RELIABILITY-BASED CONDITION ASSESSMENT OF

EXISTING HIGHWAY BRIDGES

A Dissertation
Presented to
The Academic Faculty

By

Naiyu Wang

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

Georgia Institute of Technology

August, 2010
 Reliability-based Condition Assessment of Existing Highway Bridges

Approved by:

Dr. Bruce R. Ellingwood, Advisor Dr. Kenneth M. Will

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

Dr. Abdul-Hamid Zureick Dr. Donald W. White

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

Dr. James I. Graig

School of Aerospace Engineering
Georgia Institute of Technology

Date Approved: May 17, 2010

 ACKNOWLEDGEMENTS

My years at Georgia Tech have been, and will always be, one of the most

valuable experiences in my life. It is the people with whom I have the privilege to work

that have made this journey so special and memorable.

I cannot express enough my appreciation to my co-advisor Dr. Bruce R.

Ellingwood. Beyond his wealth knowledge, passion for research, and dedication to

excellence, he has showed me what it means to be a professional engineer, an outstanding

professor and an exceptional researcher. Working with him has shaped me in so many

ways and his influence will no doubt propagate beyond my Ph.D. study and serve me

well for many years to come. I also would like to extend my gratitude to my co-advisor,

Dr. Abdul-Hamid Zureick, who has encouraged me with his knowledge, patience,

enthusiastic support, and guidance throughout the study. I attribute my valuable

experience in Tech and research contributions to them.

I am grateful to my committee members - Dr. Kenneth M. Will, Dr. Don White,

and Dr. James I. Craig - for their insight, advice and critical assessment of my

dissertation work. The research described in this dissertation was supported, in part, by

the Georgia Department of Transportation under grant RP05-01. This support is

gratefully acknowledged.

I also like to thank wonderful fellow graduate students, Quanwang, Chuang-

Sheng, Guoqing, Murat, Ozan, Kursat, Eun, Soravit, Laura, Bo, Rulan and Alan, for

iii
offering me technical knowledge, lively discussions and friendship along the road. I

extend a special note of appreciation to Curtis for providing me with reliable

experimental data for the calibration of my simulations and for his willingness to

coordinate on many other tasks of this research project. Thanks also due to my best

friends, sisterhood, Li and Ming, for continuing to be so thoughtful and encouraging.

My ultimate support is always from my families: Xuefang, Dingyi, Xiaoyu,

Xiangge and Ranran, without whom I could not stand, and from my husband, Yong, an

unexpected reward from this journey, who has brought immeasurable happiness to my

life.

iv
 TABLE OF CONTENTS

ACKNOWLEDGEMENTS iii

LIST OF TABLES ix

LIST OF FIGURES x

SUMMARY xii

CHAPTER 1: INTRODUCTION 1

1.1 Background 1

1.2 Research Objective and Scope 3

1.3 Outline of the Thesis 4

CHAPTER 2: REVIEW OF CONDITION EVALUATION PRACTICES FOR

EXISTING BRIDGES 7

2.1 Reliability-based Bridge Condition Assessment Framework 7

2.2 Current AASHTO Guidelines for Bridge Evaluation 10

2.2.1 Bridge Rating by ASR, LFR and LRFR 10

2.2.2 Probability Models and Supporting Data for Reliability-based

Bridge Rating 19

2.3 Survey of Bridge Rating Practices in the United States 25

2.3.1 Development of Survey 25

2.3.2 Synthesis of Survey Response 26

2.3.3 Comparison of Rating Methods through Sample Bridges 35

2.4 Bridge Evaluation in Foreign Countries 37

v
 2.4.1 Bridge Rating in Canada 37

2.4.2 Bridge Rating in the United Kingdom 38

2.4.3 Bridge Rating in Australia 41

2.5 Critical Appraisal of Current Bridge Evaluation Practice 43

2.6 Closure 45

CHAPTER 3: IMPROVED FRAMEWORK FOR BRIDGE RATING PRACTICE

– A PROTOCOL WITH THREE LEVELS OF ASSESSMENT 47

CHAPTER 4: SELECTION, TESTING AND FINITE ELEMENT MODELING OF

THE SAMPLE BRIDGES 52

4.1 Selection of Bridges for Testing and Analysis 52

4.1.1 Reinforced Concrete Bridge – Straight (ID: 129-0045) 54

4.1.2 Reinforced Concrete Bridge – Skewed (ID: 015-0108) 55

4.1.3 Pre-stressed Concrete Girder Bridge (ID: 223-0034) 56

4.1.4 Steel Girder Bridge (ID: 085-0018) 57

4.2 Development of Finite Element Models of the Selected Bridges 59

4.3 Validation of Finite Element Models of the Selected Bridges through

Diagnostic Load Tests 62

4.4 Closure 64

CHAPTER 5: LEVEL-ONE ASSESSMENT: CORRELATION BETWEEN

CONDITION RATING AND CAPACITY RATING 67

5.1 Introduction 67

5.2 Stochastic Bridge Deterioration Process 68

5.3 Synthesis of Conditional Rating Data from NBI 75

vi
 5.5 Correlation between Condition Rating and Resistance Model 76

5.5 Calibration of Condition Factor, φc 77

5.6 Closure 79

CHAPTER 6: LEVEL-TWO ASSESSMENT: RATING USING COMPONENT-

LEVEL INFORMATION 80

6.1 Rating Using In Situ Material Strength 81

6.2 Rating Using Realistic Load Distributions 83

6.3 Rating of Deep Reinforced Concrete Bridge Components 84

6.4 Closure 88

CHAPTER 7: LEVEL-THREE ASSESSMENT: BRIDGE SYSTEM

RELIABILITY 90

7.1 Virtual Proof Load Tests of Sample Bridges 91

7.1.1 Push-down Analysis 92

7.1.2 Finite Element Analysis Based Bridge System Reliability 94

7.2 Rating Using Service Proving Load History 98

7.3 Closure 100

CHAPTER 8: SUMMARY, CONCLUSIONS AND FUTURE WORK 104

8.1 Summary 104

8.2 Research Conclusions 105

8.3 Recommended Future Work 107

REFERENCES 110

vii
 LIST OF TABLES

Table 2.1 Statistical Parameters of Component Resistance [Nowak, 1999] 23

Table 2.2 Statistical Parameters of Dead Load [Nowak, 1999] 23

Table 3.1 Responding States 26

Table 2.4 Summary of Sample Bridge Flexural Rating for Interior Girders 36

Table 2.5 Summary of Sample Bridge Shear Rating for Interior Girders 36

Table 4.1 Truck Weight (lb) Details for RC T-Beam Bridge (ID: 129-0045) Test 64

Table 4.2 Comparison of the Maximum Deflections Measured in the Test and
Predicted by FE Analysis 64

Table 5.1 NBIS Instruction for Superstructure Condition Rating 70

Table 5.2 Condition Factor, φc (AASHTO LRFR, 2005, Table6-2) 70

Table 5.3 Random Variable for MC Simulation (Enright and Frangopol, 2000) 74

Table 5.4 Proposed Condition Factors 79

Table 6.1 Compression Tests of Cores from the RC Bridge (ID 129-0045) 83

Table 6.2 Comparison of the Moment Distribution Factors for Interior Girders 85

Table 6.3 Shear Ratings (tons) for RC Bridge (ID 085-0018) 87

Table 6.4 Shear Ratings (tons) for HS20 Vehicle 88

Table 7.1 Analysis of Bridge Capacity, Determined as the Point of First Yield 97

Table 7.2 Load Rating at Component Level vs System Level (HS20 Operating
Rating) 97

Table 7.3 Comparison of Rating Factors Computed Before and After Considering
Service Load History for RC Bridge (ID: 129-0045) 103

viii
 LIST OF FIGURES

Figure 2.1 Load and Resistance Factor Rating Procedure [AASHTO LRFR, 2005] 16

Figure 2.2 LRFD Design Live Load (HL-93) 19

Figure 2.3 State of Georgia Legal Loads 20

Figure 3.1 Improved Bridge Rating Framework 49

Figure 4.1 Bridge Categories Identified by Decade of Construction 53

Figure 4.2 Posted Bridges Identified by Decade of Construction 53

Figure 4.3 Straight T-Beam Bridge (ID: 129-0045, Gordon County) 55

Figure 4.4 Skew T-Beam Bridge (ID: 015-0108, Bartow County) 56

Figure 4.5 Pre-Stress Bridges (ID: 223-0034, Paulding County) 57

Figure 4.6 Steel Girder Bridge (ID: 085-0018, Dawson County) 58

Figure 4.7 FEM of the Gordon County Bridge (ID: 129-0045) 61

Figure 4.8 Schematic of Concrete Reinforced T-beam Bridge 63

Figure 4.9 RC T-beam Bridge Girder Displacements Due to Truck Loadings 66

Figure 5.1 Mean and COV of g(t) of Time-variant Bending Resistance with Different
Corrosion Rate rcorr (Enright and Frangopol, 2000) 73

Figure 5.2 Time-dependent Mean and Coefficient of Variation of Bridge Degradation

