earth and the unsecured energy supply, the efficient
use of available energy resources is one of the key approaches in the modern
society and industry. Companies today are becoming increasingly interested in
J. Wang (&) S. Li
Department of Industrial and Manufacturing System Engineering, Huazhong University
of Science and Technology, Wuhan, People’s Republic of China
e-mail: [email protected]
J. Liu
School of Mechanical Engineering and Automation, Beihang University,
Beijing, People’s Republic of China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1055
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_112,
Springer-Verlag Berlin Heidelberg 2013
1056 J. Wang et al.
measuring and reducing the environmental footprint of their products and activi-
ties. The manufacturing industry, with its about 75 % of the world’s yearly coal
consumption, 20 % of global oil consumption, 44 % of the world’s natural gas
consumption and 42 % of all electricity produced (IEA 2007), is one of the main
energy consumer and largest emitters of carbon dioxide (CO2). The pressures
coming from energy prices, environmental regulations with their associated costs
for CO2 emissions and the changing purchasing behavior of customers make the
manufacturing industry adopt new methodology and techniques for a sustainable
manufacturing (Bunse et al. 2011).
Energy efficient manufacturing (Rahimifard et al. 2010), which aims to inte-
grated manage the energy efficiency and the production performance, can be
beneficial to industrial companies for economic, environmental or societal aspects
by reducing energy consumption and maintaining the system throughput. Although
energy intensive industries (e.g. steel, cement, pulp and paper, chemicals) remain
in the focus (Solding and Petku 2005), research finds challenges for small and
medium sized enterprises and the non-energy intensive industries (e.g. discrete
mechanical manufacturing industry). They should be not neglected but was
lacking of research attention in the past (Ramírez et al. 2005). Studies show that
there is a significant potential to improve energy efficiency in discrete manufac-
turing. Even with already available technologies, improvements of 10–30 % are
likely to be achieved (Devoldere et al. 2007; Herrmann et al. 2011).
The introduction of energy consumption as a parameter to support the decision
making process may help to forecast and manage the energy costs associated to the
production plan while maintaining the suitable throughput. As a very effective
approach and tool for problem solving and optimizing in manufacturing systems
design, operation and control, discrete event simulation (DES) provides engineers
with a flexible modeling capability for extensive analysis of a production flow and its
dynamic behavior. Currently, the main parameters measured in DES are throughput,
utilizations, and time-span. A review of commercially available manufacturing
simulation tools (e.g. Plant Simulation, Arena, Quest et al.) reveals that they do not
support the energy evaluation for production schedules. With the development of
real time electrical signal monitoring technologies, the information-rich energy data
can be collected and analyzed in the ICT systems (Vijayaraghavana and Dornfeld
2010). A holistic energy consumption model is required to simulation applications
for discrete manufacturing system. In this paper, a multi-granularity energy con-
sumption model will be constructed to simulate and control the discrete machining
manufacturing system for energy management purpose.
112.2 Related Work
Planning and operating energy-efficient production systems require detailed
knowledge on the energy consumption behavior of their components, energy
consumption of production processes. Most of research work focuses on the
112 Multi-Granularity Model for Energy Consumption 1057
energy consumption models of machine tools in manufacturing system based on
the machine components in various running states.
Gutowski et al. (2006) broke down the energy consumption of machine tools
according to functional components such as computer and fans, servos, coolant
pump, spindle, tool changer and so on. The electrical energy requirements of each
component are measured for understanding the energy consumption source.
A modular modeling approach to simulate the power consumption of produc-
tion systems is proposed in Dietmair and Verl (2009) which have been detailed for
process, component and system aspects. The basic structure of a generic energy
consumption model in the form of a digraph with discrete states and transitions
was derived by analyzing the activity of the components.
From the manufacturing of a unit product (Rahimifard et al. 2010), the theo-
retical energy is calculated by the Arena simulation tool using appropriate math-
ematical models representing various processes. This calculated data is
complemented with actual (real) data related to the auxiliary energy and indirect
energy, recorded by advance metering devices and commercial energy manage-
ment systems used within empirical studies. The idle energy is not considered
Weinert et al. (2011) developed a planning system for the detailed prognosis of a
production system’s energy consumption with a concept of Energy Blocks. The
methodology is based on the representation of production operations as segments
of specific energy consumption for each operating state of the production equip-
ment. Energy-saving potentials are addressed and the approach is applied on the
basis of a simulation based evaluation of the energy consumption of a job-shop
manufacturing system for a predetermined production program.
The SIMTER project (Heilala et al. 2008) calculated energy use for specific
equipment or process activities using equipment energy specifications (often
obtained from equipment manuals or vendors) coupled with equipment operation
data (e.g. number of hours the equipment is in different modes of operation). While
such calculations are not very precise, they can indicate the order of magnitude of
energy use.
He et al. (2012) proposed a modeling method of task-oriented energy con-
sumption for machining manufacturing system. An event graph methodology has
been exploited to model the energy consumption driven by tasks in production
processes. But the stochastic failures of the machine are not considered in the
model.
Johansson et al. (2009) indicated the potential use of utilizing DES in combi-
nation with LCA data to generate requirements specification for designing sus-
tainable manufacturing systems and decreasing CO2 emissions. The traditional
idle, busy, and down state are used for energy consumption modeling.
A process module based on state depiction of machine/energy with technical
building services (TBS) module is proposed by Herrmann et al. (2011). An energy
oriented manufacturing system simulation environment is developed using
AnyLogic. When starting a simulation run, single consumption profiles of pro-
duction machines lead to cumulative load curves for the manufacturing system.
1058 J. Wang et al.
TBS-related energy demand of the actual production equipment (e.g. compressed
air) serves as input for appropriate partial TBS-models.
It is observed that there can be a significant amount of energy savings when
non-bottleneck (i.e. underutilized) machines/equipments are turned off when they
will be idle for a certain amount of time. By clarifying the energy consumption
state as start up energy, make part energy, idle energy and turn off energy, several
dispatching rules were proposed to analysis the energy consumption decreasing
effect of especially underutilized manufacturing equipment (Mouzon et al. 2007).
A Finite-State Machines were used to describe the energy consumption model
(Le et al. 2011). A number of different operation states which are linked to the
status of machine components and power consumption. Each operation state is
defined according to the functionality of the machine.
Mori et al. (2011) classified energy consumption of machine tool into constant
power consumption regardless of the running state, power consumption for cutting
by the spindle and servo motor, and the power consumption to position the work
and to accelerate/decelerate the spindle to the specified speed. Most of the current
energy consumption models are for energy statistical purpose and can not be used
for energy consumption control of production process. The different state parti-
tions of machine are specific and not flexible, which have not a holistic view from
the simulation aspect.
112.3 A Multi-Granularity State Chart Model
for Energy Simulation
In order to setup an energy consumption model of machine for manufacturing
system simulation, the model will have appropriate granularity, scalability and
lower computational loading. The energy simulation model should not only sup-
port the energy audit for different view (e.g. plant view, process view and product
view), but also can be used to control the state of the manufacturing system/
facilities in order to reducing the system level energy consumption, for example
changing the state of the facilities to a lower energy mode or power off mode.
112.3.1 Energy Consumption Profile Definition
for Simulation
From the experimentation of current researches, the energy consumption profile of
a manufacturing facility has many distinct states during a working cycle. Here the
energy consumption profile of a CNC machine is analyzed and the method can be
extended to other equipment, such as robotics, conveyor and AGV. The following
112 Multi-Granularity Model for Energy Consumption 1059
factors should be considered when defining the energy consumption profile for
simulation application.
• The partition method of energy consumption process of equipment can be
applied to different facilities used in manufacturing industry.
• Several energy consumption states can be merged to one state or vice versa for
different simulation granularity object.
• The energy consumption state should accommodate important concerned energy
resources (i.e. electricity, gas, heat, and coal) in simulation of production
system.
• Some instantaneous state with higher energy consumption must be included in
order to evaluate the energy efficiency of the control strategy by simulation.
From the literatures (Rahimifard et al. 2010; Gutowski et al. 2006; Dietmair and
Verl 2009; Weinert et al. 2011; Heilala et al. 2008; He et al. 2012; Johansson et al.
2009; Mouzon et al. 2007; Le et al. 2011; Mori et al. 2011), it is known that a
typical electrical consumption profile of a CNC machine has a number of different
operational states which are arose by the activities of its components and deter-
mine the power consumption. In each state, other types of energy resources can be
appended to different state according to the practical requirement for specific
machine. In this paper, the states are clarified to the following types with different
characteristics and simulation intention.
• Power off: The machine power if off and all the energy resources are not
consumed.
• Shut down: The machine will consume some energy to be shut off even if the
duration of this state is very short.
• Warm up: The electrical switch is on and some peripheral equipments of the
machine are start up. Although the warm up time is short, the required energy is
higher.
• Power on: This is an idle state with no material removal. The whole machine
consumes the basic energy in this state. This state can serve as a lower energy
saving mode when no production activity take place.
• Start up: This state is the transition between the power on and production modes.
The main components of the machine (e.g. the spindle and coolant system) will
change to working state. This state is an acceleration process which consumes
higher amplitudes of power consumption with short duration.
• Stand by: This is also an idle state. All drives and pumps of the machine are in
stand-by but with no material removal.
• Production: This is a working state with the material removal process. There is
some short duration with no material remove because of interspaces between the
machining paths. In this paper, the duration of this state is defined as the time
from the product loaded into the machine until it is dropped out the machine.
• Maintenance: The machine is maintained according to preventive maintenance
(PM) schedule or stochastic failure (SF). The energy type and quantity can be
defined for maintenance activities if they are concerned issue.
1060 J. Wang et al.
Start up Stand by
Idle (t2)
Condition 1
Warm up Condition 4
(t1) Condition 2
Power on
Part
Condition 3 Part
arrive
finished
Setup Busy
Power Shut Production
off down
PM or SF
Repaired
Down
Maintenance
Fig. 112.1 State chart model for energy consumption simulation and control
112.3.2 A Multi-Granularity State Chart Model for Energy
Consumption Management
Based on the analysis of the energy consumption profile, a holistic state chart model
for energy consumption simulation and control of discrete manufacturing system is
shown in Fig. 112.1. The model has a multi-granularity form in order to adopt the
different requirements of simulation and control. The dot line in the model will be
used to control the machine state in manufacturing process for energy saving
purpose. The state changing conditions can be related with the arrival time of the
next part or the current state has lasted for a predefined duration. Some states have a
constant duration in production process. For example, execution time of the warm
up (t1) and start up state (t2) will be a constant for a specific machine. The model
has a nested structure which makes it general enough to be extended and modified
for different application scenarios. This modeling method also can be applied to
other faculties such as the conveyor, robot and AGV.
The characteristic and the parameters of each state are summarized in
Table 112.1. The practical duration of state can be constant or stochastic, and the
consumed energy type can be obtained for the specific machine. After the simu-
lation, the required energy and the throughput for a shift of a production plan can
112 Multi-Granularity Model for Energy Consumption 1061
Table 112.1 The characteristics of the energy consumption state
Energy profile Duration characteristic Main energy types
State Sub-state
Setup(Power off) Constant, stochastic No
Idle Warm up Constant Electricity
Power on Stochastic Electricity
Busy Start up Constant Electricity
Stand by Stochastic Electricity, heat, gas
Production Constant, stochastic Electricity, heat, gas
Down (maintenance) Stochastic No or Electricity
be reported. For a specific amount of part production, the energy and the overall
makespan time can also be obtained for decision making. The model has the
following classical using scenarios.
• For a coarse granularity simulation aiming to energy audit, only three states can
be reserved, i.e. power on (idle), production (busy) and maintenance (down),
which are supported by most of the current simulation software. By endowing
with energy consumption data for each state, the energy quantity can be accu-
mulative by multiplying the state duration and its required energy amount
during the simulation process.
• For both the energy audit and energy saving control simulation, all the states in
Fig. 112.2 can be used. Particularly, after the busy state, if the machine queue is
not null (i.e. there are parts waiting for to be machined), the next part will be
machined at once. Otherwise the machine will choose a suitable state (i.e. stand
by, power on, shut down) in order to reduce the energy consumption from the
view of the system level considering the arrival time of the next part or the idle
state lasting for a predefined duration.
Fig. 112.2 The five states
model for energy Idle
consumption control
Warm up Power on
No part waiting Part
and lasting waiting
duration is end
Part Part
arrive finished
Power Shut Production
off down
Setup Busy
1062 J. Wang et al.
• For a finer granularity energy consumption management purpose, the production
state can be further decomposed to more sub-states and then the detailed energy
consumption in production mode will be visualized.
112.4 Experimentation
The experiment is partly based on a case provided by Mouzon et al. 2007, Le et al.
2011. The data herein presented are for the purposes of demonstration of our
method and do not necessarily imply an actual plant floor data.
One hour production ability will be evaluated for a single CNC center. The
inter-arrival time and service time of parts are exponentially distributed with a
mean of 20 and 6 s, respectively. The initial condition of the machine is assumed
to be power off. The warm up takes 5 s and consumes 4 unit powers per second.
The production and power on (i.e. idle) power are 6 and 2 unit powers per second,
respectively. Shutting down a machine takes 2 s, consuming 1 unit powers per
second. The state chart model for energy audit and energy saving control is shown
in Fig. 112.2. When a part production is finished, the machine state will be
changed to different state according to the following rule. If there are parts waiting
for machining, the machine will be changed to production state at once. Otherwise,
the machine will be idle at power on state. When the idle (power on) state lasts for
a predefined duration (e.g. 5 s), the machine will be shut down for energy saving
purpose until there is part arrival for machining.
An ARENA simulation model (Fig. 112.3) has been developed for the state
chart model of the CNC machine with energy saving control strategy. Apparently,
Fig. 112.3 The arena model for energy consumption simulation and control
112 Multi-Granularity Model for Energy Consumption 1063
Table 112.2 Performances of two scenarios in 1 h production
Performance indicator Part throughput Energy consumption
Strategy No control 173 11917
Control 162 8743
Performance result 6.4 % reduce 26.6 % saving
Fig. 112.4 The energy
consumption ratio of five
states in two scenarios
Fig. 112.5 The state time
ratio of five states in two
scenarios
if a machine has more states, the simulation flow module and their logic relation
will be more complex. The five states in Fig. 112.2 are all included in the simu-
lation model by using ARENA StateSet module. By changing the condition in
some modules, the model in Fig. 112.3 can be only used to collect energy con-
sumption data with no energy saving control. That is to say, the machine will stay
at power on state when there is no part to be machined.
Table 112.2 shows the performance of the above two scenarios for 1 h pro-
duction. From the Table 112.2, the state control scenario will have 6.4 %
decreasing in thought with 26.6 % energy saving.
Figures 112.4 and 112.5 show the energy consumption ratio and the state time
ratio of five states in two strategies. Apparently, the power off state in Fig. 112.5
has a relative longer time but have no energy consumption in Fig. 112.4 under the
energy saving control strategy.
1064 J. Wang et al.
112.5 Conclusion
In order to evaluate the energy consumption of the discrete manufacturing system,
a multi-granularity state chart model is proposed for energy consumption process
of the manufacturing equipment. The general energy consumption states are
analyzed and composed for different granular simulation applications. The method
will be extended to other manufacturing faculties and a multi equipments plant
floor will be studied in the future.
References
Bunse K, Vodicka M, Schönsleben P, Brülhart M, Ernst FO (2011) Integrating energy efficiency
performance in production management—gap analysis between industrial needs and scientific
literature. J Clean Prod 19(6–7):667–679
Devoldere T, Dewulf W, Deprez W, Duflou B (2007) Improvement potential for energy
consumption in discrete part production machines. In: Proceedings of the 14th CIRP
international conference on life cycle engineering (LCE), Tokyo, pp 311–316
Dietmair A, Verl A (2009) A generic energy consumption model for decision making and energy
efficiency optimization in manufacturing. Int J Sustain Eng 2(2):123–133
Gutowski T, Dahmus J, Thiriez A (2006) Electrical energy requirements for manufacturing
processes. In: Proceedings of 13th CIRP international conference on life cycle engineering.
Leuven, May 31st–June 2nd, pp 1–5
He Y, Liu B, Zhang XD, Liu XH (2012) A modeling method of task-oriented energy
consumption for machining manufacturing system. J Clean Prod 23(1):167–174
Heilala J, Vatanen S, Tonteri H, Montonen J, Lind S, Johansson B, Stahre J (2008) Simulation-
based sustainable manufacturing system design. In: Proceedings of the 40th winter simulation
conference, pp 1922–1930
Herrmann C, Thiede S, Kara S, Hesselbach J (2011) Energy oriented simulation of manufacturing
systems—concept and application. CIRP Ann Manuf Technol 60(1):45–48
IEA (2007) Tracking industrial, energy efficiency and CO2 emissions. http://www.iea.org/
Johansson BJ, Skoogh A, Mani M, Leong SK (2009) Discrete event simulation as requirements
specification for sustainable manufacturing systems design. In: Proceedings of the PerMIS’09
conference, Gaithersburg
Le CV, Gan O, Chee XM, Zhang DH, Luo M, Chan HL, Lewis FL (2011) Intelligent energy audit
and machine management for energy-efficient manufacturing. In: Proceedings of the IEEE 5th
international conference on cybernetics and intelligent systems (CIS), pp 142–147
Mori M, Fujishima M, Inamasu Y, Oda Y (2011) A study on energy efficiency improvement for
machine tools. CIRP Ann Manuf Technol 60(1):145–148
Mouzon G, Yildirim MB, Twomey J (2007) Operational methods for minimization of energy
consumption of manufacturing equipment. Int J Prod Res 45(18):4247–4271
Rahimifard S, Seowa Y, Childs T (2010) Minimizing embodied product energy to support energy
efficient manufacturing. CIRP Ann Manuf Technol 59(1):25–28
Ramírez CA, Patel M, Blok K (2005) The non-energy intensive manufacturing sector: an energy
analysis relating to the Netherlands. Energy 30:749–767
Solding P, Petku D (2005) Applying energy aspects on simulation of energy-intensive production
systems. In: Proceedings of the 37th conference on winter simulation, pp 1428–1432
Vijayaraghavana A, Dornfeld D (2010) Automated energy monitoring of machine tools. CIRP
Ann Manuf Technol 59(1):21–24
Weinert N, Chiotellis S, Seliger G (2011) Methodology for planning and operating energy-
efficient production systems. Ann CIRP 60(1):41–44
Chapter 113
Analysis on System Archetype of High
College and University Competitiveness
Based on Hierarchical Structure
Li-qing Li and Ying-ting Yu
Abstract The paper combined hierarchical structure with system archetype
analysis technology of Peter M. Senge, with the research on the spot in the con-
crete high colleges and universities and systematical analysis, proposed the key
influencing factors of high college and university competitiveness and analyzed
the hierarchical relation of them, then constructed the key variable feedback
system archetypes based on the influencing factors and analyzed it systematically
and qualitatively, at last put forward corresponding management countermeasure,
which has certain theoretical and practical significance to analyzed the problem of
improving the competitiveness of high college and university.
Keywords Hierarchical structure Competitiveness of high college and univer-
sity System archetype
113.1 Introduction
Systems archetypes analysis was an effective method that can grasp the structure
of system. Masters of modern management Senge (1992) built 9 system archetypes
in the book of the Fifth Discipline—the Art & Practice of the Learning Organi-
zation. Regarded the system archetype as a key tool to analyze the issue of
organization and management, and made it to be the core content of the learning
organization theory, but in the book, he did not discuss how to build systems
archetypes. To improve the competitiveness of high college and university is a
system engineering project. Based on the method of system archetype, the authors
analyzed the issue of how to enhance the competitiveness of high college and
L. Li (&) Y. Yu
Jiangxi Science & Technology Normal University, Nanchang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1065
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_113,
Ó Springer-Verlag Berlin Heidelberg 2013
1066 L. Li and Y. Yu
university, and provided an effective method to research the complex problem in
the management with the analytic hierarchy principle.
113.2 Analyze the Influencing Factors of High College
and University Competitiveness Based
on Hierarchy Structure
In the twenty-first century, competition became an unavoidable practical problem
of many social organizations. With the evolution of the overall external envi-
ronment of higher education, the problem of competition that ignored for a long
time between high colleges and universities is gradually appearing. Under the new
situation of economic and technological globalization, China will integrate with
the world in educational development deeply, which will bring both opportunities,
but also have to face great challenges. There are different competitive advantages
in different industries (Zhu and Liu 2007). To high college and university, its
competitiveness is the organic integration and harmonization of various factors,
resources and capabilities, a comprehensive capacity that promote comprehensive,
healthy and orderly development of high college and university. This capability is
the result of an effective integration and interaction of multiple factors.
According to in-depth interviews, literature reading and system analysis on the
specific high college and university, the paper identified the main influencing factors
of competitiveness: school reputation, image of school, scientific research, teaching
faculty, social resources, the number of students, among them the most directly
factor is school reputation, and school reputation is influenced by school image,
scientific research, teaching faculty, social resources, student number. Therefore, the
competitiveness of high college and university is expressed by C = F(S), S = f(s1,
s2, s3, s4), S denotes as school reputation, s1, s2, s3, s4 denotes as image of school,
scientific research, teaching staff, social resources, the number of students. The
hierarchy structure of influencing factors of as showed as Chart 113.1.
Competitivenes
School reputation
School Scientific research Social Teaching Student
image & development resources faculty number
Chart 113.1 Hierarchy structure of influenced factors of high college and university
competitiveness
113 Analysis on System Archetype of High College 1067
113.3 Build System Archetype of High College
and University Competitiveness Based
on Hierarchy Structure
From the Chart 113.1, top-layer variable is the competitiveness of high college and
university, the first layer variable is school reputation that affects the top-layer
variable directly (Jia and Ding 2002), the causal structure is positive feedback,
v1 v2 v1, the variables in the second layer are key factors to affect the
school reputation which affected by other factors that include some positive factors
and some negative factors, the former improve the competitiveness, the latter
restrict the development of high college and university competitiveness. The paper
built the feedback system archetypes based on the second layer variables, analyzed
its complex relationship, and propose the effective countermeasures to enhance the
high college and university competitiveness.
113.3.1 Limits to Growth Feedback Archetype
Based on School Image-Oriented
To high college and university, image of school is an invisible business card, at the
meanwhile (Sun and Liu 2010), it also is the valuable educational resources, its
essence of image of school is the high college and university objective reality
external performance, however, improve the perfect school image, some colleges
pour large money on school size expansion or new campus construction. In fact, in
that way, it isn’t useful to add luster to school image, but also may be caught in
huge debt, which has brought a series of negative impact to high college and
university. Therefore, through systematic analysis, the paper built feedback
archetype based on the school image-oriented showed as Chart 113.2.
hardware facilities investment plan based on own
+ development needs
-1
+school image
new+ campus expasion
competitiveness +
+1
+ +
school reputation capital investment
-2
- +
debt ration + amount of loan -
-
-3 research award education fund
research & -4
development +
discipline construction
+
+ achievement + investment
+ key disciplines
Chart 113.2 Limits to growth feedback archetype based on school image-oriented
1068 L. Li and Y. Yu
From Chart 113.2, the feedback archetype based on the school image-oriented
is composed of a positive feedback and four negative feedbacks, in the chart, the
left feedback loop is a positive feedback loop that promote system development,
the right feedback loop is a negative feedback loop that strict system development.
The positive feedback loop reveals the mutual promotion between the school
image, school reputation, and school competitiveness, however, in order to
improve school image or reputation, some high college or university spend most
large money on expanding new campus, regardless of the school affordability of a
large number of loans. The results of a heavy debt ratio seriously affect school
teaching, research and the normal operation of the subject building work. All of
these lead a direct damage to school reputation and school image, and have a bad
affect on the promotion of competitiveness.
Therefore, in order to effectively promote the healthy and orderly development
of high college and university, each school must think over and over before
investment, blindly to follow the example is not conducive to the enhancement of
competitiveness.
113.3.2 Shifting Burden Feedback Archetype Based
on Scientific Research & Development
Achievements-Oriented
With introducing competition into the education, scientific research & develop-
ment becomes more and more important to improve the teaching quality, school
reputation and social influence (Li 2011). Colleges and universities which used
been a single place to impart knowledge turned into the base of production and
innovation and became an important pillar of national economic and technological
development. First prize winner of the State Technological Invention Award,
President of Central South University, academician Boyun Huang said: ‘‘in
addition to teach the most advanced knowledge, the more important function of
higher education is creating new knowledge.’’ Therefore, scientific research &
development becomes the most important evaluation index of measuring the
colleges and universities competitiveness.
More and more high colleges and universities lay emphasis on scientific
research & development. It directly influence academic standards, it also repre-
sents the school’s overall strength and competitiveness. However, research &
development achievements become a huge invisible pressure to high college and
university teachers. Although the moderate pressure of research can not only make
teachers concentrate on scientific research, maintain a strong scientific research
ambition, but also enhance teachers’ level of business as soon as possible, too
much scientific research pressures also do harm to teachers. As it is showed on
Chart 113.3, shifting burden feedback archetype based on research & development
pressure-oriented, revealed universities did not correctly handle the question of
113 Analysis on System Archetype of High College 1069
Chart 113.3 Shifting academic corruption
burden feedback archetype activities:rob awakened, symptoms solution
based on research & plagiarism and etc +
development pressure-
- -1 +
oriented
+2
research pressure
academic corruption
+
- -2
cultivation of internal -
research team
fundamental solution
competitiveness
+ +
+1 + achievement
research
+
school reputation
research & development pressure. In Chart 113.3, positive feedback 1 shows
promotion function between scientific research achievements and competitiveness,
positive feedback 2, negative feedback 1, negative feedback 2 is a shifting burden
archetype.
The numbers of paper and research project become the key index to measure
the level of scientific research, school rankings and teacher promotion. Academic
corruption becomes more and more serious under this background, for example,
plagiarize others’ paper after modifying it, one paper but multiple submissions, or
publish other’s paper after translating it into Chinese paper for its own purposes.
The management countermeasure for shifting burden feedback archetype is that
focus on the fundamental solution. However, research labor is spiritual production,
long cycle and slowly effective. To high college, effective way to improve sci-
entific research & development is that constructing research team, encouraging
college teachers to improve the quality of the research and enhance research
competitiveness.
113.3.3 Vicious Competition Feedback Archetype Based
on Teaching Faculty-Oriented
‘‘Teacher is essential for that a school whether can be able to train qualified
personnel to the socialist construction or not.’’ said Deng Xiaopeng (Li and Xia
2011). To a high college or university, teacher is the foundation to cultivate
talented person, it is the root to form school characteristics and advantages, and it
also is the pledge to keep sustainable development. Some school according to the
own development needs, spend large money on attracting talents to keep up with
the pace of development of the schools, but some high colleges blindly follow,
regardless of their own needs whether it is reasonable. This is a typical vicious
competition, as shown as in Chart 113.4.
1070 L. Li and Y. Yu
+
school B's threateness
school A's development
from school A
+ +activity by school
- B:investment in talent
school A's teaching faculty
-
+
school A's threateness
from school B +
activity by school + school B's development
A:investment in +
talent +
school B's teaching faculty
Chart 113.4 Vicious competition feedback archetype based on teaching faculty- oriented
From Chart 113.4, we can see clearly that two universities each see their
welfare as depending on a relative advantage over the other. Whenever one side
gets ahead, the other is more threatened, leading it to act more aggressively
reestablish its advantage, which threatens the first, increasing its aggressiveness,
and so on, each side sees its own aggressive behavior as a defensive response to
the other’s aggression, but each side acting ‘‘in defense’’ results in a buildup that
goes far beyond either side’s desires.
The vicious competition feedback archetype management principle is that to
look for a way for both sides to ‘‘win’’, or to achieve their objectives. In many
instances, one side can unilaterally reverse the vicious spiral by taking overtly
aggressive ‘‘peaceful’’ actions that cause the other to feel less threatened.
113.3.4 Success to the Successful Feedback Archetype
Based on Social Resources-Oriented
Development of social resources is the internal need of high colleges and univer-
sities, and the only way to improve educational level and quality. Due to the
scarcity of resources, a fierce competition for the limited social resources is showed
between the different colleges and universities. The more successful one becomes,
the more support it gains, thereby starving the other, as show as Chart 113.5.
Therefore, the management principle is looking for the overarching goal for bal-
anced achievement of both choices. In some cases, break or weaken the coupling
between the two, so that they do not compete for the same limited resource.
113.3.5 Limited to Growth Feedback Archetype
Based on the Number of Students-Led
Student is the lifeblood for university normal operation, affected by the popular-
ization of higher education, the competitiveness of snatching students becomes
113 Analysis on System Archetype of High College 1071
school A's reputation
+
school A's social resources
+
+
govement support
school A's competitiveness allocation to A instead
of B
- - B's social resources
school
school B's competitiveness
+
+
school B's reputation
Chart 113.5 Success to the successful feedback archetype based on social resources–oriented
more and more intense. However, different universities have different capacity for
students. Reasonable student number can promote health and orderly development,
too many students do may be harm to the sustainable development of the uni-
versity, as show as Chart 113.6, in the left of the chart, positive feedback causal
relationship reveal student number play an important role in improving the school
reputation and competitiveness; on contrary, in the right of the chart, the negative
feedback reveals too many students isn’t conducive to the improvement of the
university.
+ teaching task
+ workload
the needs of campus +
difficulty of management
+student number
+ dormitary -
+the needs of new
research investment
+ campus expansion
competitiveness + +
+ school reputation
- capital investment internal crisis -
+
amount of loan -
teaching quality +
+ research achievement
Chart 113.6 Limited to growth feedback archetype based on student number-oriented
1072 L. Li and Y. Yu
First of all, too many students lead to more dormitories and campus expansion
which force many schools to pour into blindly, and lead to loan burden. Secondly,
too many students lead to difficulty of management, which result in more and more
internal crisis (Li and Jiang 2011). Finally, too many students will bring more
teaching load for teachers. Due to everyone’s energy is limited, too much teaching
task is bound to teachers may be difficult in being focus on research. Therefore,
high colleges and universities have to be consistent with their own reasonable
capacity in enrollment.
113.4 Conclusion
With the research on the spot in the concrete high colleges and universities and
systematical analysis, the paper proposed the key influencing factors of high
college and university competitiveness, combined hierarchical structure with
system archetype analysis technology of Peter M.Senge, constructed the key
variable feedback system archetypes based on the influencing factors, at last put
forward corresponding management countermeasure, which has theoretical and
practical significance for improving high college and university competitiveness.
References
Senge PM (1992) The fifth discipline the art & practice of the learning organization. Century
Business Publishing House, London
Zhu J, Liu Z (2007) The evaluation index system of high college and university competitiveness.
Prod Forces Res 02:67–68
Jia R, Ding R (2002) System dynamics—complex analysis of feedback dynamics. Higher
Education Press, Beijing
Sun J, Liu Z (2010) The pressure of research influenced the university teachers. Teachers 5:28–29
Li L (2011) Analysis on the evaluation of hospital competitiveness, countermeasures producing
and the effect implementation simulation—take three A-level comprehensive hospital in
Jiangxi province as an example. Economic Science Press, Beijing
Li L, Xia L (2011) Analysis on the influencing factors and countermeasures of high college and
university competitiveness based on feedback archetype generating set. In: 2011 IEEE 18th
international conference on industrial engineering and engineering management
(IE&EM2011)
Li L, Jiang M (2011) Analysis on key variable-oriented typical archetype of high college and
university competitiveness based on system dynamics. In: 2011 IEEE 18th international
conference on industrial engineering and engineering management (IE&EM2011)
Chapter 114
Analysis on the Operation Effects
of Logistics Park Based on BP Neural
Network
Jun Luo
Abstract With the rapid development of the construction and operation of
Logistics Park, the operation effects and the development level of Logistics Park
will become the focus of attention. In this paper, the factors affecting the operation
effects of logistics park are proposed firstly. Then a set of evaluation metrics of the
operation effects of logistics park is given. Besides, based on BP neural network, a
model for calculating the operation effects is built. Finally, a case study has been
studied with the model.
Keywords BP neural network Logistics park Operation effects
114.1 Introduction
With the development of logistics, logistics park has become a kind of emerging
logistics management way. In Japan, Germany and other developed countries
logistics park has developed rapidly. The construction of logistics park of our
country began in Shen Zhen city in 1990s, and other cities also began the con-
struction of logistics park rapidly. By 2008 in September, according to statistics,
there had been 475 logistics parks in our country. Among them, 122 logistics parks
have operated already, 219 were under construction, and 134 were being planned
(China Federation of Logistics & Purchasing).
In the western developed countries, the rate of return on investment of logistic
park is about 6–8 %. The income gained by the investors of logistics park usually
comes from the return of rental and the land appreciation. In China, the vacancy
rate of logistics park is more than 60 %, and even some logistics park is used for
J. Luo (&)
School of Economics and Management, Wuyi University, Jiangmen, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1073
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_114,
Springer-Verlag Berlin Heidelberg 2013
1074 J. Luo
other purposes. With the rapid development of the construction and operation of
logistics park, the operation effects and the development level of logistics park will
become the focus of attention. In this paper, a set of evaluation metrics of the
operation effects of logistics park will be built, and the method of BP neural
network will be used for analysis of the operation effects of logistics park.
114.2 The Set of Evaluation Metrics
114.2.1 The Factors Affecting the Operation Effects
of Logistics Park
In the guidance of the government planning, logistics park is the large place in which
several sorts of modern logistics facilities and several logistics organizations layout.
Through sharing the infrastructure and the supporting service facilities, logistics
park can give full play to its whole advantage and complementary advantages.
The intensification and scale of logistics can promote the sustainable development of
the city (Zhang 2004; Richardson Helen 2002; Marian 2006). When we plan the
logistics park, we should consider regional economic level, customer industry,
distribution of retail industry and the entire functional orientation.
The factors affecting the operation effects of logistics park have two aspects
which are external factors and internal factors. The external factors include gov-
ernment’s support and relative policy, the economic situation and the market
environment. The internal factors mainly refers to the own operation ability of
logistics park.
The policy environment mainly reflects the government’s support to the
development of logistics park. The local governments have provided some policies
to logistics park, but these policies are not comprehensive and perfect. On the
whole, in our country, the policy support to the development of logistics park is
not enough. Soon, this situation will be changed, as we pay more and more
attention to logistics and promulgate some related policy. The market demand of
logistics park mainly includes target market’s service demand, the adaptability of
logistics park service and the matching degree of supply and demand of logistics
park service. These factors directly affect the operation effects of logistics park.
The service ability of logistics park its own mainly includes: transportation,
warehousing, distribution, packing and sorting, circulation processing, market
development and maintenance, informatization level and management ability, etc.
The internal factors are the foundation of the operation effects of logistics park.
114 Analysis on the Operation Effects of Logistics Park 1075
114.2.2 The Set of Evaluation Metrics of the Operation
Effects of Logistics Park
According to the influencing factors and the set of metrics of related documents
(Mingming 2010; Dai 2010; Zhong 2009), a set of evaluation metrics of the
operation effects of logistics park is built, which includes the economic benefits,
the condition of the enterprises in the park, park ability and social benefits four
parts, as is shown in Table 114.1.
114.3 The Operation Effects of Logistics Park Evaluation
Model Based on BP Neural Network
In 1985, the BP neural network model was brought out by D. Rumelhart from
Stanford University. As BP neural network can solve the nonlinear problem well, it
has become one of the most widely applied neural networks. BP algorithm solves
the connection weight problem of the hidden layer in the multi-lever network
model, improves the learning and memory function of neural, and especially solves
the XOR problem. The BP neural network model is the prior to connection model
that constitutes of input layer, output layer and some hidden layer (Yin 2003; Liu
and Lu 2011; Hagan et al. 2002).
Table 114.1 The set of evaluation metrics
First level index Second level index
Economic benefits Return on capital employed(X1)
Debt-to-asset ratio(X2)
Asset maintaining and increase ratio(X3)
The condition of the enterprises in the Number of enterprises in the park (X4)
park Gross asset of the enterprises in the park(X5)
Annual gross income of the enterprises in the
park(X6)
Satisfaction degree of the enterprises in the park(X7)
Loyalty of the enterprises in the park(X8)
Park ability Storage area(X9)
Annual freight Volume(X10)
Delivery capacity(X11)
Processing capacity(X12)
Estate service capacity(X13)
Informatization level(X14)
Goods damage rate(X15)
Social benefits Number of new employment(X16)
Influence on the urban traffic(X17)
Full Load Rate(X18)
Energy saving and emission reduction(X19)
1076 J. Luo
The BP neural network can also deal with qualitative and quantitative knowl-
edge. It’s operation is very fast, and has strong learning and forecast ability.
Therefore this paper using BP neural network model evaluates the operation
effects of logistics park. The specific procedure is as follows:
(1) The number of neurons: In this paper, the BP neural network will use three
layer structures, namely, input layer, hidden layer and output layer.
a. Input layer node: The number of input layer node is the set of evaluation
metrics. There are 19 input nodes.
b. Hidden layer node: The number of hidden layer node is related to the
number of input layer node, the character of sample data and the character
of the unsolved problem. To determine the number of hidden layer node,
pffiffiffiffiffiffiffiffiffiffiffiffi
we usually use the experience formula: q ¼ n þ m þ a. Among them, n is
the number of input layer node, m is the number of output layer node,
a = 1, 2,…, 10. Through several tests, 10 is the optimal number of hidden
layer node.
c. Output layer node: The results of the evaluation are output layer node.
According the analysis, the number of input layer node is 19; the number of
hidden layer node is 10; and the number of output layer node is 1.
(2) The initialization of weight value and threshold value: According to the set of
metrics, the index is divided into two kinds of indexes, namely, qualitative
index and quantitative index. Dealing with qualitative index, we generally use
the expert scoring method, and for the quantitative index we use normalized
processing. Generally, weight value and threshold value of initialization is the
random number from -1 to 1.
(3) The positive information transmission: In this paper, we use sigmoid function
to process network transmission, and purelin function to process transmission
is in output layer. After confirming the number of each layer node and
transmission function, we initialize the BP network again.
P
The output vector of hidden layer is y ¼ f1 wij xi þ ai , the output vector of
P
output layer is y ¼ f2 wjk yj þ aj .
(4) The reverse error transmission: Calculation the error E of network. If error
E is less than the previously set error e, the network training process is over,
and the output value is approximated the expected value. Otherwise we pro-
ceed the reverse error transmission of output layer and hidden layer node.
(5) Confirm
P the final evaluation results. Calculating the global error function
E¼ ek , if E \ e the training process is over. According to the final output
value results, the greater the output value, the better the operation effects of
logistics park. From very good to very bad, the output value is divided to six
levels, very good (0.9–1), good (0.8–0.9), preferably (0.6–0.8), general
(0.4–0.6), bad (0.2–0.4), very bad (0–0.2).
114 Analysis on the Operation Effects of Logistics Park 1077
114.4 The Model Training and Testing
According to the set of evaluation metrics and the BP neural network theory, we
establish the model steps. Using the initial, training and simulation functions of
Matlab7 neural network tool box, it can quickly complete the network training
process.
(6) Selection of sample data: Let the 19 indexes of the set of metrics as the input
node, the simulation data of the set of metrics of front five logistics park
W1–W5 as the training sample, and the back three logistics park W6–W8 as the
testing. Normalizing the input data, the input data is shown as Table 114.2.
(7) Determination of network structure: The number of input layer node is 19, the
number of hidden layer node is 10, and the number of output layer node is 1.
The network structure figure is shown as Fig. 114.1. The transmission function
of hidden layer node is sigmoid, and the transmission function of output layer
node is purelin.
(8) Model training: The training time is 265, target error is 0.001, learning rate is
0.01, using Matlab to calculate the algorithm.
After 800 times training, the network overall error is in the range of tar-
get allowable error. The prediction error figure is shown as Table 114.3. The
training is over.
Table 114.2 Normalization input datas
Index W1 W2 W3 W4 W5 W6 W7 W8
X1 0.046 0.035 0.058 0.035 0.044 0.040 0.031 0.052
X2 0.043 0.037 0.055 0.037 0.041 0.026 0.036 0.040
X3 0.038 0.033 0.052 0.029 0.049 0.036 0.037 0.049
X4 0.037 0.023 0.048 0.032 0.033 0.041 0.042 0.033
X5 0.033 0.030 0.050 0.030 0.036 0.046 0.046 0.026
X6 0.030 0.026 0.042 0.029 0.042 0.048 0.044 0.022
X7 0.045 0.032 0.047 0.031 0.037 0.039 0.053 0.048
X8 0.044 0.034 0.049 0.033 0.040 0.051 0.058 0.046
X9 0.085 0.092 0.067 0.111 0.078 0.056 0.099 0.088
X10 0.075 0.072 0.055 0.094 0.069 0.053 0.082 0.086
X11 0.068 0.085 0.056 0.087 0.062 0.062 0.076 0.079
X12 0.071 0.078 0.057 0.094 0.065 0.061 0.065 0.070
X13 0.074 0.081 0.055 0.105 0.059 0.067 0.063 0.092
X14 0.068 0.075 0.060 0.092 0.071 0.072 0.061 0.093
X15 0.060 0.066 0.058 0.084 0.063 0.078 0.059 0.085
X16 0.050 0.055 0.043 0.020 0.049 0.060 0.031 0.014
X17 0.049 0.049 0.046 0.016 0.046 0.053 0.033 0.012
X18 0.051 0.050 0.050 0.030 0.059 0.069 0.044 0.038
X19 0.033 0.047 0.052 0.011 0.057 0.042 0.040 0.027
1078 J. Luo
. .
. .
. .
Input layer Hidden layer Output layer
Fig. 114.1 The network structure figure
Table 114.3 Forecast error table
Sample W1 W2 W3 W4 W5
Expect output 0.613 0.411 0.513 0.783 0.535
Network forecast 0.607 0.395 0.509 0.788 0.533
Forecast error 0.006 0.016 0.004 -0.005 0.002
(9) Model testing: Using the network which has already been trained, we can get
network output value of the three samples. The network output value of W6 is
0.488, W7 is 0.752, W8 is 0.613. The network output of W7 is the better one.
114.5 Conclusion
In this paper, we have studied the influence factor of the operation effects of logistics
park, and establish the set of evaluation metrics of the operation effects of logistics
park. Through the BP neural network model, the operation effects of logistics park
have been analyzed, then the manager of the logistics park can find the shortage of the
operation process, and furthermore, can improve the operation of logistics park.
114 Analysis on the Operation Effects of Logistics Park 1079
References
China Federation of Logistics & Purchasing. China society of logistics. The second national
logistics park (base) survey report. http://b2b.toocle.com
Dai H (2010) Study on the operating model of Logistics Park based on the game theory. Wuhan
University of Technology, Wuhan
Hagan MT, Demuth HB, Beale M (2002) Neural network design. China Machine Press, Beijing
Liu H, Lu H (2011) Study on risk evaluation for manufacturers’ lean supply chain collaboration
based on BP neural network. Logist Technol 30(3):103–105
Marian S (2006) Logistics park development in Slovak Republi. Transport ll(3):197–200
Mingming Ni (2010) Research on the operating conditions of Logistics Park based on fuzzy
comprehensive evaluation. Value Eng 10:27–28
Richardson Helen L (2002) 3PL today: a story of changing relationships. Transp Distrib
43(9):38–40
Yin N (2003) The application design of BP neural network. Inform Technol 27(6):18–20
Zhang X (2004) The research on layout planning of logistics park, vol 6. China Supplies Press,
Beijing
Zhong J (2009) Construction of evaluation index system on economic operation of multi-service
Logistics Park. Logist Eng Manag 31:7
Chapter 115
Application of Ant Colony Algorithm
on Secondary Cooling Optimization
of Continuous Slab
Ji-yun Li and Hong-xing Pei
Abstract Continuous casting secondary cooling water is one of the key factors to the
quality of slab. Reasonable surface temperature maximum rate of cooling and surface
temperature rise speed of every secondary cooling stage can reduce the factors that
causing inside and surface crack of the slabs. The optimized model of continuous
casting secondary cooling was established according to metallurgy criterion (include
Goal surface temperature, Straightening spot temperature, surface temperature
maximum rate of cooling, surface temperature rise speed, fluid core length etc.) and
equipment constraints request. The water of continuous casting secondary cooling
was optimized by ant group algorithm to improve the quality of slabs.
Keywords Ant colony algorithm Continuous casting Optimization Water
assignment of secondary cooling
115.1 Introduction
The continuous casting of steel solidification technology is important innovation,
Second cooling is continue strengthening cooling to slab that out from mould to
accelerate the cooling process of slab (Laitinen et al. 2003). Through the
improvement of secondary cooling system, optimize the secondary cooling water,
can realize the uniform cooling slab, then get higher inner and surface quality of
the casting slab. Therefore, the optimization to second cooling is one of the
J. Li (&)
Department of Information Engineering, Henan Polytechnic,
Zhenzhou, China
e-mail: [email protected]
H. Pei
Physical Engineering College, Zhengzhou University,
Zhengzhou, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1081
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_115,
Springer-Verlag Berlin Heidelberg 2013
1082 J. Li and H. Pei
important measures of efficient continuous casting technology. Highly effective
casting has become important technologies to optimize steel industry structure of
our country (Natarajan and El-Kaddah 2004).
115.2 System Optimization Model
Slab, in the continuous casting process, under certain assumptions (Ying et al.
2006), ignores heat slab width direction, can be simplified as one-dimensional heat
transfer, the solidification and heat transfer equation (Radovic and Lalovic 2005).
oT o2 T
qC ¼k 2
ot ox
where: q is density of every phase of steel, kg/m3, C is specific heat capacity of
every phase of steel, J/(kg•K), k is thermal conductivity of every phase of steel,
W(mK).
Slab along the direction of the casting speed is separated into (0 to n) cross-
sections. Get the integral of each different spatial position in a unit on heat transfer
partial. The result is the ordinary differential equations relative to temperature T on
the time derivative. Derivative that temperature on the time calculated through
chase method to ordinary differential equations. By the time derivative of the
temperature on the slab surface temperature can be obtained (Lotov et al. 2005).
Secondary cooling system determined by continues casting metallurgical cri-
teria, device constraints, heat transfer model of different kinds of steel type. The
aims that integrate optimization secondary cooling system are to let the temper-
ature distribution of slab rationalization. This can get the best slab quality and
yield. The optimization method is: The value of objective function constructed by
metallurgical criteria is Minimum. Assuming second cooling paragraphs water and
converted into integrated heat transfer coefficient under constraint of industrial
conditions. Substitute these into simulation model of heat transfer calculation as
third boundary condition. To obtain transfer function distribution of second
cooling zone that meet the various metallurgical criteria and determine the dis-
tribution of secondary cooling water quantity.
Optimization model of the system expressed by M. Control vector k = [k1,k2,
…,kn]T. Where, n is the number of segments of cooling water section. Optimi-
zation model determined by metallurgical criterion and equipment constraints
(Bergh and Engelbrecht 2006). Optimal control parameters based on the perfor-
mance guideline that establishment of comprehensive evaluation on object func-
tion. And the object function must optimize according to certain rules.
Optimization model in the derivation use the following symbols:
f f [0
f ¼
0 f 0
115 Application of Ant Colony Algorithm on Secondary Cooling 1083
115.2.1 Optimal Model Determined by Metallurgical Criteria
1. Target temperature of surface
Tz, target temperature of slab surface, is determined by steel type and pro-
duction technology. The actual surface temperature T(h,z) should close enough
to target temperature of slab surface.
2
J1 ¼ Tðh;zÞ Tz ð115:1Þ
2. Straightening point temperature
Control T(tc,h), surface temperature of slab, at straightening point above the Tc,
brittle temperature (Santos et al. 2006), when strong cold. Avoid brittle ‘‘pocket
area’’.
( )2
J2 ¼ Tc Tðtc;hÞ ð115:2Þ
3. The maximum cooling and surface temperature rise rate
Control cool rate less then Td (C/m) to avoid the slab surface temperature in
the low ductility zone led to crack propagation. When slab out from model,
temperature rise rate must be controlled less than Tr (C/m) to prevent the
solidification front within the slab under the action of the tension cracks (Lan
et al. 2002)
2 2
oTðh;zÞ oTðh;zÞ
J3 ¼ Tr þ Td ð115:3Þ
oz oz
4 Liquid core lengths
Liquid hole in the slab must be completely solidified before straightening point
Td.
2
J4 ¼ ½ðLm Ld Þ ð115:4Þ
5 Bulging of slab
Bulging of slab solidification will cause front tensile stress, may lead to
breakage and segregation solidification front. Surface temperature T(h,z) of slab
should be controlled less than 1100 C (Gutjahr 2002) to prevent large bulging.
J5 ¼ f½ðTðh;zÞ 1100Þ g2 ð115:5Þ
115.2.2 Optimal Model Determined by Equipment Constraints
Casting speed and actual water of secondary cooling sections are in a certain range
in the production.
1084 J. Li and H. Pei
J6 ¼ f½ðVmin VÞ 2 þ ½ðV Vmax Þ 2 g ð115:6Þ
X
n
J7 ¼ f½ðxi ximax Þ 2 þ ½ðximin xi Þ 2 g ð115:7Þ
i¼1
Normalize the Formula 115.1–115.7 can obtain Optimization model of system.
Xn
Ji Jimin
J¼
J
i¼1 imax
Jimin
115.3 Ant Colony Algorithm
115.3.1 Principle of Ant Colony
Ants release pheromones in action. The information less volatilize in the shorter
path. The pheromone effect as a signal to the other’s actions, And then the
pheromone left by the original pheromone is enhanced by latecomers. The results
of continue cycle is that more ants visit the path, more probability of path chosen.
Within a certain period of time, the shorter path will be visited by the more ants.
Thus there will be more pheromone in the shorter path. More pheromone means a
shorter path, which means better answer (Gao and Yang 2006).
115.3.2 Algorithm Described and Procedures to Ant Colony
Optimization of continuous casting secondary cooling can be described as a
minimization problem. It means that find an optimal solution in feasible solution
set, so objective function M has a minimum value, m ants are randomly placed on
the structural map of the nodes. The ants move random according to the phero-
mone in path of the current point. The movement of ants is limited by the con-
straints w. The program of the arithmetic as follows:
N according to the first section of the secondary cooling zone to calculate the
actual water surface temperature
1. Calculate the surface temperature of the nth section in second cooling zone
according to actual water flow;
2. Initialize the parameter; Set the maximum number of iterations Nmax, and
number of ants K;
3. Construct solutions for the ants in accordance with the following formula
(Wang et al. 2008);
115 Application of Ant Colony Algorithm on Secondary Cooling 1085
8 a b
>
< P½si;j ðtÞ ½gai;j ðtÞ b ; j 2 tabuk
Pki;j ðtÞ ¼ ½si;j ðtÞ ½gi;j ðtÞ
>
: j2tabuk
0 ; others
where tabuk is node set that ant k has scanned at point ck.
4. Update pheromone
To every ants that completed construct solution, the pheromone volatilize in
accordance with the following formula.
si;j ðt þ 1Þ ¼ ð1 qÞ si;j ðtÞ
where, q ð0; 1Þ, is evaporation coefficient of pheromone.
Let ^s S, is the best feasible solution found so far. Increase the pheromone that
^s0 path in accordance with the following formula.
si;j ðt þ 1Þ ¼ si;j ðtÞ þ q Ds
5. Let smin[ 0 is the minimum pheromone value requirements.
si;j ¼ max smin ; si;j ;
6. Repeat steps 3–5 until all the ants converge to one path or reached the maxi-
mum number of iterations Nmax.
7. Let n = n+1. Repeat steps 1–6 until the last section of the secondary cooling
zone.
115.4 Optimization Results and Analysis
The secondary cooling of continues slabs are optimized according to the actual
production equipment, process parameters, physical properties parameters of steel.
The restriction are straightening point temperature greater than 900 C, surface
cooling rate less than 200 C/m, surface temperature rise rate along casting less
than 100 C/m, metallurgy length 21.58 m. No-optimized and optimized surface
temperature of slab is as Fig. 115.1.
After optimization, the maximum cooling rate and surface temperature rise rate
are both lower than before. The maximum cooling rate is drop from 152 to 72 C/m.
The maximum surface temperature rise rate is drop from 34 to 12 C/m. The
surface temperature distribution is flat. These reduce the stress factors that induced
slab inner and surface cracking.
1086 J. Li and H. Pei
Fig. 115.1 No and 1200
optimized surface
)
temperature of slab 1150
Surface Temperature (
1100
No-optimized
1050
1000
950 Optimized
900
0 5 10 15 20 25
Distance to Meniscus (m)
115.5 Concluding Remarks
The optimized model of continuous casting secondary cooling was established
according to metallurgy criterion (include Goal surface temperature, Straightening
spot temperature, surface temperature maximum rate of cooling, surface temper-
ature rise speed, fluid core length etc.) and equipment constraints request. It is in
favor of continues casting real-time control that fast optimization to system model
using ant colony.
Use ant colony to optimize continues casting second cooling water flow. It is
tends to reasonable after optimization that surface temperature maximum rate of
cooling and surface temperature rise speed of every secondary cooling stage.
These reduce the stress factors that induced slab inner and surface cracking. It
meets the criteria for the improvement of metallurgical slab cooling process and
improves product quality requirements.
Acknowledgments The correspondent of paper is Hong-xing Pei. His research area is
Mechanical and Electronic Engineering, application of computer and automation.
References
Bergh van den F, Engelbrecht AP (2006) A study of particle swarm optimization particle
trajectories. Inf Sci 176(6):937–971
Gao S, Yang J (2006) Swarm intelligence algorithm and its application. China Water Resources
and Electric Press, Beijing
Gutjahr WJ (2002) ACO algorithms with guaranteed convergence to the optimal solution. Inf
Process Lett 82(3):145–153
Laitinen E, Lapinb AV, Piesk J (2003) Asynchronous domain decomposition methods for
continuous casting problem. J Comput Appl Math 154(2):393–413
Lan CW, Liu CC, Hsu CM (2002) An adaptive finite volume method for incompressible heat flow
problems in solidification. J Comput Phys 178:464–497
115 Application of Ant Colony Algorithm on Secondary Cooling 1087
Lotov AV, Kamenev GK, Berezkin VE (2005) Optimal control of cooling process in continuous
casting of steel using a visualization-based multi-criteria approach. Appl Math Model
29(7):653–672
Natarajan TT, El-Kaddah N (2004) Finite element analysis of electromagnetic and fluid flow
phenomena in rotary electromagnetic stirring of steel. Appl Math Model 28(1):47–61
Radovic Z, Lalovic M (2005) Numerical simulation of steel ingot solidification process. J Mater
Process Technol 160:156–159
Santos CA, Spim JA Jr, Maria CF et al (2006) The use of artificial intelligence technique for the
optimisation of process parameters used in the continuous casting of steel. Appl Math Model
26(11):1077–1092
Wang S, Gao L, Cui X (2008) Study on multi-depots vehicle routing problem and its ant colony
optimization. Syst Eng Theory Prac 2:143–147
Liu Y, Cao T, Xi A (2006) Control model for secondary cooling in continuous slab casting.
J Univ Sci Technol Beijing 28(3):290–293
Chapter 116
Application of the Catastrophe
Progression Method in Employment
Options for Beijing, Shanghai
and Guangzhou
Qun Yuan, Ting Chen and Yang Gao
Abstract Aimed at the choice problem of employment city among Beijing,
Shanghai and Guangzhou, working conditions, living conditions and other indi-
cators about employment in the various cities are studied in this paper. To get the
comprehensive evaluation score situation of the three cities, the catastrophe pro-
gression method was applied to establish the corresponding index system and
make a specific analysis and comprehensive evaluation. It provides a theoretical
basis to choose the employment of the three cities. The application of the catas-
trophe progression method is more appropriate and accurate than the fuzzy
mathematics method without giving a weight to the evaluation indicators.
Keywords Catastrophe progression method City system Employment
Normalization formula
116.1 Introduction
Beijing, Shanghai and Guangzhou are China’s three first-tier cities, there are
thriving and talent. From the economy, the culture to the standard of living, the three
cities are the elites of China’s cities. Thousands of people arrived here dreaming to
create their own place in the world. So which city is the right city to work in? All the
data in the article is rooted in the three cities ‘‘statistic almanac’’ (2011).
Q. Yuan (&) T. Chen Y. Gao
Kunming University of Science and Technology, Kunming, China
e-mail: [email protected]
T. Chen
e-mail: [email protected]
Y. Gao
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1089
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_116,
Springer-Verlag Berlin Heidelberg 2013
1090 Q. Yuan et al.
116.2 The Basic Idea of the Catastrophe Theory
and Evaluation Steps
116.2.1 The Basic Idea of the Catastrophe Theory
Catastrophe is a process that its stationary state suddenly and discontinuously
changes to a new steady state. Catastrophe theory is an emerging discipline of
catastrophe study created by the French mathematician R. Thom in 1972. It is a
mathematical theory based on singularity theory, topology and stability theory to
discuss the compliable relationship between those control variables and the
characteristics of system state variables. It is recognized as ‘‘a revolution in math
after calculus’’. The basic features of this theory is make an classification of the
critical points of the system based on a potential function of the system, study the
features of state changing nearby the classified critical points, then sum up a
number of elementary catastrophe models as the basis to explore the natural and
social catastrophic phenomena (Dou 1994). It has been a useful tool in studying in-
continuous phenomena since it was established. By means catastrophe progression
method that derived from the catastrophe mode of catastrophe theory can be
widely applied in multi-criteria decision problem.
The total evaluation index is divided by multi-hierarchical contradictory groups
at first, then with the catastrophe fuzzy subordinate function which is from the
combination of catastrophe and fuzzy mathematics, the total function value can be
got through recursion operation with the unitary formula, finally comprehensive
evaluation results obtained.
116.2.2 Evaluation Steps of Catastrophe Progression Method
(1) Firstly, establish the catastrophe assessment indexes system. According to the
purpose of the evaluation system, the total indicators is divided into multi-hier-
archical contradictory groups and the system arranged in tree structure of the level
of purpose, obtain the more concrete quantifying indexes. Some indexes may have
to be decomposed further. The decomposition will be stopped until the metered
indexes are obtained. But the number of control variables in catastrophe system
should not be more than 4, so the sub-indexes of a single index had better not be
more than 4 (Liang et al. 2008).
(2) Determine the index system of each levels catastrophe system model.
R. Thom had classified a special kind of mapping f: Rn ? R of singularities,
namely smooth mapping f: Rn ? Rd a r parameter family (for any limited n and
all r 2 4) locally equivalent to one of the seven kinds of structural stability family
of functions, called the seven elementary catastrophe (Zhang et al. 2009). And
there are several commonly used catastrophe systems: cusp catastrophe system,
swallowtail catastrophe system and butterfly catastrophe system. The feature of
116 Application of the Catastrophe Progression Method 1091
several basic catastrophe models and each model’s details is showed in
Table 116.1.
The f(x) is the potential function of the state variable x. The coefficient a, b, c, d
indicated the control variable of system in Table 116.1. State variable and control
variables of system potential function are two contradictory aspects, and their
relations are showed in Table 116.1. Usually, the primary control variable is write
in front and the secondary control variable in the back. If the total evaluation index
in this system is only divided into two sub-indexes, it is called cusp catastrophe
system. If the total evaluation index is divided into three sub-indexes or four sub-
indexes, the system can be regarded as swallowtail catastrophe system or butterfly
catastrophe system.
Deduce the normalization formula from bifurcation equation. By the equation
in Table 116.1, the following are the procedure: let the potential function of the
catastrophe system be f(x). According to catastrophe theory, the critical points of
the potential function form an equilibrium surface, The equation of the surface is
obtained by a first derivative f0 (x) = 0 and the singular point set of the balanced
curved surface can be obtained through a second derivative f00 (x) = 0. So by
f0 (x) = 0 and f00 (x) = 0, we can obtain the equation of the set of bifurcation points
of the catastrophe system by eliminate x in the equation (Li 2004).
The decompose difference equation reflects the relationship between the control
variables and the state variables is received by the state variables. The normali-
zation formula is derived from the decomposition forms of the equation of the
bifurcation point set. Through the normalization formula, the different states of the
control variables of the catastrophe system are transformed into states (Yu 2008).
In the normalization formula, the value range of x and those control variables
are form 0 to 1. and we must follow two principles in the calculation: comple-
mentary and non-complementary (He and Zhao 1985). The complementary prin-
ciple implies that the control variables complement each other so that each of them
tends to reach the average value, which is calculated by x = average (xa, xb, xc,
xd). And the non-complementary principle indicates that the control variables of a
system, such as a, b, c and d cannot make up each other’s shortage (Shi et al.
2003). Therefore, when finding the value of the state variable x using the nor-
malization formulas, the smallest of the state variable values corresponding to the
control variables is chosen as the state variable value of the whole system (Huang
2001).
(3) Implement comprehensive evaluation by the normalization formula. Based
on the theory of fuzzy multi-objective decision making, to the same solution, in a
variety of target, if put A1, A2,…,Am for fuzzy goal, the ideal strategy should be
C = A1\A2\…\Am, and its membership functions is l(x) = lA1(x)^lA2(x)^…
^lAm(x), which lAi(x) is the membership function of Ai (Wan et al. 2006).
Different program for C1,C2, …, Cm, the membership functions of the program
Ci can regard as lci(x). Because the initial mutations are series the bigger the better
type, so total membership functions are also the bigger the better. So if
lci(x) [ lcj(x), it means the program Ci is better than Cj. Then we can get the
1092
Table 116.1 Several basic catastrophe models
Type Cusp catastrophe system Swallowtail catastrophe system Butterfly catastrophe system
Model f ð xÞ ¼ x4 þ ax2 þ bx f ðxÞ ¼ 1=5x5 þ 1=3ax3 þ 1=2bx2 þ cx f ðxÞ ¼ 1=6x6 þ 1=4ax4 þ 1=3bx3 þ 1=2cx2 þ dx
Diagram
x x x
a b a b c a b c d
Variables a, b a, b, c a, b, c, d
The equation of bifurcation a ¼ 6x2 a ¼ 6x2 a ¼ 10x2
b ¼ 8x3 b ¼ 8x3 b ¼ 20x3
c ¼ 3x4 c ¼ 15x4
d ¼ 5x5
pffiffiffi pffiffiffi pffiffiffi
Normalization formula xa ¼ a xa ¼ a xa ¼ a
pffiffiffi pffiffiffi pffiffiffi
xb ¼ 3 b xb ¼ 3 b xb ¼ 3 b
p ffiffiffi p ffiffiffi
xc ¼ 4 c xc ¼ 4 c
p5
ffiffiffi
xd ¼ d
Q. Yuan et al.
116 Application of the Catastrophe Progression Method 1093
result of advantage and disadvantage ordering of every evaluate target by the score
of the total evaluation index (Liang et al. 2008).
116.3 Establishment of a Catastrophe Index System
According to the different levels and the application of analytic hierarchy process
and categories of the evaluation indicators, the establishment of assessing indi-
cators system for employment options is shown in Table 116.2. The original data
is adopted by ‘‘Statistic Almanac’’ (2011) to ensure the accuracy (Shanghai
Municipal Bureau of Statistics 2011; Guangzhou Municipal Bureau of Statistics
2011; Beijing Municipal Bureau of Statistics 2011).
Based on the requirement of catastrophe theory, the primary control variable
write in front and the secondary control variable write in the back. According the
divided of catastrophe system, from Table 116.2, the third-class indexes from top to
bottom are swallowtail catastrophe system, swallowtail catastrophe system, cusp
catastrophe system, butterfly catastrophe system, butterfly catastrophe system,
swallowtail catastrophe system, swallowtail catastrophe system, cusp catastrophe
system, swallowtail catastrophe system, cusp catastrophe system, swallowtail
catastrophe system, which consist of 32 indexes and record these as x1, x2,
x3,…,x32; The second-class are cusp catastrophe system, swallowtail catastrophe
system, cusp catastrophe system and butterfly catastrophe system respectively,
which consist of 11 indexes and record these as y1, y2, y3…y11; The first-class is
butterfly catastrophe system which consist of 4 indexes and record these as z1, z2,
z3, z4 (Gao et al. 2008).
116.4 The Process of Calculation
According to the requirements of the catastrophe progression method, at first
normalize the control variables of the catastrophe model. In this system, there is a
positive index (the larger the better) and negative index (the smaller the bet-
ter),among the indices x4, x6, x18, x19, x20, x22, x17, x21, x30, x31, x32 are negative
index and by using Eq. (1) form 0 to 1, the other indices are positive index and by
using Eq. (2) shape 0–1 (Liang et al. 2008). If the value of the control variable is
between 0 and 1, data processing is not needed (Yao et al. 2008).
They can be used directly to catastrophe progression calculation. Finally,
according to the normalization formula of each catastrophe program, we can
combine upward gradually of the indicators, until the highest assessment is real-
ized (Chen 2004).
1094 Q. Yuan et al.
Table 116.2 The indexes of employment of options
First-class index Second-class index Third-class index
Work Employment 1. New employment posts (million)
conditions opportunity 2. Urban units in average salary (yuan)
3. The employment agency success rate (%)
Employment 4. Labor dispute number (count)
security 5. Minimum wage for employees (yuan)
6. Unemployment rate (%)
Living Infrastructure 7. Public investment (billion yuan)
conditions 8. Municipal construction investment (billion yuan)
Environment 9. Good rate of environmental air quality (%)
10. Per capita green area (m2)
11. Green coverage rate (%)
12. The environmental protection investment proportion
of GDP(%)
Transportation 13. Per capita roads length (km)
14. Every per capita operation public vehicles number
15. Every per capita taxi operation vehicle number
16. Passenger volume (millions miles)
Living costs Housing 17. Lease contract dispute number (count)
18. The average selling prices of residential (yuan/square)
19. Average house lease price index
Consumption 20. Resident consumption level (yuan for each person)
21. Consumer spending as a percentage of disposable
income (%)
22. Consumer spending per person (yuan)
Livelihood Social security 23. Basic annuities minimum standards (yuan)
24. Minimum living standard of urban residents (yuan)
Health care 25. Per population doctor number
26. Per population number of beds
27. The fitness activity places number
Education 28. Regular institutions of education number
29. Every ten thousand people collect books
Urban security 30. The criminal case number of knowledge
31. A civil case number of knowledge
32. The traffic accident number
For The-Larger-The-Better indices:
xmax x
xi ¼ ð1Þ
xmax xmin
For The-Smaller-The-Better indices:
x xmin
xj ¼ ð2Þ
xmax xmin
116 Application of the Catastrophe Progression Method 1095
For example of Beijing, the calculating process is as follows:
(1) According to complementary catastrophe model, the second-class index of
employment opportunity gets the catastrophe function which is calculated by
normalization formula:
pffiffiffi p
3
ffiffiffi p4
ffiffiffi
Ux1 ¼ 0; Ux2 ¼ 1; Ux3 ¼ 1;
Average these results:
Uy1 ¼ ðUx1 þ Ux2 þ Ux3 Þ=3 ¼ 0:6667
Similarly:
Uy2 ¼ 0:5757 Uy3 ¼ 0:4523 Uy4 ¼ 0:75
Uy5 ¼ 0:9453 Uy6 ¼ 0:3245 Uy7 ¼ 1; Uy8 ¼ 0:5
Uy9 ¼ 0:9788 Uy10 ¼ 0:9182 Uy11 ¼ 0:3639
(2) According to complementary catastrophe model, the first-class index of the
working of city gets the catastrophe function which is calculated by normalization
formula:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi p3
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Uy1 ¼ 0:66671; Uy2 ¼ 0:5757; Uy3 ¼ 4 0:4523;
Average these results:
Uz1 ¼ Uy1 þ Uy2 þ Uy3 =3 ¼ 0:8242
Similarly: Uz2 ¼ 0:8557; Uz3 ¼ 0:7848; Uz4 ¼ 0:874;
(3) According to complementary catastrophe model, the total index of
employment opinions gets the catastrophe function which is calculated by nor-
malization formula:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi p
3
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Uz1 ¼ p0:8242
ffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi Uz2 ¼ p 0:8557;
ffiffiffiffiffiffiffiffiffiffiffi
Uz3 ¼ 4 0:7848 Uz4 ¼ 3 0:874
Average these results:
Uz ¼ ðUz1 þ Uz2 þ Uz3 þ Uz4 Þ=4 ¼ 0:945
Similarly, the ultimate catastrophe function of Shanghai and Guangzhou and the
catastrophe functions of the subsystems of other two cities can be realized (Tan et al.
1999). And the results of calculation are summarized in Tables 116.3 and 116.4.
116.5 The Results of Evaluate Indexes
As can be seen in the Table 116.3, the results of the optimization Beijing [ Gu-
angzhou [ Shanghai. As the capital of our country, Beijing’s total scores are the
1096 Q. Yuan et al.
Table 116.3 The results of Position 1. Beijing 2. Guangzhou 3. Shanghai
the optimization
Work conditions 0.8242 0.8751 0.69
Living conditions 0.8557 0.7173 0.8237
Living costs 0.7848 0.7487 0.7495
Livelihood 0.874 0.8549 0.6002
highest in the three cities. But its working is not as good as Guangzhou, Gu-
angzhou which is china’s leading opening up the city, and in 2010 succeeded in
held the 16th Asian games, success made many employment opportunities and
improve the quality of life. we can know the low standards of Shanghai is mainly
due to the town average salaries is lower and the unemployment rate is higher
compared with other city from the working of Table 116.4, and the inadequate of
livelihood is mainly due to doctors and beds per capita (Gao et al. 2005).
116.6 Conclusion
(1) Urban system evaluation is a multi-level, multi-index and multi-criteria
comprehensive evaluation problem. The inherent logic relation among indexes is
used in catastrophe progression instead of the weight factor, so that the evaluation
only need to consider the comparative importance of the indicators, avoiding the
deficiency of static evaluation methods and human subjectivity, thus the results
obtained are more quantitative and more subjective. The catastrophe function
progression method adopts the normalization formula which of the calculation is
small, so that making it easy to program and master. It is much more effectively
and precise than general fuzzy mathematics to solve multi-objective evaluation
problems for decision-making (Chen and Chen 2011).
Table 116.4 The low Number Shanghai Beijing Guangzhou
standards of Shanghai
Employment opportunity 0.3333 0.6667 0.7901
Employment security 0.5171 0.5757 0.639
Infrastructure 0.6667 0.4523 0.2974
Environment 0.5897 0.75 0.4656
Transportation 0.4431 0.9453 0.4779
Housing 0.4624 0.3245 0.8254
Consumption 0.5492 1 0.2042
Social security 0.407 0.5 0.6616
Health care 0 0.9788 0.8372
Education 0.5 0.9182 0.3458
Urban security 0.6667 0.3639 0.58
Total score 0.945 0.9325 0.9004
116 Application of the Catastrophe Progression Method 1097
(2) Through the evaluation analysis of three cities, establish the employment
options index system of three cities, and according to the basic principle of
catastrophe theory, carries on sorting to each influence factor by the importance of
its objectives, respectively for the first-class index and the second-class index, and
choose the third-class index able to represent the second-class index to calculate
and analysis, and carries on the normalized computation gradually upwards, finally
gets the evaluation results. It can be found the deficiencies of the cities from the
evaluation results of the urban employment and provide the reasonable reference
for the selection of employment city.
(3) Because the method can only handle within four control variables, it is not
suitable for more than four control variables decision problem. Therefore the total
indicators need to be divided into multi-hierarchical contradictory groups (Zhang
2009). There are various problems about score problem of the lowest indicators, all
levels of decomposition problem, order of importance between the indexes of the
same level and the complementary and the non complementary relationship
between the indexes and so on, the catastrophe progression method needs to be
researched and improved further.
References
Beijing Municipal Bureau of Statistics (2011) The national bureau of statistics survey office in
Beijing, Beijing statistics book 2011. China Statistics Press, Beijing
Chen ML (2004) The application of catastrophe model to comprehensive evaluation. J Univ Sci
Technol Suzhou (Nat Sci) 21(4):23–27
Chen JC, Chen ZN (2011) The application of catastrophe to China real estate industry
competitive power evaluation of the city. Decis Making 19(8):254–256 (In Chinese)
Dou XF (1994) The application of catastrophe theory in economic field. University of Electric
Science and Technology of China, Chengdu
Gao MS, Dou LM, Zhang N, Kan JG (2005) Cusp catastrophic model for instability of coal pillar
burst damage and analysis of its application. J China Univ Min Technol 4(34):432–437 (In
Chinese)
Gao K, Li M, Wu C (2008) Application of catastrophe progression method in forecasting
spontaneous combustion of blasted muck pile of sulfide ore. Met Mine 2(2):21–22
Guangzhou Municipal Bureau of Statistics (2011) Guangzhou statistics book 2011. China
Statistics Press, Beijing
He P, Zhao ZD (1985) Catastrophe theory and its application. Dalian University of Technology
Press, Dalian
Huang YL (2001) Application of catastrophe progression method to sustainable usage of water
resource. Arid Environ Monit 15(3):167–170
Li HW (2004) Application of the catastrophe progression method in evaluation index system of
eco-city. Environ Assess 2004(9):44–48
Liang GL, Xu WJ, He YZ, Zhao TX (2008a) Application of catastrophe progression method to
comprehensive judgment of slope stability. Rock Soil Mech 29(7):1895–1899
Liang GL, Xu WJ, He YZ, Zhao TX (2008b) Application of catastrophe progression method to
comprehensive judgment of slope stability. Rock Soil Mech 29(7):1895–1899 (In Chinese)
Shanghai Municipal Bureau of Statistics (2011) Shanghai statistics book 2011. China Statistics
Press, Beijing
1098 Q. Yuan et al.
Shi YQ, Liu YL, He JP (2003) Further study on some questions of catastrophe evaluation method.
Eng Diurnal Wuhan Univ 36(4):132–136
Tan YJ, Chen WY, Yi JX (1999) Principle of system engineering. National University of Defense
Technology Press, Changsha, pp 341–348
Wan WL, Yang CF, Wang DJ (2006) Application of catastrophe theory evaluation method in
assessment of economic profit and productivity of mine. Min Eng 4(2):5–7
Yao DQ, Guo XC, Tu SW (2008) The application of catastrophe progression method on the
decision-making planning alternatives for through highways. In: 2008 International confer-
ence on intelligent computation technology and automation, Changsha, CA
Yu L (2008) Create Chinese characteristics of urban evaluation system. China Dev 8(4):89–95
Zhang JX (2009) Livable city evaluation and countermeasures about Henan. Bus Econ (6):93–95
(In Chinese)
Zhang TJ, Ren SX, Li SG, Zhang TC, Xu HJ (2009) Application of the catastrophe progression
method in predicting coal and gas outburst. Min Sci Technol 28(4):431–434
Chapter 117
Cooperation Relationship Analysis
of Research Teams Based on Social
Network Analysis and Importance
Measures
Zheng-sheng Han and Zhi-qiang Cai
Abstract To overcome the disadvantages of qualitative methods in team
relationship analysis, we introduce an adjacency matrix to modeling the cooper-
ation relationships between team members and use the importance measures to
evaluate the effect of a member on the team and the performance of the whole
team. First of all, the weight adjacency matrix is developed based on adjacency
matrix to describe the research team. In the weight adjacency matrix, a node means
a member in the team, while the weight describes the cooperation relationship
between members. Then, the social network analysis and importance measures are
used to estimate the cooperation of members. Finally, the case study of a research
team from Northwestern Polytechnic University is implemented to show the
cooperation relationship results of the proposed methods.
Keywords Cooperation relationship Importance measure Research team
Social network analysis
117.1 Introduction
A team is a group of professionals working together for a common task, where a
variety of disciplines may be represented. Because of the stimulation of shared
knowledge and the advantage of a greater diversity of skills, teamwork offers
benefits to all team members to achieve higher reputations (Young 1998).
Z. Han (&) Z. Cai
School of Mechatronics, Northwestern Polytechnical University, Xi’an, China
e-mail: [email protected]
Z. Cai
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1099
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_117,
Ó Springer-Verlag Berlin Heidelberg 2013
1100 Z. Han and Z. Cai
In science, research used to be implemented by single researcher. Nowadays,
research team plays a very importance role in the student education and research
innovation. It will be imperative for researchers to break the mentality of a single
laboratory/single research focus and develop an interdisciplinary research team
aimed at addressing real world challenges (Dodson et al. 2010). However, there
are also some potential conflicts in research teams with poor relationships between
team members, especially in huge research teams. It is very interesting to analyze
the cooperation relationships of research teams to improve the performance of the
whole team.
For several years, researchers have been focused on the cooperation relation-
ships which affect potential benefits and costs of teams much. Ghobadi and
D’Ambra (2012) conceptualized and implemented the multi-dimensional construct
of cross-functional competition, and presented an instrument for measuring this
construct. Dekker et al. (2008) investigated whether members of virtual teams
from the U.S., India, and Belgium perceived the same cooperation behaviors to be
critical for team functioning as Dutch members. LePine et al. (2011) reviewed the
theoretical and empirical research on the role of personality in team cooperation,
and found that team member personality is associated with various aspects of team
functioning and effectiveness. Pagell and LePine (2002) reported a qualitative
study aimed at identifying factors in operational systems that influence team
effectiveness, including work organized around the team’s output, opportunities
for cooperation, novel problems to solve in work, and management trust in teams.
Salmi (2010) analyzed case studies in the context of industrial business networks
to explore the collaboration in international research teams and interest in business
between international customers and their suppliers. Adams et al. (2005) explored
recent trends in the size of scientific teams and in institutional collaborations
according to the data derived from 2.4 million scientific papers written in 110 top
U.S. research universities over the period 1981–1999.
However, most of the research works use the qualitative methods to describe
the cooperation relationships in teams and to reflect their functions in team works.
These results can provide little practicable suggestions for team leader in team
managements. In this paper, we introduce an adjacency matrix to modeling the
cooperation relationships between team members and use the social network
analysis and importance measures to evaluate the effect of a member on the whole
team.
The paper in all has 5 Sections and is organized as follows. In Sect. 117.2, the
adjacency matrix is introduced to model the cooperation relationships in research
team. The social network analysis and importance measures are applied to analyze
the cooperation relationship in research team in Sect. 117.3. In Sect. 117.4, a case
study is presented to show the implementation process of the proposed methods.
Finally, Sect. 117.5 concludes the contributions of this work.
117 Cooperation Relationship Analysis of Research Teams 1101
Fig. 117.1 Examples of adjacency matrix
117.2 Methodology
117.2.1 Adjacency Matrix
In mathematics and computer science, an adjacency matrix is a means of
representing which nodes of a graph are adjacent to which other nodes (Wikipedia
2012a). Specifically, the adjacency matrix of a finite graph G with n nodes is a
n n matrix. The non-diagonal variable aij may be 1 to represent an edge from
node i to node j. It may also be 0 to represent none edge from node i to node j. The
diagonal variable aii is always 0. There exists a unique adjacency matrix for each
isomorphism class of graphs, and it is not the adjacency matrix of any other
isomorphism class of graphs. If the graph is undirected, the adjacency matrix is
symmetric. Figure 117.1 shows some examples of adjacency matrix (Wikipedia
2012a).
117.2.2 Modeling Research Team with Weight
Adjacency Matrix
Because the traditional adjacency matrix can only represent the connectivity
between nodes, we present a new weight adjacency matrix (WAM) to describe the
characteristics of research team. It can evaluate the cooperation relationships
between team members quantitatively.
A WAM also represents which nodes of a graph are adjacent to which other
nodes. The diagonal variable aii is still 0. However, the non-diagonal variable
aij ; 1 aij 1 can represent an edge from node i to node j with weight. If aij ¼ 1,
it shows that node i and node j are connected with absolute positive relationship. If
0\aij \1, it shows that node i and node j are connected with part positive
1102 Z. Han and Z. Cai
1
0. 5 ⎡0 0 0 0.5 ⎤
⎢0 0 0 −1
⎥
2
-1
4 ⎢ ⎥
⎢0 0 0 1 ⎥
⎢ ⎥
⎣ 0. 5 − 1 1 0 ⎦
1
3
Fig. 117.2 An example of WAM
Member
1
0.5 -0.3
⎡ 0 0.5 0 −0.3 ⎤
⎢ 0.5 0 0.8 −1
⎥
Member -1 Member ⎢ ⎥
2
Research
4 ⎢ 0 0.8 0 1 ⎥
team ⎢ ⎥
1
⎣ −0.3 −1 1 0 ⎦
0.8
Member
3
Fig. 117.3 An example of WAM for research team
relationship. If aij ¼ 0, it shows that node i and node j are isolated with none
relationship. If aij ¼ 1, it shows that node i and node j are connected with
absolute negative relationship. Figure 117.2 shows an example of WAM.
Usually, a research team is represented with a WAM. In the WAM, a node i
means a member i in the team, while the weight aij describes the cooperation
relationship between member i and member j. The research team can also be
represented with the corresponding graph to get a more direct understanding.
Figure 117.3 shows an example of WAM for research team.
117.3 Analysis of Research Team
117.3.1 Social Network Analysis of Research Team
Social network analysis (SNA) is the mapping and measuring of relationships and
flows between people, groups, organizations, computers, URLs, and other
117 Cooperation Relationship Analysis of Research Teams 1103
connected information/knowledge entities. It provides both a visual and a math-
ematical analysis of human relationships (Krebs et al. 2012).
SNA has emerged as a key technique in modern sociology. It has also gained a
significant following in anthropology, biology, communication studies, economics,
geography, information science, organizational studies, social psychology, and
sociolinguistics (Wikipedia 2012b). There are some popular measures of node and
network in SNA, as follows.
Degree—the number of direct edges a node has. It means how many people this
person can reach directly.
Density—the ratio of the number of edges in the network over the total number
of possible edges among all pairs of nodes.
In the WAM of research team, the degree of member i is calculated as in
(117.1). The density of the whole research team is computed as in (117.2).
X
n
ai ¼ aij ð117:1Þ
j¼1
,
X
n
a¼ ai ðn ðn 1ÞÞ ð117:2Þ
i¼1
117.3.2 Importance Measures of Research Team
Components importance measures were first introduced by Birnbaum (1969). The
Birnbaum importance measure quantifies the contributions of individual compo-
nents to the system performance. Based on this achievement, a wide range of
importance measure definitions have been proposed in the engineering field. The
Fussell-Vesley importance measure was introduced by Vesley (1970) and used by
Fussell (1975) in the context of fault tree analysis which depends on the current
reliability of basic events.
Assuming that objective n-component system S under study is a binary system,
the system has corresponding restrictions in analysis. First of all, the system S
consists of n components which are described as fC1 ; C2 ; . . .; Ci ; . . .; Cn g. Sec-
ondly, the system and all components in it can only have two mutual states, where
Ci ¼ 0 represents the function state and Ci ¼ 1 means that the component Ci is
failure.
The Birnbaum importance measure represents which impact has a component
on the reliability of a system. So the Birnbaum importance measure is also called
reliability importance. It can be calculated as (117.3), where RðÞ represents the
reliability function of system and components (Birnbaum 1969).
1104 Z. Han and Z. Cai
oRðSÞ
IðBMÞSCi ¼ ð117:3Þ
oRðCi Þ
If the failure function of system and components is written as FðÞ, the Birn-
baum importance could also be measured as (117.4).
oRðS ¼ 0Þ oð1 FðSÞÞ oFðSÞ
IðBMÞSCi ¼ ¼ ¼ ð117:4Þ
oRðCi ¼ 0Þ oð1 FðCi ÞÞ oFðCi Þ
From the aspect of probability distributions, (117.3) can be transformed as
(117.5), which denotes the system reliability decrease when the component Ci
degrades from function to failure state.
oRðSÞ
IðBMÞSCi ¼ ¼ PðS ¼ 0jCi ¼ 0Þ PðS ¼ 0jCi ¼ 1Þ ð117:5Þ
oRðCi Þ
According to (117.5), in the WAM of research team, the Birnbaum importance
of member i is calculated as (117.6) and (117.7).
BPi ¼ ðajai ¼ n 1Þ a ð117:6Þ
BLi ¼ a ðajai ¼ 0Þ ð117:7Þ
BPi is the positive importance. Its physical meaning is the density increase of
research team when member i is full connected with others. The one with highest
BPi should be paid attention for deep cooperation.
BLi is the independent importance. It means the density decrease of research
team when member i is full independent with others.
117.4 Case Study
We take a research team M in Northwestern Polytechnical University, China as an
example to implement the case study. This research team has 6 full time faculties,
including 2 professors, 3 associate professors and 1 assistant professor. Their
research interests focus on mechanical engineering and have published more than
30 papers last year.
According to the concept of WAM, each member is represented with a node, such
as professor 1 = node 1, professor 2 = node 2, associate professor 1 = node 3,
associate professor 2 = node 4, associate professor 3 = node 5, assistant professor
1 = node 6. All members are connected with edges.
The weight of each edge represents the cooperation relationship between two
members. So we introduce the co-author index to estimate the weight of each edge.
The co-author index is calculated as (117.8).
117 Cooperation Relationship Analysis of Research Teams 1105
Fig. 117.4 The practical
WAM for research team M
Research Node 1
team M
Node 3 Node 4
Node 5 Node 6
Node 2
⎡ 0 0.47 0.68 0.61 0.58 0.74 ⎤
⎢0.47 0 0.37 0.31 0.28 0.24
⎥
⎢ ⎥
⎢ 0.68 0.37 0 0.32 0.39 0.81 ⎥
⎢ ⎥
⎢ 0.61 0.31 0.32 0 0.63 0.29 ⎥
⎢ 0.58 0.28 0.39 0.63 0 0.18 ⎥
⎢ ⎥
⎣⎢ 0.74 0.24 0.81 0.29 0.18 0 ⎦⎥
!
the number of papers published
by both node i and node j
aij ¼ ! ð117:8Þ
the number of papers published
by node i or node j
Finally, the established WAM of research team M is shown in Fig. 117.4
Based on (117.1) and (117.2), the degree of each member and the density of the
whole research team M are listed in Table 117.1.
From Table 117.1 it is clear that professor 1 has the highest degree which
verifies that he is the head and center of the team. Professor 2 has the lowest
degree. This is because she paid more time on international cooperation which
can’t be revealed in this model. The rest members have similar degree. The Team
density is 0.46, which shows that there are stable cooperation relationships in this
team.
1106 Z. Han and Z. Cai
Table 117.1 SMA results of research team M
Member Node Degree
Professor 1 Node 1 3.08
Professor 2 Node 2 1.67
Associate professor 1 Node 3 2.57
Associate professor 2 Node 4 2.16
Associate professor 3 Node 5 2.06
Assistant professor 1 Node 6 2.26
Team Density
Team M 0.46
Table 117.2 Birnbaum importance results of research team M
Node a a|ai = n-1 a|ai = 0 BPi BLi
Node 1 0.46 0.588 0.255 0.128 0.205
Node 2 0.46 0.682 0.349 0.222 0.111
Node 3 0.46 0.622 0.289 0.162 0.171
Node 4 0.46 0.649 0.316 0.189 0.144
Node 5 0.46 0.656 0.323 0.196 0.137
Node 6 0.46 0.643 0.309 0.183 0.151
Then the Birnbaum importance of each member in research team M is com-
puted with (117.6) and (117.7). The results are shown in Table 117.2. From
Table 117.2, it is clear that professor 2 has the highest BPi value, which means that
if she is full connected with others, the density of research team will get the
biggest increase. So, the team should paid attention to professor 2 for more
cooperation. Professor 1 has the highest BLi value, which means that if he leaves
the research team, the density of research team will get the biggest decrease.
117.5 Conclusion
In this paper, we introduce a weight adjacency matrix to modeling the cooperation
relationships between team members and use the social network analysis and
importance measures to evaluate the effect of a member on the whole team. The
case study of a research team from Northwestern Polytechnic University is
implemented to verify the effectiveness of the proposed methods.
Acknowledgments The authors gratefully acknowledge the financial support for this research
from the National Natural Science Foundation of China (Grant No. 71101116).
117 Cooperation Relationship Analysis of Research Teams 1107
References
Adams JD, Black GC, Clemmons JR, Stephan PE (2005) Scientific teams and institutional
collaborations: Evidence from U.S. universities, 1981–1999. Res Policy 34(3):259–285
Birnbaum ZW (1969) On the importance of different components in a multi-component system,
Multi-variate analysis 2. Academic Press, New York, pp 581–592
Young CA (1998) Building a care and research team. J Neurol Sci 160(S1): S137–S140
Dekker DM, Rutte CG, Van den Berg PT (2008) Cultural differences in the perception of critical
interaction behaviors in global virtual teams. Int J Intercult Relat 32(5):441–452
Dodson MV, Guan LL, Fernyhough ME et al (2010) Perspectives on the formation of an
interdisciplinary research team. Biochem Biophys Res Commun 391(2):1155–1157
Fussell BJ (1975) How to hand-calculate system reliability characteristics. IEEE Trans Reliab
24(3):169–174
Ghobadi S, D’Ambra J (2012) Competitive relationships in cross-functional software develop-
ment teams: How to model and measure? J Syst Softw 85(5):1096–1104
Krebs V (2012) Social network analysis, a brief introduction. http://www.orgnet.com/sna.html
LePine JA, Buckman BR, Crawford ER, Methot JR (2011) A review of research on personality in
teams: Accounting for pathways spanning levels of theory and analysis. Human Resour
Manag Rev 21(4):311–330
Pagell M, LePine JA (2002) Multiple case studies of team effectiveness in manufacturing
organizations. J Oper Manag 20(5):619–639
Salmi A (2010) International research teams as analysts of industrial business networks. Ind Mark
Manag 39(1):40–48
Vesely WE (1970) A time-dependent methodology for fault tree evaluation. Nucl Eng Des
13(2):337–360
Wikipedia (2012a) Adjacency matrix. http://en.wikipedia.org/wiki/Adjacency_matrix
Wikipedia (2012b) Social network analysis. http://en.wikipedia.org/wiki/Social_network_
analysis
Chapter 118
Establishment of Construction Standard
System Based on the Complex System
Theory
Zhi Sun and Shou-jian Zhang
Abstract Construction standard system is an organic whole, which is formed
scientifically by construction standards within certain range. The systematic effect
of standard system can make the construction activities get the optimum benefit.
Construction standard system has dynamic, open and nonlinear characteristics in
structure, environment and behavior aspects, and it is a typical complex system.
This paper uses system science theory to reveal the complex adaptive character-
istic of the standard system, provides the mathematical form of the system, and
analyzes the emergent property mechanism of construction standard system.
Finally, the structure modeling method of complex system is given according to
the complex characteristic of standard system. This paper provides a new method
and theory bases for the establishment of standard system.
Keywords Emergent property Complex adaptive system Complex system
Construction standard system (CSS)
118.1 Introduction
The quantity of construction standards has sharply increased along with the rapid
development of construction industry. Most of these standards were formulated
according to the foretime condition, and it caused some issues between con-
struction standards after several years’ development, such as discordance,
imperfect supporting, unreasonable content, mutual repeated, and conflict. The
Z. Sun (&) S. Zhang
School of Management, Harbin Institute of Technology, Harbin, China
e-mail: [email protected]
S. Zhang
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1109
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_118,
Ó Springer-Verlag Berlin Heidelberg 2013
1110 Z. Sun and S. Zhang
standard system structure becomes more and more complicated, and it is hard to
guarantee the normal operation and function properly.
Construction standard system has been investigated in a number of research
studies. Existing research in this important area focused on the function and
institution of the system. There are two trends of the construction standard system
in the future: the developed countries’ standard system will become the global
standard with their business expansion around the world, the other trend is the
mature standard system will still maintain the primary position along with their
own construction industry development (Bredillet 2003). The reasonable supply of
construction standard system can satisfy the building quality requirement of the
enterprises and the consumers, and it also can bring the ideal economic and social
benefits (Ofori and Gang 2001). Both ‘‘mandatory article’’ and ‘‘recommended
standard’’.
Come along at the present stage in China. This management mode will be
replaced by the ‘‘technical regulation’’ and ‘‘technical standard’’ management
mode in order to adapt the new requirement of the market environment (Mu 2005;
Yang 2003). Construction standards are divided into 18 professional categories in
China. Each category’s structure can be described by four levels: synthesis, base,
general, and specialized level (Wang 2007). Construction standard system should
contain the analysis of standard current situation, reasonable standard management
system, flexible operating mechanism, demand of standard in 5–10 years and the
harmonizing relationship of ‘‘technical regulation’’ and ‘‘technical standard’’. This
is the development trend of the construction standard system (Mu 2005).
‘‘So generally, project is more complex, the cost is higher, and the time limit for
the project is longer’’ (Baccarini 1996). Along with the development of economic,
the subject of investment of a construction project is diversification, and it makes
huge investment project possible. More and more complicated management
objects of organization, technology, time, quality go along with that, and it
enhances the demand of construction standard. The standard system reflects the
features of complex system gradually. Despite of the significant contributions of
the above research studies, there is no reported research that focused on: (1)
complex features of standard system; (2) emergent property of complex system;
(3) establishment method of construction standard system based on complex
system theory. Accordingly, there is an urgent need for additional studies to
address these there critical research gaps, as shown in Fig. 118.1.
118.2 Analysis of Construction Standard System
Complexity
Complex system is formed by a certain number of interaction elements. The inter-
action of each element can make system generate the self-organized behavior as a
whole (Wu 2006). Standard is a scientific organic whole constituted by associated
118 Establishment of Construction Standard System 1111
Fig. 118.1 The establishment model based on the complex system theory
Relationship
Subsystem of Elements
External
Environment
Input Output
The ComplexSystem of
Construction Standard System
Three Elements in Boundary
Feedback Loop
Fig. 118.2 The general concept of construction standard system
standards within a certain range. The system effect can make the economic activities
get the best benefit. Construction standard system is an interdependence, mutual
constraint, mutual complement and cohesion organic whole, which is formed by
interconnected standards in construction field.
Figure 118.2 shows that the construction system is a complex system, and it is
constituted by many elements and subsystem. The system is a multilevel, multiple
target system formed by a number of construction standards. The complexity
factors of construction standard can be described as external factor and internal
factor. The external complexity is the condition of system complexity including
open and dynamic. The internal factor is the reason of complexity, which contains
complexity of elements, organizational relationship and information.
1112 Z. Sun and S. Zhang
118.2.1 External Complexity Factor
Construction standard system is an open system, and it causes the complexity of
the system. An open system has the exchanges of material, energy, information
and knowledge with the environment. The way and degree of a system open, the
way of system and environmental interactions will influence the complexity of
the system. The elements and the subsystem of the standard system exist within the
scope of certain constraints and space–time. The system has various contacts with
all aspects of the construction field, and there are information, function, and benefit
exchange between them. These dynamics cause the construction standard system
to adjust and update constantly through the structure of the system, due to there is
dynamics in feature, extent, quality requirements and technology development of
the subjects of regulation. This discrete change is one of the most important
reasons for the complexity system.
118.2.1.1 The Dynamics in the Area of Construction Laws
and Regulations
Along with the development of economy and the field of construction, the laws
and regulations of construction field must adapt to the development demand of
construction activities. Specially, regulations about safety of people’s life and
property require the elements of the standard system to adjust accordingly.
118.2.1.2 The Development of Information and Communication
Platform
The development of the Internet makes the information, energy and material
exchange of construction standardization work in different way. It also brings out
higher request on the speed of standard update and the standard system optimi-
zation. The platform makes the standard update cycle shorten greatly, and adds the
complexity level of the system.
118.2.1.3 The Dynamics of Technology Environment Development
Technological innovation is transforming the economy at an accelerating pace
along with the arrival of knowledge-based economy. As the main carrier of the
technology transfer, construction standard need to update more rapidly in order to
satisfy the request of technical progress. The more the standard system includes or
relates discipline knowledge and technical type, the more complex of construction
technology requirement is, and the projects need more and more high technology
118 Establishment of Construction Standard System 1113
Fig. 118.3 External
complexity factors of
construction standard system
Development of Information
Technology Technology
Changes of
System
Structure
Complexity of
CSS
Adjustment
Increase of
of Standard
Standards
Content
Laws and
Regulations
equipment, new materials and knowledge. These requirements enhance the com-
plexity of standard system during the standard update.
The dynamic change of external factor of construction standard system is the
origin of system adjustment. The dynamic of system causes the unsteadiness of the
system. The disturbance of external elements is one of reasons of the complex
system, as shown in Fig. 118.3.
118.2.2 Internal Complexity Factor
The internal complexity of construction standard system is caused by the com-
plexity of standard system structure, the quantity of participating subject giant, the
different objects of the system and the complexity of the feedback information, as
shown in Fig. 118.4.
Fig. 118.4 Internal
complexity factors of
construction standard system Complexity of Complexity of
Standard System main Participants
Structure of CSS
Complexity of Complexity of
Each Subsystem’s the Feedback
Object Information
Internal Complexity of
the System
1114 Z. Sun and S. Zhang
118.2.2.1 The Complexity of Standard System Structure
Construction standard system is constituted by a number of elements and subsystem.
Each subsystem can be described by many sub elements, such as housing building
standard system, urban construction standard system. Generally, the more levels
system has, the more complexity system will get. There are many factors increase
the internal complexity of the system, such as the interconnect interact between
different levels and sub elements, the complex link between technology index,
subjects and basic terminology. The standard system optimization is a dynamic
process, and the update of each subsystem also shows dynamic characteristic.
118.2.2.2 The Complexity of Main Participants of Construction
Standard System
The range of construction activities is very wide, and it not only contains the civil
engineering and equipment installation, but also contains the materials, cost,
infrastructure and other professional fields. These participants of the system
influence each other in space–time, due to the establishment of system need the
cooperation of participants. The establishment of construction standard system
needs to cross multiple professional. The cross of the professional between
transverse and longitudinal is very complicated, for example, the housing building
subsystem needs the building design, structure, and foundation professional to
combine in their respective stage. Each elements of the subsystem enters or exits
the system in their own life cycle. So the complexity of the standard system
embodies how to coordinate different professional standards in their different life
cycle organically, thereby give full play to the function of the system, and reduce
the conflict and contradiction between different standards.
118.2.2.3 The Complexity of Each Subsystem’s Object
Construction standard system can be divided into 18 majors by different fields.
Each major of the system has different subsystem object. The combination of the
standards in each subsystem can evolve new system with different objects. The
diversification and the pertinence of the system’s function causes the complexity
of construction standard standard system, finally complexity of the system reflects
the complexity of each subsystem’s aim.
118.2.2.4 Complexity of the Feedback Information
The feedback information of the construction standard system comes from owner,
design institute, builder, supplier of the materials and the supervision units, and it
also comes from different project stage, such as feasible research, design, bid
118 Establishment of Construction Standard System 1115
inviting and construction process. The management process also generates large
quantity of feedback information, such as quality control, investment control,
progress control and the contract management. The feedback information from
different units enhances the complexity of the information search during the
system optimization process. The feedback information from related units may
have contradiction, and the benefit may have consistency sometimes. The
dependency and relevancy of feedback information from different related units,
process, and environment can enhance the difficulty of the information collection
and analysis of the information demanders.
118.2.2.5 The Analysis of Complex Adaptive Character
of Construction Standard System
The research on the construction standard system is stuck in early stage of system
science, which means the concept of elements, subsystem and the structure. The
standards in system is completely passive, the purpose of their existence is to
realize some task or function of the system, and these standards have no their own
objects and orientations. The system can’t ‘‘grow’’ or ‘‘evolve’’ in the environ-
mental interactions, and it only can make fixed reaction according to the fixed
mode even though there’s some communication with external environment. The
elements of the construction standard system can be regarded as an active and
adaptive agent with their own purpose and initiative by using the theory of
complex adaptive system.
Holland provided four characteristics of the study on the adaptive and evolu-
tionary process about the concept of ‘‘agent’’: aggregation, nonlinearity, flows and
diversity. He also gave three mechanisms: tag, internal, building blocks (Holland
1995; Jin and Qi 2010).
The constitutors of the construction standard system can analyze the demand of
standard function, forecast the development direction of technology and take
action according to the predetermined object. The constitutors can form an
aggregate of organization in some particular field by ‘‘adhere’’, and this aggregate
becomes the organization of standardization finally. The new aggregate develops
in the environment, which has huge demand for the aggregate. The whole process
can be considered as the motion of a subsystem. This aggregation relationship
doesn’t mean that every organization can adhere together. These organizations that
comply with subsystem development goals, be helpful for standardization field,
have the professional relativity can have this kind of aggregation relationship. This
common object, conditions and rules of the choice are endowed with a cognizable
form, which is called ‘‘tag’’.
‘‘Nonlinearity’’ and ‘‘flows’’ are two characteristics of construction standard
system. There is the exchange current of information, function and benefit between
standards, subsystems, levels and standardization activities as previously men-
tioned. Moreover, the unblocked degree and turnover frequency of the ‘‘flows’’ are
in a high level due to the complexity of the system. The elements and their
1116 Z. Sun and S. Zhang
characteristics are constantly changing, which is caused by the existence of the
flows of information, function and benefit under the repeated interactions with the
system. The changes don’t follow the simple, passive, unidirectional causality and
linear relationship, but are the active adaptive relationship. The previous experi-
ence can affect the future activities, and the changes are the complexity rela-
tionship of interaction and mutual entanglement actually.
The diversity of construction standard system is very common, such as the
diversity of system dimensionality, objects of standardization and standard form.
The primary reason is the diversity and instability of system, which is caused by
the nonlinear function between internal system and the environment.
The constituters of construction standard system have the ability of foreseeing
environment change and adjusting their own behavior. The constituters have the
unique internal mechanism and decision model during the process of environment
adapting and stress reaction. The way of reaction is judged by the internal
mechanism. The whole process can be called ‘‘internal’’ for the system.
As a complexity system, the construction standard system is formed by the
different combination way of many simple parts, which is defined as ‘‘building
blocks’’. The complexity of the system doesn’t depend on the number or size of the
building blocks, but depends on the recombination of the original building blocks.
In many cases, the old internal is often considered as the building blocks to
generate new internal by the recombination.
118.3 The Analysis of Emergent Property Mechanism
of Construction Standard Complexity System
The elements form the system according to some way, and it will produce the
specific attribute, characteristics, behavior, function, which the whole system has
but the parts don’t have. The system theory calls this emergent property (Miao
2006). The most important thing to describe the construction standard system by
using the system theory is to grasp the whole emergence of the system.
118.3.1 The Qualitative Description of Construction
Standard System
There are a large number of elements involved in construction standard system,
and the relationships between them are very complicated. On one hand the par-
ticipants can’t understand it deeply. On the other hand, the research on system
can’t be extended to the analysis of the whole system operation process, due to the
dynamic of the system. The cognition and analysis on system should pick up the
core influence variables from the huge number and multifarious qualitative
118 Establishment of Construction Standard System 1117
Fig. 118.5 The structure of
abstract construction standard
X
system
x1 Rx
x2 Rx x3 Xn
y11 R 1y y12 R1y y13 Y1m … Ym
R12 z R12 z
z121 z122 z123 Z12e … Ze
aspects, and get the related functional factors (Bo et al. 2002; Farley and Lin 1990;
Berndsen and Daniels 1994).
According to the description of the complexity system, the paper uses the
mathematical expression to describe a construction standard system with tertiary
structure qualitatively:
The definition: the system X means a whole formed by n relational elements
x1 ; x2 ; . . .; xn , measured as X ¼ fXn ; Rx g, and Xn ¼ fxi ji ¼ 1; 2; . . .; n; n 2g. Rx
is the relation of these elements, called the soft structure of the system (Miao
2005). Meet the following conditions:
P
n
(1) 9 a subsystem Y ¼ fYm ; Ry g ; Y Xn ; Ym ¼ Yim ; and xi ¼ fYim ; Riy g
i¼1
Yim ¼ fyij jj ¼ 1; 2; . . .; m; m 2g. Again the tertiary of the subsystem: Z ¼
P
n P
m
fZe ; Rz g; Z Ym ; Ze ¼ Zije ; Zije ¼ fzijk jk ¼ 1; 2; . . .; e; e 2g;
i¼1 j¼1
yij ¼ fZije ; Rijz g The structure is shown in Fig. 118.5:
0 1
a11 . . . a1n
B . .. .. C
(2) Zm ¼ AXnT , in there A ¼ @ .. . . A Xn ¼ ½xli ; ðl ¼ 1; 2; . . .; LÞ,
am1 amn MN
and L 2.
(3) Xn \ Ym 6¼ £, if the x2 in Xn relates to the y3 in Ym , Xn \ Ym ¼ ½x2 ; y3
The above expression means that the construction standard system also has the
hierarchical, nonlinear and coupling characteristic.
1118 Z. Sun and S. Zhang
118.3.2 The Emergent Property Mechanism of Construction
Standard System
The emergent property of construction standard system comes from the elements,
structure and the environment to the system. The effect of elements, scale,
structure and environment make the whole emergent property of the system (Bo
et al. 2002).
(1) The construction standard system is formed by the laws, regulations,
standards and other elements, so the origin of the whole emergent property comes
from each element. The emergent of system is restrained by the characteristic of
the element, and that means the random combination of these elements can’t form
the system. (2) There’s the relationship between the emergent and the scale of the
system. The scale of the system is the essential condition of the complexity, and
it’s hard to emerge the complexity of a system from simplicity without enough
elements. (3) The level and the characteristic of each element is the material base
of the whole emergent property, and they only provide the objective possibility for
generating the emergent. The interaction, mutual inspire, restraining and com-
plement with each other between the different ways of every elements can generate
the whole emergent, which is called structure effect, and it is the core source of the
whole emergent property. (4) The external environment provides the necessary
resources and constraint condition for the generating process of emergent. The
construction standard system can get the resources from interaction with the
external environment. The resources can help the system exploit system space,
form the system boundary and establish the channel for exchanging material,
energy and information with the construction activities. It also can make the
system adapt new environment, enhance the anti-jamming capability. These
exchanging results generate the emergent property of the system finally.
According to definition of complexity system and the mechanism of emergent
property, we can deduce the emergent, which the system has but the elements
don’t have, mechanism abstractly.
According to the mathematical definition of the construction standard com-
plexity system, system X ¼ fXn ; Rx g; Xn ¼ fxi ji ¼ 1; 2; . . .; n; n 2g is the
subsystem of X, Rx is the correlation set of elements x1 ; x2 ; . . .; xn ; xi ¼ fXi ; Ri g,
S
n
and xi is in the Xi ¼ Xn .
i¼1
S
n S
n
The emergent of the system is reflected in xi X, because Xi ¼ Xn , when
i¼1 i¼1
S
n S
n
Ri Rx , xi X is true.
i¼1 i¼1
Because:
118 Establishment of Construction Standard System 1119
S
n
(1) Ri 6¼ Rx
i¼1
S
n
Prove: suppose Ri ¼ Rx , only Rx ¼ [ and Ri ¼ [, or Rx 6¼ [ and Ri ¼ 0.
i¼1
That means there’s no soft structure between two subsystems, and it’s
inconsistent with definition of construction system: X ¼ fXn ; Rx g, so usually
Sn
have: Ri 6¼ Rx
i¼1
S
n
(2) Ri Rx is not correct
i¼1
Prove: because xi is the subsystem of arbitrary system X, xi Xis true for
S
n
arbitrary system xi , so Ri Rx
i¼1
S
n S
n S
n
It’s knowable that because of Xi ¼ Xn , Ri Rx , xi X finally.
i¼1 i¼1 i¼1
Through the process of argumentation, the system has the function that the parts
don’t have, and this function comes from the soft structure of the system, which
means the structure relationship of the parts.
118.4 Conclusion
The complexity system theory applies a new thinking and a rationale to the
establishment of the construction standard system. The complexity of the standard
system is caused by the internal and the external factor. External factor contains
the dynamic of laws and technology environment, and the development of infor-
mation technology. The internal factor contains the level of system, participants of
the system, feedback information and complexity of the subsystem objects. This
paper uses the theory of CAS, describes the complexity of the standard system.
The construction standards form the system according to some objects base on the
theory of emergent property. The system has the function and the characteristic
which the parts or the total of part don’t have.
Acknowledgments Foundation item: Project of the national twelfth-five year research program
of China (2012BAJ19B03)
References
Baccarini D (1996) The concept of project complexity—a review. Int J Project Manage
14(4):201–204
Berndsen R, Daniels H (1994) Causal reasoning and explanation in dynamic economic system.
J Econ Dyn Control 18:251–271
1120 Z. Sun and S. Zhang
Bo L, Zhang S, Li Y (2002) The qualitative representation and inference of the complex systems.
Syst Eng Theory Pract 12:15–21 (in Chinese)
Bredillet CN (2003) Genesis and role of standards: theoretical foundations and socio-economical
model for the construction and use of standards. Int J Proj Manage 21:463–470
Farley AM, Lin KP (1990) Qualitative reasoning in economics. J Econ Dyn Intell 14:435–450
Holland J (1995) Hidden order: how adaptation builds complexity. Addison-Wesley, New York
Jin H, Qi W (2010) Research on the theory and applying of complex system brittleness.
Northwestern Polytechnical University Press, Xi’an, p 120 (in Chinese)
Miao D (2005) On system thoughts (4): careful investigation going deep into system. Chin J Syst
Sci 13(2):1–5 (in Chinese)
Miao D (2006) On systematic thoughts (6): to focus the attention on the emergent properties of
the whole. Chin J Syst Sci. 2006 14(1):1–6 (in Chinese)
Mu X (2005) Comprehensive description and prospect of national engineering construction
standard. Spec Struct 1:90–92 (in Chinese)
Ofori G, Gang G (2001) ISO 9000 certification of Singapore construction enterprises: its costs
and benefits and its role in the development of the industry. Eng Constr Archit Manage
8(2):145–157
Wang C (2007) The analysis of construction standard system. Architect 5:111–116 (in Chinese)
Wu S (2006) Study on the synergic mechanism and methods of construction project management
based on complex system theory. Tian Jin University, Tian Jin, pp 24–25 (in Chinese)
Yang J (2003) The research on the system of construction regulations and standards. Harbin
Institute of Technology, Harbin, pp 15–18 (in Chinese)
Chapter 119
Location Selection of Coal Bunker Based
on Particle Swarm Optimization
Algorithm
Qing-an Cui and Jing-jing Shen
Abstract Center location selection of coal bunker is an important and practical
problem in coal mine production. Because of the complex relationship between
influence variables and optimization goal, it is frequent to reach a local optimi-
zation point rather than the global one by using linear programming. This paper
combines nonlinear programming model and the algorithm of particle swarm
optimization (PSO) to optimize the location selection of coal bunker in the coal
mine transportation system. Firstly the coal bunker’ center location selection
problem is formalized and thereby the nonlinear programming model is con-
structed by minimizing the entire cost of the system. Secondly, the optimization
model is solved by using the PSO algorithm and therefore the global optimization
is reached. Finally the method mentioned above is verified by a typical coal bunker
location selection example.
Keywords Coal bunker Location selection Nonlinear programming Particle
swarm optimization
119.1 Introduction
Coal bunker usually contains bottom coal bunker, district coal bunker, section coal
bunker, ground bunker, tunneling bunker (Wang 1983). As the main cavern
(Zhang 2010), coal bunker has an important role in the process of coal
Q. Cui (&) J. Shen
Department of Management Engineering, Zhengzhou University,
Zhengzhou, China
e-mail: [email protected]
J. Shen
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1121
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_119,
Ó Springer-Verlag Berlin Heidelberg 2013
1122 Q. Cui and J. Shen
transportation. With the development of mechanization and production concen-
tration levels of the transportation system in the coal industry, the importance of
locating coal bunker is more and more significant in the transportation system. As
the coal underground needs to be transported and carried through the system to the
ground, one of the effective methods is to transport the coal mine underground to
some special coal bunkers firstly, then to carry the coal to the ground in con-
centration, it is not only saves transportation cost, but also does improve trans-
portation efficiency. The theme of this article is to study how to select location
establishment of coal bunker, thus making the total cost of transportation least. It is
important and difficult to research the location problem of coal bunker in the
underground transportation system; the study of domestic existing literature is a
little less, mostly focuses on coal construction and maintenance. For example,
Zhen and Wang (Hua 2008) use particle swarm algorithm to research the problems
of selecting monitoring stations in safety monitoring and underground transpor-
tation optimization system, and optimist coal layout to reduce the transportation
fees, thus to enhance the transportation efficiency; Wei and Ma (Ma et al. 2012)
introduce the construction technique of large deep coal bunker, and emphatically
expounds key operation of protection system, and puts forward the corresponding
measures for quality assurance to direct the similar project construction of coal. It
is also a very practical application problem. On one hand, the selection of coal
center is in relation to security of an underground transportation system; on the
other hand, it links to the cost and efficiency of transportation. Therefore, it has the
important practical significance to research the selection of coal bunker in this
paper.
At present, it has won successful applications in the research of coal trans-
portation problem based on genetic algorithm (Wu and Shi 2004), ant colony
algorithm (Zhou 2006), simulated annealing algorithm (Qin and Shi 2007) and
linear programming (Wang 2011). As the location selection of coal bunker in
logistics system belongs to the discrete and combinatorial optimization problem,
and has the properties of NP hard problem (An 2007). As it is not easy to realize
the global optimization with linear programming, the value is mostly for local
optimization, not enough to reflect the complex relationship between variables, but
it is not for the nonlinear programming. Whereas, in the current stage, it doesn’t
have the most suitable algorithm that can ensure it gets the optimal solutions, and
only can approximate to achieve optimal solutions in large probability. Therefore,
according to this problem, this paper combines nonlinear programming model with
the particle swarm optimization algorithm to optimize coal bunker in the trans-
portation system.
Firstly, this paper introduces the basic principle of particle swarm optimization
algorithm, then describes the location of coal bunker underground, and combines
nonlinear programming model with particle swarm optimization algorithm to
modeling and optimize it, and finally gets the optimization results.
119 Location Selection of Coal Bunker Based on Particle 1123
119.2 Methodology
Particle Swarm Optimization (Kennedy and Eberhart 1995), which is known as
PSO, put forward based on simulating birds’ feeding behavior by Kennedy and
Eberhart in 1995. It has characteristics of strong robustness (Niknam et al. 2009), it
is bionics optimization algorithm that can handle complex optimization problem,
and is a random search method reference to natural selection and biological par-
ticle swarm mechanism (Wang et al. 2009).
In the particle swarm optimization algorithm, each solution of optimization
problem links to a particle (Ciurana et al. 2009), every particle has its position and
speed, and then search for the optimal value according to the individual extremism
Pgest and global extremism ggest in the space.
The basic steps of particle updating are proposed to be summarized as follows
(Wang and Li 2011):
(1) For each particle that selected as xi ¼ ðxi1 ; xi2 ; . . .xid Þ, the initial value is given
with random method and initial flight rate as vi ¼ ðvi1 ; vi2 ; . . .vid Þ:
(2) Particles are flying in the search space, the flight rate of each step updates
according to formula:
vid ¼ x vid þ /1 rnd1 ðpid xid Þ þ /2 rnd2 ðpid xid Þ ð119:1Þ
Among them, x is for inertial weight, generally its init value is 0.9, and can be
decreased to 0.4 with the linear increasing of iteration number (Huang 2011),
which allows it to focus on global search firstly, and fast convergence in a certain
area. /1 ; /2 are learning factors and normal numbers, and usually the value is 2,
rnd1 ðÞ; rnd2 ðÞ are random functions, the values range in (0, 1). Pid is for particles’
best position on the history, Pgd is for the position of particle which has best fitness
in all particles.
(3) Calculate the next step of the particle’s location with the rate of current
particle, namely
xid ¼ xid þ vid ð119:2Þ
(4) Return to the second step to repeat calculating, until it reaches the limit value
set or the evaluated number of function values is greater than that of the
biggest function set.
1124 Q. Cui and J. Shen
119.2.1 Solving the Location of Coal Bunker Based
on Particle Swarm Optimization Algorithm
Hypothesis that there exist n working surfaces in a range of underground, each
working point is set as q1 ; q2 ; . . .; qn respectively, and the coal production capacity
of corresponding unit time is for m1 ; m2 ; . . .; mn respectively. mi is called as the
productive capacity of working point qi , considering the actual problem, the coal
bunker only can be built on a particular point, or in a place edge of G. At present,
the core problem is how to find point q, thus make the total cost as FðqÞ minimum.
The coal bunker set belongs to the problem of cost minimum, namely that
solving the minimization problem of transportation to minimize the cost. This
paper selects a certain number of working points as resource points, and chooses a
certain number of alternative points as the coal bunkers from alternative ones. In
order to set a model easily, it makes the following assumptions: (Liu and Zhu
2005)
(1) The cost of resource point to point of the optional coal bunker is a linear
function;
(2) The capacity of the coal bunker must meet requirements;
(3) Select coal bunker only in a certain range.
Here, the amount of resource points is called as m, the set is as
N ¼ f1; 2; . . .; mg; the total number of alternative locations is remembered as
q; M ¼ f1; 2; . . .; qg is the set; Aj is for the total supply of each resource point
j; Mk is the max capacity for alternative point kðk 2 MÞ; Cjk is charge for unit
transportation between candidate location and resource point jðj 2 NÞ; ljk is
for transportation distance between optional point k and resource point j; Fk is
the fixed expenses (including the cost of basic investment and fix manage-
ment) of candidate location k on its construction; xjk is the coal production in
alternative point k from resource point j; Ck is Management cost of unit
circulation in alternative point k; Wk is the decision variable whether optional
point k is selected to be coal bunker. Dk is for the minimum demand of
optional point k.The single objective optimization model on coal location is
shown below:
!
X q Xm Xm
F ¼ min Cjk ljk Xjk þ Ck Xjk þ Fk ð119:3Þ
k¼1 j¼1 j¼1
Constraint conditions are as follows:m
X
Xjk Aj ð119:4Þ
k¼1
X
m
Xjk Mk ð119:5Þ
j¼1
119 Location Selection of Coal Bunker Based on Particle 1125
X
m
Xjk Dk ð119:6Þ
j¼1
Among them, Formulas (119.4) is said that the total coal materials from resources j
to the alternative points are not more than supply capacity; Formulas (119.5) is
said that the total supply from the points of all coal resource to coal bunker k is not
more than its maximum capacity; Formulas (119.6) is said that supply amount
from the points of all coal resource is not less than its demand.
119.2.2 Building Model of Coal Bunker Location Based
on Particle Swarm Optimization Algorithm
The main problem of coal bunker location in logistics system is to build some
certain coal bunkers from candidate ones under the supply of a series of resource
points. In other words, to optimize which coal bunker will be distributed from
these resource points.
Therefore, particle position can be structured as following steps: For m resource
points, q candidates of coal bunker, the current position of particle k is
Xk ¼ fx1k ; x2k ; . . .; xmk g, xjk ðj ¼ 1; 2; . . .; mÞ is that transport coal resource to j,
therefore, xjk values interval in ½1; q. Select the adapt function as following form:
X
m X
m
fk ¼ Cjk ljk Xjk þ Ck Xjk þ Fk Wk ð119:7Þ
j¼1 j¼1
fk is the adapt value of particle k; and the best adapt value of global is as:
gbest ¼ minffk g ð119:8Þ
Pbest is the position owing the best adapt value that particle k has experienced, Pbest
is obtained by the following formulas :
Pbest ðtÞ; if f ðXk ðt þ 1ÞÞ f ðPbest ðtÞÞ
Pbest ðt þ 1Þ ¼ ð119:9Þ
Xk ðt þ 1Þ; if f ðXk ðt þ 1ÞÞ \ f ðPbest ðtÞÞ
Implementation procedure of algorithm is as follows:
(1) Initialization. Set the random position of particle swarm xk , speed vk and the
maximum iterating times T, the initial particle randomly generates in the
feasible domain of solution space;
(2) Calculate the adopt value of each particle according to formulas (119.7);
(3) For each particle, compare the adaptive value with the one on the best position
Pbest has experienced, thus to update the current best particle position
according to (119.9);
1126 Q. Cui and J. Shen
(4) For each particle, compare the value on the best position Pbest has experienced
with the one on the global value gbest , thus to update the global best particle
position;
(5) Update the value of the inertia weight w;
(6) The rate and position of the particles is iteration according to (1, 2);
(7) Update fbest , if it reaches maximum iterating times T, the cycle is over, and
output optimal particle, its best position gbest and the value of optimal fitness;
otherwise the cycle is not over, and return to step (2) to continue to operating.
119.3 Results
In order to validate the feasibility and effectiveness of particle swarm optimization
algorithm in solving the coal bunker setting, use the following example to analysis
and test.
Hypothesis that there are three resource points in coal underground logistics
system, the supplements of resource are D1 ; D2 ; D3 , the values respectively are 60,
70, 80. Coal bunker’s alternative points own four ones, the fixed cost, maximum
capacity and minimum demands are shown in Table 119.1. Various rates are
shown in Tables 119.2, 119.3 and 119.4.
Through using particle swarm optimization algorithm, the best choice for coal
bunker and the results of coal transport distribution are shown such as in
Table 119.5.
From the data in Table 119.5 and Fig 119.1, it can see that particle swarm
optimization algorithm can quickly and effectively be used to get optimal solution
or approximate optimal solution of the problem of coal bunker location, and it also
has reference value to solve the similar problem about optimization.
Table 119.1 Alternative points of coal bunker
Alternative points L1 L2 L3 L4
Fixed costs 90 80 110 70
Maximum capacity 40 30 60 20
Minimum demands 30 25 55 10
Table 119.2 Unit goods rate from resource point j to coal bunker’s alternative point k
k
j 1 2 3 4
1 7 8 13 13
2 10 10 9 8
3 11 8 11 9
119 Location Selection of Coal Bunker Based on Particle 1127
Table 119.3 Unit distance from resource point j to coal bunker’s alternative point k
k
j 1 2 3 4
1 3 5 5 4
2 4 3 4 3
3 5 4 31 5
Table 119.4 Management fee rate of unit circulation in coal Bunker’s alternative points
Alternative points L1 L2 L3 L4
Management fee rate of unit circulation 80 90 95 85
Table 119.5 The quantity of goods from resource point j to coal bunker’s alternative point k
k
j 1 2 3 4
1 30 0 0 0
2 0 25 0 10
3 0 0 55 0
Fig. 199.1 Distribution
relationship of coal bunker D3
between resource points and L3
alternative points D1
L2
D2
L1
L4
119.4 Conclusion
This paper solves the problem of coal bunker selection in underground logistics
system on the case study based on Particle Swarm Optimization, and establishes
the corresponding nonlinear model, thus to realize the cost minimization. The case
shows that this method can be effectively applied in single target problem of
location selection about coal bunker, thus to provide a new kind of optimization
algorithm for coal bunker selection in coal underground logistics system.
Acknowledgments This paper is sponsored by Education Natural Science Foundation of Henan
Province (No.2111A410003).
1128 Q. Cui and J. Shen
References
An LJ (2007) Research and application of the genetic algorithm optimization in logistics
distribution vehicle scheduling (in Chinese). Shanghai maritime university, China
Ciurana J, Arias G, Ozel T (2009) Neural network modeling and particle swarm optimization
(PSO) of process parameters in pulsed laser micromachining of hardened AISI H13 steel.
Mater Manuf Process 24(3):358–368
Hua, Z, Wang C (2008) Application of ant colony algorithm in the optimization of mine (in
Chinese). Coal J 33(3):353–356
Huang MM (2011) Particle swarm optimization based method for logistics center location
problem (in Chinese). Comput Eng Appl 47(4):212–214
Kennedy J, Eberhart RC (1995) Particle swarm optimization.In: Proceedings of IEEE
international conference on neural network, Australia. IEEE Computer Society Press,
pp 1942–1948
Liu L, Zhu JR (2005) The research of optimizing physical distribution routing based on genetic
algorithm (in Chinese). Comput Eng Appl 27:227–229
Ma WF, Chen JS, Zhao JC (2012) Large-scale coal bunker engineering practice in deep shaft (in
Chinese). Shanxi Archit 07:117–118
Niknam T, Amiri B, Olamaei J, Arefi A (2009) An efficient hybrid evolutionary optimization
algorithm based on PSO and SA for clustering. J Zhejiang Univ Sci A 10(4):512–519
Qin J, Shi F (2007) Bi-level simulated annealing algorithm for facility location (in Chinese). Syst
Eng 25(2):36–40
Wang L (1983) Analysis and calculation of capacity of underground coal bunker (in Chinese), J
Huainan Min 01:113–122 ? 35
Wang XY (2011) Application of linear programming used in the plan of coal production (in
Chinese). Coal Econ 10:28–31 ? 34
Wang JQ, Li SQ (2011) Researching the effectiveness of particle swarm optimization in
searching solution for linear programming (in Chinese). Manuf Autom 33(6):88–91
Wang Q, DuanMu JS, Xu L (2009) Selection of distribution center’s location based on particle
swarm optimization in military logistics (in Chinese). Comput Eng Des 30(15):3597–3599
Wu J, Shi ZK (2004) Selection of distribution center’s location based on genetic algorithm (in
Chinese). J South China Univ Technol (Nat Sci Edn) 32(6):71–74
Zhang S (2010) Design and construction of underground coal mine (in Chinese). Sci Technol Inf
36:84
Zhou T. (2006) Research on improved genetic algorithm for TSP problem (in Chinese).
Microelectron Comp 23(10):104–106, 110
Chapter 120
Mechanism of Firm’s Response
to Innovation Policy in Industrial Cluster:
Based on Echo Model
Yong-an Zhang and Chen-guang Li
Abstract This paper presents a study of the behavior in firms receiving the
government’s policy support. It is emphasized that firm’s attitude is an important
factor to policy planning. Resource-based theory is also adapted to firm’s response
to the innovation policy of industrial cluster. In the perspective of complex
adaptive system, policy responding has suffered from the influence of some factors
such as comprehension, firm’s scale, entrepreneur, benefit, innovative capability,
demands for R&D and neighbor firms. Based on echo model, the paper discusses
the mechanism of cluster firm’s response to innovation policy. Furthermore, it
describes the important role of various factors play in the process of policy
responding and trigger condition. Finally, the behavior matching modes and self—
adaptive responding flow have been analyzed. Findings show that the fundamental
of firm’s response to policy is resource—depending and self—organizing evolu-
tion ensures policy responding better and better.
