PERFORMANCE BASED GROUPING AND FRAGILITY

ANALYSIS OF BOX-GIRDER BRIDGES IN CALIFORNIA

A Dissertation
Presented to
The Academic Faculty

by

Sujith Mangalathu

In Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy in the
School of Civil and Environmental Engineering

Georgia Institute of Technology

May 2017

COPYRIGHT © 2017 BY SUJITH MANGALATHU

 PERFORMANCE BASED GROUPING AND FRAGILITY
ANALYSIS OF BOX-GIRDER BRIDGES IN CALIFORNIA

Approved by:

Dr. Reginald DesRoches, Advisor Dr. Jamie E. Padgett, Co-advisor

School of Civil and Environmental Department of Civil and Environmental
Engineering Engineering
Georgia Institute of Technology Rice University

Dr. Brani Vidakovic Dr. Lauren Stewart

School of Industrial and Systems School of Civil and Environmental
Engineering Engineering
Georgia Institute of Technology Georgia Institute of Technology

Dr. Iris Tien

School of Civil and Environmental
Engineering
Georgia Institute of Technology

Date Approved: March 13, 2017

 Dedication

To those who welcomed me with a warm smile during my journey

&

To those who are working hard to make the world a better place
 ACKNOWLEDGEMENTS

This dissertation has been made with the help of many individuals and I am grateful for
their contribution and support. I thank my primary advisor Prof. Reginald DesRoches, who
granted me the privilege to work in his research group. I am grateful to him for the
discussions and advice that helped me sort out the technical details of my work. I also thank
my co-advisor Prof. Jamie Padgett for her comments and suggestions.

I would like to thank the support I have received from the California Department of
Transportation to pursue my doctoral studies. I would like to express my sincere gratitude
to Cliff Roblee, for his time and patience in providing insight regarding the bridge
inventory in California

I would like to thank my thesis committee Dr. Brani Vidakovic, Dr. Lauren Stewart and
Dr. Iris Tien. My special thanks to Dr. Vidakovic considering the many hours I spent in his
office and his lectures discussing probability theory and statistics. With his enthusiasm, his
inspiration, and his great efforts to explain things clearly and simply, he helped to make
statistics fun for me. I also thank Prof. Leonardo Osario (Rice University) for his
professional guidance. He provided encouragement, sound advice, good company, and lots
of good ideas.

My stay at Georgia Tech was a journey of learning and inspiration. The technical
discussion I had with fellow students Farahnaz Soleimani, Parsa Banihashemi, Edwin Lim,
Stephen Hsu, and Liu Xi helped me in the dissemination of knowledge in the subject and I
thank them for sharing their time. I will always remember the great conversations we had
over lunch, and I know that we will break bread together many more times. I also wish to
acknowledge Ellen Cormack for her motherly care and advice. Her door is always open for
me, and she always has time for my concerns.

I would like to thank the former research members of Prof. DesRoches group – Dr. Karthik
Ramanathan and Dr. Jong-Su Jeon. I admire the patience of Dr. Jeon in teaching me the
OpenSees and bridge design philosophies. I appreciate his help in writing research papers,
sharing his knowledge, and his active interest in my research. I also thank my friend Jiqing
Jiang for her support and discussions. I also thank her for helping me to not to lose
confidence and pointing the need for hard work.

I thank my friends Brian Terranova, Ni Pengpeng, Adris Ayala, Yasin Ogras, Jenni Tipler,
Yesudas, Gokulnath, Ramanandan, Ashkan, Anver Hisham, Jayakrishnan, Andrea,
Dharanidharan, Kiran, Anoob, Yesudas, Prabhash, Nikhil, Gayathri for all the
encouragement and support. I believe that I am tremendously fortunate to establish the
friendship with these outstanding people. I appreciate Brian, Ni, and Gokulnath for helping
me get through the difficult times, and for all the emotional support, entertainment, and
caring they provided. I am not sure I would be at Georgia Tech without their
encouragement.

I also thank my friends and mentors Dr. Vipin Unnithan, Dr. Robin Davis and Shinto Paul
for their inspiration and guidance. The conversations, advice, friendships, and fun helped
shape my development as a researcher and a good human being.

I thank my non-Georgia tech friends at Atlanta, Karthika, Arun, Vinod, Indu, Gireesh,
Kutty, Mahesh and Subin for the countless fun and activities. They helped me in numerous
ways which are impossible to describe. It’s not easy to forget the hiking trips and poker
games we had over weekends.

I am thankful to my mother Valsala Kumari, my father Sivasubramanian Pillai and my

sister Suchithra MS for their unconditional love, encouragement, and support. No words to
explain their sacrifices and admiration. Their support has been unconditional all these
years; they have given up many things for me to be at Georgia Tech. I also extend the
thanks to my entire family for providing a loving environment for me.

I would be remiss if I didn’t acknowledge the contribution and support I have received
from Surya. Over the years she has expressed confidence in my abilities and supported me
unconditionally. My love and appreciation is extended to her.

I am deeply indebted towards the movies for providing me a recreational and artistic mind
during my Ph.D, particularly the movies of Christopher Nolan, Martin Scorsese, Giuseppe
Tornatore, Majid Majidi, Ki-Duk Kim, Asghar Farhadi, Ang Lee and Clint Eastwood. The
movies - The Lives of Others, Life is Beautiful, Cinema Paradiso, Hotel Rwanda, The
Prestige - has always a special place in my heart.

I also acknowledge the legendary soccer players Pele, Maradona and Messi for entertaining
and motivating me since childhood. I also acknowledge my soccer colleagues at Georgia
Tech.

I also extend my thanks to the dedicated and silent service of many at Georgia Tech
(Georgia Tech Police, CRC Staff, administrative staff, etc.). Their service helps my stay at
Georgia Tech a nice and pleasant one.

I also appreciate the dedicated and patient service of people around the globe to make the
world a better place.

As the adage goes - like any journey, it's not what you carry but what you leave behind.
 Disclaimer

The current study has been supported by the California Department of Transportation

(Caltrans), USA through Project P266, Task 1780: Production development of generation-

2 fragility models for California bridges. Any opinions, findings, and conclusions or

recommendations expressed in this thesis are those of the author and do not necessarily

reflect the views of the Caltrans.

 TABLE OF CONTENTS

ACKNOWLEDGEMENTS iv
LIST OF TABLES vii
LIST OF FIGURES ix
LIST OF SYMBOLS AND ABBREVIATIONS xi
SUMMARY xv
CHAPTER 1 INTRODUCTION 1
1.1 Problem Description and Motivation 1
1.2 Research Objectives 5
1.3 Dissertation outline 6
CHAPTER 2 EXISTING RESEARCH ON GROUPING OF BRIDGE CLASSES
AND BRIDGE FRAGILITY 8
2.1 Grouping of bridge classes 9
2.2 Fragility curves 12
2.2.1 Expert opinion 12
2.2.2 Empirical methods 12
2.2.3 Analytical methods 13
2.2.4 Parameterized fragility curves 16
2.3 Fragility curves for concrete bridge classes in California 19
2.4 The need to go beyond HAZUS-based grouping and fragility curves 22
2.5 Uncertainty treatment in fragility analysis 27
2.6 Closure 32
CHAPTER 3 MODELING OF BRIDGE COMPONENTS 34
3.1 Superstructure 35
3.2 Substructure 36
3.2.1 Bents 36
3.2.2 Columns 37
3.2.3 Idealization of bridge columns 43
3.2.4 Validation of bridge columns 47
3.3 Abutments 47
3.4 Bearings 52
3.5 Shear Keys 53
3.6 Pounding 54
3.7 Foundation 55
3.8 Closure 57
CHAPTER 4 PERFORMANCE BASED GROUPING OF BRIDGE CLASSES 59
4.1 Review of ANOVA, ANCOVA, and Kruskal–Wallis approach 60
4.2 ANOVA based grouping 61
4.3 ANCOVA based grouping 63
4.4 KW-based grouping 65
4.5 Case Study: Two-and Three-Span Box-Girder Bridges in California 66

iv
4.6 Comparison of various grouping techniques 72
4.6.1 Case 1: Significant per ANCOVA 72
4.6.2 Case 2: Significant per ANOVA and ANCOVA 75
4.6.3 Case 3: Significant per KW 76
4.7 Grouping of bridge classes 79
4.8 Conclusion 86
CHAPTER 5 CALIFORNIA BRIDGE INVENTORY 88
5.1 Bridge classification based on BIRIS 89
5.2 Box girder bridge class statistics 89
5.3 Abutments 90
5.4 Bearings 95
5.5 Box girder deck 96
5.5.1 Span length 96
5.5.2 Deck width 97
5.5.3 Deck cross-section properties 100
5.6 Columns 102
5.6.1 Column height 102
5.6.2 Column cross-section 103
5.6.3 Column material properties 106
5.6.4 Column reinforcement details 106
5.7 Foundations 107
5.8 Other uncertain parameters 111
5.8.1 Damping 111
5.8.2 Mass factor 111
5.8.3 Shear key acceleration 111
5.8.4 Gap 112
5.8.5 Earthquake direction 112
5.9 Closure 112
CHAPTER 6 SYSTEM AND COMPONENT FRAGILITY CURVES FOR BOX-
GIRDER BRIDGES 114
6.1 Fragility Framework 114
6.1.1 Probabilistic seismic demand models 116
6.1.2 Capacity models 117
6.1.3 Abutments 128
6.2 Fragility methodology 130
6.2.1 Ground motion suite 133
6.2.2 Material and geometric uncertainties and parameterized stochastic bridge
models 134
6.3 Fragility curves for multi-span continuous concrete single frame box-girder
bridges 134
6.3.1 Trends based on design era 140
6.3.2 Trends based on spans 142
6.3.3 Trends based on abutment type 143
6.3.4 Trends based on column cross-section 144
6.3.5 Trends based on number of columns per bent 146
6.4 HAZUS comparison 147

v
6.5 Closure 152
CHAPTER 7 PARAMETERIZED FRAGILITY CURVES: LASSO APPROACH
154
7.1 Regression models 156
7.1.1 Linear Regression 156
7.1.2 Stepwise regression 157
7.1.3 Ridge regression 157
7.1.4 Lasso regression 158
7.1.5 Elastic net 159
7.2 Case-study bridges: numerical modeling, uncertainties, ground motion suite,
and demand parameters 159
7.3 Comparison of the regression models 163
7.3.1 Investigation of penalty factor 164
7.3.2 Comparison of the regression models 167
7.4 Sensitivity of input parameters to the seismic demand model 170
7.5 Multi-Parameter fragility curves 172
7.6 Conclusions 179
CHAPTER 8 CONCLUSIONS AND FUTURE WORK 181
8.1 Summary and Conclusions 181
8.2 Research Impact 185
8.3 Recommendations for future work 187
APPENDIX A. EARTHQUAKE RECORDS USED FOR FRAGILITY ANALYSIS
189
APPENDIX B. FRAGILITY CURVES IN TERMS OF PGA 207
APPENDIX C. COMPONENT FRAGILITY CURVES FOR BRIDGE CLASSES
210
REFERENCES 230

vi
 LIST OF TABLES

Table 2.1 HAZUS grouping and fragility relationships for bridge classes in California. 23
Table 3.1 – Idealization of bridge columns. 44
Table 4.1 – Uncertainty distribution considered in the bridge models 70
Table 4.2 – p–values by ANCOVA, ANOVA and KW test 71
Table 4.3 – p–values from ANOVA. 80
Table 4.4 – Results of the grouping for two-span box girder bridges. 82
Table 4.5 – Results of the grouping for multi-span bridges 83
Table 5.1 – Bridge classes in California and their proportion in the overall inventory. 89
Table 5.2 – Distribution of parameters for abutments. 94
Table 5.3 – Distribution of parameters for bearings. 95
Table 5.4 – Distribution of span length and span ratio (approach span/main span) for box
girder bridges. 98
Table 5.5 – Distribution of deck width for box girder bridges. 99
Table 5.6 – Deck cross-section properties. 100
Table 5.7 – Box-girder deck slab thickness (MTD, 2008). 101
Table 5.8 – Column height distribution. 103
Table 5.9 – Distribution of column cross-sections. 104
Table 5.10 – Statistical distribution of column material properties. 106
Table 5.11 – Statistical distribution of column reinforcement details. 107
Table 5.12 – Distribution of foundation rotational stiffness (10 kip-in/rad).
6
109
Table 5.13 – Distribution of foundation translational stiffness (kip/in). 110
Table 5.14 – Distribution of other uncertain parameters. 112
Table 6.1 – Engineering demand parameters for bridge components monitored in NLTHA.
116
Table 6.2– Component level damage state descriptions – Component Damage Thresholds
(CDT). 118
Table 6.3 – General description of BSSTs along with CDTs. 119
Table 6.4 – General definition of column capacity limit states. 120
Table 6.5 – Design details for columns in Era 11 (pre-1971). 122
Table 6.6 – Design details of columns in Era 22 (1971-1990). 123
Table 6.7 – Design details for columns in Era 33 (post-1990). 124
Table 6.8 – Summary of limit states for Era 11 columns. 125
Table 6.9 – Summary of limit states for Era 22 columns. 126
Table 6.10 – Summary of limit states for Era 33 columns. 127
Table 6.11– Statistical summary of ductility values for various design eras. 128
Table 6.12– Median value of CDT for abutment seat. 129
Table 6.13– Summary of CDT values for various bridge components. 130
Table 6.14 – Nomenclature adopted in the current study. 135
Table 6.15 – Fragility values for two span continuous concrete box-girder fragilities with
diaphragm abutments. 137
Table 6.16 – Fragility values for two span continuous concrete box-girder fragilities with
seat abutments. 137

vii
Table 6.17 – Fragility values for multi-span continuous concrete box-girder fragilities with
diaphragm abutments. 138
Table 6.18 – Fragility values for multi-span continuous concrete box-girder fragilities with
seat abutments. 138
Table 6.19 – Comparison of bridge classes. 148
Table 7.1 – Uncertainty distribution considered in the bridge models. 162
Table 7.2 – Bridge component demand parameters. 163
Table 7.3 – Estimated coefficients and test error for COL for the bridge with diaphragm
abutments by various regression techniques 168
Table 7.4 – Limit state models of various bridge components. 175

viii
 LIST OF FIGURES

Figure 2.1 – PEER framework for PBEE (Porter, 2003). 8

Figure 2.2 – HAZUS grouping of California box-girder bridge inventory. 10
Figure 2.3 – Generation of PSDMs through cloud approach (Ramanathan, 2012). 15
Figure 2.4 – Generation of PSDMs through IDA approach (Ramanathan, 2012). 16
Figure 2.5– Uncertainty sources for system demand and capacity (Ji et al., 2007). 28
Figure 3.1 – Illustration of major bridge components. 34
Figure 3.2 – Numerical modeling of various bridge components. 36
Figure 3.3 – Finite element discretization of bents. 37
Figure 3.4 – Fiber based discretization of the columns. 38
Figure 3.5 – Effectively confined core for a) circular hoop reinforcement and, b)
rectangular hoop reinforcement (Mander et al. 1988). 39
Figure 3.6 – Typical cross-sections noted from the bridge plan review (Caltrans,
2017). 41
Figure 3.7 – Circular confined core concrete sections. 41
Figure 3.8 – Typical rectangular cross sections. 42
Figure 3.9 – Oblong columns. 42
Figure 3.10 – Comparison of experimental and numerical results for era 33 columns
(Lehman and Moehle, 2000) a) Specimen No. 415 b) Specimen No. 815. 47
Figure 3.11 – Various types of abutments. 48
Figure 3.12 – Unusual abutment types (Caltrans, 2017). 49
Figure 3.13 – Modeling of the abutments. 51
Figure 3.14 – Types of bearings: (a) rocker bearing and (b) elastomeric bearing
(Mangalathu et al. 2016). 52
Figure 3.15 – Modeling of various bearings. 53
Figure 3.16 – Modeling of shear key. 53
Figure 3.17 – Pounding damage in bridges during the 1994 Northridge earthquake: (a)
barrier rail damage and, (b) connector collapse (Muthukumar and DesRoches, 2006).
54
Figure 3.18 – Analytical Model for pounding between deck and abutment back wall.
55
Figure 3.19 – Bridge foundation types (Priestley et al. 1996). 56
Figure 3.20 – Modelling of foundations. 57
Figure 4.1 – Illustration of PSDM and grouping strategy. 61
Figure 4.2 – ANOVA hypothesis. 62
Figure 4.3 – General layout of a two–span concrete box–girder bridge. 67
Figure 4.4 – Histograms and descriptive statistics for case 1: a) 3SDC, b) 3SDR, c)
box plot of 3SDC and 3SDR, and d) ANCOVA regression lines of 3SDC and 3SDR.
73
Figure 4.5 – Data analysis for case 2: a) histogram of 2SSC, b) histogram of 2SSR, c)
box plot of 2SSC and 2SSR, and d) ANCOVA regression lines of 2SSC and 2SSR. 75
Figure 4.6 – Data analysis for case 3: a) histogram of 2SSS, b) histogram of 2SSP, c)
box plot of 2SSS and 2SSP, and d) ANCOVA regression lines of 2SSS and 2SSP. 76
Figure 4.7 – a) Histogram of the PGA values of the ground motion suite, b)
Acceleration response spectrum of the ground motion suite. 79

ix
Figure 4.8 – Proposed classification scheme. 83
Figure 5.1 – Illustration of major bridge components. 90
Figure 5.2 – Distribution of abutments for various eras. 91
Figure 5.3 – Distribution of abutments for various bridge eras. 92
Figure 5.4 – Percentage distribution of bearings based on design eras. 95
Figure 5.5 – Cross-section details of box-girder bridges. 100
Figure 5.6 – Column cross-sections for various design eras of box girder bridges. 105
Figure 5.7 – Bridge foundation types (Priestley et al. 1996). 107
Figure 5.8 – Statistical distribution of foundation for various eras. 108
Figure 6.1 – Schematic representation of the NLTHA procedure used to derive the
PSDMs. 115
Figure 6.2 – Illustration of a typical PDSM. 117
Figure 6.3 – Schematic of the fragility framework. 132
Figure 6.4 – Response spectra for the selected ground motions. 133
Figure 6.5 – Illustration of change in median value and relative vulnerability. 136
Figure 6.6 – System and component fragility curves for bridge classes S-E1-S22-C-D
and S-E2-S22-C-S. 140
Figure 6.7 – Plot of median values of bridge classes across the design eras for a) S-
E(1/2/3)-S22-C-D, and b) T-E(1/2/3)-S34-C-S. 141
Figure 6.8 – Plot of median values of bridge classes across the number of spans for
bridge classes S-E2-S(22/34)-C-D and S-E2-S(22/34)-O-D. 143
Figure 6.9 – Plot of median values of bridge classes across the type of abutments for
the selected bridge classes. 143
Figure 6.10 – Plot of median values of bridge classes across various column cross-
sections for the selected bridge classes. 145
Figure 6.11 – Plot of median values of bridge classes across the number of columns
per bent for the selected bridge classes with a) diaphragm abutments and b) seat
abutments. 147
Figure 6.12– Comparison of HAZUS and selected bridge class fragilities for a) Era 11
bridges with single column bent and b) Era 22 and Era 33 bridges with single column
bent. 150
Figure 6.13– Comparison of HAZUS and selected bridge class fragilities a) Era 11
two-span bridges with two- and multi-column bents and b) Era 22 and Era 33 two-
span bridges with single and multi-column bents. 151
Figure 7.1 – Numerical modeling of various bridge components. 161
Figure 7.2 – Comparison of MSE with . 166
Figure 7.3 – Shrinkage of regression coefficients with . 167
Figure 7.4 – Radar plot depicting the comparison of accuracy of fit obtained from the
various regression models. 170
Figure 7.5 – Sensitivity of input parameters. 172
Figure 7.6 – System and component fragility curves for moderate damage state: a)
diaphragm abutment bridge, b) seat abutment bridge. 176
Figure 7.7 – Sensitivity of fragility curves to input parameters for diaphragm abutment
bridge for various limit states 177
Figure 7.8 – Sensitivity of fragility curves to input parameters for seat abutment
bridge for various limit states 178

x
 LIST OF SYMBOLS AND ABBREVIATIONS

ABS absolute error

ABA active abutment displacement
ABP passive abutment displacement
ABT transverse abutment displacement
ANOVA analysis of variance
ANCOVA analysis of covariance
ATC applied technology council
Asp area of transverse reinforcement bar
Asx area of transverse reinforcement bar in x direction
Asx area of transverse reinforcement bar in y direction
a regression coefficient - intercept
as acceleration for shear key capacity
Bc breadth of rectangular/oblong cross-section
Bd deck width
BRD bearing displacement
BSST bridge system damage state
b regression coefficient - slope
C circular cross-section
C structural capacity
CF confinement factor
COL curvature ductility
CSP bridge sub-class with a circular single-column bent and two
abutments on piles
CDF cumulative distribution function
CDT component damage threshold
CSM capacity spectrum method
CSUS central and southeastern United States
Caltrans California department of transportation
D Diaphragm abutment bridge
D structural demand
DEC deck displacement
DV decision variable
DM damage measure
ds diameter of spiral bar centers
ED earthquake direction
EDP engineering demand parameter
Era 11/E1 bridges constructed before 1970
Era 22/E2 bridges constructed between 1970 to 1990
Era 33/E3 bridges constructed after 1970
FNT foundation translation
FNR foundation rotation
fc unconfined strength of concrete

xi
fcc maximum concrete stress
fyh yield strength of the transverse reinforcement
G(EDP|IM) CDF of EDP conditioned on IM
G(DM|EDP) CDF of DM conditioned on EDP
G(DV|DM) CDF of DV conditioned on DM
g acceleration due to gravity
H column height
H0 null hypothesis
H1 alternate hypothesis
Ha abutment height
Hal alternate hypothesis
JPSDM joint probabilistic seismic demand model
IDA incremental dynamic analysis
IM ground motion intensity
KW Kruskal-Wallis test
Kpa stiffness of abutment piles
Kb stiffness of bearing
Kfr rotational stiffness of foundation piles
Kft translational stiffness of foundation piles
ke effective confinement factor
L span length
Lc width of rectangular/oblong cross-section
LN log-normal distribution
M multi column bent
MSE mean square estimate
m mass factor
N total sample size/number of simulations
NLTHA nonlinear time history analysis
NSP nonlinear static procedures
O oblong cross-section
PC pre-stressed concrete
PGA peak ground acceleration
PSDM probabilistic seismic demand model
R rectangular cross-section
RC reinforced concrete
S single column bent
S11 single span bridge
S22 two span bridge
S34 three and four span bridge
S5x bridges with more than five spans
Sa-1.0s spectral acceleration at a period of 1.0 s
Sd median estimate of the demand as a function of the IM
Sc median estimate of the capacity
s center to center spacing or pitch of spiral or circular hoop
SDOF single-degree of freedom
SST total sum of squares

xii
T two column bent
UST superstructure unseating displacement
(•) standard normal cumulative distribution function
 displacement
t gap between the deck and shear key
 regression coefficients
d|IM dispersion of the demand conditioned on the IM
a abutment displacement in active direction
b bearing displacement
d deck displacement
fnd foundation displacement
p abutment displacement in passive direction
seat seat displacement
t abutment displacement in transverse direction
u unseating displacement
key shear key displacement
y yield displacement
cc maximum concrete strain
 penalty factor
 mean of fragility curves
μ mean value
μas coefficient of friction for abutments on spread footing
μb coefficient of friction of bearing pad
μ column curvature ductility
DV mean annual frequency of a decision variable
 longitudinal reinforcement ratio
cc ratio of area of longitudinal reinforcement to the area of core
of section
f foundation rotation
σ standard deviation
 damping ratio
 dispersion of fragility curves
2SSC two span seat-abutment bridge with circular cross section and
single-column bent
2SSR two span seat-abutment bridge with rectangular cross section
and single-column bent
2SSS two span seat-abutment bridge with abutments resting on
spread footing
2SSS two span seat-abutment bridge with abutments resting on
piles
2SDPS two span diaphragm-abutment single-column bent bridge
with pre-stressed concrete superstructure
2SDRS two span diaphragm-abutment single-column bent bridge
with reinforced concrete superstructure

xiii
3SDC three span seat-abutment bridge with circular cross section
and single-column bent
3SDR three span seat-abutment bridge with rectangular cross
section and single-column bent
3SCMCB three span seat-abutment bridge with rectangular cross
section and multi-column bent
3SRMCB three span seat-abutment bridge with circular cross section
multi-column bent bridge

xiv
 SUMMARY

Fragility curves play a critical role in regional seismic risk assessment and are a key

component of tools used to support emergency response and preparedness in California

following an earthquake. To have an accurate assessment of regional damage, it is critical

to provide fragility curves that best represent the bridge inventory. However, it is

impractical to develop unique fragility curves for each structure across a regional portfolio.

One strategy that has been adopted to address this challenge is to group bridges into classes

with similar design or structural performance. Traditionally, this grouping has been

performed based on a relatively subjective identification of sub-classes. However, such an

identification leads to a number of bridge classes and unwarranted grouping. This work

suggests a performance based grouping methodology to group the box-girder bridges in

California, and is the first systematic approach in sub-binning bridge classes for the

regional risk assessment. The proposed grouping and analytical fragility methodology is

used to derive fragility relationships for single frame box girder bridges in California. This

work concludes with the application of machine learning techniques for the generation of

bridge-specific fragility curves.

xv
 CHAPTER 1 INTRODUCTION

1.1 Problem Description and Motivation

Highway bridges constitute the key components of transportation networks, yet

recent earthquakes have demonstrated that these bridges are one of the most vulnerable

components of transportation networks. Damage to bridges from an earthquake can pose a

serious threat to immediate recovery efforts and cause communities to incur large

economic losses. Recovery efforts and loss estimation can typically be calculated via

damage functions of structures called fragility functions (Kircher et al., 2006). A fragility

curve is defined as a conditional probability that gives the likelihood that a structure or

component of that structure will meet or exceed a certain level of damage for a given

ground motion intensity (IM). This enables the realistic estimation of economic losses, as

well as planning for emergency responses and ascertaining the need for retrofitting.

Fragility curves can be generated by empirical (Shinozuka et al., 2000) or analytical

approaches (Basöz and Mander, 1999; Banerjee and Shinozuka, 2007; Gardoni et al., 2003;

Zhong et al., 2008; Nielson 2005; Mangalathu et al. 2015a). Empirical methods have been

used to develop fragility curves in regions where extensive earthquake records are

available. Empirical curves are based on observed damage from previous earthquakes.

These types of fragility curves tend to be the most realistic, but are very specific to a

particular earthquake and structure and thus have limited application (Jeong and Elnashai,

2007). The limitations of the empirical approach motivate the generation of fragility curves

using analytical approaches. Analytical fragility curves are generated using numerical

simulations that account for material, geometric, and ground motion uncertainties.

1
 California is a state with a high seismic hazard and a history of damaging

earthquakes. Various researchers (Mackie and Stojadinovic, 2001; Mackie and

Stojadinović, 2006; Zhang and Huo, 2009) have developed fragility curves for bridges in

California that have proved to be valuable in understanding the behavior and seismic

vulnerability of bridges. However, the developed fragility curves are structure-specific, and

therefore are only beneficial for the risk assessment of a specific bridge (Ramanathan et

al., 2015). Fragility curves proposed by Basöz and Mander (1999) are remotely applicable

to bridge classes in California and are used in HAZUS (HAZUS-MH, 2003). However, the

HAZUS fragility relationships were developed based on simplified two-dimensional

analysis, a limited number of bridge parameters, and damage states based on a limited set

of field observations. The effects of interior supports, framing systems, and design

standards were not addressed in HAZUS fragilities.

This issue was partially addressed in the work of Ramanathan et al. (2015) by

generating fragility curves that are applicable to a portfolio of bridges in California. These

curves considered uncertainties in attributes such as span length, column height, number

of spans, superstructure type, and material properties. However, a detailed review of bridge

plans from in-house databases obtained from the California Department of Transportation

(Caltrans) shows that their study addresses only a specific class of bridges with a specific

column shape (circular), bearing type (elastomeric), and abutment type (abutment on piles).

To approximately cover the entire range of the California bridge inventory, their study

should be extended to various configurations of bridge components and classes. Such an

extension supports regional risk assessment for transportation networks and helps agencies

such as Caltrans plan their emergency management response.

2
 The proposed research strives for the generation of fragility curves for various

classes of box-girder concrete bridges in California. Each class of bridge systems relies on

a combination of configurations of components such as the number of spans, column cross-

section, abutment type, type of interior support (number of columns per bent), and design

era. Such a classification leads to numerous subclasses for a particular bridge type, which

makes it cumbersome to compute the seismic fragility for each subclass. It is also not clear

whether all such combinations can yield distinct bridge performance classes, or which

parameters need to be most critical to the establishment of the distinct subclasses.

Additionally, it is not yet clear whether such a detailed classification would result in a better

refinement of the vulnerability assessment. The initial study is directed towards the

sensitivity of bridge attributes and the configurations on their fragility curves.

Various studies (Saiidi et al., 1996; Jangid, 2004; Nielson and DesRoches, 2006;

Padgett and DesRoches, 2007) have been conducted to evaluate the sensitivity of fragility

curves to various input parameters. However, there is a lack of understanding of the effects

of various bridge attributes on the fragility curves and the necessity of this understanding

to creating groups of bridge classes. Mangalathu et al. (2015b) addressed this issue through

the application of a statistical technique called Analysis of Covariance (ANCOVA) on the

Probabilistic Seismic Demand Model (PSDM). The research carried out underscored the

necessity to create distinct bridge subclasses, as well as the importance of structural

attributes on fragility curves. However, their study is limited to two- and three-span single

frame bridge configurations constructed before 1970. The effects of span, design era,

number of frames, and other attributes have not been addressed in their study. The grouping

methodology suggested by them requires extensive numerical simulations. The limitation

3
of their work is addressed in this study through a simple method based on Analysis of

Variance (ANOVA). ANOVA is a statistical technique to compare the population and

means, and a grouping methodology based on ANOVA is suggested in this current study.

The methodology also helps to identify which bridge attributes significantly impact the

seismic bridge response and hence fragility curves. The proposed approach and the

methodology are relevant and applicable to other bridge systems too.

The proposed grouping methodology classifies California box-girder bridge

inventory Representative bridge systems (called RBS- hereafter) so that it can fairly

represent fairly the entire bridge inventory. Real bridges in the California inventory are

designed to meet site-specific requirements. While most, in a broad engineering sense,

conform well to one of the proposed RBS classes, many real bridges have unique design

features and/or combinations that do not ‘fit’ neatly within the proposed RBS class

definitions. To improve fit (or modeling fidelity), one could add a new RBS class to directly

capture the unique features but only at the cost of increasing the number of RBS classes

requiring analysis. At the extreme, this strategy would lead to the development of bridge-

specific models, which is not warranted and is beyond the scope of this work. The current

study generates fragility curves for the single frame box-girder bridge classes using the

multi-phase analytical approach developed by researchers (Nielson, 2005; Padgett and

DesRoches, 2008b; Ramanathan et al., 2015; Jeon et al., 2015).

The proposed grouping and analytical fragility methodology is used to derive

fragility relationships for single frame box-girder bridges in California. The study is limited

to straight bridges and the effects of skew, curvature, and unbalanced frame will not be

addressed.

4
1.2 Research Objectives

The limitation in the HAZUS grouping of bridge classes and the corresponding

fragilities are identified in the subsequent sections. This research aims to improve the

grouping of bridge classes in California by suggesting a performance-based grouping

methodology. Another objective of the study is to make substantial improvements in the

fragility relationships of box-girder bridges in California based on advanced modeling

techniques and the available experimental data.

Specifically, this research will attempt to:

• Identify whether it is rational to go beyond the existing HAZUS grouping and fragility

relationships.

• Suggest a performance based grouping strategy (instead of more traditional subjective

grouping) to group bridge classes with statistically similar performance and damage

measures. The proposed methodology will account for the effects of design eras, cross-

sections, number of spans, number of frames, abutment types, span continuity, and pier

types in grouping the bridge classes.

• Perform a detailed plan review of various bridge classes in Californian through review

of the in-house database of bridge plans (hereafter, BIRIS) assembled by Caltrans

engineers.

• Use an advanced numerical modeling procedure to capture the seismic response of

various bridge components.

• Generate statistically significant yet nominally identical bridge models accounting for

the various geometric, material, system and ground motion uncertainties.

5
• Generate a refined set of fragility curves at system and component level for various

box-girder bridge classes in California. These improved fragility curves will help

determine the relative vulnerability of various bridge classes. They will also assist the

bridge owners such as Caltrans in spending their resources judiciously in their

development of retrofitting strategies.

• Extend traditional single-parameter fragility curves to multi-parameter fragility curves

using advanced statistical and machine learning techniques such as Lasso, Ridge, and

elastic net. Such study will provide insight in quantifying whether the variation of

uncertain parameters should be explicitly treated or remain neglected; it eliminates

those parameters which have minimal influence on the seismic demand and reduces

unnecessary and exhaustive efforts in statistical sampling.

1.3 Dissertation outline

The research is organized into seven subsequent chapters with the following contents:

Chapter 2 presents an overview of existing literature on the grouping of bridge classes and

the seismic fragility assessment of bridges.

Chapter 3 provides extensive details about the modeling strategies of various bridge

components: superstructure, columns, abutments, foundations, bearings, and shear keys.

The chapter also presents the integration of various component models to generate global

analytical model of the bridge for fragility analysis

Chapter 4 investigates the application of various performance based grouping strategies

such as analysis of variance (ANOVA), Analysis of Covariance (ANCOVA) and Kruskal–

6
Wallis (KW) towards the grouping of structures of similar performance. The chapter also

outlines the grouping of box-girder bridges in California.

Chapter 5 presents an in-depth study of the California bridge inventory using the in-house

database called BIRIS, assembled by Caltrans engineers. The chapter also presents

statistical distributions of various bridge attributes in California.

Chapter 6 discusses the multi-phase framework adopted in the generation of analytical

fragility curves for box girder bridges in California. The chapter also describes the system

and component fragility curves for single frame multi-span box girder bridges in

California. Insights are provided on the relative performance of various bridge classes and

the importance of various design attributes.

Chapter 7 explores the application of regression and machine learning techniques for the

generation of multi-parameter seismic demand models and fragility curves.

Chapter 8 presents the conclusions from the present research, key contributions and the

suggestions for future research.

7
 CHAPTER 2 EXISTING RESEARCH ON GROUPING OF

BRIDGE CLASSES AND BRIDGE FRAGILITY

Probabilistic Performance-based Earthquake Engineering (PBEE)

framework (Cornell and Krawinkler, 2000; Moehle and Deierlein, 2004) has evolved to be

the next-generation framework in risk mitigation decision making for structure and

infrastructure systems. The framework presented by the Pacific Earthquake Engineering

Research Center (PEER) is the widely accepted robust methodology for PBEE. The PEER

framework calculates the performance of a structure in a probabilistic manner by the

rigorous treatment of uncertainties. The underlying approach is shown in Figure 2.1, and

the framework assumes that the performance assessment of components as discrete Markov

process, where the conditional probabilities between parameters are independent (Moehle

and Deierlein, 2004).

Figure 2.1 – PEER framework for PBEE (Porter, 2003).

The methodology expressed in Figure 2.1 can be expressed in a mathematical form

as expressed in Equation 2.1.

 DV ( DV )   G( DV | DM )  dG( DM | EDP . dG( EDP | IM . d ( IM ) (2.1)

8
where DV is the mean annual frequency of a decision variable (DV, e.g., repair cost,

downtime, loss), G(DV|DM) is the cumulative distribution function (CDF) of DV

conditioned on damage measure (DM, e.g., damage to structural or non-structural

components), G(DM|EDP) is the CDF of DM conditioned on engineering demand

parameter (EDP, e.g., curvature ductility, bearing displacement), G(EDP|IM) is the

demand model describing the CDF of EDP conditioned on ground motion intensity

measure (IM, e.g., peak ground acceleration), and (IM) is the seismic hazard model

describing the mean annual frequency of exceeding an IM. The convolution of

G(DM|EDP) and G(EDP|IM) yields fragility curves. A fragility curve can thus be defined

as a conditional probability that gives the likelihood that a structure or component will

meet or exceed a certain level of damage for a given ground motion intensity (IM). The

fragility curves intended for regional risk assessment require grouping of bridge classes

with similar performance during earthquakes.

2.1 Grouping of bridge classes

Any existing bridge has its own structural characteristics due to its location, soil

conditions, geometric and material properties, and construction technique. However,

bridges with similar structural properties are expected to show statistically similar

performance under a given earthquake loading, and so the bridges with similar performance

can be grouped together. The existing literature on the grouping of bridge classes is given

in this section.

HAZUS (HAZUS-MH, 2003) is, by far, the most comprehensive document in

grouping the bridge classes and seismic vulnerability estimation. HAZUS grouped the

bridge classes with similar damage/loss characteristics and suggested fragility relationships

to the grouped bridge classes. HAZUS classified the bridge classes based on seismic

design, number of spans, span length, bent type, span continuity, and span discontinuity,

9
and is shown in Figure 2.2. The HAZUS grouping was based on engineering judgment,

past experience, and expert opinion. The effects of evolution in seismic design philosophy,

column cross-section, and number of frames are not addressed in HAZUS. The limitations

of HAZUS grouping are discussed later in this chapter.

Figure 2.2 – HAZUS grouping of California box-girder bridge inventory.

Moschonas et al. (2008) classified bridges in Greece according to piers (single

column circular, single column rectangular, multi-column, or wall-type), deck type (slab,

box-girder, or simply supported precast-prestressed beams connected through continuous

RC slab) and pier-to-deck connections (monolithic bearings or combination). Although 36

combinations are possible, the authors further reduced the bridge classes to 11 classes

(those with five or more bridges). Based on past earthquake data and the previous research,

Avsar et al. (2011) classified the highway bridges in Turkey that were constructed after

1990. The important structural attributes identified by the authors are span number (single

or multiple), bent (single or multiple), and skew angle (negligible or significant, chosen to

be >30).

Ramanathan et al. (2015) grouped the bridge classes in California based on limited

parameters such as abutment type (diaphragm abutment or seat abutment), number of

columns per bent (single or multiple), superstructure type (box girder, I-girder, T-girder,

10
or slab bridges) and design era (pre-1971, 1971-1990, or post-1990). However, the

grouping was based on engineering judgment and doesn’t consider the effects of number

of spans, span discontinuity, span length, or cross-section. Thus, the grouping covers only

a limited portion of California bridge inventory.

Mangalathu et al. (2016) suggested a performance-based grouping approach using

a statistical technique called Analysis of Covariance (ANCOVA). This technique compares

the probabilistic seismic demand models (PSDMs) of different bridge classes. A PSDM is

defined as the probability distribution of structural demands (D) conditioned on the ground

motion intensity measure (IM). This works presents the first systematic and reliable

methodology for grouping bridge classes for performing regional risk assessments. The

authors demonstrated their grouping methodology through case studies of two-span and

three-span box-girder bridges in California with various design attributes. Their research

showed the importance of binning of bridge classes through the comparison of fragility

curves for bridge classes with different design attributes as identified by the ANCOVA.

However, the focus of the grouping was only on two-span and three-span box-girder

bridges and doesn’t consider the effect of number of spans, foundation type, design era, or

span discontinuity.

Mehr and Zaghi (2016) used analysis of variance (ANOVA) to compare the

response of single-frame and multi-frame bridges and group them accordingly. The authors

investigated the effects of number of frames, soil type, substructure system, valley shape,

intensity of ground motion, and design capacity-to-demand ratio on single and multi-frame

bridges. Using the ANOVA results, the authors concluded that a multi-frame system is

more robust than a single frame system from a seismic perspective. The authors also noted

that the application of ANOVA is yet to be advanced in civil engineering.

11
 The existing bridge groupings suffer some limitations and cannot be used to group

the California bridge inventory. Also the structural attributes chosen to classify the bridges

vary depending on the type of bridge, bridge location, and research intention. California

has close to 29,000 bridges, which vary in age. In order to obtain a reliable estimate of the

risk associated with the California bridge inventory, it is crucial to group the entire bridge

class that yields similar performance or suffers similar damage following a seismic event.

2.2 Fragility curves

Past decades have witnessed the development of several fragility curve generation

methodologies; a brief summary of their evolution is given in this section.

2.2.1 Expert opinion

The earliest attempt to develop a fragility curve was based on expert opinion (ATC,

1985). A panel of 42 earthquake engineering experts was asked to provide probability

estimates and the results were presented in the form of damage probability matrices, later

converted to vulnerability functions and restoration curves. However, the technique was

wholly subjective and depended on the number of experts queried. The limitations of

expert-opinion fragility curves, coupled with actual damage data from earthquakes,

motivated the generation of empirical fragility curves (Basoz and Kiremidjian, 1998;

Shinozuka et al. 2000).

2.2.2 Empirical methods

Basoz and Kiremidjian (1998) assembled data regarding damage to bridges from

the 1989 Loma Preita and the 1995 Northridge earthquakes in California and analyzed

them to obtain the relationships between bridge damage and ground motion. The authors

generated the fragility curves afterwards thorough logistic regression techniques.

12
Shinozuka et al. (2000) used bridge damage data from the 1995 Kobe earthquake and used

Maximum Likelihood Method to estimate the parameters of a lognormal probability

distribution describing the fragility curves. These types of fragility curves tend to be the

most realistic, but are very specific to a particular earthquake and structure and thus have

limited application (Jeong and Elnashai, 2007). The limitations of the empirical approach

motivated the generation of fragility curves by analytical approaches.

2.2.3 Analytical methods

Analytical methods can be used to generate fragility curves where earthquake data

is not available. Various researchers have employed analysis techniques, with varying

levels of sophistication: Yu et al. (1991) generated analytical fragility curves of bridges

through Elastic Response Spectrum Analysis of single-degree of freedom (SDOF) models,

and Hwang et al. (2000) extended this approach by quantifying the uncertainties in capacity

and demand assessments. With this advancement in the modeling capabilities, researchers

moved to nonlinear static procedures (NSP). Capacity spectrum method (CSM) and N2

method are the different types of NSP. Developed by Fajfar (2000), N2 method combines

the pushover analysis of a multi-degree of freedom (MDOF) system with the response

spectrum analysis of an equivalent SDOF system. HAZUS uses the fragility relationships

suggested by Mander and Basöz (1999), which are based on CSM. Further details on the

fundamental assumptions and limitation of HAZUS fragilities are given in the next section.

Other researchers resorted to the more reliable but computationally expensive

nonlinear time history analysis (NLTHA) (Banerjee and Shinozuka, 2007; Choi, 2002;

Gardoni et al., 2003; Kim and Shinozuka, 2004; Mackie and Stojadinovic, 2001; Mackie

and Stojadinović, 2005; Mangalathu et al., 2016a; Mangalathu et al., 2015; Padgett, 2007;

Ramanathan et al., 2015). Recalling the fragility curve as the probability that the seismic

demand (D) placed on a component exceeds the capacity (C), the probability can be

13
computed using Equation 2.1, assuming lognormal distribution for the D and C ( Cornell

et al., 2002)

 ln( S / S ) 
P[ D  C | IM]    d c
 (2.1)
  d / IM   c2
2


where Sd is the median estimate of the demand as a function of the IM, Sc is the

median estimate of the capacity, d|IM is the dispersion of the demand conditioned on the

IM, c is the dispersion of the capacity, and (•) is the standard normal cumulative

distribution function. Sd and d|IM can be estimated from the PSDMs. As previously

mentioned, PSDM is defined as the probability distribution of structural demands (D)

conditioned on the ground motion intensity measure (IM). NLTHA employs analysis of

bridges with different ground motion intensities to obtain the PSDMs. PSDMs can be

generated by a cloud approach (Figure 2.3) or an incremental dynamic analysis approach

(Figure 2.4). The cloud approach consists of selecting ground motions that represent the

hazard at a region and carrying out NLTHA on the bridge samples. This technique is

limited because it makes a prior assumption about the PSDM.

In the IDA approach, ground motions are scaled successively until significant

reduction (collapse) of the primary load-bearing elements in the structural system. Hence,

IDA can offer the transition of the structural response from elastic to inelastic behavior,

finally leading to global dynamic instability, and the accurate and reliable estimates of the

global collapse capacity of the structure. The overall formulation of IDA was proposed by

(Vamvatsikos and Cornell, 2002). A significant drawback with this approach is that the

process involves scaling of ground motions without altering the frequency content of the

ground motions. These scaling approaches could lead to unrealistic time histories that

might not be representative of the seismic hazard of the site.

14
Figure 2.3 – Generation of PSDMs through cloud approach (Ramanathan, 2012).

15
 Figure 2.4 – Generation of PSDMs through IDA approach (Ramanathan, 2012).

2.2.4 Parameterized fragility curves

As stated in Ghosh et al. (2013), single-parameter demand models and fragility

curves have some limitations: (1) the inability to account for the influence of uncertainty

16
(modeling) parameters on structural performance during earthquakes without extensive re-

simulations for each new set of parameter combinations; (2) the inability to explicitly

address the effect of uncertainty parameters on fragility curves; and (3) the inability to

incorporate field instrumentation data resulting from monitoring of highway bridges to

enable updating of fragility estimates. To alleviate the limitations of single-parameter

fragility curves, recent research has been focused on multi-parameter parameterized

fragility curves that can account for the variation in the design details or geometric

parameters of bridges (Seo and Linzell, 2012; Dukes, 2013; Ghosh et al., 2013; Kameshwar

and Padgett, 2014; Park and Towashiraporn 2014; Jeon et al., 2015; Mangalathu et al.,

2015; Mangalathu et al. 2017c). Assuming that the input variables are statistically

independent, a multi-parameter demand model of each bridge component (demand

parameter) is constructed. Samples obtained from this demand model are compared with

those of the associated limit state model to obtain the binary survival-failure vector. This

vector is used to perform a logistic regression analysis to determine the regression

coefficients and thus develop the multi-parameter fragility curve in the component.

Seo and Linzell (2013) used response surface models to generate parameterized

fragility curves for curved steel bridges. The authors identified the critical range of the

important bridge components using a statistical screening approach.

Dukes (2013) developed multi-parameter demand models of bridge components for

two example bridges using the response surface method and then derived parameterized

fragility curves for the bridges using logistic regression. The parameterized fragility curves

were used to produce bridge-specific fragilities by substituting a specific value for each of

six design parameters: longitudinal reinforcement ratio; volumetric transverse ratio; aspect

ratio; span length-to-column height ratio; deck depth-to-column diameter ratio; and deck

width. This framework was developed for use in the seismic design process in the design

17
of new bridges. It produces fragility curves without the need to create the curves

deterministically with new simulations for each new bridge design.

Ghosh et al. (2013) used a multi-parameter demand model to account for the effect

of all uncertain parameters in the generation of fragility curves. The authors used four

surrogate modeling techniques (polynomial response surface models, multivariate adaptive

regression splines, radial basis function networks, and support vector machines) to

determine the best-fitting parameterized demand models involving the uncertain input

parameters. To achieve this goal, these authors selected for their case study multi-span

simply supported concrete bridges that were not seismically designed, which are typical in

the central and southeastern United States (CSUS). They then used ten parameters

associated with material and geometric uncertainties, along with an IM (eleven predictor

variables), to develop demand models. This work concluded that the MARS model

provided the most accurate estimates of component responses with the fewest predictive

errors. Using the MARS model and logistic regression, parameterized fragility curves were

developed for the component and system level.

Kameshwar and Padgett (2014) suggested a parameterized fragility based multi-

hazard risk assessment for highway bridges subjected to earthquake and hurricane events.

The authors used stepwise logistic regression with a non-linear logit function to generate

the parameterized fragility curves. The significant parameters were identified by the

authors using a sequential forward selection scheme. The authors demonstrated the

proposed approach with the case studies on multi-span simply supported concrete bridges

in South Carolina.

Park and Towashiraporn (2014) estimated the probabilistic seismic damage to

track-on steel-plate-girder (TOSPG) bridges in Korea, accounting for variations in the

number of spans, pier height, and earthquake magnitude. The authors used response surface

18
modeling to create second-degree polynomials for the estimation of seismic damage. The

study revealed that span length does not significantly affect seismic damage to bridges.

Jeon et al. (2016) employed a Bayesian framework to generate multi-parameter

fragility estimates. The framework includes the selection of a bridge class, characterization

of bridge attributes such as material and geometric uncertainties, creation of numerical

component models, construction of multi-parameter demand models using a Bayesian

parameter estimation method, and development of bridge-specific fragility models using

logistic regression and one-dimensional fragility curves using a Monte Carlo integration.

Additionally, the Bayesian approach used in the suggested framework enables the

identification of significant uncertainty parameters affecting seismic demands without

performing numerous structural analyses required for design of experiments. The authors

demonstrated their approach through a study of two classes of curved bridges commonly

found in California: two-frame and three-frame reinforced concrete box-girder bridges

with single column bents, diaphragm abutments, and in-span hinge(s).

Stefanidou and Kappos (2016) suggested a methodology for the generation of

bridge-specific fragility curves in which the limit state of the bridges is explicitly defined,

accounting for the effects of varying geometry, material properties, reinforcement, and

loading patterns. The methodology can account for the uncertainty in capacity, demand,

and damage state definition. The authors used nonlinear static analysis and IDA to estimate

the demand and capacity, and reduced sampling techniques for the uncertainty treatment.

2.3 Fragility curves for concrete bridge classes in California

California is a state with a high seismic hazard and a history of damaging

earthquakes. Various researchers have generated fragility curves for bridges in California,

using either an empirical (Başöz and Kiremidjian, 1996; Shinozuka et al., 2000) or

19
analytical approach. This section details the existing analytically based research on the

generation of fragility curves of various concrete bridge classes in California.