Function g(t) 74

Figure 5.3 Average Condition Rating as Function of Time for Non-interstate

Highway Bridges (Bolukbasi et al, 2004) 76

Figure 5.4 Time-dependent Mean and COV of Bridge Flexural Capacity 77

Figure 5.5 Optimal Condition Factors for Different Condition Ratings 78

ix
Figure 6.1 Development of Arch Action in Deep Beam 87

Figure 7.1 Push-down Analysis of the RC Bridge (ID: 129-0045) 94

Figure 7.2 Lognormal Fit of System Resistance of the RC Bridge (ID: 129-0045) 96

Figure 7.3 Structural Reliability Models for Bridge Proof Load Test 101

Figure 7.4 Influence of service load on updated distribution of structural resistance

for the RC Bridge (ID: 129-0045) 101

Figure 7.5 Updated Failure Probabilities and Reliability Indices for RC Bridge
(ID: 129-0045) 102

Figure 7.6 Updated Rating Factors Respect to HL-93 at Inventory Level for the RC
Bridge (ID: 129-0045) 102

x
 SUMMARY

Condition assessment and safety verification of existing bridges and decisions as

to whether bridge posting is required are addressed through analysis, load testing, or a

combination of methods. Bridge rating through structural analysis is by far the most

common procedure for rating existing bridges. The American Association of State

Highway and Transportation Officials (AASHTO) Manual for Bridge Evaluation (MBE),

First Edition permits bridge capacity ratings to be determined through allowable stress

rating (ASR), load factor rating (LFR) or load and resistance factor rating (LRFR); the

latter method is keyed to the AASHTO LRFD Bridge Design Specifications, which is

reliability-based and has been required for the design of new bridges built with federal

findings since October, 2007. A survey of current bridge rating practices in the United

States has revealed that these three methods may lead to different ratings and posting

limits for the same bridge, a situation that carries serious implications with regard to the

safety of the public and the economic well-being of communities that may be affected by

bridge postings or closures.

To address this issue, a research program has been conducted with the overall

objective of providing recommendations for improving the process by which the

condition of existing bridge structures is assessed. This research required a coordinated

program of load testing and finite element analysis of selected bridges in the State of

Georgia to gain perspectives on the behavior of older bridges under various load

xi
conditions. Structural system reliability assessments of these bridges were conducted and

bridge fragilities were developed for purposes of comparison with component reliability

benchmarks for new bridges. A reliability-based bridge rating framework was developed,

along with a series of recommended improvements to the current bridge rating methods,

which facilitate the incorporation of various in situ conditions of existing bridges into the

bridge rating process at both component and system levels. This framework permits

bridge ratings to be conducted at three levels of increasing complexity to achieve the

performance objectives, expressed in the terms of reliability, that are embedded in the

LRFR option of the AASHTO Manual of Bridge Evaluation. This research was

sponsored by the Georgia Department of Transportation, and has led to a set of

Recommended Guidelines for Condition Assessment and Evaluation of Existing Bridges

in Georgia.

xii
 CHAPTER 1
INTRODUCTION

1.1 BACKGROUND

Bridge structures in the State of Georgia are at risk from aging, leading to

structural deterioration from service demands from increasing traffic and heavier loads,

from aggressive environmental attack and other physical mechanisms, and from deferred

maintenance. Condition assessment and evaluation of existing bridges may be prompted

by changes in traffic patterns; concern about faulty building materials or construction

methods; discovery of a design/construction error after the structure is in service; concern

about deterioration discovered during routine inspection; and damage following extreme

load events. A condition assessment may be conducted to develop a bridge load rating,

confirm an existing load rating, increase a load rating for future traffic, or determine

whether the bridge must be posted in the interest of public safety. The Bridge Inventory

Management System in the State of Georgia lists 8,9881 bridges, which are monitored by

the Georgia Department of Transportation (GDOT). While rating calculations have yet

to be performed on 1,587 of these bridges, it has been determined that approximately

1,982 (or 22%) of them require posting. Posting or other restrictive actions may have a

severe economic impact on the state economy, which depends on the trucking industry

for distribution of resources and manufactured goods. The economics of upgrading or

posting a bridge makes it imperative that condition assessment criteria and methods

1
As of August 1, 2009

1
(either by analysis or by testing) be tied in a rational and quantitative fashion to public

safety, functional requirements and economics.

Condition assessment and safety verification of existing bridges, and decisions as

to whether posting is required are addressed through analysis, load testing, or a

combination of methods. Of these, bridge rating by structural analysis is by far the most

common (and most economical) method. Load testing may be indicated when analysis

produces an unsatisfactory result, when the analysis cannot be completed due to lack of

design documentation or information, or when structural deterioration of the bridge

renders the traditional analysis methods questionable or inapplicable. Until recently, the

customary rating process used in most states has been described in the American

Association of State Highway and Transportation Officials (AASHTO) Manual for

Condition Evaluation of Bridges, Second Edition, which allows ratings to be determined

through either allowable stress methods (ASR) or load factor methods (LFR). In recent

years, the State of Georgia has utilized the LF method for the majority of those bridges in

the state that have been rated. A third (and more recent) rating procedure found in the

Manual for Condition Evaluation and Load and Resistance Factor Rating (LRFR) of

Highway Bridges is keyed to the AASHTO Load and Resistance Factor Design (LRFD)

method, which is defined in the AASHTO LRFD Bridge Design Specification, Fourth

Edition. The LRFR method is being introduced in bridge maintenance, and some states

are beginning to use it to determine bridge ratings. The AASHTO Manual for Bridge

Evaluation (MBE), First Edition (2008) has included all three methods. These three

competing rating methods may lead to different rated capacities and posting limits for the

same bridge, a situation that cannot be justified from a professional engineering

2
viewpoint and has potentially serious implications with regard to the safety of the public

and the economic well-being of businesses and individuals who may be affected by

bridge postings or closures. Moreover, analytical methods with a fixed format are

designed to be applicable to the entire highway bridge population. While they may be

conservative for many bridges, they also may fail to properly consider all the risks facing

a particular bridge, since the condition rating and capacity rating in the current rating

practice is relatively weak. The cost and social impact of failing to meet a performance

objective when evaluating an existing bridge, especially in terms of posting unnecessarily

or failing to post when necessary, can be very large. Accordingly, the economics of

upgrading or posting a bridge makes it imperative to determine condition assessment

criteria and methods (either by analysis or by testing) that are tied in a rational and

quantitative fashion to public safety and functional objectives.

The Georgia Department of Transportation has a need for condition assessment

tools that can be used with confidence to determine whether or not to post certain existing

bridge structures. To address this need, the Georgia Institute of Technology has

conducted a multi-year research program, sponsored by the GDOT, aimed at making

improvements to the process by which the condition of existing bridge structures in the

State of Georgia is assessed. The research reported in this dissertation provides the

technical basis for improving bridge rating practices in Georgia.

1.2 RESEARCH OBJECTIVES AND SCOPE

The research in this dissertation is aimed at developing a practical reliability-

based analysis framework for assessing bridge load-carrying capacity and for developing

3
rational inspection/maintenance strategies and policies. To accomplish this objective, the

following research tasks have been conducted:

• Critically appraise current bridge condition assessment procedures.

• Summarize and assemble applicable structural reliability analysis tools to

support condition assessment of bridge systems.

• Select, test and perform in-depth analysis of sample bridges that are

representative of bridges that are of most concern to the GDOT.

• Develop a reliability-based bridge capacity assessment framework which

provides practical tools for incorporating available in situ data into reliability

based rating analysis.

The scope of this dissertation is limited to bridges on the state primary and

secondary system, including reinforced concrete tee, prestressed and steel girder

bridges, which are subjected primarily to permanent gravity loads and vehicular

loads. Interstate, wood or historical bridges and railway bridges are excluded.

1.3 OUTLINE OF THE THESIS

This thesis is organized into eight chapters.

Chapter 2 reviews the reliability bases for current bridge condition evaluation and

AASHTO guidelines on bridge assessment, and summarizes the results of a survey of

state DOTs conducted to investigate the engineering practices on bridge rating

nationwide. In addition, documents used for bridge rating in the United Kingdom,

Australia, and Canada were obtained to gain an international perspective on the subject.

4
 Chapter 3 presents a general framework for bridge safety evaluation that directly

addresses the deficiencies in current practice noted in Chapter 2. This framework has

three levels of assessment of increasing complexity. In the first level, the deterministic

member-based format of the AASHTO LRFR method is kept, and the correlation

between visual condition rating and the capacity evaluation is established. The second

level allows for the incorporation of site-specific data obtained from material tests,

diagnostic load test and from in-depth structural analysis in rating calculations. In the

third level, bridge system reliability is evaluated by incorporating proof load test results

and routine inspection records regarding bridge performance history. This framework

highlights the learning process in rating a given bridge and provides clear incentives to

obtain quantitative in situ measurements in routine bridge inspections.

Four bridges typical of bridges of concern in rating and posting are summarized in

Chapter 4 and the load testing and finite element analysis of these bridges is described.

Finite element models of these bridges were developed to assist the design of the load tests

and in the interpretation of the results. The bridge test results, in turn, were used to validate

and improve the finite element modeling.