Keywords Echo model Industrial cluster Innovation policy Policy
responding
120.1 Introduction
The industrial cluster (cluster) has a tremendous role in promoting the develop-
ment of all sectors of society on the regional economy (de Oliveira Wilk and
Fensterseifer 2003). It is no doubt about the positive role of regional innovation
policy promotes enterprise technology research and innovation. Specially, in the
Y. Zhang (&) C. Li
School of Economics and Management, Beijing University of Technology, Beijing, China
e-mail: [email protected]
C. Li
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1129
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_120,
Ó Springer-Verlag Berlin Heidelberg 2013
1130 Y. Zhang and C. Li
industrial cluster, innovation policies conducts firms overcome the drawback of
limited resource and innovation bottleneck by reciprocal investment and resource
redistribution. However, asymmetry of policy information, longer return cycle,
higher cost and standards of accreditation lead to the innovation policy responding
effect is not very good. In recent years, scholars acknowledge that system theory is
very helpful to policy studies about innovation policy towards societal challenges,
rather than economic growth objectives only (Weber and Rohracher 2012).
Moreover, due to the participation and influence of governments, S&T institutes,
enterprises and other agents, policy issues are increasingly complex (Kern 2012).
The implementation effect of innovation policy depends on two agents—gover-
nance and firms. On the one hand, government should be make a fair, reasonable,
humane support policies based on firm’s actual R&D requirement. On another,
firms should respond positively after clear the policies provide foundational sup-
port (as money, materials, human-resource, etc.) and create a favorable environ-
ment by timely analyzing the policies’ purpose and meaning. A new approach has
being demand pressingly. From the macro point of view, we have to consider the
mechanism of policy acting on innovative performance. And from the micro,
the innovative performance is related with entrepreneurial attitude to the policy.
The complex adaptive system (Holland 1995) and its echo model could research
the issue from macro and micro aspect better. That is the reason the paper choose
CAS theory. Firstly, we generalize the framework of innovation policy responding.
Then, summarize the responding process. Finally, we find out key factors by
literature analysis and build the Echo Model through advocating five mechanisms
of firm responding innovation policy in Industrial cluster.
120.2 Research Framework
Using the echo model, researcher could solve the issue about mutual dissemination
and absorption of firm’s resource innovation by resource-based theory (Holland
1995). This paper sums up the echo model research framework as shown in
Fig. 120.1. The model provides an analyze method of multi-agent interaction
mechanisms for the government and industrial cluster’s firms. The mechanism is
divided into two segments naming control and tagging. The control segment
includes selective response, resource acquisition, conversion, evolution and con-
ditions of replication. The tagging segment is composed by offensive, defensive
and adhesion.
For the process of policy response, the government realizes resource allocation
by policy planning and implementing. Simultaneously, firms are responding to the
policy for obtaining resource and engaging in innovation. These resources are
usually divided into the information, funds, intellectual results, human resource
and physical tools (Filippetti and Archibugi 2011). In this paper, innovation policy
provides market information, funds, project items, human resource. The infor-
mation includes guidance, specifications, market situation, technological advances,
120 Mechanism of Firm’s Response to Innovation Policy 1131
Causality
Analyzing
Firms of R&D R&D
Industrial Cluster Performance Behavior
Condition
Echo Model
Selective response
INTERACT
COMPLEX Resource acquisition
ADAPTIVE Resource conversion Mechanism
SYSTEM
Intelligent evolution
Adhesion
Innovation Policy
Factors
Government Analysis
Other Effects
Fig. 120.1 Research framework
and R&D results. The funds are a collection of investment, subsidies, tax relief,
etc. The projects provided by the government are stimulating cluster’s firms R&D
passion which has the form of major projects, fund projects, technology projects,
and special projects. General, technology and R&D staff as human resource is
supplying skill persons who are needed to Innovation R&D activities. Among
them, the information and project providing by the government is an interactive
resource which also could be shared. To be emphasized, the information only has
the situation ‘have’ or ‘None’, and the remaining resource can be measured.
120.3 Response Mechanism Analysis Using Echo Model
After created echo model (Holland 1995) uses it to research group prisoner’s
dilemma and success to apply adaptive function in the process of multi-agent
interaction for gaming. This paper will define agents as cluster’s firms and gov-
ernment in the analysis of innovation policy response mechanism.
120.3.1 Key Factors
It is difficult to analyze the factors of innovation policy response directly
(Coronado et al. 2008). And from the process of cluster’s firms respond innovation
policy perspective, a new thinking about key factors mining this paper proposed.
1132 Y. Zhang and C. Li
Based on the actual response to innovation policy, Scholars has been summarized
seven key factors (such as shown in Table 120.1).
(1) Cognition ability: which is the ability to learn and understand, and the correct
cognition of innovation policy is a prerequisite for cluster firm’s response to
the policy. Based on questionnaire survey, scholars believe that carefully
Table 120.1 Factors Affecting Firms Response to Innovation Policy
Factors Literature analysis
Arguments Variables Scholars
1 Cognizing policy correctly is a Benefit qualification Bessant (1982),
precondition, and analyzing Fichman (1997),
clause carefully and object Waarts(2002)
clearly is crucial
2 Larger firms often actively Assets, employees, Souitaris (2002), Jeroen
respond to policy, because resources, liabilities (2010), Paunov
they need to strengthen (2012)
innovation capability,
develop new market, and
cooperate
3 Innovative entrepreneurs support RiskAppetite, Claver (1998),
innovation policy actively. firmaffiliation, location Horwitch and
An inert, questioning and Mulloth (2010),
inimical entrepreneurs hinder Colwell (2010)
the policy implementation
4 Direct invest and projects of Funds, investment, cost, Simpson (1996), Huang
policy affect firms’ R&D and skillmennum, patent, (2007), Gilbert and
technology innovation which profitability, tax relief Katz 2011
reduces cost and increases
income. The expected return
is key factor whether a firm
responds policy
5 Innovation policy response Utilization and conversion Acs (1992), Mowery
willing depends on rate of resource and (2001), Audrets
knowledge spillovers knowledge, yield (2005), Grace and
resource utilization, and Shen (2010)
innovation results promoting
effect
6 Policies effect meets the actual Direct invest of policy, Pavitt (1989), Breschi
R&D requirement (including R&D subsidy, staff (2000), Palmberg
technical persons, R&D welfare, R&D (2004), Holmen
motivation and support), and efficency, resource (2007)
firms responds policy depend
on it could promote
technology exchange and
R&D activities
7 Policy implementation Cooperate willing, Hadjimanolis (2001),
experience of cluster’s inter- distance, information Beneito (2003)
enterprise affects the response exchange frequency
process, the cluster’s firms
will follow adjacent firms
benefited
120 Mechanism of Firm’s Response to Innovation Policy 1133
analyze and clearly understand the range of agents and limitations is very
important.
(2) Firm’s size: It limits innovation activities and capabilities, firms with different
sizes have different attitude to policy response. Large scale firms tend to
actively respond to innovation policy. And SMEs prefer subsidies and tax
policy. The studies usually use the variables of assets, employees, resource,
etc.
(3) Entrepreneur’s attitude: It is a direct impact on whether a firm responds to
policy. The studies usually use the variables of firms’ diversification and
position in the network of government and firms.
(4) The results and income: It is a key factor to firms’ response policy. The studies
usually use the variables of policy funding amount, project limits, innovation
cost, Skillman incremental, patent and other results of incremental, new
product profitability, tax relief and other income.
(5) Innovation capability: Policy can help to enhance firms’ innovation capability
is an influencing key factor of firms’ response policy. The studies usually use
the variables of yield, utilization and conversion rate of resource and
knowledge.
(6) R&D Requirement: Innovation policy focuses on supporting firms’ R&D
innovation and R&D require meeting is also a factor of firms’ response. The
studies usually use the variables of direct invest of policy, R&D subsidy, staff
welfare, R&D efficiency and resource.
(7) Inter-Enterprise: To measure the interaction between cluster firms and imitate
innovation behaviors, cooperation and communication is so important. The
studies usually use the variables of cooperate willing, distance and information
exchange frequency.
120.3.2 Response Mechanism and Equations
The mechanisms of firm’s response to innovation policy echo model is summa-
rized, and each mechanism has its specific trigger conditions. In order to descript
clearly, the model based on the classic Holland trigger rules to establish the
matching pattern of the policy supply and firm’s needs. The echo model chro-
mosome structure contains five mechanisms as shown in Fig. 120.2.
(1) Selection: Cluster firm’s response to innovation policy is a continuous
‘stimulate-react’ process, which according condition matching to ensure firm’s
attitude (Nybakk and Hansen 2008). Entrepreneur is the main factor of
response. With resource-based theory, we think, the expected resource pro-
viding of policy influences entrepreneur whether respond the policy. On the
one hand, they expect the amount of resource responding will be more than
other activities. On the other, they also hope the benefits from responding will
satisfy next innovation activity. At the moment t in the process, for Ri, a firm
1134 Y. Zhang and C. Li
Tagging Control
Offensive Defensive Adhesion Selection Acquisition Conversion Evolution
Matching
Fig. 120.2 Echo model chromosome of firm’s response to innovation policy
t
has amount of resource ARi , resource acquisition from response is AtRi and from
others is MAtRi , resource of innovation use is A~ t , Innovation policy should
Ri
trigger two conditions as in
At
Dt1 ¼ MARti [ 1;
Ri
t1 A
AtR þA ~ t1 ð120:1Þ
Dt2 ¼ i
~
A
Ri
t
Ri
[1
Ri
Firms will select to respond the policy, as in
t1 A
AtRi AtRi þ A ~ t1
Ri Ri
Dt1 Dt2 ¼ [1 ð120:2Þ
MAtRi ~t
A Ri
It is emphasized that resource is not only cluster firm’s innovation guarantee but
the key point of response to policy. Compare to get income, resource acquisition
meeting innovation require is the prime purpose of firm response to policy. So
government needs to enhance resource supply in policy planning.
(2) Acquisition: Cluster firm’s response to innovation policy is related to gov-
ernment resource supply. The resource which firms select themselves contains
information (R1), funds (R2), projects (R3) and human resource (R4) (Kang
and Park 2012). At the moment t in the process, a firm has resource A t and
~ t t ~ t ~ tþ1
consumes A . When ARi ARi \ARi , it will decide to respond to policy.
Furthermore, because there is competition and cooperation relation between
firms in cluster, so resource could disseminate from one firm to another. The
MAtRi denotes resource acquisition from other firms. If we do not consider the
policy’s effect, a firm maximum requires A t A ~ t þ MAt at the t ? 1
Ri Ri Ri
moment. According to gaming theory, the government only needs to provide
resource more than t ? 1 moment above mentioned, the firm will respond to
innovation policy. The amount of resource is acquired from response as in
120 Mechanism of Firm’s Response to Innovation Policy 1135
~t ; A
AtRi ¼ ð1 kÞ½minðA t A
~ t þ MAt Þ ð120:3Þ
Ri Ri Ri Ri
Where, kdenotes the unexpected shrinkage of resource in the responding pro-
cess (0 k\1).
(3) Conversion: In the process of resource converses to innovative results and
income, different firm has different technology capability influencing con-
version ratio. The trend of innovative results is Logistic curve (Rodríguez-
Pose 1999). Traditional research shows that innovation results as in
Y ¼ aX C ð120:4Þ
where, adenotes technology results conversional coefficient, C denotes innovative
cost. It is classic that income converted is relationship with resource, innovative
capability and cost. However, the process is not a continuous and no technology
effect in formula (120.4). So we reference Gilbert and Katz’s standpoint to
improve the formula (Gilbert and Katz 2011). We assume the policy responding
process referred to n type’s resource and conversional time t becomes shorter and
shorter gradually with experience unceasing accumulating as in 1 enht ,
parameter h [ 0. Then, every innovative profit I(T) converted from resource
through technology T as in
8 1 9
< Z h i=
IðTÞ ¼ max a nhI ðT þ 1ÞeðnhþrÞt dt nc
n : ;
0 ð120:5Þ
nhIðT þ 1Þ
¼ max a nc
n nh þ r
Where, r denotes a changing parameter of resource convert to profits by
technology. To all of firms, if a 0, no response activity because of no profits. So
I [ 0 and a [ 0 is the condition of firms responding to innovation policy.
(4) Evolution: Firms as the agents who could realize evolution by experience
accumulating from the study process of response. There are three kinds of
evolution patterns based on firm’s size. The first is passive evolution, which
describes some firms depend on innovation policy intensively. Because they
need the subsidy of policy to operate, so they have to respond policy. The
process is a passive response, but after response their experience is accumulate
to a database. And different responding patterns to different policy. The second
is imitating evolution, which describes some firms depend on innovation
policy generally. In this style, response process is decided by entrepreneurs
who always follow large scale firms’ experience in the cluster. The third is
activity evolution, which describes some firms could select to respond policy
or not themselves. Firms with this style often have their policy analysis sector
which charges for statistic analyzing, evaluating the policy and establishing
indicators of response. Firms also need to give government a feedback. The
1136 Y. Zhang and C. Li
evolution process is often from pattern 1–3. Pe denotes the probability of
policy response experience as in
8
< 0\p\0:4 first
Pe ¼ p 0:4\p\0:7 second
:
0:7\p\1 third
The probability of experience growing is random, which describes firms
responding process is complex adaptive.
(5) Adhesion: By firm’s experience accumulating, there are two connection styles.
One names inertia connection, which describes some firms respond to inno-
vation policy sustained through experience judging. Another is one time
connection which describes some firms attempt to respond policy. The
advantages of inertia connection are directly response without energy con-
suming and reduce the cost (Horwitch and Mulloth 2010). But some reason as
qualification makes firm hasn’t response and opportunity cost increasing. The
selection of connection style is related to resource acquisition, conversion and
the mechanism of selection. We assume that innovation policy j is responded
by a firm, its sustained cost Cjk00 , attempt cost Cjk0 , and Cjk00 \Cjk0 . The intensity
of inertia and attempt connection separately is l00jk and l0jk (l00jk þ l0jk ¼ 1). If
unresponsive time t0 at t time slice in the actual response process and
Cjk00 t [ Cjk0 t0 . Then, reduce l00jk while increase l0jk . When the intensity equals
1, the process is stop. On the contrary, reduce l0jk while increase l00jk . If it
appears inertia connection continuously, firms and policy is adhesion.
120.3.3 Analysis of the Response Process
In the industrial cluster firms interaction is so frequency, the process of a firm
responding to innovation policy needs to consider other firms’ influences and
experience accumulation. The response may looks like a ‘stimulate-react’ and
gaming activity with multiple variables of seven key factors affecting as shown in
Fig. 120.3.
The matching mechanisms of policy response are a multi-agents process. The
matching driver reflects the needs meeting of each other. And, the judgment of the
trigger conditions should be based on agents’ situation changes.
1137
benefit
policy
policy
attract
could
could
from
firm
firm
The
The
the
Terminate Terminate
Game decision
Judging from
experience
Cognition
Terminate
-making
The effect of factors
120 Mechanism of Firm’s Response to Innovation Policy
Project demand
Transformation
Subsidy welfare
R&D efficiency
Innovation cost
Funds amount
Responding the policy
Invest Require
Qualification
Achievements
Information
Other firms
Cooperation
Utilization
Tax relief
Resource
Resource
Location
Tax relief
Liability
Willing
Subsidy
Profit
Profits
Asset
……
Profits
Risk
Staffs
……
……
Fig. 120.3 Innovation policy response process
Policy’s funds is attracting Responding effect is helpful
Firm meets policy’s demand Firm’s assets is ample Firm walks away from dilemma
The income expected is greats Promoting innovation capability
Resource is firm’s require
No
The responding risk is receivable Policy’s effect is obviously
Firm accords project’s condition Firms responding effect is good
Matching
condition
Policy’s subsidy is adequate Information exchange is extensive Giving impetus to firm R&D
Policy’s tax relief is moderate Firm is willing to R&D innovation Increasing firm’s performance
Firm wishes to cooperation Increasing profit by responding
Start
Yes
1138 Y. Zhang and C. Li
120.4 Conclusion
This study attempts to use echo model research on the response to innovation
policy. It puts forward research frame based on CAS theory and establishes Echo
model by response mechanisms analyzing. It is emphasized that firms in cluster
pay attention to policy resource supplying. And other firms’ experience also
influences the policy response. Cluster firms response is a continuous ‘stimulate-
react’ and gaming process. As experience accumulating, self-organizing evolution
makes firms’ policy cognitive capability and resource utilization increase gradu-
ally. This mechanism is important guarantee for benign response. On the one hand,
government should focus on firms’ requirement of innovation while cluster
development planning. It is best to let firms participate in policy-making for
ensuring the process to be impartial, opening and fair. Because of firms self-
adaptive, the requirement of resource could be controlled within a certain range to
achieve a balanced. It is not conducive to innovation with more and more resource,
so government should seek a reasonable resource range of policy allocating in
order to avoid the resource surplus. The other hand, the firm should avoid to blind
pursuit benefit of policies and encourage staffs to analyze policy. Meanwhile,
policy response could be incorporated into the innovative planning. Firms should
respond to policy according their capability, and acknowledge the pros and cons of
policy. The study also has some significance to government administration.
REFERENCES
Acs Z, Audretsch D, Feldman M (1992) Real effects of academic research: a comment. Am Econ
Rev 82:363–367
Audretsch DB, Lehmann EE, Warning S (2005) University spillovers and new firm location. Res
Policy 34(7):1113–1122
Beneito P (2003) Choosing among alternative technological strategies: an empirical analysis of
formal sources of innovation. Res Policy 32(4):693–713
Bessant JR (1982) Influential factors in manufacturing innovation. Res Policy 11(2):117–132
Breschi S, Malerba F, Orsenigo L (2000) Technological regimes and Schumpeterian patterns of
innovation. Econ J 110:388–410
Claver E, Llopis J, Garcia D, Molina H (1998) Organizational culture for innovation and new
technological behavior. J High Techol Manag Res 9:55–68
Colwell K, Narayanan V K (2010) Foresight in economic development policy: Shaping the
institutional context for entrepreneurial innovation. Futures 42(4):295–303
Coronado D, Acosta M, Fernandez A (2008) Attitudes to innovation in peripheral economic
regions. Res Policy 37:1009–1021
de Oliveira Wilk E, Fensterseifer JE (2003) Use of resource-based view in industrial cluster
strategic analysis. Int J Oper Prod Manag 23:995–1009
Fichman RG, Kemerer CF (1997) The assimilation of software process innovations: an
organizational learning perspective. Manag Sci 43:1345–1363
Filippetti A, Archibugi D (2011) Innovation in times of crisis: national systems of innovation,
structure, and demand. Res Policy 40:179–192
120 Mechanism of Firm’s Response to Innovation Policy 1139
Gilbert RJ, Katz ML (2011) Efficient division of profits from complementary innovations. Int J
Ind Organ 29(4):443–454
Grace TRL, Shen YC, Chou J (2010) National innovation policy and performance: comparing the
small island countries of Taiwan and Ireland. Technol in Societ 32(2):161–172
Hadjimanolis A (2000) A resource-based view of innovativeness in small firms. Technol Analy
Strat Manag 12:263–281
Holland JH (1995) Hidden order: how adaptation builds complexity (Helix Books), Addison-
Wesley Publishing Company, New York
Holmen M, Magnusson M, Mckelv M (2007) What are innovative opportunities? Indus Innov
14:27–45
Horwitch M, Mulloth B (2010) The interlinking of entrepreneurs, grassroots movements, public
policy and hubs of innovation: the rise of cleantech in New York city. J High Technol Manag
Res 21(1):23–30
Huang CY, Shyu JZ, Tzeng G H (2007) Reconfiguring the innovation policy portfolios for
Taiwan’s SIP Mall industry. Technovation 27(12):744–765
Jeroen PJ, de Jong, Freel M (2010) Absorptive capacity and the reach of collaboration in high
technology small firms. Res Policy, 39(1):47–54
Kang K, Park H (2012) Influence of government R&D support and inter-firm collaborations on
innovation in Korean biotechnology SMEs. Technovation 32:68–78
Kern F (2012) Using the multi-level perspective on socio-technical transitions to assess
innovation policy. Technol Forecast Social Change (Contains Special Section: Emerging
Technologies and Inequalities) 79:298–310
Mowery DC, Nelson RR, Sampat BN, Ziedonis AA (2001) The growth of patenting and licensing
by U.S. universities: an assessment of the effects of the Bayh–Dole act of 1980. Res Policy
30(1):99–119
Nybakk E, Hansen E (2008) Entrepreneurial attitude, innovation and performance among
Norwegian nature-based tourism enterprises. Forest Policy Econ 10:473–479
Palmberg C (2004) The sources of innovations—looking beyond technological opportunities.
Econo Innov New Technol 13:183-197
Paunov C (2012) The global crisis and firms’ investments in innovation. Res Policy 41(1):24–35
Pavitt K, Robson M, Townsend J (1989) Technological accumulation, diversification and
organisation in UK companies, 1945–1983. Manag Sci 35:81–99
Rodríguez-Pose A (1999) Innovation prone and innovation averse societies. Growth Change
30:75–105
Simpson RD, Bradford R L (1996) Taxing Variable Cost: Environmental Regulation as Industrial
Policy. J Environ Econom Manag 30(3):282–300
Souitaris V (2002) Technological trajectories as moderators of firm-level determinants of
innovation. Res Policy 31(6):877–898
Waarts E, Van Everdingen YM, Van Hillegersberg J (2002) The dynamics of factors affecting the
adoption of innovations. J Prod Innov Manag 19:412–423
Weber KM, Rohracher H (2012) Legitimizing research, technology and innovation policies for
transformative change: combining insights from innovation systems and multi-level
perspective in a comprehensive ‘failures’ framework. Res Policy (Special Section on
Sustainability Transitions) 41:1037–1047
Chapter 121
Modeling and Simulation
of Troubleshooting Process
for Automobile Based on Petri Net
and Flexsim
Wei-zhen Liao, Kai-hu Hou, Yu-jie Zheng and Xiao-ling He
Abstract In this paper, Yunnan Y Automotive Company is taken as a research
subject. Firstly, the troubleshooting methods from shallower to the deeper were
summarized to find the origin of the defects quickly and efficiently. Then, the
model of troubleshooting method was built by Petri net. What is more, the model
was corresponded to single-queue and multiserver model and the optimal value of
C corresponding to the optimal number of the quality engineer was calculated
adopting marginal solving method. Finally, the results of calculation were simu-
lated and verified using the simulation software Flexsim. Results indicate the
troubleshooting method can detect the fault source efficiently.
Keywords Fault source M/M/C queuing model Modeling and simulation
Petri net Troubleshooting
121.1 Introduction
Domestic studies on automotive failure: among the literatures, some scholars
(Song and Yao 2009) used thousands of car failure data about auto after-sales
service, from the perspective of statistical theory analysis to establish the reli-
ability of theoretical models; some scholars (Luo and Zhu 2005) used a new
pattern of support vector machine identification method to analysis and forecast
short-term after-sales failure data, it was more reliable; in addition, reasoning
W. Liao (&) K. Hou Y. Zheng X. He
Department of Industrial Engineering, Kunming University of Science and Technology,
Kunming, People’s Republic of China
e-mail: [email protected]
K. Hou
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1141
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_121,
Springer-Verlag Berlin Heidelberg 2013
1142 W. Liao et al.
method which was widely used in the automotive fault diagnosis expert system is
fault tree and fuzzy set theory (Kong and Dong 2001; Ji 2003; Su 2011). In
conclusion, most of the literature on the automotive failure analysis is simply
about the after-sales data, although the created model has a certain reference value,
but there is a strong lag; the fault tree, fuzzy set analysis methods can not be a
good handling with fuzzy and concurrency of car fault feature extraction; in
addition, these literature did not have simulation, their reliability needs to be
elegant.
To make a breakthrough of three aspects which are mentioned above, this paper
focuses on the process of cars production, to identify a process-oriented and
experience-oriented approach that can gradually find fault source; combines with
Petri Net and Flexsim to finish the modeling and simulation, handling with fuzzy,
parallelism and concurrency of cars fault; using queuing theory to make quanti-
tative analysis is more reliable.
121.2 About Petri Nets
Petri Net is a system model that uses P-element to represent the state, uses
T-element to represent the changes and associates resources (material, informa-
tion) flowing. Overall, it contains the state (Place), change (Transition), and flow.
So its mathematical definition (Su 2011) is a triple N = (P, T; F)
P = {p1, p2,…,pn} is Place set, n is the number of the Place;
T = {t1, t2,…,tm} is Transition set, m is the number of Transition;
F is a Set of ordered pair that consists of a P-element ana a T-element. And it
meets F ðP TÞ [ ðT PÞ:
The characteristics of Petri Nets are mainly reflected in two aspects: first, it is
realizability, the Petri Net systems must ensure that each Transition meets the laws
of nature, so it can be achieved; the most prominent feature of Petri Nets is suitable
for description and analysis of asynchronous concurrent systems on the various
levels of abstraction.
121.3 Car Troubleshooting Method
In order to solve quality problems, the old and the new seven tools of quality
management have been widely used in all aspects of business operations. These
methods in the practical application have their respective strengths and focus, but
they do not meet the authors’ requirements: to identify the sources of the failure
efficiently, step by step, from easy to difficult (Fig. 121.1 and Table 121.1).
According to the production of cars, combined with the experience of the staffs
of the Yunnan Y Automotive Company, a new troubleshooting method is born.
1143
framework. It is used to promote the problem-solving. It can be decomposed into
the following four steps: step 1—the correct process; step 2—the right tools; step
It is the most effective method to determine the defect source of a standardized
Operators know how Tools are able to
operate properly
to communicate
Right parts?
Need for craft
changes?
Preventive
Operators understand maintenance checks
quality requirements
Yes
Yes
Station layout
Whether there are efficiency
visual aids using appropriate error
Ending
No
No
proofing technology
Operators whether to
Troubleshooting for
Troubleshooting for
accept a correct Tool equipment
training devices when there is
no short circuit
3—the right parts; step 4—the need for craft changes.
third step
fourth step
protection
correct process
Fig. 121.3 Problems need to be shooting in the second step
Operators are frequent
121 Modeling and Simulation of Troubleshooting Process
Tools have an
Fig. 121.2 Problems need to be shooting in the first step
changing alternate system
Tools will be
Operators
right tools
connected to the dark
Fig. 121.1 Diagram for troubleshooting process
understanding the lamp system
product standard
Pneumatic tools
Right tools?
quality problems
Found the car
correct ?
connected to alternate
Process
Executing by operating air duct
requirements
Yes
Yes
Tools whether already
wear
enforcing the order of
operations
All shifts using the
No
No same tools
The correct guidance
documents
All tools are calibrated
Troubleshooting for
Troubleshooting for
Executive key point
second step
Tools are setting the
first step
right torque
Beginning Work standardization Use the right tools and
devices
1144 W. Liao et al.
right parts
cumulative tolerances
The need of changing
Fixed logistics routes
Is there material pull
The material on the
meet requirements
Parts existing of
Parts positioned
Supplier’s Parts
corresponding
right parts are
parts’ design
Part number
identified
properly
cards
Fig. 121.4 Problems need to be shooting in the third step
The need
for craft changes
execution of correct Assembly sequence need to increase the dark lights, error
craft in documents need to change proofing , interlock
Fig. 121.5 Problems need to be shooting in the fourth step
121.4 Modeling and Analysis Combine with Petri Nets
for Automotive Troubleshooting Process
121.4.1 Modeling with Petri Nets
Figure 121.6 is built models with basic Petri Net (Ren and Hao 2010; Su and Shen
2007; Xue et al. 2006). From Fig. 121.6, we know, nodes of the system are too
much and model is too huge (Bourjij et al. 1993). In order to better express the
logic relationship, so we references to the thought of Colored Petri Net (Wu and
Yang 2007): through introduction of color transition and replacement reduces the
complexity of system Petri Network model, making model intuitive and simple.
modeling by using CPN-tools (Vinter et al. 2003) simplified as Fig. 121.7.
From Fig. 121.7, readers can stick out a mile to the entire car troubleshooting
process; connection with Fig. 121.6, readers can get detailed reference, also can
grasp from the overall and specific well for the whole troubleshooting process.
121 Modeling and Simulation of Troubleshooting Process 1145
Table 121.1 The meaning of place and transition in Fig. 121.6
Meaning
Place elements
p1 Appear car quality problems
p11, p13, …, p119 No error state by searching 10 problems in Fig. 121.2
p12, p14, …, p120 Error state by searching 10 problems in Fig. 121.2
p2 The car still have failure after the first step
p21, p23, …, p215 No error state by searching 8 problems in Fig. 121.3
p22, p24, …, p216 Error state by searching 8 problems in Fig. 121.3
p3 The car still have failure after the second step
p31, p33, …, p325 No error state by searching 13 problems in Fig. 121.4
p32, p34, …, p326 Error state by searching 13 problems in Fig. 121.4
p4 The car still have failure after the third step
p41, p43, p45 No error state by searching 3 problems in Fig. 121.5
p42, p44, p46 Error state by searching 3 problems in Fig. 121.5
p5 The car trouble shooting, returned to normal
Transition elements
t11, t12, …, t110 Search the corresponding 10 problems in Fig. 121.2 one by one
t21,t22, …, t210 Solve the problem of the first step
t3 Check if the car can run normally
t41, t42, …, t48 Search the corresponding 8 problems in Fig. 121.3 one by one
t51, t52, …, t58 Solve the problem of the second step
t5 Check if the car can run normally
t61, t62, …, t613 Search the corresponding 13 problems in Fig. 121.4 one by one
t71,t72, …, t713 Solve the problem of the third step
t8 Check if the car can run normally
t91, t92, t93 Search the corresponding 3 problems in Fig. 121.5 one by one
t101, t102, t103 Solve the problem of the fourth step
t10 Start the car to show normal
Fig. 121.6 Modeling of p11 t 21 p21 t 51 p31 t 71
automotive troubleshooting t 11 p12 p22 p32
t 41 t 61
based on Petri Net p2
p1 t 12 p13 t 22 t 42
p23
t 52 p3 t 62 p33 t 72
p14 t3 p24 t5 p34
…
…
…
p11 p21 p32
…
…
…
t 110 p912 t 48 p521 t 613 p532
0 6 6
t 210 t 58 t 713
t8
p4
p5 t 93 t 92 t 91
p46 p45 p43
p41
p44 p42
t 10
t 103 t 102 t 101
1146 W. Liao et al.
Fig. 121.7 Modeling of automotive troubleshooting based on Colored Petri Net
first quality second quality third quality fourth quality
engineer engineer engineer engineer
Car Of
1 1 1 1 troubleshooting is
Fault car completed
...
...
...
...
n m h k
Fig. 121.8 Queuing model of car troubleshooting
121.4.2 With the Queue Theory Knowledge to Determine
the Optimal Parameters of the Model
(Service Desk C)
In Petri Nets model, if regarding problems that every step need to search as object
(customers) waiting for service, the quality engineer as the service desk, and more
than one quality engineer can work simultaneously, then the whole process can be
simplified into the following multi-server single queue tandem queuing system
model:
Assuming that the arrival of customers (fault cars) obey the Poisson, the quality
engineer‘s checking time obeys negative exponential distribution, makes each step
independent, then this model is the M/M/C queuing model. In the situation that the
probability of cost and fault car basically turns to stabilized, to determine the
optimal number of service desk can reduce costs and maximize the benefits.
Seeking the optimal number of service desk is obtained in Fig. 121.8 n, m, h, k.
To improve the reliability and accuracy of the calculation, using the following
references (Ai et al. 2007) to seek the method of M/M/C model optimal service
desk number C.
In steady-state case, the expectations of the unit time full cost (service costs and
waiting costs):
z ¼ c0s c þ cw L ð121:1Þ
where c is the number of service desk; c0s is the unit time cost of each desk; cw is
the unit time costs of each customer stay in the system; L is the system’s customer
121 Modeling and Simulation of Troubleshooting Process 1147
average number Ls or the queue’s customer average Lq , the service desk number
setting up has the deep impact on it. Because c0s and cw can get the statistics
through the actual situation, so (121.1) is the function zðcÞ about c, the purpose is
through getting the optimal solution c makes zðcÞ minimize.
c is an integer, using the marginal analysis method:
z ð c Þ z ð c 1Þ
ð121:2Þ
z ð c Þ z ð c þ 1Þ
Substituting ‘z’ of formula (121.1) into formula (121.2), then
0
cs c þ cw Lðc Þ c0s ðc 1Þ þ cw Lðc 1Þ
ð121:3Þ
c0s c þ cw Lðc Þ c0s ðc þ 1Þ þ cw Lðc þ 1Þ
Simplify formula (121.3), then
0
.
Lðc Þ Lðc þ 1Þ cs cw Lðc 1Þ Lðc Þ ð121:4Þ
According to the related analysis to thousands of cars failure data of Yunnan Y
Automotive Company collected at the scene, the car failure source in the four steps
of the searching, which belongs to the first step is about 25 %, the second step is
about 40 %, about 30 % is the third step, the last step is only 5 %. Due to the
influence of external factors, every step’s average service time is different: 240,
300, 300, 600 s, and obey exponential distribution.
Yunnan Y Automotive Company’s internal data shows that: the costs of each
searching incurred due to delays in other processes about 8 Yuan, the services
costs to each time set an quality engineer (salaries and equipment wear and tear)
for about 37 Yuan per hour.
121.4.2.1 The Determination of the Numbers of the Optimal Service
Desks in the First Step
According to the statistics, in the first step, the fault cars arrived time obeys the
average arrival rate of 26 times per hour to the Poisson distribution; service time is
negative exponential distribution which average service rate is 15 times per hour.
So, c0s ¼ 37 Yuan/quality engineer, cw ¼ 8 Yuan/time, k ¼ 26, l ¼ 15,
k=l ¼ 1:73, assuming the number of quality engineers to c, makes c respectively
as 1, 2, 3, 4, 5. According to the Wq l value of multi-server desks (Ai et al. 2007),
with linear interpolation algorithm to find the corresponding value of Wq l, as
shown in the Table 121.2.
Substituted Ls into the formula (121.4), obtained the following data in
Table 121.3 by the formula (121.1):
From Table 121.3, 128.09 Yuan is the lowest total cost, the corresponding c
equals to 3, so the result coming out is the lowest cost that needs to set three
quality engineers.
1148 W. Liao et al.
Table 121.2 Average number of customers (L) in the system
c 1 2 3 4 5
k=cl 1.730 0.865 0.577 0.433 0.346
Wq l – 3.393 0.235 0.054 0.012
Ls ¼ lk Wq l þ 1 – 7.600 2.136 1.823 1.751
Table 121.3 Data calculated by marginal analysis
The number of quality The number of car waiting L(c) - The total costs
engineers c to search Ls(c) L(c ? 1)*L(c) - (every hour) z(c)
L(c - 1)
1 ? ? ?
2 7.6000 5.4640–? 134.80
3 2.1360 0.313–5.4640 128.09
4 1.8230 0.116–0.313 162.58
5 1.7510 – 199.01
121.4.2.2 The Determination of the Numbers of the Optimal Service
Desks in the Second Step
Easy to know that fault cars reached in the second step still obey to Poisson
distribution (Winston 2004), if N is Poisson random variables, then
EðNÞ ¼ var N ¼ k. So the fault cars from the first step into the second step reduce
25 %, then k ¼ 26 75 % ¼ 19:5 20, service time also obeys negative expo-
nential distribution, l ¼ 12 times per hour.
It can be gained the smallest total cost is 109.30 yuan, corresponding to the
optimal service desks c = 2, this step sets two quality engineers is best.
121.4.2.3 The Determination of the Numbers of the Optimal Service
Desks in the Third and the Fourth Step
Similarly, it is known that the fault cars reached in the last two steps obey Poisson
distribution with k ¼ 26 35% ¼ 9:1, k ¼ 1:3 respectively; service time also
obeys index distribution, Service rate are l ¼ 12, l ¼ 6 times per hour. Because of
the average arrival rate k less than service rate l, so the last two steps only need to
set up 1 quality engineer.
According to the above theoretical calculation result, it is known that first step
should set three quality engineers, the second step shall set up two quality engi-
neers, the last two steps shall just set one quality engineer that can make the benefit
of the system to achieve optimal.
121 Modeling and Simulation of Troubleshooting Process 1149
Fig. 121.9 Modeling plan
121.5 Flexsim Simulation Analysis
121.5.1 Modeling and Parameter Settings
According to the previous section, fault car’s arrival interval time is about 138 s,
average service time is exponentially distributed, their means, respectively, as 240,
300, 300, and 600 s (Chen et al. 2007). Using Flexsim to modeling, the plan is
shown in the Fig. 121.9.
As shown in Fig. 121.9: using a Source as fault car’s generator; four Queues to
achieve car’s cache; Processor represents quality engineer, there are different
proportion of car finished troubleshooting inflow to Sink, so uses three Flow Node
to provide path, achieve the shunting of the first three steps.
121.5.2 Analysis of Simulation Data
121.5.2.1 Confirm the Simulation Time
According to the work schedules of Yunnan Y Automotive Company, the com-
pany belongs to the mode of two shifts 1 day, 8-hour one shift, two for 16 h, so the
length of simulation time is 16 h.
121.5.2.2 Confirm the Simulation Method
Using independent replications (Lin 2003) to simulate, namely, use of different
random variables and select independent starting state to simulate R times.
1150 W. Liao et al.
121.5.2.3 Confirm the Simulation Times of Established Interval
Utilization of different random variables, after R simulations, generated R
observed values (average values ^hr ), using the R observed values to conduct point
estimation values is as follows:
^ 1X R
^
h¼ hr ð121:5Þ
R r¼1
The sample standard deviation:
XR
ð^
hr ^
hÞ2
S2 ¼ ð121:6Þ
r¼1
R1
The estimate value ^
h:
XR 2
S2 1
^ 2 ð^
r hÞ ¼ ¼ ^hr ^h ð121:7Þ
R ðR 1ÞR r¼1
In order to confirm the optimal times of simulation, assuming that a parameter’s
interval estimation within the scope of a particular, half-length interval is less than
a certain value e, we need to simulate at least R times to meet the required half
interval length. Just starts our first R0 simulations, general 4–5 times. Through
formula (121.6) obtains standard deviation S0 , when the half-length interval less
than e, can be expressed as follows:
^ ð^
h:l: ¼ ta=2;R1 r hÞ e ð121:8Þ
Formula (121.8) substituted the formula (121.7), then:
ta=2;R1 S0
R ð121:9Þ
e
Owing to ta=2;R1 za=2 (za=2 is the a=2 quantile of the standard normal dis-
tribution), so R meets the following minimum integer, and R R0 , then:
za=2 S0 2
R ð121:10Þ
e
Utilization of formula (121.5) to find out the point estimation value of each
Processor in 5 different random variables simulation is shown in Table 121.4. It is
clear that each Processor’s utilization remains relatively low, so the three Pro-
cessors of the first step should be reduced to two, then analysis the simulation. The
data obtained from simulation analysis is shown in Table 121.5:
Five simulations is shown in Table 121.5, the data gained from Table 121.5
substituted the formula (121.6) to (121.10), followed by count r ^2 ð^hÞ, S20 , ðz0:025e S0 Þ2 .
121 Modeling and Simulation of Troubleshooting Process 1151
Table 121.4 Simulation data of first model
Times Processor
1 2 3 4 5 6 7
1 (%) 77.70 65.50 55.74 88.72 83.39 71.33 16.93
2 (%) 72.02 58.28 47.33 80.79 73.68 82.70 21.47
3 (%) 72.04 61.21 45.20 77.87 69.90 72.37 22.05
4 (%) 69.53 60.00 47.78 79.66 71.15 77.03 20.40
5 (%) 70.64 62.62 54.54 80.83 71.91 81.58 11.21
Mean 72.39 61.52 50.12 81.57 74.01 77.00 18.41
Table 121.5 Simulation data of second model
Times Processor
1 2 3 4 5 6
1 (%) 89.66 81.84 78.49 68.69 70.84 21.86
2 (%) 99.28 98.25 82.59 81.25 90.40 23.22
3 (%) 91.05 86.26 84.17 77.84 70.32 22.06
4 (%) 90.47 86.47 79.69 70.79 73.27 21.29
5 (%) 79.83 71.00 78.31 69.92 70.23 25.33
Means (%) 90.06 84.76 80.65 73.70 75.01 22.75
Table 121.6 Simulation times had been determined
^ ðE 03Þ
^2 ðhÞ
r 0.96 1.93 0.14 0.61 1.51 0.05
S20 ðE 03Þ 4.77 9.63 0.68 3.05 7.56 0.26
z0:025 S0 2 7.34 14.8 1.05 4.69 11.6 0.40
e
Simulation times R 8 15 5 5 12 5
Ultimately to determine how many times need to be simulated if 95 % processor
busy rates of the confidence interval between ±0.05. The result is shown in
Table 121.6:
In the whole model, the highest simulation times determine the final simulation
times. From Table 121.6, knowing that should simulate more than 12 times, if we
takes 15 times, 10 more times need to be simulated to insure that processors’ busy
rates could be drawn between the confidence interval ±0.05. The resulting data of
the entire simulation process is shown in Table 121.7.
Confidence intervals can know from Table 121.7, half-length interval of 95 %
confidence interval for all processors is less than 0.05. Finally it is turned out to be
confidence in the 95 %: busy rates of Processor1 will fall in the interval
0.8843 ± 0.0425; busy rates of Processor2 will fall in the interval 0.8321 ± 0.0363;
busy rates of Processor3 will fall in the interval 0.8126 ± 0.0375; busy rates of
Processor4 will fall in the interval 0.7387 ± 0.0421; busy rates of Processor5 will
1152 W. Liao et al.
Table 121.7 Data analysis of the whole simulation process
Value Processor
1 2 3 4 5 6
Mean 88.43 83.21 81.26 73.87 72.58 21.35
Standard deviation 0.0549 0.0797 0.0392 0.0544 0.0739 0.0447
Confidence interval ±0.0425 ±0.0363 ±0.0375 ±0.0421 ±0.0407 ±0.0428
fall in the interval 0.7258 ± 0.0407; busy rates of Processor6 will fall in the interval
0.2135 ± 0.0428.
Combined with the calculation of third section and results of two different
models’ simulation, in the first step two quality engineers are more reasonable; the
efficiency of Processor6 is still low, to further improve the efficiency and benefits
of system, should train the technology-packed quality engineers. The problems of
the last two steps will be finished by one quality engineer.
121.6 Conclusion
Automotive troubleshooting method that is described in this paper is a more
universal method to find out the fault source. It is from simple to complex, is
trending to standardization procedures to solve problems.
Modeling with Petri Net and Colored Petri Net, from global and local trou-
bleshooting on cars have a good control and master; effectively solve the parallel
reasoning of the car fault detection system; reduce the complexity of the fuzzy
inference reasoning method.
And collecting data in real-time from car production process, the use of queuing
theory related knowledge obtains the optimal number of quality engineers.
Modeling and Simulation with Flexsim have great reference value for reason-
able arrangements to quality inspectors. Using of the troubleshooting method
makes all activities of the enterprises with strong timeliness.
References
Ai Y et al (2007) Operations research, vol 11. Tsinghua University Press, Beijing, pp 336–337 (in
Chinese)
Bourjij A, Zasadzinski M, Darouach M, Krzakala G, Musset M (1993) On the use of hybrid Petri
Nets for control process safety: application to a steam2boilers network simulator. In: IEEE
international conference on systems, man and cybernetics, no. 2
Chen G, Wu H, Chen Y (2007) Industrial engineering and system simulation, vol 6. Metallurgical
Industry Press, Beijing, pp 79–253 (in Chinese)
Ji C (2003) Development of automotive diagnostic system based on fault tree. Veh Power
Technol 1:52–57 (in Chinese)
121 Modeling and Simulation of Troubleshooting Process 1153
Kong F, Dong Y (2001) Failure diagnosis and reasoning based on fault tree knowledge. Automot
Eng 23(3):209–213 (in Chinese)
Lin Z (2003) Theory and application of system simulation. Press by Canghai Bookstore,
Nanchang, Jiangxi, China, p 357 (in Chinese)
Luo Z, Zhu S (2005) A new e-insensitivity function support vector inductive regression algorithm
and after-sales service data model forecast system. Comput Sci 32(8):134–141 (in Chinese)
Ren J, Hao J (2010) Petri network-based modeling analysis. J Xi’an Aerotech College
28(3):50–52 (in Chinese)
Song L, Yao X (2009) Research on probability model of vehicle quality based on the broken-
down number per thousand cars. J Chongqing Technol Business Univ 26(6):543–547 (in
Chinese)
Su C (2011) Modeling and simulation for manufacturing system. Mechanical Industry Press,
Beijing, p 120 (in Chinese)
Su C, Shen G (2007) Development for system reliability modeling and simulation based on
generalized stochastic Petri net (GSPN). Manuf Inf Eng China 36(9):45–48 (in Chinese)
Vinter RA, Liza W, Henry Machael L, et al (2003) CPN tools for editing, simulting, and
analysing coloured petri net. In: Proceeding of the applications and theory of petri nets
presented at the 24th international conference, Eindhoven, The Netherlands
Winston WL (2004) Operations research introduction to probability models, 4th edn. A Division
of Thomson Learning Asia Pte Ltd, Belmont, pp 333–336
Wu H, Yang D (2007) Hierarchical timed coloured petri-net based approach to analyze and
optimize medical treatment process. J Syst Simul 19(4):1657–1699 (in Chinese)
Xue L, Wei C, Chen Z (2006) Modeling design and simulation of hybrid systems review and
analysis. Comput Simul 23(6):1–5 (in Chinese)
Chapter 122
Modeling and Simulation of Wartime
Casualty Surgical Treatment
Kai Zhang, Rui-chang Wu, Yun-dou Wang, Xiao-feng Zhang
and Hai-jian Du
Abstract The objective of this paper is to model and simulate the wartime casualty
surgical treatment with a discrete simulation tool (Simio) based on treatment process
analysis and medical data. Firstly, the surgical treatment process is analyzed. Then, a
3D visual simulation mode is built with Simio. Seven scenarios about different
casualty arrival rates are used to test the surgical capability of the field hospital of the
PLA. The results show that two hundred casualties may reach the maximum
throughput in the field hospital equipped with one operation table. The modeling and
simulation of wartime casualty surgical treatment contributes to obtaining the
system performance indicators, and simulation model developed can support
medical resources estimation and allocation optimization.
Keywords Casualty Modeling Simulation Wartime
122.1 Introduction
Warfare has changed significantly in modern time. Range and accuracy of the
lethal modern weapon systems are far more effective than ever, and the army has
transformed into modular units that are smaller, more deployable and flexible.
K. Zhang (&) R. Wu Y. Wang X. Zhang H. Du
Institute of Medical Equipment, Academy of Military Medical Sciences, Tianjin, China
e-mail: [email protected]
R. Wu
e-mail: [email protected]
Y. Wang
e-mail: [email protected]
X. Zhang
e-mail: [email protected]
H. Du
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1155
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_122,
Ó Springer-Verlag Berlin Heidelberg 2013
1156 K. Zhang et al.
The field characteristics of dispersion, rapid military operations, increased
mobility, rapid task organization, and extended lines of communications make the
battlefield more lethal than ever. These directly impact the medical service.
The medical facility must adjust to these impacts (Nuhut and Sabuncuoglu 2002).
As known to all, the operating room (OR) is the most demanding department in
field hospital of the PLA. The process of the operating room directly influences the
treatment efficiency of the medical treatment facility. Since the OR treatment
process is dynamic and contains many stochastic elements, simulation is used in
this research to model and analyze the related problems.
The objective of this paper is to model and simulate the wartime casualty
surgical treatment. Firstly, the surgical treatment process is analyzed. Then, a 3D
visual simulation mode is created with Simio, a quite new simulation platform.
Seven scenarios about different casualty arrival rates are used to test the surgical
capability of the medical aid station. The results show that two hundred casualties
may reach the maximum throughput in the field hospital equipped with one
operation table. The modeling and simulation of wartime casualty surgical treat-
ment contributes to obtaining the system performance indictors, and simulation
model developed can support medical resources estimation and allocation
optimization.
122.2 Casualty Surgical Treatment Process
Casualties are treated at medical facilities organized into a series of echelons in
wartime. The facilities at the forward echelons have the greatest mobility but least
surgical capability (Fleet marine force manual (FMFM) 1990). Each facility
establishes some treatment areas and treats casualties based on treatment range,
rules and capability, which are defined by treatment rules of the army. The field
hospital, equipped with necessary operation resources, has the surgical capability.
The casualty surgical treatment process in the field hospital is shown in Fig. 122.1.
When casualties arrive at the facility, they are distributed to different treatment
areas after triage. The casualties, who immediately need operation disposition, are
sent to preoperative room, and the others, to the areas of Lab, X-ray, serious or
minor injury treatment, etc. In addition, some casualties flow between these areas
and could then get to the preoperative room. When the personnel and equipment
Arrival of Preoperative Operationg Postoperative
patient Triage Room Room Ward
Other treatment
areas
Fig. 122.1 Flowchart of patient movement in the field hospital
122 Modeling and Simulation of Wartime Casualty Surgical Treatment 1157
required to treat the casualty are available, the casualties are delivered to OR and
then flow to postoperative ward.
In the OR, the casualties would receive operation disposition based on their
traumatic conditions. The treatment process in the OR could be considered as a
series of treatment tasks connected with each other, which could then be named
operation treatment task sequence (Zhang and Wu 2011a). There are 2 types of
treatment tasks according to the relative order between each other:
(1) Sequential tasks are those performed one after another.
(2) Concurrent tasks are those completed simultaneously.
The operation treatment task sequence is shown in Fig. 122.2. This task
sequence is obtained by literature investigation and expert consultation.
So, a casualty surgical treatment process could be considered as this casualty
flowing through the above operation treatment task sequence and all casualty
surgical treatment processes actually make up this sequence.
Basic consumable
preparation
Basic operating Equipment
preparation
Anesthesia equipment OR team
Treatment area setup
preparation preparation
Basic OR instruments
preparation
Operation table preparation
Infusion
blood
... Patient preparation
... ... in the table
Infusion
fluids
Clean and prepare
instruments
Treatment Transfer
The end
paper patient
Room clean
Fig. 122.2 Treatment task sequence of an operating room
1158 K. Zhang et al.
122.3 Modeling and Simulation
122.3.1 Simulation Scenario
The simulation scenarios provided are a typical medical support context in which a
field hospital of the PLA provides emergency treatment to the casualties involved
according to the treatment rules defined by the PLA. In the baseline mode, each
patient arrives randomly with an exponential time between arrivals with a mean of
14 min. They would receive first aid treatment within 10 min after injury and then
be evacuated to the regiment aid station and get to the preoperative room randomly
following an uniform distribution with parameters 0.5–3.5 h. This time internal is
defined by the treatment rules of the PLA. There is 1 operation table equipped with
necessary medical personnel and resources in the OR. In order to test the OR
capability, the casualty arrival rate would successively increase by 25 % in other
scenarios.
122.3.2 Modeling and Simulation Tool
Selection of a proper modeling and simulation tool is critical to the outcome of
data needed for analysis. In this paper, Simio is selected as the modeling and
simulation platform for its various advantages. Simio is a quite new simulation
tool, which has functions of visual, interactive, and interpretative modeling. Using
Simio, modeling is based on describing system’s objects and evolution of system
behavior by interaction of these objects. Its graphics and extended capabilities are
able to help the researchers easily model the system and determine how an existing
or a proposed system will perform. In addition, Simio provides the most advanced
real-time 3D technology, which strengthens the interaction of simulation (Zhang
and Wu 2011b; Pegden 2008; Dennis Pegden 2009).
122.3.3 Evaluation Measures for Modeling and Simulation
The main focus of modeling and simulation is to valuate the system’s surgical
treatment capability. For the system, the average casualty wait length and time for
operation, operation time, and mortality rate must be accepted by the treatment
rules. Since the model developed is a baseline model, and only the casualty arrival
rate is changed in other scenarios, the same metrics to measure the performance of
system would be used. This allows us to collect similar data in each of the
simulations and compare data obtained from several runs of the simulation. Once
the data are collected, statistical analysis is performed and the results are used in
the analysis of different allocation of the operation room.
122 Modeling and Simulation of Wartime Casualty Surgical Treatment 1159
122.3.4 Medical Parameters
122.3.4.1 Casualty Types and Generation
The casualty types in this simulation research mostly come from the U.S. army
Deployable Medical System (DEPMEDS) PC Code and are adjusted by the subject
experts of the PLA. These PC codes occur during deployment and combat
operations and range from snake bites, to severe hearing impairment, to more
serious injuries (James et al. 2005; Deployable Medical System (DEPMEDS)
2003). The casualties needing operation treatments involve in 87 PC codes. In the
simulation, casualties are randomly generated based on an exponential distribu-
tion. The casualty cumulative probability distribution obtained from historical
accounts of ground operations and adjusted by factors such as recent of operation
and medical advances is used for simulation model to indentify a certain PC Code
for each injury event.
122.3.4.2 Casualty Survival Probability
The wartime casualty survival probability data are obtained by expert question-
naires. After preliminary analysis, casualties are identified and designated as
having either a high (H), medium (M), or low (L) risk of mortality according to the
severity of life-threatening. In addition, the casualty survival probability data are
fitted by the Weibull survival function with MATLAB. Then the survival functions
based on types of medical treatment facility and treatment delays are obtained
(Zhang and Wu 2011; Mitchell et al. 2004).
In a certain medical treatment facility, the casualty survival model based on a
treatment delay would be obtained by the functions known. A certain type of
casualty starts treatment at c0, and this time point is between c1 and c2
(c1 \ c0 \ c2), then the casualty survival model based on c0 treatment delay is:
SðtÞc0 ¼ Pr½T [ t
¼ ððc0 c1 ÞÞ exp ðt=a1 Þ^ b1
þ ððc2 c0 Þ=ðc2 c1 ÞÞ exp ðt=a1 Þ^ b2 ð122:1Þ
Using this model and the function parameters fitted, a certain type of casualties’
survival probability at any point and time during their treatment processes could be
obtained.
122.3.4.3 Casualty Treatment Data
The casualty treatment process is a continuous one composed of a series of
treatment tasks required to treat that specific type of casualty. Each type of
casualty is linked to a set of treatment tasks, and each treatment task is linked to
the resources required to accomplish that task. These tasks could be connected
1160 K. Zhang et al.
together based on their relative order and compose a casualty treatment sequence.
Actually, each casualty treatment task sequence is a subset of the OR treatment
task sequence. When this casualty arrives at the operating room, he/she would flow
through the OR treatment task sequence. The treatment task sequences are mostly
obtained from the U.S. army treate file. The treatment time, personnel and nec-
essary equipment and supplies are obtained by consulting with experts and
researching books.
122.3.5 Modeling Process
122.3.5.1 Modeling Casualty Treatment Task Sequence
The treatment process could also be considered as a series of treatment tasks
connected by a series of junctions. A new junction object from scratch and a
treatment task object sub-classed and redesigned from the standard Time Path
object are developed with Simio (Jeffrey and Roberts 2011; Dennis Pegden 2009).
Linking the junction and treatment task objects together and setting the object
properties, the casualty treatment task sequence is developed as shown in
Fig. 122.3.
Fig. 122.3 Casualty treatment task sequence in Simio
122 Modeling and Simulation of Wartime Casualty Surgical Treatment 1161
Fig. 122.4 Typical treatment task logic
122.3.5.2 Modeling Resources Consumed by Casualty Treatment Task
The medical personnel are modeled with Worker object, medical equipment with
Resource object, and time consumed by the task with triangle distributions got by
consulting with subject experts. The logic of each treatment task is designed using
the graphical process flows. Figure 122.4 depicts typical treatment task logic.
When a casualty flows to this task process, the Search and Decide steps are used to
decide whether this task is required by the casualty from a casualty treatment task
data table which would be described below. The Set Row and the next Decide
steps are used to link the task to the required resources in the table. Then, the
Seize, Delay and Released step are used together to model the resources to be
seized, delayed and released (Simio user’s manual 2009; Dennis 2009).
122.3.5.3 Modeling Casualty Survival Situation
There are four types of risk of mortality and five types of internal of treatment
delay. Each type of mortality risk and internal of treatment delay is distinguished
by the Decide step. The survival model is used following the next Decide step to
determine the casualty survival situation. If the casualty is still alive, he/she would
then flow the next treatment process. These logics are shown in Fig. 122.5.
122.3.5.4 Setting Simulation Data
In addition to entering data directly into the modeling objects, a casualty table,
including casualty types, composition of proportions, litter conditions, treatment
chances and priorities, is defined to set all casualties’ basic information, and a
treatment task table, including casualty types, task types, task time and treatment
probabilities, is defined to set all casualties’ treatment information, which is shown
in Fig. 122.6.
1162 K. Zhang et al.
Fig. 122.5 Casualty survival situation modeling logic
122.3.5.5 Achieving Simulation Results
Though Simio platform has powerful statistical functions and makes most statis-
tical data automatically, the surgical system still needs some special data statistics.
So, some statistic elements are created to record the treatment data in Simio, as
shown in Fig. 122.7.
In addition, some process logics, accompanied with statistical elements, are
created to trace the simulation data.
122.3.5.6 Visual Simulation Model of Casualty Treatment
The 3D casualty, medical personnel, equipment, and operating room objects are
developed by 3D modeling software and imported to create the realistic 3D
casualty treatment model with Simio, which is shown in Fig. 122.8 (Dennis 2009).
122.3.6 Results and Discussion
Seven scenarios about different casualty arrival rates are built in the experiment
window within Simio. The simulation time lasts for 34 h, and the first 10 h is not
122 Modeling and Simulation of Wartime Casualty Surgical Treatment 1163
Fig. 122.6 Casualty treatment data tabe
Fig. 122.7 Statistic elements created to record simulation data in Simio
1164 K. Zhang et al.
Fig. 122.8 Simio model with animation 3D
used for collecting data. The scenario 1 is the baseline model, and the arrival rate is
increased by 25 % after each scenario. We take 100 replications for each scenario
and the results are within the 95 % confidence interval.
All the important performance indicators of the system are obtained. Parts of
mean value of average data are shown in Table 122.1.
Two hundred casualties have long been considered as the maximum throughput
in the field hospital researched in this paper. Though, long time has passed, the
performance indicators still reflect this situation. As shown in Table 122.1, the
Table 122.1 Performance indicators of the field hospital with one operation table
Scenario Casualty Mortality Operation Wait length for Wait time for Operation
arrival rate (%) number operation operation (h) time (h)
number
1 103.46 1.42 8.40 0.25 0.53 1.23
2 128.21 1.48 10.50 0.46 0.74 1.24
3 160.89 1.99 12.94 0.89 1.20 1.20
4 201.07 1.89 14.83 1.37 1.48 1.21
5 249.97 2.60 16.78 2.81 2.25 1.23
6 314.26 3.03 18.26 4.97 3.06 1.25
7 392.91 3.15 18.41 9.48 3.59 1.29
122 Modeling and Simulation of Wartime Casualty Surgical Treatment 1165
average casualty arrival number is 201.07 ± 2.72 in scenario 4. In this situation,
after consulting with subject experts, we consider that the wait length and wait
time for operation, and the mortality rate may reach the maximum, which could be
borne by the casualties. In scenario 5, 6 or 7, more operation tables should be
established.
122.4 Conclusion
The objective of this paper is to model and simulate the wartime casualty surgical
treatment. Firstly, the surgical treatment process is analyzed. Then, a 3D visual
simulation mode is built with Simio simulation platform. Seven scenarios about
different casualty arrival rates are used to test the surgical capability of the medical
aid station. The results show that two hundred casualties may reach the maximum
throughput in the field hospital equipped with one operation table. The modeling
and simulation of wartime casualty surgical treatment contributes to obtaining the
system performance indictors, and simulation model developed can support
medical resources estimation and allocation optimization.
References
Dennis Pegden C (2009) Now bring simulation in-house to support good decision making. (2009-
12-19) [2010-02-10]. http://www.simio.com/resources/white-papers/bring-simulation-in-
house/Now-Bring-Simulation-In-house-to-Support-Good-Decision-Making.htm
Dennis Pegden C (2009) Introduction to Simio for beginners. (2009-12-19) [2010-02-10] http://
www.simio.com/resources/white-papers/Introduction-to-Simio/introduction-to-simio-for-
beginners-page-1.htm
Dennis Pegden C, How Simio objects differ from other object-oriented modeling tools [EB/OL].
(2009-12-19) [2010-02-10] http://www.simio.com/resources/white-papers/How-Simio-
Objects-Differ-From-Others/how-simio-objects-differ-from-other-object-oriented-modeling-
tools.htm
Deployable Medical System (DEPMEDS) PC codes (2003), The PCs and their accompanying
treatment briefs are updated on a quarterly basis by the Joint Readiness Clinical Advisory
Board (JRCAB). For the most up-to-date information on PCs and specific treatment briefs,
refer to the JRCAB website at: http://www.armymedicine.army.mil/jrcab/d-prod.htm
Fleet marine force manual (FMFM) (1990) 4–50, Health Service Support. Department of the
Navy, Headquarters United States Marine Corps, Washington
James M, Zouris G, Walker J (2005) Projection of patient condition code distributions based on
mechanism of injury. ADA434281, Naval Health Research Center, San Diego
Jeffrey AJ, Roberts SD (2011) Simulation modeling with SIMIO: a workbook. Smiley Micros,
Pennsylvania November pp 203–222
Mitchell R, Galarneau M, Hancock B, Lowe D (2004) Modeling dynamic casualty mortality
curves in the tactical medical logistics (TML+) planning tool. Naval Health Research Center,
San Diego, p 8
Nuhut O, Sabuncuoglu I (2002) Simulation analysis of army casualty evacuations. Simulation
8(10):612–625
1166 K. Zhang et al.
Pegden D (2008) SIMIO: a new simulation system based on intelligent objects. In: Proceedings of
the 2008 winter simulation conference, Institute of Electrical And Electronics Engineers, New
Jersey, pp 2293–2300
SIMIO LCC (2009), Simio user’s manual (2009-12-19) [2010-02-10]. http://www.simio.com/
resources.htm
Zhang K, Wu R (2011a) Using Simio for wartime casualty treatment simulation. In: The 3rd
IEEE international symposium on it in medicine and education, IEEE Press, December,
pp 315–318
Zhang K, Wu R (2011b) Using simio for wartime optimal placement of casualty evacuation
assets. In: The 3rd international symposium on information engineering and electronic
commerce, ASME Press, July, pp 277–280
Zhang K, Wu R (2011) Research on survival model of casualty in wartime. In: The 3rd IEEE
international symposium on it in medicine and education, IEEE Press, pp 625–628
Chapter 123
Modeling of Shipboard Aircraft
Operational Support Process Based
on Hierarchical Timed Colored Petri-Net
Ting Wang, Bo-ping Xiao, Lin Ma and Yan-kun Tian
Abstract The operational support of shipboard aircraft is a complicated process,
reasonable planning of aviation support resources and reduce the operational
support time plays a vital role of the tasks persistent of the aviation support system.
The paper introduces the basic principles of Petri nets, points out the defects of the
basic Petri net in the operational support process modeling and puts forward a
simulation model of shipboard aircraft operational support process based on
Hierarchical Timed Colored Petri-Net. The approach simplifies the hierarchical
modeling, make up for the lack of a Petri net time performance analysis, and
distinguish the different operational support resources. The operational support
process of shipboard aircraft and the modeling method based on HTCPN has been
given in this paper. It shows the application of Petri net in the analysis and
evaluation of the shipboard aircraft and is important to optimize the shipboard
aircraft operational support process.
Keywords Aviation support system HTCPN Petri-Net Operational support
Shipboard aircraft
123.1 Introduction
The aircraft carrier is the combat platform at sea with the most powerful combat
effectiveness in the world now. It plays a significant role due to its unique char-
acteristics like integrating the sea and air routes, combining ships and planes,
controlling the air and sea and rapid deployment. The main reason for aircraft
carrier becomes an important force in naval battle and land combat is its unique
T. Wang (&) B. Xiao L. Ma Y. Tian
School of Reliability and System Engineering, Beihang University, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1167
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_123,
Ó Springer-Verlag Berlin Heidelberg 2013
1168 T. Wang et al.
weapon equipment—shipboard aircraft. The shipboard aircraft is the power of the
aircraft carrier. The operation and management of aircraft carrier is very complex,
starting and recycling need a series of procedures like the lifting, guiding,
arranging, refueling, loading ammunition, landing, catapulting, and blocking,
which are implemented by the aviation security system. Therefore, the rational
planning of aviation security resources and reducing the security time of using
aircraft carrier play vital roles in the sustainability of the aviation security system
task.
The security process of using the carrier aircraft is the typical discrete system.
The process modeling methods commonly used for simulation modeling in dis-
crete system are: CPM/PERT methods, IDEF3 method, Petri network method,
activity-based random network method. The concepts of the place, transition and
arc in the Petri network correspond to the activities, states, and rules in the security
process of equipment. Applying Petri network can better show the use and
maintenance process of the equipment. But the basic Petri network is difficult to
describe a variety of security resources, unable to model the security task exe-
cution time, moreover, the system is complex, the model is large, hence it is easy
to become chaotic. As the air security system have more security resources, higher
requirements of the task duration, and complex and variable security tasks, the
Timed Colored Petri network concept is introduced, which models the security
process of applying shipboard aircraft, and lays foundation for building compre-
hensive security model of complex equipment and systematic analysis.
123.2 Basic Principle of Petri-Net Modeling
Petri network is put forward by German scholar C. A. Petri in 1962 as a process
modeling and analysis tool. It is a tool for describing the graphical and mathe-
matical processes, which provides a powerful means for studying complex systems
with parallel, asynchronous, distributed and stochastic characteristics. After four
decades’ development, the Petri network has been widely used in various fields to
simulate, analysis and control of the system such as the design, artificial neural
networks, network performance analysis in parallel program and so on.
123.2.1 Basic Elements of Petri Network
The basic elements of the PDM workflow model include Store house, Transition,
Token and directional arc (Sun et al. 2011).
(1) Store house: Represents the conditions that is, promoting factors of processes,
and is shown by a circle ‘‘s’’. When the conditions are met, the corresponding
end nodes of directional arc with this store house as the starting point will be
123 Modeling of Shipboard Aircraft Operational Support Process 1169
activated. The introduction of store house is used to strictly distinguish the
abilities and the real implementation of the activities.
(2) Transition: represents the tasks, which is the activities in the PDM workflow,
and is shown by a box ‘‘h’’. For example, document countersignature and the
design change. However, when an activity is not completed within the set
time, the system will deliver the timeout warning to the operating users, and
the process will be suspended. In order to better describe the PDM workflow,
transition of zero delay-instantaneous transition as auxiliary structure is added
in the model. To make the model easy to be understood, this paper will apply
‘‘h’’ to represent the transition of zero delay.
(3) Token: represents the resources and data that can be used, which is a sign that
a certain condition is met, and is shown by the black spots in the circle. When
the process is moving forward, the token will transfer from one store house to
another.
(4) The directional arc: it is used to connect store house and transition, repre-
senting the order of the implementation of activities. When the starting node of
the directional arc is finished, the system will conduct process navigation
according to the definition of the directional arc and the follow-up activities
will be activated.
(5) Routing: routing means the sequencing description of activities and connec-
tion through the process of business implementation, and the various activities
are associated through routing. There are four basic routing structures in PDM
workflow: sequential routing, parallel routing, conditional routing, and cycle
route. In order to describe some basic routings in the workflow model, several
corresponding structured components are constructed in a functional network,
i.e.: serial component, parallel component, condition selecting component and
cycle component.
(a) Serial component
Serial relationships define the activities performed in a fixed order. For
example, in the shipboard aircraft transporting process, the carrier aircraft
must be transported from the hangar to a lift, and then transported by the lift to
the deck, which is just as shown in Fig. 123.1.
(b) Parallel component
If several tasks can be executed simultaneously or in any order, it can be called
as the parallel relationship. It mainly applies two basic workflow primitives:
And-Split and And-Join. For instance, in the deck support process, when the
shipboard aircraft gets to the support point, a variety of deck support activities,
such as refueling and charging. If these activities have no relations, they can
be seen as different branches of the whole process. The implementation of the
Fig. 123.1 Serial component
Hangar Lift
transport transport
1170 T. Wang et al.
And-Split refuel And-Join
To
To support
take off
point
point
charge
Fig. 123.2 Parallel component
two activities is not in chronological order, the ‘‘take off’’ activity will not be
triggered only when the two activities are completed. It is just shown as in
Fig. 123.2.
(c) Condition selecting component
Corresponding to conditional routing, it is used to define the split activities
with mutual restraint and exclusive relations between each other. This kind of
split activity often conducts the ‘‘single choice’’ or ‘‘multiple choices’’ based
on the specific implementation situation. The condition selecting component
also requires two basic work-flow primitives: OR-Split and OR-Join. The
relationships of condition selection can be divided into two kinds: one is the
implicit or split selection, that is, it is not known in advance that the trigger
order of the activities determines which split is triggered; and the other is the
explicit or split selection, that is, determine which branch is triggered
according to the activity property before split. The operational support process
mainly adopts explicit or split selection logic. In the refueling process, do
pressure refueling or gravity refueling will be determined according to the
aircraft model. After refueling is completed, move to the next step as shown in
Fig. 123.3.
(d) Cycle component
Cycle component is used to characterize the repeated execution of a task. In
this component, an explicit or OR-Split is used. For example, in the tractor
repair process, if the repair is successful, go to the next step, that is used for
aircraft transporting. If the repair is unsuccessful, continue to repair, until
pressure
OR-Split refuel OR-Join
Prepare Refueling
refueling complete
gravity
refuel
Fig. 123.3 Condition selecting component
123 Modeling of Shipboard Aircraft Operational Support Process 1171
Fig. 123.4 Cycle component OR-Split
repair detection
available. Here does not consider the tractor maintenance grading and repair
strategies, only discuss the maintenance activities as a unity. It is shown as in
Fig. 123.4.
123.2.2 HTCPN Model
Hierarchical timed colored Petri net (HTCPN) not only extends the color of the
model and the execution time of activities, but also hierarchical modeling oper-
ational support processes, which combine data structures and hierarchical
decomposition (Zheng et al. 2011).
Aviation support system is very complicated; the model created with traditional
Petri net will be a huge scale and have a large amount of notes. This will not only
make the process of modeling complex, but also make the analysis of the model
characteristics difficult. Therefore the hierarchy Petri network models can be
introduced, that is, use corresponding subnet in a large model to replace the
transition needs to refine. The transition that contains the subnet is represented by
the double box ‘ ’. The design process of hierarchy Petri network model can be
divided into two stages. First: define tasks at the top of the entire workflow
structure; second: to determine the details of tasks description in the lower level
(Zhao et al. 2009).
In the basic Petri network model, the transition is only with the feature of
‘‘transient’’, which means its trigger is not time-consuming. When studying the
operational support process, time is the parameter must be considered, as many
analysis quantitative indicators like maintainability and support are described in
time value, such as MTTR. HTCPN brings in the concept of time and the modeling
task execution time, hence they can be simulated to obtain the time performance of
shipboard aircraft using the support HTCPN model, and estimate the support time
of aviation support system and the utilization rate of the support resources so as to
provide the basis for the optimization of applying the support process.
In the work of the equipment support, contents like support equipment and
support personnel must be considered. HTCPN defines the color of the places and
enhance the arc expression ability so as to uniformly model different support
resources and avoid utilizing the large and complex support models (Yang et al.
2010; He et al. 2010).
1172 T. Wang et al.
123.3 The Support Process Analysis of Shipboard
Aircraft’s Aviation Support System
Aviation support system is constituted by a variety of subsystems, components,
parts and equipment assembly, and it is designed to provide the overall of carrying,
taking off and landing, maintenance and supplies for shipboard aircraft. It includes
lifts, catapults, tractors, jet bias board, the Fresnel lens, arresting wires, island-type
superstructure, flight deck and hangar, etc. The system can be divided into com-
mand, transporting, deck support, landing, catapult, and other subsystems (Yao
et al. 2009; Wang et al. 2005).
The aviation support system studied in this paper is composed of the trans-
portation system, deck support system, taking off system and landing system
(Zhang 2010) (Fig. 123.5).
The transportation system includes shipboard aircraft’s tractors and lifts as well
as weapons’ tractors and lifts. To guarantee the plane transportation of shipboard
aircraft on the hangar deck, the hangar is equipped with the tractors of the ship-
board aircraft. To guarantee the shipboard aircraft’s lifting transportation between
the hangar and flight deck, the shipboard aircraft’s lifts are set. To guarantee the
plane transportation of weapons between the hangar deck and flight deck, the
weapons’ tractors and lifts are set.
The deck support system is mainly for the detection and maintenance services of
shipboard aircraft, including preparation before start, inspection after the task, and
preparation for another start. The main equipment contains supply facilities of air
power, compressed air, nitrogen, air oil and fuel and weapons mounting equipment.
Landing system mainly ensures the safe landing of the shipboard aircraft. In
order to complete the safe landing of the shipboard aircraft, the landing system is
equipped with facilities like optical landing aid device, arresting wires and
arresting nets.
Weapons transport
Lift Deck Deck
Traction in place
transport transport support
Hangar Take off
aircraft
transport
transport landing do a task
Fig. 123.5 The operational support process of shipboard aircraft
123 Modeling of Shipboard Aircraft Operational Support Process 1173
Take-off system mainly includes the facilities such as catapults, jet bias boards.
The main task is to provide sufficient power to make the shipboard aircraft take off
smoothly in a shorter distance.
The task of the aviation support system is triggered by the shipboard aircraft’s
task, and the number of shipboard aircrafts can be flexibly set according to the task.
123.4 Petri Network Modeling for the Aviation
Support System
HTCPN modeling is used to establish support modeling for applying shipboard
aircraft as shown in Fig. 123.6. This model fully reflects the characteristics of
hierarchical modeling, simply and clearly shows that working processes of avia-
tion support system, and lays the foundation for further analysis and evaluation of
the operational support processes (Song 2008).
123.4.1 The Top-level Model of the Aviation Support System
The operational support process of shipboard aircraft from the hangar to the take-
off point first needs to transport shipboard aircraft from hangar to deck support
point, meanwhile, the ammunition must be transported from ammunition depot to
the deck support point, and then conduct deck support for the shipboard aircraft so
as to complete deck support and tract the shipboard aircraft to take-off point. This
process can be divided into four relatively independent modules, and the opera-
tional support process is sub-divided according to hierarchical Petri network
theory (Fig. 123.7).
In the figure, the black token means shipboard aircraft, and the red token
represents ammunition. It should be noted that the token in the figure is only for
more vividly expressing the relationship between support resources and major
equipment, not for showing the number of tokens. During the actual modeling
process, the amount of preset resources can be simulated and the allocation of
resources can be balanced through analyzing the simulation results.
123.4.2 Transportation Process Modeling
The transportation of shipboard aircraft can be divided into three stages. The first
stage is the transportation of shipboard aircraft in the hangar, which is the time
transporting shipboard aircraft from the hangar to the lift; the second stage is the
time transporting shipboard aircraft from the hangar to the deck, that is, the trans-
portation process of the lifts; the third stage is the transporting time of the deck 1.
1174
p3 p4
t3
p1 p2 p5 p6
t1 t2 t4
t20 t27
t19 t28
t14 t13 t12 t11
t22 t34
t29
t17 t10
t30
t18 t21 t23 t25 t26
t5 t6 t7 t8 t9 AND-Join
t24
t15
t32
t16
OR-Split OR-Join
T. Wang et al.
Fig. 123.6 The petri-net model of aviation support system
123 Modeling of Shipboard Aircraft Operational Support Process 1175
p3 p4
t3
p1 p2 p5 p6
t1 t2 t4
Fig. 123.7 The top-level model of the aviation support system. t1: transport shipboard aircraft
from hangar to shipboard support point. t2: shipboard support. t3: Weapons transport. t4: tract the
shipboard aircraft to take-off point
The transporting time here means the time from the lift to the protection time to
time, excluding from the protection point to the take-off time, and to protect the
support point excluding the time from the support point to the landing point and
from the landing point to the support point. The number of shipboard aircraft’s
tractors in hangar, the number of lifts, the number of shipboard aircraft’s lifts on the
deck and the number of shipboard aircrafts need transporting directly affect the
transporting time. The more the tractors and lifts are, the less time needed is.
However, due to space and weight constraints, unlimited increase in the number of
tractors and lifts is impossible. Considering from the other aspect, it will also cause a
waste of resources. Therefore, in order to meet the conditions required, a reasonable
number must be determined.
The model of transportation process is shown as Fig. 123.8.
The model also takes the failure of support equipment into account, and
incorporates the support equipment maintenance activities into the shipboard
aircraft’s transporting sub-module in dominant or branch ways. The green token in
the figure means the hangar tractor and deck tractor while the blue token represents
lift. It should be noted that in order to more clearly show the utilization of support
resources, this paper only uses a token to show the support device, but in the actual
situation there should be more standby support equipment.
Ammunition transporting process is similar to transporting process of shipboard
aircraft, and hence it will not be repeated here. It should be noted that the quantity
of the weapons delivery should be measured by weight, which can be split. The
transporting of the shipboard aircraft is as a whole, which cannot be split. This
should be considered when conducting simulation.
123.4.3 Deck Support Process Modeling
The deck support system includes jet fuel system, aviation power system, air
supply system, and deck support facilities. The major functions include pressure
refueling and gravity refueling for shipboard aircraft; responsible for the support of
the preparation before deck and shipboard aircraft flight, and aviation power
1176 T. Wang et al.
p23
t20
OR-Split
p24 p22
t19
p2
p15 p14 p13
t14 t13 t12 t11
p21 p12
p20
t17 t10
OR-Split
p19 p11
t18
p1 p7 p8 p9 p10
t5 t6 t7 t8 t9
p16 p18
t15
OR-Split
p17
t16
Fig. 123.8 The model of transportation process. t5: tie the aircraft to the hangar tractor. t6: untie
the aircraft from the hangar. t7: transport by the tractor. t8: tie to the lift. t9: untie the hangar
tractor. t10: transport by the lift. t11: tie to the deck tractor. t12: untie the lift. t13: transport by the
deck tractor. t14: untie the deck tractor. t15: check the hangar tractor. t16: repair the hangar
tractor. t17: check the lift. t18: repair the lift. t19: check the deck tractor. t20: repair the deck
tractor
supply before the second start; power supply for the maintenance of flight deck and
shipboard aircraft in the hangar; aviation power supply for guaranteeing the ship
aviation maintenance, and related cabin maintenance; guaranteeing the flight deck
utilizes aviation power to start the shipboard aircraft; guaranteeing the centralized
storage, management and charge/discharge maintenance for aviation batteries,. In
addition, it is also responsible for preparation before flight, maintenance of the
required gas including filling oxygen and nitrogen to the shipboard aircraft,
guaranteeing the gas filling for the shipboard aircraft’s wheels and making sure the
cooling of electronic equipment in shipboard aircraft when the power is on.
Guarantee the routine maintenance of the deck support equipment and the wash,
hydraulic maintenance, safe ground and the snow removal of the flight deck.
Suppose all the processes can be conducted at the same time except filling
oxygen, and every support site is equipped with the same set of jet fuel system, the
deck support model can be established as shown in Fig. 123.9.
123 Modeling of Shipboard Aircraft Operational Support Process 1177
123.5 Simulation and Analysis
Petri network has powerful analytical techniques and means. Analysis of work-
flow’s behavior, status, and performance can be solved through the nature of Petri
network (such as accessibility, safety, livability, etc.); moreover, the analysis
techniques of Petri network can be used to calculate various performance indi-
cators of the model, such as response time, latency time and share.
CPN-Tools is a Petri network modeling and simulation platform developed by
the Petri network Research Center for the University of Aarhus, Denmark. It is
featured with fast simulation speed and powerful network grammar checker. It
supports for Linux and Windows operating system, supports hierarchical modeling
and analysis of Timed Colored Petri network and supports secondary development.
After modeling by Petri network, the features of the system can be analyzed to
check the characteristics of the actual system. CPN-Tools support the state
equation analysis and time simulation. By assigning the corresponding model’s
transition, arc, and the place, it can be clearly learnt the overall situation of the
p35 p34
t27
p36
t28
p26 p27 p44
t22 t34
p45
p38 p37
t33
t29
p43
p39
AND-Split
t30
p25 p28 p29 p32 p33
t21 t23 t25 t26
p30 p31 AND-Join
t24
p42
t32
p41 p40
t31
Fig. 123.9 The model of deck support process. t21: repair the deck support. t22: refuel. t23:
charge. t24: load ammunition. t25: oxygenate. t26: complete deck support. t27: check the refuel
equipment. t28: repair the refueling equipment. t29: check the charging equipment. t30: repair the
charging equipment. t31: check the loading equipment. t32:repair the loading equipment. t33:
check the oxygenating equipment. t34:repair the oxygenating equipment
1178 T. Wang et al.
operational support, parameters like the average support delay time and the sup-
port resources’ utilization can be obtained, and the operational support time of the
shipboard aircraft can be analyzed according to the simulation results so as to
achieve the optimization of support resources (Song et al. 2007).
123.6 Conclusion
Based on the operational support feature of shipboard aircraft, this paper utilizes
the hierarchical Timed Colored Petri Network (HTCPN) to establish the process
model of shipboard aircraft to achieve a simple hierarchical modeling, which
makes up for the shortage of time performance analysis of Petri network, achieves
the distinction for support resources, provides a reference for the research of the
support process of aviation support system, and plays a significant role in the tasks
persistence of aviation support system.
References
He J-B, Su Q-X, Gu H-Q (2010) Virtual maintenance process model based on extended Petri net
(in Chinese). Comput Simul 27(3):254–257
Song G-M, Wang D-S, Song J-S (2007) The equipment maintenance support resource
management model based on timed colored Petri net (in Chinese). J Syst Simul
19(1):233–236
Song K (2008) Research on modeling and simulation technology of integrated logistics support
system based on Petri net (in Chinese). Graduate School of National University of Defense
Technology, Changsha
Sun B, Wang Y, Guo Y (2011) Process modeling and analysis of maintenance support command
based on Petri net (in Chinese). Command Control Simul 33(1):113–117
Wang K, Feng J-L, Zhang H-X (2005) Modeling based on Petri nets about operation maintenance
and support of military aircraft (in Chinese). J Acad Equip Command Technol 16(6):15–17
Yang C, Yang J, Hu T (2010) Method of maintainability prediction of complex equipment based
on CPN simulation (in Chinese). J Eng Des 17(1):25–29
Yao X-L, Feng L-H, Zhang A-M (2009) Systems vulnerability assessment of aircraft guarantee
system based on improved FPN (in Chinese). Electri Mach Control 13(3):464–470
Zhang W (2010) Research of analysis approach of the mission sustainability on the carrier based
aircraft support system (in Chinese). Beihang University, Beijing
Zhao X-M, Gao X-J, Hang G-D (2009) Modeling for communication equipment maintenance
support system based on HTPN (in Chinese). J Jilin Univ Inf Sci Ed 27(4):412–417
Zheng Z, Xu T, Wang X (2011) A method of determining equipment maintenance resources
based on colored timed Petri net (in Chinese). Ship Sci Technol 33(2):131–133
Chapter 124
Modeling and Simulation
of a Just-in-Time Flexible
Manufacturing System Using Petri Nets
Yue Cui and Yan-hong Wang
Abstract The modeling and simulation issues of the flexible manufacturing
system under Just-in-Time environment is addressed in this paper. A typical
flexible manufacturing system has been used as the study case, and its Petri nets
model with Kanban has been presented. Since bottleneck or hunger resources in
the manufacturing system usually have bad influence on the production process,
more attentions were paid to the bottleneck identification and digestion in support
of the proposed modeling and simulation mechanism in this paper. The machine
utilization, under the premise of meeting custom needs just-in-time, is used as the
main measure, while the trigger priority and the kanban numbers are two main
adjusted artifices. Therewith, a large number of numerical simulations are inves-
tigated and detail discussions are proposed further. The simulation results show
that the proposed Petri nets based modeling technique, as while as the bottleneck
identification and digestion strategies, are feasible and effective.
Keywords FMS JIT Modeling Petri nets Simulation
124.1 Introduction
FMS is a manufacturing mode that combines computer information control system
and material automatic storage and transportation system (Du 2010). JIT approach
to production was originated by Toyota in 1970s in their car assembly plants and
the core content is eliminating manufacturing wastes by producing only the right
amount and combination of parts at the right place at the right time (Araz et al.
2006; Zhang et al. 2012). The advantage of JIT flexible manufacturing system is
Y. Cui (&) Y. Wang
Department of Information Science and Engineering, Shenyang
University of Technology, Shenyang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1179
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_124,
Ó Springer-Verlag Berlin Heidelberg 2013
1180 Y. Cui and Y. Wang
not only embodies the JIT ideology but also enhances the flexibility of production
systems. However, flexible manufacturing system is an extremely complex dis-
crete event dynamic system; it is difficult to be described with traditional math-
ematical models.
Petri nets with its perfect mathematical theory as the foundation, has strong
modeling capabilities to describe the parallel, synchronous, conflict relations and
plays an important role in the system modeling and simulation. It has also been
applied in the modeling of flexible manufacturing systems (Colombo et al. 1997;
Mao and Han 2010). On the other hand, with the manufacturing system has
become increasingly complex, especially under today’s challenge environments,
the auxiliary analysis software becomes a prerequisite for the application of Petri
nets. ExSpect (Voorhoeve 1998), the Executable Specification Tool, is a powerful
modeling and analysis of language and software tools based on timed colored Petri
nets. It is widely used in transportation systems, workflow modeling and main-
tenance support system (Qu et al. 2009; van der Aalst and Waltmans 1991;
Vanit-Anunchai 2010; University of Aarhus 2005).
This paper addresses the modeling and simulation issues of the flexible man-
ufacturing system under Just-in-Time environment basing Petri nets and ExSpect,
a common simulation software platform. A typical flexible manufacturing system
has been used as the study case, and its Petri nets model with Kanban has been
built. Since bottleneck or hunger resources in the production process are com-
monly occurred cases, and they often have bad influence on the production process
of JIT flexible manufacturing system (Zhang and Wu 2009), more attentions were
paid on the bottleneck identification and digestion in support of the proposed
model and simulation mechanism. The machine utilization, under the premise of
meeting custom needs just-in-time, is the main measure of the problem, while the
trigger priority and the kanban numbers are two main adjusted artifices. At the end
of the paper, a large number of numerical simulations are investigated to verify the
effective of the proposed model and detail discussions are proposed further.
124.2 Petri Nets Model of the Single-Kanban System
One of the major elements of JIT philosophy is the kanban system (Al-Tahat et al.
2009). The kanban system is an information system which controls the production
quantities in every process. Figure 124.1 shows a Petri nets model of the single-
kanban system (Di Mascolo et al. 1991) and describes the production process of
three adjacent processing units. In a single kanban system, a production line could
be divided into several stages and there are a fixed number of kanbans at every
stage. The production of a part cannot start until a kanban indicates that this part is
needed by the following downstream station (Matzka et al. 2012).
124 Modeling and Simulation of a Just-in-Time Flexible Manufacturing System 1181
Td
F1 F2 F3
D
W1 W2 W3
2 3 2 K
I1 O1 I2 O2 T2 I3 O3 T3
T0 T1
K N1 K N2 K N3
Fig. 124.1 Petri nets model of the single-kanban system
124.3 Modeling of JIT Flexible Manufacturing System
124.3.1 A JIT Flexible Manufacturing System Case
This paper takes a typical JIT flexible manufacturing system given in (Raju et al.
1997) as a case for addressing the modeling and simulation problem. The JIT
flexible manufacturing system consists of five machining centers (from M1 to M5)
and a load/unloads station (LUS), and they were connected by automatic guided
vehicles (AGV) network. It caters to a variety of part types. In this paper, three part
types are processed in this JIT flexible manufacturing system.
124.3.2 Petri Net Model of JIT Flexible
Manufacturing System
The proposed JIT flexible manufacturing system modeling strategy adopted a
hierarchical modeling methodology. First, The Petri net model of each part type is
made separately. Then, these three models are linked by merging the common
Fig. 124.2 The JIT flexible
manufacturing system model tx
p29 px1 p40 p45 px2 p58 p63 px3 p76
part 1 part 2 part 3
p1 p2 p3 p4 p5 p6 p7
1182 Y. Cui and Y. Wang
Fig. 124.3 User demands subsystem model
resource places to yield the system net. The interpretations of places and transi-
tions are given in (Raju et al. 1997). Figure 124.2 shows the model.
Here main elements in the system are defined as:
px1, px2, px3: num, //Input requirements
p29, p45, p63: num, //Input of raw materials
p40, p58, p76: num, //Output products
p1, p2, p3, p4, p5: num, //Machines
p6: num, //Fixture
p7: num, //AGV
This Petri net model contains four sub-system, they are named as tx, part1,
part2 and part3, respectively representing user demands subsystem, part1 pro-
cessing subsystem, part2 processing subsystem and part3 processing subsystem.
Among them, tx randomly generates user demands. A Poisson arrival pattern with
a different mean arrival time for each part variety is considered in the present
study. Each part has 10 demands as example in this paper, as an example for the
system modeling, the tx model shown in Fig. 124.3.
This paper takes part1 as an example to introduce the processing subsystem, its
model shown in Fig. 124.4. Being a demand-driven system, the functioning of the
JIT flexible manufacturing system starts with the arrival of a demand. When a
demand arrives, the system directly delivered the part to the user from output
buffer. Then, the system begins to produce the same number of semi-finished or
finished products to compensate for output buffer.
124.4 Simulation and Results
This paper use ExSpect as the simulation platform for our JIT flexible manufac-
turing system to illustrate the performance of the proposed modeling mechanism.
Since bottleneck or hunger resources in the production process are commonly
occurred cases, and they often have bad influence on the production process of JIT
flexible manufacturing system, more attentions were paid on the bottleneck
identification and digestion in support of the proposed model and simulation
mechanism. The machine utilization, under the premise of meeting custom needs
124 Modeling and Simulation of a Just-in-Time Flexible Manufacturing System 1183
Fig. 124.4 The processing subsystem model for Part1
just-in-time, is the main measure of the problem, while the trigger priority and the
kanban numbers are two main adjusted artifices. The simulation is running by the
concurrent execution of the system net. A large number of simulations have been
done, and the results data are recorded.
124.4.1 Initial Settings
The initial conditions of the system include the number of resources place token is
one, the number of output buffer token is one, kanban number is zero, raw material
is infinite and the average arrival time of the three part type demands are 10, 12
and 15 s respectively.
Simulation data include that the processing time of three parts, the difference of
takt time, the total time, the machines utilization and the machines average
utilization.
124.4.2 Simulation Results
Among them, takt time is an important factor in simulation. In order to produce
only what the customers need just-in-time, the supplier has to adapt his production
quantity to the customers’ orders and produce to the takt time. The takt time is
used to synchronize the pace of production with the pace of sales.
Takt time = Available working time per day/Customer demand rate per day
(Matzka et al. 2012).
The difference of takt time is the difference between largest takt time of the part
type and minimum takt time of the part type. The difference is smaller, the syn-
chronization is better. However, because of bottleneck resources and hunger
resources in the production process often impact the synchronization of
1184 Y. Cui and Y. Wang
Table 124.1 Simulation Data
Performance Simulation
Index
1 2 3 4 5 6
P40 (s) 6052 5077 3872 3655 3670 3620
P58 (s) 3167 2517 3562 3170 3410 4200
P76 (s) 3797 3407 3142 3440 3470 3390
Difference of takt time 288.5 249 73 48.5 26 81
Total time 6632 5457 4892 4960 5290 5515
P1 (%) 61.37 69.27 83.20 94.35 94.52 91.21
P2 (%) 46.97 81.82 96.48 86.69 76.08 95.19
P3 (%) 33.47 51.31 52.33 76.81 93.01 81.41
P4 (%) 84.74 96.11 93.93 91.94 90.08 87.04
P5 (%) 49.00 73.30 91.99 90.73 85.07 90.66
P6 (%) 27.14 32.99 36.79 38.10 37.43 37.53
P7 (%) 78.71 47.42 52.59 54.39 53.36 52.63
Average utilization 55.11 74.36 83.59 88.10 87.75 89.10
production, it must eliminate bottlenecks and hunger issues. So this paper gives the
definitions of bottleneck resources and hunger resources. Bottleneck resources
refer to the machine resources that utilization exceed 10 % of the machines
average utilization, and hunger resources refer to the machine resources that uti-
lization below 10 % of the machines average utilization.
Simulation results were shown in Table 124.1 and the detail analysis were
given bellow.
(1) Simulation 1. The simulation is run in the initial conditions; data show that the
difference of takt time is too large; it means that the production synchroni-
zation is too weak. By comparing machines utilization and machines average
utilization, it can obtain hunger resource is P3, bottleneck resources are P4 and
P7 respectively.
(2) Simulation 2. In JIT flexible manufacturing system, resources are divided into
two classes: fixed resources (P1–P5) and variable resources (P6, P7). Because
of the former is the machine resource with high cost, it is not allowed to
increase arbitrarily the number of resources. While the latter resource cost is
low, it allowed be added properly its resources quantity. This paper adds one
token in P7 for eliminating the influence of the bottleneck P7. The simulation
results show the difference of takt time is decreased and bottleneck P7 is
eliminated, it means that the production synchronization is improved. But
there are still bottleneck resource P4 and hunger resource P3.
(3) Simulation 3. In JIT flexible manufacturing system, priority is divided into two
kinds: resources priority and processing subsystem priority. Application
simulation 1 utilization to set resources priority, (P1–P7): 0.61, 0.47, 0.33,
0.85, 0.49, 0.27, and 0.79. Through respectively calculating the ratio of three
parts production time and the total time to set three processing subsystem
priority, they are 0.91, 0.48, and 0.57. Resources priority can be set in the
corresponding transitions, processing subsystem priority is set in all
124 Modeling and Simulation of a Just-in-Time Flexible Manufacturing System 1185
transitions. When two kinds of priority will be set in the same transition, they
should be added together. The priority is greater, the transition to be inspired
more early. Results show that the difference of takt time are decreased, uti-
lization are increased. But there are still bottleneck resources P2, P4, and
hunger resource P3.
(4) Simulation 4. Increasing one kanban of P3, results show that synchronization
continues to strengthen. By analyzing the data, it can obtain bottleneck affect
has been eliminated, but there are still hunger resource P3.
(5) Simulation 5. Increase two kanbans of P3, results show that the synchroni-
zation is best, hunger resource P3 has been eliminated, but hunger resource P2
is appeared again.
(6) Simulation 6. The above conditions remain unchanged, increase one kanban of
P2, results show that synchronization has been weakened, but the basic
elimination of bottlenecks and hunger problem.
Through simulation and analysis of the simulation data, the bottleneck
resources and hunger resources can be identified. Via adjust the number of kanban
and the system priority, the system can effectively solve the bottleneck and the
hunger problems, and optimize the performance indicators.
124.5 Conclusion
The main trend of the system simulation is the integration of modeling and sim-
ulation. Petri nets give a convenient method for the flexible manufacturing system
modeling and simulation. This paper presents our modeling and simulation
mechanism of the flexible manufacturing system under Just-in-Time environment.
By virtue of strong modeling capabilities of timed Petri nets, the model of the JIT
flexible manufacturing system can describe the complex production process com-
pletely. In support of ExSpect environment, through simulation and data analysis, it
can identify bottleneck resources and hunger resources. By setting the system pri-
ority and the number of kanban, the bottlenecks or hunger facilities are settled, and
performance of the system is thereby improved by ameliorating the machine utili-
zation and takt time in manufacturing processes. Therefore, the manufacturing
process can run as a smooth and orderly mode with the premise of meeting custom
needs in just-in-time manner.
Acknowledgments This research work is partly supported by the Scientific Research Fund
given by the Liaoning Education Department (LS2010112).
1186 Y. Cui and Y. Wang
References
Al-Tahat MD, Dalalah D, Barghash MA (2009) Dynamic programming model for multi-stage
single-product Kanban-controlled serial production line. J Intell Manuf 23:37–48
Araz ÖU, Eski Ö, Araz C (2006) A multi-criteria decision making procedure based on neural
networks for Kanban allocation. Springer, Berlin, pp 898–905
Colombo AW, Carelli R, Kuchen B (1997) A temporised Petri net approach for design, modelling
and analysis of flexible production systems. Adv Manuf Technol 13:214–226
Di Mascolo M, Frein Y, Dallery Y, David R (1991) A unified modeling of Kanban systems using
Petri nets. Int J Flexible Manuf Syst 3:275–307
Du X (2010) Development of flexible manufacturing system (FMS). Sci Technol Assoc Forum
5:35 (In Chinese)
Mao Y, Han W-G (2010) Research and implementation of FMS scheduling based on Petri nets.
J Chin Comput Syst 31(5):1001–1005 (In Chinese)
Matzka J, Di Mascolo M, Furmans K (2012) Buffer sizing of a heijunka Kanban system. J Intell
Manuf 23(1):49–60
Qu C, Zhang L, Yu Y, Liang W (2009) Development of material maintenance organization
modeling and simulation environment based on ExSpect domain library. J Syst Simulat
21(9):2772–2775 (In Chinese)
Raju KR, Reddy KRB, Chetty OVK (1997) Modelling and simulation of just-in-time flexible
systems. Sadhana 22(1):101–120
van der Aalst WMP, Waltmans AW (1991) Modelling logistic systems with ExSpect. Eindhoven
University of Technology, The Netherlands
Vanit-Anunchai S (2010) Modelling railway interlocking tables using coloured Petri nets. Coord
Model Lang, pp 137–151
Voorhoeve M (1998) ExSpect language tutorial. Eindhoven University of Technology,
Eindhoven
University of Aarhus (2005) Sixth workshop and tutorial on practical use of colored Petri nets and
the CPN tools. University of Aarhus, Aarhus
Zhang R, Wu C (2009) Bottleneck identification procedures for the job shop scheduling problem
with applications to genetic algorithms. Adv Manuf Technol 42:1153–1164 (In Chinese)
Zhang X, Li P, Yan C (2012) Shallow discussion just-in-time (JIT) production mode. Guide Bus
3:257 (In Chinese)
Chapter 125
Numerical Simulation of External-
Compression Supersonic Inlet
Flow Fields
Ping Wang, Hong-wei Wang, Si-dong Wei, Xue-shan Liu,
Qing-guo Zhang and Xin Hua
Abstract In this paper, Method of CFD is used to simulate 2D flow fields for a
certain external-compression supersonic inlet. It describes the methods of mesh
generation, boundary conditions determination, convergence techniques of gov-
erning equation, and analyzes simulation result. The numerical results match well
with the theory.
Keywords Convergence Flow fields Numerical simulation Supersonic inlet
125.1 Introduction
The airplane inlet is an important component of aviation propulsion systems, its
main functions are to transform kinetic energy of high-speed airflow into potential
energy, and to provide necessary air to engine. The airplane inlet flow charac-
teristics affect engine performance greatly. Inlet design and research are mainly
based on wind tunnel experiments, aerodynamics and computational fluid
dynamics (CFD). Hereinto, wind tunnel experiments need a long cycle, high cost,
and there are many limits such as air flow temperature, pressure, speed, and model
size. Currently, CFD method has become a popular aerodynamic analysis tools
(Ju et al. 2005; Wang 2004; Zhang and Bai 2008; Zhu et al. 2002; Zhang and Wu
2008). CFD method is used in the paper to simulate 2D flow fields for a certain
external-compression supersonic inlet; it describes the method of mesh generation,
P. Wang (&) H. Wang S. Wei X. Liu
Aeronautical Mechanical Engineering Department, Aviation University
of Air Force, Changchun, People’s Republic of China
e-mail: [email protected]
Q. Zhang X. Hua
Disciplinal Department, Aviation University of Air Force, Changchun,
People’s Republic of China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1187
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_125,
Ó Springer-Verlag Berlin Heidelberg 2013
1188 P. Wang et al.
boundary conditions determination, convergence techniques of governing
equation, and analyzes three kinds of operating mode of external-compression
supersonic inlet.
125.2 Calculation Model and Mesh Generation
In this paper, the external-compression supersonic Inlet is with 3 wave system.
Before entering the channel, supersonic flows get through two oblique shock
waves and a normal shock wave, and then the flows reduce to subsonic flow,
furthermore velocity drop, pressure increase in the diffuser, and ultimately achieve
the requirements of the combustion chamber. Figure 125.1 shows 2D structure of
inlet used for calculation and schematic of the shock wave system. The design
operating mode of the inlet is as follows: flight altitude H = 11000 m, stream
Mach number M = 2.6, attack angle a = 0.
Mesh generation is the key issue for calculation of the inlet flow field, a rea-
sonable grid make for the computational speed and accuracy. The inlet CFD model
is defined with structured mesh. Wall functions are adopted for wall surface, initial
grid is defined for Y ? & 500 and the number of fluid cells is 19,500. After
adapted, final Y ? & 150 and the number of fluid cells are 20,675. The sensitivity
to grid adaption has been investigated to reduce effect of mesh density. Results
show that adaption of a solution grid with static pressure gradients \40 and
Y ? \ 500 does not significantly change the internal CFD results under the same
type operating condition. The boundary conditions are given in Fig. 125.2. AGFE
is defined as non-reflecting pressure boundary, ED pressure-outlet, ABCD and EH
no-slip adiabatic wall. Many experiments and calculations (Ju et al. 2005; Wang
2004; Zhang and Bai 2008; Zhang and Wu 2008) have proved the reasonableness
of the above choices
Fig. 125.1 Inlet structure
and the shock wave system 0.4
0.3
Y(m)
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
X(m)
125 Numerical Simulation of External-Compression Supersonic Inlet Flow Fields 1189
Fig. 125.2 Mesh and 0.6
G
boundary conditions 0.5
F
0.4
Y(m)
E
H
0.3
0.2 D
C
0.1
A B
0
-0.5 0 0.5
X(m)
125.3 Numerical Methods and Turbulence Model
Finite volume method is employed to solve N–S equation. Coupled implicit
arithmetic with second order discretization and ideal gas properties is used to
model the inlet. Combined with wall functions, RNG k e turbulence model with
eddy viscosity correction is adopted. The formula of standard RNG k e model
for solving the flow effective viscosity is described as:
leff ¼ l þ lT : ð125:1Þ
where l is molecular kinematics viscosity of the fluid and lT is RNG model
turbulent viscosity calculated. The formula of RNG k e model with eddy vis-
cosity correction for solving the flow effective viscosity is described as:
pffiffiffi pffiffiffiffiffiffi2
leff ¼ l þ lT ð125:2Þ
125.4 Convergence Techniques
The numerical calculation of supersonic inlet is not easy to convergence, which is
the generally accepted view (Wang 2004; Zhang and Wu 2008). The main reason
is supersonic flow inside inlet while subsonic flow outside inlet, and there exists
interaction between pressure gradient and viscosity in flow field, accompanied
with shock waves, to get a stable normal shock wave flow field, iteration is much
slower than the whole supersonic flow field. For the non-convergence problems,
this paper has tried two methods and compared them. The one is to select first
order upwind discretization scheme until convergence, then select the QUICK
scheme, and the other is a direct QUICK scheme calculation until convergence, the
results show that the first method calculation uses 7 h in computer with 2 GHZ
CPU, and the second method uses 32 h, besides, the calculation results are con-
sistent. That is to say, we can adopt the first method when calculating other similar
problems. During the calculation process, the residuals are difficult to down, order
of magnitude is always maintained at 10-2–10-3, we can concern about the
1190 P. Wang et al.
amount of change to determine convergence, such as exit flow turnover, total
pressure recovery coefficient, etc., when the value changes in each iteration step
become very small and no real meaning, that is convergent (Hu and Wang 2008;
Huang and Wang 2007; Li 2009; Zhao and Zhou 2001; Zhou et al. 2009).
A reasonable initialization of flow field affects significantly the accuracy of
results and convergence rate. Initialization method in this article is as follows: To
begin with, full multigrid initialization (FMG initialization) is adopted, which is
effective for the flow with great pressure or velocity gradient, and then, the cal-
culation results of inlet design operating mode are used to extrapolates the flow
field else, Numerous tests have been performed about the computational scheme in
the case, and results show that calculation efficiency could be improved more than
5–8 times (Carlson 1994; Lakshmanan and Nagarajan 2010; Launder and Spalding
1974; Lee and Edwin 1981; Morgan 2004).
Fig. 125.3 Contours of the critical operating mode. a Contours of Mach number. b Contours of
static pressure (Unit of pressure in atm)
125 Numerical Simulation of External-Compression Supersonic Inlet Flow Fields 1191
125.5 Calculation Results
When the flight number M is constant, the flow status of inlet changes with the
flow capacity of inlet export. The inlet flow fields including critical, supercritical
and subcritical operating mode have been simulated with the Mach number of
design operating mode. The boundary conditions of critical operating mode are as
follow: flight altitude H = 11000 m, flow Mach number M = 2.6, angle of attack
a = 0, the export anti-pressure P = 2.78 atm. Figure 125.3 shows the calculation
result.
Two oblique shock waves and a normal shock wave intersect just at the leading
edge of outer walls in critical operating mode. After the first oblique shock wave,
Mach number is reduced to 1.7, after the second oblique shock wave, it decreases
to 1.3, the flows are still supersonic, after going through a normal shock wave M
decreases 0.93, the flows become subsonic. Static pressure before and after every
shock has a mutation, inside the pipeline flow pressure increases and velocity is
reduced.
The intake conditions of supercritical and subcritical modes are the same as
critical operating mode, the calculation results of critical operating mode are
utilized to extrapolate the flow field else, Figs. 125.4 and 125.5 show the
numerical results.
As shown in Fig. 125.4, the normal shock wave moved into channel in the
supercritical operating mode, it is still supersonic flow in the initial segment, after
normal shock wave flow pressure increases and velocity is reduced. As shown in
Fig. 125.4 Contours of the
supercritical operating mode.
a Contours of Mach number.
b Contours of static pressure
(Unit of pressure in atm)
1192 P. Wang et al.
Fig. 125.5 Contours of the subcritical operating mode. a Contours of Mach number. b Contours
of static pressure (Unit of pressure in atm)
Fig. 125.5, oblique shock waves and normal shock wave intersect before the
entrance in the subcritical operating mode. Low-energy flow inpour into the
channel so as to increase total pressure loss, serious cases may lead to an unstable
operating mode. The total pressure recovery coefficients under the three modes are
as follows: 0.849 in the critical operating mode, 0.741 in supercritical operating
mode, and 0.827 in subcritical operating mode. As far as total pressure loss is
concerned, the most favorable location of normal shock is just at the leading edge
of entrance.
125.6 Conclusions
Finite volume method is employed to solve N–S equation. Combined with wall
functions, RNG k e turbulence model with eddy viscosity correction is adopted.
Numerical techniques such as FMG initialization, extrapolation and reasonable
125 Numerical Simulation of External-Compression Supersonic Inlet Flow Fields 1193
calculation process can significantly improve efficiency. Numerical results of three
operating mode show that the most favorable location of normal shock is just at the
leading edge of entrance. Numerical results are consistent with the theory better,
which indicate that the simulation method is reasonable, and it is valuable for inlet
design and research.
References
Carlson JR (1994) Computational prediction of isolated performance of an axisymmetric nozzle
at Mach number 0.90. NASA, TM-4506
Hu H, Wang Q (2008) Numerical study on combination flow field of axisymmetric convergent-
divergent nozzle under transonic and supersonic conditions. J Aerosp Power 23(6):1402–1406
Huang H, Wang Q (2007) Numerical investigation on internal and external flows for
axisymmetric convergent-divergent nozzles on over expansion state. J Aerosp Power
22(7):1070–1073
Ju Y, Zhou C, Wang Z (2005) Numerical analysis of flow field of supersonic axisymmetric
double cone-shaped inlet. J Nanjing Univ Sci Technol 8(4):411–413
Lakshmanan T, Nagarajan G (2010) Experimental investigation of timed manifold injection of
acetylene in direct injection diesel engine in dual fuel mode. Energy 35(8):3172–3178
Launder BE, Spalding DB (1974) The numerical computation of turbulent flows. Comput
Methods Appl Mech Eng 3:269–289
Lee CGT, Edwin E Jr (1981) Experimental analytical investigation of axisymmetric supersonic
cruise nozzle geometry at Mach number from 0.6 to 1.3. NASA, TP1953
Li Y (2009) Introduce of the method RANS. Aerodyn Res Exp 28(3):5–10
Morgan MJ (2004) Pressure transient characterization test star-2 propulsion system fuels
manifold. AIAA 2004-3666
Wang F (2004) Computational fluid dynamics analysis, 1st edn. Tsinghua University Press,
Beijing (Chinese)
Zhang J, Wu S (2008) Numerical simulation of supersonic inlet flow fields based on CFD.
Comput Simul 6(6):75–78 (Chinese)
Zhang S, Bai T (2008) Numerical simulation and analysis of 2-D supersonic Inlet. J Projectiles
Rockets Missiles Guidance 6(3):203–209
Zhao J, Zhou L (2001) Numerical investigation of internal and external supersonic flows of
nozzle. J Propul Technol 22(4):295–298
Zhou W, Huang J, Zhou R (2009) Improvement of laval nozzle calculation model and simulative
verification. J Aerosp Power 24(11):2602–2606 (Chinese)
Zhu Y, Li T, Li C (2002) Applications of CFD in the design of supersonic Inlet. Aircr Des
124:11–17 (Chinese)
Chapter 126
Ontology-Based Multi-Enterprise
Heterogeneous Model Fusion Method
Hong-xiu Wang
Abstract For the multi-enterprise collaborative modeling environment, the
semantic conflicts of concept in the merger from the local model into the overall
model or in the integration from the lower model into the upper model, the
semantic-based multi-enterprise heterogeneous model fusion method is proposed.
Moreover, the semantic similarity among the model instances is analyzed from the
various levels, and based on the semantic similarity, a series of rules of the model
merging are proposed, and then model integration is completed. Finally, the
similarity matching tool is developed to realize based on semantic similarity
analysis.
Keywords Heterogeneous Model integration Multi-enterprise model
Ontology
126.1 Introduction
In the collaborative enterprise modeling, the model is completed by more than one
person in the project team; they will apply their own terms to create a model
instance, resulting in the semantic conflict in the merger from the partial model to
the overall model. The main problems are: the one on the same physical appli-
cation of different terms to describe, the same terminology to describe different
content, three different definitions of the granularity of the process, activity.
For this type of semantic heterogeneity, the related research work focus on
building a unified dictionary based on meta-data (Castano et al. 2005; Missikof
and Schiappelli 2003), but less involved in the essence of the information
H. Wang (&)
Department of Industrial Engineering, Tianjin Polytechnic University, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1195
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_126,
Springer-Verlag Berlin Heidelberg 2013
1196 H. Wang
semantics, and can not fundamentally solve the problem (Cui et al. 2001; Mike
et al. 1998). On the basis of the work on the research at home and abroad, proceed
from solving the semantic heterogeneity of shared modeling the collaborative
modeling, proposed an enterprise ontology-based concept of constraints to solve
the consistency problem in enterprise modeling.
Ontology achieves the effective semantic understanding and communication
between people or application systems (Horrocks et al. 2003; Pulido et al. 2006).
In engineering applications, the ontology can support semantic interoperability. It
provides the mechanism described in the explanation of the objective world.
Semantic interoperability requirements of the data easier to understand, and can be
easily defined mapping between the data of known and unknown (Athena 2004;
Berre et al. 2007).
126.2 Multi-Model Fusion Method
126.2.1 Two Assumptions
Generally, the enterprise model is composed of multiple views, and its structure is
complex. In order to achieve the integration of partial models, first of all assume
that the two conditions have been met: First, the model is divided into view
established; Second, during the merger have been identified belong to the same
view of part of the model can be combined with a view of the upper model.
126.2.2 The Formal Definition of the Enterprise Model
According to the references (Vernadat 2002) on the definition of enterprise model
and (Thiagarajan et al. 2006) on the definition of the business model, it is present
that the business model formal definition is as follows:
Definition 1 Enterprise model EM = (E, R), where E is composed of a collection
of elements for the model, R is the relationship between these elements in the
model.
Modeling element E = {{Attribute}, {range}, {Subelem}}, Attribute is
description of the properties of the element characteristics, range is range of the
property, Subelem is set of elements of sub-concepts.
126 Ontology-Based Multi-Enterprise Heterogeneous Model 1197
126.2.3 Similarity Definition
Definition 2 The formal definition of the similarity of two concepts x, y :
Sim(x,y)[[0…1]
Sim(x,y) = 1?x = y: The two entities are equivalent
Sim(x,y) = 0 : The two entities do not intersect, unrelated
Sim(x,y) = Sim(y,x) : the symmetry
Matching relations between concepts, we can make a judgment according to their
similarity.
In this paper, Sim, (c1, c2) expressed the similarity function between the two
concepts c1 and c2.
Set a threshold t, when Sim (c1, c2) C t, we believe that c1 and c2 is similar.
126.2.4 Analysis of Ontology-Based Model Semantic
Similarity
Basis for the formal definition of the enterprise model, in order to calculate
comprehensively, accurately the similarity between the concepts, Respectively, it
is calculated based on the name, concept properties, the subset of concept. Finally,
it is given the right value to merge the similarity.
(1) Calculation of concept name similarity. Assumptions two concepts A and B,
the similarity of their names is calculated as:
Nðthe longest substring between Aname and Bname Þ
simname ðAname ; Bname Þ ¼
ðNðAname Þ þ NðAname ÞÞ
ð126:1Þ
If the concept has an alias, in addition does computing the concept name
similarity, but also to calculating the similarity of the concept alias. Using the
formula (126.1), the final name similarity is
mþ1 X
X nþ1
SimnameZ ¼ wij Simname ðAi ; Bj Þ n; m 0 ð126:2Þ
j¼1 i¼1
P nP
mþ1 þ1
Among them, wij ¼ 1, n is the number of the alias of the concept A, m is
j¼1 i¼1
the number of the alias of the concept B. When n = 0, m = 0, SimnameZ ¼ Simname
(2) Calculation of similarity based on concept attributes
Based on the attributes, the theoretical basis for calculation of the conceptual
similarity is: if the attributes of the two concepts are the same, then the two
concepts are the same; if the two concepts have similar properties, these two
1198 H. Wang
concepts are similar. Each concept in the ontology is to be described and limited
by a set of attributes. The attribute set definition is given in the following.
Definition 3 Let A = {A1[V1], A2[V2],…, An[Vn]}, A is a set of properties. Ai is
the attribute name; Vi is the range of Ai. The definition is the set of attributes were
classified into the attribute set level and the attribute value level. The calculation of
similarity of the property is divided into two parts of the set of attributes and
attributes values to conduct investigations. Let C1 and C2 are the concept asso-
ciated attribute set of the objects o1 and o2. The similarity of the attribute set is:
1 jC1 \ C2 j
SimattrS ¼
jdistðo1 ; o2 Þ 1j jC1 \ C2 j þ ajC1 C2 j þ ð1 aÞjC2 C1 j
ð126:3Þ
There may be different values in the instances of two objects in the common
property. Therefore, the value of the similarities and differences in the common
property need inspect. Let Ai jC1 \ C2 j. Ai(o)[v] represents that the value of the
instance o on attribute Ai is v, and the upper and lower bounds of the statistical
range of the Ai values are expressed as Low(Ai), High(Ai). The similarity of the
attribute value is:
1 \C2 j
jCY
jAi ðo1 Þ½v1 Ai ðo2 Þ½v2 j
SimattrV ¼ 1 ð126:4Þ
i¼1
jLowðAi Þ HighðAi Þ þ 1j
According to Ai specific data types, the specific meaning of its statistical range
is different. For example, for the numerical data type, the difference between the
maximum and minimum can be used in the actual value of the attribute. For
Boolean data type, 0, 1 value is processed. For string type, if the attribute values of
two instances are the same, similarity is 1, otherwise 0.
In the end, the similarity of two instances in the characteristics of the attribute
set is the superposition of these two aspects. The formula is
Simattribute ¼ SimattrS SimattrV ð126:5Þ
In addition, a concept may have multiple attributes and the effects and the
extent described of each attribute on the concept are different. Therefore, if each
attribute is involved, the amount of the calculation will be greatly increased. When
the attribute similarity is calculated, the attributes need be classified, and focusing
on the calculation of the business attributes.
(3) Calculation of similarity based on a set of concept
In the ontology, the meaning of a concept can consist of the meaning of its
direct sub-concepts. The combination of all sub-concepts can describe the meaning
of the concept. Thus, the similarity between the upper concepts can be obtained by
calculating the similarity between the sub-concepts. This method is flexible and
126 Ontology-Based Multi-Enterprise Heterogeneous Model 1199
extensible. Let A, B for the two upper concept in the ontology, similarity between
A and B using following formula:
P P
max Sðai ; bj Þþ max Sðbj ; ai Þ
ai 2A bj 2B bj 2B ai 2A
Simsub ðA; BÞ ¼ ð126:6Þ
NðAÞ þ NðBÞ
N(A) indicates the number of sub-concepts of A, N(B) indicates the number of
sub-concepts of B. S(a,b) is calculated by using the instance-based method, for-
mulated as:
PðA \ BÞ PðA; BÞ
SimðA; BÞ ¼ ¼ ð126:7Þ
PðA [ BÞ PðA; BÞ þ PðA; BÞ þ PðA; BÞ
which P(A,B) indicates the probability that this concept is sub-concepts both A
and B when a concept randomly is selected from the ontology.
A;B A;B .
PðA; BÞ ¼ ðNðU1 Þ þ NðU2 ÞÞ ðNðU1 Þ þ NðU2 ÞÞ ð126:8Þ
Ui indicates the set of underlying concepts in the ontology i, N(Ui) indicates the
number of the concepts in Ui. N(UA,B1 ) indicates the number of the concepts both
belong A and B in the ontology 1. N(UA,B 2 ) indicates he number of the concepts
both belong A and B in the ontology 2. At this point, the similarity of A and B is
obtained.
(4) Comprehensive computation of similarity
This three kinds of similarities are comprehensively computed, the formula of
the final comprehensive similarity as follows:
SimðA; BÞ ¼ wname SimnameZ ðA; BÞ þ wattribute Simattribute ðA; BÞ þ wsub Simsub ðA; BÞ
which, wname þ wattribute þ wsub ¼ 1.
126.2.5 E Model Merging Rules Based on Semantic
Similarity
Setting a threshold for the above four kinds of similarity, the threshold is usually
determined by experts or analysts. When the calculated similarity is greater than
the threshold, they are called name similarity, attribute similarity, subset similarity
and comprehension similarity. Where, the model merging rules are defined based
on these four similarity relations. According to these rules, and then the overall
model is generated.
Rule 1: if the two model instances are comprehension similarities, the one is
kept, another is deleted in the model merging.
1200 H. Wang
Rule 2: if the two model instances are name similarities and the similarity is
less than 1, but the attribute and the subset are not similar, two models are kept in
the model merging.
Rule 3: if the name similarity of the two model instances is equal to 1, but the
attribute and the subset are not similar, two models are kept in the model merging.
At the same time, the name of a model is modified.
Rule 4: if two models are name similarity and attribute similarity, but their
subsets are not similar, two models are kept in the model merging.
Rule 5: if two models are name similarity and subset similarity, but their
attributes are not similar, two models are kept in the model merging.
Rule 6: if two models are attribute similarity and subset similarity, but their
names are similar, one model is kept in the model merging.
Rule 7: if the two models are attribute similarity, but their names and subsets
are not similar, two models are kept in the model merging.
Rule 8: if the two models are subset similarity, but their names and attributes
are not similar, two models are kept in the model merging.
Rule 9: if the two models are not subset similarity, name similarity, attribute
similarity, two models are kept in the model merging.
126.3 The Tool System of Model Knowledge Matching
The function of the system of model knowledge matching consists of ontology
editing, database/owl transformation and concept matching. The ontology editing
module has completed the editing and maintenance functions of ontology, estab-
lishing ontology tree and all knowledge stored in the database. Database/owl
transformation module has completed to transformation the knowledge stored in
Fig. 126.1 Concept matching
126 Ontology-Based Multi-Enterprise Heterogeneous Model 1201
the database into the standard expressed in OWL ontology. Concept matching
module has completed to compute the similarity of the inputted concept.
According to the precious method of similarity calculation, when two concepts
and the weight of each similarity input, the multi-layer similarities and the total
matching are computed. Figure 126.1 show the matching result between ‘‘Quo-
tation’’ and ‘‘Payment application form’’.
126.4 Conclusion
In this paper, the model merging method from the partial model to the whole
model is studied. Moreover, the semantic similarity among the model instances is
analyzed from the various levels, and based on the semantic similarity, a series of
rules of the model merging are proposed, and then model integration is completed.
Finally, a prototype system of the model knowledge matching is developed.
And a case is described to validate the modeling method proposed in this paper.
References
Athena (2004) Advanced technologies for interoperability of heterogeneous enterprises networks
and their applications. FP6-2002-IST-1, integrated project description of Work, pp 19–36
Berre A, Elvesæter B, Figay N, et al (2007) The ATHENA interoperability framework. In:
Proceedings of the 3rd international conference on interoperability for enterprise software and
applications, enterprise interoperability II. Springer, Madeira, pp 569–580
Castano S, De Antonellis V, De Capitani di Vimercati S (2005) Global viewing of heterogeneous
data sources. IEEE Trans knowl Data Eng 13(2):277–297
Cui Z, Jones D, O’Brien P (2001). Issues in ontology-based information integration. http://
www.csd.abdn.ac.uk/ebiweb/papers/cui.pdf
Horrocks I, Peter F, Van Harmelen Frank (2003) From SHIQ and RDF to OWL: the making of a
web ontology language. J Web Seman 1(1):7–26
Mike U, Martin K, Stuart M (1998) Enterprise ontology. Knowl Eng Rev 13(1):31–98
Missikof M, Schiappelli F (2003) A controlled language for semantic annotation and
interoperability in e-business applications. In: Proceedings of the second international
semantic web conference, Sanibel Island, pp 1206–1211
Pulido JRG, Ruiz MAG, Herrera R et al (2006) Ontology language for the semantic web: a never
completely updated review. Knowl Based Syst 19(7):489–497
Thiagarajan RK, Srivastava AK, Pujari AK (2006) BPML: a process modeling language for
dynamic business models. In: Proceedings of the 4th IEEE international workshop on
advanced issues of e-commerce and web-based information systems, Califoria, pp 222–225
Vernadat F (2002) UEML: towards a unified enterprise modeling language. Int J Prod Res
40(17):4309–4321
Chapter 127
Outpatient Scheduling in Highly
Constrained Environments: A Literature
Review
Xiao-dan Wu, Mohammad T. Khasawneh, Juan Hao
and Zhan-ting Gao
Abstract This paper provides a comprehensive survey of research on scheduling
in outpatient services. An effective scheduling system has the goal of matching
demand with capacity so that resources are better utilized, especially in highly
constrained environments. This paper presents a general problem formulation and
modeling considerations. It also provides taxonomy of methodologies used in the
literature. The current literature fails to develop general guidelines that can be
applied to design outpatient scheduling systems. Therefore, we identify future
research directions that provide opportunities to expand the existing knowledge
and close the gap between theory and practice. Our paper presents a literature
review about four primary aspects: allocation of outpatient resources (R), outpa-
tient appointment model (A), patient preferences (P), and research methodology
for outpatient scheduling (M) under highly constrained environments. The models
presented are focused on three outpatient appointment models (i.e., the traditional
model, carve-out model, and advanced access model).
Keywords Allocation of outpatient resources Highly constrained environments
Outpatient appointment model Outpatient scheduling research methodology
Patient preferences
X. Wu (&) J. Hao Z. Gao
School of Management, Heibei University of Technology, Tianjin, China
e-mail: [email protected]
J. Hao
e-mail: [email protected]
Z. Gao
e-mail: [email protected]
M. T. Khasawneh
State University of New York at Binghamton, Binghamton, NY 13902, USA
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1203
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_127,
Ó Springer-Verlag Berlin Heidelberg 2013
1204 X. Wu et al.
127.1 Introduction
Nowadays appointment reservations can be made in advance in many health
centers and outpatient clinics in an effort to improve the utilization of resources.
However, the randomness of outpatient appointment models should be emphasized
in scheduling under constrained conditions because of the uncertainty associated
with patient demand, arrival patterns, and preferences. Typically, for some
healthcare services, most appointment time-slots are prepared for those randomly
arriving patients (i.e., walk-in patients) and only a small portion is available for
scheduled patients. Under the condition of appointments in advance, for example,
some patients may accept any time offered while others may ask to be treated at
certain times that are convenient to them. Some may not mind the waiting time.
Some patients may only choose to see their own doctor. Therefore, one of the
primary areas of research can focus on how to arrange all patients and allocate
available resources according to their changing preferences.
In 2008, Gupta and Denton (2008) talked about various challenges and
opportunities facing appointment scheduling in healthcare. They introduced
scheduling from three aspects, namely primary care, specialty clinics, and surgery
appointments, with a focus on four factors: arrival process, service process,
patients and provider preferences and incentives, and performance measures. In
addition to providing specific application areas and research issues, another
researcher (Erdogan and Denton 2009) presented a review on planning and
scheduling problems that have been developed for three stages: preoperative stage,
intra-operative stage, and postoperative stage. Furthermore, the authors also pro-
vide new insights for future areas of research. From the perspective of appointment
scheduling, we introduce patient preferences and resource allocation in an overall
outpatient scheduling methodology.
Our work presents a literature review about the above four aspects (i.e., RAPM)
under highly constrained environments. The models in our paper are mainly about
three outpatient appointment models (i.e., traditional model, carve-out model, and
advanced access model) and several types of appointment systems. Patient pref-
erences are divided into three aspects: assigned appointment time, appointment at
convenient time for patient, and walk-ins. Outpatient resources include healthcare
providers, staff, and equipment. Those scheduling models in the literature are
either cost-oriented or revenue-oriented. Simulation, dynamic programming, and
heuristic methods, etc., are the main methodologies used in outpatient scheduling.
This paper is organized as follows. In the first section, we introduce the topic
and purpose of this research. Section 127.2 describes the relevant literatures. In
Sect. 127.3, we introduce the various outpatient appointment models and compare
them. Section 127.4 reviews patient preferences. The allocation of outpatient
resources and research methodology of outpatient scheduling are discussed in
Sects. 127.5 and 127.6, respectively. Finally, in Sect. 127.7, open challenges and
possible future research directions are discussed. We also provide related litera-
tures in the appendix.