HAZUS (2003) generated fragility curves of various bridge configurations based

on seismic design, span length, bent type, and span discontinuity. However, HAZUS

fragility relationships were developed on the basis of a limited number of parameters and

simplified two-dimensional analysis, and did not account for the uncertainties in geometric

attributes for bridge classes such as the number of spans, span length, deck width, and

column height. A critical review of HAZUS fragilities follows this section.

Mackie and Stojadinović (2005) improved HAZUS fragility relationships by

reflecting the variation in bridge design parameters, including the skewness, span length,

span to column height ratio, and column to superstructure dimension ratio. However, their

models are applicable to a smaller subset of bridges, such as single-frame multi-span

continuous box-girder bridges with a single column bent.

Ramanathan (2012) generated fragility curves that are applicable to a portfolio for

various classes of bridges by accounting for uncertainties in attributes such as span length,

column height, number of spans, superstructure type, and material properties. They

addressed the evolution in seismic design philosophy by grouping the bridge classes into

three eras: pre-1971 bridges (Era 11, hereafter), 1971-1990 bridges (Era 22, hereafter), and

post-1990 bridges (Era 33, hereafter). Although the study provides valuable insight

regarding the bridge fragilities, it has some limitations, which are noted below:

• The grouping of bridge classes in the study was carried out based on a traditional

subjective approach that relies on engineering judgment. Such subjective grouping has

been criticized by more recent research which favors performance-based grouping

(Mangalathu et al., 2016a; Mangalathu et al., 2016b).

20
• The study considers that only one type of abutment footing (abutment on piles) is

possible in California. However, the plan review of California bridges revealed various

abutment footings such as abutment on spread footing, abutment on piles, and tall

cantilever footing. Recent studies (Mangalathu et al., 2015; Mangalathu et al., 2016a)

have noted that bridge fragilities are significantly influenced by the abutment footing

type.

• The study was limited to specific bridge classes in the California bridge inventory. For

example, the study on box-girder bridges was limited to two-span bridges, yet bridges

with more than two spans are common in California.

• The study only addressed a specific class of bridges with circular column shape. It has

been noted from the plan review that various cross-sections such as circular,

rectangular, and oblong (interlocking spirals) are present in the California bridge

inventory.

• The study assumed that the bearings in seat abutments are elastomeric. However,

rocker-type bearings are common in Era 11 bridges (Mangalathu et al., 2016a).

• The study considers only flexural mode of failure for columns in bridges constructed

during Era 11, although the lap-splice mode of failure is common in bridges from that

era.

• The study was limited to straight (non-flared) columns. It is common in post-1970

bridges (Era 22 and Era 33) to flare the columns in the upper region to provide support

to the cap beam under eccentric live load for architectural reasons. The response of the

bridge columns to seismic loading is significantly affected by the flares (Sanchez et al.,

1997).

21
• The study didn’t consider the effects of frames, pier-type columns, and spread-type

footing for foundations in the fragility analysis.

• The capacity estimates or limit state models in Ramanathan (2012) were preliminary

estimates (Mangalathu et al., 2016a) and need improvement.

2.4 The need to go beyond HAZUS-based grouping and fragility curves

HAZUS is the most comprehensive document for grouping bridge classes and

estimating seismic vulnerability. HAZUS groups bridge classes with similar damage/loss

characteristics and suggested fragility relationships to the grouped bridge classes. This

section summarizes the HAZUS grouping and fragility relations, and discusses their merits

and faults. Figure 2.2 shows the HAZUS-based grouping for the selected California box-

girder bridge inventory; Table 2.1 shows the grouping and fragility relationships suggested

by HAZUS.

22
 Table 2.1 HAZUS grouping and fragility relationships for bridge classes in
California.
Year Description (acronym in Fragility values in terms of Sa-1.0 s
Class
built HAZUS) Slight Moderate Extensive Complete Dispersion
Major bridge
HWB1 < 1975 0.40 0.50 0.70 0.90 0.6
Length > 150 m
Major bridge
HWB2  1975 0.60 0.90 1.10 1.70 0.6
Length > 150 m
HWB3 < 1975 Single span 0.80 1.00 1.20 1.70 0.6
HWB4  1975 Single span 0.80 1.00 1.20 1.70 0.6
Multi-column bent,
HWB6 < 1975 0.30 0.50 0.60 0.90 0.6
Simple support, Concrete
Multi-column bent,
HWB7  1975 0.50 0.80 1.10 1.70 0.6
Simple support, Concrete
Single column, Box-girder,
HWB8 < 1975 035 0.45 0.55 0.80 0.6
Continuous concrete
Single column, Box-girder,
HWB9  1975 0.60 0.90 1.30 1.60 0.6
Continuous concrete
Continuous concrete
HWB10 < 1975 0.60 0.90 1.10 1.50 0.6
(not HWB8/ HWB9)
Continuous concrete
HWB11  1975 0.90 0.90 1.10 1.50 0.6
(not HWB8/HWB9)
Multi-column bent, Simple
HWB18 < 1975 0.30 0.50 0.60 0.90 0.6
support, Prestressed concrete
Multi-column bent, Simple
HWB19  1975 0.50 0.80 1.10 1.70 0.6
support, Prestressed concrete
Single-column, Box-girder,
HWB20 < 1975 Prestressed concrete 0.35 0.45 0.55 0.80 0.6
continuous
Single-column, Box-girder,
HWB21  1975 Prestressed concrete 0.60 0.90 1.30 1.60 0.6
continuous
Continuous concrete (not
HWB22 < 1975 0.60 0.90 1.10 1.50 0.6
HWB20/HWB21)
Continuous concrete (not
HWB23  1975 0.90 0.90 1.10 1.50 0.6
HWB20/HWB21)
HWB28 All other bridges that are not classified 0.80 1.00 1.20 1.70 0.6

The salient features noted from the critical review of HAZUS grouping and fragility

relationships for bridges in California are noted below:

• HAZUS classifies bridges in two design eras, pre-1975 and post-1975. However,

bridge design philosophies in California were significantly influenced by the historic

1971 San Fernando and the 1989 Loma Prieta earthquakes. The extensive damage from

the 1989 Loma Prieta earthquake forced Caltrans to solicit the Applied Technology

23
 Council (ATC) to conduct a detailed study and provide recommendations for design

standards, performance criteria, and practices. The recommendations from ATC

described in ATC-32 were incorporated in Caltrans design manuals (Caltrans, 2010).

A study by Ramanathan et al. (2015) showed that fragility curves are highly influenced

by these design philosophies; seismic vulnerability decreases with the evolution in

column design philosophy. Therefore, it is necessary to separate the post-1975 bridge

class based on the evolution in seismic design philosophy.

• HAZUS classifies the bridge classes that are not addressed in the main classification as

the bridge group. The other bridge group represents the high-risk bridge inventory.

This classification leads to a situation where multi-frame bridges that are not addressed

explicitly in the main group are in the non-classified group, although the seismic

vulnerability of slab bridges is much lower than the vulnerability of continuous box-

girder bridges (Mehr and Zaghi, 2016). Therefore, HAZUS classifications significantly

overestimate the seismic vulnerability and loss assessment for multi-frame bridges.

• Although HAZUS classifies bridges based on abutment type (monolithic versus non-

monolithic, which is inferred as diaphragm versus seat abutments based on recent

seismic notions), HAZUS does not suggest explicit fragility relationships based on

abutment type. Previous studies (Mangalathu et al., 2016a; Ramanathan et al., 2015)

have noted that the demand models and fragilities for various components and bridge

systems differ drastically depending on the abutment style. Further, Ramanathan et al.

(2015) indicated that diaphragm abutments are less vulnerable than seat abutments in

pre-1990 bridges, but that the trend is reversed in post-1990 bridges.

• HAZUS fragility relationships were developed using a limited number of parameters

and simplified two dimensional analyses, and did not account for uncertainties in

geometric and material attributes for bridge classes such as the number of spans, span

24
 length, deck width, and column height. Also, other researchers (Porter, 2010;

Ramanathan, 2012) have criticized the capacity spectrum method (CSM) of structural

analysis used in HAZUS. The capacity spectrum method estimates the capacity of

bridge in the form of a pushover curve of the column and the demand in the form of a

response spectrum. The inability to account the higher-mode contributions and

vulnerability of other components leads to a non-reliable estimation of fragility curves.

• HAZUS considers the vulnerability of bridges to be governed by columns alone. As

pointed out by Ramanathan (2012), columns are not always the critical components;

neglecting the damage to bearings, abutments, and shear keys underestimates the bridge

vulnerability.

• HAZUS suggests the same fragility relationships for bridge classes HWB10 and

HWB22, and for HWB11 and HWB23. It can be inferred from these grouped fragility

relationships that the type of superstructure (reinforced versus pre-stressed concrete) is

not a significant parameter for the bridge fragilities. This is consistent with similar

conclusions noted in recent studies (Mangalathu et al., 2015a; Mangalathu et al.,

2016a). However, slab bridges, T-girder, and box-girder bridges are classified in the

same group and Ramanathan (2012) showed that these bridge classes do not have

similar fragility curves.

• A comparison of fragility relationships of bridge classes HWB8 and HWB10, HWB9

and HWB11, HWB18 and HWB20, and HWB19 and HWB21 shows that single-

column bents (SCBs) are more vulnerable than multi-column bents (MCBs).

Ramanathan et al. (2012) and Mangalathu et al. (2016a) showed that SCBs are less

vulnerable than MCBs for two- and three-span box-girder bridges.

• While comparing the fragility relationships of bridge classes HWB22 and HWB23, and

HWB10 and HWB11 for moderate, extensive, and complete damage states, the effect

25
 of design eras does not have an influence on the fragility relations. Such a conclusion

contradicts the research explicitly focusing on the effect of design eras (Mangalathu et

al. 2015a; Ramanathan, 2012).

• HAZUS suggests the same fragility relationships for single-span bridges irrespective

of the design eras (HWB3 and HWB4). Although it might hold for bridges with

diaphragm abutments, it is clearly not the case for bridges with seat abutments as there

is an increase in the seat-width provision for newer era bridges. Since span-unseating

or bearing displacement is the critical component for single-span seat abutment bridges,

new era single-span seat abutment bridges are less vulnerable than their counterparts

from previous eras, because of increased seat width.

• HAZUS fragility relationships suggest that simply-supported bridges are more

vulnerable than continuous bridges. The study by Ranf et al. (2007) utilized damage

data collected from the Nisqually Earthquake in 2011 to reveal that this is not true for

lower damage states. As there is not enough data for higher damage states, it is not

certain whether the HAZUS fragility relationships (that is, that simply-supported

bridges are more vulnerable than continuous bridges) hold for higher damage states.

The study also indicated that HAZUS fragility relationships overestimate the damage

for simply-supported bridges.

• Although HAZUS classifies bridges without considering the number of frames, Mehr

and Zaghi (2016) used three-dimensional nonlinear time history analysis to show that

single frame bridges do not have similar fragility curves to multi-frame bridges.

• An extensive plan review of the California bridge inventory revealed various column

cross-sections such as rectangular, circular, and oblong. These cross-sections occupy a

major portion of concrete bridges in California and recent studies (Mangalathu et al.,

26
 2016a; 2016b) have shown that bridges with circular and rectangular column cross-

sections have different seismic demands and fragilities.

• Although HAZUS classifies the bridges based on length (length > 150 m and length <

150 m), it is not clear whether the length is per frame or the total length of the bridge.

• As pointed out by Ramanathan (2012), there is a mismatch between overall bridge

functionality in HAZUS and the damage state definitions used in the fragility analysis.

Such a discrepancy could cause problems for Departments of Transportation officials

in emergency response decisions.

• HAZUS classified California bridges as either single-span or multi-span. The effect of

number of spans is accounted in HAZUS by a modification factor. Per HAZUS

modification factor, three-span bridges are less vulnerable than their counterparts. A

study is needed to verify the HAZUS modification factor.

• The effect of pier shaft foundation type is not addressed in HAZUS.

Given the key points noted from the critical review of HAZUS bridge classification,

it is clear that HAZUS groupings and fragility relationships need significant improvement.

Also, it is rational to advance the grouping of bridge classes from a traditional perspective

relying on the judgment of engineers to one that is performance-based.

2.5 Uncertainty treatment in fragility analysis

As stated previously, the prevalent approach for the generation of fragility curves

is the convolution of demand models with capacity models. Demand models are usually

obtained by conducting non-linear time history analysis (NLTHA) on bridge models (Choi,

2002; Gardoni et al., 2003; Kim and Shinozuka, 2004; Mackie and Stojadinovic, 2001;

Mackie and Stojadinović, 2005; Banerjee and Shinozuka, 2007; Padgett, 2007;

27
Mangalathu et al., 2015; Ramanathan et al., 2015; Mangalathu et al., 2016a). In the

probabilistic seismic performance assessment, bridge models are generated by conducting

sampling across the uncertain input parameters (Padgett, 2007; Ramanathan et al., 2015,

Mangalathu et al., 2016a; Mangalathu et al., 2015a). It is highly likely that various

uncertainties such as geometric, material, or component response parameters exist due to

structure-to-structure variation in the generation of fragility curves, especially if the

fragility curves are intended for the regional risk assessment of bridges (Mangalathu et al.,

2016a). The source of uncertainties can either be due to lack of knowledge (epistemic) or

due to inherent randomness (aleatoric). As it is impossible to eliminate uncertainties, the

risks due to uncertainty must be properly evaluated and limited.

Figure 2.5– Uncertainty sources for system demand and capacity (Ji et al., 2007).

Uncertainties can present both in seismic demand and capacity (Ji et al., 2007,

Figure 2.5). Capacity is a property of the system to withstand maximum force or

displacement without failure. Researchers have attempted to determine the extent to which

uncertainties affect the seismic demand, capacity, and fragilities. The uncertainties in

ground motions can be accounted for by including many records of ground motions to

cover as many frequencies and seismic energies as possible (Ji et al., 2007). However, the

28
number of records needed to have a reliable estimate of the fragility is not well defined

(Haselton et al., 2012). Celik and Ellingwood (2010) noted that uncertainties in ground

motion dominated overall uncertainty in structural response in the case of gravity load

designed reinforced concrete frames. Their study concluded that other sensitive parameters

affecting the seismic response of reinforced concrete frames are damping, concrete

strength, and joint cracking strain. The uncertainty in the capacity is usually accounted for

by modeling the capacity or limit state as a random variable (Ji. et al. 2007). For example,

HAZUS suggests a dispersion measure to account for variability in the damage state. In

the case of bridges, researchers have attempted to assess the sensitivity of seismic demand

or have evaluated fragility to the parameter uncertainty (Dicleli and Bruneau, 1995;

Nielson and DesRoches, 2006; Padgett, and DesRoches, 2007; Padgett et al., 2010; Ghosh

et al. 2013, Jeon et al., 2015).

Dicleli and Bruneau (1995) investigated the response of single-span simply

supported and continuous slab-on-girder steel bridges using linear elastic and nonlinear

inelastic analysis. Based on elastic spectral analysis, they noted that bearing forces, both in

longitudinal and transverse direction, were proportional to span length. The authors

concluded from the inelastic time history analysis of bridges that 3-lane bridges are less

vulnerable than 2-lane bridges. Another conclusion from their study is that the bridge

response was significantly influenced by the stiffness with which the steel bearings are

modeled. However, this conclusion was contradicted by the research on simply supported

steel-girder bridges conducted by Ala Saadeghvaziri and Rashidi (1998), in which the

bridge response was dependent on the stiffness of the bearings only in the transverse

direction, but was inconsequential in the longitudinal direction.

Nielson and DesRoches (2006) carried out an experimental design to ascertain the

significance of geometric, material, and structural uncertainties of multi-span simply

supported steel girder bridges in the central and southeastern United States (CSUS).

29
Nonlinearities in the abutments, bearings, columns, and bent caps were explicitly

considered in their study using detailed 3-D nonlinear models. The authors used a statistical

analysis called ANOVA to identify the significant parameters for bridge samples subjected

to seismic loading. The study revealed that damping ratio and loading direction are the

most important parameters affecting the seismic response of bridges. The study also noted

that column ductility and bearings deformations are sensitive to the type and stiffness of

bearings.

Padgett and DesRoches (2007) extended the procedure used by Nielson and

DesRoches (2006) to retrofitted bridges in CSUS and concluded that fragility curves

developed with sensitive parameters are nearly identical to those developed with all

potential sources treated as variables. Their study illustrated that preliminary screening of

parameters could reduce the simulation and computational efforts for the generation of

fragility curves. It has been noted from their study that the uncertainty in ground motion

and gross geometry overshadows the contributions of other sources of uncertainty

attributed to the bridge modeling. The sensitivity study was further extended by Padgett et

al. (2010) to identify the effect of liquefiable soil and modeling parameters on the seismic

reliability of critical components of steel bridges in CSUS. Although such studies are

valuable for providing critical insights, most of them are rooted in rigorous statistical

analysis based on experimental design and demand exhaustive computational efforts. Also,

Mangalathu et al. (2017a) demonstrated that underlying assumptions in statistical methods

have a significant influence on the identification of critical parameters. Kunnath et al.

(2006) investigated the effect of foundation flexibility and soil-structure interaction on the

seismic demands. However, the demand model is conditioned only on one parameter (IM)

and hence it is difficult to estimate the sensitivity of other parameters on seismic demand.

Ghosh et al. (2013) used a multi-parameter demand model to account for the effect of all

uncertain parameters in the generation of fragility curves. The authors used four surrogate

30
modeling techniques, including polynomial response surface models, multivariate adaptive

regression splines, radial basis function networks, and support vector machines to

determine the best-fitting parameterized demand models involving the uncertain input

parameters. However, as noted by a previous study (Padgett, 2007), it is highly unlikely

that all of the uncertain parameters have a significant influence on the seismic demand

model.

Kameshwar and Padgett (2014) identified the significant parameters that can affect

the performance of bridges exposed to earthquake and hurricane hazards through a

sequential forward selection scheme. The authors discussed the earthquake and hurricane

risks to multiple-span simply supported (MSSS) concrete girder bridges in South Carolina

based on variations in column diameter, column slenderness, and length of spans.

Jeon et al. (2016) employed a Bayesian framework to screen the parameters that

have a significant influence on the seismic fragility estimate of curved bridges in

California. The authors identified that parameters such as damping ratio, mass factor,

longitudinal gap, backfill type, pile stiffness, column longitudinal reinforcement ratio,

rotational and transverse stiffness of footing, bridge angle, main span length, side span-to-

main span length ratio, and column height have a significant influence on the seismic

demand of bridges. The study concluded that seven parameters (the earthquake direction

factor, concrete strength, rebar yield strength, coefficient of friction, and shear modulus of

elastomeric bearing pads, transverse gap, and abutment height) have little impact on all

demand models. However, the framework suggested by authors is computationally

expensive because they performed a set of stepwise regressions until the reduced model

satisfied an acceptable value.

The random nature of earthquakes, and differences in geometric, material and

structural attributes of bridges cause significant uncertainties in the estimation of seismic

31
demand and fragilities of bridge classes. The literature review on the uncertainty treatment

of input parameters in bridge fragilities suggests the sensitivity parameters vary depending

on the type of bridge and its location. Although these studies have been invaluable in

acquiring an understanding of the effects of significant variables on the seismic demand

and fragilities of bridges, there is still a need for identifying the relative impact of each

uncertain input variable and the level of treatment needed for these variables in the

estimation of seismic demand models and fragility curves. As the current study focuses on

the box-girder bridges in California, a sensitivity study is needed for the box-girder bridges

in California that can (1) identify the variables that exhibit strongest influences on seismic

demand and seismic fragilities; (2) provide insight in quantifying whether the variation of

uncertain parameters should be treated explicitly or be neglected; 3) eliminate the

parameters which have a minimal influence on seismic demand and reduce unnecessary

and exhaustive efforts in statistical sampling; (4) identify parameters which could reduce

the uncertainty in demand models and fragility curves by more explicit evaluation of the

uncertainty distribution (e.g., by developing an extensive database); and (5) help bridge

owners (such as California Department of Transportation) spend their resources judiciously

(e.g. data collection, field investigations, censoring) on parameters that have significant

influences on bridge fragilities.

2.6 Closure

California is a state with a high seismic hazard and a history of damaging

earthquakes. It has close to 29,000 bridges with varying ages and design parameters based

on the year of their construction. However, it is cumbersome and time-consuming to

develop unique fragility curves for each structure across a regional portfolio. One strategy

that has been used to address this challenge is to group bridges into classes based on similar

design or structural performance. Traditionally, this grouping has been conducted based on

a relatively subjective identification of sub-classes. It is not known whether the subjective

32
classification yields distinct seismic performances (or seismic demand) between the

grouped bridge classes. The literature review suggests the need for a performance-based

grouping that can cover the entire California bridge inventory.

Fragility curves, which are probabilistic tools used to assess seismic damage to

highway bridges, can be generated based on expert opinion, empirical approach, and

through numeric methods. A review of the current methods for generating fragility curves

is given in this chapter. Caltrans’ current deployment of ShakeCast uses HAZUS-based

bridge fragility models developed in the 1990s to support loss estimation by the Federal

Emergency Management Agency. By necessity, these early models were derived with

simplified analysis methods, compared to a limited set of damage observations, and use a

bridge taxonomy based on the limited data fields available in the National Bridge Inventory

(NBI). HAZUS fragility relationships suffer major drawbacks which are discussed in detail

in this chapter. It has been noted that using HAZUS relationships leads to non-realistic

estimation of the seismic risk. Also, HAZUS framework is not well aligned with Caltrans

seismic design philosophy or the California bridge inventory. There is a need for the

generation of fragility curves that can lead to a realistic estimation of seismic risk in

California. High fidelity three-dimensional analytical models will be used in the current

research to develop fragility curves for highway bridge classes.

The literature review also reveals the need for a sensitivity study on box-girder

bridges that can identify the significant input parameters and the relative impact of these

parameters on seismic demand models and fragilities. Such a sensitivity study would help

Caltrans to spend their resources (e.g. data collection, field investigations, censoring)

judiciously on parameters that have a significant influence on bridge fragilities.

33
 CHAPTER 3 MODELING OF BRIDGE COMPONENTS

This section presents the various bridge components and the adopted numerical

modeling strategies. Consistent with the previous work of Ramanathan et al. (2015),

bridges are classified into pre 1971 design era (Era 11, hereafter), 1971-1990 design era

(Era 22, hereafter), post 1990 design era (Era 33, hereafter) based on evolutions in the

seismic design philosophy.

Highway bridges, as illustrated in Figure 3.1, have a number of bridge components.

The components can be primarily classified into as superstructure and substructure. The

super structure includes the girders, deck slab, and parapet, while the substructure consists

of abutments, footing, bents (beam and columns), bearings, and shear key.

Figure 3.1 – Illustration of major bridge components.

A typical layout of the numerical modeling for a two-span bridge is shown in Figure

3.2. The bridge is modeled in three dimensions to capture responses in both the longitudinal

and transverse directions, as well as their interactions. The modeling is carried out with

the finite element package OpenSees (Mazzoni et al. 2006), incorporating geometric and

material nonlinearities. Rayleigh damping is used in the non-linear time history analysis.

34
The ground motions representative of seismic risk in California are used in this study. The

two horizontal components of the ground motions are assigned simultaneously to the

longitudinal and transverse direction of the bridge, and the orientation is assigned

randomly. The effects of vertical acceleration and spatially variable ground motions are

not considered in this study. The following section explains the numerical modeling of

various bridge components.

3.1 Superstructure

The superstructure of bridges is modeled as a spine with elastic beam-column

elements since it is expected that the deck will remain elastic during earthquake loading.

Transverse deck elements are modeled using elastic beam-column elements (rigid and

massless) and are connected to the columns using rigid links to ensure the moment and

force transfer between members. Translational and rotational springs are added to the base

of the column to simulate the behavior of the footing. Zero length elements capturing the

response of the abutment back fill soil and bi-directional force (abutment piles or frictional

surface) are connected in parallel and are connected to the transverse deck elements in the

case of diaphragm abutments. The abutment pile or friction surface model is selected based

on the type of footing, whether the abutment is resting on piles or on a spread footing. In

the case of cantilever abutments, the wall stem flexure is connected in series with the bi-

directional force springs. Bearing pad elements and pounding elements are also modeled

with zero length spring elements and are connected in parallel.

The following section presents detailed modeling considerations for various bridge

components. In general, the bridge components are divided into superstructure and

substructure.

35
 The superstructure (bridge deck) typically remains elastic during an earthquake.

The superstructure in this study is modeled using elastic beam-column elements and is

shown in Figure 3.2.

Figure 3.2 – Numerical modeling of various bridge components.

3.2 Substructure

California bridges have different pier types (single column bents (SCB), multi-

column bents (MCB), pier walls or pile shafts), footing types (spread footing, shaft pile, or

footing on piles), and abutment types (abutment on piles or abutment on footing).

3.2.1 Bents

The bents are modeled using a combination of displacement based beam column elements
and rigid links to cause moment and force transfer between the members of the bent. The

36
finite element discretization of a single column and two-column bent is shown in Figure
3.3.

Figure 3.3 – Finite element discretization of bents.

3.2.2 Columns

Columns are one of the most vulnerable components in the event of an earthquake; the
majority of bridge seismic failures in the past have been attributed to column failures.
Displacement-based beam column elements with fiber-defined cross-sections are used in
this study to model the columns (Figure 3.3). Fiber cross-sections have the distinct
advantage of specification of unique material properties for different locations across a
member’s cross-section. For instance, confined concrete is used to represent the concrete
behavior in the core section of the column, while unconfined concrete is used to represent
the unconfined cover concrete. The Chang and Mander (1994) model is used to define the
monotonic stress-strain curves of confined and unconfined concrete. Material models for
the concrete section are shown in Figure 3.4, and Figure 3.5 shows the effectively confined
core concrete area for the circular and rectangular section. Suppose, fc represents the
unconfined strength of concrete, the maximum concrete stress (fcc) and corresponding
strain (cc) can be calculated as (Mander et al. 1988):

37
 f cc  f c  f l k e
fl
 f c (1  k e )
fc
(3-1)
 f c CF
fl
 cc   co (1  k 2 )
fc

where fl is the effective lateral confinement factor, ke and k2 are coefficients that are
functions of the concrete mix and lateral pressure, and co is the unconfined strain in
concrete.

Figure 3.4 – Fiber based discretization of the columns.

The effective lateral confining pressure (fl, Equation 3.2) for a circular column can

be calculated as:

1 4 Asp
fl   s f yh ; s  (3-2)
2 dss

where Asp is the area of transverse reinforcement bar, fyh is the yield strength of the

transverse reinforcement, s is the center to center spacing or pitch of spiral or circular hoop,

and ds is the diameter of spiral bar centers. This is shown in Figure 3.5. The ratio of the

38
area of the effectively confined concrete core to the area of core of section enclosed by the

perimeter lines of the perimeter spiral or hoop (Equation 3.3), ke can be calculated as

s'
1
2d s
ke  (3.3)
1   cc

where cc is the ratio of area of longitudinal reinforcement to the area of core of section. k2

is given as 5k1.

ds ds
cover concrete cover concrete ds-s'/2
(spalls off)
A A
ineffectively s' s
confined core
45˚
effectively B B
¼s'
confined core
Section B-B Section A-A
a)

bc bc
effectively w'
confined ineffectively bc-s'/2
core confined
dc
core
Y dc-s'/2 Y s' s
cover Z Z
concrete
Section Z-Z (spalls off) Section Y-Y
b)

Figure 3.5 – Effectively confined core for a) circular hoop reinforcement and, b)
rectangular hoop reinforcement (Mander et al. 1988).

For a rectangular bc  dc (Figure 3.5), the area of effectively confined concrete core to the

area of core of section (ke) for a rectangular section can be calculated as

39
  n
(w' ) 2  s'  s' 
1   i 1  1  
 
ke  
i 1 6bc d c  2bc  2d c 
(3.4)
1   cc

The lateral confining stress of concrete in x and y directions (Figure 3.5) is given

as

Asx
f l x   x f yh ; x 
dc s
(3.5)
Asy
f l y   y f yh ; y 
bc s

where Asx and Asy are the total area of transverse bars running in the x and y direction.

The Menegotto and Pinto (1973) model, later modified by Filippou et al. (1983), is

used to add isotropic strain hardening for steel (Figure 3.4).

3.2.2.1 Columns Types

Various cross sections such as circular, rectangular, wide perimeter, and wide

circular are noted from the plan review. The typical details of some column cross-sections

are shown in Figure 3.6. The wide perimeter cross-section corresponds to the section in

which the reinforcement is laid around the perimeter, while the wide circular cross-section

corresponds to the multi-circular reinforcement pattern. Regular cross-sections are

classified as circular, rectangular, or oblong depending on the cross-section shape of their

concrete core.

40
 Figure 3.6 – Typical cross-sections noted from the bridge plan review (Caltrans,
2017).

3.2.2.2.1 Circular columns

Figure 3.7 – Circular confined core concrete sections.

Figure 3.7 represents the common choice for columns with circular distribution of

longitudinal reinforcement contained within transverse spirals or hoops. The area outside

the core concrete can be circular, hexagonal, octagonal, or any other shape. In the case of

a circular cross-section, flexural strength, shear strength, and moment capacity are

independent of the direction of loading.

41
3.2.2.2 Rectangular columns

Rectangular columns of various sizes are noted from the plan review of bridges.

Figure 3.8 shows some of the typical rectangular cross sections noted from the plan review.

It is noteworthy to mention that the strength and stiffness of the rectangular columns

depend on the direction of loading.

Figure 3.8 – Typical rectangular cross sections.

3.2.2.1.3 Oblong columns

The superior performance of the spiral reinforcement over the rectilinear tie

transverse reinforcement led to the development of columns with interlocking spirals (also

called oblong columns, Figure 3.9). The confinement factor of a single column fiber is

applicable to the oblong columns as most of the confined concrete in double-spirals is

confined by a single spiral and the interlocking region is a relatively small area (Correal et

al. 2007).

Figure 3.9 – Oblong columns.

42
3.2.3 Idealization of bridge columns

As the California bridge inventory consists of wide range of column cross-sections,

some idealizations are used in this study. The idealization is carried out in such a way as

to mimic the bridge inventory noted from plan review and is detailed in Table 3.1. The

idealization of the cross-sections is carried out based on the area of confined concrete, as

the core concrete will continue to carry stress at higher strains. The cover concrete becomes

ineffective once the compressive strength is attained in the flexural deformation, and the

effect of cover concrete is neglected in idealizing the cross-sections.

43
 Table 3.1 – Idealization of bridge columns.
Column cross-section from the bridge inventory Idealized cross-section shape Remarks*
No idealization
needed

Do = Di

CF= circular
Idealized to
circular cross-
section. Area of
Symmetric

core concrete
remains the same
Dco = Dci
co = c i
CF= circular
No idealization
needed
Loo = Lco
Boo = Bco
CF= circular

44
 Idealized to oblong
cross-section. Area
of core concrete
remains the same
Loc = Loi
Boc = Boi
co = ci
CF= circular
Idealized to
rectangular cross-
section.
L = Lr
B = Br
CF= rectangular

Idealized to
Wide perimeter

rectangular cross-
section.
Ll = Lc
Bl = Bc
CF= rectangular

Idealized to
rectangular cross-
section. (Area of
two sections
remains the same)
Blc = Brc
CF= rectangular

45
 Idealized to
rectangular cross-
section. (Area of
two sections
remains the same)
Brcc = Bric
CF= rectangular
Idealized to
rectangular cross-
section. (Area of
two sections
remains the same)
Bro= Bio
CF= rectangular
Planning to
exclude as the
Wide circular

count is less than

5% of total
inventory

46
3.2.4 Validation of bridge columns

Figure 3.10 shows the comparison of the numerical model with the experimental results

for Era 33 columns. The geometric and reinforcement details are obtained from Lehman

and Moehle (2000), and the modeling details are outlined in section 3.2.2. It is seen from

the comparison that the numerical model is able to capture the key responses fairly well.

a) b)
b)
Figure 3.10 – Comparison of experimental and numerical results for era 33
columns (Lehman and Moehle, 2000) a) Specimen No. 415 b) Specimen No. 815.

3.3 Abutments

Abutments can be classified in two basic types: diaphragm abutments and seat

abutments (Ramanathan, 2012). Diaphragm abutments are cast monolithic with the

superstructure. As they engage the backfill soil during seismic action, diaphragm

abutments provide a good source of energy dissipation and reduce the likelihood of span

unseating. Seat abutments provide a bearing support to the superstructure, which is

restrained longitudinally by the abutment backwall and transversely by the piles and the

shear key. The stiffness and resistance to the seismic action increases when the deck is in

contact with the abutment backwall in the longitudinal direction. However, as the

47
superstructure moves away from the abutment, the resistance depends primarily on the

bearing pads, which makes it susceptible to unseating. The backwall of the seat abutment

is typically designed to fail under impact and passive response, before damaging forces are

transmitted to the lower portion of the abutment.

Figure 3.11 – Various types of abutments.

The major configurations of abutments noted from the review of bridge plans in

various design eras are shown in Figure 3.11. Abutment can be on piles, on spread footing

or cantilever. Abutments resting on piles are the major configuration for both diaphragm

and seat abutments. However, the review of bridge plans also revealed a variety of unusual

abutment wall details (Figure 3.12) which seemed to be designed to provide a weak link in

the stem wall above the footing which could either translate along a construction joint or

rotate on an intermediate section of the stem wall. These details were found on about 15-

20% of the diaphragm-type abutments.

48
 Figure 3.12 – Unusual abutment types (Caltrans, 2017).

Abutment responses to seismic actions include earth pressure response and

structural response. The earth pressure on the abutment is due to the longitudinal response

of the bridge deck and includes passive and active resistance. Passive resistance is provided

by the backfill soil and piles/friction surface (depending on the abutment footing type); it

develops when the abutment moves toward the backfill soil. Piles/friction surface alone

contribute the active resistance, which is activated when the abutment moves away from

the backwall soil. The passive response of the abutment backwall is simulated using the

hyperbolic soil model proposed by Shamsabadi and Yan (2008) and is given in Figure 3.13.

The model is based on experimental testing of bridge abutments conducted at the

University of California Los Angeles with 5.5 ft. high backwalls and typical non-cohesive

and cohesive backfill soils. The test results were then extended to develop closed form

solutions for the abutment backfill soil response for a range of backwall heights based on

a series of analyses using the limit-equilibrium method which implements mobilized

logarithmic-spiral failure surfaces coupled with a modified hyperbolic soil stress strain

behaviour. Fult is the maximum abutment force developed at maximum displacement ult.

49
Equation 3.6 presents the closed form solution for the force displacement response of the

backfill soil, where F is the force expressed in kip/ft width of the backwall,  is the

displacement expressed in inches, and H is the height of the backwall expressed in feet.

8
F ()  H1.5 Granular backfills
1  3
(3.6)
8
 H Cohesive backfills
1  1.3

The maximum displacement of the backwall is 0.05H and 0.1H (expressed in

inches) for granular (sandy soils) and cohesive (clay soils) backfills, respectively, and

substitution of these values in Equation 3.6 yields the ultimate force in the abutment. The

backfill soil response is modeled using HyperbolicGapMaterial provided by OpenSees,

which corresponds to the model proposed by Shamsabadi and Yan (2008). The force-

displacement response of the HyperbolicGapMaterial is given in Equation 3.7.


F ()  (3.7)
1 
 Rf
K max Fult

where, the failure ratio Rf is 0.7, Kmax is secant stiffness at half of the maximum abutment

force.

50
 Figure 3.13 – Modeling of the abutments.

Piles provide longitudinal and transverse stiffness to the abutments when abutments

rest on piles. The trilinear force-deformation response of the pile, along with the associated

modeling parameters, are presented in Figure 3.13. The initial yield parameters (1, F1) are

determined following the design recommendations of the Caltrans 2014 draft of bridge

design aids on ‘Permissible Horizontal Loads for Standard Plan and Steel HP Piles’

(Caltrans, 2015). The plastic yielding parameters (2, F2) are calculated based on results of

modeling various pile systems simulated in LPILE (Caltrans, 2015). The hysteretic

behavior of piles is captured using the Hysteretic material in OpenSees with the hysteretic

parameters pinchX and pinchY as 0.75 and 0.5 (Ramanathan, 2012). In contrast, for

abutments supported on spread footings, a frictional response model is used. The maximum

force (Fs) is calculated as the product of the coefficient of friction (f) and the dead load

reaction on the abutment. For cantilever abutments, the bi-linear model is obtained from

numerical modelling of various cantilever type abutments in OpenSees.

51
3.4 Bearings

The bearings most commonly used for seat abutment bridges are rocker or

elastomeric bearings, as shown in Figure 3.14. These bearings are characterized by

different response mechanisms. Rocker bearings shown in Figure 3.14 (a) are considered

vulnerable due to non-ductile transverse keeper plate failure and longitudinal instability

(Mander et al., 1996). The elastomeric bearings shown in Figure 3.14(b) usually transfer

horizontal forces using friction, and their behavior is characterized by sliding. Elastomeric

bearings decouple the superstructure from the substructure, and thus the superstructure is

susceptible to large deformations. Additionally, a special type of bearing called friction

bearing can also be found in Era 11 bridges. The predominant difference between the Era

11 and the later design eras is the replacement of steel rocker bearings with elastomeric

steel bearings. This shift in the bearing type is due to advancement in the seismic design

philosophies.

(a) (b)

Figure 3.14 – Types of bearings: (a) rocker bearing and (b) elastomeric bearing
(Mangalathu et al. 2016).

The elastomeric bearing is assumed to be elasto–plastic and the yield force, Fy, is

obtained by multiplying the normal force by the coefficient of friction (Figure 3.15). The

52
rocker bearing is modeled following the experimental work of Mander et al. (1996). This

material model includes a frictional component, and its longitudinal force-deformation

behavior is shown in Figure 3.15. An elasto–plastic behavior is assumed for the friction

bearings. The bearings are modeled using the Steel01 material in OpenSees.

Figure 3.15 – Modeling of various bearings.

3.5 Shear Keys

Figure 3.16 – Modeling of shear key.

Shear keys help restrain the relative transverse movement between the deck and the

bridge abutments. A shear key can fail through four mechanisms in the event of an

earthquake, namely shear friction, flexure, shear, and bearing (Megally et al. 2002). The

shear key designs are categorized as isolated (emerging designs) or non-isolated

53
(conventional designs) (Caltrans 2017). Since the isolated shear key is a new type of design

and does not exist in the current inventory, this study will focus only on the non-isolated

shear keys. The nonlinear model of the shear key is also depicted in Figure 3.16. Fcap

denotes the capacity of the shear key, which is computed as the product of the dead-load

reaction and the acceleration (Caltrans 2015). Megally et al. (2002) conducted a series of

experiments on the shear keys and found that Δmax minus Δgap equal to 3.5 in. is the

deformation at which the capacity of the shear keys essentially degrades to zero.

3.6 Pounding

Seismic pounding is the impact between the bridge decks, between deck and

abutment, or between the adjacent decks in a multi-frame bridge in the longitudinal

direction. Impact occurs when the relative displacement between adjacent decks or deck

and abutment exceeds the gap between them. Significant pounding damage was noticed at

the I-5/SR-14 interchange during the 1994 Northridge earthquake (Muthukumar and

DesRoches, 2006). Some instances of pounding damage, ranging from superficial to

complete collapse, are shown in Figure 3.17.

Figure 3.17 – Pounding damage in bridges during the 1994 Northridge earthquake:
(a) barrier rail damage and, (b) connector collapse (Muthukumar and DesRoches,
2006).

The contact element developed by Muthukumar and DesRoches (2006) is used to

model the pounding between superstructures and abutments. This material model explicitly

54
accounts for the loss of hysteretic energy (Figure 3.18). The maximum deformation, m, is

assumed to be 1.0 in. The yield deformation, 1, is assumed to be 0.10m. The stiffnesses,

k1 and k2, are recommended to be 1022.3 kip/in/ft and 351.755 kip/in/ft, respectively.

Figure 3.18 – Analytical Model for pounding between deck and abutment back wall.

3.7 Foundation

The foundation provides a means to transmit service and ultimate loads from the

structure to the underlying soil. Foundations can be classified as either shallow or deep. As

the name implies, the loads from the structure are transferred to the underlying soil at a

shallow depth for shallow foundations. Deep foundations are used when soil conditions are

not favourable to shallow foundations and transfer the load through piles. The type of

foundation for a particular bridge is determined by various factors such as soil conditions,

overhead clearance, existing utilities, and proximity to existing facilities such as buildings

and railroads (Caltrans 2017). The possible types of footings are shown in Figure 3.19.

Foundations are modeled using elastic translational and rotational springs (Figure 3.20)

and are added at the base of the column.

55
Figure 3.19 – Bridge foundation types (Priestley et al. 1996).

56
 Figure 3.20 – Modelling of foundations.

3.8 Closure

The numerical modeling of various bridge components, and the integration of

various component models, to generate a global analytical model of the bridge for fragility

analysis. Displacement-based beam column elements are used to model the columns.

Translational and rotational springs are added to the base of the column to simulate the

behavior of the footing. Zero length elements capturing the response of the abutment back

fill soil and bi-directional force (abutment piles or frictional surface) are connected in

parallel and to the transverse deck elements in the case of diaphragm abutments. Bearing

pad elements and pounding elements are also modeled with zero length spring elements

and are connected in parallel to abutment springs in the case of seat abutments.

Interested readers are directed to the study from our research group members

(Ramanathan et al. 2015; Soleimani 2017), for the validation of the global analytical model

of the bridges. Ramanathan et al. (2015) compared the efficacy of the nonlinear dynamic

analysis of the bridge models with the recorded ground motion data. The authors compared

the response of a two span reinforced concrete box-girder bridge built in 1971 (Meloland

Road Overpass) with the recorded ground motion during the 1970 Imperial Valley

Earthquake. The bridge is instrumented with twenty-six channels of accelerometers and

57
peak ground acceleration (PGA) values of the ground motions that strike the bridges were

of intensity, 0.32 g, 0.30 g, and 0.23 g in the longitudinal, transverse, and vertical

directions, respectively. The authors concluded that the analytical model yield comparable

results to the sensor data for the Meloland bridge. Note that the lack of data prevents the

comparison of the bridge model with the ground motions that can cause significant

nonlinear response. Soleimani (2017) compared the dynamic response of bridge columns

with the full scale shake table test on single column bridge bent (Schoettler et al. 2012) to

the various ground motions. It is noted from their study that the analytical model can predict

the shear force and deformation of the column fairly well for various damages states.

58
CHAPTER 4 PERFORMANCE BASED GROUPING OF BRIDGE

CLASSES

One of the critical aspects of seismic risk assessment of highway bridge

infrastructure systems is the generation of fragility curves that are applicable to a class of

bridges. California, a state with a high seismic hazard and a history of damaging

earthquakes, has close to 29,000 bridges that vary in age based on their construction.

However, it is cumbersome and time-consuming to develop unique fragility curves for each

structure across a regional portfolio. One strategy that has been adopted to address this

challenge is to group bridges into classes with similar design or structural performance.

The bridges in a particular class (or group) are expected to have similar performance or

damages during an earthquake. The identification of specific bridge parameters or bridge

design attributes that yield distinct seismic performance to bridges is an important step in

this procedure.

Traditionally, this grouping has been performed based on a relatively subjective

identification of sub-classes. However, such subjective identification leads to a number of

bridge classes with unwarranted grouping (HAZUS, 2013). The limitation of the HAZUS

grouping is discussed in detail in the literature review of this thesis. This chapter explains

the various performance based grouping strategies such as analysis of variance (ANOVA),

analysis of Covariance (ANCOVA) and Kruskal–Wallis (KW) approach. Although these

approaches are used widely in many disciplines such as biology, medical, and industrial

engineering (Vidakovic 2011; Vickers 2005), the application, relevance, and advantages

of these grouping techniques for the grouping of bridge classes for probabilistic seismic

performance assessment have not been fully explored. The selected methods have different

approaches and underlying assumptions to grouping structures, which are reviewed and

compared on the basis of statistical power to identify attributes that dictate distinct bridge

59
sub-classes of structural performance. The comparison is based on case studies of two-span

and three-span bridges in California. The assumptions underlying each approach are

investigated in detail in this chapter. It is noteworthy to mention that the comparison is

conducted in order to select a rational grouping procedure. A performance-based grouping

approach is suggested in this chapter to group the California bridge inventory. A review of

the application of ANOVA, ANCOVA, and KW in grouping the bridge classes is given in

the next section.

4.1 Review of ANOVA, ANCOVA, and Kruskal–Wallis approach

This section briefly introduces the ANOVA, ANCOVA, and KW methods for

grouping bridge classes. The grouping is achieved by comparing bridge responses of

various bridge sub-classes and by identifying whether they are statistically different. The

response considered in this study includes (1) maximum column curvature ductility, (2)

maximum abutment passive deformation, (3) maximum abutment active deformation, (4)

maximum abutment transverse deformation, (5) maximum unseating deformation, and (6)

maximum bearing deformation. The maximum responses are obtained from a set of

NLTHAs. Figure 4.1 shows the scatter plot of seismic demand or response (D) of two

typical bridge groups with the IM, along with the probability distribution of their seismic

demands. PSDMs are usually obtained by performing a linear regression for a pair of D

and IM, using a suite of N ground motions. In a mathematical form, PSDM can be written

as

ln( S d )  ln(a )  b ln( IM ) (4.1)

where a and b are the regression coefficients and, Sd is the median estimate of the demand

in terms of an IM. The coefficients a and b are obtained by performing a linear regression

analysis on D and IM pairs in the log–transformed space. Dispersion, d/IM, is evaluated

based on statistical analysis of D and IM pairs:

60
 N

  ln d  ln ( S d ) 
2
i
 d / IM  i 1
(4.2)
N 2

where di is the demand for the ith ground motion.

Figure 4.1 – Illustration of PSDM and grouping strategy.

4.2 ANOVA based grouping

Analysis of covariance (ANOVA) is generally utilized to determine whether there

are any significant differences between the means of two or more independent, unrelated

groups (Miller Jr 1997; Keselman et al., 1998; Vidakovic 2011; Mangalathu et al., 2017a;

Mangaalthu et al., 2017b). It tests the hypothesis ( H 0 :  D1   D2 ,...,   Dk ) that the mean seismic

responses of different bridge classes are equal (Figure 4.2) and can group the bridge classes

accordingly. The assumptions underlying ANOVA are: (1) the responses are mutually

independent, (2) homogeneity of response variance, and (3) samples are mutually

independent. If the assumptions are violated, ANOVA is not a powerful test and the results

may be incorrect or misleading.

Let Di1, Di2,…,Din, be the samples of bridge responses and n be the sample size,

then N   i 1 ni is the total sample size and k the number of groups to be compared.
k

ANOVA can be used to test the hypothesis (Figure 4.2)

61
 H 0 :  D1   D2  ...   Dk
(4.3)
H1  ( H 0 ) c (or i   j for atleast one pair i, j )

The total sum of squares (SST) is represented as a sum of the treatment sum of

squares (SSTr) and the sum of squares due to error

SST  SSTr  SSE

k ni k k ni (4.4)
 (D
i 1 j 1
ij  D) 2   n (D
i 1
i i  D ) 2   (Dij  Di ) 2
i 1 j 1

The test H 0 (treatment effect is zero) is based on the F–statistic,

SSTr / (k  1)
F (4.5)
SSE / ( N  k )

with k-1 and N – k degrees of freedom. The null hypothesis is rejected if the F - statistic is

large compared to (1 - ) quantile of an F – distribution. The results can be inferred more

easily though p - value.

p  value 1  Fcdf  F , k  1, N  k  (4.6)

The p–value is the evidence against a null hypothesis or the probability that the

variation between groups occurred by chance. The p–value can be interpreted as the

probability of such an 'extreme' value of the test statistic when H0 is true.

Hypothesis
H0: D1 =  D 2 ,…,=  D k
Frequency

Ha:  D 1 ≠  D 2 ≠,…,  D k

… n
 2

Figure 4.2 – ANOVA hypothesis.

62
4.3 ANCOVA based grouping

Analysis of covariance (ANCOVA) mainly checks whether two or more population

means are equal, similar to analysis of variance (ANOVA). However, ANCOVA

formulation includes the bridge response, bridge sub-class, and ground motion intensity

(covariate). The first step in performing ANCOVA is to establish a linear regression

relationship between the bridge response and the intensity measure for different bridge sub-

classes (or PSDMs) and to identify whether slopes of the regression lines are significantly

different. If the slopes are not significantly different, a regression line is drawn through

each group of points with the same slope. The intercept of the regression lines is then

checked, and if they are different, it can be concluded that the response of the bridge sub-

classes is different (Figure 4.1). The statistical significance of treatment is more likely to

be identified in an ANCOVA F-test than ANOVA. Two important assumptions in

ANCOVA, in addition to the assumptions in ANOVA, are the independence of the

treatment effect and the homogeneity of regression slopes.