Chapter 5 presents the level-one assessment which is basically consistent with the

current AASHTO LRFR method, but with one significant adjustment: a new method is

introduced to correlate visually-based bridge condition ratings from routine periodic

inspections with structural capacity. A revised set of values of φc tied to the AASHTO

LRFR rating equations are developed to be consistent with the structural reliability-based

philosophy embodied in the AASHTO LRFR and to incorporate recent developments in

5
bridge resistance degradation modeling and comprehensive databases of bridge condition

rating history.

In the level-two assessment in Chapter 6, bridge resistance models at the

component level can be “customized” for an individual bridge by incorporating available

site-specific knowledge. This level of analysis reflects the fact that each bridge is unique

in its as-built condition, and provides bridge engineers with an option to account for this

uniqueness to achieve a better evaluation of the bridge performance when such effort is

believed to be warranted. Tools are provided for incorporating the structural component

knowledge gained from in-situ material tests, diagnostic load tests, and improved

mechanical models for structural component analysis into the bridge safety evaluation

process.

Chapter 7 presents the level-three assessment, which focuses on bridge safety at

the system rather than component level and provides additional perspective on the

(unknown) level of conservatism furnished by the current generation of reliability-based

condition evaluation and rating procedures which are member-based. The possibilities of

incorporating proof load test results and successful service performance history into the

bridge rating framework are explored.

Finally, Chapter 8 summarizes the major research findings and conclusions, and

outlines future research needs.

6
 CHAPTER 2
REVIEW OF CONDITION EVALUATION PRACTICES FOR
EXISTING BRIDGES

This chapter reviews current rating procedures for performing condition

assessments of existing bridge structures. The review emphasizes current practices in the

United States, but practices in several other industrialized countries are also summarized

to provide additional context. This review is aimed at achieving a general perspective on

technical issues associated with condition assessment methodologies used for bridges and

other civil infrastructure applications.

2.1 RELIABILITY-BASED BRIDGE CONDITION ASSESSMENT FRAMEWORK

Uncertainties arise from variations in loads, material properties, dimensions,

natural and manmade hazard, insufficient knowledge, and human errors in design and

construction [Ellingwood, et al, 1980; 1982; Ellingwood and Galambos, 1982]. The

uncertain nature of the data makes structural reliability theory a logical and powerful tool

for performing quantitative performance assessments of existing structures. Moreover,

probability-based limit states analysis provides a clear link between theoretical research

and in-service experience, and also provides a theoretical basis for utilizing in situ data in

the bridge evaluation process. Recent advances in bridge design and rating in the United

States and elsewhere have a reliability basis.

The starting point for a quantitative evaluation of structural reliability is the

description of the limit state of concern (flexural failure, instability, etc) by an expression

7
relating the resistance and load variables described above, derived from principles of

structural mechanics. This expression, denoted the limit state function, is given by,

G ( X ) = G ( X 1 , X 2 ,L X m ) = 0 (2-1)

in which X = (X1, X2, …, Xm) = vector of random resistance and load variables. The limit

state is defined, by convention, as when G(X)<0. Thus, the limit state probability is,

Pf = ∫ f x ( x1 , x 2 , L x m ) dx1 dx 2 L dx m (2-2)

in which f x (x) = joint probability density function of X and the domain of the multi-fold

integration is that region of x where G(X) < 0. The limit state probability, Pf, is the

quantitative metric of structural performance that is consistent with the uncertainties in

structural resistance and loads.

Modern probability-based limit states design approaches, including the AASHTO

LRFD Bridge Design Specifications [2007], have adopted the reliability index, β, as a

measure of reliability instead of Pf. For typical structural engineering situations, the

reliability index is in the range of 2 to 4.5. The reliability index is related, in a first-order

sense, to the limit state probability by Pf = Φ (-β) for well-behaved limit state functions

that are typical of those found in bridge design and condition assessment.

The target reliability index of 3.5 in the AASHTO LRFD Specifications for new

bridge structures was determined by calibration to a spectrum of traditional bridge

design situations (vintage 1985 and earlier) involving steel, reinforced and pre-stressed

concrete construction. Gravity load situations were considered in this calibration

exercise. A group of experts from the material specification committees participated in

assessing the results of this calibration, and selecting target reliabilities. This target index

of 3.5 was adoped directly in LRFR for capacity checking at the design load level. For

8
AASHTO legal load or State legal loads, the target index was chosen to be 2.5 by

judgment [Moses, 2001; Minervino, et al, 2004]. In the latter case, the implied

acceptable annual failure rate of an existing bridge would be at least an order of

magnitude higher than a newly constructed bridge. The design load checking and legal

load checking are comparable to inventory and operating level checking, which are

discussed later in section 2.2.1.

When an existing bridge structure is evaluated, the knowledge gained from

additional in situ data about the existing bridge or its components through field

inspection, load testing, material tests, or traffic surveys, if available, could be applied to

refine the probabilistic models of related random variables. Inspection, therefore, should

lead to an improvement in the prior estimate of failure probability discussed above. The

theoretical basis for incorporating additional information is provided by Bayes theorem

[Ang and Tang, 2007]. The updated (posterior) failure probability of an existing

bridge, Pf' , can be expressed as [Madsen, 1987; Ellingwood, 1996]:

P[G ( X ) < 0 ∩H ]
Pf' = P[G ( X ) < 0 | H ] = (2-3)
P[ H ]

in which, H is an event describing the available site-specific knowledge, such as the

result of a bridge inspection or a proof load test. It is clear that strong stochastic

dependence between the events [G ( X ) < 0] and [ H ] will produce a tighter updated

distribution, giving more confidence about the estimated random vector X . If little is

learned by the inspection or condition assessment, the correlation is weak; the prior and

posterior estimates of the reliability will be nearly the same.

9
 It should be emphasized that current condition assessment procedures for bridges

do not utilize the valuable information regarding in situ condition that is reflected by the

updating process summarized in Eq (2.3). The proposed methodology in Chapter 4 will

remedy this deficiency in current practice.

The structural reliability theory framework described above provides a conceptual

platform for the codified limit state bridge design and evaluation. The probabilistic

models of the major random variables involved in the bridge reliability analysis will be

discussed next.

2.2 CURRENT AASHTO GUIDELINES FOR BRIDGE EVALUATION

2.2.1 Bridge Rating by ASR, LFR and LRFR

Until 1970, the sole design philosophy embedded within AASHTO Standard

Specifications for Highway Bridges was Allowable Stress Design (ASD). The allowable

stress is established as a fraction of the load carrying capacity of a structural element

(usually the yield or fracture strength in tension or point of instability in compression),

and the structural action (stress in tension, bending or compression) from the applied

loads may not exceed this allowable limit. Detailed procedures for rating existing bridges

based on the ASD method first appeared in 1970 in the AASHTO Manual for

Maintenance Inspection of Bridges.

Beginning in the early 1970's, as the design of reinforced concrete and steel

structures was reformulated in terms of "ultimate strength" for concrete and "plastic"

design for steel, the load analysis formerly used in ASD was modified as well, with

adjustments to the load factors to reflect the relative uncertainty and predictability of

different loads, such as vehicle loads, wind and earthquake effects. The new design

10
philosophy was referred to as Load Factor Design (LFD) and was incorporated in the

Manual for Condition Evaluation of Bridges (MCE), which was published by AASHTO

in 1994 to replace the earlier Manual for Maintenance Inspection of Bridges. Although

the 1994 manual contains some guidance for allowable stress rating (ASR), it clearly

emphasized the load factor rating (LFR) method. Many State DOTs continue to use the

1994 Manual, with 1995, 1996, 1998 and 2000 interim revisions, in their bridge rating

work1.

In 1994, the AASHTO Bridge Subcommittee voted to adopt the AASHTO LRFD

Bridge Design Specifications and in 1998 designated LRFD as the primary design

method for highway bridges. The LRFD Bridge Design Specifications (the latest edition

is dated 2010) represented the first effort by AASHTO to integrate modern principles of

structural reliability and the probabilistic and statistical models of loads and resistance

into the design of highway bridges. LRFD introduced the reliability-based limit states

design philosophy to achieve a more uniform and controllable safety levels for each

applicable limit state. To extend this philosophy to the evaluation of existing bridges,

AASHTO released the 2003 Guide Manual for Condition Evaluation and Load and

Resistance Factor Rating (LRFR) of Highway Bridges, which presents the first bridge

load rating method in the United States to have a structural reliability basis.

At the present time, the ASR, LFR and LRFR methods of bridge rating are all

included in AASHTO Manual for Bridge Evaluation (MBE), First Edition, 2008 and are

1
Wang, N., Ellingwood, B.R., Zureick, A. and O’Malley, C. (2009). “Condition assessment of existing
bridge structures: Report of Task 1 – Appraisal of state-of-the-art of bridge condition assessment.” Report
of Project GDOT No. RP05-01, Georgia Department of Transportation, Atlanta, GA (ftp://ftp.dot.
state.ga.us/DOTFTP/Anonymous-Public/Research_Projects/) 05/2010.

11
current use by State DOTs. A summary of these procedures and a critical appraisal of

their relative merits are presented in this section.