127 Outpatient Scheduling in Highly Constrained Environments 1205
127.2 Problem Definition and Formulation
The uncertainty in the service related variables represent the primary challenge in
outpatient scheduling. Highly constrained environments determine the conditions
which should be considered in the selection of outpatient appointment models,
patient preferences, allocation of outpatient resources, and outpatient scheduling
research methodology. With different appointment models, highly variable pref-
erences of treatment time, and various scheduling methodologies, we can derive
different optimal policies for making appointments. Under optimal conditions,
those four highly constrained environments (Fig. 127.1) should be considered and
all service related processes must be quantified. Early studies in the literature
provide significant research about surgical scheduling, with outpatient cost and
revenue as objective functions. However, to the best of the authors’ knowledge,
there is limited literature on the scheduling process under highly constrained
environments. As stated earlier, there are a significant number of constraints in the
outpatient scheduling process, such as patient preferences and allocation of out-
patient resources.
Figure 127.1 shows that highly constrained environments mainly consist of four
elements: outpatient appointment model, patient preferences, allocation of out-
patient resources, and outpatient scheduling research methodology. Patients arrive
to the system with the following distributions: uniform, empirical, and lognormal.
Traditional model requires patients to accept any scheduled or service time pro-
vided to them. The work (Fries and Marathe 1981) considers patients’ waiting
Fig. 127.1 Highly constrained environments of outpatient Scheduling
1206 X. Wu et al.
time, as well as the idle time and overtime of providers and staff as performance
metrics or objective functions and utilizes dynamic programming and queuing
models to maximize system capacity.
For instance, according to different outpatient appointment models, different
patients’ preferences may lead to different impact on the objective function under
the same constraints. The distribution and utilization of outpatient resources can be
analyzed as a function of different modes of capacity allocation considering the
changing preferences of different patient groups. Finally, different results can be
obtained when choosing different outpatient scheduling methodologies, with the
ultimate goal being to find a better research methodology of outpatient scheduling,
which further motivates the research presented in this paper.
127.3 Outpatient Appointment Model
There are several ways for determining the appropriate outpatient appointment
model. When an appointment has been made, patients have to wait for a long time,
and sometimes their treatments will be postponed. Thus, the patient may not be
able to see their own doctor which could be the output of an inefficient outpatient
appointment scheduling system. This could also result in poor communication
between patients and doctors and could lead to unnecessary costs. Based on this
fact, the paper attempts to select the appropriate outpatient appointment model. As
stated earlier, there are generally three outpatient appointment models: traditional
model, carve-out model, and advanced access model (Murray and Tantau 2000).
In the advanced access model, every doctor has available appointment slots,
which improves the availability of outpatient services. When existing capacity is
unable to meet patient demand, advanced access models become more advanta-
geous. This model can better balance supply and demand. Though patients make
diverse choices when making reservations, this method only takes into consider-
ation reservations made in advance and walk-ins, which will reduce the variability
in patient types. Moreover, the model can increase the effective utilization of
resources, especially bottleneck resources, such as expensive equipment in out-
patient clinics. Table 127.1 illustrates the daily capacity available on physicians’
schedules in the three access models described in the paper, where TPA refers to
the proportion of appointments; WU refers to walk-ins and the urgent; C-R refers
to cost-revenue; and CP refers to capacity.
Table 127.1 Outpatient Model TPA (%) WU CP C-R
appointment model
TM 100 N L Higher-L
COM [50 Y H H–H
AAM \50 Y Higher L-Higher
127 Outpatient Scheduling in Highly Constrained Environments 1207
In the traditional model (TM) (Bowers and Mould 2005; Guo et al. 2004; Gupta
and Wang 2007; Hassin and Mendel 2008; Huang 2008; LaGanga and Lawrence
2007; Murray and Tantau 2000; Muthuraman and Lawley 2008; Ogulata et al.
2009; Turkcan et al. 2010), the schedule is completely booked in advance; same-
day urgent care is either ignored or added on top of existing appointments. In a
carve-out model (COM) (Chakraborty et al. 2010; Chao et al. 2003; Fries and
Marathe 1981; Gallucci et al. 2005; Green and Savin 2007; Kaandorp and Koole
2007; Patrick et al. 2008), appointment slots are either booked in advance or held
for same-day urgent care; same-day non-urgent requests are satisfied in future
time. In advanced access model (AAM) (Murray and Berwick 2003; Murray and
Tantau 1999; Green et al. 2006; Kim and Giachetti 2006; Klassen and Rohleder
1996; Liu et al. 2010; Qu and Shi 2011), where practices focus on doing today’s
work today, there is true capacity. The majority of appointment slots are open for
patients who call that day for routine, urgent, or preventive visits.
Our paper focuses on the overall process of scheduling, which stresses the
proportion of scheduled patients. Table 127.2 emphasizes that every patient is
assigned to a time block, and that the number of patients a doctor can serve in a
certain period of time is fixed. The types of block appointment systems are shown
in Table 127.2.
The work (Fries and Marathe 1981) shows that a multiple-block system is more
feasible when the number of patients change. It was found that it is better to
expand the size of the reservation model. The literature also gave the appropriate
weight to patient’s waiting time, doctor’s idle time and overtime, in an effort to
compare different booking systems using those performance metrics. The paper
(Patrick et al. 2008) applied dynamic programming to schedule multi-priority
patients in the diagnosis of resources. The size of patients in a multi-block
appointment system is not carried out using a dynamic programming model. The
study found that the impact of patients with different priority on outpatient costs is
large.
The paper (Hassin and Mendel 2008) studied the patient no-show rate in the
single-block system, and investigated the degree to which the model is influenced
by outpatient costs and revenue. In the multiple-block system, the patient’s waiting
time cost, no-show costs, and service costs with no-show rate are examined. The
study found that patient no-show cost has a smaller impact than the services cost
on outpatient scheduling. Through the introduction of the above three studies
(Fries and Marathe 1981; Hassin and Mendel 2008; Patrick et al. 2008), it can be
clearly seen that the previous literature studied single-block and multiple-block
systems, with the primary objective function being the overall cost, with variations
in the outpatient factors considered.
1208
Table 127.2 The type of appointment systems
Type Representation Explanation References
Single-block n n patients arrive at same time Hassin and Mendel (2008) and Kim and
Giachetti (2006)
Individual-block/Fixed-interval 1 1 1 1 1 An initial-block of 1 patient Green and Savin (2007) and LaGanga and
Lawrence (2007)
Multiple-block/Fixed-interval m m m m m patients individually at intervals equal to Cayirli et al. (2006) and Patrick et al. (2008)
the mean service times of patients
Single-block/Individual-block n m patients with intervals set equal to twice the Green et al. (2006)
1 1
mean service time, m [= 1
Single-block/Multiple-block n An initial-block of n patients, others call Gupta and Wang (2007) and Kaandorp and
m m m patients 1-at-a-time, n [= 1 Koole (2007)
Variable-sized multiple-block n1 An initial-block of n0 patients, m0 -at-a-time patients Chakraborty et al. (2010) and Fries and Marathe
n2 n3
with intervals set equal to twice the mean service (1981)
time, n0 [ m0
X. Wu et al.
127 Outpatient Scheduling in Highly Constrained Environments 1209
127.4 Patient Preference
In recent years, extensive studies in the literature emphasized the importance of
taking customer preferences into account using various choice models in different
application areas. In particular, the discrete choice model has received major
attention in economics, marketing, and operations literature, with significant
research conducted to develop methods that simulate choice probabilities.
Therefore, we will introduce the previously mentioned models including: Inde-
pendent demand model (IDM), the multinomial logit model (MNL), the random
utility-maximization model (RUM), and the Independent from irrelevant alterna-
tives (IIA) (Gupta and Wang 2007). This way, our paper makes a simple intro-
duction about the importance of patient preference in scheduling systems.
Patient preferences (PP) mainly include the following aspects. Some patients
want to be treated at the exact day they make appointments (Bowers and Mould 2005;
Chao et al. 2003; Fries and Marathe 1981; Green et al. 2006; Klassen and Rohleder
1996; Muthuraman and Lawley 2008; Ogulata et al. 2009; Turkcan et al. 2010), while
others prefer their treatment to be performed when it is convenient since they do not
mind waiting for some time (Chakraborty et al. 2010; Erdogan and Denton 2009;
Green and Savin 2007; Guo et al. 2004; Hassin and Mendel 2008; LaGanga and
Lawrence 2007; Liu et al. 2010). Some patients only prefer random arrivals (Gupta
and Denton 2008; Gupta and Wang 2007; Kaandorp and Koole 2007; Qu and Shi
2011). PP adds to the complexity of the mathematical modeling process in which
better optimized scheduling policies can be set and optimal number of patients of
urgent reservations can be found to improve revenue (Gupta and Wang 2007).
In recent years, there have been few studies that focus on patient preferences in
healthcare systems, unfortunately. The work (Talluri and Ryzin 2004) analyzed
consumer choice model in revenue management, where they demonstrate that
some patient preferences make the calculation of the optimization criteria simpler.
The study (Zhang and Cooper 2005) considered a seat-allocation problem in which
there are multiple flights between the same start and destination points. The paper
(Gupta and Wang 2007) examined outpatient services with a Markov decision
process model according to patient preferences with revenue being the objective
function. When there is more than one doctor in the outpatient clinic, patient
preferences differ, thereby making the optimization criteria even more complex.
It is observed that many of the published studies considered the situation when
patients accept any appointment time offered to them. As expected, the dynamics
of the scheduling system varies with different patient preferences. The paper
(Patrick et al. 2008) divided outpatients into several priorities without considering
patient preferences, wherein each patient is offered sometime to accept the CT
scan appointment. The scheduling manager even has the right to reject patients or
go overtime to complete the scan. However, in the literature, patients can choose to
accept an offered time or select the appropriate time for them. The paper (Qu and
Shi 2011) studied the effect that patients’ preferences have on outpatient
appointment scheduling in advanced access model using Markov chains.
1210 X. Wu et al.
127.5 The Allocation of Resources in Outpatient Clinics
Outpatient resources include many elements, such as providers, staff, and equip-
ment. Furthermore, there might be a single department or multiple departments in
outpatient clinics. The paper (Chao et al. 2003) proposed multi-block appointment
and scheduling system based on patients’ waiting time, provider’s available
appointment slots, and other factors. The focus was to determine a reasonable
distribution of outpatient resources, which was not found to be proportional with
other factors.
The paper summarizes the relevant literature regarding outpatient resources,
including slack capacity (SC) (Chakraborty et al. 2010; Huang 2008), the penalty
(P) (Chakraborty et al. 2010; Fries and Marathe 1981; Gupta and Wang 2007; Kim
and Giachetti 2006; Liu et al. 2010; Patrick et al. 2008), cost (C) (Fries and
Marathe 1981; Hassin and Mendel 2008; Kim and Giachetti 2006; Klassen and
Rohleder 1996; Liu et al. 2010; Muthuraman and Lawley 2008; Patrick et al.
2008), providers and staff (PS) (Fries and Marathe 1981; Gupta and Denton 2008;
Gupta and Wang 2007; Guo et al. 2004; Hassin and Mendel 2008; Huang 2008;
Klassen and Rohleder 1996), revenue (R) (Bowers and Mould 2005; Chakraborty
et al. 2010; Fries and Marathe 1981; Gupta and Denton 2008; Gupta and Wang
2007; Kim and Giachetti 2006; Liu et al. 2010; Muthuraman and Lawley 2008;
Patrick et al. 2008), equipment resources (ER) (Bowers and Mould 2005; Guo
et al. 2004; Gupta and Denton 2008; Patrick et al. 2008). While some studies
focused on costs and revenue to evaluate the scheduling process, others considered
slack capacity. In those studies (Huang 2008), slack capacity is said to be the idle
time of staff or equipment, wherein idle time varies between departments. The
paper (Chakraborty et al. 2010) studied both the idle time and overtime with
consideration given to no-show rate and service time distribution based on out-
patient services’ revenue.
We can see in the paper that only four studies focus on outpatient equipment.
The work (Guo et al. 2004) emphasized the efficiency of equipment under certain
scheduling processes. The study (Bowers and Mould 2005) focused on the utili-
zation of equipment shared by outpatient and inpatient clinics. The paper (Gupta
and Denton 2008) mentioned the distribution problem of special facilities in
outpatient clinics, and showed the efficiency when taking into consideration spe-
cial patient conditions. The work (Patrick et al. 2008) mainly focused on the
scheduling process of patients with different priorities. In equipment related
research (Bowers and Mould 2005; Guo et al. 2004; Gupta and Denton 2008;
Patrick et al. 2008), the evaluation should not be focused only on facility effi-
ciency, but should also consider facility planning and department layout.
127 Outpatient Scheduling in Highly Constrained Environments 1211
127.6 Outpatient Scheduling Research Methodology
There are numerous methodologies to solve this scheduling problem, with many
using simulation (Bowers and Mould 2005; Cayirli and Viral 2003; Guo et al.
2004; Hassin and Mendel 2008; Huang 2008; Klassen and Rohleder 1996; Ogulata
et al. 2009) and regression analysis (Gallucci et al. 2005; Hassin and Mendel 2008;
Kim and Giachetti 2006; LaGanga and Lawrence 2007). Heuristic methods (Green
et al. 2006; Gupta and Wang 2007; Liu et al. 2010) and the curve fitting
approaches (Muthuraman and Lawley 2008; Qu and Shi 2011) are relatively rare.
The variables can involve many aspects in the scheduling process. Revenue and
cost are widely used as performance metrics, although other measures such as
resource utilization have also been used. We know that genetic algorithms
(Kaandorp and Koole 2007) and local search algorithms (Turkcan et al. 2010) are
used only in two studies. However, we think that genetic algorithms can be a
promising research tool.
The work (Fries and Marathe 1981) used dynamic programming to determine
the optimal block sizes for the next period given that the number of patients
remaining to be assigned is known. They present an approximate method to apply
the dynamic results to generate a schedule for the static version.
The paper (Ogulata et al. 2009) used simulation to analyze different conditions,
such as the percentage of unaccepted patients, treatment delay, number of patients
waiting in queue, normal capacity usage ratio, and slack capacity usage ratio,
which radiology patients need to schedule. Simulation has shown that in systems
with high frequency, the percentage of unaccepted patients is mostly determined
by maximum waiting parameter rather than slack capacity; the treatment delay is
completely determined by the slack capacity; and the main factor that affects the
treatment delay is maximum waiting time.
A study (Qu and Shi 2011) assessed the impact of patient preferences and
providers/staff capacity using mathematical modeling. They included different
patient choices which belong to the IIA method. The work (Gupta and Wang 2007)
assumed patients choices were subject to IDM, MNL, RUM methods. Different
models used linear programming to study outpatient revenue as a target and to find
community cost according to bound heuristic methods.
The paper (Turkcan et al. 2010) proposed genetic algorithms to plan and
schedule the entire chemotherapy cycle. The main factors are the height variation
of the resource requirements, such as treatment time, nursing time, and pharmacy
time, so that limited resources are fully utilized. The work (Liu et al. 2010) also
used a dynamic heuristic method to study patients’ no-show behavior. The sim-
ulation analysis shows that the method is more suitable when the number of
patients exceed the outpatient clinic’s actual capacity.
1212 X. Wu et al.
127.7 Open Challenges and Future Research
We mainly pay close attention to patient preferences and the allocation of
resources in the outpatient scheduling process. In recent years, this emerged as one
of the main challenges with significant potential for various research opportunities.
Research to find an optimal outpatient scheduling process that is suitable for the
Chinese healthcare system is relatively rare, and suitable scheduling rules for
medical treatment, especially given the domestic characteristics, were not found.
As stated earlier, most researchers consider cost and revenue as their goals.
However, from the customer perspective, patients’ waiting time is the most
important issue to consider in an outpatient scheduling system.
Most of the literature discussed focuses on the uncertainty of patient prefer-
ences. Little efforts have been spent on managing the uncertainty in outpatient
demands. For instance, current literature focuses on patients waiting times, when
in fact the time spent in waiting lists may be more important from a healthcare
outcomes perspective. Despite the rich literature considering the allocation and
utilization of outpatient resources (e.g., idle time and providers and staff overtime,
patient’s waiting time) is still largely unexplored.
Due to different considerations, the optimal percentage of appointment time-
slots in each category is not researched. The determination of optimal variable-
sized multiple-blocks in appointment systems in this paper has been proposed.
Also, most of the literature focused on using simulation and dynamic program-
ming. Another research area that needs further study is optimal algorithms for
outpatient resources allocation. This has the potential to incorporate multiple
factors that impact patient satisfaction, such as the patient diagnostic process.
Acknowledgments This work is supported by Natural Science Foundation of Hebei
(F2009000111) and Social Science Foundation of National Education Department (12YJC
630235).
References
Bowers J, Mould G (2005) Ambulatory care and orthopedic capacity planning. Health Care
Manag Sci 8:41–47
Cayirli T, Viral E (2003) Outpatient scheduling in health care: a review of literature. Prod Oper
Manag 12(4):519–549
Cayirli T, Viral E, Rosen H (2006) Designing appointment scheduling systems for ambulatory
care services. Health Care Manag Sci 9:47–58
Chakraborty S, Muthuraman K, Lawley M (2010) Sequential clinical scheduling with patient no-
shows and general service time distributions. IIE Trans 42(5):354–366
Chao X, Liu L, Zheng S (2003) Resource allocation in multisite service systems with intersite
customer flows. Manag Sci 49(12):1739–1752
Erdogan SA, Denton BT (2009) Surgery planning and scheduling: a literature review. Med Decis
Making 30:380–387
127 Outpatient Scheduling in Highly Constrained Environments 1213
Fries BE, Marathe VP (1981) Determination of optimal variable-sized multiple-block appoint-
ment systems. Oper Res 29(2):324–345
Gallucci G, Swartz W, Hackerman F (2005) Impact of the wait for an initial appointment on the
rate of kept appointments at a mental health center. Psychiatr Serv 56(3):344–346
Green LV, Savin S (2007) Providing timely access to medical care: a queuing model. Harvard
Business School, Harvard
Green LV, Savin S, Wang B (2006) Managing patient service in a diagnostic medical facility.
Oper Res 54(1):11–25
Guo M, Wagner M, West C (2004) Outpatient clinic scheduling: a simulation approach. In:
Proceedings of the 2004 winter simulation conference
Gupta D, Denton B (2008) Appointment scheduling in health care: challenges and opportunities.
IIE Trans 40:800–819
Gupta D, Wang L (2007) Revenue management for a primary care clinic in the presence of
patient choice. Oper Res 56(3):576–592
Hassin R, Mendel S (2008) Scheduling arrivals to queues: a single-server model with no-shows.
Manag Sci 54(3):565–572
Huang YL (2008) An alternative outpatient scheduling system: improving the outpatient
experience. PhD Dissertation, University of Michigan
Kaandorp GC, Koole G (2007) Optimal outpatient appointment scheduling. Health Care Manag
Sci 10:217–229
Kim S, Giachetti RE (2006) A stochastic mathematical appointment overbooking model for
healthcare providers to improve profits. IEEE Trans Syst 36(6):1211–1219
Klassen KJ, Rohleder TR (1996) Scheduling outpatient appointment in a dynamic environment.
J Oper Manag 14:83–101
LaGanga LR, Lawrence SR (2007) Clinic overbooking to improve patient access and increase
provider productivity. Decis Sci 38:251–276
Liu N, Ziya S, Kulkarni VG (2010) Dynamic scheduling of outpatient appointments under patient
no-shows and cancellations. Manuf Serv Oper Manag 12(2):347–364
Murray M, Berwick DM (2003) Advanced access: reducing waiting and delays in primary care.
J Am Med Assoc 289(8):1035–1040
Murray M, Tantau C (1999) Redefining open access to primary care. Manag Care Quart
7(3):45–55
Murray M, Tantau C (2000) Some-day appointment: exploding the access paradigm. An FPM
classic. Fam Pract Manag 7(8):45–50
Muthuraman K, Lawley M (2008) A stochastic overbooking model for outpatient clinical
scheduling with no-shows. IIE Trans 40(9):820–837
Ogulata SN, Cetik MO, Koyuncu E, Koyuncu M (2009) A simulation approach for scheduling
patients in the department of radiation oncology. Med Syst 33:233–239
Patrick J, Putmeran M, Queyranne M (2008) Dynamic multi-priority patient scheduling for a
diagnostic resource. Oper Res 56(6):1507–1525
Qu XL, Shi J (2011) Modeling the effect of patient choice on the performance of open access
scheduling. Int J Prod Econ 129:314–327
Talluri K, Ryzin GV (2004) Revenue management under a general discrete choice model of
consumer behavior. Manag Sci 50(1):15–33
Turkcan A, Zeng B, Lawley MA (2010) Chemotherapy operations planning and scheduling.
Optimization Online, pp 1–30
Zhang D, Cooper WL (2005) Revenue management for parallel flights with consumer choice
behavior. Oper Res 53(3):414–431
Chapter 128
Realization of 3D Reconstruction of CAD
Model Based on Slicing Data
Ming Li and Quan-qing Li
Abstract This paper points out that the reverse engineering is the important
technology to realize product innovation based on the prototype. It puts forward a
new method to realize 3D reconstruction that takes slicing data of prototype as
original data, under the commercial CAD modeling software environment, and
briefly introduces the system developed by authors. The working process of this
system is to read in the slicing data of the prototype, after pretreatment and feature
recognition, to output feature data and realize 3D reconstruction under the
SolidWorks environment, at last to construct the CAD solid model. It lays a good
foundation for modifying the model so as to realize the product innovation. The
innovation of this system lies in: according to slicing data, it can directly construct
the CAD solid model. This paper also analyzes the key problems of the recon-
struction process, and indicates that this technology has obvious advantages for
mechanical manufacture filed.
Keywords 3D reconstruction CAD model Feature recognition Reverse
engineering
M. Li (&) Q. Li
Zhengzhou Institute of Aeronautical Industry Management, Zhengzhou,
People’s Republic of China
e-mail: [email protected]
Q. Li
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1215
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_128,
Ó Springer-Verlag Berlin Heidelberg 2013
1216 M. Li and Q. Li
128.1 Introduction
128.1.1 Basic Concept of Reverse Engineering
Reverse engineering is a science technology that is widely used in mechanical
manufacture, modern-design-method, computer hardware, computer software and
computer graphics. Reverse engineering is the important means of digestion and
absorption, innovation for imported products (Honsni and Ferreira 1994; Dasch-
bach et al. 1995; Puntambekar et al. 1994; Motavalli and Bidanda 1994; Chen and
Lin 1997; Abella et al. 1994; Liu et al. 1998; Liu and Huang 1992). The essential
difference between reverse engineering and copy technology is that the product
model constructed by reverse engineering is CAD model. After acquire CAD
model, it can be modified and re-designed in order to realize the innovation
purpose. The working process of reverse engineering is shown in Fig. 128.1.
128.1.2 Key Technology of Reverse Engineering
It is known from Fig. 128.1, digitizing the part and constructing CAD model are
the two key technology of reverse engineering (Luan et al. 2003).
Digitizing the part refers to adopt some measuring methods and equipments to
acquire the geometry coordinate of the part. Presently the measuring methods that
are used in industrial are coordinate measuring machines (CMM), laser beam
scanning, industrial computed tomography and layer-layer cutting image. Using
these methods can get every layers of slicing data of the prototype.
The method to construct CAD model that is commonly used in the reverse
engineering technology at home and abroad is: to recognize the border to the
slicing data automatically or manually; 3D dots are grouped according to the
feature single principle; to carry on surface modeling to each group dots; the last is
solid modeling (Chow et al. 2002; Huang et al. 2001). That is to link up each
surface to form a complete part. The surface modeling theory and algorithm has
basically ripe, but the research of oriented solid modeling of reverse engineering is
still not reach the ideal practical level, therefore, in constructing CAD model is
looking forward to have a breakthrough as soon as possible.
OK Drawing
Existing Digitized Finished
Prototype Data CAD modal Check Manufacture
Part
No
Fig. 128.1 The working process of reverse engineering
128 Realization of 3D Reconstruction of CAD Model Based on Slicing Data 1217
128.2 Functions of Reverse Engineering System Based
on Slicing Data
Reverse engineering system based on slicing data, short for SdRe system is the
reverse engineering system software of constructing products 3D model based on
slicing data, which is developed by us. SdRe system includes three function
modules: slicing data processing; feature recognition: 3D reconstruction.
128.2.1 Slicing Data Processing
The data obtained after digitizing the part is the bitmap image of all slicing layers
of the part prototype. Processing these slicing images have two steps: filtering out
noise and extracting borders. In image filtering technology, global filtering tech-
nology requires to know the signal or noise statistic model in advance, which for
the slicing images is almost impossible, so the SdRe system uses the local filtering
technology, which uses the local operators to do local treatment for the images in
turn.
The SdRe system is with a filter function library including a lot of filter
function. The user can select appropriate function for various image qualities and
modify the filtering parameters, in order to get the best filtering result. Then the
system extracts borders after filtering image, the image information is constructed
as the ring chain that is composed of interconnected pixels. Aiming at the slicing
data of object prototype in reverse engineering must be a ring of respective closed
and mutually disjoint; the system has the effective arithmetic to extract borders for
slicing data of object prototype in reverse engineering. While extracting borders
the system’s parameters can be modified by man–machine conversation to elim-
inate futility data like air holes and chips. Moreover, it can also distinguish
between the convex feature and concave feature for extracted border. The concave
means the solid surface materials to be removed, such as a hole, while the convex
means the solid surface to possess materials, such as a cylinder. After extracting
borders, the data become each closed ring composed of ordered dots, which is
named data ring.
128.2.2 Feature Recognition
SdRe system uses feature model to construct CAD model, so the work of stage 2 is
the feature recognition. Data ring after feature recognition is constructed the data
that expresses object prototype feature, that is named feature ring. This is the core
model of the system.
1218 M. Li and Q. Li
128.2.3 3D Solid Reconstruction
SdRe system selects the commercial CAD modeling software, such as SolidWorks,
as system support software of 3D modeling. SolidWorks is 3D modeling software
oriented computer. Its function is powerful, the cost is effective and it easily realize
the interfaces with the programming language and other commonly used CAD
software (Wen 2004). Moreover, on the support of commercial CAD modeling
software, the 3D solid model that is reconstructed can be modified, it can output the
part drawings and assembly drawings, as well as the files to be read by other
commonly used CAD software, it still can output the STL files for the rapid pro-
totyping (Liu 2004; Schreve et al. 2006). Using SolidWorks as support software, can
save the research work to realize basic function, to focus on the key technology of
reverse engineering. Solidworks provides the soft interface with programming
language. Using this interface, SdRe system develops the interface model. Under
SolidWorks environment, to run this model, input the feature data produced by
feature recognition model, the CAD model of the product can be reconstructed.
128.3 Feature Recognition
The feature recognition model is the core of SdRe system. It completes feature
recognition function, in order to realize 3D solid reappear. SdRe system read in the
data of data ring, output the feature ring data for SolidWorks modeling.
128.3.1 Feature Types
SdRe system to recognize the feature from the three levels of line, surface and
solid, the feature types are shown in Fig. 128.2.
From the standpoint of feature body, the features that are constructed by SdRe
system have two types: extruding body and extruding body of layer change. The
extruding body is the feature that the shape and size of cross-section do not
change; it is the equal cross-section body, such as the cylinder, prism. The
extruding body of layer change is the body with variable cross-section, the size and
shape of its cross-section are changed. The feature body of containing free surface
is one of the extruding body of layer change. From the standpoint of feature plane,
SdRe system can construct plane, cylinder surface, cone surface and free-form
surface, each surface can be outside surface, also can be inner surface.
From the standpoint of feature line, SdRe system can construct straight line, arc,
circle and free curve, as well as the polygon by their combination.
SdRe system recognizes the feature body and feature line with explicit rec-
ognition form, and the feature surface with implicit recognition form, for the
128 Realization of 3D Reconstruction of CAD Model Based on Slicing Data 1219
Fig. 128.2 Feature types
Product model
Feature body
Extruding body extruding body of layer
change
Feature surface
plane cylinder surface free-form surface
Feature line
straight line arc circle free curve
feature surface recognition is included in the recognition of the feature body and
feature line.
128.3.2 Feature Recognition Manners
As mentioned above, the data ring actually is the data that a certain feature of
object prototype is reflected in a certain cross-section. Some data ring that
expresses the same feature surface of the object prototype, distributes in different
cross-section are congregated, that is formed a new data chain, named solid ring.
Solid ring is actually the data ring collection that expresses the feature surface of
the object prototype. The solid ring after feature recognition is feature ring.
From the standpoint of automation, feature recognition has two manners:
interactive recognition and automatic recognition (Li et al. 2003). Automatic mode
can recognize the two types of feature line of straight line and free curve and all
feature body; interactive mode can recognize all feature line and feature body.
From the standpoint of checking the recognition results, in the interactive recog-
nition there are two manners of relatively recognition and absolute recognition.
For relative recognition, the system to the results of interactive recognition carries
on fitting operations by means of least squares, in order to check the recognition
correctness, if the error exceeds the specified threshold value (the value may be
modified by user), then give warning information, and the user decides to rec-
ognize again or ignore the system warning. Using absolutely identify way, directly
accept the artificial recognition results, no longer inspection. From the standpoint
of working process, there are direct and indirect recognition. When carry on
indirect recognition, the first is to recognize feature line, and then to recognize
feature body, it is mainly used for recognizing the complex feature body. Direct
recognition is to recognize the feature body directly, is mainly used for
1220 M. Li and Q. Li
recognizing the simple feature body. If need to recognize the complex feature
body, the system will guide users to recognize feature line.
128.3.3 Sub-Function Models
In order to carry on the feature recognition work efficiently and conveniently, it
has carried on the function decomposition and model partition to the system, and
developed the corresponding sub-function models. The structure of sub-function
model system is shown in Fig. 128.3. Its main function models are briefly
described as follows:
(1) Constructing solid ring
The process to construct the solid ring is the matching process of the data ring.
That is according to certain principles and algorithms, all data rings of the
same feature surface of the object prototype will be combined together to
construct a solid ring. In order to calculate simply and improve efficiency, this
software uses the data ring to match directly, that is using dot image of data
ring to match, rather than the dot image of data ring is vectorized into the
geometry image at first, then to match them again. In this way, it has spared
the vectorization work of thousands of data ring dots in hundreds of images,
greatly improving the processing speed.
Using dot image to match directly, at present there is no literature to introduce
the algorithm. The data ring matching method that is constructed by our
system, from the three aspects of shape, location and size of dot image to
judge the matching relation of dot image. Completely matching in certain error
range, compose the extruding body; incompletely matching, compose the layer
changed extruding body; cannot match at all, that illustrates the dot images are
belong to different feature surface separately. The modeling coefficient that
expresses the matching relations can be adjusted through modifying the sys-
tem variables by the system operator. Modeling coefficient size has determined
the feature modeling precision and the feature number.
(2) Images display
In the different stages of the system running, all kinds of images will be dis-
played in the screen timely for operator to understand the situation and control
the operation. After the solid ring is constructed, the operator can choose solid
ring to display wholly, partly or single ring displays. The operator can be used
the mouse to choose some one solid ring to edit or model. If need to recognize
feature line, it will display the plane figure of the data ring, the operator can drag
and move the mouse with the method of open Windows to choice the dot group.
(3) Solid ring editing
The construction of solid ring is completed automatically by system according
to the modeling coefficients. The system operator can use the solid ring editing
functions which is provided by the system to edit the solid ring, to carry on the
128 Realization of 3D Reconstruction of CAD Model Based on Slicing Data 1221
input/output
model
solid ring
whole
constructing solid
ring model
solid ring
solid
single
image display
model
plane
data
ring
delete
data ring
combine solid
editing model
ring
separate
solid ring delete
Feature recognition
editing model
Straight
circle line
polygon arc
closed free Free
curve curve
feature recognition closed Straight
model polyline line
cylinder
extrude
feature prism
editing model
cone
layer
change
extrude free-surface
body
Fig. 128.3 The sub-function models of feature recognition
operations of deleting, separating and combining for the solid ring. Deleting
solid ring, that is to modify the data of the original data ring in order to
eliminate futility data like air holes and chips. Separating is from a solid ring
(compound solid ring) to be separated another solid ring, one ring is became
two rings. Combining is two solid rings will be combined into one ring. When
combining, SdRe system will be according to the shape of the space position
and the plane position of the two rings to judge whether the solid rings can be
combined, if the system think them unfavorable combine, will give warning,
1222 M. Li and Q. Li
please the operator confirm or give up; if the system think them cannot
combine, it will give error messages, and refuse to combine.
(4) Data Ring Editing
The deleting operation for the data of solid ring can be done, that is to delete
the current data ring (slicing) that is composed of the solid ring, in order to
eliminate redundant data.
(5) Feature recognition
After solid ring is constructed, in order to construct the feature ring, the
geometric recognition to the data composed of solid ring need be completed.
The system uses the above various methods to recognize the feature line and
feature body.
(6) Feature ring editing
Feature ring that has constructed already is displayed in the form of a tree, in
order to do the editing work for it when necessary. This editing work is mainly
to meet the requirements of SolidWorks. The editing work includes two
contents, one is to adjust the modeling order of the feature ring, another is to
adjust the corresponding relation of data point in surface body.
128.4 Conclusion
3D reconstruction method that puts forward in this paper is to directly construct the
CAD solid model of the prototype based on slicing data of the prototype and under the
commercial CAD modeling software environment, this is a new method. Previously,
the reverse engineering method generally is to construct the local surface model of
the prototype first, then to match and joint the surfaces and get the whole surface
model. In the process of the surface matching and jointing, it is very complicated to
deal with the problems of surface tearing and overlap. The method that research in
this paper avoids these problems. The method is to realize 3D reconstruction in
commercial CAD software environment, so it will save much time to develop the
additional modeling software or models. In addition, in the mechanical manufacture
field, the most of parts are composed of the regular surfaces, so this method has
unique advantages in the reverse engineering of mechanical manufacture field.
References
Abella RJ, Daschbach JM, Mcnichols RJ (1994) Reverse engineering industrial application.
Comput Ind Eng 26(2):381–385
Chen L-C, Lin GCI (1997) An integrated reverse engineering approach to reconstructing free-
form surfaces. Comput Integr Manuf Syst 10(1):49–60
Chow J, Xu T, Lee S-M, Kengskool K (2002) Development of an integrated laser-based reverse
engineering and machining system. Int J Adv Manuf Technol 19:186–191
Daschbach J, Abella R, McNichols R (1995) Reverse engineering: a tool for process planning.
Comput Ind Eng 29:(1–4):637–640
128 Realization of 3D Reconstruction of CAD Model Based on Slicing Data 1223
Honsni Y, Ferreira L (1994) Laser based system for reverse engineering. Comput Ind Eng
26(2):387–394
Huang X, Du X, Xiong Y (2001) Modelling technique in reverse engineering. China Mech Eng
12(5):539–542 (in Chinese)
Li D, Wang M, Liu Y (2003) Research on interacted-modeling method for reverse engineering.
China Mech Eng 14(19):1677–1680 (in Chinese)
Liu Y (2004) Research on CAD modeling key technology of reverse engineering based slicing
feature. Zhejiang University, Hangzhou (in Chinese)
Liu Z, Huang C (1992) Reverse engineering design. China Machine Press, Beijing, p 116 (in
Chinese)
Liu Y, Hang J, Wan Y (1998) Reverse engineering and modern design. J Mach Des 16(12):1–4
(in Chinese)
Luan Y, Li H, Tang B (2003) Reverse engineering and its technologies. J Shan Dong Univ (Eng
Sci) 33(2):114–118 (in Chinese)
Motavalli S, Bidanda B (1994) Modular software development for digitizing systems data
analysis in reverse engineering application: case of concentric rotational parts. Comput Ind
Eng 26(2):395–410
Puntambekar NV, Jablokow AG, Joseph Sommer III H (1994) Unified review of 3D modal
generation for reverse engineering. Comput Integr Manuf Syst 7(4):259–268
Schreve K, Goussard CL, Basson AH, Dimitrov D (2006) Interactive feature modeling for reverse
engineering. J Comput Inf Sci Eng 6:422–424
Wen X (2004) Reverse engineering technique of complex surface product based prototype. Mech
Electr Inf 4(8):35–37 (in Chinese)
Chapter 129
Recommender System Based ‘‘Scenario-
Response’’ Types Post-Disaster
Emergency Supplies Planning
Gang Kou, Xin Zhao and Daji Ergu
Abstract In recent years, frequent outbreaks of unexpected natural disasters have
caused great threats and losses to people’s property and personal safety. Owe to
the reality that natural disaster has the characteristics of wide affecting-areas, long
duration, different needs of emergency supplies and inadequate rescue resources, it
is strongly important to achieve effective planning for limited emergency supplies.
The questions of how to ensure that each disaster-affected area gets emergency
supplies matched with its losses and all demands of emergency supplies from
different disaster areas met have become critical tasks of natural disaster emer-
gency management. Based on the necessity and urgency of the emergency supplies
as well as the real-time data of the scenario evolution of the disaster, this paper
proposes an algorithm which combines recommender system and social tagging
with allocation management to establish ‘‘scenario-response’’ type’s poster-
disaster emergency suppliers planning in order to reach higher performance in
post-disaster recovery.
Keywords Group decision making Post-disaster supplies planning Recom-
mender system Scenario-response Social tagging
G. Kou (&) X. Zhao D. Ergu
School of Management and Economics, University of Electronic Science
and Technology of China, Chengdu, China
e-mail: [email protected]
X. Zhao
e-mail: [email protected]
D. Ergu
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1225
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_129,
Springer-Verlag Berlin Heidelberg 2013
1226 G. Kou et al.
129.1 Introduction
The rapid development of economic globalization not only deepens the level of
national industrialization and urbanization, but also increases the property losses
and causalities which brought out by large-scale unexpected natural disasters.
When such an event happens, all disaster areas are in great demands for emergency
supplies (Fiedrich et al. 2000; Kevin and Liu 2004). Generally, different affected
areas by different natural disasters such as typhoon, flooding, drought or earth-
quake etc. may have exactly different needs for different supplies. As for allocation
management of emergency supplies in large-scale natural disasters, irrational
distribution of resources usually leads to further expanse of personnel and property
losses and deterioration of threats(Bakuli and Smith 1996; Zheng 2007), thus,
more effective way based on the real-time data of the affected area is needed to
achieve optimized post-disaster emergency supplies planning, which could ensure
its fairness and rationality, greatly help reconstruction after disaster, speed up the
recovery of order of daily life and production (Mezher et al. 1998).
At present, the allocation strategies of emergency supplies coping with large-
scale unexpected natural disasters have the problems described as follows: (1) don’t
take the actual emergency supplies needs of disaster areas into consideration, the
needless materials which don’t need are over-supply while the much-needed sup-
plies are on the contrary, which result in unnecessary waste of precious emergency
supplies; (2) don’t lay emphasis on the actual needs of different emergency supplies
from different areas(Toregas et al. 1971; Yuan and Wang 2009).
In a word, the existing post-disaster supplies plans are rough and simple, the
great imbalance of allocation makes them cannot respond well to the demands for
emergency supplies of disaster-affected areas (Chang et al. 2007; Mailler et al.
2003). Therefore, advanced technologies are gradually introduced into post-
disaster supplies planning.
129.2 Related Works
Because of the necessity and urgency of building more structure-optimized, high-
performance poster-disaster emergency supplies planning, many scholars have
carried out some researches and have achieved certain success. For instance, Zohar
and Albert (2008) developed a system dynamics simulation for complex high-tech
environment and calculated the expected net benefits; Kaan (Ozbay et al. 2004)
proposed mathematical programming models with probabilistic constraints to
address incident response and resource allocation problems; Barbarosoglu and Arda
(2004) introduced a two-stage stochastic programming framework for transportation
129 Recommender System Based ‘‘Scenario-Response’’ Types 1227
planning in disaster response; Chiu and Zheng (2007) developed a model formu-
lation and solution for real-time emergency response in no-notice disasters.
Therefore, it can be concluded that most of the studies on planning of post-
disaster emergency supplies paid more attention on the combination of allocation
management with available transportation strategies (Gwo-Hshiung et al. 2007;
Fang et al. 2007). However, it is obvious to see that numerous disaster areas
usually have different needs for emergency supplies, including their quantity and
category. The planning should be combined with the actual losses of the areas.
In this paper, based on the real-time data of the affected area and the evolution
of the scenario, an algorithm that combines recommender system and social tag-
ging with allocation management is proposed to establish a ‘‘scenario-response’’
type’s poster-disaster emergency suppliers planning in order to reach higher per-
formance in disaster emergency management and post-disaster recovery.
129.3 Methodology
The method of combining recommender system and social tagging information
with post-disaster supplies planning is described in detail as follows.
(1) Assume the number of areas which suffer different losses in a large-scale
unexpected natural disaster event is n, the set of the disaster areas is marked as I;
(2) Assume all of the disaster areas are considered as an integrated group G. Based
on the actual damages and losses of the areas, k-means algorithm is applied to
divide the group into m different clusters, which makes high similarity in the
internal cluster and low similarity among the clusters, as shown in Fig. 129.1.
(3) For each cluster in the target group, looking at two different aspects: the data
of damages and losses and tag information of emergency supplies, which are
described in details in the following.
Fig. 129.1 Divides the target
Target group
group into different clusters
K-means
Cluster 1 Cluster 2 Cluster m
1228 G. Kou et al.
129.3.1 The Aspect of the Data of Damages and Losses
(a) First, the average strategy is applied to aggregate the data of damages and
losses of the group member in every cluster, which results in a single vector
representing the integral damage degree of the disaster areas. Assume there are
n1 areas in the cluster; the first area’s vector of damages and losses is
R1 ¼ ða11 ; a12 ; a13 ; . . .; a1i Þ, the second area’s is R2 ¼ ða21 ; a22 ; a23 ; . . .; a2i Þ …
where i indicates the number of the appointed attributes that are used to
appraise the losses and damages of disaster-affected areas. Calculate the single
vector that represents the integral damage degree of the disaster areas in the
same cluster and denoted as Rcluster1, each number in the vector is the average
damages and losses for the attribute i by all of the areas in the cluster.
Xn1
Rcluster1 ¼ R =n1
j¼1 j
ð129:1Þ
(b) Second, after getting the single vector that represents the integral damage
degree of the disaster areas in the same cluster, consider it as a virtual disaster
area v1 . To provide a recommendation list to show the needed quantity of the
emergency supplies of the areas in the cluster, it is necessary to figure out the
ratio in the whole dataset under the same attribute of every cluster, the ratio is
marked as rm;i , representing cluster m‘s ratio in the whole dataset under
attribute i. Suppose that we have already known the total amount tj of the
emergency supplies through an emergency rescue site, j is the number of the
category of the emergency resources. The recommendation list drawn from the
specific data can be obtained by following formula.
Rc ¼ r m;i tj ð129:2Þ
129.3.2 The Aspect of Tag Information of Emergency
Supplies
(a) Classify the whole attributes which could represent the damages and losses of
the disaster areas into different categories in accordance with their intrinsic
correlation. Each category corresponds to different class of emergency sup-
plies, for example, collapsed and damaged houses belong to the loss of
buildings which could indicate the demand for emergency tent in an unex-
pected event;
129 Recommender System Based ‘‘Scenario-Response’’ Types 1229
(b) According to the real-time data of damages and losses of the disaster areas in
each cluster, figure out the tag information of the needed emergency supplies
in order, select the top-5 frequency tags as the most tag set which could shown
the demands for the rescue resources of the clusters;
(c) Based on the ratio of the frequency of the tags deduced from the above steps,
we can calculate another different recommendation list for quantity and cat-
egory of emergency supplies which are needed by the clusters Rt .
What is to be addressed here is that the aspect of the data of damages and
losses can be helpful in figuring out the similarity of demands for emergency
supplies of different areas, while the tag information is used to figure out the
differences between different areas in the same cluster, therefore it manages to
compensate the shortage of simple post-disaster supplies planning strategies,
which can fully reflect the needs of the disaster areas.
(d) Calculate the average of the two recommendation list, and then the quantity of
the emergency supplies which allocated to each cluster can be easily obtained.
129.3.3 The Allocation Strategy Within the Same Cluster
It is true that even in the same cluster, different areas may have different demands
for the quantity and category of the emergency supplies. Therefore, it is also
important to find out good way to solve this problem. In the following section, the
tag information of the disaster areas is used, and the detailed steps are proposed.
129.4 Case Study
In this section, a case study is conducted to explain the detailed steps of the
proposed strategy for allocation planning of post-disaster emergency supplies.
Because of the existed unit inconsistencies in the original dataset, it is necessary
to standardize the data first; the preprocessed dataset is shown in the Table 129.1.
Step1: Divide the fourteen disaster areas into three different clusters according
to their actual data of damages and losses. The result is shown as follows:
Cluster1: Anhui, Fujian, Jiangxi, Shandong, Henan, Hubei, Hunan, Shanghai,
Jiangsu, Hainan, Yunnan
Cluster2: Guangdong, Guangxi
Cluster3: Zhejiang
Step2: Average strategy is used to calculate the single vector that represents the
integral damage degree of the disaster areas in the same cluster.
1230 G. Kou et al.
Table 129.1 The standardized dataset
Attributes
Areas Victim Death Shift Victim Drought Collapsed Damaged Economic
crop houses houses loss
Shanghai 0.0006 0.0049 0.0085 0.0021 0.0000 0.0000 0.0000 0.0075
Jiangsu 0.0378 0.0071 0.0063 0.0932 0.0051 0.0088 0.0055 0.0101
Zhejiang 0.2059 0.0523 0.2874 0.1230 0.1119 0.0423 0.0935 0.0964
Anhui 0.0616 0.0487 0.0145 0.0830 0.1241 0.0664 0.0743 0.0998
Fujian 0.0535 0.0613 0.2268 0.0465 0.0459 0.0478 0.0744 0.0866
Jiangxi 0.0375 0.0340 0.0179 0.0350 0.0472 0.0818 0.0569 0.0508
Shandong 0.0042 0.0001 0.0071 0.0042 0.0005 0.0118 0.0104 0.0019
Henan 0.0000 0.0371 0.0207 0.0008 0.0001 0.0292 0.0045 0.0040
Hubei 0.0016 0.0055 0.0003 0.0020 0.0001 0.0001 0.0073 0.0020
Hunan 0.0331 0.0728 0.0241 0.0352 0.0915 0.0585 0.0261 0.0524
Guangdong 0.2762 0.3014 0.1528 0.2349 0.4533 0.4259 0.4580 0.1330
Guangxi 0.1825 0.0938 0.1744 0.2297 0.0993 0.1446 0.1345 0.2466
Hainan 0.0670 0.0717 0.0561 0.0545 0.0167 0.0025 0.0090 0.1522
Yunnan 0.0385 0.2096 0.0031 0.0559 0.0044 0.0804 0.0457 0.0568
Step3: Suppose the emergency supplies collected from an emergency rescue
site are mineral water, instant food, relief tents and prevention medicines, the tag
corresponding to each supply is water, food, tents and medicine. Here, t1 ; t2 ; t3 ; t4
indicate the quantity of the four kinds of supplies in the rescue site.
Step4: Calculate the ratio of the data that represents the damages and losses of
the disaster areas.
Step5: Classify the whole attributes which could represent the damages and
losses of the disaster areas into different categories in accordance with their
intrinsic correlation.
The number of the attributes used in this case is up to 8, which can be classified
into three categories, as shown in Table 129.2.
Step6: From Table 129.2, it is easy to figure out that the three categories of
attributes are independently corresponding to prevention medicines, instant food
and mineral water, relief tents. Therefore, the recommendation list for different
Table 129.2 Virtual disaster area for every cluster
Attributes
Clusters Victim Death Shift Victim Drought Collapsed Damaged Economic
crop houses houses loss
Cluster1 0.030491 0.050255 0.035036 0.037491 0.030509 0.035209 0.028555 0.047645
Cluster2 0.22935 0.1976 0.1636 0.2323 0.2763 0.28525 0.29625 0.1898
Cluster3 0.2059 0.0523 0.2874 0.123 0.1119 0.0423 0.0935 0.0964
129 Recommender System Based ‘‘Scenario-Response’’ Types 1231
Table 129.3 The ratio of the data represents the damages and losses of the disaster areas
Attributes
Clusters Victim Death Shift Victim Drought Collapsed Damaged Economic
crop houses houses loss
Cluster1 0.065468 0.167429 0.072086 0.095447 0.072865 0.097059 0.068263 0.142717
Cluster2 0.492441 0.658328 0.3366 0.591409 0.659885 0.786335 0.708216 0.568527
Cluster3 0.442091 0.174244 0.591314 0.313144 0.26725 0.116606 0.223521 0.288756
Table 129.4 Classify the whole attributes into three categories
Attributes Category of damages and losses
People suffered Crop destruction Losses
Victim Death Shift Victim Drought Collapsed Damaged Economic
crop houses houses loss
Table 129.5 Average score of the three categories attribute
Cluster Category of damages and losses
People suffered Crop destruction Crop destruction
Cluster1 0.101661 0.084156 0.10268
Cluster2 0.49579 0.625647 0.687692
Cluster3 0.40255 0.290197 0.209628
quantity for different emergency supplies can be obtained. Acquire the average score
of the three categories attributes of the three clusters. The recommendation list Rc is:
0 1 0 1
0:101661 0:084156 0:10268 t4
Rc ¼ @ 0:49579 0:625647 0:687692 A @ t1 þ t2 A
0:40255 0:290197 0:209628 t3
Step7: As for every cluster, figure out the frequency of the tag of the demands
for the emergency supplies. For simplicity, cluster2 is used as an example,
including Guangdong, Guangxi based on the standardized dataset; it is easy to get
the tag information corresponding to the urgency of the emergency supplies.
Guangdong: tents, tents, instant food, medicine, medicine, medicine;
Guangxi: instant food, medicine, medicine, instant food;
Thus, the frequency of the tags can be obtained.
Step8: The similar method described in Step6 can be used to obtain the
recommendation list Rt . Acquire the average score of Rc and Rt , then the final
recommendation list can be obtained (Tables 129.3, 129.4, 129.5).
Step9: The allocation within the cluster could rely more heavily on the prin-
ciples and tag rank.
1232 G. Kou et al.
129.5 Conclusion
Effective planning of post-disaster emergency supplies is critical for improving the
level of natural disaster emergency management. Based on the existed researches,
an algorithm is proposed to establish a ‘‘scenario-response’’ type’s poster-disaster
emergency suppliers planning in disaster emergency management and post-
disaster recovery by combining recommender system and social tagging with
allocation management. A case study is conducted to show the simplicity and
feasibility of the proposed management strategy.
Future work will lay emphasis on more mature strategy and more complex
situation. The distance between emergency rescue site and disaster areas is an
important factor in affecting the allocation of emergency supplies, thus it should be
taken into consideration. Otherwise, the data of damages and losses of the disaster
areas are changing with the rescue activities; therefore, building a dynamic post-
disaster supplies planning system based on the idea of ‘‘scenario-response’’
strategy is also important.
References
Bakuli DL, Smith JM (1996) Resource allocation in state-dependent emergency evacuation
networks. Eur J Oper Res 89:543–555
Barbarosoglu G, Arda Y (2004) A two stage stochastic programming framework for
transportation planning in disaster response. J Oper Res Soc 55:43–53
Chang M-S, Tseng Y-L, Chen J-W (2007) A scenario planning approach for the flood emergency
logistics preparation problem under uncertainty. Transp Res Part E Logist Transp Rev
43:737–754
Chiu Y-C, Zheng H (2007) Real-time mobilization decisions for multi-priority emergency
response resources and evacuation groups: Model formulation and solution. Transp Res Part E
Logist Transp Rev 43:710–736
Fang L (2007) Dynamic multi-objective emergency relief logistics: a decision support system
framework. In: Proceedings of 2007 IEEE international conference on grey systems and
intelligent services, pp 779–783
Fiedrich F, Gehbauer F, Rickers U (2000) Optimized resource allocation for emergency response
after earthquake disasters. Saf Sci 35:41–57
Gwo-Hshiung T, Hsin-Jung C, Tsung Dow H (2007) Multi-objective optimal planning for
designing relief delivery systems. Transp Res Part E Logist Transp Rev 43:673–686
Kevin F, Liu R (2004) Agent-based resource discovery architecture for environmental emergency
management. Expert Syst Appl 27:77–79
Mailler R, Lesser V, Horling B (2003) Cooperative negotiation for soft real-time distributed
resource allocation. In: Proceedings of the second international joint conference on
autonomous agents and multi-agent systems, pp 224–239
Mezher T, Chedidy R, Zahabi W (1998) Energy resource allocation using multi-objective goal
programming: the case of Lebanon. Appl Energy 61:175–192
Ozbay K, Xiao W-H, Iyigun C, Baykal-Gursoy M (2004) Probabilistic programming models for
response vehicle dispatching and resource allocation in traffic incident management. In: I &
SE-working paper, pp 4–34
129 Recommender System Based ‘‘Scenario-Response’’ Types 1233
Toregas C, Swain R, Revelle S, Bergman L (1971) The location of emergency service facilities.
Oper Res 19:l363–1373
Yuan Y, Wang D-W (2009) Path selection model and algorithm for emergency logistics
management. Comput Ind Eng 56:1081–1094
Zheng Y-J (2007) Distributed cooperative planning and scheduling for disaster management. In:
Proceedings of the IEEE international conference on automation and logistics, August,
pp 1172–1176
Zohar L, Albert IG (2008) Resource allocation under uncertainty in a multi-project matrix
environment: is organizational conflict inevitable? Int J Project Manage 26:773–788
Chapter 130
Research on the Simulation Case
of Traffic Accident
Chao Wei, Xiang Gao and Miao-xin Nie
Abstract In order to accurately reconstruct the traffic accident, it needs to make
the traffic accident simulation. In order to determine the choice of simulation
parameters, it needs to extract the data when the traffic accident occurs. According
to the actual accident status and logic analysis, we should verify the correctness of
input parameters from the positive and negative aspects and get the simulation
result with the accident reconstructing software PC-CRASH. The final simulation
results should confirm the evidence of accident scene.
Keywords Traffic accident reconstruction Simulation PC-CRASH
Parameters
130.1 Introduction
Each traffic accident has its own unique. Variables of people, vehicles, road and
environment have forced that each accident reconstruction should also consider the
particularity on the basis of universality. The traffic accidents in this study are
those shelved for many years with controversy.
C. Wei (&) X. Gao
School of Automotive and Traffic Engineering, Jiangsu University, Zhenjiang, China
e-mail: [email protected]
M. Nie
Zhejiang Industry and Trade Professional College, Wenzhou, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1235
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_130,
Ó Springer-Verlag Berlin Heidelberg 2013
1236 C. Wei et al.
130.2 Overview of the Accident
On January 21, 2007, ten past eighteen, 1369 km ? 300 m of 202 line, snowy,
road had the ice and snow. The road was two-lane straight asphalt pavement, drove
in different directions. Mr. Wang drove Santana LX sedan (with a passenger
Mr. Zhang) from north to south. When driving to 1369.3 km in 202 lines, it had a
collision with Mr. Lv in agricultural tricycle (with a passenger) from south to
north. After the accident, the Santana sedan fell into the drain outside the slope,
while the agricultural tricycle stopped in the road. The accident has caused
damaged in varying degrees of Santana sedan and agricultural tricycle, and caused
Mr. Wang and the passenger Mr. Zhang died on the spot.
130.3 Problems to be Solved
The problems are the speed of Santana sedan and agricultural tricycle when the
accident occurred.
130.4 Simulation Process
130.4.1 Automobile Damage Condition
The central part in the right side of Santana sedan has severe hollow deformation,
the right side of the ceiling has inward deformation, and the whole car has bending
(See Fig. 130.1). The deformation zone is from the upper of right front wheel to
the upper of right rear wheel with a length of 1170 mm and a distance of
360–1930 mm (the latter is the height after the deformation, higher than the ori-
ginal height size) from the ground. Within the deformation zone, the depth is about
870 mm, and the deepest area is 2570 mm to the rear. Within the scope of
110–310 cm from the front to the bear with the distance of 15–158 cm from the
ground to the front and rear fender panels at the right side and the two doors, it has
the overall impacting hollow, the right side has severe quasi-impacted marks, and
the vehicle skin has left many scratches. Among them, the front corner of B
column has the similar hollow with the front wheel shape of agricultural tricycle.
For the Santana sedan, the shroud of the engine has deformation, the gearbox
handle has damaged; the combined lamp in the right front has broken off, the front
and rear windshield and right window has broken; there are many paints flakes off
in the front and rear doors and the front and rear fender panel at the right side.
130 Research on the Simulation Case of Traffic Accident 1237
Fig. 130.1 Santana sedan
Fig. 130.2 Agricultural
tricycle
Within the whole width of the vehicle, the front of agricultural tricycle has
inhomogeneous deformation, and the front right deformation is more severe than
the left; the down edge at the right front door and window has backward dislo-
cation of 70 mm, the down edge of right door has backward dislocation of
170 mm; the down edge of left front door and window glass has backward dis-
location of 15 mm; the whole front especially the front wheel hit the cover and has
the hollow; the left front combination has broken off, and the right front combi-
nation cover has broken off; the windshield and right window has broken off; The
two rearview has off (See Fig. 130.2); the right front corner of the head has hollow
and deformation with obvious fold deformation, impacting cracks and scratch
marks; The down edge of left corner has hollow with fold deformation and scratch
marks; The left and right paints have broken off; The front wheel has broken off
from the fork tray.
1238 C. Wei et al.
Santana sedan
Agricultural tricycle
Fig. 130.3 Pre-crash station and post-crash station
130.4.2 Simulation Analysis
The accident process mainly adopts the establishment of momentum conservation
model and functional conservation model, and then realizes with the computer
simulation. The simulation software is PC-CRASH, and the required technical
parameters should refer to the finished original data and actual use data. The crash
process is shown in Fig. 130.3.
The test has proved that after the collision of the traffic accident, the two rear
wheels of Santana sedan turn smoothly, while the rotation of two front drive
steering wheels has friction with the resistance of variable drive axle. The collision
focus area of Santana sedan firstly lies in the junction of B pillar at the right side
and right front door, and the car shows the arcuate shape after the deformation. In
this case, the front and rear wheels especially the right wheel has turned a steering
angle respectively to right and left.
The accident scene sketch and vehicle inspection can show that the front fork of
agricultural tricycle has steered to the tray and broken. The front end of two
longitudinal beams has collision with the wheel to break and touched the road
forming the friction defect. Even though the agricultural tricycle has no front
wheel brake in the structural design, in this case, it can be seen that the front wheel
braking is effective.
The accident scene sketches, survey notes and photos can show that after the
collision, the left rear wheel of agricultural tricycle has left the braking traces, and
the braking is effective. The vehicle inspection can show that whether the engine is
working, the two rear wheels of agricultural tricycle have brake. Besides, the left
and right rear wheel can be locked when the vacuum tank and piping are perfect,
and the braking force of the left and right wheels is balanced. Thus it suggests that
the braking of agricultural tricycle is normal and effective.
After the collision, the Santana sedan has across the road, curbs of 12 cm, green
belt and slope (slope length is 326 cm, and the slope is about 58 %), and felt into
the drain (Ladder, the width of bottom edge is 100 cm, the side length is 60 cm,
and the slope is 47 %).
130 Research on the Simulation Case of Traffic Accident 1239
Because the curb of 12 cm and the green belt can hinder the movement of
Santana sedan, it has set a low wall with the height of 12 cm representing the curb
and green belt in the simulation process.
The deformation parts, features and size of the two vehicles can show that the
front of right middle size of the vehicle was firstly contacted and had collision with
the front wheel of agricultural tricycle, and the angle of the two vehicle direction at
the collision is near 90°. The smaller deformation in the front of agricultural
tricycle can show that the main part in the collision is the front wheel, which
means that the collision has a larger rebound effect.
The instant speed and road condition before the collision can determine the
moving distance of the agricultural tricycle in the transverse direction after the
collision.
The relative position and the relative road direction status before the collision
can determine the moving direction after the collision. The moving distance is
mainly influenced by the collision speed of agricultural tricycle, the braking or not,
the bending status, road condition and the conditions outside the road of curb and
green belt.
Based on the above principle analysis, consulting the deformation size, feature
and accident scene sketches of Santana sedan and agricultural tricycle; input the
data of road traces, vehicle quality, vehicle technical parameters and road adhesion
coefficient into the software.
Gradually set the parameters of collision initial speed, initial direction angle
and initial position of Santana sedan and agricultural tricycle, simulate the
moving trajectory, status and result, and make repeated comparison with the cite
situation, especially the final stationary position, and we can get the speed at the
collision:
Vagricultural tricycle 43:4 km=h
VSantana 50:7 km=h
Because the Santana sedan has severe sideslip when close to the collision, the
longitudinal axis is inconsistent with the speed direction. Component velocity of
the head along the longitudinal axis is as follows:
0
VSantana 8:5 km=h
In addition, the total mass of the agricultural tricycle is heavier than that of
Santana sedan; the road vertical speed of the former is larger than the latter.
Besides, with the influence of curb and ditches, it seems that each wheel of the
Santana sedan has free rotation after the collision, the speed direction and angular
speed when the vehicle is close to stop has greatly influenced the stop position and
direction, and the agricultural tricycle speed with larger mass and adopted the
braking is relatively stable.
1240 C. Wei et al.
Because the accident scene has no road traces before the accident collision, it
can not infer the vehicle moving status and trajectory. The initial speed and
direction in the simulation analysis are to realize the relative status (collective
speed, direction and the contact position and direction), but it does not represent
the actual speed, for the driver may adopt the braking and steering operation from
the initial time to collision contact.
The tricycle braking traces are in coincident with the impacting point. It needs a
period of time for the braking from zero to maximum or from the beginning of
braking to the appearance of braking trace in the braking process. In other words,
the tricycle speed of the driver reaction, implemented operation and braking
response is larger than that in the collision. With the analysis, we can get that the
agricultural tricycle speed at 13 m (or 1 s) before the collision is about 45 km/h,
and that at 26 m (2.1 s) before the collision is about 46.5 km/h. Because the
Santana sedan has the sideslip status before the collision, it can not infer the speed
before the accident.
130.5 Conclusion
With the above analysis and computer simulated result, we can infer that the
Santana sedan speed was larger than 50.7 km/h at the collision, the component
speed along the head longitudinal axis was about 8.5 km/h; The agricultural tri-
cycle speed was about 43.4 km/h at the collision.
The agricultural tricycle speed at 13 m (or 1 s) before the collision was about
45 km/h, and that at 26 m (or 2.1 s) before the collision was 46.5 km/h.
Accident simulation should accurately extract the accident parameters. After
obtaining the simulation results, we should take use of the result to verify the
accident process to have a perfect combination of logic and evidence.
Acknowledgments My heartfelt gratitude goes to my doctoral tutor Xiang Gao in Jiangsu
University for the valuable instructions, and I feel grateful for the help of doctoral tutor Hong-guo
Xu in Jilin University.
Chapter 131
Regional Brand Development Model
Under the Perspective of System Theory
Lai-bin Wang
Abstract The regional brand development is to enable regional brands sustained,
orderly, stable and coordinated development under the constraints of the four
elements of the government, business, social and other intermediary organizations,
from this, the structural model of the regional brand development is formed. The
article argues that the regional brand building and development must proceed from
the following points: First, in the social subsystem, focus on the cultivation of
social cultural, change people’s traditional concept, nurture a kind of social culture
of brand share, and strengthen the basic research for technological innovation;
second, in the enterprise subsystem, improve product quality, good brand mar-
keting, management and technological innovation; third, government subsystem,
in terms of policy, capital, infrastructure, branding promotion, offer support;
fourth, in other intermediary organizations subsystem, give full play the coordi-
nating role of industry associations, human resources, financial resources, legal
advice, and other aspects are inseparable from intermediary organizations.
Keywords Development model Regional brand Resource system theory
131.1 Introduction
The regional brand concept was first proposed by Keller et al. (1998), the location
like the products or services as branding, brand name is usually the actual name of
this region. The brand makes people aware of the existence of the region, and
related associations. Keller et al. (1998) also believes that the region can be like as
the brand of product or service (Kevin 1999). Rosenfeld (2002) believes that the
L. Wang (&)
Department of Political and Management Science, Chizhou University,
Chizhou, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1241
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_131,
Ó Springer-Verlag Berlin Heidelberg 2013
1242 L. Wang
implementation of the regional brand strategy based on industry clusters in less
developed countries is a kind of way to enhance the competitiveness (Rosenfeld
2002; Simon 2007). In the development process of regional brands, Allen etc.
believe that traditional brand theory is applied to the regional context, two espe-
cially critical issues must be considered, namely the management of stakeholder
groups and government leaders to play the role (Simon 2007; Malcolm 2006;
Eraydn 2010).
Domestic studies of the regional brand theory date from 2002. With the vigorous
development of industrial clusters in China, in the industrial clusters of Zhejiang,
Jiangsu, Guangdong, Fujian and other brand-name products gathered growth and
brand regionalization phenomenon, theoretical studies of the regional brand begins
to be the part of the attention of scholars. Xia Zengyu explores the regional brand
building. After 2005, the regional brand research articles were published exten-
sively, and researches involving the content, nature and characteristics of the
regional brands, the formation mechanism and affecting factors (Sun 2009), regional
brand values and interests regional brand and corporate brand, the effect of the
regional brand, the regional brand equity and assessment (Yang 2005), regional
brand governance mechanisms, regional brand building in the role of government
(Cai 2008), dynamic evaluation and coordinated development of regional brands,
the regional independent brand cultivation mode, regional brand management and
other aspects of a more systematic study (Zeng 2000).
The current study mainly uses case studies, mostly stays in the qualitative
phase, lack of quantitative research tools and methods, and seriously impact on the
establishment of the theoretical framework of regional brands, therefore, on the
basis of qualitative research, to develop a regional brand quantitative measurement
tools which enhance the quantitative study of the regional brand, is an important
direction for future research in this field.
Regional brand studies are now beyond the narrow boundaries of the marketing
application, involving disciplines such as sociology, history, political which is
little attention in the past marketing and branding. So disciplines involved with
regional branding have extensive features and present the trend of multi-
disciplinary.
131.2 Connotation of the Regional Brand
Since the establishment of a socialist market economy, have emerged around a
number of products suitable for local production, and gradually formed a large-
scale industrial clusters and business clusters, and thus become regional brand
being characteristics of region. The regional brand is the ‘‘industrial products’’
which have considerable size and strong production capacity, higher market share
131 Regional Brand Development Model 1243
and influence within an administrative (geographic) area, is an area of industry
group to build a unified brand. In addition to general brand unique, easy to dis-
tinguish and irreplaceable features, it has non-exclusive unique features, as well as
non-competitive, external, regional, cluster and dispersed such.
Regional brand is the outward manifestation of the regional competitiveness,
the strength of the competitiveness of the regional brand can reflect the level of
regional economic development and government performance level. Therefore,
regional brand is not only the practical issues of corporate, is also a scholar of great
concern academic proposition.
It is generally believed that the regional brand is a product of regional economic
development, is the industry within the region of considerable size and strong
manufacturing capabilities, higher market share and influence of business and
business-owned brand goodwill sum. It consists of two elements: regional and a
good brand effect. In the backdrop of economic globalization, regional brand can
make people will a certain image and its association with this region, there is a link
up to drive regional economic development through regional brands to create and
disseminate. The regional competition has become an important form of compe-
tition in the market; the regional brand has become a contemporary economic
development and significant features. Regional brand is the inevitable outcome of
the regional economic development, reflecting the core competitiveness of the
regional economy, representing the subject and the image of a region.
In addition to general brand characteristics, regional brand has its own unique
characteristics: First is symbiotic. Regional brand is a public brand and resource
within a regional. Therefore, it has positive externalities; the utility has a good
symbiotic which can be shared by enterprises in the region.
Second is the persistence. Regional brand is the result of collaboration of many
enterprises in the region, the extraction of a number of brand essences, have a solid
foundation, and brand effect more extensive and lasting.
Third is regional. Regional brand has a certain degree of regional and rooted. It
deeply affected by local cultural, economic, political, regional factors and overall
performance.
131.3 Formation Process of the Regional Brand
Regional brands have distinct characteristics and formation processes. Regional
brand in China under the reform, opening up and the market economy environ-
ment, the formation process has a strong era, nationality. The formation process of
the regional brand is the main formation of the regional brand building elements,
including regional competitive industries to choose, the formation of industrial
clusters, regional brand creation and the formation of brand economy chain.
The choice of regional advantage industry is the basic link of the regional
brand, has a direct impact on the quantity and quality of the regional brand
building. In a way, it is the premise of the formation of industrial clusters. The
1244 L. Wang
formation of industrial clusters is the basis of regional brands. The continuous
development of the regional brand will inevitably bring about the brand economy
chain. At the same time, the creation of the regional brand permeates every aspect
of being.
131.4 System Theory and Regional Brand Development
131.4.1 System Theory
In this paper, the regional brand building still is a regional industrial clusters as the
foundation, regional industry cluster stakeholders will naturally become the main
body of the regional brand building, and these stakeholders is to constitute the
elements of the regional industrial clusters, exist the complex contact among them;
regional brand building is a strategic development to execution and then to the
outside world to exchange feedback rating system, regardless of the strategy
establishment, the follow-up implementation, the final regulatory assessment for
feedback to the strategic plan amendment is influence each other, so the regional
brand building based industry cluster is a complex system works.
131.4.2 Whole Course Regional Brand Development Model
(1) Input stage
In different brand-building phase, the investment of resources is also different.
In the start-up period, due to the extremely limited resources of the enterprise, the
lack of first-class industrial clusters, this requires the government to increase
investment, make the related policies, and other intermediary organizations in
technology, human resources, and financial resources need give strong support.
Resources focus on investment to a single market to achieve rapid breakthrough of
a brand, is a universal model for a regional brand to be established firmly in a
market (Fig. 131.1).
In the period of rapid development, with the region brands is becoming well-
known in the market, followed by rapid growth in the size of the industry cluster,
government support and policy makers should be biased in favor of SMEs, to
develop a reasonable market access system, and other intermediary organizations
should strengthen resource inputs.
At maturity, the regional brand has a strong competitive position in traditional
markets; status in the industry is about to changes from the pursuer to the tran-
scender and the leader. The input resources are relatively abundant, but because of
loss of the development of benchmarking, will have a more technical or market
131 Regional Brand Development Model 1245
Fig. 131.1 Whole course
regional brand development • Enterprise Resource
model • Government Resource
• Intermediary Organization
Input Industry guild, Advisory body
• Selforganization and
Helerorganization
Conversion • Coordination Strategy
Process • Coopetition Game
• Allwin
• Management Mechanism
Output
trial and error behavior. The right strategic choices of region brand development
will lead to greater efficiency; otherwise it will bring huge economic losses.
In the recession phase of the regional brands, in order to re-establish the brand
market image, you must change the development strategy, the liberal market
access system, excellent products and good social public relations which can make
regional brand reemerge to the former presence.
(2) Transformation process
The transformation process is the integration process of the resources, between
enterprises, between enterprises and governments, enterprises and other interme-
diary organizations by competing collaborative strategy, heter-organization and
self-organization, yield the greatest returns on investment resource.
Self-organization theory is called by a joint name including the dissipative
structure theory, collaborative theory, super-cycle theory, and chaos and fractal
theory. It is generally believed that the self-organizing mechanism is through the
various elements of the system of ‘‘competing-Synergy’’. Synergy can be divided
into positive and negative, positive synergy is Pareto improvement by repeated
game, resulting in ‘‘social promotion’’, the negative synergy produce ‘‘social
inserting’’. Visible, the results of competing is not really expected, in fact, a case in
point is that a large number of clusters decline and disappear in external com-
peting, which is manifestation of non-adaptability of the industry cluster and the
regional brand.
Similar regional brand should not be conflict and contradiction, but should be a
mutual learning, learn from each other, enhance each other the fraternal competing
synergies, therefore, must be designed new means of communication, change the
concept of the development of regional brands interested parties (enterprises,
1246 L. Wang
industry associations, government), and guide the development of similar regional
brand moving in the benign interaction competing relationship.
(3) Output stage
After reasonable resources transformation process, business, government and
other intermediary organizations achieve a win–win situation, regional brand of
scientific management system, to promote local economic and socio-economic
development, mainly reflected the knock-on effect in the regional brand.
131.5 Systematic Regional Brand Development Mode
Government, business, social and other intermediary organizations, regional brand
building as the core, constitutes an interrelated complex system. The core of the
regional brand development system is composed of four subsystems that is the
government, business, social and other intermediary organizations, the basic
structure shown in Fig. 131.2. Shown in Fig. 131.2, a regional brand development
system not only has the characteristics of the general system, the interaction
mechanism between the internal structure and systems, is much more complex
than the general system. The sustainable development of the system depends not
only on the coordinated development of the various subsystems, but also depends
on the degree of coordination between the various subsystems.
Fig. 131.2 Systematic
regional brand development
model Government
subsystem
Regional
Social enterprise
Brand subsystem
Subsystem development
intermediary
organizations
subsystem
131 Regional Brand Development Model 1247
131.5.1 Government Subsystem
The government plays an extremely important role in regional brand building and
development process. From the perspective of systems theory, government sub-
system, use of local advantages of resources, take a variety of marketing tools,
establish and promote regional brand, and in combination with the needs of the
regional brand development process, make scientific support policies.
First of all, the government must conduct regional image marketing. In addi-
tion, the government should rationalize the ideas and mechanism in macroscopic
level of the regional economic development, strengthen macro-guidance and
promotion, improve relevant mechanisms, and develop appropriate policies and
measures. The government need encourage brands to create famous brands,
implement brand strategy and corporate incentives, and make preferential policies
and incentives for brand-name enterprises to play a designer demonstration effect.
131.5.2 Enterprise Subsystem
Enterprises play a leading role in regional brand building, by nurturing large
enterprises in particular in a comparative advantage industry in the region, as the
core, and corresponding formation of a series of supporting vendors, large-scale
enterprise communities can be formed in the region, through the corporate com-
munity acts to promote the development of competitive industries and industrial
clusters, and form a certain reputation and influence of regional brands in the
market.
In the enterprise subsystem, the enterprises of different sizes create quality
products in the use of local comparative advantage resources, increase brand
awareness and reputation, and lay a solid foundation for the construction of the
regional brand. Then, this enterprise make use of modern marketing management
concepts, methods, strategies and means to improve their own brand influence at
the same time, strengthen the regional brand promotion. In addition, it is necessary
for enterprise to improve the technological innovation capability. If no R&D
innovation, always behind the others, the production would be not market
advantage. The enterprise can improve production efficiency, reduce operating
costs and improve management of resources, only through innovation of product
design and production technology, establishment and improvement of the tech-
nological innovation system, the active use of advanced technology to transform
traditional industries; only with high-tech, regional brand has a high brand value
and market competitiveness.
1248 L. Wang
131.5.3 Intermediary Organizations Subsystem
The development of regional brands requires a large number of professional
intermediaries in professional support services, such as policy coordination
withered, standard setting, trademark use, legal services, technical support, man-
agement consulting, disciplinary mechanism, as well as qualified guarantee (such
as loan credit guarantee), etc. Thus, industry associations, chambers of commerce
and other intermediary institutions should give full play to its functions.
Industry Association plays an important coordination and communication role
between enterprises. The Association is composed of voluntary corporate, non-
governmental organizations, has connection function in the construction of
regional economic between the enterprise and the market, business and govern-
ment. Association does depth analysis of the regional brand development, in
consultation with relevant government departments to help enterprises solve dif-
ficulties and problems of brand development process, through the development of
a regional brand development plans and unified applicable policies of regional
brand, the exchange of experience in regional brand building in the industry and
strive to create a higher value of regional brands.
131.5.4 Social Subsystem
Long-term development of the regional brand cannot be separated from local
economic development, support services, is also affected by the impact of local
social and cultural environment. In the beginning of the construction of regional
brand, creating a brand-sharing culture has an important role in the development of
the regional brand. It is necessary to establish and perfect a set of brand sharing
mechanism that is shared investment, shared interests, to coordinating operation,
the formation of clusters acting criteria, and accompanying cost estimates and
cost-sharing mechanism, with a strict barrier to entry, put an end to the subversive
opportunistic behavior to the regional brand image.
Social progress is also very important to the development of regional brands.
We should speed up reform of the scientific research system, and construction to
meet the requirements of the market economy, the investment system, scientific
research and development mechanisms and the growth of talent and incentives.
The same time increase the distribution system reform to meet the requirements of
the market economy; enhance technological innovation through technology shares
and other forms of power. To create a social environment of the growth of tech-
nology professionals, scientific and technological personnel to be able to get
enough market-oriented, institutionalized incentives for technological innovation
to develop and retain a sufficient number of qualified personnel.
131 Regional Brand Development Model 1249
131.6 Conclusion
Through the above measures, the contradiction in these subsystems can be
effectively avoided, and promote the coordinated development between the vari-
ous subsystems, finally achieve economic benefit, social benefit and environment
efficiency unification, so as to realize the sustainable development of regional
brand system.
Of course, the article on regional brand development structure model is only
limited to theoretical studies, large amounts of data investigation and analysis on
the basis of use of systems analysis tools such as system dynamics on empirical
research, which can be better to provide a reference for the development of
regional brands.
Acknowledgments I would like to thank Anhui Education department humanities and social
science fund (SK2012B338) for their support in this research.
References
Cai L (2008) The application of system dynamics in the research of sustainable development.
China Environmental Science Press, Beijing, pp 28–30 (in Chinese)
Eraydn HA (2010) Environmental governance for sustainable tourism development: collaborative
networks and organization building in the antalya tourism region. Tour Manag 31:113–124
Keller KL (1998) Strategic brand management. Prentice-Hall, Upper Saddle River
Kevin LK (1999) Effective long-run brand management: brand reinforcement and revitalization
strategies. Calif Manag Rev 41(3):102–124
Malcolm SA (2006) Bangkok: the seventh international conference on urban planning and
environment. In: Place branding, pp 11–14
Rosenfeld SA (2002) Just clusters: economic development strategies that reach more people and
places, regional technology strategies. North Carolina, Carrboro
Simon A (2007) Competitive identity: the new brand management formations, cities and regions.
Palgrave Macmillan, New York, pp 25–41
Sun L (2009) Progress analysis of foreign regional brand theory. Foreign economic and management,
pp 40–49 (in Chinese)
Yang G (2005) Targeting model of sustainable development in ecotourism. Hum Geogr 5:74–77
(in Chinese)
Zeng R (2000) System analysis of harmonization development among population, resource,
environment and economy. Syst Eng Theory Pract 20(12):1–6 (in Chinese)
Chapter 132
Research on Bayesian Estimation
of Time-Varying Delay
Meng Wang, Ying Liu and Ji-wang Zhang
Abstract Time delay estimation is one of key techniques to array signal pro-
cessing, and it has already had several mature algorithms. According to its dif-
ferent scenes, time delay estimation can be transferred to the estimation of
coefficients of adaptive filter, which is on the basis of parameter model of adaptive
filter. The simulations of Bayesian methods including Extended Kalman Filter,
Unscented Kalman Filter and Bootstrap Particle Filter show that under Gaussian
nonlinear system, EKF and UKF can estimate time-varying delay effectively.
Besides, algorithms of UKF perform better than that of EKF, which are only
subject to Gaussian system. In the nonlinear non-Gaussian system, BSPF is able to
estimate time delay exactly.
Keywords Time delay estimation Extended Kalman Filter Unscented Kalman
Filter Bootstrap Particle Filter
132.1 Introduction
Time delay, which is resulted by different transmission distance of signals, refers
to the time difference accepted by different homologous receivers. The earliest
method of time delay estimation is Generalized Cross Correlation (GCC) algo-
rithm put forward by Knapp and Carter in 1976 (Knapp 1976).
Traditional methods of time delay estimation like GCC algorithm and higher
order cumulant approach can effectively estimate fixed delay under certain cir-
cumstances (Xie et al. 2008). Nevertheless, Extended Kalman Filter (EKF) and
Unscented Kalman Filter (UKF) that adopts Unscented Transformation (UT) are
M. Wang (&) Y. Liu J. Zhang
Department of Electrical Information Engineering, University of Beijing Transportation,
Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1251
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_132,
Ó Springer-Verlag Berlin Heidelberg 2013
1252 M. Wang et al.
widely used in location and tracking in nonlinear dynamic system. EKF achieves
filtering by first-order linearization (Taylor series expansions), which inevitably
results in extra error and leads to divergence in strong nonlinear system (Crassidis
2005). UKF applies unscented transformation so as to transfer mean and covari-
ance nonlinearly and substitutes Jacobian matrix of EKF with simple mathematics
(Ma and Yang 2009).
UKF algorithm is of high precision, but it can only be used in the system that
noise obeys Gaussian distribution. As a sub-optimal estimation algorithm, particle
filter is commonly applied in nonlinear and non-Gaussian systems. The thesis
simulates EKF, UKF and particle filter and analyzes their performance according
to different scenes, producing relatively good estimation.
132.2 Signal Model of Time Delay Estimation
Assuming that s(t) represents signals from the same mobile transmitter, then at t,
the signals received by two independent base stations can be formulated as
follows:
(
r1 ðtÞ ¼ sðtÞ þ v1 ðtÞ
ð132:1Þ
r2 ðtÞ ¼ Asðt sðtÞÞ þ v2 ðtÞ
In order to facilitate analysis, the said formula is simplified. A is amplitude
ratio; Sðt sðtÞÞ stands for delayed signal; sðtÞ denotes time-varying delay; v1(t)
and v2(t) refer to noise interference of tow signals, which is assumed to be
independent Gaussian white noise.
By using parameter model of adaptive filter, the problem of time delay esti-
mation can be resolved by the procedure presented in Fig. 132.1 (Ching and Chan
1988). r1(k) and r2(k) are samples of r1(t) and r2(t). If sampling time T = Dd,
output of FIR filter is:
Fig. 132.1 Block diagram p
for adaptive estimation of two
time delays
z (k ) = ∑ ξ r (k − i)
i 1
i =− p
r1 (k ) FIR
+
e( k )
∑
-
r2 (k )
132 Research on Bayesian Estimation 1253
X
p
zðk Þ ¼ ni r1 ðk iÞ p!1 ð132:2Þ
i¼p
The minimum quadratic sum of error e(k) = z(k) - r2(k) can be achieved by
adjustment of n. If r1(t) and r2(t) are expressed as formula (132.1), then in
according with sampling theorem, expression is as follows:
sin pði sðkÞÞ
ni ¼ A sin cði sðkÞÞ ¼ A ð132:3Þ
pði sðkÞÞ
In practice, with tolerance for certain truncation error, error can be basically
ignored, as long as p is larger than maximum time delay sðkÞmax , for instance
p [ sðkÞmax þ5 , the error can be basically ignored. By this way, process of
estimation becomes less complicated without considering signals’ waveform.
sðkÞ is regarded as state variable. Providing that the transmitter moves at a
constant velocity in a straight line with noise (Gaussian white noise) disturbance,
the state equation and observation equation of the system are:
8
> sðkÞ ¼sðk 1Þ þ ðk 1Þ=100 þ wðk 1Þ
<
X
p
ð132:4Þ
>
: r 2 ð k Þ ¼ A sin cði sðkÞÞ r1 ðk iÞ þ mðkÞ
i¼p
wðk 1Þ and mðkÞ represent system noise and observation noise respectively.
r1 ðk iÞ is a number sequence of a known waveform. In such a case, signal model
of time delay estimation is completed.
132.3 Bayesian Filtering Techniques
132.3.1 Bayesian Estimation
Signal processing model of Bayesian filtering can be expressed by state equation
and observation equation:
(
xk ¼fk ðxk1 ; wk1 Þ
ð132:5Þ
yk ¼hk ðxk ; vk Þ
k denotes number of time series; fxk ; k 2 N g means state sequence;
fwk1 ; k 2 N g and fmk ; k 2 N g are noise series of independent identical distribu-
tion; fk ðÞ is state transfer function of the system; hk ðÞ is observation function of
the system. State function can be depicted by transfer probability matrix as
pðxk jxk1 Þ; k 0. Likewise, observation function can be depicted as pðyk jxk1 Þ;
1254 M. Wang et al.
k 0. Bayesian filtering is mainly applied to estimate xk on the basis of known
observation vector y1:k ¼ ðy1 ; y2 ; . . .; yk Þ and initial distribution pðx0 jy0 Þ ¼ pðx0 Þ.
The essence of Bayesian filtering is to use all the information known to con-
struct the posterior probability density of the system’s state variables, that is, to
predict prior probability density of the state by system models. And then use the
recent measurements to amend it, in this way, get the posterior probability density.
Aiming at the state space model, the best estimation of the state can be obtained
through confidence pðxk jy1:k Þ of different value in recursive calculation on xk by
measurement of data y1:k (Fu and Cui 2009).
The results of filtering pðxk jy1:k Þ can be attained by prediction and updating.
Assuming that probability distribution pðxk jy1:k1 Þ is obtained at time k 1, the
state transfer probability matrix pðxk jxk1 Þ will help to produce the prior proba-
bility distribution at k:
Z
pðxk jy1:k1 Þ ¼ pðxk jxk1 ; y1:k1 Þ pðxk1 jy1:k1 Þdxk1 ð132:6Þ
Assuming that the above formula in a system model which obeys a first-order
Markov random process, which satisfies the following equation:
pðxk jxk1 ; y1:k1 Þ ¼ pðxk jxk1 Þ ð132:7Þ
New observation data yk is available at k. Based on Bayesian principle, prior
probability distribution can be updated by force of measurement model pðyk jxk Þ so
as to reduce expected results of filtering:
pðyk jxk Þpðxk jy1:k1 Þ
pðxk jy1:k Þ ¼ ð132:8Þ
pðyk jy1:k1 Þ
and
pðyk jxk Þ ¼ pðyk jxk ; y1:k1 Þ ð132:9Þ
Z
pðyk jy1:k1 Þ ¼ pðyk jxk Þpðxk jy1:k1 Þdxk ð132:10Þ
Formulas (132.6) and (132.8) represent two basic steps of prediction and
updating, the recursive computation of which contributes to optimal Bayesian
estimation.
If noise wk1 and vk is zero-mean white Gaussian noise of independent dis-
tribution with known parameters, the state equation fk ðxk1 ; xk1 Þ is a given linear
equation of xk1 and wk1 , the observation equation hk ðxk ; mk Þ is a given linear
equation of xk and vk . The optimal solution pðxk jy1:k Þ can be achieved by Kalman
Filter, in the setting of consecutive xk (Kalman 1960).
In many cases, fk ðxk1 ; wk1 Þ and hk ðxk ; mk Þ are nonlinear, and noise wk and vk is
non-Gaussian, under which Kalman Filter won’t work well. Some extended
algorithms of Kalman filtering, like Extended Kalman Filtering and Unscented
132 Research on Bayesian Estimation 1255
Kalman Filter can be employed in nonlinear conditions. Furthermore, Algorithms
based on Bayesian theory like particle filter can be used in non-Gaussian noise
cases.
132.3.2 Introduction of EKF, UKF and BSPF
Kalman Filtering features minimum mean squared error under linear system
estimation. Through recursion and iteration, its updating can be completed by
calculation of estimated value and current input value, which is beneficial to real-
time processing. EKF is a classic algorithm used in nonlinear estimation. It adopts
linear transformation of Taylor expansion to approximate nonlinear models, and
combines Kalman Filtering to estimate. EKF algorithms are simple and less cal-
culated, but they can only work in the weak nonlinear Gaussian condition.
Actually, the approximation of probabilistic statistical characteristics of ran-
dom quantity by limited parameters is easier than that of arbitrary nonlinear
mapping function. Great importance has been attached to approximation of non-
linear distribution by sampling, solutions to nonlinear problems, like Unscented
Transformation (UT) (Kastella 2000; Gordon et al. 1993; Julier and Uhlmann
2004). UKF employs Kalman Filtering frame and uses UT to process mean and
covariance, instead of linearizing the nonlinear function. UKF doesn’t need der-
ivation of Jacobian matrix, without ignoring the higher order term, so its nonlinear
distribution statistics is of high precision. Though the calculation of UKF is as less
as that of EKF, its performance is better than that of EKF.
Particle filter achieves approximation of probability density function pðxk jyk Þ
by a pairs of random samples that are transmitted in state space, and has
sample mean instead of integral operation so as to produce state minimum
variance estimation. These samples are called particles. Importance density
function is one of the key techniques, which exerts a direct impact on the
effectiveness of the algorithm. Besides, the number of particles will become
increasingly less along with iteration, which is the phenomenon called ‘‘particle
shortage’’. Two effective solutions to particle shortage are selection of optimal
importance density function and adoption of resampling methods. From an
application perspective, most importance density functions adopt pðxk jxk1 Þ
which can easily achieved by sub-optimal algorithms. The resampling methods
are to increase the number of particles by the resample of particles and
probability density function denoted by corresponding weight. Common
resampling methods include random resampling, stratified resampling and
residual resampling,etc. BSPF is built on optimal importance density function
and resampling. Particle filter serves as the main filtering tool for the nonlinear
non-Gaussian system.
1256 M. Wang et al.
132.4 Three Algorithms of Time-Varying Delay
132.4.1 The Estimation Based on EKF
State equation and observation equation of time-varying estimation:
8
> sðkÞ ¼sðk 1Þ þ k 1=100 þ wðk 1Þ
<
X
p
ð132:11Þ
>
: 2r ð k Þ ¼h ð
k ks ; k Þ þ v ð k Þ ¼ A sin cði sðkÞÞr1 ðk iÞ þ vðkÞ
i¼p
wk ; vk are respectively stand for the statistics of system noise and observed
noise, in addition, Rwk ; Rvk are the covariance matrix.
The following are steps of the time-varying delay estimation.
(1) Initialization: assuming that the time-delay state value is equal to s0 , as k = 0,
and the initial variance is p0 .
(2) Time prediction:
(
^sðkjk 1Þ ¼^sðk 1Þ
ð132:12Þ
Pðkjk 1Þ ¼Pðk 1Þ þ Rw ðk 1Þ
(3) Updating Measurement:
8 T T 1
>
< K ðkÞ ¼Pðkjk 1Þ½H ðkÞ H ðkÞPðkjk 1Þ½H ðkÞ þRv ðkÞ
^sðkÞ ¼^sðkjk 1Þ þ KðkÞ½r2 ðkÞ hk ð^sðkjk 1Þ; kÞ ð132:13Þ
>
:
PðkÞ ¼½I K ðkÞH ðkÞPðkjk 1Þ
" #
ohk ðsðkÞ; kÞ o Xp
H ðk Þ ¼ ¼ sin c ð i sðkÞ Þr ð k i Þ
o^sðkÞ ^sðkÞ¼^sðkjk1Þ o^sðkÞ i¼p
1
^sðkÞ¼^sðkjk1Þ
After recursion operation, the estimated value of ^s1:k can be acquired.
132.4.2 The Time-Delay Estimations Based on UKF
The signal model follows the formula (132.4). N is the dimension of sk , N = 1.
The steps of UKF estimation include:
(1) Initialization: assuming that the time-delay state value is s0 , as k = 0, and the
initial variance is p0 .
(2) Calculation of the Sigma point set and the corresponding weight.
132 Research on Bayesian Estimation 1257
8
> ð0Þ
>
> vk1 ¼sk1 ; i ¼ 0
>
< ðiÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vk1 ¼sk1 þ ðN þ kÞPk1 ; i ¼ 1; . . .; N ð132:14Þ
>
> pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i
>
>
: vðk1
iÞ
¼sk1 ðN þ kÞPk1 ; i ¼ N þ 1; . . .; 2N
i
8 m
< x0 ¼k=ðN þ kÞ
>
xc0 ¼k ðN þ kÞ þ 1 þ a2 þ b ð132:15Þ
>
: m
xi ¼xci ¼ 1=½2ðN þ kÞ; i ¼ 1; . . .; 2N
In the above formulas, k ¼ a2 ðN þ kÞ N is a scalar; k is used to set distance
from the sigma point to the average point. a, which is usual a very small positive
number, was responsible for controlling the high-order information errors in
nonlinear transformation. b is applied to inputting the prior information of random
variables. If the state variable is of single variable, the parameters will be first set
to a ¼ 1; b ¼ 0; k ¼ 2 and xi represent those corresponding weights,
P2N
i¼0 xi ¼ 1:
(3) Time estimation:
Substituting the sample points into the state equation to figure out the
predicted point set:
ðiÞ ðiÞ
vkjk1 ¼ vk1jk1 þ ðk 1Þ=100 ð132:16Þ
` Calculating the mean value and variance based on the predicted point set.
X
2N
ðiÞ
^skjk1 ¼ xm
i vkjk1 ð132:17Þ
i¼0
X
2N h ih iT
Pkjk1 ¼ xci vi;kjk1 ^skjk1 vi;kjk1 ^skjk1 ð132:18Þ
i¼0
(4) Updating the measurements:
Calculating the measured point set in the basis of nonlinear mapping:
X
p
wkjk1 ¼ h vkjk1 ¼ A sin c i ^
vkjk1 r1 ðk iÞ ð132:19Þ
i¼p
` Calculating mean value, variance and covariance of the measured points
set:
X
2N
ðiÞ
^rkjk1 ¼ xm
i wkjk1 ð132:20Þ
i¼1
1258 M. Wang et al.
X
2N h i h iT
Prr
kjk1 ¼ xci wi;kjk1 ^rkjk1 wi;kjk1 ^rkjk1 ð132:21Þ
i¼0
X
2N h i h iT
Psr
kjk1 ¼ xci vi;kjk1 ^skjk1 vi;kjk1 ^rkjk1 ð132:22Þ
i¼0
´ Updating the state and the variance value, calculating the gains of filtering:
1
Kk ¼ Psr rr
kjk1 Pkjk1 ð132:23Þ
^sik ¼ ^skjk1 þ Kk rk ^rkjk1 ð132:24Þ
T
Pk ¼ Pkjk1 Kk Prr
kjk1 ðKk Þ ð132:25Þ
(5) Applying the recursive operation to calculate the value of ^s1:k .
132.4.3 The Estimation Based on BSPF
(1) Initialization: set k = 0 and sample.
1
si0 pðs0 Þ; wi0 ¼ ; i ¼ 1; 2; . . .; N: ð132:26Þ
N
(2) Importance sampling,
sik p sk jsik1 ; i ¼ 1:; 2; . . .; N ð132:27Þ
(3) Calculating the weight and then normalizing:
wik ¼ wik1 p rk jsik ð132:28Þ
wik
ik ¼
w ð132:29Þ
P
N
wik
i¼1
(4) Calculating posterior probability:
X
N
pðsk jr1:k Þ ¼ ik d sk sik
w ð132:30Þ
i¼1
132 Research on Bayesian Estimation 1259
(5) Estimating the time-delay value at the moment of k:
Z X
N
^sk ¼ E½sk ¼ sk pðsk jr1:k Þdsk ¼ ik sik :
w ð132:31Þ
i¼1
N
(6) Resampling to get a new set of points si
k i¼1
and then conducting stratified
sampling.
132.5 Simulation Results
To compare the performance of these three methods—EKF, UKF, BSPF, we
applied these three filtering methods in one experiment. The system noise sepa-
rately obeyed normal distribution (N ð0; 1Þ) and uniform distribution. The mea-
sured noise obeyed normal distribution. The initial state value of simulation was
set to 0.1, and the variance was made equal to 1, the number of sample points for
simulation was 500, moreover we had taken 1000 particles in BSPF. As to UKF,
the parameters were a, b and i which were respectively equal to 1, 0 and 2.
Simulation 1: the simulation of the system noise conformed to Gauss distri-
bution (N ð0; 1Þ) (Figs. 132.2, 132.3).
EKF filtering result
40 Real signal
20 EKF filtered estimation
0 10 20 30 40 50 60 70 80 90 100
UKF filtering result
40 Real signal
20 UKF filtered estimation
0 10 20 30 40 50 60 70 80 90 100
Bootstrap Particle filtering result
40
Real signal
20 BSPF filtered estimation
0 10 20 30 40 50 60 70 80 90 100
Fig. 132.2 The filtering results under the Gauss noise
1260 M. Wang et al.
EKF error
10
Estimation error of EKF
0 3σ interval
-10
0 10 20 30 40 50 60 70 80 90 100
UKF error
5
Estimation error of UKF
0 3σ interval
-5
0 10 20 30 40 50 60 70 80 90 100
Bootstrap Particle filtering error
5
Estimation error with BSPF
0 3σ interval
-5
0 10 20 30 40 50 60 70 80 90 100
Fig. 132.3 The absolute filtering error under the Gauss noise
EKF filtering result
40
20 Real signal
0 EKF filtered estimation
0 10 20 30 40 50 60 70 80 90 100
UKF filtering result
40
20 Real signal
0 UKF filtered estimation
0 10 20 30 40 50 60 70 80 90 100
Bootstrap Particle filtering result
40
20 Real signal
0 BSPF filtered estimation
0 10 20 30 40 50 60 70 80 90 100
Fig. 132.4 The filtering results under uniform distribution
132 Research on Bayesian Estimation 1261
EKF error
50
0
Estimation error of EKF
3σ interval
-50
0 10 20 30 40 50 60 70 80 90 100
UKF error
50
0
Estimation error of UKF
3σ interval
-50
0 10 20 30 40 50 60 70 80 90 100
Bootstrap Particle filtering error
5
Estimation error with BSPF
3σ interval
0
-5
0 10 20 30 40 50 60 70 80 90 100
Fig. 132.5 The absolute filtering error under the uniform distribution
In order to compare the estimating performance of time-varying delay in EKF,
UKF and BSPF, we defined the average variance of one independent experiment:
!1=2
1X T
MSE ¼ ð^xk xk Þ2
T k¼1
The symbol T represented one time step length, MSE of the algorithms in the
figure above respectively equal to 6.9328 (EKF), 6.6033 (UKF) and
6.4577(BSPF). Therefore, according to filtering results and squared error, we could
find that the three filtering methods can all get comparatively good results. The
reason went like this: firstly, hypothetic model was the Gaussian; secondly, the
nonlinearity of model was weak.
Simulation 2: the simulation of system noise conformed to uniform distribution
(U [0, 1]) (Fig. 132.4).
Figure 132.5 showed that the MSE were respectively 125.5421 of EKF,
127.0202 of UKF, and 8.5351 of BSPF. From these data, it was found that EKF
and UKF could not estimate the true values of time-varying delay accurately. The
MSE of EKF and UKF were two orders higher than that of BSPF, which could
deduce a conclusion that in both EKF and UKF, the system must be confined to the
Gauss model for a good performance; in the meanwhile, the BSPF has advantages
in estimation of nonlinear and non-Gaussian systems.
1262 M. Wang et al.
132.6 Conclusion
This paper had introduced Bayesian filtering theory and the algorithm steps of
EKF, UKF and BSPF. Then we had separately simulated in a Gauss nonlinear
system and a non-Gauss nonlinear, which was in order to compare the results of
EKF, UKF and BSPF. It was found that in Gauss nonlinear system, EKF, UKF and
BSPF could all perform well because of the weak nonlinearity. But in non-Gauss
nonlinear system, both EKF and UKF could no longer conduct estimation with
accuracy. They would cause much higher MSE than BSPF which proved to be
more suitable for estimating a non-Gauss nonlinear system.
Acknowledgments This paper is supported by National Natural Science Foundation of China
(Grant No. 61172130).
References
Ching PC, Chan YT (1988) Adaptive time delay with constraint. IEEE Trans Acoust Speech Sig
Process 36(4):599–602
Crassidis JL (2005) Kalman filtering for integrated GPS and inertial navigation. In: AIAA
guidance, and control conference and exhibit, San Francisco: AIAA, 2005–6052
Fu W, Cui Z (2009) Based on improved extended kalman filter static target tracking.
Optoelectronics 36(7):24–27
Gordon N, Salmond DJ, Smith AFM (1993) Novel approach to nonlinear and non-Gaussian
Bayesian state estimation. IEEE Proceedings-F (S0-7803-2914-7) 140(2):107–113
Julier SJ, Uhlmann JK (2004) Unscented filtering and nonlinear estimation. Proc IEEE
92(3):401–422
Kalman RE (1960) A new approach to linear filtering and prediction problems. Trans ASME-J
Basic Eng 35–45
Kastella K (2000) Finite difference methods for nonlinear filtering and automatic target
recognition[J]. Multi-target multi-sensor tracking Appl advan 3:233–258
Knapp CH (1976) The generalized cross correlation method for estimation of time delay. IEEE
Trans Acoust Speech Sig Process 24(8):320–327
Ma Y, Yang S (2009) Based on the combination of UKF navigation error state estimation.
J Huazhong Univ Sci Technol 37:219–222
Xie J, Wu C, Fu S (2008) Study on passive location and time delay estimation method. Ships
Electron 31(6):26–29
Chapter 133
Research on Design and Analysis Platform
for Modular Spacecraft
Xiao-wen Zeng, Zhao-xia He and Hao Luo
Abstract In spacecraft structure design process, modular technology is very
important for improving efficiency and reducing cost. Based on the modular design
method, a deployable on-orbit spacecraft modular configuration was established.
The influence of simultaneously deployment solar panel and modular structures
under the situation of weightlessness on the spacecraft’s gesture was simulated and
analyzed with the assistance of virtual prototype technology. The simulated result
shows that the deployment of module structures is vital to the angle of spacecraft’s
gesture and its velocity. Within the Eclipse development environment, an inte-
grated platform which includes design and analysis for spacecraft was developed
with J2EE. In order to acquire rapid response to the whole process, which includes
structural design, digital prototype assembly and dynamics analysis, the structure
design, assembly and dynamics analysis of spacecraft were well integrated.
Keywords Integrated platform Modular design Spacecraft Virtual prototype
technology
133.1 Introduction
With the development of space technology, larger and more complex structure of
the spacecraft system is needed in space exploration. Therefore, when modern and
advanced spacecraft is designed, some factors must be considered, such as the size
X. Zeng (&)
School of Computer, Hunan Institute of Science and Technology, Yueyang, China
e-mail: [email protected]
Z. He
School of Construction Machinery, Chang’an University, Xi’an, China
e-mail: [email protected]
H. Luo
School of Mechanical Engineering, Northwestern Polytechnical
University, Xi’an, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1263
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_133,
Ó Springer-Verlag Berlin Heidelberg 2013
1264 X. Zeng et al.
and complexity. Since modular spacecraft structure design concept was put for-
ward in NASA’s Goddard Space Flight Center (Bartlett 1978) in the seventies of
the twentieth century, Modular, Adaptive, Reconfigurable System (MARS) (Jaime
2005), modular spacecraft design concepts for on-orbit deployment based on
MARS (Sugawaraa 2008), concepts and technology for on-orbit servicing (Rod-
gers and Miller 2005) in modular spacecraft design have been experienced. It can
be seen that more and more attention was paid to modular spacecraft design
concept. Recently, there are mainly two kinds of modular spacecrafts for on-orbit
deployment-Hexpak (Hicks et al. 2005, 2006) and Panel Extension Satellite
(PETSAT) (Higashi et al. 2006). References (Hicks et al. 2005; Hicks et al. 2006;
Higashi et al. 2006; Larry and Rolland 2013 ; Edward 2013; Deborah and Grau
2013; Jon et al. 2002; Murata et al. 2002) elaborate a lot of research on spacecraft
structures for on-orbit deployment and many kinds of mechanical interfaces.
However, related dynamics analysis on such mechanism configuration is rarely
reported.
In this paper, based on the idea of modular design, a spacecraft module con-
figuration for on-orbit deployment assembled with unified hinge mechanism is
designed, which could deploy different modules according to task demands. With
the assistance of virtual prototype technology, spacecraft attitudes influenced by
the deployment sequence of solar panels and modular spacecraft mechanism are
simulated and studied in the state of weightlessness in space. On this basis, based
on J2EE Technology, a design and analysis platform is developed for modular
spacecraft design in Eclipse environment. The configuration design process, model
and data association demand in assembly and simulation and a variety of design
and simulation software integration requirements are considered in this platform.
133.2 Design of Modular Structure and Dynamics
Simulation of Deployment
133.2.1 Rapid Design and Assembly of Spacecraft Structure
Driven by Parameters
Modular design is usually directed at the same series of spacecrafts and each
module’s geometry dimensions are relatively invariant. In order to increase the
flexibility of structure design, each module is designed driven by parameters.
Therefore the design and assembly speed is improved to realize rapid response.
The designed spacecraft consists of two parts-spacecraft body and solar panels,
as shown in Fig. 133.1. The spacecraft body is made up of five similar modules
which can be lay flexibly according to the functional requirements. Each module
could be installed with different equipment and instruments according to the task
requires, as shown in Fig. 133.2. There are positioning pins in the modules which
are connected with hinges and driven by motors. The positioning pins are used to
133 Research on Design and Analysis Platform for Modular Spacecraft 1265
Fig. 133.1 Spacecraft configuration and assembly form
Fig. 133.2 Equipment and
instruments installed in the
module
locate each module in the process of deployment. The solar panels are also con-
nected with hinges and driven by torsion springs. Each solar panel is made up of 4
pieces of rigid boards of which the geometric dimensions are 1750, 1500 and
30 mm.
The solar panels are folded before released, parallel to each other and fixed on
the spacecraft body. When released on orbit, the solar panels mounted on both
sides deploy at the same time driven by the torsion springs. Table 133.1 shows
mass properties of spacecraft modules and solar panels.
In the process of analysis, the spacecraft body’s coordination is fixed to module 1.
The X direction is parallel to each module when deploys, the Y direction is parallel
1266 X. Zeng et al.
Table 133.1 Mass properties of spacecraft modules and solar panels
Mass properties Modules names
Module 1 Module 2 Module 3 Module 4/5 Solar array 1/2/3/4
Mass (kg) 510.73 413.15 378.91 223.90 183.47
Jx (kg m2) 87.03 74.44 67.87 57.01 53.44
Jy (kg m2) 97.89 86.54 81.71 60.09 34.43
Jz (kg m2) 170.53 148.85 138.28 110.27 19.07
to solar panels and the Z direction is perpendicular to the spacecraft body and solar
panels.
133.2.2 On-Orbit Deployment Dynamics Simulation Based
on Virtual Prototype Technology
In the simulation, the power output of the motors could be applied by two methods
which are torque and rotational speed. In this paper, a constant speed is applied to
simulate the driving force of the motors. The solar panels are connected with
hinges and driven by torsion springs. Torque calculation formula of the torsion
spring is as follow (Bai et al. 2009):
T ¼ T 0 Kh ð133:1Þ
where T0 is Initial torque of the torsion spring, K is stiffness of the torsion spring
and h is the deployment angle of the solar panels.
Closed Cable Loops (CCL) is currently the most common synchronous
deployment control mechanism (Tianshu et al. 2000; Yuan et al. 2009), which is
made up of the grommets fixed to hinges, grommet guides and soft cable. It is a
synchronous transmission device to realize the inside and outside solar panels
deploy at the same time. The basic principle is shown in Fig. 133.3 where L is the
distance of two grommets and r is the radius of the grommet.
When the unfolding angles of the adjacent solar panels are the same, the
mechanism doesn’t work. But when the solar panels don’t move synchronously,
the angles of the two grommets don’t equal that would causes the upside edge tight
and the downside edge loose. The tight edge would be stretched and the force F
Fig. 133.3 Principle of CCL
133 Research on Design and Analysis Platform for Modular Spacecraft 1267
will be applied on the grommet. So there will be a torque caused by F in the
grommet which makes the solar panels tend to synchronize. The principle for-
mulas are as follows:
T 0 ¼ K 0 hi hj ð133:2Þ
F ¼ K 0 r Dh ð133:3Þ
where T 0 is the controlling torque provided by CCL, K 0 is the equivalent torsional
stiffness of CCL, hi and hj are the two adjacent solar panels’ unfolding angles
correspondingly.
The process of deployment is simulated by using the analysis software Adams.
The spacecraft body and solar panels are regard as rigid bodies. The solar panels
are connected to spacecraft body by hinges and so do as the solar panels. In the
dynamics simulation, the motors drive all the modules to deploy of which speeds
are set to 30 red/s. For all the solar panels are driven by torsion springs, when the
inside solar panel deployed to the angle 90°, the deployment angle of the outside
solar panel will be 180°. Therefore, it could be concluded from formula (133.1)
that the torque of the outside solar panel is 1 time larger than that of the inside one.
The stiffness of the torsion spring is set to 0.1 N m/(°) and damping influence is
ignored. So the pre-tightening torque of the inside and outside solar panel is 9 and
18 N m correspondingly. The solar panels’ synchronous deployment is controlled
by the connection joint in Adams. The transmission ratio of the connection joint is
set as 1:2 and that of the adjacent rotation joint is set as 1:1.
The virtual prototype of the designed spacecraft is shown as Figs. 133.1 (the
initial state) and 133.4 (the final state of the deployment). There are 3 deployment
sequences for the spacecraft: spacecraft body and solar panels deploy at the same
time (Named as sequence 1); solar panels deploy before spacecraft body
(Sequence 2); spacecraft body deploys before solar panels (Sequence 3). All of the
3 deployment sequences are simulated. The attitude angles affected by the
deployment sequences are shown as Figs. 133.5 and 133.6.
As shown in Fig. 133.5, the attitude angles of the X direction are very small.
The value of sequence 3 is the largest and it is 2.15°. Sequence1 costs 9.8544s,
which is less than the others. The attitude angles of the Z direction are very similar.
Fig. 133.4 The spacecraft’s
final state of the deployment
1268 X. Zeng et al.
Fig. 133.5 The attitude angles of X direction
Form Fig. 133.6, it could be seen that the attitude angles of the Y direction are
very large, no matter what the sequence is. The deployment of the solar panels has
almost no influence on the attitude angle change.
Fig. 133.6 The attitude angles of Y direction
133 Research on Design and Analysis Platform for Modular Spacecraft 1269
Fig. 133.7 Flexible and multi-level design and analysis flow of modular spacecraft
133.3 Design and Analysis Platform for Modular
Spacecraft
133.3.1 Flexible and Multi-level Design and Analysis Flow
of Modular Structure
Multi-level design and analysis flow is made up of systematic flow of spacecraft
and modular part analysis flow, including various types of structure design and
analysis, such as configuration design, system assembly, static and dynamic
analysis etc. Figure 133.7 is a sketch map to explain the flexible and multi-level
design and analysis flow.
The main flow controls the execution sequence of sub flows and manages the
association of analysis data. But it doesn’t participate in analysis process.
At the aspect of platform realization, by using WFCP-net’s function of multiple
instance management, multiple flows could be managed to realize multiple
working conditions and flexible flow regression function.
The main flow’s management engine operates as a server on the serve computer
that could make all the parallel sun flow distribute on different client computers.
This distributed flow can obviously improve execution efficiency.
The sub flows are responsible for the scheduling and execution of analysis
tasks. By associating main and sub flows, the states of sub flows could be con-
trolled. Client programs control sub flows’ operation and analysis tasks are
1270 X. Zeng et al.
Fig. 133.8 The sketch map of simulation data of spacecraft deployment on-orbit
completed by specific analysis software. At the same time, client programs can
also control the states of analysis tasks.
The design and analysis flow is very flexible in which flows’ operation process
could be changed according to researchers’ judgments. The analysis flow could be
retreated to specified analysis step. The existing designed models and analyzed
results could be reused.
133.3.2 Automatic Association of the Design and Analysis
Data Which is Reusable and Centralized
The designed model and analysis data includes the initial CAD models, experi-
ment data, simulation models and simulation results etc. All of the models and
result data are associated in the platform. So the operators could view and obtain
these data quickly.
(1) Automatic association of flow tasks’ models and data.
133 Research on Design and Analysis Platform for Modular Spacecraft 1271
Design and analysis data management is to manage uniformly by concentrating
design and analysis data distributed in the clients to data server, which associates
and transfers models and data of different flows automatically.
According to different task demands of spacecraft systems, designed models
and analysis data could be obtained and reused quickly by controlling the work
condition of analysis flow and the data version. In this way, the design and analysis
knowledge could be reused.
(2) Automatic obtainment of designed models and analysis data.
The platform could be visited by Web browser which is clear and intuitive. The
researchers can view and download any analysis data that includes various initial
data, intermediate process data and analysis result data. The analysis data could be
shown in many forms, such as thumbnails, pictures and visual animations. Fig-
ure 133.8 is the sketch map of simulation data of spacecraft deployment on-orbit.
133.4 Conclusion
Based on the concept of modular spacecraft design, a modular spacecraft con-
figuration is designed for the deployment on-orbit. By the way of virtual proto-
typing technology, the designed spacecraft is simulated in the analysis software
Adams. Considering the space’s weightlessness environment, the effects of the
spacecraft’s attitudes caused by the different deployment sequences of the modules
are studied. It is concluded that the deployment of the spacecraft modules affects
the attitude angle more than solar panels.
Based on Multi-level design and analysis (D&A) flow management, D&A data
management and reused technology of D&A knowledge facing to tasks, a D&A
platform is established for the modular spacecraft. In this platform, modeling,
assembly and a variety of mechanical analysis are managed by being concentrated
together to realize the fast response of modeling and analysis.
Acknowledgments This work is supported by the Scientific Research Fund of Hunan Provincial
Education Department under the grant No. 10C0756.
References
Bai Z, Tian H, Zhao Y (2009) Dynamics simulation of deployment of solar panels in different
layouts based on ADAMS. J Syst Simul 21(13):3976–3977
Bartlett RO (1978) NASA standard multi-mission modular spacecraft for future space
exploration. In: American astronautical society and deutsche gesellschaft fuer luft-and
raumfahrt, goddard memorial symposium 16th, Washington, DC, AAS, pp 8–10
Deborah MW, Grau J (2013) Modular spacecraft standards: supporting low-cost, responsive
space. In: AIAA, 2004–6098
Edward F Jr (2013) Multi-mission modular spacecraft (MMS). AIAA-88-3513
1272 X. Zeng et al.
Hicks M, Enoch M, Capots L (2005) In: AIAA 3rd responsive space conference
Hicks M, Enoch M, Capots L (2006) In: AIAA 4th responsive space conference, Los Anglees.
Paper No: RS4-2006–3006
Higashi K, Nakasuka S, Sugawara Y (2006) In: 25th international symposium on space
technology and science. Paper No: 2006-j-02
Jaime E (2005) AIP Conf Proc 746:1033–1043
Jon M, Jim G, David G (2002) Space frame: modular spacecraft building blocks for plug and play
spacecraft. In: 16th Annual/USU conference on small satellites
Larry M, Rolland S (2013) Options for enhanced multi-mission modular spacecraft (MMS)
maneuver capability. AIAA-80-1292
Rodgers L, Miller D (2005) Synchronized Position Hold, Engage, Reorient, Experimental
Satellites (SPHERES), working paper. Massachusetts Institute of Technology, Cambridge
Satoshi M, Eiichi Y (2002) M-Tran: self-reconfigurable modular robotic system. IEEE/ASME
Trans Mechatron
Sugawaraa Y, Saharab H, Nakasukab S (2008) Acta Astronaut 63:228–237
Wang T, Kong X, Wang B (2000) The research on principle and function of closed loop
configuration of solar arrays. J Astronaut 21(3):29–38
Yuan AN, Song GU, Guang JIN (2009) Analysis and simulation of deployment motion of
satellite solar array. Chinese J Opt Appl Opt 1(2):29–35
Chapter 134
Research on H-Point of Driver Based
on Human Physical Dimensions
of Chinese People
Lin-lin Sun, Fan-sen Kong, Duo-nian Yu, Bao Lan
and Xiang-bei Wang
Abstract The H-point is an important reference point in the general arrangement of
the design of car body, which determines the convenience and comfortableness of the
driver. The referred standard in the arrangement of the H-point is the agreeable
position curve in SAE J1517. The curve, however, does not apply to Chinese, so the
Chinese manikin was used. In order to guarantee the comfortableness and vision,
H-point position of the 5th percentile woman, 10th, 50th, 90th and 95th percentile
man sitting models were adjusted, and the coordinates of points were recorded.
MATLAB software was applied to process these data, so that the corresponding
H-point trajectory equations were fit, and curves of the equations were plotted in the
same coordinate system. The suitable H-point position regions are areas enclosed by
these curves. Using CATIA, American H-point regions were calculated in the same
way. Finally, the obtained two areas were compared and analyzed.
Keywords Comfortableness H-point H-point trajectory curve equation
Manikin
134.1 Introduction
In order to reduce the degree of fatigue of driving and riding when arranging the
interior, the requirements for human comfort posture must be met in the design,
which is the basis of the layout of the human body and seat design. The
L. Sun (&) F. Kong X. Wang
Department of Industrial Engineering, Jilin University, Changchun, China
e-mail: [email protected]
F. Kong
e-mail: [email protected]
D. Yu B. Lan
College of Automotive Engineering, Jilin University, Changchun, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1273
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_134,
Ó Springer-Verlag Berlin Heidelberg 2013
1274 L. Sun et al.
comfortableness of driver or occupant is a very important performance indicator in
the body layout. The comfortableness covers widely, which includes the sitting,
the dynamic, the thermal and the contact etc., and have comprehensive, subjective,
individual characteristics. The Comfort is multidimensional, and is dependent on a
variety of factors such as direction, intensity, frequency, time, individuals. The
concept of comfort defined in the ergonomics is that there being no uncomfortable
state (Hockenberry 1979). From a physiological point of view, the comfort should
be a minimization of energy engaged in physical activity under certain constraints
(SAE 1988). Thus, the minimum uncomfortable sitting posture is just the sitting
comfort.
Actual automotive point H has a great significance in the auto body layout
design: After the driver or occupant takes a seat with a normal driving position or
the posture seat, most of the weight will be borne through the buttocks by a
cushion, the other part be supported by the backrest via the human back, while
only a small part will be applied to the floor through the left foot heel. Under this
posture, the pilot body when operating often moves through the axis of the hori-
zontal level around the H point. Therefore, the actual H-point position in the auto
body determines the convenience and comfortableness of the pilot operation,
which is why it has become a reference point in the body internal dimension.
In the 1980s, Nancy LPhilippart etc. from General Motors Corporation in the
United States proposed a H-point position curve model suitable for the driver,
which were used to predict the driver’s H-point position. In 1985, the model was
included in SAE J1517 by the American Society of Automotive Engineers. It is a
set of H-point position curves when a different percentile driver is driving suitably,
in which each one is characterized by horizontal and vertical position relationship
between the H-point position and orientation reference point. Congenial H-point
position curves in the SAE J1517 were generated statistically based on the driver’s
body and test data in the United States in the 1960s, in which body size, layout
constraints and posture and other factors implied. These factors are used to
determine the H-point adjustment range and to meet the required fitness theoret-
ically. However due to be constrained by the time and region, this method has
some shortcomings, mainly in the following aspects:
(1) Congenial H points on a curve or a straight line, only corresponding to an
H-point in a H30 value.
(2) Recommended H point position curves given based on the human data
published by the SAE does not apply to Chinese human body.
(3) Recommended H-point position curves with the ankle fixed at 87 degrees a
as one of prerequisites, thus increasing a redundancy for the layout of the pedal,
but also limiting the layout space of other elements.
(4) Recommended H-point position curves with the comfortable body posture as a
consideration, obviously factors to be considered too small (Huang et al. 2000).
The shortcomings mentioned above will be considered significantly in the study
of the H-point layout method. The Chinese manikin will be used of for layout
design. For the ankle constrained angle range is 78°–105°, and in addition to the
driver’s comfort considered, to, requirements for the vision of the dashboard and
134 Research on H-Point of Driver 1275
direct forward visibility and traffic light vision also were taken into account in the
design process, so that the H-point design is not only meet the comfort but also
ensuring a good vision.
134.2 Methodology
134.2.1 Comfortableness
Studies have shown that comfort or not are affected essentially by the subjective
feeling when maintaining a specific posture, in which the joint angle has a very
important influence on the subjective feelings of the comfort. Of course, there are
also a number of other factors that are likely to affect the feeling of comfort, such
as the contact pressure distribution between the person and the seat (Bubb and
Estermann 2000). The comfort and fatigue degree during the body driving and the
ride are related to the posture determined by the joint angles in the design.
Therefore, various manikin can be positioned according to a comfortable joint
angle so that its vision, encompasses, comfortableness etc. be able to evaluated
(Ren et al. 2006). The major joint angle adjustment range of drivers’ in a com-
fortable posture has been described as shown in Fig. 134.1 and Table 134.1
Each joint angle range of activities can be divided by the preferred angle editor
in CATIA manikin posture analysis module, and then the current position will be
able to be assessed globally and locally by the system. Edit Preferred angle
‘‘button’’ clicked after entering the module, the corresponding angle to a2–a7 in
Table 134.1 were found in the manikin. The scope of activities of each angle is
divided into two areas; one is into the area within the comfort range marked with
Fig. 134.1 Comfortable posture
1276 L. Sun et al.
Table 134.1 Human joint angle range under a comfortable sitting
a1 a2 a3 a4 a5 a6 a7
20°–30° 95°–120° 95°–135° 78°–105° 0°–50° 80°–170° 170°–190°
green, another falling into the region outside the comfort indicated by orange. For
the comfortable driver sitting angle, it focuses mainly on the plane of the side
view, and the comfortable angle of the other views given little studies. So in order
to prevent the angle change in other views caused by the manikin posture changes,
the freedom of each joint in the other directions was locked using the Lock
function. After locked, the parts cannot perform any operation in the locked
freedom direction, the angle value does not change with other joints’, thus
ensuring only the angle in Table 134.1 impacted in the adjustment process.
Because a1 is related with the seat’s backrest angle but not with a joint in the
body, we can not constrain it by using the above methods, but at the hinge pairs in
this section, and the rotation angle range of the kinematic pair is defined as 20°–
30°, so that the backrest angle can only be adjusted within the range of comfort
without out of range.
After defining the comfort range of different angles, open the dialog box of the
manikin posture assessment and analysis in a manikin posture analysis module, the
system provides two kinds of display modes analyzing for postures, a list type as
shown in Fig. 134.2, and the other a chart type shown in Fig. 134.3. Values in
Angle (angle) item in the list indicate the zero angle values of parts and positions
under a certain degree of freedom. The result (evaluation results) was expressed as
a percentage. It was used for indicating the comfort degree of postures, the higher
the score the more comfortable it is. Score (score) item indicates that the one when
the angle in the preferred area. In the chart, the color of the various parts complied
with the one set at its preferred angle region. When the angle is in the different
regions, it represent the color bar of the parts will be displayed in different colors.
If the color of the preferred angle region not set, the moving parts would not have
the appropriate colors.
134.2.2 Concept of Point H
H is the hinge point of the human body and thigh, i.e. the hip point (hip Point). In
determining the man-machine interface geometric size relationship of the auto
body, it is often with this point as the basis of positioning of the body. The actual H
is the midpoint of the left and right H-point marking connections on the manikin
when the 3D H-point manikin placed in the car seat according to the specified
steps. It indicates the position of the hip joint in the car after the driver or occupant
seated (SAE 1998, 2002).
134 Research on H-Point of Driver 1277
Fig. 134.2 Dialog box of assessment and analysis for manikin posture (List type)
134.2.3 Determination of H-Point Trajectory
Curve Equation
CATIA assembly module was used to import the human sitting posture model and
the parameterized dashboard vision design model into the same environment
(CATIA Object Manager 2000; CATIA 2000), as shown in Fig. 134.4.
Clicking the ‘‘open horizons window’’ button in the human modeling module in
Ergonomics, the system will pop up a vision window, in which images is within sight
of the manikin. Then entering the body of the analysis module of the model, opening
the ‘‘manikin posture assessment and analysis’’ dialog box and selecting the chart
1278 L. Sun et al.
Fig. 134.3 Dialog box of assessment and analysis for manikin posture (Chart type)
Fig. 134.4 Manikin imported
pattern, finally into the electronic prototype module and opening the motion Simu-
lation dialog box, the posture control of the manikin can be realized through the
adjustment of three driving commands of the dialog box. The size to be adjusted was
constrained by using the windows of the vision and the manikin posture assessment.
When the color bars in the window of the manikin posture assessment are all into
green and the images in the window of vision able to meet the requirements of the
134 Research on H-Point of Driver 1279
dashboard vision and the front vision, the right value of the slide adjusted by the first
two drive commands is the H point coordinates. The entire adjusting interface was
shown in Fig. 134.5.
The gender, percentile and nationality of dummies can be modified in Properties
of the manikin, and the modified position of the model and the related settings
remained unchanged, so the manikin with different percentiles can be easily and
quickly studied in the same file to obtain H-point coordinates of each percentiles.
Fig. 134.5 Adjustment interface of H point
1280 L. Sun et al.
With the heel point as the origin of coordinates, 10 mm in the Z direction as a
step, it can be found the range of the X coordinates of H-points of each percentiles
meeting the vision and comfortableness. The curves of the each equation above
were plotted in the same coordinate system, indicating the boundary curves of
H-point adjustment range of 5th, 10th, 50th, 90 h and 95th percentile manikin
(Sundin et al. 1966), as shown in Fig. 134.6.