In ANCOVA, the model is assumed as

ln( S d )ij   ln( IM )   si  bs (ln( IM )ij  ln( IM ))   ij , i 1,..., at ; j 1,..., n (4.7)

where ln(Sd) is the response variable, ln(IM) the overall population mean of the intensity
1
measure, at the number of treatments, n the common sample size, ln( IM )   ln( IM )ij
at n i , j
the

overall mean of ln(IM)’s, bs the regression slope, and si the treatment effect. The error

term ij is assumed to be an independent normal distribution with zero mean and variance

2. ln( IM )i 
1
 ln( IM )ij is the ith bridge attribute effect for the ln(IM)’s. The means ln( S d )
n j

and ln( S d )i are defined analogously to ln( IM ) and ln( IM ) i , respectively. The sum of

squares and mixed-product sums can be calculated as

63
 at n at n
Sln( IM ) ln( IM )   (ln( IM )ij  ln( IM )) 2 , Sln( IM ) ln( Sd )   (ln( IM ) ij  ln( IM ))(ln( S d ) ij  ln( S d ))
i 1 j 1 i 1 j 1

at n at
Sln( Sd ) ln( Sd )   (ln( S d )ij  ln( S d )) 2 , Tln( IM ) ln( IM )   (ln( IM ) i  ln( IM )) 2
i 1 j 1 i 1
at at
Tln( IM ) ln( Sd )   (ln( IM )i  ln( IM ))(ln( Sd )i  ln( Sd )), Tln( S
i 1
d ) ln ( Sd )
  (ln( S d ) i  ln( S d )) 2
i 1
(4.8)
at n at n
Qln( IM ) ln( IM )   (ln( IM )ij  ln( IM )i ) 2 , Qln( IM ) ln( Sd )   (ln( IM ) ij  ln( IM ) i )(ln( S d ) ij  ln( S d ) i ),
i 1 j 1 i 1 j 1
at n
Qln( Sd )ln ( Sd )   (ln( S d )ij  ln( S d )i ) 2
i 1 j 1

The estimator of the variance is

2
SSE Q ln( IM ) ln( Sd )
s 2  MSE  , where SSE  Qln( Sd ) ln( Sd )  (4.9)
(at (n  1)  1) Qln( IM ) ln( IM )

If there are no treatment effects, i.e.,  si  0 , then the model is

ln( S d )ij   ln( IM )  bs (ln( IM )ij  ln( IM ))   ij , i 1,..., at ; j 1,..., n (4.10)

In this case, the error sum of squares is

S 2ln( IM ) ln( Sd )
SSE '  Sln( Sd ) ln( Sd )  , with at n  2 degrees of freedom (4.11)
Sln( IM ) ln( IM )

Thus, the test H 0 :  si  0 (treatment effect is zero) is based on the F–statistic,

( SSE ' SSE ) / (a t  1)

F (4.13)
SSE / (at (n  1)  1)

which has an F–distribution with at  1 and at ( n  1)  1 degrees of freedom.

The p–value of the F–statistic (upper trail area of the F–distribution) can be calculated as

p  value 1  Fcdf  F , at  1, at (n  1)  1 (4.14)

64
where Fcdf is the cumulative distribution function of the F–distribution. The p–value is the

evidence against a null hypothesis, H 0 :  si  0 ; in other words, the effect of treatment in

Equation 4.7 is zero. The inclusion of ground motion intensity in the grouping of bridge

sub-classes has several advantages: there is a higher probability that the test will reject a

false H0; there is a reduction in bias caused by chance differences between groups; and

there is a conditionally unbiased estimate of treatment effects (Mangalathu et al. 2016a;

Huietema 1980). ANCOVA assumes the homogeneity of regression slopes; and if the

relationship assumed in the ANCOVA comparison is non–linear, there will be a significant

reduction in the statistical power (Owen and Froman, 1998). This limitation can be averted

in some cases by a transformation of the data.

4.4 KW-based grouping

The stringent assumptions on the response data placed in ANOVA and ANCOVA

are often difficult to satisfy, which has led the development of non–parametric and semi–

parametric approaches. The non–parametric approaches help to relax the restrictive

assumptions on the ANOVA and ANCOVA. The Kuruskal-Wallis (KW) is a non–

parametric version of ANOVA. KW tests place no restriction on the population data, and

are applicable to complex experiments and messy sampling plans. The KW test statistic

(H) for k bridge sub-classes with sample sizes, n1,…,nk is (Kruskal and Wallis, 1952)

1  k Ri 2 ( N  1) 2 
H'    N  (4.15)
S 2  i 1 ni 4 

where, N   ni , and Ri is the sum of the ranks for the ith sample, and

1  k ni
( N  1) 2 
S2   R N
2
 (4.16)
N  1  i 1
ij
j 1 4 

65
where Ri is the sum of the ranks for the ith sample across the sample size j. If there are no
ties in the data, Equation 4.15 can be simplified to

ni ( N  1) 
2
12 k
1 
H 
N ( N  1) i 1 ni  Ri 
 2 

(4.17)

which approximately follows a  distribution with k  1 degrees of freedom. KW tests the

2

hypothesis that responses from different bridge sub-classes have identical distribution

functions against the alternate hypothesis that the samples differ only with respect to the

median, if at all. Despite all its advantages over ANOVA and ANCOVA, the power of the

KW is restricted in the case of normally distributed data (Vickers, 2005).

4.5 Case Study: Two-and Three-Span Box-Girder Bridges in California

Two-span and three-span box girders bridges, which possess a major portion of the

California bridge inventory (Ramanathan, 2012) are the subject bridges in this study. The

selected bridges were designed and constructed prior to 1970. A typical layout of a two-

span box-girder bridge is shown in Figure 4.3. The current study adopts six different bridge

attributes such as (1) bearing type (elastomeric or rocker bearings), (2) column cross–

section (circular or rectangular), (3) abutment configuration (abutment on piles or on

spread footing), (4) abutment backfill (clay or sand), (5) interior bent type (single–column

or multi–column), and (6) material type of superstructure (reinforced or prestressed

concrete).

66
 Figure 4.3 – General layout of a two–span concrete box–girder bridge.

The analytical modeling of the bridges is explained in detail in Chapter 3. A number

of sources of uncertainties (aleatoric or epistemic) are present in the selected class of

bridges, and are given in Table 4.1, which shows the mean value (), standard deviation

(), and the associated probability distribution of various input variables. These variables

are determined based on an extensive plan review of California bridges. In addition to the

uncertainties related to structures, the uncertainty in ground motions are accounted for

using the suite of ground motions assembled by Baker et al. (2011). The entire suite of

ground motions are scaled by a factor of two (Ramanathan, 2012) to have sufficient

67
response data of IMs higher than Palmdale spectrum (the highest probabilistic design

hazard level in California), and thus the expanded suite of 320 ground motions is used for

the current study. Additionally, the spectral acceleration at 1.0 sec (Sa–1.0s) is the optimal

intensity measure for the class of concrete box–girder bridges (Ramanathan, 2012), and is

adopted as the IM in the current study.

As mentioned before, the objective of this study is to identify whether all of the

grouping techniques lead to similar sub-classes. To compare the responses with different

bridge attributes, the total simulation is split amongst the bridge attributes. For example, to

identify the significance of cross–section on seismic response, 50% of the simulations

(160) are carried out for bridges with rectangular columns, while the remaining simulations

are performed for bridges with circular columns. The response variable and the ground

motion intensity measure are transformed into the lognormal space to produce a linear

relationship between the two (Cornell et al. 2002; Managalathu et al. 2016b). Using the

relationship between the transformed response variable and intensity measure, ANOVA,

ANCOVA, and KW grouping techniques are carried out. In all the grouping methods, p–

values are computed to interpret the results of the hypothesis test. A smaller p–value

indicates stronger evidence for rejecting the null hypothesis (H0), and a cut–off p–value of

0.05 (Mangalathu et al. 2016a) is adopted in the current study. For example, at a cut–off

value of 0.05, if the p–value is less than 0.05, it can be concluded with a 95% degree of

confidence that the variation in the demand measure is not due to random chance, but due

to the influence of the different bridge attributes. p–values less than 0.05 are highlighted in

Table 4.2. The table indicates that the various demand parameters are typically sensitive to

different bridge attributes. Among six bridge attributes, the type of interior bent is a

significant parameter for all cases. It can be explained that the transverse moment demands

in single column bents are higher than the longitudinal moment demands and have less

redundancy than multi–column bents (Priestley et al. 1996).

68
 The grouping pattern identified by the three methods is the same for 80% of the

cases. ANOVA and KW yield the same identification of significant bridge attributes in

96% of the cases. The results of ANCOVA are different from those of ANOVA and KW

in approximately 17% of cases. ANCOVA identifies more design attributes as significant

in comparison to ANOVA and KW tests, which is shown in italics in Table 4.2. While

comparing the ANOVA and KW grouping methods, ANOVA identifies more relevant

parameters. The reason for such a discrepancy is investigated in detail in the following

section.

69
 Table 4.1 – Uncertainty distribution considered in the bridge models
Distribution
Parameter Units
Type μ σ
Concrete compressive strength (fc) MPa Normal 29.03 3.59
Reinforcing steel yield strength (fy) MPa Lognormal 465.0 37.30
Span length (L)
Two-span mm Lognormal 31775 8738
Approach to main span ratio (three-span bridge) – Normal 0.57 0.13
Deck width (Bd)
Single column bent mm Lognormal 9780 1980
Multi-column bent mm Lognormal 11970 2418
Column height (H) mm Lognormal 6625 865
Abutment backwall height (Ha)
Diaphragm abutments
On piles mm Lognormal 3234 488
On spread footings mm Lognormal 2925 1056
Seat-type abutments
On piles mm Lognormal 2186 441
On spread footings mm Lognormal 2186 441
Abutments on piles - Lateral capacity/deck width (Kpa)
Diaphragm abutment N/mm Lognormal 1120 404
Seat-type abutment N/mm Lognormal 1498 540
Abutments on spread footing
Coefficient of friction (as) – Normal 0.40 0.075
Yield displacement (y) mm Uniform 19.0 13.4
Elastomeric bearing pad
Stiffness per deck width (Kb) N/mm/m Lognormal 908 327
Coefficient of friction for bearing pad (b) – Normal 0.30 0.10
Rocker bearing
Coefficient of friction (l, longitudinal direction) – Normal 0.04 0.01
Coefficient of friction (t, transverse direction) – Normal 0.10 0.02
Gap (g)
Longitudinal (btw. deck and abutment wall) mm Lognormal 23.5 12.5
Transverse (btw. deck and shear key) mm Lognormal 12.8 2.58
Mass factor (m) Uniform 1.25 0.007
Damping () Normal 0.045 0.0125
Acceleration for shear key capacity (as) g Lognormal 1.00 0.20
Longitudinal reinforcement ratio () (%) Uniform 2.25 0.52
Pile group – pile cap and piles
Translational stiffness (Kft)
Single column – 1% long. rebar N/mm Normal 297716 140101
Single column – 3% long. rebar N/mm Normal 245178 105076
Multi column – 1.5% long. rebar N/mm Normal 140101 105076
Rotational stiffness (Kfr)
Single column – 1% long. rebar N-m/rad Normal 4.5109 1.1109
Single column – 3% long. rebar N-m/rad Normal 6.8109 1.1109

70
 Table 4.2 – p–values by ANCOVA, ANOVA and KW test
Column (μϕ) Passive (δp) Active (δa) Transverse (δt) Unseating (δs) Bearing (δb)
Ty
Bridge attribute ANC ANO ANCO ANO ANCO ANO ANCO ANO ANCO ANO ANCO ANO
pe KW KW KW KW KW KW
OVA VA VA VA VA VA VA VA VA VA VA VA
Column cross–
Diaphragm abutment

0.000 0.000 0.000 0.834 0.667 0.567 0.897 0.720 0.631 0.908 0.723 0.686 – – – – – –
section
Interior bent 0.000 0.000 0.000 0.038 0.233 0.326 0.046 0.315 0.417 0.000 0.000 0.000 – – – – – –
Abutment 0.000 0.000 0.000 0.494 0.354 0.322 0.382 0.359 0.332 0.000 0.021 0.033 – – – – – –
– – – – – –
Two–span bridges

Backfill 0.458 0.635 0.856 0.000 0.000 0.000 0.000 0.000 0.000 0.018 0.209 0.387
Superstructure 0.438 0.775 0.875 0.715 0.464 0.639 0.832 0.467 0.680 0.703 0.230 0.269 – – – – – –
Column cross–
0.043 0.317 0.343 0.149 0.446 0.554 0.463 0.727 0.800 0.005 0.048 0.064 0.263 0.466 0.509 0.116 0.223 0.318
section
Seat–type abutment

Bearing 0.005 0.003 0.002 0.043 0.008 0.001 0.124 0.031 0.003 0.008 0.002 0.000 0.000 0.000 0.000 0.002 0.000 0.008
Interior bent 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.007 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Abutment 0.302 0.988 0.943 0.252 0.052 0.038 0.469 0.125 0.085 0.116 0.507 0.548 0.600 0.200 0.209 0.194 0.494 0.315
Backfill 0.045 0.032 0.012 0.148 0.401 0.478 0.145 0.486 0.537 0.549 0.892 0.831 0.722 0.844 0.887 0.641 0.667 0.747
Superstructure 0.631 0.413 0.321 0.039 0.609 0.686 0.084 0.583 0.667 0.036 0.643 0.646 0.001 0.438 0.451 0.351 0.600 0.600
Column cross–
0.010 0.135 0.070 0.690 0.882 0.873 0.709 0.897 0.859 0.426 0.405 0.470 – – – – – –
section
Diaphragm
abutment

Interior bent 0.000 0.000 0.000 0.000 0.093 0.151 0.000 0.077 0.128 0.000 0.000 0.000 – – – – – –
Abutment 0.095 0.520 0.356 0.317 0.298 0.529 0.440 0.368 0.620 0.441 0.954 0.953 – – – – – –
Three–span bridges

Backfill 0.799 0.637 0.771 0.000 0.002 0.002 0.000 0.007 0.005 0.228 0.568 0.626 – – – – – –
Superstructure 0.155 0.347 0.498 0.419 0.850 0.807 0.314 0.763 0.723 0.877 0.593 0.551 – – – – – –
Column cross–
0.097 0.451 0.459 0.582 0.548 0.625 0.858 0.853 0.881 0.001 0.027 0.025 0.637 0.907 0.935 0.669 0.569 0.914
Seat–type abutment

section
Bearing 0.258 0.378 0.482 0.005 0.043 0.357 0.000 0.002 0.048 0.006 0.066 0.173 0.000 0.000 0.003 0.128 0.168 0.754
Interior bent 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Abutment 0.003 0.104 0.065 0.002 0.334 0.739 0.000 0.153 0.376 0.000 0.174 0.158 0.000 0.000 0.000 0.014 0.050 0.034
Backfill 0.000 0.001 0.001 0.388 0.565 0.952 0.918 0.958 0.836 0.142 0.272 0.259 0.001 0.059 0.072 0.023 0.117 0.058
Superstructure 0.172 0.961 0.678 0.088 0.776 0.815 0.067 0.885 0.994 0.005 0.321 0.327 0.004 0.718 0.779 0.083 0.732 0.931

71
4.6 Comparison of various grouping techniques

ANCOVA shows that some of the design attributes are significant, while ANOVA and KW

tests identify the same attributes as non–significant, as noted in Table 4.2. This non–significance

can be interpreted in two ways. Firstly, the lack of attribute significance may be obtained because

there is indeed very little or no true effect of the design attribute on the seismic response.

Alternatively, the finding suggesting a lack of significance can be attributed to the grouping

technique having less power to identify the effects of the design attribute (Huitema, 1980). To

evaluate the power and efficiency of various grouping techniques, the current study selects three

bridge sub–classes: (1) three-span diaphragm abutment bridges grouped by the cross–section shape

based on the curvature ductility response, (2) two-span seat-type abutment bridges grouped by the

cross–section shape based on the transverse abutment response, and (3) two-span seat-type

abutment bridges grouped by the abutment type based on the passive abutment response. The three

sub classes are carefully chosen such that one of the grouping techniques identifies the design

attribute as non–significant.

4.6.1 Case 1: Significant per ANCOVA

For the three-span diaphragm abutment bridge, two sub–classes are formed based on the

curvature ductility response according to the column cross–section: (1) with circular cross–section

(hereafter, 3SDC) and (2) with rectangular cross–section (hereafter, 3SDR). The ANCOVA result

(p–value of 0.010) reveals that the cross–section is significant for this bridge, while the ANOVA

(p–value of 0.135) and KW (p–value of 0.070) tests conclude it is not significant. Figure 4.4 shows

the data analysis results of the curvature ductility for 3SDC and 3SDR. The power of ANOVA and

the ANCOVA depends on the three or four assumptions and is evaluated initially. One of the

critical assumptions in both ANOVA and ANCOVA is that the data is normally distributed. The

normality of 3SDC and 3SRC is checked using the Kolmogorov–Smirnov test (KS test)

(Kolmogorov, 1933), which identifies the null hypothesis that the data is normally distributed. If

72
the p–value is less than  = 0.05, the null hypothesis is rejected. Thus, there is enough evidence

that the data do not follow a normally distributed population. The p–value of normality check for

3SDC and 3SDR is 0.067 and 0.099, respectively, and hence one fails to reject the null hypothesis.

The assumption that each observation is mutually independent is valid because of the random

sampling and pairing of the bridge and ground motions. The validity of the assumption of the

independence of the treatment effect lies in the fact that the different bridge attributes are

independent of each other. ANCOVA has one additional assumption regarding the homogeneity

of regression slopes. As the demand measure increases with the intensity of ground motions, it

implies a monotonic relationship between the two.

Figure 4.4 – Histograms and descriptive statistics for case 1: a) 3SDC, b) 3SDR, c) box plot
of 3SDC and 3SDR, and d) ANCOVA regression lines of 3SDC and 3SDR.

The homogeneity of regression slopes is checked in the current study using the F-test for

the equality of slopes (Vidakovic, 2005) and also seems to hold. ANCOVA and ANOVA have a

73
greater power than the KW test when the above assumptions are satisfied (Vidakovic, 2005;

Vickers 2005).

Figure 4.4 (c) shows box plots of the curvature ductility for 3SDC and 3SDR. The mean

value of intensity measure in the logarithmic scale for 3SDC and 3SDR is –1.14 and –1.21,

respectively. The ground motions are randomly assigned in this study because it is computationally

intensive and almost impractical to produce responses for all the cases with the same suite of

ground motions. The experiment has been designed in a way that the design attributes are randomly

sampled in the 320 simulations from a practical perspective. Such an assignment also helps assess

the combined effect of two or more different design attributes; for example, the combined effect

of the abutment type and column cross-section. The question arising from this case is whether the

column response of 3SDC and 3SDR can be compared as there is a clear variation in the IM and

the column response is highly related to IM. It is possible that the results are skewed if the column

responses are compared as the mean values of IM are different. ANOVA and KW tests neglect

such variation in IM and such a comparison leads to erroneous results. The mean value of the

curvature ductility for3SDC and 3SDR is shown in dashed lines in Figure 4.4(b). It can be seen

from Figure 4.4 (c) and Figure 4.4(d) that the mean values are very close; their comparison without

reflecting the effect of IM leads to erroneous conclusions. ANOVA and KW tests neglect the

variation of IM in their comparison and this ignorance leads to p–values higher than 0.05. On the

other hand, ANCOVA compares the treatment means after adjusting the variation in IM, i.e., what

would be the response means, if the two bridge sub-classes have the same IM. In other words,

ANCOVA tests the null hypothesis whether the ‘adjusted population’ means are equal. It can also

be formulated as whether the regression intercept of the two bridge sub-classes is equal

(Mangalathu et al. 2016a; 2017a). Under the assumption of the homogeneity of regression slopes,

the difference between the intercept means is equal to the difference between the adjusted means.

The regression lines of the curvature ductility for 3SDC and 3SDR are plotted in Figure 4.4(d).

ANCOVA yields the cross–section as significant (p–value of 0.010) because of the difference

74
between the intercepts. The adjusted and unadjusted means are equal only when the IM means are

same. In such a scenario, ANOVA and ANCOVA yield similar results.

4.6.2 Case 2: Significant per ANOVA and ANCOVA

Figure 4.5 – Data analysis for case 2: a) histogram of 2SSC, b) histogram of 2SSR, c) box
plot of 2SSC and 2SSR, and d) ANCOVA regression lines of 2SSC and 2SSR.

The two-span seat-type abutment bridges grouped by the cross–section shape based on the

transverse abutment response are 1) with circular cross–sections (hereafter, 2SSC) and (2) with

rectangular cross–sections (hereafter, 2SSR). ANOVA and ANCOVA identify this cross-section

shape as a significant design attribute with p–value of 0.005 and 0.048, respectively. The KS test

shows a p–value greater than 0.015 and 0.016 for 2SSC and 2SSR, and hence fails to reject the

null hypothesis that they are normally distributed. As seen from Figure 4.5(c) and Figure 4.5(d),

there is a statistically significant difference between the mean values, and both ANOVA and

ANCOVA method can capture this difference. The reason why the KW test suggests it as non-

75
significant might be due to the limited power when the data is normally distributed, which has also

been pointed out by other researchers (Vickers 2005; Kvam and Vidakovic 2007).

4.6.3 Case 3: Significant per KW

Figure 4.6 – Data analysis for case 3: a) histogram of 2SSS, b) histogram of 2SSP, c) box
plot of 2SSS and 2SSP, and d) ANCOVA regression lines of 2SSS and 2SSP.

The two span seat-type abutment bridges are categorized by the abutment type based on

the passive response of the abutments (1) with spread abutments (hereafter, 2SSS) and (2) with

abutment on piles (hereafter, 2SSP). The results (Figure 4.6) underscore the importance of

ANCOVA–based grouping compared to other methods. The KW identifies abutment type as a

significant design attribute, while other grouping methods suggest it as non–significant. The mean

value of IM (in a logarithmic scale) for the passive abutment action of 2SSS and 2SSP is –1.31

and –1.13, respectively, and the comparison of the adjusted means identifies this attribute as non–

significant (Figure 4.6d). Although ANOVA identifies it as insignificant, the KW test fails here.

76
 All of the above cases identify the statistical power of ANCOVA to provide an unbiased

estimator compared to ANOVA and KW methods by accounting for the variation in IMs. Although

not shown here, similar conclusions are also drawn for the other cases. To evaluate whether the

bridge sub-classes identified by ANCOVA yield a distinct response, fragility curves are developed

in the next section.

The following points can be inferred from the comparison of various grouping techniques:

1. ANCOVA compares the linear regression between the component response and the

intensity measure. In a logarithmic space, it corresponds to probabilistic seismic demand

models (PSDMs). The PSDMs are one of the core steps in the generation of fragility curves.

Thus, ANCOVA compares the PSDMs and the group’s bridge sub-classes based on the

difference in PSDMs. ANOVA compares the responses based on mean value of the

responses.

2. ANOVA and ANCOVA yield similar results when the mean value of ground motions

associated with the groups to be compared is the same. If there is a variation in the IMs

mean value, ANCOVA is more likely to catch the significant parameter than ANOVA.

3. The normality assumptions in ANOVA and ANCOVA seem to be satisfied in the case of

seismic responses of bridges. Hence, the statistical power in identifying the significant

attributes is greater in ANOVA and ANCOVA than the KW. It seems from the current

study that the KW is not a reliable performance based grouping approach for bridges.

4. The pairwise comparison is difficult in the case of ANCOVA than ANOVA.

5. Where the ground motions associated with the bridge groups to be compared are the same,

ANOVA yields the same results as ANCOVA.

77
 In light of the inferences from the various grouping strategies, the current study adopts

ANOVA from the point of pair-wise comparisons and the extent of numerical simulations needed

to group the California bridge inventory.

The grouping strategy adopted in the current study is given below:

Step 1: Select possible combinations of bridge configurations.

Step 2: Using Latin Hypercube Sampling method, select N ground motions from the suite of

ground motions assembled for the fragility analysis. The ground motions are selected based

on the distribution of Sa-1.0s of ground motions.

Step 3: Analyze each bridge configuration in OpenSees for the selected N ground motions.

Step 4: Collect output of interest (or response) including curvature ductility, bearing displacement,

abutment active/passive/transverse displacement, etc.

Step 5: Conduct an ANOVA to evaluate the sensitivity of each component to the variation in

bridge configurations. The results can be inferred more easily though p–value. The p–value

is the evidence against a null hypothesis or the probability that the variation between groups

occurred by chance. The p–value can be interpreted as the probability of such an 'extreme'

value of the test statistic when H0 is true.

Step 6: Perform Fischer Method on the ANOVA output to group the bridge configurations that

have a statistically similar response. The Fisher Method compares all pairs of groups while

controlling the individual error rate. It identifies the group with the highest sensitivity and

checks a null hypothesis whether the mean values of other groups match with the most

sensitive one. If there is a match, they will be grouped together. If not, it will check the

group with the second highest sensitivity and check whether the mean value of the

78
 remaining groups matches the group with second highest sensitivity. The procedure is

repeated until all the members are grouped.

4.7 Grouping of bridge classes

ANOVA-based grouping strategy is used to group the bridge classes in California. To

group the bridge classes, the uncertain parameters in Table 4.1 are kept at their mean values.

Nonlinear time history analysis (NLTHA) is carried out for the bridge models for the selected

ground motions and the maximum response of the various bridge components are recorded. Thirty

ground motions are selected from the expanded suite of 320 ground motions by Latin Hypercube

Sampling (LHS) for the grouping of bridge classes and is chosen based on a sensitivity study. The

histogram of the peak ground acceleration (PGA) and the acceleration response spectrum of the

ground motion suite are shown in Figure 4.7.

Figure 4.7 – a) Histogram of the PGA values of the ground motion suite, b) Acceleration
response spectrum of the ground motion suite.

Various demand parameters such as column curvature ductility (μϕ, -), passive abutment

displacement (δp, mm), active abutment displacement (δa, mm), transverse abutment displacement

79
(δt , mm), bearing displacement (δb, mm), and superstructure unseating (c, mm) are used to group

the bridge classes. As mentioned before, the results are inferred in terms of p–value. A smaller p–

value refers to stronger evidence for rejecting the null hypothesis (H1). If the p–value is less than

0.05, it can be concluded that not all of the population means are equal. The sensitivity of the

seismic demand on two-span bridge configurations to the various design eras and bent

configurations are evaluated using ANOVA and is given in Table 4.3. It is clearly observed from

Table 4.3 that all of the demand parameters in rigid abutments are highly sensitive to the design

eras and bent configurations, and hence cannot be grouped together. In the case of seat abutments,

column curvature ductility and bearing displacement are the most sensitive parameters affected by

the design eras and bent configurations. The Fischer Method is carried out on ANOVA to group

the bridges that have similar seismic demands; those results are shown in Table 4.3.

Table 4.3 – p–values from ANOVA.

p–value
Bridge type
μϕ δp δa δt δb
Rigid 0.000 0.001 0.002 0.001 –
abutments
Seat abutments 0.037 0.857 0.888 0.503 0.050

Table 4.4 presents the Fischer Method grouping results for two span box-girder bridges

and Table 4.5 shows the grouping results of Era 11 bridges to number of spans. The inferences

obtained from the sensitivity study results, presented in Table 4.4 and Table 4.5, are summarized

below:

1. For the two-span seat and diaphragm abutment bridges, Era 11 shows behavior that is

distinct from the other design eras (Table 4.4). It can be inferred that the changes in the

seismic design philosophy from Era 22 to Era 33 don’t significantly change the seismic

demand of bridge components.

80
2. In the case of two-span bridge configurations, the seismic demand of columns (μϕ) is the

component greatly influenced by the design eras and number of columns per bent. It

requires special attention as column vulnerability governs bridge vulnerability in most

bridge configurations (Mangalathu et al. 2016a).

3. By comparing the seismic demand on abutments, it can be inferred that the bridge design

philosophy and number of columns per bent have more influence on the diaphragm

abutment bridge than the seat abutment bridge. It might be due to the integral connection

of the diaphragm abutment bridges at the ends which causes the abutment to share a

significant portion of the seismic demand. In the case of seat abutment bridges, the seismic

demand on the abutments is less influenced by the design eras and number of columns per

bent.

4. The seismic demand of single–column bent, two– column bent, and multi–column bent

(bent with greater than two columns) bridges are statistically different from the selected

two span bridge configurations (Table 4.5) and thus cannot be grouped together from a

seismic demand perspective.

5. In the case of diaphragm abutment bridges, two-span bridges have distinct seismic

demands that are distinct from three-to six-span bridge configurations for all bridge

components. Although not shown here, similar conclusions have also been noted for

bridges with seat abutments.

6. The bridges with diaphragm abutments and seat abutments have different seismic demand

for all of the design eras and bent configurations.

81
 Table 4.4 – Results of the grouping for two-span box girder bridges.
Column ductility Abutment passive Abutment active Abutment Bearing/ Unseating
Bridge configurations transverse displacement
Mean* Grouping# Mean Grouping Mean Grouping Mean Grouping Mean Grouping
Era 11 - 1 column bent * 0.86
value B C 0.00 B 0.08 Method
B 0.75 MethodC – – –
Era 11 - 2 column bent (µ)
1.96 A 0.39 A 0.45 A 1.58 A – – –
– – –
Diaphragm abutments

Era 22 - 1 column bent 0.10 D 0.00 B 0.08 B 0.67 C

Era 22 - 2 column bent 1.13 B 0.45 A 0.51 A 1.17 B – – –
Era 22 - 3 column bent 0.90 B C 0.42 A 0.48 A 1.00 B C – – –
Era 22 - 4 column bent 0.73 B C 0.48 A 0.53 A 0.90 B C – – –
Era 33 - 1 column bent 0.10 D 0.00 B 0.09 B 0.67 C – – –
Era 33 - 2 column bent 1.13 B 0.45 A 0.51 A 1.17 B – – –
Era 33 – 3 column bent 0.91 B C 0.42 A 0.48 A 1.01 B C – – –
Era 33 - 4 column bent 0.73 B C 0.48 A 0.53 A 0.90 B C – – –
Era 33 - 5 column bent 0.65 C 0.49 A 0.54 A 0.83 B C – – –
Era 11 - 1 column bent 1.26 B -0.02 A 0.01 A -0.05 A 0.94 B
Era 11 - 2 column bent 1.99 A 0.32 A 0.35 A 0.65 A 1.60 A
Era 22 - 1 column bent 0.51 C 0.07 A 0.10 A -0.02 A 0.93 B
Seat abutments

Era 22 - 2 column bent 1.26 B 0.36 A 0.39 A 0.42 A 1.22 A B

Era 22 - 3 column bent 1.03 B C 0.31 A 0.33 A 0.33 A 1.09 B
Era 22 - 4 column bent 0.92 B C 0.25 A 0.28 A 0.23 A 1.04 B
Era 33 - 1 column bent 0.51 C 0.07 A 0.10 A -0.03 A 0.93 B
Era 33 - 2 column bent 1.26 B 0.36 A 0.39 A 0.42 A 1.22 A B
Era 33 – 3 column bent 1.04 B C 0.31 A 0.33 A 0.33 A 1.09 B
Era 33 - 4 column bent 0.92 B C 0.25 A 0.27 A 0.23 A 1.04 B
Era 33 - 5 column bent 0.76 B C 0.23 A 0.24 A 0.06 A 0.99 B

*Mean is shown in logarithmic scale. #Bridge configurations that do not share common alphabet cannot be grouped together as the seismic demand of these bridges are statistically
different.

82
 Table 4.5 – Results of the grouping for multi-span bridges
Column ductility Abutment passive Abutment active Abutment Bearing/Unseating
Bridge configurations transverse displacement
Mean* Grouping# Mean Grouping Mean Grouping Mean Grouping Mean Grouping
Era 11 – 2 span * 1.26
value A -0.02 A 0.01 AMethod -0.05 AMethod 0.93 A
Era 11 – 3 span (µ)
0.82 A B -0.49 A B -0.47 A B -0.70 A B 0.61 A B
Era 11 – 4 span 0.55 B -0.88 B -0.87 B -0.85 B 0.40 B b B
Era 11 – 5 span 0.53 B -0.99 B -0.97 B -0.97 B 0.36 B
Era 11 – 6 span 0.52 B -1.04 B C -1.02 B -1.03 B 0.36 C
B

*Mean is shown in logarithmic scale. #Bridge configurations that do not share common alphabet cannot be grouped together as the seismic demand of these bridges are statistically
different.

Figure 4.8 – Proposed classification scheme.

83
 It is noteworthy to mention that only the abutment types, design code eras, interior

support and number of columns per bent, column cross-section, span-range and frame-

system are adopted as design attributes in this study to group bridge classes. These

attributes are selected based on the current sensitivity study, insights from the previous

research on the bridge’s seismic responses for regional risk assessment (Shinozuka et al.

2000; Mackie and Stojadinovic 2001; Avsar et al. 2001; Choi 2002; Nielson 2005; Padgett

2007; Ramanathan 2012; Moschonas et al. 2008; Banerjee and Shinozuka 2008; Mehr and

Zaghi 2016; Managalathu et al 2016a, 2016b; Zelaschi et al. 2016) and the input from

Caltrans (Caltrans, 2017). Figure 4.8 shows the proposed grouping; classification is carried

out based on the abutment type, column cross–section, pier type, number of spans, span

continuity and seismic design. They are explained below:

• Abutment types: It has been noted that the response of rigid abutments is different

from seat type abutments and hence cannot be grouped together.

• Column cross-section: On the basis of the column cross-section shape, the bridges are
classified into circular, rectangular, and oblong bridges. Soleimani et al. (2017) showed
different demands for bridges with circular, rectangular, and oblong cross sections.

• Interior support and number of columns per bent (Pier type): Sensitivity results

showed that the number of columns greater than three in a multi-column support does

not significantly impact response and therefore does need to be considered as separate

classes. The responses for bridge models having from 3-5 columns-per-bent were most

consistent (MCB, hereafter) and can be grouped together. Responses for 2-column

bents (TCB, hereafter) were statistically different from MCB and hence grouped

separately. The response of bridges supported on single-column bents (SCB, hereafter)

was shown to be distinct for each era and therefore cannot be grouped with other

support systems. Although pier-wall supports (PW, hereafter) were not explicitly

84
 modelled, engineering judgment indicated this support system would also yield distinct

responses.

• Design code era: Sensitivity results showed that bridges built or rebuilt within either

of the two later design code eras (i.e. Era 22 and Era 33) had statistically similar

responses and could be grouped as ‘E22/E33’ Era for purposes of establishing demand

models (note: capacity models, particularly for the columns, are different for these eras

and hence the fragilities). Era 11 bridges were shown to have distinct response (also

capacity) and require the development of separate demand models. Mixed-era bridges

are not addressed in this study.

• Span range: Single-span bridges need to be treated as a separate class due to their

unique and limited combination of demand parameters. Sensitivity studies considered

single-frame systems having more spans (from 2-span to 6-span) to determine if any

range could be grouped. Depending on other factors such as era and abutment, the

responses varied from being similar for all span ranges to being distinct for various

combinations. As a practical compromise, 2-span bridges (S22, hereafter) were to be

treated separately, while span-groups of two (i.e. three- and four-span bridges (S34,

hereafter), and spans greater than 4 (S4x, hereafter) were adopted for longer bridges.

• Multi-Frame System: The sensitivity studies also considered various simplified

framing configurations. Responses for 2-frame systems were clearly unique, but the

distinctions between bridges with a higher numbers of frames were less clear. The ‘two-

frame’ system was therefore retained as a distinct class. Larger numbers of frames such

as three frames four frames et al., are combined into the ‘multi frame’ It is noteworthy

to mention that very less number of multiple frame system are noted from the extensive

plan review and such a compromise is reasonable.

85
4.8 Conclusion

Regional risk assessment relies on fragilities that are applicable to a portfolio of

structures, as it is cumbersome and time consuming to generate fragility curves for each

structure in a specific region. Also the generation of each structure-specific fragility curve

is not warranted as some structures have similar performance or fragilities. Currently, the

grouping of structures is typically conducted based on engineering judgment and there is a

lack of systematic strategy for binning/grouping structures. These limitations can be

addressed by performance based grouping techniques, which lead to more reliable sub-

classes of bridges relative to the traditional subjective lumping of bridges. The current

study explores the various performance based grouping techniques such as analysis of

variance (ANOVA), analysis of Covariance (ANCOVA) and non–parametric Kruskal-

Wallis test (KW) for the grouping of structures of similar performance. The selected

grouping methods have different underlying assumptions and approaches in grouping the

structures. The comparison of various grouping techniques is carried out with the case

study of two span and three span concrete box girder bridges in California with seat-type

and diaphragm abutments.

In the light of these studies, a new performance-based grouping methodology utilizing

analysis of variance (ANOVA) is suggested in this chapter. Thirty ground motions are

selected by Latin hypercube sampling from the suite of ground motions assembled for

California for the nonlinear time history analysis (NLTHA) of bridge models. Consistent

with the ground motions, 30 three-dimensional bridge models are created in OpenSees and

the maximum component responses are noted for each NLTHA. ANOVA is carried out for

the recorded maximum bridge responses to determine if the mean responses are statistically

similar. If they are not, the bridges are grouped by pairwise comparison using the Fischer

method.

86
 The insights from the performance based grouping method, Caltrans engineering

experience regarding bridge-design features affecting seismic performance, and the

population size of the sets of bridges sharing a particular combinations of design features

are used to group the of box-girder bridge classes in California.

87
 CHAPTER 5 CALIFORNIA BRIDGE INVENTORY

To have a reliable estimate of the vulnerability or fragility of highway

bridges in California, it is necessary to understand and characterize the California bridge

inventory. This chapter presents an in-depth study of the California bridge inventory using

the in-house database called BIRIS, assembled by California Department of Transportation

(Caltrans) engineers.

Bridge design philosophy in California has been significantly influenced by the

historic 1971 San Fernando and the 1989 Loma Prieta earthquakes. Based on the unique

design attributes and the evolution in seismic design philosophy, California bridges can be

separated by three design eras: Era 11 (pre-1971), Era 22 (1971-1990), and Era 33 (post-

1990) (Ramanathan, 2012). In Era 11 seismic design philosophy, seismic forces were

proportional to the dead weight of the structure. Bridges were designed for a lateral seismic

force equal to 6% of the dead weight of the structures. Column shear failure and pull-out

of the longitudinal reinforcement were predominant due to the lack of ductility, as was

revealed in the 1971 San Fernando earthquake. After the 1971 San Fernando earthquake,

capacity design principles were added to the design standards. The lateral load-carrying

capacity of the bridges was increased by a factor of 2 or 2.5 and the aspects of fault

proximity, site conditions, dynamic structural response, and ductile details were considered

in the design of bridge columns. However, column shear failure in the plastic hinge regions

was typical in Era 22 due to the lack of confinement in this zone. The extensive damage

from the 1989 Loma Prieta earthquake forced Caltrans to solicit the Applied Technology

Council (ATC) to conduct a detailed study and to provide recommendations for design

standards, performance criteria, and practices. The recommendations from ATC described

in ATC-32 were incorporated in the Caltrans design manuals and led to the Era 33 design

details. The fundamental emphasis in Era 33 design philosophy was on displacement-based

88
or capacity design approach, which ensures a ductile failure mode in the columns. A

detailed review of the design details pertinent to various design eras is presented in this

chapter.

5.1 Bridge classification based on BIRIS

BIRIS is the bridge inventory assembled by Caltrans engineers for the purpose of

having a unified database for bridges; it contains bridge types, materials, operational

conditions, geometric data, functional details and other data. Table 5.1 shows the

distribution of various bridge classes in California obtained from BIRIS. Bridges are

classified based on the material and the bridge type. Box girder bridges account for the

majority of the California bridge inventory. The current study is limited to concrete box

girder, which accounts for more 30% of the California bridge inventory.

Table 5.1 – Bridge classes in California and their proportion in the overall
inventory.
Materials Total
Concrete Steel Mixed
Box girder 7839 23 166 8028
Tee Girder 2901 0 16 2917
I girder 1015 2133 603 3751
Slab 5703 15 64 5782
Culvert 3307 264 22 3593
Others 837 378 798 2013
Total: 26084

5.2 Box girder bridge class statistics

The various components of a three-span box girder bridge are illustrated in Figure

5.1. An extensive plan review was carried out to identify the design attributes and are

explained below:

89
 Figure 5.1 – Illustration of major bridge components.

5.3 Abutments

Abutments can be classified into two basic types: seat abutments and diaphragm

abutments (Ramanathan 2012). Diaphragm abutments are cast monolithic with the

superstructure. As the diaphragm abutments readily engage the backfill soil during the

seismic action, it provides a great source of energy dissipation and reduces the likelihood

of span unseating. Seat-type abutments provide a bearing support to the superstructure,

which is restrained longitudinally by the abutment backwall and transversely by the shear

key. Mixed abutments are supported by a diaphragm abutment at one end and seat abutment

at other end. The distribution of the abutments for various design eras are shown in Figure

5.2. Rigid abutments are the most common type in Era 11, while seat abutments are more

common in Era 33. Era 22 contains an approximately equal mix of rigid and seat abutments.

90
 13% 1%
9%

45%
86% 46%

Era 22
Era 11
1% 7%
Diaphragm
Seat
92% Mixed

Era 33

Figure 5.2 – Distribution of abutments for various eras.

Based on the support type, the abutment can be (1) on piles, (2) on spread footing

or (3) cantilever type. The distribution of the abutment based on the support type is shown

in Figure 5.3. Abutments resting on piles are the major configuration for both diaphragm

and seat-type abutments.

91
 Diaphragm abutments (Era 11) Seat abutments (Era 11)
5%
Resting Resting
on piles on piles
31%
34% Spread 44% Spread
61% footing footing
Cantileve Cantilev
r type 25%
er type

Diaphragm abutments (Era 22) Seat abutments (Era 22)

0%
6%
Resting Resting on
on Piles Piles
29%
Spread Spread
footing Footing
71% Cantilever 25% 69%
type Cantilever
type

Diaphragm abutments (Era 33) Seat abutments (Era 33)

0% 8%
Resting on Resting
Piles on Piles
Spread 46% Spread
100% footing Footing
46%
Cantilever Cantilev
type er type

Figure 5.3 – Distribution of abutments for various bridge eras.

92
 Table 5.2 shows the statistics summary of abutment parameters for various design

eras and abutment types. The development of abutment parameter values was based on

manual review of bridge details found in plans downloaded through the BIRIS search. It is

seen from Table 5.2 that the abutment backwall height varies depending on the design era,

abutment type, and the abutment support type. Caltrans 2014 draft of bridge design aids

(BDA, hereafter) on ‘Permissible Horizontal Loads for Standard Plan and Steel HP Piles’

(Caltrans, 2017) was used to establish a typical value and representative range of pile

capacity for various types of standard piles in both sands and clays. BDA defines the

permissible load corresponding to a deflection of 0.25-inch. The mean value of

approximately 25 kips per pile was determined from BDA for a 5-foot cutoff, and the

representative range of 15-50 kips per pile showed the variability to be approximately a

factor of two above and below the central value. These ranges of values are used to derive

the abutment pile stiffness and are given in Table 5.2. To derive the coefficient of friction

for abutments on spread footing, concrete-soil and concrete-concrete friction coefficient

were considered (Potyondy, 1961), and the range is modeled as a normal distribution with

mean of 0.40 and standard deviation of 0.075. A slip as little as 0.04-0.10 inch can mobilize

the full friction in an abutment on spread footing and hence yield displacement is assumed

to be uniformly distributed with an upper bound of 0.10 in. and lower bound of 0.04 in.

93
 Table 5.2 – Distribution of parameters for abutments.
Parameter Design Type of abutment Units Distribution
era Type Parameters† Lower Upper
§ Mea Standard bound Bound
n () Deviatio (L) (U)
n ()
On piles feet LN 2.35 0.15 8.0 14.0
Diaphragm On spread feet LN 2.20 0.35 4.5 18.0
Cantilever feet U 25.0 8.33 20.0 30.0
On piles feet LN 1.95 0.20 5.0 10.5
Era11
Seat On spread feet LN 1.95 0.20 5.0 10.5
Cantilever feet U 25.0 8.33 20.0 30.0
On piles feet LN 2.39 0.20 6.5 13.0
Diaphragm On spread feet LN 2.37 0.09 9.5 12.5
Abutment
Cantilever feet - - - - -
backwall
On piles feet LN 2.45 0.18 9.5 20.0
height Era22
Seat On spread feet LN 2.50 0.09 10.5 14.5
Cantilever feet - - - - -
On piles feet LN 2.45 0.18 9.5 20.0
Diaphragm On spread feet - - - - -
Cantilever feet - - - - -
On piles feet LN 2.63 0.22 10.5 23.5
Era33
Seat On spread feet LN 2.58 0.14 11.0 19.0
Cantilever feet - - - - -
Abutment pile
All
stiffness Diaphragm On piles kip/ft LN 1.79 0.35 2.5 12.0
Eras
(lateral
capacity All
Seat On piles kip/ft LN 2.08 0.35 4.0 16.0
/deck width) Eras
Coefficient of All Diaphragm
On spread _ N 0.40 0.075 0.25 0.55
friction Eras /Seat
Yield All Diaphragm
On spread in. U 0.75 0.02 0.50 1.0
displacement Eras /Seat
Abutment
All Diaphragm
backfill soil All types B Equally split among all simulations
Eras /Seat
(clay vs sand)
§
N = normal, LN = lognormal, U = uniform, and B = Bernoulli distribution.
† and  are the parameters of the distribution. These denote mean and standard deviation for a normal and uniform distribution, and mean
and standard deviation of the associated normal distribution (in log space) in the case of a lognormal distribution.

94
 Table 5.3 – Distribution of parameters for bearings.
Design Type of Parameter Units Distribution
era bearings Type§ Parameters† Lower Upper
Mean Standa bound Bound
() rd (L) (U)
Deviati
on ()
Coefficient of friction
- N 0.04 0.01 0.02 0.06
(longitudinal direction)
Rocker
Coefficient of friction
- N 0.10 0.02 0.06 0.14
(transverse direction)
Stiffness per feet of
kip/in/ft LN 0.40 0.35 0.70 3.0
Era11 Elastomeric deck width
Coefficient of friction - N 0.30 0.10 0.10 0.50
Coefficient of friction - U 0.50 0.03 0.20 0.80
Friction
Yield displacement In. U 0.07 0.0003 0.04 0.10
Stiffness per feet of
kip/in/ft LN 0.77 0.52 0.7 6.0
Era22 Elastomeric deck width
Coefficient of friction - N 0.30 0.10 0.10 0.50
Stiffness per feet of
kip/in/ft LN 0.00 0.45 0.4 2.5
Era33 Elastomeric deck width
Coefficient of friction - N 0.30 0.10 0.10 0.50
§
N = normal, LN = lognormal, U = uniform, and B = Bernoulli distribution.
† and  are the parameters of the distribution. These denote mean and standard deviation for a normal and uniform distribution, and mean
and standard deviation of the associated normal distribution (in log space) in the case of a lognormal distribution.

5.4 Bearings

Era 11 bridges Era 22/ Era 33 bridges

10%
Rocker
Elastomeric Rocker
Friction
20% 100% Elastomeric

70%

Figure 5.4 – Percentage distribution of bearings based on design eras.

Bridges with seat abutments are rest on bearings at the abutments. Era 11 consists

of various bearing such as rocker bearings, elastomeric bearings, and friction bearings

95
(Figure 3.14), while Era 22 and Era 33 consists only of elastomeric bearings. The motion

associated with elastomeric and friction bearings are based on sliding. On the other hand,

motion is characterized by rocking in case of rocker bearings. The distribution of the

bearings of various eras is shown by percentage in Figure 5.4. The statistical distribution

of the uncertain parameters for the bearings is given in Table 5.3.

5.5 Box girder deck

The important uncertain parameters in the modeling of superstructure are the span

length, deck-width, and type of girder (reinforced or pre-stressed). The distribution of these

parameters is derived based on an extensive plan review of bridges pertinent to the design

eras. As explained in Chapter 4, bridges are grouped based on the span length as S11 (single

span), S22 (two-span), S34 (three and four spans), and S5x (spans greater than 4). Also,

Era 22 and Era 33 bridges are grouped together from a demand perspective.

5.5.1 Span length

Span length is a critical parameter that governs the seismic responses of bridges;

the parameters of the span length distribution for box girder bridges are given in Table 5.4.

The span length is defined as a function of the design era, type of superstructure, and the

number of spans. Interested readers are directed to the memo on the span length models for

box girder bridges (Roblee, 2016a) and a brief summary is given in this section. Some

notable trends in the distribution of span length are given below:

• In general, the span value associated with reinforced concrete (RC) girder is less

than the pre-stressed concrete (PC) girder.

• The mean span value for multi-span PC bridge models is 135-feet for 2-span

bridges and 155-feet for 3- to 6-span bridges, independent of the design era.

96
 However, the mean span length of RC bridge models is not independent of design

era.

• Two-span RC bridges are somewhat longer than other span ranges for each era.

• In the case of single-span bridges, both RC and PC have somewhat shorter span

length than their multi-span counterparts.

The ratio of the approach span to the main span for multi-span bridges is defined

as the span ratio and its properties are also given in Table 5.4. The span ratio model for PC

bridges has a higher mean (0.75) than that for RC bridges (0.60), but the overall range is

comparable (0.35-1.00 for RC, 0.40-1.00 for PC).

5.5.2 Deck width

Deck width parameters are determined based on the extensive plan review of

bridges (Roblee 2016b) and are given in Table 5.5. Deck width distribution is a function

of the design era and number of columns per bent. Era 22 and Era 33 are combined for

deck width distribution as the plan review suggested similar trends in Era 22 and Era 33.