Allowable Stress Rating (ASR) and Load Factor Rating (LFR)

The rating factors in both ASR and LFR are determined by [AASHTO MCE,

1994]:

C - A1 D
RF = (2-4)
A2 L(1 + I )

in which RF is the rating factor for the live load carrying capacity (expressed as a

multiple of the design live load effect (from a rating vehicle) that can be carried by the

bridge), C is the capacity of the structural member, D and L are, respectively, the dead

and live load effect on the member, I is the impact factor to be used with the live load

effect, A1 is the factor on dead load, and A2 is the factor on live load. The Rating Factor

(RF) determined from Eq (2-4) is used to compute the rating of the bridge in tons as

[AASHTO MCE, 1994]:

RT = ( RF ) × W (2-5)

where RT is the bridge member rating in tons, and W is the nominal weight (tons) of the

rating truck used in determining the live load effect (L).

Both ASR and LFR methods rate bridges at two levels: Inventory and Operating.

The Inventory rating level generally corresponds to the customary design level of

allowable stress or strength, but reflects the existing bridge and material conditions with

regard to structural deterioration. Load ratings based on the Inventory level allow a

comparison of the estimated capacity of an existing bridge with the capacity for a new

bridge, and therefore result in a live load which can safely carried by the existing bridge

12
structure for an indefinite period of time. In contrast, load ratings based on the Operating

rating level generally describe the maximum permissible live load to which a structure

may be subjected during a limited period of time. Allowing an unlimited number of

vehicles to use the bridge at the Operating level may shorten the life of the bridge

[AASHTO MCE, 1994]. Rating at the Operating level generally is the basis for decisions

regarding traffic restriction and load posting.

Although the rating factor format for ASR and LFR is the same, the load factors

( A1 , A2 ) and the calculation of the capacity (C) used in Equation (2.1) are different. In

ASR, A1 = A2 =1.0 for both Inventory and Operating level rating; C depends on the rating

level desired, with the higher value of C used for Operating level. In the LFR procedure,

A1 = 1.3, while A2 equals 2.17 for Inventory rating and equals 1.3 for Operating level

rating; the nominal capacity C is the same regardless of the rating level desired.

Load and Resistance Factor Rating (LRFR) Procedure

The general LRFR rating equation is (AASHTO LRFR, 2003):

C - γ CD DC - γDW ± γ P P
RF =
γ L LL(1 + IM )
C = φφC φ S Rn (2-6)
φ sφ c ≥ 0.85

in which C is the structural capacity, Rn is the nominal member resistance, DC is the

dead-load effect of structural components and attachments, DW is the dead-load effect of

wearing surfaces and utilities, P is the permanent loading other than dead loads (post-

tensioning for example), LL is the live-load effect, IM is the dynamic load allowance,

γ DC is the load factor applied to the weight of structural components and attachments,

γ DW is the load factor for wearing surfaces and utilities, γ P is the load factor for

13
permanent loads other than dead loads, and γ L is the live-load factor. The resistance

factor ( φ ) accounts for the general uncertainties in the resistance of a bridge member in a

satisfactory condition and is the same as that used in LRFD bridge design. The condition

factor ( φC ) accounts for increasing uncertainties in bridge member resistance once its

condition deteriorates, and takes a value of 0.85 for members in poor condition, 0.95 for

members in fair condition, and 1.0 for members in good condition. The system factor

( φ S ) accounts for the level of redundancy in the structure. Bridges that are less redundant

or non-redundant are assigned a lower system factor and therefore have lower rated

capacities.

The LRFR method supports bridge evaluation for three general limit states that

were introduced in the LRFD Bridge Specification: the strength-limit state (flexural or

shear capacity), the service-limit state (deflections and rotations) and the fatigue limit

state. The strength limit state is fundamental for public safety and is the main determining

factor for bridge posting, closure and repairing. Service and fatigue limit states are

applied selectively to bridges.

In the LRFR method, bridges are evaluated in a three-step approach for each limit

state, as shown in Figure 2.1: design load rating (HL93), legal load rating

(AASHTO/state legal trucks), and permit load rating (overweight trucks). An initial

check first is performed using the HL-93 design load (Figure 2.2) using the dimensions

and properties corresponding to the present in situ condition of a bridge. The bridge is

rated using the same live and dead load factors as those used in the LRFD Bridge Design

Specifications, which were calibrated to ensure a safety index of 3.5 (discussed

subsequently in section 2.1.2 of this chapter). This check measures the performance of

14
the existing bridge in comparison to the expected performance of a new bridge, and

serves as an initial screening check; a bridge resulting in a RF at this level larger than 1.0

requires no further analysis for any legal loads that result in member forces lower than

the HL-93 design load. For example, the HL-93 load is designed to represent the

member forces caused by the AASHTO legal loads through a single load case. Therefore

any State legal loads that are equal to or less than the AASHTO legal load are covered by

a HL-93 design load analysis. On the other hand, if a state has legal loads that surpass

the AASHTO legal loads, those states must verify that HL-93 load case incorporates

those legal loads.

If the bridge fails to pass the HL-93 design load check, a follow-up evaluation is

performed using the AASHTO/State legal trucks (illustrated for the State of Georgia in

Figure 2.3). The live load factor used at this level is calibrated to a safety index of 2.5 and

varies in accordance with local truck traffic conditions at the bridge site (ADTT). The

safety criteria, in comparison with the 3.5 in the previous step, are less conservative and

reflect the substantial cost impact of strengthening an existing bridge or restricting traffic,

as well as the shorter future service period expected compared to the 75 years that is

typical for the design of a new bridge [Nowak, 1999; Moses, 2001]. The ratings

determined using the legal loads are generally used as the basis for determining whether

to post or strengthen the bridge.

In certain cases, a permit load rating may be performed to check the safety (and

serviceability) of the bridge for vehicles above the legally established weight limit. This

procedure is only necessary when an overweight vehicle is to use a bridge, and it is only

allowed for bridges that yield RF ≥1 at the previous legal load rating levels. The permit

15
live load factors were derived to account for the possibility of the simultaneous presence

of one or more non-permit heavy trucks on the bridge when the permit vehicle crosses the

span, as well as the site-specific traffic conditions described by the ADTT.

Figure 2.1 Loads and Resistance Factor Rating Procedure [AASHTO LRFR, 2005]

16
 A comparison of the ratings used in the LRFR method (Eq.2.6) with those in the

LFR/ASR method (Eq.2.4) shows three key improvements. First, LRFR attempts to

reflect the in situ bridge resistance systematically and objectively through the use of the

system factor ( φ S ) and the condition factor ( φC ). In the LFR/ASR methods, the condition

of the bridge, its redundancy, and any deterioration at the time of evaluation must be

factored into the estimation of the capacity term (C) in a completely subjective manner.

Second, the LRFR method considers dead load from factory-made members, cast-in-

place members and wearing surfaces separately, with each assigned an independent dead

load factor to account for the different degrees of variability in these components of dead

load (discussed subsequently in section 2.1.2 of this chapter). In the LFR/ASR methods,

all permanent loads are combined in calculating the dead load effect (D), to which an

overall dead load factor is applied; adjustments that might be indicated by available in

situ dead load measurements are difficult to handle in the rating process. Third, the

LRFR method has provided a set of live load factors that ranges from 1.4 to 1.8,

depending on the bridge’s in situ traffic condition indicated by ADTT, for rating

calculations at the legal load level. This improvement allows site-specific traffic data to

be incorporated into the load rating process, which is a major advantage from applying

probability-based structural reliability theory in existing bridge condition assessment and

offers an important enhancement of the LRFR method over the traditional stress-based

rating approaches.

The LRFR method further simplifies the bridge rating process by requiring the

use of the HL-93 design load as the starting point in the rating and as a screening check

for all other AASHTO/State legal loads. The HL-93 live load envelopes all types of legal

17
loads in the United States and provides a uniform reliability check for various span

lengths with just this one load model. Otherwise, to achieve a uniform reliability for

highway bridges using LRFR, rating calculations have to be applied to all three

AASHTO legal loads individually, with each controlling short, medium, or long spans

respectively [NCHRP 12-28, 2001; Minervino, et al, 2004]. In contrast, the HS-20

design load checking used in the ASR/LFR process does not envelope current trucks on

the highway system and the ratings determined with this vehicle do not provide uniform

reliability for bridges of varying span lengths. Finally, permit vehicles that are

significantly heavier than the AASHTO/State legal loads may have very different

configurations. While the LRFR method provides procedures and live load factors

specific to permit vehicles ratings for bridges that have been demonstrated to have

adequate capacity for AAASHTO/State legal loads, the LFR/ASR methods provide no

guidance on permit checking.

Despite these improvements, the LRFR procedure has not been widely adopted

for rating or posting bridges in the United States. A survey of State Departments of

Transportation on bridge evaluation practices (presented in section 2.3) has revealed a

number of issues and concerns with the LRFR method. Addressing these issues will

facilitate the adoption of the LRFR, in a modified form, and provide an improved bridge

rating methodology. Such improvements are the subject of the current research program,

and are presented in Chapter 3, following the survey.