MATLAB software was used to fit the curves, H-point trajectory curve equa-
tions (134.1) of the each percentile Chinese manikin were the following:
8
>
> Z95th1 ¼ 0:003864X 2 þ 4:547X 1042
>
>
>
>
>
> Z95th2 ¼ 0:001547X 2 þ 21:94X 7496
>
>
>
> Z90th1 ¼ 0:003795X 2 þ 4:349X 950:6
>
>
>
>
>
> Z90th2 ¼ 0:008559X 2 þ 11:19X 3372
>
>
>
>Z
> 2
< 50th1 ¼ 0:002326X þ 2:113X 159:9
>
Z50th2 ¼ 0:006999X 2 þ 8:536X 2308 ð134:1Þ
>
>
>
> 2
Z10th1 ¼ 0:01581X 19:58X þ 6273
>
>
>
>
>
> Z10th2 ¼ 0:1397X 2 þ 183:6X 60020
>
>
>
>
>
> Z ¼ 0:01055X 2 14:53X þ 5171
>
> 5th1
>
>
>
>
> Z5th2 ¼ 0:02228X 2 þ 23:51X 5810
>
:
Fig. 134.6 Adjustment scope of H point suitable for Chinese body
134 Research on H-Point of Driver 1281
Fig. 134.7 Adjustment scope of H point suitable for American body
For the American manikin, the same method can be applied to find the H-point
ranges of each percentiles meeting the vision and the comfortableness, as shown in
Fig. 134.7. The fit H-point curve equations (134.2) were:
8
>
> Z95th1 ¼ 0:003887X 2 þ 5:167X 1392
>
>
>
> Z ¼ 0:01078X 2 þ 16:59X 1608
>
> 95th2
>
>
>
>
> Z90th1 ¼ 0:003547X 2 þ 4:535X 1120
>
>
>
> Z90th2 ¼ 0:0125X 2 þ 19:27X 7134
>
>
>
> 2
> Z50th1 ¼ 0:0001384X 0:7642X þ 881:9
>
>
>
>
>
> Z50th2 ¼ 0:03041X 2 þ 46:31X 17330
>
> 8
>
> > 0:0108X 2 12:53X þ 3879ðX\638Þ
>
> >
<
<
Z10th1 ¼ 0:003364X 2 5:297X þ 2292ðX [ 638Þ ð134:2Þ
>
> >
>
>
> :
>
>
>
>
>
> Z10th2 ¼ 0:09494X 2 þ 136:9X 49050
>
> 8
>
>
>
> >
> 0:05769X 2 71:67X þ 22580ðX\627Þ
>
> <
>
>
>
> Z5th1 ¼ 0:05522X 2 72:69X þ 24190ðX [ 627Þ
>
> >
>
>
> :
>
>
>
>
>
> Z5th2 ¼ 0:02075X 2 þ 24:39 6749
>
>
:
1282 L. Sun et al.
It can be found through the comparison of Figs. 134.6 and 134.7 that congenial
H-point curves of the human of Chinese and Americans have a large differences.
Thereby if the population differences ignored and only a single standard used to
design in R & D process, it is bound to produce defects, adversely affecting the
quality of the product.
134.3 Conclusion
Combined with requirements of the vision and the comfortableness, the H-point
range of the driver has been researched and the boundary curves of the 5th, 10th,
50th, 90th, 95th percentile driver’s H-point range obtained. Only the joint angle
was divided into the comfortable and the uncomfortable when evaluating the
comfortableness. In the uncomfortable angle range, regardless of their proximity to
the comfort zone, the same score was adopted to assess. Thus it brings an
inconvenience for the flexibility of the scoring system. In conclusion, the rela-
tionship between the joint angle and the comfortableness should be improved
further.
References
Bubb H, Estermann S (2000) Influence of forces on comfort feeling in vehicles. SAE Paper, 2000-
01-2171
CATIA V5 (2000) Knowledge advisor user’s guide, Dassault Systems
CATIA Object Manager (2000) Interactive user access reference manual, Dassault Systemes
Hockenberry J (1979) Comfort, the absence of discomfort. CP. News. Human Factors Society,
April 1979
Huang J, Long L, Ge A (2000) Optimization on H point in car body packaging. Automot Eng
22(6):368–372 (in Chinese)
Ren J, Fan Z, Huang J (2006) An overview on digital human model technique and its application
to ergonomic design of vehicles. Automot Eng 28(7):647–651 (in Chinese)
SAE (1988) Recommended Practice J826. Devices for use in defining and measuring vehicle
seating accommodation, May 1988
SAE (1998) SAE J1516-1998, Accommodation Tool Reference Point
SAE (2002) SAE J826-2002, H-point machine and design tool procedures and specifications
Sundin A, Örtengren R, Sjöberg H (1966) Proactive human factors engineering analysis in space
station design using the computer Manikin Jack, SAE
Chapter 135
Research on Modeling Framework
of Product Service System Based
on Model Driven Architecture
Xin-zhi Zhao and Xin Cai
Abstract Produce Service System (PSS) has attracted much attention in recent
years for its providing new methods to the combination of product manufacturing
with service. Model building of PSS has become a basic question of related
researches. This article aims to propose a modeling framework able to characterize
elements and structures of PSS. First, present research achievements on PSS
modeling are analyzed. Second, PSS spatial structure and application model facing
to its whole life cycle are proposed. Third, a 4-layer modeling framework of PSS is
put forward; the meta-model of PSS is defined. At last, PSS single-view modeling
is discussed, so is the application of this PSS modeling framework based on MDA.
Keywords Model driven architecture Meta-model Product service system
135.1 Introduction
PSS is developing rapidly as a new kind of manufacturing paradigms, which is highly
integrated of product and service and wholly optimized and formed in the mode of
extended producer responsibility in which manufacturing enterprise is responsible to
product service in its life cycle (Gu et al. 2009). For its systematic mode of ‘‘product
and service’’, PSS has become a main solution of manufacturing enterprise extending
to service enterprise. It is widely discussed and studied in international academic and
industrial field. How to build PSS model should be the first question to be solved
among related studies. PSS includes so many elements, such as product, service,
X. Zhao
College of Management and Economics, Tianjin University, Tianjin, China
X. Cai (&)
Weichai Power Co., Ltd, Weifang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1283
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_135,
Ó Springer-Verlag Berlin Heidelberg 2013
1284 X. Zhao and X. Cai
manufacturing process, manufacturing resource, participation subject, value
delivery and commerce mode etc., that it needs a systematic framework to conduct its
model descriptions. This framework should be able to achieve the following:
describe, manipulate and organize every possibly existing enterprise entities, system
organizations, products, services, processes and supporting technologies; provide
different application views to describe a whole PSS model; offer modeling language
with enough expression abilities to describe all kinds of applications and the rela-
tionships among them; support model reuse in its system.
Present widely applied frameworks, for example CIM-OSA, ARIS, GRAI/GIM,
GERAM, IEM, PERA, Zachman, TOGAF, UML, IBM overall enterprise archi-
tecture framework and NIIIP etc. (Xu et al. 2007), mainly adapt to single enter-
prise modeling but hard to PSS modeling for its higher demands of expression
abilities, openness and maneuverability. However, the Model Driven Architecture
(MDA) technology can meet the demands much better, which describes models
from different levels of information structures, business logics and information
expressions with meta-model definition and single-view modeling.
135.2 Main Achievement of Existing PSS Modeling
Lin-yan Sun built an architecture based on the amalgamation of commerce mode
and producing organization style which integrates services and manufacturing.
Manufacturing enterprises cooperated with each other by the way of exchanging
manufacturing-process-level service; producer service enterprises offered
customers products and services through sharing manufacturing enterprises and
customers business-process-level service covering whole life cycles. With inte-
gration of product manufacturing, service offering and consuming, intellectual
capital, human capital and industrial capital were amalgamated together to con-
struct a value adding aggregation. Sun believed this architecture was a new kind of
commerce mode and also a new way of producing organization (Li et al. 2010).
Ping-yu Jiang analyzed the structure of PSS from the view of engineering
science and divided it into configuration system, scheduling system and service
supporting system. This structure focused on service transition but ignored inter-
face and product/service terminals (Jiang and Zhu 2008).
Meier proposed a hierarchy covering driving force, business environment,
business mode, capacity management, core business service, organization struc-
ture, society technology network, cost risk control and other relevant factors
mainly from the view of value network which was fairly completed but not clear
enough on expression of factors interaction (Meier et al. 2010).
Gu et al. 2009 put forward an architecture of product service life cycle man-
agement, which was actually a model of life cycle management of PSS based on
the platform of life cycle management system.
135 Research on Modeling Framework of Product Service System 1285
Qi et al. 2010, starting from the background of modern manufacturing services,
on the basis of analyzing the content of manufacturing services of PSS life cycle,
proposed a wheel structure of service-Embedded manufacturing. In fact, this was a
content structure of all kinds of manufacturing services in the process of PSS life
cycle.
Many existing researches try to explain the whole system with single structure,
while product service system is a complicated system consisting amounts of ele-
ments, which means single-structure-explanation is an unrealistic thing. Besides,
present researches on output forms usually stay at the level of ‘‘product and
service’’ without regarding it as an integrated system, which makes them too
simple to be completed enough. Therefore, to build the whole PSS model, it should
be started from the systematic integration of product and service according to its
spatial and temporal characteristics.
135.3 Analysis of PSS Structure
According to the structure characteristics of PSS and shortages of present
frameworks, PSS structure is analyzed from the angle of PSS spatial structure and
application model of PSS life cycle, which is a foundation for further modeling
research in this paper.
135.3.1 Spatial Structure of PSS
Present researches rarely focused on PSS structure. It was usually considered as
simple ‘‘product and service’’ which manifests as product and service being rel-
atively independent, or core products combining with additional service. These
descriptions were too simple to reflect the essential characteristics of PSS, which
was an important defect resulting in lack of further research.
As shown in Fig. 135.1, PSS is a system providing a unity of product and
service as terminal by which customers and other subjects can participates in PSS.
PSS has interfaces to interact with subjects, ports to access networks, and networks
to connect PSS platforms and other terminals. With the ports, PSS enables all
kinds of interaction between product and product, product and service, service and
service, product and platform, service and platform, etc. The essence of different
PSS systems lies in configuration of these structure elements.
1286 X. Zhao and X. Cai
End Products Terminal Services
Interface
Expand
Extended End Terminal
Products Services
Core End Port Core Terminal
Products Services
Network
Platform
Fig. 135.1 Spatial structure form of PSS
135.3.2 Application Model of PSS Life Cycle
PSS is a product highly integrating serving with manufacturing, ultimately aiming
at customer’s maximum demand, taking the life cycle of product and service as
temporal scale and material supplier to final customer as spatial scale with service
running through all activities of manufacturing services network to realize con-
tinuous value adding. Its application model contains full application of producing
service and serving production, as shown in Fig. 135.2. The former refers to all
kinds of services around manufacturing process, such as technical services,
information services, logistics services, management consulting, financial services,
human resource services, legal services, etc. while the latter refers to before-sale,
in-sale and after-sale activities including installation, commissioning, mainte-
nance, upgrades and remanufacturing, etc.
Marketing Design Manufacturing Purchasing Delivery Maintenance
Logistics
Consulting
Financial
Law
……
Information
Resource Allocation
Process Control
Infrastructure
Fig. 135.2 Figure of application model of PSS life cycle
135 Research on Modeling Framework of Product Service System 1287
135.4 PSS Modeling Framework Based on MDA
Above analysis of system structure is only conceptual representation of PSS,
which cannot be directly used for modeling. In the following part, according to
PSS system structure, this article manages to construct a PSS modeling framework
based on MDA via construction of meta-model, meta–meta-model, and single-
view models based on meta-model.
135.4.1 Model Driven Architecture
In the field of MDA, Object Management Organization (OMG) technology defined
a kind of 4-layer modeling framework which contains meta–meta-model layer
(M3), meta-model layer (M2), model layer (M1), and run-time layer (M0). Among
these, meta–meta-model layer contains the required elements of defining modeling
language. Meta-model layer defines the structure and syntax of modeling lan-
guage. Model layer defines a specific model of a system. Run-time layer includes
running status of objects of a model (OMG 2009). MDA based on meta-model has
been applied in the field of enterprise modeling but not been seen in PSS modeling
(Wan et al. 2012; Li et al. 2008).
135.4.2 4-Layer Modeling Framework of PSS
Figure 135.3 shows a 4-layer PSS modeling framework based on MDA. In this
framework, meta-model layer defines the semantic relationship of various ele-
ments, and model layer is a kind of instantiation of PSS. Due to the complexity of
PSS system structure, it is difficult to build a model from one single dimension.
Therefore, this paper proposes a PSS system model with single-view model based
on meta-model in model layer from different dimensions, such as organization,
process, product/service, communication, control, knowledge, and quality etc. All
views are unified together through the intrinsic correlation of meta-model. Data
layer mainly contains run-time information and data for PSS to describe its run-
ning status.
135.4.3 Meta-Model of PSS
Shown in Fig. 135.4 is meta-model of PSS, which defines the basic elements of
PSS including: (1) terminals, meaning the ‘‘products ? services’’ combination
delivered to customers; (2) interface, meaning the interaction interface between the
1288 X. Zhao and X. Cai
Meta-
meta- Meta-meta-Model
Model
Meta- Meta-Model
Model
Product & Process Resource Information Communica-
Model Service Model Model Model tion Model
Model
Knowledge Organization Quality
……
Model Model Model
Data the Information of Models
Fig. 135.3 The 4-layer modeling framework of PSS
Customer
IT Manufacturing Service
Resource Resource Resource
Service Service
Service Product Acceptor Provider Resource
Participator
Connector
Terminal Platform Service
Management
Interface
Network
Service Manufacturing Service
Scene Process
Coordination Services Strategies
Financial Logistics Design Processing Assembly Mainten-
Business Organism of Delivery Remanufact-
Services Service Services Service Service ance
Model Production Service uring Service
Services
Fig. 135.4 The meta-model of PSS
system and customer; (3) connector, which managing connections like products
and products, services and service, products and services, or product/service and
PSS platform; (4) platform, main part of the PSS, including process, participants,
resources, business model, production organization, service logic and application,
etc. The meta-model also defines the inherent relationship among various ele-
ments, including two basic relationships affecting and containing.
135.4.4 Single-View Model of PSS
Single-view model of PSS mainly includes product/service model, process model,
organization model, resource model, information model, communication model,
knowledge model, quality model etc. The first six models are the core ones. MDA
135 Research on Modeling Framework of Product Service System 1289
Meta-Model
Model
Instantiation
Design Standard Meta- Design
Requir- Model PSS
ement Custom Reference
Meta-Model Model Modeling
Framework
Fig. 135.5 The application of the PSS modeling framework based on MDA
emphasizes independence of each other among platforms, which means if the
internal consistency of semantics and syntax of different models is ensured, any
modeling methodology can be applied. Zhang et al. (2010) summarized the mod-
eling methodologies of process model and organization model. In particular, there
are varieties of modeling techniques applied to product and service modeling,
which mainly focus on only product or service but few on their integration. It is
indeed a kind of segmentation which results in the loss of characteristic information
of PSS. Therefore, the product/service model must provide a clear definition of
product, service, interface and network which is completed enough to meet the
demands of PSS model being systematic. Sadek (2009) provided an integrated
modeling framework for product/service without specific modeling technology.
UML and XML are believed to be able to meet related requirements in application.
135.4.5 The Application of the PSS Modeling Framework
Based on MDA
The PSS modeling framework proposed in this paper aims to provide a method-
ology for PSS modeling in PSS structure analysis and design.
In the process of PSS analysis and design, as shown in Fig. 135.5, PSS mod-
eling framework can be applied for building visual model, acquiring semantic
information of system structure, analyzing status data of system elements or some
other activities. Reuse a standard meta-model or customize a new one at first.
Second quote or refer to the reference models to proceed with single-model design.
After instantiation at last, a PSS model can be obtained.
135.5 Conclusion
PSS is a multi-elements, multi-subjects complicated manufacturing system.
Aiming at its system modeling, this paper presented a modeling framework based-
on MDA technology. Also, here proposed PSS structure form and application
1290 X. Zhao and X. Cai
model, put forward a 4-layer modeling framework, defined its meta-model, and
discussed PSS single-view modeling methodology. This framework is able to
describe the complex relationships among various PSS elements and support PSS
analyzing and designing. On this basis, customization and change of meta-model
shall be further discussed to support PSS modeling better.
References
Gu X, Li X, Qi G, Ji Y, Tang R, Jiang P (2009) Theory and key technology of product service
system. J Zhejiang Univ (Eng Sci) 12(43):2237–2243
Jiang P, Zhu Q (2008) Product service systems and research perspective. Manuf Autom 12(30)
Li Q, Li W, Xu D (2008) Meta-model-based enterprise model expression. J Tsinghua Univ
(Sci Technol) 4(48):1209–1212
Li G, Sun L, Gao J (2010) Architecture and implementation of service-embedded manufacturing.
Sci Technol Prog Policy 27(7):45–50
Meier H, Roy R, Seliger G (2010) Industrial product service systems IPS2. CIRP Ann Manuf
Technol 59:607–627
OMG (2009) OMG Model Driven Architecture [EB/OL]. http://www.omg.org/mda/
specs.htm#MOF. Accessed 24 Feb 2011
Qi E, Shi X-G, Li X (2010) Survey of research advancement on modern manufacturing services.
Ind Eng J 13(5):1–7
Sadek T (2009) Einmodellorientierter Ansatzzur Konzeptentwicklung industrieller Product-
Service Systems. Doctoral thesis, Ruhr-University Bochum, Institute: Product and Service
Engineering
Wan H, Shu Z, Li H, X Luo (2012) Theory of meta-model and its application in the development
and design of enterprise architecture. Syst Eng Theory Pract 4(32):847–853
Xu H, Xue H, Liu M, Wu S (2007) The status and trends of CIM reference architecture. Manuf
Inf Eng China 2007 5(36):17–21
Zhang X, Zhao X, Cai X (2010) Research on panoramic time-space-oriented enterprise modeling
integration architecture. In: 17th international conference on industrial engineering and
engineering management, IE and EM
Chapter 136
Research on the Civil Aircraft Customer
Service System Simulation Based
on SD Model
Jun Xu and Cui-xia Bi
Abstract Customer service ability is one of the key competitiveness in market
competition for civil aircraft manufacturers who participate. To establish the
mature reliable civil aircraft customer service system based on the service strategy
is an important aspect to reflect the customer service skills. Based on the idea of
system dynamics, this paper constructed a system dynamics model (SD model) of
the customer service system of COMAC, and through the simulation of the system,
analyzed the key factors which affect the running of the customer service system,
to provide the basis for decision-making and measures.
Keywords Civil aircraft Customer service system Simulation System
dynamic model
136.1 Introduction
Customer service ability is one of the key competitiveness in market competition
for civil aircraft manufacturers. To establish a mature reliable civil aircraft cus-
tomer service system based on the service strategy is an important aspect to reflect
the customer service skills, and has a close relationship with the well development
of China’s civil aircraft industry.
The customer service system of COMAC, is involving COMAC, customer
service center, suppliers, airlines and other subjects. These different subjects relate
to each other, and interact with each other, which together form a large compli-
cated system. In the situation that we cannot completely achieved quantitative
J. Xu C. Bi (&)
School of Economics and Management, Beihang University, Beijing, China
e-mail: [email protected]
J. Xu
Shanghai Aircraft Customer Service Co. Ltd., Shanghai, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1291
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_136,
Springer-Verlag Berlin Heidelberg 2013
1292 J. Xu and C. Bi
data, in order to analyze the working principle and the system function of the large
and complex system better, this article chooses using system dynamics model to
research problems. The article will construct SD model of the customer service
system of COMAC, and through the simulation of the system, to analyze the key
factors which affect the running of the customer service system, to provide the
basis for decision-making and measures.
136.2 Establish the SD Model of Civil Aircraft Customer
Service System
System dynamics is an efficient approach to understanding the behavior of com-
plex system. It deals with internal feedback loops and time delays that affect the
behavior of the entire system (Wang 1994), and has been applied extensively in
multiple fields (Lyneis 2000; Tan and Wang 2010; Ovalle and Marquez 2003; Gao
et al. 2006; Angerhofer and Angelides 2000; Kim 2003).
Customer service system of COMAC as the subsystem of the social economic
system, is a organic whole, its external system is the environment, which mainly
including the social, political, economic and other external environment, also
including systems of suppliers, logistics system, COMAC, airlines and other cli-
ents systems, which closely related with its operation.
COMAC customer service internal system is included management system,
technology system, security system, the evaluation system. The relationship of
each subsystem and relationship between systems and external environment are
shown in Fig. 136.1.
Fig. 136.1 System structure
of customer service system of
Supplier
COMAC system
external environment
support
management system
system
Airlines
system
technique evaluation
system system
COMAC
tit f
Logistics
System
136 Research on the Civil Aircraft Customer Service System Simulation 1293
Customer service system of COMAC is an open, dynamic and complex time-
varying system, the main internal factors is including the demand of customer
service, the progress level of science and technology (such as infrastructure con-
struction level, inventory management ability, modernization of the organization
operation and management, information level and working staff), customer service
capability (include fast response time, service satisfaction) and so on. The main
external factors are social and economic development level, COMAC profit level,
the investment of COMAC for customer service center, the service level of the
supplier and the supply capacity, logistics development level, etc.
In order to simulate COMAC customer service system, we must comprehensive
considerate the internal and external influence factors.
In this paper, the SD software Vensim PLE 32 (Ventana Simulation Environ-
ment Personal Learning Edition) was applied to establish the customer service
system based on its practical operation and powerful design tools.
136.3 Establish the SD Causality Diagram of COMAC
Customer Service System
By the comprehensive analysis of the causality relationship between the internal
and external factors of COMAC customer service system, this paper give an SD
causality diagram of COMAC customer service system as shown in Fig. 136.2.
Fig. 136.2 The SD causality diagram of COMAC customer service system
1294 J. Xu and C. Bi
Fig. 136.3 SD flow figure of COMAC customer service system
This paper used flow figure to research the system dynamic simulation of
COMAC customer service system. This paper based on the analysis of the cau-
sality diagram of COMAC service system to determine the flow figure.
In Fig. 136.3, this paper use the customer response time to stand for the service
ability level of COMAC customer service system, use the information level to
stand for the customer service center investment in logistic and information level
construction, use the supplier delivery time to stand for the supplier availability
(Hui and Jha 1999; Jenkins 1999). This paper is based on these hypotheses, to
carry on the SD simulation experiment. Specific variables and equation set of this
paper as follow.
136.3.1 Equation of State
In the model there are two state variables that are GDP value and profit of cus-
tomer service center. Two equations of state as follow.
• GDP value = GDP initial value ? GDP increment
• customer service center profit = profit initial value ? profit increment
State variable is connected with its initial value and growth rate. The growth
rate is description by rate equation.
136 Research on the Civil Aircraft Customer Service System Simulation 1295
136.3.2 Equation of Flow Rate
In the model there are two flow rate variables that are GDP increment and profit
increment. Two equations of flow rate as follow.
• GDP increment = GDP value * GDP growth rate
• profit growth = Airline service de mand * average profit of single service.
136.3.3 Assistant Equations
Assistant equations can make the rate equation be expressed briefly. In this paper
the assistant equations as follow.
• The civil aircraft traffic demand
In this paper the civil aircraft traffic demand can be obtained by regression
analysis based on the proportion of annual air transportation in GDP value.
According to the data of China Statistical Yearbook from 2000 to 2010, by
simple linear regression we got the formula as follow.
y ¼ 0:0081 x þ 565:08; R2 ¼ 0:9756 ð136:1Þ
We can see the goodness of fit is greater than 0.95, it is means that the fitting
precision is high, the equation is available. The unit of airplane traffic volume is
hundred million passenger-kilometers.
• Profit of COMAC
Suppose the relationship between the plane traffic demand and the operating
profit is linear, the operating profit ratio is referred to the ratio of BOEING, and set
it to be 8 %. The profit of COMAC can be obtained by the formula as follow.
The Profit of COMAC = The civil aircraft traffic demand * the operating profit
ratio.
• Airline service demand
Suppose the relationship between the airplane traffic demand and the service
demand is linear. And also suppose the service demand of unit traffic demand is 0.3.
• The investment of COMAC for the customer service center
Suppose the investment takes up about 20 % proportion in the profit of
COMAC.
• The investment for the construction of customer service ability
1296 J. Xu and C. Bi
This investment = the investment of COMAC for the customer service cen-
ter ? profit of the customer service center * the investment ratio for customer
service ability construction
Suppose the ratio is 30 %.
• Information level
Information level = the investment for customer service ability construction
* the investment ratio for information construction
• Order processing time
Order processing time = square of airlines service demand/Information level
• Own inventory
Own inventory = the investment for customer service ability construction
* 0.5 * 0.8
Own inventory is the spare parts of airlines demand which can be met by their
own stock. We suppose there is about 50 % of the investment for service ability
construction of the customer service center to be used for spare parts inventory
level construction. And suppose there is about 80 % of the own inventory can meet
the demand.
• Amount of goods purchased
Amount of goods purchased = airlines service demand- own inventory
• Logistics delivery time
Logistics delivery time = Square of own inventory/information level
• Supplier delivery time
Supplier delivery time = Amount of goods purchased * Supplier average unit
delivery time
In this paper the unit delivery time is supposed to be 2.
• Customer response time
Customer response time = Supplier delivery time ? Logistics delivery
time ? Order processing time
136.3.4 Model Constant
• GDP initial value
136 Research on the Civil Aircraft Customer Service System Simulation 1297
This paper chose the value of 2010 to be the initial value; it is about 39 trillion
RMB.
• The initial value of Customer service center profit
This paper chose the value of 2010 to be the initial value. In this year the profit
is -790, the income is about it is about 95.33 million RMB.
• GDP growth ratio
This paper chose the average growth ratio between 1980 and 2010 to be the
GDP growth ratio. By calculating it can be determined to be 10 %.
• Average profit ratio of single service
Suppose average profit ratio of single service is 15 %.
• The investment ratio for information construction
Suppose this ratio is 20 %.
136.4 The System Simulation Results Analysis
This paper based on the above assumptions to carry on the simulation of the
model, took the customer response time to be the target variable, to examine the
response time change level in different conditions. In this paper result 1 is obtained
based on the original hypothesis. Result 2 is obtained based on the condition that
the ratio of investment for information construction is raised to be 40 % and the
other conditions remain unchanged. Result 3 is obtained based on the condition
that the ratio of investment for inventory construction is raised to be 60 % and the
other conditions remain unchanged. Result 4 is obtained based on the condition
that supplier average unit delivery time changed to be 2.5 and the other conditions
remain unchanged. Result 5 is obtained based on the condition that the ratio of
investment of COMAC changed to be 40 % and the other conditions remain
unchanged. Result 6 is obtained based on the condition when the ratio of invest-
ment for inventory construction has been raised to be 0.5 and the other conditions
remain unchanged. In the research the choice of time span for 5 years, and the
result as follow.
We can see from Fig. 136.4 that, the best results are result 1 and result 2, the
better results are result 5 and result 6, the curve of the result 3 has a transitory
decline at first and then rose again as same as result 4. So we can get some
inference from this.
• The result 2 shows that the information construction level has large contribution
to the customer response time, so the service center should strengthen the
construction for it. Here the information network not merely be the construction
1298 J. Xu and C. Bi
Fig. 136.4 Contrast figure of
customer response time
of information network system, but also be the logistic network system. For the
customer service center of COMAC, if it has its own logistic team, it can save
the customer response time and rise customer satisfaction.
• We can see from the result 5 and result 6 that at present the construction for
basic ability should be strengthened, but after the time when basic ability had
met the demand, more investment will not bring about more contribution.
• We can see from result 3 that when the center input too much on the inventory
construction, the repay may not homologous be more, but may be bad. That
because when the investment for inventory is too much, the spare parts inven-
tory will be too much, the pressure on the management of inventory will be too
high, this may lead to a negative influence for the operation of the center. That is
in accordance with the principle that Inventory and not the more the better.
• We can see from result 4 that the supplier average unit delivery time does
negative influence for the customer service ability. So in the process of the
customer service center development, it should be put more attention on the
choice of suppliers.
136.5 Conclusion
Insufficient and inaccuracy data, difficulty of quantifying the relationship are
knotty problems in the research of social and economic problems study. The
structure of the SD model is based on feedback loop, the existence of the multiple
feedback loops make the system behavior pattern is no sensitive to parameters. So
the system behavior can be similar so long as the parameters are in accepted scope.
Just because this principle this paper gave the SD simulation of the civil aircraft
customer service system, and the result is objective and credible.
136 Research on the Civil Aircraft Customer Service System Simulation 1299
References
Angerhofer BJ, Angelides MC (2000) System dynamics modeling in supply chain management:
research review. In: Proceedings of the 2000 winter simulation conference
China Statistical Yearbook (1980–2010) China statistical yearbook. Beijing (in Chinese)
Gao J, Lee JD, Zhang Y (2006) A dynamic model of interaction between reliance on auto mation
and cooperation in multi-operator multi-automation situations. Ind Ergon 36: 511–526
Hui SC, Jha G (1999) Data mining for customer service support. Inf Manag 38:1–13
Jenkins D (1999) Customer relationship management and the data ware house. Call Center
Solutions, Norwalk, pp 10–22
Kim SW (2003) An investigation of information technology investments on buyer–supplier
relationship and supply chain dynamics. Michigan State University, Michigan
Lyneis JM (2000) System dynamics for market forecasting and structural analysis. Syst Dyn Rev
2:68–77
Ovalle OR, Marquez AC (2003) The effectiveness of using e-collaboration tools in the supply
chain: an assessment study with system dynamics. J Purchasing Supply Manag 9:151–163
Tan YY, Wang X (2010) An early warning system of water shortage in basins based on SD
model. Proc Environ Sci 2:399–406 (in Chinese)
Wang Q-f (1994) System dynamics. Tsinghua University Press, Beijing, pp 1–25 (in Chinese)
Chapter 137
Research on the Modeling Method
of Wargaming for Equipment Support
on Computer
Xiao-ming Du, Gui-qi Wang, Ping Gu and Lu Gao
Abstract The study of modeling is a core problem to actualize the function of
wargaming for equipment support on computer. The chessman is an information
carrier to show the situation of support force and also an assignment carrier to
implement the support activity in the wargaming system based on the computer.
Firstly, the method of modeling which is based on the period of chessman’s life is
put forward on the basis of analyzing the state of chessman’s life. Then, the
method is described, and the framework of model for wargaming is established by
the method. Finally, a case is introduced to explain the application of the method.
Keywords Chessman Equipment support Modeling Wargaming
137.1 Introduction
The wargaming (Peter 1990) for equipment support on computer is the application
of wargame’s principle in equipment support, by using which, the commander
using the wargame map and units representing the real battle field and military or
using the computer simulation model (Yang 2007; Peng et al. 2008) in terms of the
rule and the principle of probability theory to command the activity of equipment
support in the war for verifying and improving the equipment support project.
The activity of equipment support is a complex system, and how to study the
model for wargaming is a core problem to carry out the function of wargaming
based on computer. The scientific property of model connects with the capability
of wargaming flow and result. In the text, the method of modeling which is based
on the period of chessman’s life is firstly put forward on the basis of analyzing the
state of chessman’s life to establish the system of model for wargaming.
X. Du (&) G. Wang P. Gu L. Gao
Command and Management Department, Shijiazhuang Mechanical Engineering College,
Shijiazhuang, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1301
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_137,
Springer-Verlag Berlin Heidelberg 2013
1302 X. Du et al.
137.2 The Analysis for the State of Chessman’s Life
137.2.1 The Concept of Chessman
The function of chessman is to show different classes of army and weapons, and
the commander who uses the system of wargaming could inquire the ability
parameters which are evaluated by the level of the army’s training or the capability
of the unit’s equipment through the chessman (James 1997). The chessman’s
parameters of the wargaming for equipment support are composed of the value of
movement, defense, support, attack, and are composed of the information of
support object and unit’s code, and so on, as shown in Fig. 137.1.
The chessman is an information carrier to show the situation of military and is
an assignment carrier to implement the action in the wargaming system which is
based on computer (Liu et al. 2008). The forms and movements of chessman are
basically function achieved for the system working.
137.2.2 The Analysis for the Cycle of Chessman’s Life
Anything has a process from produce to perish. The chessman’s process which
involves produce change perishes; illuminate the flow of the wargaming. The
chessman’s movement is the carrier for wargaming (Ross 2006).
Firstly, the chessman’s entity is made by experts in military affairs through
generalizing the information and the attribute of equipment support forces and
combat forces. When the wargaming begin, the chessman is working on the
purpose of commanders who operate the command platform on computer and
working under the trigger conditions. The movements of chessman involve
mobility deployment maintenance regress, and so on. At the same time, the
chessman’s status messages are changing along with the movements of chessman,
and are displaying on the situation display platform for the commanders who want
to know it on real-time. In course of wargaming, if the chessman is exposed to the
firepower strike from enemy force, then the damage or perishes of the chessman is
coming, as shown in Fig. 137.2.
The supply state The morale state
The supply value The maneuver value
A battalion
support unit
Fig. 137.1 The chessman
137 Research on the Modeling Method of Wargaming 1303
Fig. 137.2 The chessman’s Concept
life
Produce
Data
Maneuverability
The Deployment
Chessman’s Activity
Life Maintenance
Return
Damage
Transformation
Perish
137.3 The Development of the Model System
for Wargaming
137.3.1 The Method of Modeling
The activity of wargaming for equipment support is a complex system; include the
support entity, the combat entity, the interactive relationship between the entity
and the correlation of the entity. So we need an effective method of modeling to
describe the entity, operating modes of system, complicate battlefield environ-
ment, relationship between the entity and the arbitrage formulae.
The method of modeling which is based on the period of chessman’s life is put
forward on the basis of analyzing the state of chessman’s life to establish the
system of model for wargaming. In course of modeling process, the chessman is a
main. The relevant model is established by analyzing the chessman’s life and the
variation state of chessman in different phases, and the homonymic model archi-
tecture is developed to describe the wargaming from the chessman’s life based on
the method. The entity model is coming with the produce of the chessman, to
describe the static and dynamic attribute of the entity, to model the support force
and the combat force. The structuring model is coming with the trigger conditions
of the chessman’s business activities, to describe the subjection relationship and
correlation relationship between the chessmen, to ascertain the support relation-
ship so as to the chessman’s business activities is under the right order. The
behavior model which is the core in the modeling is coming with the development
of the chessman’s business activities; it is composed of maneuverability model
1304 X. Du et al.
deployment model and so on. The information model and the interactive formulae
model are coming with the change of the chessman’s state, to describe the process
how the chessman’s state is changed and how the message and information is
transferred when the chessman is working. Otherwise the probability model and
the terrain environment model are developed to describe the haphazard and the
environment.
137.3.2 The Development of the Model
• The entity model is described the structure of entity, the attribute of entity and
the correlation of the entity (Xin et al. 2010). The wargaming entity model is
usually composed of the ability parameter, the structure and the state of the
military force. For example the support force entity model include the infor-
mation of the support units, such as the ability to maneuver support recovery,
the object and the range of operation, to be grounded on the realization of the
dispatching order. The combat force entity model is the object of the support
force, and it isn’t central model relative to the support force entity model, so the
model is simply described the equipment information, the comeback parameter
of combat, the real-time state and so on. Otherwise the equipment entity model
is developed to describe the equipment information, such as maintenance type,
mean time to repair.
• The structuring model is developed to describe the subjection relationship and
the correlation relationship between the chessmen, and to build up the organi-
zational relations of the chessman, the support force correlation, and the rights
of wargaming seat (Peng et al. 2009). The model’s function is to establish a
relationship between the commander’s order and the chessman’s movement.
The model is described the rights formulae for the wargaming seat to develop
the maneuver relationship between the wargaming seat and the chessman, is
defined the trigger conditions to develop the order’s produce and implement.
The subjection relationship and the correlation relationship are developed by
defining the relationship of chessman and the attribute of chessman to imple-
ment the dispatching order and the return order.
• The maneuverability model is developed to describe the process, the chessman’s
movement to the destination after incepting the relevant order. It is based for
quantifying the maneuverability ability of the chessman, synthetically referring
to the influencing factors of environment and enemy’s situation, to estimate the
case of chessman’s maneuver on the road. The model of wargaming is described
the attribute of force, the type of maneuver, the geography information,
the formulae to manage the haphazard. The attribute of force is composed of the
entity’s type, the force’s level, the maneuver ability, the real-time state, etc.
The type of maneuver is composed of the maneuver mode, the beginning
137 Research on the Modeling Method of Wargaming 1305
rapidity, the destination coordinate, the real-time point information. The func-
tion of geography information is to provide the battlefield environment data for
the maneuverability model, it include the weather parameter and the landform
influencing factors. The formulae to manage the haphazard is developed to
describe the case, when the haphazard happen, the chessman automatically
manage it, for example when the chessman is attacked, the formulae may
operate to tell the chessman to remain in concealment firstly, and retaliate upon,
then wait for the commander’s order, but not to sequentially move on the road.
• The deployment model is described when the chessman’s state accord with the
trigger.
• Composed of the support force attribute, the deployment formulae, the
deployment time, the information of operation site, etc.
• The model is the most important model in the behavior model, it is described the
maintenance business process (Xu 2008). When the degree of damage equip-
ment and the level of maintenance unit are accordant the chessmen which
substitute the force begin a maintenance activity to the damage equipment.
According to the class and the amount of damage equipment, the hours of
maintenance task is counted, then the value of maintenance force ability is
established by integrating the utilization of time and the grade of enemy force.
The chessman take turns to maintain the damage equipment until the list of task
is clear. If all of the tasks are achieved, then the maintenance model is over, if
not the wargaming is going on while the formulae judge the grade of enemy
force and the availability of the equipment.
• The information model is developed to describe the process, the information
transmitting and exchanging in the wargaming activity, the logic relation
between the chessmen and the data. The function of the model is to manage the
data transmitting between the entity models, the structuring model, the behavior
model, the interactive formulae model on the computer. In the wargaming the
information includes the command message, the feedback message, state change
message, etc.
• The interactive formulae model is developed to describe the process, the
chessman’s state changing when the chessman’s state accord with the trigger
conditions and the interaction effect happen (Liu et al. 2011). The interactions in
the wargaming for equipment support mostly include the value translation
between support and combat, the value of the support force change under the
enemy force. The model’s parameter is composed of the correlation type, the
trigger formulae, the support value, the combat value, the coefficient translation.
• The probability model is described the haphazard in the battlefield. In the tra-
ditional handwork wargaming, the designer makes use of the probability number
list and the dice to simulate the effect of the haphazard. But in the modern
wargaming system based on computer, the haphazard is simulated by estab-
lishing the probability model through the probability function.
1306 X. Du et al.
Fig. 137.3 The flow of
The wargaming begining
wargaming with focus on
chessman’s life
Dispatch to produce
support chessman
N
Y
Receive the support
order
Maneuver to
destination
The trigger conditions of support
N Y
The support begining
The chessman’s
state update
137.4 A Case
There is a case which is an equipment support activity in the wargaming to verify
the method for establishing the framework of the model integrality.
The flow of an equipment support activity: the commander assign a mainte-
nance unit to maintain many damage equipments after judging the situation in the
battlefield. When the chessmen arrive at the destination, the activity is operated
according to the trigger conditions, and the value of the chessman’s state is
changing at the same time (Fig. 137.3).
The analysis for modeling: after the commander make the dispatching order,
the chessmen are produced by distilling the state of the support units and the
combat units, the process is making the entity model. The commander make the
task order to the unit chessman to arrive at the combat unit chessman which one
needs support, the process is making the structuring model. After the unit chess-
man receives the task order, and maneuvers, deploys, maintains, returns, the
process is making the behavior model. When the unit chessman begin the support
activity, the transfer of the value between the chessmen is doing, the process is
making the interactive formulae model. In the whole activity, the data and
information are transferring at all times; the process is making the information
model (Fig. 137.4).
137 Research on the Modeling Method of Wargaming 1307
The wargaming begining
Dispatch to produce The entity
support chessman model
N
Y
Receive the support The structuring
order model
The
the Maneuver to maneuverability
destination model
Datas
transfering
The deployment
The trigger conditions of support model
N Y
The
maintenance
The support begining model
The interactive
The chessman’ s formulae model
state update
The information
model
Fig. 137.4 A corresponding relationship between the flow of wargaming and the framework of
model
137.5 Conclusion
The method modeling which is based on the period of chessman’s life is put
forward in the text, we establish the framework of the model for wargaming on
computer, and a case is introduced to explain the application for the method. But
the method has limitations, we should summarize the other methods on modeling
technique at home and abroad to amply design the models for the wargaming
system under the framework system of model based on computer.
References
James FD (1997) The complete wargames handbook. William Morrow & Company, New York
Liu JH, Xu XD, Xu XH (2008) Study on human-computer interaction platform for computer
wargame. In: Proceedings of Chinese control and decision conference, Chinese IEEE,
pp 2233–2238
Liu X, Long GJ, Chen C (2011) Modeling of entity’s data regulations of wargame. Ordnance Ind
Autom 30(Suppl. 8):35–39
1308 X. Du et al.
Peng CG, Liu BH, Huang KD (2008) The study of wargames based on HLA. In: Asia simulation
conference-7th international conference on system simulation and science and scientific
computing, Beijing, China
Peng CG, Ju RS, Yang JC, Huang KD (2009) Analysis of technology of modern wargaming.
J Syst Simul 21(Suppl 2):97–100
Peter PP (1990) The art of wargaming. Naval Institute Press, Annapolis
Ross D (2006) Designing a system on system wargame. U.S. Air Force Research Lab,
pp 149–153
Xin T, Wei W, Zhang MJ (2010) Wargame system modeling and CLIPS-based rule description
method. In: International conference on computer application and system modeling, IEEE
Chinese, pp 572–577
Xu XD (2008) Study and realization of human-computer interaction platform for computer
wargame. Northeastern University, Boston
Yang NZ (2007) Wargame, war game, simulation. The Publishing Company of PLA, Beijing
Chapter 138
Risk Sharing Proportion of Cooperation
Between the Banks and Guarantee
Agencies Based on Elman Neural Network
Jun Liang and Qiang Mei
Abstract Considering the problems such as weak practicality generated from the
application of the mathematical model to calculate risk sharing proportion between
banks and guarantee agencies. This paper puts forward that Elman neural network
model can be adopted to study risk sharing proportion between banks and guar-
antee agencies. The computing process is as followed. First of all, selecting the
existing sample to train network model, and then proving network availability
through the tests, finally inputting the actual data operations to obtain the evalu-
ation results. The result indicates that Elman neural network model exhibits more
effective performance than traditional mathematical model on estimating the risk
sharing proportion in practice.
Keywords Banks Guarantee agencies Neural network Risk sharing
proportion
138.1 Introduction
In the process of cooperation between banks and guarantee agencies, mathematical
model is universally applied to calculate the risk sharing proportion (Wang and
Zou 2011; Fu and Zhao 2006). But in this way there are unreasonable hypothesis
and unfavorable application blocking the using in practice which requires an extra
evaluation method to assess risk sharing proportion. This paper suggests a way to
J. Liang (&) Q. Mei
School of Management, Jiangsu University, Zhenjiang,
People’s Republic of China
e-mail: [email protected]
Q. Mei
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1309
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_138,
Ó Springer-Verlag Berlin Heidelberg 2013
1310 J. Liang and Q. Mei
study the cooperation of banks and guarantee agencies on risk sharing proportion
with the application of Elman neural network model (Jia et al. 2012, Liu et al.
2011).
Neural network model has the ability of self-organizing, applicability, self-
learning (Dong et al. 2007). Different from mathematical model, neural network
model particularly suitable for issues which needs considering a variety of factors
and imprecise and vague information processing. Neural network model is rep-
resented by the network topology, node characteristics, and learning rules instead
of particular logic function (Chen et al. 2005). The application of this method
greatly relieves the modeling difficulty and reduces modeling time and also it
significantly decreases the interference of human factors and effectively reduce the
number of auxiliary variables and intermediate variables which will make the
hypothesis be more reasonable. What’s more, provided the ample study samples,
reasonable network structure and well-designed training parameters trained based
on samples, neural network can automatically extract knowledge rules from his-
torical data to operate accurately the simulation among variables complex map-
ping relationships (Wang et al. 2010, You et al. 2012), so as to overcome the
limitations of traditional mathematical models.
138.2 Elman Neutral Networks Model
138.2.1 Basic Principle
The neural network model has already been widely applied in assessment and
prediction of various economic indicators and phenomena (Li 2002). Elman neural
network is a feedback neural network. Compared with feed forward neural net-
work, it has many advantages such as fast approaching, and high calculation
accuracy (Li et al. 2011). In assessment of the proportion of risk sharing, math-
ematical model computes the extreme point of income of banks and guarantee
agencies to determine the reasonable proportion. While the Elman neural network
model, to calculate the optimal fitted values of risk sharing proportion within the
reasonable range of error by fitting the complex relationships between the vari-
ables in the cooperation of banks and guarantee agencies (Li and Su 2010).
The planned process includes the selection of the existing sample to train
network model, the test of proved network availability, and finally the obtaining of
assessment result by inputting of actual data (Zhao et al. 2005, Hou et al. 1999).
This paper employs the segmentation method to form the risk sharing pro-
portion. According to the data acquired from the 12th national joint session of
guarantee agencies for small and middle-sized enterprises, most domestic banks, at
present, are not willing to share the risks with the guarantee agencies. Even if some
banks think over to share risk, the sharing proportion is generally less than 10 %.
But referring to the experiences and data analysis of the cooperation of banks and
138 Risk Sharing Proportion of Cooperation 1311
Table 138.1 List of risk sharing proportion
Fuzzy comprehensive evaluation levels V1 V2 V3 V4
Risk sharing proportion [15 %, 20 %] [10 %, 15 %) [5 %, 10 %) [0 %, 5 %)
guarantee agencies at home and abroad, it shows that bank are likely to share more
risks after assessing risks of guarantee agencies in the process of cooperation. The
upper limit of bank risk sharing proportion is configured as 20 %. Therefore, this
paper adopts the upper limit.
Meanwhile with reference to the present situation of guarantee agencies in
Jiangsu province, risk sharing proportion is theoretically segmented into four
levels as V1[15 %,20 %], V2[10 %,15 %), V3[5 %,10 %), V4[0 %,5 %) of bank
risk sharing proportion. V1 stands for the less risky guarantee agencies, banks tend
to share the highest [15 %, 20 %] level of risks. The V2 represents comparably
secure guarantee agencies, banks are willing to share [10 %, 15 %) of the risks.
The V3 stands for the generally risky guarantee agencies, banks can share [5 %,
10 %) of the risks. V4 represents the most risky guarantee agencies that banks are
usually unwilling to share the risks which means banks will share [0 %, 5 %) of
the risks. Specific data is shown in Table 138.1.
Based on the rating method calculated by Elman neural network model and the
evaluation value interval in Table 138.1, the level of the evaluation value of
guarantee agencies can be conjectured, and then decided the risk sharing pro-
portion that banks are willing to share with guarantee agencies.
138.2.2 Fulfillment Process
The process of evaluating the proportion of the specific risk shares by Elman
neural network model are as following (Cong and Xiang 2001, FECIT Science and
Technology Product Development Center 2005):
(1) Select parameters. Referring to researches and experiences at home and
abroad and considering the analysis result of the data of the guarantee agencies
in Jiangsu province, selecting capital, asset ratio, guarantee business profit-
ability, guarantee compensation rate, compensatory loss rate, margin ratio, re-
guarantee proportion, and willingness of cooperation, which is standardized as
input parameters of Elman neural networks.
(2) Determine the target output model. The level of risk sharing proportion
between banks and guarantee agencies are divided into four levels, using the
following array to indicate target value:
1312 J. Liang and Q. Mei
V1 :ð1; 0; 0; 0Þ
V2 :ð0; 1; 0; 0Þ
V3 :ð0; 0; 1; 0Þ
V4 :ð0; 0; 0; 1Þ
(3) Input sample data to train Elman neural network. The network should be
ensured to meet the evaluation requirements of cooperation of banks and
guarantee agencies.
(4) Input test samples. The trained model Establish learning network to evaluate
the error, according to European norm theory.
(5) Input the evaluation indicators to calculate by Elman neural network.
(6) Acquiring the bank risk sharing according to the output vector-valued. The
largest dimension of the output vector s is the risk rank which the bank wants
to share. For example, output (0.6, 0.3, 0.1, 0.1), risk sharing for V1, namely
[15 %, 20 %].
138.3 Empirical Analysis
Before empirical analysis, the data input and the collection of original data require
to be processed in specific patterns to meet the request of the model. So the first
mission is to unify and standardize the data format (FECIT Science and Tech-
nology Product Development Center 2005; Song and Bai 2010).
(1) Design of input and objective vector
Elman neural network’s input parameters consist of seven indexes including
the capital, guarantee scale, guarantee business profitability, margin ratio, the
asset-liability ratio, compensatory loss and cooperation aspiration. The origi-
nal data of the number of the vector from a level, in order to prevent partial
neurons to supersaturated, so before inputting to the neural network, these data
should be standardized, and then provide system for the corresponding oper-
ation. Here sample data are standardized between [0, 1].
(2) Using Elman model for the evaluation of risk sharing proportion between
banks and guarantee agencies.
One of the most important inputs of Elman neural network model is learning
sample. The same data, according to different ways of learning training, will
produce different outcomes, so the set of study sample directly affects the output.
Table 138.2 gives the input vector of the twelve group study sample data (stan-
dardized sample data), and output vector corresponding is the bank risk sharing
proportion, it is known that risk allocation proportion is set to four interval [0 %,
5 %), [5 %, 10 %), [10 %, 15 %) and [15 %, 20 %].
138 Risk Sharing Proportion of Cooperation 1313
Table 138.2 Data samples of risk sharing proportion
Number Sample characteristics V1 V2 V3 V4
1 0.392 0.875 0.612 0.000 0.396 0.260 0.000 1 1 0 0 0
2 0.551 0.875 0.589 0.020 0.355 0.486 0.000 1 1 0 0 0
3 0.435 0.875 0.566 0.018 0.330 0.434 0.000 1 1 0 0 0
4 0.391 0.875 0.640 0.175 0.176 0.146 0.000 1 0 1 0 0
5 1.000 0.875 0.564 0.200 0.049 0.294 0.000 1 0 1 0 0
6 0.184 0.000 0.661 0.211 0.341 0.605 0.000 0 0 1 0 0
7 0.108 0.000 0.573 0.251 0.643 0.134 0.000 1 0 0 1 0
8 0.184 0.000 0.661 0.272 0.233 0.161 0.000 0 0 0 1 0
9 0.092 0.875 0.611 0.475 0.301 0.020 0.000 0 0 0 1 0
10 0.049 0.875 0.592 0.582 0.317 0.222 0.788 0 0 0 0 1
11 0.051 0.000 0.658 0.652 0.344 0.062 0.000 0 0 0 0 1
12 0.091 0.000 0.564 0.846 0.370 0.533 0.067 0 0 0 0 1
The formal evaluation of neural network will be started after determining the
learning sample. Four steps are required which include network creation, network
training, error inspection and network output. These steps will be illustrated in the
following parts.
The first step: network creation.
The three layers of network are considered to be a fairly effective solution to
recognize the general pattern issues. The three layers comprise an input layer,
hidden layer and output layer. Among the three layer network, the number of
hidden neurons is two times of the number of the input layer plus 1.
According to the principles above, this paper designs the network according to
the following way: The number of neurons inputting layer network is 8, so that the
number of neurons in hidden layer is 17, and output layer number of neurons is 4.
For the convenience of analysis, the following data structure can be used to
create he network model. The standardized network ensures the input vector in
range of [0, 1], hidden neurons in the transfer function of using tansig tangent
function, the output layer neural function using logsig logarithm function, so that
the output model could satisfy the network of output requirements.
threshold ¼ ½0 1; 0 1; 0 1; 0 1; 0 1; 0 1; 0 1; 0 1
net ¼ newelmðminmaxðPÞ; ½17; 4; f0 tansig0 ; 0 logsig0 gÞ
Among them, threshold defines the input vectors to the maximum value and the
minimum value. Network parameter is shown in Table 138.3.
The second step: network training.
Traingdx function will be called through the following codes.
Table 138.3 Network parameters
Training function Learning function Performance function
trainlm learngdm mse
1314 J. Liang and Q. Mei
Fig. 138.1 Training records
of risk sharing proportion
net:trainParam:epochs ¼ 500
net:trainParam:goal ¼ 0:01
net ¼ trainðnet; P; TÞ
The P and T respectively are set for input vector and target vector. The
Fig. 138.1 shows that, with the increase of the training intensity, convergence
speed gets higher. Network did not meet requirements, until the 181th training.
The third step: the error inspection.
The test is taken to display whether the network can meet the requirements of
the evaluation. Four groups of new data are selected as test data, as shown in
Table 138.4.
Test result:
Y1 ¼ ð0:0000 0:0977 0:4264 0:0031Þ
Y2 ¼ ð0:9994 0:0054 0:1293 0:0000Þ
Y3 ¼ ð0:0000 0:2146 0:0002 0:9992Þ
Y4 ¼ ð0:0360 0:0072 0:9991 0:0000Þ
The level is planned to be determined by the largest element from numerical
vector according to Elman neural network analysis, so it is suitable to use the
evaluation results of vector ?-norm to calculate the errors. These errors of the
assessment were 0.0023, 0.0006, 0.0008 and 0.0009. Obviously these errors are
within the acceptable limits (-0.003, +0.003) in the statistical scale. So it is
acknowledged that, the network could meet with the requirements of the risk
sharing proportion between bank and guarantee agencies after training.
The fourth step: network output.
138 Risk Sharing Proportion of Cooperation 1315
Table 138.4 Test data for risk sharing proportion
Data Assessment index data Risk sharing proportion
Test data 1 0.160 0.000 0.801 0.628 0.140 0.709 0.000 1 [10 %, 15 %)
Test data 2 0.511 0.875 1.000 0.222 0.160 1.000 0.223 1 [15 %, 20 %]
Test data 3 0.084 0.000 0.109 0.754 0.103 0.309 0.000 0 [0 %, 5 %)
Test data 4 0.065 0.000 0.831 0.000 0.119 0.457 0.000 0 [5 %, 10 %)
Table 138.5 Risk sharing proportion of banks
Guarantee gencies V1 V2 V3 V4 Risks sharing proportion
1 0.453 0.003 0.935 0.016 [5 %, 10 %)
2 0.042 0.957 0.006 0.002 [10 %, 15 %)
3 0.008 0.029 0.034 0.998 [0 %, 5 %)
4 0.118 0.016 0.469 0.008 [5 %, 10 %)
5 0.054 0.001 0.863 0.079 [5 %, 10 %)
6 0.025 0.090 0.003 0.979 [0 %, 5 %)
7 0.014 0.574 0.001 0.252 [10 %, 15 %)
8 0.006 0.002 0.999 0.383 [5 %, 10 %)
After inspection above, the estimated result of the risk sharing proportion is
accurate. The risk sharing proportion for the eight guarantee agencies are shown in
Table 138.5, which was calculated by the same method.
138.4 Conclusion
Elman neural network is used to estimate the risk sharing proportion between
guarantee agencies and banks. The process contains selection of the existing
sample to train network model, tests to prove network availability and finally input
of the actual data to calculate evaluation results. The application of artificial neural
network is predicted to perform effectively and scientifically to estimate the risk
sharing proportion and guarantee magnification for better practical generalization.
References
Chen X, Li J, Li Y, Ni Y (2005) Achievements of study on application of artificial neural network
to beverage industry. Beverage Ind 1:8–22 (in Chinese)
Cong S, Xiang W (2001) Design and selection of construction, parameters and training method of
BP network. Comput Eng 10:36–38 (in Chinese)
Dong P, Chen Y, Wang P (2007) Application of the steam coal quality predication technique base
on elman network. J China Coal Soc 12:1312–1315 (in Chinese)
FECIT Science and Technology Product Development Center (2005) Neural network theory and
MATLAB7 realization. Publishing House of Electronics Industry, pp 131–255 (in Chinese)
1316 J. Liang and Q. Mei
Fu J, Zhao H (2006) Research on risk sharing mechanism of commercial banks with guarantee
institutions on the base of credit assurance. J Syst Manag 6:565–570 (in Chinese)
Hou X, Chen C, Yu H, Wang T, Ji S (1999) Optimum method about weights and thresholds of
neural network. J Northeast Univ 4:447–450 (in Chinese)
Jia W, Zhou R, Zhang Z, Wang Z, Guo J (2012) Research on gear fault diagnosis based on elman
neural net. Comput Meas Control 5:1173–1175 (in Chinese)
Li X (2002) The establishment of economy forecasting model based on GMDH and artificial
neural network. Forecasting 6:63–66 (in Chinese)
Li X, Su X (2010) A new method for forecasting shield’s disc-cutters wearing based on elman
neural network. J Liaoning Tech Univ (Natural Science) 6:1121–1124 (in Chinese)
Li J, Ren Z, Liu Y (2011) Research on fault diagnosis system of mine ventilator based on elman
neural network. Coal Mine Mach 8:250–253 (in Chinese)
Liu N, Chen Y, Yu H, Fan G (2011) Traffic flow forecasting method based on elman neural
network. J East China Univ Sci Technol (Natural Science Edition) 2:204–209 (in Chinese)
Song S, Bai J (2010) The local impact analysis in artificial neural networks. Value Eng 7:144–145
(in Chinese)
Wang Z, Zou G (2011) The risk allocation mechanism in small and medium-sized enterprises
financing. J Tianjin Norm Univ (Social Science) 5:57–60 (in Chinese)
Wang L, Wang T, Chu Y (2010) Application of B-spline interpolation in system dynamics’
model based on BP artificial neural networks. Value Eng 14:153–154 (in Chinese)
You M, Ling J, Hao Y (2012) Prediction method for network security situation based on elman
neural network. Comput Sci 6:61–76 (in Chinese)
Zhao Q, Liu K, Pang Y (2005) A new training method of Elman and it’s application investigation
in system identification. Coal Mine Mach 5:73–74 (in Chinese)
Chapter 139
Simulation Analysis on Effect
of the Orifice on Injection Performance
Yu-lan Li, Xiang-bi An and Da-hai Jiang
Abstract The injector is one of the precision components for a diesel engine, and
it is inevitable to wear fault during utilization. For the fault of the orifice expansion
and the orifice obstruction, they are essentially changing the structure parameters.
In order to analyze the effect of the orifices on injection performance, the simu-
lation model of a certain type diesel injector was established based on AMESim.
And a simulation for a whole injection cycle of this injector was performed, thus
the injection characteristics and the relevant information about motion of the
needle valve was obtained. The effect on the velocity of the needle valve, the flow
rate and the volume of the fuel oil injection, etc. had been analyzed by changing
the number, or the diameter of orifices, and setting different diameters for each
orifice. The analysis would provide some references in structure design, optimi-
zation, testing data analysis and fault diagnosis.
Keywords AMESim Diesel injector Orifice Simulation analysis Working
process
139.1 Introduction
The injector is one of the precision components for a diesel engine, and it is
inevitable to wear fault during utilization. The orifice expansion and the orifice
obstruction are two of the most common fault phenomena.
The orifice expansion is due to constant spray and erosion of high-pressure fuel
oil flow on the orifices during the injector working. It drops the injection pressure,
Y. Li (&) X. An D. Jiang
Automobile Engineering Department, Academy of Military Transportation,
Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1317
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_139,
Ó Springer-Verlag Berlin Heidelberg 2013
1318 Y. Li et al.
shortens the injection distance, which leads the diesel atomizing worse and
increasing the carbon deposit in cylinder.
The orifice obstruction is due to half or complete block caused by the nozzle
corrosion during long-term storage for the diesel engine, or some solid impurity
particles mixing into the fuel oil, or the carbon deposit caused by bad combustion
accumulating around the orifices and making the orifices be in half blocking state
(Jin 2008).
These two kinds of fault phenomena are essentially changing the structure
parameters of the orifice, from the point of view of the physical mechanism. It is
difficult to record the parameter variable in real time during the working process of
the injector. So simulation is the common method for analysis on the injector
(Lv et al. 2009).A hole-type injector model is built based on AMESim to simulate
an injection cycle, and to analyze the effect of the orifices on the injection per-
formance, which can provide some references in structure design, optimization,
test data analysis and fault diagnosis.
139.2 Phenomena Analysis on Effection of the Orifices
on Injection Performance
The combustion process of the traditional diesel engine is mainly diffusion com-
bustion, and its combustion heat release rules and its fuel economy depends on the
fuel injection spray and the spread mix. So it is a high requirement for the spray
quality (Zhou 2011). The fuel injection spray process is very complex. As the fuel
spraying into the cylinder, the processes of fuel bunch rupturing, the fuel droplets
colliding and polymerizing, the fuel droplets running up against the cylinder wall,
and the fuel droplets evaporating spread, all accomplish in tiny space and time
scale (Xie 2005). The diameter and number are important parameters for the fuel
injection system in the diameter. The diameter of orifice has a great influence on
the fuel injection column shape, spray quality, fuel and air mixing state (Ma et al.
2008).
It is in favor of the fuel mixture formation to decrease the orifice diameter, but it
will also prolong the injection duration, in condition of the same cam lift and the
same injector open pressure. The average diameter and the heterogeneity of the
fuel oil droplets increase, the injection flow rate increases, and the fuel oil injection
duration is shorten, with the increasing of the orifice diameter (Jia et al. 2003).
However, smaller orifice diameter improves the low speed performance, mean-
while leads worse emission of NOX (Zhang et al. 2008). Therefore, the smaller
injector orifice diameter is more advantageous to forming the injection rectangle.
And the larger orifice diameter is benefit for reducing the diesel engine noise,
vibration index and emission levels, with being in line with the ideal fuel injection
law requirements of continuous acceleration injection until the quick broken fuel
oil injection process at the end (Wang et al. 2012).
139 Simulation Analysis on Effect of the Orifice on Injection Performance 1319
It is inclined to cause the fuel oil mist adhering to the cylinder wall and
producing more soot with too few orifices. It causes higher temperature inner the
cylinder and it is inclined to cause interference and overlap of the fuel oil bunch,
thus producing more NOX and soot, with too many orifices (Zhou et al. 2008;
Ding et al. 2008; Wu et al. 2010; Zhou et al. 2008).
139.3 Simulation Model and Injection Process Analysis
LMS Imagine.Lab AMESim offers a complete simulation suite to model and
analyze multi-domain, intelligent systems and to predict their multi-disciplinary
performance. The software creates a physics based model of the system, which
doesn’t require a full 3D geometer representation. AMESim can be used to study
the system or the components of steady state and dynamic characteristics. It adopts
top-down modeling method to achieve the complex system being modularized and
the abstract system being materialized. Now AMESim has been used to analyze
the fuel oil injection performance for the injector in practice (Boudy and Seers
2009; Wen and Zhang 2010; Zeng et al. 2008).
The simulation model including physical model of mechanical motion and fluid
movement is built according to the working principle of the injector. The model
can mainly be divided into the volume unit, the movement unit and the leakage
unit, in order to be convenient for analysis.
The concentrated volume of the injector is mainly in the nozzle, which is a
pressure chamber formed by clearance between the needle valve and the valve
body. In AMESim, the model of a conical poppet valve shown as in Fig. 139.1a is
used to signify the needle valve, and simulate certain valve by setting corre-
sponding structure parameters.
The movement unit of the injector points to the movement parts includes the
valve body and the mandril. In AMESim, the models are as shown in Fig. 139.1b
and c are used to signify the two components separately.
In order to signify the fuel oil leakage during the working process for the
injector, a model shown as Fig. 139.1d is used in AMESim.
The model of diesel injector composed of the three units above and other
necessary auxiliary components, which are as shown in Fig. 139.2. The model
Fig. 139.1 Main model units of the injector. a Model of conical poppet valve; b Model of
mandril; c Model of valve body; d Model of leakage unit
1320 Y. Li et al.
Fig. 139.2 Model of diesel
injector
simulates with the hypothesis that the fuel oil in the inlet passage is to motionless
at the beginning of the injection, because of stickiness force. That is the pressure in
the whole injector and the density of the fuel oil to be equivalent. The simulation
process computes a whole injection cycle, including the needle valve opening
time, the fuel oil injection duration tine and the needle valve closing time.
The basic parameters are four orifices with the same diameter of 0.28 mm. Each
parameter is set to basic parameter except the control parameter. In order to
analyze how the orifices affect the injection performance, the model batch runs
taking different parameter as control parameter respectively.
139.3.1 Analysis on the Injection Performance
with Different Number of Orifices
Setting model with 7, 6, 5, 4 or 3 orifices respectively, the curves of each per-
formance with different number of orifices are draw in the same graph.
The orifice number does not affect the injection time, but the injected volume is
slightly more with more orifices. It exerts great influence on the change process of
the injected volume. Usually, it takes shorter time to reach the maximum injected
volume and finish the injection process with more orifices. However, the influence
weakens after the orifice number increasing some value. The results are as shown
in Fig. 139.3.
139 Simulation Analysis on Effect of the Orifice on Injection Performance 1321
Fig. 139.3 Curves of
injected volume with
different number of orifices
Fig. 139.4 Curves of flow
rate with different number of
orifices
The fuel oil flow rate reaches maximum rapidly and then drops to zero quickly
with advisable orifice number, which closes to the ideal fuel oil flow rate curve.
Meanwhile, too few orifices leads to increase slowly after the injection, keep short
at the maximum flow rate, and take too long dropping to zero, which does not
conform to the requirements that begins and stops supplying fuel oil to the
combustion chamber quickly. The results are as shown in Fig. 139.4.
The needle valve rises following the law of slow first and then rapid, which also
being an ideal state of the injector. The needle valve rises rapidly at start time and
stays long at the maximum displacement. The results are as shown in Fig. 139.5.
139.3.2 Analysis on the Injection Performance
with Different Needle Valve Diameter
Shown as the curves in Fig. 139.6, the beginning injection time is brought forward
and the injection during time is longer with increasing of the needle valve diameter.
That is because the pressure-bearing surface area of the needle valve increases with
1322 Y. Li et al.
Fig. 139.5 Curves of the
needle valve lift with
different number of orifices
Fig. 139.6 Curves of flow
rate with different needle
valve diameter
the increasing of the needle valve diameter, which makes the volume in the pressure
chamber decrease correspondingly. So the pressure in the chamber increases fast,
and the needle valve opens earlier. The pressure falls slowly after the needle valve
opening which makes the needle valve keep the maximum displacement for longer.
At the end of the injection, the needle valve takes its seat quickly under the force of
the spring preload, and the injection flow rate falls to zero rapidly. But too large
needle valve diameter can make the pressure in the chamber fall slowly and produce
pressure wave that exceeding the needle valve opening pressure at the time of the
needle valve taking its seat, compelling the needle valve go up again as a result to
generate a twice-injection. However, too small poppet diameter can intensify the
volatility at the beginning injection time.
139.3.3 Relation Between Different Parameters
The effect of the orifice on injection performance is relevant to the orifice number
and orifice diameter. With different diameter for each orifice, the injected volume
and the injection flow rate computes as equivalent to convert the corresponding
number of diameter of the same orifices in the model.
139 Simulation Analysis on Effect of the Orifice on Injection Performance 1323
Fig. 139.7 Curves of the
pressure at the orifices with
different needle valve
diameter
Fig. 139.8 Curves of the
force on top of the needle
valve with different needle
valve diameter
Shown as the curves in Fig. 139.7 and 139.8, it is exactly the same form for the
curves of the pressure at the orifices and the curves of the force on top of the
needle valve. In other words, it will get same value with normalizing the corre-
sponding data.
139.4 Conclusion
The simulation model of a traditional diesel hole-type injector is built up based on
AMESim. By simulating an injection cycle with different needle valve parameters,
and analyzing the effect of the orifice on the injection working process, the con-
clusions are given below:
(1) It computes fast and accurately by simulating the injection performance based
on AMESim for the injector. And it offers important references for well
matching with the engine, designing and optimizing the injector.
1324 Y. Li et al.
(2) The parameters such as the flow rate, the injected volume, the velocity and the
lift of the needle valve, which character the injection performance. The
parameters are not mutually independent but have inherence relations with
each other. And they are coincident to each other. It is to examine how each
parameter satisfies the diesel engine features in certain aspect that analyzes the
curves of each parameter independently.
(3) The orifice number affects the duration of the injection process, especially
with too few orifices to meet the requirement. The injection flow rate increases
slowly at first, and drops to zero taking quite a long time after reaching the
maximum value, which does not accord with the requirement of the instan-
taneous injection, shown as the movement of the needle valve that it
residences too long at the maximal displacement, going against the throughout
distance and the spray column cone angle meeting required values.
(4) The orifice diameter has little effect at the beginning of the needle valve
opening, while has great effect on stopping supply fuel oil during later of the
injection process.
References
Boudy F, Seers P (2009) Impact of physical properties of biodiesel on the injection process in a
common-rail direct injection system. Energy Convers Manage 50(12):2905–2912
Ding J, Su T, Yang Z (2008) Optimized matching of injector by thermodynamic simulation.
Small Intern Combust Engine Motorcycle 37(1):31–33 (in Chinese)
Jia G, Pang H, Hong D (2003) Effects of fuel injection design parameters on diesel engine
emission characteristics. Diesel Engine, July, pp 35–38 (in Chinese)
Jin J (2008) Reason and elimination for diesel injector breakdown. Farm Mach Maintenance
36:26–27 (in Chinese)
Lv F, Cai Y, Li X, Li X (2009) Effect of injector specifications on combustion process of didiesel
engines. Tractor Farm Transp 6:82–85 (in Chinese)
Ma T, Li J, Wang D (2008) Improvement of the structure of a diesel engine’s fuel injector. Diesel
Engine 30(1):32–33 (in Chinese)
Wang L, Zhang Z, Liu P (2012) Simulation and experimental study on injection characteristics of
electronic control injector. Small Internal Combustion Engine and Motorcycle 41(2):14–16
(in Chinese)
Wu J, Wang M, Ma Z, Xu B, Liu Y, Wu R (2010) The effect of fuel injector parameters on
formation of mixture and combustion characteristics of diesel engine-based on fire numerical
simulation. J Agric Mech Res 18:202–205 (in Chinese)
Wen Y, Zhang Z (2010) Study on simulation of the diesel injector based on AMESim. Auto Mob
Sci Technol 6:38–41 (in Chinese)
Xie M (2005) Calculation of combustion engine, 2nd edn. Dalian University of Technology
Press, Dalian (in Chinese)
Zeng D, Yang J, Huang H, He W (2008) Working process simulation of an injector based on
AMESim. Small Intern Combust Engine Motorcycle 38:5–8 (in Chinese)
Zhang X, Song X, Yao H (2008) Research into the effect of injector structure on the performance
and emission of electric controlled diesel engine. J Yangzhou Polytech Coll 12(4):22–25 (in
Chinese)
139 Simulation Analysis on Effect of the Orifice on Injection Performance 1325
Zhou M, Long W, Leng X, Du B (2008) Simulation research on the effect of fuel injector
parameters on diesel’s combustion characteristics. Vehicle Engine 176: 21–26
Zhou B (2011) Internal combustion engine, 3rd edn. China Machine Press, Beijing (in Chinese)
Chapter 140
Simulation and Optimization Analysis
of a Wharf System Based on Flexsim
Na-qing Lin and Xiao-yan Zhai
Abstract From the viewpoint of production efficiency, a general cargo wharf
system optimization analysis is discussed in this paper. Wharf of Huangpu Port is
taken as an example. A simulation model of the general cargo handling system is
built by applying Flexsim simulation technology, which simulates the real oper-
ation process of wharf once being run. Then the analyses on the system bottle-
necks, problems and causes are carried out by figuring out the system output index
such as throughput, equipment occupying rate, average stay time of truck etc., on
the basis of which, a series of optimization approaches to solve the problems are
proposed.
Keywords Flexsim General cargo wharf Optimization analysis Simulation
140.1 Introduction
Modern logistics becomes important considerations in the government’s ‘‘twelfth
five-year plans’’, which emphasizes further improvement of transportation system
so as to upgrade modernization level of coastal port groups, and which proposes
further construction of organic links between railway, road, port, airport and city
transportation in order to accelerate the development of comprehensive transpor-
tation hubs. Wharf logistics plays a critical role in goods’ collecting and distrib-
uting, which once is optimized, large profit is at heel.
N. Lin (&) X. Zhai
The School of Management, Guangdong University of Foreign Studies,
Guangzhou, China
e-mail: [email protected]
X. Zhai
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1325
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_140,
Ó Springer-Verlag Berlin Heidelberg 2013
1326 N. Lin and X. Zhai
Huangpu Port is located on the estuarine area of the Pearl River, in the south-
east of Guangzhou City, which is a branch company of Guangzhou Port Group
Co., Ltd. Its business involves import and export bulk and general cargo handling,
storage and transportation. Taking responsibility of more than 60 % cargos col-
lecting and distributing, General Cargo Wharf of Huangpu nowadays encounters
some problems. Combining with research program of Guangzhou Port Business
Management System Construction, a simulation model for the practical operation
of general cargo wharf of Huangpu Port is built based on Flexsim technology. By
running the 3D model, system bottlenecks are recognized. And then, a series of
proposals are designed, which finally are validated by system simulation.
140.2 Status Quo of Huangpu Port and Wharf System
Simulation
140.2.1 Problem Description
Throughput of Huangpu Port reaches more than 28 million tons a year. For the
extensive economy hinterland, covering the Pan-Pearl River Delta region and
trading relationship with more than 60 countries and regions, it is one of the
important trade ports in South China. General Cargo Wharf is regarded as char-
acteristic wharf of Huangpu Port. However, in the recent year, handling capacity is
difficult for breakthroughs. 4 Problems are summarized as follows:
(1) Operation dispatching is based on experience management. Experience is
accumulated by operation planners or instructors who work on their job for a
long time, so that decision making is random and lack of scientific backing.
(2) Wharf service is simple, which mainly covers storage and transportation of
domestic cargo business, especially cargo handling, storing and transportation
of cargos in Pearl River Delta.
(3) Wharf information technology is in low level, and business process is slow.
Most business data is collected and handled manually. For the wide varieties
of cargo, bills of document are comprehensively delivered.
(4) Traffic jams occur frequently on the roads of wharf, which annoys customers,
some of whom complain a lot of it and some of whom just stop further
cooperation.
140.2.2 Building Wharf System Simulation Model
Flexsim is one of the most popular simulation software in the world, which
contains technology of 3-dimensional image processing, simulation, artificial
intelligence and data processing (Shi and Wang 2011), which is tailored to serve in
140 Simulation and Optimization Analysis of a Wharf System 1327
manufacturing and logistics industry. In this paper, taking general cargo wharf of
Huangpu Port as an example, a practical and visual wharf simulation system is
developed for decision making of wharf operation by Flexsim simulation tech-
nology, which simulates wharf operation by importing real data and which outputs
effective index to help managers recognize key problems and make good decision.
General cargos consist of steel, mechanical equipment, and cargos with pack-
aging, among which, steel takes up more than 50 %. In this paper, steel is chose to
be representative of cargos in the model. The equipments taken into use mainly
include gantry crane, jib crane, fork lift and trailer. Most cargos arrive at wharf by
water and leave by truck, in order to simplify the model, which is treated as the
only way of cargo flow in this paper.
The main part of wharf operation system can be simplified as a queuing system
of G/G/1. All the service wharf offers could be looked as service counter, and truck
or ship are service targets (Gao 2011), while truck service would be the focus in
this paper, which could be specified into detail service of truck scale, check-in and
loading (Fig. 140.1).
Assume that k is average arrival rate per minute, l is the average service rate
per minute, so that is
(1) The average minutes a truck spends on queuing
k
Wq ¼ ð140:1Þ
lð l k Þ
(2) The average minutes a truck stays in the system comprises time of running
(Tt), queuing and being served
1
Ws ¼ Tr þ Wq þ ð140:2Þ
l
(3) The rate if a truck arrives but couldn’t be served timely and has to wait
k
Pw ¼ ð140:3Þ
l
Fig. 140.1 Truck service
process sketch of general Wharf
cargo wharf from huangpu
port Truck Truck
arrives Truck Scale
leaves
Yard or
Check-in Shipside
1328 N. Lin and X. Zhai
We can build a practical and visual wharf system simulation model by applying
G/G/1 queuing mathematics model and Flexsim simulation technology.
140.2.3 Simulation Result and Analysis
By running the model for 7200 min (5 days), we can get the following data:
berthing time of ship is 23.4 h; and if a truck carries 60 tons, wharf handling
capacity is 3444 tons daily (Table 140.1). The counters of truck scale and check-in
are too busy, particularly truck scale, busy at 98 % of time (Table 140.2); Vacancy
rate of crane and fork lift is high, and there is large space for storage or other
operation on the yard (Table 140.3)
The time a truck spends in the system includes running on the road of wharf,
waiting for service and being served. Truck arrival time interval has Poisson
distribution, and the average interval is 12 min. Assume that it takes 14 min for
running on the road per truck. During 5 days, 287 trucks arrive, being served and
leave, so that is
1
k¼ 0:08
12
l ¼ 287 ð7200 287 14Þ 0:09
Plugging them into formula (140.1–140.3), we can get: it takes 134.65 min for
a truck to queue for service; it takes 159.74 min for a truck to stay on wharf; and
the rate that a truck arrives but have to wait for service is 0.92.
By investigation, we find that: Truck scale works for weighing and measuring
for truck and goods on it, which then print a weight note for goods loaded on each
Table 140.1 Throughput of Stay time Average content Input Output
model 0
Entry road 4.40 0.35 449 –
Exit road 4.00 0.16 – 287
Table 140.2 Other data Vacancy rate
report of model 0
Truck scale 0.02
Check-in 0.48
Table 140.3 Vacancy rate of Gantry crane Jib crane Fork lift
equipment of model 0
Vacancy rate 0.65 0.95 0.94
140 Simulation and Optimization Analysis of a Wharf System 1329
truck. Truck license number and weight note number are typed by hand, the
computer working for truck scale is aging and slow. Besides, there is little
information sharing between truck scale and other departments including sched-
uling department and check-in office. The arrival truck must go for truck scale to
have weight record of empty truck and then go for the check-in office. Check-in
officer will check the truck’s pick-up document and search for the corresponding
release sheet and then keep record on the tally sheet when the truck is loaded. For
all the documents are written by hand, it’s troublesome to find the history record,
which leads to average 10 min for truck check-in. All of these may result in traffic
jams near truck scale system. By the analysis above, we can focus on the handling
effectiveness on truck scale and check-in system. Proposals for optimization are as
follows.
140.3 Optimization Analyses
140.3.1 Constructing Business Management System
and Sharing Digital Information Between
Departments
Taking advantage of the current strategy of Guangzhou Port for constructing
business management system, upgrading the effectiveness of truck scale is
attainable. In the business management system, it is necessary to bind weight note
number with release sheet number. As a result, it’s possible to get the tally sheet
when weight note is printed. After check-in officer check the truck’s pick-up
document, the truck could go for yard or shipside for loading. In this way, the time
check-in officer searching release sheet in a piles of sheets is removed. Truck stays
for 2 min at check-in office, efficiency increasing 150 %. Faster computer is put
into use for truck scale, the efficiency of which will raise 100 %. By modifying the
corresponding parameter and running the new model again, we can get:
Truck arrival time interval is subject to Poisson distribution, arrival rate per
minute
1
k¼ 0:08
12
service rate per minute
l ¼ 407 ð7200 407 14Þ 0:27
Plugging them into formula (140.1–140.3), we can get: it takes 1.64 min for a
truck to queue for service; it takes 19.33 min for a truck to stay on wharf; and the
rate that a truck arrives but have to wait for service is 0.3.
1330 N. Lin and X. Zhai
Table 140.4 Handling capacity of model 1
Stay time Content Input Output
Entry Road 3.83 0.36 684 –
Exit Road 0.25 4.00 – 407
Table 140.5 Vacancy rate of equipment and gantry crane in model 1
Gantry crane Jib crane Fork lift
Equipment in model 1
Vacancy rate 0.44 0.84 0.83
Gantry crane in model 1
1 2 3 4 5 6 7 8
Vacancy rate 0.35 0.59 0.34 0.59 0.30 0.65 0.27 0.43
We can see each index is improved. Handling capacity reaches 4884 tons daily
(Table 140.4). For a truck carries 60 tons, and the system runs for 5 days, we can
come up with throughtput0 as follows:
throughtput0 ¼ 407 60 5 ¼ 4884
The occupying rate of gantry crane, jib crane and fork lift are increasing 60, 240
and 217 % respectively (Table 140.5). So, promoting information technology
arguably weakens the resistance from truck scale and check-in service on wharf
operation while in the other way, it demands more equipment. If arrival interval of
truck and ship are constant, truck can run without any obstacles and reach shipside.
More and more trucks arrive at shipside, if the handling of gantry cranes is not
effective, traffic jams would come up.
Therefore, we can conclude that model 1 can settle the traffic jams at a period of
time, which improves the occupying rate of equipment as well as enlarges wharf
throughput. However, when the handling capacity reaches a certain scale, equip-
ments are too occupied. Model 1 is not effective any more. At this time, the
planner should arrange more handling equipments so as to meet the demand of
larger throughput.
140.3.2 Reengineering Truck Business Process
and Improving Wharf Service Quality
(1) When the owner of cargo is applying for release sheet, he or she could apply
for an IC card according to their own need. Truck arrives at wharf with IC
card. At the truck scale, information of release sheet, weight note and cargo
storage position could be called up by swiping IC card. Loading admission
notice is printed and the information is also sent to scheduling department,
140 Simulation and Optimization Analysis of a Wharf System 1331
employees from which would dispatch operators to the corresponding storage
position.
(2) After finishing loading, the truck returns to truck scale for loaded weight.
Truck scale save the record and print exit admission note as well as tally sheet.
(3) Entrance guard checks the exit note and information on the business man-
agement system, if both match, truck can leave, and the truck business process
is over.
In this process, it’s unnecessary for the truck to check in. IC card is the cer-
tificate for picking up cargos. Business system identifies the validation of IC card,
and recognizes whether there is stock, which would save the time for truck (Cao
et al. 2009). And there is no need for check-in officer to keep record on the paper
tally sheet, and then type in the system. Tally sheet is generated automatically.
Modifying the parameter, running the new model, we can get:
After truck business process reengineering, there is no queuing phenomenon
with throughput of 6252 tons daily (throughtput1). For a truck carries 60 tons, and
the system runs for 5 days, we can come up with throughtput1 as follows:
throughtput1 ¼ 521 60 5 ¼ 6252
However, the owner of cargo may take the cost of IC card into consideration,
and compare it with cargo value, not all the owner choose to buy IC card. For the
limit length, this paper discuss situation which is assumed that all the owners
would purchase IC cards.
140.3.3 Assigning Wharf Equipments Reasonably
and Increasing the Operation Efficiency
Along with upgrading information system, the handling capacity is increasing
rapidly. New problems would come up if operation equipments remain the same
number as before. For the limit of wharf resource, equipment should be dispatched in
a reasonable way. At this time, scheduler can look at the data report of simulation
model, and arrange more resource on the weakest link whose occupying rate is high.
From report of model 1, the occupying rate of gantry crane 1, gantry crane 3, gantry
crane 5, gantry crane 7 and gantry crane 8 are more than 50 % (Table 140.6). It’s
reasonable to add more gantry crane on the corresponding berth. For there are 13
gantries on the wharf and 5 of them are not taken into use. At this time, they work.
Base on the designing of model 1, 5 gantry cranes are added. Modifying the
parameter, a new model is designed, by running which for 5 days, we can get
Tables 140.7 and 140.8.
By further improvement on Model 1, handling capacity (throughtput2) can
reach 6492 tons daily (Table 140.8). For a truck carries 60 tons, and the system
runs for 5 days, we can come up with throughtput2 as follows:
1332 N. Lin and X. Zhai
Table 140.6 Handling Stay time Content Input Output
capacity of model 2
Entry road 3.81 0.37 707 707
Exit road 0.28 4.00 521 521
Table 140.7 Handling Stay time Content Input Output
capacity of model 3
Entry road 3.97 0.38 684 684
Exit road 4.00 0.28 541 541
Table 140.8 Vacancy rate of Gantry crane Jib crane Fork lift
equipments on model 3
Vacancy rate 0.54 0.84 0.83
throughtput2 ¼ 541 60 5 ¼ 6492
Efficiency of wharf operation is advanced while the occupying rate of gantry
cranes is reduced by 18 %. The road loads near water is decreased, but as handling
capacity rises, the road to truck scale is inadequate for the amount of traffic, which
makes it hard for truck to reach the shipside area. New problems occur again.
140.3.4 Expanding Wharf Function and Strengthening
Regional Cooperation
According to modern logistics philosophy, traditional ports is going to be the
center of cargo transport, collecting, appreciation, assembly and distribution,
which control the speed cargo circulation (Li et al. 2010). Port logistics service
quality does not only depend on the efficiency of handling, but more on the
connection with its economical hinterlands. Approaches discussed above can settle
wharf operation in a short time. Nevertheless, in order to achieve the long term
profit, it‘s insufficient. After all, wharf system content is limited. In the context of
constant land area, traffic jams are inevitable along with the increase of
throughput.
In the long run, Huangpu Branch Company should seek to expand its wharf
function and try to develop multimodal transport and other regional cooperation.
Compared with traditional single means of transportation, multimodal transport
consists of two or more transportation modes, which plays a vital part in reducing
transportation cost and enhancing enterprises’ competitiveness (Gao et al. 2010).
Further more, in the wake of proposal for constructing and improving compre-
hensive transportation system, multimodal transport is highly promising. Huangpu
140 Simulation and Optimization Analysis of a Wharf System 1333
Port should make full use of its advantage from water transport and railway
transport, which could reduce the loads of wharf in a wide range and attain cost
advantage. It’s a good way to strengthen regional cooperation and try to build a
strong network connecting with waterway, highway, railway and skyway in order
to embrace an encouraging future (Ding et al. 2010). In addition, Huangpu Port
should reposition its wharf function. Through the analysis above, we can see that
the yard space and yard equipments are not fully utilized. It’s considerable to
select proper cargo which are strongly connected with the economical hinterlands
and offer supply chain service such as transportation between the upstream and
downstream, distribution processing on the wharf yard etc.
140.4 Strategy Analysis
According to Model 0, when the throughput is 3444 tons per day, bottlenecks
come up in front of the wharf logistics system operation. At this time, invoke
model 1. It’s a good way to solve traffic jams as well as handling capacity shortage
by standardizing business system and reducing manual labor. When the throughput
reaches 4884 tons per day, invoke model 2. Model 2 takes advantage from Model 1
and accelerates the handling efficiency in order to meet strong demand. Sources on
wharf are so limited, which requires reasonable distribution in dispatch agency.
Running the models, we can get indexes of operation, which are helpful to be the
guide for distributing more sources to the weak part which index suggests that the
vacancy rate is lower. When throughput becomes 6252 tons a day, the system
encounters bottleneck again. However, we can still use the model to carry out the
latter data analysis.
In this case, 80 % of cargoes are picked up directly from ship side by truck,
20 % of cargoes are picked up at the storage yards. And the utilization rate of
storage yard is only 0.05 %. The service function of Huangpu Port remains in
traditional handling-loading and uploading, which leads to insufficiency of storage
yard and its equipments. The situation will be changed if Huangpu Port extends
wharf service function (Hu et al. 2006). It is considerable to take full use of
equipments and offer potential logistics service to customers. For example, with
the help of source advantage and economic environment superiority, Huangpu port
can provide warehousing, transit shipment and distribution, etc. In addition, highly
dependency of road transportation is Huangpu’s weak point. Assume the port area
is constant, traffic jams will be inevitable result. Only when Huangpu Port expands
new profit source (Chen et al. 2005), can it solve the congestion fundamentally.
Huangpu port can have the aid from goodness on water transportation as well as
train transportation and promotes regional cooperation (Huang 2012), which forms
a strong system of seaway, road, and train as well as air network in order to make
full use of wharf sources.
1334 N. Lin and X. Zhai
140.5 Summary and Outlook
The simulation model provided in this paper is helpful for schedule manager to
make short-term decision. In the long run, Huangpu Port should make great effort
to facilitate the construction of regional transportation network (Tang 2011), and
carry out more regional operation. In addition, it should try to expand new service
field, for example, make full use of its advantage and offer tailored logistics
service for customers, select proper cargo and carry out supply chain service on the
wharf area etc.
References
Cao Q, Huang L, Song Y (2009) Optimization of truck distributing business process and system
design in port. Logist Technol 11:118–120
Chen Z, Cao X, Yan X (2005) Research on correlation relation between the development of
Guangzhou Port and Guangzhou. City Econ Geogr 25(003):373–378
Ding W, Zhang L, Li J (2010) The Construction of hub-and-spoke logistics networks and its
empirical research. China Soft Sci 08:161–168
Gao P (2011) Research on modeling and operation optimization on port logistics network system.
Dalian University of Technology, Dalian
Gao P, Jin C, Deng L (2010) A review on connection issues on port multimodal transport system
and their modeling. Sci Technol Manag Res 23:234–238
Hu F, Li J, Wu Q (2006) On the strategies for the utilization of Pearl river coastland and the
development of Guangzhou harbor. Urban Probl 130(2):31–35
Huang X (2012) Accelerate transition and construct new Guangzhou port. Port Econ 11:11–13
Li J, Chen Y, Zhai J (2010) Research on port supply chain system dynamics simulation model.
Comput Eng Appl 46(35):18–21
Shi Y, Wang J (2011) Optimization of production system by simulation based on UML and
Flexsim. In: International conference on management and service science (MASS), IEEE
China, vol 30, pp 1–4, Aug 2011
Tang S (2011) The interaction and development between Guangzhou port and city. Port Econ
6:48–51
Chapter 141
Simulation Design of Piezoelectric
Cantilever Beam Applied on Railway
Track
Hui Zheng and Fengjing Zheng
Abstract To achieve energy conservation, this paper optimized the size of
piezoelectric cantilever applied on railway track. In this paper, by modeling the
piezoelectric cantilever through ABAQUS simulation software, author calculated
resonant frequency by modal analysis. Considering the characteristic of track
vibration, the paper designed four kinds of piezoelectric cantilever beams for
absorbing different vibration in various frequency bands.
Keywords Green design Piezoelectric cantilever beam Rail Simulation
141.1 Introduction
In recent years, as wireless sensor networks widely used in rail transport, power
supply of wireless sensor network nodes causes for concern. Now wireless sensor
network node power supply is battery powered. Because of the limited life of the
commonly used chemical batteries, they need regular replacement. So this way
will bring a heavy workload, high cost and serious waste. For a large area of
wireless sensor networks, such as roadbed monitoring, battery replacement is
difficult. Therefore, the power supply and management of the wireless sensor
network node problems are urgent to be solved.
At the same time, the train produces a large number of vibration energy during
operation and radiant this energy by wheel-rail noise resulting in a lot of energy
loss and noise. If we can use this energy for power supply of wireless sensor
network nodes, We not only can solve the power supply of wireless network and
will be able to achieve the purpose of energy saving and environmental protection.
H. Zheng (&) F. Zheng
School of Mechanics, Tianjin University of Science & Technology,
Tianjin, People’s Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1335
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_141,
Ó Springer-Verlag Berlin Heidelberg 2013
1336 H. Zheng and F. Zheng
Now with the rapid development of electronic technology, the new ambient
vibration energy of a piezoelectric vibration generation came into being. Com-
pared to other micro-generation devices, piezoelectric self-generating device has a
simple structure, no heat, no electromagnetic interference. Therefore, piezoelectric
power generation device is widely used in different areas. Collecting the rail
vibration energy with piezoelectric vibrator will be new ideas to solve this problem
for wireless sensor network power.
141.2 The Mechanism of Piezoelectric Device Generator
Based on the piezoelectric effect of piezoelectric materials, piezoelectric materials
produce the deformation because of the incentive and its surface will release of
absorb charge, achieving the mechanical vibrations to electrical energy. Piezo-
electric ceramic is cut into the harmonic oscillator with specific mechanical vibra-
tion, known as the piezoelectric vibrator. The core component of the piezoelectric
generator is a piezoelectric vibrator, usually the main use of the piezoelectric
vibrator first piezoelectric equations (Wang and Su 2011), namely:
(
D ¼ dT þ eT E
S ¼ sE T þ d t E
where: D is the electric displacement, d is the piezoelectric constant, T is the
stress, eT is the dielectric constant, E is the electric field, S is the strain, sE is the
elastic compliance constant, dt is the transpose of d.
Piezoelectric ceramic material shows the piezoelectric effect, when it is
polarized. The external force applied to the piezoelectric ceramic material can
cause its deformation, resulting in charge. Charge after storage will be able to load
acting by circuit. Therefore, piezoelectric power generation device with the
rational design will produce repeated deformation through the absorption of
vibration energy environment to achieve the mechanical energy to electrical
energy conversion. These devices can long-term supply power for the wireless
network sensors, micro-power systems.
The piezoelectric generator modes depending on the polarization direction of
piezoelectric materials mainly make use of d33 and d31 modes. d33 electricity
generation mode stress and voltage direction are in the Z direction. The direction
of external force and voltage are the same. Generally, circular piezoelectric
vibrator use d33 mode, and d31 power mode collect vibration energy and achieve
electrical energy conversion by making use of external incentives. The stress
direction is in the X-axis direction, and the voltage is in the Z direction. The
direction of external forces and the piezoelectric voltage is perpendicular. Rect-
angular piezoelectric vibrator use d31 mode, and the model structure are simple.
Although d33 piezoelectric mode open circuit voltage is higher than the d31 mode,
the d31 mode is much better than the d33 mode on the performance of the
141 Simulation Design of Piezoelectric Cantilever Beam 1337
collected charge amount. The collection of environmental vibration of the
piezoelectric power generation device mainly use in d31-mode (Chu et al. 2008).
141.3 The Theory of Piezoelectric Cantilever
Vibration Power
Collection of environmental low-frequency vibration energy usually uses the
inertia of the free vibration of cantilever beams piezoelectric oscillator. Cantilever
support ways can produce the greatest deflection and supple coefficient with a
lower resonant frequency. Cantilever has a very wide frequency range, the reso-
nant frequency is from tens of hertz to several megahertz. Generating capacity of
the power generation is weak, but it can generate free vibration. The vibration has
longer duration and can be used to absorb the ambient vibration energy, to provide
continuous supply of electrical energy. For the cantilever type piezoelectric
oscillator, the output voltage U is:
3 t 2
U¼ g31 Ep u
16 L
2
3 L
U ¼ g31 F
4 tw
where: g31 is the piezoelectric constant of piezoelectric material; Ep is piezo-
electric modulus of elasticity; U is displacement of the cantilever free end; L; w; t
respectively, the piezoelectric ceramic chip length, width and thickness; F the
applied load of the free end.
141.4 The Structure of Piezoelectric Cantilever
The structure of piezoelectric cantilever is shown in Fig. 141.1. Piezoelectric
ceramic wafer is affixed to the substrate, and the reduced frequency mass is
attached to the free end of cantilever piezoelectric vibrator. The two parts con-
stitute a spring mass system. The role of the mass is to reduce the natural fre-
quency of the cantilever piezoelectric oscillator structure. Substrate acts as the
lower electrode at the same time. The top of the piezoelectric ceramic wafer is
covered with metal film as upper electrode. When the cantilever fixed end is
excited by vibration, the cantilever will produce bending vibration up and down,
causing the cantilever deformation and generate the charge and output voltage.
In this model, piezoelectric ceramics is PZT-5H. Substrate material is phosphor
bronze. Structural parameters are in Table 141.1.
With the ABAQUS simulation software, through modal analysis, the piezo-
electric cantilever structure of the first vibration mode is bending vibration up and
1338 H. Zheng and F. Zheng
Fig. 141.1 The structure of piezoelectric cantilever
Table 141.1 Geometrical parameters of piezoelectric cantilever
Material q/(kgm-3) E/G Pa Poisson’s ratio
Phosphor bronze 8920 106 0.35
PZT-5H 7500 61 0.31
down that meets the piezoelectric vibrator collection of mechanical vibration
energy requirements. The author mainly study various natural frequency of pie-
zoelectric cantilever. Though comprehensive consideration of resonance and
generate electricity requirement, the author Optimized the structure of piezo-
electric cantilever.
141.5 The Analysis of Track Vibration Feature
Track vibration is due to mutual collisions of the wheel and rail, when the train is
running. In the propagation of the vibration, the high frequency part decays faster
than the low frequency part of the frequency of the vibration which will change
over distance. Horizontal vibration attenuates faster than the vertical vibration.
Track vibration is complex synthesis of transverse, longitudinal wave, surface
wave. Because this kind vibration is affected by a variety of complex factors, it’s
vibration mechanism and the communication pattern fluctuate. Therefore, only
through a large number of measured data for statistical analysis to consider the
combined effects of various factors can we get the feature of track vibration. GAO
Guang-yun (Gao et al. 2007) tested Qinhuangdao-Shenyang passenger railway
tracks vibration. The test results show that: rail vibration frequency is about
100 Hz, and in the 70–130 Hz the amplitude is relatively large. Based on this, this
paper installed oscillator on rail base, shown in Fig. 141.2.
141 Simulation Design of Piezoelectric Cantilever Beam 1339
Fig. 141.2 Piezoelectric
vibrator installation diagram
141.6 Structural Design
The output current of double crystal parallel piezoelectric oscillator is significantly
higher than the output current of the single crystal and double crystal series
(Qi et al. 2011), so system selects the parallel double-crystal cantilever piezo-
electric oscillator shown in Fig. 141.1. Track vibration has a strong vibration
response in a wide frequency range. In order to maximize the absorption of the rail
vibration energy, the paper designs four piezoelectric cantilever structures to
absorb the vibration energy of the different bands.
As we know that track vibration has relatively large amplitude in frequency
range 70–130 Hz. Therefore, the resonant frequency of the cantilever structure
preferably falls within this frequency range for the greatest degree of recovery of
vibration energy. Frequency may be divided into 70–85, 85–100, 100–115,
115–130 Hz four sections; so we need four cantilevers, and their resonance fre-
quency is just within the four frequency bands. Taking median principle, we get
four target frequencies: 77.5, 92.5, 107.5, 122.5 Hz.
According to the structural characteristics of the piezoelectric crystal, thickness
of piezoelectric crystal adopted in the paper is 0.1 or 0.2 mm. Taking into account
that the track vibration acceleration is very large (about 100 g), the width is
20 mm in order to ensure the strength of the cantilever. According to the research
(Liu et al. 2011), output voltage of piezoelectric cantilever beam decreases with
increasing the length of the mass. Therefore, mass-length identified as the smaller
10 mm. Now we need to design four kinds of piezoelectricity cantilever with
resonant frequency 77.5, 92.5, 107.5, 122.5 Hz. In order to ensure the strength and
power requirements, we set the substrate thickness as 0.4, 0.6, 0.8 mm, and length
is set between 60 and 80 mm. Mass not greater than 6 g.
(1) Modal Analysis
With ABAQUS simulation software, the paper models the cantilever beam and
study the structure modal shown in Fig. 141.3.
By modal analysis, we get a variety of resonant frequency of the cantilever
structure. When the thickness of piezoelectric crystal is 0.1 mm, the natural fre-
quency of cantilever beam is shown in Figs. 141.4 and 141.5. (when the length of
cantilever is 70 mm, the natural frequency is very low.)
1340 H. Zheng and F. Zheng
Fig. 141.3 First-order vibration mode diagram of the cantilever structure
Fig. 141.4 Length of 115
piezoelectric oscillator is 110 0.4mm
105 0.6mm
50 mm, the cantilever 0.8mm
frequency(Hz)
100
vibration frequency 95
90
85
80
75
70
65
60
0 1 2 3 4 5 6
mass(g)
Fig. 141.5 Length of 90 0.4mm
0.6mm
piezoelectric oscillator is 85 0.8mm
frequency(Hz)
50 mm, the cantilever 80
vibration frequency 75
70
65
60
55
50
0 1 2 3 4 5 6
mass(g)
When the thickness of piezoelectric crystal is 0.2 mm, the natural frequency of
cantilever beam is shown in Figs. 141.6, 141.7 and 141.8.
From the above modal analysis, we can see that with the increasing in the
length of the cantilever beam and the mass of mass block, the cantilever’s resonant
frequency will reduce. Base on this, we design the size of the cantilever.
Now we design a kind of cantilever which has a resonant frequency 77.5 Hz.
First, we draw a line along the frequency 62.5 Hz in each map shown in
Figs. 141.4, 141.5, 141.6, 141.7 and 141.8, then we get a point of proximity and
ten points of intersection shown in Table 141.2.
141 Simulation Design of Piezoelectric Cantilever Beam 1341
Fig. 141.6 Length of 0.4mm
120 0.6mm
piezoelectric oscillator is 115 0.8mm
60 mm, the cantilever 110
frequency(Hz)
vibration frequency 105
100
95
90
85
80
75
70
0 1 2 3 4 5 6
mass(g)
Fig. 141.7 Length of 110
0.4mm
piezoelectric oscillator is 105 0.6mm
frequency(Hz)
0.8mm
60 mm, the cantilever 100
vibration frequency 95
90
85
80
75
70
0 1 2 3 4 5 6
mass(g)
Fig. 141.8 Length of 95
piezoelectric oscillator is 0.4mm
90 0.6mm
60 mm, the cantilever
frequency(Hz)
0.8mm
vibration frequency 85
80
75
70
65
0 1 2 3 4 5
mass(g)
According to the study (Shan et al. 2010), the output voltage of the cantilever
will increase with the increase of the length of the cantilever. So compare pro-
gram3 and program4, we choose program4. Compare program2 and program5, we
choose program5. Compare program6 and program7, we choose program7.
Compare program8 and program10, we choose program10. Compare program9
and program11, we choose program11. At the same time (Liu et al. 2011), the
output voltage of the cantilever will increase with the decrease of the thickness of
the cantilever. So Compare program4 and program5, we choose program5.
Compare program10 and program11, we choose program10. Cantilever substrate
material is phosphor bronze, the optimum thickness ratio of about 0.5 (Shan et al.
2010). So the thickness ratio of program5 is 0.33, and the thickness ration of
1342 H. Zheng and F. Zheng
Table 141.2 Size chart of cantilever
Number Length Mass Thickness of the Substrate thickness
piezoelectric crystal
1 50 0.6 0.1 0.4
2 50 1.9 0.1 0.6
3 50 4.1 0.1 0.8
4 60 0 0.1 0.8
5 60 1 0.1 0.6
6 50 4.2 0.2 0.4
7 60 1.2 0.2 0.4
8 60 3.2 0.2 0.6
9 60 6 0.2 0.8
10 70 0.47 0.2 0.6
11 70 1.8 0.2 0.8
Table 141.3 Size chart of cantilever
Number Length Mass Thickness of the Substrate thickness
piezoelectric crystal
1 50 0.6 0.1 0.4
2 60 1 0.1 0.6
3 70 0.47 0.2 0.6
program7 is 1. Therefore, we choose program5. Finally, we get three programs as
following Table 141.3:
For getting the optimal size, we further design of the structure. For getting the
maximum output voltage, we assume that the mass is 0 in program1 and 2.
In program1, the mass, thickness of the piezoelectric crystal and Substrate
thickness is 0.6 g, 0.1, 0.4 mm respectively. We assume that the mass is 0. The
frequency is shown in Fig. 141.9:
From the figure, we can get the maximum length is 54 mm.
In program2, the mass, thickness of the piezoelectric crystal and Substrate
thickness is 1 g, 0.1, 0.6 mm respectively. We assume that the mass is 0. The
frequency is shown in Fig. 141.10:
From the figure, we can get the maximum length is 59 mm.
So for the cantilever with resonant frequency 77.5 Hz, we have three structures
shown Table 141.4.
Now we build this three piezoelectric cantilever models, On condition that the
acceleration being 10 m/s2 and the vibration frequency being 77.5 Hz, ABAQUS
simulation is used to conduct simulation. The result is shown in Table 141.5.
It can be seen from Table 141.5 that program3 can output the most voltage. So
we choose program3 for absorbing vibration energy with 77.5 Hz.
Apply the same approach, we can find the other three the cantilever structure.
Results are summarized in Table 141.6.
141 Simulation Design of Piezoelectric Cantilever Beam 1343
Fig. 141.9 The curve of 90
frequency with length 85
80
frequency(Hz)
75
70
65
60
55
50
45
50 55 60 65 70
length(mm)
Fig. 141.10 The curve of 100
frequency with length 95
90
frequency(Hz)
85
80
75
70
65
60
55
50
50 55 60 65 70
length(mm)
Table 141.4 Size chart of cantilever
Number Length Mass Thickness of the Substrate thickness
piezoelectric crystal
1 54 0 0.1 0.4
2 59 0 0.1 0.6
3 70 0.47 0.2 0.6
Table 141.5 The output voltage of three programs
Program number Output voltage(V)
1 0.57
2 2.03
3 5.55
Table 141.6 Four kinds of cantilever structure
Length Mass Thickness of piezoelectric crystal Substrate thickness Resonant frequency
70 0.47 0.2 0.6 77.5
70 0 0.2 0.8 92.5
61 0 0.2 0.6 107.5
61 0 0.2 0.8 122.5
1344 H. Zheng and F. Zheng
141.7 Conclusion
By modal analysis and orbital vibration characteristics analysis, a new design
concept for designing piezoelectric cantilever size is suggested. After a compre-
hensive analysis, four kinds of cantilever beams matching track vibration char-
acteristics were designed.
Acknowledgments This work was supported by the Natural Science foundation of Tianjin,
China, unfer grant 10JCYBJC06800.
References
Chu J, Du X, Piao X (2008) Foreign development of piezoelectric generators for MEMS.
Piezoelectrics Acoustooptics 30(1): 22–25
Gao G, Li Z, Feng S, Sun Y (2007) Experimental results and numerical predictions of ground
vibration induced by high-speed train running on Qin-Shen railway. Rock Soil Mech
28(9):1817–1827
Liu H, Han S, He P, Zu Y, Zhou X (2011) Analysis and simulation of power generation capacity
of piezoelectric cantilever harvesting vibration energy. Min Process Equip 39(3):98–102
Qi J, Li Z, Yang H, Ye Q (2011) The development of subway vibration energy harvester system.
Energy Conservation Technology 29(4):315–322
Shan X, Yuan J, Xie T, Chen W (2010) Modeling and simulation of power generation with
piezoelectric unimorph cantilever. J Zhejiang Univ (Eng Sci) 44(3):528
Wang J, Su W (2011) Piezoelectric vibration theory and application. Science Press, Beijing,
pp 15–17
Chapter 142
Simulation of Scenic Spots Dynamic
Pricing Based on Revenue Management
Zhi-ping Duan, Shu-lian Yang and Fu-ping Zhang
Abstract The reasonable pricing of the scenic spots ticket involves many aspects,
and it is a complicated and changeable process. In this paper, according to the idea
of revenue management, scenic spot ticket dynamic pricing model is constructed,
simulation calculation is performed by using particle swarm optimization ant
colony algorithm. The simulation result shows that the dynamic pricing based on
revenue management can bring more profit and this study will provide a scientific
means for the scenic spots tickets pricing.
Keywords Ant colony optimization Dynamic pricing Particle swarm
optimization Revenue management
142.1 Introduction
At present, the scenic spot ticket pricing is muddledness and lack of scientific basis
in China. ‘‘Comparison method’’ or ‘‘follow-the-leader method’’ is the common
adopted pricing approaches by scenic operators. In this cases many scenic spots
cannot reach the expected passenger flow volume, and economic benefit of scenic
spots is influenced seriously (Lu et al. 2008). Revenue management is a set of
system management ideas and methods, it makes the scientific forecasting and
optimization techniques combine together with the modern computer technology
organically and faultlessly, its core is based on market segmentation. Revenue
Z. Duan F. Zhang
College of Economics and Management, Shandong University
of Science and Technology, Qingdao, China
S. Yang (&)
Computer basic courses department, Shandong Institute
of Business and Technology, Yantai, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1345
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_142,
Ó Springer-Verlag Berlin Heidelberg 2013
1346 Z. Duan et al.
management is to sell the right product to the right customer at the right time for
the right price through the right channel, so as to achieve the maximum profit. In
other words, revenue management is the art and science of predicting real-time
customer demand and optimizing the price and availability of products according
to the demand. This paper attempts to build the dynamic pricing model of the
scenic tickets based on the idea of revenue management. The simulation calcu-
lation carries out by using the particle swarm optimization of ant colony algorithm,
these works can provide scientific means for the scenic spot tickets pricing.
142.2 Revenue Management and Pricing Method
The revenue management area encompasses all work related to operational pricing
and demand management. This includes traditional problems in the field, such as
capacity allocation, overbooking and dynamic pricing, as well as newer areas, such
as oligopoly models, negotiated pricing and auctions.
An American airline is considered the main pioneer in this field (Duan et al.
2008). Recent years have seen great successes of revenue management, notably in
the airline, hotel, and car rental business. Currently, an increasing number of
industries are exploring to adopt similar concepts (Nagae and Akamatsu 2006;
Weatherford 1997).
Applying revenue management method to set ticket price is a typical market-
oriented pricing. It depends more on the relationship between supply and demand
rather than cost. Based on the EMSR (Expected Marginal Seat Revenue) theory,
one ticket revenue equals to the ticket price and the probability of being sold if
does not consider the cost.
The dynamic pricing method takes tourists willing pay (Wi ) as the basis of
setting price. We take the single scenic spot for instance, the scenic spot ticket
price which the consumers can accept set forth as follows.
Z 1
pi ¼ wi ¼ Fðp; zÞdp
0
In the formula: p stands for the costs from the Starting point to the scenic spots,
z stands for Socio-economic characteristics of the population. Because there are
many consumers, the total revenue TR of the scenic spots is equal to the sum of
different price that the consumers can accept.
X
n
TR ¼ pi
i¼1
142 Simulation of Scenic Spots Dynamic Pricing Based on Revenue Management 1347
142.3 Dynamic Pricing Modeling by Revenue
Management Theory
Revenue management pricing method is the typical market pricing method, which
reflects the game relation of interests of the tourists and the scenic spots in the
market
Firstly, we suppose that the capacity of a scenic spots is M. The advance sales
cycle is T. The price of each cycle keeps invariability at a period of time. If
booking tickets 4 weeks advance and each week as a pricing cycle, then the cycle
numbers are 4. Assume that the reserve price of visitors obeys a certain probability
distribution FðpÞ and it is changeless in the whole sales cycle, the sales price is pt
in each sales cycle, only when the reserve price is lower than the current prices
tourists will buy the tickets, so the probability of a ticket been bought by an arrived
tourist is 1 Fðpt Þ, thereby Dt ðpt Þ ¼ Mt ½1 Fðpt Þ is the demand function of
cycle t. Here Mt means the potential market size of the cycle t, the price pt is the
decision variable. The ticket revenue management of scenic spot aims to determine
the optimal price for different sale cycle in limited sales cycles of ½0; T 1, so as
to make the total revenue of scenic spots tickets maximize. Dynamic pricing model
can be expressed as follow.
X
T 1
Max TR ¼ pt Dt ðpt Þ
t¼0
8
>
> X
T 1
>
> Dt ðpt Þ M t ¼ 0; 1; 2; ; T 1
>
>
>
> t¼0
>
>
>
> pt b p t ¼ 0; 1; 2; 1
< aP
s:t pt 0 t ¼ 0; 1; 2; ; T 1
>
>
>
> ab
>
>
>
>
>
> a0
>
>
:
b0
The first restriction shows that the total number of tickets can’t exceed the
maximum capacity of the scenic spot, and the second says that the price in each
cycle cannot exceed a guiding price range providing by the state or department. a
and b express the lower limit and upper limit ratio of price change .
1348 Z. Duan et al.
142.4 The Solution Model of Particle Group:
Ant Colony Algorithm
Because the analytical solution of model is difficult to get, so optimization algo-
rithm are used to obtain the optimal solution. Traditional intelligent algorithm such
as genetic algorithms have defects as slow convergence speed, easy to get into
local optimal solution with combinatorial optimization decision problems (Tang
et al. 2010). This paper adopt particle swarm optimization ant colony algorithm to
obtain the solution model.
The ant colony algorithm was first proposed by Italian scholars Dorigo as a
bionic intelligent optimization algorithm (Dorigo et al. 1996; Dorigo and Gam-
bardella 1997). It aims to achieve the purpose of optimization by simulates the
process of ants looking for the shortest path between food source and their nest and
it has been applied to various engineering problems in recent years. It has the
properties of distributed computing, strong enlightening, positive feedback, par-
allelism (Dorigo 1999). However, it is known that the result of ant colony algo-
rithm relies heavily on the selection of algorithm parameters. If the parameter
setting is not properly, it is very easy to cause many problems such as increased
calculation, the slow solution speed, the long solution time and it is likely cannot
obtain the optimal solutions.
It is just because of the shortages of ant colony algorithm, many scholars have
proposed a variety of hybrid optimization algorithm for the improvement of the ant
colony algorithm, using particle swarm optimization ant colony algorithm is one of
them. For example, Li Shiyong and Wang Qing studied the extensive particle
swarm ant colony algorithm for continuous space optimization (Shi-yong and
Wang 2009), Yu Xue-Cai and Zhang Tian-wen applied multiple colony ant algo-
rithm based on particle swarm optimization to solve the vehicle routing problem
with time windows (Yu and Zhang 2010), Ye Rong and Zhao Lingkai studied the
Localization Algorithm for Wireless Sensor Network Based on Ant Colony Opti-
mization-Particle Swarm Optimization (ACOPSO) (Rong and Zhao 2011).
Particle swarm optimization ant colony algorithm was first suggested by
Kennedy and Eberhart (Kennedy and Eberhart 1995). Compared to other evolu-
tionary algorithms based on heuristics, the advantages of particle swarm optimi-
zation ant colony algorithm consist of easy implementation and a smaller number
of parameters to be adjusted. Therefore, it has been widely employed for com-
binatorial optimization problems (Kathiravan and Ganguli 2007; Hetmaniok et al.
2012).
Based on above references, this paper applies the particle swarm optimization
ant colony algorithm on the optimization of three control parameters (n; q; q0 ), the
solving procedure is as follows.
Step1: Particle swarm initialization, select n (n is the pricing cycle) particles
randomly, every particle contains three parameters (n; q; q0 ), here n [ ½1; 5 ran-
domly, q andq0 [ ½0; 1 randomly;
142 Simulation of Scenic Spots Dynamic Pricing Based on Revenue Management 1349
Step2: Ant colony algorithm initialization, place n ants containing only their
parameters randomly on n nodes, according their respective variable value to solve
fitness function value;
Step3: External loop calculator reset;
Step4: Internal loop calculator reset;
Step5: Perform ant colony algorithm on each ant containing respectively
parameters (n; q; q0 ), update pheromone;
Step6: If internal loop conditions are not met loop to step5, otherwise loop to
step7;
Step7: Update local pheromone, record the results of each ant, update n; q; q0
by ant colony algorithm;
Step8: If external loop conditions are not met loop to step4, otherwise loop to
step9;
Step9: Output the optimal solutions.
142.5 Model Simulation
This paper takes QingDao LaoShan JuFeng scenic spot as an example, Laoshan
scenic spot’s seasonality is very obvious, winter has less tourists, so the present
pricing of off-season is different from of busy-season. Assume that every April to
October as busy-season, the pricing is 95 yuan/people, every November to next
March as off-season, the pricing is 65 yuan/people. Take 4 weeks as a pricing
cycle; by using matlab as a simulation tool, simulations are performed for off-
season and busy-season respectively. Off-season starts from November and busy-
season starts from April, 8 weeks data are used. The benefits of the dynamic
pricing and the fixed pricing are compared. Simulation price in integer forms,
simulation results show as Tables 142.1 and 142.2.
The simulation results showed that based on revenue management the dynamic
pricing revenue are obviously higher than the fixed pricing whether in the off-
Table 142.1 Profit comparison for dynamic pricing and fixed pricing in off-season
Cycle Fixed price Dynamic price Fixed price Dynamic price Income increased
(yuan) (yuan) profit (yuan) profit (yuan) rate (%)
1 65 57 398580 417353 4.71
2 65 55 384215 402196 4.68
3 65 52 391365 408311 4.33
4 65 50 371280 387691 4.42
5 65 51 380510 396529 4.21
6 65 50 367575 382168 3.97
7 65 52 338000 351114 3.88
8 65 50 295295 306723 3.87
1350 Z. Duan et al.
Table 142.2 Profit comparison for dynamic pricing and fixed pricing in peak-season
Cycle Fixed price Dynamic price Fixed price Dynamic price Income increased
(yuan) (yuan) profit (yuan) profit (yuan) rate (%)
1 95 91 1517340 1619305 6.72
2 95 92 1552965 1658101 6.77
3 95 95 1588495 1690476 6.42
4 95 97 1778780 1913612 7.58
5 95 102 1879480 2023260 7.65
6 95 100 1782675 1916910 7.53
7 95 96 1737835 1866956 7.43
8 95 94 1756930 1887294 7.42
season or in peak -season, and the peak-season revenue is obviously higher than
off-season.it is mainly because of the number of visitors to the off-season is a lot
less than peak-season, and the dynamic price for off-season are lower than the
fixed price, but it is just the reverse for peak-season. The highest tickets price up to
102 yuan, and the time is just corresponding to the May 1 vacation. It denotes that,
when there are less visitors one obvious way to win more customers would be to
reduce the price. But the reality is the opposite in more tourists’ conditions. In this
way, the tourists flows have adjusted to a certain extent and the overall income will
be always higher than the fixed pricing mode. The whole tourist routes of laoshan
scenic spot up to six, the revenue management based dynamic pricing strategy will
bring considerable income for scenic spots.
142.6 Conclusion
The revenue management theory is used for building the dynamic pricing model of
scenic spots, simulation is performed by using the particle swarm optimization ant
colony algorithm, the scenic spots income tends to maximize based on the existing
by dynamic adjusting the tickets price. This will provide a scientific means for the
scenic spots tickets pricing. However, the scenic spots ticket pricing is a com-
plicated and changeable process, along with the change of ticket price, the con-
sumption behaviour of the tourists will change accordingly. The further researches
are very necessary for the more accurate demand forecasting and dynamic pricing.
Acknowledgments Financial support: National social science fund (11BJY121); The education
ministry humanities and social science research project (09YA790128); Shandong province soft
science research plan (2012RKB01209).
142 Simulation of Scenic Spots Dynamic Pricing Based on Revenue Management 1351
References
Dorigo M (1999) Ant algorithms for discrete optimization. Artif Life 5(3):137–172
Dorigo M, Gambardella LM (1997) Ant colony system: a cooperative learning approach to the
traveling salesman problem. IEEE Trans Evol Comput 1(1):53–56
Dorigo M, Maniezzo V, Colorni A (1996) The ant system: optimization by a colony of
cooperating agents. IEEE Trans Syst Man Cybern PartB 26(1):1–13
Duan Z, Li J, Lv Z (2008) Management in the Chinese scenic area of ticket pricing. Price Theory
Pract 06:35–38 (in Chinese)
Hetmaniok E, Słota D, Zielonka A (2012) Application of the ant colony optimization algorithm
for reconstruction of the thermal conductivity coefficient. Swarm Evol Comput 7269:240–248
Kathiravan R, Ganguli R (2007) Strength design of composite beam using gradient and particle
swarm optimization. Compos Struct 81(4):471–479
Kennedy J, Eberhart RC (1995) Particle swarm optimization. In: International conference on
neural networks, pp 1942–1948
Lu R, Liu X, Song R, Pan L (2008) A study on admission fee fixing model in china’s tourist
attractions. Tourism Tribune 23(11):47–49 (in Chinese)
Nagae T, Akamatsu T (2006) Dynamic revenue management of a toll road project under
transportation demand uncertainty. Netw Spatial Econ 6(3–4):345–357
Rong Y, Zhao L (2011) Localization algorithm for wireless sensor network based on ant colony
optimization-particle swarm optimization (ACOPSO). Comput Meas Control 3(19):732–735
(in Chinese)
Shi-yong LI, Wang Q (2009) Extensive particle swarm ant colony algorithm for continuous space
optimization. J Test Meas Technol 23(4):319–325 (in Chinese)
Tang L, Zhao L, Zhang Y (2010) Research on multi-period dynamic pricing model and algorithm
for fresh foods. J Syst Manag 19(2):140–146 (in Chinese)
Weatherford LR (1997) Using prices more realistically as decision variables in perishable-asset
revenue management problems. J Comb Optim 1(3):277–304
Yu X, Zhang T (2010) Multiple colony ant algorithm based on particle swarm optimization.
J Harbin Inst Technol 42(5):766–769 (in chinese)
Chapter 143
Study of Adaptive Noise Cancellation
Controller
Cui-jian Zhao and Su-jing Sun
Abstract One important problem in the adaptive noise cancellation controller is
responses feedback. This paper studies in the theory and proposes an improved
adaptive noise cancellation controller, which is used spectral line enhancement.
The simulation provides that the new cancellation controller was more efficient
and had robust performance.
Keywords Adaptive
Noise cancellation controller Responses feedback
Spectral line enhancement
143.1 Introduction
In the application of engineering, the most classical method for eliminating the
noise from signals is Wiener filtering (Dai 1994; Shen 2001; Hassoun 1995). But
design this filter must know the information of signals and noise. Begin from
60 years, with the developing of adaptive filtering theory, this problem becomes
not so important (He 2002). Adaptive filter can set apart the signals and noises
without information of them. Then, this technology was applied in many fields
(Wu 2001). However, there are still two concern questions—interfered reference
channel and responses feedback. These problems are both not solved satisfied
(Haykin 1994; Zhang and Feng 2003; Jiang et al. 2001). This paper is mostly work
over the question of responses feedback, and proposed a new improved adaptive
noise cancellation controller.
C. Zhao (&)
Department of Electrical and Information Engineering,
Shijiazhuang University, Shijiazhuang, China
e-mail: [email protected]
S. Sun
Institute of Information, Shijiazhuang Tiedao University, Shijiazhuang, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1353
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_143,
Ó Springer-Verlag Berlin Heidelberg 2013
1354 C. Zhao and S. Sun
Original import
Signal s + n0 Output e
source +
−
Noise n1 Adaptive y
source filter
Reference import
Error e
Adaptive noise cancellation
Fig. 143.1 Normal adaptive noise cancellation controller
143.2 Theory of Adaptive Noise Cancellation
Figure 143.1 is a normal adaptive noise cancellation controller (Yang and Zhou
1998; Larimore et al. 1978; Evinson 1946). It has two sensors. The signal will be
add irrelevance noise n0 when it was transported to the first sensor. Incorporate
signal s ? n0 were transmitted to noise cancellation controller from ‘‘original
import’’. The second sensor accepted the noise signal n1, which is irrelevant to
signal but is relevant to noise n0 with some way. The signal out from second sensor
was called ‘‘reference import’’ (Doherty and Porayath 1997).
Suppose s, n0, n1, y are fixed in statistic. Their average values all are 0. s and n0,
n1 are irrelevant, but n1 and n0 are relevant. The error of output is:
e ¼ s þ n0 y ð143:1Þ
If square formula (143.1) and get mathematic expectation. s and n0, n1are
irrelevant, so:
h i
E e2 ¼E s2 þ E ðn0 yÞ2 þ 2E½sðn0 yÞ
h i ð143:2Þ
¼E s2 þ E ðn0 yÞ2
Adjust the adaptive filter let E[e2] achieve the least. Because the power of signal
E[s2] is not affected, E[(n0 - y)2] will be the least. Ideal instance, E[(n0 - y)2] = 0.
Then y = n0, e = s, it indicate that the least power of output make output no any
noise.
143 Study of Adaptive Noise Cancellation Controller 1355
143.3 The Reasons of Responses Feedback and Solved
Project
143.3.1 Reasons of Responses Feedback
In Fig. 143.1, the method of adjusting parameters of adaptive filter is similar to the
steepest descent method, which is a kind of optimization algorithms. Actually, this
method is use instantaneous gradient r(k) ^ instead of factual gradient r(k),
namely:
rðkÞ ¼ 2E½eðkÞX ðkÞ
ð143:3Þ
¼ 2E n0 ðkÞ W ðkÞT X ðkÞ X ðkÞ
r^ðkÞ ¼ 2eðkÞX ðkÞ
ð143:4Þ
¼ 2sðkÞX ðkÞ 2 n0 ðkÞ W ðkÞT X ðkÞ X ðkÞ
While the formula of iterating weight is:
W ðk þ 1Þ ¼W ðkÞ þ lr^ðkÞ
ð143:5Þ
¼W ðkÞ 2leðkÞX ðkÞ
When the adaptive filter converges to steady state, system output signal e(k)
will be equal to the signal s(k). Then the weight will be changed based on the
signal s(k). So it can said that the signal s(k) have feedback through adaptive filter.
The next system output signal is affected by the frontal one. The formula (143.4)
indicated that the instantaneous gradient r(k) ^ is added an error vector:-
2 s(k)X(k), which should be equal zero. This will bring on surplus mean square
error. Thereby, system output signal e(k) will different from input useful signal
s(k). This will bring to distortion of signal, namely, responses feedback
phenomenon.
143.3.2 Solved Project
The former Refs. (Zhen-ya He et al. 2000; Cichocki and Unbehauen 1996; Miller
et al. 1990), adopted the method of variable step size algorithm commonly to
counteract responses feedback phenomenon. This paper is based on another
opinion to improve the iterating weight. Sequentially, it will weaken or eliminate
the effect of responses feedback.
From (143.3)–(143.5) we can see that the surplus mean square error is accretion
because of error vector -2 s(k)X(k), which exist in instantaneous gradient. Then
the formula of iterating weight (143.5) will be rewrite:
1356 C. Zhao and S. Sun
W ðk þ 1Þ ¼ W ðkÞ 2l½eðkÞ ^sðkÞX ðkÞ ð143:6Þ
where, ^s(k) is the estimate value of input useful signal s(k). In ideal condition,
^s(k) = s(k). Then the error vector will be eliminated completely from instanta-
neous gradient. Based on this theory, we designed another adaptive noise can-
cellation controller in the following text.
143.4 Improved Adaptive Noise Cancellation Controller
Based on the idea of above and adaptive spectral line enhancement, propose a new
adaptive noise cancellation controller show in Fig. 143.2. It is composed of two
parts, one is high frequency noise cancellation the other is low frequency noise
cancellation.
In the first part high frequency noise cancellation, by the theory of adaptive
spectral line enhancement we know, when the mixed signal get across the delay
z-4, the output of adaptive filter 1 is only contain narrowband weight, if delay
time is longer than the reciprocal of broadband bandwidth while shorter than that
of narrowband. The cause is the continuance of narrowband autocorrelation is
longer than that of broadband.
Then the reference input can be regarded as predictive estimate signal of error
e1. Use the difference e2 to adjust adaptive filter 1 in order to cancellation high
frequency noise.
High frequency noise cancellation
Signal +
source
+ n + −
0 e1
Adaptive
z− Δ Filter 1
Noise e2
filter − +
+
Noise Adaptive − Output
source Filter 2
e3
− Adaptive
Filter 3
+
Low frequency noise cancellation
Fig. 143.2 Adaptive noise cancellation
143 Study of Adaptive Noise Cancellation Controller 1357
Fig. 143.3 Input signal
143.5 Simulation
In order to validate the validity of adaptive noise cancellation controller proposed
in this paper, the following is simulation of it.