97
 Table 5.4 – Distribution of span length and span ratio (approach span/main span) for box girder bridges.
Span length distribution Span ratio distribution
Type of
Parameters† Type Parameters†
superstructure: §
No. Mi Stand
Design Span Reinforced Unit Lower Upper Standa Lower Upper
of x Typ ard
era range (RC) or s Mean bound Bound Mean rd bound Bound
spans % e§ Devia
Pre-stressed () (L) (U) () Deviati (L) (U)
tion
(PC) on ()
()
50 RC feet N 80 25 35 130 - - - - -
S11 1
50 PC feet N 110 35 40 180 - - - - -
75 RC feet N 95 20 55 140 - - - - -
S22 2
25 PC feet N 135 35 75 230 - - - - -
3 55 RC feet N 90 25 50 160 N 0.60 0.20 0.35 1.00
Era11 S34
4 45 RC feet N 90 25 50 160 N 0.60 0.20 0.35 1.00
5 80 RC feet N 90 20 60 125 N 0.60 0.20 0.35 1.00
S5x
6 20 RC feet N 90 20 60 125 N 0.60 0.20 0.35 1.00
10 RC feet N 80 25 35 130 - - - - -
Era22
35 PC feet N 130 35 50 220 - - - - -
S11 1
15 RC feet N 105 40 35 200 - - - - -
Era 33
40 PC feet N 130 35 50 220 - - - - -
5 RC feet N 95 20 55 140 - - - - -
Era22 S22
35 PC feet N 135 35 75 230 - - - - -
2
10 RC feet N 135 35 85 200 - - - - -
Era 33 S22
50 PC feet N 135 35 75 230 - - - - -
10 RC feet N 90 25 50 160 N 0.60 0.20 0.35 1.00
3
20 PC feet N 155 45 75 250 N 0.75 0.20 0.40 1.00
Era22 S34
5 RC feet N 90 25 50 160 N 0.60 0.20 0.35 1.00
4
10 PC feet N 155 45 75 250 N 0.75 0.20 0.40 1.00
5 RC feet N 110 35 55 190 N 0.60 0.20 0.35 1.00
3
30 PC feet N 155 45 75 250 N 0.75 0.20 0.40 1.00
Era 33 S34
5 RC feet N 110 35 55 190 N 0.60 0.20 0.35 1.00
4
15 PC feet N 155 45 75 250 N 0.75 0.20 0.40 1.00
10 RC feet N 90 20 60 125 N 0.60 0.20 0.35 1.00
5
10 PC feet N 155 35 95 240 N 0.75 0.20 0.40 1.00
Era22 S5x
5 RC feet N 90 20 60 125 N 0.60 0.20 0.35 1.00
6
5 PC feet N 155 35 95 240 N 0.75 0.20 0.40 1.00
15 RC feet N 125 35 75 165 N 0.60 0.20 0.35 1.00
5
30 PC feet N 155 35 95 240 N 0.75 0.20 0.40 1.00
Era 33 S5x
5 RC feet N 125 35 75 165 N 0.60 0.20 0.35 1.00
6
20 PC feet N 155 35 95 240 N 0.75 0.20 0.40 1.00
§
N = normal, LN = lognormal, U = uniform, and B = Bernoulli distribution.
† and  are the parameters of the distribution. These denote mean and standard deviation for a normal and uniform distribution, and mean and standard deviation of the associated normal
distribution (in log space) in the case of a lognormal distribution.

98
 Table 5.5 – Distribution of deck width for box girder bridges.
Column Distribution Cell distribution
type (no. Parameters
Design of Lower Upper
Mix % Standard % % %
era columns) Unit Type§ Mean bound Bound Type Type Type
Deviation distr. distr. distr.
() (L) (U)
()
25 feet N 26.5 1.5 22 30 3 cell 100 - - - -
1 50 feet N 34 1.2 30 38 3 cell 70 5 cell 30 - -
25 feet N 40 1.5 38 46 3 cell 40 5 cell 60 - -
15 feet N 34 2.0 30 38 3 cell 50 5 cell 50 - -
2 25 feet N 41 5 38 48 3 cell 25 5 cell 75
Era 11 15 feet N 58 26 48 74 5 cell 25 7 cell 50 9 cell 25
10 feet N 48 18 38 56 5 cell 65 7 cell 35 - -
3 15 feet N 66 9 56 74 7 cell 50 9 cell 50 - -
5 feet N 80 9 74 92 9 cell 70 11 cell 30 - -
5 feet N 60 34 38 72 5 cell 25 7 cell 35 9 cell 40
4
10 feet N 88 34 72 106 9 cell 25 11 cell 75 - -
15 feet N 28 1.2 22 30 3 cell 100 - - - -
20 feet N 34 4 30 38 3 cell 85 5 cell 15
1
55 feet N 42 2 38 46 3 cell 75 5 cell 25 - -
10 feet N 50 14 46 60 3 cell 30 5 cell 50 7 cell 20
20 feet N 43 7 36 50 3 cell 40 5 cell 60
Era
2 15 feet N 57 8 50 66 5 cell 80 7 cell 20
22/Era
10 feet N 73 22 66 88 5 cell 25 7 cell 50 9 cell 25
33
10 feet N 59 18 50 68 5 cell 50 7 cell 50
3 15 feet N 79 20 68 88 7 cell 50 9 cell 50
10 feet N 98 20 88 108 7 cell 20 9 cell 40 11 cell 40
5 feet N 75 32 58 90 5 cell 25 7 cell 40 9 cell 35
4
15 feet feet 107 38 90 128 9 cell 40 11 cell 35 13 cell 25
§
N = normal, LN = lognormal, U = uniform, and B = Bernoulli distribution.
† and  are the parameters of the distribution. These denote mean and standard deviation for a normal and uniform distribution, and mean and standard deviation of the associated normal
distribution (in log space) in the case of a lognormal distribution

99
5.5.3 Deck cross-section properties

Figure 5.5 illustrates the typical cross-section of box girder bridges. The height of

the box girder is a function of the span length; the acceptable depth-to span ratios are 0.055

and 0.04 for RC and PC concrete boxes, respectively. The cross-section details noted from

the plan review are presented in Table 5.4. It has been noted that the wall thickness (twall)

is constant across the design eras and type of superstructure (RC or PC). A constant value

is adopted for the bottom flange thickness (tbot) for each design era and is given in Table

5.6. The top flange thickness is a function of the center-to-center spacing of the girders

(MTD, 2008) and is given in Table 5.7.

Figure 5.5 – Cross-section details of box-girder bridges.

Table 5.6 – Deck cross-section properties.

Deck cross-section properties Era 11 Era 22 Era 33

Bottom flange thickness (tbot, in.) 6.0 6.5 7.0

Wall thickness 12.0 12.0 12.0

100
Table 5.7 – Box-girder deck slab thickness (MTD, 2008).

101
5.6 Columns

As mentioned before, the column details vary depending on the design era due to

the changes in design philosophy. The design philosophies adopted for California bridges

have been significantly influenced by the historic 1971 San Fernando and the 1989 Loma

Prieta earthquakes. In seismic design philosophy during Era 11, seismic forces were

proportional to the dead weight of structures. Bridges were designed to resist a lateral

seismic force equal to 6% of the dead weight of the structures. After the 1971 San Fernando

earthquake, capacity design principles were introduced in the seismic design standards.

The lateral load-carrying capacity of the seismically designed bridges increased by a factor

of 2 or 2.5. In addition, several aspects, including fault proximity, site conditions, dynamic

structural response, and ductile details were considered in the design of bridge columns.

The extensive damage from the 1989 Loma Prieta earthquake forced Caltrans to solicit the

Applied Technology Council (ATC) to conduct a detailed study and provide

recommendations for design standards, performance criteria, and practices. The

recommendations described in ATC-32 were incorporated in the Caltrans design manuals,

leading to the Era 33 columns. The design attributes and statistical properties of bridge

columns are identified by an in-depth review of bridge plans pertinent to the design eras

for the chosen box girder bridge classes.

5.6.1 Column height

In this study, column heights are measured as the height between the underside of

the bridge deck and the top of column footing by manual plan review. The basic statistics

of the column height are provided in Table 5.8. As noted in Table 5.8, the median column

height increases from Era 11 to Era 33.

102
 Table 5.8 – Column height distribution.
Design Units Distribution
era Type Parameters† Median Lower Upper
§
Mean () Standard Deviation () bound (L) Bound (U)
Era11 ft LN 3.06 0.13 21.5 16.5 28.0
Era 22 ft LN 3.14 0.16 23.2 17.0 32.0
Era 33 ft LN 3.22 0.18 25.0 17.5 36.0
§
N = normal, LN = lognormal, U = uniform, and B = Bernoulli distribution.
† and  are the parameters of the distribution. These denote mean and standard deviation for a normal and uniform distribution, and
mean and standard deviation of the associated normal distribution (in log space) in the case of a lognormal distribution

5.6.2 Column cross-section

Various cross sections such as circular, rectangular, and oblong are noted from the

plan review. Era 11 consists of circular and rectangular cross-sections while Era 22 and

Era 33 contain circular and oblong cross-sections. As seen in Figure 5.6, where the cross-

section details of various design eras are presented, each era consists of a wide range of

cross-sections. The percentage distribution of the Era 11 columns is given in Table 5.9.

103
 Table 5.9 – Distribution of column cross-sections.
Column
Design Type of Diameter Length Breadth
cross- Mix %
era bent (D, in.)* (Lc, in.)* (Bc, in.)*
section
48 - - 10
Circular 60 - - 20
72 - - 20
Single - 72 36 10
- 96 36 15
Rectangular
- 120 36 10
- 96 48 15
Era 11
36 - - 15
Circular 48 - - 20
60 - - 5
Multi - 42 30 20
- 48 36 25
Rectangular
- 60 42 10
- 96 36 5
60 - - 10
66 - - 25
Circular
72 - - 15
Single 84 - - 10
- 72 48 10
Oblong - 96 48 15
Era 22 - 99 66 15
48 - - 35
60 - - 10
Circular
66 - - 20
Multi
72 - - 10
- 60 36 10
Oblong - 72 48 15
60 - - 5
66 - - 15
Circular
84 - - 20
Single 108 - - 10
- 72 48 5
- 96 48 10
Oblong
- 99 66 25
- 108 72 10
Era 33 48 - - 30
60 - - 10
Circular
66 - - 25
Multi 84 - - 5
- 72 48 15
Oblong - 99 66 10
- 126 84 5
*Refer figure 5.6

104
Figure 5.6 – Column cross-sections for various design eras of box girder bridges.

105
5.6.3 Column material properties

The bridge classes considered in this study use concrete as the construction material

and the statistical properties of the concrete are given in Table 5.10. Following the

recommendations of Choi (2002), the compressive strength of concrete is modeled using a

normal distribution. The statistical properties of the yield strength of the reinforcing steel

are also presented in Table 5.10.

Table 5.10 – Statistical distribution of column material properties.

Parameter Design Units Distribution
era Type§ Parameters† Lower Upper
Mean Standard bound Bound
() Deviation (L) (U)
()
Concrete compressive 3.90 0.48 2.94 5.19
Era11 ksi N
strength (4.23) (0.52) (4.86) (5.36)
Era 22/
ksi N 4.55 0.56 3.43 5.67
Era33
Steel yield strength 57.3 4.5 49.0 67.0
Era11 ksi N
(69.0) (5.5) (58.0) (80.0)
Era 22/
ksi N 69.0 5.5 58.0 80.0
Era33
§
N = normal, LN = lognormal, U = uniform, and B = Bernoulli distribution.

80% simulations with 3.9 ksi and remaining 20% with 4.23 ksi, same the case for reinforcing steel yield strength.
† and  are the parameters of the distribution. These denote mean and standard deviation for a normal and uniform
distribution, and mean and standard deviation of the associated normal distribution (in log space) in the case of a
lognormal distribution.

5.6.4 Column reinforcement details

The statistical properties of the column reinforcement identified from the plan

review of bridges are given in Table 5.11. In Era 11, the column shear reinforcement

consisted of #4 transverse stirrups spaced at 12 in. on center regardless of the column size

or the size of the longitudinal reinforcing bars. Uniform distribution is assumed for the

longitudinal and transverse reinforcement ratio.

106
 Table 5.11 – Statistical distribution of column reinforcement details.
Distribution
Parameters†
Design Lower Upper
Parameter Units Standard
era Type§ Mean bound Bound
Deviation
() (L) (U)
()
Longitudinal steel
All - U 2.00 0.33 1.0 3.0
reinforcement ratio
Era11 - #4 @ 12 in. irrespective of the cross-section
Transverse steel
Era 22/
reinforcement ratio - U 0.85 0.07 0.4 1.3
Era33
§
N = normal, LN = lognormal, U = uniform, and B = Bernoulli distribution.
† and  are the parameters of the distribution. These denote mean and standard deviation for a normal and uniform
distribution, and mean and standard deviation of the associated normal distribution (in log space) in the case of a
lognormal distribution.

5.7 Foundations

Figure 5.7 – Bridge foundation types (Priestley et al. 1996).

The foundation provides a means to transmit service and ultimate loads from the

structure to the underlying soil. Foundations can be classified into two types: shallow

107
foundations and deep foundations. As the name implies, the loads from the structure are

transferred to the underlying soil at a shallow depth for shallow foundations. Deep

foundations are provided when soil conditions are not favorable to shallow foundations

and transfer the load through piles. The type of foundation for a particular bridge is

determined by various factors such as soil conditions, overhead clearance, existing utilities,

and proximity to existing facilities such as buildings and railroads. The possible types of

footings are shown in Figure 5.7 and the statistical distribution of foundations across the

design eras is shown in Figure 5.8. The rotational and translational stiffness of the

foundations are presented in Tables 5.12 and 5.13, respectively.

Era 11 26% Era 22

Spread Spread
7% 11%
footing footing
41%
Footing Footing
67% on piles 48% on piles
Shaft Shaft
piles piles

Era 33
17% Spread
footing
25%
Footing
58% on piles
Shaft
piles

Figure 5.8 – Statistical distribution of foundation for various eras.

108
 Table 5.12 – Distribution of foundation rotational stiffness (106 kip-in/rad).
Type of Type of Foundation Transverse to bridge direction Trans/Long stiffness ratio
bent footing fixity Parameters Parameters
Design Lower Upper Lower Upper
Mix % Standard Standard
era Type§ Mean bound Bound Type§ Mean bound Bound
Deviation Deviation
() (L) (U) () (L) (U)
() ()
pile fixed LN 25.0 2.5 10 62.5 LN 1.5 1.5 1.0 2.25
single 100
spread fixed LN 25.0 2.5 10 62.5 LN 1.5 1.5 1.0 2.25
pinned 37.5 LN 2.5 2.5 1.0 6.3 LN 1.0 1.5 0.67 1.50
Era 11 pile
fixed 12.5 LN 4.0 2.5 1.6 10.0 LN 1.0 1.5 0.67 1.50
multiple
pinned 37.5 LN 2.5 2.5 1.0 6.3 LN 1.0 1.5 0.67 1.50
spread
fixed 12.5 LN 4.0 2.5 1.6 10.0 LN 1.0 1.5 0.67 1.50
pile fixed 50 LN 80.0 2.5 32.0 200.0 LN 1.5 1.5 1.0 2.25
single
spread fixed 50 LN 50.0 2.5 20.0 125.0 LN 1.3 1.3 1.0 1.70
pinned 25 LN 12.0 2.5 4.8 30.0 LN 1.0 1.5 0.67 1.5
Era 22 pile
fixed 25 LN 18.0 2.5 7.2 15.0 LN 1.0 1.5 0.67 1.5
multiple
pinned 25 LN 12.0 2.5 4.8 30.0 LN 1.0 1.5 0.67 1.5
spread
fixed 25 LN 18.0 2.5 7.2 15.0 LN 1.0 1.5 0.67 1.5
pile fixed 90 LN 190.0 2.5 76.0 475.0 LN 1.15 1.15 1.00 1.32
single
spread fixed 10 LN 50.0 2.5 20.0 125.0 LN 1.15 1.15 1.00 1.32
pinned 50 LN 20.0 2.5 8.0 50.0 LN 1.20 1.25 0.96 1.50
Era 33 pile
fixed 0 LN 30.0 2.5 12.0 75.0 LN 1.20 1.25 0.96 1.50
multiple
pinned 50 LN 20.0 2.5 8.0 50.0 LN 1.20 1.25 0.96 1.50
spread
fixed 0 LN 30.0 2.5 12.0 75.0 LN 1.20 1.25 0.96 1.50

109
 Table 5.13 – Distribution of foundation translational stiffness (kip/in).
Type of Type of Foundation Transverse to bridge direction Trans/Long stiffness ratio
bent footing fixity Parameters Parameters
Design Lower Upper Lower Upper
Mix % Standard Standard
era Type§ Mean bound Bound Type§ Mean bound Bound
Deviation Deviation
() (L) (U) () (L) (U)
() ()
pile fixed LN 1250.0 2.5 500.0 3125.0 LN 1.0 1.0 1.0 1.0
single 100
spread fixed LN 1250.0 2.5 500.0 3125.0 LN 1.0 1.0 1.0 1.0
pinned 37.5 LN 625.0 2.5 250.0 1562.5 LN 1.0 1.0 1.0 1.0
Era 11 pile
fixed 12.5 LN 625.0 2.5 250.0 1562.5 LN 1.0 1.0 1.0 1.0
multiple
pinned 37.5 LN 625.0 2.5 250.0 1562.5 LN 1.0 1.0 1.0 1.0
spread
fixed 12.5 LN 625.0 2.5 250.0 1562.5 LN 1.0 1.0 1.0 1.0
pile fixed 50 LN 2000.0 2.5 800.0 5000.0 LN 1.0 1.0 1.0 1.0
single
spread fixed 40 LN 2000.0 2.5 800.0 5000.0 LN 1.0 1.0 1.0 1.0
pinned 25 LN 1000.0 2.5 400.0 2500.0 LN 1.0 1.0 1.0 1.0
Era 22 pile
fixed 12.5 LN 1000.0 2.5 400.0 2500.0 LN 1.0 1.0 1.0 1.0
multiple
pinned 25 LN 1000.0 2.5 400.0 2500.0 LN 1.0 1.0 1.0 1.0
spread
fixed 12.5 LN 1000.0 2.5 400.0 2500.0 LN 1.0 1.0 1.0 1.0
pile fixed 65 LN 2500.0 2.5 1000.0 6250.0 LN 1.0 1.0 1.0 1.0
single
spread fixed 10 LN 2500.0 2.5 1000.0 6250.0 LN 1.0 1.0 1.0 1.0
pinned 35 LN 1000.0 2.5 400.0 2500.0 LN 1.0 1.0 1.0 1.0
Era 33 pile
fixed 0 LN 1000.0 2.5 400.0 2500.0 LN 1.0 1.0 1.0 1.0
multiple
pinned 35 LN 1000.0 2.5 400.0 2500.0 LN 1.0 1.0 1.0 1.0
spread
fixed 0 LN 1000.0 2.5 400.0 2500.0 LN 1.0 1.0 1.0 1.0

110
5.8 Other uncertain parameters

The distribution of other uncertain parameters is given in this section.

5.8.1 Damping

Bavirisetty et al. (2003) estimated the 2nd and 98th percentile of damping ratios in bridges

to be 0.02 and 0.07 respectively. The recommendations of Feng et al. (1999) for tall buildings are

extended to bridges by the researchers (Nielson 2005; Padgett 2007) and the damping uncertainty

is modeled using a normal distribution. Based on these studies, damping is modeled as normal

distribution (mean = 4.5%, standard deviation = 1.25%) and is shown in Table 5.14.

5.8.2 Mass factor

Mass sources such as parapets and barrier rails, variable deck slab thickness, electric poles

and other equipment, re-pavement procedures, and variation in material densities are not

considered in the OpenSees bridge model; because of this, it has been decided to account for these

mass sources explicitly with the addition of a mass factor. The mass factor is assumed to be

uniformly distributed with bounds of 0.95 and 1.15.

5.8.3 Shear key acceleration

Per Caltrans, designs of shear keys are categorized as either isolated, as new emerging

designs, or non-isolated, as older conventional designs. Since the isolated design is a new type of

design that does not appear in the existing inventory, the current study will focus only on the non-

isolated type. The non-isolated shear keys were designed to withstand dead-load reaction for 0.3

to 0.5g. However, they have been shown in reality to be on the order of 3 times stronger (Caltrans,

2017). Based on the recommendation from Caltrans, shear key is modeled using lognormal

distribution, with a mean value of 1.0 and standard deviation of 0.20.

111
5.8.4 Gap

Lognormal distribution is adopted for the gap between the deck and the superstructure with a
median value of -0.20 and standard deviation of 0.50. Further, the gap between the superstructure
and shear keys is assumed to be uniformly distributed between 0 and 1.5 in.

5.8.5 Earthquake direction

The ground motions considered in the present study have fault-normal and fault-parallel

components. Mackie et al. (2011) concluded that there is a negligible effect of the angle of

incidence on the mean ensemble response of bridge components and hence the incident angle is

not considered as a major source of uncertainty in the study. As such, the two horizontal

components of ground motions are assigned simultaneously to the longitudinal and transverse

direction of the bridge and the orientation is assigned randomly. The effects of vertical acceleration

and spatially-variable ground motions are not considered in this study.

Table 5.14 – Distribution of other uncertain parameters.

Design Parameter Units Distribution
era Typ Parameters† Lower Upper
e§ Mean Standard bound Bound
() Deviatio (L) (U)
n ()
Damping - N 0.045 0.0125 0.02 0.07
Mass factor U 1.05 0.0033 0.95 1.15
Shear key acceleration g LN 0.0 0.20 0.8 1.20
All
eras Gap
deck and superstructure in. LN -0.20 0.50 0.30 2.20
superstructure and shear key in U 0.75 0.19 0.00 1.50
§
N = normal, LN = lognormal, U = uniform, and B = Bernoulli distribution.
† and  are the parameters of the distribution. These denote mean and standard deviation for a normal and uniform distribution, and mean and
standard deviation of the associated normal distribution (in log space) in the case of a lognormal distribution.

5.9 Closure

This chapter presents the extensive plan review and analysis of the California box girder bridge

inventory using the BIRIS assembled by Caltrans engineers. The bridges are divided into three

112
eras: Era 11 (pre 1971), Era 22 (1971 – 1990), and Era 33 (post 1990), based on the seismic design

principles and column detailing. Details pertinent to abutments, bearing, foundations,

superstructure, and columns are gathered across the three design eras to aid in the development of

stochastic finite element models for the generation of probabilistic seismic demand models

(PSDMs) and fragility curves. Also, the bridge design details and physical characteristics

identified in this chapter help to capture the vulnerabilities associated with various components.

The input parameters are assumed to be uncorrelated in the current study and further studies are

needed to address the correlation effects.

113
 CHAPTER 6 SYSTEM AND COMPONENT FRAGILITY CURVES

FOR BOX-GIRDER BRIDGES

The economic and social impacts of earthquakes on civil infrastructure have

increased awareness of potential seismic hazards and the associated vulnerability of structures.

The evaluation of economic losses after earthquakes is primarily based on the prediction of

structural damage. These damage assessments can be represented probabilistically by fragility

curves, a statistical function that gives the conditional probability of exceeding a certain damage

state given a certain ground motion intensity measure (IM). Component and system fragility curves

can be useful in prioritizing both post-earthquake emergency responses and field inspections.

Generating seismic fragility curves involves the convolution of demand models and

capacity models. This chapter explains the fragility framework, including the formulation of

probabilistic seismic demand models (PSDMs) and capacity models. The methodology presented

in this section is used in this study to develop system and component fragility curves for single

frame multi-span box-girder bridges in California. Comparisons are conducted among various

bridge classes to assess the relative vulnerability of each class, and are presented in detail in this

chapter.

6.1 Fragility Framework

The multiphase framework used by numerous researchers (Nielson 2005, Padgett 2007,

Ramanathan et al. 2015, Jeon et al. 2015) is adopted in the current study to shed light on the

fragilities of various bridge classes and the effects of various bridge components on bridge

fragilities. This methodology also helps to generate system as well as component fragilities. The

parameters listed in Chapter 5 are varied to capture uncertainties in bridge classes. Input variables

are sampled across the range of parameters presented in the Chapter 5 using Latin Hypercube

114
Sampling technique in order to generate statistically significant yet nominally identical three–

dimensional bridge models. The variables are randomly paired with the selected suite of ground

motions. The two orthogonal components of the ground motions are randomly assigned to the

longitudinal and transverse direction of the bridge axis. A set of 320 simulations of nonlinear time

history analyses (NLTHAs) is performed for all bridge-ground motion pairs to monitor the

maximum response of various bridge components. Figure 6.1 shows the schematic of the

procedure that was adopted to capture the demand of various bridge components due to the ground

motions.

Figure 6.1 – Schematic representation of the NLTHA procedure used to derive the
PSDMs.

115
Table 6.1 – Engineering demand parameters for bridge components monitored in NLTHA.
Component Engineering Notation Units
demand parameter

Columns Curvature ductility c -

Abutment seat Displacement seat Inches
Unseating Displacement u Inches
Elastomeric bearing Displacement b Inches
Deck Displacement d Inches
pads
Foundation translation Displacement fnd Inches
Foundation rotation Rotation f
Passive abutment Displacement p Inches
Active abutment Displacement a Inches
response
Transverse abutment Displacement t Inches
response
Shear key Displacement key Inches
response

The current study considers the vulnerability of multiple components: columns, abutment

seat (seat type abutments), elastomeric bearings, joint seal, restrainer cables (retrofitted bridges),

deck displacement, foundations, abutments, and shear keys. The engineering demand parameters

(EDP) representing the above components are indicated in Table 6.1. The following section

explains the estimation of the seismic demand and capacity models.

6.1.1 Probabilistic seismic demand models

Fragility curves require the convolution of the demand model and capacity models. As

mentioned before, the seismic demands on bridge components are obtained by the three-

dimensional NLTHAs of bridge models. The peak response of the components (di, e.g., column

curvature ductility, bearing deformations, and abutment deformations) is recorded for each

NLTHA. Based on Cornell et al. (2002), PSDMs are defined as the linear regression of pairs of D

and IM in the log-transformed space, as illustrated in Figure 6.2. These can be written as

ln( Sd )  ln(a)  b ln( IM ) (6.1)

116
 where a and b are the regression coefficients. The coefficients a and b are obtained by

performing a linear regression analysis on D-IM pairs in the log-transformed space. Dispersion,

d/IM, is evaluated based on statistical analysis:

  ln d  ln ( S d ) 
2
i
 d / IM  i 1
(6.2)
N 2

Figure 6.2 – Illustration of a typical PDSM.

6.1.2 Capacity models

The development of a probabilistic seismic demand model using the results of the NLTHA

forms the demand side of the fragility formulation. The next crucial step in the fragility formulation

is the formulation of capacity, or limit state, models. The capacity models are described by a two-

parameter lognormal distribution with median, Sc and dispersion, βc. Discrete damage states are

defined for each component corresponding to the significant change in its response and consequent

to its own performance and the performance of the bridge at both the global and system levels. A

general description of the component damage thresholds (CDT) and bridge system-level damage

states (BSST) is given in Table 6.2 and 6.3, respectively (Ramanathan et al. 2015).

117
 The bridge components are categorized as either primary or secondary. Primary

components include the columns and abutment seat; these are the components that affect the

vertical stability and load carrying capacity of the bridge. Secondary components are those whose

failure will not force the closure of the bridge; this includes abutment deformations, shear key

displacement, and others. As the failure of a primary component affects the load carrying capacity

and stability of the bridge system, CDTs of primary components map directly into BSSTs. Only

two broad CDTs, CDT-0 and CDT-1, are defined for the secondary components and these map

directly into BSST-0 and BSST-1. CDTs and BSSTs were developed in close collaboration with

Caltrans. The number of components used to integrate the system fragility varies based on the

BSST under consideration. Such mapping ensures similar consequences in terms of repair and

traffic implications at the system level.

Table 6.2– Component level damage state descriptions – Component Damage Thresholds
(CDT).

CDT-0 CDT-1 CDT-2 CDT-3

Component Aesthetic Repairable minor Repairable Component

No damage functional major
damage states damage replacement
damage functional
damage

118
 Table 6.3 – General description of BSSTs along with CDTs.

BSST-0 BSST-1 BSST-2 BSST-4

Bridge system damage
states
MINOR MODERATE EXTENSIVE COMPLETE

ShakeCast Inspection
Low Medium Medium-High High
Priority levels

Open to Limited Emergency Closed (until

Open to normal
Likely Immediate Post- public traffic – vehicles only – shored/braced) –
public traffic – No
Event Traffic State speed/weight/lane speed/weight/lane potential for
Restrictions
restrictions restrictions collapse
Traffic Operation
Implications
Is closure/detour needed? Very unlikely Unlikely Likely Very likely
Are traffic restrictions Unlikely Likely Very Likely Very Likely -
needed? Detour

Emergency Repair
Implications
Is shoring/bracing Very unlikely Unlikely Likely Very likely
needed?
Is roadway leveling Unlikely Likely Very Likely Very Likely -
needed? Detour
Component Damage
Range
Primary components CDT-0 to 1 CDT-1 to 2 CDT-2 to 3 Above CDT-3
Secondary components CDT-0 CDT-1 NA NA

A significant contribution of the present study is the suggestion of capacity limit states for

columns (CCLS) based on extensive experimental review. A review of the existing research

pertinent to various design eras was conducted to collect the experimental data for bridge columns,

and statistical analysis was carried out to suggest the CCLS for bridge columns. Such an exercise

helps to support seismic risk evaluation of bridges in California by developing a new generation

of more accurate and useful bridge fragility models for incorporation into the ShakeCast

earthquake alerting system developed by the California Department of Transportation (Caltrans).

As mentioned before, the intention of the current study is to suggest fragility relationships for

damage states that range from minor spalling of concrete to complete bridge collapse. The

definitions of these limit states and their operational consequences are given in Table 6.4.

119
 Table 6.4 – General definition of column capacity limit states.

Column capacity Component state Component damage Component repair

limit states (CCLS)

CCLS-0 (Slight) None or aesthetic EQ-related minor Seal and paint

cracking

CCLS-1 (Moderate) Minor repairs Minor spalling of Epoxy inject, minor

needed cover concrete removal/patch

CCLS-2 (Extensive) Major repairs Exposed core, Major removal/patch.

needed, but function confinement yield Add jacket.
intact

CCLS-3 (Complete) Irreparable damage, Bar bucking, large Remove/Replace

function drift, core crushing column (or bridge)
compromised

6.1.2.1 Experimental data analysis for various design eras

The design philosophies for California bridges have been significantly influenced by the

historic 1971 San Fernando and the 1989 Loma Prieta earthquakes. Based on unique design

attributes and evolutions in seismic design philosophy, California bridges are categorized into

three design eras: Era 11 (pre-1971), Era 22 (1971-1990), and Era 33 (post-1990). In Era 11

seismic design philosophy, seismic forces were proportional to the dead weight of structures.

Bridges were designed to resist a lateral seismic force equal to 6% of the dead weight of the

structures. Shear failure and pull-out of the longitudinal reinforcement in columns were

predominant due to the lack of ductility, as was revealed by the 1971 San Fernando earthquake.

After the 1971 San Fernando earthquake, capacity design principles were introduced in the

seismic design standards. The lateral load carrying capacity of seismically-designed bridges

increased by a factor of 2 or 2.5. In addition, several aspects, including fault proximity, site

conditions, dynamic structural response, and ductile details, were considered in the design of

120
bridge columns. However, shear failure in the columns was still observed in Era 22 columns due

to the insufficient confinement in the plastic hinge regions.

The extensive damage from the 1989 Loma Prieta earthquake forced Caltrans to solicit the

Applied Technology Council (ATC) to conduct a detailed study and provide recommendations for

design standards, performance criteria, and practices. The recommendations described in ATC-32

were incorporated in Caltrans design manuals, leading to the distinctive Era 33 columns. The

fundamental emphasis in the Era 33 design philosophy was on the displacement-based or capacity

design approach, which ensures a ductile failure mode in the columns. The design details pertinent

to various design eras are given in Table 6.5, 6.6, and 6.7 respectively.

Table 6.8, 6.9, and 6.10, respectively, summarize the extensive literature review on

experimental investigations of bridge columns. The tables also present the geometric features of

the test column, and the values of displacement and curvature ductility limit states for various

levels of damage.

121
 Table 6.5 – Design details for columns in Era 11 (pre-1971).

Design details Typical column reinforcement layout

Lateral seismic force

6% of structural dead weight

Column shear reinforcement

#4 @ 12 in. irrespective of column size

Other notable features

• Lap-splice of column longitudinal bars near
footing.
• Inadequate development length of the column
longitudinal bars into the footing without any standard
hooks.
• Concept of ductility was absent (failure modes are
brittle, shear or lap-splice failure)
• Strength of column degrades rapidly once yield
moment is attained Caltrans (2017)
• Column bars are embedded into footing without
90 degree hook
• Splice length of longitudinal bars was too short

122
 Table 6.6 – Design details of columns in Era 22 (1971-1990).

Design details Typical column reinforcement layout

Lateral seismic force

2 to 2.5 times higher than era 11. Based on ATC-6

guidelines.

Column shear reinforcement

Spacing: 3-6 in. However, confinement in the plastic hinge

region is absent.

Other notable features

• Bar lap-splice is not permitted at the location of
maximum moment.
• Increase in the negative moment reinforcement in
footing and pile caps without any shear
reinforcement.
• Joint reinforcement between the column and the
bent cap, and the column and the footing was
absent.
• Column shear in the plastic hinge region is typical.
• Due to the poor flare details, shear failure was seen Caltrans (2017)
in columns with flares.

123
 Table 6.7 – Design details for columns in Era 33 (post-1990).

Design details Typical column reinforcement layout

Lateral seismic force

Based on ATC-32. Capacity based design to ensure

the ductile mode of failure.

Column shear reinforcement

Spacing: <6 times the longitudinal bar diameter.

Other notable features

• Usage of column flares was very minimal.

• Tight confinement reinforcement was
provided in the column plastic hinge
zones.
• Improvised flare details were provided by
isolating the flare from the superstructure,
i.e., a gap of 2–4 in.
• No lap splices were provided in the plastic
hinge zones.
• Shear reinforcement was provided in the
footing and pile caps.
• Joint reinforcement was provided between
the column and the bent cap, and the Caltrans (2017)
column and the footing.

124
 Table 6.8 – Summary of limit states for Era 11 columns.
Section properties Displacement ductility Curvature ductility Reference

Colu
Lon
mn Colu
Colum g.
Cross- dia Length * mn Scal Design CCLS- CCLS- CCLS- CCLS- CCLS- CCLS- CCLS- CCL
n Steel
section mete breadth (in.) heigh e code 0 -1 -2 -3 -0 -1 -2 S--3
model ratio
r t (ft.)
(%)
(in.)
R-I Rectangle - 28.75 19.25 12.0 0.4 Caltrans 2.55 0.80 1.00 2.00 3.00 0.80 2.00 6.00 10.00
Sun et al.
R-5 Rectangle - 28.75  19.25 12.0 0.4 Caltrans 5.00 0.80 2.00 3.00 4.00 0.80 2.50 3.50 6.00 (1993)
SRPH-
Circular 24 - 7.0 0.4 Caltrans 5.4 0.80 1.00 1.50 2.00 0.80 1.00 1.54 2.56 Hose et al.
6
(1997)

col #2 Circular 24 - 6.0 0.4 ACI 1.06 0.80 1.50 2.00 4.00 0.80 2.26 3.52 8.56 Priestley et al.
(1996)
col-1 Circular 24 - 12 0.4 Caltrans 2.53 0.80 1.00 1.50 4.00 0.80 1.00 2.31 8.87
Chai et al.
col-3 Circular 24 - 12 0.4 Caltrans 2.53 1.00 1.50 3.00 5.00 1.00 2.31 6.25 11.49 (1991)
T1 Circular 10 - 3.33 0.28 Caltrans 2 0.80 2.00 3.00 5.00 0.80 2.93 4.85 8.71
Jaradat et al.
T2 Circular 10 - 3.33 0.28 Caltrans 1.1 0.80 2.00 4.00 5.00 0.80 2.93 6.78 8.71 (1998)
T3 Circular 10 - 3.33 0.28 Caltrans 2 0.80 2.00 3.00 4.00 0.80 2.93 4.85 6.78
S1 Circular 10 - 5.83 0.28 Caltrans 2 0.80 1.00 1.50 2.00 0.80 1.00 2.33 3.65
S2 Circular 10 - 5.83 0.28 Caltrans 1.1 0.80 2.00 3.00 5.00 0.80 3.65 6.31 11.62
S3 Circular 10 - 5.83 0.28 Caltrans 1.1 1.00 2.00 3.00 4.00 1.00 3.65 6.31 8.96
S1 Circular 20 - 5.0 0.33 Washington 0.99 0.8 1.82 5.04 8.27 0.8 2.48 8.28 14.10
Ranf et al.
S3 Circular 20 - 5.0 0.33 Washington 0.99 0.8 1.34 2.45 3.56 0.8 1.61 3.61 5.61 (2006)
S15 Circular 20 - 5.0 0.33 Washington 0.99 0.8 2.05 4.03 6 0.8 2.89 6.46 10.01
C2 Circular 20 - 5.0 0.33 Washington 0.99 0.8 1.47 2.65 3.83 0.8 1.85 3.97 6.10
C4 Circular 20 - 5.0 0.33 Washington 0.99 0.8 1.52 4.6 7.68 0.8 1.94 7.49 13.04
C3R Circular 20 - 5.0 0.33 Washington 0.99 0.8 1.32 3.51 5.7 0.8 1.58 5.52 9.47

125
 Table 6.9 – Summary of limit states for Era 22 columns.
Section properties Displacement ductility Curvature ductility Reference

Colu Lengt
Colu Long.
mn h*
Column Cross- mn Design Steel CCLS- CCLS- CCLS- CCLS- CCLS- CCLS- CCLS- CCL
diam bread Scale
model section height code ratio -0 -1 -2 -3 -0 -1 -2 S--3
eter th
(ft.) (%)
(in.) (in.)

SRPH-1 Circular 24 - - 0.4 Caltrans 2.7 0.77 1.00 6.00 8.00 0.77 2.77 7.00 9.26 Hose et al.

SRPH-2 Circular 24 - - 0.4 Caltrans 5.4 0.72 1.50 3.00 4.50 0.72 1.60 4.15 7.80 (1997)

SRPH-3 Circular 24 - - 0.4 Caltrans 5.4 0.69 2.50 5.00 7.30 0.69 3.10 5.80 9.23
SRPH-7 Circular 24 - - 0.4 Caltrans 5.4 0.75 2.00 4.00 6.00 0.75 1.60 4.56 7.32
Sanchez
MG-2 Circular 28.8 - 8.8 0.4 Caltrans 2.0 0.86 2.3 2.88 3.45 0.86 4.48 6.04 7.57 et al.
24 (1997)
RDS-2 Oblong - 13.0 0.4 Caltrans 2.0 2.0 6.0 8.0 10.0 3.75 14.77 20.28 25.79
36
Asad and
PEER- Xiao
Circular 16 - 6.0 0.33 Caltrans 1.17 1.3 5.08 4.16 5.55 1.75 11.27 8.95 12.46
COL-1 (2005)
PEER-
Circular 16 - 6.0 0.33 Caltrans 1.17 1.11 2.80 4.17 5.55 1.277 5.534 8.984 12.46
COL-3
Stone and
Flexure Circular 60 - 30.0 1.0 Caltrans 2.0 1.0 3.00 5.00 6.60 3.87 6.74 12.49 17.08 Cheok
Shear Circular 60 - 15.0 1.0 Caltrans 2.0 1.0 4.00 6.00 10.00 3.09 7.26 11.44 19.79 (1989)

126
 Table 6.10 – Summary of limit states for Era 33 columns.
Section properties Displacement ductility Curvature ductility
Colu Long.
Column Length *
Column Cross- mn Scal Design Steel CCLS- CCLS- CCL CCL CCLS- CCLS- CCL CCL Reference
diamete breadth
model section heigh e code ratio -0 -1 S--2 S--3 -0 -1 S--2 S--3
r (in.) (in.)
t (ft.) (%)
328 Circular 24 - 6.0 0.5 Caltrans 2.72 1.40 3.51 4.91 9.12 2.07 7.64 11.36 22.51
328T Circular 24 - 6.0 0.5 Caltrans 2.72 1.57 4.71 7.84 10.20 2.51 10.81 19.12 25.35 Calderone et
al. (2001)
828 Circular 24 - 16.0 0.5 Caltrans 2.72 1.26 4.41 4.41 7.77 1.91 12.90 12.90 24.63
1028 Circular 24 - 20.0 0.5 Caltrans 2.72 1.81 5.81 6.46 9.04 3.94 18.47 20.82 30.19
ISL 1.0 Oblong - 12  17.5 4.83 0.2 Caltrans 2.00 0.8 1.5 5.6 9.6 0.50 2.24 12.40 22.31

ISL 1.5 Oblong - 12  20.25 6 0.2 Caltrans 2.00 1.5 2.4 7.5 10.4 2.35 4.78 18.53 26.35

ISH 1.0 Oblong - 10  14.5 7.62 0.2 Caltrans 2.90 0.9 1.4 3.6 4.7 1.00 2.19 8.72 11.99 Correal. et al.
(2007)
ISH 1.25 Oblong - 10  16.75 8.15 0.2 Caltrans 2.80 0.7 1.4 3.7 4.7 1.00 2.22 9.22 12.26

ISH 1.5 Oblong - 10  15.62 8.79 0.2 Caltrans 2.90 1 1.6 2.2 4.7 1.00 2.86 4.72 12.46

ISH1.5T Oblong - 10  16.75 8.79 0.2 Caltrans 2.90 1.0 1.7 2.8 3.8 1.00 3.18 6.60 9.71
415 Circular 24 - 8.0 0.33 Caltrans 1.5 1.0 2.7 5.0 8.0 1.00 5.04 10.59 17.78
407 Circular 24 - 8.0 0.33 Caltrans 0.75 1.0 2.5 3.0 6.0 1.00 4.60 5.80 12.99
Lehman and
430 Circular 24 - 8.0 0.33 Caltrans 1.5 1.0 3.0 5.0 7.0 1.00 5.88 10.59 15.41 Moehle
(2000)
815 Circular 24 - 8.0 0.33 Caltrans 1.5 1.0 2.0 3.0 5.0 1.00 3.47 5.94 10.89
1015 Circular 24 - 8.0 0.33 Caltrans 1.5 1.0 2.0 3.0 5.0 1.00 3.47 5.94 10.89
Hose et al.
SRPH-4 Circular 24 - 7.0 0.4 Caltrans 5.4 1.00 2.50 8.00 - 1.00 2.00 14.60 - (1997)

Orozco et al.
VP-2 Circular 16 - 6.0 0.4 Caltrans 1.17 1.0 3.4 7.0 8.3 1.00 5.10 10.80 13.85
(1999)
RDS-1 Oblong - 24 ´ 36 13 0.4 Caltrans 1.64 2.0 4.0 8.0 12.0 2.30 4.50 10.90 17.30 Sanchez et al.
(1997)
RDS-6 Oblong - 24 ´ 36 13 0.4 Caltrans 2.00 2.0 6.0 8.0 12.0 1.50 5.30 11.40 17.40
H/D (6) Circular 24 - 12.0 0.5 Caltrans 2.10 0.75 4.50 6.0 18.0 0.75 8.71 12.01 38.43
Shanmugam
(2009)
H/D (3) Circular 24 - 6.0 0.5 Caltrans 2.10 0.75 4.33 13.75 17.0 0.75 5.03 16.42 20.36

127
 Table 6.11 presents the summary of the ductility values for various design eras.

Column curvature ductility (µ) is chosen as the EDP for columns and the median values

for various design eras are highlighted in Table 6.11.

Table 6.11– Statistical summary of ductility values for various design eras.

Design No. of Properties Displacement ductility Curvature ductility

era experiments
CCLS- CCLS CCLS CCLS CCLS- CCLS- CCLS- CCLS-
0 -1 -2 -3 0 1 2 3
Mean 0.82 1.58 2.93 4.56 0.82 2.25 4.99 8.57
Median 0.80 1.51 3.00 4.00 0.80 2.30 5.20 8.80
Era 11 18 lower
0.80 1.00 1.50 2.00 0.80 1.00 1.54 2.56
bound
upper
1.00 2.05 5.04 8.27 1.00 3.65 8.28 14.10
bound
Mean 1.02 3.02 4.65 6.29 1.32 5.64 8.22 11.49
Median 1.00 2.65 4.59 6.30 1.00 5.00 8.00 11.00
Era 22 9 Lower
0.69 1.00 2.88 3.45 0.69 1.60 4.15 7.32
bound
Upper
2.00 6.00 8.00 10.00 3.75 14.77 20.28 25.79
bound
Mean 1.20 2.98 5.45 7.86 1.45 5.63 11.73 18.05
Era 33 21 Median 1.00 2.68 5.00 8.13 1.00 5.00 11.00 17.50
lower
0.70 1.40 2.20 3.80 0.50 2.00 4.72 9.71
bound
upper
2.00 6.00 10.00 12.00 3.94 18.47 23.67 30.19
bound

6.1.3 Abutments

In addition to columns, seat is also considered a primary component in the current

study to account for the unseating potential. Bridge seat widths chronologically increased

from the 4 – 12 inch range in Era 11, to the 12 – 24 inch range in Era 22, to greater than 24

inches in Era 33. Table 6.12 gives the median CDT values for abutment seats for various

128
design eras. The values were developed based on previous studies (Ramanathan, 2012) and

discussion with Caltrans engineers.

Table 6.12– Median value of CDT for abutment seat.

Design era Units CDT-0 CDT-1 CDT-2 CDT-3
Era 11 Inches 0.5 1.0 2.0 3.0
Era 22 Inches 1.0 4.5 10.0 15.0
Era 33 Inches 1.5 4.5 14.0 21.0

The CDT values for other components such as superstructure deck, abutment

displacement in the passive, active, and transverse directions, foundation displacements,

shear keys, and joint seals are consistent with previous research conducted by Ramanathan

(2012). Interested readers are directed to Ramanathan (2012) for a more detailed

explanation of the CDT values of these components. Table 6.13 provides the summary

values of the CDT and CCLS value for various bridge components. The capacity models

are described by a two-parameter lognormal distribution with median, Sc and dispersion,

βc. βc is assigned as 0.35 in a subjective manner due to lack of sufficient information and

is adopted as the same across the components and the respective damage states.

129
 Table 6.13– Summary of CDT values for various bridge components.
Components EDP Units Median value, Sc c
CDT- CDT- CDT- CDT
0 1 2 -2
Primary Components
Columns
Era 11 Curvature NA 0.8 2.0 5.0 8.0 0.35
ductility
Era 22 Curvature NA 1.0 5.0 8.0 11.0 0.35
ductility
Era 33 Curvature NA 1.0 5.0 11.0 17.0 0.35
ductility
Abutment seat
Era 11 Displacement Inches 0.5 1.0 2.0 3.0 0.35
Era 22 Displacement Inches 1.0 4.5 10.0 15.0 0.35
Era 33 Displacement Inches 1.5 4.5 14.0 21.0 0.35
Secondary Components
Joint Seal Displacement Inches 2.0 5.0 NA NA 0.35
Bearings Displacement Inches 1.0 4.0 NA NA 0.35
Shear keys Displacement Inches 1.0 5.0 NA NA 0.35
Deck Displacement Inches 4.0 12.0 NA NA 0.35

Bent foundation
Translation Displacement Inches 1.0 4.0 NA NA 0.35
Rotation Rotation Radian 1.5 6.0 NA NA 0.35

Abutments
Passive Displacement Inches 3.0 10.0 NA NA 0.35
Active Displacement Inches 1.5 4.0 NA NA 0.35
Transverse Displacement Inches 1.0 4.0 NA NA 0.35

6.2 Fragility methodology

The probability that the seismic demands (D) placed on a component exceed the

capacity (C) conditioned on a chosen intensity measure (IM) can be assessed by a fragility

function. Assuming a lognormal distribution for the D and C, component fragility curves,

defined here as the probability of reaching or exceeding a specified damage state for a

particular component, are estimated as

130
  ln( S / S ) 
P[ D  C / IM]    d c
 (6.3)
  d2/ IM   c2 

where, Sd is the median estimate of the demand as a function of the IM, Sc is the

median estimate of the capacity, d/IM is the dispersion of the demand conditioned on the

IM, c is the dispersion of the capacity, and (•) is the standard normal cumulative

distribution function. As presented in equation 6.3, a component fragility curve is

computed through a convolution of a PSDM and a limit state model. A set of component

fragility curves computed in equation 6.3 has to be integrated into a system fragility (or

bridge fragility), which is facilitated through the development of joint probabilistic seismic

demand models (JPSDMs) (Nielson 2005, Padgett 2007, Ramanathan et al. 2015, Jeon et

al. 2015). The JPSDM recognizes the correlation between various components. If the

vector demands, Xi, placed on the n components of the system are expressed as

X   X 1 , X 2 ,..., X n  , then the vector, Y  ln  X  represents the vector of component demands

in the log-transformed space. The JSPDM is formulated in this space by assembling the

vector of means  Y and the covariance matrix,  Y . A Monte Carlo simulation (106 in

the current study) in which samples are drawn from both the demand and capacity models

is used to estimate the probability of demand exceeding the capacity value for each IM.

This procedure is repeated for the increasing value of the IM, and regression analysis is

used afterwards to estimate the lognormal parameters, median, and dispersion, which

characterize the bridge fragility. The system fragilities are helpful to measure the

correlation between various bridge components and determine which component is

dominating the overall system vulnerability. The system or bridge fragilities are useful for

risk assessment on a regional level, where it would be computationally burdensome to

calculate the probability of failure of each bridge component, and then combine them to

form the probability of failure of the system. Moreover, it is found that using the fragility

131
of any single bridge component to represent the overall vulnerability of the bridge would

likely result in a significant underestimation of that vulnerability (Nielson 2007). Figure

6.3 shows a schematic of the fragility framework adopted in the current study.

Figure 6.3 – Schematic of the fragility framework.

The input for the fragility framework is presented in the following section.

132
6.2.1 Ground motion suite

It is important to have a wide range of ground motions with a large variation of

peak ground accelerations to ensure the evaluation of a sufficient range of bridge responses.

The current study utilizes the ground motions from the NGA-2 database assembled by

Caltrans (2017). These motions were developed specifically for this project and consist of

320 ground motions. The ground motion details are given in Appendix A, and the response

spectra for the ground motions are presented in Figure 6.4. The work of Ramanathan (2012)

indicated that spectral acceleration at 1.0 sec (Sa-1.0s) is the optimal intensity measure for

the class of concrete box-girder bridges. Based on this observation, the current study

selected Sa-1.0s as the IM.