18
 Figure 2.2 LRFD Design Live Loads (HL-93)

2.2.2 Probability Models and Supporting Data for Reliability-based Bridge Rating

As noted previously, the LRFD option in the AASHTO Manual for Bridge

Evaluation [2003] is the first bridge load rating method in the United States to be based

on modern principles of structural reliability and limit states design. The essential

ingredients of a reliability-based design and evaluation include probabilistic models of

the structural resistance and loads and a method for analyzing the reliabilities (or,

conversely, the limit state probabilities) that are relevant to each bridge limit state. This

section provides a brief summary of such methods and tools, as they have been applied to

developing the AASHTO LRFD Bridge Specifications and the MBE and are expected to

19
be relevant to the current research program to develop improved bridge rating methods.

Details are available in the archival literature [Nowak, 1999; Moses, 2001].

Structural Resistance Models

The capacity of a bridge depends on the strength of its components and

connections. The strength, R, is a random variable having uncertainties that fall into three

categories [Ravindra and Galambos 1978; Moses, et al, 1987; Tabsh et al, 1992]: material

properties, M, including material strength, modulus of elasticity, cracking stress and

chemical composition; fabrication, F, including geometry, dimensions and section

modulus; and structural modeling, P, reflecting assumptions and approximate analysis

methods. The mean and coefficient of variation for M, F and P are usually determined by

material tests, simulations, observations of existing structures and engineering judgment.

Figure 2.3 State of Georgia Legal Loads

20
 In the development of the ASHTO LRFD Bridge Design Specifications [Nowak,

1999], R was determined as the product of the nominal resistance Rn and the three above-

mentioned parameters, M, F and P:

R = MFPRn (2-7)

As a product of random variables that are assumed to be statistically independent, the

resistance is modeled by a lognormal distribution with mean, µ R , and coefficient of

variation (COV), VR , computed as follows:

µ R = Rn µ M µ F µ P
(2-8)
V R = (VM2 + V F + V P2 )1 / 2

in which µ M , µ F and µ P are the means of M, F and P and VM , VF and VP are the

COVs of M, F and P, respectively. The statistical parameters of R used in the

development of the LRFD Specifications for different types of structural components

(steel girders, composite and non-composite, reinforced concrete T beams and pre-

stressed concrete AASHTO-Type girders) in different failure modes (bending and shear)

are presented in Table 2.1.

Dead Load Model

Dead load is the weight of structural members, nonstructural components and

attachments, and traffic wearing surfaces. Because of the different degrees of variability,

one must consider the components of bridge dead load from factory-made members (steel

and pre-cast concrete), cast-in-place members (T-beams, slabs), and wearing surfaces

(asphalt) separately. Generally speaking, dead loads can be predicted more accurately

than live loads, as long as accurate records have been kept and the as-built condition

agrees with the available drawings. In the study by Moses and Verma [1987], the bias

21
(defined as the ratio of the mean to nominal load) and COV of bridge dead loads were

taken to be 1.0 and 0.10 respectively. Later in the AASHTO LRFD calibration [Nowak

1999], the dead load was divided into four components and each component was modeled

with a normal distribution. Finally, Ghosn [2000] used 1.0 and 0.09 for the dead load

bias and COV respectively in his study. These components of dead load are listed in

Table 2.2 along with their statistical parameters; the “miscellaneous” category is the dead

load portion from railings and luminaries.

Live Load Model

Bridge live load is produced by vehicles moving on the bridge. Variability in live

load arises from uncertainties in vehicle weight, vehicle position, average daily truck

traffic (ADTT), calculations of live load effect (including distribution of live load to

supporting girders), and the likelihood of several heavy vehicles being on the bridge at

the same time [Moses and Verma, 1987]. Traditionally, the static and the dynamic effects

of the live load are considered separately and assumed to be statistically independent

[Nowak, 1993; 1999].

Based on weigh-in-motion (WIM) data, Moses and Verma [1987] identified

several variables to provide a simplified model for determining the maximum expected

single truck load effect:

M = aW.95 mHIg (2-9)

in which M is the predicted maximum dynamic live load effect; a is a constant which

relates M to a reference loading model (taken as an AASHTO/legal rating vehicle); W.95 is

the 95th percentile characteristic value of 75-year maximum truck weight, assumed to be

a random variable to reflect the possible errors (epistemic uncertainty) in load estimation

22
and site-to-site differences; the variable m reflects the influence of the dominant vehicle

type and configuration at a site; the variable H reflects the overload events due to the

multiple vehicle presence, such as side by side or following vehicles, and also reflects the

probability that truck weight exceeds the 95th percentile in combination with closely

spaced vehicles; variable I is the dynamic impact allowance and variable g is girder

distribution factor. Except for the constant a, all of the variables in Eq. (2-9) are random

variables with statistics based on studies and data collected on a number of sites.

Table 2.1 Statistical Parameters of Component Resistance [Nowak, 1999]

Type of Structure FM P R
λFM VFM λP VP λR VR
Non-composite steel girders
Moment (compact) 1.10 0.08 1.02 0.06 1.12 0.10
Moment (non-com.) 1.09 0.08 1.03 0.06 1.12 0.10
Shear 1.12 0.08 1.02 0.07 1.14 0.11
Composite steel girders
Moment 1.07 0.08 1.05 0.06 1.12 0.10
Shear 1.12 0.08 1.02 0.07 1.14 0.11
Reinforced concrete
Moment 1.12 0.12 1.02 0.06 1.14 0.13
Shear w/ steel 1.13 0.12 1.08 0.10 1.20 0.16
Shear w/o steel 1.17 0.14 1.20 0.10 1.40 0.17
Prestressed concrete
Moment 1.04 0.05 1.01 0.06 1.05 0.08
Shear w/ steel 1.07 0.10 1.08 0.10 1.15 0.14

Table 2.2 Statistical Parameters of Dead Load [Nowak, 1999]

Component Bias Factor C.O.V

Factory-made members 1.03 0.08
Cast-in-place members 1.05 0.10
Asphalt 3.5 inch* 0.25
Miscellaneous 1.05 0.10
* mean thickness

23
 The live load model used to calibrate the AASHTO LRFD Bridge Design

Specifications is based on the weigh-in-motion data of 10,000 trucks taken at a site in

Ontario in 1975, which included axle weights, gross weight and axle spacing for each

vehicle [Nowak, 1999]. These 10,000 data points were assumed to define the upper 20%

of the truck traffic at the site over a period of about two weeks. By finding the maximum

bending moment and shear forces for each Ontario truck on different spans ranging from

10 ft (3 m) to 200 ft (60 m), the cumulative distribution functions (CDFs) of live load

effect for various span lengths were obtained. Bridges with both simple spans and two

continuous equal spans were considered. These CDFs were then extrapolated to a full

lifetime (75 years) consisting of some 75 million truck load events and the 75-year

maximum live load was fitted by a normal distribution.

Static and dynamic load effects were studied separately [Tabsh and Nowak,

1991]. On the basis of a finite element study of bridges with various span lengths, it was

found that the ratio of the mean value of the 75-year maximum live load (without

dynamic impact) to nominal (HL-93) live load is dependent on the bridge span and its

COV is about 12%. The study also concluded that dynamic impact was dependent on

three major factors: bridge dynamics, vehicle dynamics and road roughness; the mean

value of the dynamic load factor does not exceed 0.15 for a single truck and 0.10 for two

trucks side by side, and its COV is about 80%. For the static and dynamic combined load

effect, the mean of this 75-yr maximum live load with respect to the design load model

(HL93 in Figure 2.1) fell in the range 1.0-1.2, depending on span length, and the COV

was found to be about 0.18.2

2
Imai and Frangopol (2001) found that the maximum bridge live load was best modeled by a Type I
distribution of extreme values. Bhattacharya et al. (2006) also found that the Type I distribution fits the

24
2.3 SURVEY OF BRIDGE RATING PRACTICES IN THE UNITED STATES

2.3.1 Development of Survey of Bridge Rating Practices

As part of NCHRP Project 12-46 that developed the AASHTO LRFR Guide Manual

[2003], a survey questionnaire had been mailed to State Bridge Engineers in May, 1997,

asking for current practices and views on technical issues pertaining to the inspection,

evaluation and load rating of bridges. The responses to this questionnaire were valuable in

developing the rating criteria in the AASHTO LRFR Guide Manual. However, in the

intervening years, the state of bridge evaluation practices in the United States has continued

to evolve. Accordingly, a follow-up questionnaire was prepared that requested additional

information on a subset of topics covered in the older survey, with specific emphasis on

bridge capacity evaluation practices that may have changed in the intervening years and

would be of particular interest to the current research to develop improved bridge rating

procedures and a set of Recommended Guidelines. The questionnaire was sent out to all

states in November, 2005, and after four months, forty one responses (Table 2.3) were

received and reviewed. A copy of this survey questionnaire can be found in Appendix A

to this report.