Input signal show in Fig. 143.3. It is composed of direct current signal and low
frequency alternate current signal:
sðtÞ ¼ 1 þ 2sinð0:1tÞ
Noise signal is consisted of low frequency alternate current (narrowband) signal
and stochastic (broadband) signal:
nðtÞ ¼ 2sinð2tÞ þ 0:5PðtÞ
In original import, the additive correlative noise n0 is brought by noise source n,
which is through a noise filter. Here we use nonlinear IIR filter (Zhang and Feng
2003), namely:
n0 ð k Þ ¼ n ð k Þ þ f ð n0 ð k 1Þ Þ
Here:
f ðn0 ðk 1ÞÞ ¼ exp ðn0 ðk 1Þ 1Þ2 =2r2 =2
exp ðn0 ðk 1Þ þ 1Þ2 =2r2 =2
Figure 143.4 shows this noise filter. r2 is 3. The mixed signal was shown in
Fig. 143.5.
The result of common adaptive noise cancellation controller is show in
Fig. 143.6. It effect is not so perfect. Figure 143.7 is the result of adaptive noise
cancellation controller proposed in this paper. Compare Fig. 143.6 with
Fig. 143.7, we can see that the improved adaptive noise cancellation controller is
better than common one obviously. Its wave is much smoother, there are lesser
burrs and their amplitudes are very small.
1358 C. Zhao and S. Sun
Fig. 143.4 Noise filter n n0
+
+
z−1
f n0 k −1 ))
) )
Fig. 143.5 Mixed signal
Fig. 143.6 Effect of
common cancellation
controller
Fig. 143.7 Effect of
cancellation controller in this
paper
From the simulation we can see that the adaptive noise cancellation controller
proposed in this paper can wipe off noise effectively, and the result is better than
the common.
143 Study of Adaptive Noise Cancellation Controller 1359
References
Cichocki A, Unbehauen R (1996) Robust neural networks with On-line learning for blind
identification and blind separation of sources. IEEE Trans Circuits Syst 43:894–906
Dai Y-s (1994) Weak signal detection methods and instruments. Defense Industry Press, Beijing,
pp 50–51
Doherty JF, Porayath R (1997) A robust echo canceler acoustic environments. IEEE Trans
Circuits Syst 44:389–398
Evinson NL (1946) The wiener RMS error criterion in filter design and pre diction. J Math Phys
25:261–278
Hassoun MH (1995) Fundamentals of artificial neural networks. The MIT Press, Cambridge,
pp 126–150
Haykin S (1994) Neural networks. Macmillan College Publishing Company, New York,
pp 260–264
He Z-y (2002) Adaptive signal processing. Science Press, Beijing, pp 67–68
Jiang M-f, Zheng X-l, Peng C-l (2001) The new variable step-size LMS-type algorithm and its
application during adaptive noise cancellation. Signal Process 17(3):282–286
Larimore MG et al (1978) Adaptive canceling using SHARF. Proceedings of 21st Midwest
symposium on circuits., pp 30–32, Aug 1978
Miller WT, Sutton RS, Werbos P (1990) Neural networks for control. MIT Press, Cambridge,
pp 255–260
Shen F-m (2001) Adaptive signal processing. Xidian University Press, Xi’an, pp. 80–81
Wu W (2001) Study of adaptive noise cancellation in enhancement of speed. M.S. thesis, Xidian
University, Xi’an
Yang J-x, Zhou S-y (1998) Simulation of adaptive noise canceller based on neural network.
J Date Acquis Process 13:74–77
Zhang Q, Feng C-q (2003) Variable step-size LMS algorithm and its application in adaptive noise
cancellation. Modern Electron Technol 14:88–90
Zhen-ya He Ju, Liu J, Yang L-x (2000) Blind separation of images using Edge worth expansion
based ICA algorithm. Chin J Electron 3(8):436–439
Chapter 144
Study of Cost-Time-Quality in Project
Failure Risk Assessment Based
on Monte Carlo Simulation
Xing Pan and Zi-ling Xin
Abstract In order to analyze project failure risk, the quality factor, added on the
basis of cost-time joint risk assessment, are showed by the degree of deviation and
expressed by 2-norm firstly. Secondly, considering the cost-time-quality factors
jointly, the joint distribution model of cost-time based on the Monte Carlo model is
established; meanwhile, the definition of project failure risk value is given. Last,
an example is given based on Program Evaluation Review Technique (PERT) to
simulate and analyse project failure risk through Monte Carlo Simulation (MCS).
Keywords Cost-time-quality Engineering project failure risk Monte Carlo
Simulation The degree of deviation
144.1 Introduction
Management of project typically includes the three aspects: cost, time and quality
(Oisen 1971). Project’s goal is to achieve the expected quality performance
requirement within the specified time and the approved budget. Cost, time and
quality influence each other.
In 1996, Babu and Suresh adapted the continuous scale from Zero to One to
specify quality attained at each activity. They developed the optimization models
and presented an illustrative example (in press) (Babu and Suresh 1996). In 2006,
Xu, Wu and Wang determined the conditional percentile ranking of the schedule (or
cost) values with the integration method, which combined Monte Carlo multiple
simulation analysis technique, regression analysis and statistical analysis (in press)
(Xu et al. 2006). Gao, Hu Cheng and Zhong built the mathematics model of their
X. Pan Z. Xin (&)
School of Reliability and System Engineering, Beihang University,
Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1361
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_144,
Springer-Verlag Berlin Heidelberg 2013
1362 X. Pan and Z. Xin
synthesis optimization considering time, cost and quality (in press) (Gao et al. 2007).
In 2009, XU Zhe, WU Jin-jin and JIA Zi-jun formed the marginal p probability
distribution functions and the conditional p probability distribution functions about
cost and schedule and analyzed the simulation outputs (in press) (Xu et al. 2009). In
2012, Kim, Kang and Hwang proposed a mixed integer linear programming model
that considers the PQLC for excessive crashing activities (in press) (Kim et al.
2012). The main idea of Monte Carlo Simulation is to estimate the amount people
interested in through simulating system reliability and risk behaviors randomly
(Dubi 1998, 2000; Yang and Sheng 1990; Marseguerra and Zio 2000).
This paper builds the model with Arena the software and analyses the results
with Excel to calculate the value of project failure risk, which relates to the
knowledge of probability theory and MCS.
144.2 Quality Assessment of a Project
Quality has an important influence on the risk of project. In this paper, the author
proposes a method to quantify the quality. The corresponding quality will grad-
ually increase when cost and time are increasing, and vice versa reduced. The
failure risk reaches the maximum when cost and time are smallest, which led to the
minimum of project’s quality. On the contrary, the failure risk reaches the mini-
mum when cost and time are biggest, which led to the maximum of project’s
quality.
Figure 144.1 is a scatter plot of the cost and time data. Horizontal and vertical
axes are the dimensionless cost C and time T, which are shown in the figure.
XðCX ; TX Þ is a cost-time point of the project, OðCO ; TO Þ is the minimum value of
all the data points, while AðCA ; TA Þ is the maximum. It shows that O and A defines
the scope of cost and time value. Therefore, the quality of the project is defined as:
the degree of deviation from the value point of project relative to the minimum of
cost and time. Within the definition of deviation from the farther, the higher the
quality of the project; deviation from the closer, the lower the quality of the
project. The formula of any quality values in Fig. 144.1 are defined as follows:
Fig. 144.1 Cost and time A
scatter plot
O
144 Study of Cost-Time-Quality in Project Failure Risk Assessment 1363
kOXk2
QX ¼ ð144:1Þ
kOXk2 þkXAk2
kOXk2 and kXAk2 are 2-norm, namely distance, of OX and XA, defined respec-
tively as follows:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kOXk2 ¼ ðCX CO Þ2 þðTX TO Þ2 ð144:2Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kXAk2 ¼ ðCA CX Þ2 þ ðTA TX Þ2 ð144:3Þ
As seen above, quality of the project is quantitated with the numeral from 0 to 1.
144.3 Project Failure Risk Assessment
144.3.1 Cost-Time Joint Failure Risk Assessment Model
Assume simulation times k ¼ 1; 2; . . .N, C ð1Þ ; C ð2Þ ; . . .; CðNÞ and T ð1Þ ; T ð2Þ ; . . .; T ðNÞ
are the output results of the total cost C and time T from N times simulation. aij is
the frequency of results which fall into the shaded area in0. Then the joint
C1 ; Ci
probability distribution of the project in the region is as follow:
T1 ; Tj
Xi1 X
j1
F Ci ; Tj ¼ P C Ci \ T Tj ¼ agh N ð144:4Þ
g¼1 h¼1
Pi1 Pj1
g¼1 h¼1 agh is the cumulative frequency of shaded area in Fig. 144.2. So the
joint failure risk probability is as follow:
PCT ¼ PðCi ; Tj Þ ¼ PðC [ Ci [ T [ Tj Þ ¼ 1 FðCi ; Tj Þ ð144:5Þ
144.3.2 Project Failure Risk Value
In this paper, considering the three factors of cost, time and quality, the failure risk
value of the project is given by the following formula:
Ri ¼ PCTi Qi ð144:6Þ
Ri is the failure risk value from the ith time of simulation, PCTi is the joint failure
risk probability of cost and time from the ith time of simulation and Qi is the
quality value from the ith time of simulation.
1364 X. Pan and Z. Xin
Fig. 144.2 Joint frequency C
statistics schematic of cost
Cm+1
and time
Cm
C j+1
aij
Cj
C3 i −1 j −1
∑∑ agh
g =1 h =1
C2
C1
T1 T2 T3 Ti Ti+1 Tn Tn+1 T
144.3.3 Example of a Network Plan
(1) Example of Network Plan
This paper will combine a project network plan with nine activities (including a
virtual activities) to analyze. Figure 144.3 is the network plan for the project.
0 is the data of cost and time from the activities of network plan. Assuming that
the cost, duration and quality of the estimates from each activity are random
variables, which subject to the triangular probability distribution TRIAða; m; bÞ. A
is the most optimistic value; m is the most likely value; b is the most pessimistic
value (Table 144.1).
(2) Assessment of Project Failure Risk Value
Cost and time joint failure risk probability. The 2500 groups of output data from
project cost and schedule simulation can be got through 2500 independent repeated
simulation by software Arena.
Fig. 144.3 Example of a 2
network plan
1 3 7
4
144 Study of Cost-Time-Quality in Project Failure Risk Assessment 1365
Table 144.1 Network plan
Activity Cost of activity Time of activity
Good Mode Bad Good Mode Bad
Begin node – – – – – –
?` 5.1 6 8.4 2.55 3 4.2
?´ 3.15 4.5 5.4 1.4 2 2.4
?ˆ 8 10 12.5 4 5 6.25
`?˜ 7.65 9 11.7 4.25 5 6.5
` ?˜ 0 0 0 0 0 0
´?Þ 10.88 16 19.2 4.76 7 8.4
ˆ?Þ 13.87 19 25.65 5.84 8 10.8
˜?þ 11.68 16 21.6 5.84 8 10.8
Þ?þ 3 4 5.2 4.95 6.6 8.58
End node þ – –
0 is the results of data (Table 144.2).
The sample mean and standard deviation of cost and time can be got by
statistical analysis. Both the frequency distribution for normality goodness of fit
test, the test results show that the cost and time values follow a normal distribution.
0 is the frequency histogram of the results of cost and schedule outputs
(Fig. 144.4).
The joint failure risk probability table can be got through statistics and com-
puting, as shown in 0. Values in the table are the probability of project cost and
time failure risk (Table 144.3).
Analysis of quality factor. According to formula (144.7) and formula (144.8),
the dimensionless cost and time value are as follows:
Ci
Ci0 ¼ ; i ¼ 1; 2. . .n ð144:7Þ
lc
Table 144.2 Simulation results
Results Cost of activity Time of activity
Good Mode Bad Good Mode Bad
1? 2 5.107 6.486 8.377 2.554 3.247 4.19
1? 3 3.157 4.345 5.394 1.401 1.933 2.395
1? 4 8.005 10.157 12.477 4.009 5.076 6.242
2? 5 7.670 9.434 11.697 4.259 5.239 6.496
3? 5 0 0 0 0 0 0
3? 6 10.926 15.268 19.158 4.782 6.702 8.391
4? 6 13.935 19.212 25.584 0 8.132 10.745
5? 7 11.729 16.264 21.546 5.863 8.133 10.754
6? 7 3.013 4.050 5.185 4.968 6.661 8.5642
Sum 63.541 85.216 109.418 27.836 45.122 57.778
1366 X. Pan and Z. Xin
Fig. 144.4 The frequency histogram of the results of cost and time
Table 144.3 Joint failure risk probability
Total time value (M) Total cost value
76 77 … 87 … 92
12 1 1 … 0.9996 … 0.9996
13 1 1 … 0.9996 … 0.9992
: : : : : : :
19 1 1 … 0.866 … 0.842
20 1 1 … 0.5132 … 0.4308
: : : : : : :
23 1 0.9996 … 0.154 … 0.0012
24 1 0.9996 … 0.1536 … 0
Tj
Tj0 ¼ ; j ¼ 1; 2. . .n ð144:8Þ
lt
Ci and Tj are the cost and time value of the i and j groups of data. Ci0 and Tj0 are the
dimensionless data. According to formula (144.1) to (144.3), the quality value of
each group can be got. In this paper CO ¼ 0:90, TO ¼ 0:59, CA ¼ 1:07, TA ¼ 1:19.
Then we can get the ordinal position of each data after sorting the joint failure
risk probability values from small to large. According to the formula (144.9), the
location of the corresponding failure risk probability of the p-th confidence per-
centile value is determined, while the estimation of joint failure risk probability of
corresponding confidence percentile can be got.
h p i
k ¼ int 2500 ð144:9Þ
100
For example, the 95th of confidence percentile data is located on k = 2375
order and the cost is 870,508 yuan with the time of 20.7264 months. The joint
failure risk probability is 0.1568. 0 shows the 80th, 85th, 90th, 95th confidence
144 Study of Cost-Time-Quality in Project Failure Risk Assessment 1367
Table 144.4 Results of percentile estimates
Percentile estimates Failure risk probability Cost Time Quality
80th 0.39 85.02 20.11 0.70
85th 0.33 85.93 21.66 0.82
90th 0.23 86.72 20.81 0.76
95th 0.16 87.05 20.73 0.76
Table 144.5 Results of percentile estimates
Cost Time Quality C–T joint failure risk probability Project failure risk value
1 81.17 23.06 0.83 0.9676 0.8021
2 81.67 22.01 0.79 0.968 0.7687
3 81.17 21.90 0.79 0.9688 0.7581
4 78.22 21.70 0.74 0.9988 0.7384
5 80.58 21.24 0.74 0.9908 0.7321
6 83.00 22.06 0.82 0.8968 0.7285
7 81.95 21.19 0.75 0.9688 0.7259
8 81.98 21.16 0.75 0.9688 0.7245
9 81.80 21.00 0.74 0.9688 0.7131
10 82.77 21.76 0.80 0.898 0.7118
percentile of the joint failure risk probability of cost and time, as well as the
corresponding cost and time values (Table 144.4).
Project failure risk value. According to the formula (144.6), the project failure
risk value of 2500 groups of data can be calculated.
0 shows the statistics results of randomly selected 10 groups of data
(Table 144.5).
It can be seen from the above data that the quality factor, which has a direct
impact on the failure risk of a project, cannot be simply ignored.
144.4 Conclusion and Discussion
Study of ‘‘risk’’ or ‘‘failure’’ but not ‘‘success’’ is a common method today. Project
failure risk in engineering can be as low as possible to minimize losses by esti-
mation of failure risk value.
This paper establishes a quantitative method of quality factors and gets the
quality formula. Cost and time joint failure risk probability is estimated by Monte
Carlo Simulation, statistical analysis and confidence estimation methods. Finally
the project failure risk value can be calculated and analyzed by adding quality
value.
The improvement of this paper is given as follows.
1368 X. Pan and Z. Xin
The quantification of the quality factor. As we know quality factor in the actual
situation is very complex, whether which can be estimated by the formula is not
known. The author tries to find a better quantitative method for quality.
The project failure risk value of cost, time and quality is not the joint proba-
bility value. So the next step is to study the joint risk assessment methods for the
three factors of a project to better predict the actual situation.
Acknowledgments This work is supported by the National Natural Science Foundation of China
under Grants No. 70901004/71171008 and the Fundamental Research Funds for the Central
Universities.
References
Babu AJG, Suresh N (1996) Project management with time, cost, and quality considerations.
J Oper Res 88:320–327
Dubi A (1998) Analytical approach & Monte Carlo methods for realistic systems analysis. Math
Comput Simul 47:243–269
Dubi A (2000) Monte Carlo applications in system engineering. Wiley, Chichester
Gao XF, Hu CS, Zhong DH (2007) Study synthesis optimization of time-cost-quality in project
management. Syst Eng Theory Pract 10:112–117
Kim JY, Kang CW, Hwang IK (2012) A practical approach to project scheduling: considering the
potential quality loss cost in the time–cost tradeoff problem. Int J Proj Manage, Korea
30:264–272
Marseguerra M, Zio E (2000) System unavailability calculations in biased Monte Carlo
simulation: a possible pitfall. Ann Nucl Energy 27:1577–1588
Oisen RP (1971) Can project management be defined? Proj Manag Q 1:12–14
Xu Z, Wu JJ, Wang YQ (2006) Confidence percentile estimation to cost and schedule integration
based on Monte Carlo multiple simulation analysis technique. J Syst Simul 18:3334–3337
Xu Z, Wu JJ, Jia ZJ (2009) Estimation of risk joint probability to cost and schedule integration
based on joint probability distribution theory. J Syst Eng 24:46–53
Yang WM, Sheng YX (1990) System reliability digital simulation. BeiHang University Press,
Beijing
Chapter 145
Study on Innovation and Practice
of Independent Auto Companies Lean
Digital Factory Building
Yu-chun Wang, Li-fang Wang, Ze-yong Xu, Zhen-kui Li
and Feng-qi Wang
Abstract Lean Thinking extracted on the basis of the lean production is a theory
suitable for all industries, can prompt managers to rethink the business process, to
eliminate waste and to create value. It has entered the various fields of design,
manufacturing, logistics, procurement, sales and operations management so far.
The digital technology is a key technology to realize the knowledge-based,
automated, flexible enterprises and their rapid response of the market. It has now
achieved good economic benefits in optimize the design, fault diagnosis, intelli-
gent detection, system management, scheduling optimization, resource allocation
and other aspects in various industries. In this paper, lean digital factory solutions
is proposed, on the base of analyzing the problems of multi-production line, multi-
plant, multi-brand, short cycle, low-cost and so on in a domestic independent car
manufacturer, lean digital manufacturing framework model is built based on
applications of the information technology and digital technology, advantages of
lean digital manufacturing in creating value, improving resource utilization,
enhancing the competitiveness of enterprises are verified through case studies,
experience and effective measures of developing enterprises’ lean digital manu-
facturing are prompted ultimately.
Keywords Lean Virtual manufacturing Simulation modeling Structure
model
Y. Wang (&) L. Wang Z. Xu Z. Li F. Wang
FAW Car Co. Ltd., High and New Technology Development Zone, 4888 Weishan Road,
Changchun, 130012 Jilin Province, People’s Republic of China
e-mail: [email protected]
L. Wang
Department of Industry Engineering, Jilin University Nanling College, Changchun,
Jilin Province, People’s Republic of China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1369
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_145,
Ó Springer-Verlag Berlin Heidelberg 2013
1370 Y. Wang et al.
145.1 Purpose, Meaning and Research Status
The core of Lean Thinking is to reduce costs by completely rule out the waste. The
lean, that is, be concise and economic, do not invest the extra factors of produc-
tion, just at the right time to produce the necessary number of market need for
product (or the next process urgently needed products) but all business activities
must be beneficial and effective. Automotive manufacturing digital is a general
term of applying digital technology to factory design and building, product process
planning and actual manufacturing and management processes, increase manu-
facturing efficiency and product quality, reducing manufacturing costs, optimizing
design and improving manufacturing processes through information modeling,
simulation, analysis and information processing. It includes digital product defi-
nition (Product), digital process planning (Process), digital factory layout planning
(Plan), workshop production of digital management (Production) and digital
technology manufacturing resource (the Resource, including digital equipment
(CNC machining centers, robots, etc.), tools, tooling and operators). Factories
carry out a comprehensive digital activities are called Digital Factory. Factory lean
digital is applying lean thinking to digital factory construction and operation,
thereby creating products that meet user requirements: the right product, right
time, right price, the right to meet user requirements. Lean digital manufacturing
to achieve the overall upgrade in T, Q and C, T (Time) refers to the enterprises to
continuously adapt to the international advanced enterprise product development
speed to bring the fierce competition; Q (Quality) refers to quality improvements
of the whole process from drawings to physical vehicles. C (Cost) refers to
advancing late product design to the product development process, thus avoiding
late design changes, repeated transformation and production preparation caused
massive waste of the cost.
Research shows that digital technology has been widely used in many advanced
domestic and foreign enterprises so far. Many research institutes and enterprises
have adopted the digital factory program within different ranges like: CIM Institute
of Shanghai Jiaotong University applied EM-plant and Deneb virtual factory
simulation software in the technological transformation projects in the engine
factory production line of Shanghai Volkswagen Automotive Co., Ltd.; digital
factory platform is used as the digital base for aviation manufacturing enterprises
in Modern Design and Integrated Manufacturing Technology Laboratory of the
Northwestern Polytechnical University; In the aspect of production engineering
and manufacturing process management, Tecnomatix company, a world leader,
applied eMPower series of their software products including the industry’s leading
virtual factory simulation software eM-Plant (SIMPLE++) in factories of all sizes
(and even large-scale multinational corporations) and the production line model-
ing, simulation and optimization of production systems; products’ processes are
verified by EM-Assemble software in the simulation assembly process of the early
product development; U.S. Dassault has designed and built a car production line,
cut in half than in the past with traditional CAD technology cycle; GM has applied
145 Study on Innovation and Practice of Independent Auto Companies 1371
DENEB software to the luxury car factory assembly optimization design. These
digitization projects have very good economic benefits (Liu 2002; Li et al. 2008;
Zhai et al. 2004; Tecnomatrix Corporation Website (2013); http://www.longsea-
tech.com/eMPower.htm; Liao et al. 2004; Beit-On 2002). Many domestic enter-
prises on the understanding and application of digital still at primary stage, the
realization of digital manufacturing also need to invest a lot of time, personnel
funding and more scientific, comprehensive planning (Shao et al. 2000; Pi 2002).
145.2 Digital Manufacturing Problems Faced
by Autonomous Enterprises
FAW Car Co., Ltd. (hereinafter referred to as FAW Car) is one of the important
independent enterprises, in ‘‘the 12th development strategy’’ implementation
process, it is facing multiple challenges and pressures: (1) a number of production
line and factories, including local factories, remote factories and overseas factories
and so on; (2) Development and production of multi-brand (both co-branded and
own brand) vehicle models; (3) short cycle-market competition requires compa-
nies to achieve rapid product development and mass production; (4) low cost -
requires companies to fully identify the manufacturability of the product before the
manufacturing, to avoid design changes at the later stage. With the development of
the company, products are gradually updating and developing, market competition
is requiring the company to adjust the structure, change the mode of production,
change from manual to automatic, reduce design changes, respond to abnormal
situations and solve resource waste and project tardiness problems.
145.3 Framework Model of Lean Digital Factory
in Independent Auto Enterprises
In order to be able to survive in the intense global competition in the market and
development, FAW Car have identified the strategic objectives of building a
digital manufacturing system and have made the decision of changing original
extensive growth mode: in early production preparation, extensively use virtual
manufacturing software simulation and provide a reliable technology program
basis for late production preparation; extensively use information network tech-
nology to construct multi-functional information systems to provide tools for the
factory management and office automation, to ensure quality, schedule and cost
optimization of the production preparation and volume manufacturing. Established
a framework of lean digital factory model composed by one goal, two mains and
two basics (as shown in Fig. 145.1).
1372 Y. Wang et al.
Fig. 145.1 Framework model of lean digital manufacturing
The one goal is to build lean digital manufacturing system for the overall
objective to completely exclude the unreasonable waste; the two mains are man-
agement information system and virtual design and manufacturing system, they
can achieve and enhance Lean Thinking and industrial engineering, including: (1)
digital Lean design and manufacturing (include 1, virtual product design the
process is Product Design-CAE Engineering Analysis-CAE Process Analysis—
virtual product testing; 2, process virtual evaluation and simulation, includes
process simulation of stamping, welding robot simulation with the logistics sim-
ulation, painting and offline programming, assembly ergonomics and the logistics
simulation, matching tolerance allocation management simulation engine
machining and manufacturing simulation; etc.), (2) informational lean manage-
ment and manufacturing (the development of management information systems,
such as MES production control systems, ERP/DRP Enterprise Resource Man-
agement system, PDM collaborative development and management system, the
DMS/CRM dealer customer relationship management systems, production oper-
ations, knowledge management systems, collaboration and decision support
145 Study on Innovation and Practice of Independent Auto Companies 1373
systems, realized the informationalization, networking and platform) make lean,
IE and IT able to penetrate and throughout the corporate culture, management and
operational process. The two basics are Lean Production and Industrial Engi-
neering, they are the guiding ideology for the two main’s formation and
enhancement, they can be deeply applied in various fields to form the lean design,
lean manufacturing and lean management. The core of the industrial engineering is
the TQC, it is theory methods, tools and means of Scientific, rational planning and
management of manufacturing systems, the technologies usually used are the study
of methods, plant layout, knowledge management, simulation technology, digital,
group technology, information systems development, cost and profit analysis,
organizational studies, matching production, inventory control and analysis, the
planned network technology, value analysis, work measurement, objectives
management and so on.
145.4 Lean Digital Manufacturing Cases
To ensure effective establishment of lean digital manufacturing systems, FAW Car
has researched the latest technology in terms of enterprise information manage-
ment, product design, manufacturing technology, logistics management and
quality management at home and abroad the latest technology research, deeply
used CAE, database, network technology, information technology and virtual
simulation technology in various fields and has achieved remarkable results. There
are many successful cases in the product design, stamping, and welding technol-
ogy, and logistics planning (Fig. 145.2).
145.4.1 The Application Case of Information System
in Stamping Operation
The main production materials of mass production of stamping are the mold,
equipment and plates, production process embodies the salient features of the
downtime, causes and data volume. Since 2010, FAW Car stamping plant started
building the production and operation of information technology knowledge
management system model (Fig. 145.3) based on lean thinking, knowledge
management methods and production operations management experience accu-
mulated over the years. The model is divided into three levels: data management,
information management knowledge management, specifically divided into
building ‘‘working platform’’, forming ‘‘business experts’’ and ‘‘management
consultant’’. The stamping operation of knowledge management system infor-
mation platform including daily management of production, management of
production data statistical analysis, mold management, problem management,
1374 Y. Wang et al.
Fig. 145.2 Stamping plant production information system built schematic
Fig. 145.3 The first program plan of floor hand-line
production planning and inventory management and other information manage-
ment modules integrated the business and management processes in stamping
plants has been completed and used.
Since the use of information systems, mold downtime has decreased by 3.3 %
in 2010–2012, comprehensive and effective stroke has enhanced 0.4 times/min,
monthly reduced by an average waste 80.
In the process of development and use of the information platform, trained
workers in workshops for using the system 40 times, improved the management
process, optimized, cured the process more than 50 simultaneously, standardized
shutdown description, analysis, and measures more than 500 languages,
145 Study on Innovation and Practice of Independent Auto Companies 1375
accumulated mold problems, downtime analysis more than 1,000, determined the
core business of the stamping plant, provided a platform for business accumulation
and inheritance of the workshop, provided support for the workshop Lean Pro-
duction Management.
This case illustrates that using means of information in the manufacturing of the
product management process can effectively improve work efficiency, reduce
waste, the accumulation of the core business, improve the quality of staff to ensure
the high efficiency and low-cost of the production run.
145.4.2 The Application Case of Digital Manufacturing
in Welding Technology and Logistics Planning
FAW Car began to apply digital simulation technology to welding technology and
logistics planning, the software included ProcessDesigner (process planning)
module, ProcessSimulate (process simulation) module, Plant (Logistics Simula-
tion) module and RobCAD (simulation and offline programming) module. Plan-
ning 2 factory welding shop production capacity of 200,000/year, continuously
and randomly product more than four models. So, digitally simulate various
welding technology programs. The typical case is floor welding logistics simula-
tion and verification.
Process plan improved by technology personnel is shown in Fig. 145.3, A and
B are front and rear floor cache areas, the simulation showed that if the location
appliances are manually dragged to D area, workers’ work load is moderate
(balanced production of 50–70 %, the limit case in 70–85 %), the plan is rea-
sonable (Fig. 145.4).
This case illustrates that, if digital factory technology is applied to welding
technology and logistics design, the process and logistics design could be opti-
mized, labor efficiency could be improved and the waste of workshop’s reform in
production could be reduced.
Fig. 145.4 The developed program plan of floor hand-line
1376 Y. Wang et al.
145.5 Effective Way to Construction and Innovation
of Lean Digital Manufacturing System
145.5.1 Lean Thinking is the Basis of Lean Digital
Manufacturing System
FAW Car’s rapid development process, first of all, is the process of learning and
developing lean thinking, it established a philosophy of ‘‘to be correct at the first
time’’, pursued ‘‘zero defect’’, took ‘‘customers first’’ as the origin and manufactured
first-class products. Companies to make a profit, must have a long-term strategic
vision, focus on investing in new technology and talent training, thereby, could
reduce design, manufacturing defects and thoroughly eliminate wasteful aspects,
guarantee mutual benefit among businesses, employees and partners. FAW Car takes
lean thinking and digital factory building as part of the strategy ‘‘to create one million
international passenger cars business units’’, enhances independent innovation and
system capacity, shifts from extensive management to fine management, shifts from
‘‘Fuzzy management’’ and ‘‘chaos management’’ into a ‘‘precision management’’.
145.5.2 Integrated Innovation is the Only Way to Achieve
Lean Digital Manufacturing
FAW Car’s principle is ‘‘self-development, open and cooperative’’, Began to build
self-management mode—Red production system (that is, of HPS, the Hongqi
Production System) from 2007, lean thinking was extended from the production
management to all areas of product development, quality control, procurement,
technology, logistics, and production organization, therefore, all parts played a
synergistic effect, the brand premium and the ability to support the one million-
scale system are enhanced continuously. FAW Car belongs to the main body of the
international auto market, only to constantly summarize and analyze its own
deficiencies and problems, capacity and conditions it should have, speed and
objective to fill the gaps, to unify ideas, to open minds and to strive to build lean
digital factory, in an open competitive environment, to develop self-development
on the road, can it achieve survival, be solid, stronger and bigger.
145.5.3 Step Implementation is the Effective Guarantee
of Realizing Lean Digital Manufacturing
The FAW Car did not have a digital plant technology capability before and could
only rely on products partners the MAZDA to complete large-scale production line
145 Study on Innovation and Practice of Independent Auto Companies 1377
process planning (such as welding and assembling the M1, M2 production line
planning), controlled technology resulted in very high manufacturing costs. ‘‘Inde-
pendent’’ determine the characteristics of the FAW Car can not completely copy the
digital factory technology of advanced foreign enterprises, therefore, the ‘‘Digital
manufacturing’’ could only be divided into different areas, different stages and dif-
ferent degrees to be planned, implemented, and ultimately to achieve the general-
ization and integration of various fields of information transmission, use and
management of the manufacturing system. The TECNOMATIX/eMPower body
planning systems and logistics planning systems were introduced, completed mod-
eling of the ‘‘digital manufacturing’’, establishment of work ideas, technology library
architecture design, repository architecture design and other preliminary works in the
first stage; The second stage was the ‘‘digital manufacturing’’ island style applica-
tions, gradually improved the ‘‘digital manufacturing’’ system functions and pro-
cesses to create the conditions and lay the foundation for realizing the following
‘‘digital’’ planned projects. These two stages have been completed. In the third stage,
data management platform will be unified, the digital factory software applications
proficiency, through technology of island and network technology in all areas will be
strengthened, accordingly, will the interfaces and sharing will be realized.
145.6 Conclusion and Prospect
In this paper, the importance of FAW Car Digital Manufacturing System is illus-
trated on the basis of the model and cases, useful lessons are provided for other auto
manufacturers: (1) lean ideas, information and digital technology is an important
means for the auto enterprises to promote product updates, develop production and
improve the international competitiveness; (2) lean digital manufacturing system is
the effective guarantee to ensure multiple production lines, multi-plant, multi-
brand, short cycle and low-cost production; (3) if independent auto enterprises carry
out comprehensive lean digital factories, it will continuously promote the enterprise
economic growth mode to transfer from extensive and technology-introduction type
to intensive and innovation type; (4) enable enterprises to achieve 4 development
through lean digital manufacturing system: efficiency is continuously increased,
cost is continuously reduced; quality is continuously improved; ability is contin-
uously increased; the system’s core ability is enhanced, the core competitiveness of
enterprises is formed, its own excellent is built.
References
Beit-On H (2002) Delmia-tecnomatix—the duel for the digital factory. Promising Market, 5
http://www.longsea-tech.com/eMPower.htm
Li S, Yang T, Chen B et al (2008) Digital factory technology and application in aeronautical
product R&D. Aeronaut Manufact Technol 19:46–49
1378 Y. Wang et al.
Liao F, Zhang L, Xiao T et al (2004) An interactive assembly process planner. Tsinghua Sci
Technol 9(2):219–226
Liu H (2002) Study on equipment layout of multi-production line plant (Chinese). Shanghai Jiao
Tong University, vol 2
Pi X-Z (2002) Research and application of assembly line balancing and simulation technology
(Chinese). Shanghai Jiao Tong University, vol 1
Shao L, Yan J-Q, Ma D-Z, Zhong T-X et al (2000) Virtual integrated design of production line.
Institute of Computer Integrated Manufacturing, Shanghai Jiao Tong University, vol 6
Tecnomatix Corporation website (2013) Available at www.tecnomatix.com
Zhai W-B, Chu X-N, Ma D-Z, Jin Y, Yan J-Q et al (2004) Planning process modeling of a virtual
factory. Institute of Computer Integrated Manufacturing, Shanghai Jiao Tong University, vol
38(6), pp 862–865
Chapter 146
The Application of Kernel Estimation
in Analysis of Crime Hot Spots
Yan-yan Wang, Zhi-hong Sun, Lu Pan, Ting Wang and Da-hu Zhang
Abstract In order to analyze crime hot spots, we use Kernel estimation. The
choice of Kernel function and Band-width is critical in kernel density estimation,
which decides the accuracy of the estimation. We choose Gauss kernel and further
obtain the optimal Band-width in the sense of square error MISE. Using Kernel
estimation, not only can we calculate the density of crime in the region, but also
accurately show the areas with the relative high-crime density and get the maxi-
mum point according to the information about the previous criminal spots. Last we
use Kernel estimation to predict Peter Sutcliffe ‘‘the Yorkshire Ripper’’ 11th
criminal location based on the previous criminal locations in the Serial murders.
Finally we can get the range of the criminal hot zone: Longitude:
53.6875–53.8125 N; Altitude: 1.775–1.815 W. In fact, the coordinate of Peter’s
11th criminal location is (53.817 N, 1.784 W). From this, it can be seen that our
estimation is relatively accurate.
Keywords Band-width Crime hot spots Kernel estimation Kernel function
MISE
146.1 Introduction
In space, the phenomenon of high concentration of crime naturally will be related
to the expression of crime hot spots in the map (Rossmo 2000). In the process of
crime mapping, the text address of crime hot spots by geo-coding can be changed
Y. Wang (&) Z. Sun L. Pan T. Wang
Department of Foundation, Air Force Service College, Xuzhou, Jiangsu, China
e-mail: [email protected]
D. Zhang
Department of Aviation Oil, Air Force Service College, Xuzhou, Jiangsu, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1379
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_146,
Ó Springer-Verlag Berlin Heidelberg 2013
1380 Y. Wang et al.
into the coordinate point. Criminologists observe the spatial distribution of crime,
such as gregariously, regularly, discretely or randomly. Through the analysis of
crime hot spots, we can look for the gather place of Point group and the identifiable
range, further understanding the formation reasons and its possible impact. The
commonly analysis methods in the crime hotspots are Grid-counting Method,
nearest neighbor distance Method and Moran’s I way. The advantage of Kernel
estimation method is to regard crime space as a core site. Not only can we cal-
culate the density of crime in the region, but also accurately show the areas with
the relative high-crime density.
146.2 Model
Rosenblatt and Parzen proposed a very important estimation—Kernel Density
Estimation, which is a kind of non-parametric density estimations (Chen 1989).
The characteristic is that there is not a definite function form and we cannot
calculate density function without parameter. In the K-means, a kernel function
(probability density function) (Venables and Ripley 2002), used to indicate the
spot distribution in the neighborhood, is set at each data point. The density
function of the point in the spot can be regarded as the total contributions of known
point’s kernel density, which is made by all the given points to the point (Lai
1996). Therefore, for any point X, the contribution law made by the given point Xi
not only depends on the distance from X to Xi but also on the shape and the span of
the kernel density (Band-width) (Worton 1989).
Definition 1 Suppose X1; X2 ; ; Xn are the samples of X, which are subject to
f ðxÞ. Let us define the function as
^ 1 X n
x xi
f ðxÞ ¼ K ð146:1Þ
hn nhn i¼1 hn
It is called K-means of the density function f ðxÞ.
Kð:Þ Kernel function is a preset probability density function, in which the Band-
width is hn and n is sample size.
146.2.1 Selecting Kernel Function
In estimating the density function about X, let us suppose each sample, just like a
small irradiative light bulb, serves X, the function of which is related to the distance
in a certain sense (situation) from sample X1 to Xn , i.e., the farther the distance, the
weaker the light intensity. When it is taken into the consideration, the kernel
function should be chosen according to the distance from X to each sample Xi ,
146 The Application of Kernel Estimation 1381
which decreases with the growing distance. The kernel function KðxÞ is insensible
to ^fhn ðxÞ, therefore each kernel function with satisfying conditions
R is suitable.
Kernel function is symmetrical about the origin and satisfies kðlÞdl ¼ 1: Ep-
anechnikov kernel, Bisquare kernel and Gauss kernel are in common use (Silver-
man 1986). We might as well suppose
2
1 x
K ð xÞ ¼ pffiffiffiffiffiffi exp ð146:2Þ
2p 2
P
n h i
ðxxi Þ2
Then ^fhn ðxÞ ¼ h1n n p1ffiffiffiffi
2p
exp 2h 2 .
n
i¼1
146.2.2 Selecting Band-Width
The choice of Band-width is critical in kernel density estimation, which decides
the slipperiness of density shape, i.e. it has an effect on the accuracy of the
estimation.
Generally, the coefficient of the Band-width hn diminishes with the augment of
the sample size n. When the kernel function KðxÞ is fixed, the generated shape
intends to be smooth only to cover up the density structure if the Band-width is
oversize. And the generated shape intends to be sharp which cannot indicate the
internal regularity. In addition, the density should be considered in deciding hn . In
data compact district, hn tends to be small and in data rare faction, hn to be large.
Only after many experimental investigations can the Band-width be determined.
As the density is continuous theoretically, we usually obtain the optimal Band-
width in the sense of square error MISE (Sheather and Jones 1991).
Z þ1
MISE = E ffn ðxÞ f ðxÞg2 dx
1
Z þ1 2 Z þ1 Z þ1 ð146:3Þ
1 1
l2 kðlÞdl ff 00 ðxÞg2 dxh4n þ k2 ðlÞdl
4 1 1 1 nhn
We adjust the parameters to minimize MISE and get Band-width hn estimation
" R þ1 #1=5
k2 ðlÞdl
1
hn ¼ R þ1 R þ1 :
n 1 ff 00 ðxÞg dx f 1 l2 kðlÞdlg2
2
We use sample standard deviation Sn to replace unknown parameter r in order
to get normal reference Band-width. Then substituting ^fhn ðxÞ for f ðxÞ, we obtain
1
hn ¼ 1:06Sn n5 ð146:4Þ
Finally we can get the kernel function
1382 Y. Wang et al.
" #
^ 1 Xn
1 ðx xi Þ2
f ð xÞ ¼ 4
pffiffiffiffiffiffi exp 2 ð146:5Þ
hn 1:06 Sn n5 i¼1 2p 1:1236 S2n n5
Using this function, we can get the maximum point Xnþ1 according to fXi gni¼1
the information about the previous criminal spots. And spot represented by the
maximum point is just the most probable criminal location.
146.3 The Application of Kernel Estimation
Serial murders have serious social impact and make people in great horror. So it is
extremely critical to predict the offender’s hideout by the offender’s crimes in
solving the case.
Peter Sutcliffe was born in Bingley, West Riding of Yorkshire on 2nd June in
1946. He was a ferocious serial killer and committed over 20 crimes just within
6 years, including 13 murders and a series of vicious attacks. He was nicknamed
‘‘the Yorkshire Ripper’’ because of his vicious criminal means.
The victims’ information is listed below (Table 146.1).
We can predict Peter’s eleventh criminal location through the Eq. (146.5) based
on the previous criminal locations. And we get the coordinate of the 11th location
(53.7975 N, 1.5652 W). Actually, we can get the range of the criminal hot zone:
Longitude: 53.7620–53.8125 N
Altitude: 1.5124–1.5876 W
In fact, the coordinate of Peter’s 11th criminal location is (53.7997 N,
1.54917 W). From this, it can be seen that our estimation is relatively accurate
(Fig. 146.1).
Table 146.1 The victims’ information
Data Name Location
Latitude Longitude
5 July 1975 Anna Rogulskyj 53.867 -1.911
August 1975 Olive Smelt 53.725 -1.863
27 August 1975 Tracy Browne 53.914 -1.937
30 October 1975 Wilma McCann 53.816 -1.531
January 1976 Emily Jackson 53.808333 -1.53333
9 May 1976 Marcella Claxton 53.841667 -1.4925
5 February 1977 Irene Richardson 53.841667 -1.4925
23 April 1977 Bradford 53.8 -1.75206
26 June 1977 Jayne MacDonald 53.64932 -2.43999
1 October 1977 Jean Jordan 53.4286 -2.2582
14 December 1977 Marilyn Moore 53.799722 -1.54917
Source http://en.wikipedia.org/wiki/Peter_Sutcliffe
146 The Application of Kernel Estimation 1383
Fig. 146.1 Peter’s predicted next possible criminal spot based on Kernel Density Estimation
146.4 The Application of Kernel Estimation
Grid-Counting Method
We take the crimes which Peter committed in Bingley (West Riding of Yorkshire),
his former hideout, from 1975 to 1977 as example.
Peter is one of the Marauder offenders. His criminal spots, centered on his
stable hideout, were scattered around. Also it is found that his criminal spots were
not regularly distributed but focused on some area.
From Fig. 146.2, we find that more crimes were committed in the elliptic
region, up to 5 (2 crimes committed in one certain point), which coincide with the
criminologist David Canter’s opinion that the criminal chooses his familiar loca-
tion to commit the crimes. The offender repeated committing his crimes in the
small area, which reflects that he desired to gain control over the criminal spots.
Once he succeeded in committing, the offender would become confident in his
scheme and repeat his crime at the same spot (Becker et al. 1988). So, more police
force should be laid out in this area.
146.5 Further Research Trend
The police agency is much interested in finding out the next criminal spot of the
offender in serial criminal cases.
Given a series of crimes at the locations X1 ; X2 ; . . .; Xn committed by a single
serial offender, we are to estimate the probability P Xnext jX1; X2 ; . . .Xn of the next
1384 Y. Wang et al.
Fig. 146.2 Peter’s clustered criminal spots
criminal spot Xnext . Based on Bayesian Model from Mike O’Leave (Levine and
Block 2011). Once again, we can use Bayes’ Theorem get the expression formula
P Xnext jX1; X2 ; . . .Xn
ZZZ
/ PðXnext jz; aÞPðX1 jz; aÞPðX2 jz; aÞðXn jz; aÞHðzÞpðaÞdzð1Þ dzð2Þ da:
In reality, it is not easy to estimate the value of HðzÞand the value of pðaÞ.
Moreover, even the estimation values of HðzÞand pðaÞhas been given, it is still not
so easy for us to get the result of the triple integral (Scott 1992).
In solving the problem, we aim to discretize the continuous process and then
use the numerical method to get the probability P Xnext jX1; X2 ; . . .Xn .
146.6 Conclusions
Through the grid mesh and refinement of the area, we set up an optimization model
and Bayesian Model to construct ‘‘Geographic profiling’’ pointing to search the
offender’s hideout.
146 The Application of Kernel Estimation 1385
In Bayesian Model, by using ‘‘maximum likelihood estimation method’’ to
estimate parameter Z, we can obtain the longitude and altitude of Peter’s hideout
(53.8063 N,1.77044 W), which is also close to his actual hideout.
We can use the internal relation among the criminal spots in serial cases to
predict the offender’s next probable criminal location after predicting his hideout.
First, we can use the Bayesian Model to predict the criminal probability in any
spot X and further PðX jZ; aÞ ¼ 4a12 EXP ð 4ap2 jX Z j2 Þto identify the function on
distance and criminal frequency as negative exponential decay function. This
function reflects the characteristic, i.e., the farther the spot is away from the
offender’s hideout, the smaller the criminal probability is.
The prediction obtained by using the Bayesian Model, a conceptual model, is
inaccurate. Compared with the former, the prediction obtained by using the Kernel
Density Estimation Model is more accurate. Furthermore, the model is established
on the actual criminal spots which can induce higher accuracy. In Peter’s cases, the
coordinate of the predicted criminal spot is (53.7375 N, 1.795 W) and the coor-
dinate of the actual criminal spot is (53.817 N, 1.784 W). From this, it can be seen
that the prediction obtained by using the Kernel Density Estimation Model is more
accurate and credible.
References
Becker RA, Chambers JM, Wilks AR (1988) The new S language. Wadsworth & Brooks/Cole
(for S version), Pacific Grove
Chen X (1989) Non-parametric statistics. Science and Technology Press, Shanghai, pp 283–296
Lai Z (1996) Residential burglary crime Map and crime location analysis in Taipei. Resource and
Environment Institute of Geography, University of Taipei, Taipei
Levine N, Block R (2011) Bayesian journey to crime estimation: an improvement in geographic
profiling methodology. Prof Geogr 63(2):9
Rossmo DK (2000) Geographic profiling. CRC, New York
Scott DW (1992) Multivariate density estimation. Theory, practice and visualization. Wiley,
New York
Sheather SJ, Jones MC (1991) A reliable data-based bandwidth selection method for kernel
density estimation. J Roy Stat Soc B 53(3):683–690
Silverman BW (1986) Density estimation. Chapman and Hall, London
Venables WN, Ripley BD (2002) Modern applied statistics with S. Springer, New York
Worton B (1989) Kernel methods for estimating the utilization distribution in home-range
studies. Ecology 70:164–168
Chapter 147
The Research of Full-Bridge and Current
Double Rectifier Switched-Mode Power
Supply for Vehicle
Yi-lin Yin, Rui-peng Chen and Xi-ming Cheng
Abstract The switched-mode power supply (SMPS) has many advantages, for
instance high efficiency of transformation, small volume and etc. As the most
directive and effective way to decrease the size of the switching converter, high
frequency, however, increases the switching wastage. Therefore, the soft-switch-
ing technology has been invented and developed to reduce switching wastage. This
paper presents the high frequency of full-bridge switching converter using zero-
voltage switching technology and current double rectifier technology. The steady
state model and small signal model are built by PWM switching technology.
Peaking-current control mode is adopted as the control strategy. The simulation
circuit of full-bridge and current double rectifier switch power is designed on
PSIM platform, and the simulation is finished. The simulation proves that the
model is right and the control strategy is effective.
Keywords The switched-mode power supply Current double rectifier ZVS
soft-switching PSIM
147.1 Introduction
Electric vehicles have become an important topic throughout the automotive
industry and are developing at a high speed. As the main part of DC switching
power supply, DC/DC SMPS will be widely applied in electric vehicles.
Y. Yin (&) R. Chen X. Cheng
School of Mechanical Engineering, Beijing Institute of Technology, Beijing, China
e-mail: [email protected]
R. Chen
e-mail: [email protected]
X. Cheng
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1387
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_147,
Ó Springer-Verlag Berlin Heidelberg 2013
1388 Y. Yin et al.
Fig. 147.1 The structure of
the full-bridge and current
double rectifier SMPS
Nowadays the switching power supply development tendency is: high efficiency,
low loss, miniaturization, integration, intellectualization, redundant reliability
(Middlebrook and Cuk 1977; Hua and Lee 1993). In order to reduce switch losses,
noises and improve power density, soft-switching technology under the principle
of zero-current switching (ZCS) and zero-voltage switching (ZVS) is widely used
in many applications (Liu et al. 1987; Theron and Ferreira 1995; Canesin and
Barbi 1997; Dudrik et al. 2006; Liu and Lee 1990).
The SMPS consists of three basic topological structures: buck converter, boost
converter and buck-boost converter. In this paper, the full-bridge and current
double rectifier SMPS is designed on the basis of buck converter and adopts
current double rectifier in the secondary side of pulse transformer. The structure of
the full-bridge and current double rectifier SMPS is shown in Fig. 147.1.
147.1.1 Modeling of the SMPS
The models of SMPS are built for steady state analysis, transient analysis and
designing of SMPS. There are many modeling methodologies of SMPS, including
method of state-space averaging, the model of PWM switch and so on (Hua and
Lee 1995; Smith and Smedley 1997; Chen et al. 1991).
As in the case of continuous conduction mode (CCM), the model of the PWM
switch in DCM represents the dc and small-signal characteristics of the nonlinear
part of the converter which consists of the active and passive switch pair (Vor-
perian 1990). The dc and small-signal characteristics of a PWM converter are then
obtained by replacing the PWM switch with its equivalent circuit model in a
manner similar to obtaining the small-signal characteristics of linear amplifiers
whereby the transistor is replaced by its equivalent circuit model.
Thus this paper adopts the model of PWM switch to build the steady state
model and small signal model of full-bridge and current double rectifier SMPS.
The simplified equivalent circuit structure by the model of PWM switch is shown
in Fig. 147.2.
147 The Research of Full-Bridge and Current Double Rectifier 1389
Fig. 147.2 The simplified
equivalent circuit structure A C L 2 vo
NS 1 d ESR
⋅ vin R
+
NP Co
−
P
The model of simplified equivalent circuit is obtained as
8
NS L diL
>
< NP d vin ¼ 2 dt þ vo
vo ¼ vC þ ESR Co dvdtC ð147:1Þ
>
: i ¼ C dvC þ vo
L o dt R
After adding disturbance, the model of steady state can be derived as
8N
< NPS D Vin ¼ Vo
V ¼V ð147:2Þ
: o Vo C
IL ¼ R
The small signal model in complex frequency domain is needed to design the
compensation network and the model will be different on the basis of different
control mode. This paper adopts the peak-value current control mode which
includes voltage and current dual close-loop feedback system. In the peak-value
current control mode, the circuit includes current detecting circuit and ramp
compensation circuit which are shown in Fig. 147.3
The ramp compensation voltage vCT is supplied by oscillation circuit and ve is
supplied by feedback compensate network.
Fig. 147.3 current detecting 1: N
circuit and ramp
compensation circuit v in KA KC
+ • • •
− Cin
KB K D
+
Ri
−
Ri 2
Ri 2 0.8 V
Ri1
−
v CT vP PWM d
+ Ri 1 Comparator
+
− ve
Ri 3 Ri4
1390 Y. Yin et al.
(a) Np ⋅N
Ramp − Sa ⋅
N p⋅ N Compensation N s ⋅ Ri
(ve − 0.8) ⋅
N s ⋅ Ri
iLP1
iL1
iLV 1
s d d ⋅ Tsw s1−d Tsw
(b) Np ⋅N
N p⋅ N
Ramp − Sa ⋅
(ve − 0.8) ⋅
Compensation N s ⋅ Ri
Ns ⋅ R i
iLP 2
iL 2
iLV 2 Tsw 2 Tsw
s1− d sd
Fig. 147.4 The current waveform of two filter inductors. a Current waveform of L1, b current
waveform of L2
The current waveform of two filter inductors in the circuit is shown in Fig. 147.4.
The upward slope of current of filter inductor is sd and the downward slope is
s1d . Therefore, the transfer function of the peak-value current control mode can
be derived from Fig. 147.3
R 1
H 1 ð SÞ ¼ Fp ð SÞ F h ð SÞ ð147:3Þ
R0i 1 þ RTLsw0 :p Qp
where 8
> 1 þ xSZ1
>
> Fp ð SÞ ¼
>
> 1 þ xSP1
>
>
>
>
>
> 1
>
>
> Fh ð SÞ ¼
> 2
>
> 1 þ xn Qp þ xS 2
S
>
> n
>
> N
< R0 ¼ s
Ri
i
2Np N ð147:4Þ
>
>
>
> Ns
>
> 0 Np Vin Vo
>
> Sd ¼
>
> L
>
> 2
>
>
>
> 1
>
> Qp ¼
>
> Sa =R0i
>
: p ð1 DÞ 1 þ S0 1
2
d
147 The Research of Full-Bridge and Current Double Rectifier 1391
Table 147.1 The technical Technical indicator Value
indicators of SMPS for
vehicle Input voltage (V) 60–90
Output voltage (V) 14
Output ripple voltage (V) B100
Output power range (W) 100–800
Transformer turns ratio 3:2
Switching frequency (kHz) 50
Minimal output power in CCM (W) 200
147.2 Designing of the SMPS
The first step of designing SMPS, including power circuit and control circuit, is to
choose the right structure of SMPS, control method and related SMPS technology
according to the technical indicators of SMPS. The technical indicators of this
paper are illustrated in Table 147.1
147.2.1 Designing of Power Circuit
The designing of Power circuit of SMPS based on the technical indicators mainly
includes the choosing of output capacitors and filter inductors, designing of
parameters of soft-switching.
Essentially, full-bridge and current double rectifier SMPS is the derivation of
BUCK converter. The minimum of filter inductors Lm must be obtained, using the
model of steady state in (147.2)
Vo N p
Lm ¼ RM Tsw 1 ð147:5Þ
VinM Ns
where VinM stands for maximum input voltage of SMPS and RM stands for load
impedance of SMPS in discontinuous current mode (DCM) and can be derived as
follows:
Vo2
RM ¼ ð147:6Þ
Pm
where Pm is the minimum output power in CCM.
Finally, inductor with inductance L = 15 lH is selected as the filter inductor.
The key parameter of choosing output capacitor is equivalent series resistance
(ESR) rather than capacitance, because ESR of output capacitor has a much bigger
impact on the ripple of output voltage than the capacitance. Thus, the ESR of
output capacitor is obtained firstly according to the ripple of output voltage and the
1392 Y. Yin et al.
output capacitor is selected according to obtained ESR and manufacturer’s data
specification sheets.
The capacitance Co can be ignored only when Co satisfies the following
relationship:
1
Co ð147:7Þ
fc ESR
where fc is the operating frequency of output capacitor.
The ripple of output voltage cannot exceed the maximum value DVM when the
ESR satisfies the following relationship:
DVM
ESR ð147:8Þ
DICM
where DICM is the maximum ripple voltage through the output capacitor Co .
The fc and DICM can be derived as follows:
(
fc ¼ T2sw
ð147:9Þ
N
DICM ¼ VLo Tsw 1 V2VinMo Nps
Combining Eqs. (147.7), (147.8), and (147.9), yields
8 DVM LVinM Ns
< ESR
Vo Tsw ðVinM Ns 2Vo Np Þ
ð147:10Þ
: C Vo Tsw2 ðVinM Ns 2Vo Np Þ
o 2DVM LVinM Ns
The initial value of ESR and Co can be obtained, using expression (147.10) and
Table 147.1.
ESR 40:1 mX Co 63 lF
When choosing capacitors, it is necessary to take into consideration the loss
factor tan d which is provided in manufacturer’s data specification sheets because
the ESR of capacitor decreases with the increasing of capacitance.
tan d ¼ 2pfc C ESR ð147:11Þ
Thus, two capacitors with withstand voltage 63 V and capacitance 1500 lF are
selected as the output capacitor.
147.2.2 Designing of Soft-Switching
Soft-switching technology can effectively improve work environment of
power converter and greatly decrease its power consumption (Hua et al. 1994;
Cho et al. 1994; Jiang et al. 2003; Schutten and Torrey 2003). Soft switching
includes ZVS and ZCS. Since using MOSFET as the switch, this paper adopts ZVS.
147 The Research of Full-Bridge and Current Double Rectifier 1393
The power circuit of full-bridge and current double rectifier SMPS using
MOSFET is shown in Fig. 147.5.
When the voltage drop of diode is ignored, the total time tCD of charging or
discharging of the capacitor CCD in the leading leg can be derived as follows:
vin
tCD 2CCD ð147:12Þ
ipk
where iPK is the peak current though the primary winding of transformer. To make
sure the soft switching of leading leg is achieved, tCD must be less than com-
mutation dead time tdeadCD .
RDELCD
tCD tdeadCD ¼ 5 1011 ð147:13Þ
1:5ðvCS vADS Þ þ 1
During switching of lagging leg, the transformer is not impacted but only the
resonant inductance Lr is impacted. Therefore, to make sure the capacitor CAB in
the lagging leg is fully charged or discharged, the resonant inductance Lr must
satisfies the following expression.
8
< Lr 2CAB2 v2in
iAB
ð147:14Þ
: iAB ipk Ns ð0:5dÞTsw vo
Np L
The total time tAB of charging or discharging of the capacitor CAB in the lagging
leg can be derived as follows:
rffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2CAB vin
tAB ¼ 2Lr CAB arcsin ð147:15Þ
Lr iAB
L1
KD KB
N p : Ns Co ESR
D2 − +
+ Lr R
vin −
C in vo
D1
KC KA
L2
Fig. 147.5 The power circuit of SMPS
1394 Y. Yin et al.
To make sure the soft switching of lagging leg is achieved, tAB and tdeadCD must
satisfies the following expression.
tAB tdeadAB tAB þ tr
RDELAB
tdeadAB ¼ 5 1011 1:5ðvCS ð147:16Þ
vADS Þþ1
147.2.3 Designing of Control Circuit
The designing of control circuit mainly includes type selection of control chip and
designing of its outside circuits. UCC3895 is selected as the control ship and the
peak-value current control mode is selected as the control mode. The control
circuit is shown in Fig. 147.6.
The feedback compensate network can be designed as
8
> Sa =R0i
> 1 1 Tsw 1
>
> R1 C2 ¼ RCo þ L0 Co ð1 DÞ 1 þ S0d 2
>
>
>
< R4 C4 ¼ ESR Co
xcross ¼ RCTRR 4
1 þ R5
H1 ð0Þ ð147:17Þ
>
> R C R
>
>
1 3 2 6
>
> H1 ð0Þ ¼ RR0 h 1 i
>
: i Sa = R 0
1þRTsw ð1DÞ 1þ
L0 S0 2
i
1
d
147.3 Simulation Results
The whole simulation circuit construction of the designed full-bridge and current
double rectifier SMPS is shown in Fig. 147.7.
The circuit is designed with power outputs ranging from 100 to 800 W. The
designed SMPS is simulated in PSIM and the simulated waveforms are illustrated
in Figs. 147.8, 147.9, 147.10 and 147.11.
When the output power is 100 W, the crossover frequency fcross of the open
loop transfer function and the dominant pole of slope compensation network are
set about 63 kHz. Thus, when the output power is below 100 W, the phase margin
will decline and even become negative because the crossover frequency fcross
moves to higher frequency, which will lead to system instability, as is shown in
Fig. 147.8. The current sensor is designed when the output power is 800 W. If the
output power exceeds 800 W, current limiter and protecting circuit will start to
work and then the output power will be limited by limiting input peak current and
the duty ratio of controller, so the output voltage will be below 14 V, as can be
seen in Fig. 147.11
147 The Research of Full-Bridge and Current Double Rectifier 1395
vi
+ R5 R6
− Output
R i2 R i2 vo
R3
Optocoupler
R4
R i1 EAN EAP
EAOUT SS
R bias
R i3 RAMP OUTA R1
U C SS
REF C OUTB
R i4 R i1 C 2
− GND C PGND
C
+ SYNC 3 VCC 4
CT 8 TL431
Slope CT 9 OUTC
Compensation RT
RT 5 OUTD R2
R DELAB
DELAB CS R ADS 1
R DELCD
DELCD ADS
11. 7. 25 R ADS 2
Fig. 147.6 The control circuit
Fig. 147.7 The whole simulation circuit construction
1396 Y. Yin et al.
Fig. 147.8 Po = 70 W, R = 2.8 X
Fig. 147.9 Po = 100 W, R = 1.96 X
The ripple of output voltage is mainly determined by the ESR of output
capacitor Co .
Figure 147.12 shows the ripple of output voltage. The ripple of output voltage
is always below 100 mV in a period of input voltage vin , so the selection of output
147 The Research of Full-Bridge and Current Double Rectifier 1397
Fig. 147.10 Po = 800 W, R = 0.254 X
Fig. 147.11 Po = 852 W, R = 0.23 X
capacitor Co is right. The ripple of output voltage is the biggest under the maxi-
mum input voltage, as is shown in Fig. 147.12.
1398 Y. Yin et al.
Fig. 147.12 The ripple of output voltage
147.4 Conclusion
In this paper, a kind of full-bridge and current double rectifier SMPS for vehicle is
analyzed and modeled. Based on the models, the circuit of the SMPS is designed
and soft-switching technology is also used to reduce the switching losses. By
simulation in PSIM, the models and designed circuits are validated. The result
shows that the models have very high accuracy and the designed circuit is right.
References
Canesin CA, Barbi I (1997) Novel zero-current-switching PWM converters. IEEE Trans Ind
Electron 44(3):372–381
Chen Q, Lofti AW, Lee FC (1991) Design trade-offs in 5 V output off-line zero-voltage PWM,
telecommunications energy international conference—INTELEC. Proceedings VPEC,
pp 89–99
Cho JG, Sabate JA, Lee FC (1994) Zero voltage and zero current switching full-bridge PWM,
power electronics and ECCE Asia (ICPE & ECCE), IEEE 8th international conference on
converter for high power applications. IEEE PESC Rec, pp 1585–1590
Dudrik J, Spanik P, Trip ND (2006) Zero-voltage and zero-current switching full-bridge DC–DC
converter with auxiliary transformer. IEEE Trans Power Electron 21(5):1328–1335
Hua G, Lee FC (1993) Soft-switching PWM techniques and their applications. IET Power
Electron Appl Euro Conf 3:87–92
Hua GC, Lee FC (1995) Soft-switching techniques in PWM converters. IEEE Trans Ind Electron
42(6):595–603
Hua G, Leu CS, Jiang Y, Lee FC (1994) Novel zero voltage switching PWM converters. IEEE
Trans Power Electron 9(2):213–219
147 The Research of Full-Bridge and Current Double Rectifier 1399
Jiang JC, Zhang WG, Shen B (2003) Analysis and design of a novel ZCT-PWM converter. IEEE
Trans Power Electron 1:126–130
Liu KH, Lee FCY (1990) Zero-voltage switching technique in DC/DC converters. IEEE Trans
Power Electron 5(3):293–304
Liu KH, Oruganti R, Lee FCY (1987) Quasi-resonant converters-topologies and characteristics.
IEEE Trans Power Electron PE-21(1):62–71
Middlebrook D, Cuk S (1977) A general unified approach to modeling switching converter power
stages. Int J Electron 6:521–550
Schutten MJ, Torrey DA (2003) Improved small-signal analysis for the phase-shifted PWM
power converter. Trans Power Electron 18(2):659–669
Smith KM, Smedley KM (1997) A comparison of voltage-mode soft-switching methods for
PWM converters. IEEE Trans Ind Electron 12(2):376–386
Theron PC, Ferreira JA (1995) The zero voltage switching partial series resonant converter. IEEE
Trans Ind Appl 31(4):879–886
Vorperian V (1990) Simplified analysis of PWM converters using the model of the PWM switch,
part I(CCM) and II (DCM). Trans Aerosp Electron Syst 26(3):490–496
Chapter 148
The Research of Industrial Optimization
of Beijing CBD Ribbon Based on Fitness
Function Mode
Youliang Zhang and Gang Zong
Abstract Beijing CBD, guided by modern service industry, is international
business areas with international finance as the leading role. By fitness function
model, it researches the scale-free characteristics of industrial network of CBD
Ribbon, on the basis of analyzing its network node by cluster analysis. The results
show that the modern service industry is closely linked to the traditional leading
industries, and the stimulative correlation relationship with the rest of the industry,
which will promote the continuous optimization of regional industrial structure.
Keywords CBD ribbon Cluster analysis Fitness function model
148.1 Introduction
There are numerous nodes in the urban industrial system, the generation of pillar
industry and leading industry and the evolvement of forerunner industry and sunset
industry have certain rules. To analyze urban industrial clustering structure by
means of fitness function model in order to discover its rules in evolvement and
features can accelerate the development of urban industrial structure as well as
promote the optimization of urban industry.
Y. Zhang (&) G. Zong
School of Economics and Management, Beijing University of Technology,
Beijing, People’s Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1401
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_148,
Ó Springer-Verlag Berlin Heidelberg 2013
1402 Y. Zhang and G. Zong
148.1.1 Fitness Function Model
There are few abnormal nodes in many realistic networks. These nodes’ growth
rate of degree depends on not the age of nodes but their competitive capacity.
Moreover, they don’t acquire new sides complying with the principle of degree
preferential attachments. They may connect to only a few sides in the early
evolution of time step for these nodes (Population Division, Department of Eco-
nomic and Social Affairs, United Nations 2010). According to the principle of
degree preferential attachments, the probability that they may obtain new sides
should be very small. Because of some other reasons, however, they have a greater
probability to get new sides. This phenomenon reflects the fitness is becoming rich.
As a result, people put forward fitness function model based on this phenomenon.
The following is the evolutionary pattern of its network model (Gomez-Gardenes
et al. 2006):
1. Increase: Begin with a network included nodes, bring in one new node each
time and connect it to existing nodes, fitness of each node is selected in
accordance with probability distribution.
2. Preferential connection: The probability of connection between a new node and
an existing node, the node’s degree, the node’s degree and fitness meet the
following relationship:
Y g ki
¼ Pi ð148:1Þ
i
gj k j
j
More specifically, utilize fitness function model in the optimal development of
urban area industry:
1. Increase: In initial process, after years of development, it has formed industrial
system which is comprised of mi relative independent industries. These
industries may be in the position of forerunner industry, leading industry, pillar
industry and sunset industry respectively in the present industrial structure.
2. Preferential connection: In the next time step, there will be a new industry in
urban area industrial system. The new industry will demonstrate a strong ability
to adapt. A great deal of social resource will concentrate on this new industry
due to general optimism and particular attention about it. As a result, this
industry will develop rapidly in the short term and establish extensive contact
with other industries; in the following time step, all evolvement of urban area
industrial system will comply with this industrial increasing principle.
148.1.2 Evolutionary Analysis of the Model
Evolving in accordance with fitness function model, the prominent feature of
urban area industry is not only able to establish relevance with the original
148 The Research of Industrial Optimization of Beijing CBD Ribbon 1403
Fig. 148.1 The increase of
industrial network based on
fitness function model B B
A A
M
M3
M1 M2
industry but derive the new industry rapidly, and even become the new leading
industry and pillar industry (Mo et al. 2008; Wang et al. 2006). It increases the
proportion of it in the national economy, and promotes optimization of urban area
industrial structure and increase of overall economy constantly, as shown in
Fig. 148.1.
A, B represent leading industry, pillar industry, and M represents emerging
industry and its derivative industry. In the long term, the industrial optimization
based on the fitness function model confirms to the stream and principle of
industrial development. But under the effect on path dependence, urban area has
already formed several industries so this process will be limited. There is a con-
stant gambling process between emerging industry and original industry (Liu
2009). If they have no mutual promoting relationship, participation of the
emerging industry must squeeze the other’s living space and emerge competitive
relationship. Meanwhile, the original and relative outdated industries are not
willing to withdraw from the market. Consequently, the government needs to give
enough initial support before the maturity of emerging industries; as the maturing
of the emerging industry, its relevant industry enters into market gradually and the
industrial structure will be optimized constantly.
Certainly, the participation of emerging industry has limit. The new partici-
pants, comparing with other industries, have to invest on numerous personnel,
resources and capital in order to obtain greater competitive advantages (Luo 2005).
Or else the emerging industry may abort so as to make adverse effect on long term
development.
148.2 Methodology
148.2.1 The Network Analysis of Industrial Optimization
in CBD Ribbon
We can get Table 148.1 through classifying and disposing the date of Beijing CBD
Ribbon’s node relation. Utilizing Ucinet software, we can generate corresponding
network relation figure (viz. Fig. 148.2). The node represents the industry’s
1404 Y. Zhang and G. Zong
Table 148.1 The list of No. Node name
network node of industries in
CBD ribbon Bb2 Manufacturing
Bb3 Electric power, fuel gas and water production and supply
industry
Bb4 Construction industry
Bc1 Transportation, storage and postal industry
Bc2 Information transmission, computer services and software
industry
Bc3 Wholesale and retail industry
Bc4 Accommodation and catering industry
Bc5 Financial industry
Bc6 Real estate industry
Bc7 Lease and business services industry
Bc8 Scientific research, technology services and geological
exploration industry
Bc9 Water conservancy, environment and public facilities
management industry
Bc10 Resident services and other services industry
Bc11 Education
Bc12 Health, social security and social welfare industry
Bc13 Culture, sport and recreation
Fig. 148.2 The industrial
network construction figure
of olympic ribbon
designation and the line represents the relations among industries. The scale of
network represents quantity of total participants; therefore the scale of industrial
network in this Ribbon is 16.
As shown in the Fig. 148.1, we can see clearly that this network has significant
scale-free network characteristic that the connection situation (number of degrees)
among each node has asymmetrical distributivity, and most of nodes have relative
less number of degrees and only few of them have higher number of degrees. In
the CBD Ribbon industrial structure, retail business and traditional service
148 The Research of Industrial Optimization of Beijing CBD Ribbon 1405
Fig. 148.3 The network node’s correlation matrix of each industry in CBD ribbon
industry still have relative higher number of degrees, but the proportion of CBD
Ribbon’s modern service industry grows constantly in national economy, and its
industrial correlation degree has significant improvement, emerging a trend to
become the leading industry gradually (Newman 2001; Barabási 2001; Liljeros
et al. 2001).
148.2.2 The Structural Analysis of Complicated Network
of CBD Ribbon Industry
According to the relationship matrix of CBD Ribbon’s each industry, we can
generate corresponding network node’s correlation matrix (Table 148.1) and each
industrial network node’s cluster analysis tree (Fig. 148.3) by utilizing Uncinet
software. According to the correlation coefficient of the modern service industry
(Bc7, Bc2) and related industries, we can figure out that the correlation coefficient
of Bc7 and Bc5 is 0.75, similarly, the correlation coefficient of Bc2 and Bc5 is
0.42. It shows that the correlation between modern service industry and financial
industry is the most significant due to the higher correlation coefficient (Refiner
1998; Yook and Jeong 2002). As shown in the each industrial network node’s
cluster analysis tree (Fig. 148.4), we can clearly see that modern service industry
has intimate relation with traditional leading industry and mutual promoting
1406 Y. Zhang and G. Zong
Fig. 148.4 The network node’s cluster analysis tree of each industry in CBD ribbon
relation with other industries. Therefore modern service industry shows strong
adaptive capacity and will optimize the regional industrial structure constantly.
148.3 Conclusions
This article carries on evolutionary analysis which analyzed 16 industries of
Beijing CBD Ribbon on fitness function model supported by the construction of
urban industry. As a result, it verifies the scale-free characteristic of inner
industrial relation. In the CBD Ribbon, traditional industries like the retail and
traditional services have greater centrality. However there is a gradual increasing
trend in the relevancy of modern service industry. Moreover it has a tight con-
nection with traditional leading industry as well as other industries.
148 The Research of Industrial Optimization of Beijing CBD Ribbon 1407
References
Barabási AL (2001) The physics of the web [DB/OL]. http://physicsweb.org/articles/world/14/7/
9/1. Physics Web
Gomez-Gardenes J, Echenique P, Moreno Y (2006) Immunization of real complex communi-
cation networks. Eur Phys J B 49:259–264
Liljeros F, Rdling CR, Amaral LAN et al (2001) The web of human sexual contacts. Nature
411:234–243
Liu J (2009) Overall network analysis lecture notes. Gezhi Press, Shanghai, p 2
Luo J (2005) Social network analysis lecture notes. Social Science Literature Press, Beijing, p 4
Mo H, Wang J, Jin F (2008) The complexity research of transportation network. Prog Geogr Sci
27(6):112–120
Newman MEJ (2001) The structure of scientific collaboration networks. Natl Acad Sci
98:404–409, 455–459
Population Division, Department of Economic and Social Affairs, United Nations (2010) World
urbanization prospects the 2009 revision, New York
Refiner S (1998) How popular is your paper? An empirical study of the citation distribution. Eur
Phys J B 4:131–134
Wang X, Li X, Chen G (2006) Complex network theory and application. Tsinghua University
Press, Beijing, p 5
Yook S-H, Jeong H (2002) Modeling the Internet’s large-scale topology. Natl Acad Sci
99(21):13382–13386
Chapter 149
A Study on the User Acceptance Model
of SNS Websites Based TAM
Dan Jin and Mei-mei Zhou
Abstract Based on Davis’s TAM, combining with the user satisfaction theory in
information system and motivation theory, and the SNS user behavior character-
istics, this study proposed the user acceptance model on SNS websites. In this
model, Perceived Usefulness and Perceived Ease-of-Use were retained, and Per-
ceived Enjoyment and Perceived Connectivity were added. In addition, the
external variables affecting these key factors were subdivided. The questionnaire
was designed and Structural Equation was used to validate the empirical
hypothesis. The results showed that TAM could apply to user acceptance on SNS
website basically, and Perceived Enjoyment and Perceived Connectivity were all
positively correlated with Willingness, also, the subdivision of external variable
reflected the importance of user activity.
KeyWords Activity level Perceived enjoyment SNS websites Technology
acceptance model
149.1 Introduction
SNS websites is a booming Internet applications, based on the theory of ‘‘Six
Degrees of Separation’’, taking the customer relationship as the core, designed to
help people build social networks. The large number of Internet users makes a
great contribution to the rapid development of SNS sites. Various SNS websites
rise rapidly after 2006, and competition is fierce, causing serious phenomenon of
homogeneity.
D. Jin (&) M. Zhou
School of Management and Economics, Beijing Institute of Technology, Beijing, China
e-mail: [email protected]
M. Zhou
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1409
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_149,
Ó Springer-Verlag Berlin Heidelberg 2013
1410 D. Jin and M. Zhou
MicroBlog, a Twitter-like service, has been rapidly developed, and the regis-
tered users have increased dramatically.SNS websites users are coincident with
those of MicroBlog to a great extent. Double pressures of homogenous competi-
tion and MicroBlog’ rising, makes SNS websites face a great challenge. SNS
websites need to absorb new users continuously and retain old ones. SNS websites
are facing with the problem of user acceptance.
149.2 Literature Review
The user acceptance problem of SNS websites belongs to information technology
acceptance. The user’s act in accepting technology is regarded as one of the most
mature research field of information systems (Davis 1989). The most representa-
tive theories are theory of rational act (TRA), theory of planning behavior (TPB),
and technology acceptance model (TAM).
149.2.1 Technology Acceptance Model
Based on rational behavior theory, Davis (1986) put forward TAM. TAM adopts
the well-established causal chain of beliefs ?attitude ? intention ? behavior
which has become known as the Theory of Reasoned Action (TRA). Based on
certain beliefs, a person forms an attitude about a certain object, on the basis of
which he or she forms an intention to behave with respect to that object. The
intention to behave is the sole determinant of actual behavior (Fig. 149.1). In TAM
applications, two key factors, Perceived Usefulness and Perceived Ease-of-Use,
can effectively explain users’ behavior intention.
149.2.2 Other Related Theory
Papacharissi and Rubin (2000) summarized motivation of Internet usage in the
following: interpersonal communication, killing time, achieving information,
Fig. 149.1 Technology Perceived
acceptance model Usefulness
External Attitude Behavior System
Variables Intention Use
Perceived
Ease of Use
149 A Study on the User Acceptance Model 1411
convenience and entertainment. Through the online questionnaire survey in a
German SNS website, Schaefer (2008) discussed participating motivation and
running mode, considered the participating motivation includes keeping in touch,
searching for information, entertainment, communicating, managing existing
relationship and so on. From these motivations, some other factors may also affect
SNS websites usage, apart from Perceived Usefulness and Perceived Ease-of-Use.
Wixom and Todd (2005) proposed that the user satisfaction of information
system could be united with Davis’s TAM. Information quality and system quality
influence information satisfaction and system satisfaction respectively, and
information satisfaction and system satisfaction exert influence on Perceived
Usefulness and Perceived Ease-of-Use respectively. Delone and Mclean (2003)
added service quality in the improved D&M information system success model,
and proposed that the SERVQUA scale in marketing field could be referred to
measure service quality. As studying the relationship between intimate degree and
diving behavior, Patrick Rau (2008) and others measured the member behavior of
participating in SNS websites with member activity level. These theories could
provide a foundation for further study on external variables subdivision in TAM.
149.3 Model and Hypothesis
149.3.1 User Acceptance Model of SNS Websites
In TAM, the external variables have not been fractionized, which will make
against further analysis the influencing factors on user acceptance to SNS web-
sites.SNS websites provide an interactive platform for friends, on which integrates
basic Internet applications, such as log, photo, video, community and game, and
meet users’ sociality demands through online interactive among friends, infor-
mation sharing, participating in activities and other ways. Webster and Martocchio
(1995) thought entertainment was an intrinsic motivation using computers in
workplaces. Scholars put forward Perceived Enjoyment as doing Internet empirical
study, which refers to entertainment degree by using SNS websites. Moreover,
Perceived Connectivity refers to being connected with friends in passions and is
not confined by time or location. During using SNS websites, users may have the
feeling of satisfaction or happiness, which makes users to accept SNS websites in
further (Shin 2008).
Based on TAM, the user acceptance model for SNS websites is shown in
Fig. 149.2. Perceived Usefulness and Perceived Ease-of-Use are retained in the
model, and Perceived Enjoyment and Perceived Connectivity are added as another
two key factors affecting users to accept SNS websites. In addition, external
variables of the above factors are further divided into the information quality,
quality system, active degree and the service quality and related factors.
1412 D. Jin and M. Zhou
Fig. 149.2 Prototype of user Service Quality
acceptance model for SNS
Individuation Perceived
websites
Communication Enjoyment
Innovation
Activitylevel
Participation
Sharing Perceived
Connectivity
Sociability
Game Interactive
InformationQuality
Accuracy Behavior
Perceived Willingness
Timeliness Usefulness
Integrity
System Quality
Security Perceived
Ease-of-Use Social Impact
Interface
149.3.2 Assumptions of Model
The related assumptions in TAM are still established in this model. And new
assumptions about Perceived Enjoyment and Perceived Connectivity are proposed.
(1) Perceived Ease-of-Use and related assumptions.
Based on customer satisfaction theory and with the previous empirical data
support, (Seddon 1997) also confirmed that system quality has a positive effect
on Perceived Ease-of-Use. In addition, the system security is much important
to users, which means to promise personal information safety. Moreover, both
user interface and interactive process will affect users to accept SNS websites.
H1: System quality positive effect on this term
H1a: Security positive effect on it
H1b: Interface positive effect on it
H2: Perceived Ease-of-Use positive effect on this item
(2) Perceived Usefulness and related assumptions
H3: Information Quality positive effect on this item
H3a: Accuracy positive effect on it
H3b: Timeliness positive effect on it
H3c: Integrity positive effect on it
H4: Activity level positive effect on this item
H4a: Participation positive effect on it
149 A Study on the User Acceptance Model 1413
H4b: Sharing positive effect on it
H4c: Sociability positive effect on it
H4d: Game interactive positive effect on it
H5: Perceived Ease-of-Use positive effect on this item
H6: Perceived Usefulness positive effect on this item
(3) Perceived Connectivity and related assumptions
H7: Activity level positive effect on this item
H7a: Participation positive effect on it
H7b: Sharing positive effect on it
H7c: Sociability positive effect on it
H7d: Game interactive positive effect on it
H8: Perceived Connectivity positive effect on this item
(4) Perceived Enjoyment and related assumptions.
As a community website where users’ activities are based on group, SNS
website should pay much attention to users’ activity involvement, which can
tell the influence of Perceived Connectivity to Perceived Enjoyment.
H9: Service quality positive effect on this item
H9a: Individuality positive effect on it
H9b: Communication positive effect on it
H9c: Innovation positive effect on it
H10: Activity level positive effect on this item
H10a: Participation positive effect on it
H10b: Sharing positive effect on it
H10c: Sociability positive effect on it
H10d: Game interactive positive effect on it
H11: Perceived Connectivity positive effect on this item
H12: Perceived Enjoyment positive effect on this item
(5) Social impact and related assumptions
Social impact refers to others influence on individual for using SNS websites.
Similar to the TRA subjective enorms, people are very conscious of others
view about their particular behavior willing. Whether individual uses SNS
websites would be influenced by others recommendation or evaluation.
H13: Social impact positive effect on this item
149.4 Research Methods
149.4.1 Survey Questionnaire Design
In order to verify the model assumptions, a questionnaire is used to collect data. The
questionnaire uses a standard 7-point Likert-typescale. The 7 point are ‘‘completely
disagree’’, ‘‘relatively disagree’’, ‘‘some disagree’’, ‘‘not sure’’,’’ some agree’’,
‘‘relatively agree’’ and ‘‘completely agree’’. According to actual situation,
1414 D. Jin and M. Zhou
respondent shave the right to choose from 1 (completely disagree) to 7 (completely
agree). As the final questionnaire data is analyzed and tested by using structural
equation model, it need to meet the requirements that structural equation model to
observed variables and characteristics of measuring factors, so three or more items
are used to measure each factor. All the measure items use the mature scales which
have been used in empirical study by researchers at home and abroad, and are
adjusted according to SNS websites characteristics and questionnaire’s semantic
environment. The questionnaire has 18 variables to measure, including 13 external
variables, such as accuracy, timeliness, integrity, security, interface, individuality,
innovation, communication, participation, sociability, sharing, game interactive and
social impact, and 5 internal variables, as Perceived Usefulness, Perceived Ease-of-
Use, Perceived Enjoyment and Perceived Connectivity and Willingness. Above all,
the total number to be measured is 58.
149.4.2 Pre-test Questionnaire
In order to ensure the effectiveness of the survey questionnaire, it is necessary to
pretest it before actually using it. This paper uses SPSS 18 to do reliability and
validity analysis on the test results. The reliability analysis results showed that
Cronbach’s Alpha value of every measured factor was greater than 0.5, while the
overall questionnaire reliability was 0.959, indicating the internal consistency of
the questionnaire was acceptable. In the validity analysis, separate factor analysis
was made on corresponding measured items of independent variables, mediating
variables and dependent variables. The factor analysis result showed that other
factors were all better classified in the corresponding dimension apart from two
factors measured items, sociability and social impact of the independent variables.
So sociability factor and social impact factor were deleted.
149.4.3 Data Collection
According to the pre-test results, the questionnaire were modified in final two
ways, electronic and paper questionnaire. On the principle of simple random
sampling, questionnaires were distributed. 170 electronic questionnaires and 68
paper questionnaire were distributed, altogether 202 questionnaires were returned.
After moving the data obviously not meeting the requirements, 153 valid ques-
tionnaires were remained.
149 A Study on the User Acceptance Model 1415
149.5 Model Analysis
149.5.1 Second Order Confirmatory Factor Analysis
Second order confirmatory factor analysis (CFA) is put forward as there is high
degree relevance among original first order factors in the first order CFA, and the
first order CFA can be in agreement with sample data. In the model, information
quality, system quality, service quality and activity level are measured with the
multidimensional method, so AMOS 17 would be used respectively during their
second order factor analysis. The analysis results show, all the first order factors to
second order factors of information quality, system quality, service quality, and
activity level, the load capacity value lie between 0.5 and 0.95, and the Signifi-
cance Probability and C.R. are greater than 1.96, and decision criteria is achieved
totally. Meanwhile, compared with the goodness-of-fit standard, the whole
goodness-of-fit reaches the basic standard. So we can conclude that the first order
factors of information quality, system quality, service quality and activity level
would measure these second order factors well.
149.5.2 Model Analysis and Revision
According to the validity inspection and analysis results of second order CFA
above, social impact and sociality factors are removed from the original hypothesis
model, the goodness-of-fit indexes of the model are summarized as Table 149.1.
The ratio Chi-square/freedom degree of the model is 1.828 \ 2, RMSEA is
0.074 \ 0.080, two goodness-of-fit indexes meet the standard. But other goodness-
of-fit indexes, such as GFI (0.658 \ 0.9), AGFI (0.62 \ 0.8), CFI (0.828 \ 0.9),
TLI (0.816 \ 0.9) and NFI (0.689 \ 0.8), are not up to the standards, thus the
model need to be modified and optimized.
AMOS offers two model modification indexes, in which modification index MI
is used for model expanding, and the critical ratio C.R. is used for model
restricting. According to the value of critical ratio and modification indexes, fol-
lowing the principle of modifying one parameter once, the final revised model is
shown in Fig. 149.3.
The path of regression coefficients of each factor of the revised full model has
increased, the value of C.R. is larger than 1.96, some of the path of significant
probability are larger than 0.05 that are sufficiently close to the standard, which
shows the modification effect well. Table 149.2 is the path coefficient of the
revised model for the first time.
1416 D. Jin and M. Zhou
Table 149.1 Goodness-of-fit indexes of full model
Absolute goodness-of-fit index Relative goodness-of-fit index
v2 df GFI AGFI RMSEA CFI NFI TLI
(1, 5) [=0.9 [=0.8 \0.08 [=0.9 [=0.8 [=0.9
1.828 0.658 0.621 0.074 0.828 0.689 0.816
Fig. 149.3 SNS websites ActivityDegree
user acceptance model based
Participation Perceived
on TAM
Enjoyment
Sharing
GameInteractive
Perceived
Information Quality Connectivity
Accuracy Behavior
Intention
Timeliness Perceived
Integrality Usefulness
System Quality
Security Perceived
Ease-of-Use
Interface
The goodness-of-fit of the revised model is shown as Table 149.3. The value of
Chi square (1468.758) and freedom (835) improve markedly than the preceding
ones. The ratio Chi-square/degree of freedom (1.759 \ 2) and the value of
RMSEA (0.071 \ 0.08) reach the standard both. Several other goodness-of-fit
indexes meet the standard basically. For numbers of latent variables exist in the
model, the relationship between factors is relatively complex and some indexes
may be influenced greatly by the sample size, the revised model can be regarded as
the final model.
149.5.3 Test Results of Model Hypothesis
The model set up in this study originally uses 13 level 1 hypothesis from H1 to
H13. As system quality, information quality, service quality, and activity level use
multidimensional measure method, H1–H6 each exists level 2 hypothesis. The
data analysis results support the rest hypothesis apart from H4c, H5, H7c, H9, H9a,
H9b, H9c, H10c and H13.
Table 149.2 The path coefficient of revised full model
Path Estimated coefficient Standard deviation Critical ratio Significant probability Standard coefficient
(S.E.) (C.R.) (p)
connectedness \— activity 0.562 0.121 4.627 *** 0.535
usefulness \— information 0.913 0.232 3.940 *** 0.641
easiness \— system 0.702 0.111 6.340 *** 0.735
entertainment \— activity 0.758 0.120 6.311 *** 0.620
usefulness \— activity 0.343 0.137 2.492 0.013 0.315
entertainment \— connectedness 0.322 0.102 3.147 0.002 0.277
willingness \— usefulness 0.229 0.084 2.731 0.006 0.221
willingness \— entertainment 0.340 0.084 4.070 *** 0.368
willingness \— easiness 0.213 0.084 2.530 0.011 0.177
willingness \— connectedness 0.307 0.096 3.205 0.001 0.285
easiness4 \— easiness 1.342 0.121 11.089 *** 0.919
149 A Study on the User Acceptance Model
easiness3 \— easiness 1.411 0.125 11.283 *** 0.942
easiness2 \— easiness 1.000 0.717
easiness1 \— easiness 1.128 0.099 11.415 *** 0.790
willingness4 \— willingness 1.034 0.076 13.592 *** 0.850
willingness3 \— willingness 0.981 0.068 14.359 *** 0.876
willingness2 \— willingness 0.880 0.070 12.591 *** 0.813
willingness1 \— willingness 1.000 0.864
usefulness4 \— usefulness 0.944 0.100 9.440 *** 0.729
usefulness3 \— usefulness 0.946 0.103 9.225 *** 0.716
usefulness2 \— usefulness 1.000 0.803
usefulness1 \— usefulness 1.193 0.117 10.235 *** 0.779
entertainment4 \— entertainment 1.000 0.870
entertainment3 \— entertainment 1.147 0.061 18.711 *** 0.966
entertainment2 \— entertainment 1.016 0.060 16.805 *** 0.918
(continued)
1417
Table 149.2 (continued)
1418
Path Estimated coefficient Standard Critical Significant Standard coefficient
deviation(S.E.) ratio(C.R.) probability(p)
entertainment1 \— entertainment 0.923 0.078 11.784 *** 0.762
connectedness3 \— connectedness 1.214 0.183 6.651 *** 0.811
connectedness2 \— connectedness 1.000 0.685
connectedness1 \— connectedness 0.473 0.149 3.171 0.002 0.296
Where *** p \ 0.001
D. Jin and M. Zhou
149 A Study on the User Acceptance Model 1419
Table 149.3 Revised goodness-of-fit index of full model
Absolute goodness-of-fit Index Relative goodness-of-fit Index
x2/df GFI AGFI RMSEA CFI NFI TLI
(1, 5) [=0.9 [=0.8 \0.08 [=0.9 [=0.8 [=0.9
1.759 0.717 0.680 0.071 0.863 0.734 0.852
149.6 Conclusion and Prospect
This study constructs SNS websites user acceptance model based on TAM, and
model hypothesis is verified by the structure equation model. The conclusion can
be drawn as follows: (1) TAM is basically suitable to SNS websites user accep-
tance study, but no evidence supports the causality between Perceived Usefulness
and Perceived Ease-of-Use. (2) Both Perceived Enjoyment and Perceived Con-
nectivity have a positive correlation with Usage Willingness, and Perceived
Connectivity further affects Perceived Enjoyment. (3) The subdivision of external
variables reflects the importance of user activity, the activity level influences
Perceived Usefulness, Perceived Connectivity and Perceived Enjoyment simulta-
neously, while the influence of service quality to Perceived Enjoyment is deleted
for path of regression coefficients is too little.
Based on above study, some constructive suggestions are proposed to SNS
service providers: perfect amusement and e-commerce functions to enhance user’
viscidity, pay attention to interface operation to optimize user’ experience; provide
service in information filtering, sorting and pushing, study deeply on promoting
user’ activity level.
Acknowledgments Supported by National Natural Science Foundation of China (No.
70802008), Beijing Municipal Natural Science Foundation (No. 9112011).
References
Davis FD (1986) A technology acceptance model for empirically testing new end user
information systems. Cambridge, MA
Davis F (1989) Perceived usefulness, perceived ease of use, and user acceptance of information
technology. MIS Q 13(3):319–340
Delone WH, McLean ER (2003) The DeLone and McLean model of information systems
success: a ten-year update. J Manage Syst 19(4):9–30
Papacharissi Z, Rubin AM (2000) Prediction of Internet use. J Broadcast Electron Media 44:4–13
Patrick Rau P-L, Gao Q, Ding Y (2008) Relationship between the level of intimacy and lurking in
online social network services. Comput Human Behav 24:2757–2770
Schaefer C (2008) Motivations and usage patterns on social network sites. Institute of
Information Systems and Management, Kaiserstr
Seddon PB (1997) A respecification and extension of the DeLone and McLean Model of IS
Success. Inf Syst Res 8(3):240–253
1420 D. Jin and M. Zhou
Shin D (2008) What do people do with digital multimedia broadcasting? Path analysis of
structural equation modeling. J Mobile Commun 6(1):258–275
Webster J, Martocchio JJ (1995) The differential effects of software training previews on training
outcomes. J Manage 21(4):757–787
Wixom BH, Todd TA (2005) A theoretical integration of user satisfaction and technology
acceptance[J]. Inf Syst Res 16(1):65–102
Chapter 150
Augmented Reality Based Factory Model
Comparison Method
Wei-wei Sun, Jian-feng Lu and De-zhong Li
Abstract Through factory digital mock-up, Digital Factory (DF) technology can
save enormous time and cost in factory planning. A problem of the digital factory
mock-up maintenance is checking the digital models with the real factory. This
paper introduces a method using Augmented Reality (AR) technology to compare
the 3D models with the real object in real time. Compared to other measures, this
method have the benefit of the cost saving. An experiment demonstrates the
proposed method is given at the end of the paper.
Keywords Digital factory mock-up Model comparison Augmented Reality
ARToolKit
150.1 Introduction
The factory digital mock-up creates a visual simulation platform for product
design and processes planning which has been the key point to optimize processes
and offer optimal production scheme (Bracht and Masurat 2005). It works as the
foundation of Digital Factory (DF) technology which is widely used in many fields
such as aviation, automobile manufacturing, chemical industry and electronic
products. Model calibration is a significant issue in the application of factory
W. Sun (&) J. Lu
CIMS Research Center, Tongji University, Shanghai, China
e-mail: [email protected]
J. Lu
e-mail: [email protected]
D. Li
CDHAW, Tongji University, Shanghai, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1421
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_150,
Ó Springer-Verlag Berlin Heidelberg 2013
1422 W. Sun et al.
digital mock-up. Previous methods of model comparison perform inefficient and
high-cost, which result in a deep impact on the development of DF. With the
expectation of seeking a cost-effective way, Augmented Reality (AR) is proposed
to be used in factory digital model comparison (Azuma 1997; Azuma et al. 2001).
150.2 Factory Digital Mock-Up
Factory digital mock-up is a complex of digital archive for the whole life-cycle of
a factory. It includes not only 3D model of the factory, but also all the design
documents, construction documents, and maintenance information. DF technol-
ogy, the company to meet the challenges of the twenty-first century an effective
means (Liu 2009), which integrates techniques of computer, virtualization, emu-
lation, and networks, plays a significant role in keeping competitive advantage for
enterprises. It operates in a collaborative way under 3-D visualization environment
and interactive interface. Based on the actual data and models the planned products
and production processes can be improved using virtual models until the processes
are fully developed and extensively tested for their use in the real factory. DF is a
comprehensive approach in factory layout planning, which consists of the 3-D
model design of plant (that contains workshop structure, equipments and facilities,
material flow and other resources for production) and processes optimization
(Zhang et al. 2006). And Factory digital mock-up acts as the prerequisite for
operative information concerned.
Factory digital mock-up shows its advantages: engineers make assessment by
optimizing the plant layout and resolving conflicts between different parts, then
avoid loss due to irrational design, and make data and information of equipments and
process flow optimum coordinate with the factory building (Yu et al. 2003). When
applied in the aspect of automobile industry, it coordinates materials resource
(components and modules of automobile), equipment (machine tools and facilities),
workshop (area), and process flow (automobile manufacturing processes) into an IT
system. While in the field of pharmaceutical and chemical industry, factory digital
mock-up makes it possible to increase product innovation and flexibility.
150.3 Methodology
150.3.1 Method of Factory Digital Mock-Up Comparison
One of the most basic problems currently limiting factory digital mock-up
applications is updating. Regular calibrating factory models taking real factory for
reference is necessary. There are several model comparison methods prevailed:
laser scanning, laser ranging, photograph-visual inspection comparison.
150 Augmented Reality Based Factory Model Comparison Method 1423
Laser scanner firstly gets the outline integrated data of the object rapidly by
Omni-directional scanning, and then generates point cloud records after precisely
construction, editing, and modification by computer. Accurate as the data is, the
method cannot be widely used because of the high cost. Furthermore, the instru-
ment is unable to display data instantly.
Another approach for updating factory digital mock-up is laser ranging. People
can easily obtain elevation and other information of the object, relative position,
for example, using the hand-held laser distance meter. However, this kind of
method is not appropriate for objects which are precise and complicated.
Photograph-visual inspection means to make a comparison between real factory
and the photograph of the factory. It is feasible, however, not accurate.
These methods above are commonly used at present. However, more efforts
should be paid to explore new resort which is inexpensive and accurate. Therefore,
Augmented Reality based factory model comparison method has been proposed.
AR system can present a view of blended scene of real factory environment and
digital mock-up. Then the information files can be easily changed to correct the
model without complicated manual operation.
150.3.2 Augmented Reality and ARToolKit
Augmented Reality (AR) technology can enhance user’s perception of the real
world by providing information from computer system. It has the following three
characteristics: combine real and virtual, interactive in real time, and registered in
3-D (Azuma 1997; Azuma et al. 2001). AR system has been applied in medical,
manufacturing, visualization, path planning, entertainment, military and many
other aspect (Quan et al. 2008).
ARToolKit is an AR application development tool kit based on C/C++ lan-
guage. It has successfully developed indoor registration technology with fiducial
mark pattern tracking system. On the condition of controlled environment, it has
achieved a fine tracking result. The kit includes camera calibration and mark
making tool which can compound Direct3D, OpenGL graphics and VRML scenes
into the video stream and support a variety of display devices.
150.3.3 Augmented Reality Based Factory Model
Comparison Method
Augmented Reality based factory model comparison method is put forward on the
foundation of model-loaded procedure in ARToolKit to blend virtual objects
seamlessly with a real factory environment in 3-D. The mainly work flow is shown
in Fig. 150.1.
1424 W. Sun et al.
Fig. 150.1 AR based model
comparison method work Train mark
process
Identify
Lord Model
Modify
N
Blend
End
Take the case of a section of a cross fire fighting sprinkler and a pipe support in
a classroom (as Fig. 150.2 shows) which is too high to measure, just meet the
requirements of the experiment object selection.
The operating system is Windows XP with Microsoft Visual studio 2008
development environment. An ordinary CMOS camera (pixels 320 9 240) with
2.0 USB interface and a printed fiducial mark is enough.
Besides provided mark patterns, other patterns also can be designed and trained
according with the instruction in ARToolKit. There are some limitations to purely
computer vision based AR systems. The larger the physical pattern, the further
away the pattern can be detected. What’s more, the simple pattern is better. Taking
the height of experimental subject into account, a proper coordinate axis offset
should be set to get a clearly vision. The mark was fixed in the bottom right-hand
corner of the object which is also shown in Fig. 150.2.
DAT files contain the name, size display, rotation and other information of the
models. Some new models can be matched with modifying the data as well as
model updating. When multiple patterns tracked associate with different 3D object
needed, DAT files also can be easily fixed to load more than one pattern.
The virtual model of the classroom was created by Microstation in real pro-
portion and converted into WRL by 3DSMAX. The mark shown in Fig. 150.2 has
been trained to match the digital model. The initial vision of the blending scene
was rendered as soon as the camera identified the mark, which is shown in
Fig. 150.3.
The model and the vision of real object can be rotated synchronously when the
camera moved. While immersed in a view of real world, some subtle changes in
DAT files can make the model get closer to the object. Then the position can be
determined with acceptable precision and accuracy in real time. The modified
model was shown in Fig. 150.4.
150 Augmented Reality Based Factory Model Comparison Method 1425
Fig. 150.2 The real object
scanning
Fig. 150.3 The initial vision
of blending
Fig. 150.4 The modified
vision of blending
1426 W. Sun et al.
150.4 Conclusion and Future Work
From the conclusions drawn from experiment above, Augmented Reality based
factory model comparison method has the following advantages over the existing
methods:
(1) Flexibility. The method operated in a direct way which has avoided laser
scanning, ranging, or other manual work.
(2) Low-cost. It can do the work well without professional instrument and also
result in time-saving.
(3) Accuracy. The real object was blended with the virtual model seamlessly, that
lead to errorless result.
(4) Real-time. The real world and the digital model were rendered and combined
in real time, and it is possible to reduce time with the use of the method by
working more efficiently.
Certainly, there are also some limitations. The method cannot work without
document database of the object, while Laser scanning is better. And Laser ranging
does well in presenting data like elevation and relative position for some big
facilities. It is more efficiently to make a proper combination of these approaches
in the real factory.
Since visual interface has advanced the performance, more effort needed to
perfect the AR system, which includes tracking visible natural features without
prepared marks (Neumann and You 1999), model loaded automatically, project of
the controller interface with CAD and so on. The author will do further study in
this area to support the factory digital modeling techniques in future.
References
Azuma RT (1997a) A Survey of augmented reality. Presence; Teleoper Virtual Environ
6(4):355–385
Azuma RT (1997) A survey of augmented reality. Presence Teleoper Virtual Environ 6(4): 355–38
Azuma RT, Baillot Y, Behringer R et al (2001) Recent advances in augment reality. IEEE
Comput Graph Appl 21(6):34–47
Azuma R, Baillot Y, Behringer R, Feiner S, Julier S, MacIntyre B (2001) Recent advances in
augmented reality. IEEE Comput Graph Appl 21(6):34–47
Bracht U, Masurat T (2005) The digital factory between vision and reality. Comput Ind
56(4):325–333
Liu D (2009) About digital factory [DB/OL]. www.ai.vogel.com.cn
Neumann U, You S (1999) Natural feature tracking for augmented reality. IEEE Trans Multimed
1(1):53
Quan H, Wang C, Ling J (2008) Survey of vision-based augmented reality technologies. Robot
30:379–384
Yu C, Guo G, Liu J (2003) Digital Mock-up (DMU) technology in the development of product
innovation. Sichuan Ordnance J pp 45–48
Zhang H, Fan L, Ma Y (2006) The technology and application of digital factory (Chinese).
Beijing, China Machine Press, pp 28–29
Chapter 151
An Evaluation Method Based on Cloud
Model for the Credit of High-Tech SMEs
Guo-qiang Zhou, Xue-qing Wang, Rui Liu and Li-guo Sun
Abstract In the process of resolving financing difficulties of high-tech small and
medium enterprises (SMEs) in China, the evaluation of credit risk of high-tech
SMEs becomes a very challenging problem for the bank. This paper proposes a
novel evaluation method based on cloud model to measure the credit risk of
Chinese listed high-tech SMEs. Finally, an example is provided for illustrative
purpose, and the indexes system of credit evaluation is established of 25 key
factors, embedded within five broad categories: credit quality, organizational level,
operation level, R&D level and network position. This research shows that it is a
better way to use this method to realize transforming qualitative terms described in
a natural language to distribution patterns of quantitative values, especially for
high-tech SMEs.
Keywords Cloud model Credit evaluation Credit risk Credit scoring
High-tech SMEs
151.1 Introduction
In recent years, the credit of high-tech small and medium enterprises (SMEs) has
been gaining much more importance according to their high growth in financial
world (Derelioglu and Gürgen 2011). However, the credit guarantee risk is very
G. Zhou (&) X. Wang
College of Management and Economics, Tianjin University,
Tianjin, People’s Republic of China
e-mail: [email protected]
R. Liu
Nuclear and Radiation Safety Center, Ministry of Environmental Protection,
Beijing, People’s Republic of China
L. Sun
China Bohai Bank, Tianjin, People’s Republic of China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1427
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_151,
Ó Springer-Verlag Berlin Heidelberg 2013
1428 G. Zhou et al.
high in SMEs due to their particular characteristics, which lead to a low credit
scoring in general (Chen et al. 2006, 2010). The evaluation of credit risk of high-
tech SMEs becomes a very challenging problem for the bank. Therefore, it is
essential to develop an accurate credit scoring model for high-tech SMEs for the
efficient management of bank.
Most well-known evaluation models use probability or fuzzy set theory to hold
randomness or fuzziness respectively, such as the decision trees (Frydman et al.
1985), artificial neural networks (Jensen 1992), genetic algorithm (Walker et al.
1995), etc. Among all of these methods, only cloud model based models consider
both aspects of uncertainty. Cloud model is the innovation and development of
membership function in fuzzy theory (Di et al. 1999), which transforms qualitative
terms described in a natural language to distribution patterns of quantitative values
(Deyi et al. 1995). It is successfully used in spatial analysis (Haijun and Yu 2007),
target recognition (Fang et al. 2007), intelligent algorithm improvement (Yunfang
et al. 2005) and so on.
Therefore, in this paper, we propose an evaluation method based on cloud
model for the credit of high-tech SMEs. As credit evaluation is a typical multi-
attribute evaluation problem, it is more significant in applying this novel approach
to credit evaluation so as to demonstrate its usefulness.
151.2 Methodology
151.2.1 Basic Concepts of Cloud Model
Suppose that r is the language value of domain U, and mapping CrðxÞ : U ! ½0; 1,
8x 2 XðX UÞ, x ! CrðxÞ, then the distribution of CrðxÞ in U is called the
membership cloud of r, or cloud in short, and each projection is called a cloud drop
in the distribution. If the distribution of CrðxÞ is normal, it is named normal cloud
model.
151.2.2 Numerical Characteristics of the Cloud Model
The normal cloud model is determined by the following three parameters:
expectation Ex, entropy En, and hyper entropy He.
Expectation Ex represents the values that mostly stand for this qualitative
concept, generally it is the value x that corresponds to the gravity center of the
cloud, it should belongs to this qualitative concept hundred percent.
Entropy En is the measurement of the qualitative concept fuzzy degree, it
determines the range of the cloud and about 99.74 % cloud drops fall within the
151 An Evaluation Method Based on Cloud Model 1429
Fig. 151.1 Numerical
characteristics of the cloud
model
range between ½Ex 3En; Ex þ 3En, and it reflects the numerical value range
acceptable by concept and represents the margin with double-sided property.
Hyper entropy He is the entropy of entropy, it reflects the dispersion degree of
the entropy of concept (Lv et al. 2009).
A cloud model can be denoted with vector CðEx; En; HeÞ. The numerical
characteristics of the cloud Model are shown in Fig. 151.1.
151.2.3 Normal Cloud Model
Let U be a quantitative universal set and r be the qualitative concept related to
U. If x 2 U, which is a random realization of the concept r, and x satisfies
0 0
x NðEx; En 2 Þ, where En NðEn; He2 Þ, and the certainty degree of x on r is
ðxExÞ2
0
l¼e 2ðEn Þ2 ð151:1Þ
Then the distribution of x on U is a normal cloud (Li et al. 2009), and every x is
defined as a cloud drop. Given the three parameters Ex, En, He, the normal cloud
model can be generated (Deyi and Yi 2005).
Input: Ex, En, He, and the number of the cloud drops n.
Output: n of cloud drops x and their degree l.
Step 1. Generate a normally distributed random number En0i with expectation
En and variance He2 , i.e., En0i ¼ NORMðEn; He2 Þ.
Step 2. Generate a normally distributed random number xi with expectation Ex
and variance En02 02
i , i.e., xi ¼ NORMðEx; Eni Þ.
Step 3. Calculate
ðxi ExÞ2
2En02
li ¼ e i ð151:2Þ
Step 4. xi with certainty degree of li becomes one cloud drop in the domain.
Step 5. Repeat Steps 1 to 4 until n cloud drops are generated.
1430 G. Zhou et al.
151.2.4 Cloud Model-Based Credit Evaluation Algorithm
In the evaluation method based on cloud model, gravity center of cloud can be
denoted as:
T ¼ab ð151:3Þ
In the type, a means the position of gravity center of cloud, depicting with the
expectation value Ex, then if the expectation value Ex changes, the position of
gravity center of cloud also corresponds of change; b means the height of gravity
center of cloud, depicting with the heavy value of power, which takes often value
(0.371).
Therefore, the variety that passes gravity center of cloud can reflect the variety
of system information status, the concrete step of the evaluation method based on
cloud model is as follows:
Step 1: Confirming index system and index power weight.
Step 2: Denoting the cloud model of each index. Denotation of accurate-
number type and language-description type are different in cloud model. Withdraw
a set of sample n to constitute to make policy matrix, so the index of accurate-
number type can be denoted as follow:
Ex1 þ Ex2 þ þ Exn
Ex ¼ ð151:4Þ
n
maxðEx1 ; Ex2 ; . . .; Exn Þ minðEx1 ; Ex2 ; . . .; Exn Þ
En ¼ ð151:5Þ
6
And the index of language-description type can be denoted as follow:
Ex1 En1 þ Ex2 En2 þ þ Exn Enn
Ex ¼ ð151:6Þ
En1 þ En2 þ þ Enn
En ¼ En1 þ En2 þ þ Enn ð151:7Þ
Step 3: Denoting status of the system. n indexes can be depicted with n cloud
models, therefore the evaluation system containing n indexes can be denoted with
n dimension comprehensive cloud, T ¼ ðT1 ; T2 ; . . .; Tn Þ, Ti ¼ ai bi ði ¼ 1; 2;
. . .; nÞ, and when the status of evaluation system occurrences variety, the gravity
center changes to T 0 ¼ ðT10 ; T20 ; . . .; Tn0 Þ.
Step 4: Measuring the variety of cloud gravity center based on power-added
deviation degree. Suppose that each index of the ideal status of a system is given,
then the vector of cloud gravity center can be depicted as T 0 ¼ a bT
¼ ðT10 ; T20 ; . . .; Tn0 Þ, a ¼ ðEx01 ; Ex02 ; . . .; Ex0n Þ, b ¼ ðb1 ; b2 ; . . .; bn Þ, bi ¼ wi 0:371,
and the normal status of vector of n dimension comprehensive cloud gravity center
is denoted as T ¼ ðT1 ; T2 ; . . .; Tn Þ.
Generally power-added deviation degree can be used to measure the variety of
cloud gravity center between ideal status and normal status. The vector of
151 An Evaluation Method Based on Cloud Model 1431
Fig. 151.2 Five-scale
evaluation set based on cloud
model
n dimension comprehensive cloud gravity under normal status is normalized to
T G ¼ ðT1G ; T2G ; . . .; TnG Þ, among them,
(
ðTi Ti0 Þ=Ti0 ; Ti \Ti0
TiG ¼ ði ¼ 1; 2; . . .; nÞ ð151:8Þ
ðTi Ti0 Þ=Ti ; Ti Ti0
Therefore, power-added deviation degree hð0 h 1Þ is denoted as:
X
n
h¼ wi TiG ð151:9Þ
i¼1
Under the ideal status, h ¼ 0.
Step 5: Confirming the evaluation set based on cloud model. Generally, the
more numbers of evaluation scales, the more accurate of the evaluation results.
According to the feature of high-tech SMEs, five-scale evaluation set is adopted
(see Fig. 151.2):
V ¼ ðv1 ; v2 ; v3 ; v4 ; v5 Þ
¼ ðbadðCÞ; generalðBÞ; goodðAÞ; verygoodðAAÞ; bestðAAAÞÞ
151.3 Application Example
151.3.1 Confirming Index System and Sample Data
The index system of credit evaluation of high-tech SMEs includes total 25 indexes,
5 major type as follows: credit quality U1 —register capital U11 , history credit
condition U12 , equipment level U13 and guarantee U14 ; Organizational level U2 —
business strategy U21 , organization system U22 , stability of management team U23 ,
stability of R&D team U24 , business proposal U25 ; Operation level U3 —turnover
1432 G. Zhou et al.
ratio of accounts receivable U31 , turnover ratio of total assets U32 , return on total
assets ratio U33 , operating profit ratio U34 , income growth ratio U35 , profit growth
ratio U36 , liquidity ratio U37 , debt asset ratio U38 , after-sales service U39 ; R&D
level U4 —R&D input U41 , intellectual property rights U42 , R&D character U43 ;
Network position U5 —market share U51 , public relations U52 , industry trend U53 ,
geography position U54 .
Select ten high-tech SMEs in the second-board Market in China as test sample
set S ¼ fSi ji ¼ 1; 2; . . .; 10g, including electronics information, medical apparatus,
biological pharmacy, etc. Finance data comes from database in CSMAR Solution
(2012), and other qualitative indexes are descripted by the expert evaluation
languish. Take company S1 (NO. 300002) in the test sample set as example, the
status of each credit evaluation index are shown in Tables 151.1, 151.2, 151.3,
151.4 and 151.5.
Table 151.1 Status of credit quality U1
Status U11 U12 U13 U14
1 37920 33470815 General Good
2 37920 33470815 General Good
3 37920 33470815 Good Good
4 37920 252764848 Good Very good
Ideal 100000 300000000 Best Best
Table 151.2 Status of organizational level U2
Status U21 U22 U23 U24 U25
1 1.701 Good Very good Good Good
2 1.789 Good Good Good Good
3 1.660 General Good Good Good
4 1.912 Good Very good General Very good
Ideal 2.000 Best Best Best Best
Table 151.3 Status of operation level U3
Status U31 U32 U33 U34 U35 U36 U37 U38 U39
1 2.450 0.307 0.121 0.420 0.545 0.266 12.400 0.069 Good
2 0.436 0.073 0.028 0.378 –0.267 –0.245 11.178 0.068 General
3 0.776 0.188 0.075 0.407 0.613 0.816 11.305 0.067 Good
4 1.563 0.255 0.091 0.358 –0.277 –0.577 6.884 0.095 Good
Ideal 1.000 0.100 0.030 0.500 0.200 0.100 10.000 0.050 Best
Table 151.4 Status of R&D level U4
Status U41 U42 U43
1 19435331 66500 General
2 29018320 76110 Bad
3 38857798 1076920 Bad
4 12724180 49268370 Bad
Ideal 50000000 50000000 Best
151 An Evaluation Method Based on Cloud Model 1433
Table 151.5 Status of network position U5
Status U51 U52 U53 U54
1 General Very good Very good Best
2 General Good Best Best
3 Bad Very good Very good Best
4 General Good Very good Best
Ideal Best Best Best Best
151.3.2 Denoting the Cloud Model of Each Index
Normalize the evaluation languish set (bad, general, good, very good, best) to (0,
0.25, 0.50, 0.75, 1), and thus the policy matrix A1 –A5 is constituted as follow:
2 3
37920 33470815 0:25 0:50
6 37920 33470815 0:25 0:50 7
A1 ¼ 64 37920 33470815 0:50 0:50 5
7
37920 252764848 0:50 0:75
2 3
1:701 0:50 0:75 0:50 0:50
6 1:789 0:50 0:50 0:50 0:50 7
A2 ¼ 64 1:660 0:25 0:50 0:50 0:50 5
7
1:912 0:50 0:75 0:25 0:75
2 3
2:450 0:307 0:121 0:420 0:545 0:266 12:400 0:069 0:50
6 0:436 0:073 0:028 0:378 0:267 0:245 11:178 0:068 0:25 7
6
A3 ¼ 4 7
0:776 0:188 0:075 0:407 0:613 0:816 11:305 0:067 0:50 5
1:563 0:255 0:091 0:358 0:277 0:577 6:884 0:095 0:50
2 3 2 3
19435331 66500 0:25 0:25 0:75 0:75 1
6 29018320 76110 0 7 6 0:25 0:50 1 17
A4 ¼ 6
4 38857798
7 A5 ¼ 6 7
1076920 0 5 4 0 0:75 0:75 15
12724180 49268370 0 0:25 0:50 0:75 1
Expectation Ex and entropy En of each index cloud model are calculated by the
above policy matrixes (see Tables 151.6, 151.7, 151.8, 151.9 and 151.10).
Table 151.6 Expectation Ex and entropy En of credit quality U1
Parameter U11 U12 U13 U14
Ex 37920 88294323.3 0.375 0.563
En 0 36549005.5 0.042 0.042
Table 151.7 Expectation Ex and entropy En of organizational level U2
Parameter U21 U22 U23 U24 U25
Ex 1.766 0.438 0.625 0.438 0.563
En 0.042 0.042 0.042 0.042 0.042
1434 G. Zhou et al.
Table 151.8 Expectation Ex and entropy En of operation level U3
Parameter U31 U32 U33 U34 U35 U36 U37 U38 U39
Ex 1.306 0.206 0.079 0.391 0.154 0.065 10.442 0.075 0.438
En 0.336 0.039 0.016 0.010 0.148 0.232 0.919 0.005 0.042
Table 151.9 Expectation Ex and entropy En of R&D level U4
Parameter U41 U42 U43
Ex 25008907.25 12621975 0.063
En 4355603 8200311.667 0.042
Table 151.10 Expectation Ex and entropy En network position U5
Parameter U51 U52 U53 U54
Ex 0.188 0.625 0.813 1
En 0.042 0.042 0.042 0
Table 151.11 The power weights of indexes
wi wi1 wi2 wi3 wi4 wi5 wi6 wi7 wi8 wi9
0.36 0.58 0.14 0.07 0.21 – – – – –
0.16 0.11 0.19 0.19 0.19 0.33 – – – –
0.23 0.17 0.17 0.07 0.07 0.09 0.09 0.12 0.12 0.11
0.13 0.25 0.59 0.17 – – – – – –
0.12 0.56 0.11 0.26 0.07 – – – – –
151.3.3 Confirming the Power Weight of Indexes
The power weights of indexes are shown in Table 151.11.
151.3.4 Result and Analysis
Through cloud model computation, the credit evaluation value of high-tech SMEs:
X
5
PS1 ¼ ðwi PUi Þ ¼ 0:487
i¼1
Then the credit evaluation value is input into the five-scale evaluation set based
on cloud model (see Fig. 151.3). It will activate two cloud objects: A and B, but
the activation degree of A is far larger than B, so the credit evaluation of company
S1 (NO. 300002) obtains A.
151 An Evaluation Method Based on Cloud Model 1435
Fig. 151.3 Result of credit
evaluation of company S1
151.4 Conclusion
In this paper, credit evaluation applications in high-tech SMEs are discussed, and
an evaluation method based on cloud model is formulated. The application of this
model is also illustrated. After this research, we find that it is a better way to use
this method to realize transforming qualitative terms described in a natural lan-
guage to distribution patterns of quantitative values, especially for high-tech
SMEs.
Acknowledgments This work was partially supported by the National Natural Science Foun-
dation of China (Grants No. 71172148) and Soft Science Research Projects of the Ministry of
Housing and Urban-Rural Construction (Grants No. 2011-R3-18). The authors are also grateful to
the referees for their helpful comments and valuable suggestions for improving the earlier version
of the paper.
References
Chen X, Han W, She J (2006) The credit risk, company quality and growth illusion. Econ Theory
Bus Manage 12:62–67
Chen X, Wang X, Wu DD (2010) Credit risk measurement and early warning of SMEs: an
empirical study of listed SMEs in China. Decis Support Syst 49:301–310
CSMAR Solution (2012) CSMAR database. http://www.gtarsc.com
Derelioglu G, Gürgen F (2011) Knowledge discovery using neural approach for SME’s credit risk
analysis problem in Turkey. Expert Syst Appl 38(8):9313–9318
Deyi L, Yi D (2005) Artificial intelligence with uncertainty. Chapman & Hall, Boca Raton
Deyi L, Haijun M, Xuemei S (1995) Membership clouds and membership cloud generators.
J Comput Res Dev 32(6):15–20
Di K, Deyi L, Deren L (1999) Cloud theory and its applications in spatial data mining knowledge
discovery. J Image Graph 4A(11):930–935
Fang W, Yanpeng L, Xiang L (2007) A new performance evaluation method for automatic target
recognition based on forward cloud. In: Proceedings of the Asia simulation conference,
pp 337–345
1436 G. Zhou et al.
Frydman H, Altman EI, Kao DL (1985) Introducing recursive partitioning for financial
classification: the case of financial distress. J Finance 40(1):269–291
Haijun W, Yu D (2007) Spatial clustering method based on cloud model. In: Proceedings of the
fourth international conference on fuzzy systems and knowledge discovery, no. 7, pp 272–276
Jensen HL (1992) Using neural networks for credit scoring. Manag Finance 18(6):15–26
Li DY, Liu CY, Gan WY (2009) A new cognitive model: cloud model. Int J Intell Syst
24:357–375
Lv P, Yuan L, Zhang J (2009) Cloud theory-based simulated annealing algorithm and application.
Eng Appl Artif Intell 22:742–749
Walker RF, Haasdijk E, Gerrets MC (1995) Credit evaluation, using a genetic algorithm. In:
Intelligent systems for finance and business, Wiley, US, pp 39–59
Yunfang Z, Chaohua D, Weirong C (2005) Adaptive probabilities of crossover and mutation in
genetic algorithms based on cloud generators. J Comput Inf Syst 1(4):671–678
Chapter 152
The Structural Optimum Design
of Erected Circular Medicine-Chest Based
on Non-Intervention Motion
Zhi-qiang Zhang, Chao Yun, Xiang-quan Liu and Li-yong Wang
Abstract In this paper, the following studies are completed, including the analysis
of the mechanical structure of the erected circular medicine-chest as well as its
working principle, discussion of the non-intervention motion conditions of drug
containers according to their different motion phases, establishment of the optimum
functions of the chain transmission based on the conditions of the non-intervention
motion of the containers, obtention of the minimum circular radius of drug con-
tainers by application of Matlab as well as the obtention of the design parameters of
chain transmission, the simulation model building with UG and ADAMS, and the
accomplishment of the motion simulation to prove the success of optimum design
based on the non-intervention motion condition of drug containers.
Keywords Erected circular medicine-chest
Optimum functions Non-
intervention motion conditions Motion simulation
152.1 Introduction
Currently, with the limitation regarding technology, the drug storing system in a
hospital pharmacy mainly consists of ordinary shelves, unable to realize the dense
storage. In most of the hospital pharmacy, the facilities are old, working envi-
ronment poor and pharmacists’ working intensity high. In addition, there are other
problems such as storage complexity, space waste, low working efficiency, etc.
Z. Zhang (&) C. Yun
Institute of Robotics, Beijing University of Aeronautics and Astronautics,
100191 Beijing, China
e-mail: [email protected]
Z. Zhang X. Liu L. Wang
School of Mechanical and Electrical Engineering, Beijing Information Science and
Technology University, 100192 Beijing, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1437
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_152,
Ó Springer-Verlag Berlin Heidelberg 2013
1438 Z. Zhang et al.
The introduction of automated pharmacy system may help to make a better overall
plan for the pharmacy, including reducing the drug storage area, effectively car-
rying out the standardized and automated management of the pharmacy, thus
improving drug dispensing and reducing patients’ waiting time. The automated
pharmacy system may be connected to HIS, putting all the working process of the
working staff under the supervision (Li et al. 2007; Liu et al. 2009; Zhao 2009).
The typical equipment is erected circular Medicine-chest, which is originated
from the digital-controlled erected circular inventory for the management and
storage of accessories and tools in large factories. Basing on the stereoscopic
inventory, when adding safety security, humanization design and connecting with
the HIS system of hospital, it can be sued for the pharmacy management in the
hospital. This is a semi-automatic system, in which the drug dispensary works
should be done artificially. This type of automatic system is adaptable to the
package of drugs, and can be used as the supplementary equipment to the drug
dispensing machine for box-packed drugs as effective assisting (Zhao et al. 2008).
152.2 The Working Principle of the Erected Circular
Medicine-Chest
In operation, after two rate reductions, one by reducer, the other by the first chain
transmission, the motor of the erected circular medicine-chest drives the two
driving chain wheels on the synchronizing shaft, connected to which are two
driven wheels fixed on the two half axles accordingly. Driven by the chains, the
support rods and balance bars fixed on them make all the drug containers move
circularly. Drugs in the movable containers, after receiving the dispensing order,
are conveyed along the shortest path and reached the outlets within the shortest
time.
152.3 The Analysis of the Conditions in Regard
to Non-Intervention Motion of Neighbouring
Drug Containers
As shown in Fig. 152.1, the pitch of chain 084 A: 12.7 mm; the span between the
two support rods connected to the same container: 18 pitches; the span between
the support rods respectively connected to the neighbouring containers: 4 pitches;
the number of containers (evenly arranged along the chains): 12.
The pitch radiuses of both driving wheels and driven wheels: r; the overhang of
the support rods: c; the width and height of the containers W: and h, respectively;
the space between the centers of the pivots of the two neighbouring drug
152 The Structural Optimum Design of Erected Circular Medicine-Chest 1439
Fig. 152.1 The structure of
the erected circular medicine-
chest
containers: H; the fixation range for the connection beam: A; the safety space
between the connection beam and the drug container: S.
As shown in Fig. 152.2, with the center of driven wheel as the origin of
coordinate, the reference coordinate is built; O1, O2, O3, O4 are, respectively, the
joint points of the transmission chain with the Supporting Rods No. 1, 2, 3 and 4
for the Drug Containers A1, A2, whose coordinates are ðx1 ; y1 Þ, ðx2 ; y2 Þ, ðx3 ; y3 Þ,
ðx4 ; y4 Þ accordingly. A1 is the center of the pivot of Container 1 and Supporting
Rods No. 1 and 2, while A2 is the center of the pivot of Container 2 and Supporting
Fig. 152.2 The motion
initial positions of the two
neighbouring drug containers
1440 Z. Zhang et al.
Rods No. 3 and 4. The coordinates of A1, A2 are ðX1 ; Y Þ1 , ðX2 ; Y2 Þ, respectively.
The wheels do counterclockwise rotation.
Suppose located in the pitch circle is Point B, which is on the same horizontal
line with Center O of the chain wheel, coinciding with O1, O2, O3, O4 are on the
same vertical line. At the beginning of the container motion, O1 moves unclock-
wise along the circle with r as the radius, while O2, O3, O4 move upward verti-
cally. When O2 reaches the same height as that of Point O, O1 and O2 move along
the same circle. h0 is the included angle between OO1 and OO2 when O1, O2 move
along the same circle simultaneously; h1 is the angle between OB and X axis; h2 is
the included angle between OO2 and OO3 when O2, O2 move along the same circle
simultaneously.
Suppose, at the motion of Containers 1 and 2, DX and DY are the horizontal
space and the vertical space of the Centers of Pivots A1 and A2, the condition of
nonintervention motion is (Zheng 1992):
if 0 DX w,
jDY j [ h; ð152:1Þ
When DX [ w, there must be no intervention between the two drug containers.
152.4 The Optimum Design of Chain Transmission Based
on Non-Intervention Motion
152.4.1 The Mathematical Model of the Optimum Design
of Chain Transmission
152.4.1.1 The Idea of the Optimum Design of Chain Transmission
The floor area of the erected circular medicine-chest refers to the projected area of
the outline of the steelwork, whose length is affected by such factors as the length
of the drug containers, while whose width by such factors as the turning radius and
the width of the drug containers. After the design of the drug containers, the
following references may be given, including the width w and height h of the
container, as well as the pitch of the chain. In the design of the chain transmission,
available are the best overhang of the support rods c and the pitch radius of the
chain wheel r based on the optimum design, thus also available is the minimum
turning radius of the container, whose outline of the steelwork is smallest in width.