Figure 6.4 – Response spectra for the selected ground motions.

133
6.2.2 Material and geometric uncertainties and parameterized stochastic bridge models

Geometric and material uncertainties considered in the current study are discussed

in detail in Chapter 5. As mentioned before, most of the parameters were chosen based on

the plan review of more than 1,000 bridges, using in-house database obtained from

Caltrans. Having identified all key modeling assumptions and the uncertainty distribution

of the bridge models, the next step was to develop representative bridge models that could

capture the entire range of material and geometric uncertainties. Statistically significant yet

nominally identical three-dimensional bridge models were generated by sampling across

the range of parameters using Latin Hypercube Sampling technique, and were then paired

randomly with the selected suite of ground motions. 320 analytical bridge models were

generated, consistent with the number of ground motions. These were then paired randomly

to create the bridge model-ground motion pair. NLTHA was performed on each case and

the peak component demands are noted for each. The responses from the NLTHA were

used to generate the PSDMs. The PSDMs were then convolved with capacity models to

generate the fragility curves.

6.3 Fragility curves for multi-span continuous concrete single frame box-girder

bridges

The intent of this research is to generate fragility curves for multi-span continuous

concrete box-girder (MSCC-BG) bridge classes in California in order to identify the

relative vulnerabilities of various bridge groupings. Some nomenclatures were adopted for

this study and are detailed in Table 6.14. For example, S-E1-S22-R-D corresponds to

single-column bent (S) in design era Era11 (E1) with two spans (S22), rectangular cross-

sections, (R) and diaphragm abutments (D).

134
 One simple technique to evaluate the differences in fragility curves is to evaluate

the relative change in the median value of the fragility curves. An increase in the median

value means a less vulnerable structure, while a decrease in the median value indicates a

more vulnerable structure, and is illustrated in Figure 6.5. In Figure 6.5, bridge 2 is less

vulnerable than bridge 1 for the limit state BSST-3.

Table 6.14 – Nomenclature adopted in the current study.

Design attributes Nomenclature
Bent type
Single column bent S
Two column bent T
Multi column bent M
Design era
Era 11 (pre 1970) E1
Era 22 (1971-1990) E2
Era 33 (post 1990) E3
Span range
Two span S22
Three-four span S34
Column cross-section
Rectangle R
Circular C
Oblong O
Abutment type
Diaphragm D
Seat S

Fragility curves were generated for the 72 single frame box-girder bridge classes

and are presented in Table 6.15 - 6.18. These classifications cover more than 75% of the

California box-girder bridge inventory. Tables 6.15 – 6.18 also give the dispersion, , a

single value of dispersion characterizing the fragility across the four limit states. Appendix

B documents the fragility values in terms of PGA. The system and component fragility

curves for two bridge classes, S-E1-S22-C-D and S-E2-S22-C-S, are presented in Figure

6.6. It is seen from Figure 6.6 that the column is the most vulnerable component for bridges

135
with diaphragm abutments. In the case of bridges with seat abutments, columns and bearing

each contribute to the system vulnerability. The median and dispersion values for the

component fragility curves for the bridge classes are documented in Appendix C.

Figure 6.5 – Illustration of change in median value and relative vulnerability.

136
Table 6.15 – Fragility values for two span continuous concrete box-girder fragilities
with diaphragm abutments.
Design Bridge class BSST-0 BSST-0 BSST-0 BSST-0 *
era        
S-E1-S22-C-D 0.12 0.60 0.30 0.61 0.70 0.63 1.06 0.62 0.61
S-E1-S22-R-D 0.17 0.63 0.41 0.64 0.82 0.65 1.18 0.66 0.65
T-E1-S22-C-D 0.08 0.63 0.20 0.66 0.47 0.71 0.72 0.70 0.68
Era 11
T-E1-S22-R-D 0.10 0.69 0.21 0.73 0.43 0.75 0.62 0.76 0.73
M-E1-S22-C-D 0.07 0.82 0.17 0.87 0.44 1.11 0.69 1.11 0.98
M-E1-S22-R-D 0.06 0.88 0.15 1.06 0.41 1.36 0.69 1.36 1.17
S-E2-S22-C-D 0.15 0.68 0.63 0.68 1.45 0.69 1.93 0.69 0.68
S-E2-S22-O-D 0.16 0.64 0.71 0.64 2.36 0.73 3.07 0.65 0.67
T-E2-S22-C-D 0.11 0.62 0.42 0.61 0.97 0.66 1.27 0.66 0.63
Era 22
T-E2-S22-O-D 0.15 0.58 0.55 0.57 1.17 0.57 1.47 0.57 0.57
M-E2-S22-C-D 0.10 0.61 0.33 0.60 0.67 0.74 0.86 0.74 0.68
M-E2-S22-O-D 0.10 0.61 0.33 0.60 0.67 0.74 0.86 0.74 0.68
S-E3-S22-C-D 0.15 0.68 0.63 0.68 1.91 0.70 2.86 0.72 0.69
S-E3-S22-O-D 0.16 0.64 0.71 0.64 3.10 0.67 4.39 0.63 0.65
T-E3-S22-C-D 0.11 0.62 0.41 0.61 1.27 0.66 1.84 0.66 0.64
Era 33
T-E3-S22-O-D 0.15 0.58 0.55 0.57 1.49 0.57 2.06 0.58 0.57
M-E3-S22-C-D 0.10 0.63 0.33 0.60 0.86 0.75 1.20 0.74 0.68
M-E3-S22-O-D 0.10 0.63 0.33 0.60 0.86 0.75 1.20 0.74 0.68

Table 6.16 – Fragility values for two span continuous concrete box-girder fragilities
with seat abutments.
Design Bridge class BSST-0 BSST-0 BSST-0 BSST-0 *
era        
S-E1-S22-C-S 0.08 0.61 0.15 0.59 0.29 0.58 0.42 0.58 0.59
S-E1-S22-R-S 0.08 0.52 0.15 0.54 0.27 0.53 0.38 0.52 0.53
T-E1-S22-C-S 0.06 0.57 0.11 0.60 0.21 0.59 0.30 0.59 0.59
Era 11
T-E1-S22-R-S 0.08 0.57 0.14 0.57 0.25 0.57 0.35 0.55 0.57
M-E1-S22-C-S 0.04 0.63 0.08 0.62 0.18 0.60 0.26 0.60 0.61
M-E1-S22-R-S 0.09 0.58 0.15 0.58 0.27 0.58 0.37 0.58 0.58
S-E2-S22-C-S 0.11 0.58 0.53 0.57 0.99 0.65 1.32 0.65 0.61
S-E2-S22-O-S 0.11 0.51 0.53 0.49 1.14 0.55 1.58 0.55 0.52
T-E2-S22-C-S 0.08 0.51 0.38 0.49 0.77 0.55 1.08 0.56 0.53
Era 22
T-E2-S22-O-S 0.08 0.45 0.37 0.44 0.80 0.55 1.10 0.56 0.50
M-E2-S22-C-S 0.07 0.49 0.36 0.48 0.73 0.59 1.02 0.62 0.54
M-E2-S22-O-S 0.08 0.54 0.42 0.51 0.91 0.66 1.30 0.67 0.60
S-E3-S22-C-S 0.11 0.57 0.53 0.56 1.25 0.63 1.88 0.63 0.60
S-E3-S22-O-S 0.12 0.50 0.53 0.49 1.28 0.53 1.91 0.53 0.51
T-E3-S22-C-S 0.08 0.51 0.38 0.50 0.94 0.53 1.41 0.54 0.52
Era 33
T-E3-S22-O-S 0.08 0.45 0.37 0.43 0.91 0.52 1.32 0.50 0.48
M-E3-S22-C-S 0.07 0.49 0.36 0.49 0.89 0.58 1.35 0.58 0.53
M-E3-S22-O-S 0.08 0.54 0.42 0.52 1.07 0.60 1.64 0.61 0.56

137
 Table 6.17 – Fragility values for multi-span continuous concrete box-girder
fragilities with diaphragm abutments.
Design Bridge class BSST-0 BSST-0 BSST-0 BSST-0 *
era        
S-E1-S34-C-D 0.12 0.61 0.33 0.65 0.83 0.70 1.33 0.69 0.66
S-E1-S34-R-D 0.06 0.89 0.25 1.11 1.00 1.32 2.04 1.31 1.16
T-E1-S34-C-D 0.06 0.72 0.15 0.77 0.36 0.89 0.56 0.89 0.82
Era 11
T-E1-S34-R-D 0.07 0.75 0.17 0.79 0.41 0.91 0.66 0.92 0.84
M-E1-S34-C-D 0.01 1.04 0.04 0.98 0.19 1.13 0.40 1.13 1.07
M-E1-S34-R-D 0.01 1.04 0.04 0.98 0.19 1.13 0.40 1.13 1.07
S-E2-S34-C-D 0.15 0.56 0.59 0.54 1.34 0.71 1.79 0.71 0.63
S-E2-S34-O-D 0.19 0.67 0.93 0.66 3.04 0.93 4.26 0.87 0.78
T-E2-S34-C-D 0.09 0.62 0.39 0.59 0.82 0.89 1.12 0.89 0.75
Era 22
T-E2-S34-O-D 0.11 0.71 0.56 0.73 1.79 1.12 2.64 1.16 0.93
M-E2-S34-C-D 0.09 0.54 0.35 0.50 0.68 0.71 0.93 0.71 0.61
M-E2-S34-O-D 0.03 0.82 0.47 0.73 2.00 1.59 4.00 1.66 1.20
S-E3-S34-C-D 0.15 0.55 0.58 0.54 1.78 0.71 2.63 0.69 0.62
S-E3-S34-O-D 0.20 0.67 0.93 0.66 4.47 0.94 6.44 0.82 0.77
T-E3-S34-C-D 0.09 0.62 0.39 0.60 1.12 0.89 1.72 0.89 0.75
Era 33
T-E3-S34-O-D 0.11 0.73 0.56 0.73 2.59 1.12 4.34 1.12 0.93
M-E3-S34-C-D 0.08 0.54 0.35 0.50 0.92 0.70 1.40 0.71 0.61
M-E3-S34-O-D 0.03 0.86 0.47 0.74 3.74 1.47 9.02 1.50 1.14

Table 6.18 – Fragility values for multi-span continuous concrete box-girder

fragilities with seat abutments.
Design Bridge class BSST-0 BSST-0 BSST-0 BSST-0 *
era        
S-E1-S34-C-S 0.10 0.50 0.20 0.49 0.38 0.49 0.54 0.49 0.49
S-E1-S34-R-S 0.05 0.64 0.13 0.61 0.29 0.58 0.44 0.56 0.60
T-E1-S34-C-S 0.05 0.62 0.11 0.62 0.22 0.60 0.31 0.59 0.61
Era 11
T-E1-S34-R-S 0.02 0.77 0.07 0.73 0.20 0.67 0.31 0.65 0.70
M-E1-S34-C-S 0.01 1.15 0.02 0.99 0.11 0.82 0.22 0.81 0.94
M-E1-S34-R-S 0.01 1.15 0.02 0.99 0.11 0.82 0.22 0.81 0.94
S-E2-S34-C-S 0.10 0.66 0.52 0.65 0.96 0.80 1.34 0.81 0.73
S-E2-S34-O-S 0.11 0.62 0.65 0.66 1.52 0.85 2.27 0.90 0.76
T-E2-S34-C-S 0.03 0.75 0.28 0.66 0.63 1.11 1.02 1.20 0.93
Era 22
T-E2-S34-O-S 0.05 0.67 0.39 0.63 1.02 0.88 1.64 0.88 0.77
M-E2-S34-C-S 0.07 0.66 0.32 0.66 0.57 0.88 0.79 0.92 0.78
M-E2-S34-O-S 0.01 1.06 0.28 0.72 0.78 0.94 1.43 0.90 0.90
S-E3-S34-C-S 0.10 0.65 0.52 0.65 1.30 0.77 2.05 0.77 0.71
S-E3-S34-O-S 0.11 0.63 0.65 0.66 1.82 0.76 2.87 0.79 0.71
T-E3-S34-C-S 0.03 0.72 0.28 0.66 0.88 1.02 1.56 0.98 0.84
Era 33
T-E3-S34-O-S 0.05 0.67 0.40 0.63 1.26 0.76 2.06 0.69 0.69
M-E3-S34-C-S 0.07 0.66 0.32 0.66 0.76 0.85 1.18 0.87 0.76
M-E3-S34-O-S 0.01 1.01 0.28 0.72 1.13 0.78 1.93 0.69 0.80

138
S-E1-S22-C-D S-E2-S22-C-S

139
Figure 6.6 – System and component fragility curves for bridge classes S-E1-S22-C-D
and S-E2-S22-C-S.

6.3.1 Trends based on design era

The plot of the median value of bridge fragility curves for two bridge classes, S-E-

S22-C-D and T-E-S34-C-S is presented in Figure 6.7, and the trend is similar across the

different bridge classes. The following are the salient inferences noted from bridge

fragilities across the design eras:

• In general, Era 11 bridges are more vulnerable than Era 22 and Era 33 bridges. This

trend is valid irrespective of the type of abutments, cross-section, and number of

columns per bent.

• Era 22 bridges are more vulnerable than Era 33 bridges. The lower vulnerability of

Era 33 is due to the high ductility that is associated with Era 33 bridge columns.

This highlights the importance of seismic design detailing.

• Across the design eras for a particular abutment type, it is seen that single column

bents are more vulnerable than the multi-column bents. This trend is consistent with

a previous study (Mangalathu et al. 2016a). Amongst the bridges with multi-

column bents, two-column bents are the least vulnerable.

140
• In general, diaphragm abutment bridges are less vulnerable than seat abutment

bridges for all design eras. The lower vulnerability of the diaphragm abutment

bridges might be attributed to the complete engagement of the superstructure and

the abutments in the load transfer mechanism.

a) 3.50
3.00 Era 11 Era 22 Era 33
BSST-i(i=0,1,2,3)

2.50
2.00
1.50
1.00
0.50
0.00
BSST-0 BSST-1 BSST-2 BSST-3

Bridge System Damage States

b) 2.00
1.80 Era 11 Era 22 Era33
BSST-i(i=0,1,2,3)

1.60
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
BSST-0 BSST-1 BSST-2 BSST-3

Bridge System Damage States

Figure 6.7 – Plot of median values of bridge classes across the design eras for a) S-
E(1/2/3)-S22-C-D, and b) T-E(1/2/3)-S34-C-S.

141
6.3.2 Trends based on spans

The variation of the median value of fragility curves for bridge classes with number

of spans is presented in Figure 6.8. The following inferences can be made from the data

presented in Tables 6.15 – 6.18 and Figure 6.8:

• The relative vulnerability between the bridge classes with different spans depends on

the column cross-section, abutment type, design era, and number of columns per bent.

For example, it is seen in Figure 6.8 that the multi-span Era 22 bridge class with circular

cross-section resting on diaphragm abutments (S-E2-S34-C-D) is more vulnerable than

its two-span counterpart S-E2-S22-C-D for all limit states. The percentage change in

median values between two-span and multi-span bridges are 1%, 7%, 8%, and 8% for

BSST-0, -1, -2, and -3 respectively. However, the trend is different for the Era 22 bridge

class with oblong cross-section resting on diaphragm abutments. In this case, multi-

span bridges are less vulnerable than two-span bridges, and the relative vulnerabilities

of two-span bridges (S-E2-S22-O-D) compared to multi span bridges (S-E2-S34-O-D)

are 19%, 32%, 29%, and 39% for BSST-0, -1, -2, and -3 respectively.

• The differences in vulnerabilities for two-span and multi-span bridges underscore the

necessity to account for the number of spans in the generation of fragility curves, which

is not currently captured by the existing HAZUS classifications. It also substantiates

the performance-based grouping of the bridge classes suggested in Chapter 4.

142
 5.00
S-E2-S22-C-D S-E2-S34-C-D
4.00

BBST-i(i=0,1,2,3)
S-E2-S22-O-D S-E2-S34-O-D
3.00

2.00

1.00

0.00
BSST-0 BSST-1 BSST-2 BSST-3
Bridge system Damage States

Figure 6.8 – Plot of median values of bridge classes across the number of spans for
bridge classes S-E2-S(22/34)-C-D and S-E2-S(22/34)-O-D.

6.3.3 Trends based on abutment type

3.50
S-E1-S22-C-D S-E1-S22-C-S
3.00 S-E2-S22-C-D S-E2-S22-C-S
S-E3-S22-C-D S-E3-S22-C-S
BBST-i(i=0,1,2,3)

2.50
2.00
1.50
1.00
0.50
0.00
BSST-0 BSST-1 BSST-2 BSST-3
Bridge system Damage States

Figure 6.9 – Plot of median values of bridge classes across the type of abutments for
the selected bridge classes.

The following inferences can be deduced from the comparison of fragilities for

bridge classes with various abutment types:

143
• For all design eras, diaphragm abutment bridges are less vulnerable than seat abutment

bridges. A general trend is shown in Figure 6.9 for selected bridge classes. The lower

vulnerability of the diaphragm abutment bridges might be attributed to the complete

engagement of the superstructure and the abutments in the load transfer mechanism.

• The columns are the most vulnerable component of bridges with diaphragm abutments.

However, in the case of seat abutment bridges, bearings as well as columns contribute

to the overall vulnerability. This highlights the need for adequate seat width in the case

of bridges with seat abutments.

• The vulnerability of bridges reduces with the evolution of column design philosophy.

The trend is the same irrespective of the type of cross-section and number of columns

per bent.

• HAZUS suggests the same fragility relationships for bridges with seat and diaphragm

abutments. The fragility curves generated in the current study outline the need to

account for the type of abutments. For example, the bridge class with seat abutment (S-

E3-S22-C-S) is 36%, 19%, 53% and 53% more vulnerable than the bridge class with

diaphragm abutment (S-E3-S22-C-D) for the limit states BSST-0, -1, -2, and -3,

respectively. Note that the only difference between the bridge class S-E3-S22-C-S and

S-E3-S22-C-D is the type of abutments.

6.3.4 Trends based on column cross-section

As mentioned in Chapter 5, Era 11 bridges consist of circular and rectangular column

cross-sections, while Eras 22 and 33 utilize circular and oblong cross-sections. The

144
following inferences are noted from the comparison of median fragilities for bridges

classes (Tables 6.15 – 6.18, and Figure 6.10) with various cross-sections:

5.00
T-E1-S34-C-D T-E1-S34-R-D
T-E2-S34-C-D T-E2-S34-O-D
4.00 T-E3-S34-C-D T-E3-S34-O-D
BBST-i(i=0,1,2,3)

3.00

2.00

1.00

0.00
BSST-0 BSST-1 BSST-2 BSST-3
Bridge system Damage States

Figure 6.10 – Plot of median values of bridge classes across various column cross-
sections for the selected bridge classes.

• In the case of bridges from Eras 22 and 33, columns with oblong cross-sections are less

vulnerable than columns with circular cross-sections. This trend is the same

irrespective of the type of abutment and number of columns per bent.

• No general trend is observed in the case of Era 11 bridges. For example, bridges with

rectangular cross-sections performed better in the case of two-span bridges with single

column bent and multi-column bent bridges with diaphragm abutments. However, the

trend is reversed in the case of two-span bridges with two-column bents resting on

diaphragm abutments. Also, there is no trend observed in the case of bridges resting on

seat abutments.

145
6.3.5 Trends based on number of columns per bent

The general trend of the change in the median value of the fragility curves is

presented in. The following inferences can be drawn from the comparison of the median

values of fragility curves:

• Multi-column bents are more vulnerable than bridges with single column bents. The

increased vulnerability of bridges with multi-column bents is mainly due to the bridge

width and modeling assumptions, as a significant portion of multi-column bents are

pinned at the base, while single-column bents have a significant amount of rotational

restraint (see Chapter 5).

• For a given design era and number of spans, multi-column bents with circular cross-

section are more vulnerable. For example, M-E2-S34-C-D is 34%, 194%, and 331%

more vulnerable than M-E2-S34-O-D for the limit states BSST -1, -2, and -3,

respectively. BSST-0 for the two bridge classes is significantly low.

• It is seen that the vulnerability of single, two-column, and multi-column bridges

decreases with the evolution is seismic design philosophy. Also, there in a

tremendous reduction in the vulnerability of bridges from Eras 22 and 33, compared

to their Era 11 counterparts.

• Further, the differences in the median value of bridge classes with different numbers

of columns per bent underscore the necessity to capture the number of columns per

bent in the fragility curves. Such a classification is not currently available in HAZUS.

146
 5.00
a) S-E2-S34-C-D
S-E2-S34-O-D
4.00
T-E2-S34-C-D

BBST-i(i=0,1,2,3)
3.00 T-E2-S34-O-D
M-E2-S34-C-D
2.00 M-E2-S34-O-D

1.00

0.00
BSST-0 BSST-1 BSST-2 BSST-3
Bridge system Damage States

b) 2.50 S-E2-S34-C-S
S-E2-S34-O-S
2.00 T-E2-S34-C-S
BBST-i(i=0,1,2,3)

T-E2-S34-O-S
1.50
M-E2-S34-C-S
1.00 M-E2-S34-O-S

0.50

0.00
BSST-0 BSST-1 BSST-2 BSST-3
Bridge system Damage States

Figure 6.11 – Plot of median values of bridge classes across the number of columns
per bent for the selected bridge classes with a) diaphragm abutments and b) seat
abutments.

6.4 HAZUS comparison

A detailed discussion on the classification, assumptions, and methodologies of the

HAZUS fragilities is given in Chapter 2. HAZUS utilized a subjective classification of

bridges based on seismic design, span length, bent type, and span discontinuity. However,

147
the current study utilized a performance-based grouping strategy. HAZUS failed to

consider the variability of geometric and material attributes, which this study incorporated.

Despite the differences between the present study and HAZUS, Sa (1.0s) is adopted as the

intensity measure in both cases. Also the number of damage states characterizing the bridge

system vulnerability is similar in both studies. Table 6.19 presents the median values of

HAZUS fragility curves corresponding to slight (s), moderate (m), extensive (e), and

complete (c) damage states. HAZUS suggests a single value of dispersion (ds) across all

bridge classes and damage states. The equivalent bridge class notations between HAZUS

and the current study are presented in Table 6.19 to facilitate comparison. The comparison

of HAZUS and the present study’s selected bridge class fragilities are shown in Figures

6.12 and 6.13.

Table 6.19 – Comparison of bridge classes.

Bridge class notation Median fragilities ds

Present study HAZUS s m e c
S-E1-S22/S33-C/R-D/S HWB8/HWB20 0.35 0.45 0.55 0.80 0.60
S-E2-S22/S33-C/O-D/S
HWB9/HWB21 0.60 0.90 1.30 1.60 0.60
S-E3-S22/S33-C/O-D/S
T-E1-S22/S33-C/R-D/S
HWB10/HWB22 0.60 0.90 1.10 1.50 0.60
M-E1-S22/S33-C/R-D/S
T-E2-S22/S33-C/R-D/S
M-E3S22/S33-C/R-D/S
HWB11/HWB23 0.90 0.90 1.10 1.50 0.60
T-E2-S22/S33-C/R-D/S
M-E3-S22/S33-C/R-D/S

By comparing the HAZUS fragilities with the fragilities generated in this study, the

following inferences can be made:

148
• For the lower limit states BSST-0 and BSST-1, bridges are more vulnerable than is

currently predicted by HAZUS. This trend is the same across all of the bridge classes

considered in the present study.

• Single column bent seat abutment bridges in Era 11 are more vulnerable than is

currently predicted by HAZUS (Figure 6.12). For example, for the respective limit

states BSST -1, -2, and -3, bridges in the S-E1-S22-C-S class are 338%, 200%, 90%,

and 90% more vulnerable than HAZUS predicts.

• For all bridge classes, HAZUS either over-estimates or under-estimates the fragilities

for higher limit states BSST-2 and BSST-3. The over-estimation or under-estimation

depends on the column cross-section, design era, number of spans, and number of

columns per bent.

• HAZUS fragilities are unreliable for Era 11 two-span and multi-span bridges (Figure

6.13) resting on seat and diaphragm abutments for all the limit states. For example,

HAZUS overestimates the bridge fragilities by 650%, 543%, 340% and 330%,

compared to T-E1-S22-R-S for the limit states BSST -1, -2, and -3 respectively.

• Era 33 two-column bridges with diaphragm abutments are less vulnerable than is

predicted by HAZUS.

• The dispersions obtained in the current study are close to the HAZUS values, except

for classes of multi-column bent bridges, where the dispersions are higher than HAZUS

values.

• Based on the current study, it is reasonable to conclude that existing HAZUS bridge

groupings and their associated fragilities lead to a non-realistic estimation of seismic

fragilities for bridges in California.

149
Figure 6.12– Comparison of HAZUS and selected bridge class fragilities for a) Era
11 bridges with single column bent and b) Era 22 and Era 33 bridges with single
column bent.

150
 Figure 6.13– Comparison of HAZUS and selected bridge class fragilities a) Era 11
two-span bridges with two- and multi-column bents and b) Era 22 and Era 33 two-
span bridges with single and multi-column bents.

151
6.5 Closure

This chapter presents the multi-phase framework for the generation of fragility

curves, and the fragility curves for selected bridge classes. The framework includes the

selection of a bridge class, characterization of bridge attributes such as material and

geometric uncertainties, creation of numerical component models, non-linear time history

analysis, convolution with limit-state models, and generation of system fragility curves

through joint probabilistic seismic demand models (JPSDMs). The variations in the

material and geometric properties of the bridges were accounted for on the basis of the plan

review of California bridges. A set of numerical bridge models reflecting these

uncertainties and nonlinear responses of various bridge components was created using

OpenSees, and each of the bridge models obtained from a Latin hypercube sampling (LHS)

was randomly paired with one of ground motions. The response data of each component,

monitored from dynamic analyses, was used to develop the associated probabilistic seismic

demand models (PSDMs). An important aspect presented in this chapter is the suggestion

of limit-state models based on the literature review of experimental studies on bridge

columns.

Bridge component and system fragilities were generated for 72 bridge classes and

are presented in detail in this chapter. The generated fragility curves were compared with

the existing HAZUS fragility curves. The following are some of the significant findings

presented in this chapter:

• The seismic vulnerability for all bridge classes reduced with evolutions in column

design philosophy (ductile detailing).

• Multi-column bents are more vulnerable than single column bents. The increased

vulnerability of multi-column bents is mainly due to the bridge width and modeling

assumptions.

152
 • Across the various design eras, diaphragm abutment bridges are less vulnerable than

seat abutment bridges.

• Columns are the most vulnerable components in the case of bridges with diaphragm

abutments. In seat abutment bridges, columns and bearings are significant

contributors to the overall vulnerability.

• There is a wide disparity between the fragility curves generated in the present study

and the existing HAZUS fragility curves. Based on the current study, it is reasonable

to state that existing HAZUS bridge groupings and their associated fragility curves

lead to a non-realistic estimation of the seismic fragilities.

153
CHAPTER 7 PARAMETERIZED FRAGILITY CURVES: LASSO

APPROACH

This chapter explores the relative impact of various uncertain input

variables and level of treatment needed for these variables in the estimation of seismic

demand models and fragility curves. As seen in the previous chapters, the seismic fragility

of bridges has been expressed with one-dimensional (1-D) fragility curves developed with

low-degree polynomial demand models (linear regression models) conditioned a single-

parameter (IM). It is difficult to estimate the sensitivity of seismic demand models to input

parameters in traditional single parameter fragility analysis, as the demand model is

conditioned only on IM. Also, as stated in Ghosh et al. (2013), single-parameter demand

models and fragility curves have some limitations: (1) the inability to account for the

influence of uncertainty (modeling) parameters on structural performance during

earthquakes without extensive re-simulations for each new set of parameter combinations,

(2) the inability to explicitly address the effect of uncertainty parameters on fragility

curves, and (3) the lack of flexibility to incorporate field instrumentation data resulting

from monitoring of highway bridges to enable updating of fragility estimates.

Recently, to alleviate the limitations of the single-parameter demand models,

logistic regression in conjunction with multi-parameter demand models comprising various

predictor variables has been gradually increased in the realm of seismic vulnerability and

loss estimation of bridges (Seo and Linzell, 2012; Dukes, 2013; Dukes et al. 2017 Ghosh

et al., 2013; Kameshwar and Padgett, 2014; Park and Towashiraporn 2014; Mangalathu et

al., 2015; Jeon et al., 2017; Mangalathu et al. 2017c). Assuming that the input variables are

statistically independent, a multi-parameter demand model of each bridge component

(demand parameter) is constructed. Samples obtained from this demand model are

compared with those of the associated limit-state model to obtain the binary survival-

154
failure vector. This vector is used to perform a logistic regression analysis to determine the

regression coefficients and thus develop the multi-parameter fragility curve in the

component. This chapter (1) identifies the variables that exhibit strongest influences on the

seismic demand and seismic fragilities, (2) quantifies the relative impact of various sources

of uncertainty on the seismic response of bridges, (3) compares the regression-based

response surface models such as linear, stepwise, Lasso, Ridge, and elastic net in the

estimation of seismic demand models, and (4) suggests a parameterized fragility

methodology that accounts for the effect of uncertain input variables in the seismic demand

models as well as seismic fragilities. Such a study (1) provides insight in quantifying

whether the variation of uncertain parameters should be treated explicitly or be neglected,

(2) eliminates the parameters which have a minimal influence on the seismic demand and

reduces unnecessary and exhaustive efforts in statistical sampling, (3) identifies the

parameters that can reduce the uncertainty in demand models and fragility curves by more

explicit evaluation of the uncertainty distribution (e.g., by developing an extensive

database), and (4) helps bridge owners, such as California Department of Transportation,

spend their resources judiciously (e.g. data collection, field investigations, censoring) on

parameters that have a significant influence on bridge fragilities.

The generation of the parameterized fragility curves for the bridge classes discussed

in Chapter 6 is beyond the scope of this thesis. It is noteworthy to mention that the

geometric, material, and structural uncertainties, ground motions, and fragility curves in

this chapter are used to demonstrate the approach and are not consistent with the earlier

chapters. Note that the purpose of this chapter is to suggest a methodology and demonstrate

its application. The approach is explained with a case study of two-span box girder bridges

in California. A brief review of various regression models, such as linear, stepwise, Lasso,

Ridge, and elastic net, is given in the next section. The efficiency of these models in

155
estimating the seismic demand is then compared using the mean square error (MSE) and

absolute error (ABS) in predicting the seismic demand models.

7.1 Regression models

The following subsections describe the various regression models used for

estimation of seismic demand models. Five regression models such as linear regression,

stepwise regression, Ridge regression, Lasso regression, and elastic net regression are used

in this paper; this subsection describes their relevance to seismic demand modeling.

7.1.1 Linear Regression

The linear regression (or least squares fitting) is the simplest and most commonly

applied form of regression technique and provides a solution to the problem of finding the

best-fitting straight line through a set of points. In the case of the seismic demand model

for bridges with input vector, X = (X1, X2,..., Xp) and a real-valued seismic demand (output

from NLTHA) D, the linear regression has the following form:

p
D  0   X j  j (7.1)
j 1

where j’s are the unknown parameters or coefficients and p is the number of input

parameters. With a training data set ( x1 , d1 ),..., ( xN , d N ) the most popular method for the

estimation of  is to minimize the residual sum of squares (RSS, Equation 7.2). The training

set data is explained in section 7.2.

N
RSS (  )   (d i  f( xi )) 2
i 1
N p (7.2)
  (d i   0   xi j  j ) 2
i 1 j 1

156
 The least square estimates of the parameter  have the smallest variance among all

linear unbiased estimates (GuassMarkov Theorem).

7.1.2 Stepwise regression

In the stepwise regression approach, only a subset of the input variable is retained

and the rest of the variables are eliminated by a selection criterion. Forward stepwise

regression is used in this paper. Forward stepwise regression starts with the intercept, and

subsequently adds the variables into the model that most improve the fit. The improvement

in fit is often based on the F statistic (Equation 7.3) in which the variables are sequentially

added until the model attains the largest value of F.

RSS ( ˆ )  RSS (  )
F (7.3)
RSS (  ) / ( N  k  2)

The parameter estimate ˆ in Equation 7.3 is with k inputs and the estimate  is with

the addition of a predictor. The F-ratio stopping rule doesn’t attempt to find the best model,

as the stopping rule provides only local control of the model search (Friedman et al. 2001).

7.1.3 Ridge regression

The least square estimates often have low bias but suffer the drawback of large

variance (Hoerl and Kennard, 1970). The search for a biased estimator with smaller mean

square error (MSE) and significant reduction in variance led to the development of Ridge

regression (Tibshirani, 1996). Regression coefficients are shrunk by the Ridge regression

by imposing a penalty on their size; the Ridge regression minimizes a penalized residual

sum of squares (Equation 7.4)

157
  N p p

ˆridge  argmin  (di   0   xij  j ) 2     j2  (7.4)
  i1 i 1 j 1 

The  in Equation 7.4 corresponds to the shrinkage parameter, as the larger the

value of , the greater the shrinkage (towards zero) of the regression coefficients. The

inputs have to be standardized before applying the Ridge regression because the Ridge

solutions are not equivariant under the scaling of the input variables. Also, the intercept 0

has not been penalized in the Ridge regression.

7.1.4 Lasso regression

Lasso regression is similar to Ridge regression, but can do variable selection

(Tibshirani, 1996). Lasso regression minimizes the residual sum of squares, subjected to a

constraint based on the sum of absolute values of regression coefficients (Equation 7.5).

The Lasso estimate ˆlasso is defined as

 N p p

ˆlasso  argmin  (di   0   xij  j )2     j  (7.5)
  i1 i 1 j 1 

In Lasso regression, the penalty is imposed on  p

1
j , and because of the form of

the Lasso penalty, Lasso regression does variable selection and shrinkage of regression

coefficients. In other words, the most significant variables are retained while the

insignificant variables are removed from the model. Bias is more controllable in Lasso

regression when compared to Ridge regression (Tibshirani, 1996).

158
7.1.5 Elastic net

Zou and Hastie (2005) suggested another regularization and variable selection

technique called elastic net, which is given as

 N p
1  p p

ˆelastic  argmin  (di  0   xij  j )2    j2    j  (7.6)
  i1 i 1 2 i 1 j 1 

The penalty is imposed on  p

1
j and  1
1
 j2 in elastic net regression and provides

a bridge between Lasso regression and Ridge regression. It is particularly useful for

analyzing high dimensional data. In Ridge, Lasso, and elastic net,  needs to be specified

by the analyst.

The relative advantages of the above-mentioned regression techniques in the

evaluation of seismic demand model will be discussed in Section 7.3. More detailed

descriptions of the various regression techniques can be found in Friedman et al. (2001).

7.2 Case-study bridges: numerical modeling, uncertainties, ground motion suite,

and demand parameters

Two-span box girder bridges are the most common type of highway bridge

inventory in California and account for more than 35% of the box girder bridge inventory.

Two-span bridges with seat and diaphragm abutments are selected for the case study in this

paper, and the selected bridges were designed and constructed prior to 1970. Figure

7.1shows the numerical modeling of various bridge components; three-dimensional

numerical modeling is carried out with the help of the finite element package OpenSees. A

detailed description of the modeling strategy is given in Chapter 3.

159
 Different sources of geometric, material, and system uncertainties are included in

this study. Table 7.1 shows the mean value (), standard deviation (), and the associated

probability distribution of various input variables used in the current study. Mangalathu et

al. (2016a) identified the input variables based on an extensive plan review and hence

mimic the California bridge inventory. The current study selects the suite of ground

motions developed by Baker et al. (2011), which was proposed as part of the PEER

Transportation Research Program. The suite comprises 120 pairs of broadband ground

motions and 40 pairs of near-fault ground motions. The entire suite of ground motions are

scaled by a factor of two (Ramanathan, 2012) to have sufficient response data of IMs higher

than the Palmdale spectrum (the highest probabilistic design hazard level in California),

and thus the expanded suite of 320 ground motions is used for the current study. The

spectral acceleration at 1.0 sec (Sa–1.0s) is adopted as the intensity as the IM in the current

study based on the work of Ramanthan (2012).

The input variables are sampled across the range of parameters presented in Table

7.2 using Latin Hypercube Sampling technique to generate statistically significant yet

nominally identical three–dimensional bridge models. The variables are randomly paired

with the selected suite of ground motions. The two orthogonal components of the ground

motions are randomly assigned to the longitudinal and transverse direction of the bridge

axis. A set of NLTHAs (320 simulations) is performed for all bridge-ground motion pairs

to monitor the maximum response of various bridge components. The various demand

parameters considered in this study are presented in Table 7.2.

160
Figure 7.1 – Numerical modeling of various bridge components.

161
 Table 7.1 – Uncertainty distribution considered in the bridge models.
Distribution
Parameter Units
Type μa σ*
Concrete compressive strength (fc) MPa Normal 29.03 3.59
Reinforcing steel yield strength (fy) MPa Lognormal 465.0 37.30
Span length (L) mm Lognormal 31775 8738
Deck width (Bd) mm Lognormal 9780 1980
Column height (H) mm Lognormal 6625 865
Abutment backwall height (Ha)
Diaphragm abutments
On piles mm Lognormal 3234 488
On spread footings mm Lognormal 2925 1056
Seat-type abutments
On piles mm Lognormal 2186 441
On spread footings mm Lognormal 2186 441
Abutments on piles - Lateral capacity/deck width (Kpa)
Diaphragm abutment N/mm Lognormal 1120 404
Seat-type abutment N/mm Lognormal 1498 540
Elastomeric bearing pad
Stiffness per deck width (Kb) N/mm/m Lognormal 908 327
Coefficient of friction for bearing pad (b) – Normal 0.30 0.10
Gap (g)
Longitudinal (btw. deck and abutment wall, l) mm Lognormal 23.5 12.5
Transverse (btw. deck and shear key, t) mm Lognormal 12.8 2.58
Mass factor (m) Uniform 1.25 0.007
Damping () Normal 0.045 0.0125
Acceleration for shear key capacity (as) g Lognormal 1.00 0.20
Longitudinal reinforcement ratio () (%) Uniform 2.25 0.52
Pile group – pile cap and piles
Translational stiffness (Kft)
Single column – 1% long. rebar N/mm Normal 297716 140101
Single column – 3% long. rebar N/mm Normal 245178 105076
Multi column – 1.5% long. rebar N/mm Normal 140101 105076
Rotational stiffness (Kfr)
Single column – 1% long. rebar N-m/rad Normal 4.5109 1.1109
Single column – 3% long. rebar N-m/rad Normal 6.810 9
1.1109
Multi column – 1.5% long. rebar N-m/rad Normal 0 0
Superstructure box type
Reinforced vs. cast-in-place prestressed concrete(BT) Bernoulli Equally split**
Abutment backfill type (sand vs. clay, ST) Bernoulli Equally split**
Ground motion direction (fault parallel)
Longitudinal vs. transverse (ED) Bernoulli Equally split**
Column diameter (1524 mm vs. 1828.8 mm, D) Bernoulli Equally split**
*µ and  denotes the mean and standard deviation of the distribution. ** 160 simulations are carried out
by one type and the remaining 160 simulations are carried out by other type and are chosen randomly

162
 Table 7.2 – Bridge component demand parameters.
Bridge component Demand parameter Abbreviation Units
Column Curvature ductility COL –
Abutment Passive abutment displacement ABP mm
Active abutment displacement ABA mm
Transverse abutment displacement ABT mm
Deck Displacement DEC mm
Bearing Superstructure unseating UST mm
displacement
Bearing deformation BRD mm
Foundation Translation FNT mm
Rotation FNR rad

7.3 Comparison of the regression models

As the variable of interest ranges over several orders of magnitude (Table 7.1), the

demand and input variable are transformed into the logarithmic space (Cornell et al. 2000;

Mangalathu et al. 2016b). Also, because Ridge, Lasso, and elastic net regression models

are sensitive to the scaling of input variables, the input variables are transformed (after the

logarithmic transformation) into standard space (zero mean and unit variance). It is

assumed that all of the input parameters are independent of each other and are therefore

non-correlated. As mentioned before, Ridge, Lasso, and elastic net models identify the

variables that have less influence on the regression model by penalizing the regression

coefficient associated with the variable to zero. The regression coefficient of the input

variable that has a minimal variance on the regression is penalized and the penalization

procedure depends on the type of regression formulation (Equation 7.4, 7.5 and 7.6). A

detailed description of the penalization procedure and algorithms can be found in Friedman

et al. (2001). In general, the larger the penalty applied, the greater the shrinkage of

regression coefficients (or setting the regression coefficient associated with the least

163
significant variables in the regression model to zero). However, an increase in penalization

beyond a certain point might lead to the removal of variables that have a significant

influence on the demand model. Hence, an investigation is carried out to find the optimal

penalty factor.

7.3.1 Investigation of penalty factor

The performance or penalization of Ridge, Lasso, and elastic net regression models

depends on the shrinkage factor  (Equation 7.4, 7.5 and 7.6). Thus, this study is carried

out to (1) understand the variation of MSE with  and (2) determine the optimal value of

 for the estimation of seismic demand models. The MSE for a vector D with n predictors

can be estimated as
1 n
MSE   ( Di  Di )2
n i 1
(7.7)

As mentioned before,  controls the amount of shrinkage of regression coefficients.

MSE is estimated in this study through tenfold cross validation: the data is fitted on nine-

tenths of the data, and the prediction error is computed for the remaining data. An in-depth

discussion on the cross-validation techniques can be found in Friedman et al. (2001). Note

that  = 0 corresponds to the linear regression model, and in the case of  = , all the

regression coefficients except intercept are shrunk to zero. The results of the investigation

of shrinkage factor are presented in Figure 7.2 and Figure 7.3. Figure 7.2 shows some

bridge components for the variation of MSE with , while Figure 7.3 shows the shrinkage

of the regression coefficients with . Figure 7.3 measures the number of regression

coefficients (or significant parameters) retained in the model with the increase in . The

following inferences obtained from Figure 7.2 and Figure 7.3 are summarized below.

164
 • MSE decreases with  until reaching an optimal value for Ridge, Lasso and elastic

net regressions and starts increasing beyond this value.

• The MSE of Ridge, Lasso, and elastic net regressions decreases with the increase of

 up to the optimal value, and thus the Ridge, Lasso, and elastic net regressions

produce better performance until the optimal .

• The optimal  varies depending on the component demand parameter. For example,

in the case of diaphragm abutment bridges, the optimal  for COL is 0.014 in the case

of Lasso, while it is 0.032 for ABP.

• The optimal  is different for different regression methods.

• The shrinkage of regression coefficients increases with an increase in ; however, it

also increases the MSE of the demand model once  is beyond the optimal .

• As the recommended values of  (Friedman et al. 2001) are very low (in the order of

10-2 to 10-3), Ridge regression is not able to shrink the regression coefficients.

Although only selected component demand parameters are shown in Figure 7.2, similar

observations are also noted for other demand models.

165
 Figure 7.2 – Comparison of MSE with .

The optimal value of  is adopted as 0.01 in the current study because it is found to

be the lower bound for all the optimal  values in the estimation of seismic demand

parameters. However, such an estimate fails to identify many insignificant variables (or

fails to shrink the regression coefficient), and thus overestimates the shrinkage coefficients.

Nevertheless, the criteria adopted for the selection of  is that the MSE of the models

should (1) be less than the linear regression, (2) shrink highly insignificant regression

coefficients, and (3) serve as a uniform value of  for all the demand parameters. The

optimal choice of  for each demand model is beyond the scope of this paper.

166
 Figure 7.3 – Shrinkage of regression coefficients with .

7.3.2 Comparison of the regression models

The criteria adopted for evaluating the model-fitting procedure must be (1) able to

more accurately predict the future data and (2) a simple model with less number of

regression coefficients (Zou and Hastie, 2005). To evaluate the accuracy in predicting the

future data, the current database is split into two: training set and test set. 75% of the data

is assigned to the training set and is used to fit the model (Friedman et al. 2001). The

remaining 25% of the data is used as the test set to estimate the accuracy of the fitted model.

The assignment of data to the training set and test set is carried out randomly. The reason

to split the data into training set and test set is to avoid the over-fitting of data (Friedman

et al. 2001), which is a common problem if we use the entire data as training set. Table 7.3

167
shows the coefficients from five different regression models for COL with the training set

for the diaphragm abutment bridge. The estimated fit is used to check the error in the test

set. To achieve this goal, two types of error, MSE and absolute error (ABS), are selected

in this paper. The ABS for a vector Yˆ with n predictors can be estimated as

1 ˆ
ABS  (Yi  Yi ) (7.7)
n

Table 7.3 – Estimated coefficients and test error for COL for the bridge with
diaphragm abutments by various regression techniques

Parameter Linear Stepwise Ridge Lasso Elastic

regression regression regression regression net
Intercept 0.376 0.376 0.376 0.376 0.376
Sa–1.0s 1.129 1.120 1.119 1.117 1.118
ST -0.034 -0.035 -0.031 -0.033
BT -0.007 -0.005
L 0.179 0.176 0.180 0.173 0.177
H 0.033 0.033 0.022 0.028
Bd 0.169 0.162 0.163 0.126 0.146
D -0.243 -0.232 -0.235 -0.199 -0.219
 -0.411 -0.415 -0.395 -0.407 -0.407
Ha 0.003 0.000
Kpa -0.055 -0.058 -0.028 -0.032
Kft 0.022 0.033
Kfr -0.002 -0.017 -0.003
fc -0.810 -0.899 -0.228 -0.066 -0.072
fy 0.734 0.821 0.151
m 0.115 0.107 0.117 0.104 0.110
 0.006 0.010 0.005 0.009
ED 0.014 0.015 0.001 0.007
MSE (test set) 0.474 0.471 0.459 0.444 0.447
Std. MSE 0.079 0.079 0.077 0.075 0.075
ABS (test set) 0.534 0.527 0.528 0.518 0.519
Std. ABS 0.049 0.049 0.048 0.047 0.047

The standard error of the mean value of the MSE (std.MSE) and ABS (std.ABS) is
estimated as
standard deviation (MSE)
Std.MSE 
n
standard deviation (ABS) (7.8)
Std.ABS 
n

168
 Table 7.3 also gives the MSE, Std.MSE, ABS, and Std.ABS value of the various

regression models. The comparison of MSE and ABS helps to demonstrate how well the

fit explains a given set of test data. The results from Table 7.3 show that (1) linear and

Ridge regressions identify all the input variables as significant; (2) stepwise regression is

the one which identifies the least number of significant parameters; (3) Lasso regression

model has the minimum of MSE, Std. MSE, ABS, and Std. ABS, and thus is the best fit

amongst the selected regression models; (4) among Lasso, Ridge, and elastic net, Lasso

regression identifies more insignificant parameters; (5) these three regression methods

identify that fy, Kft, and BT have less effect on the seismic demand model for COL, and

thus it can be deduced that these input parameters are the least significant parameters; (6)

all the regression methods identify Sa–1.0s as the most significant parameter; and (7)

standard error for all of the models are fairly similar. MSE and ABS are compared for the

different regression models for the various demand parameters and are plotted in Figure

7.4. Lasso regression is the one having the least MSE and ABS for all the demand models.

The MSE and ABS associated with COL, UST, and BRD are low for all of the regression

models, which shows the good predictive capability. However, the MSE and ABS for ABA

and ABP are higher for all the models. Previous studies (Ramanathan 2012; Mangalathu et

al. 2016a) pointed out that columns and bearings are the components that govern the system

fragility, and thus the adopted models are good enough to capture the seismic demand of

COL and UST. Note that the results shown in Table 7.3and Figure 7.4 are for  = 0.01 and

the results can be significantly improved by selecting a better optimal value for . However,

the selected  yields good results and is able to remove insignificant parameters from the

regression model. In addition, stepwise regression might result in the data over-fitting and

instability (Vidakovic 2011), so Lasso regression is adopted as the regression model for

further part of this study. Lasso regression leads to low values of MSE and ABS and has

the ability to shrink the regression coefficients.

169
 Figure 7.4 – Radar plot depicting the comparison of accuracy of fit obtained from
the various regression models.

7.4 Sensitivity of input parameters to the seismic demand model

The regression coefficients from Lasso regression are also a measure of sensitivity

of input parameters to the seismic demand model or fragilities because the input parameters

170
are converted to a standard space (after a logarithmic transformation of the data) in Lasso

regression. A positive regression coefficient for a particular variable indicates that the

seismic demand increases with a positive increase in the variable, and a negative regression

coefficient shows that an increase in that variable reduces the seismic demand. Figure 7.5

shows the sensitivity of demand models for bridges with diaphragm and seat abutments. It

is noted from Figure 7.5 that the significant parameters for both bridges tend to vary from

component to component. In general, the IM (here, Sa–1.0s) has the greatest influence on the

demand model for all the demand parameters. Span length (L) is the second most sensitive

parameter for all the demand parameters. Soil type has a significant influence on the ABA

and ABP in the case of the diaphragm abutment bridges. The influence of soil type is due

to the fact that diaphragm abutments are stiffer than the adjacent bent and attract a large

portion of the seismic demand. Similar conclusions are also noted in previous studies

(Ramanathan 2012; Mangalathu et al. 2016a). COL is significantly affected by the

reinforcement ratio () and the column diameter (D) for the two case-study bridges. L and

IM are the most significant parameters for UST.

Concrete strength (fc), steel strength (fy), damping ratio (), superstructure box type

(BT), earthquake direction (ED), abutment height (Ha), acceleration for shear key capacity

(as), gap between the deck and shear key (t), and coefficient of bearing (µb) all have a

minimal impact on all the seismic demand models. In addition, Jeon et al. (2017) used a

Bayesian parameter estimation method coupled with stepwise regression to identify

significant parameters affecting the seismic response of curved concrete box girder bridges.