The synthesis of the survey responses in Section 2.3.2 is presented without

identifying the state or the respondent. These responses often were presented in sentence

fragments; in that case, an attempt has been made to complete the view expressed in the

comment with a minimum of editing. The survey questions fall into several general

categories: when to load rate a bridge, when to update existing ratings, how to rate, when

experimental measurements of live load effect properly. Finally, Galambos, Ellingwood et al (1982) used
the Type I distribution to model the 50-year maximum live load for building structures.

25
to post, and other performance issues (connections, fatigue, and scour). The following

synthesis of the survey responses is organized around those categories.

Table 2.3 Responding States

2.3.2 Synthesis of Survey Response

When to rate?

In order to comply with FHWA regulations all states either perform a load rating

analysis, or make a professional judgment as to the load capacity of their bridges. Most

states are working toward 100% load rating, and most of those responding reported to

have rated between 80% and 100% of their bridges. This intention is summarized by the

response from a Western state: “Our goal is to rate all state owned bridges to determine

the maintenance requirements and bridge load carrying capacities and to comply with the

National Bridge Inventory System (NBIS). Also modeling all bridges will help in

26
overload permit evaluations.” Only five states have 60% or fewer of their bridges rated.

One Western state is in the process of updating all of its ratings to include both Inventory

and Operating and presently has over 90% rated at Inventory levels and approximately

5% rated at Operating levels. It is the policy of most states to rate all new bridges when

they are designed or constructed. Existing unrated bridge structures are being evaluated

and rated, as circumstances and resources permit. The rating of existing bridge structures

in general begins with those for which design documents are available, and then

continues to bridges without them. The rating of bridges without plans is typically

performed in one of four ways: using plans from a similar bridge built at about the same

time; by load testing the bridge; using results of load tests from a similar bridge structure;

or by professional judgment.

When to update ratings?

As to when to update existing load ratings, the following is quoted from the

response provided by a Midwestern state and is indicative of other responses:

1. There is a physical change in the condition of a bridge or a structural member,

e.g., physical alteration in the structure; new beam or new deck, rusting or

spalling or damage occurred to the structural member(s) resulting in section loss;

change in the wearing surface; change in the super-imposed dead loads; excessive

deflection or settlement observed; occurrence of an accident;

2. There is a request to re-evaluate the rating of a structure for a vehicle different

than what was previously used such as for single trip permit load ;

3. There is a change from the method of analysis used for previous rating;

4. Special circumstances dictate re-analysis of the structure.

27
 5. There is a change of the rating method (e.g. switch from ASR to LFR), rating

software or the truck weight regulations.

A western state remarked that all their load ratings are being recalculated because: “the

previous ratings were done by different individuals and are not consistent.” Some other

states have specific policies on this issue such as revising load rating when “overlay

changes more than 2 inches”, “steel section losses are more than 1/16 inches” or “primary

member condition rating on the inspection report has changed by more than one point

since last routine inspection if the initial rating was 5 or lower.”

What method to use in rating?

Thirty one (31) of the responding states reported that the LFR method is their

primary rating method, but that they occasionally used the ASR method in cases where

the LFR method does not appear to be applicable. The remaining ten (10) states reported

to use a combination of ASR, LRF, and LRFR depending on what specifications

governed the design of the bridge.

Nine (9) of the responding states reported that they were currently using LRFR on

either all of their bridges or those bridges designed by AASHTO LRFD Bridge Design

Specifications. Five (5) states were in the process of evaluating the suitability of LRFR,

and were planning a transitioning from LFR to LRFR. Most of the remaining states cited

either the lack of resources or readily available software as reasons for not making the

transition, but noted that they would change to LRFR if mandated. Several specifically

28
said they were waiting for VIRTIS3 to offer a version that incorporated the LRFR method

before they considered switching from LFR.

Several states which have considered transitioning to LRFR raised some

significant questions and issues. One Midwestern state suggested that “the proposed

updates to the guidelines do not inspire confidence in the manual.” A Western state

responded “we are concerned with the high load factors; if we can not lower these factors

through WIM (weight in motion) data, we may use older load rating methods on older

bridges.” Two other Western states simply stated “it was too uncomfortable with the

LRFR method to use it” and “not fully confident in this document.” The strongest

opposition to transitioning to LRFR came from an Eastern state, which observed: “Too

much work for no value. Ratings for concrete and timber do not correlate to real world.

For timber, LRFR requires a “fudge” factor to get reasonable results for posting. For

reinforced concrete bridges, the change from ASR to LFR resulted in a reduction of

approximately 20% in posting values and changing from LFR to LRFR will result in

another 15% to 20% reduction in the posting limits. On the other hand, with LFR and

LRFR, “posting values for steel bridges increase.” This Eastern state also had serious

questions as to the applicability of LRFR and its ability to perform its main function of

providing a uniform reliability for all bridge structure types. A similar concern was

expressed by a Midwestern state, which also doubted whether LRFR was suited for all

bridge types. These apprehensions about the transition from the older methods to the

LRFR method warrant further investigation. A summary of an investigation aimed at

examining the differences in ratings using these methods through illustrative rating

3
VIRTIS is a bridge rating software package, developed by Cambridge Systematics, Inc.

29
calculations performed for four sample bridges selected from the Georgia bridge

inventory will be presented in Section 2.3.3.

Most of the states using LFR employ the HS 20-44 vehicle for both Inventory and

Operating ratings at the design load level. Some states use the full set of AASHTO Legal

vehicles, HS-20, H-20, type 3, type 3-3, and type 3S2, for legal load ratings. In some

other states, the AASHTO vehicles are modified and designated as “state legal loads.”

These modifications typically consist of a scaled-up load and/or a redistribution of the

load between the cab and trailer. There are also a few states with unique legal loads, such

as logging trucks or other highly used regional vehicles.

When to post?

Answers to the survey question regarding the decision as to when to post a bridge

had the widest variation of any of the answers. Twenty (20) of the responding states

reported that they post a bridge when its Legal loads exceed the Operating level rating.

Georgia and four other states use the Operating rating as the posting limit for bridges on

the state system and the Inventory rating for bridges on the local system.

Some other states have more detailed policies regarding the posting limit, such as:

“use Operating rating for bridges having a condition rating larger than 5, otherwise, use

Inventory level rating”; “for fracture critical member use Inventory rating , for others use

some value in between the Operating and Inventory levels based on engineer’s

judgment”; “use Operating rating for concrete members and the average of the two for

steel members ”; or, “post when the Operating rating is exceeded, or when the Inventory

rating is exceeded and posting will have minimal economic effects” One Eastern state

specified that all structures need to be considered for posting if the structure’s Inventory

30
capacity rating is less than 30 tons for HS20 vehicle, 35 tons for 3S2 vehicle or 18 tons

for the H20 vehicle, or when the gross tonnage of a “4 Axle” vehicle exceeds the

structure’s Operating level capacity.” In another Eastern state, a bridge will not be posted

if “the bridge can carry H15 at Inventory level and HS20 and all state Legal loads at the

Operating level.” Several states don’t have specific criteria for posting, but will consider

it if the structure has a rating factor less than 1.0 at the Inventory level for HS-15 vehicles

or if the structure shows signs of major deterioration. There is no consensus among the

states as to whether to post a bridge at Operating level or Inventory level ratings.

Engineering judgment sometimes is used either to post a bridge whose rating would not

normally entail posting, or to not post a bridge that is calculated to require posting.

As to what percentage of the state bridge inventory has been posted, twenty (20)

of the responding states reported posting fewer than 4% of their bridges, fourteen (14)

reported that between 5 to 19% were posted, and the remaining seven (7) have posted

over 20% of their bridges. This survey question was poorly phrased, however, leading

some states to report the total percentage of posted bridges while others reported the

percentages of state and local bridges separately. The percentage of posted bridges on

local roads is typically anywhere from 10 to 100 times the percentage of posted bridges

on state roads.

As to whether serviceability or fatigue limit states are considered when setting up

the posting limits, twenty four (24) states, including Georgia, do not consider either;

sixteen (16) consider serviceability, and the vast majority of states generally do not

consider fatigue. Those that consider serviceability do so only for steel or pre-stressed

concrete girder bridges.

31
When to load test a bridge?

Fourteen (14) out of 41 of the responding states, had performed some form of

load testing for the purpose of load rating as a part of bridge evaluation practice. Five (5)

other states reported that they had once performed very few load tests for the reason of

academic research only. The remaining states have never used load testing as a tool for

bridge condition assessment; one Eastern state remarked that the reason is “testing is too

time consuming and expensive.”

Most of the load tests have been performed on structures that were in good

condition but required posting according to standard rating analysis, on special

construction such as FRP bridges, on those bridges without available plans or design

documentations, or on those with serious deterioration that prevented an accurate

theoretical strength calculation. One Western state noted that they performs test on

bridges “deemed to be high risk, or fracture-critical.” The benefit of load testing results

is best summarized by the response from one Western state:

1) To allow bridges to remain in service without traffic restriction

2) To avoid unnecessary repairs and needless replacement

3) To avoid repairs to bridges scheduled for replacement

4) To get more accurate load distribution factors, and

5) To compare calculated stresses with actual stresses

One other common use of load testing is in evaluation of overload permits. Two of the

states that perform load tests do so extensively to prevent having to perform costly

repairs, replacement or posting due to “unreliable AASHTO rating factors.”