If the length remains the same, the floor area will be reduced, also reduced will be
the mechanical deformation of the steelwork.
152 The Structural Optimum Design of Erected Circular Medicine-Chest 1441
152.4.1.2 The Design Variables and Objective Functions
Based on the analysis of the condition of nonintervention motion, h1 is a
independent variable, if the coefficient k is introduced, and c ¼ kr, then k and r
will be variables, so:
X ¼ ½x1 x2 x3 T ¼ ½k r h1 T
To make available the smallest width of the outline of the steelwork, the
Turning Radius L is taken as the objective function, then:
L ¼ c þ r ¼ r ð 1 þ kÞ ð152:2Þ
Based on the above, the expression of the objective function of the optimum
design is:
f ðxÞ ¼ x2 ð1 þ x1 Þ
152.4.1.3 Constraints
The vertical nonintervention motion of the containers on the left and those on
the right:
2L w þ A þ 2s
In the practical design, selected are the following data:
W = 420 mm, A = 30 mm, s = 87 mm
*L ¼ c þ r ¼ r ð1 þ kÞ;
) rð1 þ kÞ 0:312 m
` Since the container forces on the support rods, the overhang of the support
rod should not be too long in order to ensure the enough strength of the chain. In
the design:
0:15 m c 0:25 m
le 0:15 m r k 0:25 m ð152:3Þ
´ The diameter of the chain pitch circle:
p
d¼ ð152:4Þ
sin 180
z
1442 Z. Zhang et al.
: the pitch of the chain, here p ¼ 12:7 mm, z : 40 60; 60;
0:081 m r 0:121 m ð152:5Þ
ˆ The space between the centers of the pivots of the two neighbouring drug
containers:
{ p ¼ 12:7 mm,
; H (the space between the centers of the pivots of the two neighbouring drug
containers moving vertically) must meet the following condition:
H h þ h0
h0 : the gap between the two neighbouring drug containers moving vertically,
generally h0 3 mm.
Then, according to the structure of the chain, H ¼ np.
n: the number of the pitches whose total length amounts to h, i.e. the height of
the container.
Here h ¼ 275 mm, and it may be calculated: n 22. If n ¼ 22,the total number
of the pitches will be 264.
)H ¼ 279:4 mm
˜ h0; h2 :
n1 360 6:35
h0 ¼ ¼ 36 arc sin ð152:6Þ
z r
n2 360 6:35
h2 ¼ ¼ 8 arc sin ð152:7Þ
z r
n1: the number of the pitches whose total length amounts to the distance
between O1 and O2; n1: the number of the pitches whose total length amounts to
the distance between O2, O3.
Þ Based on Formulae (152.3) and (152.4), it may be deduced:
1:24 k 3:08
þ The scope of: 0 h1 p þ 2h0 þ h2
152.4.1.4 The Mathematical Model
Omitted here are the track equations of the centers of Pivots and that of the guide
rail of the balance bars
From the above, Variables x1 , x2 and x3 may be substituted in the formula, then
the optimum mathematical model of the chain transmission based on the
152 The Structural Optimum Design of Erected Circular Medicine-Chest 1443
conditions of the Non-intervention Motion of the containers may be expressed as
follows (Liu 2005; Peng 2007; Wang 2005):
min f ðxÞ ¼ x2 ð1 þ x1 Þ
g1 ðxÞ ¼ 0:312 x2 ð1 þ x1 Þ
g2 ðxÞ ¼ 0:15 x2 x1
g3 ðxÞ ¼ x1 x2 0:25
When 0 h1 h0 ,
g4 ðxÞ ¼ DX1
g5 ðxÞ ¼ DX1 420
g6 ðxÞ ¼ 275 Y1 þ x2 x3 393:7
When h0 \h1 h0 þ h2 ,
g4 ðxÞ ¼ x2 þ R cos ðx3 h0 =2Þ
g5 ðxÞ ¼ x2 R cos ðx3 h0 =2Þ 420
g6 ðxÞ ¼ 275 R sin ðx2 h0 =2Þ þ x2 x3 393:7
When h0 þ h2 \h1 p,
g4 ðxÞ ¼ DX3
g5 ðxÞ ¼ DX3 420
g6 ðxÞ ¼ 275 DY3
When p\h1 h0 þ ðh2 þ pÞ=2,
g4 ðxÞ ¼ DX4
g5 ðxÞ ¼ DX4 420
g6 ðxÞ ¼ 275 DY4
When p\h1 2h0 þ h2 ,
g4 ðxÞ ¼ DX5
g5 ðxÞ ¼ DX5 420
1444 Z. Zhang et al.
g6 ðxÞ ¼ 275 DY5
g7 ðxÞ ¼ 1:24 x1
g8 ðxÞ ¼ x1 3:08
g9 ðxÞ ¼ 0:081 x2
g10 ðxÞ ¼ x2 0:121
g11 ðxÞ ¼ x3
g12 ðxÞ ¼ x3 p 2h0 h2
152.4.2 The Optimization Based on Matlab
It may be seen from the mathematical model, the optimum design belongs to that
of the constrained nonlinear optimization (Su et al. 2004). The Matlab functions to
solve the above problem of the constrained nonlinear optimization are FMINCON.
The calculation results based on Matlab are:
k ¼ 1:7609; r ¼ 113 mm; Lmin ¼ 312 mm
According to formula (152.8), z = 55.88, rounded for z = 56, then
p
d¼ ¼ 226:5 mm
sin 180
z
i.e. r ¼ 113:25 mm
According to Formula (152.2), if c = 199.5 mm, the turning radius of the
container L = 312.75 mm.
The results of the optimum design include the pitch of the chain: 12.7 mm, the
overhang of the support rod: 199.5 mm, the number of the pitches whose total length
amounts to the height of the container: 22, the total number of the pitches: 264, the
diameter of the pitch: 226.5 mm, the teeth number of the chain wheel: 56, the length
of the chain: 3352.8 mm, the theoretical center distance: 1320.8 mm
152.5 The Simulation Analysis Based on UG and ADAMS
According to the results of the optimum design, by applying UG/Model, built up
are the models of the chain wheels, chain, containers and support rods etc. The
solid model for the analysis of the container motion is built up as shown in
Fig. 152.3 (Lijun and Qin 2002).
152 The Structural Optimum Design of Erected Circular Medicine-Chest 1445
Fig. 152.3 The solid model
for the analysis of the
container motion
According to the test results of the motion simulation intervention, based on the
conditions of the Non-intervention Motion of the containers, feasible are the
results that the turning radius of the container L = 312.75 mm and the horizontal
overhang of the support rods c = 199.5 mm.
152.6 Conclusion
The analysis is made on the working principle of the erected circular medicine-chest,
built up are the conditions in regard to non-intervention motion of neighbouring drug
containers, structure optimization is implemented by the application of Matlab,
feasibility of the above mentioned structural optimum design is established by the
analysis based on modeling and simulation.
Acknowledgments This work was completed with the support of the project of national natural
science foundation of china (No. 51105041)
References
Li C, Wang W, Yun C (2007) Status quo and new development of automated pharmacy. Robot
Technol Appl China 5:27–32
Lijun XJ, Qin W (2002) ADAMS tutorial example. Beijing Institute of Technology Press, Beijing
Liu W (2005) Design of vertical carousel automation based on solid works. Shandong University,
Jinan
Liu X, Yun C, Zhao X, Wang W, Ma Y (2009) Research and design on automated pharmacy.
Chin J Mach Des 26(7):65–68
Peng G (2007) Study and design of vertical circulation cubic garage based on CAE. Shandong
University, Jinan
Su J, Zhang L, Liu B (2004) MATLAB toolbox application. Electronic Industry Press, Beijing
Wang J (2005) Non-intervention motion conditions for vertical circulation of parking equipment
carrier. Machinery 32(3):P22–P23
1446 Z. Zhang et al.
Zhao T (2009) Hospital pharmacy automation is the inevitable trend of pharmacy development.
Capital Med China 16(24):31
Zhao X, Yun C, Liu X, Wang W, Gao Z (2008) Research on the automated pharmacy system.
Chin J Sci Instrum 29(4):193–195, 200
Zheng Z (1992) Chain transmission design and the application. China Machine Press, Beijing
Chapter 153
Application of Noise Estimator
with Limited Memory Index on Flexure
Compensation of Rapid Transfer
Alignment
Wei-dong Zhou and Yu-ren Ji
Abstract In order to solve the flexure compensation problem in rapid transfer
alignment, the error equations are simplified by noise compensation method firstly.
Due to the time variant characteristics of flexure process in time domain, which
leads to the fixed noise statistical characteristics cannot follow the variation of
actual environment, the noise estimator with limited memory index is proposed.
By limiting the memory length of obtained data, too old historical data is giving up
and the accuracy of online noise estimator is improved. The final simulation
verifies that the method proposed have higher accuracy and faster convergence
speed than conventional methods.
Keywords Flexural deformation Limited memory Noise compensation
Transfer alignment
153.1 Introduction
Transfer alignment is an important technology of moving base alignment. In the
process of transfer alignment, the maneuver of carrier is required and the running
environment is getting complicated. As a result, the flexural deformation is one of
the most important error sources (Kain and Cloutier 1989; Spalding 1992; Wendel
et al. 2004; Xiao and Zhang 2001).
There are mainly two methods for the compensation of flexural deformation, the
model construction method and noise compensation method. Because of the high
complexity of physical modeling, most researches are turned to the experimental
modeling, which uses the markov process to describe the random flexural
W. Zhou Y. Ji (&)
Department of Automation, Harbin Engineering University, Harbin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1447
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_153,
Ó Springer-Verlag Berlin Heidelberg 2013
1448 W. Zhou and Y. Ji
deformation. The flexure motion is modeled as the third-order markov model
driven by the white noise through analysis of experimental data in literature (Kain
and Cloutier 1989). On the base of above job, the markov process is further
divided into high frequency mode and low frequency mode by literature (Spalding
1992). In literature (Xiao and Zhang 2001), the elastic deformation of aircraft wing
is modeled as two-order markov process. The model construction method is
realized by computing the Markov parameters based on the similarity principle of
power spectral density (Jones et al. 1993; Lim and Lyou 2001; Ross and Elbert
1994; Robert 1996). However, modeling of Markov process will lead to rapid
increase of state dimension and computation burden, so a suboptimal filter without
the third-order markov model is proposed in literature (Kain and Cloutier 1989),
where the strength of system noise is enhanced to compensate the uncertainty of
flexural deformation, which is just the noise compensation method. Comparing to
model construction method, the complexity of noise compensation method is
reduced and the robustness is improved, but the accuracy also declines. It is mainly
attributed to the variation of actual noise because of the environmental change
while the noise statistics is set to a fixed value in the filter. Adaptive filter can be
used to solve this problem (Sage and Husa 1969; Qi and Han 2008; Xiong et al.
2006, 2007). A noise estimation algorithm based on maximum likelihood is pro-
posed for linear system in literature (Mohamed 1999). Noise variance estimator is
designed based on EM principle in literature (Bavdekar et al. 2011). An nonlinear
noise estimation algorithm is proposed based on maximum-posterior likelihood in
literature (Zhao and Wang 2009). Aiming at the rapid time-variant characteristics
of flexural deformation, the limited memory index is combined with the noise
estimator. By limitation of memory length, the old history data is abandoned to
improve the precision of online noise estimator.
In this paper, the system error equation is simplified by noise compensation
method firstly. Then the adaptive filter with limited memory index is designed.
Finally the effectiveness of the algorithm is verified by simulation.
153.2 System Error Model
153.2.1 Velocity Error Model
The velocity differential equation of master inertial navigation system (MINS) is
given by
V_ mn ¼ Cmn fmm 2xnie þ xnen Vmn þ gnm ð153:1Þ
The velocity differential equation of slave inertial navigation system (SINS) is
given by
n
V_ sn ¼ Csn fss 2x ~ nen Vsn þ gns
~ ie þ x ð153:2Þ
153 Application of Noise Estimator 1449
where the relations between the variables can be defined by
fss ¼fms þ fls þ asf þ rs dV ¼ Vsn Vmn Vrn
ð153:3Þ
dV_ ¼V_ sn V_ mn V_ rn x
~ nie xnie x
~ nen xnen gns gnm
By inserting (153.3) to the difference between (153.1) and (153.2), leading to
V_ sn V_ mn ¼ Csn fss Csn Cms Csm fss fls asf rs 2xnie þ xnen Vsn Vmn
ð153:4Þ
The lever-arm velocity and its differential equation are given by
Vrn ¼Cin Csi xsis r s
ð153:5Þ
V_ rn ¼Cin xini Vri þ Csn xsis xsis r s þ x_ sis r s
where the term of xsis xsis r s þ x_ sis r s is lever-arm acceleration, which can be
written as
fls ¼ xsis xsis r s þ x_ sis r s ð153:6Þ
So (153.5) is given by
V_ rn ¼ Csn fls xnin Vrn ð153:7Þ
The Coriolis term is given as
n
2xie þ xnen Vsn Vmn xnin Vrn þ dV ð153:8Þ
When physical misalignment wa and measurable misalignment wm are small,
their direction cosine matrix can be written as
Cms ¼ I wm
ð153:9Þ
Csm ¼ I þ wa
Using (153.7) (153.8) and (153.9), reducing two-order small terms, (153.4) can
be given by
n
dV_ ¼ Csn ðwm wa Þ fss 2x ~ nen dV þ Csn asf þ rs
~ ie þ x ð153:10Þ
153.2.2 Attitude Error Model
Differential equation of direction cosine matrix of wm can be written as
C_ sm ¼ Csm xsms ð153:11Þ
1450 W. Zhou and Y. Ji
where
xsms ¼ w_ m ð153:12Þ
Expansion and differentiation of (153.11) results in
C_ sm ¼ Cnm xnmn Csn þ Cnm Csn x
~ sns ð153:13Þ
By inserting (153.11), (153.12) and (153.13) leading to
w_ m ¼ Cns xnmn Csn þ x
~ sns ð153:14Þ
thus
w_ m ¼ Cms Csm xsnm þ x
~ sns ð153:15Þ
where
~ sns ¼ xsis x
x ~ sin ð153:16Þ
~ sin is instruction angular velocity computed by SINS, xsis is measured angular rate
x
which can be written as
xsis ¼ xsim þ xsf þ es ð153:17Þ
Using (153.16), (153.17), (153.15) becomes
w_ m ¼ Cms Csm xsnm þ xsf þ es þ xsim x
^ sin ð153:18Þ
where
xsim x
^ sin xsnm x
~ sns ð153:19Þ
According to (153.9), by inserting (153.19) to (153.18) leading to the attitude
error model
w_ m ¼ ðwm wa Þx
~ sns þ xsf þ es ð153:20Þ
153.2.3 Inertial Instrument Error Model
The error model of accelerometer and gyro are composed by constant drift and
white noise, which can be written as
r ¼ rc þ xa ; r_ c ¼ 0 ð153:21Þ
e ¼ ec þ xg ; e_ a ¼ 0 ð153:22Þ
153 Application of Noise Estimator 1451
153.2.4 Analysis of System Error Model
According to the error model of small misalignments, the state equation and
measurement equation are linear equation with additive noise, whose discrete
general formula can be given by
(
xk ¼Uk1 xk1 þ wk1
ð153:23Þ
zk ¼Hk xk þ vk
where xk is the state of n 1 vector and zk is the observable variable of m 1
vector. Process noise wk and measurement vk are zero mean white noise and
uncorrelated, whose prior statistical characteristics can be expressed as
(
E½wk ; wTj ¼Qk dkj ;
ð153:24Þ
E½vk ; wTj ¼Rk dkj
where dkj is the function of Kronecker-d. The flexure process af and xf are chosen
as the state variables
X ¼ ½ dV wm wa r e af xf ð153:25Þ
When Markov model is used to describe the flexure process, the dimension of
state will increase rapidly. If the east channel and west channel are considered and
the two-ordered Markov model is adopted, the required state variables will be 8. If
all the three channels are described by three-ordered Markov model, the required
state variables will be 18. So the system model needs to be simplified and the state
equation can be written by
xk ¼ Uk1 xk1 þ DUk1 þ wk1 ð153:26Þ
where DUk1 can be written by
" #
Csn asf
DUk1 ¼ ð153:27Þ
xsf
So the process noise is adjusted as
wk1 ¼ DUk1 þ wk1 ð153:28Þ
In time domain, the uncertainty caused by flexure process is presented as zero
mean oscillation curve, which means the process noise obeys the statistical
characteristics of zero mean and unknown variance. So the compensated noise
needs to be estimated online.
1452 W. Zhou and Y. Ji
153.3 Design of Adaptive Filter Based on Noise Estimator
with Limited Memory Index
153.3.1 Classic Kalman Filer
When error model is built accurately and the system noise can be obtained cor-
rectly, the classic Kalman filter can be written by
(1) Set of initial value
^x0 ¼ Eðx0 Þ
ð153:29Þ
P0 ¼ E ðx0 ^x0 Þðx0 ^x0 ÞT
(2) Time and measurement update
^xk;k1 ¼ Uk1^xk1 ð153:30Þ
Pk;k1 ¼ Uk1 Pk1 UTk1 þ Qk1 ð153:31Þ
^zk;k1 ¼ Hk ^xk;k1 ð153:32Þ
(3) State update
1
Kk ¼ Pk;k1 HkT Hk Pk;k1 HkT þ Rk ð153:33Þ
Pk ¼ ½I Kk Hk Pk;k1 ð153:34Þ
^xk ¼ ^xk;k1 þ Kk zk zk;k1 ð153:35Þ
153.3.2 Noise Estimator Based On Limited Memory Length
According to the analysis of Sect. 153.2, the variance of compensated noise is
unknown, so the noise estimator will be designed in this section. The process noise
estimator based on maximum likelihood principle can be given by
Xk n h T i o
^k ¼ 1
Q Ki zi hi^xi;i1 zi hi^xi;i1 KiT Ui1 Pi1 UTi1 þ Pi
k i¼1
ð153:36Þ
Because the observability needs to be increased by adopting required maneu-
vers, the running environment will become complicated and the statistical char-
acteristics of compensated noise vary rapidly, as a result the effectiveness of too
old history data will become weak and even negative. So the length of history data
153 Application of Noise Estimator 1453
is limited to improve the accuracy of process noise estimator, which will be
expressed as follows. Firstly the weighted coefficient fbi g is given by
X
m
bi ¼ 1; bi ¼ bi1 b ð153:37Þ
i¼1
where m is the memory length, b is forgetting factor. According to (153.37), bi can
be further deduced as
bi1 bi
bi ¼ ð153:38Þ
1 bm
By inserting (153.38) to (153.36), leading to
X
k n h T i o
^k ¼ 1
Q bkþ1i Ki zi hi^xi;i1 zi hi^xi;i1 KiT Ui1 Pi1 UTi1 þ Pi
k i¼1
ð153:39Þ
So the process noise estimator based on limited memory length is given by
n h i o
^ k1 þ 1 b Kk zk Hk ^xk;k1 zk Hk ^xk;k1 T K T Uk1 Pk1 UT þ Pk
^ k ¼bQ
Q k k1
1 bm
bm bmþ1 n h T i T
Kkm zkm Hkm^xkm;km1 zkm hkm^xkm;km1 Kkm
1 bm
Ukm1 Pkm1 UTkm1 þ Pkm
ð153:40Þ
where the memory length m can be adjusted according to actual environment.
153.4 Simulation and Analysis
The rapid transfer alignment of shipboard aircraft is simulated, where the swing
maneuver of ship is driven by sea wave. To obtain better observability, velocity/
attitude matching is selected whose measurement equation is given by
z ¼ Hx þ v ð153:41Þ
According to Sect. 153.2, after compensation of process noise, the state can be
given by
X ¼ ½dV wm wa r e ð153:42Þ
1454 W. Zhou and Y. Ji
So the observing matrix can be written as
I22 023 028
H¼ ð153:43Þ
032 I33 038
The initial position of ship is 45.6° north latitude and the velocity is Vn = 10 m/s.
The ship move northward and the swinging model driven by the sea wave can be
expressed as
hx ¼ hxm sinðxx t þ hx0 Þ
hy ¼ hym sin xy t þ hy0 ð153:44Þ
hz ¼ hzm sinðxz tÞ þ hz0
The swinging amplitude hxm , hym , hzm are 5°, 4°, 2°, the frequency xx , xy , xz are
0.18, 0.13, 0.06 Hz and initial angle hx0 , hy0 , hz0 are all set 0°. The gyro constant drift
of SINS is set 0.05°/h and accelerometer offset is set 10-3 g. The variance (vari-
ances) of white noise of gyro and accelerometer are set respectively (0.001°/h)2 and
(10-4 g)2. Misalignments wx , wy , wz are 150 , 300 , 1°. Simulation time is 100 s.
The two-ordered Markov model is adopted for the true flexural deformation and
the model coefficients of three channels are set bx ¼ 0:1, by ¼ 0:2, bz ¼ 0:4,
simultaneously the variance of white noise are (0.05/h)2 and (10-3 g)2. Scheme 1
takes the same model but with different coefficients, which are set bx ¼ 0:2,
by ¼ 0:3, bz ¼ 0:5. Noise compensation method is used in Scheme 2, but the
statistical characteristics of noise is set a fixed value, in which the compensation
coefficient is set 1.5. The process noise estimator with limited memory index is
used in Scheme 3, the memory length m is set 10 and forgetting factor b is set 0.3.
The filter frequency is set 5 Hz. Estimation results of three schemes are shown
from Figs. 153.1, 153.2, and 153.3.
There are slight deviations in model coefficients between true model and
Scheme 1. However, it can be seen from the simulation results that in Scheme 1
20
Scheme 1
15
Scheme 2
10 Scheme 3
5
ψx/'
-5
-10
-15
-20
0 10 20 30 40 50 60 70 80 90 100
Time/s
Fig. 153.1 Estimation error of misalignment wx
153 Application of Noise Estimator 1455
10
5
0
ψy/' -5
-10
-15
-20 Scheme 1
-25 Scheme 2
Scheme 3
-30
0 10 20 30 40 50 60 70 80 90 100
Time/s
Fig. 153.2 Estimation error of misalignment wy
20
10
0
-10
-20
ψ z/'
-30
-40
-50 Scheme 1
Scheme 2
-60
Scheme 3
-70
0 10 20 30 40 50 60 70 80 90 100
Time/s
Fig. 153.3 Estimation error of misalignment wz
the standard deviations of three misalignments are 0.320 , 0.670 , 2.190 . But limited
to the high computation burden, the convergence speed declines. In Scheme 2,
because of the reduction of state, the convergence speed is improved. However,
the cost is the declining of filter accuracy since the fixed noise statistical charac-
teristics cannot follow the variation of actual environment. The standard deviations
of three misalignments are 1.430 , 3.560 , 12.530 . On the basis of Scheme 2,
Scheme 3 uses noise estimator with limited memory index to real-time track the
system noise. The standard deviations of three misalignments are 0.280 , 0.760 ,
2.550 , which are better than Scheme 2 and faster than Scheme 1.
153.5 Conclusion
After model building of rapid transfer alignment, the error equations are simplified
by noise compensation method. Aiming at the time variant characteristics of
flexure process in time domain, the noise compensation problem of flexural
1456 W. Zhou and Y. Ji
deformation is transformed to the problem of online estimation of system noise,
which is dealt with the noise estimator based on limited memory length. The final
simulation shows that, when external interference cannot be obtained accurately,
the method proposed by this paper can provide a new idea for the compensation of
flexural deformation.
References
Bavdekar VA, Deshpande AP, Patwardhan SC (2011) Identification of process and measurement
noise covariance for state and parameter estimation using extended Kalman filter. J Process
Control 21(4):585–601
Jones D, Roberts C, Tarrant D (1993) Transfer alignment design and evaluation environment. In:
IEEE proceedings of aerospace control systems, pp 753–757
Kain JE, Cloutier JR (1989) Rapid transfer alignment for tactical weapon application. In:
Proceedings of the AIAA guidance, navigation and control conference, Boston, pp 1290–1300
Lim Y-C, Lyou J (2001) An error compensation method for transfer alignment. In: Proceedings
of IEEE conference on electrical and electronic technology. TENCON, vol 2, pp 850–855
Mohamed AH (1999) Adaptive Kalman filtering for INS/GPS. J Geodesy 73(4):193–203
Qi S, Han J-D (2008) An adaptive UKF algorithm for the state and parameter estimation of a
mobile robot. Acta Automatica Sinica 34(1):72–79
Robert MR (1996) Weapon IMU transfer alignment using aircraft position from actual flight tests.
In: Proceedings of IEEE position location and navigation symposium, pp 328–335
Ross CC, Elbert TF (1994) A transfer alignment algorithm study based on actual flight test data
from a tactical air-to-ground weapon launch. In: Proceedings of IEEE position location and
navigation symposium, pp 431–438
Sage AP, Husa GW (1969) Adaptive filtering with unknown prior statistics. In: Joint automatic
control conference, Colombia, pp 760–769
Spalding K (1992) An efficient rapid transfer alignment filter. In: Proceedings of the AIAA
guidance, navigation and control conference, pp 1276–1286
Wendel J, Metzger J., Trommer GF (2004) Rapid transfer alignment in the presence of time
correlated measurement and system noise. In: AIAA guidance, navigation, and control
conference and exhibit, Providence, RI, pp 1–12
Xiao Y, Zhang H (2001) Study on transfer alignment with the wing flexure of aircraft. Aerosp
Control 2:27–35
Xiong K, Zhang HY, Chan CW (2006) Performance evaluation of UKF-based nonlinear filtering.
Automatica 42(2):261–270
Xiong K, Zhang HY, Chan CW (2007) Author’s reply to ‘‘comments on ‘performance evaluation
of UKF-based nonlinear filtering’’’. Automatica 43(3):569–570
Zhao L, Wang X (2009) Design of unscented Kalman filter with noise statistic estimator. Control
Decis 24(10):1483–1488
Chapter 154
A Cumulative SaaS Service Evolution
Model Based on Expanded Pi Calculus
Jun He, Tong Li and De-hai Zhang
Abstract SaaS emphasizes the concept of ‘‘customization’’ and it can provide a
higher level of service customizability, dynamic adaptability and customer
transparency. By abstracting software functions into ‘‘services’’, expanding the
grammar of the typical Pi calculus, and identifying the mapping between the
progress theory and service evolution, this paper proposes a cumulative evolution
model of SaaS service based on an expanded Pi calculus. It first analyses four atom
cumulative evolution models and then presents the integration among them. The
four atom cumulative evolution models are respectively the sequential, the reverse,
the simultaneous and the consecutive and for each of the models, the formula is
accordingly demonstrated. The model this paper proposes is supposed to con-
tribute to the customizability and dynamic adaptability of SaaS services so that the
evolutionary process of the services should be more transparent for customers and
thus the customers should have better on-line experiences.
Keywords SaaS Expanded Pi calculus Cumulative evolution Model
154.1 Introduction
Software as a Service (SaaS) is an internet-based software service supply and
delivery model (Wei 2011). It can provide services exactly as customers appoint,
namely ‘‘services provided as appointed’’. Different from the customer groups the
traditional software services are faced with, customer groups nowadays are more
J. He (&) T. Li D. Zhang
School of Software, Yunnan University, Kunming, Yunnan, China
e-mail: [email protected]
J. He T. Li
Yunnan Provincial Key Lab of Software Engineering, Kunming, Yunnan, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1457
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_154,
Springer-Verlag Berlin Heidelberg 2013
1458 J. He et al.
mixed and harder to approach as their requirements are open to more rapid
changes and variations. To meet their needs, a new revolutionary software service
that permits an enhanced customizability and dynamic adaptability is called for.
SaaS software is just the right thing that comes in. Compared to traditional
software, it can orchestrate much more complex evolution processes and it can
deal with the frequent corrections and replacements caused by both universal and
particular needs of the customers simultaneously. However, whether the post-
evolution services are customer-friendly or not poses challenges. Therefore, SaaS
has to be evolved on a coarsest-grained layer so that the evolution process is
unknown and transparent to customers to guarantee a better on-line experience on
their part. For the moment, the biggest challenges posed by the evolution of SaaS
services range from the problem of the coarsest-grain, transparency to
progressiveness.
A number of researches on the evolution of SaaS software have been con-
ducted. However, they are mainly focused on the issue of ‘‘multi-tenant versus one
instance’’ for the customization of work procedures and measures for data safety
(Luo and Wu 2011; Liang et al. 2010; Bezemer and Zaidman 2009). One research
(Liang et al. 2011) brings forward, from the perspective of work flow, an evolution
model and its method that supports a supreme work flow; one research is on
evolution model and data dependence (Liu et al. 2010); another research, though
yet to be expressed in formulas, proposes the concept of services in evolution to
describe the incessant changes of services and a possible solution to the cooper-
ation across services (Ramil and Lehman 2002).
With the expanded Pi calculus (EPI) as its descriptive formula, this paper
proposes a cumulative model to support the evolution of SaaS services so as to
bring forward a theoretical formula that will be applicable to an automatic evo-
lution of services in prospect.
154.2 SaaS Services Formularized in Expanded Pi
Calculus
154.2.1 Expanded Pi Calculus
The Pi Calculus (Milner 1999) is the calculus model brought forward by Milner
and others on the basis of Calculus of Communication System (CCS) to depict and
analyze the top structure in change. It is often used to describe inter-process
interactions and concurrent communicative operations. The formula can be put in
several ways (Sangiorgi and Walker 2003). The one my paper adopts is in its
ordinary form. It can be defined as following (Liao et al. 2005):
Definition 154.1 Suppose N is a set of an infinite number of names, and i, j, k, l,
m, n,… are the names of the N; capital letters A, B, C,… denote different
154 A Cumulative SaaS Service Evolution Model Based on Expanded Pi Calculus 1459
processes; and capital letters P, Q, R,… stand for the expressions of the processes,
then the basic grammar is as in (154.1).
P ::¼ 0jij:PjjðiÞ:Pjs:PjP þ QjPjQj
ð154:1Þ
ðiÞPj½i ¼ jPjAði; j; . . .; kÞ
Besides:
1. 0 indicates 0 process, which means no operation ever occurs.
2. P ? Q indicates a selective execution of process P or Q
3. Prefixed expression differentiates no positive prefix or negative prefix. ij is seen
as a negative prefix, ij: P denotes: to output the name j at the interface i, and
then execute P; jðiÞis a positive prefix, jðk=iÞ:P indicates: to input a random
name k at the interface j, then execute Pfk=ig (k=i means replacing i with k).
The symbol i in the prefixed expression is a pass, a link that connects two
reciprocal interfaces i and i; s:P is seen as a dumb prefix used to indicate the
invisible operations out of process.
4. PjQ is the expression for parallel processes, it means process P and process
Q are executed concurrently.
5. ðiÞP is the expression for restrictive processes, it means the new pass i permits
no ulterior operation; whereas P as interior communication is allowed to go
through the pass i.
6. ½i ¼ jP is the expression for identical processes, it means when i and j are
identical passes, their operations are the same as P, or the process is a 0 process.
7. Aði; j; . . .; kÞ is the symbol of process, i; j; . . .; k refers to any of the random
processes.
These are just a brief review of the definitions of the Pi calculus; further
explorations into it are provided in other papers (Milner 1999; Sangiorgi and
Walker 2003). The Pi calculus has proven to be a very useful formula when it
comes to applications like the description of procedural activities, structuring of
models etc. However, it can scarcely be used to depict services as it cannot denote
the proprietary relationship between a sub-service and the service sequence the
sub-service belongs to, nor can it properly describe the restrictive conditions that
determine transfers from service to service. Due to some original understandings
of the characteristics of SaaS service, and inspirations coming from the perceptive
literature (Zhou and Zeng 2009), I first propose in this paper an expansion of the
typical Pi calculus so as to present the cumulative model in the end (Zhou and
Ceng 2007).
Definition 154.2 The expansion of the typical Pi calculus is mainly in two
aspects:
1. Use the symbol ‘‘!’’, an indicator of two-tuple relationship, to describe the
proprietary relationship in which the collection N includes an infinite number of
names. A ! B signifies that the name A belongs to B. The employment of the
proprietary relationship can properly describe the belonging of a sub-process to
1460 J. He et al.
the process of an upper level, for example: i ! Si means the process i belongs
to the process sequence Si , and thus the analysis of the relationship of processes
can be limited to a certain layer.
2. Use the symbol ‘‘[h]’’ to describe the restrictive conditions of process transfers,
while h is the symbol of the conditions responsible for transfers, and it can be
taken as ½i ¼ jP, an expression expanded but identical, i.e. ½hP ½i ¼ jP.
Symbols ! and ? respectively indicate the output and input of data; i!x indicates
the output of data x through data channeli; and j?yindicates the input of data
y through data channel j. Expressions ½hi !x and uj ?x are used to indicate the
conditions that invoke the transfer of processes as well as the emission and
½hi !x indicates the process i sends the data x under the
reception of data. And
condition of hi ; uj ?x indicates the process j receives the data x under the
condition of uj .
154.2.2 Description of SaaS Services
When SaaS software functions are abstractedly taken as services, the evolution of
services can be applied to realize the dynamic change and maintenance of soft-
ware. As the evolution is basically the changes in structure, property and operation
of SaaS services (Papazoglou 2008), so to build an organizational structure that
encompasses all these factors can not only illustrate the relationship between an
upper and a lower service layer, define the property of services and the type of
operations, demonstrate the mapping relationships among them, but also illumi-
nate the organic integration of the service structures and the mechanism of service
solicitations SaaS software involves. Based on the analysis of the evolution of
services and the mapping relationship embodied by the expanded Pi calculus, the
following definition is made:
Definition 154.3 Suppose there are SaaS services whose sequences are as many as
n, it can be put as Si , and i ¼ f1; 2; . . .; ng. Si is a set of atom services, then it
follows Si ¼ frjr ! Si g. The atom service r in the expression is equal to the
process in the expanded Pi calculus.
Definition 154.4 A service schema of the SaaS services can be defined as a
quintuple, as in (154.2).
X
s ¼ ðS; A; E; C; f Þ ð154:2Þ
In it:
1. S indicates the set of atom services of the SaaS services,S ¼ fS1 ; S2 ; . . .; Sn g.
154 A Cumulative SaaS Service Evolution Model Based on Expanded Pi Calculus 1461
S
n
2. A ¼ Ai is the set of the service operations that each of the atom services
i¼1
provides, Ai ¼ fai1 ; ai2 ; . . .; ain g indicates the set of the service operations that
the atom service Si provides.
3. E indicates the set of the sequential execution processes provided by the atom
service S, an execution process refers to a partly sequence that consists of
nuclear services, for example: S1 S2 S3 S4 S5 is a partial
sequence.
4. C is a partial sequence of a service operation, for example: ðS1 ; a11 Þ
ðS2 ; a21 Þ ðS3 ; a31 Þ ðS4 ; a41 Þ ðS5 ; a51 Þ is a partial sequence of a
service operation.
5. The expression f : A ! A is to define the operational mapping function of the
services from an upper layer to a lower one, and f justifies the expression
f ðS; EÞ ¼ C, and f can be defined as a recursive function, it indicates a SaaS
service can be multiplied into a number of layers and grains according to actual
customer needs. Such a rendering of the definition can greatly improve the
applicability of the formula.
154.3 The Cumulative Evolution Model of SaaS Services
In the running process of a SaaS software, customer needs and commercial logic
keep changing. Though the changes are only relevant to some of the customers, the
evolution process is supposed to be transparent and cumulative to all users
undistinguished. To improve the service customizability and dynamic adaptability
of a SaaS software, this paper proposes an cumulative evolution model of the
SaaS, and the model is discussed in terms of four atom evolutions and the inte-
gration of them. The four nuclear evolutions are: the cumulative evolution in
sequential order, the cumulative evolution in reverse order, the cumulative evo-
lution in parallel order, and the cumulative evolution in corrective order.
Definition 154.5 The Service Schema proposed in Definition 154.4 can be used to
P
express a SaaS service: s ¼ ðS; A; E; C; f Þ. The capitalized A, B, C,… each
indicates one of the service serials represented by services i, j, r,… The symbol f,
as a nuclear service, can be evolved into a mapping function in operation, standing
for the mapping relationship of the service sequences prior to and post an
evolution.
1462 J. He et al.
154.3.1 The Sequential Cumulative Evolution
The sequential cumulative evolution means to add the service r to the sequence S,
and the sequence can be transformed into a new service sequence as a result. The
sequential cumulative evolution is an atom form for other evolutions.
The service sequence
S before evolution can be expressed as:
i ! Si ½hi !x uj ?x j ! Sj , in which: A ¼ i ! Si ½hi !x 0,
B ¼ uj ?x j ! Sj .
The added service can be expressed as r ! Sr , thus, the sequential cumulative
evolution of the service sequence S is:
fSIE : i ! Si ½hi !x ½ur ?x r ! Sr ½hk !y uj ?y j ! Sj ð154:3Þ
in which:
A ¼ i ! Si ½hi !x 0; C ¼ ½ur ?x r ! Sr ½hk !y; B ¼ uj ?x j ! Sj
With the running of the evolution of service sequence, changing are the data
receiving and sending conditions responsible for the happening of evolution. This
again suggests that the evolution process of a service sequence is always dynamic
and incessant.
154.3.2 The Reverse Cumulative Evolution
The reverse cumulative evolution is the opposite of the sequential cumulative
evolution.
The service sequence S before evolution can be expressed as in (154.4).
i ! Si ½hi !x uj ?x j ! Sj hj !y ½uk ?y k ! Sk ð154:4Þ
in which:
A ¼ i ! Si ½hi !x 0; B ¼ uj ?x j ! Sj hj !y 0; C ¼ ½uk ?y k ! Sk
The cumulative service in reverse can be expressed as j ! Sj .
The cumulative evolution of service sequence S in reverse can be expressed as
in (154.5).
1
fSIE : i ! Si ½hi !z ½uk ?z k ! Sk ð154:5Þ
in which:
A ¼ i ! Si ½hi !z 0; B ¼ ½uk ?z k ! Sk
154 A Cumulative SaaS Service Evolution Model Based on Expanded Pi Calculus 1463
154.3.3 The Parallel Cumulative Evolution
The parallel cumulative evolution supports the concurrent running of two services,
of which one is an inset and both services can be executed as parallels. Once one
service is added as an inset, then the number of messages sent or received by the
prepositive and postpositive service of either of them will increase from 1 to 2.
The service sequence S prior to the evolution can be expressed as in (154.6).
i ! Si ½hi !x uj ?x j ! Sj ½hk !y
ð154:6Þ
½uk ?y k ! Sk ½hk !y
In which:
A ¼ i ! Si ½hi !x 0; B ¼ uj ?x j ! Sj ½hk !y 0;
C ¼ ½uk ?y k ! Sk ½hk !y
The cumulative service can be expressed as r ! Sr , and u, v indicate the inlet
and outlet of message the service r is responsible for.
The parallel cumulative evolution of the service serial S thus can be expressed
as in (154.7).
h 0i
fPIE : i ! Si ð½hi !xj hi !uÞ
h i
ðð uj ?x j ! Sj hj !yÞjð½ur ?u ð154:7Þ
h 0i
r ! Sr hr !vÞÞ ð½uk ?yj uk ?vÞ k ! Sk
In which:
h 0i h i
A ¼ i ! Si ð½hi !xj hi !uÞ 0; B ¼ uj ?x j ! Sj hj !y 0;
h 0i
C ¼ ð½uk ?yj uk ?vÞ k ! Sk
154.3.4 The Adaptive Cumulative Evolution Model
When a service can no longer appeal to customers, it entails the replacement of the
old service with either one that is adapted or one that is brand new.
The service sequence S before evolution can be expressed as in (154.8).
i ! Si ½hi !x uj ?x j ! Sj hj !y ½uk ?y k ! Sk ð154:8Þ
In which:
1464 J. He et al.
A = i ! Si ½hi !x 0; B ¼ uj ?x j ! Sj hj !y 0; C ¼ ½uk ?y k ! Sk ;
The service that needs to be adapted is j ! Sj , the service through adaptation is
0
j ! Sj0 , the messages received and sent are transformed into u and v.
The service sequence S in its adaptable cumulative evolution model thus can be
expressed as in (9).
h i 0
fCIE : i ! Si ½hi !u uj0 ?u j ! Sj0 ½hk !v ½uk ?v k ! Sk ð154:9Þ
In which:
h i 0
A ¼ i ! Si ½hi !u 0; B ¼ uj0 ?u j ! Sj0 hj0 !u 0; C ¼ ½uk ?v k ! Sk
154.3.5 The Integration of the Cumulative Evolution Models
The sequential cumulative evolution, the reverse cumulative evolution, the parallel
cumulative evolution and the adaptive cumulative evolution are the four atom
evolution models that underlie the ultimate cumulative evolution of SaaS service
in question. Every evolution process of it can be realized by variedly configuring
the four atom evolutions. Among them, as the sequential cumulative evolution and
the parallel cumulative evolution involve the question of branching, it is necessary
to analyze the interrelationship of the evolution sequences and its evolution out-
comes. The integrated evolution process can be expressed as (Figs. 154.1, 154.2):
The service sequence S before integration can be expressed as in (154.10).
i ! Si ½hi !x uj ?x j ! Sj hj !z ½uk ?z k ! Sk ð154:10Þ
Fig. 154.1 Integration 1
i i
i x x
x m
m
u
j y y
z j j n
k z z
v
k k
154 A Cumulative SaaS Service Evolution Model Based on Expanded Pi Calculus 1465
Fig. 154.2 Integration 2
i
u
x
i i
u
x x m m
y v
j j n
j n
z z
w
k z
k
w
k
In which:
A ¼ i ! Si ½hi !x 0; B ¼ uj ?x j ! Sj ½hk !z 0; C ¼ ½uk ?z k ! Sk
When the sequential integration runs before the parallel integration:
i ! Si ½hi !x ½um ?x m ! Sm ½hm !y uj ?y
j ! Sj hj !z ½uk ?z k ! Sk ð154:11Þ
i ! Si ½hi !x ½um ?x m ! Sm ð½hm !xj
h 0 i h i
hm !uÞ ðð uj ?y j ! Sj hj !zÞjð½un ð154:12Þ
h 0i
?v n ! Sn hn !vÞÞ ð½uk ?zj uk ?vÞ k ! Sk
When the sequential integration runs after the parallel integration:
h 0i
i ! Si ð½hi !xj hi !uÞ ðð uj ?x j ! Sj
h i
hj !zÞjð½un ?u n ! Sn hn !wÞÞ
h 0i ð154:13Þ
ð½uk ?zj uk ?wÞ k ! Sk
1466 J. He et al.
h 0i
i ! Si ð½hi !xj hi !uÞ ðð½um ?x m !
h i
Sm hm !y uj ?y j ! Sj hj !zÞj
ð½um ?u m ! Sm hm !v ½un ?v n ! ð154:14Þ
h 0i
Sn hn !wÞÞ ð½uk ?zj uk ?wÞ k ! Sk
When the parallel integration runs before the sequential integration, the evo-
lution process would be more complex as it may induce redundant services.
Therefore, the elimination of the redundant services has to be considered.
Whether formula (154.13) and (154.14) are equal in value has to be verified.
Theorem 154.1 When the sequential cumulative evolution and the parallel
cumulative evolution are simultaneously executed in the service sequence S, then
the outcomes of the two evolution sequences are the same.
Proof The interactive simulation theory (Milner 1999). To differentiate the out-
comes of the two evolutions, the services of integration sequencing 2 are to be
represented by i0 ; j0 ; k0 ; m0 ; m00 ; n0 , and m0 and m00 are equal in value.
Suppose (S, T) is a system that indicates the transfer of symbols, T represents
the transfer of services from the service S
T ¼ fði; mÞ; ði0 ; m0 Þ; ðm; nÞ; ðm; iÞ; ðm0 ; nÞ; ðm00 ; n0 Þg
And suppose F is a two-tuple of S, then
F ¼ fði; i0 Þ; ðj; j0 Þ; ðm; m0 Þ; ðm; m00 Þ; ðn; n0 Þg
The transfer of the service i within T is identical with the transfer of the service
i0 within T, and the transfer of the service j within T is identical with the transfer of
the service j within T, similarly, all of the services in pair within F can match and
simulate mutually.
The reflexivity, symmetry, and transmissibility of S can be verified. According
to the definition of compulsory equaling of values as stipulated in the theory of
mutual simulation, it can be concluded that the results of the two evolution
sequences are the same.
Deduction To inset services as numerous as n between every two random
services (as well as two services in immediate sequence) of the service sequence S
either in parallel order or in sequential order, the results of their evolutions do not
reflect the difference in their operation orders.
154 A Cumulative SaaS Service Evolution Model Based on Expanded Pi Calculus 1467
154.4 Conclusion
SaaS is an internet-based software service supply and delivery model. With the
development of the cloud computing technology, researches on SaaS are getting
momentum (Wu et al. 2011; Ardagna et al. 2007). But the literature that is focused
on the evolution of SaaS service remains inadequate. As the evolution of SaaS
software involves a higher level of service customizability and dynamic adapt-
ability than that the traditional software requires, the evolution has to be executed
in a coarsest-grained, transparent and gradual way (Weber et al. 2008). The paper
proposes a cumulative evolution model of SaaS service on the basis of an
expanded Pi calculus. By expanding the typical Pi calculus, it is possible to
multiply the proprietary relationships in the collection of names and the restrictive
conditions that determine the transfers of processes, and thus probable is the
overall orchestrating of the evolution of SaaS services. Following the analysis of
the cumulative evolution, four atom cumulative evolutions are discussed. After
that, the possible integrations of them and the conditions under which they are
equal in value are also demonstrated. A relevant research to be furthered should be
on the layering of the services and the testing of the functions of the model.
Acknowledgments Foundation item: National Natural Science Fund Project (No. 60963007);
Software College of Yunnan University Construction Fund Project (No. 2010KS01).
References
Ardagna D, Comuzzi M, Mussi E (2007) Paws: a framework for executing adaptive web service
processes. IEEE Softw 24:39–46
Bezemer C-P, Zaidman A (2009) Multi-tenant SaaS applications maintenance dream or
nightmare. Position Paper 4:88–89
Liang S, Shuai L, Zhong L (2010) TLA based customization and verification mechanism of
business process for SaaS. Chin J Comput 33(11):2056–2058 (in Chinese)
Liang G, Jian C, Chen J (2011) Self-evolving for process model of software as a service. Comput
Integr Manuf Syst 17(8):1603–1608 (in Chinese)
Liao J, Tan H, Liu J (2005) Based on Pi calculation of web services composition description and
verification. Chin J Comput 33(4):635–643 (in Chinese)
Liu S, Wang H, Cui L (2010) Application of SaaS based on data dependency of the progressive
pattern evolution method. In: The first national conference on service computing (CCF NCSC
2010) essays, pp 127–129
Luo X-l, Wu Q-l (2011) Research of business logic framework for SaaS software service SaSed
on mass customization. Telecommun Sci 32:26–28 (in Chinese)
Milner R (1999) Communicating and mobile systems: the Pi calculus. Cambridge University
Press, Cambridge
Papazoglou M (2008) The challenges of service evolution. In: Proceedings of the 20th
international conference on advanced information systems engineering, pp 1–15
Ramil JF, Lehman MM (2002) Evolution in software and related areas. In: ACM
Sangiorgi D, Walker D (2003) The Pi calculus: a theory of mobile processes. Cambridge
University Press, New York
1468 J. He et al.
Weber B, Reichert M, Rinderle-Ma S (2008) Change patterns and change support features-
enhancing flexibility in process-aware information systems. Data Knowl Eng 64(3):438–466
Wei G (2011) Overview of SaaS theory and application. Agric Netw Inf 26:69–70 (in Chinese)
Wu X, Wang M, Zhang W (2011) Overview of cloud computing development. Sci Technol Vane
209:49–52 (in Chinese)
Zhou J, Ceng G (2007) Based on the CPi calculus grid service behavior research. Comput Sci
34(6):13–18 (in Chinese)
Zhou J, Zeng G (2009) A mechanism for grid service composition behavior specification and
verification. Future Gen Comput Syst 25(3):378–383
Chapter 155
Permanence and Extinction of Periodic
Delay Predator–Prey System with Two
Predators and Stage Structure for Prey
Wei-wei Zheng and Er-dong Han
Abstract In this paper, a periodic predator–prey delay system with
Beddington–DeAngelis and Holling IV functional response is proposed and
analyzed, where prey has stage structure and all three species are density dependent.
Using the comparison theorem and analytical method, sufficient conditions of the
permanence and extinction of the predators and prey species are obtained. In addi-
tion, sufficient conditions are derived for the existence of positive periodic solutions
of the system. According to the conclusions of the theorems, two examples are given
to check the correctness of the main results.
Keywords Beddington–DeAngelis functional response Delay Extinction
Permanence Stage Structure
155.1 Introduction
The aim of this paper is to investigate the permanence and extinction of the
following periodic delay three species predator–prey system with Holling IV and
Beddington–DeAngelis functional response and stage-structure for prey
W. Zheng (&) E. Han
School of Science, Xi’an Polytechnic University, Xi’an 710048,
Shaanxi, People’s Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1469
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_155,
Ó Springer-Verlag Berlin Heidelberg 2013
1470 W. Zheng and E. Han
8 Rt
> 0 bðsÞds h1 ðtÞx1 ðtÞ
>
> x 1 ðtÞ ¼ aðtÞx 2 ðtÞ bðtÞx 1 ðtÞ aðt s1 Þe ts1
x2 ðt s1 Þ y1 ðtÞ;
>
> a 1 ðtÞ þ x21 ðtÞ
>
> Rt
>
> 0 bðsÞds h2 ðtÞx2 ðtÞ
>
< x2 ðtÞ ¼ aðt s1 Þe ts1 x2 ðt s1 Þ cðtÞx22 ðtÞ y2 ðtÞ;
a2 ðtÞ þ bðtÞx2 ðtÞ þ cðtÞy2 ðtÞ
>
> p1 ðtÞx1 ðt s2 Þ
>
> y01 ðtÞ ¼ y1 ðtÞ q1 ðtÞ þ g1 ðtÞy1 ðtÞ ;
>
> a1 ðtÞ þ x21 ðt s2 Þ
>
>
>
> p2 ðtÞx2 ðt s3 Þ
: y0 ðtÞ ¼ y2 ðtÞ q2 ðtÞ þ g2 ðtÞy2 ðtÞ ;
2
a2 ðtÞ þ bðtÞx2 ðt s3 Þ þ cðtÞy2 ðt s3 Þ
ð155:1Þ
where x1 ðtÞ and x2 ðtÞ denote the densities of immature and mature prey species at
time t, respectively; y1 ðtÞ and y2 ðtÞ denote the densities of the predators that prey
on immature and mature prey at time t , respectively; aðtÞ; bðtÞ; cðtÞ; gi ðtÞ;
hi ðtÞ; pi ðtÞ; qi ðtÞ; ai ðtÞ; i ¼ 1; 2; bðtÞ; cðtÞ are all continuous positive x- periodic
functions; si ; i ¼ 1; 2; 3 are positive constants.
We can refer to (Liu and Yan 2011; Zhu et al. 2011) to get biological signif-
icance of all parameters and assumptions explanation of (155.1).
The initial conditions for system (155.1) is as follow
xi ðsÞ ¼ /i ðsÞ; yi ðsÞ ¼ wi ðsÞ [ 0; /i ð0Þ [ 0; wi ð0Þ [ 0;
ð155:2Þ
i ¼ 1; 2; s 2 ½s; 0:
where s ¼ maxfs1 ; s2 ; s3 g; / ¼ ð/1 ; /2 ; /3 ; /4 Þ 2 C ½s; 0; R4þ0 , here we let
R4þ0 ¼ fðx1 ; x2 ; x3 ; x4 Þjxi 0; i ¼ 1; 2; 3; 4g.
For the convenient of the following discussion, to a continuous x periodic
function f ðtÞ, we set
Z
1 x
mðf Þ ¼ f ðtÞdt
x 0
Meanwhile, we add definition
Z t
def
BðtÞ ¼ aðt s1 Þ exp bðsÞds
ts1
According to the analysis of the above, we get the following system (155.3).
155 Permanence and Extinction of Periodic Delay 1471
8
> h1 ðtÞx1 ðtÞ
>
> x_ 1 ðtÞ ¼ aðtÞx2 ðtÞ bðtÞx1 ðtÞ BðtÞx2 ðt s1 Þ y1 ðtÞ;
>
> a 2
1 ðtÞ þ x1 ðtÞ
>
>
>
>
>
> 2 h2 ðtÞx2 ðtÞ
>
< x_ 2 ðtÞ ¼ BðtÞx2 ðt s1 Þ cðtÞx2 ðtÞ a ðtÞ þ bðtÞx ðtÞ þ cðtÞy ðtÞ y2 ðtÞ;
2 2 2
>
> p1 ðtÞx1 ðt s2 Þ
>
> y_ 1 ðtÞ ¼ y1 ðtÞ q1 ðtÞ þ g1 ðtÞy1 ðtÞ ;
>
> a1 ðtÞ þ x21 ðt s2 Þ
>
>
>
>
>
> p2 ðtÞx2 ðt s3 Þ
: y_ 2 ðtÞ ¼ y2 ðtÞ q2 ðtÞ þ g2 ðtÞy2 ðtÞ :
a2 ðtÞ þ bðtÞx2 ðt s3 Þ þ cðtÞy2 ðt s3 Þ
ð155:3Þ
The dynamic behavior of the predator–prey system with delay and stage
structure for prey has been long discussed [see (Liu and Yan 2011; Zhu et al. 2011;
Li and Qian 2011; Li et al. 2011)]. Recently, Wang (Xiong and Li 2008) study the
model of this type, by using software Maple, the authors got the corresponding
numeric results of the conclusions. We can refer to (Hao and Jia 2008; Wang and
Chen 2010; Liu and Yan 2011; Naji and Balasim 2007; Cai et al. 2009; Zhao and
Lv 2009; Feng et al. 2010; Zhang et al. 2011; Liu and Yan 2011) to have in-depth
understanding of more research achievement of those models. Furthermore, the
research concerning the stage structure while being time delays predator–prey
periodic systems are quite rare. Indeed, recently, Kar (Ta and Nguyen 2011),
Huang (Chen and You 2008) and Chen (Huang et al. 2010) project those systems
on permanence and extinction. To keep the biological variety of ecosystem, the
dynamic behavior of biotic population is a significant and comprehensive problem
in biomathematics. So it is meaningful to investigate the system (155.3).
In the next section, we state the main results of this paper. Sufficient conditions
of the permanence and extinction of the system (155.3) are proved in Section III.
The conclusions we obtain further promote the analysis technique of Huang (Chen
and You 2008) and Chen (Huang et al. 2010).
155.2 Statement of the Main Results
Theorem 155.2.1 Suppose that
p1 ðtÞx1 ðt s2 Þ
m q1 ðtÞ þ [ 0;
a1 ðtÞ þ x21 ðt s2 Þ
ð155:4Þ
p2 ðtÞx2 ðt s3 Þ
m q2 ðtÞ þ [ 0;
a2 ðtÞ þ bðtÞx2 ðt s3 Þ
hold, where x1 ðtÞ; x2 ðtÞ is the unique positive periodic solution of system (155.4)
given by Lemma 155.2.2 [see (Chen and You 2008)]. Then system (155.3) is
permanent.
1472 W. Zheng and E. Han
Theorem 155.2.2 Assume the condition (155.4) hold, there is at least a positive
x- periodic solution of system (155.3).
Theorem 155.2.3 Suppose that
p1 ðtÞx1 ðt s2 Þ
m q1 ðtÞ þ 0;
a1 ðtÞ þ x21 ðt s2 Þ
ð155:5Þ
p2 ðtÞx2 ðt s3 Þ
m q2 ðtÞ þ 0;
a2 ðtÞ þ bðtÞx2 ðt s3 Þ
hold, then any solutions of system (155.3) with initial condition (155.2) satisfies
lim yi ðtÞ ¼ 0; i ¼ 1; 2:
t!þ1
155.3 Proof of the Main Results
We need the Lemma155.3.1–155.3.4 to proof Theorem 155.2.1.
Lemma 155.3.1 There exist positive constants Mx and My , such that
lim sup xi ðtÞ Mx ; lim sup yðtÞ My ; i ¼ 1; 2:
t!þ1 t!þ1
for all solutions of system (155.3) with initial condition (155.2).
Proof Let ðx1 ; x2 ; y1 ; y2 Þ be a solution of system (155.3) with initial conditions
(155.2), so we have
( 0
x1 ðtÞ aðtÞx2 ðtÞ bðtÞx1 ðtÞ BðtÞx2 ðt s1 Þ;
0 :
x2 ðtÞ BðtÞx2 ðt s1 Þ cðtÞx22 ðtÞ
By using Lemma 155.2.2, the following auxiliary equation:
( 0
u1 ðtÞ ¼ aðtÞu2 ðtÞ bðtÞu1 ðtÞ BðtÞu2 ðt s1 Þ;
0 ð155:6Þ
u2 ðtÞ ¼ BðtÞu2 ðt s1 Þ cðtÞu22 ðtÞ:
has a globally asymptotically stable positive x periodic solution x1 ðtÞ; x2 ðtÞ .
Let ðu1 ðtÞ; u2 ðtÞÞ be the solution of (155.6) with initial condition ðu1 ð0Þ; u2 ð0ÞÞ ¼
ðx1 ð0Þ; x2 ð0ÞÞ. According to the comparison theorem [see (Hao and Jia 2008)], we
have
xi ðtÞ ui ðtÞði ¼ 1; 2Þ; t 0 ð155:7Þ
By (155.4), there exists a e [ 0 which is sufficiently small, such that
155 Permanence and Extinction of Periodic Delay 1473
pi ðtÞ x1 ðtÞ þ e
m qi ðtÞ þ [ 0: ð155:8Þ
ai ðtÞ
Thus, from the global attractive of x1 ðtÞ; x2 ðtÞ ; for every given eð0\e\1Þ;
there exists a T1 [ 0; such that
ui ðtÞ xi ðtÞ \e; t T1 : ð155:9Þ
On the basis of (155.7) and (155.9), we have
xi ðtÞ\xi ðtÞ þ e; t [ T1 : ð155:10Þ
Let Mx ¼ max xi ðtÞ þ e; i ¼ 1; 2 , we have
t2½0;x
lim sup xi ðtÞ Mx ; i ¼ 1; 2:
t!þ1
For, t T1 ; from system (155.3) and (155.10), we can obtain
0 pi ðtÞxi ðt s2 Þ
yi ðtÞ yi ðtÞ qi ðtÞ þ gi ðtÞyi ðtÞ
ai ðtÞ
pi ðtÞ
yi ðtÞ qi ðtÞ þ u ðt s2 Þ þ e gi ðtÞyi ðtÞ :
ai ðtÞ i
Consider the following equation:
0 pi ðtÞ
vi ðtÞ ¼ vi ðtÞ qi ðtÞ þ u ðt s2 Þ þ e gi ðtÞyi ðtÞ : ð155:11Þ
ai ðtÞ i
From the Lemma 155.2.2 of (Chen and You 2008), (155.11) has a unique x-
periodic solution yi ðtÞ [ 0; i ¼ 1; 2:. Similarly to the above analysis, there exists a
T2 [ T1 , such that for the above e, we have
yi ðtÞ\yi ðtÞ þ e; t [ T2 :
Let Mx ¼ max fyi ðtÞ þ e; i ¼ 1; 2g, then
t2½0;x
lim sup yi ðtÞ My ; i ¼ 1; 2:
t!þ1
To the same argument of Lemma 155.3.1, we can easily get Lemma 155.3.2.
Lemma 155.3.2 There exists positive constant gix \Mx ; i ¼ 1; 2, such that
lim inf xi ðtÞ [ gix ; i ¼ 1; 2:
t!þ1
Lemma 155.3.3 Assumed that (155.4) holds, then there exists two positives
constants giy ; i ¼ 1; 2, such that any solutions ðx1 ðtÞ; x2 ðtÞ; y1 ðtÞ; y2 ðtÞÞ of system
(155.3) with initial condition (155.2) satisfies
1474 W. Zheng and E. Han
lim sup yi ðtÞ [ giy ; i ¼ 1; 2: ð155:12Þ
t!þ1
Proof Assume that condition (155.4) is establish, there exists a constant e0 [ 0,
1
and e0 [ min fxi ðtÞg, such that
2 t2½0;x
m ue0 ðtÞ [ 0; m we0 ðtÞ [ 0; ð155:13Þ
where
p1 ðtÞ x1 ðt s2 Þ e0
ue0 ðtÞ ¼ q1 ðtÞ þ q1 ðtÞe0 ;
a1 ðtÞ þ x1 ðt s2 Þ e0
p2 ðtÞ x2 ðt s3 Þ e0
we0 ðtÞ ¼ q2 ðtÞ þ q2 ðtÞe0 :
a2 ðtÞ þ bðtÞ x2 ðt s3 Þ e0 þ cðtÞe0
Take the equation below with a parameter e [ 0 into account:
8
>
> 0 h1 ðtÞ
> x
< 1 ðtÞ ¼ aðtÞx 2 ðtÞ bðtÞ þ 2e x1 ðtÞ BðtÞx2 ðt s1 Þ;
a1 ðtÞ
ð155:14Þ
>
> h ðtÞ 2
> x02 ðtÞ ¼ BðtÞx2 ðt s1 Þ cðtÞ þ 2e 2
: x ðtÞ:
a2 ðtÞ 2
By Lemma155.2.2, system (155.14) has a unique positive x- periodic solution
x 1e ðtÞ; ex 2e ðtÞ ; which is global attractive. Let ðx1e ðtÞ; x2e ðtÞÞ; be the solution of
e
(155.14) with initial condition xie ð0Þ ¼ xi ð0Þ; i ¼ 1; 2. Then, for the above e0 ,
there exists a sufficiently large T4 [ T3 , such that
e0
xie ðtÞ xie ðtÞ \ ; t T4 :
4
We have xie ðtÞ ! xi ðtÞ in ½T4 ; T4 þ x; as e ! 0. Then, for e0 [ 0, such that
e0
xie ðtÞ xi ðtÞ \ ; t 2 ½T4 ; T4 þ x; 0\e\e0 :
4
So, we can get
e0
xie ðtÞ xi ðtÞ xie ðtÞ xie ðtÞ þ xie ðtÞ xi ðtÞ \ :
2
Since xie ðtÞ; xi ðtÞ are all x-periodic, hence
e0
xie ðtÞ xi ðtÞ \ ; i ¼ 1; 2; t 0; 0\e\e0 :
2
Choosing a constant e1 ð0\e1 \e0 ; 2e1 \e0 Þ, we have
155 Permanence and Extinction of Periodic Delay 1475
e0
xie ðtÞ xi ðtÞ ; i ¼ 1; 2; t 0: ð155:15Þ
2
Assuming (155.12) is false, then there exists / 2 R4þ , such that, under the initial
condition ðx1 ðhÞ; x2 ðhÞ; y1 ðhÞ; y2 ðhÞÞ ¼ /; h 2 ½s; 0. We have lim sup yi ðt; /Þ
t!þ1
\e1 ; i ¼ 1; 2.
So, there exists T5 [ T4 , such that
yi ðt; /Þ\2e1 \e0 ; t T5 : ð155:16Þ
By using (155.16), from system (155.3), for all t T6 T5 þ s1 , we can obtain
0 h1 ðtÞ
x1 ðtÞ aðtÞx2 ðt; /Þ bðtÞ þ 2e1 x1 ðt; /Þ BðtÞx2 ðt s1 ; /Þ;
a1 ðtÞ
h2 ðtÞ 2
x02 ðtÞ BðtÞx2 ðt s1 ; /Þ cðtÞ þ 2e1 x ðt; /Þ:
a2 ðtÞ 2
Let ðu1 ðtÞ; u2 ðtÞÞ be the solution of (155.14), with e ¼ e1 and ðx1 ðT6 ; /Þ;
x2 ðT6 ; /ÞÞ, then
xi ðt; /Þ ui ðtÞ; i¼ 1; 2;t T6 :
From the global attractive of x1e1 ðtÞ; x2e1 ðtÞ , here we let e ¼ e20 , there exists
T7 T6 ; such that
e0
ui ðtÞ xie1 ðtÞ \ ; i ¼ 1; 2; t T7 :
2
So, we have
e0
xi ðt; /Þ ui ðtÞ [ xie1 ðtÞ ; i ¼ 1; 2; t T7 :
2
Hence, by (155.15), we can obtain
xi ðt; /Þ xi ðtÞ e0 ; i ¼ 1; 2; t T7 : ð155:17Þ
Therefore, by (155.16) and (155.17), for, t T7 þ s2 ; such that
y01 ðt; /Þ
" #
p1 ðtÞ x1 ðt s2 Þ e0
y1 ðt; /Þ q1 ðtÞ þ g1 ðtÞe0
a1 ðtÞ þ x1 ðt s2 Þ e0
¼ ue0 ðtÞy1 ðt; /Þ;
1476 W. Zheng and E. Han
y02 ðt; /Þ
" #
p2 ðtÞ x2 ðt s3 Þ e0
y2 ðt; /Þ q2 ðtÞ þ g2 ðtÞe0
a2 ðtÞ þ bðtÞ x2 ðt s3 Þ e0 þ cðtÞe0
¼ we0 ðtÞy2 ðt; /Þ:
ð155:18Þ
Integrating both sides of (155.18) from T7 þ s2 to t and from T7 þ s3 to t,
respectively, so we can get
Z t
y1 ðt; /Þ y1 ðT7 þ s2 ; /Þ exp ue0 ðtÞdt;
T7 þs2
Z t
y2 ðt; /Þ y2 ðT7 þ s3 ; /Þ exp we0 ðtÞdt:
T7 þs3
Thus, from (155.16), we obtain yi ðt; /Þ ! þ1; i ¼ 1; 2; t ! þ1. It is a
contradiction of the Lemma 155.3.1. So the proof of the theorem 155.3.3 is
complete.
Lemma 155.3.4 Under the condition (155.4), there exist positive constants
siy ; i ¼ 1; 2, such that any solutions of system (155.3) with initial condition
(155.2) satisfies
lim inf yi ðtÞ [ siy ; i ¼ 1; 2: ð155:19Þ
t!þ1
Proof of Theorem 155.2.1 By Lemma 155.3.2 and 155.3.3, system (155.3) is
uniform weak persistent. Further, from the Lemma 155.3.1 and 155.3.4, system
(155.3) is persistent.
Proof of Theorem 155.2.2 From the proof of Lemma 155.3.1–155.3.4 in
Theorem 155.2.1, using the same method, we can proof the Theorem 155.2.2. Here
we omit the detail of certificate process.
Proof of Theorem 155.2.3 Actually, by (155.5), for any given positive constant
eðe\1Þ; there exist e1 [ 0 ð0\e1 \eÞ and e0 [ 0, we get the following (155.20).
!
p1 ðtÞ x1 ðt s2 Þ þ e1 e
m q1 ðtÞ þ 2 g1 ðtÞe mðq1 ðtÞÞ\ e0 ;
a1 ðtÞ þ x1 ðt s2 Þ þ e1 2
!
p2 ðtÞ x1 ðt s3 Þ þ e1 e
m q2 ðtÞ þ g2 ðtÞe mðq2 ðtÞÞ\ e0 :
a2 ðtÞ þ bðtÞ x1 ðt s2 Þ þ e1 2
ð155:20Þ
155 Permanence and Extinction of Periodic Delay 1477
Since
0
x1 ðtÞ aðtÞx2 ðtÞ bðtÞx1 ðtÞ BðtÞx2 ðt s1 Þ;
0
x2 ðtÞ BðtÞx2 ðt s1 Þ cðtÞx22 ðtÞ:
For the above e1 ; there exists a T ð1Þ [ 0, such that
xi ðtÞ\xi ðtÞ þ e1 ; t T ð1Þ : ð155:21Þ
It follows from (155.20) and (155.21) that for t max T ð1Þ þ s2 ; T ð2Þ þ s3 ,
p1 ðtÞx1 ðt s2 Þ
m q1 ðtÞ þ g1 ðtÞe \ e0 ;
a1 ðtÞ þ x21 ðt s2 Þ
ð155:22Þ
p2 ðtÞx2 ðt s3 Þ
m q2 ðtÞ þ g2 ðtÞe \ e0 :
a2 ðtÞ þ bðtÞx2 ðt s3 Þ
First, there exists a T ð2Þ [ max T ð1Þ þ s2 ; T ð2Þ þ s3 , such that yi ðT ð2Þ Þ\
eði ¼ 1; 2Þ. Otherwise, by (155.22), we have
e y1 ðtÞ
Z
t
h1 ðtÞx1 ðs s2 Þ
y1 ðT ð1Þ þ s2 Þ exp q1 ðsÞ þ q1 ðsÞe ds
T ð1Þ þs2 a1 ðsÞ þ x1 ðt s2 Þ
n o
ð1Þ ð1Þ
y1 ðT þ s2 Þ exp e0 t T s2 ! 0:
As t ! þ1. Similarly, we can get
n o
e y2 ðtÞ y2 ðT ð1Þ þ s3 Þ exp e0 t T ð1Þ s3 ! 0; t ! þ1:
So we have e\0, which is contradictions.
Second, we will prove that
yi ðtÞ e expfMðeÞxg; i ¼ 1; 2; t T ð2Þ ; ð155:23Þ
where
pi ðtÞx1 ðt sj Þ
MðeÞ ¼ max qi ðtÞ þ þ gi ðtÞe; i; j ¼ 2; 3; i 6¼ j ;
t2½0;x ai ðtÞ þ x1 ðt sj Þ
is a bounded constant for e 2 ½1; 0. Otherwise, then there exists a T ð3Þ T ð2Þ , we
can obtain
yi ðT ð3Þ Þ [ e expfMðeÞxg; i ¼ 1; 2:
By the continuity of yi ðtÞ, then there must exists T ð4Þ 2 T ð2Þ ; T ð3Þ , such that
yi ðT ð4Þ Þ ¼ e and yi ðtÞ [ e, for t 2 T ð4Þ ; T ð3Þ . Let P1 be the nonnegative integer
such that T ð3Þ 2 T ð4Þ þ P1 x; T ð3Þ þ ðP1 þ 1Þx . From (155.22), we have
1478 W. Zheng and E. Han
e expfMðeÞxg
(Z )
T ð3Þ
ð3Þ ð4Þ h1 ðtÞx1 ðt s2 Þ
\y1 ðT Þ\y1 ðT Þ exp q1 ðtÞ þ g1 ðtÞe dt
T ð4Þ a1 ðtÞ þ x21 ðt s2 Þ
(Z ð4Þ Z T ð3Þ )
T þP1 x
h1 ðtÞx1 ðt s2 Þ
¼ e exp þ q1 ðtÞ þ g1 ðtÞe dt
T ð4Þ T ð4Þ þP1 x a1 ðtÞ þ x21 ðt s2 Þ
(Z ð3Þ )
T
h1 ðtÞx1 ðt s2 Þ
\e exp d2 ðtÞ þ qðtÞe dt
T ð4Þ þP1 x a1 ðtÞ þ x21 ðt s2 Þ
e expfMðeÞxg:
We can see that this is a contradiction. Similarly, from the second equation of
(155.22), we have
e expfMðeÞxg\y2 ðT ð3Þ Þ e expfMðeÞxg:
Which is also contradiction, so (155.23) holds. By the random of the parameter
e; we know yi ðtÞ ! 0; i ¼ 1; 2; t ! þ1. So we complete the proof of Theorem
155.2.3.
155.4 Examples and conclusion
Example 1 From system (155.3), cause aðtÞ ¼ 4; bðtÞ ¼ 2=3, cðtÞ ¼ 9 exp
f0:3gð1 expf0:3gÞ; q1 ðtÞ ¼ 1=4 cos t; q2 ðtÞ ¼ 1=5 cos t; h1 ðtÞ ¼ 5; h2 ðtÞ
¼ 6; p1 ðtÞ ¼ 2 þ cos t; p2 ðtÞ ¼ 3 þ cos t; a1 ðtÞ ¼ 1; a2 ðtÞ ¼ 1=4ð1 expf0:3gÞ;
bðtÞ ¼ 7 þ 2 cos t; cðtÞ ¼ 1; s1 ¼ s2 ¼ s3 ¼ 0:6;gi ðtÞ; i ¼ 1; 2 are any arbitrary
nonnegative continuous 2p-periodic functions.
The above parameters conditions satisfy Theorem 155.2.1, so system (155.3) is
permanent and admits at least a positive 2p-periodic solution. From Fig. 155.1, we
can see that the density restriction of the predators have a major impact on the
stability of the predator–prey system. When predator species have no crowding
effect, the predator species is at high density; and with crowding effect, the
predator species is at low density.
Example 2 Assuming that the conditions of example 1 are established. Causing
q1 ðtÞ ¼ 5=4 cos t; q2 ðtÞ ¼ 1=2 cos t, those parameters satisfy the Theorem
155.2.3. So any positive solution of system (155.3) satisfies
lim yi ðtÞ ¼ 0; i ¼ 1; 2:
t!þ1
Figure 155.2 shows that two predators are extinction and the immature and
mature preys are permanent.
155 Permanence and Extinction of Periodic Delay 1479
Fig. 155.1 The growth curve
of system (1.3) with initial
condition ðx1 ðhÞ; x2 ðhÞ;
y1 ðhÞ; y2 ðhÞÞ ¼
ð1 expf0:5g;
0:5; 1; 2Þ; g1 ðtÞ ¼
2 þ sin t; g2 ðtÞ ¼ 2 þ sin t;
0 t 50; 0:6 h 0
Fig. 155.2 The growth curve
of system (1.3) with initial
condition ðx1 ðhÞ; x2 ðhÞ;
y1 ðhÞ; y2 ðhÞÞ ¼
ð1 expf0:5g; 0:5; 1; 2Þ;
0 t 50; 0:6 h 0
From the Theorem 155.2.1–155.2.3, we can get a conclusion: the death rate and
the density restriction of the two predator population have a great extent influence
on the dynamic behavior of the system.
Acknowledgment This paper is supported by the Natural Science Fund of Shaanxi Provincial
Education Administration Bureau (Grant No. 11JK0502) and the Doctor’s Research Fund of
Xi’an Polytechnic University.
1480 W. Zheng and E. Han
References
Cai LM, Li XZ, Yu JY (2009) Dynamics of a nonautonomous predator-prey dispersion-delay
system with Beddington-DeAngelis functional response. Chaos Solutions Fractals
40:2064–2075
Chen FD, You MS (2008) Permanence, extinction and periodic solution of the predator-prey
system with Beddington-DeAngelis functional response and stage-structure for prey.
Nonlinear Anal Real World Appl 9:207–221
Feng JF, Zhu L, Wang HL (2010) Stability of a ecosystem induced by motual interference
between predators. Procedia Environ Sci 2:42–48
Hao M, Jia JW (2008) The existence of positive periodic solutions of a food-chain system with
Beddington-DeAngelis functional response and stage structure for prey. J Shanxi Norm Univ
(Nat Sci Edn) 22(1):22–26 (Chinese)
Huang CY, Zhao M, Zhao LC (2010) Permanence of periodic predator-prey system with two
predators and stage-structure for prey. Nonlinear Anal Real World Appl 11:503–514
Li HG, Qian J (2011) Existence of periodic solution in a Beddington-DeAngelis predator-prey
system with stage-structure and time delay. J Hubei Norm Univ (Nat Sci) 31(3):81–84
(Chinese)
Li YK, Teng ZD, Mehbubare R (2011) Periodic solutions for a delayed nonautonomous stage-
structure predator-prey model. J Xinjiang Univ (Nat Sci Edn) 28(1):57–64 (Chinese)
Liu GR, Yan JR (2011) Positive periodic solutions of neutral predator-prey with Beddington-
DeAngelis functional response. Comput Math 61:2317–2322
Liu GR, Yan JR (2011) Positive periodic solutions of neutral predator-prey model with
Beddington-DeAngelis functional response. Comput Math Appl 61:2317–2322
Liu GR, Yan JR (2011) Positive periodic solutions of neutral predator-prey model with
Beddington-DeAngelis functional response. Comput Math Appl 61:2317–2322
Naji RK, Balasim AT (2007) Dynamical behavior of a three species food chain model with
Beddington-DeAngelis functional response. Chaos Solutions Fractals 32:1853–1866
Ta VT, Nguyen TH (2011) Dynamics of species in a model with two predators and one prey.
Nonlinear Anal 74:4868–4881
Wang L, Chen SY (2010) Existence of positive periodic solutions for the predator-prey difference
system with Beddington-DeAngelis functional response and time delays. J Southwess
Jiaotong Univ (Engl Edn) 18(3):260–270 (Chinese)
Xiong YB, Li HZ (2008) Existence and global asymptotic stablity of periodic solution in a food-
chain system with delays and Beddington-DeAngelis functional response. J Syst Sci Math Sci
28(3):288–301
Zhang JF, Li WT, Yan XP (2011) Hopf bifurcation and turing instability in special homogeneous
and inhomogeneous predator-prey models. Appl Math Comput 218:1883–1893
Zhao M, Lv SJ (2009) Chaos in a three-species food chain model with a Beddington-DeAngelis
response. Chaos Solutions Fractals 40:2305–2316
Zhu HT, Zhu WD, Zhang ZD (2011) Persistence of competitive ecological mathematic model.
J Control South Univ Forest Technol 31(4):214–218 (Chinese)
Chapter 156
Design of Military Logistics Management
System Based on Internet of Things
Technology
Xue-jiang Wei, Jian-fei Feng, Shao-juan Feng and Yu-zhu Zhai
Abstract Since the capability of information collection, information processing
and information using is the bottleneck of the development of military logistics,
this paper puts forward ‘‘three-layer architecture’’ of military logistics manage-
ment system based on Internet of Things technology after analyzing the operation
process of military logistics, and introduces its functions which include demand
management, decision-making assistant, procurement management, warehouse
management, transportation & distribution management.
Keywords Internet of things Military logistics Military logistics management
system RFID
156.1 Introduction
The ultimate objective of military logistics operation is distributing military supplies
to all combat troops actively at the right time and places (O’hanlon 2009). The
achievement of this objective is greatly influenced by the capability of information
collection, information transmission, information analyzing, information processing
and information using, which has been the bottleneck of the development of military
logistics (Jiang 2009). This situation may be changed by developing and applying
Internet of Things technology in military logistics. Based on the Internet of Things
technology, some appropriate objectives of military logistics can be easily achieved
by optimizing military logistics operation process and constructing military logistics
management system (Yang and Pan 2011; Li 2011a).
X. Wei (&) J. Feng S. Feng
Public Acquisition Department, Wuhan Economics Academy, Wuhan, China
e-mail: [email protected]
Y. Zhai
Wuhan Economics Academy, Wuhan, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1481
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_156,
Ó Springer-Verlag Berlin Heidelberg 2013
1482 X. Wei et al.
156.2 Analysis of Military Logistics Operation Process
The military logistics operation process can be divided into three phases: demand
applying, material acquisition, transportation & distribution (Zhang et al. 2007), as
shown in Fig. 156.1.
(1) Demand applying. When combat troops need some materials, they apply the
demand including material names, amounts, quality standards, deadlines and
destinations to military logistics command.
(2) Material acquisition. When military logistics command receives combat
troops’ material demands, it makes material acquisition decision promptly
after contracting demand with inventory. If the inventory is enough, it will
give an outing order to military depot; otherwise it will give a material pro-
curement mission to military procurement agency which will purchase those
needed materials from suppliers.
(3) Transportation & distribution. During the period of carrying out material
acquisition mission, military logistics command decides how to distribute the
supplies. If it is dangerous for civil logistics enterprise, military logistics
command will select military transportation agency to transport the supplies;
otherwise it will select civil logistics enterprise. If the supplies are only
distributed to one place, they will be transported to the destination directly;
otherwise they will be transported to distribution center firstly and be dis-
tributed to combat troops by distribution center later.
Fig. 156.1 Military logistics operation process
156 Design of Military Logistics Management System 1483
156.3 Design of Military Logistics Management System
156.3.1 Objective of Military Logistics Management System
Military logistics management system aims at providing an information manage-
ment platform to military logistics command, military procurement agency, military
depot and military transportation agency, and combining them with suppliers and
logistics enterprises by information flow. It also aims at achieving automatic
information collection, intelligent information processing, scientific resource
configuration and maximum support efficiency by applying Internet of Things
technology and Geographic Information System (GIS).
156.3.2 Architecture of Military Logistics Management
System
Architecture of military logistics management system includes three layers:
sensing layer, network layer and application layer (Li 2011b; Qu 2010; Su 2011;
Lin 2011; Chen and Jiang 2011; Zhu 2010), as shown in Fig. 156.2.
156.3.3 Functions of Military Logistics Management System
156.3.3.1 Military Logistics Decision-Making Assistant
Military logistics decision-making assistant is provided to military logistics com-
mand, which includes demand management, potential support capability evaluation
and decision-making assistant of material support plan (Ruan and Lu 2011).
(1) Demand management. After combat troops apply material demands through
internet or by Personal Digital Assistant (PDA), military logistics management
system integrates these demands with same material and destination auto-
matically accorded with the common rule and request of demand management
by the assistance of computers.
(2) Potential support capability evaluation. When receiving a supply mission,
military logistics management system analyzes the information provided by
supplier database, inventory database and transportation database intelligently,
and evaluates the potential support capability provided by suppliers, military
depots, military transportation agencies and logistics enterprises automatically
by the assistance of GIS application support service, cloud computing service
and intelligent analysis service.
1484 X. Wei et al.
Fig. 156.2 Architecture of military logistics management system
(3) Decision-making assistant of material support plan. Based on the evaluation
of potential support capability, military logistics management system can
assist the commander to make decision after analyzing the situation of military
logistics, the distribution of resources and the condition of transportation.
156 Design of Military Logistics Management System 1485
156.3.3.2 Procurement Management
Procurement management is provided to military procurement agency, which
includes supplier selection and E-Procurement management (Huang 2006).
(1) Supplier selection. With the assistance of Geo-database and GIS application
support service, military logistics management system searches those suppliers
which accord with demands intelligently following the rules set by com-
mander, and transforms the result into report automatically.
(2) E-procurement management. Through military logistics management system,
all material procurement activities such as promulgating procurement infor-
mation, bidding, evaluating bids, contracting and payment can be carried out
online.
156.3.3.3 Warehouse Management
Warehouse management is provided to military depot, which includes storage
location assignment, out-warehouse and in-warehouse management, inventory
management, monitoring and control to warehouse’s surroundings (Wu 2011).
(1) Storage location assignment. Military logistics management system provides
graphical interfaces to users, and assigns storage location intelligently
according to the information of materials including deposit periods, bulk &
weight, destination and so on, in order to improve the utilization ratio of
storage location and speed of inbound & outbound operation.
(2) Out-warehouse & in-warehouse management. Military logistics management
system transmits out-warehouse & in-warehouse information through local
area network (LAN) of military rear depot, and collects materials information
automatically by using Radio Frequency Identification technology (RFID).
(3) Inventory management. Military logistics management system provides
graphical interface to users, collects information and refreshes data automat-
ically and periodically, calculates the threshold value of inventory with the
assistance of intelligent analysis service, and advises manager to apply for
material while inventory is under threshold value.
(4) Monitoring and control of warehouse’s surroundings. Military logistics
management system will give an alarm while incident happens such as in-
break, fire and leak of dangerous materials, sense the temperature and
humidity of warehouses and activate adjustments automatically while tem-
perature or humidity is out of normal scope.
156.3.3.4 Transportation Management
Transportation management is provided to military transportation agency, which
includes vehicle monitoring, transportation planning and routing planning.
1486 X. Wei et al.
(1) Vehicle monitoring. Military logistics management system provides a graph-
ical interface to users to monitor vehicle on electronic map with the appli-
cation of GPS and the assistance of GPS support service.
(2) Transportation planning. Military logistics management system provides
assistance to users to select transportation mode and vehicles according to
destination, material type, material amount and material bulk & weight.
(3) Routing planning. Military logistics management system gives advice about
routing planning to commander, and provides guidance to vehicles while some
selected routes are interrupted.
156.3.3.5 Distribution Management
Distribution management is provided to distribution center, which includes dis-
tribution planning and loading planning (Li 2011c).
(1) Distribution planning. Distribution center may distribute one batch materials
to several combat troops and demand included several types of materials
which are applied by one combat troop may be grouped in different batches, so
distribution planning is one of the most important tasks of distribution center.
Military logistics management system provides assistance to distribution
center by making distribution plans automatically based on contrasting
received materials with combat troops’ demands.
(2) Loading planning. Military logistics management system makes loading
decisions that include loading orders and loading modes based on intelligent
analysis service.
156.3.3.6 Military Logistics Operations Monitoring and Control
This function is provided to military logistics command. Military logistics man-
agement system records and reviews the operation process to the commander, so
the commander can adjust the plan while emergency happens.
156.4 Conclusion
Military logistics management system which is based on Internet of Things aims at
improving the capability of information collection, information processing and
information using. Sensing layer provides functions of information collection
and collaborative information processing; Network layer provides the function
of information transmitting just-in-time; Application layer provides functions
which include decision-making assistant, procurement management, warehouse
156 Design of Military Logistics Management System 1487
management, transportation management and distribution management to each
department of military logistics. To achieve the objective, it needs not only to
solve the technical problems, but also to solve those problems such as standard-
ization of material code, network security and maximization of economic benefit.
References
Chen FZ, Jiang DH (2011) Design and implication on the intelligent logistics delivery system
based on internet of things. Microelectron Comput 28(8):19–21
Huang FY (2006) Design and implication of E-purchase system. J Shangqiu Vocat Tech Coll
5(5):34–36 (in Chinese)
Jiang LW (2009) Research on models and strategies of military logistics capability optimization.
Beijing jianotong University, pp 8–9 (in Chinese)
Li ZS (2011a) The influence of internet of things to logistics development. Logist Sci-Tech
3:77–78 (in Chinese)
Li HW (2011b) Planning and design of logistics information platform based on internet of things.
Inf Technol 9:13–16 (in Chinese)
Li YB (2011c) The research on optimizing the intelligentized logistics distribution system by
building the internet of things. Logist Eng Manage 33(7):56–57 (in Chinese)
Lin ZX (2011) Unified information system of logistics pallet circulation based on internet of
things. Sci Technol Res 7:190–201 (in Chinese)
O’hanlon ME (2009) The science of war. Princeton University Press, Princeton, pp 141–145 (in
Chinese)
Qu XL (2010) Internet of things technology-based logistics management system for emergency.
Comput CD Softw Appl 13:107–109 (in Chinese)
Ruan ZYL, Lu L (2011) Networking technology based on the logistics of emergency. Jiangsu
Commer Forum 9:60–62 (in Chinese)
Su YH (2011) Design and implementation of logistics and vehicle monitoring system based on
the internet of things. Comput Digit Eng 7:75–78 (in Chinese)
Wu XZ (2011) Analysis on the application of internet of things to the field of storage logistics and
the prospect. Chin Bus Market 6:36–39 (in Chinese)
Yang YQ, Pan H (2011) Discussion on reengineering strategies of logistics management
information system based on internet of things. Comput Mod 12:98–101 (in Chinese)
Zhang FL, Wang C, Huang J (2007) Military logistics operations model based on process. Logist
Sci-Tech 4:97–99 (in Chinese)
Zhu WH (2010) Realization of whole-process intelligent supply chain distribution service based
on the internet of things. Logist Technol 7:172–173 (in Chinese)
Chapter 157
Simulation and Optimization of the Steel
Enterprise Raw Material System
Lin-wei Xu, Pei-qing Wang, Shang-lun Chen and Xing-li Zhong
Abstract This paper uses discrete system simulation method to simulate the
production scheduling material factory, verify the feasibility of the material fac-
tory system design, find out the system weakness, optimize the design and
scheduling schemes, save investment, reduce the cost of operation. With the
understanding of production and operation in detailed, we use system simulation to
build material factory simulation model, the system simulation not only provides a
powerful data analysis, and supports virtual reality 3D animation. On the opti-
mization problem of belt conveyor route, we compare the A * algorithm and
depth-first recursion algorithm, the best algorithm was obtained. About the yard
optimization problem, we use unfixed type and variable tonnage stock pile, and use
search mechanism to dispatch belt conveyor route and reclaimer, combining
Optimization module to optimize the yard. At the same time we can get the input
of coal field to formulate purchase plan.
Keywords Materials factory Simulation Optimization
157.1 Introduction
With the further adjustment of China’s steel industry structure, steel enterprises
gradually develop to be large-scale; the delivery and storage capacity of the material
factory has become one of the bottlenecks which restrict the production scale. The
major domestic steel enterprises have risen a new turn of energy saving and emission
L. Xu (&)
CISDI Chongqing Iron Steelmaking Plant Integration Co. Ltd., Chongqing 400013, China
e-mail: [email protected]
P. Wang S. Chen X. Zhong
Logistics Department, CISDI Engineering Co., Ltd., Chongqing 400013, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1489
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_157,
Ó Springer-Verlag Berlin Heidelberg 2013
1490 L. Xu et al.
reduction and eliminate backward production capacity. Further optimizing design
which aims to improve the production capacity of material factory is in research and
practice. With the social economic development, the land resources are becoming
more and more expensive, steel enterprises can not expand its area of material
factory indefinitely. With limited resources how to optimize design, explore the
potential, and improve the production capacity of material factory is a common
problem for all designers of raw material plant. Simply increasing the investment to
expand raw material scale can increase its production capacity, but apparently it isn’t
the economic way; Compression investment, streamline processes, improve the
efficiency of equipment and explore production capacity is also exposed to the risk
of insufficient capacity. We need an accurate evaluation method that can give
quantitative data to balance the two sides and provide the useful data for the deci-
sion-making (Sun and Xu 2009a, b; Hopp and Spearman 2002).
Material factory has much equipment, scattered layout, complex process and its
subsystem intersect with each other, and it is a typical of random discrete event
system, Mathematical analytical method cannot give a comprehensive analysis and
optimization. Thus the current design often uses empirical coefficient method. In
order to avoid wasting a lot manpower, material and time, system simulation is the
inevitable choice to support and optimize design. The system simulation allows us
to observe the dynamic change of the system model, identify bottlenecks, modified
parameters repeatedly and find the best parameters, and then optimize the per-
formance of the entire system. The system simulation of material factory makes us
to accurately evaluate the quality of the design in its planning stage, improve the
reliability of the actual production, and achieve better social and economic ben-
efits, at the same time, we can simulate various scheduling schemes to optimize the
inventory of the material factory and guide the production.
157.2 Material Factory System Design
This paper simulate a typical coal yard where store eight kinds of coking coal, two
kinds of thermal coal, two kinds of blind coal and three kinds of injecting Mixed
Coal. There are A, B, C, D four material strip feeder. Each material strip feeder has
a track bed with two bucket-wheel stacker reclaimer above it.
157.2.1 Input System
The input system mainly including the pier and the rail car dumper input system.
The main raw material is transported from the sea to the pier and then transported
into the material factory through belt conveyor, the other material unload by the
train car dumper and then transported by the belt conveyor too. The pier input
157 Simulation and Optimization of the Steel Enterprise Raw Material System 1491
system designed to one conveyor line. The rail input system consists of car dumper
and relevant delivery system that designed to two lines.
157.2.2 Output System
The coking coal output system is designed to two lines that mainly transported coking
coal from coal yard to coking coal blending bunker, transported blind coal from coal
yard to sintering blending bunker and transported injecting mixed coal from coal
yard to the blast furnace injection blending bunkers. The thermal coal output system
is designed to one line that mainly transported thermal coal from coal yard to power
plant. The blast furnace output system and the coking coal output system use the same
transport lines.
157.3 Equipment of the Material Factory
and Simulation Module
Establish each simulation module and the simulation model shown in Fig. 157.1.
In the following section this paper will introduce various equipment and simula-
tion module.
Fig. 157.1 The simulation model diagram
1492 L. Xu et al.
157.3.1 The Belt Conveyor Module
The belt conveyor is a material handing machine that continuous transfer material
in a certain line, also known as a tape machine. Belt conveyor can be horizontal,
tilt and vertical transport, the space can also be composed of transport lines,
transport lines are generally fixed. Belt conveyor transport capacity, long distance,
but also in the transportation process of a number of processes operating at the
same time, so a wide range of applications. The simulation software has a standard
fixed belt conveyor module, so set the parameters then can use it.
157.3.2 The Stacker-Reclaimer Module
The stacker-reclaimer is are widely used in building materials, mining, coal,
power, metallurgical, chemical, cement and other industries. Because of there is no
corresponding standard module in the software, so this module need to completely
customize. The stacker-reclaimer movement is very complex, it not only walk,
rotate arm, but also the bucket wheel need to be rotated, so the stacker-reclaimer
need to customize a variety of kinematics, Import 3D models, the module shown in
Fig. 157.2.
157.3.3 The Bunker Module
There are a variety of bunkers to provide raw material for the production system in
the raw material system, it is the end for the transport system, but for the pro-
duction it is the beginning. Each bunker consumption rates and demand are not the
Fig. 157.2 Stacker and reclaimer
157 Simulation and Optimization of the Steel Enterprise Raw Material System 1493
Fig. 157.3 The coal pile simulation module
same. The bunker that contains a lot of codes is a Satellite control centre in
simulation system. The bunker will call the belt conveyor line and scheduling
module of stacker-reclaimer according to the level of the material. Central control
module will determine the task priority. Finally, the bunker module appoints belt
conveyor line and stacker-reclaimer. Then the belt conveyor line and stacker-
reclaimer finish the task.
157.3.4 The Coal Pile Module
Coal pile is the most important module in the simulation model and the amount of
code it contains is also the largest, because it is not only required the coal but also
the supplier, so it is not only an active judgment entity but also a passive entity.
When it is in the active state, it sends a message to the port module or railway
station module. Then it will call the belt conveyor and stacker-reclaime, and then
the module will call the central control module to determine the task priority,
finally finishes the transport task. When it is in a passive state, it only receives the
message sent by the central control module to tell which bunker does the coal send
to. Its three-dimensional entity shows in Fig. 157.3.
157.3.5 The Central Control Module
The role of this module is to determine which task execute first based on task priority
parameters, it can be a no-display graphical entity; the other is the interrupt module,
its role is mainly to interrupt the non-critical tasks to free the resource, the freed
resource will meet the emergence situation based on the number of interrupt within a
certain time.
1494 L. Xu et al.
Fig. 157.4 Tape machine route table
157.4 Optimization of the Material Factory
157.4.1 Belt Conveyor Route Optimization
The belt conveyor route is very complex; there are a lot of transfer stations. Some
of the transfer station between the belt conveyors is for several belt conveyor
routes, so there is not only one belt conveyor route from one operating point to
another, therefore it is necessary to determine which is the shortest route. Although
there are some routes in the simulation process, but if one or a few belt conveyor
of some routes has been occupied, so we must call the shortest route based on the
available route situation. The shortest path algorithm is Dijkstra algorithm, A*(A
start) algorithm, depth-first algorithm. Dijkstra algorithm is A* algorithm of the
special case, also is the lowest efficiency case. If we just need to find a path, depth-
first algorithm can quickly find out the route and jump out of the circulation;
however, if we search for all the routes, and then compare all the paths, so the
efficiency is very low. In the simulation model, each task need to call the search
path algorithm, so if the shortest path algorithm efficiency is low, which will affect
the simulation speed. A* algorithm is actually a heuristic search (Manuel and
Johan 2004; Dong et al. 2003; Chow 1990; Scholl 1999; Zhao et al. 2000); A*
algorithm uses a best-first search and finds a least-cost path from a given initial
node to one goal node. It uses a distance-plus-cost heuristic function to determine
the order in which the search visits nodes in the tree. The A* can be implemented
more efficiently—roughly speaking, no node needs to be processed more than once
(Clymer and Cheng 2001; Solow 2005). If we use general function to package the
157 Simulation and Optimization of the Steel Enterprise Raw Material System 1495
Fig. 157.5 Stock pile layout optimization table
A*, we call this function then we can get a belt convey or route. As shown in
Fig. 157.4, through this route table, the coal entities can reach its destination.
157.4.2 Storage Yard Optimization
Each strip feeder has a certain amount of stock pile, in order to optimize, each
stock pile set a unique number and tonnage, if the consumer want to find the right
stock pile by the unique number, so it must search the storage yard module. In
other words, we can change the unique number and tonnage of each stock pile in
storage yard module, that mean the stock pile martial changed. By set the unique
number constraints and tonnage constraint and the objective function in the
optimize module, run the simulation we can get the best storage layout. As shown
in Fig. 157.5.
157.5 Simulation Results and Conclusions
By way of simulation we can get the utilization of all kinds of equipment, as
shown in Table 157.1, by analyzing utilization of the equipment, we can find out
the bottlenecks and redundancy of the system, and then we can find out system
defects and reduce redundant equipment to reduce costs.
1496 L. Xu et al.
Table 157.1 A part of utilization of tape machine
Convey Idle (%) Blocked (%) Conveying (%)
G303_BW1400 35.54 0.00 64.46
G304SHR_BW1400 35.84 0.00 64.16
B304_BW1600 37.60 0.00 62.40
B302_BW1600 39.19 0.00 60.81
G104_BW1600 40.43 0.00 59.57
G102_BW1600 40.61 0.00 59.39
G103SHR_BW1800 40.61 0.00 59.39
G101_BW1600 40.62 0.00 59.38
B305SHR_BW1800 38.23 0.00 58.40
P131_BW1200 41.84 0.00 58.16
P207_BW1200 41.86 0.00 58.14
P205_BW1200 42.60 0.00 57.40
P107_BW1200 42.63 0.00 57.37
P108SHR_BW1400 42.68 0.00 57.32
P206SHR_BW1400 42.68 0.00 57.32
P203_BW1200 42.72 0.00 57.28
P105_BW1200 42.74 0.00 57.26
P204R_BW1200 43.16 0.00 56.84
Table 157.2 Inventory of coal pile
Object Minimum inventory Maximum inventory Average inventory
Stock pile b1 33162 33600 33212.87
Stock pile b2 24741 25200 25016.22
Stock pile b3 23934 25200 24613.88
Stock pile b4 32731 33600 33315.41
Stock pile b5 24365 25200 24759.07
Stock pile b6 24141 25200 24653.88
Stock pile b7 33037 33600 33362.62
Stock pile b8 22921 25200 24187.45
Stock pile b9 24611 25200 24858
Stock pile c1 14235 15150 14667.43
Stock pile c2 17346 18225 17796.27
Stock pile c3 10314 11175 10849.64
Stock pile c4 10727 11175 11011.88
Stock pile c5 15150 15150 15150
Stock pile c6 5438 5625 5607.377
Stock pile c7 10713 11175 10934.45
Stock pile d1 5439 5625 5566.625
Stock pile d2 5225 5700 5505.055
Stock pile d3 10307 11175 10774.34
Stock pile d4 11175 11175 11175
Stock pile d5 13827 15150 14468.55
Stock pile d6 10345 11175 10790.21
Stock pile d7 10724 11175 10918.79
Stock pile d8 9843 11175 10619.38
157 Simulation and Optimization of the Steel Enterprise Raw Material System 1497
Table 157.3 Input of storage yard
Begin time End time Type Quantity delivered Source
204493.9498 213501.039 5 3000 Railway
229834.1575 238175.9075 8 5000 Ship
239054.1902 247395.9402 2 5000 Ship
310582.79 319589.8792 1 3000 Railway
364302.9018 380978.0684 7 10000 Ship
396078.4561 404420.2061 3 5000 Ship
485135.0921 494142.1813 1 3000 Railway
We can also get the maximum, minimum and average inventory of each bunker
and stock pile. At the same time, so we can get minimum safe stock to reduce the
costs (Table 157.2).
Get the input information of the storage yard by ways of simulation is earlier to
make procurement plan. Management staff can make out the corresponding pur-
chasing transport plan. Table 157.3 is input information of the storage yard.
Finally by means of using simulation technology we can make quantitative
evaluation and analysis of the material factory and optimize the design and
scheduling. This paper also showed that we can use simulation technology to
establish the random complex large production scheduling system model. If we
change the Parameters and strategy, we can get the simulation results quickly, and
by using optimization function of simulation software, we can get the optimal
configuration of the system resources.
References
Chow W (1990) Assembly line design methodology and applications. Marcel Dekker, New York
Clymer JR, Cheng DJ (2001) Simulation-based engineering of complex adaptive systems using a
classifier block. In: The 34th Annual Simulation Symposium, Seattle, WA, pp 243–250,
22–26 April 2001
Dong J, Xiao T, Zhao Y (2003) Application of genetic-Tabu Search Algorithm in sequencing
mixed-model assembly lines industrial engineering and management. Ind Eng Manag
8(2):14–17
Hopp WJ, Spearman ML (2002) Factory physics—foundations of manufacturing management.
Tsinghua University Press, Beijing
Manuel L, Johan M (2004) Business processing modeling, simulation and design. Prentice Hall,
New Jersey
Scholl A (1999) Balancing and sequencing of assembly lines. Physica-Verlag, Heidel-berg
Solow D (2005) On the challenge of developing a formal mathematical theory for establishing
emergence in complex systems in Complexity. Wiley, New York, vol 6, no 1, pp 49–52
Sun H, Xu L (2009) Optimization of scheduling problem for auto mixed model assembly line. In:
Proceedings of the first international workshop on intelligent systems and applications,
2009.5, vol 3, pp 2015–2017
1498 L. Xu et al.
Sun H, Xu L (2009) Simulation and optimization for noshery service system. In: Proceedings of
the first international symposium on information engineering and electronic commerce,
2009.5, pp 721–723
Zhao W, Han W, Luo Y (2000) Scheduling mixed-model assembly lines in JIT production
systems. J Manag Sci China 3(4):23–28
Chapter 158
Comprehensive Evaluation
and Optimizing for Boarding Strategies
Da-wei Sun, Xia-yang Zheng, Zi-jun Chen and Hong-min Wang
Abstract In this paper, we focus on the need for reducing boarding time for
airlines. Therefore existing researches devoted to designing boarding routines and
studying boarding strategies in existence. A model based on cellular automata is
developed for calculating the integrated boarding time and testifies that the
Reverse-Pyramid way is one of the most effective strategies. Aiming at giving an
optimal boarding strategy, this paper combines a new evaluating criterion with
some further analysis of Reverse-Pyramid and finally concludes that Reverse-
Pyramid strategy which is divided into 5 groups and has more groups with a
particular proportion is the best. Somehow the present paper solves the neglect of
passengers’ satisfaction and time spent on organizing before boarding in existing
researches and gives some recommendations to airlines at last.
Keywords Aircraft boarding Aisle interference Cellular automata Evaluating
criterion
158.1 Introduction
How much time is usually demanded for different necessary tasks when the air-
plane is landed, which are departure, fuel, baggage loading and unloading,
catering, and passengers’ boarding? According to reports from Boeing, the pas-
sengers’ boarding is the most time-consuming task, around 60 % of the total
(Capelo et al. 2008).
Because the plane makes money for airline only when it is in motion, reducing
the boarding time is helpful to not only arrange more scheduled flight but also be
D. Sun (&) X. Zheng Z. Chen H. Wang
Department of Electric and Electronic Engineering, North China Electricity Power
University, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1499
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_158,
Ó Springer-Verlag Berlin Heidelberg 2013
1500 D. Sun et al.
beneficial to finance and customer’s satisfaction. So in order to reduce the
boarding time, many researches and the present paper build the boarding strategy,
which is a group of rules that aim at boarding all passengers as quickly as possible.
For the former researches, in 2002, Van Landeghem and Beuselinck compared
the random strategy and back-to-front strategy. And results showed that the ran-
dom strategy performed better (Van Landeghem and Beuselinck 2002). Then in
2005, Pieric and Kai modeled the aisle interference by grid simulation (Ferrari and
Nagel 2005). The same year, Menkes and Briel firstly came out the reverse pyr-
amid strategy (Menkes et al. 2005).
For the present paper, a model based cellular automata is presented to study the
different strategies. An integrated standard is given to estimate the strategies
including the total boarding time, interference waiting time per passenger and time
spent on organizing before boarding. Finally the paper optimizes the reverse
pyramid strategy and gives some recommendations.
158.2 Comparing Boarding Strategies
158.2.1 Model
158.2.1.1 Assumptions
(1) Interferences are the major reasons lead to wait. There are two main kinds of
interferences which are aisle interference and seat interference. The aisle inter-
ference shows that the process of placing baggage will delay the passengers who
followed by them. And for the second kind, when a passenger wants to settle in the
window seat, he may block other passengers of the same line. We call this ‘seat
interference’. But in the model, we ignore the seat interference, because many
researches made clear that Outside-In strategy is better than Back-To-Front and
random strategy mainly because Outside-In strategy avoiding the seat interference.
The paper mainly studies the strategies which avoid seat interference themselves
on the basis of predecessors, so the seat interference doesn’t influence the results.
(2) Assumptions for the passengers: they will not happen to have other’s seat or
miss his seat. And they will obey our arrangement of boarding. Also the paper
doesn’t consider the passengers’ coming late.
(3) Assumptions for the planes: the boarding gate is in the top of the cabin. And
the business class seats and the first class seats are far less than economy class
seats. That allows us to only consider the boarding of passenger in economy class.
158.2.1.2 Introduction of Three Strategies
See Table 158.1 for introduction of three strategies.
158 Comprehensive Evaluation and Optimizing for Boarding Strategies 1501
Table 158.1 Introduction of three strategies
Name: Back-to-Front Picture
Advantages
Somehow avoid the conflict of gangway because the back
passenger will not be obstructed by the front passenger
Name: Outside-In
Advantages
Totally avoid the conflict of seat and fully make use of the
space of gangway to place the luggage
Name: Reverse-Pyramid
Advantages
Have the advantages of the former two strategies
158.2.1.3 Describe the Cabin
Because we want to describe the individual behavior, we decompose the cabin into
many units. Just like Fig. 158.1. The figure follows these assumptions: (1) All the
seats are treated as the same size and each seat is considered as one unit which is
arranged very closely. (2) The gangway has the same width of a seat and it only
gangway
Fig. 158.1 Cabin
1502 D. Sun et al.
allows passengers to stand as single line. (3) The only entrance is at the top of the
economy class.
This figure is the small size of plane that can have a capacity of 100 passengers.
158.2.1.4 Describe the Behavior of Passengers
All the passengers fit the following three rules.
(1) There are 100 passengers and everyone’s target seat is his own seat and will
not have a wrong seat. (2) If no one arranges the passengers, they will board
randomly. (3) Passengers board continuously.
The individual behavior fits the following five rules. (1) If no one stops a
passenger, he will walk directly to the ‘gangway unit’ that is the nearest one to his
target seat. (2) Once the passenger arrive the nearest gangway unit, he will spend
some time placing the luggage. (3) After placing the luggage, he will move to the
target seat. (4) It is a discrete model, so when the timer adds a unit of time, every
passenger can move to the empty unit near his present unit. (5) If timer adds a unit
of time and the next unit that the passenger wants to move in is busy, the passenger
will stay at his present unit until the next unit is empty.
158.2.1.5 Describe Time Spending on Placing Luggage for Every
Passenger
Most researches consider the aisle interference as a probability event, without
considering the influence of increasing and decreasing of the luggage. Although
some considered it that the time grew linear with the quantity of luggage existed in
the trunk. But these researches didn’t account for the luggage capacity of the
plane.
For these reasons, we draw support from literature (Shang et al. 2010).
k
Tbag ¼ nlug
ðc þ 1Þ ðnbin þ nlug Þ ð158:1Þ
s.t.:0 ðnbin þ nlug Þ c
In the formula, Tbag means the time using for placing the luggage. nbin means
the present amount of baggage that already be placed in the luggage trunk. Nlug
means the number of bags that will be put in the luggage trunk. c means the
number of capacity of luggage rack for one row of seat. k is correction coefficient.
According to (Shang et al. 2010), c = 4, k = 20 (Trivedi 2002; Kiwi 2006).
In this hyperbolic model, set DT to stand for the unit time of placing luggage.
ntotal stands for the total number of capacity of luggage rack for one row of seat.
So DT ¼ fðntotal Þ is a hyperbola (see Fig. 158.2). We can find that when the
spare of suitcases become smaller, the time of placing luggage is longer. When
ntotal [ 4, the time of placing luggage approaches infinite (Bohannon 1997).
158 Comprehensive Evaluation and Optimizing for Boarding Strategies 1503
Fig. 158.2 Hyperbola
Table 158.2 Probability
Number of bags Zero One Two
A passenger 20 % 70 % 10 %
But not everyone will bring luggage and they have different number of bags.
The paper refers to the report from ‘Data 100 Market Research’ and gets the
probability for every passenger as in Table 158.2 (Merry 1998).
158.2.2 Results
For each result, we simulate 200 times and average the results to become the
finally result. And we calculate the root-mean-square deviation of every result and
find that all the root-mean-square deviations are smaller than the 5 % of the
corresponding results. So we can say the results are credible (Fig. 158.3).
Compare the results of strategy 1, 2, 3. We can see the strategy 3 is best because
its boarding steps and waiting steps are the least. Strategy 1 is similar to the
strategy of Back-To-Front. Strategy 2 is similar to the strategy of Outside-in. And
strategy 3 is similar to the strategy of Reverse-Pyramid. Therefore, we can see the
advantage of Reverse-Pyramid. For the same reason, when we compare the
strategy 4, 5, 6, 7, we can get the same laws. So we can conclude that Reverse-
Pyramid is the best one on the hand of total steps and waiting steps. So we
concentrate on the strategy of Reverse-Pyramid.
1504 D. Sun et al.
No. How to dividethe group Steps of Waiting
boarding steps
1 5
7
1
3 196.5 42.5
gangway
8 4
6 2
2 1
2 gangway
6 5 190 39
8 7
4 3
3 1
3 7 4 185 36
gangway
8 6
5 2
1
4 gangway
2 19 9 38.5
2
1
3 1
5 gangway
4 2 201 45
4 2
3 1
2 1
6 gangway
4 3 194 42
4 3
2 1
3 1
7 gangway
4 3 189 37
4 2
2 1
Fig. 158.3 Results
158.3 Optimal Number of Groups
The higher the number of groups of seats in the airship, the longer it takes to
organize a line before boarding. Most researches didn’t explain how they chose the
number of groups. The present paper calculates the financial lose because of the
time spent on organizing a line and waiting time. So using the economic indicator,
we find the optimal number.
First we assume that all the 100 passengers sit in waiting hall by a line. Finally,
different strategies require different orders of queues. For example, if one strategy
requires every one boarding in a certain order, that’s to say, the number of groups
is 100, the queue must be formed in order one by one. But, if one strategy makes
the passengers’ boarding randomly, all passengers will stand the position closest to
his seat in waiting hall. Finally, we get the total average steps using for organizing
a line for different number groups. Some results are showed in Table. 158.3.
We treat the steps using for organizing a line the same as the average waiting
steps, because the two kinds of steps reflects the satisfaction of passengers. And
according to the literature (Li 2010) and The Civil Aviation Act, Air China’s flying
Table 158.3 Average number of steps
Number of groups 0 5 100
Organizing steps 0 105 2514
158 Comprehensive Evaluation and Optimizing for Boarding Strategies 1505
Table 158.4 Optimal number of groups
Number of groups 0 5 100
Total steps 226 211 217
hours is 736770 in 2007 and retained profits are 3773 million RMB. Also airline
pays every passenger 200 RMB for 4 h aircraft delay. So according to the ratio of
money, we get the ratio of waiting step and boarding step which is 1/25.605. With
all the data, we can calculate the optimal number of groups. The result is that 5 is
the most optimal number and there are some results in Table. 158.4.
All the results in Sect. 158.3 are for Reverse-Pyramid strategy.
158.4 Optimal Reverse-Pyramid Strategy
Compare with the four strategies in Fig. 158.4, we find the strategy 13 is the best.
These four strategies are all Reverse-Pyramid. The difference between the four
strategies is shape of the second, the third and the fourth group. The second group
has some window seats and some aisle seats. If we change the proportion of the
two kinds of seats, the results will change. For the same reason, changing the third
and the fourth groups’ proportion of two kinds of seats will change the final
results. Comparing these four strategies, we infer that it is better to arrange more
groups that have the proportion of 7/3 approximately (7 is window seat and 3 is
aisle seat).
Steps of Waiting
No. How to dividethe group
boarding steps
4 3 2 1
5 4 3 2
11 gangway
5 4 3 2
209 44.9
4 3 2 1
4 3 2 1
5 4 3 2
12 gangway
5 4 3 2
207.5 44.7
4 3 2 1
4 3 2 1
5 4 3 2
13 gangway
5 4 3 2 205.5 44.2
4 3 2 1
3 2 1
5 4 3 2
14 gangway
5 4 3 2 209.5 45.4
3 2 1
Fig. 158.4 Comparison of strategies
1506 D. Sun et al.
158.5 Conclusion
The present paper simulates a model to calculate the total boarding steps, waiting
steps and organizing steps. First the model compares the three boarding strategies.
Finding that the best one is Reverse-Pyramid, the model’s results are the same as
the other researches and that somehow proves the reliability of the model. Then a
new way of evaluating boarding strategies comes out and shows that dividing into
5 groups is the best choice. Finally an optimal Reverse-Pyramid is given. And the
paper recommends airline that for the similar structure as Fig. 158.1, it’s better to
use Reverse-Pyramid strategy which is divided into 5 groups and arranges more
groups that have the proportion of 7/3 approximately. For the other type of planes,
one method is to divide the bigger plane into the structures as Fig. 158.1 shows.
Another way is to change the parameter in Fig. 158.1 to fit the special type plane.
Both methods are easy to achieve.
References
Bohannon RW (1997) Comfortable and maximum walking speed of adults aged 20-79 years,
reference values and determinants. Age and Aging 26:15–19
Capelo E, de Castro Silva JL, van den Briel MHL, Villalobos JR (2008) Aircraft boarding fine-
tuning. In: XIV international conference on industrial engineering and operations
management
Ferrari P, Nagel K (2005) Robustness of efficient passenger boarding strategies for airplanes.
Transp Res Board 1915:44–54
Kiwi M (2006) A concentration bound for the longest increasing subsequence of a randomly
chosen involution. Discrete Appl Math 154(13):1816–1823
Li X (2010) Airlines flight delays analysis. Friends Account 2(2):41–43
Menkes HL, Briel VD, Villalobos JR, et al (2005) America West Airlines develops efficient
boarding strategies. Interface 35(3):191–201
Merry MM (1998) The role of computer simulation in reducing airplane turn time. Aero
Magazine 1
Shang HY, Lu HP, Peng Y (2010) Aircraft boarding strategy based on cellular automata.
Tsinghua Univ (Sci & Tech), vol 50, No 9
Trivedi KS (2002) Probability and statistics with reliability, queuing and computer science.
Wiley, New York
Van Landeghem H, Beuselinck A (2002) Reducing passenger boarding time in airplanes: a
simulation based approach. Eur J Oper Res 142:294–308
Chapter 159
Modeling for Crime Busting
Da-wei Sun, Xia-yang Zheng, Zi-jun Chen
and Hong-min Wang
Abstract The paper models for identifying people in a 83-workers-company who
are the most likely conspirators. The train of thought is that: (1) get a priority list
for valuing the suspicious degree of the workers, (2) get a line separating con-
spirators from nonconspirators, (3) get the leader of the conspiracy. The paper first
sets different values of suspicious degree for messages with various features in
order to value the suspicious degree of everybody. Secondly, we optimizes the
primary figure by using a formula based on weighted average method. Thirdly, we
worked through each individual on the better priority list from both ends. Then, the
paper used some methods of semantic analysis to better distinguish possible
conspirators from the others and finally got the priority list. Next, the discriminate
line is determined by using probability theory and clustering analysis theory. At
last, get the leaders by the priority list and discriminate line.
Keywords Mathematic model Crime busting Social network Text analysis
159.1 Restatement of the Problem
The present paper is for investigating a conspiracy to commit a criminal act. Now,
we realize that the conspiracy is taking place to embezzle funds from the company
and use internet fraud to steal funds from credit cards of people who do business
with the company. All we know is that there are 83 people, 400 messages (sent by
the 83 people), 15 topics (3 have been deemed to be suspicious), 7 known con-
spirators, 8 known non-conspirators and there are three managers in the company.
D. Sun (&) X. Zheng Z. Chen H. Wang
Department of Electric and Electronic Engineering,
North China Electricity Power University, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1507
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_159,
Springer-Verlag Berlin Heidelberg 2013
1508 D. Sun et al.
159.2 Model and Results
159.2.1 Step1
Our goal is to get a table to explain the suspicious degree of different messages,
and then we can get a preliminary priority list on the basis of the table.
We consider that every message connects two workers. According to the sus-
picious degree of the message’s topic, we add a reasonable weight to each worker.
The weight is related not only to the suspicious degree of the message’s topic, but
also to the suspicious degree of the speaker and listener.
In Table 159.1, ‘c.’ means conspirator. This table gives the value to Qm that we
use in the formula (159.1). The corner mark ‘m’ means the number of a message.
Mentioned in the section of sensitivity analysis, ‘(A#)’ stands for the number of
the cell in the table. This table is one of our foundations of the model. We can see
the Qm is decided by three factors: the speaker, the listener and the topic of No. m
message.
X
N
Yi ¼ Rim Qm ð159:1Þ
m¼0
‘Yi’ is an intermediate variable for worker i and we will use ‘Yi’ in formula
(159.2). ‘N’ means total amount of the messages and in our case N = 400. ‘m’
means the No. of a message. ‘Qm’ means the weight that No. m message gives.
The value of ‘Qm’ is given by Table 159.1. Rim means the relation between No. m
message and worker i. If the No. m message is sent from or sent to worker i,
Rim = 1. If not Rim = 0. At first we use ‘Yi’ to present the suspicious degree for
every workers.
The primary priority list is the result and it is made by traversal all messages
with Matlab.
It is obvious that there are two special points (20, 12) and (56, 58). Our goal is
to distinguish conspirators and non-conspirators, so there should be one special
point to divide the figure into two parts in theory. So we don’t satisfy with this
figure, and there must be some other factors that we didn’t take into consideration.
Table 159.1 Values for Qm
Suspicious topics Normal topics
Spoken to Known Unknown Known Known Unknown Known
Weight c. worker non-c. c. worker non-c.
Spoken from
Known c. 15(A11) 10(A1) 2(A12) 4(A15) 2(A16) 1(A16)
Unknown 10(A2) 6(A3) 4(A4) 2(A17) 1(A8) 0(A9)
worker
Known non-c. 2(A13) 4(A15) 0(A14) 1(A17) 0(A10) 0(A18)
159 Modeling for Crime Busting 1509
159.2.2 Step2
Our goal is to optimize the priority figure we have got in step1 using method of
weighted average.
We think ‘Yi’ (mentioned in formula (159.1)) to stand for the suspicious degree
is not reliable. Through a formula we create, we can evaluate the suspicious degree
of each worker by using the method of weighted average.
PM
SdðkÞ Ai ðkÞ
Wi ¼ k¼1 Yi ð159:2Þ
Ai Sdmax
‘Wi’ means the crime suspicious degree of worker i. ‘k’ means the No. of the
topic. M means the total amount of topics and in our case, M = 15. ‘Yi’ is given
by formula (159.1). Sd(k) stands for the influence of linguistics for No. k topic.
Before step4, Sd(k) = 1 when topic k is not suspicious and Sd(k) = 2 when topic
is the suspicious. And we will optimize the value of Sd(k) at step4 and finally give
many different values for Sd(k) according to different topic number. Sdmax is
defined by
Sdmax ¼ maxfSd1 Sd2 . . .SdM g ð159:3Þ
Ai is the total amount of text topics worker i gets. For example, if David
receives only one message and sent no message. And the only message includes
three text topics, ADavid = 3. Ai(k) is the total amount of text topics of number k.
That is to say
X
M
Ai ¼ Ai ðkÞ ð159:4Þ
k¼1
We can get Wi, which is the crime suspicious degree of worker i, by searching
all messages. Then we get the priority list by ranking Wi.
Figure 159.2 is better than Fig. 159.1, because there is only one inflection
point. But the slope of both sides of the inflection point is not different largely. We
want to optimize the priority figure further more in the step3.
159.2.3 Step3
Our goal is to optimize the better priority figure we have got in step2 using method
of iteration.
We find a disadvantage of step1 and step2: We didn’t consider the difference
between the unknown workers. In fact, some of the unknown workers are crimi-
nals and the others are not. So they should be considered differently. In another
word, different unknown worker have different influence for other’s ‘Wi’. We
consider these different influences in step3 to optimize our priority list.
1510 D. Sun et al.
Fig. 159.1 Primary priority figure
Fig. 159.2 Better priority figure
159 Modeling for Crime Busting 1511
Fig. 159.3 Final priority figure
In Fig. 159.1, point (1, 3) worker and point (68, 93) worker are considered
differently in this step. We treat point (68, 93) worker as a known criminal and
treat point (1, 3) worker as an innocent. And use the way of step1 again. So we get
another priority list.
But we think it’s unconvincing that let two unknown worker to be a criminal
and an innocent apart in each time. Because if you see the Fig. 159.1, you will find
the point (68, 93) worker has obvious difference from the near one in crime
suspicious degree. But point (1, 3) worker has tiny difference from the near one. So
we match the 67 points a curve and use slopes of the two vertex tangents to
describe how many unknown workers should be treated as criminals and inno-
cents. So we do these continually and finally we will places all unknown workers
in two categories: criminals and innocents. We will get a priority list at last
(Fig. 159.3).
159.2.4 Step4
Our goal is that: consulting the literatures about the text analysis methods, we try
to optimize our priority figure by this method.
We set ‘Sd(k)’ which means the suspicious degree for the No. k topic. We think
that the topics talked more frequently in the group of conspirators should be added
a larger value for ‘Sd(k)’. We use the following formula (159.5) and (159.6) to
describe ‘Sd(k)’
1512 D. Sun et al.
Fig. 159.4 Final priority figure with linguistics
Sdmaxi
SdðkÞiþ1 ¼ uðkÞi uavei 10 þ SdðkÞi ð159:5Þ
umaxi
pðkÞh
uðkÞ ¼ h pðkÞj ð159:6Þ
j
‘P(k)h’ means the times that topic k is talked by criminals. ‘h’ means the
amount of criminals. ‘j’ means the total amount of people, which means j = 83.
uðkÞ describes the frequency degree of topic k in conversation of criminals.
umaxi ¼ maxfuðkÞi g; uavei ¼ averagefuðkÞi g; Sdmaxi ¼ maxfSdðkÞi g. In step2, we
elect unknown workers to be criminals and innocents continually. So we set cir-
culating in Matlab and the corner mark ‘i’ means the times we circulate.
We use the text analysis idea to value Sd(k), and finally change the ‘Wi’ (degree
of crime suspicious for worker i).
With the considering of influence from text analysis method to ‘Sd(k)’, we get
the final priority figure with linguistics. We get our final priority list by the priority
figure (Fig. 159.4).
159.2.5 Step5
In step5, our goal is to locate a discriminate line to help distinctly categorize the
unknown workers using the ideas of cluster analysis and method of hypothesis
testing.
159 Modeling for Crime Busting 1513
We find a variable ‘AW1’ to describe the degree of conspiring for the group of
conspirator. ‘AW’ is defined by the following formula (159.7)
!
X67
AW1x ¼ Wi þ W83K K ð83 xÞ ð159:7Þ
i¼x
‘AW1’ means the average weight for the group of conspirators, and it stands for
the degree of conspiring. The ‘x’ is the abscissa of the point where the discriminate
line located at. ‘Wi’ is defined in formula (159.2). ‘K’ is the amount of the known
conspirators. We consider the suspicious degrees (‘Wi’) of all known conspirators
are all same and the value of these suspicious degrees is same as the most right
point in Fig. 159.4.
(a) Draw the Fig. 159.5 that shows the changing of ‘AW1x’ by growing of the
‘x’.
(b) For the same reason, we can define ‘AW2’ to describe the degree of con-
spiring for the group of non-conspirator, using formula (159.8).
!
X
i¼x
AW2x ¼ Wi þ WL L ðL þ XÞ ð159:8Þ
1
‘L’ is the amount of the known non-conspirators. We consider the suspicious
degrees (‘Wi’) of all known non-conspirators are all same and the value of these
suspicious degrees is same as the most left point in Fig. 159.4. And we can draw
Fig. 159.6 shows the changing of ‘AW2x’ by growing of the ‘x’.
Fig. 159.5 Changing of ‘AW1x’ by growing of the ‘x’
1514 D. Sun et al.
Fig. 159.6 Changing of ‘AW2x’ by growing of the ‘x’
We finally give different discriminate lines and evaluate the solution according
to different probability of first type error and second type error to fit different
requires of the police.
The first type error in our model is let the conspirators get away with crime. We
describe the probability of first type error by ‘P1%’ using formula (159.9)
0x 1
R2
B f ðxÞdxC
B C
P1% ¼ 1 B xx2 C 100% ðx1 x x2 Þ ð159:9Þ
@R A
f ðxÞdx
x1
The f(x) is the function of the curve we match in Fig. 159.5. X1 and X2 is left
and right point’s abscissa.
The second type error in our model is let the non-conspirators be treated as
conspirators. We describe the probability of first type error by ‘P2%’ using formula
(159.10)
0x 1
R
B gðxÞdxC
Bx C
P1% ¼ 1 B x12 C 100% ðx1 x x2 Þ ð159:10Þ
@R A
gðxÞdx
x1
159 Modeling for Crime Busting 1515
The g(x) is the function of the curve we match in Fig. 159.6. And we finally
find that P2%, the probability of second type error, is decreasing with the
increasing of ‘x’.
If we change the value of ‘x’, the probability of the first and second type error
will change too. At last we find when ‘x’ = 55, the value of P1% ? P2% is the
smallest. So we recommend police to locate discriminate line at the point whose
abscissa is 55 in Fig. 159.4 (Li and Zhu 2008; Guo and Zhu 2005).
159.2.6 Step6
We try to find the boss of the crime group using the concept of point centrality in
study of social network analysis (Ma and Guo 2007).