Their work concluded that fc, fy, ED, µb, Ha, and t are the least significant parameters on

the seismic demand model of the bridges. The current sensitivity study shows that other

variables, except for IM, also have a significant influence on the seismic demand. The

demand model conditioned only on the IM, as used in traditional fragility analysis, might

not lead to a realistic estimation of the seismic demand and the fragility curves. Hence, a

171
fragility methodology accounting for the influence of significant parameters in the demand

model as well as the fragility is given in the next section.

Figure 7.5 – Sensitivity of input parameters.

7.5 Multi-Parameter fragility curves

Recently, a number of studies (Seo and Linzell, 2012; Dukes, 2013; Ghosh et al.,

2013; Kameshwar and Padgett, 2014; Park and Towashiraporn 2014; Jeon et al., 2015;

Mangalathu et al., 2015) generated parameterized component and system fragility curves

of highway bridges using multi-parameter demand models in conjunction with logistic

regression technique. A modified approach stemming from the previous research is

172
suggested in this paper. Unlike the previous studies, the proposed approach helps to

identify the relative impact of the various input parameters on the seismic demand as well

as fragilities. Also, the proposed approach removes the less significant variable from the

generation of seismic demand model and the seismic fragilities without much

computational effort. The outline of the proposed approach is given below, for uncertain

input parameters (x1,…,xn, IM) and the demand measures (d1,…,dns).

Step 1: Evaluate the linear regression coefficients (li) by performing Lasso regression

analysis for each component (ki, i = 1,…,m) with the input parameters (x1,…,xn,

IM), assuming that the input variables are statistically independent.

Step 2: Generate a large number of demand estimates (N, 1 million in this study) for each

component, ki, using their respective Lasso regression model by generating N

values of randomly generated input parameters based on their probabilistic

distribution.

Step 3: Generate N capacity values for a specific damage state for each bridge component

based on the assumed distribution of the limit states (Table 7.1).

Step 4: Obtain the binary survive-failure (N  1) vector by comparing the capacity values

(step 3) with the demand values (step 2).

Step 5: Conduct a Lasso logistic regression on the survive-failure vector to determine the

kth component probability model, conditioned on the input parameters as

nl
 k ,0  k ,IM ln(IM)  k j ln(x j )
j 1
e
PFk|IM, x1 , x2 ,.. xn  nl
, nl  n (7.9)

l
 k,0  k,IM ln(IM)   k j ln(x j )
1 e j 1

173
 where k,0, k,sa, and k,j’s (j = 1,…,nl) are the Lasso logistic regression coefficients

of the kth bridge component. This step helps to identify the sensitivity of bridge

component fragility curves to the uncertain input parameters.

Step 6: Assuming that the bridge failure is a series system (the system fails if one or more

components fail), estimate the binary survive-failure vector, and conduct a Lasso

logistic regression to obtain the system failure. This step helps to identify the

sensitivity of bridge system fragility to the uncertain input parameters

ns
 SYS ,0  SYS ,IM ln(IM)   SYS , j ln(x j )
j 1
e
PFSYS |Sa , x1 , x2 ,.. xn  ns
, ns  n (7.10)
 SYS , j ln(x j )
s
 SYS ,0  SYS ,IM ln(IM) 
1 e j 1

where SYS,0, SYS,sa, and SYS,j’s (j = 1,…, ns) are the Lasso logistic regression

coefficients for the system failure.

Step 7: For a particular bridge with significant input parameters, x1 ,..., x ns , the classical one-

dimensional fragility curves can be obtained as

ns
 SYS ,0  SYS ,IM ln(IM)   SYS , j ln( x j )
j 1
e
PFSYS |IM    ...  ns
f ( x1 )... f ( xns ) dx1 ...dxns
x1 x2 xns  SYS ,0  SYS ,IM ln(IM)   SYS , j ln( x j )
1 e j 1

(7.11)

where f(x1),…, f(x9) are the probability density parameters for parameters,

x1 ,..., xns .

The limit state models for the various bridge components are given in Table 7.4,

and are consistent with the limit states presented in Chapter 6. The limit states were derived

174
in such a way as to align with the California Department of Transportation (Caltrans)

design and operational experience. This will facilitate Caltrans’ evaluation of repair–

related decision variables, repair cost, and repair time.

Table 7.4 – Limit state models of various bridge components.

Median value, Sc
Component c
Slight Moderate Extensive Complete

Column curvature ductility (COL, –) 0.80 2.0 5.00 8.00 0.35

Passive abutment response (ABP, mm) 76 254 – – 0.35
Active abutment response (ABA, mm) 38 102 – – 0.35
Transverse abutment response (ABT, mm) 25 102 – – 0.35
Deck displacement (DEC, mm) 102 305 – – 0.35
Bearing displacement (BRD, mm) 25 76 – – 0.35
Superstructure unseating (UST, mm) – – 152 229 0.35

Based on the approach mentioned in this section, multi-parameter demand model

and fragility curves are generated for the bridges with diaphragm and seat abutments for

various limit states. Figure 7.6 shows the single parameter fragility curves (conditioned on

Sa-1.0s) for the diaphragm and seat abutments for the moderate damage state. It is clear from

Figure 7.6 that the bridge fragility is mostly dominated by column fragilities for the

moderate damage state. ABT and BRD, respectively, are the second most vulnerable

components in the case of diaphragm abutment bridges and seat abutment bridges.

Although not shown here, similar conclusions are observed for other limit states. Interested

readers are directed to previous studies (Mangalathu et al. 2015, Jeon et al. 2017) for the

comparison of the fragility curves through the multi-parameter demand model with the

traditional single-parameter demand model.

175
 Figure 7.6 – System and component fragility curves for moderate damage state: a)
diaphragm abutment bridge, b) seat abutment bridge.

As mentioned before, the proposed approach also helps identify the sensitivity of

the fragility curves to the input parameters. Figures 7.7 and 7.8 show the sensitivity of the

fragility curves to the input parameters for the diaphragm and seat abutment bridges,

respectively. The uncertainty in IM dominated over all the other uncertainties in the

fragility curves. In the case-study bridges where the fragility of COL determines the bridge

fragility, Sa–1.0s, L, , D, Bd, and m are the most sensitive parameters for the diaphragm

abutment bridges for system fragility, while Sa–1.0s, L, , D, Bd, m, and l are the sensitive

parameters in the case of seat abutment bridges. fy, , ED, and Kfr are the least significant

parameters for all the components for all the damage states. Another advantage of the

proposed approach is the ability to identify the sensitivity of input variables based on the

limit state under consideration. For example, the sensitivity of IM is higher for ABP at the

moderate damage state when compared to the slight damage state in the case of diaphragm

abutment bridges. It is deduced from the comparison of the magnitude of regression

coefficients.

176
Figure 7.7 – Sensitivity of fragility curves to input parameters for diaphragm
abutment bridge for various limit states

177
Figure 7.8 – Sensitivity of fragility curves to input parameters for seat abutment
bridge for various limit states

178
7.6 Conclusions

This chapter identifies the relative impact of various uncertain input parameters on

the seismic response of various bridge components. The efficiency of various regression

models such as linear, stepwise, Ridge, Lasso, and elastic net in the generation of seismic

demand models are evaluated in the initial part of the paper. The comparison is carried out

for two-span box girder bridges with seat and diaphragm abutments. Nonlinear time history

analysis (NLTHA) is carried out for the bridge models accounting for the material,

geometric, and system uncertainties. Various demand parameters such as curvature

ductility demand, abutment displacements in the passive, active, and transverse directions,

deck displacement, foundation translation and rotation, superstructure unseating

displacement, and elastomeric bearing displacement are recorded for each NLTHA. Multi-

parameter demand models are generated for each demand parameter by using 75% of the

data from the NLTHA. The efficiency of the regression methods are compared in terms of

the mean square error (MSE) and absolute error (ABS) in predicting the remaining 25% of

the NLTHA data. It is observed that the Lasso regression model is the most effective with

regard to lowest MSE and ABS in predicting the data. Also, the Lasso regression model is

able to remove insignificant variables from the demand model. As the Lasso regression

coefficients are a measure of the sensitivity of the input variables, the results of the Lasso

regression is used to identify the significance of the input variables in the estimation of

seismic demand models for various bridge components. Ground motion intensity measure

(IM, here, spectral acceleration at 1 sec) and span length (L) are identified as the most

sensitive variables in the demand models for various demand parameters. In general, the

steel strength (fy), damping ratio (), superstructure box type (BT), earthquake direction

(ED), abutment height (Ha), acceleration for shear key capacity (as), gap between the deck

and shear key (t), and coefficient of bearing (µb) all have a minimal impact on all the

seismic demand models.

179
 The sensitivity results reveal that the generation of demand models without

considering the uncertainties in the significant parameters might lead to inaccurate

estimates of the demand models. To include the effect of significant uncertain variables in

the generation of demand models and fragility curves, a multi-parameter fragility

methodology using Lasso regression is suggested in this paper. The proposed fragility

approach helps to identify the relative impact of the various input parameters on the

demand as well as fragility curves of various structural components. Hence, the proposed

approach provides additional perspectives for the decision makers or the owners in

prioritizing their resources in the database development or structural monitoring. The

sensitivity study on fragility curves indicates that IM, L, reinforcement ratio (), column

diameter (D), deck width (Bd), mass factor (m), and gap between the deck and shear key

(l) are the parameters significantly affecting on the bridge fragilities. In general, the

component as well as system fragilities are minimally affected by fy, , ED, and foundation

rotational stiffness (Kfr). The finding of the sensitivity study is helpful in evaluating the

level of uncertainty treatment required for each variable. Depending on the user

requirement and whether the intention is to have system or component vulnerability, the

proposed approach notifies whether the uncertainty in a particular variable should be

treated explicitly or be neglected.

Although the findings observed from this study are based on the case studies of

two-span concrete box girder bridges in California, the methodology is relevant and

applicable to other structures as well. As the level of uncertainty treatment is a common

challenge in regional risk assessment, the proposed approach helps to identify whether the

uncertainty for a particular variable needs to be treated explicitly or be neglected. The

identification of level of uncertainty treatment required for a particular variable also helps

bridge owners spend their resources judiciously and develop a more reliable database of

the uncertain input variables for the seismic risk assessment.

180
 CHAPTER 8 CONCLUSIONS AND FUTURE WORK

8.1 Summary and Conclusions

Regional seismic risk assessment relies on fragilities that are applicable to a

portfolio of structures, as it is time consuming and impractical to generate fragility curves

for each individual structure in a specific region. Also the generation of structure-specific

fragility curves is not warranted as some structures have similar performance or fragilities.

The grouping of structures is carried out in most cases based on engineering judgment and

there is a lack of systematic strategy for binning/grouping structures. In the existing

grouping methodology, HAZUS grouping is the widely accepted approach. However, a

critical review of HAZUS based on recent research revealed many drawbacks and showed

that the HAZUS bridge grouping and associated fragilities leads to an unrealistic estimation

of the seismic demand and the associated losses. These limitations in the traditional

engineering judgment based grouping can be addressed by performance based grouping

techniques. The performance based grouping leads to more reliable sub-classes of bridges

relative to the traditional subjective lumping of bridges. The current study explores various

performance based grouping techniques such as Analysis of Variance (ANOVA), Analysis

of Covariance (ANCOVA) and non-parametric Kruskal-Wallis test (KW) towards the

grouping of structures of similar performance. Based on the insights from the comparison,

a new performance grouping methodology based on ANOVA is suggested in this study.

This study improved bridge classifications of HAZUS by considering various structural

attributes such as column cross-section, design era, number of spans, abutment type, pier

type and span continuity. Although the current study groups the box-girder bridge classes

181
in California, the proposed grouping approach can also be applied to other regions by fine-

tuning the grouping based on the evolution in seismic design philosophy and other

attributes which are different from California.

A major task in the current research was to understand and characterize the

California bridge inventory. California is a state with a high seismic hazard, a history of

damaging earthquakes, and has close to 29,000 bridges which vary in age based on their

construction. Bridge plans pertinent to various design eras and structural configurations

were reviewed in detail, and descriptive statistics were calculated for the material,

geometric and structural parameters of the box-girder bridges in California. Such a

characterization helps to make the fragility models applicable to a wide geographic area.

More than 1000 bridge plans were reviewed for this process with the help of California

Department of Transportation (Caltrans) in-house database.

Numerical bridge models accounting for geometric, material, and system

uncertainties are created in OpenSees. The numerical models incorporate a high degree of

detail with respect to the component modeling strategies and their ability to capture the

damage due to the seismic demand. The input variables are sampled across the range of

parameters using the Latin Hypercube Sampling technique to generate statistically

significant yet nominally identical three–dimensional bridge models. The variables are

randomly paired with the selected suite of ground motions. The two orthogonal

components of the ground motions are randomly assigned to the longitudinal and

transverse direction of the bridge axis. A set of NLTHAs (320 simulations) is performed

for all bridge-ground motion pairs to monitor the maximum response of various bridge

components. Various demand parameters considered in the current study are curvature

182
ductility, abutment displacements in the passive, active and transverse direction,

superstructure unseating displacement, and elastomeric bearing displacement. The demand

models are convolved with the capacity models to generate the component fragility curves.

A significant contribution of the present study is to suggest the capacity limit states (CCLS)

for columns based on extensive experimental review. Pertinent to various design eras,

literature review was carried out to collect the experimental data for bridge columns and

statistical analysis was carried out to suggest the CCLS of bridge columns. Such an exercise

helps to develop a new generation of more accurate and useful bridge fragility models for

incorporation into the ShakeCast earthquake alerting system developed by the California

Department of Transportation (Caltrans) and to support seismic risk evaluation of bridges

in California. System fragility curves are generated afterwards using Monte Carlo

simulations and joint probabilistic seismic demand models (JPSDMs) incorporating the

correlation between the components.

Bridge component and system fragilities are generated for 72 bridge classes and the

selected bridge classes cover more than 75% of the California box-girder bridge inventory.

The following are some of the notable findings from the fragility analysis:

• The seismic vulnerability of all the bridge classes reduced with the evolution in

column design philosophy (ductile detailing).

• Multi-column bents are more vulnerable than the single column bents. The increased

vulnerability of the multi-column bents is mainly due to the bridge width and

modeling assumptions.

• Across the various design eras, diaphragm abutment bridges are less vulnerable than

seat abutment bridges for all the design eras.

183
 • Columns are the most vulnerable component in the case of bridges with diaphragm

abutments. However, in the case of seat abutment bridges, bearings also contribute

significantly to the overall vulnerability in addition to the columns.

• Comparison with HAZUS fragilities revealed the wide disparity between the fragility

curves generated in the present study and the HAZUS fragilities. Based on current

study, HAZUS bridge grouping and associated fragilities leads to an-unrealistic

estimation of the seismic fragilities.

Another contribution in the present study is the generation of bridge-specific fragility

curves through machine learning techniques. The framework includes the selection of a

bridge class, characterization of bridge attributes such as material and geometric

uncertainties, creation of numerical component models, construction of multi-parameter

demand models using the Lasso regression method, and development of parameterized

(multi-dimensional) fragility models using logistic regression. The parameterized fragility

models are used (1) to produce bridge-specific (one-dimensional) fragility curves when the

uncertainty parameters are available and (2) to develop bridge-class (one-dimensional)

fragility curves using a Monte Carlo integration. The advantage of the Lasso regression

model is that it is able to remove insignificant variables from the demand model. As the

Lasso regression coefficients are a measure of the sensitivity of the input variables, the

results of the Lasso regression are used to identify the significance of the input variables in

the estimation of seismic demand model for various bridge components. Ground motion

intensity measure (IM, here, spectral acceleration at 1 sec) and span length (L) are identified

as the most sensitive variables in the demand models for various demand parameters. In

general, the steel strength (fy), damping ratio (), superstructure box type (BT), earthquake

184
direction(ED), abutment height (Ha), acceleration for shear key capacity (as), gap between

the deck and shear key (t), and coefficient of bearing (µb) have a minimal impact on all

the seismic demand models. The proposed fragility approach helps to identify the relative

impact of the various input parameters on the demand as well as fragility curves of various

structural components. Hence, the proposed approach provides additional perspectives for

the decision makers or the owners in prioritizing their resources in the data base

development or structural monitoring.

8.2 Research Impact

This study presents a performance based grouping methodology to group the box-

girder bridge classes in California and generated component and system fragility curves

for single frame multi-span box girder bridges in California. This resulted in a significant

number of contributions which are as follows:

• The grouping has been traditionally performed based primarily on engineering

judgment and prior experience. This work (1) presents an overview of various statistical

techniques such as analysis of variance (ANOVA), analysis of covariance (ANCOVA),

and Kruskal Wallis (KW) test for grouping the bridges of similar performance; (2)

compares the groupings that emerge from the various grouping techniques; and (3)

identifies the method that has more statistical power in creating bridge sub-classes of

distinct structural performance. The grouping is achieved by comparing the structural

responses of bridge classes obtained from the non-linear time history analysis of

bridges.

185
• A simple performance based grouping methodology based on ANOVA is suggested in

this work. The ANOVA-based grouping provides insight into the potential bridge

attributes which significantly affect the seismic response and fragility curves, which to

date has not been assessed thoroughly. The method can be used to group bridge classes

depending upon the user requirement, i.e. whether the user would like to assess the

system or component vulnerability. The grouping of bridge classes yields a more

reliable estimation of the seismic vulnerability of various bridge classes. The proposed

method helps to save considerable computational effort and simulation compared to the

traditional subjective grouping.

• Extensive details are provided regarding the variability in the material, structural and

geometric attributes of box-girder bridges in California. Such a characterization helps

to make the fragility models applicable to a wide geographic area.

• Column capacity limit states are suggested based on the literature review of

experimental studies on bridge columns. The proposed limit states are more reliable,

and help to develop a new generation of more accurate and useful bridge fragility

models for incorporation into the ShakeCast earthquake alerting system.

• Fragility curves and relative vulnerabilities are evaluated for the 72 box-girder bridge

classes in California. The selected bridge classes cover more than 75% of the California

box-girder bridge inventory. The generate fragility curves underscore the necessity to

go beyond the traditional engineering judgment based grouping.

• The study suggests a methodology to evaluate the sensitivity of demand models and

fragility curves to the uncertain input parameters. To include the effect of significant

uncertain variables in the generation of demand models and fragility curves, a multi-

186
 parameter fragility methodology using Lasso regression is suggested in this work. The

proposed approach provides additional perspectives for the decision makers or the

owners in prioritizing their resources in the data base development or structural

monitoring.

8.3 Recommendations for future work

Potential areas in which this work can be extended through additional research include the

following:

• Bridge foundations and abutments may be founded on liquefiable soil, and

significant damage can be seen on bridges in regions with high liquefaction

potential. Also, the random and non-liner behavior of soil can significantly affect

the behavior of bridges. Further studies are needed to explore the effect of

liquefaction, ground deformation hazard and soil-structure interaction on the

fragility curves.

• The performance based strategy should be extended to other bridge class such as I-

girder, T-girder, slab bridges, to mention a few.

• The study focused on the seismic risk assessment due to main-shock ground

motions, and doesn’t account for aging and deterioration mechanism. The effect of

aftershock, aging and deterioration, and material degradation should be

investigated further.

• The current study is limited to machine learning techniques such as Lasso, Ridge

and Elastic net. Future work should investigate other machine learning techniques

such as Random Forest, Support Vector Machine, to name a few. Also the

187
 application of machine learning for fragility curves is limited to two-span box-

girder bridges with circular columns in Era 11. Future work can focus on expanding

the use of the method and tool to other bridge types common in California.

• The current study has adopted relevant capacity models for the bridge type and

design era of interest. In theory, the subgrouping of bridges presented in this paper

may also have slight variations in component capacities. However, such refined

capacity estimates that are attribute-dependent for a range of damage states and

components are still rather lacking and typically require extensive additional

experimental testing for model building or validation. As a result, the current study

has adopted capacity estimates appropriate for the bridge class and era, where

uncertainty in these capacity models is generally a result of the attribute variation

within the class. Further studies are needed for the development of refined capacity

models per damage state and component associated with variations in bridge

attributes. Currently these variations are generally reflected in the uncertainty in a

general capacity model per component, overall bridge class, and design era.

188
APPENDIX A. EARTHQUAKE RECORDS USED FOR FRAGILITY ANALYSIS

Aria
PEER
Scal Sa Vs30 s
Earth Record Rjb Rrup
e (1.0, (m/sec Inten Earthquake Name Year Station Name Magn Mechani
quake Sequence (km) (km)
Fact g) ) sity itude sm
no: Number
or (m/s
ec)
strike 91.2
1 0007 1.18 0.04 219.31 0 "Northwest Calif-02" 1941 "Ferndale City Hall" 6.6 slip 91.15 2
strike 56.8
2 0009 1.17 0.06 213.44 0.1 "Borrego" 1942 "El Centro Array #9" 6.5 slip 56.88 8
strike 27.0
3 0020 1.10 0.35 219.31 0.5 "Northern Calif-03" 1954 "Ferndale City Hall" 6.5 slip 26.72 2
4 0051 1.02 0.05 280.56 0 "San Fernando" 1971 "2516 Via Tejon PV" 6.61 Reverse 55.2 55.2
61.7
5 0056 0.90 0.03 235.00 0.1 "San Fernando" 1971 "Carbon Canyon Dam" 6.61 Reverse 61.79 9
"Cedar Springs 92.5
6 0058 0.87 0.02 477.22 0 "San Fernando" 1971 Pumphouse" 6.61 Reverse 92.25 9
"Gormon - Oso Pump 46.7
7 0065 1.20 0.07 308.35 0.1 "San Fernando" 1971 Plant" 6.61 Reverse 43.95 8
"LA - Hollywood Stor 22.7
8 0068 0.92 0.16 316.46 0.7 "San Fernando" 1971 FF" 6.61 Reverse 22.77 7
9 0070 1.12 0.37 425.34 0.3 "San Fernando" 1971 "Lake Hughes #1" 6.61 Reverse 22.23 27.4
28.9
10 0078 1.04 0.14 452.86 0.3 "San Fernando" 1971 "Palmdale Fire Station" 6.61 Reverse 24.16 9
"Wheeler Ridge - 70.2
11 0092 0.81 0.01 347.67 0 "San Fernando" 1971 Ground" 6.61 Reverse 68.38 3
12 0122 0.81 0.11 249.28 0.1 "Friuli Italy-01" 1976 "Codroipo" 6.5 Reverse 33.32 33.4
13 0126 2.11 1.33 259.59 5.7 "Gazli USSR" 1976 "Karakyr" 6.8 Reverse 3.92 5.46
14 0126 1.03 0.65 259.59 5.7 "Gazli USSR" 1976 "Karakyr" 6.8 Reverse 3.92 5.46

189
 strike
15 0160 2.27 1.01 223.03 6.1 "Imperial Valley-06" 1979 "Bonds Corner" 6.53 slip 0.44 2.66
strike
16 0160 1.11 0.49 223.03 6.1 "Imperial Valley-06" 1979 "Bonds Corner" 6.53 slip 0.44 2.66
strike 10.4
17 0161 1.01 0.26 208.71 0.4 "Imperial Valley-06" 1979 "Brawley Airport" 6.53 slip 8.54 2
strike 10.4
18 0162 0.92 0.15 231.23 0.9 "Imperial Valley-06" 1979 "Calexico Fire Station" 6.53 slip 10.45 5
strike 21.6
19 0172 1.04 0.08 237.33 0.3 "Imperial Valley-06" 1979 "El Centro Array #1" 6.53 slip 19.76 8
strike 12.5
20 0174 2.45 0.58 196.25 2 "Imperial Valley-06" 1979 "El Centro Array #11" 6.53 slip 12.56 6
strike 12.5
21 0174 0.91 0.22 196.25 2 "Imperial Valley-06" 1979 "El Centro Array #11" 6.53 slip 12.56 6
strike 17.9
22 0175 0.91 0.16 196.88 0.4 "Imperial Valley-06" 1979 "El Centro Array #12" 6.53 slip 17.94 4
strike
23 0179 2.13 1.14 208.91 1.4 "Imperial Valley-06" 1979 "El Centro Array #4" 6.53 slip 4.9 7.05
strike
24 0179 0.86 0.46 208.91 1.4 "Imperial Valley-06" 1979 "El Centro Array #4" 6.53 slip 4.9 7.05
strike
25 0180 2.25 1.33 205.63 1.7 "Imperial Valley-06" 1979 "El Centro Array #5" 6.53 slip 1.76 3.95
strike
26 0180 1.09 0.65 205.63 1.7 "Imperial Valley-06" 1979 "El Centro Array #5" 6.53 slip 1.76 3.95
strike
27 0181 2.37 1.15 203.22 1.8 "Imperial Valley-06" 1979 "El Centro Array #6" 6.53 slip 0 1.35
strike
28 0181 1.05 0.51 203.22 1.8 "Imperial Valley-06" 1979 "El Centro Array #6" 6.53 slip 0 1.35
strike
29 0182 2.27 1.53 210.51 1.7 "Imperial Valley-06" 1979 "El Centro Array #7" 6.53 slip 0.56 0.56
strike
30 0182 0.98 0.66 210.51 1.7 "Imperial Valley-06" 1979 "El Centro Array #7" 6.53 slip 0.56 0.56
strike
31 0183 2.24 0.78 206.08 1.6 "Imperial Valley-06" 1979 "El Centro Array #8" 6.53 slip 3.86 3.86
strike
32 0183 0.90 0.31 206.08 1.6 "Imperial Valley-06" 1979 "El Centro Array #8" 6.53 slip 3.86 3.86

190
 "El Centro Differential strike
33 0184 1.85 0.79 202.26 2.1 "Imperial Valley-06" 1979 Array" 6.53 slip 5.09 5.09
"El Centro Differential strike
34 0184 1.02 0.44 202.26 2.1 "Imperial Valley-06" 1979 Array" 6.53 slip 5.09 5.09
strike 30.3
35 0188 0.96 0.04 316.64 0.1 "Imperial Valley-06" 1979 "Plaster City" 6.53 slip 30.33 3
strike 31.9
36 0191 0.88 0.06 242.05 0.3 "Imperial Valley-06" 1979 "Victoria" 6.53 slip 31.92 2
37 0285 0.99 0.27 649.67 0.4 "Irpinia Italy-01" 1980 "Bagnoli Irpinio" 6.9 Normal 8.14 8.18
46.2
38 0287 0.86 0.04 356.39 0 "Irpinia Italy-01" 1980 "Bovino" 6.9 Normal 44.62 5
22.5
39 0288 1.00 0.10 561.04 0.5 "Irpinia Italy-01" 1980 "Brienza" 6.9 Normal 22.54 6
53.1
40 0294 0.87 0.05 496.46 0 "Irpinia Italy-01" 1980 "Tricarico" 6.9 Normal 51.74 6
"Taiwan 92.0
41 0427 1.03 0.02 671.52 0 SMART1(25)" 1983 "SMART1 E02" 6.5 Reverse 91.54 4
"Taiwan 97.6
42 0432 1.01 0.05 267.67 0 SMART1(25)" 1983 "SMART1 O01" 6.5 Reverse 97.16 3
43 0436 1.01 0.02 279.97 0 "Borah Peak ID-01" 1983 "CPP-601" 6.88 Normal 82.6 82.6
"TRA-642 ETR Reactor 79.5
44 0440 0.91 0.01 324.20 0 "Borah Peak ID-01" 1983 Bldg(Bsmt)" 6.88 Normal 79.59 9
"TRA-670 ATR Reactor
45 0441 1.07 0.02 324.20 0 "Borah Peak ID-01" 1983 Bldg(Bsmt)" 6.88 Normal 80 80
16.0
46 0587 0.99 0.21 551.30 0.7 "New Zealand-02" 1987 "Matahina Dam" 6.6 Normal 16.09 9
"El Centro Imp. Co. strike
47 0721 2.27 0.66 192.05 1.1 "Superstition Hills-02" 1987 Cent" 6.54 slip 18.2 18.2
"El Centro Imp. Co. strike
48 0721 0.95 0.28 192.05 1.1 "Superstition Hills-02" 1987 Cent" 6.54 slip 18.2 18.2
strike
49 0723 2.22 1.60 348.69 3.7 "Superstition Hills-02" 1987 "Parachute Test Site" 6.54 slip 0.95 0.95
strike
50 0723 1.08 0.78 348.69 3.7 "Superstition Hills-02" 1987 "Parachute Test Site" 6.54 slip 0.95 0.95
strike 22.2
51 0724 1.06 0.16 316.64 0.6 "Superstition Hills-02" 1987 "Plaster City" 6.54 slip 22.25 5

191
 "Salton Sea Wildlife strike 25.8
52 0726 1.08 0.19 191.14 0.4 "Superstition Hills-02" 1987 Refuge" 6.54 slip 25.88 8
Reverse 23.9
53 0730 1.07 0.32 343.53 0.3 "Spitak Armenia" 1988 "Gukasian" 6.77 Oblique 23.99 9
Reverse 24.5
54 0737 0.95 0.16 239.69 0.5 "Loma Prieta" 1989 "Agnews State Hospital" 6.93 Oblique 24.27 7
"Anderson Dam Reverse 20.2
55 0739 0.91 0.16 488.77 0.8 "Loma Prieta" 1989 (Downstream)" 6.93 Oblique 19.9 6
Reverse 10.7
56 0741 2.30 1.23 476.54 5.4 "Loma Prieta" 1989 "BRAN" 6.93 Oblique 3.85 2
Reverse 10.7
57 0741 0.85 0.46 476.54 5.4 "Loma Prieta" 1989 "BRAN" 6.93 Oblique 3.85 2
"Bear Valley #14 Upper Reverse 71.3
58 0745 0.90 0.05 422.79 0.2 "Loma Prieta" 1989 Butts Rn" 6.93 Oblique 71.28 9
"Bear Valley #7 Reverse 69.3
59 0747 0.81 0.03 509.87 0 "Loma Prieta" 1989 Pinnacles" 6.93 Oblique 68.22 8
Reverse 44.1
60 0748 0.99 0.14 627.59 0.2 "Loma Prieta" 1989 "Belmont - Envirotech" 6.93 Oblique 43.94 1
Reverse
61 0753 2.47 1.24 462.24 3.2 "Loma Prieta" 1989 "Corralitos" 6.93 Oblique 0.16 3.85
Reverse
62 0753 1.20 0.61 462.24 3.2 "Loma Prieta" 1989 "Corralitos" 6.93 Oblique 0.16 3.85
Reverse 10.9
63 0764 1.06 0.39 308.55 0.7 "Loma Prieta" 1989 "Gilroy - Historic Bldg." 6.93 Oblique 10.27 7
Reverse 12.8
64 0767 2.11 0.67 349.85 2.1 "Loma Prieta" 1989 "Gilroy Array #3" 6.93 Oblique 12.23 2
Reverse 12.8
65 0767 1.09 0.35 349.85 2.1 "Loma Prieta" 1989 "Gilroy Array #3" 6.93 Oblique 12.23 2
"Hollister - South and Reverse 27.9
66 0776 1.77 1.26 282.14 2.2 "Loma Prieta" 1989 Pine" 6.93 Oblique 27.67 3
"Hollister - South and Reverse 27.9
67 0776 1.04 0.74 282.14 2.2 "Loma Prieta" 1989 Pine" 6.93 Oblique 27.67 3
Reverse
68 0779 1.58 1.19 594.83 7.9 "Loma Prieta" 1989 "LGPC" 6.93 Oblique 0 3.88
Reverse
69 0779 1.08 0.82 594.83 7.9 "Loma Prieta" 1989 "LGPC" 6.93 Oblique 0 3.88

192
 "Salinas - John and Reverse 32.7
70 0800 1.00 0.10 279.56 0.2 "Loma Prieta" 1989 Work" 6.93 Oblique 28.66 8
"Saratoga - W Valley Reverse
71 0803 2.27 1.37 347.90 1.3 "Loma Prieta" 1989 Coll." 6.93 Oblique 8.48 9.31
"Saratoga - W Valley Reverse
72 0803 0.84 0.51 347.90 1.3 "Loma Prieta" 1989 Coll." 6.93 Oblique 8.48 9.31
strike
73 0821 2.42 1.87 352.05 1.8 "Erzican Turkey" 1992 "Erzincan" 6.69 slip 0 4.38
strike
74 0821 1.08 0.83 352.05 1.8 "Erzican Turkey" 1992 "Erzincan" 6.69 slip 0 4.38
75 0825 2.29 1.39 567.78 6 "Cape Mendocino" 1992 "Cape Mendocino" 7.01 Reverse 0 6.96
76 0825 1.12 0.68 567.78 6 "Cape Mendocino" 1992 "Cape Mendocino" 7.01 Reverse 0 6.96
19.9
77 0827 0.95 0.17 457.06 0.3 "Cape Mendocino" 1992 "Fortuna - Fortuna Blvd" 7.01 Reverse 15.97 5
78 0828 2.36 1.93 422.17 3.8 "Cape Mendocino" 1992 "Petrolia" 7.01 Reverse 0 8.18
79 0828 1.05 0.86 422.17 3.8 "Cape Mendocino" 1992 "Petrolia" 7.01 Reverse 0 8.18
strike 68.6
80 0860 1.16 0.11 328.09 0.3 "Landers" 1992 "Hemet Fire Station" 7.28 slip 68.66 6
strike 26.9
81 0880 1.01 0.09 355.42 0.4 "Landers" 1992 "Mission Creek Fault" 7.28 slip 26.96 6
"Morongo Valley Fire strike 17.3
82 0881 0.94 0.20 396.41 1.2 "Landers" 1992 Station" 7.28 slip 17.36 6
strike 41.4
83 0897 1.04 0.03 635.01 0.1 "Landers" 1992 "Twentynine Palms" 7.28 slip 41.43 3
strike 23.6
84 0900 2.18 0.92 353.63 0.9 "Landers" 1992 "Yermo Fire Station" 7.28 slip 23.62 2
strike 23.6
85 0900 0.88 0.37 353.63 0.9 "Landers" 1992 "Yermo Fire Station" 7.28 slip 23.62 2
"Beverly Hills - 12520 18.3
86 0952 0.88 0.26 545.66 3 "Northridge-01" 1994 Mulhol" 6.69 Reverse 12.39 6
"Beverly Hills - 14145 17.1
87 0953 1.18 1.15 355.81 4.5 "Northridge-01" 1994 Mulhol" 6.69 Reverse 9.44 5
53.4
88 0966 1.00 0.08 324.79 0.1 "Northridge-01" 1994 "Covina - W Badillo" 6.69 Reverse 53.21 5
"Downey - Co Maint 46.7
89 0968 0.97 0.15 271.90 0.6 "Northridge-01" 1994 Bldg" 6.69 Reverse 43.2 4

193
 53.9
90 0975 0.91 0.09 362.31 0.1 "Northridge-01" 1994 "Glendora - N Oakbank" 6.69 Reverse 53.71 4
"Jensen Filter Plant
91 0982 1.74 2.48 373.07 5.3 "Northridge-01" 1994 Administrative Building" 6.69 Reverse 0 5.43
"Jensen Filter Plant
92 0982 0.93 1.32 373.07 5.3 "Northridge-01" 1994 Administrative Building" 6.69 Reverse 0 5.43
"Jensen Filter Plant
93 0983 1.93 1.93 525.79 6.5 "Northridge-01" 1994 Generator Building" 6.69 Reverse 0 5.43
"Jensen Filter Plant
94 0983 1.07 1.07 525.79 6.5 "Northridge-01" 1994 Generator Building" 6.69 Reverse 0 5.43
41.1
95 0984 1.05 0.14 301.00 0.4 "Northridge-01" 1994 "LA - 116th St School" 6.69 Reverse 36.39 7
36.6
96 0990 0.98 0.15 365.22 1.1 "Northridge-01" 1994 "LA - City Terrace" 6.69 Reverse 35.03 2
26.7
97 0998 1.00 0.18 315.06 1.4 "Northridge-01" 1994 "LA - N Westmoreland" 6.69 Reverse 23.4 3
33.9
98 1001 0.98 0.19 285.28 0.7 "Northridge-01" 1994 "LA - S Grand Ave" 6.69 Reverse 29.52 9
"LA - Sepulveda VA
99 1004 1.66 1.42 380.06 7 "Northridge-01" 1994 Hospital" 6.69 Reverse 0 8.44
"LA - Sepulveda VA
100 1004 0.92 0.78 380.06 7 "Northridge-01" 1994 Hospital" 6.69 Reverse 0 8.44
22.4
101 1006 1.09 0.25 398.42 1.6 "Northridge-01" 1994 "LA - UCLA Grounds" 6.69 Reverse 13.8 9
102 1013 2.33 1.46 628.99 1.8 "Northridge-01" 1994 "LA Dam" 6.69 Reverse 0 5.92
103 1013 1.13 0.71 628.99 1.8 "Northridge-01" 1994 "LA Dam" 6.69 Reverse 0 5.92
"Mojave - Oak Creek
104 1037 0.96 0.03 422.73 0 "Northridge-01" 1994 Canyon" 6.69 Reverse 75.64 75.8
105 1044 1.71 1.71 269.14 5.7 "Northridge-01" 1994 "Newhall - Fire Sta" 6.69 Reverse 3.16 5.92
106 1044 0.91 0.91 269.14 5.7 "Northridge-01" 1994 "Newhall - Fire Sta" 6.69 Reverse 3.16 5.92
107 1052 0.97 0.50 508.08 1.8 "Northridge-01" 1994 "Pacoima Kagel Canyon" 6.69 Reverse 5.26 7.26
108 1054 2.12 2.47 325.67 3.1 "Northridge-01" 1994 "Pardee - SCE" 6.69 Reverse 5.54 7.46
109 1054 1.17 1.37 325.67 3.1 "Northridge-01" 1994 "Pardee - SCE" 6.69 Reverse 5.54 7.46
"Rancho Palos Verdes - 52.1
110 1061 1.14 0.07 580.03 0.1 "Northridge-01" 1994 Hawth" 6.69 Reverse 48.02 8

194
111 1063 1.86 2.72 282.25 7.5 "Northridge-01" 1994 "Rinaldi Receiving Sta" 6.69 Reverse 0 6.5
112 1063 0.91 1.33 282.25 7.5 "Northridge-01" 1994 "Rinaldi Receiving Sta" 6.69 Reverse 0 6.5
"Simi Valley - Katherine 13.4
113 1080 2.32 1.66 557.42 4.1 "Northridge-01" 1994 Rd" 6.69 Reverse 0 2
"Simi Valley - Katherine 13.4
114 1080 1.06 0.76 557.42 4.1 "Northridge-01" 1994 Rd" 6.69 Reverse 0 2
115 1084 1.62 2.24 251.24 6 "Northridge-01" 1994 "Sylmar - Converter Sta" 6.69 Reverse 0 5.35
116 1084 1.01 1.40 251.24 6 "Northridge-01" 1994 "Sylmar - Converter Sta" 6.69 Reverse 0 5.35
"Sylmar - Olive View
117 1086 1.75 1.14 440.54 5 "Northridge-01" 1994 Med FF" 6.69 Reverse 1.74 5.3
"Sylmar - Olive View
118 1086 0.97 1.12 440.54 5 "Northridge-01" 1994 Med FF" 6.69 Reverse 1.74 5.3
"Wrightwood - Nielson 81.6
119 1097 0.98 0.03 506.00 0 "Northridge-01" 1994 Ranch" 6.69 Reverse 81.54 9
strike 11.3
120 1101 1.87 1.58 256.00 2 "Kobe Japan" 1995 "Amagasaki" 6.9 slip 11.34 4
strike 11.3
121 1101 0.97 0.82 256.00 2 "Kobe Japan" 1995 "Amagasaki" 6.9 slip 11.34 4
strike
122 1106 1.69 2.34 312.00 8.4 "Kobe Japan" 1995 "KJMA" 6.9 slip 0.94 0.96
strike
123 1106 1.16 1.61 312.00 8.4 "Kobe Japan" 1995 "KJMA" 6.9 slip 0.94 0.96
strike
124 1107 0.97 0.33 312.00 1.7 "Kobe Japan" 1995 "Kakogawa" 6.9 slip 22.5 22.5
strike 70.2
125 1109 0.91 0.03 609.00 0.1 "Kobe Japan" 1995 "MZH" 6.9 slip 69.04 6
strike
126 1111 2.31 0.66 609.00 3.4 "Kobe Japan" 1995 "Nishi-Akashi" 6.9 slip 7.08 7.08
strike
127 1111 1.06 0.30 609.00 3.4 "Kobe Japan" 1995 "Nishi-Akashi" 6.9 slip 7.08 7.08
strike
128 1114 2.31 2.15 198.00 1.8 "Kobe Japan" 1995 "Port Island (0 m)" 6.9 slip 3.31 3.31
strike
129 1114 1.13 1.05 198.00 1.8 "Kobe Japan" 1995 "Port Island (0 m)" 6.9 slip 3.31 3.31
strike 28.0
130 1115 1.02 0.18 256.00 0.6 "Kobe Japan" 1995 "Sakai" 6.9 slip 28.08 8

195
 strike 19.1
131 1116 1.02 0.27 256.00 0.8 "Kobe Japan" 1995 "Shin-Osaka" 6.9 slip 19.14 5
strike
132 1119 2.27 1.86 312.00 3.9 "Kobe Japan" 1995 "Takarazuka" 6.9 slip 0 0.27
strike
133 1119 1.11 0.91 312.00 3.9 "Kobe Japan" 1995 "Takarazuka" 6.9 slip 0 0.27
strike
134 1120 1.62 2.09 256.00 8.7 "Kobe Japan" 1995 "Takatori" 6.9 slip 1.46 1.47
strike
135 1120 1.02 1.31 256.00 8.7 "Kobe Japan" 1995 "Takatori" 6.9 slip 1.46 1.47
strike 27.7
136 1121 0.91 0.37 256.00 1.1 "Kobe Japan" 1995 "Yae" 6.9 slip 27.77 7
strike 65.5
137 1154 1.00 0.12 612.78 0.1 "Kocaeli Turkey" 1999 "Bursa Sivil" 7.51 slip 65.53 3
strike 15.3
138 1158 2.01 0.98 281.86 1.3 "Kocaeli Turkey" 1999 "Duzce" 7.51 slip 13.6 7
strike 15.3
139 1158 0.92 0.45 281.86 1.3 "Kocaeli Turkey" 1999 "Duzce" 7.51 slip 13.6 7
strike 31.7
140 1162 1.06 0.14 347.62 0.3 "Kocaeli Turkey" 1999 "Goynuk" 7.51 slip 31.74 4
strike 30.7
141 1166 0.94 0.21 476.62 0.4 "Kocaeli Turkey" 1999 "Iznik" 7.51 slip 30.73 3
strike
142 1176 2.33 0.90 297.00 1.3 "Kocaeli Turkey" 1999 "Yarimca" 7.51 slip 1.38 4.83
strike
143 1176 1.14 0.44 297.00 1.3 "Kocaeli Turkey" 1999 "Yarimca" 7.51 slip 1.38 4.83
Reverse
144 1197 1.49 1.51 542.61 5.9 "Chi-Chi Taiwan" 1999 "CHY028" 7.62 Oblique 3.12 3.12
Reverse
145 1197 1.02 1.04 542.61 5.9 "Chi-Chi Taiwan" 1999 "CHY028" 7.62 Oblique 3.12 3.12
Reverse
146 1231 1.11 2.34 496.21 9.3 "Chi-Chi Taiwan" 1999 "CHY080" 7.62 Oblique 0.11 2.69
Reverse 28.4
147 1234 0.92 0.21 665.20 1 "Chi-Chi Taiwan" 1999 "CHY086" 7.62 Oblique 27.57 2
Reverse
148 1244 2.35 1.73 258.89 3 "Chi-Chi Taiwan" 1999 "CHY101" 7.62 Oblique 9.94 9.94

196
 Reverse
149 1244 1.04 0.77 258.89 3 "Chi-Chi Taiwan" 1999 "CHY101" 7.62 Oblique 9.94 9.94
Reverse 47.7
150 1289 1.07 0.26 484.97 0.3 "Chi-Chi Taiwan" 1999 "HWA041" 7.62 Oblique 43.37 6
Reverse 16.7
151 1486 1.10 0.18 465.55 0.4 "Chi-Chi Taiwan" 1999 "TCU046" 7.62 Oblique 16.74 4
Reverse
152 1492 2.22 2.27 579.10 2.9 "Chi-Chi Taiwan" 1999 "TCU052" 7.62 Oblique 0 0.66
Reverse
153 1492 0.96 0.98 579.10 2.9 "Chi-Chi Taiwan" 1999 "TCU052" 7.62 Oblique 0 0.66
Reverse
154 1503 1.90 2.22 305.85 7.7 "Chi-Chi Taiwan" 1999 "TCU065" 7.62 Oblique 0.57 0.57
Reverse
155 1503 0.93 1.09 305.85 7.7 "Chi-Chi Taiwan" 1999 "TCU065" 7.62 Oblique 0.57 0.57
Reverse
156 1505 1.51 1.06 487.34 3.3 "Chi-Chi Taiwan" 1999 "TCU068" 7.62 Oblique 0 0.32
Reverse
157 1505 1.04 0.73 487.34 3.3 "Chi-Chi Taiwan" 1999 "TCU068" 7.62 Oblique 0 0.32
Reverse
158 1507 2.05 1.43 624.85 9.5 "Chi-Chi Taiwan" 1999 "TCU071" 7.62 Oblique 0 5.8
Reverse
159 1507 1.00 0.70 624.85 9.5 "Chi-Chi Taiwan" 1999 "TCU071" 7.62 Oblique 0 5.8
Reverse 13.4
160 1509 1.85 2.11 549.43 6.4 "Chi-Chi Taiwan" 1999 "TCU074" 7.62 Oblique 0 6
Reverse 13.4
161 1509 0.90 1.03 549.43 6.4 "Chi-Chi Taiwan" 1999 "TCU074" 7.62 Oblique 0 6
Reverse
162 1510 1.99 0.69 573.02 3 "Chi-Chi Taiwan" 1999 "TCU075" 7.62 Oblique 0.89 0.89
Reverse
163 1510 1.03 0.36 573.02 3 "Chi-Chi Taiwan" 1999 "TCU075" 7.62 Oblique 0.89 0.89
Reverse 10.9
164 1513 1.39 0.88 363.99 7.7 "Chi-Chi Taiwan" 1999 "TCU079" 7.62 Oblique 0 7
Reverse 10.9
165 1513 0.93 0.59 363.99 7.7 "Chi-Chi Taiwan" 1999 "TCU079" 7.62 Oblique 0 7
Reverse 11.4
166 1517 1.06 1.99 665.20 20.3 "Chi-Chi Taiwan" 1999 "TCU084" 7.62 Oblique 0 8

197
 Reverse
167 1549 2.42 1.37 511.18 9.3 "Chi-Chi Taiwan" 1999 "TCU129" 7.62 Oblique 1.83 1.83
Reverse
168 1549 0.95 0.54 511.18 9.3 "Chi-Chi Taiwan" 1999 "TCU129" 7.62 Oblique 1.83 1.83
Reverse
169 1551 1.04 0.45 652.85 1.7 "Chi-Chi Taiwan" 1999 "TCU138" 7.62 Oblique 9.78 9.78
strike 12.0
170 1602 2.23 2.16 293.57 3.7 "Duzce Turkey" 1999 "Bolu" 7.14 slip 12.02 4
strike 12.0
171 1602 1.09 1.05 293.57 3.7 "Duzce Turkey" 1999 "Bolu" 7.14 slip 12.02 4
strike
172 1605 2.36 1.51 281.86 2.9 "Duzce Turkey" 1999 "Duzce" 7.14 slip 0 6.58
strike
173 1605 1.15 0.74 281.86 2.9 "Duzce Turkey" 1999 "Duzce" 7.14 slip 0 6.58
strike 45.1
174 1620 1.12 0.02 411.91 0 "Duzce Turkey" 1999 "Sakarya" 7.14 slip 45.16 6
strike 34.6
175 1626 1.07 0.05 649.67 0.2 "Sitka Alaska" 1972 "Sitka Observatory" 7.68 slip 34.61 1
strike 50.8
176 1627 1.07 0.03 432.58 0.1 "Caldiran Turkey" 1976 "Maku" 7.21 slip 50.78 2
26.4
177 1628 0.97 0.27 306.37 0.9 "St Elias Alaska" 1979 "Icy Bay" 7.54 Reverse 26.46 6
strike 49.9
178 1636 1.08 0.13 302.64 0.4 "Manjil Iran" 1990 "Qazvin" 7.37 slip 49.97 7
"Banning - Twin Pines strike 83.4
179 1767 0.97 0.02 667.42 0 "Hector Mine" 1999 Road" 7.13 slip 83.43 3
strike 74.9
180 1782 1.03 0.08 436.14 0.1 "Hector Mine" 1999 "Forest Falls Post Office" 7.13 slip 74.92 2
strike 31.0
181 1794 0.92 0.28 379.32 0.6 "Hector Mine" 1999 "Joshua Tree" 7.13 slip 31.06 6
"Nenana Mountain "TAPS Pump Station strike 104.7 104.
182 2093 1.08 0.02 382.50 0 Alaska" 2002 #09" 6.7 slip 3 73
strike
183 2111 0.88 0.09 341.56 0.1 "Denali Alaska" 2002 "R109 (temp)" 7.9 slip 42.99 43
"TAPS Pump Station strike
184 2114 2.40 1.79 329.40 1.9 "Denali Alaska" 2002 #10" 7.9 slip 0.18 2.74