32
 Only one Southern state among the fourteen states that performed load tests used

the provisions in Chapter 8 of the AASHTO LRFR Guide Manual (2003) to guide their

load testing practices, although there is one other state that “follows NCHRP, Nov 1998-

No.234, Manual for Bridge Rating through Load Testing, which is consistent with

Chapter 8 of the LRFR Guide Manual.” One Western state reported that the reason for

not using the LRFR Guide Manual is that “we are not yet sufficiently comfortable with

it.” Two other western states, having performed load tests prior to the issuance of the

Guide Manual, have also developed their own guidelines and testing procedures, which

were reported to be in the process of being compared with the LRFR Guide Manual.

Some states perform and analyze the load tests themselves, while states that do not have

their own guidelines usually leave the testing and interpretation entirely to the

Universities to which they contract the work. One Eastern State “uses the load test to

determine live load distribution, which is then applied to LFR formula to update load

rating factors.” Another Western state has a load testing protocol that involves taking

“strain transducer measurements when the structure is under various loads. A model of

the bridge is produced based on the strain transducer measurements. This model is then

used to predict responses of the bridge to design loads and over-loads.”

Other performance issues - connections, fatigue and scour

Thirty seven (37) of the forty one (41) responding states do not assess the capacity

of connections on a regular basis. Connections are routinely inspected in most states;

however, they are checked for adequate capacity only if engineers suspect that the

connections may govern the load rating of a bridge. For the four exceptions, one Eastern

state stated that: “Our policy requires load rating of connections for all primary

33
components of a bridge unless the district Bridge Engineer concludes that the connections

would not control the rating of the member.” A Western state does consider connections,

but only those on continuous bridges with a splice at the piers; an Eastern state considers

all types of connections, while another Western state examines “all areas of the

structure.”

Most states normally do not compute the remaining fatigue life of a bridge unless

fatigue cracking is found during inspections, with the typical reason being lack of

sufficient truck volume data. Four states are exceptions. One Eastern state performs a

100% hands-on inspection of fatigue sensitive members; however, one can avoid this by

calculating the fatigue life of bridges with low traffic counts, and then perform 100%

hands-on inspection if the member has a remaining fatigue life of less than 10 years;

One Western state “computes remaining fatigue life based on an arms length inspection”

and performs such analysis on a 1 to 10 year cycle where the interval is usually 3 years

for fatigue prone members as determined by fatigue life; Another Western state performs

an in-depth inspection of all fracture critical members regardless of fatigue life, however,

when the remaining fatigue life is finite or expired, the frequency of inspections

increases. Finally, one Northern state performs fatigue analysis on selected bridges.

All states indicated that they perform some form of scour investigations on a

regular basis. Most investigate scour for bridges that cross wade-able waterways during

the FHWA-mandated 2 year inspections and all other bridges during a special underwater

or scour investigation every 4 to 5 years. Two states report that they perform special

scour investigations on any bridges identified as scour-susceptible following floods.

34
2.3.3 Comparison of Rating Methods through Sample Bridges

The survey of current bridge rating practices of State Departments of

Transportation, summarized in Section 2.3.2, revealed considerable differences in current

practices and concerns that the ASR, LFR and LRFR methods yielded substantially

different ratings. The AASHTO MBE allows three rating methods but does not provide

guidance as to which method should be used for specific circumstances. It is apparent

that such discrepancies would be a barrier in routine bridge rating practices.

To determine the extent to which such discrepancies might exist and to quantify

the magnitude of the rating differences that might result from the use of ASR, LFR and

LRFR methodologies for typical Georgia bridges, a rating analysis with these three

methods was performed for four typical bridges that had been identified for subsequent

load testing and advanced analysis. A detailed description of the engineering

characteristics of these sample bridges are provided in Chapter 4. Tables 2.4 and 2.5

present a summary of these rating results for flexure and for shear respectively [Wang, et

al 2009].

In general, rating results by ASR and LFR are reasonably close in all cases. The

LRFR legal load ratings for the HS20 vehicle fall between the Inventory and the

Operating level ratings computed by either the LRF or ASR method for both moment and

shear for all four bridges. In other words, the LRFR legal level ratings generally are more

conservative than the LFR/ASR Operating level ratings and more liberal than the

LFR/ASR Inventory level ratings. These results for typical Georgia bridges are consistent

with what was found in the Survey,

35
Table 2.4 Summary of Sample Bridge Flexural Rating for Interior Girders

Rating Vehicle
Bridge
Method HS20 H93
Type
Inv Opr Inv Opr
ASR 0.70 1.25
Conc. T
LFR 0.75 1.25
(Straight)
LRFR 0.93 0.65 0.84
ASR 1.36 2.17
Conc. T
LFR 1.16 1.93
(Skewed)
LRFR 1.77 1.27 1.65
ASR
Prestressed
LFR 1.54 2.57
Girder
LRFR 1.95 1.34 1.73
ASR 0.82 1.33
Steel Girder LFR 0.71 1.18
LRFR 1.08 0.72 0.93

Table 2.5 Summary of Sample Bridge Shear Rating for Interior Girders

Rating Vehicle
Bridge
Method HS20 H93
Type
Inv Opr Inv Opr
ASR 0.41 0.75
Conc. T
LFR 0.43 0.72
(Straight)
LRFR 0.61 0.45 0.58
ASR 0.94 1.44
Conc. T
LFR 0.84 1.40
(Skewed)
LRFR 1.05 0.83 1.08
ASR
Prestressed
LFR 1.43 2.39
Girder
LRFR 1.47 1.05 1.36

36
2.4 BRIDGE EVALUATION IN FOREIGN COUNTRIES

2.4.1 Bridge Rating in Canada

The provisions of Section 11 of the Ontario Highway Bridge Design Code

(OHBDC), 3rd edition pertain to the evaluation and posting of existing bridges other than

soil-steel structures4 and pedestrian bridges. Provisions are given for the condition

inspection, analytical load rating procedure, load testing and calculation of posting limit

for bridges. In contrast to the requirements in the United States, evaluation is not

mandated for every highway bridges and is not required on a periodic basis in Canada.

The OHBDC is based on the limit state design philosophy and a target reliability

index, β, of 3.5 is used for both design and evaluation. There is no explicit reduction of β

in evaluation, although a few adjustments can be applied to reduce those load factors

used for design when evaluation is performed. Provisions are provided for ultimate,

service and fatigue limit states checking, and only the ultimate limit state is specified to

be used for determining the load carrying capacity, stability and load posting of bridges;

the exceptions are masonry abutments, masonry piers and masonry retaining walls, for

which serviceability is the governing limit state. Fatigue checks are performed only if the

bridge owner wants to assess the remaining life of the bridge because of the observation

of the physical evidence of fatigue-prone details or fatigue related defects. The method

of fatigue life assessment is the same as in the AASHTO LRFR Guide Manual.

The rating process requires the use of three live load models, designated OHBEL

levels 1, 2 and 3 respectively, with different gross magnitudes and configurations. These

4
Defined as a bridge comprised of bolted structural steel plates and engineered soil, designed and
constructed so as to utilize structural interaction between the two materials. The OHBDC devotes an entire
chapter to this type of bridge, as it does for concrete and steel bridges.

37
three live load models appear to be similar to the three AASHTO legal loads. The live

load factors calibrated for bridge design are adopted in the capacity evaluation for most

general cases, with some exceptions: the live load may be reduced by 10% for bridges

with inspection intervals less than 5 years; the corresponding lane load equivalent used

in evaluation is reduced as a function of in situ traffic volume or varies according to the

road classes; live load factor may be reduced for multiple lane bridges with a certain level

of redundancy, and the dead load factor can be reduced if the nominal dead load is

carefully estimated. These reductions are not applied in controlled vehicle rating, which

is a procedure that is comparable to the AASHTO permit load checking, and is conducted

for specific vehicles for which permission must be granted prior to their using the bridge.

The OHBDC presents detailed provisions and curves for establishing posting

limits according to the rating calculations performed for the three above-mentioned live

load models. The provision regarding posting concrete bridges is similar to that in the

AASHTO LRFR Manual, that is, a concrete bridge need not to be posted if it has been

carrying normal traffic without signs of excessive cracking or deformation.

The OHBDC also states that a load test may be proposed as a part of the

evaluation procedure when the analytical approach does not accurately reflect the actual

behavior of the bridge. However, no detailed definitions and provisions are provided as to

different types of load tests, loading patterns, instrumentation or interpretation of test

results.

2.4.2 Bridge Rating in the United Kingdom

Document BD 21/01, Assessment of Highway Bridges and Structures, adopts a

limit state format with appropriate partial safety factors for condition evaluation of most

38
highway bridges except for cast iron bridges and masonry arch bridges. It is stipulated

that bridges built after 1965 should normally be evaluated for serviceability as well as for

the ultimate limit states; bridges constructed before 1965 do not need to be assessed for

service limit states. Requirements for fatigue endurance however are not included in the

standard and the reason stated is that the past stress history of each structure, which could

profoundly influence fatigue limit checking, cannot generally be determined to the

accuracy level required for assessment.