Now that we know Jerome, Dolores and Gretchen are the senior managers of the
company. If one or two, or even all of these three people are in the list of con-
spirators, we can credibly make sure that the leader or leaders come from the group
of the three managers (Estevez et al. 2007; Santos et al. 2006; Kiss et al. 2006).
If all of them are not in the primary priority list, it will become more complex.
Assuming the crime group is isolated from the other group, that is to say, it has little
connection to the outside, so we can focus only on them. From previous work, we
can obtain the criminal topics and their Sd(k). So we can calculate everyone’s
weight of point centrality by the same formula as formula (159.2). If someone’s
weight is much higher than others, we can surely know that he is the leader
(Klovdahl et al. 1994; Klovdahl 1985; Peiris et al. 2003; Svoboda et al. 2004).
159.3 Sensitivity Analysis and Model Evaluating
In our models, the value of weight in Table 159.1 and Sd(k) are defined by
ourselves through perceptual knowledge and some experiments of the example of
Investigation EZ. That is to say, the weight and Sd(k) have no certain standard, so
it is necessary for us to make sure how it will affect our results.
There are 18 weights needed in our models, named A1, A2, A3, A4,
A5……A18, which you can see them in Table 159.1. We choose A3 and A16
randomly. With previous value, we get the priority list as follows (be expressed by
the code for unknown workers):
3, 32, 15, 37, 17, 40, 10, 81, 34, 22, 31, 13 ……
First, change A3 from 4 to 5, we got the result:
3, 32, 15, 37, 40, 17, 10, 81, 4, 34, 31 ……
Then, change the value of A16 from 1 to 2, we got result:
3, 32, 15, 37, 17, 40, 10, 81, 34, 22, 31, 13 ……
The basic value of Sd(k) is defined to be 1 when topic k has nothing about
crime. Otherwise, the topic is suspicious. The problem is that we didn’t define the
1516
Table 159.2 The results
No. Known criminals Known non-criminals Criminals Similar Time to count
1 7, 18, 21, 43, 49, 54, 67 0, 2, 48, 64, 65, 68, 74, 78 7, 18, 21, 43, 49, 54, 67, 3, 32, … – 28
2 – 0, 2, 48, 64, 65, 68, 74, 78 67, 21, 54, 7, 3, 43, 81, 49, 10, … 89.5 % 30
3 7, 18, 21, 43, – 7, 18, 21, 43, 49, 54, 67, 3, 32, … 89.5 % 31
49, 54, 67
4 – – 21, 67, 54, 7, 3, 43, 32, 2, 18, 1, … 84.2 % 33
5 7, 18, 21, 43 0, 2, 48, 64, 65, 68, 74, 78 7, 18, 21, 43, 67, 54, 3, 49, 17, … 94.7 % 29
6 7, 18, 21, 43, 0, 2, 48, 64 7, 18, 21, 43, 49, 54, 67, 3, 17, … 94.7 % 25
49, 54, 67
D. Sun et al.
159 Modeling for Crime Busting 1517
maximum value of its initial value. The value will influent the results of formula
(159.2). In our models, we defined it as 2. Now, we analysis whether the small
change of the value will affect our results.
3, 32, 17, 15, 10, 37, 81, 40, 22, 16, 34, 4, 44 ……
Observing these results carefully, we can conclude that when we change the
weight or the maximum initial value of Sd(k), the results are not sensitive. We can
say that our models behave well during the process of sensitivity analysis.
Now we evaluate this model, because we don’t know whether it is stable or
accurate. We guess that our model seems rely on the initial conditions. If it relies
tightly on initial conditions, the model may not be trustful because no one can
make sure the initial condition. So we will show that how our model’s result may
change when the initial conditions change. And we only take the conspirators who
are at the top of the priority list into consideration. These are some extreme
conditions.
Condition 1. Set initial conditions as normal, this result can be the basic
standard.
Condition 2. Assume we cannot identify the conspirators;
Condition 3. Assume we cannot identify the non-conspirators;
Condition 4. Assume we cannot identify all of them;
Condition 5. Assume we can only identify some of the conspirators, such as 7,
18, 21, 43;
Condition 6. Assume we can only identify some of the non-conspirators, such
as 0, 2, 48, 64 (Marsden 2002; Newman 2003).
Through the analysis of Table 159.2, we conclude some laws:
(1) The initial conditions about ‘Known conspirators’ and ‘Unknown conspir-
ators’ will affect the results, but the effect is tolerable; (2) The more accurate initial
conditions there are, the more fast and accuracy the results have; (3) More initial
conditions means more accuracy and less time(especially for large data base),
however it also means more energy it costs; (4) Our model has great stability so
that it can be used widely, and it will show strong adaptability (Anderson and May
1992).
References
Anderson RM, May RM (1992) Infectious diseases of humans: dynamics and control. Oxford
University Press, Oxford
Estevez PA, Vera P, Saito K (2007) Selecting the most influential nodes in social networks.
Proceedings of international joint conference on neural networks. [s.n.], Orlando, FL, USA,
pp 12–17
Guo L, Zhu Y (2005) Application of social network analysis on structure and interpersonal
character of sports team. 2005 China Sports Sci Technol 41(5):10–13
Kiss C, Scholz A, Bichler M (2006) Evaluating centrality measures in large call graphs.
Proceedings of the 8th IEEE international conference on e-commerce technology and the 3rd
1518 D. Sun et al.
IEEE international conference on enterprise computing, e-commerce, and e-services. IEEE
Computer Society Press, Washington, DC, USA
Klovdahl AS (1985) Social networks and the spread of infectious diseases: the AIDS example.
Soc Sci Med 21:1203–1216
Klovdahl AS, Potterat JJ, Woodhouse DE (1994) Social networks and infectious diseases: the
Colorado Springs Study. Soc Sci Med 38:79–99
Li L, Zhu Q (2008) DSNE: a new dynamic social network analysis algorithm. J Jilin Univ (Eng
Technol Edn) 38(2):139–143
Ma Q, Guo J (2007) A study of the pattern-based clustering theories. Yanshan University, March
2007
Marsden PV (2002) Egocentric and sociometric measures of network centrality. Soc Netw
24:407–422
Newman ME (2003) Properties of highly clustered networks. Phys Rev E Stat Nonlin Soft Matter
Phys 68:026121. Epub: 21 Aug 2003. doi:10.1103/PhysRevE.68.026121
Peiris JS, Yuen KY, Osterhaus AD (2003) The severe acute respiratory syndrome. N Engl J Med
349:2431–2441
Santos EE, Pan L, Arendt D (2006) An effective anytime anywhere parallel approach for
centrality measurements in social network analysis. Proceedings of 2006 IEEE international
conference on system, man, and cybernetics. IEEE Press, 2006:8.11
Svoboda T, Henry B, Shulman L (2004) Public health measures to control the spread of the
severe acute respiratory syndrome during the outbreak in Toronto. N Engl J Med
350:2352–2361
Chapter 160
Personalized Emotion Model Based
on Support Vector Machine
Jin-bin Wu and Wan-sen Wang
Abstract Emotion deficit is an intelligent in e-learning technology research. The
main purpose of the paper is based on Support Vector Machine (SVM) through the
samples data analysis of the face area, interpupillary distance, eye spacing and
mouth curvature to build to the aversion degree, cheer degree and pleasure degree
based emotion model of personality academic emotions. All of these lay the
foundation for emotional teaching in E-Learning system.
Keywords Academic emotions Emotion deficit E-Learning Support vector
machine
160.1 Introduction
Analysis of the existing E-Learning system, we can easily find a common phe-
nomenon: the current system is often web-based information technology
‘‘boilerplate’’ are text-based teaching can be seen, usually posted on the Internet,
teaching practice, and related methods, this teaching method is indifferent learning
lack of personalized teaching guidance, we generally call ‘‘emotional deficit’’
(Zhang 2009). The importance of emotions in the E-Learning, University of
Adelaide Professor Kerry O’Regan, told that students conducted a survey of dis-
tance learning, she found that the emotion was the key to learning networks and
was the essential factor in the teaching and learning process (O’Regan 2003). In
addition, according to the psychological studies, emotional factors have an
important impact on the learning behaviors (Su and Xu 2009).
J. Wu (&) W. Wang
Information Engineering Institute Capital Normal University, Beijing 100048, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1519
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_160,
Ó Springer-Verlag Berlin Heidelberg 2013
1520 J. Wu and W. Wang
The main purpose of these researches is the introduction of emotional teaching
function in the traditional E-Learning. With the rapid development of information
technology in the field of education, e-learning has rapidly changed distance
education. But after the initial practice boom retreated, people gradually return to
rationality. E-Learning has many advantages, but due to the lack teachers to
participate and can not completely replace the classroom educational activities of
teachers and students learning is not as conceived so perfect. This emotional
missing affected the teaching effectiveness and widely used of distance education.
Meanwhile, Chinese educational psychologist Professor Qing-lin Wu pointed out
that intelligent e-learning system is not only a truly personalized teaching system
but also is emotional (Wu 2003).
160.2 Basic Conception
This research focuses on learners’ emotions and emotional modeling. In general,
the mood in psychology with a number of different classification methods, the each
different method can be divided into several different types.
Emotions have physiological and explicit characteristics: the physiological
characteristics mainly refer to the obvious physiological changes inside the body,
such as breathing, heartbeat, etc. The explicit features refer to the physical external
changes, such as smiling, frowning, etc. The explicit characteristics often refer to
the expression. Different emotional characteristics access to information is dif-
ferent. Physiological characteristics of access to information usually require a
physical means of galvanic skin contact. While expression information can be
obtain by video, audio and other non-contact means. Visible, in network learning,
expression method access to information is practical and significant.
The expression can be divided into three types, such as facial expressions, body
expressions and language expressions. Because the expression is the external
expression of emotions, so if a different expression has recognized, it can be
judged by different expression from different emotions of the people, and also by
the different expressions of the people to express different emotions. In the three
expressions, the facial expression is the most identification of emotional signs, the
most sophisticated to identify the specific mode of the different nature of emotions.
Emotional modeling is the core of this paper and important aspects of emotional
information in E-Learning. In this paper, the extraction of facial expression
characteristics, its steps, the first analyses the sample data for stability and reli-
ability, and then introduce aversion degree, cheer degree and pleasure degree to
describe the emotional state of the learners. Select a few learners to information
collection and analysis the image filter as a sample. Experiment in a variety of
environments, the rapid classification of the characteristics of small samples by
Support Vector Machine (SVM), build the aversion degree, cheer degree and
pleasure degree of sample data for modeling and analysis.
160 Personalized Emotion Model Based on Support Vector Machine 1521
OCC emotion model is proposed by a book named emotional cognitive struc-
ture, write by Ortony A, Clore G, Collins A, which is also the earliest and most
complete one of the model study of human emotions. According to the emotional
causes, OCC emotion model is divided into the following three categories: the
results of the event, the action of the agent, the object perception. The model
defines a total of 22 species basic emotions and the relationship between their
levels. The OCC model is not a basic emotion set or a clear multi-dimensional
space to express their feelings to express emotion, but with consistency cognitive
export conditions. In particular, in the model assumes that the emotional satis-
faction and dissatisfaction Agent, happy and unhappy event, and the likes and
dislikes object, they constitute the reaction of the situation in a positive or negative
tendency. The model summarizes the standards, which includes 22 kinds of
emotional type used to generate the basic structure of the rules, and these emo-
tional types are derived by different cognitive conditions.
In this paper, academic emotions, defined as three dimensions (Wang 2009),
discuss the reverse of each other in the most common of the six academic emo-
tions: interested, bored, excited, tired, happy and distress. In the basic emotional
space, A represents the interest, B is boredom, C is excited mood, D is fatigue, E is
a happy and F is distressed. The origin of coordinates is removed from the
emotional space, because the origin of coordinates is emotional state, so it does not
meet the normal human emotions.
160.3 Support Vector Machine Introduction
Support vector machine (Cores 1995) designed to solve nonlinear problems of the
small sample study and classification. On the one hand, to overcome, the least
squares method is too simple, can not distinguish between complex nonlinear
classification designs; on the other hand, support vector machine has good clas-
sification ability, but neural network had the problem of overfitting and underfit-
ting. SVM technology, the most critical is the selection of the kernel function:
different kernel functions have a great impact on the classification results. There
are several different ideas in the selected kernel function to solve practical prob-
lems: The first use of an expert transcendental knowledge for kernel function
selected; the other is the Cross-Validation method, kernel function selection
process, experimenting with different kernel functions and parameters; the third
using mixed kernel function method (Smits et al. Proposed), which is using dif-
ferent kernel functions combined to obtain better performance, it is also the basic
idea of the mixed kernel function. On the whole, the parameter selection problem,
in essence, is an optimization problem.
In this paper, using the main advantage of the SVM algorithm is to classify
training data characteristics of the facial expression modeling and has obtained good
experimental results. In the research, using libsvm (http://www.csie.ntu.edu.tw/
*cjlin/libsvm/index.html) toolbox in Matlab, it is developed by National Taiwan
1522 J. Wu and W. Wang
University Professor LinChin-jen. The aim is to design a simple, easy to use support
vector machine (SVM) to solve pattern recognition and linear regression package.
The software not only provides a compiled version of Microsoft Windows, but also
other operating systems. The executable file is open source code, facilitate others to
improve and modify; the software can solve the C-support vector (C-SVC), nu-
support vector classification (nu-the SVC), one-class SVM (distribution estimation),
epsilon-support vector regression (epsilon-SVR), nu-support vector regression (nu-
SVR) and other problems, including based on one-on-one algorithm to solve the
related many types of algorithm’s problems. Support vector machine is used to solve
the problem of pattern recognition or regression, the international scientific com-
munity has not yet formed a unified view the parameters choice and the choice of
kernel function. This also means that the parameter selection of the optimal SVM
algorithm can only use the excellent previous experience, or comparative experi-
ments, or large-scale search, or use the package cross-validation function. Also use
other algorithms to achieve optimization, the algorithms such as genetic algorithm
(Kang et al. 2011), particle swarm optimization (PSO) (Chen and Mei 2011) and cats
swarm optimization (CSO) (Wang and Wu 2011). In this paper, due to the com-
plexity of the experimental constraints, only to choose the experts and experience
the results of selected parameters.
160.4 Comprehensive Experiment
Due to the complexity of the human face, related to the study about this experi-
ment, I proposed the three main concepts are aversion degree, cheer degree and
pleasure degree. Aversion is based on the face area and interpupillary distance to
locate positioning method. Positioning and calculation of the face and the pupil of
the eye is to determine the learners in the learning process, interested in learning
the current content. Under normal circumstances, when the detected face area and
interpupillary distance is larger, which means learners leaned forward in the
learning process, learning content is relatively interested in, aversion for bigger;
On the contrary, when the changes in the hours means that learners lean back, and
not interested in learning content, and even boredom, aversion for smaller.
Similarly, cheer degree of detect eye spacing heterozygosity is to describe and
judge the cheer extend. And pleasure degree, through the mouth upturned angle to
detect degree of pleasure in the learning process.
Verify the stability of the data to prove its stability. A learner and B learner
within two hours (every 60 s tested once) detected in the normal state of learning
data. Because of space constraints, I only cite the face area and interpupillary
distance of data analysis figure, similar to other situations.
Detected sample data (the face area and interpupillary distance) really focused
within a certain range. Accordingly, we propose a hypothesis: the face area and
ongoing testing to get a sufficient amount of data, we believe that it is possible to
meet the normal distribution, if that were true, then we can change the scope of the
160 Personalized Emotion Model Based on Support Vector Machine 1523
Fig. 160.1 Face area of a
learner the normal reference
curve
already mentioned, and then detect whether the current learners in the normal
learning state. Figure 160.1 shows the results of tests of the data sample, to
demonstrate that they meet the assumption.
160.4.1 Input Variable Selection
Aversion degree of face area and interpupillary distance, and the statistical analysis
of previous data have been found that these two sets of data for normal distri-
bution, indicating that learner’s mood is relatively stable over a period of time.
We put this the face area and interpupillary distance of 120 sets of data in two
different learners classification preprocessing, select 100 as the training set, the
remaining 20 as test set. Consider a simplified classification of emotions into the
four categories, which are very interested, interested, tired and very tired. So one
On behalf of very tired, two is tired, three is interested and four is very interest,
enter is test data of human face area and interpupillary distance. Figure 160.2
shows the relationship between category labels and face area, interpupillary dis-
tance, asterisk is the distribution of sample points.
160.4.2 Data Preprocessing
Training and test set were normalized preprocessing.
1524 J. Wu and W. Wang
Fig. 160.2 Aversion degree
labels and property
distribution
160.4.3 Training and Prediction
Kernel function to select the radial basis kernel function, the function C is selected
as 1000.
160.4.4 Analysis of Experimental Results
Operating results: Accuracy = 70 % (14/20) (classification).
The final classification results are as follows:
The experimental results can be seen from Fig. 160.3, the blue represents the
classification of the actual test set, and red represents the prediction set
Fig. 160.3 Aversion degree
of classification results
160 Personalized Emotion Model Based on Support Vector Machine 1525
classification, the classification accuracy is 70 %, the basic realization of the
successful implementation of the aversion degree modeling and analysis. Suc-
cessful implement of the mapping from the face area and interpupillary distance to
four different emotions.
Other, cheer degree and pleasure degree have the similar results. Accuracy can
reach 95 %, to obtain good experimental results.
160.5 Conclusion
This paper is using fast learning classification adopting support vector machine
network of small sample nonlinear characteristics based on the OCC emotion
model, aversion degree, cheer degree and pleasure degree to establish academic
emotions model in E-Learning. The model provides the necessary basis of aca-
demic emotions, is also a useful attempt of the SVM algorithm in the field of
emotion recognition, achieved the good results.
Acknowledgments The research is supported by the National Natural Science Foundation of
China (Grant No.60970052) and Beijing Natural Science Foundation (The Study of Personalized
e-learning Community Education based on Emotional Psychology).
References
Chen W, Mei Y (2011) Research on forecasting methods for reduction ratio of pore closure in
forging stock based on PSO and SVM. Comput Eng Appl 47(27):243–245
Cores C (1995) Vapnik. Support vector networks. Mach Learn 20:273–297
[EB/OL] http://www.csie.ntu.edu.tw/*cjlin/libsvm/index.html
Kang H, Li M, Zhou P, Zhao Z (2011) Prediction of traffic flow using support vector machine
optimized by chaos higher efficient genetic algorithm. J Wuhan Univ Technol (Transportation
Science and Engineering) 35(4):649–653
O’Regan K (2003) Emotion and e-learning. JALN 7(3):78–92
Su X-p, Xu Y-x (2009) Intelligent E-learning system having emotion interaction function.
Comput Eng Des 30(15):3690–3693
Wang Z-l (2009) Artificial emotion. Mechanical Industry Press, Beijing
Wang W-s, Wu J-b (2011) Emotion recognition based on CSO&SVM in e-learning, ICNC
Wu Q-l (2003) Educational psychology—the book dedicated to the teachers. East China Normal
University Press, Shanghai
Zhang X-y (2009) Framework of an E-Learning system based on affective computing. J Hunan
Inst Sci Technol (Natural Sciences) 22(4):51–54
Chapter 161
Research on Decision-Making Behavior
Test System for Top Management Team
Based on Simulation Environment
Xue-ying Hong, Zhu-chao Yu, Zhu Wang and Yang Jiang
Abstract The decision made by Top Management Team is fatally important for
business operation. So, how to improve the quality and reliability of decision-
making seems very necessary. Starting from the Prospect Theory of behavioral
decision-making theory, this paper puts forward testing decision-making behaviors
of Top Management Team, and analyzes the specific process and methods of
decision-making. According to results of the decision-making behavior testing, the
characteristics of Top Management Team can be obtained, and so as to provide
reasonable foundation for evaluation and improvement of decision-making
behaviors.
Keywords Behavior testing Decision-making behavior Decision simulation
Top management team
161.1 Introduction
The Nobel Prize Winner Herbert Simon used to say that ‘‘management is making
decision’’, which reveals how important the decision-making is in business
administration. With the global economic integration goes further in China, drastic
market competition and rapid changes of information revolution, diversification
oriented business and close coordination oriented department all present new
challenges to executive leaders (Ancona and Nadler 1989). At the same time, team
decision-making gradually takes place of personal decision-making and is
X. Hong Z. Yu (&) Z. Wang Y. Jiang
Northeastern University of Business Administration, ShenYang 110004, LiaoNing, China
e-mail: [email protected]
X. Hong
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1527
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_161,
Ó Springer-Verlag Berlin Heidelberg 2013
1528 X. Hong et al.
becoming more and more important in business administration. By full under-
standing of Top Management Team (TMT) on decision-making behavior char-
acteristics, we can make better use of the advantages of certain teams and achieve
the effect that team decision is superior to individual decision, thus we can keep
continual development of business in the long run.
Katzenbach defined TMT as ‘‘top leaders team in organizations and institu-
tions’’, they usually have the power to solve problems, coordinate activities,
mobilize organization members and make important decisions of the company.
Because of the importance of the TMT, researchers in both domestic and overseas
do a lot of researches on behaviors of TMT. Henning Bang and his partners got the
conclusion that goal clarity and focused communication is positively related to
team effectiveness by self-report and observer data from eight top management
groups that processed 56 agenda items during meetings, thus provided instructions
on how to make decisions efficiently. At the same time, they extended three
dimensions related to team efficiency: Task performance, relationship quality, and
member satisfaction (Bang et al. 2010). From two total different aspects: Func-
tional-background and locus-of-control, Boone and Hendriks (2009) analyzed the
compositional diversity and organizational performance by collecting and ana-
lyzing the data collected from scientific and technical corporations, and they
finally provided instructions on the optimizing of team composition.
In this paper, we introduce decision simulation system into the research of
decision-making test of TMT, so we can avoid the problems that are met in
traditional surveys, such as information distortion and data ambiguous. This paper
also takes advantage of management simulation programs to create management
decision scenarios which are very similar with real market. By data collection and
result analyzing, we can obtain decision-making characteristics of TMT, and
finally help TMT to improve their decision performances.
161.2 Method
161.2.1 Behavioral Decision-Making Effect
Behavioral decision-making theory is a new theory developed to solve problems that
are difficult to rational decision-making theory; classical decision theory also can be
regarded as special case of behavioral decision-making theory when there are a lot of
hypotheses. In 1979, Kahneman and Tversky (KT) put forward the Prospect Theory
under uncertain conditions of personal decision behavior based on economic
experiments (Kahneman and Tversky 1979), thus denied the Expected Utility
Maximization Theory proposed by von Neumann and Morgenstern (1944).
Expected Utility Maximization Theory holds the opinion that the preference of a
decision maker is changeless, and his decisions can be predicted by statistical
methods. On the contrary, Prospect Theory indicates that there exists Framing
161 Research on Decision-making Behavior Test System 1529
Effect, Reference Point Effect, Deterministic Effect and other effects caused by
irrational behaviors under uncertain conditions, what is more these irrational
behaviors make individual decisions betray the Expected Utility Maximization.
Framing Effect manifests that different formulations can lead to different prefer-
ences to the same problem. Deterministic Effect manifests that decision makers have
obvious preference on certainty. In fact, these two effects are both caused by the
changing of reference point which has a big influence on individual decisions. In
different fields, different environments and among different individuals, decision
makers’ reference points may change, and we define this change as Reference Point
Effect. For example, through experiments Zhan-lei Li validated that there exist
Framing Effect, Reference Point Effect and Deterministic Effect in individual
decisions under the environment of economy, society and culture (Li et al. 2007).
In view of different behavioral decision-making effects’ different influences, in
this paper we mainly use four effects (Reference Point Effect, Framing Effect,
Fuzzy Avoid Effect and Deterministic Effect), which are extended from Prospect
Theory, to test decision behaviors of TMT under different decision scenarios and
analyze their behaviors’ characteristics.
161.2.2 Management Simulation
Management Simulation is a kind of Computer simulation system using Computer
Science, Management, Game Theory and Operational Research, and it is used to
simulate management activities of business. The earliest Management Simulation
came from some of American famous universities in 1850s, and the first university
to use it was the University of Washington. In 1957, the University of Washington
used a simulation system named High-level decision-making and management in
the course of Business guidelines (Wang 1999). According to the data of 2001,
more than half of the core members belonging to the Association to Advance
Collegiate Schools of Business (AACSB) widely use Management Simulation in
such courses as Strategic Management, Marketing, Accounting and Finance, and
there are also Management Simulation Competitions in America and some other
regions. At present, there are mainly 3 kinds of Management Simulation in the
world, these are Management Game governed by Carnegie Melton University
(CMU), MBABEST21 governed by CSIM College of Aoyama Gakuin University
and Global Management Challenge (Xiao 2001).
When taking part in these Management Simulation systems, corporate execu-
tives will form several teams. These teams compete with each other and try to
improve their own business’s performance. As a result, they can experience
management, use theory and cultivate innovative thinking (Iwai 2007).
Aiming at testing subjects’ decision-making behaviors, this system takes
advantage of parameter model used in traditional Management Simulation systems
when design testing scenarios. Besides, we create various kinds of testing
1530 X. Hong et al.
scenarios based on four basic effects. At last, we can analyze their decision-making
characteristics by the data we collected.
161.2.3 Methods of Decision-Making Testing
From the middle of 1970s, behavioral decision-making had become an indepen-
dent subject, and was widely used in areas of economy, finance and management.
In this stage, research methods includes observational method, investigation
method (mainly are questionnaire survey and interview survey) and experimental
method (psychological experiment and economical experiment). Since then, these
methods are also called general empirical research of decision-making behavior
(Huang 2006). For example, by using ‘‘Asian disease problem’’, Tversky and
Kahneman testified the existing of Framing Effect and successfully questioned
traditional invariance (Kahneman and Tversky 1981).
At present, researches of behavioral decision-making theory mainly focus on
summarizing behavioral characteristics and refining the behavioral variables, then
apply it to analysis of rational decision-making. Representative studies of this kind
of researches include such four investor psychology models as BSV model
(Barberis et al. 1998), DHS model (Daniel et al. 1998), HS model (Hong and Stein
1999), BHS model (Barberis et al. 2001), and Behavioral Asset Pricing Model
(Shefrin and Statman 1994), Behavioral Combination Model (Shefrin and Statman
2000).
Being different with general research methods, this paper’s method uses logical
processes to test TMT’s decision-making behavior under management simulation
scenarios. That is testing decision makers’ behaviors and analyzing experiment
results by controlling some variables under controlled experiment conditions, thus
reinforce the reliability of the results. Besides, high emulation situation have
directive functions in real life.
161.3 System Design
161.3.1 Overall Structure of the System
Aiming at testing decision-making behavioral characteristics of TMT, this system
includes the function of managing multitask scenarios.
There are two main subsystems: Testing subsystem and analyzing subsystem.
The overall structure of the system is shown in Fig. 161.1. As we can see, there are
two modules in testing subsystem which are module of testing online and module
of information services, and two modules in analyses system: Module of data
statistics and module of management of testing scenarios. At first, subjects log in.
161 Research on Decision-making Behavior Test System 1531
Fig. 161.1 Overall structure
of decision-making behavior Login
testing Subjects Manager
Testing online Data
Statistic
Data-
base
Management
Information of testing
services situation
Secondly, they should input their team information and then they can choose
testing scenarios. After all of these, they will enter into testing subsystem to finish
the whole processes under the guide of the system.
161.3.2 Design of Database
Database is used for depositing testing data of testing subsystem and system
parameters of simulation programs. In testing phrases, subjects input decision
variables and decision values as the simulation system asked. Then, the system
will use relevant decision parameters that are set by managers in advance to
calculate the results of simulation operation. Finally, the results will be seen in the
interface as the form of reports; in decision-making behaviors analyzing phrase,
managers take advantages of statistic analysis software to analyze all the decision
data and give the results of the subjects’ characteristics of decision-making
behaviors.
161.3.3 Design of Function Module
This system obtains results of decision-making testing and of behavioral testing by
calculating various of function modules, which includes module of parameter
setting, module of operation calculating and modules of decision behavior ana-
lyzing. Among these modules, module of parameter setting is used for creating
decision simulation environment and reducing errors to real situation; module of
operation calculating, which includes demand function, constant cost function,
variable function and so on, is used for calculating the operational results based on
1532 X. Hong et al.
the values inputted by subjects; module of decision behavior analyzing is used for
analyzing the decision data and the results of operation, and finally the testing
results by classifying subjects’ behaviors based on four effects mentioned before
can be obtained.
161.4 System Implementation
Based on decision-making testing scenarios and researches of four effects, variable
testing scenarios can be designed to cater for different demands of subjects.
According to different testing scenarios and relevant parameters this system can
provide special decision-making situation for subjects.
There are mainly four functional modules in this system: Situation manage-
ment, testing online, data statistic analysis and information services.
161.4.1 Situation Management
This function includes adding testing scenarios that have been designed, modifying
relevant marketing parameters in every progress according to different behavioral
decision-making effects, editing or deleting certain situations that are not signifi-
cance in testing phrase. Only managers have the right to modify parameters to
cater for demands.
161.4.2 Testing Online
Testing online mainly provides functions that can be used to test subjects’ deci-
sion-making behaviors. At first, subjects log in main interface, then under the
guide of the system they can implement testing needed. This system provides
individual settings, which are single-period testing and multiple rounds of simu-
lations decision testing, in view of different testing aims. In single-period testing,
system will guide subjects to another decision situation after they finished their
first decision-making and provide simulation operation results. Subjects will be
tested a lot of times under certain testing scenarios, where only some certain
parameters will be changed in order to control certainty of simulation situation,
until enough data has been collected. In multiple rounds of simulations decision
testing, subjects will be required to manage a company for several periods and
testing scenarios will be changed along with subjects’ decisions. And different
with single-period testing, subjects’ decisions will always affect next period in
multiple rounds of simulations decision testing.
161 Research on Decision-making Behavior Test System 1533
161.4.3 Data Statistic Analysis
After all the tests are over, managers will use statistical analysis software to
analyze data which subjects submitted, and final results will be obtained.
161.4.4 Information Services
There are mainly two aspects: In multiple rounds of simulations decision testing,
operational results will be provided and can be searched; after tests over, decision-
making behavioral characteristics and relevant suggestions will be provided. The
process of searching is showed in Fig. 161.2.
161.5 Summary
The trend of economic globalization has brought great challenges to companies, so
how to ensure the quality of executives in decision-making will become increas-
ingly important. Decision simulation system, which integrates application of
management science, decision science, computer technology and IT, can provide
managers with a realistic management environment and good experimental envi-
ronment for researches on decision-making behaviors. Based on the four behav-
ioral decision-making effects of prospect theory, we develop a testing system to
test executives’ decision-making behaviors. Using this system, TMT can obtain
their different decision-making preferences under the conditions of uncertainty and
understand their own decision-making behavior characteristics. Thus, they can
Fig. 162.2 Process of
searching User Login main
interface Information services
Yes Select criteria
Other
operations Searching criteria
Whether
No do further Getting the
searching result
1534 X. Hong et al.
avoid decision-making bias in a major decision and improve their decision-making
as a whole quality. What’s more, the testing system also provides practical
directions on the composition of TMT according to their different decision-making
behavioral characteristics.
Acknowledgments This work was partly supported by the National Science Foundation of
China (Project No. 71171043), the National Basic Scientific Research Expenses—National
Project Breeding Fund (Project No. N090406006) and the National Undergraduates Innovating
Experimentation Project ‘‘Under Team Task Situations Decision-Making Behavior of Business
Executives Test Platform’’ (Project No. 110105).
References
Ancona DG, Nadler DA (1989) Top hats and executive tales: designing the senior team. Sloan
Manage Rev 31(1):19–28
Bang H, Fuglesang SL, Ovesen MR, Eilerten DE (2010) Effectiveness in top management group
meetings: the role of goal clarity, focused communication, and learning behavior. Scand J
Psychol 51:253–261
Barberis N, Huang M, Santos T (2001) Prospect theory and asset prices. Q J Econ 116(1):1–53
Barberis N, Shleifer A, Vishny R (1998) A model of investor sentiment. J Financ Econ
49:307–343
Boone C, Hendriks W (2009) Manage Sci 55(2):165–180
Daniel K, Hirshleifer D, Subrahmanyam A (1998) Investor psychology and security market under
and overreactions. J Financ 53:1839–1885
Hong H, Stein J (1999) A unified theory of under reaction, momentum trading, and overreaction
in asset markets. J Financ 54:2143–2184
Huang C (2006) Discussion on behavioral decision theory and decision-making behavior of
empirical research methods (in Chinese). Economic Survey, No. 5
Iwai C (2007) Development of MBABEST21: a case-based management game. http://
www.MBABEST21.org
Kahneman D, Tversky A (1979) Prospect theory: an analysis of decision under risk.
Econometrica 47:263–295
Kahneman D, Tversky A (1981) The framing of decisions and the psychology of choice. Science
211:453–458
Li Z-L, Li H-M, Li N (2007) Experimental verification of the non-rational behavior of individual
decision-making (in Chinese). J HeBei Univ Eng 24(2):14–18
Shefrin H, Statman M (1994) Behavioral capital asset pricing theory. J Financ Quant Anal
29(3):323–349
Shefrin H, Statman M (2000) Behavioral portfolio theory. J Financ Quant Anal 35(2):127–151
von Neumann J, Morgenstern O (1944) Theory of games and economic behavior. Princeton
University Press, Princeton
Wang Q-W (1999) Economics and management computer essentials (in Chinese). Higher
Education Press, Beijing, pp 245–246
Xiao X-D (2001) Evolution and revolution of the development of modern management
simulation (in Chinese). Mod Manage 5:50–52
Chapter 162
An Exploratory System Dynamics Model
of Business Model Evolution
Xiao-lin Shao and Ju-wei Shi
Abstract The evolution of the business model is a complex dynamic process and
shows abundant dynamic characteristics, which are the integrated effects of sur-
roundings and inter structure of the system. By developing causal loop diagrams
and stock and flow diagrams, we build a scientific dynamic system model to
identify the relationships among key variables in the business model and probe its
evolutionary dynamics. As a result, we are able to lay the foundation for further
research of business model evolution.
Keywords Business model evolution Causal loop diagram Stock and flow
diagram System dynamic model
162.1 Introduction
The business model (BM hereafter) concept became prevalent in the mid 1990s
with the advent of the Internet and its massive adoption for e-commerce (Amit and
Zott 2001), rapid growth in emerging markets and interest in ‘‘bottom-of-the-
pyramid’’ issues (Seelos and Mair 2007), as well as expanding organizations
dependent on post-industrial technologies (Perkmann and Spicer 2010), and it has
been gathering momentum since then.
In fact, each firm has its unique BM from its foundation. How to deal with the
evolution of many elements and their interactions in each BM subsystem is one of
X. Shao (&)
School of Management, Zhejiang University, Hangzhou, China
e-mail: [email protected]
J. Shi
College of Mechanical and Transportation Engineering, China
University of Petroleum, Beijing, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1535
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_162,
Ó Springer-Verlag Berlin Heidelberg 2013
1536 X. Shao and J. Shi
the main problems bothering all firms which want to gain long-term survival and
sustained competitive advantage in a changing market. However, the academic
research on BM seems to lag behind practice and lacks the systematic and dynamic
view, which leads to the confusion of management. Consequently, in practice,
many managers don’t know when and how to implement BM change; even if they
take action, the results may be contrary to expectations due to the lack of system
thinking or ignore delays in the system, triggering poor dynamic behaviors in the
evolution of BM.
162.2 Literature Review
The initial research on the business model took a static approach and focused on
the related concepts, structures and elements, etc. Recently, the dynamic approach
emerges, but only emphasizes on the impetus, approach, and implementation of
the BM innovation.
162.2.1 Review on the Static Perspective of Business Model
Despite the overall surge in the literature on BM, no generally accepted definition
of the term ‘‘business model’’ has emerged. At a general level BM has been referred
to as a statement (Stewart and Zhao 2000), a description (Applegate 2000), a
representation (Morris et al. 2005), an architecture (Timmers 1998), a conceptual
tool or model (Teece 2010). Morris et al. (2005) conducted a content analysis of key
words in 30 definitions to identify three general categories of definitions which can
be labeled economic, operational, and strategic, with each comprised of a unique set
of decision variables (Stewart and Zhao 2000).
Though diversity in the available definitions poses substantive challenges for
determining what constitutes a good BM, we can find ‘‘value’’ appears most
frequently through a literature review, as shown in Table 162.1.
162.2.2 Review on the Dynamic Perspective
of Business Model
The relationship between business model and time is little discussed, and the
dynamic perspective has only recently been incorporated into research on this
topic. Research at home and abroad explores the innovative impetus of BM from
perspectives of technology, demand, competition, executives and systems (Wang
and Wang 2010), which can be seen as the influencing factors of BM evolution.
162 An Exploratory System Dynamics Model 1537
Table 162.1 The components of business model
Source Specific components Number E-
commerce/
general
Gordijn et al. Actors, market segments, value offering, value activity, 8 E
(2001) stakeholder network, value interfaces, value ports,
and value exchanges
Applegate Concept, capabilities, and value 3 G
(2000)
Morris et al. Factors related to offering, market factors, internal 6 G
(2005) capability factors, competitive strategy factors,
economic factors, growth/exit factors
Benoît and Value proposition, resources and competences, 4 G
Xavier organization
(2010)
Zhang and Lei Value proposition, value network, value maintenance, 4 G
(2008) and value realization
Overall, the dynamic perspective tends to view the BM as a ‘‘black box’’ to
examine environmental factors affecting its innovation, which ignores the inter-
actions and changes among the elements within BM as well as its evolution in the
long run.
162.2.3 Summary of the Literature Review
Through literature review, we find that there is no denying the fact that BM is a
complex system of value creation. However, scholars cannot agree on what its
components are since it emerged from business practice recently and scholars
frequently adopt idiosyncratic definitions that fit the purposes of their specific
studies, which lacks the theoretical underpins and as a result are difficult to rec-
oncile with each other.
Studies on the dynamic perspective of the BM are relatively rare, partly due to
its debatable structure and also due to its research methodology. Existing resear-
ches are almost without exception qualitative case analysis which could not por-
tray and explain the dynamic characteristics and internal mechanism of BM
evolution.
The evolution of the BM is a complex dynamic process and shows rich dynamic
characteristics, which is the integrated effect of surroundings and inter structure. In
order to identify the mechanism and probe the evolutionary dynamics of BM, we
have to build a scientific dynamic system model. Due to limited space, this article
will focus on the various components of the BM and their interaction mechanisms,
taking external environmental factors as given exogenous variables. We use the
System Dynamics (hereafter SD) approach to establish the initial exploratory
1538 X. Shao and J. Shi
model of business model evolution system (hereafter BMES) building on the
resource-based view. We use causal loop diagrams and stock and flow diagrams to
reveal the interaction between the various elements within BMES, hoping to
provide a new perspective and method for the current research.
162.3 Method
System Dynamics, initially established by Professor Jay W. Forrester at MIT in
1965, is an approach to understand the behavior of complex systems over time
(Größler 2010). The SD model and its methodology put emphasis on endogenetic
view and contend that the system’s behavioral pattern and characteristics are
determined by its inter feedback structure and feedback mechanism.
SD uses causal loops to describe the feedback structure. There are two basic
causal loops, namely reinforcing loops (also called positive feedback loops) and
balancing loops (also called negative feedback loops). Reinforcing loops have an
even number of negative links and will generate behaviors of growth, amplify,
deviation, and reinforce. Balancing loops have an odd number of negative links
and tend to produce stable, balance, equilibrium and goal-seeking behavior over
time.
Causal loop diagrams aid in visualizing a system’s structure and behavior, and
analyzing the system qualitatively. However, since mathematical relationships of
different variables cannot be revealed by causal loop diagrams, stock and flow
diagrams are introduced to perform a more detailed quantitative analysis. A stock
is the term for any entity that accumulates or depletes over time. A flow is the rate
of change in a stock (Wang 1995).
There are four steps in SD: (a) problem identification and goal setting; (b) SD
model establishment, including causal loop diagrams for qualitative analysis and
stock and flow diagrams for quantitative analysis; (c) SD model simulation and
testing; (d) strategy choosing.
Causal loop diagrams and stock and flow diagrams are the key elements of SD
(Zhang et al. 2010) and make SD different from other approaches to study complex
systems which will be mainly discussed in the next section.
162.4 Model
The various variables involved in the process of BM evolution can be divided into
two types: state variables and influencing variables. These variables and their
mechanisms constitute a complex nonlinear dynamic feedback system referred as
BMES. While the state variables are all endogenous, the influencing variables can
be both endogenous and exogenous.
162 An Exploratory System Dynamics Model 1539
The creation of the state variables is essentially based on the understanding of
BM. However, many scholars describe the concept of BM from different points of
view and have not reached an agreement. This article examines the BM based on
Penrose’s resource-based view and contends that three core variables are included
in the BM, namely the resources scale, management capabilities and value prop-
osition, which can be well consistent with the essence of the BM. In addition,
resources, capabilities and value proposition can interact with each other and the
corresponding process is able to depict the evolution of BM. Continuing expansion
in the resource scale and the improved capabilities allow the generation of
remaining management capacities and remaining resources which make it possible
and necessary for firms to discover new market opportunities and provide new
value propositions, leading to the growth in the number of new customers.
However, a number of definitions in Penrose’s theory are not clear, nor did she
consolidate all the interactions mentioned into a clear logic. In this section, a
model of BMES will be established, containing the state variables and influencing
variables with their meanings and relationships elaborated. What is worth noting is
that in this article, the system boundaries will be set within the firm while variables
coming from political, economic, cultural, social, technological and market
environment are considered as given external variables.
162.4.1 Model Structure
BMES is divided into three subsystems as shown in Fig. 162.1: the subsystem of
resources scale, the subsystem of value proposition, the subsystem of management
capability, which are respectively corresponding to the three state variables. These
three subsystems connect with each other via variables and there is only one
Fig. 162.1 The model remaining
structure of BMES management
capabilies
the the
management resources
capability scale
subsystem the resource subsystem
dropout rate
management the remaining
capabilities remaining resource resources
building management investment
investment capabilities
the customer scale
(value proposition)
subsystem
1540 X. Shao and J. Shi
rate of + + cost to gain single
return on profits customer
profit market market scale
+ promotion
sales costs - +
+
+ + required
+
management
resource customer scale capabilities
investment + - customer
+ +
defection shortage of
+ market +
management
expanding
resource management capabilities
motive
scale knowledge -
+
- + + management
remaining
resource - capabilities
+
unit ratio of - remaining + -
resource resource management
cost dropout
+ capabilities
+ -
management daily
capabilities operation cost to establish unit
complexity of + management management capabilities
investment management
Fig. 162.2 The causal loop diagram of BMES model
negative feedback between the resource scale subsystem and the management
capabilities subsystem.
In summary, the main interactions between the three subsystems of business
model can be displaced in one causal loop diagram, as shown in Fig. 162.2,
including five enhancing loops and two balancing loops.
162.4.2 The Resources Scale Subsystem
The resources scale is a state variable, referring to the resources purchased from
the external market or internally developed by firms, including physical resources
(e.g., plant, capital), intangible resources (such as patents, trademarks), and
knowledge resources (existing in personnel, files, or other similar media) (Benoît
and Xavier 2010).
The complexity of daily operation management and the required management
capability change in the same direction with the resources scale as shown in the
Fig. 162.3. But the growth of resource scale will decrease the remaining man-
agement capabilities used to expand the resource scale, so that the growth rate will
be slowed down. The negative feedback loop, ‘‘resource scale-remaining
162 An Exploratory System Dynamics Model 1541
Fig. 162.3 The resource the remaining
scale subsystem resources + the market
expanding motive
+
the resource
scale
resource acumulation resources dropout
+
unit - + Daily operation +
resource resource
cost management
resource dropout rate
investment complexity
-
profits + + remaining
+ unit resource +
Required management
ratio of management management - capabilities
resource complexity
investment capability
on profits
management capability
-
required for unit
management complexity
management capabilities-resource dropout rate-resource scale’’, reflects the con-
straint effect of management capabilities on the growth of the resource scale.
Two variables in this subsystem, ‘‘resource investment’’ and ‘‘remaining
resources’’, are respectively connected with the ‘‘profits’’ and ‘‘market expanding
motive’’ in the value proposition subsystem. Thus, a positive feedback loop is
formed, that is ‘‘resource investment—resource scale—remaining resources—
market expanding motive—the customer scale—sales—profits—resource invest-
ment’’, which reflects the positive relationship between resource scale and cus-
tomer scale.
162.4.3 The Management Capability Subsystem
Management capability is a state variable referring to the managers’ ability to
improve and reconstruct the productive services provided by resources. As a gross
variable, management capability has to deal with two tasks: one is to manage the
daily operation of existing resources; another is to manage BM change. We
assume that only after daily tasks have been finished, can the management
capabilities be used to expand value proposition.
There are two sources to increase the management capability represented in the
Fig. 162.4: one is the management capability investment, the other is the man-
agement knowledge gained from learning effects by providing more services for
increasing customers.
1542 X. Shao and J. Shi
remaining
management
management
capabilities capabilities gap
+ -
management management
knowledge increasing capabilities
achieved via single decreaseing
management
customer + management
+ capabilities -
capabilities
+
time required
increasing + +
management to establish
knowledge management resource
Amount of
established + capability dropout rate
+ management
capability + ratio of
ratio of -
number of new resource
management investment
custormers - capability on profits
investment
costs for management on profits -
establishing unit capability ratio ofr
management investment
capability market pr
omotion
on profits
Fig. 162.4 The management capability subsystem
management
market scale
knowledge
+ +
The number of
increasing customers existing customers customer defection
remaining + +
resources +
- + +
+
market
sales + Shortage of
market promotion
expanding costs Required management
motive + management capability
+ + capability +
-
costs to gain Profits +
single new +
customer unit actual
customer-servicing management
ratio of market rate of return
promotion capability capability
on sales
cost on profits
Fig. 162.5 The value proposition subsystem
162 An Exploratory System Dynamics Model 1543
162.4.4 The Value Proposition Subsystem
BM transforms the resources into products and services by management capa-
bilities and delivers certain value proposition to customers. This value-creation
process reveals the nature of BM. The value proposition will evolve which results
in the changes in market scope and existing customers. So we adopt the number of
customers to quantify value proposition
Figure 162.5 shows that the number of customers will increase infinitely due to
the enhancing loop if there are no balancing loops. However, as the customer scale
increases, the remaining management capabilities will be reduced and consequently
the growth speed of customer scale will slow down. Meanwhile, the shortage of
management capabilities, which would bring customer defection, will be high-
lighted. Besides, the total market scale is also an external constraining factor.
162.5 Conclusion
This paper builds a SD model to explore the internal structure of complex BMES
and clarifies the meanings and functions of many vague concepts in Penrose’s
theory. We are hoping to help the practitioners gain a better understanding of the
dynamic relationships between various promoting and constraining variables and
improve the systematic thinking in the BM decision-making.
Since this paper is an exploratory study of SD evolution using SD approach,
there are several limitations which should be paid attention to in the future
research.
Firstly, although the SD model of BMES has been established, the model
simulation was not carried out due to limited time and space. Therefore, the
dynamic system behavior patterns had not been fully revealed.
Secondly, this paper focused on internal structure, which is essential to the
system behavior. Consideration on the technical, social, political, and other
external factors was inadequate which should be included in future research to gain
a more comprehensive understanding.
References
Amit R, Zott C (2001) Value creation in e-business. Strateg Manag J 22:493–520
Applegate LM (2000) Emerging e-business models: lessons from the field. Harvard Business
School, Boston
Benoît D, Xavier L (2010) Business model evolution: in search of dynamic consistency. Long
Range Plan 43:227–246
Gordijn J, Akkermans J, Van Vliet J (2001) Designing and evaluating e-business models. IEEE
Intell Syst 16(4):11–17
1544 X. Shao and J. Shi
Größler A (2010) An exploratory system dynamics model of strategic capabilities in
manufacturing (in German). J Manuf Technol Manage 21(6):651–669
Morris M, Schindehutte M, Allen J (2005) The entrepreneur’s business model: toward a unified
perspective. J Bus Res (58):726–735
Perkmann M, Spicer A (2010) What are business models? Developing a theory of performative
representation. Res Sociol Organ 29:265–275
Seelos C, Mair J (2007) Profitable business models and market creation in the context of deep
poverty: a strategic view. Acad Manage Perspect 21:49–63
Stewart DW, Zhao Q (2000) Internet marketing, business models, and public policy. J Pub Policy
Mark 19(3):287–296
Teece TDJ (2010) Business model, business strategy and innovation. Long Range Plan
(43):172–194
Timmers P (1998) Business models for electronic markets. Electron Markets 8(2):1–6
Wang Q (1995) Advanced system dynamics. Tsinghua University Press, Beijing, pp 13–20
Wang X, Wang Z (2010) Literature review on the foreign business model innovation research (in
Chinese). Foreign Econ Manage 31(12):33–37
Zhang T, Lei Y (2008) Analysis of business model clarification method based on ‘3-4-8’ structure
system. J Cent Univ Finan Econ 2:79–85
Zhang B et al (2010) Introduction to SD and SD software (in Chinese). Environ Sustain Dev
2:1–3
Chapter 163
An Empirical Study on the Influence
of Independent Directors on Fund
Performance in China Fund
Management Company
Tiao-yan Hui and Fei-lan Lu
Abstract This paper makes an empirical study on the influence of independent
director on fund performance, based on the data of the fund management company
in China and its equity open-end Fund and the mixed open-end fund from 2005 to
2010. According to the descriptive statistics and linear regression analysis using
SPSS13.0 software, the results show that: To the China’s fund management
company, the independent directors’ proportion is low, and they are mainly master
and doctor degree, economic management major and from the university research
institutes; The higher of the proportion of the independent director, with senior
professional titles, and form industrial and commercial enterprises, the higher the
performance of the fund is; The education and professional of independent
directors have no significant impact on fund performance.
Keywords The fund management company Independent director Fund per-
formance Empirical study
163.1 Introduction
The independent director is directors, who not be employed by the company or its
affiliates and not be closely related to the company or its management through
significant economic, family or other ties. In 1940 \ Investment Company
Act [ was putted forward by United States, the proportion of independent
T. Hui (&) F. Lu
Department of Economics and Management, University of Xidian,
Xi’an, China
e-mail: [email protected]
F. Lu
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1545
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_163,
Ó Springer-Verlag Berlin Heidelberg 2013
1546 T. Hui and F. Lu
directors in the Investment Company should not be less than 40 %. For the
important role of the independent directors, it has received widely attention, and a
lot of scholars have done researches on it about the number of independent
directors, background characteristics, and so on. But, due to the research samples,
research time, research method, we still do not get consistent conclusions.
Securities investment fund has become an invest manner and is developed
largely in recent years. By the end of 2011, there are 914 securities investment
funds in China, the scale is 2646.465 billion which has grown up 9.17 % over the
last year. At the same time, various kinds of matters, such as Interests transmis-
sion, Tunnel Digging, Mouse storehouse take place usually. So how to improve the
governance structure of the fund management company to protect the interest of
investors is becoming emergent. The independent directors as an important gov-
ernance mechanism in fund management companies are endowed with great
expectations. According to the analysis of independent director’s structure and its
influence on fund performance, this paper hopes to promote the development of
the fund management company healthily and rapidly in China.
163.2 Literature Review
Based on the principal-agent theory, resources dependence theory, and so on,
domestic and foreign scholars make numerous studies on independent directors,
but mainly focus on the listed company and the research for the fund management
company is relatively lack. In 2001 year, the CSRC requested the fund manage-
ment company to have independent directors and further requested there should be
at least three or more independent directors which are more than the number of
directors appointed by the company’s largest shareholder, and the proportion of
dependent directors should not be lower than one-third (Wei 2005).
The representative literature on the relationship between the structure of
independent director and company performance is shown in the Table 163.1,
especially for the fund company as shown in Table 163.2.
In all, the independent directors play a positive role in company performance,
but as a result of the diversity of each research samples, research time, research
methods, the research scope, the results are very different.
163.3 Hypothesis
(1) The relationship between the proportion of independent directors and fund
performance: Director Independence is the foundation of decision-making and
supervision for the board of directors. For the fund management company, usually,
compared to association board of directors, driven by social reputation and long-
term material interests of themselves, independent directors can supervise
Table 163.1 The list of research literature about general company
Author Dependent Data sources Research conclusions
variables
Zhang et al. ROA 14 banks listed on shanghai and shenzhen stock The independent director has significantly positive effect on
(2011) 2006–2009 performance to bank.
Fan and Li ROE 109 listed companies in Guangxi 2002–2006 ROE is negative correlation with the proportion of the independent
(2009) directors, but not significant
Meng ROA, 3740A-shares listed companies The proportion of independent directors is weak related to
(2010) Tobin’s company governance performance
Q
Liu et al. ROA 141 civilian-run listed companies 2004–2007 The proportion of independent directors is not an important factor
(2009) to affect the company performance.
Li et al. Tobin’s Q 863 listed companies in manufacturing industry in cross- The independent board system has significant positive effect
(2009) section from Japan 2001–2006 toward company performance.
163 An Empirical Study on the Influence of Independent Directors
1547
Table 163.2 The research literature about the fund management company
1548
Author Dependent variables Data sources Research conclusions
Stephen and Yan Expense ratios 6228 funds in 448 fund families in The percentage of independent directors is negatively
(2007) the CRSP database at the end of related to expense ratios.
2002
Xing and Song Sharp ratio of the investment target after 25 fund companies in Shanghai The proportion of independent directors is positive
(2008) the Adjustment 2005–2009 related to company performance.
Zhou (2008) Annual net yield 53 closed-end funds in all the listed The proportion of the independent directors with
fund companies in 2006, which experience of finance, legal, financial work 5 years
set up before December 31, 2005 or more is a significant positive to fund
performance.
He (2005) Annual fund net income, Jenson index, 54 closed-end funds of 17 listed fund The proportion of the professional independent
Treynor index, Sharpe index, companies before December 31, directors is larger, the fund performance is better
Tobin’s Q, expense ratio 2002 and the expense ratio is lower
Li (2010) Fund return, fund’s annual average rate The fund companies had managed a The proportion of independent directors and with the
of return more than fund industry full year by the end of 2009 professional knowledge have a significant positive
correlation to company performance
Xiao and Peng Annual net ratio of growth,fund’s 131 fund companies raised stock The proportion of independent directors has no
(2010) average annual rate of return more funds and mixed funds to the significant influence on fund performance.
than fund industry’s median, Jensen public before January 1, 2005
index
Wu (2006) The year average revenue rate 47 stock funds in 2004 The proportion of independent directors has a
significant positive correlation to fund performance.
T. Hui and F. Lu
163 An Empirical Study on the Influence of Independent Directors 1549
managers effectually, make independent and impartial judgments, protect the
interests of the fund holder and promote the fund performance. So, we put forward
assumption 1:
H1: The proportion of independent directors is positive related to fund
performance.
(2) The relationship between the education of independent directors and fund
performance: \ Administration of Securities Investment Fund Management
Companies [ issued by the CSRC on September 27 in 2004 required that: the
board in the fund management company must have at least three independent
directors who have experience of finance, law or financial more than 5 years, and
have enough time and energy to perform their duties. This law is proposed new
requirements to maintain the independence and effectiveness of the board of
directors in the fund management company. The educational background is the
symbol of human capital and it can embody people’s ability and qualify to the
work, so the higher the proportion of independent directors with high degree, the
more it can improve the efficiency of management. So, we come up with
assumption 2:
H2: The proportion of independent directors with high degree is positive related
to fund performance.
(3) The relationship between the major of independent directors and fund
performance: The major not only decides the skills of the individual have, but also
influences their mode of thinking. Majoring in economic and management will
provide a systematical study of economic, management and financial knowledge,
which will help the independent directors to improve their ability of insight of
macro-economic environment and analysis of enterprise management decision.
Even though science and technology majors are stronger in logic, they lack of
knowledge in the enterprise management, to a certain extent, which hinders their
functions of the management decision, supervision and control. So, we put forward
assumption 3:
H3: The proportion of independent directors in economic management major is
positive related to fund performance.
(4) The relationship between the professional title of independent directors and
fund performance: The professional title represents a level of person in academic
or job, symbolizing a certain identity, affirming the ability of independent direc-
tors. Therefore, we proposed hypothesis 4:
H4: The professional title of independent directors is positive related to fund
performance.
(5) The relationship between the sources of independent directors and fund
performance: Based on the theory of check-and-balance ownership structure, the
diversity of the board source is beneficial to balance the internal power and
improve the performance of company. Fu (2008) have done empirical research on
336 companies about manufacturing in the Shanghai stock exchange, found dif-
ferences in the source of directors can lead to diversity existing in the governance
performance of listed companies (Fu 2008).This paper is divided the sources of
independent into: schools, research institutions, accounting firm, law firms,
1550 T. Hui and F. Lu
financial institutions and industrial-commercial enterprises. If the independent
directors do not have some certain experiences, their own opinion cannot influence
the board of directors to make decision, which will become ‘‘vase directors’’ or ‘‘
favor directors ’’, and it will not do good for company management and perfor-
mance. Therefore, the rich experience of independent directors has very great help
to improve the company performance. So, we put forward hypothesis 5:
H5: The independent directors come from some industries have more practi-
cality, such as accounting firm, law firms, financial institutions, which are
positively related to fund performance.
163.4 Empirical Research
163.4.1 Sample and Data
The data is based on the fund management company in China and its equity open-
end Fund and the mixed open-end fund from 2005 to 2010, eliminating the LOF
and QDII. In order to ensure the integrity of fund dates, the fund must be estab-
lished before the year of researched.
163.4.2 Definition of Variables
Research variables in this paper are composed of: dependent variables, indepen-
dent variables, and control variables. Their descriptions are shown in Table 163.3.
163.4.3 The Descriptive Statistical Analysis
The Table 163.4 indicates that, in 2005–2010 the size of independent directors is
changed little, the maximum and minimum are 6 and 2, of which more than 90 %
independent directors in the fund management company are three or four, reached
the requirement of the CSRC that the fund management company should have at
least three independent directors.
The Fig. 163.1 shows, the degree of independent directors in the fund man-
agement company is mainly doctor, the proportion of it has increasing tendency,
especially in 2010 it reaches at 44.0158 %. Master degree also goes up, but lower
than that of the doctor degree. Only the proportion of bachelor degree is reduced
year by year, from the highest 33.2258 % in 2005 to the lowest 23.9444 % in
2010, almost decrease by 10 % in six years. This shows the fund management
163 An Empirical Study on the Influence of Independent Directors 1551
Table 163.3 Definition of variable
Type Variables Introductions
Dependent Fund return An annual average rate of return of the fund more
variables than the fund industry
Fund risk The standard deviation of the return of fund
Risk-adjusted fund Adjusted sharp index, it is the ratio of fund return
performance to fund risk
Independent The proportion of The ratio of independent directors in the board
variables independent directors (PID)
The educational The proportion of bachelor (BAC)
background The proportion of master (MAS)
The proportion of doctor (DOC)
The major The proportion of economic management major
(ECO)
The proportion of law major (LAW)
The proportion of science and technology major
(SCI & TEC)
The professional title The proportion of senior title (SEN)
The sources The proportion of school and research institutions
(SRI)
The proportion of accounting firms (AF)
The proportion of law firms (LF)
The proportion of financial institutions (FI)
The proportion of industrial–commercial
enterprises (ICE)
Control Fund scale An annual average fund share (FS)
variables Fund time The time of fund established (FT)
Bull market The year of 2006, 2007, 2009, it is 1; others it is 0.
(BULL)
Bear market The year of 2008, it is 1; others it is 0. (BEAR)
Table 163.4 Descriptive statistics of size of independent directors
Years N Min Max Mean Std.
2005 35 2.00 6.00 3.5429 0.81684
2006 47 3.00 5.00 3.5319 0.62035
2007 52 3.00 6.00 3.4615 0.64051
2008 55 3.00 5.00 3.3636 0.52223
2009 58 3.00 5.00 3.3621 0.55245
2010 58 3.00 5.00 3.3793 0.58722
companies have more and more strict demand with independent director, the
pursuit of highly educated is one of the means to be an independent director.
From the Fig. 163.2, we can see that, the majority of independent directors in
the fund management company are mainly economic management and the annual
average proportion reaches up to 62.552 % in 2005–2010. The proportions of law
1552 T. Hui and F. Lu
Fig. 163.1 The degree of N
independent directors 100
80
60
40
20
0
2005 2006 2007 2008 2009 2010 Y
BAC MAS DOC
Fig. 163.2 The major of N
independent directors 150
100
50
0
2005 2006 2007 2008 2009 2010 Y
ECO LAW SCI&TEC
Fig. 163.3 The sources of N
120
independent directors
100
80
60
40
20
0
2005 2006 2007 2008 2009 2010 Y
CRI AF LF FI ICE
major are decreasing. In recent years, the irrational behaviors of some fund
management companies in the pursuit of high return, the cases lead the share-
holders to suffer loss can be found everywhere. This requires the fund management
company to add more independent directors in science and technology major, and
make its more reasonable judgments. Accordingly, the proportion of independent
directors in science and technology major has a rising trend.
Figure 163.3 shows, independent directors from the college and research
institutions make up more than half of the proportion in the fund management
company, the following is financial institutions, and its proportion rises gradually
in 2005–2010. And the others from accounting firms, law firms and industry and
commerce enterprises accounts for only about 20 % of all the independent
directors. Half of independent directors are from school and research institutions
with less company operating experiences; it will lead to some problems on the
function of independent directors in the fund management company.
163 An Empirical Study on the Influence of Independent Directors 1553
163.4.4 Regression Analysis
Regression results are show in Table 163.5, among the variables of structure
characteristics of independent director in fund management company, the pro-
portion of the independent directors has a negative impact on the fund return, but
not significantly, and at the 0.05 level, it has a significant negative correlation with
the fund risk, also it is positively related to risk-adjusted fund performance, which
is consistent with the existing literature. Now, the introduction of independent
director plays a positive role to control fund risk under the condition of less
effectively supervise and protection of the fund holder’s interests, which can
alleviate principal-agent problems between the fund holder and the fund man-
agement company. Meanwhile, relating to the internal director, the independent
director have little knowledge on the company management and it is weak in
supporting function of management decision, so it has a negative impact on the
fund return, but the positive influence of its risk control exceeds the negative
influence, thus it has positive influence to the risk-adjusted fund performance.
The education and major of independent director are not significantly influ-
enced on fund performance, whether return, risk, or risk-adjusted fund perfor-
mance. For the title of independent directors, the proportion of independent
director who has a senior professional title has a negative influence with fund risk,
which generally has a high status in society and practice experience, so they can
control risk better. Besides, considering their own reputation, the independent
director has the motivation to work harder and qualified for the role as an enter-
prise management supervisor.
The proportion of independent directors from industry and commerce enter-
prises is significantly positive correlation to fund return, and negative correlation
to fund risk. This is mainly because of the independent director come from
industry and commerce enterprise have no relationship with fund management
companies, they can better ensure the objectivity of their supervision, and long-
term practices enrich their working experience, helping fund management com-
panies make better investment decisions from the point of view outside the
financial enterprise, so they have some positive influence on the fund performance.
In addition, recent years the fund management companies in China do not have
learning effect, the fund established time has negative influence on both the fund
return and risk. It is maybe the longer time the fund operates, the more conser-
vative its operation is, so the fund’s incomes and risk are lower. Especially for the
risk control, a fund which through a long-term experience, management method
and system are relatively mature, so it has a significant negative effect on fund
risks at the 0.05 level, but is not significant to the risk-adjusted fund performance.
For the fund scale, it has a certain scale effect, the greater the scale, the higher the
fund return, but more difficulty to management, the lower is the fund’s flexibility,
that are more apparent in bear market. So, fund size is positively related to fund
risk, not significant to the risk-adjusted fund performance.
1554
Table 163.5 The result of regression analysis
Model Fund return Fund risk Adjusted performance
Coeff t Sig. Coeff t Sig. Coeff t Sig.
(C) 0.651 0.515 15.326 0.000 –0.096 0.924
PID –0.029 –0.847 0.397 –0.110 –4.237 0.000 0.035 1.903 0.057
MAS –0.063 –1.289 0.198 0.006 0.162 0.871 –0.017 –0.658 0.511
DOC –0.046 –0.880 0.379 0.015 0.381 0.703 0.006 0.210 0.834
ECO 0.012 0.243 0.808 –0.023 –0.632 0.527 –0.004 –0.150 0.881
LAW 0.057 1.173 0.241 –0.016 –0.444 0.657 0.034 1.294 0.196
SEN –0.033 –0.886 0.376 –0.079 –2.811 0.005 –0.015 –0.733 0.464
AF 0.009 0.276 0.782 0.011 0.429 0.668 –0.009 –0.474 0.635
LF –0.022 –0.569 0.569 0.014 0.493 0.622 –0.028 –1.338 0.181
FI 0.039 1.015 0.310 –0.009 –0.309 0.757 0.006 0.271 0.787
ICE 0.079 2.286 0.022 –0.072 –2.791 0.005 –0.008 –0.436 0.663
FS –0.028 –0.882 0.378 –0.088 –3.645 0.000 –0.003 –0.160 0.873
FT 0.065 1.985 0.047 0.158 6.362 0.000 0.009 0.512 0.609
BULL 0.062 1.734 0.083 0.379 14.099 0.000 0.565 29.449 0.000
BEAR –0.019 –0.540 0.589 0.670 24.670 0.000 –0.414 –21.386 0.000
Modle-fitting degree F = 1.748 F = 58.873 F = 186.686
Sig = 0.042 Sig = 0.000 Sig = 0.000
Coeff is Standardized Coefficients
T. Hui and F. Lu
163 An Empirical Study on the Influence of Independent Directors 1555
Finally, fund performance is influenced by market quotation significant in
China. To bull market, fund return, risk and risk-adjusted fund performance are
higher and its influence is significant at the 0.1 level. To bull market, fund return is
low, but the influence is not significant, and because investment is more difficult,
fund management companies often due to the loss of some stock, they can’t sell
them in time, hoping to earn profits with certain bets psychological, thus signifi-
cantly high risk is presented, and at the 0.05 level it is significantly negative to the
risk-adjusted fund performance.
163.5 Suggestions
Firstly, the proportion of independent directors should be increased. The empirical
results show that the proportion of the independent directors has positive influence
on fund performance, especially the risk control and the risk-adjusted fund per-
formance. At present the average number of independent directors in the fund
management company is 3.37, just meeting the requirement of CSRC to inde-
pendent directors in the fund management company not less than 3. In 2010, the
scale of independent director are 3, 4 and 5, corresponding the number of the fund
management company are 39, 16 and 3, and the proportion of the fund manage-
ment company are 67.24, 27.59 and 5.17 %, but the average proportion of inde-
pendent director just 3.3793, and is far less than the requirement of two-thirds in
the United States investment management company. Therefore, in the future the
fund management company in China should further increase the proportion of the
independent directors, perfect the mechanism of the hiring, and do the best to
ensure its independence.
Secondly, the professional titles of independent director should be strengthened.
The education background and major should not become the only standard to
appoint independent directors in the fund management company. Independent
director in the board of directors has the main function of supervision, exercise this
responsibilities needs more rich work experience. Education background and
major have little effect on this, the professional title is an important signal of
ability, so in the choice of independent directors should pay more attention to the
professional title in the fund management company.
Finally, the source of independent directors should be optimized (Ma 2010).
Now the practice of independent directors in the fund management company are
mostly academic directors and celebrity directors, these independent directors
often have more famous, much positions, maxed affairs, but have little time and
energy to handle corporate affairs, so they could not have effective function of
supervision. The empirical results in this paper show that the independent director
from industrial-commercial enterprises has effectively improved the company
performance. So, in the choice of the independent directors should be required
1556 T. Hui and F. Lu
some enterprise or business experience, familiar with the laws and regulations, and
possess the knowledge of capital market operation theory, it will makes them
better to perform the duties of independent directors.
Acknowledgments Supported by the National Science Fund of China (NO: 70802047).
Reference
Fan L, Li X (2009) A positivist study of the formation of board of directors in relation to business
performance— evidence from guangxi listed firms’ empirical data in nearly five years.
J Guangxi Univ Philos Soc Sci 31(3):34–37 (in Chinese)
Fu G (2008) Study on corporate governance performance and directors’ source, the degree of
economic correlation. Mark Modernization 37(27):355–358 (in Chinese)
He J (2005) Independent director, governance structure & performance of contractual-type fund
in China. Nankai Bus Rev 8(1):41–48 (in Chinese)
Li B, Zhang J, Zheng C (2009) The panel model empirical analysis of the effect of member age of
board of directors and independence on corporate performance: examples from Tokyo main
board listed companies in Japanese Manufacturing Industry. (2009LDQN14)
Liu J, Zheng L, Wan G (2009) The relationship between board structure and corporate
performance—evidence from private listed firms in China. Reform Econ Syst 27:68–73 (in
Chinese)
Li N (2010) Relevant research on China fund management company internal governance and
corporate performance. Yangzhou University, Jiangsu, pp 30–39 (in Chinese)
Ma Z (2010) China’s fund management company’s independent director system in research and
analysis. J Northeast Norm Univ Philos Soc Sci 60:43–46 (in Chinese)
Meng X (2010) Relevant research on the board of directors features and corporate governance
performance. Chongqing University, Chongqing, pp 25–37 (in Chinese)
Stephen PF, Xuemin (Sterling) Y (2007) Do independent directors and chairmen matter? the role
of boards of directors in mutual fund governance. J Corp Financ 13:392–420
Xing Q, Song F (2008) An empirical research of fund performance and fund governance. 12th
annual meeting of the academic management science in China, Editorial Department of China
Management Science, Beijing (in Chinese)
Xiao J, Peng W (2010) A research into inside governance of fund management corporation and its
effect: taking open-ended fund as a sample. Audit & Econ Res 25(1):105–112 (in Chinese)
Wei Z (2005) On independent director system of fund managing companies. Securities Mark
Herald 15:17–21 (in Chinese)
Wu X (2006) An empirical study and apocalypse on the validity of board of directors of american
corporate funds. Soochow University, Suzhou, pp 25–31 (in Chinese)
Zhou T (2008) An empirical study on the relationship between characteristics of board of
directors and fund performance in fund management company. Southwest Jiaotong
University, Chengdu, pp 43–62 (in Chinese)
Zhang N, Guan Z, Guo Z (2011) An empirical study on the relationship between director board’s
characteristics and bank performance—evidence from China’s 14 listed banks. Econ Surv
28:25–28 (in Chinese)
Chapter 164
Research on Optimal Enterprise
Contribution of Hunan Province Based
on OLG Model
Ni Yang
Abstract Optimizing the enterprise contribution is the key factor for promoting
the reforming of public pension system and insuring the dynamic balance of social
security fund. This paper has made a research on optimal enterprise contribution of
Hunan province based on OLG model. The empirical result showed that life
expectancy growth would raise the optimal enterprise contribution, while popu-
lation growth rate decline would reduce the contribution, and the latter factor made
more influence. If both two factors were introduced in the equilibrium equation,
the optimal enterprise contribution would be reduced from 20 to 10.04 %, when
life expectancy growth raised from 73.8 to 77.2 and population growth rate
declined. The research on optimal enterprise contribution provides theory basis
and the policy support for macroeconomic policy making and pension reforming
promoting.
Keywords OLG model Optimal enterprise contribution Pension reforming
Life expectancy growth Population growth rate
164.1 Introduction
Optimizing the enterprise contribution is an important component of public pen-
sion reforming, which makes sense for economic development and social progress,
and has effect on national saving, enterprises’ operating costs, families’ con-
sumption structure, labor’ supply and so on. With the development of population
structure changing and population ageing turning seriously, how to get the optimal
enterprise contribution has attract more and more attention.
N. Yang (&)
College of Economics and Management, Hunan Normal University, Changsha, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1557
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_164,
Ó Springer-Verlag Berlin Heidelberg 2013
1558 N. Yang
Hunan province has a large population, as the age population (65 years old and
above) has reached 6.35 million by the end of 2009 accounting for 9.22 % of total
population. Furthermore, the population aging of Hunan province has character-
istics of speedy growth, large scale, ‘‘growing old before growing rich’’ and so on,
which makes significant impact on dependency ratio, consumption structure and
social security. Therefore, making proper mechanism for optimal enterprise con-
tribution is an important premise for promoting the stability and development of
social security in long run.
As a frequently used tool for public pension analyzing, overlapping generations
model (OLG) could examine the influence of public pension on the whole macro
economy through analyzing micro economic agents, based on the general equi-
librium frame. The theory was advanced by Samuelson, and expanded to classic
cross time dynamic model by Diamond (1965). Many scholars discussed the
relation between social security mechanism and economic growth based the model
from different aspect. For example, Barro (1974) and Romer (1986) discussed the
influence of pay as you go pension on economic growth from bequests motivation
and personal savings respectively. Casamatta (2000) constructed a two period
OLG model to analysis the reallocation function of social security. Fanti and Gori
(2007) analyzed the effects of the regulation of wages in a standard one sector
OLG model of neoclassical growth extended to account for endogenous fertility
decisions of households and unemployment benefit policies financed at balanced
budget. Rausch and Rutherford (2010) developed a decomposition algorithm by
which a market economy with many households may be solved through the
computation of equilibria in OLG models. Moreover, many scholars explode the
influence of population changing on economic development of OECD countries,
such as Auerbach and Kotlikoff (1989), Neusser (1993), Hviding and Mérette
(1998), Fougère and Mérette (1998), etc.
Public pension reforming has been launched in 1997 in China. In recent years,
OLG model has been used for studying the transition cost, invisible debt, dynamic
effectiveness during the pension reforming. For example, Bo (2000) explored the
influence of different institutional arrangement on economic growth and Pareto
efficiency. Yuan and Song (2000) made a conclusion that the personal saving rates
in China was efficient while the macro savings rate was ineffective through con-
structing OLG model and simulation. Yuan and He (2003) discussed the possi-
bility of dynamic inefficiency in free competition economic with the help of OLG
model. Yang (2008) analyzed endowment insurance for enterprise employees
based on OLG model, to get the optimum pension replacement rate in a general
equilibrium frame. Wang et al. (2010) analyzed the economic effects of the
occupational pension system from macroeconomic capital and output, microeco-
nomic producers and microeconomic consumers based on OLG model. Li and Bai
(2006) constructed an OLG model by GAMS software, whose input variable was
population age changing, to illustrate the changing of social output, personal
consumption and government revenue arise from population ageing in China.
Huang and Sun (2005) analyzed the difference in informal institutions and con-
sumption modes in OLG model, and made a theoretical analysis on households’
164 Research on Optimal Enterprise Contribution of Hunan Province 1559
consumption in the oriental culture and belief. Liu and Hao (2011) constructed a
discrete time bilateral generation transfer model, and discussed the optimal
investment structure and economic growth pattern. Lu (2011) constructed a three
period OLG model, to explain the influence of population structure and income
growth on personal savings in China.
This paper makes a research on the optimal enterprise contribution in partially
funded endowment insurance of Hunan province based on OLG equilibrium
framework. Combining the characteristics of population structure and economic
growth in Hunan Province, parameter selection and model improvement has been
analyzed. The rest has been arrangement as follows. Section 164.2 introduced the
basic framework of OLG model and made a deriving of optimal enterprise con-
tribution. Section 164.3 made an empirical research on the selection of optimal
enterprise contribution in urban old-age insurance of Hunan province. And the
Sect. 164.4 was the conclusion and future research suggestions.
164.2 Basic Framework of OLG Model and Deriving
of Optimal Enterprise Contribution
This paper conducts research based on equilibrium framework raised by Yang
(2010). Suppose that there are many persons, many enterprises and one govern-
ment in an infinite closed economy. Personal life is finite, and experiences working
period and retirement period. At the beginning of t, there are Nt persons become
workingmen in the tth generation. The rate of population growth can be calculated
as
Nt
n¼ 1 ð164:1Þ
Nt1
The partially funded endowment insurance was introduced.
164.2.1 Basic Framework of OLG Model
Supposing everyone provide a unit nonelasticly labor while working, obtains
income equals wt, contribution rate for endowment insurance is s, consumes c1t
and saves st. Furthermore, the individual consumes c2t+1 after retired, including the
principal and interest of savings while working (1 ? rt+1)st, the payments of
funded personal account is (1 ? rt+1)gt, and the public pension is bt+1. The person
selected the pattern of savings and two periods consuming for pursuing utility
maximizing.
max Ut ¼ uðc1t Þ þ huðc2tþ1 Þ ð164:2Þ
fst ;c1t ;c2tþ1 g
1560 N. Yang
s:t: c1t ¼ ð1 sÞwt st ; c2tþ1 ¼ ð1 þ rtþ1 Þst þ ð1 þ rtþ1 Þgt þ btþ1 ð164:3Þ
Parameter h [ (0,1) presents discount factor. The utility function u(.) is a
monotonous increasing function of consumption, and a strict concave function,
satisfying u0 (.) [ 0, u00 (.) \ 0. Enterprise produces homogeneous goods in com-
petitive market, satisfying first order homogeneous production function yt = f(kt),
in which function kt presents labor capital ratio. Enterprise contribution rate for
endowment insurance is g [ (0,1).
According to Euler theorem, interest rate equals to the marginal capital output,
while (1 ? g)wt equals to the marginal labor output, satisfying:
f ðk t Þ k t f 0 ðk t Þ
rt ¼ r 0 ðkt Þ wt ¼ ð164:4Þ
1þg
bt Nt 1 ¼ gwt Nt ; gt ¼ swt ð164:5Þ
The government takes enterprises’ contribution as overall planning account,
paying as public pension for current retirees; while personal’s contribution as
accumulated personal account, which would be drawn after retirement. Further-
more, the capital market would satisfy the following equation. That’s the savings
plus personal account of the tth labor equals the capital stock at the beginning of
t ? 1 period.
st þ gt ¼ ð1 þ nÞktþ1 ð164:6Þ
164.2.2 Dynamic Equilibrium and Its Stability Conditions
Supposing the dynamic system exists a stationary equilibrium which is single,
stable and no oscillation.
A dynamic equilibrium equation could be obtained as follow.
f ðkt Þ kf 0 ðkt Þ
u0 ð1 þ nÞktþ1 þ h½1 þ f 0 ðktþ1 Þ
1þg
g
u0 ð1 þ nÞktþ1 ½1 þ f 0 ðktþ1 Þ þ ð1 þ nÞ ½f ðktþ1 Þ ktþ1 f 0 ðktþ1 Þ ¼ 0
1þg
ð164:7Þ
164 Research on Optimal Enterprise Contribution of Hunan Province 1561
164.2.3 Equilibrium Equation for Optimal
Enterprise Contribution
In order to make the state of market economy reach the optimal state, policy
parameters should be adjusted, to obtain optimal capital labor ratio.
From functions described above, we could obtain the equilibrium equation for
optimal enterprise contribution.
h i
p ð1aÞ 1þn
q a ð 1 þ n q Þ þ p ð 1 p Þ q ð 1 þ p Þ ð 1 þ n Þ
g¼ h i ð164:8Þ
1 ð1aÞ 1þn
p a ð 1 þ n qÞ p ð 1 p Þ q ð 1 þ p Þ ð 1 þ n Þ
According to the equation above, we could know that the optimal enterprise
contribution would be influenced by survival probability of retirement P, social
discount factor q, capital income share a and population growth rate n.
164.3 An Empirical Research on Optimal Enterprise
Contribution of Hunan Province
This section discussed the optimal enterprise contribution of Hunan province
through empirical research based on OLG equilibrium framework.
There are three steps for empirical. Firstly, parameters would be set according
the population structure and economic development of Hunan province, and then
social discount factor would be estimated. Secondly, the value of parameters
would be brought in the equilibrium function. And the optimal enterprise contri-
bution would be obtained with the supposing of population growth remaining
unchanged. Lastly two variables, life expectancy growth and population growth
rate decline would be introduced in the equilibrium framework, and the combined
effect would be examined.
164.3.1 Set Parameters and Estimate Social Discount Factor
Parameter capital income share a is usually equals 0.3 for developed countries.
While income of labor is lower in China, and Hunan province has a large popu-
lation density. Therefore, parameter a is 0.36.
Urban population of Hunan province is adopted as population statistics caliber.
According to the data announced by the Department of Economic and Social
Affairs of United Nations, the life expectancy growth would be increased to
80.3 years old in 2055–2060. Therefore, the length of one period is set 27 years,
which satisfying the condition ‘‘three periods time span should be equal or greater
1562 N. Yang
than life expectancy growth to guarantee survival probability P of retirement less
than or equal to 100 .
According to Hunan province statistics yearbook 2007, the urban population
growth rate could be calculated n = (2619.93/639.6) - 1 = 3.092 during
1979–2006. The life expectancy of Hunan province was announced to 73.8 years
old by Hunan province bureau, so survival probability of retirement could be
calculated as p = 73.33 %.
Furthermore, according to relevant policies, the town enterprise contribution
g = 20 %.
Parameter social discount factor could reflect social planners’ preferences.
Bringing all value of parameters in (164.8), we obtain that q = 0.5458.
164.3.2 Estimation of Optimal Enterprise Contribution
with Population Growth Rate Fixed
With the promotion of living quality and improvement of medical conditions, the
life expectancy is increasing. Because of the limitation of data available, we
assume that the life expectancy in Hunan province is the same as the whole
country.
According to the data announced by Department of Economic and Social
Affairs of United Nations, we could obtain the life expectancy of Chinese in future
30 years (the data could be seen in Table 164.1).
Bringing the values of parameters set above in (164.8), including a = 0.38,
n = 3.0962, q = 0.5458, the optimal enterprise contribution could be obtained
under different life expectancy.
According to Table 164.1, the optimal enterprise contribution increases along
with the life expectancy ascend.
And in the next 25 years, the life expectancy in China would be increased from
73.8 to 77.2 years old. If population growth rate fixed, the optimal enterprise
contribution would be promoted from 20 to 22.97 %.
Table 164.1 Estimation of optimal enterprise contribution (fixed population growth rate)
Period Life expectancy P g
2015–2020 74.7 0.7667 0.2104
2020–2025 75.6 0.8 0.2189
2025–2030 76.4 0.8296 0.2251
2030–2035 77.2 0.8592 0.2297
164 Research on Optimal Enterprise Contribution of Hunan Province 1563
Table 164 2 Estimation of optimal enterprise contribution (two variables changing)
Period Life expectancy n g
2015–2020 74.7 2.3133 0.1916
2020–2025 75.6 1.6405 0.1714
2025–2030 76.4 1.1032 0.1363
2030–2035 77.2 0.7928 0.1004
164.3.3 Estimation of Optimal Enterprise Contribution
with Population Growth Rate Decline
China has executed very strict one-child policy since 1980s, so human fertility
declined rapidly, which is the main reason for population ageing. Low fertility,
low mortality, low population growth rate, has been the main characteristic of the
population of China. The changing of population growth rate would influence
population age composition, social support rate, and then the optimal enterprise
contribution.
Table 164.2 shows the result of the optimal enterprise contribution when both
of the two variables, life expectancy growth and population growth rate decline
have been brought in the equilibrium framework. According to Table 164.2, the
optimal enterprise contribution declined while both of the two variables, life
expectancy growth and population growth rate decline have been brought in the
equilibrium framework.
In the next 25 years, the life expectancy in China would be increased from 73.8
to 77.2 years old and population growth rate would decline continuously, and then
the optimal enterprise contribution would be declined from 20 to 10.04 %. Now
that life expectancy growth would increase the optimal enterprise contribution and
the changing of two variables make the optimal enterprise contribution decline,
therefore we could obtain the conclusion that the population growth rate plays a
more important role than life expectancy growth in the optimal enterprise con-
tribution determine because of the large population bass.
164.4 Conclusion and Future Research Prospect
Accompanying with population structure changing and series problem of ageing
population, reforming public pension, pushing forward harmonious development
of social security, has becoming the focus of researches. The optimal enterprise
contribution is one of the most important indexes, plays an important role for
dynamic balance for social security fund. OLG model could bring the actions of
personal, enterprises and the government into an equilibrium framework, and it is a
useful tool for analyzing the influence of public pension reforming on macro
economy.
1564 N. Yang
This paper has discussed the optimal enterprise contribution of Hunan province
based on the OLG model. Firstly, we have introduced the basic framework of OLG
model and the deriving of optimal enterprise contribution; and then we have made
an empirical research on the optimal enterprise contribution determination of
Hunan province. Empirical results show that life expectancy growth would
increase the optimal enterprise contribution while population growth rate decline
would lower the optimal value. And the latter variable of population growth rate
plays a more important role because of the large population base in China.
However, this paper discussed the optimal enterprise contribution adjustment
caused by parameters changing based static equilibrium equations. How to con-
structing a general dynamic equilibrium system to describe the real state of pen-
sion operating and the dynamic changing of parameters based social optimization,
would be one of the future research.
References
Auerbach AJ, Kotlikoff LJ (1989) The economic dynamics of an ageing population: the case of
four OECD Countries. OECD Econ Rev 12(1):97–130
Barro RJ (1974) The impact of social security of private saving. American Enterprise Inst,
Washington DC, pp 21–35
Bo J (2000) The influence of endowment insurance system arrangement on economic growth and
Pareto efficiency. Econ Sci 27(1):78–88 (in Chinese)
Casamatta G (2000) The political economy of social security. Scand J Econo 102(3):503–522
Diamond PA (1965) National debt in a neoclassical growth model. Am Econ Rev
55(1):1126–1150
Fanti L, Gori L (2007) Fertility, income and welfare in an OLG model with regulated wages. Int
Rev Econ 54(2):405–427
Fougère M, Mérette M (1998) Population ageing and current account in selected OECD
countries. Working Papers-Department of Finance Canada, vol 4, no 1, pp 1–24
Huang S, Sun T (2005) Informal institutions, consumption modes and assumption of OLG
model—a theoretical analysis on households’ consumption in the oriental culture and belief
(in Chinese). Econ Res J 24(4):57–65
Hviding K, Mérette M (1998) Population effects of pension reform in the context of ageing. OLG
simulations for seven OECD countries, OECD Working Paper, pp 1–23
Li H, Bai X (2006) Life-cycle model and its application to research in aging China. Chin J
Population Sci 28(4):28–35 (in Chinese)
Liu Q, Hao S (2011) Theoretical analysis on uncertainty of aging issue in gift economy based on
OLG model. Stat Res 28(10):84–90 (in Chinese)
Lu D (2011) Population structure, economic growth and China’s household saving: empirical
research based on OLG model and panel data. Shanghai Finance 32(1):10–15 (in Chinese)
Neusser K (1993) Savings, social security, and bequests in an OLG model: a simulation exercise
for Austria. J Econ 7(1):133–155
Rausch S, Rutherford TF (2010) Computation of equilibria in OLG models with many
heterogeneous households. J Econ 36(2):171–189
Romer PM (1986) Increasing returns and long run growth. J Polit Econ 94(2):1002–1037
Wang X, Zhai Y, Yan H (2010) Economic effects of the occupational pension system: the
research based on general equilibrium model. Nankai Econ Stud 12(5):46–55 (in Chinese)
164 Research on Optimal Enterprise Contribution of Hunan Province 1565
Yang Z (2008) The public pension for enterprise employees, benefit replacement rate and
population growth rate. Stat Res 25(5):38–42 (in Chinese)
Yang Z (2010) OLG model analysis on public pension: principles and applications. Guangming
Daily Press, Beijing, pp 27–45 (in Chinese)
Yuan Z, He Z (2003) Dynamic inefficiency in China’s economy since 1990s. Econ Res J
24(7):18–27 (in Chinese)
Yuan Z, Song Z (2000) The age composition of population, the endowment insurance system and
optimal savings ratio in China. Econ Res J 11(1):24–32 (in Chinese)
Chapter 165
Comprehensive Experiment Design
of Production Logistics Based on CDIO
Ying-de Li and Xiu-ju Lan
Abstract Production logistics engineer in manufacturing plant is one of the
important potential employments for logistics engineering students. The training
purpose of our school is to provide some professionals who are familiar with the
production logistics and from the logistics engineering graduates. Based on the
CDIO concept, a comprehensive experiment is designed, which include the pro-
duction forecasts and orders issued, facilities planning, production line design,
production planning and scheduling, quality control and analysis, just in time and
material distribution, to simulate the production logistics in the typical manufac-
turing factory. We will put forward the experiment design, group and role allo-
cation, experiment procedures, experiment result analysis, the effect evaluation
and improvement and so on.
Keywords CDIO concept Production logistics Comprehensive and project-
based experiment Role exchange
165.1 Analysis of Teaching Problems
Production and Operation Management, Facilities Planning and Logistics are the
important courses for the students whose major is the Logistics Engineering, the
core contents of the two courses are closely related to actual operation of enter-
prises (especially the manufacturing plants) and are the core curriculum to culti-
vate production logistics professional (Zhang 2006). The two courses cover the
market demand analysis, facilities planning and layout, logistics systems analysis
and design, organization and design of flow line, production planning and control,
Y. Li (&) X. Lan
Mechanical Engineering College, Zhejiang University of Technology,
Hangzhou, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1567
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_165,
Ó Springer-Verlag Berlin Heidelberg 2013
1568 Y. Li and X. Lan
quality control, work study and time study, business process reengineering,
advanced manufacturing systems and so on, which have strong practical (Jiang
2006; Li and Chen 2008).
Among the many interesting facts we know about how experiences affect
learning, one relates especially to CDIO (Qi and Wu 2010): Engineering students
tend to learn by experiencing the concrete and then applying that experience to the
abstract. Unlike their counterparts of yesteryear, many engineering students these
days don’t arrive at college armed with hands-on experiences like tinkering with
cars or building radio (Li et al. 2011; Feng 2009). CDIO has open and accessible
channels for the program materials and for disseminating and exchanging
resources. CDIO collaborators have assembled a unique development team of
curriculum, teaching and learning, assessment, design and build, and communi-
cations professionals. They are helping others to explore adopting CDIO in their
institutions (Cheng 2006).
Current teaching model pays more attention on the basic theory (Xu 2003; Chen
and Peng 2007), the teacher mainly use the lecture notes, the students are accepted
the basic knowledge passively. It is lack of comprehensive, systematic experiment
course, teachers and students are lack of communication deeply, the students are
lack of real feeling to the knowledge, it is difficult to achieve the teaching
effectiveness. Although the teacher use some auxiliary teaching material and tools
(Xiao and Zheng 2008; Zhang and Bo 2004; Jian and Li 2008), such as cases
study, video explore, basic understanding of practice, there are some problems in
teaching model, which include as follows:
(1) In case study, there are some descriptive content, but there are lack of actual
data, the case is far away from the practices, which is hard to attract the
students interesting.
(2) The comprehensive video is lack of production logistics, the special subject of
production logistics is lack, it is difficult to integrate the curriculum content
closely.
(3) The understanding of the practice is too short a, it is difficult for students to
understand the application of professional knowledge deeply. The existing
curriculum design is limited to theoretical and lack of practices supporting.
(4) The experiment teaching software pay more attention on the solution of the
model of production and management activities, not on the production
logistics analysis which is play more roles on culturing. It is hard to achieve
the teaching purpose which is to culture the applying and innovative ability.
(5) The knowledge and emotional experience of juniors are far away from the
actual business operation, with the abstraction and boring the course knowl-
edge, which lead to the lack of study interesting for students.
In response to these problems, it is necessary to design a comprehensive
experiment on the logistics engineering and industrial engineering laboratory
platform based on the CDIO engineering education philosophy (Bartholdi and
Hackman 2008). The comprehensive experiment will integrate many courses to
improve the understanding and application to the organization, design, operation
165 Comprehensive Experiment Design of Production Logistics Based on CDIO 1569
and control for the production logistics in manufacturing factory (Frazelle 2002).
The comprehensive experiment can attract the interesting in professional courses.
165.2 Experiment Objectives Design
The experiment purpose is to help the students to learn how to use the theory
knowledge comprehensively. The experiment will provide a production process
model by independent design, analysis and optimization for students to cultivate
the independent analysis and problem-solving ability. The detail objectives include
as follows:
(1) To change the single teaching method and improve the teaching effectiveness,
which can improve the application of CDIO concept in the high education;
(2) To help the students to understand the production and operation in manu-
facturing enterprises deeply, to improve the understanding on basic theory and
methods in courses, to reduce abstraction and boring feeling, and to attract the
study interesting;
(3) To improve the recognize, understanding and application to the internal pro-
duction logistics system of manufacturing plant, to enhance recognize and
interesting on the production logistics jobs, to broaden the employment view
and choice;
(4) To improve the application ability to solve practical problems with compre-
hensive knowledge, and to culture the creativity and teamwork ability.
165.3 Experiment Syllabus Design
165.3.1 CDIO Concept and Philosophy
Engineering education and real-world demands on engineers have in recent years
drifted apart. Realizing that this widening gap must be closed, leading engineering
schools in the USA, Europe, Canada, UK, Africa, Asia, and New Zealand formed
the CDIO Initiative: A worldwide collaborative to conceive and develop a new
vision of engineering education (Zhang 2006).
CDIO is an initialism for Conceive–Design–Implement–Operate, which is an
innovative educational framework for producing the next generation of engineers.
The framework provides students with an education stressing engineering funda-
mentals set in the context of Conceiving–Designing–Implementing–Operating
real-world systems and products (Feng 2009; Cheng 2006). Throughout the world,
CDIO Initiative collaborators have adopted CDIO as the framework of their
curricular planning and outcome-based assessment.
1570 Y. Li and X. Lan
The CDIO concept was originally conceived at the Massachusetts Institute of
Technology in the late 1990’s. In 2000, MIT in collaboration with three Swedish
universities (i.e. Chalmers University of Technology, Linköping University and
the Royal Institute of Technology) formally founded the CDIO Initiative. It
became an international collaboration, with universities around the world adopting
the same framework.
CDIO collaborators recognize that an engineering education is acquired over a
long period and in a variety of institutions, and that educators in all parts of this
spectrum can learn from practice elsewhere. The CDIO network therefore wel-
comes members in a diverse range of institutions ranging from research-led
internationally acclaimed universities to local colleges dedicated to providing
students with their initial grounding in engineering.
The CDIO Initiative is rich with student projects complemented by internships
in industry, features active group learning experiences in both classrooms and in
modern learning workshops/laboratories, and rigorous assessment and evaluation
processes.
The CDIO Initiative’s goals are to educate students to master a deeper working
knowledge of the technical fundamentals, to educate engineers to lead in the
creation and operation of new products and, to educate future researchers to
understand the importance and strategic value of their work.
The collaborators maintain a dialogue about what works and what doesn’t and
continue to refine the project. Determining additional members of the collabora-
tion is a selective process managed a Council comprising original members and
early adopters.
165.3.2 Experiment Syllabus
Under the CDIO teaching concept, experiment includes many professional cour-
ses, relevant principles and theories, which includes the production demand
analysis and forecast, facility planning and layout, flow line organization and
balancing, through output analysis, production planning and schedule, quality
control and statistical analysis, Just in Time system, Kanban system and so on. The
experiment syllabus includes:
(1) To master the market demand analysis and forecast; to understand the JIT
production model; to grasp the basic production planning methods and pro-
duction analysis methods.
(2) To familiar with the general methods of facility planning and layout, to use
these methods to analyze the production logistics system; and to know well the
logistics equipment and the basic process of internal logistics and so on.
(3) To grasp the assembly line design and balancing method, know well about the
application of line balancing software and tools; to use the general standard
165 Comprehensive Experiment Design of Production Logistics Based on CDIO 1571
time method and tools; to understand the important role of the assembly line
organization and management in manufacturing system.
(4) To master the common tools and statistical software in quality control and
analysis; to understand the impact on the production of the quality fluctua-
tions; to understand the basic quality management knowledge and concepts,
such as the qualified rate, sample testing, pass-through rate, rework rate,
downgrade management and so on.
(5) To understand the organization, design, operation and control system in
manufacturing plant, to improve the interesting in study; to help students to
grasp the core operation process of manufacturing plants comprehensively and
systematically.
165.4 Experimental Instructions
165.4.1 Experiment Roles
The experiment includes the background and roles design and setting. The
experiment background is a children toys manufacturing plant, which has com-
plete organization, flexible assembly lines and production facilities, the production
mode is the JIT system. The third parts supplier can provide all the materials in the
BOM.
There are 7 roles, including one teacher and 10 students (the ID is from S1 to
S10), the initial roles setting is in the Table 165.1.
165.4.2 Experiment Procedure
The comprehensive experiment procedure flow is shown in the Fig. 165.1 and the
detailed procedure is shown in the Table 165.2.
165.5 Experiment Implementation
Experiment goes into the teaching guideline as an independent course; the
experiment needs 12 h totally, with group exchange model to allocate the time.
For example, there are about 30 students in the Logistics Engineering classes, all
the students can be divided into three groups, and there are 10 members in each
group. Each group needs 2 h in the laboratory, this procedure will repeat three
times for each group. The role setting can be exchanged at each time based on the
needs and interesting. The difficulty will increase gradually.
1572 Y. Li and X. Lan
Table 165.1 Experiment roles
Role Actor Responsibility
Customer Teacher Supplying the demand information dynamitic and negotiate with the
manufacturer
Market S1 Team leader, Collect the market demand information, capacity analysis
and confirm the orders; organize production meeting and record the
production data; responsible for the production process design and
implementation
Production S2 According to orders and production conditions, draw up the production
planning planning tasks, track and adjust the production schedule in real-
time
Assembly line S3–S7 Student 3, 4 and 5 are responsible for assembly tasks, Student 6 and 7
are responsible for the take-apart task. The allocation can be
adjusted dynamically based on the orders
Material S8 Responsible for the material supply based on the JIT and Kanban
Supplier model
Distribution S9 Optimize the delivery route and the products and disassembled parts
distribution
Quality S10 Responsible for production line quality inspection, statistical analysis
and keeping improvement
CDIO
Group and role
Basic theory
allocation
teacher team
Orders
Demand analysis
teacher
Flow line organization and
Capacity analysis
balancing
Production planing
Material supply
team
Team
2 times
Role re-
allocation
Execution at lab
Summary third time
improvement Team and teacher
teacher
Fig. 165.1 The produce flow of the experiment
165 Comprehensive Experiment Design of Production Logistics Based on CDIO 1573
Table 165 2 Experiment detail steps
Step Tasks
Step 1 Introduce the experiment theory, principle, procedures, tasks; students ask questions;
(2 h) Students are divided into three groups, each group includes 10 students, set the role
for each student;
Teachers provide the production data and orders; Students discuss and decide
whether or not to take the orders; the team leader will allocate the tasks among
the team to fulfill the order-demand
Step 2 Group take the production preparation meeting to analysis the capacity and draw up
(2 h) the detailed production planning, finally, publish the task to the assembly line
Change the laboratory layout based on the demand
After receiving the production task, assembly group put forward the layout and
organization, including task assignment, balancing, material supply, product
distribution
Suppliers give the materials supply planning to ensure the operation; Distributor
explore the product distribution planning
Step 3 Do the experiment at the lab
(6 h) First time, the teacher will participate the whole process an give some advice on-site,
students will fulfill the experiment independently at the second and third time.
The team leader (S1) is responsible for the organization
Each student should be responsible to the team leader, and complete their own tasks
The above three steps are repeated three times, the group members can change the
role ate each experiment based on their interests and practice. The difficulty will
be improved gradually
Step 4 Summary the experience and communication;
(2 h) Hand in the report document, show the PPT, do some experience exchangement;
Give some advices to improve the experiment;
Teacher reviews and scores each group and each student
165.6 Conclusion
We do some tests in the class Logistics Engineering 2008 at Zhejiang University of
technology from September 2011 to December 2011. The results show that we
have made a perfect improvement on teaching; the single teaching method has
been changed. The students generally reflected that they understand of the basic
theory through project-based experiments deeply, their interesting to study have
been increased greatly. They learned the practice knowledge and the horizons of
employment have been broadened greatly.
Comprehensive experiment course achieved a perfect teaching effectiveness
and teaching evaluation, the total score was 98.73, the teaching resources index,
teaching content index, teaching methods index, teaching effectiveness index and
teaching services index were 9.85,9.88,9.87,9.85 and 9.88 respectively. The
experiment results show that the comprehensive experiment has a strong maneu-
verability and practice value.
1574 Y. Li and X. Lan
References
Bartholdi JJ, Hackman ST (2008) Allocating space in a forward picking area of a distribution
center for small parts. IIE Trans 40:1046–1053
Chen Z, Peng Y (2007) Application of FR in production operation and management. China Educ
Guid 14:58–59
Cheng Z (2006) Study on production operation and management course construction and
teaching method for MBA. Educ Mod 9(3):3–8
Feng G (2009) Practical teaching research on the course of production and operation
management. Res Explor Lab 28(1):118–120
Frazelle EH (2002) World-class warehouse and material handling. McGraw Hill, New York
Jian X, Li Z (2008) Storage location assignment in a multi aisle warehouse considering demand
correlations. Comput Integr Manuf Syst 14(12):2447–2451
Jiang Z (2006) Industrial engineering curriculum design guidance. Machinery Industry Press,
Beijing, China
Li C, Chen Y (2008) Teaching reform of the industrial engineering curriculum design. Educ
Innov Guide 1:23–26
Li H, Fang Z, Wang Y (2011) Industrial engineering practice teaching system planning and
construction. China Electr Power Educ 10:57–60
Qi L, Wu S (2010) Industrial engineering theory teaching, laboratory teaching, curriculum design
trinity of design and implementation. China Electr Power Educ 32:112–115
Xiao J, Zheng L (2008) Storage location assignment in a multi aisle warehouse considering
demand correlations. Comput Integr Manuf Syst 14(12):2447–2451
Xu Z (2003) Course design of factory visiting in production operation and management. J Xiamen
Univ (Nat Sci) 42(10):144–147
Zhang X (2006) Industrial engineering experiments and practical tutorial. Machinery Industry
Press, Beijing, China
Zhang YF, Bo L (2004) Application of genetic algorithm in selecting accurate freight site. J Syst
Simul 16(1):168–171
Chapter 166
Improved Grey Forecasting Model
for Taiwan’s Green GDP Accounting
Shin-li Lu, Ching-I Lin and Shih-hung Tai
Abstract This paper applies the grey forecasting model to forecast the green GDP
accounting of Taiwan from 2002 to 2010. Green GDP accounting is an effective
economic indicator of human environmental and natural resources protection.
Generally, Green GDP accounting is defined as the traditional GDP deduces the
natural resources depletion and environmental degradation. This paper modifies
the original GM(1,1) model to improve prediction accuracy in green GDP
accounting and also provide a value reference for government in drafting relevant
economic and environmental policies. Empirical study shows that the mean
absolute percentage error of RGM(1,1) model is 2.05 % lower than GM(1,1) and
AGM(1,1), respectively. Results are very encouraging as the RGM(1,1) fore-
casting model clearly enhances the prediction accuracy.
Keywords Grey theory Forecasting Green GDP accounting
166.1 Introduction
Energy consumption and the threat of global warming have drawn nation and
international attention. In 1992, the Commission for Sustainable Development of
the United Nation signed the convention to pursue equilibrium between ecological
reservation and economic development. In 1997, Taiwan’s government promul-
gated an Article 10 amendment of Taiwan’s Constitution to support environmental
S. Lu
Department of Industrial Management and Enterprise Information,
Aletheia University, Taipei, Taiwan, China
C.-I. Lin S. Tai (&)
Department of Industrial Management, Lunghwa University of
Science and Technology, Taipei, Taiwan, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1575
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_166,
Ó Springer-Verlag Berlin Heidelberg 2013
1576 S. Lu et al.
and ecological protection. To implement this policy, the Executive Yuan’s
Environmental Protection Administration (EPA) invited relevant departments to
discuss Taiwan’s green Gross Domestic Product (GDP) accounting under the
System of Integrated Environmental and Economic Accounting (SEEA) led by the
United Nations and the World Bank. An initial effort has been made to collect
required data and coordinate with pertinent departments to establish a database for
green GDP accounting and then officially released Taiwan’s green GDP
accounting by the Directorate General of Budget, Accounting and Statistics
(DGBAS).
Green GDP is a very important indicator to reflect the real domestic wealth,
which tries to take into account some important determinants of human welfare
and therefore is believed to be a better indicator of a country’s welfare than the
traditional GDP. With regard to the measurement of the green GPD accounting,
Yue and Xu (2008) and Yue et al. (2009) classified as two main types. Type I
green GDP accounts GDP minus the cost of environmental quality degradation and
natural resources depletion, but it ignored the value of natural ecosystem services.
As Heal (2007) and Xu et al. (2010) pointed out that the value of direct ecosystem
services need to consider in green GDP, which is type II green GDP accounting.
Currently, the SEEA framework has been adopted for compiling Taiwan’s green
GDP accounting, which is similar to Type I green GDP accounting. The system,
introduced by UN and the World Bank, is supported by major international
organizations and environmental specialists and widely adopted by more than 20
countries including the United States, Japan, South Korea and Canada.
The depletion of Taiwan’s natural resources include the depletion of
groundwater, crude oil, natural gas, coal and gravel as determined by the net price
method, which means exploitative gain minus exploitative cost. Factors such as
over-fishing, illegal exploitation of coral reefs and excessive land development
have not been included in natural resources depletion. According to the DGBAS in
Taiwan, the natural resources depletion was reduced from NT$ 20.70 billion of
2002 to NT$ 18.19 billion of 2010. The depletion of groundwater continued to the
top of the list of 2010, decreasing by 10.95 % from 2002, to represent 80.65 % of
the natural resources depletion.
The estimation of Taiwan’s environmental degradation has been taken by the
maintenance cost method, which means the act of pollution without taking any
preventive measures is applied. Currently, the DGBAS only accounts for water, air
and solid waste pollution in calculating environmental degradation, while noise
pollution, soil pollution and greenhouse effects are temporarily left out since most
countries have not taken into account such categories. Environmental degradation
throughout Taiwan in 2010 totaled NT$ 63.4 billion, up 4.52 % compared to the
2002 of NT$ 60.66 billion. The findings show that the degradation by water pol-
lution in 2010 at NT$ 35.1 billion, the highest among three categories, to represent
53.61 % of the environment degradation. The government has stepped up its efforts
in improving the water pollution from major sources of industrial and residential
discharges. The degradation of water pollution has obviously declined 10.07 %
since 2002. With rapid industrial and commercial developments, however, the
166 Improved Grey Forecasting Model 1577
degradation of air pollution has increased 28.51 % since 2002. Through reduction,
recycling and proper disposal of the solid waste in the past years, the degradation of
solid waste is much less than those of water and air pollution.
In 2010, there was NT$ 13.61 trillion created in GDP, but negative impact on
environment was accompanied by a high economic growth. One is depletion of
natural resources totaled NT$ 18.19 billion and the other is environmental deg-
radation up to NT$ 65.47 billion. Consequently, the green GDP accounting is NT$
13.53 trillion, up by 30.97 percent compared to NT$ 10.33 trillion in 2002. The
dramatic growth in green GDP accounting may attribute to the increasing envi-
ronmental awareness of Taiwanese and the policy implementation by the gov-
ernment. Accordingly, one of main concerns in this article is to construct
forecasting model to forecast green GDP accounting of Taiwan. The proposed
model not only can reflect how much degree paid on environmental protection but
support government to draft pertinent policies for Taiwan’s environmental issues.
Time series models are widely used in predicting and acquiring management
information. A large number of observations are required to understand the pattern
and choose a reasonable mathematical model for time series process. Unfortu-
nately, only a little data are obtained over time and simultaneously we are inter-
ested in speculating succeeding observations in the future. Neither statistical
methods nor data mining techniques are suitable for exploring the small obser-
vation problem. The grey system theory, originally developed by Deng (1982),
effectively deals with limit data and uncertain information. Since then, the grey
system theory has become popular when we have only incomplete information and
also successfully applied to various fields such as transportation (Pai et al. 2007),
energy (Hsu and Chen 2003; Akay and Atak 2007), financial (Chang and Tsai
2008; Huang and Jane 2009; Kayacan et al. 2010), social and economic (Shen
et al. 2009), engineering (Li and Yeh 2008) and so on. Above mentioned articles,
the grey system theory is utilized in this work to forecast green GDP accounting of
Taiwan.
166.2 Methodology
166.2.1 Original GM(1,1) Forecasting Model
The aim of this article is to construct a green GDP accounting forecasting model
based on grey system theory. Unlike statistical methods, this theory mainly deals
with original data by accumulated generating operations (AGO) and tries to find its
internal regularity. Deng (1986) has been proven that the original data must be
taken in consecutive time period and as few as four. In addition, the grey fore-
casting model (GM) is the core of grey system theory and the GM(1,1) is one of
the most frequently used grey forecasting model. The GM(1,1) model constructing
process is described as follows:
1578 S. Lu et al.
Step 1: Denote the original data sequence.
xð0Þ ¼ xð0Þ ð1Þ; xð0Þ ð2Þ; xð0Þ ð3Þ; . . .; xð0Þ ðnÞ ; n4 ð166:1Þ
Step 2: Use AGO to form a new data series.
xð1Þ ¼ xð1Þ ð1Þ; xð1Þ ð2Þ; xð1Þ ð3Þ; . . .; xð1Þ ðnÞ ;
where xð1Þ ð1Þ ¼ xð0Þ ð1Þ and
X
k
xð1Þ ðkÞ ¼ xð0Þ ðiÞ; k ¼ 2; 3; . . .; n ð166:2Þ
i¼1
Step 3: Calculate background values zð1Þ
zð1Þ ðkÞ ¼ ð1 aÞxð1Þ ðk 1Þ þ axð1Þ ðkÞ; a 2 ð0; 1Þ ð166:3Þ
Step 4: Establish the grey differential equation.
dxð1Þ ðkÞ
þ axð1Þ ðkÞ ¼ b ð166:4Þ
dt
where a is the developing coefficient and b is the grey input.
Step 5: Solve Eq. (166.4) by using the least square method and the forecasting
values can be obtained as
8
< ^xð1Þ ðkÞ ¼ xð0Þ ð1Þ b eaðk1Þ þ b
>
a a ð166:5Þ
>
: ^ð0Þ ð1Þ ð1Þ
x ðkÞ ¼ ^x ðkÞ ^x ðk 1Þ
where
½a; bT ¼ ðBT BÞ1 BT Y ð166:6Þ
Y ¼ ½xð0Þ ð2Þ; xð0Þ ð3Þ; . . .; xð0Þ ðnÞT ð166:7Þ
2 3
zð1Þ ð2Þ 1
6 ð1Þ 7
6 z ð3Þ 17
6 7
B¼6 .. .. 7 ð166:8Þ
6 7
4 . .5
zð1Þ ðnÞ 1
166 Improved Grey Forecasting Model 1579
166.2.2 Residual GM(1,1) forecasting model
The residual modification GM(1,1) model, called RGM(1,1), first developed by
(Deng 1982). The differences between the original series, xð0Þ , and the GM(1,1)
model forecasting values, ^xð0Þ , are defined as the residual series. Hence, the
absolute values of the residual series eð0Þ can be represented as:
eð0Þ ¼ eð0Þ ð2Þ; eð0Þ ð3Þ; eð0Þ ð4Þ; . . .; eð0Þ ðnÞ ; ð166:9Þ
where
eð0Þ ðkÞ ¼ xð0Þ ðkÞ ^xð0Þ ðkÞ; k ¼ 2; 3; . . .; n: ð166:10Þ
Execute the Steps 1–5, a RGM(1,1) forecasting model can be established and
the forecasting values ^eð0Þ ðkÞ be:
be
^eð0Þ ðkÞ ¼ eð0Þ ð2Þ ð1 eae Þeae ðk1Þ ; k ¼ 3; 4; . . .; n ð166:11Þ
ae
Considering the residual modification on GM(1,1) model can improve the
predictive accuracy of the original GM(1,1) model.
166.2.3 Adaptive GM(1,1) Forecasting Model
Li and Yeh (2008) proposed the trend and potency tracking method (TPTM) to
acquire concealed information, and then construct a triangular trend and potency
(TP) function with an asymmetrical domain range. TP values of the existing data
are determined by ration rule of a triangle and represented the current datum’s
intensity close to the central location. The detailed procedure of computing TP
values is described by Li and Yeh (2008).
Moreover, the background value is the most important factor which affects the
model’s adoption and precision. Many researchers generally regard each datum as
having equal importance, and set a ¼ 0:5 in Eq. (166.3) to compute the back-
ground value. However, Li et al. (2009) discussed the influence of a and renamed
Eq. (166.3) as
zð1Þ ðkÞ ¼ xð1Þ ðk 1Þ þ axð0Þ ðkÞ; a 2 ð0; 1Þ; k ¼ 2; 3; . . .; n:
Clearly, the influence from a to the background value mainly comes from the
newest data. Therefore, the Adaptive GM(1,1), known as AGM(1,1), are presented
by Li et al. (2009) and described as follows:
Step 1–2 are same as original GM(1,1).
Step 3: Calculate the TP values by TPTM
fTPi g ¼ fTP1 ; TP2 ; . . .; TPn g; i ¼ 1; 2; . . .; n ð166:12Þ
1580 S. Lu et al.
Step 4: ak is computed by
Pk i1
i¼1 2 TPi
ak ¼ P k
; k ¼ 2; 3; . . .; n ð166:13Þ
i1
i¼1 2
Step 5: Calculate background values.
zð1Þ ðkÞ ¼ xð1Þ ðk 1Þ þ ak xð0Þ ðkÞ; a 2 ð0; 1Þ ð166:14Þ
Step 6: Establish the grey differential equation and estimate the developing
coefficient a and the grey input b by least square method to obtain forecasting
value of AGM(1,1).
166.3 Empirical Studies
To demonstrate the precision and stability of grey forecasting method, the relevant
green GDP accounting provided by DGBAS are examined in this study. The
historical annual data of original GDP accounting, natural resources depletion,
environmental degradation and green GDP accounting from 2002 to 2010 are
presented in Table 166.1.
166.3.1 Formulating the Three Compared Models
(1) Original GM(1,1)
The original data sequence is obtained as xð0Þ ¼ ½103:30; 106:18; . . .; 135:32
based on the green GDP accounting in Taiwan. The parameters of a and b of
original GM(1,1) model are estimated by the least-square method through the
Table 166.1 Values of the relevant green GDP accounting from 2002 to 2010 (NT$ Billion)
Years GDP Natural resources Environmental Green GDP
accounting depletion degradation accounting
2002 10411.63 20.70 60.66 10330.27
2003 10696.25 20.29 57.59 10618.37
2004 11365.29 21.07 67.14 11277.08
2005 11740.27 19.55 66.64 11654.08
2006 12243.47 18.58 66.14 12158.75
2007 12910.51 18.58 67.23 12824.70
2008 12620.15 18.07 65.39 12536.68
2009 12481.09 17.60 63.20 12400.28
2010 13614.22 18.19 63.40 13532.62
166 Improved Grey Forecasting Model 1581
Eqs. (166.2)–(166.4). (a ¼ 0:029, b ¼ 105:00). The original GM(1,1) forecasting
model is listed as follow:
ð0Þ ð0Þ 105:001
^x ðkÞ ¼ x ð1Þ þ 1 e0:029 e0:029ðk1Þ ; k ¼ 2; 3; . . .; n
0:029
(2) Residual GM(1,1)
The residual data sequence is built by Eq. (166.10). Repeat the Eqs. (166.2)–
(166.4) to estimate the parameters of a and b of RGM(1,1) model (a ¼ 0:128,
b ¼ 5:374). The RGM(1,1) forecasting model is listed as follow:
ð0Þ ð0Þ 5:374
^e ðkÞ ¼ e ð2Þ 1 e0:128 e0:128ðk1Þ ; k ¼ 3; 4; . . .; n
0:128
(3) Adaptive GM(1,1)
Original data sequence xð0Þ is adopted to establish the TP values by Li et al.
(2009). In order to emphasize the effect of the newest datum, the weight is changed
in the background value. Different weights, ak , are calculated as {0.52,0.66,0.79,
0.87,0.83,0.85,0.87,0.75} in accordance with TP values, {TP} = {0.45,0.54,
0.77,0.90, 0.94,0.79,0.86,0.89,0.64}. Finally, the parameters of a and b of
AGM(1,1) model are estimated by least square method. (a ¼ 0:027,
b ¼ 104:594). The AGM(1,1) forecasting model is listed as follow:
ð0Þ ð0Þ 104:594
^xA ðkÞ ¼ xA ð1Þ þ 1 e0:027 e0:027ðk1Þ ; k ¼ 2; 3; . . .; n
0:027
166.3.2 Results
The predicted results obtained by the original GM(1,1), residual GM(1,1) and
adaptive GM(1,1) model are presented in Table 166.2 and Fig. 166.1. To measure
the forecasting performance, mean absolute percentage error (MAPE), is used for
evaluation of these models. The results indicate that the RGM(1,1) has the smallest
MAPE (2.05 %) compared with original GM(1,1) and AGM(1,1) (3.25 and
2.32 %, respectively). Therefore, RGM(1,1) model not only can reduce the fore-
casting error effectively, but enhance the precision of a grey forecasting model.
However, the absolute percentage error (APE) of the GM(1,1), RGM(1,1) and
AGM(1,1) models in 2007 are 6.99, 4,39 and 5.14 %, respectively, which is higher
than its MAPE.
1582 S. Lu et al.
Table 166.2 Forecasting values and errors of green GDP accounting (NT$ 0.1*Trillion)
Years AVa GM(1,1) RGM(1,1) AGM(1,1)
FVb Error (%)c FV Error (%) FV Error (%)
2002 103.30
2003 106.18 106.42 0.23 108.94 2.60
2004 112.77 109.50 2.90 114.57 1.60 111.99 0.69
2005 116.54 112.67 3.32 117.13 0.51 115.12 1.22
2006 121.58 115.92 4.65 119.86 1.42 118.34 2.66
2007 128.24 119.28 6.99 122.74 4.29 121.65 5.14
2008 125.36 122.72 2.10 125.77 0.33 125.05 0.24
2009 124.00 126.27 1.83 128.95 4.00 128.55 3.67
2010 135.32 129.92 3.99 132.29 2.24 132.15 2.34
MAPEd 3.25 2.05 2.32
a
AV Actual value
b
FV Forecasting value
c
Error = jFVk AVk j=AVk
P
d
MAPE = 1n nk¼1 ½jFVk AVk j=AVk
Fig. 166.1 Actual values 140
and forecasting values for
green GDP accounting of 135 AGM(1,1)
Taiwan from 2002 to 2010
130
125 RGM(1,1)
Trillion
120
115 GM(1,1)
110
105
100
2002 2003 2004 2005 2006 2007 2008 2009 2010
Year
166.4 Conclusions
Numerous forecasting methods have been widely used, including the time series
analysis, regression analysis and artificial neural networks. They need a large
amount of data to construct a proper forecasting model. With the life cycle of
products, however, data collected are limit. Adopting traditional forecasting
methods with a few uncertain and insufficient data to build forecasting model is
unsuitable. Therefore, in order to obtain a highly accurate forecasting model with
limit data, Deng (1986) first presented the grey forecasting model from grey theory
to overcome the problem facing a few data. Accordingly, the goal of this paper is
166 Improved Grey Forecasting Model 1583
to forecast the green GDP accounting of Taiwan by original GM(1,1) model and
compare to residual GM(1,1) and adaptive GM(1,1) model.
To measure the performance of the GM(1,1), RGM(1,1) and AGM(1,1) models,
the criteria of MAPE is adopted. Empirical results indicate that the RGM(1,1)
forecasting model has the lowest MAPE, 2.05 %, among three models. That is,
RGM(1,1) forecasting model has a high prediction validity of forecasting the green
GDP accounting in Taiwan. The findings serve as a basis for government decision
making to make Taiwan become Green Islands both economically and
environmentally.
The results are very encouraging as they show that green GDP accounting
represented the human welfare is increasing during the last decade. More impor-
tant, natural resources depletion and environmental degradation are debit entries to
green GDP accounting, which represent negative environment impacts rising from
the economic developments achieved. Therefore, in order to pursue a high human
welfare and sustainable development of ecosystem, Taiwan government and
Taiwanese must cooperate together to execute pertinent environmental policies.
References
Akay D, Atak M (2007) Grey prediction with rolling mechanism for electricity demand
forecasting of Turkey. Energy 32:1670–1675
Chang BR, Tsai HF (2008) Forecast approach using neural network adaptation to support vector
regression grey model and generalized auto-regressive conditional heteroscedasticity. Expert
Syst Appl 34:925–934
Deng JL (1982) Grey system fundamental method. Huazhong University of Science and
Technology, Wuhan
Deng JL (1986) Grey prediction and decision. Huazhong University of Science and Technology,
Wuhan
Heal G (2007) Environmental accounting for ecosystems. Ecol Econ 6:693–694
Hsu CC, Chen CY (2003) Applications of improved grey prediction model for power demand
forecasting. Energy Convers Manage 44:2241–2249
Huang KY, Jane CJ (2009) A hybrid model for stock market forecasting and portfolio selection
based on ARX, grey system and RS theories. Expert Syst Appl 36:5387–5392
Kayacan E, Ulutas B, Kaynak O (2010) Grey system theory-based models in time series
prediction. Expert Syst Appl 37:1784–1789
Li DC, Yeh CW (2008) A non-parametetric learning algorithm for small manufacturing data sets.
Expert Syst Appl 34:391–398
Li DC, Yeh CW, Chang CJ (2009) An improved grey-based approach for early manufacturing
data forecasting. Comput Ind Eng 57:1161–1167
Pai TY, Hanaki K, Ho HH, Hsieh CM (2007) Using grey system theory to evaluate transportation
effects on air quality trends in Japan. Transp Res Part D 12:158–166
Shen VRL, Chung YF, Chen TS (2009) A novel application and grey system theory to
information security (part I). Comput Stan Interfaces 31:277–281
Xu L, Yu B, Yue W (2010) A method of green GDP accounting based on eco-service and a case
study of Wuyishan, China. Procedia Environ Sci 2:1865–1872
1584 S. Lu et al.
Yue WC, Xu LY (2008) Study on the accounting methods of Green GDP based on ecosystem
services. Ecol Econ 9:50–53
Yue WC, Xu LY, Zhao X (2009) Research of Green GDP accounting in Wuyishan City based on
ecosystem services. Ecol Econ 2:11–12
Chapter 167
Simulation Research of the Fuzzy Torque
Control for Hybrid Electrical Vehicle
Based on ADVISOR
Bo-jun Zhang, Yu Wang and Jia-tian Guo
Abstract The simulation model of the super-mild hybrid electrical vehicle is
established through the simulation software ADVISOR. The fuzzy logic torque
distribution controller of motor and engine is designed. The drive cycle selects
Urban Dynamometer Driving Schedule (UDDS). The simulation results show that
the fuzzy torque controller of motor and engine can properly distribute the torque.
The fuel economy and emission performance are improved.
Keywords ADVISOR Hybrid electrical vehicle Simulation Torque control
167.1 Vehicle Simulation Model
The simulation model of the super-mild hybrid vehicle is established through the
simulation software ADVISOR. This Model is shown in Fig. 167.1.
The backward simulation and forward simulation can be used in ADVISOR.
The backward simulation can calculate the engine and motor output power. For-
ward simulation is to be after the backward simulation, it can use the engine and
motor power along with backward simulation passing in the opposite direction.
The actual speed of vehicle is calculated by the forward simulation (Wipke et al.
1999). Every module in the vehicle simulation model contains a simulink simu-
lation module. We can modify the parameters in the M-file for data input (Zeng
et al. 2004).
The parameters in the M-file of vehicle module are modified according to the
parameters of the entire vehicle, such as:
veh_gravity = 9.81; % m/s2
B. Zhang (&) Y. Wang J. Guo
Department of Automotive and Transportation, Tianjin University of Technology
and Education, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1585
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_167,
Springer-Verlag Berlin Heidelberg 2013
1586 B. Zhang et al.
Fig. 167 1 The super-mild hybrid vehicle simulation model
veh_air_density = 1.2; % kg/m3
veh_mass = 865; % (kg)
vehicle_height = 1.380; % (m)
vehicle_width = 1.590; % (m)
veh_CD = 0.38;
veh_FA = 1.7; % (m2)
veh_wheelbase = 2.365; % (m)
167.2 Torque Fuzzy Logic Control
The super-mild hybrid electrical vehicle has no pure electric operating mode.
Motor only be used for idle starting, stopping, power compensation and recycling
braking energy, so the motor torque control will affect the vehicle performance
(Zhang et al. 2010; Deng et al. 2004; Fan and Wu 2004; Liang and Wang 2001).
The output torque of the engine can be divided into two parts: One is used to
drive the vehicle, the other is used to drive the generator for the battery charging.
The input/output torque of the motor can balance the relationship between engine
torque and vehicle required torque, which controls the engine working point along
the economic curve (Schouten et al. 2002; Lee and Sul 1998). When the engine
output torque is lower than the vehicle required torque, the motor will make up the
difference. When the engine output torque is higher than the vehicle required
torque or the vehicle is in deceleration braking condition, the extra engine output
torque or the recovery of deceleration energy can drive generator for battery
charging (Schoutena et al. 2002, 2003; Kheir et al. 2004; Poursamad and
Montazeri 2008; Baumann et al. 2000). The fuzzy controller not only improves the
vehicle fuel economy but also makes the battery SOC value to be in the high
efficiency range.
Take the differential values between the vehicle required torque and the engine
output torque (Tc), and battery SOC values to be input variables. Take the torque
167 Simulation Research of the Fuzzy Torque 1587
Fig. 167.2 Membership function
adjustment parameter to be output variables. The membership function is shown in
Fig. 167.2.
Input and output variables of the fuzzy sets are as follows:
The fuzzy set for the SOC is: {L, PL, W, PH, H}.
The fuzzy set for the Tc is: {FD, FX, W, ZX, ZD}.
The fuzzy set for the V is: {X, PX, W, PD, D}.
Where: L represents low, PL represents partial low, W represents moderate, PH
represents partial high, H represents high, FD represents negative bigness, FX
1588
B. Zhang et al.
Fig. 167.3 Clutch control
167 Simulation Research of the Fuzzy Torque 1589
Fig. 167. 4 Torque control
Fig. 167.5 Simulation results
1590 B. Zhang et al.
Table 167.1 Engine fuel economy and emissions
Torque controller Fuel economy (L/100 km) Emissions (g/mile)
HC CO NOX
Without fuzzy controller 0.47 0.422 10.82 0.113
With fuzzy controller 0.35 0.33 6.456 0.113
represents negative smallness, ZX represents positive smallness, ZD represents
positive bigness, X represents smallness, PX represents partial smallness, PD
represents partial bigness, D represents bigness.