198
 "TAPS Pump Station strike
185 2114 1.17 0.87 329.40 1.9 "Denali Alaska" 2002 #10" 7.9 slip 0.18 2.74
"Taiwan 95.9
186 3583 1.22 0.07 309.41 0 SMART1(25)" 1983 "SMART1 I08" 6.5 Reverse 95.5 8
"Taiwan
187 3594 1.04 0.06 300.22 0 SMART1(25)" 1983 "SMART1 M11" 6.5 Reverse 96.52 97
12.2
188 3744 1.06 0.40 566.42 0.6 "Cape Mendocino" 1992 "Bunker Hill FAA" 7.01 Reverse 8.49 4
"Centerville Beach Naval 18.3
189 3746 2.23 0.97 459.04 1.6 "Cape Mendocino" 1992 Fac" 7.01 Reverse 16.44 1
"Centerville Beach Naval 18.3
190 3746 1.02 0.44 459.04 1.6 "Cape Mendocino" 1992 Fac" 7.01 Reverse 16.44 1
19.3
191 3748 2.48 1.63 387.95 1.7 "Cape Mendocino" 1992 "Ferndale Fire Station" 7.01 Reverse 16.64 2
19.3
192 3748 1.21 0.80 387.95 1.7 "Cape Mendocino" 1992 "Ferndale Fire Station" 7.01 Reverse 16.64 2
20.4
193 3749 2.06 0.68 355.18 1.3 "Cape Mendocino" 1992 "Fortuna Fire Station" 7.01 Reverse 16.54 1
20.4
194 3749 0.98 0.32 355.18 1.3 "Cape Mendocino" 1992 "Fortuna Fire Station" 7.01 Reverse 16.54 1
25.9
195 3750 2.08 0.51 515.65 0.9 "Cape Mendocino" 1992 "Loleta Fire Station" 7.01 Reverse 23.46 1
25.9
196 3750 0.83 0.20 515.65 0.9 "Cape Mendocino" 1992 "Loleta Fire Station" 7.01 Reverse 23.46 1
"Thousand Palms Post strike 36.9
197 3758 1.01 0.20 333.89 0.5 "Landers" 1992 Office" 7.28 slip 36.93 3
strike 82.4
198 3882 1.22 0.02 571.63 0 "Tottori Japan" 2000 "HRS016" 6.61 slip 82.42 2
strike 88.7
199 3899 1.00 0.01 617.44 0 "Tottori Japan" 2000 "HYGH02" 6.61 slip 88.75 5
strike 28.8
200 3908 1.07 0.13 293.37 0.8 "Tottori Japan" 2000 "OKY005" 6.61 slip 28.81 2
strike 66.2
201 3915 1.23 0.08 296.96 0.1 "Tottori Japan" 2000 "OKY012" 6.61 slip 66.24 5
strike 45.7
202 3937 1.09 0.11 182.30 0.2 "Tottori Japan" 2000 "SMN005" 6.61 slip 45.73 3

199
 strike 77.8
203 3945 0.86 0.02 262.19 0 "Tottori Japan" 2000 "SMN017" 6.61 slip 77.85 5
strike 85.3
204 3946 0.99 0.05 271.29 0.1 "Tottori Japan" 2000 "SMN018" 6.61 slip 85.31 1
strike
205 3968 1.84 2.58 310.21 11.8 "Tottori Japan" 2000 "TTRH02" 6.61 slip 0.83 0.97
strike
206 3968 1.02 1.43 310.21 11.8 "Tottori Japan" 2000 "TTRH02" 6.61 slip 0.83 0.97
"Coalinga - Fire Station 70.2
207 3981 0.86 0.05 333.61 0 "San Simeon CA" 2003 39" 6.52 Reverse 69.51 3
"Greenfield - Police
208 3987 0.87 0.03 280.64 0 "San Simeon CA" 2003 Station" 6.52 Reverse 69.08 69.8
"San Luis Obispo - 48.1
209 3994 1.05 0.10 365.15 0.2 "San Simeon CA" 2003 Lopez Lake Grounds" 6.52 Reverse 48.07 1
"Templeton - 1-story
210 4031 2.28 0.76 410.66 1.9 "San Simeon CA" 2003 Hospital" 6.52 Reverse 5.07 6.22
"Templeton - 1-story
211 4031 0.96 0.32 410.66 1.9 "San Simeon CA" 2003 Hospital" 6.52 Reverse 5.07 6.22
strike
212 4040 2.28 1.74 487.40 8 "Bam Iran" 2003 "Bam" 6.6 slip 0.05 1.7
strike
213 4040 0.99 0.75 487.40 8 "Bam Iran" 2003 "Bam" 6.6 slip 0.05 1.7
"Mohammad Abad-e- strike 46.2
214 4054 0.83 0.04 574.88 0.2 "Bam Iran" 2003 Madkoon" 6.6 slip 46.2 2
84.2
215 4198 0.98 0.02 220.65 0 "Niigata Japan" 2004 "NIG008" 6.63 Reverse 83.83 8
12.8
216 4207 0.98 0.33 274.17 3.4 "Niigata Japan" 2004 "NIG017" 6.63 Reverse 4.22 1
25.8
217 4208 0.91 0.14 198.26 0.8 "Niigata Japan" 2004 "NIG018" 6.63 Reverse 21.55 4
18.0
218 4212 1.10 0.13 193.20 0.8 "Niigata Japan" 2004 "NIG022" 6.63 Reverse 17.57 3
219 4218 0.96 0.32 430.71 5.2 "Niigata Japan" 2004 "NIG028" 6.63 Reverse 0.46 9.79
220 4219 2.25 1.72 480.40 8.8 "Niigata Japan" 2004 "NIGH01" 6.63 Reverse 0.49 9.46
221 4219 1.10 0.84 480.40 8.8 "Niigata Japan" 2004 "NIGH01" 6.63 Reverse 0.49 9.46

200
 71.5
222 4222 1.05 0.04 244.84 0.2 "Niigata Japan" 2004 "NIGH05" 6.63 Reverse 70.59 2
223 4228 2.42 0.96 375.00 2.2 "Niigata Japan" 2004 "NIGH11" 6.63 Reverse 6.27 8.93
224 4228 1.11 0.44 375.00 2.2 "Niigata Japan" 2004 "NIGH11" 6.63 Reverse 6.27 8.93
"Montenegro
225 4451 1.97 1.71 462.23 3 Yugoslavia" 1979 "Bar-Skupstina Opstine" 7.1 Reverse 0 6.98
"Montenegro
226 4451 1.23 1.07 462.23 3 Yugoslavia" 1979 "Bar-Skupstina Opstine" 7.1 Reverse 0 6.98
"Montenegro
227 4456 0.93 0.42 543.26 4.6 Yugoslavia" 1979 "Petrovac - Hotel Olivia" 7.1 Reverse 0 8.01
"Montenegro
228 4458 1.95 1.06 318.74 1.8 Yugoslavia" 1979 "Ulcinj - Hotel Olimpic" 7.1 Reverse 3.97 5.76
"Montenegro
229 4458 1.05 0.57 318.74 1.8 Yugoslavia" 1979 "Ulcinj - Hotel Olimpic" 7.1 Reverse 3.97 5.76
"Joetsu Uragawaraku 22.7
230 4842 0.96 0.17 655.45 1.4 "Chuetsu-oki Japan" 2007 Kamabucchi" 6.8 Reverse 18.6 4
"Tokamachi 28.7
231 4844 0.93 0.18 640.14 0.3 "Chuetsu-oki Japan" 2007 Matsunoyama" 6.8 Reverse 23.01 5
"Kubikiku Hyakken 22.1
232 4849 0.96 0.36 342.74 0.8 "Chuetsu-oki Japan" 2007 Joetsu City" 6.8 Reverse 20.71 8
"Kashiwazaki City 11.0
233 4856 2.17 1.80 294.38 3.9 "Chuetsu-oki Japan" 2007 Center" 6.8 Reverse 0 9
"Kashiwazaki City 11.0
234 4856 0.93 0.78 294.38 3.9 "Chuetsu-oki Japan" 2007 Center" 6.8 Reverse 0 9
"Mitsuke Kazuiti Arita 20.3
235 4859 0.95 0.37 274.23 0.8 "Chuetsu-oki Japan" 2007 Town" 6.8 Reverse 11.35 3
16.2
236 4863 2.00 1.35 514.30 2.2 "Chuetsu-oki Japan" 2007 "Nagaoka" 6.8 Reverse 3.97 7
16.2
237 4863 1.17 0.79 514.30 2.2 "Chuetsu-oki Japan" 2007 "Nagaoka" 6.8 Reverse 3.97 7
"Sawa Mizuguti
238 4872 1.04 0.27 640.14 0.3 "Chuetsu-oki Japan" 2007 Tokamachi" 6.8 Reverse 21.17 27.3
239 4874 2.42 1.28 561.59 5.3 "Chuetsu-oki Japan" 2007 "Oguni Nagaoka" 6.8 Reverse 10.31 20
240 4874 1.18 0.62 561.59 5.3 "Chuetsu-oki Japan" 2007 "Oguni Nagaoka" 6.8 Reverse 10.31 20
241 4875 1.08 0.89 282.57 6.4 "Chuetsu-oki Japan" 2007 "Kariwa" 6.8 Reverse 0 12

201
 "Kashiwazaki 12.6
242 4876 2.11 1.98 655.45 8.6 "Chuetsu-oki Japan" 2007 Nishiyamacho Ikeura" 6.8 Reverse 0 3
"Kashiwazaki 12.6
243 4876 1.03 0.96 655.45 8.6 "Chuetsu-oki Japan" 2007 Nishiyamacho Ikeura" 6.8 Reverse 0 3
"Yan Sakuramachi City 18.9
244 4879 1.09 0.57 265.82 0.7 "Chuetsu-oki Japan" 2007 watershed" 6.8 Reverse 12.98 7
"Tamati Yone 11.4
245 4886 2.22 1.19 338.32 4.9 "Chuetsu-oki Japan" 2007 Izumozaki" 6.8 Reverse 0 8
"Tamati Yone 11.4
246 4886 1.08 0.58 338.32 4.9 "Chuetsu-oki Japan" 2007 Izumozaki" 6.8 Reverse 0 8
"Kashiwazaki NPP Unit 10.9
247 4894 1.36 2.15 329.00 16.5 "Chuetsu-oki Japan" 2007 1: ground surface" 6.8 Reverse 0 7
"Kashiwazaki NPP Unit 10.9
248 4894 0.97 1.53 329.00 16.5 "Chuetsu-oki Japan" 2007 1: ground surface" 6.8 Reverse 0 7
"Kashiwazaki NPP Unit 10.9
249 4895 1.33 1.51 265.50 13.3 "Chuetsu-oki Japan" 2007 5: ground surface" 6.8 Reverse 0 7
"Kashiwazaki NPP Unit 10.9
250 4895 1.03 1.05 265.50 13.3 "Chuetsu-oki Japan" 2007 5: ground surface" 6.8 Reverse 0 7
"Kashiwazaki NPP
Service Hall Array 2.4 m 10.9
251 4896 0.93 0.91 201.00 5.1 "Chuetsu-oki Japan" 2007 depth" 6.8 Reverse 0 7
55.3
252 4997 1.00 0.09 305.54 0.1 "Chuetsu-oki Japan" 2007 "FKS028" 6.8 Reverse 52.63 8
95.0
253 5003 0.80 0.01 245.88 0 "Chuetsu-oki Japan" 2007 "FKSH04" 6.8 Reverse 93.48 5
87.9
254 5064 1.03 0.03 342.36 0 "Chuetsu-oki Japan" 2007 "GNM005" 6.8 Reverse 86.23 4
83.3
255 5254 0.96 0.02 220.65 0 "Chuetsu-oki Japan" 2007 "NIG008" 6.8 Reverse 81.51 1
67.7
256 5258 1.00 0.07 229.95 0.3 "Chuetsu-oki Japan" 2007 "NIG012" 6.8 Reverse 65.54 7
10.7
257 5264 1.77 1.66 198.26 5 "Chuetsu-oki Japan" 2007 "NIG018" 6.8 Reverse 0 8
10.7
258 5264 1.11 1.04 198.26 5 "Chuetsu-oki Japan" 2007 "NIG018" 6.8 Reverse 0 8

202
 112.7 113.
259 5461 0.89 0.02 279.36 0 "Iwate Japan" 2008 "AKT006" 6.9 Reverse 8 45
58.6
260 5467 0.98 0.02 449.45 0.1 "Iwate Japan" 2008 "AKT012" 6.9 Reverse 57.37 7
48.3
261 5471 1.08 0.09 158.16 0.3 "Iwate Japan" 2008 "AKT016" 6.9 Reverse 46.77 6
96.1
262 5490 1.14 0.01 232.58 0.1 "Iwate Japan" 2008 "AKTH14" 6.9 Reverse 95.32 1
49.9
263 5648 1.12 0.04 534.71 0.1 "Iwate Japan" 2008 "IWTH16" 6.9 Reverse 48.43 7
264 5656 2.34 0.78 486.41 3.5 "Iwate Japan" 2008 "IWTH24" 6.9 Reverse 3.1 5.18
265 5656 1.14 0.38 486.41 3.5 "Iwate Japan" 2008 "IWTH24" 6.9 Reverse 3.1 5.18
266 5657 1.85 1.40 506.44 26.1 "Iwate Japan" 2008 "IWTH25" 6.9 Reverse 0 4.8
267 5657 1.02 0.78 506.44 26.1 "Iwate Japan" 2008 "IWTH25" 6.9 Reverse 0 4.8
268 5658 2.37 1.06 371.06 14.1 "Iwate Japan" 2008 "IWTH26" 6.9 Reverse 5.97 6.02
269 5658 1.15 0.52 371.06 14.1 "Iwate Japan" 2008 "IWTH26" 6.9 Reverse 5.97 6.02
20.1
270 5663 2.38 0.96 479.37 9.4 "Iwate Japan" 2008 "MYG004" 6.9 Reverse 20.17 8
20.1
271 5663 1.03 0.41 479.37 9.4 "Iwate Japan" 2008 "MYG004" 6.9 Reverse 20.17 8
13.4
272 5664 2.38 1.07 361.24 4.2 "Iwate Japan" 2008 "MYG005" 6.9 Reverse 10.71 7
13.4
273 5664 1.16 0.52 361.24 4.2 "Iwate Japan" 2008 "MYG005" 6.9 Reverse 10.71 7
48.5
274 5768 0.99 0.03 291.48 0.1 "Iwate Japan" 2008 "YMTH09" 6.9 Reverse 47.01 9
29.3
275 5774 0.94 0.19 276.30 1 "Iwate Japan" 2008 "Nakashinden Town" 6.9 Reverse 29.37 8
20.7
276 5780 1.91 0.81 345.55 1.8 "Iwate Japan" 2008 "Iwadeyama" 6.9 Reverse 20.77 8
20.7
277 5780 0.91 0.39 345.55 1.8 "Iwate Japan" 2008 "Iwadeyama" 6.9 Reverse 20.77 8
"Misato Akita City - 41.7
278 5799 1.04 0.08 552.38 0.4 "Iwate Japan" 2008 Tsuchizaki" 6.9 Reverse 39.86 2

203
 12.8
279 5818 2.35 1.24 512.26 7.3 "Iwate Japan" 2008 "Kurihara City" 6.9 Reverse 12.83 5
12.8
280 5818 1.05 0.55 512.26 7.3 "Iwate Japan" 2008 "Kurihara City" 6.9 Reverse 12.83 5
"El Mayor-Cucapah "CERRO PRIETO strike 10.9
281 5825 2.34 0.91 242.05 3.3 Mexico" 2010 GEOTHERMAL" 7.2 slip 8.88 2
"El Mayor-Cucapah "CERRO PRIETO strike 10.9
282 5825 0.94 0.37 242.05 3.3 Mexico" 2010 GEOTHERMAL" 7.2 slip 8.88 2
"El Mayor-Cucapah "MICHOACAN DE strike 15.9
283 5827 2.35 1.38 242.05 6.1 Mexico" 2010 OCAMPO" 7.2 slip 13.21 1
"El Mayor-Cucapah "MICHOACAN DE strike 15.9
284 5827 1.15 0.67 242.05 6.1 Mexico" 2010 OCAMPO" 7.2 slip 13.21 1
"El Mayor-Cucapah "El Centro - Imperial and strike 20.0
285 5837 2.27 1.22 229.25 3.7 Mexico" 2010 Ross" 7.2 slip 19.39 8
"El Mayor-Cucapah "El Centro - Imperial and strike 20.0
286 5837 0.92 0.49 229.25 3.7 Mexico" 2010 Ross" 7.2 slip 19.39 8
"El Mayor-Cucapah strike
287 5839 1.01 0.02 388.01 0 Mexico" 2010 "El Cajon - Marshall" 7.2 slip 115 115
"El Mayor-Cucapah strike
288 5864 1.01 0.08 384.66 0 Mexico" 2010 "Frink" 7.2 slip 81.63 81.8
"El Mayor-Cucapah strike
289 5970 0.82 0.01 619.00 0 Mexico" 2010 "Borrego Springs" 7.2 slip 91.9 91.9
"El Mayor-Cucapah strike 41.4
290 5972 0.91 0.11 208.71 0.8 Mexico" 2010 "Brawley Airport" 7.2 slip 41.15 8
"El Mayor-Cucapah strike 20.4
291 5975 1.87 0.60 231.23 2.4 Mexico" 2010 "Calexico Fire Station" 7.2 slip 19.12 6
"El Mayor-Cucapah strike 20.4
292 5975 0.89 0.29 231.23 2.4 Mexico" 2010 "Calexico Fire Station" 7.2 slip 19.12 6
"El Mayor-Cucapah "El Centro Differential strike 23.4
293 5985 2.19 1.22 202.26 4.3 Mexico" 2010 Array" 7.2 slip 22.83 2
"El Mayor-Cucapah "El Centro Differential strike 23.4
294 5985 0.81 0.45 202.26 4.3 Mexico" 2010 Array" 7.2 slip 22.83 2
"El Mayor-Cucapah strike 20.0
295 5991 1.76 1.01 202.85 3.6 Mexico" 2010 "El Centro Array #10" 7.2 slip 19.36 5
"El Mayor-Cucapah strike 20.0
296 5991 1.10 0.63 202.85 3.6 Mexico" 2010 "El Centro Array #10" 7.2 slip 19.36 5

204
 "El Mayor-Cucapah strike 16.2
297 5992 2.50 1.51 196.25 5.5 Mexico" 2010 "El Centro Array #11" 7.2 slip 15.36 1
"El Mayor-Cucapah strike 16.2
298 5992 1.08 0.65 196.25 5.5 Mexico" 2010 "El Centro Array #11" 7.2 slip 15.36 1
111.3 111.
299 6515 0.95 0.02 279.58 0 "Niigata Japan" 2004 "FKS016" 6.63 Reverse 3 4
109.1 109.
300 6783 1.01 0.02 265.60 0 "Niigata Japan" 2004 "TCG008" 6.63 Reverse 4 21
"Darfield New strike 14.4
301 6886 1.00 0.16 280.26 0.9 Zealand" 2010 "Canterbury Aero Club" 7 slip 14.48 8
"Darfield New strike 11.8
302 6893 2.14 0.86 344.02 2.8 Zealand" 2010 "DFHS" 7 slip 11.86 6
"Darfield New strike 11.8
303 6893 1.11 0.44 344.02 2.8 Zealand" 2010 "DFHS" 7 slip 11.86 6
"Darfield New strike
304 6906 1.79 1.82 344.02 4.7 Zealand" 2010 "GDLC" 7 slip 1.22 1.22
"Darfield New strike
305 6906 1.12 1.14 344.02 4.7 Zealand" 2010 "GDLC" 7 slip 1.22 1.22
"Darfield New strike
306 6911 2.04 1.42 326.01 3.2 Zealand" 2010 "HORC" 7 slip 7.29 7.29
"Darfield New strike
307 6911 1.13 0.79 326.01 3.2 Zealand" 2010 "HORC" 7 slip 7.29 7.29
"Darfield New strike
308 6927 2.26 1.28 263.20 2.7 Zealand" 2010 "LINC" 7 slip 5.07 7.11
"Darfield New strike
309 6927 1.11 0.62 263.20 2.7 Zealand" 2010 "LINC" 7 slip 5.07 7.11
"Darfield New strike 25.6
310 6928 0.98 0.17 649.67 0.7 Zealand" 2010 "LPCC" 7 slip 25.21 7
"Darfield New strike 35.2
311 6933 1.09 0.05 342.70 0.1 Zealand" 2010 "MAYC" 7 slip 33.54 3
"Darfield New "Pages Road Pumping strike 24.5
312 6953 2.16 0.64 206.00 1.3 Zealand" 2010 Station" 7 slip 24.55 5
"Darfield New "Pages Road Pumping strike 24.5
313 6953 1.03 0.30 206.00 1.3 Zealand" 2010 Station" 7 slip 24.55 5
"Darfield New strike
314 6962 2.23 0.85 295.74 1.6 Zealand" 2010 "ROLC" 7 slip 0 1.54

205
 "Darfield New strike
315 6962 1.09 0.42 295.74 1.6 Zealand" 2010 "ROLC" 7 slip 0 1.54
"Darfield New strike 24.3
316 6965 0.95 0.12 263.20 0.7 Zealand" 2010 "SBRC" 7 slip 21.31 4
"El Mayor-Cucapah strike 11.2
317 8161 2.49 1.67 196.88 3.2 Mexico" 2010 "El Centro Array #12" 7.2 slip 9.98 6
"El Mayor-Cucapah strike 11.2
318 8161 1.22 0.81 196.88 3.2 Mexico" 2010 "El Centro Array #12" 7.2 slip 9.98 6
"El Mayor-Cucapah "SANTA ISABEL strike 57.4
319 8163 1.02 0.02 483.02 0 Mexico" 2010 VIEJO" 7.2 slip 55.19 9
strike
320 8166 1.01 0.19 425.00 - "Duzce Turkey" 1999 "IRIGM 498" 7.14 slip 3.58 3.58

206
 APPENDIX B. FRAGILITY CURVES IN TERMS OF PGA

Chapter 6 presented the approach and methodology for the generation of fragility

curves. The median and dispersion of the fragility curves in terms of peak ground

acceleration (PGA) is documented in the subsequent tables.

Table B1 – Fragility values in terms of PGA for two span continuous concrete box-
girder fragilities with diaphragm abutments.
Design Bridge class BSST-0 BSST-0 BSST-0 BSST-0 *
era        
S-E1-S22-C-D 0.13 0.59 0.30 0.61 0.63 0.64 0.91 0.64 0.62
S-E1-S22-R-D 0.16 0.56 0.36 0.58 0.70 0.61 0.99 0.61 0.59
T-E1-S22-C-D 0.07 0.79 0.18 0.83 0.42 0.93 0.66 0.94 0.87
Era 11
T-E1-S22-R-D 0.09 0.68 0.18 0.70 0.35 0.74 0.50 0.74 0.71
M-E1-S22-C-D 0.07 0.74 0.16 0.75 0.36 1.00 0.53 1.00 0.87
M-E1-S22-R-D 0.06 0.81 0.15 0.91 0.37 1.40 0.59 1.39 1.13
S-E2-S22-C-D 0.15 0.61 0.53 0.60 1.12 0.66 1.45 0.66 0.63
S-E2-S22-O-D 0.17 0.61 0.63 0.59 1.68 0.62 2.20 0.61 0.61
T-E2-S22-C-D 0.11 0.68 0.38 0.64 0.89 0.74 1.17 0.75 0.70
Era 22
T-E2-S22-O-D 0.16 0.54 0.43 0.53 0.91 0.70 1.12 0.70 0.61
M-E2-S22-C-D 0.09 0.70 0.30 0.68 0.60 0.86 0.77 0.86 0.78
M-E2-S22-O-D 0.09 0.70 0.30 0.68 0.60 0.86 0.77 0.86 0.78
S-E3-S22-C-D 0.15 0.62 0.53 0.60 1.44 0.66 2.04 0.66 0.63
S-E3-S22-O-D 0.17 0.60 0.63 0.58 2.20 0.62 3.20 0.63 0.61
T-E3-S22-C-D 0.11 0.68 0.38 0.64 1.16 0.75 1.66 0.76 0.71
Era 33
T-E3-S22-O-D 0.17 0.53 0.43 0.54 1.12 0.69 1.49 0.69 0.61
M-E3-S22-C-D 0.09 0.70 0.30 0.69 0.77 0.87 1.07 0.87 0.78
M-E3-S22-O-D 0.09 0.70 0.30 0.69 0.77 0.87 1.08 0.87 0.78

207
Table B2 – Fragility values in terms of PGA for two span continuous concrete box-
girder fragilities with seat abutments.
Design Bridge class BSST-0 BSST-0 BSST-0 BSST-0 *
era        
S-E1-S22-C-S 0.08 0.58 0.15 0.58 0.27 0.59 0.36 0.57 0.58
S-E1-S22-R-S 0.08 0.60 0.14 0.62 0.24 0.62 0.33 0.61 0.61
T-E1-S22-C-S 0.06 0.68 0.12 0.70 0.20 0.68 0.27 0.69 0.69
Era 11
T-E1-S22-R-S 0.09 0.71 0.14 0.71 0.23 0.69 0.31 0.69 0.70
M-E1-S22-C-S 0.05 0.64 0.09 0.64 0.19 0.62 0.27 0.60 0.62
M-E1-S22-R-S 0.08 0.65 0.13 0.65 0.23 0.64 0.31 0.63 0.64
S-E2-S22-C-S 0.10 0.64 0.47 0.64 0.89 0.73 1.20 0.73 0.69
S-E2-S22-O-S 0.12 0.55 0.49 0.53 0.96 0.60 1.29 0.60 0.57
T-E2-S22-C-S 0.08 0.63 0.33 0.60 0.64 0.67 0.87 0.68 0.64
Era 22
T-E2-S22-O-S 0.09 0.55 0.33 0.52 0.65 0.57 0.87 0.58 0.55
M-E2-S22-C-S 0.07 0.60 0.31 0.55 0.62 0.65 0.86 0.68 0.62
M-E2-S22-O-S 0.07 0.63 0.37 0.59 0.79 0.80 1.12 0.82 0.71
S-E3-S22-C-S 0.10 0.66 0.46 0.63 1.10 0.71 1.63 0.72 0.68
S-E3-S22-O-S 0.12 0.55 0.49 0.53 1.11 0.57 1.61 0.57 0.56
T-E3-S22-C-S 0.08 0.62 0.33 0.60 0.77 0.65 1.13 0.64 0.63
Era 33
T-E3-S22-O-S 0.09 0.54 0.33 0.52 0.72 0.56 1.01 0.56 0.54
M-E3-S22-C-S 0.07 0.58 0.32 0.57 0.75 0.63 1.12 0.63 0.60
M-E3-S22-O-S 0.07 0.65 0.37 0.60 0.94 0.76 1.46 0.75 0.69

Table B3 – Fragility values in terms of PGA for multi-span (S34) continuous

concrete box-girder fragilities with diaphragm abutments.
Design Bridge class BSST-0 BSST-0 BSST-0 BSST-0 *
era        
S-E1-S34-C-D 0.13 0.51 0.30 0.55 0.68 0.61 1.03 0.61 0.57
S-E1-S34-R-D 0.07 0.78 0.23 0.89 0.82 1.24 1.59 1.24 1.03
T-E1-S34-C-D 0.08 0.64 0.16 0.67 0.32 0.71 0.46 0.71 0.68
Era 11
T-E1-S34-R-D 0.07 0.68 0.17 0.73 0.39 0.83 0.59 0.83 0.77
M-E1-S34-C-D 0.01 1.09 0.04 0.98 0.17 1.05 0.35 1.05 1.04
M-E1-S34-R-D 0.01 1.09 0.04 0.98 0.17 1.05 0.35 1.05 1.04
S-E2-S34-C-D 0.15 0.55 0.54 0.51 1.17 0.69 1.55 0.69 0.61
S-E2-S34-O-D 0.19 0.49 0.69 0.47 1.71 0.64 2.29 0.64 0.56
T-E2-S34-C-D 0.08 0.68 0.35 0.63 0.72 0.96 0.98 0.96 0.81
Era 22
T-E2-S34-O-D 0.12 0.62 0.45 0.58 0.95 0.87 1.27 0.88 0.74
M-E2-S34-C-D 0.08 0.64 0.32 0.60 0.59 0.86 0.80 0.86 0.74
M-E2-S34-O-D 0.02 0.79 0.42 0.61 2.00 1.82 4.13 1.81 1.26
S-E3-S34-C-D 0.15 0.54 0.54 0.51 1.54 0.70 2.28 0.69 0.61
S-E3-S34-O-D 0.19 0.50 0.69 0.47 2.29 0.65 3.44 0.67 0.57
T-E3-S34-C-D 0.09 0.69 0.35 0.63 0.98 0.95 1.50 0.95 0.81
Era 33
T-E3-S34-O-D 0.12 0.61 0.45 0.58 1.27 0.88 1.91 0.87 0.73
M-E3-S34-C-D 0.08 0.65 0.32 0.61 0.80 0.88 1.21 0.86 0.75
M-E3-S34-O-D 0.02 0.78 0.42 0.61 4.07 1.79 9.10 1.97 1.29

208
 Table B4 – Fragility values in terms of PGA for multi-span (S34) continuous
concrete box-girder fragilities with seat abutments.
Design Bridge class BSST-0 BSST-0 BSST-0 BSST-0 *
era        
S-E1-S34-C-S 0.10 0.53 0.19 0.52 0.34 0.52 0.47 0.51 0.52
S-E1-S34-R-S 0.04 0.73 0.11 0.68 0.25 0.63 0.38 0.62 0.66
T-E1-S34-C-S 0.05 0.74 0.10 0.70 0.19 0.69 0.27 0.67 0.70
Era 11
T-E1-S34-R-S 0.01 1.00 0.04 0.82 0.14 0.73 0.25 0.69 0.81
M-E1-S34-C-S 0.01 0.93 0.04 0.84 0.12 0.77 0.19 0.75 0.82
M-E1-S34-R-S 0.01 0.93 0.04 0.84 0.12 0.77 0.19 0.75 0.82
S-E2-S34-C-S 0.08 0.73 0.42 0.75 0.80 0.97 1.15 0.98 0.86
S-E2-S34-O-S 0.11 0.63 0.53 0.63 1.11 0.80 1.61 0.82 0.72
T-E2-S34-C-S 0.03 0.79 0.29 0.68 0.63 1.02 1.01 1.09 0.89
Era 22
T-E2-S34-O-S 0.07 0.68 0.42 0.64 0.90 0.82 1.31 0.79 0.73
M-E2-S34-C-S 0.06 0.75 0.26 0.70 0.46 0.87 0.64 0.89 0.80
M-E2-S34-O-S 0.01 1.56 0.13 0.90 0.48 1.19 1.01 1.18 1.21
S-E3-S34-C-S 0.08 0.72 0.43 0.74 1.10 0.91 1.76 0.95 0.83
S-E3-S34-O-S 0.10 0.62 0.53 0.63 1.34 0.73 2.04 0.71 0.67
T-E3-S34-C-S 0.03 0.78 0.29 0.68 0.85 0.93 1.44 0.89 0.82
Era 33
T-E3-S34-O-S 0.07 0.69 0.41 0.64 1.06 0.69 1.58 0.67 0.67
M-E3-S34-C-S 0.06 0.74 0.26 0.70 0.61 0.85 0.93 0.85 0.78
M-E3-S34-O-S 0.01 1.50 0.13 0.90 0.80 1.03 1.42 0.83 1.06

209
APPENDIX C. COMPONENT FRAGILITY CURVES FOR BRIDGE

CLASSES

This appendix presents the component level fragility relationships for bridge

classes mentioned in Chapter 6. Table C1 and C2 documents the median and deviation

(logarithmic standard deviation) for the components for four damage states, for

diaphragam and seat abutment bridges, respectively. When the component median value

is more than 100, the corresponding median and dispersion values are reported as 99.00

and 0.00, respectively, to indicate that the contribution of the component to the system

vulnerability is negligible. Note the IM of ground motion is Sa1.0s.

Table C1 – Component level fragility relationships for diaphragm abutment

bridges.

CDT-0 CDT-1 CDT-2 CDT-3

Bridge class        
S-E1-S22-C-D
Column 0.13 0.62 0.31 0.62 0.69 0.62 1.05 0.62
Deck-max 0.65 0.59 1.94 0.59 16.26 0.59 24.42 0.59
Fnd-tran 1.61 0.96 11.04 0.96
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.04 0.73 3.47 0.73
Ab-Act 0.50 0.73 1.38 0.73
Ab-tran 0.18 0.62 0.71 0.62
S-E1-S22-R-D
Column 0.21 0.66 0.41 0.66 0.82 0.66 1.17 0.66
Deck-max 0.73 0.59 2.02 0.59 14.49 0.59 21.12 0.59
Fnd-tran 1.05 0.98 7.59 0.98
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.91 0.71 2.65 0.71
Ab-Act 0.47 0.70 1.13 0.70

210
 Ab-tran 0.22 0.63 0.82 0.63
T-E1-S22-C-D
Column 0.08 0.71 0.20 0.71 0.46 0.71 0.72 0.71
Deck-max 0.41 0.54 1.17 0.54 9.13 0.54 13.52 0.54
Fnd-tran 2.23 0.90 11.25 0.90
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.75 0.72 2.55 0.72
Ab-Act 0.35 0.73 0.98 0.73
Ab-tran 0.11 0.58 0.43 0.58
T-E1-S22-R-D
Column 0.11 0.75 0.21 0.75 0.43 0.75 0.62 0.75
Deck-max 0.52 0.62 1.40 0.62 9.56 0.62 13.80 0.62
Fnd-tran 2.98 1.09 23.60 1.09
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.83 0.65 2.27 0.65
Ab-Act 0.45 0.66 1.05 0.66
Ab-tran 0.15 0.63 0.52 0.63
M-E1-S22-C-D
Column 0.07 1.11 0.18 1.11 0.44 1.11 0.69 1.11
Deck-max 0.41 0.70 1.10 0.70 7.20 0.70 10.32 0.70
Fnd-tran 2.05 0.82 10.22 0.82
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.77 0.65 1.91 0.65
Ab-Act 0.43 0.65 0.93 0.65
Ab-tran 0.12 0.69 0.42 0.69
M-E1-S22-R-D
Column 0.06 1.36 0.15 1.36 0.41 1.36 0.69 1.36
Deck-max 0.55 0.60 1.54 0.60 11.35 0.60 16.63 0.60
Fnd-tran 1.75 1.08 10.24 1.08
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.99 0.90 3.31 0.90
Ab-Act 0.47 0.90 1.28 0.90
Ab-tran 0.15 0.65 0.58 0.65
S-E2-S22-C-D
Column 0.23 0.69 0.95 0.69 1.44 0.69 1.92 0.69
Deck-max 0.65 0.68 1.96 0.68 16.53 0.68 24.84 0.68
Fnd-tran 1.72 1.14 12.95 1.14
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.05 0.73 3.30 0.73

211
 Ab-Act 0.50 0.74 1.36 0.74
Ab-tran 0.17 0.71 0.69 0.71
S-E2-S22-O-D
Column 0.30 0.73 1.48 0.73 2.36 0.73 3.24 0.73
Deck-max 0.73 0.63 2.52 0.63
Fnd-tran 1.03 1.00 7.89 1.00
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.94 0.78 3.35 0.78
Ab-Act 0.42 0.78 1.23 0.78
Ab-tran 0.17 0.64 0.79 0.64
T-E2-S22-C-D
Column 0.16 0.66 0.64 0.66 0.97 0.66 1.27 0.66
Deck-max 0.43 0.63 1.18 0.63
Fnd-tran 1.36 0.97 8.66 0.97
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.80 0.66 2.36 0.66
Ab-Act 0.40 0.65 1.00 0.65
Ab-tran 0.12 0.64 0.44 0.64
T-E2-S22-O-D
Column 0.25 0.57 0.83 0.57 1.17 0.57 1.49 0.57
Deck-max 0.56 0.58 1.62 0.58
Fnd-tran 1.12 1.05 10.61 1.05
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.52 0.96 8.03 0.96
Ab-Act 0.53 0.93 2.10 0.93
Ab-tran 0.16 0.59 0.59 0.59
M-E2-S22-C-D
Column 0.13 0.74 0.47 0.74 0.67 0.74 0.86 0.74
Deck-max 0.36 0.58 0.97 0.58
Fnd-tran 1.21 1.10 8.04 1.10
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.58 0.69 1.64 0.69
Ab-Act 0.31 0.69 0.73 0.69
Ab-tran 0.11 0.60 0.37 0.60
M-E2-S22-O-D
Column 0.01 13.40 0.02 13.40 0.01 13.40 0.38 13.40
Deck-max 0.61 0.50 1.78 0.50
Fnd-tran 0.59 0.74 3.18 0.74
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.79 0.58 2.35 0.58

212
 Ab-Act 0.41 0.59 1.02 0.59
Ab-tran 0.16 0.50 0.63 0.50
S-E3-S22-C-D
Column 0.23 0.69 0.95 0.69 1.92 0.69 2.82 0.69
Deck-max 0.65 0.68 1.96 0.68
Fnd-tran 1.72 1.14 12.95 1.14
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.05 0.73 3.30 0.73
Ab-Act 0.50 0.74 1.36 0.74
Ab-tran 0.17 0.71 0.69 0.71
S-E3-S22-O-D
Column 0.30 0.73 1.48 0.73 3.24 0.73 4.99 0.73
Deck-max 0.73 0.63 2.52 0.63
Fnd-tran 1.03 1.00 7.89 1.00
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.94 0.78 3.35 0.78
Ab-Act 0.42 0.78 1.23 0.78
Ab-tran 0.17 0.64 0.79 0.64
T-E3-S22-C-D
Column 0.16 0.66 0.64 0.66 1.27 0.66 1.85 0.66
Deck-max 0.43 0.63 1.18 0.63
Fnd-tran 1.36 0.97 8.66 0.97
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.80 0.66 2.36 0.66
Ab-Act 0.40 0.65 1.00 0.65
Ab-tran 0.12 0.64 0.44 0.64
T-E3-S22-O-D
Column 0.25 0.57 0.83 0.57 1.49 0.57 2.06 0.57
Deck-max 0.56 0.58 1.62 0.58
Fnd-tran 1.12 1.05 10.61 1.05
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.52 0.96 8.03 0.96
Ab-Act 0.53 0.93 2.10 0.93
Ab-tran 0.16 0.59 0.59 0.59
M-E3-S22-C-D
Column 0.13 0.74 0.47 0.74 0.86 0.74 1.20 0.74
Deck-max 0.36 0.58 0.97 0.58
Fnd-tran 1.21 1.10 8.04 1.10
Fnd-rot 99.00 0.00 99.00 0.00

213
 Ab-Pass 0.58 0.69 1.64 0.69
Ab-Act 0.31 0.69 0.73 0.69
Ab-tran 0.11 0.60 0.37 0.60
M-E3-S22-O-D
Column 0.01 13.40 0.02 13.40 0.01 13.40 0.38 13.40
Deck-max 0.61 0.50 1.78 0.50
Fnd-tran 0.59 0.74 3.18 0.74
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.79 0.58 2.35 0.58
Ab-Act 0.41 0.59 1.02 0.59
Ab-tran 0.16 0.50 0.63 0.50
S-E1-S34-C-D
Column 0.14 0.70 0.34 0.70 0.84 0.70 1.34 0.70
Deck-max 0.76 0.57 2.37 0.57
Fnd-tran 1.36 0.86 8.49 0.86
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.05 0.73 3.66 0.73
Ab-Act 0.46 0.68 1.29 0.68
Ab-tran 0.20 0.57 0.80 0.57
S-E1-S34-R-D
Column 0.06 1.32 0.24 1.32 0.99 1.32 2.03 1.32
Deck-max 1.06 0.58 3.47 0.58
Fnd-tran 1.09 0.80 6.96 0.80
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.09 0.62 3.71 0.62
Ab-Act 0.51 0.63 1.45 0.63
Ab-tran 0.29 0.60 1.30 0.60
T-E1-S34-C-D
Column 0.06 0.89 0.15 0.89 0.36 0.89 0.56 0.89
Deck-max 0.44 0.57 1.31 0.57
Fnd-tran 1.72 0.90 8.84 0.90
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.71 0.56 2.15 0.56
Ab-Act 0.35 0.57 0.90 0.57
Ab-tran 0.11 0.59 0.46 0.59
T-E1-S34-R-D
Column 0.07 0.91 0.17 0.91 0.41 0.91 0.66 0.91
Deck-max 0.54 0.53 1.57 0.53
Fnd-tran 3.86 0.93 37.46 0.93

214
 Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.86 0.63 2.58 0.63
Ab-Act 0.43 0.63 1.08 0.63
Ab-tran 0.15 0.55 0.58 0.55
M-E1-S34-C-D
Column 0.01 1.13 0.04 1.13 0.19 1.13 0.40 1.13
Deck-max 0.54 0.46 1.42 0.46
Fnd-tran 3.68 1.05 34.68 1.05
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.81 0.60 2.52 0.60
Ab-Act 0.41 0.62 1.08 0.62
Ab-tran 0.17 0.46 0.60 0.46
M-E1-S34-R-D
Column 0.01 1.13 0.04 1.13 0.19 1.13 0.40 1.13
Deck-max 0.54 0.46 1.42 0.46
Fnd-tran 3.68 1.05 34.68 1.05
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.81 0.60 2.52 0.60
Ab-Act 0.41 0.62 1.08 0.62
Ab-tran 0.17 0.46 0.60 0.46
S-E2-S34-C-D
Column 0.23 0.69 0.95 0.69 1.44 0.69 1.92 0.69
Deck-max 0.65 0.68 1.96 0.68
Fnd-tran 1.72 1.14 12.95 1.14
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.05 0.73 3.30 0.73
Ab-Act 0.50 0.74 1.36 0.74
Ab-tran 0.17 0.71 0.69 0.71
S-E2-S34-O-D
Column 0.30 0.73 1.48 0.73 2.36 0.73 3.24 0.73
Deck-max 0.73 0.63 2.52 0.63
Fnd-tran 1.03 1.00 7.89 1.00
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.94 0.78 3.35 0.78
Ab-Act 0.42 0.78 1.23 0.78
Ab-tran 0.17 0.64 0.79 0.64
T-E2-S34-C-D
Column 0.16 0.66 0.64 0.66 0.97 0.66 1.27 0.66
Deck-max 0.43 0.63 1.18 0.63

215
 Fnd-tran 1.36 0.97 8.66 0.97
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.80 0.66 2.36 0.66
Ab-Act 0.40 0.65 1.00 0.65
Ab-tran 0.12 0.64 0.44 0.64
T-E2-S34-O-D
Column 0.25 0.57 0.83 0.57 1.17 0.57 1.49 0.57
Deck-max 0.56 0.58 1.62 0.58
Fnd-tran 1.12 1.05 10.61 1.05
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.52 0.96 8.03 0.96
Ab-Act 0.53 0.93 2.10 0.93
Ab-tran 0.16 0.59 0.59 0.59
M-E2-S34-C-D
Column 0.13 0.74 0.47 0.74 0.67 0.74 0.86 0.74
Deck-max 0.36 0.58 0.97 0.58
Fnd-tran 1.21 1.10 8.04 1.10
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.58 0.69 1.64 0.69
Ab-Act 0.31 0.69 0.73 0.69
Ab-tran 0.11 0.60 0.37 0.60
M-E2-S34-O-D
Column 0.13 0.74 0.47 0.74 0.67 0.74 0.86 0.74
Deck-max 0.36 0.58 0.97 0.58
Fnd-tran 1.21 1.10 8.04 1.10
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.58 0.69 1.64 0.69
Ab-Act 0.31 0.69 0.73 0.69
Ab-tran 0.11 0.60 0.37 0.60
S-E3-S34-C-D
Column 0.23 0.69 0.95 0.69 1.92 0.69 2.82 0.69
Deck-max 0.65 0.68 1.96 0.68
Fnd-tran 1.72 1.14 12.95 1.14
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.05 0.73 3.30 0.73
Ab-Act 0.50 0.74 1.36 0.74
Ab-tran 0.17 0.71 0.69 0.71
S-E3-S34-O-D
Column 0.30 0.73 1.48 0.73 3.24 0.73 4.99 0.73

216
 Deck-max 0.73 0.63 2.52 0.63
Fnd-tran 1.03 1.00 7.89 1.00
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.94 0.78 3.35 0.78
Ab-Act 0.42 0.78 1.23 0.78
Ab-tran 0.17 0.64 0.79 0.64
T-E3-S34-C-D
Column 0.16 0.66 0.64 0.66 1.27 0.66 1.85 0.66
Deck-max 0.43 0.63 1.18 0.63
Fnd-tran 1.36 0.97 8.66 0.97
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.80 0.66 2.36 0.66
Ab-Act 0.40 0.65 1.00 0.65
Ab-tran 0.12 0.64 0.44 0.64
T-E3-S34-O-D
Column 0.25 0.57 0.83 0.57 1.49 0.57 2.06 0.57
Deck-max 0.56 0.58 1.62 0.58
Fnd-tran 1.12 1.05 10.61 1.05
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.52 0.96 8.03 0.96
Ab-Act 0.53 0.93 2.10 0.93
Ab-tran 0.16 0.59 0.59 0.59
M-E3-S34-C-D
Column 0.13 0.74 0.47 0.74 0.86 0.74 1.20 0.74
Deck-max 0.36 0.58 0.97 0.58
Fnd-tran 1.21 1.10 8.04 1.10
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.58 0.69 1.64 0.69
Ab-Act 0.31 0.69 0.73 0.69
Ab-tran 0.11 0.60 0.37 0.60
M-E3-S34-O-D
Column 0.13 0.74 0.47 0.74 0.86 0.74 1.20 0.74
Deck-max 0.36 0.58 0.97 0.58
Fnd-tran 1.21 1.10 8.04 1.10
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.58 0.69 1.64 0.69
Ab-Act 0.31 0.69 0.73 0.69
Ab-tran 0.11 0.60 0.37 0.60

217
Table C2 – Component level fragility relationships for seat abutment bridges.

CDT-0 CDT-1 CDT-2 CDT-3

Bridge class
       
S-E1-S22-C-S
Column 0.10 0.67 0.23 0.67 0.54 0.67 0.83 0.67
Hinge 0.09 0.59 0.16 0.59 0.30 0.59 0.43 0.59
Deck-max 0.53 0.57 1.58 0.57
Fnd-tran 1.53 1.07 12.21 1.07
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.39 0.64 4.31 0.64
Ab-Act 0.68 0.69 1.69 0.69
Ab-tran 0.31 0.74 1.38 0.74
Bearing 0.16 0.59 0.55 0.59
Seal 0.30 0.59 7.36 0.59
Key 7.75 1.32 59.34 1.32
S-E1-S22-R-S
Column 0.12 0.61 0.24 0.61 0.51 0.61 0.74 0.61
Hinge 0.09 0.53 0.15 0.53 0.28 0.53 0.39 0.53
Deck-max 0.49 0.51 1.33 0.51
Fnd-tran 0.97 1.02 7.44 1.02
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.35 0.73 3.94 0.73
Ab-Act 0.69 0.74 1.62 0.74
Ab-tran 0.38 0.68 1.50 0.68
Bearing 0.15 0.53 0.49 0.53
Seal 0.28 0.53 5.79 0.53
Key 5.41 1.19 32.43 1.19
T-E1-S22-C-S
Column 0.09 0.65 0.19 0.65 0.41 0.65 0.61 0.65
Hinge 0.07 0.60 0.12 0.60 0.21 0.60 0.30 0.60
Deck-max 0.33 0.55 0.87 0.55
Fnd-tran 3.40 1.08 33.47 1.08
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.32 0.68 4.57 0.68
Ab-Act 0.62 0.70 1.64 0.70
Ab-tran 0.26 0.64 0.97 0.64
Bearing 0.11 0.58 0.35 0.58
Seal 0.21 0.60 4.49 0.60

218
 Key 7.24 1.04 55.39 1.04
T-E1-S22-R-S
Column 0.12 0.69 0.23 0.69 0.45 0.69 0.63 0.69
Hinge 0.09 0.56 0.15 0.56 0.26 0.56 0.36 0.56
Deck-max 0.41 0.55 1.10 0.55
Fnd-tran 0.95 1.02 7.62 1.02
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.25 0.74 3.96 0.74
Ab-Act 0.57 0.80 1.38 0.80
Ab-tran 0.29 0.68 1.12 0.68
Bearing 0.14 0.56 0.43 0.56
Seal 0.26 0.56 4.50 0.56
Key 5.67 1.41 27.83 1.41
M-E1-S22-C-S
Column 0.04 1.19 0.11 1.19 0.29 1.19 0.48 1.19
Hinge 0.07 0.50 0.12 0.50 0.22 0.50 0.31 0.50
Deck-max 0.35 0.51 0.97 0.51
Fnd-tran 3.30 1.13 28.69 1.13
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.17 0.72 3.54 0.72
Ab-Act 0.60 0.71 1.48 0.71
Ab-tran 0.25 0.80 1.12 0.80
Bearing 0.11 0.48 0.36 0.48
Seal 0.22 0.50 4.83 0.50
Key 6.99 1.38 52.84 1.38
M-E1-S22-R-S
Column 0.13 0.71 0.25 0.71 0.49 0.71 0.70 0.71
Hinge 0.09 0.58 0.16 0.58 0.28 0.58 0.39 0.58
Deck-max 0.46 0.55 1.25 0.55
Fnd-tran 0.90 0.99 7.25 0.99
Fnd-rot 53.72 0.49 99.00 0.00
Ab-Pass 1.18 0.71 3.33 0.71
Ab-Act 0.62 0.77 1.41 0.77
Ab-tran 0.34 0.70 1.20 0.70
Bearing 0.16 0.58 0.47 0.58
Seal 0.28 0.58 4.89 0.58
Key 6.36 1.29 35.05 1.29
S-E2-S22-C-S
Column 0.16 0.66 0.66 0.66 1.01 0.66 1.34 0.66

219
 Hinge 0.16 0.63 0.82 0.63 1.97 0.63 3.06 0.63
Deck-max 0.52 0.56 1.69 0.56
Fnd-tran 1.16 1.12 10.55 1.12
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.31 0.62 3.19 0.62
Ab-Act 0.73 0.64 1.48 0.64
Ab-tran 0.38 0.52 0.98 0.52
Bearing 0.13 0.59 0.76 0.59
Seal 0.34 0.63 17.71 0.63
Key 99.00 0.00 99.00 0.00
S-E2-S22-O-S
Column 0.20 0.60 0.86 0.60 1.32 0.60 1.77 0.60
Hinge 0.15 0.54 0.66 0.54 1.44 0.54 2.14 0.54
Deck-max 0.50 0.52 1.54 0.52
Fnd-tran 0.75 0.98 5.06 0.98
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.24 0.59 3.17 0.59
Ab-Act 0.66 0.60 1.35 0.60
Ab-tran 0.39 0.53 1.09 0.53
Bearing 0.14 0.54 0.70 0.54
Seal 0.30 0.54 10.35 0.54
Key 99.00 0.00 99.00 0.00
T-E2-S22-C-S
Column 0.11 0.59 0.53 0.59 0.84 0.59 1.14 0.59
Hinge 0.12 0.56 0.52 0.56 1.14 0.56 1.70 0.56
Deck-max 0.36 0.54 1.11 0.54
Fnd-tran 1.32 0.94 9.35 0.94
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.02 0.53 2.65 0.53
Ab-Act 0.55 0.60 1.15 0.60
Ab-tran 0.25 0.49 0.71 0.49
Bearing 0.10 0.54 0.50 0.54
Seal 0.23 0.56 8.27 0.56
Key 99.00 0.00 99.00 0.00
T-E2-S22-O-S
Column 0.14 0.65 0.61 0.65 0.92 0.65 1.23 0.65
Hinge 0.12 0.46 0.48 0.46 1.02 0.46 1.50 0.46
Deck-max 0.35 0.44 1.02 0.44
Fnd-tran 0.47 0.89 2.69 0.89
Fnd-rot 99.00 0.00 99.00 0.00

220
 Ab-Pass 1.00 0.51 2.46 0.51
Ab-Act 0.46 0.42 0.90 0.42
Ab-tran 0.29 0.46 0.80 0.46
Bearing 0.10 0.46 0.47 0.46
Seal 0.22 0.46 6.89 0.46
Key 99.00 0.00 99.00 0.00
M-E2-S22-C-S
Column 0.11 0.67 0.50 0.67 0.79 0.67 1.07 0.67
Hinge 0.11 0.53 0.50 0.53 1.16 0.53 1.76 0.53
Deck-max 0.35 0.47 1.07 0.47
Fnd-tran 0.97 0.90 6.15 0.90
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.94 0.42 2.20 0.42
Ab-Act 0.54 0.42 1.06 0.42
Ab-tran 0.27 0.41 0.67 0.41
Bearing 0.09 0.52 0.49 0.52
Seal 0.22 0.53 9.46 0.53
Key 99.00 0.00 99.00 0.00
M-E2-S22-O-S
Column 0.14 0.75 0.66 0.75 1.04 0.75 1.41 0.75
Hinge 0.12 0.54 0.58 0.54 1.35 0.54 2.07 0.54
Deck-max 0.40 0.50 1.27 0.50
Fnd-tran 0.49 0.87 3.89 0.87
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.10 0.52 2.74 0.52
Ab-Act 0.63 0.52 1.28 0.52
Ab-tran 0.32 0.42 0.81 0.42
Bearing 0.10 0.58 0.58 0.58
Seal 0.25 0.54 11.29 0.54
Key 99.00 0.00 99.00 0.00
S-E3-S22-C-S
Column 0.16 0.66 0.66 0.66 1.34 0.66 1.99 0.66
Hinge 0.16 0.63 0.82 0.63 1.97 0.63 3.06 0.63
Deck-max 0.52 0.56 1.69 0.56
Fnd-tran 1.16 1.12 10.55 1.12
Fnd-rot 99.00 1.15 99.00 1.15
Ab-Pass 1.31 0.62 3.19 0.62
Ab-Act 0.73 0.64 1.48 0.64
Ab-tran 0.38 0.52 0.98 0.52

221
 Bearing 0.13 0.59 0.76 0.59
Seal 0.34 0.63 17.71 0.63
Key 99.00 0.00 99.00 0.00
S-E3-S22-O-S
Column 0.20 0.60 0.86 0.60 1.77 0.60 2.64 0.60
Hinge 0.15 0.54 0.66 0.54 1.44 0.54 2.14 0.54
Deck-max 0.50 0.52 1.54 0.52
Fnd-tran 0.75 0.98 5.06 0.98
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.24 0.59 3.17 0.59
Ab-Act 0.66 0.60 1.35 0.60
Ab-tran 0.39 0.53 1.09 0.53
Bearing 0.14 0.54 0.70 0.54
Seal 0.30 0.54 10.35 0.54
Key 99.00 0.00 99.00 0.00
T-E3-S22-C-S
Column 0.11 0.59 0.53 0.59 1.14 0.59 1.75 0.59
Hinge 0.12 0.56 0.52 0.56 1.14 0.56 1.70 0.56
Deck-max 0.36 0.54 1.11 0.54
Fnd-tran 1.32 0.94 9.35 0.94
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.02 0.53 2.65 0.53
Ab-Act 0.55 0.60 1.15 0.60
Ab-tran 0.25 0.49 0.71 0.49
Bearing 0.10 0.54 0.50 0.54
Seal 0.23 0.56 8.27 0.56
Key 99.00 0.00 99.00 0.00
T-E3-S22-O-S
Column 0.14 0.65 0.61 0.65 1.23 0.65 1.82 0.65
Hinge 0.12 0.46 0.48 0.46 1.02 0.46 1.50 0.46
Deck-max 0.35 0.44 1.02 0.44
Fnd-tran 0.47 0.89 2.69 0.89
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.00 0.51 2.46 0.51
Ab-Act 0.46 0.42 0.90 0.42
Ab-tran 0.29 0.46 0.80 0.46
Bearing 0.10 0.46 0.47 0.46
Seal 0.22 0.46 6.89 0.46
Key 99.00 0.00 99.00 0.00

222
M-E3-S22-C-S
Column 0.11 0.67 0.50 0.67 1.07 0.67 1.63 0.67
Hinge 0.11 0.53 0.50 0.53 1.16 0.53 1.76 0.53
Deck-max 0.35 0.47 1.07 0.47
Fnd-tran 0.97 0.90 6.15 0.90
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 0.94 0.42 2.20 0.42
Ab-Act 0.54 0.42 1.06 0.42
Ab-tran 0.27 0.41 0.67 0.41
Bearing 0.09 0.52 0.49 0.52
Seal 0.22 0.53 9.46 0.53
Key 99.00 0.00 99.00 0.00
M-E3-S22-O-S
Column 0.14 0.75 0.66 0.75 1.41 0.75 2.15 0.75
Hinge 0.12 0.54 0.58 0.54 1.35 0.54 2.07 0.54
Deck-max 0.40 0.50 1.27 0.50
Fnd-tran 0.49 0.87 3.89 0.87
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.10 0.52 2.74 0.52
Ab-Act 0.63 0.52 1.28 0.52
Ab-tran 0.32 0.42 0.81 0.42
Bearing 0.10 0.58 0.58 0.58
Seal 0.25 0.54 11.29 0.54
Key 99.00 0.00 99.00 0.00
S-E1-S34-C-S
Column 0.12 0.58 0.28 0.58 0.63 0.58 0.96 0.58
Hinge 0.13 0.50 0.22 0.50 0.40 0.50 0.56 0.50
Deck-max 0.68 0.48 1.99 0.48
Fnd-tran 1.24 0.81 7.07 0.81
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.47 0.72 4.39 0.72
Ab-Act 0.74 0.73 1.78 0.73
Ab-tran 0.34 0.67 1.38 0.67
Bearing 0.22 0.50 0.71 0.50
Seal 0.40 0.50 8.12 0.50
Key 3.98 1.04 18.68 1.04
S-E1-S34-R-S
Column 0.06 1.07 0.17 1.07 0.50 1.07 0.88 1.07
Hinge 0.10 0.49 0.18 0.49 0.33 0.49 0.47 0.49

223
 Deck-max 0.61 0.51 1.82 0.51
Fnd-tran 1.59 0.95 12.69 0.95
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.37 0.67 3.86 0.67
Ab-Act 0.71 0.67 1.62 0.67
Ab-tran 0.54 0.88 2.80 0.88
Bearing 0.18 0.49 0.61 0.49
Seal 0.33 0.49 8.00 0.49
Key 5.70 1.14 32.67 1.14
T-E1-S34-C-S
Column 0.06 0.90 0.13 0.90 0.29 0.90 0.44 0.90
Hinge 0.09 0.51 0.16 0.51 0.27 0.51 0.37 0.51
Deck-max 0.39 0.58 1.01 0.58
Fnd-tran 2.93 0.87 18.56 0.87
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.06 0.64 3.31 0.64
Ab-Act 0.54 0.78 1.28 0.78
Ab-tran 0.24 0.78 0.95 0.78
Bearing 0.15 0.52 0.44 0.52
Seal 0.27 0.51 4.34 0.51
Key 3.12 0.99 12.03 0.99
T-E1-S34-R-S
Column 0.02 1.10 0.08 1.10 0.26 1.10 0.49 1.10
Hinge 0.08 0.56 0.14 0.56 0.27 0.56 0.39 0.56
Deck-max 0.45 0.54 1.33 0.54
Fnd-tran 5.04 1.29 86.51 1.29
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.09 0.60 3.18 0.60
Ab-Act 0.54 0.65 1.29 0.65
Ab-tran 0.29 0.70 1.14 0.70
Bearing 0.13 0.56 0.47 0.56
Seal 0.27 0.56 7.18 0.56
Key 24.57 1.96 99.00 1.96
M-E1-S34-C-S
Column 0.01 1.36 0.02 1.36 0.11 1.36 0.24 1.36
Hinge 0.11 0.58 0.18 0.58 0.32 0.58 0.44 0.58
Deck-max 0.45 0.45 1.23 0.45
Fnd-tran 3.54 1.30 28.44 1.30
Fnd-rot 99.00 0.00 99.00 0.00

224
 Ab-Pass 1.55 1.01 4.35 1.01
Ab-Act 0.75 0.99 1.67 0.99
Ab-tran 0.22 0.60 0.71 0.60
Bearing 0.16 0.54 0.50 0.54
Seal 0.32 0.58 5.58 0.58
Key 15.07 1.94 99.00 1.94
M-E1-S34-R-S
Column 0.01 1.36 0.02 1.36 0.11 1.36 0.24 1.36
Hinge 0.11 0.58 0.18 0.58 0.32 0.58 0.44 0.58
Deck-max 0.45 0.45 1.23 0.45
Fnd-tran 3.54 1.30 28.44 1.30
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.55 1.01 4.35 1.01
Ab-Act 0.75 0.99 1.67 0.99
Ab-tran 0.22 0.60 0.71 0.60
Bearing 0.16 0.54 0.50 0.54
Seal 0.32 0.58 5.58 0.58
Key 15.07 1.94 99.00 1.94
S-E2-S34-C-S
Column 0.11 0.82 0.59 0.82 0.97 0.82 1.36 0.82
Hinge 0.22 0.64 1.12 0.64 2.67 0.64 4.16 0.64
Deck-max 0.65 0.57 2.11 0.57
Fnd-tran 1.46 1.02 12.66 1.02
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.75 0.61 4.55 0.61
Ab-Act 0.99 0.63 2.11 0.63
Ab-tran 0.39 0.48 1.05 0.48
Bearing 0.17 0.61 0.96 0.61
Seal 0.46 0.64 23.96 0.64
Key 99.00 2.32 99.00 2.32
S-E2-S34-O-S
Column 0.15 1.05 1.05 1.05 1.85 1.05 2.70 1.05
Hinge 0.21 0.56 1.00 0.56 2.29 0.56 3.50 0.56
Deck-max 0.65 0.55 2.22 0.55
Fnd-tran 0.87 0.90 5.98 0.90
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.43 0.53 3.47 0.53
Ab-Act 0.81 0.54 1.64 0.54
Ab-tran 0.50 0.60 1.43 0.60

225
 Bearing 0.17 0.59 1.01 0.59
Seal 0.43 0.56 18.68 0.56
Key 99.00 0.00 99.00 0.00
T-E2-S34-C-S
Column 0.03 1.38 0.32 1.38 0.66 1.38 1.09 1.38
Hinge 0.14 0.54 0.63 0.54 1.42 0.54 2.14 0.54
Deck-max 0.40 0.49 1.24 0.49
Fnd-tran 1.55 0.92 11.68 0.92
Fnd-rot 99.00 0.67 99.00 0.67
Ab-Pass 1.08 0.55 2.71 0.55
Ab-Act 0.58 0.56 1.20 0.56
Ab-tran 0.29 0.48 0.83 0.48
Bearing 0.11 0.49 0.57 0.49
Seal 0.28 0.54 11.00 0.54
Key 99.00 0.00 99.00 0.00
T-E2-S34-O-S
Column 0.05 1.22 0.65 1.22 1.35 1.22 2.22 1.22
Hinge 0.14 0.54 0.69 0.54 1.59 0.54 2.43 0.54
Deck-max 0.45 0.55 1.48 0.55
Fnd-tran 0.65 0.77 3.91 0.77
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.32 0.69 3.29 0.69
Ab-Act 0.68 0.69 1.35 0.69
Ab-tran 0.34 0.50 0.87 0.50
Bearing 0.11 0.58 0.67 0.58
Seal 0.29 0.54 13.16 0.54
Key 99.00 0.00 99.00 0.00
M-E2-S34-C-S
Column 0.07 0.96 0.36 0.96 0.58 0.96 0.80 0.96
Hinge 0.15 0.55 0.69 0.55 1.53 0.55 2.31 0.55
Deck-max 0.44 0.52 1.40 0.52
Fnd-tran 1.11 0.81 7.42 0.81
Fnd-rot 99.00 0.70 99.00 0.70
Ab-Pass 1.12 0.55 2.87 0.55
Ab-Act 0.62 0.60 1.28 0.60
Ab-tran 0.30 0.50 0.82 0.50
Bearing 0.12 0.58 0.66 0.58
Seal 0.30 0.55 11.72 0.55
Key 99.00 0.00 99.00 0.00

226
M-E2-S34-O-S
Column 0.01 1.17 0.31 1.17 0.89 1.17 1.82 1.17
Hinge 0.15 0.55 0.66 0.55 1.44 0.55 2.14 0.55
Deck-max 0.52 0.56 1.63 0.56
Fnd-tran 0.89 0.95 8.41 0.95
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.08 0.54 2.59 0.54
Ab-Act 0.61 0.55 1.22 0.55
Ab-tran 0.40 0.50 1.11 0.50
Bearing 0.14 0.56 0.69 0.56
Seal 0.30 0.55 10.25 0.55
Key 99.00 0.00 99.00 0.00
S-E3-S34-C-S
Column 0.11 0.82 0.59 0.82 1.36 0.82 2.16 0.82
Hinge 0.22 0.64 1.12 0.64 2.67 0.64 4.16 0.64
Deck-max 0.65 0.57 2.11 0.57
Fnd-tran 1.46 1.02 12.66 1.02
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.75 0.61 4.55 0.61
Ab-Act 0.99 0.63 2.11 0.63
Ab-tran 0.39 0.48 1.05 0.48
Bearing 0.17 0.61 0.96 0.61
Seal 0.46 0.64 23.96 0.64
Key 99.00 0.00 99.00 0.00
S-E3-S34-O-S
Column 0.15 1.05 1.05 1.05 2.70 1.05 4.55 1.05
Hinge 0.21 0.56 1.00 0.56 2.29 0.56 3.50 0.56
Deck-max 0.65 0.55 2.22 0.55
Fnd-tran 0.87 0.90 5.98 0.90
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.43 0.53 3.47 0.53
Ab-Act 0.81 0.54 1.64 0.54
Ab-tran 0.50 0.60 1.43 0.60
Bearing 0.17 0.59 1.01 0.59
Seal 0.43 0.56 18.68 0.56
Key 99.00 0.00 99.00 0.00
T-E3-S34-C-S
Column 0.03 1.38 0.32 1.38 1.09 1.38 2.15 1.38

227
 Hinge 0.14 0.54 0.63 0.54 1.42 0.54 2.14 0.54
Deck-max 0.40 0.49 1.24 0.49
Fnd-tran 1.55 0.92 11.68 0.92
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.08 0.55 2.71 0.55
Ab-Act 0.58 0.56 1.20 0.56
Ab-tran 0.29 0.48 0.83 0.48
Bearing 0.11 0.49 0.57 0.49
Seal 0.28 0.54 11.00 0.54
Key 99.00 0.00 99.00 0.00
T-E3-S34-O-S
Column 0.05 1.22 0.65 1.22 2.22 1.22 4.35 1.22
Hinge 0.14 0.54 0.69 0.54 1.59 0.54 2.43 0.54
Deck-max 0.45 0.55 1.48 0.55
Fnd-tran 0.65 0.77 3.91 0.77
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.32 0.69 3.29 0.69
Ab-Act 0.68 0.69 1.35 0.69
Ab-tran 0.34 0.50 0.87 0.50
Bearing 0.11 0.58 0.67 0.58
Seal 0.29 0.54 13.16 0.54
Key 99.00 0.00 99.00 0.00
M-E3-S34-C-S
Column 0.07 0.96 0.36 0.96 0.80 0.96 1.24 0.96
Hinge 0.15 0.55 0.69 0.55 1.53 0.55 2.31 0.55
Deck-max 0.44 0.52 1.40 0.52
Fnd-tran 1.11 0.81 7.42 0.81
Fnd-rot 99.00 0.00 99.00 0.00
Ab-Pass 1.12 0.55 2.87 0.55
Ab-Act 0.62 0.60 1.28 0.60
Ab-tran 0.30 0.50 0.82 0.50
Bearing 0.12 0.58 0.66 0.58
Seal 0.30 0.55 11.72 0.55
Key 99.00 0.00 99.00 0.00
M-E3-S34-O-S
Column 0.01 1.17 0.31 1.17 1.82 1.17 4.85 1.17
Hinge 0.15 0.55 0.66 0.55 1.44 0.55 2.14 0.55
Deck-max 0.52 0.56 1.63 0.56
Fnd-tran 0.89 0.95 8.41 0.95

228
Fnd-rot 99.00 0.51 99.00 0.51
Ab-Pass 1.08 0.54 2.59 0.54
Ab-Act 0.61 0.55 1.22 0.55
Ab-tran 0.40 0.50 1.11 0.50
Bearing 0.14 0.56 0.69 0.56
Seal 0.30 0.55 10.25 0.55
Key 99.00 1.79 99.00 1.79

229
 REFERENCES
Ala Saadeghvaziri, M., and Rashidi, S. (1998). Effect of steel bearings on seismic response of
bridges in Eastern United States. Proceedings of 6th US National Conf. on Earthquake
Engineering. Oakland, California
ATC. (1985). Earthquake damage evaluation data for California, Report No. ATC-6. Retrieved
from Applied Technology Council, Redwood City, California.
Avsar, O., Yakut A and Caner, A. (2001). Analytical Fragility Curves for Ordinary Highway
Bridges in Turkey. Earthquake Spectra, 27(4), pp: 971-996.
Baker, J. W., Lin, T., Shahi, S. K., and Jayaram, N. (2011). New ground motion selection
procedures and selected motions for the PEER transportation research program. Pacific
Earthquake Engineering Research Center.
Banerjee, S., and Shinozuka, M. (2007). Nonlinear static procedure for seismic vulnerability
assessment of bridges. Computer‐Aided Civil and Infrastructure Engineering, 22(4),
pp:293-305.
Banerjee, S., and Shinozuka, M. (2008). Mechanistic quantification of RC bridge damage states
under earthquake through fragility analysis. Probabilistic Engineering Mechanics, 23,
pp:12-22.
Basöz, N., and Mander, J. (1999). Enhancement of the highway transportation lifeline module in
HAZUS. National Institute of Building Sciences, 16(1), pp:31-40.
Basoz, N., and Kiremidjian, A. S. (1998). Evaluation of bridge damage data from the Loma
Prieta and Northridge, California earthquakes. California earthquakes (No. MCEER-
98-0004).
BDS (1990). Bridge design details, Caifornia Department of Trnasportations, Sacramento. CA.
BDA (2012). Bridge design aids, Caifornia Department of Trnasportations, Sacramento. CA.
Başöz, N., and Kiremidjian, A. S. (1996). Risk assessment for highway transportation systems:
John A. Blume Earthquake Engineering Center.
Bavirisetty, R., Vinayagamoorthy, M., and Duan, L. (2003). Dynamic Analysis, Bridge
Engineering – Seismic Design, Edited by Wai-Fah Chen and Lian Duan, CRC Press LLC,
Boca Raton, FL, ISBN: 0-8493-1683-9/02.
Bradley, B. A. (2010). Epistemic Uncertainties in Component Fragility Functions. Earthquake
Spectra 26(1), pp: 41-62.
Calderone, A., Lehman, D. E., and Moehle J. P. (2001). Behavior of reinforced concrete bridge
columns having varying aspect ratios and varying lengths of confinement, Pacific
Earthquake Engineering Research Center.
Caltrans (2017). Personal communication with the P266, Task 1780 Fragility project panel
members including Roblee C, Shantz T, Turner L. California Department of
Transportation. Sacramento.
Caltrans, S. (2010). Caltrans Seismic Design Criteria version 1.6. California Department of
Transportation, Sacramento, California.
Celik, O. C., and B. R. Ellingwood (2010). Seismic fragilities for non-ductile reinforced concrete
frames - Role of aleatoric and epistemic uncertainties. Structural Safety 32(1), pp: 1-12.
Chai, Y. H., Priestley, M. N., and Seible, F. (1991). Seismic retrofit of circular bridge columns
for enhanced flexural performance. ACI Structural Journal, 88.

230
Chang, G. A. and Mander, J. B. (1994). Seismic energy based fatigue damage analysis of bridge
columns: Part 1: Evaluation of seismic capacity. In Technical Report, US National
Center for Earthquake Engineering Research.
Cheok, G. S. and Stone, W. C. (1990). Behavior of 1/6-scale model bridge columns subjected to
inelastic cyclic loading. ACI Structural Journal 87(6).
Choi, E. (2002). Seismic analysis and retrofit of mid-America bridges. Ph.D Thesis, Georgia
Institute of Technology, Atlanta, USA.
Coffman, H. L., Marsh, M. L., and Brown, C. B. (1993). Seismic durability of retrofitted
reinforced-concrete columns. Journal of Structural Engineering, 119(5), 1643-1661.
Cornell, C., Jalayer, F., Hamburger, R., and Foutch, D. (2002). Probabilistic Basis for 2000 SAC
Federal Emergency Management Agency Steel Moment Frame Guidelines. Journal of
Structural Engineering, 128(4), pp:526-533.
Cornell, C. A., and Krawinkler, H. (2000). Progress and challenges in seismic performance
assessment. PEER Center News, 3(2), 1-3.
Correal, J. F., Saiidi, M. S., Sanders, D., and El-Azazy, S. (2007). Analytical evaluation of bridge
columns with double interlocking spirals. ACI structural journal, 104(3), 314.
Dicleli, M., and Bruneau, M. (1995). Seismic performance of single-span simply supported and
continuous slab-on-girder steel highway bridges. Journal of structural engineering,
121(10), pp:1497-1506.
Dukes, J. D. (2013). Application of bridge specific fragility analysis in the seismic design process
of bridges in california. Ph.D Thesis, Georgia Institute of Technology, Atlanta, USA.
Dukes, J., Mangalathu, S., Padgett, J. E., and DesRoches, R. (2017). Development of bridge-
specific fragility methedology to improve the seismic resilience of bridges, Earthquakes
and Structures (in review).
Esmaeily, A., and Xiao, Y. (2005). Behavior of reinforced concrete columns under variable axial
loads: analysis. ACI Structural Journal, 102(5), 736.
Fang, J. Q., Li, Q. S., Jeary, A. P., and Liu, D. K. (1999). Damping of tall buildings: its evaluation
and probabilistic characteristics. The Structural Design of Tall Buildings, 8(2), 145-153.
Friedman, J., Hastie, T., and Tibshirani, R. (2001) The elements of statistical learning. Springer
series in statistics Springer, Berlin.
Filippou, F. C., Popov, E. P., and Bertero, B. V. (1983). Effects of bond deterioration on
hysteretic behavior of reinforced concrete joints, Report No. FEMA-351. Washington
DC: SAC Joint Venture.
Fajfar, P. (2000). A nonlinear analysis method for performance-based seismic design. Earthquake
Spectra, 16(3), 573-592.
Gardoni, P., Mosalam, K. M., and der Kiureghian, A. (2003). Probabilistic seismic demand
models and fragility estimates for RC bridges. Journal of Earthquake Engineering, 7,
pp:79-106.
Ghosh, J., Padgett, J. E., and Dueñas-Osorio, L. (2013). Surrogate modeling and failure surface
visualization for efficient seismic vulnerability assessment of highway bridges.
Probabilistic Engineering Mechanics, 34, pp:189-199.

231
Haselton, C., A. Whittaker, A. Hortacsu, J. Baker, J. Bray and Grant, D. (2012). Selecting and
scaling earthquake ground motions for performing response-history analyses.
Proceedings of the 15th World Conference on Earthquake Engineering.
HAZUS-MH. (2003). Multi-Hazard Loss Estimation Methodology: Earthquake Model.
Department of Homeland Security, FEMA, Washington, DC.
Hesterberg, T., Choi, N. H., Meier, L., Fraley, C., (2008). Least angle and ℓ1 penalized
regression: A review. Statistics Surveys, 2,pp:61-93.
Hoerl, A. E., and Kennar., R. W. (1970). Ridge regression: Biased estimation for nonorthogonal
problems, Technometrics, 12,pp:55-67.
Hose, Y. D., Priestley, M. and Seible, F. (1997). Strategic relocation of plastic hinges in bridge
columns. Structural Systems Research Project, 97/05, University of California, San
Diego.
Huitema, B. E. (1990). Analysis of Covariance, in Encyclopedia of Statistics in Behavioral
Science. John Wiley and Sons, Ltd: Hoboken, New Jersey.
Hwang, H., Jernigan, J. B., and Lin, Y.-W. (2000). Evaluation of seismic damage to Memphis
bridges and highway systems. Journal of Bridge Engineering, 5(4), pp:322-330.
Jangid, R. S. (2004). Seismic response of isolated bridges. Journal of Bridge Engineering, 9(2),
pp:156–166.
Jaradat, O. A., McLean, D. I. and Marsh, M. L. (1998). Performance of Existing Bridge Columns
under Cyclic Loading Part 1: Experimental Results and Observed Behavior. ACI
Structural Journal 95(6).
Jeon, J. S. (2013). Aftershock vulnerability assessment of damaged reinforced concrete buildings
in California, PhD. Thesis, Georgia Tech, Atlanta.
Jeon, J. S., Shafieezadeh, A., Lee, D. H., Choi, E., and DesRoches, R. (2015). Damage
assessment of older highway bridges subjected to three-dimensional ground motions:
characterization of shear–axial force interaction on seismic fragilities. Engineering
Structures, 87, pp: 47-57.
Jeon, J.-S., Mangalathu, S., Song, J., and DesRoches, R. (2017). Parameterized seismic fragility
curves for curved multi-frame concrete box-girder bridges using Bayesian parameter
estimation. Journal of Earthquake Engineering (In press).
Jeong, S.-H., and Elnashai, A. S. (2007). Probabilistic fragility analysis parameterized by
fundamental response quantities. Engineering Structures, 29(6), pp:1238-1251.
Ji, J., Elnashai, A. S., and Kuchma, D. A. (2007). An analytical framework for seismic fragility
analysis of RC high-rise buildings. Engineering structures, 29(12), pp:3197-3209.
Kameshwar, S. and Padgett, J. E. (2014) Multi-hazard risk assessment of highway bridges
subjected to earthquake and hurricane hazards. Engineering Structures, 78, pp:154-166.
Keselman, H., Huberty, C. J., Lix, L. M., Olejnik, S., Cribbie, R. A., Donahue, B., Kowalchuk,
R. K., Lowman, L. L., Petoskey, M. D., and Keselman, J. C. (1998). Statistical practices
of educational researchers: An analysis of their ANOVA, MANOVA, and ANCOVA
analyses. Review of Educational Research, 68, pp:350-386.
Kim, S.-H., and Shinozuka, M. (2004). Development of fragility curves of bridges retrofitted by
column jacketing. Probabilistic Engineering Mechanics, 19(1), pp:105-112.

232
Kircher, C. A., Whitman, R. V., and Holmes, W. T. (2006). HAZUS earthquake loss estimation
methods. Natural Hazards Review, 7(2), pp:45-59.
Kowalsky, M. J. and Priestley, M. N. (2000). Improved analytical model for shear strength of
circular reinforced concrete columns in seismic regions. ACI Structural Journal 97(3).
Kolmogorov, A. N. (1933). Sulla determinazione empirica di una legge di distribuzione.
Kruskal, W. H., and Wallis, W. A. (1952). Use of ranks in one-criterion variance analysis.
Journal of the American statistical Association, 47(260),pp:583-621.
Kunnath, S. K., El-Bahy, A.,Taylor, A. W. and Stone, W. C. (1997). Cumulative seismic damage
of reinforced concrete bridge piers. Technical Report NCEER, US National Center for
Earthquake Engineering Research.
Kunnath, S. K., Larson, L., and Miranda, E. (2006) Modelling considerations in probabilistic
performance‐based seismic evaluation: case study of the I‐880 viaduct. Earthquake
Engineering and Structural Dynamics, 35, pp:57-75.
Kvam, P. H., and Vidakovic, B. (2007). Nonparametric statistics with applications to science
and engineering. John Wiley and Sons..
Lehman, D., Moehle, J., Mahin, S., Calderone, A. and Henry. L. (2004). Experimental
evaluation of the seismic performance of reinforced concrete bridge columns. Journal of
Structural Engineering, 130, pp:869-879.
Lehman, D. E. and Moehle, J. P. (2000). Seismic performance of well-confined concrete bridge
columns, Pacific Earthquake Engineering Research Center.
Mackie, K., and Stojadinovic, B. (2001). Probabilistic seismic demand model for California
highway bridges. Journal of Bridge Engineering, 6(6), 468-481.
Mackie, K. R., and Stojadinović, B. (2005). Fragility basis for California highway overpass
bridge seismic decision making. Pacific Earthquake Engineering Research Center,
College of Engineering, University of California, Berkeley.
Mackie, K. R., Cronin, K. J., Nielson, B. G. (2011). Response Sensitivity of Highway Bridges to
Randomly Oriented Multi-Component Earthquake Excitation, Journal of Earthquake
Engineering, 15(6), pp: 850-876.
Mander, J., Priestley, M., and Park, R. (1988). Theoretical Stress‐Strain Model for Confined
Concrete. Journal of Structural Engineering, 114(8), pp:1804-1826.
Mander, J. B., and Basöz, N. (1999). Seismic fragility curve theory for highway bridges. Paper
presented at the Optimizing post-earthquake lifeline system reliability.
Mangalathu, S., Jeon, J.-S., Soleimani, F., DesRoches, R., Padgett, J., and Jiang, J. (2015a).
Seismic vulnerability of multi-span bridges: An analytical perspective. Proceedings of
the 10th Pacific Conference on Earthquake Engineering, Sydney, Australia.
Mangalathu, S., Jeon, J.-S., DesRoches, R., and Padgett, J. (2015b). Analysis of Covariance to
Capture the Importance of Bridge Attributes on the Probabilistic Seismic Demand
Model. Proceedings of the 10th Pacific Conference on Earthquake Engineering, Sydney,
Australia.
Mangalathu, S., Jeon, J.-S., DesRoches, R., and Padgett, J. (2016a). ANCOVA-based grouping
of bridge classes for seismic fragility assessment. Engineering Structures, 123, 379-394.

233
Mangalathu, S., Jeon J-S., DesRoches, R., and Padgett, J. E. (2016b). Application of Bayesian
Methods to Probabilistic Demand Analyses of Concrete Box-Girder Bridges.
Proceedings of the Geotechnical and Structural Engineering Congress, Phoenix, Arizona.
Mangalathu, S., Jeon, J.-S., DesRoches, R., and Padgett, J. (2017a). Performance-based grouping
methods of bridge classes for regional seismic risk assessment: Application of ANOVA,
ANCOVA, and non-parameteric approach. Earthquake Engineering and Structural
Dynamics (In review).
Mangalathu, S., Soleimani, F., DesRoches, R., and Padgett, J. E. (2017b). ANOVA based
grouping of bridge classes. Proceedings of the 16th World Conference on Earthquake
Engineering, Santiago, Chile.
Mangalathu, S., Jeon J-S., Soleimani, F., DesRoches, R., Padgett, J. E., and Jiang, J. (2017c).
Sensitivity of Fragility Curves to Parameter Uncertainty using Lasso. Proceedings of the
16th World Conference on Earthquake Engineering, Santiago, Chile.
Mangalathu, S., Soleimani, F., DesRoches, R., and Padgett, J. E. (2017d). Bridge Classes for
Regional Risk assessment of Box-Girder Bridges in California: Improving HAZUS
Model. Proceedings of the Structures Congress, Denver, Colorado.
Mangalathu, S., Jeon J-S., DesRoches, R., (2017e) Identification of critical parameters on the
seismic performance of concrete bridges using Lasso regression, Earthquake
Engineering and Structural Dynamics (In review).
Mazzoni, S, McKenna. F., Scott, M. H., Fenves, G. L. (2006) OpenSees command language
manual. Pacific Earthquake Engineering Research (PEER) Center.
Megally, S. H., Silva, F. P., and Seible, F. (2002). Seismic Response of Sacrificial Shear Keys in
Bridge Abutments. Department of Structural Engineering, University of California, San
Diego.
Mehr, M. and Zaghi, A. E. (2016). Seismic response of multi-frame bridges. Bulletin of
Earthquake Engineering. 14, pp:1219-43.
Moehle, J., and Deierlein, G. G. (2004). A framework methodology for performance-based
earthquake engineering. Proceedings of the 13th World Conference on Earthquake
Engineering.
Moschonas, I. F., Kappos, A. J., Panetsos, P., Papadopoulos, V., Makarios, T. and Thanopoulos,
P. (2008). Seismic fragility curves for greek bridges: methodology and case studies.
Bulletin of Earthquake Engineering. 7(2), pp:439-468.
Melek, M. and Wallace, J. W. (2004). Cyclic behavior of columns with short lap splices. ACI
Structural Journal, 101.
Menegotto, M. and Pinto, P. (1973). Method of Analysis for Cyclically Loaded RC Frames
Including Changes in Geometry and Non-elastic Behaviour of Elements Under
Combined Normal Force and Bending. IABSE Congress Reports of the Working
Commission.
Miller Jr, R. G. (1997). Beyond ANOVA: basics of applied statistics. CRC Press.
Moschonas, I. F., Kappos, A. J., Panetsos, P., Papadopoulos, V., Makarios, T., and P.
Thanopoulos (2008). Seismic fragility curves for greek bridges: methodology and case
studies. Bulletin of Earthquake Engineering 7(2), pp:439-468.
MTD (2008). Bridge Memo to Designers (10-20), California Department of Transportation,
Sacramento, CA.

234
Muthukumar, S. and DesRoches, R. (2006). A Hertz contact model with non-linear damping for
pounding simulation. Earthquake Engineering and Structural Dynamics, 35, pp:811-828.
Ni, P., Mangalathu, S., Mei, G., Zhao, Y. (2017) Permeable Piles: An Alternative to Improve the
Performance of Driven Piles, Computers and Geotechnics, Vol. 84, 2017.
Nielson, B. G. (2005). Analytical fragility curves for higway bridges in moderate seismic zones.
PhD. Thesis, Georgia Tech, Atlanta.
Nielson, B. G, and DesRoches, R. (2006). Influence of modeling assumptions on the seismic
response of multi-span simply supported steel girder bridges in moderate seismic zones.
Engineering Structures, 28, pp:1083-1092.
Ohtaki, T., Benzoni, G., Priestley, M., and Seible, F. (1997). Seismic performance of a full scale
bridge column-as built and as repaired. Second National Seismic Conference on Bridges
and Highways.
Orozco, G. L. (2001). The effects of a large velocity pulse on reinforced concrete bridge columns.
Dept. of Structural Engineering, University of California, San Diego.
Owen, S. V. and Froman, R. D. (1998). Focus on qualitative methods uses and abuses of the
analysis of covariance. Research in Nursing and Health, 21(6), pp:557–562.
Park, J. and Towashiraporn, P. (2014). Rapid seismic damage assessment of railway bridges
using the response-surface statistical model. Structural Safety, 47, pp:1-12.
Padgett, J. E. (2007). Seismic vulnerability assessment of retrofitted bridges using probabilistic
methods, Ph.D Thesis, Georgia Tech, Atlanta, Georgia.
Padgett, J. E., and DesRoches, R. (2007). Sensitivity of seismic response and fragility to
parameter uncertainty. Journal of Structural Engineering, 133, pp:1710-1718.
Porter, K. (2010). Cracking an open safe: uncertainty in HAZUS-based seismic vulnerability
functions, Earthquake Spectra, 26(3), pp:893-900.
Porter, K. A. (2003). An overview of PEER’s performance-based earthquake engineering
methodology. Proceedings of Ninth International Conference on Applications of
Statistics and Probability in Civil Engineering.
Potyondy, J. G. (1961). Skin friction between various soils and construction materials.
Geotechnique, 11(4), pp: 339-353.
Priestley, M. N., Seible, F., and Calvi, G. (1996). Seismic design and retrofit of bridges. John
Wiley and Sons, New York.
Priestley, M. N. and Benzoni, G. (1996). Seismic performance of circular columns with low
longitudinal reinforcement ratios. ACI Structural Journal 93(4).
Ramanathan, K. N. (2012). Next generation seismic fragility curves for California bridges
incorporating the evolution in seismic design philosophy. Ph.D Thesis, Georgia Tech,
Atlanta, Georgia.
Ramanathan, K., Jeon, J.-S., Zakeri, B., DesRoches, R., and Padgett, J. E. (2015). Seismic
response prediction and modeling considerations for curved and skewed concrete box-
girder bridges. Earthquakes and Structures, 9(6), pp:1153-1179.
Ramanathan, K., Padgett, J. E., and DesRoches, R. (2015). Temporal evolution of seismic
fragility curves for concrete box-girder bridges in California, Engineering Structures,
97, pp:29-46.

235
Ranf, R. T., Eberhard, M. O., and Malone, S. (2007). Post-earthquake Prioritization of Bridge
Inspections. Earthquake Spectra 23(1), pp:131-146.
Ranf, R. T., Nelson, J. M., Price, Z., Eberhard, M. O., and Stanton, J. F. (2006). Damage
accumulation in lightly confined reinforced concrete bridge columns, Pacific Earthquake
Engineering Research Center.
Roblee, C. J. (2016a). Memorandum: Box Girder bridges – span length models. P266-T1780
Technical Archives File Bridge Component Models, California Department of
Transportation, Sacramento, California.
Roblee, C. J. (2016b). Memorandum: Box Girder bridges – transverse profile models. P266-
T1780 Technical Archives File Bridge Component Models, California Department of
Transportation, Sacramento, California.
Saiidi, M., Maragakis, E., and Feng, S. (1996). Parameters in bridge restrainer design for seismic
retrofit. Journal of Structural Engineering, 122(1), pp:61-68.
Saini, A. and Saiidi, M. S. (2014). Probabilistic Damage Control Approach for Seismic Design
of Bridges. University of Nevada, Reno.
Sanchez, A. V., Priestley, M., and Seible, F. (1997). Seismic performance of flared bridge
columns. UC San Diago test report.
Schoettler, M., Restrepo, J., Guerrini, G., Duck, D., and Carrea, F. (2012). A full-scale, single-
column bridge bent tested by shake-table excitation. Las Vegas, NV: Center for Civil
Engineering Earthquake Research, Department of Civil Engineering, University of
Nevada.
Seo, J., and Linzell, D. G. (2012) Use of response surface metamodels to generate system level
fragilities for existing curved steel bridges. Engineering Strutures, 52, pp:642-653.
Shamsabadi, A. and Yan, L. (2008). Closed-form force-displacement backbone curves for bridge
abutment-backfill systems. Geotechnical Earthquake Engineering and Soil Dynamics IV.
2008. 1-10.
Shanmugam, S. P. (2009). Seismic behavior of circular reinforced concrete bridge columns
under combined loading including torsion, Ph.D thesis, Missouri University of Science
and Technology.
Shinozuka, M., Feng, M. Q., Lee, J., and Naganuma, T. (2000). Statistical analysis of fragility
curves. Journal of engineering mechanics, 126(12), pp:1224-1231.
Soleimani (2017). Fragility of California bridges – development of modification factors. Ph.D
Thesis, Georgia Tech, Atlanta, Georgia.
Soleimani, F., Mangalathu, S., and DesRoches, R. (2017) A comparative analytical study on the
fragility assessment of bridges with various column shapes, in Review.
Stefanidou, S. P., and Kappos, A. J. (2017). Methodology for the development of bridge-specific
fragility curves. Earthquake Engineering and Structural Dynamics, 46(1), 73-93.
Stone, W. C. and Cheok, G. S. (1989). Inelastic behavior of full-scale bridge columns subjected
to cyclic loading. Gaithersburg, MD, National Institute of Science and Technology.
Sun, Z., Priestley, M., and Seible, F. (1993). Diagnostics and retrofit of rectangular bridge
columns for seismic loads, Department of Applied Mechanics and Engineering Sciences,
University of California, San Diego.

236
Tibshirani, R, (1996). Regression shrinkage and selection via the lasso. Journal of Royal
Statistical Society 58(1), pp:267-288.
Vickers, A. J. (2005). Parametric versus non-parametric statistics in the analysis of randomized
trials with non-normally distributed data. BMC medical research methodology, 5(1), pp:
35.
Vidakovic, B. (2011). Statistics for bioengineering sciences: with MATLAB and WinBUGS
Support, Springer Science and Business Media.
Wald, D., Lin, K.-W., Porter, K. and Turner, L. (2008). ShakeCast: Automating and Improving
the Use of ShakeMap for Post-Earthquake Decision-Making and Response. Earthquake
Spectra 24(2), pp:533-553.
Yu, O., Allen, D. L., and Drnevich, V. P. (1991). Seismic vulnerability assessment of bridges on
earthquake priority routes in Western Kentucky. Lifeline Earthquake Engineering, pp.
817-826.
Zou, H. and Hastie, T. (2005). Regularization and variable selection via the elastic net. Journal
of Royal Statistical Socirty B 67, pp:301-320.
Zhang, J. and Huo, Y. L. (2009) Evaluating effectiveness and optimum design of isolation devices
for highway bridges using the fragility function method. Engineering Structures, 31,
pp:1648-60.
Zhong, J., Gardoni, P., Rosowsky, D., and Haukaas, T. (2008). Probabilistic seismic demand
models and fragility estimates for reinforced concrete bridges with two-column bents.
Journal of engineering mechanics, 134(6), pp:495-504.
Zelaschi, C., Monteiro, R. and Pinho, R. (2016). Parametric Characterization of RC Bridges
for Seismic Assessment Purposes. Structures, 7, pp: 14-24.

237
 VITA
SUJITH MANGALATHU

Ph. D (Jan 2013 - Mar 2017) Georgia Institute of Technology, USA

(GPA 3.92/4.0) Minor: Probability and statistics
Dissertation: Performance based grouping and
fragility analysis of box-girder bridges in
California

M.Sc. (GPA 9.0/10.0) Istituto Universitario di Studi Superiori

May 2012 (IUSS) in Pavia, Italy, and University of
Patras, Greece
Specialization: Masters in Earthquake
Engineering and Engineering Seismology
(MEEES)
Dissertation (EPFL, Switzerland): Force-
deformation characteristics of masonry
spandrels

M.Tech. (GPA 8.78/10.0) Indian Institute of Technology Madras

May 2009 (IITM), India
Specialization: Structural Engineering
Dissertation: Reliability analysis of cantilever
retaining walls

PUBLICATIONS
Refereed Journal Articles

1. Ni, P., Mangalathu, S., Mei, G., Zhao, Y., “Permeable Piles: An alternative to
improve the performance of driven Piles”, Computers and Geotechnics, Vol. 84,
2017.

2. Mangalathu, S., Jeon J-S., DesRoches, R., Padgett, J. E., “ANCOVA-based grouping
of bridge classes for seismic fragility assessment”, Engineering Structures, Vol.
123, 2016.

3. Beyer, K., Mangalathu, S., “Numerical study on the peak strength of masonry
spandrels with arches”, Journal of Earthquake Engineering, Vol. 18, No. 2, 2014.

238
4. Beyer, K., Mangalathu, S., “Review of strength models for masonry spandrels”,
Bulletin of Earthquake Engineering, Vol. 11, No. 2, 2013.

5. Sujith, M. S., Menon, D., Dodagowdar, G. R., “Reliability analysis and design of
cantilever RC retaining walls against sliding Failure”, International Journal of
Geotechnical Engineering, vol. 5, No. 2, 2011.

6. Mandali, A. K. Sujith, M. S., Rao, B. N., Janardhana Maganti., “Reliability analysis

of counterfort retaining Walls”, Electronic Journal of Structural Engineering, vol.
11, No. 1, 2011.

Journal Articles (In review)

7. Mangalathu, S., Jeon J-S., DesRoches, R., Padgett, J. E., “Performance-based

grouping methods of bridge classes for regional seismic risk assessment:
application of ANOVA, ANCOVA, and non-parametric approach”, Earthquake
Engineering and Structural Dynamics.

8. Dukes, J., Mangalathu, S., Padgett, J. E., DesRoches, R., " Development of
bridge specific fragility methodology to improve the seismic resilience of
bridges", Earthquakes and Structures.

9. Jeon J-S., Mangalathu, S., Song, J., DesRoches, R., “Parameterized seismic
fragility curves for curved multi-frame concrete box-girder bridges using
Bayesian parameter estimation”, Journal of Earthquake Engineering.

10. Mangalathu, S., Jeon J-S., DesRoches, R., “Identification of Critical Parameters
on the Seismic Performance of Concrete Bridges using Lasso Regression”,
Earthquake Engineering and Structural Dynamics.

11. Ni, P., Mangalathu, S., Mei, G., Zhao, Y., “Compressive and flexural behavior of
Reinforced Concrete Permeable Piles”, Engineering Structures.

12. Nishanth, M., Dhir, P., Davis, R. Mangalathu, S., “Stochastic response of RC
buildings under high dimensional model representation”, Earthquake
Engineering and Structural Dynamics.

Conference Proceedings and Oral presentations

1. Mangalathu, S., Soleimani, F., DesRoches, R., Padgett, J. E., “ANOVA based
grouping of bridge classes”, 16th World Conference on Earthquake Engineering,
Santiago, Chile, 2017.

239
2. Mangalathu, S., Jeon J-S., Soleimani, F., DesRoches, R., Padgett, J. E., Jiang, J.,
“Sensitivity of fragility curves to parameter uncertainty using Lasso regression”,
16th World Conference on Earthquake Engineering, Santiago, Chile, 2017.

3. Mangalathu, S., Jeon J-S., DesRoches, R., Padgett, J. E., “Application of Bayesian
methods to probabilistic demand analyses of concrete box-girder bridges”,
Geotechnical and Structural Engineering Congress, Phoenix, Arizona, 2016.

4. Mangalathu, S., Jeon J-S., DesRoches, R., Padgett, J. E., “Analysis of covariance to
capture the importance of bridge attributes on the probabilistic seismic demand
model”, 10th Pacific Conference on Earthquake Engineering, Sydney, Australia,
2015.

5. Mangalathu, S., Jeon J-S., Soleimani, F., DesRoches, R., Padgett, J. E., Jiang, J.,
“Seismic vulnerability of multi-Span bridges: an analytical perspective”, 10th
Pacific Conference on Earthquake Engineering, Sydney, Australia, 2015.

6. Mangalathu, S., Jeon J-S., Soleimani, F., DesRoches, R., Padgett, J. E., “Fragility
analysis of box-girder bridges with lap-splice mode of failure”, EERI Annual
meeting, Boston, 2015.

7. Mangalathu, S., Beyer, K., “Force-deformation characteristics of masonry

spandrels with shallow arches”, Twelfth Rose School Seminar, Italy 2012.

8. Mangalathu, S., Beyer, K., “Numerical analysis of masonry spandrels with

shallow arches”, 15th World Conference on Earthquake Engineering, Portugal,
Lisbon 2012.

HONORS AND AWARDS

2015 EERI travel grant for 2015 EERI Annual meeting, Boston.

2010-2012 Erasmus Mundus Scholarship, European Union

2007 All India Rank 17 (99.98 percentile) in the national

engineering examination “Graduate Aptitude Test in
Engineering (GATE)” for admissions to graduate program

MEMBERSHIPS

American Concrete Institute (ACI), American Society of Civil Engineers (ASCE),

Earthquake Engineering Research Institute (EERI), Structural Engineering Institute
(SEI)

240
PROFESSIONAL ACTIVITIES

Technical Committee Member, SEI-ASCE’s Task Group 3 (TG3) on the risk

assessment of structural and infrastructural systems
• Machine learning technique for the reliability analysis of networks
President, EERI chapter Georgia Tech (2015 – 2017)
• Visited various schools in Atlanta to present activities and demos for the
children.
• Associated with a team in developing the website for EERI
(http://eeri.ce.gatech.edu/)
Chair, SEI chapter Georgia Tech (2016 – 2017)
• Involved in a team to organize workshops and seminars in Georgia Tech

241