25 fuzzy control rules are designed to describe the relationship between input
and output. There are some rules to illustrate its function, IF is the premise and
THEN is the conclusion:
a. IF(Tc is W) THEN (k is W)
b. IF(SOC is L) and (Tc is FX) THEN (k is W)
c. IF(SOC is PL) and (Tc is FD) THEN (k is X)
d. IF(SOC is H) and (Tc is ZX) THEN (k is X)
e. IF(SOC is PH) and (Tc is ZD) THEN (k is PD)
…
The modules of \cl[and \tc[realize the fuzzy logic torque control, as shown
in Figs. 167.3 and 167.4.
167.3 Simulation Result
The simulation results are shown in Fig. 167. 5.
The engine fuel economy and emission values are shown in Table 167.1.
The simulation results show that the battery SOC value can be maintained in the
high efficiency range. And the fuel economy of the engine has been improved, the
HC and CO emissions have been lower.
167.4 Conclusion
The ADVISOR simulation model and the fuzzy logic torque controller are
established. The fuzzy logic torque controller is realized through the clutch and
torque control module. The fuzzy torque control strategy can more effectively
distribute the motor and the engine operating range. The fuel economy of the
vehicle has been improved and the emissions have been lower.
Acknowledgments This dissertation is under the project support of Natural Science Foundation
of Tianjin (09JCYBJC04800).
167 Simulation Research of the Fuzzy Torque 1591
References
Baumann BM, Washington G, Glenn BC et al (2000) Mechatronic design and control of hybrid
electric vehicles. IEEE/ASME Trans Mechatron 5(1):58–72
Deng Y, Wang Z, Gao H (2004) Modeling and simulation of hybrid drive system on the Toyota
PRIUS based on bondgraph. J Wu han Univ Technol 2004(4):50–55
Fan J, Wu T (2004) Simulating study of the control strategy for Honda insight. J Guang xi Univ
Technol 2:18–20
Kheir NA, Salman MA, Schouten NJ (2004) Emissions and fuel economy trade-off for hybrid
vehicles using fuzzy logic. Math Comput Simul 66:155–172
Lee H-D, Sul S-K (1998) Fuzzy-logic-based torque control strategy for parallel-type hybrid
electric vehicle. IEEE Trans Ind Electron 45(4):625–632
Liang C, Wang Q (2001) Energy management strategy and parametric design for hybrid electric
family Sedan. SAE paper: 2001-01
Poursamad A, Montazeri M (2008) Design of genetic-fuzzy control strategy for parallel hybrid
electric vehicles. Control Eng Pract 16:861–873
Schouten NJ, Salman MA, Kheir NA (2002) Fuzzy logic control for parallel hybrid vehicles.
IEEE Trans Control Syst Technol 10(3):460–468
Schoutena NJ, Salman MA, Kheira NA (2003) Energy management strategies for parallel hybrid
vehicles using fuzzy logic. Control Eng Pract 11:171–177
Wipke KB, Cuddy MR, Burch SD (1999) ADVISOR user-friendly advanced powertrain
simulation using a combined backward/forward approach. IEEE Trans Veh Technol Spec
Issue Hybrid Electr Veh 1999(5):1–10
Zeng X, Wang Q, Li J (2004) The development of HEV control strategy module based on
ADVISOR2002 software. Automot Eng 26:394–396
Zhang Y, Zhou M, Wang X, Lu X, Yuan B (2010) A study on the control system of regenerative
braking for HEV. J Nat Sci Hei Longjiang Univ 27:551–556
Chapter 168
Vulnerability Analysis and Assessment
System of Natural Disaster
Jiang Shen, Jing Huang, Tao Li and Man Xu
Abstract With regard to the overall vulnerability of the complex natural disaster
system, correlation of disaster-inducing factors, disaster environment and disaster
bearing objects was analyzed, natural disaster forming efficiency was simulated,
and vulnerability mechanism of natural disaster was researched using disaster-
inducing factor-vulnerability chain and vulnerability curves. Assessment decision-
making model of natural disaster vulnerability was built. Through constructing
three index systems, natural disaster vulnerability was assessed by disaster risk
degree, vulnerability of disaster bearing objects and risk loss degree.
Keywords Natural disaster Vulnerability Disaster risk degree Risk loss
degree
168.1 Introduction
According to the research of Janssen (2005), vulnerability appeared 939 times in
2,286 authoritative publications in the past 30 years, especially in natural disaster
emergency management research and government documents, attracting
researchers and policy makers’ concern. In natural disaster research, natural
disaster vulnerability was defined as factors decided by nature, society, economy,
environment, enhancing community sensitivity facing disaster. From the disaster
point of view, vulnerability referred to characteristics easy to be damaged or
J. Shen J. Huang T. Li
College of Management and Economics, Tianjin University, Tianjin,
People’s Republic of China
M. Xu (&)
TEDA College, Nankai University, Tianjin, People’s Republic of China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1593
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_168,
Ó Springer-Verlag Berlin Heidelberg 2013
1594 J. Shen et al.
Table 168.1 Characteristics and performances of natural disaster vulnerability
Characteristics Performances
Numerous types A total of seven categories, 20 kinds, including rainstorm and flood,
drought, tropical cyclone, frost, hail, fog, sandstorms and so on
Wide range Occurred throughout the year, in the mountain, plains, plateau, island, river,
lake, sea and air
High probability Drought, floods and typhoons, and other disasters occurred every year
Long duration The same disaster often occurred quarter after quarter, year after year
Concurrent Some disasters occurred in the same period, such as thunderstorm, hail,
high wind, tornado were often a concurrent phenomenon
Significant chain Weather and climate conditions were often able to cause or worsen floods,
reaction landslide, and plant pests and so on
Severe damage 121.3 million people were killed from 1947 to 1980, caused by natural
disasters according to United Nations announcement
injured by natural disasters, reflecting the affordability of various types of disas-
ters. Various characteristics and performances of natural disaster vulnerability
were shown in Table 168.1.
The frequent occurrence of natural disasters and serious effects of secondary
disasters caused the focus of the overall vulnerability of complex systems of
natural disasters. In the large complex system, the more complex the system, the
more obvious vulnerability to natural disasters. Coupled with the dispersion of the
system, it was a must from an overall perspective to assess the vulnerability of
natural disasters.
168.2 Research Status of Natural Disaster Vulnerability
Global disaster research program had a major impact on disaster risk assessment
index system, such as Disaster Risk Index (DRI) scheme, the world first global
scale human vulnerability assessment index system with spatial resolution to the
countries. Domestic scholars focused on single disaster with risk assessment
system based on index.
(1) Natural disaster risk assessment. The Disaster Risk Hotspots Plan by
Columbia University and ProVention Union established three risk assessment
indexes and disaster risk maps of hazard-prone areas (Arnold et al. 2006).
European Spatial Planning Observation Network elaborated multiple risk
assessment index methods on the potential risk of a particular area (Greiving
2006). The U.S. Federal Emergency Management Agency and National
Institute of Building Sciences developed the HAZUS model, a standardized
national multi-hazard loss estimation method.
(2) Vulnerability assessment. Vulnerability analysis methods were mainly two
ways: the index system and the vulnerability curve. Index system method,
168 Vulnerability Analysis and Assessment System of Natural Disaster 1595
constructed the model through index optimization, weight assignment, and
assessment result of a single disaster or multi-disaster was a relative value
more applicable in the lack of quantitative parameters or data limited cir-
cumstances, such as the pressure release model (Blaikie and Cannon 1994),
the vulnerability model (Cutter 1996). The disadvantage of rough assessment
results and poor operability could be solved by vulnerability curve (Shi et al.
2009), such as the ANUFLOOD model developed by Australian Resources
and Environment Research Center (Gissing and Blong 2004).
168.3 Natural Disaster Vulnerability Mechanisms
Natural disaster vulnerability mechanism was an input–output of operational
efficiency analysis model of a complex giant system. In the process of natural
disasters, disasters were the run results of a regional disaster system, also an input–
output system. The formation of disasters as output factors could be regarded as
the working result of a regional disaster system as an input factors, including
disaster-inducing factors, disaster environment and disaster bearing objects, and
natural disaster vulnerability was the level of disasters forming efficiency. The
greater the vulnerability, the higher the disaster efficiency, and the easier to form a
serious disaster.
Through collecting the relative data of basic information of the regional context
of natural disasters and disaster bearing objects, characteristic information of risk
factors, vulnerability information and historical disaster information, based on the
analysis method of historical disaster, index system and vulnerability curve, nat-
ural disaster risk sources were identified and the refined three dimensional natural
disaster vulnerability mechanism model of was constructed as Fig. 168.1.
Fig. 168.1 Three
spatial overlapping
dimensional natural disaster Disaster forming
efficiency
vulnerability mechanism Input Output
model Disaster-
inducing
factors Vulnerability
Disaster
Interaction
environment
Disaster Simulation Disaster
bearing
objects Time
Risk degree assessment index
system of Major natural disasters
em
st
Vulnerability assessment index
sy
system of disaster bearing object
x
de
In
Assessment index system of
natural disaster risk loss degree
1596 J. Shen et al.
(1) Disaster-inducing factor-vulnerability chain. Through analyzing the internal
and kinetic characteristics of the natural disaster risk bearing objects (popu-
lation, community, region, infrastructure, environment, etc.), the expected loss
under disasters was determined, and the regularity of natural disaster vulner-
ability was described using disaster-inducing factor-vulnerability chain.
(2) Regularity model and threshold of vulnerability. In order to improve the
accuracy of risk assessment, whether disaster bearing objects were in disasters
and disaster characteristics were specific and visual using scenario simulation
method to simulate disaster scenarios and optimize exposure factors. Vul-
nerability regularity and threshold could be the modeled, such as the regularity
of the vulnerability of mechanism model in three dimensions of natural
disasters and vulnerability curve analysis, such as intensity-loss curve,
strength-loss rate curve and intensity-per unit area curve.
168.4 Assessment Decision-Making Model of Natural
Disaster Vulnerability
Assessment decision-making model of natural disaster vulnerability was con-
structed according to the requirements, guidelines, targets and layers of the disaster
loss assessment, comprehensive index system. For the main features of the major
natural disasters, an assessment index system of natural disaster risk was estab-
lished, including the target layer, rule layer and scheme layer.
Assessment of disaster risk degree, vulnerability of disaster bearing objects and
risk loss degree were included in index system of natural disaster risk. In disaster
risk degree assessment, the natural properties of the natural disasters were set as
the basic starting point, and the intensity and the likelihood of risk factors were
determined by analyzing the past frequency and intensity of risk factors. Vul-
nerability of disaster bearing objects, the loss tendency faced with natural disaster
risk, was determined by sensitive natural, social, economic and environmental
factors and interaction. Risk loss assessment evaluated the disaster loss in a certain
risk. The details included the following three assessment index systems.
168.4.1 Risk Degree Assessment Index System
of Natural Disaster
According to risk degree assessments and indexes of main meteorological disasters,
geological disasters, pests and disease disasters, risk degree assessment index sys-
tem of natural disasters was constructed based on historical data for disasters
combining each disaster with corresponding disaster bearing object types, consid-
ering the disaster intensity, the disaster probability of and environmental conditions.
168 Vulnerability Analysis and Assessment System of Natural Disaster 1597
Assessment decision-making Index system of
natural disaster vulnerability
Natural disasters Disaster bearing Risk loss
Startup risk degree assessment objects Vulnerability assessment degree assessment
Cluster Analysis
collection
Data
Observation, Dyna- Disaster formation
monitoring, Real- mic Histo- Anti- process, Disaster bearing objects
forecasting, Basic time socio- rical risk mechanism or developments forecasts
data of geographic sensing econo disas- attri- model: physical, (including future design of
geographic data monito- mic ter data bute statistics, anti-risk properties,
environmental ring data data data experimental, etc.) or scenario
factors empirical model
Economic level,
Element Disaster social Special disaster
combination
Scope Definition
abnormality bearing development, prediction, Comprehensive risk
type Of
Bayesian
analysis objects prevention preventive measures
Network
disaster- emergency
occurrence types management capability
inducing
time
Assessment
factor, major
factors anomalous
amplitude People Basic Special Comprehensive risk loss
identificat-ion Property R
duration disaster disaster
Ecology response response absolute number
Sensit- capacity capacity relative degree
ivity to Vd 1 Vd2
Disaster Disaster Classification disast-
intensity Gij Probability Pij statistics ers
Comprehensive judgment
Vs Disaster response
H Exposure Vd Classified risk loss
Ve Rij
Fuzzy Set theory
Risk degree assessment H Disaster bearing objects vulnerability(V) geographical Risk loss geographical
Target geographical distribution distribution distribution
Fig. 168.2 Assessment decision-making model of natural disaster vulnerability
(1) Disaster intensity (G) assessment. G was determined by the variability degree
of natural factors (such as the magnitude degree, the size of wind, temperature
or precipitation anomaly degree) or attribute index of natural disasters’
influence(such as seismic intensity, flood intensity).
(2) Disaster probability (P) assessment.
P was determined by the natural disaster occurrence number of the intensity in
a certain period, represented by probability or frequency (Fig. 168.2).
168.4.2 Vulnerability Assessment Index System
of Disaster Bearing Objects
Considering previous disasters and future trend, associated with social and eco-
nomic and disaster statistics system, vulnerability of disaster bearing objects was
assessed from physical exposure, sensitivity to disasters and socio-economic and
cultural disaster response capacity in favor of national, regional and community
development strategy and mitigation decision-making principles.
(1) Physical exposure (Ve) assessment. Ve indexes was divided into quantitative
and value types based on specific types and characteristics of disaster bearing
objects. The assessment process was:
Step 1: fix the minimum assessment unit.
Step 2: determine the number of disaster bearing objects in each minimum
assessment unit.
1598 J. Shen et al.
Step 3: set the influence sphere of disaster bearing objects.
Step 4: assess the physical exposure.
(2) Sensitivity to disasters (Vs) assessment. In view of the numerous disaster
bearing object types, disaster bearing objects were mainly divided into pop-
ulation, housing, crops, livestock, and the road systems for the assessment of
sensitivity to disasters.
(3) Regional disaster response capacity (Vd) assessment. Vd assessment consisted
of basic disaster response capacity index (human index, financial index and
material resources index) and special disaster response capacity index (disaster
prediction capability index, engineering disaster response capacity).
168.4.3 Assessment Index System of Regional Natural
Disaster Risk Loss Degree
(1) Risk loss assessment methods of single disaster. Adopting the analogy of
historical scenarios, physical simulation and experimental method and expert
scoring, disaster bearing objects were classified as demographic, economic
property and ecological systems to do population risk assessment, disaster risk
assessment of property loss and ecosystem loss degree assessment under
specific disaster.
(2) Risk loss assessment methods of Multi-disaster. On the basis of risk loss
assessment of single disaster, considering regional development, personal
safety and property security of residents, different natural disasters of different
power sources and characteristics were set in a regional system. The assess-
ment was divided into two levels. The first level was independent multi-
disaster risk loss assessment of the three types of disaster bearing objects
based on the assessment of the risk of loss considering risk loss and grade
assessment. The second level was integrated assessment of the three types of
bearing risk body based on the integration of assessment of the risk.
168.5 Conclusion
Certain disaster situation was formed through the interaction of disaster-inducing
factors, disaster environment and disaster bearing objects, and vulnerability was to
measure the disaster formation efficiency. In the thesis, the natural disaster vul-
nerability analysis method was put forward using vulnerability theory to analyze
the vulnerability mechanism of complex natural disaster systems with consider-
ation of comprehensive natural disasters. The proposed vulnerability assessment
system could enrich the urban disaster risk assessment system and disaster risk
management system, having important scientific significance in risk emergency
management with risk prevention as the core, city’s public safety, and the sus-
tainable development of cities.
168 Vulnerability Analysis and Assessment System of Natural Disaster 1599
Acknowledgments This paper is supported by National Natural Science Foundation of China
(Grant No. 71171143), Tianjin Research Program of Application Foundation and Advanced
Technology (Grant No. 10JCYBJC07300), Key Project of Science and Technology supporting
program in Tianjin (Grant No. 09ECKFGX00600), and FOXCONN Group.
References
Arnold M, Chen RS, Deichmann U (2006) Natural disaster hotspots case studies. Hazard
Management Unit, World Bank, Washington DC, pp 1–181
Blaikie P, Cannon T, Davis I (1994) People’s vulnerability and disasters. Nat Hazards, Routledge,
London, pp 189–19
Cutter SL (1996) Vulnerability to environmental hazards. Prog Hum Geogr 20:529–539
Gissing A, Blong R (2004) Accounting for variability in commercial flood damage estimation.
Aust Geogr 35(2):209–222
Greiving S (2006) Multi-risk assessment of Europe’s region. In: Birkmann J (ed) Measuring
vulnerability to hazards of national origin. UNU Press, Tokyo
Janssen M (2005) Scholarly network on resilience, vulnerability and adaptation with the human
dimensions of global environmental change. In: Hesse A et al (eds) Conference book for the
open meeting of the human dimensions of Global Environmental Change Research
Community, Bonn, Germany, October 2005. IHDP, pp 75–76
Shi Y, Xu S, Shi C, Sun A, Wang J (2009) A review on development of vulnerability assessment
of floods. Process Geogr 28(1):41–46
Chapter 169
Application of Actuarial Model
in the Forecast of Maternity Insurance
Fund’s Revenue and Expenditure:
A Case Study of Tianjin
Li-ping Fu, Jun Liu, Xue-zhen Chu and Jin-li Fan
Abstract To explore the way how to build up China’s urban and rural childbirth
insurance system, the crucial point is to carry out mutual helping function given by
the maternity insurance fund and make sure its sustainable use. Guided by the
principles and methods of demography and actuarial science, this paper forecasts
and calculates people who are insured by Tianjin employees’ maternity insurance
and urban and rural maternity insurance, and their fund of revenue and expendi-
ture, and draws relevant conclusions so as to provide scientific references for the
collection of Tianjin’s urban and rural unified maternity insurance fund, and for-
mulation of relative payment standards.
Keywords Balance of urban and rural Maternity insurance fund Forecast of
revenue and expenditure Actuarial science
169.1 Introduction
Basically speaking, China’s maternity insurance system is only for the employees’,
a large amount of rural women and urban non-professional women are out of the
coverage, which is beneath social justice. Therefore, in order to promote the
optimization and development of social security system, it is necessary to explore
L. Fu (&) J. Fan
Management and Economic Department, Public Resource Management Research Center,
Tianjin University, Tianjin, China
e-mail: [email protected]
J. Liu
Tianjin Health Insurance Research Association, Tianjin, China
X. Chu
Tianjin Municipal Human Resources and Social Security Bureau, Tianjin, China
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1601
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_169,
Springer-Verlag Berlin Heidelberg 2013
1602 L. Fu et al.
and build up unified maternity insurance system of urban and rural. The estab-
lishment of such unified system means the maternity insurance will cover not only
employed women, but also unemployed women, and in turn, maternity insurance
funds need to perform its mutual helping function. However, the revenue of
Tianjin employee’s maternity insurance funds is largely greater than its expendi-
ture since the year 2005. Due to too much balance of its accumulation and its
gradual annual increase, the efficiency of maternity insurance fund is quite low and
its function and effect has not been fully carried out. In a consequence, the analysis
and forecast of Tianjin maternity insurance fund’s revenue and expenditure will be
helpful to digest overmuch redundancy, which will benefit not only the rational
allocation of maternity security resources between different groups, but also the
realization of fairness between groups and the harmonious and stable development
of our society.
169.2 Actuarial Model
169.2.1 Model Construction of the Population and Insured
Group
(1) Population module (Fang and Sun 2008)
Basic data consists of regional total population and urban and rural population
structured by gender and age, the death rate of urban and rural population
differentiated by gender and age, the death rate of insured people differentiated
by gender and age and the birth rate of urban and rural differentiated by age.
Quantity of urban employed population:
Ax ¼ urban economically active population unemployed population
¼ urban economically active population ð1 bÞ ð169:1Þ
¼ working age population ð1564Þ a ð1 bÞ
According to above, Ax is the quantity of urban employed population; a is the labor
force participation rate, a = urban economically active population (both employed
and unemployed)/working age population, b is the registered urban unemployment
rate.
New birth population in t
Xn
Bt ¼ Lft;x ft;x ð169:2Þ
x¼i
In the above formula, Lft;x
is the quantity of women aged x in the year t; ft, x is the
birth rate of childbearing age women, Lit;x is the total amount of live birth by
women aged x in the year t.
169 Application of Actuarial Model 1603
As a result of Tianjin maternity insurance is limited by national family planning
policy, women who give birth under age 20 cannot enjoy the treatment, so i is
valued from 20,…, 49.
(2) Calculation module of insured people
This model not only forecasts the rural people, but also considers the differ-
ences between urban and rural from many aspects. It calculates insured
employees and retirees by the method of adding stock and incremental.
Urban employed population:
X
64
Ax ¼ Lx a ð1 bÞ ð169:3Þ
x¼15
Lx is the quantity of urban permanent resident population aged x (differentiated by
gender); a is the labor force participation rate; b is the urban registered
unemployment rate.
The quantity of employees:
Ex;t ¼ Ex1; t1 ð1 qxt e Þ þ ðAx;t Ex;t Þ d Ex1; t1 g ð169:4Þ
In the above formula, qxt e is the death rate of employees aged x in the year t
(differentiated by gender), 1 qxt e is the survival probability; d is the new insured
proportion, ex;0 is the number of new insured population aged x in base period, Ax;0
is the urban employed population aged x in base period, Ex;0 is the urban insured
employees aged x in base period, g is the ratio that employees switch to retirees,
n
REx;0 is the new added number of retirees aged x in base period, Ex;0 is the number
of employees aged x in base period.
The quantity of retirees:
REx;t ¼ REx1;t1 ð1 qxt RE Þ þ Ex1;t1 g ð169:5Þ
qxt RE is the death rate of retirees aged x in the year t (differentiated by gender),
1 qxt RE is the corresponding survival rate.
169.2.2 The Income Module
X
T X
65
Int ¼ Rð0Þ þ ðCRðtÞ ð Ex;t WAx;t jÞÞV t ð169:6Þ
t¼1 x¼15
In this module, R(0) is the surplus of fund in base period, CRðtÞ is the collection
rate in the year t, Ex;t is the average quantity of employees aged x in the year t,
WAx;t is the average pay cost wage of employees aged x in the year t, and j is the
rate of collection and payment. V t is the discount rate.
1604 L. Fu et al.
169.2.3 The Expenditure Module
X
T X
49 X
70
Ext ¼ ð Lft;x ft;x l0;x þ Lft;x jt;x o0;x ÞV t ð169:7Þ
t¼1 x¼20 x¼20
TPa10;x
l0;x ¼ Lf f
is the average birth medical expenses and allowance (quota) of
0;x 0;x
birth women aged x in base year; Lf0;x is the insured women aged x in base year; f0;x
is the birth rate of insured women aged x in base year; TPa10;x is the expenditure
fees of insured female employees aged x in base year; jt;x ¼ Tt;x =Lft;x is the family
planning level of women aged x in the year t; Tt;x is the total number of women
aged x who are under birth control; Lft;x is the quantity of insured women aged x in
TPa20;x
the year t; o0;x ¼ Lf j
is the average expenditure (quota) of family planning
0;x 0;x
women aged x in base year; j0;x is the family planning rate of insured women aged
x in base year; TPa20;x is the family planning expenditure fees of insured female
employees aged x in base year.
169.3 The Measurement and Parameter of Tianjin Urban
and Rural Unified Maternity Insurance Fund
169.3.1 The Coverage of Population
(1) Women of childbearing age
In 2010, the quantity of childbearing aged women who are birth insured
employees are 873,800, who are medically insured urban and rural residents are
1.2613 million. So the total amount is 2.1351 million, and over 446,500 women of
childbearing age are not insured (Tianjin Bureau of Statistics 2006–2011).
(2) The quantity of births
The quantity of births is 70,300 that are complied with Tianjin family planning
policy in 2010, among which 33,300 are urban residents, 37,000 are rural resi-
dents, and 16,100 are from other places (Tianjin Statistics Information 2010). In
recent years, the rate of family planning is between 98.28 % and 99.21 % in the
whole city.
(3) The quantity of birth insured people
In 2010, 28,300 women who are covered by employees’ maternity insurance
enjoyed the birth treatment and 20,500 childbearing women are subsidized by
169 Application of Actuarial Model 1605
urban and rural residents’ medical insurance, total 48,800. There are over 21,500
childbearing women who are not maternity insured (Tianjin Municipal Birth
Insurance System Documentation 2006–2011).
169.3.2 Fund Revenue
(1) The insured people
This number is calculated according to the employee’s maternity insurance,
the urban and rural medical insurance coverage range and every year’s growth
level. On the basis of 228 million insured employees in 2010, the employee’s
maternity insurance has increased gradually, with average annual growth rate
86,000. In 2010, the number of urban and rural maternity insured people is
2.2696 million according to non student adult residents (from age 20 to 60).
(2) Social insurance base
Maternity insurance is calculated in accordance with the employees’ age,
gender, the proportion of per capita base pay of last year’s social average
wage, and infer from the social average wage growth rate, then calculate as per
proportion between per capita base pay differentiated by age and gender, and
last year’s social average wage (Chi et al. 2009). Urban and rural residents’
maternity insurance is calculated as 12 Yuan per year per capita quota.
169.3.3 Fund Expenditure
(1) The number of Generational women determination
The staff maternity insurance is calculated by the summation and the change
of birth rate. Base on the summation of birth rate on the year of 2009, con-
cerning the progressive increase, 2–3 %; Urban and rural residents maternity
insurance is calculated by the summation of 2010, concerning the progressive
increase, 2–3 %; The number of Family planning of staff maternity insurance,
Urban and rural residents maternity insurance is determinate by the birth rate
on 2010, concerning the progressive increase, 3–6 %; The determination of
unemployed spouses of birth medical treatment accord to the difference of the
correlation data between male and female staff who are 20–49 years old,
referring to the group of 20–35 years old. According the female ages to cal-
culate the fertility number, considering the data coincidence to calculated the
fertility number and determinate as its 80 %.
(2) Per capita cost determination
Among staff maternity insurance, it includes family planning, antenatal care,
maternity medical expenses per capita cost by age, and then multiplied by the
corresponding number of treatment received. It calculates the average standard
neonatal care costs for 500 Yuan.
1606 L. Fu et al.
According to the requirement of the Social Insurance Act of China, maternity
benefits in accordance with the age of the maternity insurance of workers divided
by the number of days of maternity leave 30.4 for the month, and then it results by
multiplied the annual average pay wages for employees of enterprises to be.
Female staff maternity benefits calculated in accordance with currently approved
days of maternity leave (Social Insurance Act of China). Male benefit days cal-
culated in accordance with 7 days of maternity leave.
Subsidies for family planning are distributed according to a 66 % of abortion
and induced labor and an average 15 days for maternity leave. Based on the
analysis of the data from the past 6 years, the percentage of abortion and induced
labor in birth control operations in Tianjin is 45–59 % and the figure is showing a
growing momentum. So we adopted 66 % when calculating the subsidies.
The number of people who receive maternity insurance is calculated on the
basis of the birth rate of the total number of rural and urban residents who were
covered in medical care in 2010 and the payment is 80 % of the maternity
insurance of workers for people of all ages. In the charges for birth control, the
number is calculated as the number of people who are covered in medical care for
rural and urban residents multiplied by planed birth rate of corresponding worker
and the charges are the same as the staff maternity insurance fee.
169.4 Maternity Insurance Fund Forecast
According to Tianjin birth insurance database of basic data, and by using the
‘‘Tianjin statistical yearbook’’(Tianjin Bureau of Statistics 2006), ‘‘the fifth pop-
ulation census data of Tianjin’’ (Tianjin Statistics Bureau 2001) and China life
insurance industry experience life table (National Bureau of Statistics of China
Payment and Employment 2009–2011), use the fund of medical treatment insur-
ance actuarial analysis model ‘‘MIFA12, by forecasting the five years the popu-
lation growth rate (including the birth rate and mortality, net migration rate),
ginseng protect growth rates, wage growth, the total fertility rate, plan and other
important short-term level parameter value, and run outlook the birth insurance
fund operation prospect for the actuarial forecast and analysis from 2011 to 2015.’’
The feasibility assumptions anticipate reality, and take the neutral level. Model
parameters adjustment factor mainly use gradually recursion method and correc-
tion method in advance (Song 2009).
(1) Overall Balance
According to the present basic operation model calculation, since 2011, the
urban and rural birth insurance fund has been in overall revenues and expenditures
current balance level with accumulated balance continuously increased. The bal-
ance situation summary for Table 169.1.
169 Application of Actuarial Model 1607
Table 169.1 Urban and rural birth insurance fund balance
Years Fund income Fund expenditure Current Cumulative
balance balance
Staff Urban and Total Staff Urban and Total
childbirth rural amount childbirth rural amount
childbirth childbirth
2011 6.70 0.27 6.97 4.90 0.80 5.70 1.27 12.01
2012 8.03 0.28 8.31 6.62 0.82 7.44 0.87 12.88
2013 9.42 0.28 9.69 7.43 0.86 8.28 1.41 14.29
2014 10.88 0.28 11.15 8.16 0.89 9.05 2.11 16.40
2015 12.41 0.28 12.69 8.91 0.92 9.82 2.87 19.27
Unit Billion Yuan
Through Table 169.1, about the maternity insurance fund revenue part, from
2011 to 2015 total revenue increase $572 million, nearly increasing of 82.07 %,
with an average annual growth rate of 10.78 %, including staff maternity insurance
revenues grew 11.14 %, urban and rural maternity insurance revenues grew
0.62 %. In 2015, staff maternity insurance revenue share on behalf of 97.79 %,
urban–rural income accounting for 2.21 % of birth insurance.
On the side of outlay of Maternity insurance fund, the total outlay increases
average 9.94 % per year from 2011 to 2015. Including staff maternity insurance
fund increase average 11.06 %, town and country maternity insurance fund
increase average 2.37 % per year. The total outlay of fund in 2015 is 0.98 billion
dollars; staff maternity insurance occupies 90.73 %, town and country maternity
insurance occupies 9.27 %. The expenses and receipts, growth rate and the ratio of
staff maternity insurance fund exceed earning of town and country maternity
insurance fund to a large extent.
Because of the calculation, which is for the accrual basis of analysis and
forecast, the corresponding output is also accrual, so that The Fund income and
expenditure does not correspond exactly to the current cash under the system of
accounting and statistics reports. Appropriate predictions should be reflected
through the accounting statements of the current mode after 1–2 years.
(2) Maternity Insurance Fund Expenditures
Each year the staff maternity insurance amount is summarized as shown in
Table 169.2.
Through Table 169.2, during 2011–2015, the payment amount annual increase
for 13.8 %. In 2015, the staff maternity insurance cost 891 million Yuan, paid from
the main categories it’s including antenatal check fee of $34 million, accounting
for 3.82 %; Growing medical expenditure of $126 million, accounting for
14.14 %; Maternity allowance $601 million, accounting for 67.45 %; Male staff
maternity allowance of $37 million, 4.15 %; Family planning allowance of $58
million, 6.51 %; all above five accounted for 96.07 %.
The urban and rural birth insurance payments case summary in Table 169.3.
1608
Table 169.2 Staff maternity insurance fund expenditures in 2001–2015
Years Subtotal Antenatal Birth medical Birth Neonatal Male staff No spouse Family Family planning
examination cost treatment cost allowance care allowance employment growth planning allowance
2011 4.90 0.30 1.06 2.90 0.17 0.14 0.05 0.08 0.18
2012 6.62 0.32 1.18 4.18 0.18 0.26 0.05 0.09 0.37
2013 7.43 0.33 1.21 4.82 0.18 0.30 0.05 0.09 0.44
2014 8.16 0.33 1.24 5.40 0.19 0.34 0.05 0.10 0.51
2015 8.91 0.34 1.26 6.01 0.19 0.37 0.05 0.10 0.58
Unit Billion Yuan
L. Fu et al.
169 Application of Actuarial Model 1609
Table 169.3 The childbirth grant expenditure of urban and rural medical insurance in
2011–2015
Years Subtotal Prenatal Childbirth medical care Family Newborn care
examination planning
2011 0.80 0.12 0.54 0.07 0.07
2012 0.82 0.13 0.54 0.06 0.08
2013 0.86 0.14 0.57 0.07 0.09
2014 0.89 0.14 0.59 0.07 0.09
2015 0.92 0.15 0.61 0.07 0.09
Unit Billion Yuan
According to the Table 169.3, the expenditure of urban and rural childbirth
insurance has an annual average increase of 2.37 % in 2011–2015. By 2015, the
expenditure of urban and rural childbirth insurance will increase to 92 million
Yuan. Within this total prenatal examination expenditure as proportion of 16.30 %
is 14 million; Childbirth Medical Care as proportion of 66.30 % is 59 million;
Family Planning as proportion of 7.62 % is 7 million; Newborn Care as proportion
of 9.78 % is 9 million.
169.5 Conclusion
(1) The rate and balance of payments can fund in this paper and imposed or paid
based on the prediction of fund income. From 2011 to 2015, from the absolute
value to see current balance, accumulated balance continuously increased,
could realize the sustainable usage; From the relative amounts to see, birth
insurance fund income annual growth rate is 10.78 %, birth insurance fund
total spending the average annual growth of 9.94 %, after that plan, birth
insurance fund basically comply with ‘‘payment to fix the receive, balance the
basic balance’’ principle. Under this principle, deal with the fund collection
and use of scientific investigation and measurement, the determination of
reasonable seized proportion, in principle should be controlled in the pro-
portion between 0.6 % and 0.7 %, which can steadily absorb balance fund.
(2) The worker bears insurance fund collection-payment, growth rates and that
accounts for a fund income is far greater than the proportion of urban and rural
birth insurance fund income. The worker bears insurance fund income growth
rate is flat and the urban and rural birth insurance fund spending growth rate
than income growth, after that plan as a whole, urban and rural insurance fund
spending part can by as a whole fund after agreed to pay, to reflect fund each
other function.
(3) During 2011–2015, the workers’ childbirth insurance amount paid the average
annual growth of 13.8 %, the proportion of the birth grant is bigger; the urban
and rural childbirth insurance amount paid the average annual growth of
1610 L. Fu et al.
2.37 %, the proportion of birth medical treatment is larger. So it can improve
birth insurance treatment by the payment standard to effectively mediation
fund balance rate.
Project Subject: Tianjin medical insurance research board project; A research on
the issue of building up unified urban and rural maternity insurance system.
References
Chi B, Chen Z-j, Liu X-p (2009) Birth insurance policy review and audit of treatment to pay
concerned. Tianjin Social Insurance, Tianjin, pp 43–47 (Chinese)
Fang J-q, Sun Z-q (2008) The health statistics. People’s Medical Publishing House, Beijing
(Chinese)
National Bureau of Statistics of China Payment and Employment (2009–2011) Labor wage and
employment China monthly economic indicators (Chinese)
Social Insurance Act of China http://www.china.com.cn/policy/txt/2010-10/29/content_
21225907.htm
Song S-b (2009) China’s medical security system debt risk assessment and sustainability.
Economics & Management Publishing House, Beijing (Chinese)
Statistics Bulletin of the National Economic and Social Development of Tianjin Municipal in
2010. Tianjin Statistics Information (Chinese) http://www.stats-tj.gov.cn/Article/ShowClass.
asp?ClassID=44
Tianjin Bureau of Statistics (2006–2011) Tianjin statistical yearbook. China Statistics Publishing
House, Beijing (Chinese)
Tianjin Municipal Birth Insurance System Documentation (2006–2011) Tianjin Municipal
Human Resources and Social Security Bureau (Chinese)
Tianjin Statistics Bureau (2001) The Fifth Population Census Data of Tianjin
Chapter 170
Study on Process Reengineering
of Medium and Small Coal Machine
Manufacture Enterprises
Jing-wen An and Zhi-qiang Zhang
Abstract Based on the theory and method of process reengineering, this paper
implemented process engineering on JY company which is a coal machine man-
ufacture enterprise. On the basis of analysis, diagnosis and optimization on
existing process, process system and organizational structure were reengineered,
and related management system were establish. That shows that BPR is an
important way for medium and small coal machine manufacture enterprises to
standardize enterprise management, enhance organization and coordination flexi-
ble, promote enterprise competition ability.
Keywords Medium and small enterprises Coal machine manufacture enterprise
Process reengineering
170.1 Introduction
Since the 1980s, with the rapid development of the world economy and technol-
ogy, the uncertainty of the enterprise survival environment is increasing, the
competition which enterprises are facing is also becoming increasingly fierce,
which mainly reflected in the competition of variety, quality, price, time and
service. Only the one who has advantages in these five respects can survive and
develop. Enterprises used a lot of advanced management methods and manufac-
turing technology, and the comprehensive using of these technology and methods,
indeed, has improved and enhanced the enterprises’ competitiveness (Yue 2005).
However, among them, process reengineering is the most effective method to
improve the competitiveness of the enterprise in terms of the point of strategy.
J. An Z. Zhang (&)
School of Management, China University of Mining
and Technology, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1611
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_170,
Ó Springer-Verlag Berlin Heidelberg 2013
1612 J. An and Z. Zhang
Business Process Reengineering (BPR) in developed country are widely
applied, some big companies such as IBM, HP, Siemens, and other enterprises
implemented BPR, have made outstanding achievements, and business regards it
as an important strategy to achieve competitive advantage as an industrial man-
agement revolution. In our country, the implementation of BPR also has successful
cases, such as the Haier group to apply internal market chain management in the
whole group via BPR, establish business flow, logistics and the product, which is
the successful model of BPR implementation (Dong 2008). The small and med-
ium-sized enterprises are in small scale, simple organizational structure, less
management level, easier communication and simple external environment, these
features are conducive to the implementation of process reengineering.
By using BPR thoughts and methods on JY company as an example, this paper
states that the enterprises especially the small and medium sized enterprises whose
standardization management is weak and the organization flexibility is poor, after
the systematical analysis and careful arrangement and recreate process, can
effectively standardize enterprise management, enhance the organization and
coordination flexibility and promote enterprise competition ability.
170.2 Process Reengineering Theory
170.2.1 The Concept of Process Reengineering
After 1990s, 3C (Custom, Competition, Chang) have made the uncertainty of
market demand increased greatly, further more the enterprises are facing a rapid
changing and unpredictable buyer’s market, the traditional production and man-
agement pattern already unable to react to the market, it is in this circumstance, in
the beginning of 1990s, American Dr. Hamor proposed management thinking
‘‘business process reengineering’’, and then in the United States and other
industrialized countries, made the core management revolution of the ‘‘business
process reengineering’’ (Yue 2005).
Ha defined process reengineering as:’’ the fundamentally rethink and com-
pletely redesign of enterprise business process in order to improve enterprise cost,
quality, service and speed significantly.’’ BPR emphasis on improving process and
customer needs and satisfaction, the use of advanced information technology,
manufacturing technology and modern management means, maximizing the
technical function integration and management function integration, to break the
traditional functional organization structure, building a new process of organiza-
tional structure, so as to realize the improvement of enterprise in cost, quality,
service, speed, and other aspects and enhance the enterprise market reaction rate
and the market adapt greatly (Dong 2008).
170 Study on Process Reengineering of Medium and Small Coal 1613
Strategic Project Process New Process New Process Continuous
Vision Initiation Diagnosis Design Implementation Improvement
Design new processes, Continuous
BuiIdvision, Express existing Staff training,
Set up the project team, Design new improvement Of New
Selecting key process, processes, NEw salary system,
Project planning, organizational structure, process,
Ready tochange Analyze existing New IT arrangement
Identifythe target Design new human Restructure other
organization processes
resources processes
Fig. 170.1 Model figure of process reengineering implementation
170.2.2 Process Reengineering Steps
Successful implementation of enterprise process engineering and establishing
implementation plan is very importantly, the implementation of the plan is the
strategic planning implementation of process reengineering. Different scholars’
summarization of BPR implementation stage and stage classification are different,
but the general ideas are basically the same. The key is to grasp the principle and
contents of various stages (Qi and Wang 2005). The model figure of process
reengineering implementation (Wang 2005) (see Fig. 170.1).
170.2.3 Ways of Process Reengineering
170.2.3.1 New Reengineering Method
This method makes us to rethink the way of product and service providing
designing process in a white paper fundamentally. The new method will start from
the target, and gradually pour push, process should be designed to reach the
requirements of the organization. This method is profound, dramatic, high risky
and has big resistance, and may bring huge cost if the reform fails (Zeng 2008).
170.2.3.2 Gradual Transformation Method
Through the systematic analysis of the existing process, this method is to create
new process based on the existing process. This method is efficient, advance step
by step, and has a lower risk of resistance and smaller interference to the normal
operation. Many big companies at home or abroad regard continuous improvement
as important part of their enterprise culture, through the work of the hundreds of
thousands of small changes, the huge performance improvement can be gradually
accumulated.
1614 J. An and Z. Zhang
170.3 JY Company Condition
170.3.1 Company Introduction
JY company was founded in 1960s, after 30 years of development, the efforts of
several generations. JY company has owned total assets of nearly forty million
yuan, 20 mu of land, nearly one hundred sets equipment and 280 workers, and JY
company is one of professional production manufacturer of China’s coal mine
integrated production mechanization equipment.
Company mainly manages manufacture, technical development, technically
services of industrial electrical and mechanical equipment. Operation mode is
manufacture and sale primarily. The main products of JY company are mining
emulsion pump and its accessories, spary pump station and its accessories, etc.
170.3.2 Present Situation and Problem Analysis
JY company predecessor is a branch factory of one machinery company, mainly
supplies mainly corporation with all kinds of accessories of hydraulic and inde-
pendent product such as emulsion pump and spray pump. After decades of con-
struction and development, facilities, product level, personnel quality have
improved greatly. However, the head office management system reform had fin-
ished, the branch factory restructured into a company in 2008, which brought this
factory with the opportunities and challenges. At the background of the financial
crisis and the integration of the coal mine in recent years, the pump industry
competition is intense, and the weak basic management all lead to poor product
quality, slow weak customer demand response, which had made the company lost
a large market share in crisis edge.
Through the methods of field inspection, related personnel’s communication,
special problem research and material consultation, the following key problems
were found out.
170.3.2.1 Organization Structure Aspect
JY Company used strict linear function structure under original workshop man-
agement mode, Which had not adapt to the need greatly to increase production
flexibility, speed up the external market response and strengthen internal man-
agement; Departments responsibility divided was unclear, nonstandard and
imbalance, lacked of effective communication cooperation mechanism between
departments, and organizational operation efficiency is not high. In the end
responsibilities and interests were unequal, and the management ranges were too
big, which lead to low management efficiency.
170 Study on Process Reengineering of Medium and Small Coal 1615
170.3.2.2 Management Process Aspect
The responsibilities of process units were not clear, cooperation degree is not high,
and the processes were lack of flexibility, standardization, systematization. The
design and implement of Management process all existed problems, and the setting
of department and key position were not reasonable which all led some management
process missing or fuzzy, and some processes often appear as ‘‘short circuit’’ in the
implementation process, so the phenomenon of dispute over trifles and shuffle arose.
170.3.2.3 Management Aspect
Enterprise basic management is weak, because the existing extensive management
mode seriously restricted the expansion of enterprise production capacity and the
enhancement of benefit; reform task was Serious, because the company equity
structure changed, and the company’s leadership adjusted hugely, and the reform
referred to benefit pattern adjustment. So this reform will affect stability and
spread to development if carelessly a bit.
170.3.2.4 Other Aspects
Ability crisis were increasingly serious, because of the company’s development
and the change of external environment, equipment ability, management level,
workshop area, process layout and technical strength, human resources, enterprise
culture and so on all would not meet the demands of the new situation. And there
is a big gap among many domestic industry pump companies on the capital
strength technical level, management experience and product quality.
From the above analysis, we can conclude that current business process and
organization structure of JY company had been difficult to adapt to the new market
competition environment needs. Faced with such serious situation, if not decisively
reform, JY company might face survival crisis soon. Although the risk of internal
management changes was high, but in the present situation, the risks of no refor-
mation would be greater than the risks of implementing reformation. Therefore, the
implementation of the new process reengineering was at a better time.
170.4 Implementation of Process Reengineering
on JY Company
170.4.1 Goals and Principles of Process Reengineering
On the basis of the analysis of the internal and external market environment and
according to the company management status, strategic objectives and reengi-
neering ideas, the objective of process reengineering was made sure to construct
1616 J. An and Z. Zhang
‘‘smooth process, efficient organization, hard working personnel’’ process man-
agement system, ensure the strategy implementation of JY company successfully.
At the same time to ensure the smooth implementation of the process reengi-
neering, four core principles were established: with process as the core principle,
people oriented principle, customer oriented principle, quality first principle.
170.4.2 Implementation of Process Reengineering
170.4.2.1 Establishment of Framework
With the best process management practice and theory, and considering the
practical situation, JY company’s overall processes were divided into two classes:
business process and management process. According to the feature of specific
production and functional management, the first process framework of JY com-
pany was formed (see Fig. 170.2).
Meanwhile, on the basis of the framework, the second and the third process
were set up, and nine first processes and fifty-six second processes and third
processes were set up preliminary (Electricity Group Co. 2008).
170.4.2.2 Determination and Optimization of Key Processes
Through the special conference and the matrix analysis method, the determined
key processes include six key processes: production management, quality man-
agement, financial management and so on. On the basis of full understanding and
analysis to the key process, we should find and research the sickness of existing
process, and then to design process (Hui et al. 2000).
Take the process of production optimization for example.
170.4.2.3 Design Plan of Organization
Business process reengineering request to establish process firstly, and then
according to the new organizational structure of flat process to streamline
Customers Product design
Business Delivery and
exploitation and and Procurement Production
process maintaining
after sales
development
Smooth process,
Operations management Efficient
organizations,
Managem Human resource management Capable
ent personnel
process Quality management
Financial Management
Fig. 170.2 The first process framework of JY company
170 Study on Process Reengineering of Medium and Small Coal 1617
management level, reduce the management cost and delegate management, and the
new decision point is located in the business process execution place.
According to the analysis of the status and combined with external best prac-
tices and the reality of JY company, firstly, the organization frame structure design
carried through on the basis of framework system optimization of JY company and
according to ‘‘streamline organization, personnel optimization’’ principle. Sec-
ondly, the organization department functions boundary was determined and
department responsibility was written. Then the fixation of posts and staff was
determined based on the design of organizational structure. Finally, key position
introductions were established (Figs. 170.3, 170.4, 170.5 and 170.6).
Any reform and innovation of management must carry out on system level,
which is an important principle to get rid of the rule by men for modern enterprise,
especially modern china enterprise, so process reengineering as an important
management innovation is no exception. At the same time process reengineering is
an system engineering, need each aspect provide guarantee for this to set up
process management, evaluation, compensation and other enterprise management
system, to effectively ensure the smooth implementation of the process
reengineering.
170.5 Effect
From may 2010 the beginning of carrying out the above process reform plan to
may 2011, through the continuous reform optimization, temporary workers reduce
by 30 %; Organization operation efficiency and production efficiency increase
obviously, pump production cycle reduces seven days; Product quality and cus-
tomer satisfaction improve significantly. Due to the implementation of the new
assessment method and salary system, the worker enthusiasm is remarkably
improved, labor productivity increases by 20 %, the per capita wage of worker
increases by 16 %, annual output value increases by 21 % and profits growth rate
is up to 28 %. These show that the effect of BPR on JY company is obvious.
170.6 Conclusion
This paper systematically analyzed the management status of the JY company and
existing problems, used the method of combining theory with practice, and put
forward the implementing method of JY company process reengineering. Through
the study, the conclusions are as follows:
(1) The right process reengineering can make the enterprise operation efficiency
and economic benefit, product and service quality and customer satisfaction
increase hugely. Promoting process reengineering on the small and medium
1618 J. An and Z. Zhang
Responsible Department Product Department Process Name Production
Levels Process Numbers
Outsourcing
Department Product Department Quality Department Sales Department
manufacturers
Node A B C D
Beginning
contract
1 Batch purchase Release purchase
of raw material notice
Whether to have
2 production conditions
3
Release mass
production notice
Prepare
4 production plan
Arrange the
5 outsourcing
manufacturers
Release
6
production plan
Organize
7
Production
Production process Include process
8 management/Outsourcing inspection
manufacturers management
Product check list
product
9
inspection
Unqualified
If they are No product disposal
qualified process
Product
10 Yes
storage
Delivery
11
process
Creates Unit JY Company Signer Date
Fig. 170.3 Original production management process
170 Study on Process Reengineering of Medium and Small Coal 1619
Responsible
Department
Product Department Process Name Production
Levels One Level Process Process Numbers
Product Outsourcing Quality Corporate
Department
Department manufacturers Department Management Sales Department
Department
Note A B C D E
Beginning
Consulate with relevant departments
to determine
Release Release mass
1
purchase production
notice notice
2
Whether to have
production
conditions Random parts list Sales Supply
Production
plan
Prepare Release mass
3 production production
plan notice
Outsourcing review list (Manufacturer and
contract assessment)
If there is Arrange Review
4 outsourcing Yes outsourcing record
arrangements ?
Review
No
Release
production No
5
plan
6 Organize Yes Whether to
production pass?
Include process
Production process
inspection
7 management/Outsourcing
manufacturers management Product check
list
8 product
inspection
If they are No Unqualified product
qualified disposal process
Product
9 storage, Yes
ERP
operation
Delivery
10 process
Creates Unit JY Company Signer Date
Fig. 170.4 Optimized production management process
1620 J. An and Z. Zhang
Material feeding Finishedparts
workshop storage
Production Standard part
M echanical
department storage
processing
m anager
workshops
Tool storage
Production
Finance Planning
department Consumable
department
storage
General General Assembly outsourcing
manager Shop products
Sales
incoming
department
Equipment
Department
Technical Technical
m anager department
Quality
laboratory
department
Labour
union
Fig. 170.5 Original organization structure
sized enterprises similar to JY company is the need for small and medium
sized enterprise to change development mode, realize leaping development,
increase enterprise flexible, improve economic and social benefits and realize
strategic objective.
(2) BPR so far is just a thought, but not be called a theory. Because as a kind of
innovation theory, BPR is far from mature, internal mechanism of BPR and
deep understanding of the essential rule are far from established. And
advanced thinking and theory are not enough to bring the success to practice.
Not perfect method system and lack analysis tools are all obstacles factors to
effective BPR in practice. Therefore enterprises in practical application avoid
blind imitation by all means, and should combine BPR with IE and other
management methods, only that can we guarantee the success of the process
reengineering (Hanuner and ChamPy 1993).
170 Study on Process Reengineering of Medium and Small Coal 1621
Structure
General section
office
Machining
Enterprise section
management Assembly
department section
secretary semi product
Dispatching
Labor union center warehouse
Tool storage
Product
Production department Material
President
vice manager storage
Supply Standard
department partts Storage
Sales vice Sales Finishedparts
manager department storage
general
manager Chief Financial laboratory
accountant department
Gage
storage
Quality
Quality vice department Inspection
manager group
Equipment
security Design group
department
Technology
group
Chief Technical
engineer center Reproduction
room
Electrical
group
Fig. 170.6 Optimized organization structure
(3) The practice of JY company business process reengineering proved that the
method of this paper for enterprise process reengineering has certain directive
significance, and can reduce mistakes, improve the efficiency, ensure smooth
completion of the business process reengineering (Mei and Teng 2004; Huang
and Mei 2003).
References
Dong W (2008) Study on process reengineering of production management on HC machinery
plant. Nanjing University of Aeronautics and Astronautics, Nanjing
Hanuner M, ChamPy J (1993) Reengineering the corporation a manifesto for business revolution.
Harper Collins Publishers Inc, New York
Huang A, Mei S (2003) Process management principle and the remarkable process modeling
study. Ind Eng Manage J 2:46–50
Hui F, Zhang T, Sun L (2000) Analysis on the key success factors of BPR practice. Ind Eng J
1(1):10–13
1622 J. An and Z. Zhang
Mei S, Teng JTC (2004) Process reengineering—theory, method and technology. Tsinghua
University Press, Beijing
Qi E, Wang H (2005) Business process reengineering based on value chain. Ind Eng J 8(1):77–79
Wang P (2005) Process reengineering. CITIC Publishing House, Beijing
Wanbei Coal and Electricity Group Co., LTD (2008) New mode of coal mine management
Yue J (2005) Reshaping the production process is an effective way to improve the competition of
enterprises. J Inner Mongolla Finan Econ Coll 3:77–79
Zeng W (2008) Study on production process management mode optimization in Renhe casting
factory, Lanzhou University, Lanzhou
Chapter 171
The Increase and Decrease Connecting
Potential Analysis on Urban and Rural
Residential Land of Tianjin
Guo-bin Lin and Shuo Hao
Abstract Research purpose: probe into the technical routes and methods of CUR
(connecting the increase of urban construction with the decrease of rural resi-
dential land) potential calculation. Research method: quantitative analysis method.
Results: according to this calculation method and technical route, the empirical
analysis on CUR potential of Tianjin, the CUR potential coefficient is 1.25 and the
CUR potential balance is 4936.60 hm2 in as yearly planning goals. This states that
Tianjin can meet the demand of new town construction land occupying plough in
target planned years through CUR performance. Research conclusions: to find the
technical route and measure and calculated methods of CUR potential based on
overall plan of land utilization its result would reflect the actual area, tightly
integrating with land use control, be useful data references for other places
applying for CUR experimental unit, distributing CUR quota, developing CUR
items, lay the foundation for working out land reclamation planning and provide
quantitative data and references for land and resources management departments
to develop, innovation CUR policy.
Keywords CUR Calculation methods Technical route Potential calculation
171.1 Introduction
In June of 2008, the Ministry of land and resources issued ‘‘connecting the increase
in Urban construction with the decrease in Rural residential land management
approach’’, and it symbolized that our country CUR (connecting the increase of
G. Lin
School of Management and Economics, Tianjin University, Tianjin, China
S. Hao (&)
The Postgraduate Department, Tianjin Polytechnic University, Tianjin, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1623
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_171,
Ó Springer-Verlag Berlin Heidelberg 2013
1624 G. Lin and S. Hao
urban construction with the decrease of rural residential land) pilot work which
was formally incorporated into the course that runs lawfully. With the further
development of the CUR, CUR with the Chinese characteristic of land use in
China has gradually become the research focus in the field of land management
and preliminary already formed with CUR policy interpretation (Shao and Li
2009; Feng et al. 2011; Li and Wang 2009; Gong 2012), CUR pattern design
(Wang and Zhu 2007; Li et al. 2007; Qv et al. 2011; Wang and Wang 2009), CUR
benefit evaluation (Mai 2008; Gan and Zhou 2008; Yuan 2011), CUR potential
analysis (Xu and Wang 2009; (Yu 2011) and so on as the core content of the
theoretical system. However the CUR potential and space layout analysis research
is few. The few studies are mostly lack of potential calculation of system thinking,
ignoring the potential to link with the general plans for land use and the calculation
results with the overall land use planning are out of line, bad practicality. In the
future, based on the overall land use planning of CUR potential calculation will be
one of the most important research directions. It can provide useful data references
for other places applying for CUR experimental unit, distributing CUR quota,
developing CUR items, lay the foundation for working out land reclamation
planning and provide quantitative data and references for land and resources
management departments to develop, innovate CUR policy.
171.2 The Theoretical Foundation of CUR
China’s CUR policy was introduced when urbanization and industrialization were
advanced ceaselessly, the construction of new socialist countryside is just unfol-
ded. The core of this policy includes two aspects: The former is mainly for the
expansion of the city and the latter, in the present stage of our country, is mainly
through consolidation rural residential land to achieve. The CUR potential analysis
is theoretical and practical work and must have corresponding theoretical basis
guiding to make it more scientific, forward-looking and practical. This paper
argues that related theoretical basis mainly includes land use planning theory,
sustainable development theory, location theory, and rent theory.
171.2.1 Land Use Planning Theory
Land-use planning is ahead of schedule arrangement for a certain area of land use
in the future, which is land resource allocation and reasonable organization of
land-use comprehensive technical and economic measures in time and space on
the basis of the regional social economics development and land natural history
characteristics (Wang and Han 2002). The land use planning on the one hand as
science and technology is the productive force to provide support for human
reasonable and orderly use for land, on the other hand as the land in the sectors of
171 The Increase and Decrease Connecting Potential 1625
the national economy in the means of distribution industries, belonging to the
category of production relation. In CUR, the rural residential land consolidation is
the one involving engineering technique, management, administrative, economic,
and legal and other fields, and complicated system engineering, in which the
establishment of rural residential land consolidation planning is one of the most
important contents. In planning the preparation of the process, land planning plays
an important guiding role of theory. Its preparation should include the following
aspects: target; current analysis; feasibility analysis and evaluation; selection
model; preparation and evaluation of different alternatives depend on the economic
and social, environmental analysis; choosing satisfactory solution; drawing and
writing program.
171.2.2 Sustainable Development Theory
Sustainable development not only satisfy the demand answers and not harm future
generations demand, but also conforms to the local population interests and to the
interests of the global population. Here mainly includes the following means:
(1) Efficient development. Not only refer to the efficiency of economic sense, but
also contains natural resources and environment in the profits and losses of the
composition.
(2) Sustainable development. The economic and social development cannot
exceed the carrying capacity of the environment and resource, and we must
make the natural resource consumption rate be lower than the rate of regen-
eration resources.
(3) Equitable development. Contains the generation of transverse intergenera-
tional equity and intergenerational vertical equity, and people living in con-
temporary word cannot damage offspring’s survival and development
conditions due to their own development needs.
(4) Common development. The Earth is a complex giant system, between various
subsystems are interaction, as long as a sub system happen problems, will
directly or indirectly affect other subsystems and influences the whole system
function, therefore, the sustainable development is common development.
Coordinating the contradiction between supply and demand of land is the
eternal theme of sustainable utilization of land resources. In the CUR work, we
must always adhere to the concept of sustainable development. That is to say the
work must ensure the use of land on the basis of the ecological security and social
stability and ensure the land ecological environment allows the limits of land
consolidation.
1626 G. Lin and S. Hao
171.2.3 Location Theory
Location theory is about the place of human economics activities and the theory of
space economic ties. The positions of the spatial distribution of Social economic
activity include geographical location, economic position and traffic position.
These positions connecting organically work together to regional space, forming
the superiority of difference of land location. Land is the places of all human
activities. Different human activities can produce different types of land using. Plot
has not only azimuth and distance attributes, but also social economic activities
and the spatial distribution law which closely ties to geographical elements.
About the economic benefit of the land, the influence of the location factor
mainly is embodied in the following respects:
(1) Accessibility. Locations with good accessibility can enter the location strongly.
(2) The distance from the central of business district. The closer from the central
business district, the location is better and the land using efficiency is higher.
(3) Materialized labor inputs. The more social materialized labor is used, the
value of land using is greater and the economic benefit is much higher.
(4) Agglomeration benefits and its complementary with each other. Cluster can
make the enterprise to get comprehensive benefit. When multiple or related
enterprises get together, it forming the mutual complement of organic whole,
they can get more profit than scattered arrangement.
The location of the project is the embodiment and application of the location
theory. On the one hand, the rural residential areas in the CUR turns into cultivated
land again and the cultivated land for building a new area turns into urban con-
struction purposes, realizing the change of the space position of the land use,
making the land use become more reasonable; On the other hand, for the choosing
of the new site, also need the region whose location is in good, and can gain
maximum benefit.
171.2.4 Rent Theory
Differential rent theory provides theory basis for the analysis of urban and rural
construction land increase or decrease the peg operation mechanism. Differential
rent theory is the part of the excess profit a rebound by the operating better land
belonging to the land owners. The difference of land’s natural conditions and the
monopoly to the right of land using combine differential rent. In accordance with
the formation condition, the differential rent can be divided into differential rent I
and differential rent II. Differential rent I produced in different land fertile degree
and geographical location and differential rent II because in the same block con-
tinuous investment leads to higher labor productivity. Differential rent I is the basis
and the premise of differential rent II.
171 The Increase and Decrease Connecting Potential 1627
In the CUR work, one of key jobs is to set up the reasonable corresponding
relation between new build (CUR demand area) and removed the old district (CUR
supply area). Among them, the CUR supply areas mainly develop consolidation of
rural construction land potential, provide town construction land index, and the
CUR demand as key areas of urban construction. In theory, differential rent lower
land unsuitable for construction, so we should reclaim the differential land rent
lower local rural construction land as plough and high places of differential rent for
the priority of the arrangement for the new building.
171.3 The CUR Potential Measure Technical Route
CUR potential is to calculate possibility and the size of a regional implementation
CUR policy, according to the prediction of the target planed year demand of urban
construction land and land consolidation potential of rural residential area. Mea-
suring CUR potential is the foundation of developed for CUR plan, is the basis of
the CUR index distribution. In this paper, we used CUR potential coefficient and
CUR potential balance to represent the size of the CUR potential. In numerical, the
CUR potential coefficient is equal to the ratio between CUR supply and the demand
in target years, as followed (171.1). The CUR potential balance is equal to the
difference between CUR supply and demand in target years, as followed (171.2).
r ¼ Sg =Dg ð171:1Þ
Qy ¼ Sg Dg ð171:2Þ
In (171.1) and (171.2): r is CUR potential coefficient, Qy is CUR potential
balance, Sg is CUR supply ability in target years and Dg is CUR demand in target
years.
Mainly consists of the following three steps to calculate the CUR potential:
Step 1: Calculate urban construction land CUR demand based on the urban con-
struction land utilization and planning in target years.
Step 2: Calculate consolidation of rural residential land CUR supply ability based
on the rural dweller dot utilization and planning in target years.
Step 3: Calculate the CUR potential.
171.3.1 Urban Construction Land CUR Demand Calculation
The CUR demand is depending on the urban construction land demand through the
CUR model to meet based on all levels general land use planning scheme and the
regional economic social development. In this paper the calculation of CUR
demand is as followed (171.3):
1628 G. Lin and S. Hao
Dg ¼ Dz Dx ð171:3Þ
In (171.3), Dg is CUR demand, Dz is additional construction land quota
occupied farmland index in target years, and Dx is new construction land occu-
pation of cultivated land control indexes which determined by ‘‘The overall land
use planning (2006–2020)
171.3.2 The Calculation of Consolidation of Rural
Residential Land CUR Supply Ability
The consolidation of rural residential land CUR supply ability is that in various
practical constraints, the capacity of consolidation rural residential land providing
land to CUR work based on the overall planning of land use. The calculation of
CUR demand is as followed (171.4):
Sg ¼ a Sq ð Sx Sh Þ ð171:4Þ
In (171.4): Sg is the ability of CUR supply, Sq is rural residential land read-
justment potential, Sx is Rural residential area of current situation, Sh is rural
residential planning area and a is newly increased cultivated land coefficient.
The concrete calculation process is as follows:
Step 1: Calculation of rural residential land readjustment potential, based on the
rural residents utilization.
Step 2: Calculation of rural residential planning in target years CUR supply ability
based on the land use of rural residential area planning.
171.4 Empirical Analysis
171.4.1 The General Situation of Research Area
Tianjin is the biggest coastal open city in North China, within the eastern Eurasian
Continental Bridge bridgehead, and is located in the northeast of the north China
plain, the Bohai economic center, which are good location conditions. At the end
of 2008, the city’s population was 9,688,700, of which agricultural population was
3,806,000 and non agricultural population was 5,882,700. According to Tianjin
2008 current land-use change survey results, the city’s land area was
1,191,731.9 hm2. Among that, agricultural land area was 692,670.95 hm2,
accounting for 58.12 % of the total land and plough area was 441,089.72 hm2
among agricultural land, representing the city’s total land area of 37.01 %;
The total area of land for construction was 368,188.81 hm2, accounting for
171 The Increase and Decrease Connecting Potential 1629
30.90 % of the total land area, in which rural residential land was 88,192.45 hm2,
representing the city’s total land area of 4.70 %; The size of unused land was
130,872.15 hm2, accounting for 10.98 % of the total land area and mainly in Baodi
District, Ninghe County, Dagang District, Wuqing District, Jinghai County and
other places.
171.4.2 Tianjin Town Construction Land CUR Demand
Estimates
CUR demand in Tianjin district and county is as shown in Table 171.1, according
to the formula (171.3), ‘‘Tianjin city land uses overall planning (2006–2020)’’ and
the second land survey data of Tianjin city. We can see from it that in target planed
years, Tianjin by CUR to meet the need of the town construction land of cultivated
land occupied index will be 19,905.19 hm2, CUR needs at most will be Wuqing
District, as high as 3,454.1 hm2, accounting for 17.35 % of all CUR demand,
followed by Jinghai County will be 3,295.13 hm2 accounting for 16.55 % of all
CUR demand, CUR needs the least will be Binhai New Area, only 515.28 hm2,
accounting for 2.59 % of all CUR demand.
Tianjin Binhai New Area should pay future efforts to become North China
portal opening to the outside world, a high level of modern manufacturing and
research conversion base, the northern international shipping center and the
international logistics center, and gradually become the economic prosperity,
social harmony, environmental beautifully and ecological livable city. Therefore,
the ‘‘Tianjin city land uses overall planning (2006–2020)’’ gives more new con-
struction land occupation of farmland index, to provide security for its economic
Table 171.1 CUR demand table in Tianjin district and county
District New construction land of New construction land of CUR Proportion
and county cultivated land occupied cultivated land occupied demand (%)
name index (hm2) control index (hm2) (hm2)
Binhai 2723.48 2228.17 515.28 2.59
New
Area
Dongli 3572.79 2145.1 1427.69 7.17
Xiqing 3018.17 1482.6 1535.57 7.71
Jinnan 3558.03 1482.6 2075.43 10.43
Beichen 2967.41 1198.7 1768.71 8.89
Wuqing 5094.40 1640.3 3454.10 17.35
Baodi 3488.97 1703.4 1785.57 8.97
Ninghe 2925.19 883.3 2041.89 10.26
Jinghai 4430.73 1135.6 3295.13 16.55
Jixian 2826.03 820.2 2005.83 10.08
Total 34605.19 14700.00 19905.19 100.00
1630 G. Lin and S. Hao
development, strategic position, while in rural residential land readjustment
potential larger region gives fewer new construction land occupation of cultivated
land index. For the part of new construction land occupation of farmland indexes
cannot meet the needs of social and economic development of the area, the needs
must rely on the CUR work to meet. This suggests that the CUR policy has now
become the city of Tianjin land important macro-control means.
171.4.3 Calculation Consolidation of Rural Residential Land
of Tianjin CUR Supply Ability
There are two steps to calculate consolidation rural residential land in Tianjin CUR
supply ability according to (171.4) and above listed 2.2 calculated steps.
(1) Calculation of Tianjin rural residential land consolidation potential in the
planning target years. This study uses the method from Song et al. (2006) to
calculate Tianjin rural residential land readjustment potential, the calculation
is as (171.5) and (171.6):
Si ¼ Sx ðAt MtÞR ð171:5Þ
Mt ¼ G=Qt ð171:6Þ
In (171.5) and (171.6): Si is rural residential land consolidation potential, Sx is
current situation of rural residential area, At is the average standard of homestead
land in the target years, Mt is households in the target years, R is the proportion of
residential land in the target years, G is rural population in the target years, Qt is
household scale in the target years.
Calculate Tianjin rural residential land consolidation potential according to the
(171.5), (171.6) after determining household scale, the average homestead area,
the proportion of residential land in Tianjin counties and the results would be in
Table 171.2. From Table 171.2, rural settlements obviously regional differences
exist in Tianjin districts consolidation potential. Suburban districts and Binhai
New Area developed area, rural residential land readjustment potential is relatively
lower, because of the more developed economies, higher urbanization level, the
higher rural residential land saving and intensive use level. Conversely, some
undeveloped economy, urbanization level is lower, and along with the economic
and social development in the future, the rural residential land consolidation
potential is higher. Dongli district rural residential land readjustment potential is
minimum as 0 hm2 according to the ‘‘Tianjin city land uses overall planning
(2006–2020)’’. In addition, Binhai New Area rural residential land consolidation
potentiality is lower, as 3,540.61 hm2 and Beichen is 695.67 hm2. Baodi is the
highest as 13,910.79 hm2, followed by Jixian and Wuqing, which are 13,228.82
and 11,935.65 hm2 respectively.
Table 171.2 RURAL residential land readjustment potential in Tianjin district and county (unit: hm2)
District and county name Arrangement potential
Binhai New Area 3540.61
Dongli 0
Xiqing 6970.12
Jinnan 7837.01
Beichen 5695.67
Wuqing 11935.65
Baodi 13910.79
Ninghe 5141.57
171 The Increase and Decrease Connecting Potential
Jinghai 9676.31
Jixian 13228.82
Total 77936.56
Label Dongli district rural residents are without arrangement potential because it is planning rural residential area is 0 hm2 in the target years, according to
the Tianjin city land uses overall planning (2006–2020)
1631
1632 G. Lin and S. Hao
(2) Calculation of CUR supply ability in Tianjin in the target years. According to
the CUR case in Tianjin districts, Tianjin CUR rural dweller dot reclamation
newly cultivated land coefficient between 0.80 and 0.92 in the demolition of
the old district. Considering relevant expert opinions, this study identified each
involving agricultural district newly cultivated land coefficient is 0.85, in
addition to Jixian. Comprehensive determination of newly increased cultivated
land coefficient is 0.5, because Jixian is located in mountainous terrain district,
which is the northernmost ecological area of Tianjin, the main target land use
is ecological conservation and bearing the development of tourism and many
rural residents is not suitable for land reclamation. By (171.4), we can cal-
culate each distinct and county CUR supply ability as Table 171.3. Tianjin
CUR supply capacity is 24,841.79 hm2, among them, Baodi 6,966.34 hm2,
Jixian 4,506.96 hm2 followed by Ninghe, Wuqing and Jinghai; The CUR and
supply capacity of Jinnan, Binhai New Area and Beichen is lesser, Dongli is
0 hm2. The space link differences of each area county’s CUR ability of
reflecting the differences of the various districts and counties in Tianjin eco-
nomic development level, the industrialization, the urbanization process. The
counties’ supply capacity with high levels in the industrialization, the
urbanization level, economical and intensive utilization of land is poor;
Otherwise, is strong.
171.4.4 The Potential Estimates of Tianjin CUR in Planning
Target Years
According to (171.1) and (171.2), we can calculate the CUR ability of the Tianjin
each area county’s planning target years, specifically seen in Table 171.4. From
the Table 171.4, we can see the Tianjin’s potential CUR coefficient of planning
target years is 1.25, a CUR for 4,936.60 hm2 potential balances, which shows
through the CUR work, Tianjin can meet the demand of new town planning target
years construction land occupying cultivated land, and there’s a balance potential
index. However we can see from the various districts and counties’ CUR potential
coefficient, different areas’ county CUR has bigger space potential differences,
such as Jixian, Ninghe, Baodi, Jinghai and other rare counties have bigger CUR
potential, but Dongli, Beichen, Jinnan Xiqing districts, this four suburban areas,
and Binhai New Area’s CUR potential is small. In order to effectively regulate the
imbalance of CUR potential between each district and county and realize
the coordinated development among different levels regions, we should divide the
whole city into different CUR areas and distribute the balance CUR potential
index, according to the CUR potential coefficient.
Table 171.3 CUR supply ability in Tianjin district and county (unit: hm2)
District and county name CUR supply ability
Binhai New Area 287.97
Dongli 0
Xiqing 1064.8
Jinnan 216.41
Beichen 524.34
Wuqing 3715.19
Baodi 6966.34
171 The Increase and Decrease Connecting Potential
Ninghe 4054.72
Jinghai 3505.07
Jixian 4506.96
Total 24841.79
1633
1634 G. Lin and S. Hao
Table 171.4 Cur potential Area county name CUR potential CUR potential
coefficient and cur potential coefficient balance (hm2)
balance of each district and
country Binhai New Area 0.56 -227.31
DongLi 0 -1427.69
Xiqing 0.69 -470.79
Jinnan 0.10 -1859.02
Beichen 0.30 -1244.37
Wuqing 1.08 261.09
Baodi 3.90 5180.77
NingHe 1.99 2012.83
JingHai 1.06 209.94
Jixian 2.25 2501.13
Total 1.25 4936.60
171.5 Conclusion
The CUR is an effective method in speeding up new countryside construction,
developing urban and rural together, optimizing the structure of land using and
improving the land intensive using. In the process of implementing, the potential
estimation is the basis and the premise of land layout measure; need to choose a
scientific and reasonable index to represent the size of the potential. Building the
CUR potential measurement method and the technical route is based on the
general land use, and the calculated result can reflect the actual area, tightly
integrated in land use control, which would be useful data references for other
places applying for CUR experimental unit, distributing CUR quota, developing
CUR items, laying the foundation for working out land reclamation of planning,
and providing quantitative data and references for land and resources management
departments to develop, innovative policy for CUR.
References
Feng JM, Chen LQ, Song X (2011) CUR policy analysis. Anhui Agric Bull 14:12–13 (in
Chinese)
Gan LC, Zhou BT (2008) Based on the CUR of the rural construction land consolidation benefits
analysis. Land Econ 10:42–46 (in Chinese)
Gong MF (2012) Analyzing the advantages and disadvantages of CUR. Rural Econ Sci Technol
01:10–11 (in Chinese)
Li WJ, Wang L (2009) Analyzing the advantages and disadvantages of CUR. Inf Land Resour
4:34–37 (in Chinese)
Li ZJ, Fan ZA, Gao MX (2007) The rural settlement arrangement mode and countermeasures in
CUR policy—TaiAn in Shandong province as an example. J Shandong Agric Univ 1:32–36
(in Chinese)
Mai XS (2008) CUR economic analysis—Shapingba in Chongqing as an example. Master thesis
in Southwest University (in Chinese)
171 The Increase and Decrease Connecting Potential 1635
Qv YB, Zhang FR, Jiang GH, Li L, Song W (2011) Rural residential areas of land consolidation
potential and CUR partition research. Resour Sci 33:134–142 (in Chinese)
Shao SJ, Li XS (2009) Rural and urban construction land increase or decrease peg reading. Law
Soc 10:290–291 (in Chinese)
Song W, Zhang FR, Chen XW (2006) Our country rural residential areas potential consolidation
measuring method. Guangdong Land Sci 5:43–47 (in Chinese)
Wang HL, Wang X (2009) Liaoning province CUR mode analysis. Land Resour 6:48–49 (in
Chinese)
Wang J, Zhu YB (2007) In CUR operation pattern discussion. Ural Econ 8:29–32 (in Chinese)
Wang WM, Han TK (2002) The land use planning learning. China Agriculture Press, Beijing,
p 10 (in Chinese)
Xu WD, Wang ZR (2009) In CUR policy Shandong province of rural construction land
consolidation potential and key area. Shandong Land Resour 1:23–25 (in Chinese)
Yu YQ (2011) The CUR potential analysis—Nasi in Xinjiang as an example. Econ Res 23:83–89
(in Chinese)
Yuan HZ (2011) CUR implementation evaluation index system construction analysis. China
Collect Econ 3:102–103 (in Chinese)
Chapter 172
Study on Re-Evaluation of Technological
Innovation Efficiency Based on the C2R
Improvement Model in Zhongguancun
High-Tech Enterprises
Jing-wen An, Sen Zhang and Guang-lin Sui
Abstract To begin with, this paper studied the relative efficiency of the
Innovation Efficiency of 10 major High-tech industries in Zhongguancun. The
study found that 7 of the 10 high-tech industries in Zhongguancun are relatively
effective in their Innovation Efficiency. They are industries of electronic infor-
mation, advanced manufacturing, new energy, new materials, modern farming,
ocean engineering and nuclear application. Then this article introduced the virtual
optimization of DMU based on the C2R model, which re-evaluated the relative
effectiveness of the above-mentioned seven industries. Then this paper gave some
suggestions to improve the innovation efficiency of these industries.
Keywords Zhongguancun High-tech Industries Data envelopment analysis
DEA Virtual decision making units
172.1 Introduction
The innovation efficiency of the high-tech industries in Zhongguancun demon-
stration industrial park is the conversion efficiency of input–output of production
factors. It reflects the capacity of the industrial park to utilize technological
resources and develop high-tech industries, as well as the quality and standard of
the Zhongguancun innovation system. The Zhongguancun National Self-innova-
tion Demonstration Park (hereafter referred to as ‘‘Zhongguancun’’) is China’s first
self-innovation park, a leading area of high-tech industries and the hotbed of the
strategic newly emerging industries. In the new stages of the twenty-first century, a
re-evaluation of the innovation efficiency of the Zhongguancun high-tech
J. An S. Zhang (&) G. Sui
School of Management, China University of Mining and Technology
Beijing, Beijing, China
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1637
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_172,
Springer-Verlag Berlin Heidelberg 2013
1638 J. An et al.
industries can help government better plan new industries, utilize resources, raise
efficiency and promote industrial restructuring.
172.2 DEA Model
172.2.1 C2R Model
Data Envelopment Analysis is a nonparametric comprehensive evaluation method
to analyze the relative effectiveness of the DMU featured by high input and output
(Wei 2004). It has been widely used to assess the efficiency of technology inno-
vation within the same category of DMU (Che and Zhang 2010; Cheng and Chen
2009). Usually DEA can be divided into C2R model and B2C model, both of which
conduct horizontal comparison and analysis to different DMUs at the same time.
The C2R model is as follows (Xu et al. 2009):
8 !
>
> X m X s
>
> min h e
S þ S þ
¼ v d ðeÞ
>
>
>
> i¼1 r¼1
>
>
>
> s:t:
>
>
>
>
>
> X n
< X j k j þ S ¼ h X0
ðDÞ j¼1 ð172:1Þ
>
>
>
> X n
>
>
>
> Y j kj Sþ ¼ Y 0
>
>
>
> j¼1
>
>
>
> kj 0; j ¼ 1; 2 ; n
>
>
: þ
S 0; S 0
Within the (D) model, h represents the effective value of DMU0, which is the
T
effective use of the input against output. Xj ¼ x1j ; x2j ; ; xmj ; j ¼ 1; 2; . . .; n
T
represents the input of DMU, Yj ¼ y1j ; y2j ; ; ysj ; j ¼ 1; 2; . . .; n represents the
output of DMU. S +, S- stands for the slack variable, which is supposedly bigger
than zero. kj stands for composition ratio of DMUj within the DMU composition
(Wang and King 2009).
172.2.2 Improved DEA-C2R Model
In actual DEA evaluation processes, most of the DMU are relative effective, and
only a few of the DMUs are invalid. This is because there are too many indexes and
too few DMUs, making the analysis result less practical (Duan 2007). In this case, it
172 Study on Re-Evaluation of Technological Innovation 1639
should be made further analysis to the relative effective DMUs to evaluate the
efficiency. There are many ways of sequencing in the DEA evaluation, and this paper
adopts the virtual unit method (Duan 2007; Hua and Tao 2011; Liu and Song 2010).
Within the virtual unit method, a virtual decision making unit DMUn+1 is
introduced to replace the normal DMU0 within the General model of constraint
conditions so as to distinguish the different degrees of different DMUs. Suppose
the input and output of DMUn+1 is (xi,n+1,yk,n+1), xi;nþ1 ¼ min xij ði ¼ 1; ; mÞ
1jn
yk;nþ1 ¼ max ykj ðk ¼ 1; ; sÞ. The virtual decision making unit DMUn+1 is the
1jn
best decision making unit among the valid DMUs. It’s compared the efficiency
value of virtual DMU with the efficiency value of other DMUs. If the DMU show a
value that is approximate to the virtual DMU, then the value is high. The evalu-
ation process can be achieved by inputting valid DMU, or by introducing a virtual
DMUn+1. The result can be calculated through (De1) (Liu and Song 2010).
8
>
> min½h eð^eT s þ eT sþ Þ
>
>
>
> X
nþ1
>
>
>
>
> s:t: kj xj þ s ¼ h x0 ; j ¼ 1; 2; ; n þ 1; j 6¼ j0
>
>
< j¼1
ðDe1 Þ X n þ1 ð172:2Þ
>
> kj yj sþ ¼ y0 ; j ¼ 1; 2; ; n þ 1; j 6¼ j0
>
>
>
> j¼1
>
>
>
>
>
> k j 0; j ¼ 1; 2 ; n þ 1
>
: þ þ þ
s ðs1 ; s2 ; ; sþ
s Þ 0; s ðs1 ; s2 ; ; ss Þ 0
The sequence of efficiency value from (D)e is the sequence of quality of the
decision making unit DMUj0. If the DMU value B1, the bigger the value is, the
better the quality of this DMUj0.
The DMU efficiency value (De2) can be calculated after the introduction of the
virtual DMUn+1.
8
>
> max½a þ eð^eT s þ eT sþ Þ
>
>
>
> X
nþ1
>
>
>
>
> s:t: kj xj þ s ¼ h x0 ; j ¼ 1; 2; ; n þ 1; j 6¼ j0
>
>
< j¼1
ðDe2 Þ X n þ1 ð172:3Þ
>
> kj yj sþ ¼ y0 ; j ¼ 1; 2; ; n þ 1; j 6¼ j0
>
>
>
> j¼1
>
>
>
>
>
> k j 0; j ¼ 1; 2 ; n þ 1
>
: þ þ þ
s ðs1 ; s2 ; ; sþ
s Þ 0; s ðs1 ; s2 ; ; ss Þ 0
The sequence of efficiency value from (D)e is the sequence of quality of
the decision making unit DMUj0. If the value is C1, the smaller the value is, the
smaller the gap between it and the virtual DMU, thus the better the quality of
the DMUj0.
1640 J. An et al.
This paper firstly evaluates the DMU through C2R and B2C model within
the DEA method. Then the paper introduces the virtual processing unit DMUn+1.
The effectiveness of the DMU is calculated through the efficiency evaluation of the
valid DMU.
172.3 Case Study of the Effectiveness of the Innovation
Efficiency of Zhongguancun High-tech Industries
172.3.1 Establishment of the Indicator System
This paper studied the innovation efficiency of the Zhongguancun innovation
system and the research results of relative scholars, took into consideration the
representativeness and accessibility of these indicators. As a result, the paper
selected 6 indicators as the evaluation criteria of the innovation efficiency of
Zhongguancun high-tech industries. Among these six indicators, three are input
indicators and three are output indicators (He et al. 2010; Wang 2008; Xie 2011;
Quan et al. 2008).
Among the input indicators are: proportion of technology staff in the industry I1,
total expenditure of the technological innovation I2, proportion of technology
expenditure in the total revenue I3. I1 stands for the input intensity of technology staff,
which is the ratio of technology staff against the total staff. I2 stands for the activity of
technological innovation of the industry. I3 stands for the level and intensity of the
industry independent innovation (Zhang and Liu 2008; Li and Xie 2010).
Among the Output indicators are: quantity of the patent accredit O1, proportion
of new product sales revenue accounted for product sales revenue O2, proportion
of new product sales revenue accounted for gross value of industrial output O3. O1
stands for the industry innovation important output value; O2 stands for the rate of
new product sales, the contribution degree of industry enterprise technology
innovation into new products on enterprise value creating; O3 stands for the
transformation ability of the industry technology innovation (Zhang and Li 2009).
172.3.2 Selection of Data
All of the evaluation indicators used in this article in assessing the innovation
efficiency of Zhongguancun high-tech industries are objective. The data used here
are mainly from the Annual Book of Zhongguancun High-tech Industrial Park and
the Annual Book of Zhongguancun National Demonstration Park of Self-inno-
vation. Part of the data is from statistical data of the website of Zhongguancun
National Demonstration Park of Self-innovation from 2006 to 2010 (http://
www.zgc.gov.cn/tjxx/). And part of the data is through calculation of these
existing data. Therefore, these data are highly objective and credible.
172 Study on Re-Evaluation of Technological Innovation 1641
This paper treats the high-tech industries of Zhongguancun as a high-input and
high-output system. Decision variables (DMUj, j = 1, 2,…,10) are the 10 high-
tech industries of Zhongguancun Demonstration Park. Because of the different
time-lag between input and output of the innovation (Wang et al. 2009), the DEA
efficiency analysis only uses the average figure of the indicator data in the Annual
Book of the 11th five-year plan period.
172.3.3 Evaluation of DEA-B2C Model
This paper utilizes the DEA input–output returns to scale B2C model, puts the data
of Table 172.1 into the model and gets the result through DEAP 2.1. The innovation
efficiency value of the 10 key high-tech industries is shown in Table 172.2. Since
the DEA method is a relative evaluation method, the DEA value in Table 172.2 just
stands for its degree of the relative effectiveness (Wang et al. 2009).
It can be seen from Table 172.2 that the biggest overall DEA value is 1, the
smallest being 0.454 and the average value being 0.913. Among the innovation
efficiency evaluation of Zhongguancun high-tech industries, 7 industries (elec-
tronic information, advanced manufacturing, new energy, new material, and
modern agriculture, ocean engineering and nuclear application) have an innovation
efficiency value of 1. This means that 70 % of the DMUs are effective while 30 %
(environment protection, biomedicine and aerospace) of these are not. Generally
speaking, most of the high-tech industries are relative effective in terms of inno-
vation efficiency.
172.3.4 Evaluation of the Improved DEA-C2R Model
It can be seen from the results of the C2R and B2C model that industries with
relative effective innovation efficiency account for a bigger share. In order to
Table 172.1 Statistical indicators of the 10 high-tech industries of Zhongguancun in the 11th 5-
year plan period
Industries I1 (%) I2 (billion Yuan) I3 (%) O1 (item) O2 (%) O3 (%)
Electronic information 39 36.159 13 2764.00 61 83
Biomedicine 25 1.945 8 378.20 41 45
New material 25 2.493 7 676.00 66 88
Advanced manufacturing 25 4.133 6 1022.00 42 51
Aerospace 40 1.751 37 51.80 53 62
Modern agriculture 24 0.427 6 71.00 70 97
New energy 30 2.643 4 439.80 77 103
Environment protection 34 1.029 15 227.60 63 113
Ocean engineering 27 0.075 14 11.60 20 24
Nuclear application 45 0.275 12 67.20 79 69
1642 J. An et al.
Table 172.2 Innovation efficiency value of Zhongguancun high-tech industries
Field of technology Overall Pure technical Scale Returns to
efficiency efficiency efficiency scale
Electronic information 1.000 1.000 1.000 crs
Biomedicine 0.718 0.973 0.738 irs
New material 1.000 1.000 1.000 crs
Advanced 1.000 1.000 1.000 crs
manufacturing
Aerospace 0.454 0.600 0.757 irs
Modern agriculture 1.000 1.000 1.000 crs
New energy 1.000 1.000 1.000 crs
Environment 0.957 1.000 0.957 drs
protection
Ocean engineering 1.000 1.000 1.000 crs
Nuclear application 1.000 1.000 1.000 crs
Average value 0.913 0.957 0.945
distinguish the efficiency value of these industries, a virtual unit DMU11 was
introduced to re-evaluate the innovation efficiency, as is shown in Table 172.3.
Suppose e = 10-6, a C2R model based on the input of Archimedes infinitesimal
C R model is established. The C2R model of DMU1 is as follows:
2
8
>
> min½h eðs þ
1 þ s2 þ s3 þ s1 þ s2 þ s3 Þ
þ þ
>
>
>
> s:t:0:39k1 þ 0:25k2 þ 0:25k3 þ 0:24k4 þ 0:3k5 þ 0:27k6 þ 0:45k7 þ 0:24k8 s
>
> 1 ¼ 0:24
>
>
>
> 361:59k 1 þ 24:93k 2 þ 41:33k 3 þ 4:27k 4 þ 26:43k 5 þ 0:75k 6 þ 2:75k7 þ 0:75k 8 s
2 ¼ 0:75
>
>
< 0:13k þ 0:07k þ 0:06k þ 0:06k þ 0:04k þ 0:14k þ 0:12k þ 0:04k s ¼ 0:04
1 2 3 4 5 6 7 8 3
ðDe1 Þ þ
>
> 2764k1 þ 676k2 þ 1022k3 þ 71k4 þ 439:8k5 þ 11:6k6 þ 67:2k7 þ 2764k8 s1 ¼ 2764
>
>
>
> 0:61k1 þ 0:66k2 þ 0:42k3 þ 0:7k4 þ 0:77k5 þ 0:2k6 þ 0:79k7 þ 0:79k8 s2 ¼ 0:79 þ
>
>
>
>
> 0:83k1 þ 0:88k2 þ 0:51k3 þ 0:97k4 þ 1:03k5 þ 0:24k6 þ 0:69k7 þ 1:03k8 sþ
> 3 ¼ 1:03
>
>
:
kj 0; j ¼ 1; 2 ; 8sþ ðsþ
1 ; sþ þ
2 ; s3 Þ 0; s
ðs 1 ; s2 ; s3 Þ 0
ð172:4Þ
After calculation by Matlab, the following results are innovation efficiency of
Zhongguancun high-tech industries. See Table 172.4.
It’s known that virtual evaluation unit is the best decision making unit.
Therefore, we can rank the innovation efficiency of Zhongguancun high-tech
industries in the following sequence: new energy [ modern agriculture [ new
material [ electronic information [ nuclear application [ advanced manufactur-
ing [ ocean engineering. In terms of economies of scale, industries of new energy,
modern agriculture and ocean engineering are in the best condition, while all other
industries have witnessed an increasing trend of returns to scale.
Based on the DEA-C2R, it can get the invalid input indicator slack variable
value and output indicator slack variable value of the 7 high-tech industries in
Zhongguancun Demonstration Park. See Table 172.5 (the input residual value and
insufficient output value is zero, which is nothing in the table).
172 Study on Re-Evaluation of Technological Innovation 1643
Table 172.3 Indicators of virtual DMUs
DMU I1 (%) I2 (million) I3 (%) O1 (item) O2 (%) O3 (%)
DMU11 24 75 4 2764 79 103
Table 172.4 Evaluation of DEA efficiency C2R when combined with virtual DMUs
DMU h h (after 1–4 5 6–7 8 9 1
P
n
h kj
(initial) improvement) j¼1
Electronic information 1 0.615 0.000 1.000 1.000 1.626
New material 1 0.820 0.000 0.854 0.854 1.042
Advanced 1 0.510 0.000 0.532 0.532 1.042
manufacturing
Modern agriculture 1 0.942 0.000 0.942 0.942 1.000
New energy 1 0.999 0.187 0.813 1.000 1.000
Ocean engineering 1 0.253 0.000 0.253 0.253 1.000
Nuclear application 1 0.533 0.000 1.000 1.000 1.876
Table 172.5 Value of input–output slack variables
DMU S1- S2- S3- S1+ S2+ S3+
Electronic information 0.61 0.31 0.30 0.24
New material 0.05 0.18 0.61 0.02 0.00
Advanced manufacturing 0.06 0.07 0.16 0.05
Modern farming 0.01 0.15 0.92 0.07 0.00
New energy 0.13 0.06 0.68 0.03 0.00
Ocean engineering 0.02 0.19 0.25 0.03
Nuclear application 0.18 0.98 0.41
Table 172.6 Input residual value and insufficient output value
DMU S1- (%) S2- (billion Yuan) S3- (%) S1+ (item) S2+ (%) S3+ (%)
Electronic information 22.1767 40 18 20
New material 1.987 23 1685.476 15
Advanced manufacturing 0.2695 90 447.468 38
Modern agriculture 0.3315 19 2531.99 44
New energy 60 2.568 2324.2 20
Ocean engineering 80 25 688.147 21
Nuclear application 0.0717 24 2696.8 34
According to the projection analysis theory, it got the input residual value and
the insufficient output value of the seven high-tech industries of Zhongguancun
Demonstration Park, which is relatively invalid. See Table 172.6 (the input
residual value and insufficient output value is zero, which is nothing in the table).
1644 J. An et al.
172.4 Conclusion
Through the DEA analysis of innovation efficiency of the high-tech industries in
the Zhongguancun industrial park, It can be seen that:
172.4.1 Electronic Information
The efficiency value of the electronic information industry is 0.615 as evaluated
through the DEA, ranking the fourth in the seven high-tech industries, with an
economy of scale of 1.626 and an increasing trend. In 2010, electronic information
industry accounted for 46.29 % of the Demonstration Park. It also accounted for
the largest proportion of technological expenditure in the 11th five-year plan
period, almost 2.5 times as the other 9 fields. This shows that electronic infor-
mation industry is No. 1 pillar industry of Zhongguancun Industrial Park, with the
most active innovation but relatively low innovation efficiency. According to
Table 172.6, on condition that the input does not change, it should reduce the
technological expenditure by 22.1767 billion Yuan, and the input intensity be
reduced by 40 %. And while maintaining a constant input, it should raise the
proportion of sales revenue of new products in total sales revenue and total
industrial output by 18 and 20 % respectively, an effective efficiency.
172.4.2 Advanced Manufacturing
As the second largest industry in Demonstration Park, the advanced manufacturing
industry has an evaluation efficiency of 0.51 after DEA evaluation, ranking the 6th
with a scale efficiency value of 1.042 and a growing scale. Its revenue accounts for
11.89 % of the Zhongguancun Demonstration Park. Therefore, it should, in
accordance with the plan of upgrading manufacturing industrial clusters, with the
output unchanged, cut technological expenditure by 269.5 million Yuan, or reduce
90 % of its input intensity. And while maintaining a constant input, it should
increase 447.468 items of patents and raise the proportion of new products sales
revenue in total sales revenue by 38 %.
172.4.3 New Energy
As a growth point of the industries in the Demonstration Park, the new energy
industry has an efficiency value of 0.999 after DEA evaluation, ranking the first,
172 Study on Re-Evaluation of Technological Innovation 1645
with a scale efficiency of 1 and a constant return to scale. In 2010, the revenue of
the new energy industry accounts for 10.93 % of the industrial park. Therefore, in
accordance with the plan of developing new energy, with the output unchanged, it
should reduce the proportion of technological staff by 60 %, cut the technological
budget by 2.568 billion Yuan. While remaining a constant input, it should increase
the patent authorization by 2324.2 pieces and the proportion of new products sales
revenue by 20 %.
172.4.4 New Material
As a fast growing industry, the new material industry gets an efficiency evaluation
value by 0.820 % after DEA evaluation, ranking the third, with a scale efficiency
of 1.042 and increasing trend. In 2010, the new material industry accounted for
6.73 % of the total revenue in the industrial park. So, while maintaining a constant
output, it should cut the technological budget by 1.987 billion Yuan; reduce the
budget input intensity by 23 %. And while maintaining a constant input, it should
increase patent authorization by 1685.476 items and the proportion of new prod-
ucts sales revenue by 15 %.
172.4.5 The Modern Agriculture Industry
The modern agriculture industry gets an efficiency evaluation value by 0.942,
ranking the second, with a scale efficiency of 1 and a constant return to scale. In
2010, modern agriculture industry accounted for 0.76 % of the total revenue in the
Demonstration Park. Though the proportion is small, this industry is essential to
the people’s wellbeing. So, while maintaining a constant output, it should cut the
technological budget by 331.5 million Yuan; reduce the budget input intensity by
19 %. And while maintaining a constant input, it should increase patent authori-
zation by 2531.99 items and the proportion of new products sales revenue by
44 %.
172.4.6 Nuclear Application
The nuclear application industry gets an efficiency evaluation value by 0.533,
ranking the 7th, with a scale efficiency of 1 and an increasing return to scale. In
2010, modern agriculture industry accounted for 0.17 % of the total revenue in the
Demonstration Park. Despite a tiny proportion, this industry is strategically
important to the national economy. So enough attention should be paid to this
industry. While maintaining a constant output, it should cut the technological
1646 J. An et al.
budget by 71.7 million Yuan; reduce the budget input intensity by 25 %. And
while maintaining a constant input, it should increase patent authorization by
2696.8 items and the proportion of new products sales revenue by 34 %.
172.4.7 Ocean Engineering
The ocean engineering industry gets an efficiency evaluation value by 0.253,
ranking the 8th (the last place) with a scale efficiency of 1 and a constant returns to
scale. In 2010, modern agriculture industry accounted for 0.17 % of the total
revenue in the Demonstration Park. Given its strategic importance, enough
attention should be paid to this industry. While maintaining a constant output, it
should cut the proportion of the technological staff by 80 %; reduce the budget
input intensity by 24 %. And while maintaining a constant input, we should
increase patent authorization by 2696.8 items and the proportion of new products
sales revenue by 34 %.
References
Che W, Zhang L (2010) Shanghai city cooperative efficiency evaluation—based on industry data
DEA analysis. Sci Technol Prog Policy 03:20–25
Cheng H, Chen Y (2009) DEA based in Hubei Province high-tech industrial innovation efficiency
analysis. Sci Technol J 12:115–116
Duan Y (2007) Data envelopment analysis: theory and application. Shanghai Science and
Technology Press, Shanghai
He M, Shane K, Liu Y (2010) Innovation ability evaluation analysis of Zhongguancun science
and technology garden based on DEA. Sci Technol Prog Policy 9:106–109
Hua Z, Tao L (2011) Safety efficiency evaluation of China’s coal enterprises based on DEA
model and its improved model. Coal Econ Res 5:49–53
Li N, Xie Z (2010) China regional innovation system innovation efficiency evaluation—positive
analysis based on DEA. Ind Technol Econ 8:122–126
Liu W, Song M (2010) Coal listed company performance evaluation—based on DEA-C *2R
model perspective. Manag Eng 1:1–4
Quan J, Yao L, Shi B (2008) National high-tech zone sustainable development capacity
evaluation based on DEA. Soft Sci 1:75–77
Wang D (2008) Technology innovation performance evaluation of the East Lake High-tech Zone
based on DEA. J Chongqing Acad Arts Sci Nat Sci Ed 2:84–87
Wang P, Wang L, Wang L (2009) The manufacturing industry in Hunan Province based on DEA
innovation efficiency analysis. Sci Technol Manag Res 06:172–175
Wang X, King S (2009) Cooperative innovation efficiency evaluation of DEA. Stat Decis Making
3:54–56
Wei Q (2004) Data envelopment analysis. Science Press, Beijing
Xie Z (2011) Study on the national high-tech zone technology innovation efficiency influence
factors. Manag Sci Res 11:52–58
172 Study on Re-Evaluation of Technological Innovation 1647
Xu J, Sun L, He Z (2009) DEA based on China’s inter-provincial mode of high technology
industrialization development and relative advantage industry choose. Sci Technol Prog
Policy 02:30–33
Zhang G, Li B (2009) Regional innovation system efficiency evaluation based on DEA. Manag
Sci 5:47–48
Zhang G, Liu T (2008) New and high area dominant industry selection analysis based on DEA.
Technol Econ Manag Res 02:19–21
Chapter 173
Research on the Strategic Management
of Small and Medium Enterprises
in China
Xin Zhu and Ying Li
Abstract The goal of this paper is to analyze the status and the need strategic
management in SMEs in China, and make analysis on the problems and coun-
termeasures for the implementation of strategic management in SMEs. Firstly, we
make an analysis about the importance of SMEs for China’s economic develop-
ment, and analyze the need for SMEs to implement strategic management. Sec-
ondly, we make an introduction about the step of the implementation of strategic
management for the SMEs, and divide the implementation process into five steps.
Then, we make a discussion about the existing problems during implementation of
strategic management for the SMEs in recent China. We find that the lack of skills,
the lack of consideration of macro environment, and the speculative mentality are
the main obstacles for the implementation of strategic management for SMEs.
Finally, we put forward the corresponding suggestions and recommendations for
the implementation of strategic management for SMEs.
Keywords Strategic management SMEs, recommendations
173.1 Introduction
Since the reform and opening policy in 1978, the number and scale of Small and
Medium Enterprises (SMEs) have prospered for a long time with the rapid
development of private economy in China. The SMEs increasingly become an
important pillar of China’s economic development. According to the introduction
of National Development and Reform Commission, the number of SMEs in China
X. Zhu (&) Y. Li
School of Management, Guangxi University of Technology, Liuzhou, China
e-mail: [email protected]
Y. Li
e-mail: [email protected]
E. Qi et al. (eds.), The 19th International Conference on Industrial Engineering 1649
and Engineering Management, DOI: 10.1007/978-3-642-38391-5_173,
Ó Springer-Verlag Berlin Heidelberg 2013
1650 X. Zhu and Y. Li
has reached more than 4200 million, accounting for over 99.8 % of total number of
enterprises. The number of SMEs registered by the business sector reached 460
million, and the number of self-employed households reached more than 3800
million in the end of 2009. The value of final goods and services created by the
SMEs is account for 60 % of the gross domestic product. The goods produced by
the SMEs accounts for the 60 % of the community’s total sales. The tax revenue
by SMEs turned over more than half of the total. The SMEs also supply more than
80 % of the total job positions. We can get the conclusion that the SMEs are
playing an increasingly important role for the economic development in China.
The development status of SMEs becomes an important indicator of economic
vitality for a country or region all around the world.
However, it should be noted that the development of SME faced with unprec-
edented difficulties due to a variety of factors. There are serious problems for their
ideas, technologies and equipment, management structure and other aspects. Par-
ticularly in its strategic management, the majority of SMEs lack of clear strategic
positioning, and their strategic management are in chaos. With China’s accession to
WTO and economic globalization, information technology conditions, the SMEs
exposure to a dynamic ultra-competitive environment the same as the large
enterprises (Hennart 2001). This environment is unpredictable, treacherous, and for
the self-limited conditions, often in competition for SMEs at a disadvantage. This
environment is unpredictable, treacherous, and for the self-limited conditions,
SMEs are often at a disadvantage in the competition. This determines the depth of
the environment for SMEs to capture the opportunities, avoid threats, exceed and
develop appropriate strategies to be possible to grasp the business destiny.
173.2 The Need for SMEs to Implement Strategic
Management
The SMEs will face many strategic choices which can affect the fate of the
corporations in their growth and development, just the same as the large busi-
nesses. In the course of business, there are strategic management issues. SME
managers, especially senior management across the enterprise in the business
planning process of development requires far-sighted to think, to make appropriate
choices. These options relate to the long-term business interests and global
interests, only to make the right strategic choices to be effective in guiding the
development of enterprises.
173.2.1 The Implementation of the Strategic Management
is the Need for SMEs to Survive
With the development of social productivity, market supply and demand has
changed from past over-supply shortage. SMEs had a ‘‘small boat U-turn’’
173 Research on the Strategic Management 1651
advantage in the increasingly fierce competition in the market has gradually dis-
appeared. The survival of SMEs should not only to consider the current pressure,
but also to consider the future long-term environmental impact. Therefore, the
strategic management of SMEs must be on the agenda, careful analysis of the
external environment and internal business environment, accurate positioning and
positioning the industry to continuously optimize the development strategy,
otherwise the survival of SMEs will be increasingly difficult.
173.2.2 The Implementation of the Strategic Management
is the Need for SMEs to Develop
The large enterprises usually developed on the basis of SMEs through the careful
design their own development strategies, accurate self-positioning, the correct
direction of investment, etc., only to the original to the continuous development of
SMEs. Practice shows that the growth process in SMEs, strategic management is
the management of the most important, only to seize the small and medium
enterprise development strategies in the development of the road can make the
right strategic choices.
173.2.3 The Implementation of Strategic Management
is the Need for SMEs to Improve Core
Competitiveness
Core competence is to win in a competitive market, the key to SMEs succeed in
the attack of many large enterprises to survive, they must have the unique core
competence, which can not be duplicated, and strategic management is to play
small and medium enterprises effective way of core competence. SMEs under the
guidance of strategic management thinking focus on human, financial and material
resources form the core technology of its own intellectual property in order to
enhance their core competitiveness.
173.2.4 The Implementation of Strategic Management
is the Need for Innovative SMEs
With the rise of high-tech industries and knowledge economy, the world economy
has entered a new era. In this economic situation that traditional industries are
facing integration, the traditional mode of operation are facing challenges, and the
new business areas and ways are emerging (Rugman and Verbeke 2001). Only the
1652 X. Zhu and Y. Li
SMEs with continuous innovation can be invincible. In the new economic situa-
tion, the SMEs should treat strategic management as the guiding ideology, and
play their advantages as much as possible. The SMEs should continue their
innovation in areas such as in the mode of operation, technology, product devel-
opment, service production process.
173.3 The Step of the Implementation of Strategic
Management for the SMEs
Taking into account the characteristics of SMEs, the steps of the implementation
strategic management should include analysis of business environment, industry
and market position, identify strategic objectives, business strategy formulation
and selection, implementation and evaluation for strategy aspects.
173.3.1 Analysis of Business Environment
The business environment includes both the external environment and internal
environment. The purpose of the external environmental analysis is to understand
the enterprise’s survival and development of a significant impact on the various
factors including the macro environment, industry environment, and the compet-
itive environment outside the enterprise (Williamson 1999). For SMEs, the
external environment can not control, but can take corresponding measures with
the different types of external influences. SMEs have to grasp the status and trends
of macroeconomic environment and the industrial environment, which can enable
their business strategy to have strong adaptability.
Internal environment analysis is the enterprise’s own environment, including
the operation of existing enterprises, business performance, product development
and marketing, management ability, all kinds of resources to conduct in-depth
analysis to understand they will have on future activities which affect the business.
The analysis of the internal environment should have a clear understanding for the
business advantages and disadvantages. The supporting impact is advantage, and
the impeding inferior is disadvantage. We should know the advantage the com-
panies have, and should clarify the disadvantages for the further development.
In this way, we can get the right strategic direction on how to avoid weaknesses for
the long-term development for SMEs.
173.3.2 Positioning of Industry and Market
We should make comprehensive consideration of their internal and external
environment when the SMEs conduct industry and market positioning. We should
173 Research on the Strategic Management 1653
start from the situation of the enterprises themselves, and select the proper
industries and markets for the survival and development of small and medium
enterprises. For SMEs in general should choose the less trade monopoly or near
perfectly competitive market. In addition, the business scope should not be too
broad, should concentrate their limited resources and human well specialization
(Ghemawat 2003). When the development of SMEs to a certain scale, in order to
expand, you can try to diversification, but must carefully consider their own
abilities, or they might have disastrous consequences.
173.3.3 Identify the Strategic Objectives
The strategic goal means the expected results to be achieved in the scope of their
business during certain period of time under the enterprises’ operation thought. The
division of the contents of strategic objectives will be different with different cate-
gories standards. It can be divided into departmental goals, and job goals according
to target level. It can also be divided into long-term goals, medium-term objectives
and short-term goals according to the length of time. The SMEs should be according
to their capabilities, and can not be too high or too low, it should be through the
efforts of the enterprise can achieve In determining the strategic objectives.
173.3.4 Formulate and Choose the Business Strategy
Business strategy is means of channels and concrete action plans adopted to
achieve the strategic goals. Management strategy to be addressed include business
how to allocate and use enterprise resources, how to coordinate actions among
various departments, how to improve overall competitiveness of enterprises to
achieve the desired strategic effects (Almeida 1996). SMEs in the development
and selection of management strategy must be combined with the characteristics of
the enterprises themselves, according to the enterprise, outside the specific envi-
ronment to develop viable business strategy, the choice of the number of options to
select the best of its ability to pay attention not to go to the pursuit of satisfaction
strategy difficult to achieve the optimal strategy.
173.3.5 Implementation and Evaluation of the Strategy
The Implementation and evaluation of the strategy is the key link to achieve
strategic objectives. The SMEs are different from large enterprises, so we should
pay more attention for the strategy control in the implementation of the strategy.
That means we should control the speed, direction, deviation of the implementation
1654 X. Zhu and Y. Li
of the strategy. At the same time, we should clarify the scope of responsibility of all
organizations, so that every department, every employee behavior and corporate
coherent overall strategy. In the strategy implementation process, we also con-
stantly checking the implementation, correct the problem in time. We should make
accurate evaluation for the implementation of the strategic objectives, and make
certain adjustment according to strategic objectives and strategic programs. The
implementation of the strategy process is a continuous improvement process. The
SMEs should be on implementation of the strategy process and results of lessons
learned in a timely manner, so as to achieve satisfactory results.
173.4 The Problems of Implementation of Strategic
Management for the SMEs
173.4.1 The Lack of Strategic Development
and Implementation Skills
Some of our SMEs lack strategic thinking, and don’t have adequate understanding
of what is the strategy, the value of strategic for enterprise. They treat strategy
intangible thing, and the short-term behaviors are very common without long-term
goals. However, more and more enterprises realize the importance of strategic with
the deep of reform and opening up, the frequency of economic activities, the
promotion of their own understanding (Hedlund 2007). But many of the managers
of SMEs lack enough awareness of what kind of strategy on the development, how
to develop strategies and how effective implementation of the strategic and other
issues. Some SMEs treat corporate profits as a strategic objective, but lack of depth
thought on where’s the money earned, earn whose money and so on. The company
is in machine industry today, may enter the health care products, real estate and
other industries tomorrow, which significantly increase business risk.
173.4.2 SMEs Strategy is Not with the State’s Political
and Economic Environment
A considerable part of the SME leaders think macro-economic and political
environment has little to do with production and operations (Chang and Singh
2000). Some of them even don’t have time to take account of these issues. China’s
SME owners like to deal with urgent business personally, and are difficult to find
time for strategic thinking. Their efforts are how to survive their enterprises and
how to solve the employees’ food problem. Thus changes in domestic and inter-
national political and economic situation and little is known about the general
direction of existing problems. The scanty information, shallow understanding and
173 Research on the Strategic Management 1655
they are not good at closely linking the macro political and economic environment
and production and operations. Their grasp of policy and the economy is relatively
slow, which missed the best opportunity of development. Some enterprises even as
the direction and policy guidance is inconsistent, leading to bankruptcy and
insolvency risk.
173.4.3 The Speculative Mentality is Serious and Like
Opportunism
During China’s 30 years of market economic reforms, many speculators have
gotten great interest in short time due to policy loopholes and laws are not perfect.
In this way, many small business owners develop a quick success, ignoring the
rules of the bad habits (Scott 2006). They believe that success relies on speculation
and luck, and don’t take a strategic perspective on the development of enterprises.
There are lots of meteor type business because of the lack of speculative oppor-
tunities and intense competition in the market. Such kinds of enterprises have
created a brilliant closing down of many small and medium enterprises.
173.5 The Recommendations to Promote the Strategic
Management of SMEs
Whether to implement the strategic management of the enterprise is ultimately
determined by enterprises. Can be expected, with the SME owners and entrepre-
neurs on the environment is complex and dynamic nature of awareness-raising, our
newly revised state constitution to protect private property on the terms of the
deeply rooted among the growing number of SMEs will abandon the short-term
behavior, lofty ideals, in order to implement and strengthen the strategic man-
agement as an inherent requirement of enterprise development. We put forward the
following recommendations, in order to make strategic management been widely
used to enable the small and medium enterprises.
173.5.1 Strong Leadership Training and Enhance
Their Own Capacity
Corporate strategy is often the values of a company’s top leader. So, First of all,
business leaders must be trained to improve their strategy awareness and skills.
Now business leaders are increasingly aware of the importance of training.
However, most of them think that only the subordinates need training to improve
1656 X. Zhu and Y. Li
strategy execution. They do themselves or as busy no time to participate in specific
training or simply think they don’t need training. As everyone knows most need
training in business strategy is to business leaders themselves. Training method
can be flexible. For example, enterprises have plans to allow the operator to ‘‘go
out’’ to the advanced business study tour or through target training. We can send
them out to supplement the theoretical knowledge management systematically. We
can also ask some related experts and scholars come to update their management
philosophy and knowledge. We have to develop their thinking, enhance their
management, strengthen the political and economic policy studies and enhance
their ability to grasp the political and economic direction.
173.5.2 Encourage the Staff Involved in Developing
Corporate Strategies
The staff participation is the key for whether the corporate strategy can be carried
out perfectly. When employees do not agree with the strategic decisions, there will
be resistance, decreased satisfaction, which will impact on productivity directly. If
a company’s employees do not understand how to be different with companies, do
not understand what business value created, they will be difficult to understand
face multiple choices (Brandenburger and Nalebuff 1995). If the sales staff does
not know strategy, they will not know who to sell. If engineers do not know
strategy, they will not know what outputs. If the employee participate in strategic
management, understand business strategy development through, it is easy to
recognize strategy, understand strategic in their daily work.
173.5.3 Strengthen Human Resource Management
Modern enterprise competition is the talent competition. Any effective strategy is
to develop and implement by people. Therefore, the level of capacity of enterprises
will directly determine the effect of the strategy. There are certain problems for the
human resources management of China’s SMEs such as the lack of scientific and
effective introduction of talents, training and use of mechanisms; personnel
appointments and irrational allocation of human resources; human resource per-
formance evaluation and incentives imperfect. There are two main reasons for the
low efficiency of productivity in SMEs: the lack of staff capability in itself; the
second is the lack of system and means to full use the human capacity. We can rely
on new concepts, establish the right talent, strengthen staff training, individual
talent among staff. We can establish scientific personnel selection system and
equal competition, and establish a scientific distribution system. We should
improve the incentive mechanism, so as to improve the level of human resource
management of SMEs.
173 Research on the Strategic Management 1657
173.5.4 We Should Make Use of Outer Brain and Establish
Thinking Tanks
There are more and more factors to consider for enterprises’ business development
strategy. Enterprise business development strategy more and more factors to
consider. Strategies must also change according to market changes more and more
frequently. We should grasp the changing situation, develop appropriate strategies
and countermeasures. It is not enough to make decision with individual mind, and
should focus as much as possible of social intelligence (Glimstedt et al. 2006). The
thinking tank has the following functions:
Innovate business thinking. We can break the company’s own mindset through
the introduction of foreign brains Resources. We can eliminate blind spots in pro-
duction, management, sales, service, research and other areas to provide new ideas,
new knowledge, new information, new ideas, new methods, and new strategies.
Enhance enterprise intelligence. Leadership and employees are fixed constants,
but the outer brain resources are infinite variables. The establishment of think
tanks outside the brain will result in enhanced business intelligence advantage.
Mechanism in the enterprise has created a unique highlight of the new intelligence,
which will effectively improve the business identify problems and solve problems
(Mulcaster 2009). Competition in the market committed fewer errors, and create
more opportunities.
Develop the interface of enterprise. The introduction of outer brains can not
only enhance business intelligence with external intelligence, but also can inte-
grating the human, material and social relations owned by outer brains. We can
expand the interface through the project or joint system of enterprise management
platform. This flexible operation mode can maximize the company’s virtual
resources and make up the short board for business (Markides 1999).
173.6 Conclusion
Some SME managers believe that only the large enterprises need strategic man-
agement, while SMEs do not need. Such kind of idea is completely wrong. The
Viability of the SMEs is tough, because the SMEs can not to compete with larger
enterprises in terms of technology, personnel, capital and other aspects. SMEs
without clear strategic management thinking will easily get lost in the market and
defeated with the increasingly fierce international and domestic market competi-
tion. Therefore, how to implement the strategic management of small and medium
enterprises, how to analyze their strengths and weaknesses, how to correctly select
the management strategy have become the key for the healthy and rapid devel-
opment of SMEs in the future.
Acknowledgment This research is supported by the Foundation of Guangxi Educational
Committee (Grant No: 201012MS121).
1658 X. Zhu and Y. Li
References
Almeida P (1996) Knowledge sourcing by foreign multinationals: patent citation analysis in the
US semiconductor industry. Strateg Manag J 17:155–165
Brandenburger AM, Nalebuff BJ (1995) Theright game: use game theory to shape strategy. Harv
Bus Rev 73(4):57–71
Chang S, Singh H (2000) Corporate and industry effects on business unit competitive position.
Strateg Manag J 21:739–752
Ghemawat P (2003) The incomplete integration of markets, location-specificity, and international
business strategy. J Int Bus Stud 34:138–152
Glimstedt H, Lazonick W, Xie H (2006) Evolution and allocation of stock options: adapting
US-style compensation to the swedish business model. Eur Manag Rev 3:1–21
Hedlund G (2007) A model of knowledge management and the N-form corporation. Strateg
Manag J 15(3):73–91
Hennart JF (2001) Theories of multinational enterprise. In: Rugman AM, Brewer TL (eds), The
Oxford handbook of international business. Oxford University Press, New York, pp 127–149
Markides C (1999) A dynamic view of strategy. Sloan Manag Rev 40:55–63
Mulcaster WR (2009) Three strategic frameworks. Bus Strategy Ser 10:68–75
Rugman AM, Verbeke A (2001) Subsidiary-specific advantages in multinational enterprises.
Strateg Manag J 22(3):237–250
Scott AJ (2006) Entrepreneurship, innovation and industrial development: geography and the
creative field revisited. Small Bus Econ 26:1–24
Williamson OE (1999) Strategy research: governance and competence perspectives. Strateg
Manag J 20(12):1087–1108
You might also like
- Disruptive Innovation in Mining Industry 4.0No ratings yetDisruptive Innovation in Mining Industry 4.0841 pages
- A Review of Institutional Theory and EntrepreneurshipNo ratings yetA Review of Institutional Theory and Entrepreneurship1,388 pages
- Applications of Mathematical Modeling, Machine Learning, and Intelligent Computing For Industrial Development (Madhu Jain, Dinesh K Sharma, Rakhee Kulshrestha Etc.) (Z-Library)No ratings yetApplications of Mathematical Modeling, Machine Learning, and Intelligent Computing For Industrial Development (Madhu Jain, Dinesh K Sharma, Rakhee Kulshrestha Etc.) (Z-Library)425 pages
- Fuzzy Hierarchical Model For Risk Assessment - Principles, Concepts, and Practical ApplicationNo ratings yetFuzzy Hierarchical Model For Risk Assessment - Principles, Concepts, and Practical Application171 pages
- Dynamic and Advanced Data Mining For Progressing Technological Development - Innovations and Systemic Approaches (Ali & Xiang 2009-11-25)No ratings yetDynamic and Advanced Data Mining For Progressing Technological Development - Innovations and Systemic Approaches (Ali & Xiang 2009-11-25)516 pages
- Predicting Human Decision-Making From Prediction To Action100% (1)Predicting Human Decision-Making From Prediction To Action152 pages
- Analytics in Finance and Risk Management (Nga Thi Hong Nguyen)No ratings yetAnalytics in Finance and Risk Management (Nga Thi Hong Nguyen)396 pages
- Zhu Joe - Quantitative Models For Performance Evaluation and Benchmarking. Data Envelopment Analysis With Spreadsheets - 2008 PDFNo ratings yetZhu Joe - Quantitative Models For Performance Evaluation and Benchmarking. Data Envelopment Analysis With Spreadsheets - 2008 PDF274 pages
- Fuzzy Information & Engineering and Operations Research & ManagementNo ratings yetFuzzy Information & Engineering and Operations Research & Management558 pages
- Business Analytics With R and Python (David L. Olson, Desheng Dash Wu, Cuicui Luo Etc.) (Z-Library)100% (1)Business Analytics With R and Python (David L. Olson, Desheng Dash Wu, Cuicui Luo Etc.) (Z-Library)201 pages
- Pub - Handbook of Computational Intelligence in Manufact PDF0% (1)Pub - Handbook of Computational Intelligence in Manufact PDF516 pages
- 2015 - Guarnieri - Decision Models in Engineering and Management100% (2)2015 - Guarnieri - Decision Models in Engineering and Management322 pages
- Transfer Pricing (Part 1) - TP Key Principles Analysis and MethodsNo ratings yetTransfer Pricing (Part 1) - TP Key Principles Analysis and Methods76 pages
- (2022) (Sindhwani Et Al) Multi-Criteria Decision Modelling - Applicational Techniques and Case StudiesNo ratings yet(2022) (Sindhwani Et Al) Multi-Criteria Decision Modelling - Applicational Techniques and Case Studies223 pages
- An Introduction To Management Science Quantitative Approach 16th Edition PDFNo ratings yetAn Introduction To Management Science Quantitative Approach 16th Edition PDF12 pages
- Computational Intelligence In9ComplexDecision100% (2)Computational Intelligence In9ComplexDecision398 pages
- Models, Methods, Concepts SC Applications of The Analytic Hierarchy ProcessNo ratings yetModels, Methods, Concepts SC Applications of The Analytic Hierarchy Process334 pages
- Management and Intelligent Decisionmaking in Complex Systems An Optimizationdriven Approach Ameer Hamza Khan DownloadNo ratings yetManagement and Intelligent Decisionmaking in Complex Systems An Optimizationdriven Approach Ameer Hamza Khan Download36 pages
- Vincent Charles, Juan Aparicio, Joe Zhu - Data Science and Productivity Analytics-Springer International Publishing - Springer (202No ratings yetVincent Charles, Juan Aparicio, Joe Zhu - Data Science and Productivity Analytics-Springer International Publishing - Springer (202441 pages
- Production Planning and Control in Semiconductor ManufacturingNo ratings yetProduction Planning and Control in Semiconductor Manufacturing106 pages
- Industrial Engineering in The Big Data e 1No ratings yetIndustrial Engineering in The Big Data e 1500 pages
- Journal Pre-Proof: International Journal of Production EconomicsNo ratings yetJournal Pre-Proof: International Journal of Production Economics42 pages
- Computational Intelligence Paradigms Inn PDFNo ratings yetComputational Intelligence Paradigms Inn PDF280 pages
- Computational Models of Complex SystemsNo ratings yetComputational Models of Complex Systems199 pages
- Intelligent Soft Computation and Evolving Data Mining Integrating Advanced Technologies Premier Reference Source First Edition Leon Shyue-Liang Wang Instant DownloadNo ratings yetIntelligent Soft Computation and Evolving Data Mining Integrating Advanced Technologies Premier Reference Source First Edition Leon Shyue-Liang Wang Instant Download52 pages
- Big Data, Cloud Computing, and Data Science Engineering PDF100% (1)Big Data, Cloud Computing, and Data Science Engineering PDF222 pages
- Data Science and Productivity Analytics - UnknownNo ratings yetData Science and Productivity Analytics - Unknown564 pages
- Industrial Engineering Operation Management100% (1)Industrial Engineering Operation Management271 pages
- Advanced Mathematical Techniques in Science and Engineering PDFNo ratings yetAdvanced Mathematical Techniques in Science and Engineering PDF250 pages
- Operations Management and Systems EngineeringNo ratings yetOperations Management and Systems Engineering253 pages
- Data Science and Digital Transformation in The Fourth Industrial Revolution-Springer (2021)No ratings yetData Science and Digital Transformation in The Fourth Industrial Revolution-Springer (2021)234 pages
- Proceedings of International Conference On Recent Trends in Machine Learning, Iot, Smart Cities and ApplicationsNo ratings yetProceedings of International Conference On Recent Trends in Machine Learning, Iot, Smart Cities and Applications969 pages
- Mathematical Methods For Maintenance and Operation Cost Prediction Based On Transfer Learning in State GridNo ratings yetMathematical Methods For Maintenance and Operation Cost Prediction Based On Transfer Learning in State Grid17 pages
- Industrial Engineering in The Industry 4.0 Era: Fethi Calisir Hatice Camgoz Akdag Editors100% (4)Industrial Engineering in The Industry 4.0 Era: Fethi Calisir Hatice Camgoz Akdag Editors276 pages
- Data Mining - Concepts, Methods and Applications in Management and Engineering DesignNo ratings yetData Mining - Concepts, Methods and Applications in Management and Engineering Design328 pages
- Elements of Manufacturing, Distribution and Logistics QuantitativeNo ratings yetElements of Manufacturing, Distribution and Logistics Quantitative308 pages
- (Sandra Huber, Martin Josef Geiger, Adiel TeixeiraNo ratings yet(Sandra Huber, Martin Josef Geiger, Adiel Teixeira317 pages
- (Lecture Notes in Management and Industrial Engineering) Fethi Calisir Emr100% (2)(Lecture Notes in Management and Industrial Engineering) Fethi Calisir Emr500 pages
- Sensitivity Analysis in Simplex Method - GJS0% (1)Sensitivity Analysis in Simplex Method - GJS24 pages
- Models Methods For Project Selection: Concepts From Management Science, Finance and Information TechnologyNo ratings yetModels Methods For Project Selection: Concepts From Management Science, Finance and Information Technology15 pages
- Big Data Analytics For Intelligent Manufacturing Systems A ReviewNo ratings yetBig Data Analytics For Intelligent Manufacturing Systems A Review15 pages
- 2015 Book ResearchAdvancesInIndustrialEnNo ratings yet2015 Book ResearchAdvancesInIndustrialEn122 pages
- Expert Systems With Applications: Massimo Bertolini, Davide Mezzogori, Mattia Neroni, Francesco ZammoriNo ratings yetExpert Systems With Applications: Massimo Bertolini, Davide Mezzogori, Mattia Neroni, Francesco Zammori29 pages
- Industrial Revolution of The 21st CenturyNo ratings yetIndustrial Revolution of The 21st Century74 pages
- The Analytic Hierarchy Process: Planning, Priority. Setting, Resource AllocationNo ratings yetThe Analytic Hierarchy Process: Planning, Priority. Setting, Resource Allocation2 pages
- Mental Models I Find Repeatedly Useful - by Gabriel Weinberg - MediumNo ratings yetMental Models I Find Repeatedly Useful - by Gabriel Weinberg - Medium30 pages
- GROUP 3 Business Plan Chapter 1 and 2 PDFNo ratings yetGROUP 3 Business Plan Chapter 1 and 2 PDF9 pages
- Tết - Vietnamese Lunar New Year by SlidesgoNo ratings yetTết - Vietnamese Lunar New Year by Slidesgo61 pages
- Coffee Board, Karnataka, ... Vs Commissioner of Commercial Taxes On 11 May, 1988No ratings yetCoffee Board, Karnataka, ... Vs Commissioner of Commercial Taxes On 11 May, 198824 pages
- UGRD-REM6251-2027LITET Midterm Quiz 1 - Attempt ReviewNo ratings yetUGRD-REM6251-2027LITET Midterm Quiz 1 - Attempt Review6 pages
- Financial Aspects of Marketing ManagementNo ratings yetFinancial Aspects of Marketing Management10 pages
- Securing ChatGPT: Best Practices for Protecting Sensitive Data in AI Language ModelsFrom EverandSecuring ChatGPT: Best Practices for Protecting Sensitive Data in AI Language ModelsNo ratings yet