No reduction in target reliability index from the corresponding values for design

is explicitly stated in BD21/01; however, several adjustments are made to the live load

model that have the effect of reducing the level of conservatism in the evaluation of

existing bridge structures. In the UK, the bridge design live load model consists of a

uniform distributed load (UDL) and a knife edge load (KEL) with the intensities of both

components decreasing with bridge span lengths. The design load was derived by

estimating the worst credible values of relevant loading parameters from available

statistics. Adjustments are suggested in the evaluation to scale down this design load

model for bridge situations that are less onerous than the above worst case scenario,

while maintaining a consistent reliability level for the whole network; detailed scaling

curves for the live load adjustment factors are provides in the document. Furthermore, in

the absence of definite information about material characteristics in estimating the

resistance of bridge component, the document assigns a set of values to materials which

should be used in the initial assessment, mostly according to the construction period of

the bridge. Structures which cannot sustain the assessment live loading after the

checking calculation, and which are not scheduled for immediate replacement or

39
strengthening should be reevaluated for the other three lower live load models for

posting; posting provisions can be found in the document.

Document BA 54/94, Load testing for bridge assessment, presents general

instructions on load testing practices. The document states that the role of load testing

primarily is to seek out the hidden reserves of strength, and the bridges most likely to be

involved are those which contain features where such reserves may be found. Load tests

are broadly divided into the two categories - Proving load tests and Supplementary load

tests - which are analogous to the AASHTO Proof load test and Diagnostic load test,

respectively. Because there is a risk of collapse during a proving test, or of damage to

essential elements of the structure, such tests therefore are limited in the document only

to those bridges which, on the basis of their analytical assessment, would have been

closed to traffic or demolished. Bridges that previously have been subjected to proving

tests need to be thoroughly inspected and reassessed at more frequent intervals. The

document also emphasizes that extreme care has to be taken to extrapolate the results of

tests carried out with fairly low levels of loading to those likely to occur at the ultimate

limit state.

Instructions provided in BA 54/94 are rather general; detailed guidance on

loading patterns and magnitude, testing procedures, and test results interpretations are not

provided. Cautionary notes are provided concerning the effectiveness and the accuracy of

load testing as a means of load capacity evaluation of existing bridges. Concerns

expressed include: whether a static test load can adequately represent the ultimate limit

state loading condition; whether a bridge deck should be fully loaded or partially loaded,

in view of the fact that the collapse mode of a partially loaded deck may be different from

40
that when the whole deck is loaded as was intended in the design; and whether the benefit

of a test is warranted, considering the risk to personnel.

2.4.3 Bridge Rating in Australia

Section 7 of the Australian Bridge Design Standard provides rating guidelines

with a commentary. The concept of rating is based on the limit state design philosophy

and both serviceability and ultimate limit states are considered. The ultimate action is

defined as an action that has a 5% probability of being exceeded during the design life,

which represents an average return interval of 2000 years; while the survivability action

is defined as one having 5% probability being exceeded per year, corresponding to a

return interval of 20 years.

The rating for strength is carried out for all strength limit states, e.g. moment,

shear, compression, at all potential critical sections, with the lowest rating factor

determined being the rating factor for the bridge. At the service limit state, a structure is

checked for vibration and deflection. When a bridge is checked for the fatigue limit state,

the cumulative fatigue damage at the critical details of the bridge must be carefully

assessed, from which the nominal fatigue life of a bridge can be estimated. For the

purpose of rating, the cumulative fatigue damage is defined as the sum of the damage in

all previous years; the nominal fatigue life is considered having been reached when the

cumulative damage sums to unity5, in which case, a program of inspection should be

initiated to ensure that fatigue cracks are detected and suitably repaired before they

endanger the bridge’s ability to carry its applied loads.

5
Cumulative damage is assessed using the Palmgren-Miner linear damage accumulation model.

41
 A bridge may be rated at each limit state, to a general rating capacity or to

specific loading cases, using the same partial factor checking format as specified in the

Standard. For the general rating case, which is comparable to the AASHTO design load

level rating, the live load models and the corresponding load factors are the same as those

used in the design of a new bridge. For specific loading cases, the live load can either be

a legal load vehicle or an exceptional load, the former case being comparable to the

AASHTO legal load level rating, and the latter being comparable to permit load

checking. In all cases, the effects of the rating loads for the specific loading cases are

determined using the gross weight and the configurations specific to the vehicles under

consideration. Since the possibility of overloading at this step is unlikely, a reduced live

load factor is permitted. Where the rating for a bridge is less than required for current

general access vehicles, consideration shall be given to applying a posted limit on the

bridge. Detailed regulation on establishing the limits for specific vehicles however is not

presented in the document.

Two types of nondestructive test are defined in the Australian Standard: static

proof load test and static performance load test, which are comparable to the two types of

AASHTO load tests. The difference between the two types of test is in the magnitude of

loading, and in the manner and the level of confidence in which the capacity of the bridge

to carry the live load is determined from the test results. The Standard identifies the load

test as an effective method of evaluating the performance and structural capacity of a

bridge or bridge type. The document suggests that the proof test loading should be

applied incrementally from a base load of 50% of the theoretical rated ultimate capacity

in order to protect the bridge and the testing personnel, and the load response should be

42
continuously monitored to ensure that the bridge is behaving in an elastic manner. It also

suggests that a numerical model of the structure should be developed prior to the test to

assess the ultimate capacity, failure mode and the elastic limit under different loading

configurations and to determine the maximum load needed for the test. The Standard

provides some detailed formulas for updating ratings after a successful load test and also

emphasizes that the adoption of the load testing results should only to apply to bridges of

similar structural form, taking into consideration material properties and conditions.

2.5 CRITICAL APPRAISAL OF CURRENT BRIDGE RATING PRACTICE

Three different existing rating methods are currently utilized by state Departments

of Transportation in the United States in their bridge rating work, as revealed in the

survey responses. The ASR, LFR and LRFR methods are based on different design

philosophies and therefore often produce different rating results and lead to different

posting limits for the same structure; sometime the difference can be significant. One

weak point of the current practice is that no clear policy is provided in the AASHTO

rating guidelines as to which method should be used for specific circumstances. The

LRFR method is relatively new; while having the most rational basis of the three, it still

need to be tested and validated through research and practice for the bridge engineering

community to develop confidence in its use. The large number of Inventory, Operating,

and Legal loads clutters the analysis and rating process with many redundant

calculations, especially in the ASR and LFR related procedures. Consequently, this

situation causes differences in interpretations and practices from different DOTs

regarding what triggers posting and whether to use an Inventory or Operating rating to

43
post and in which circumstances. These issues should be better stipulated in rating

guidelines for safe practice and for consistent and unambiguous implementation.

None of the current AASHTO manuals provides clear guidance as to when to

revise existing load ratings. Therefore states DOTs normally make their decisions on

revising a current rating based on judgment and on what has been observed during the

field inspection. Most bi-annual inspections are visual and any insight that might be

obtained from such an inspection on existing safety or load-carrying capacity necessarily

will be qualitative rather than quantitative in nature. Either a way must be found to better

quantify what a visual inspection reveals or a more sophisticated inspection strategy,

including informative and non-invasive inspection technologies and optimal inspection

intervals, should be encouraged, so that the decisions based on inspection data are well-

substantiated.

The survey of the state Departments of Transportation also revealed that most

states rely solely on analytical methods to evaluate the load-carrying capacity of existing

bridges. Load testing as an effective alternative has been largely ignored. Due to the

conservative nature of the analytical rating methods, this inevitably leads to some

unnecessary bridge repairs, replacements or postings. The fact that most state DOTs do

not perform any kind of load testing likely can be traced to a lack of guidance on load test

to address practical issues including: under what circumstances a load test will be a good

option for bridge rating, and under what circumstances one should choose a diagnostic vs

a proof load test, and further, how to design practical load test procedures. Engineers

should be provided general guidance as to whether a load test is worthwhile, considering

the cost of a test, as well as specific instructions on field data acquisition and

44
interpretation. The current load test guidelines in Section 8 of the AASHTO LRFR Guide

Manual (2003) do not provide engineers with enough details to bridge the gap between

the concept and practice of load testing.

In the current bridge evaluation practice in the US, two major parameters used to

describe a bridge’s present condition are the condition rating (on a scale of 1 to 10) and

capacity rating (on a scale of 0 to 1). The condition rating is based on visual inspection

data and measures deterioration level of a bridge. The capacity rating, on the other hand,

is performed in practice according to inventory data to check whether a given bridge

meets the current design standards. One weak point of current practice is that the

bridge’s capacity rating is computed based on its design documentation and is not

properly coupled with the bridge’s deterioration state. Bridges are inspected periodically

and the condition rating assigned by inspector on the basis of visual inspection identifies

whether deterioration is occurring and, ideally, at what level. These inspection data

clearly should be considered in computing the capacity rating and real time reliability of

the bridge.

2.6 CLOSURE

In general, modern bridge rating procedures worldwide have adopted reliability

principles as their basis. They have utilized the limit state philosophy to allow the safety

checking to be performed in a deterministic manner without an explicit structural

reliability assessment. The reliability indices for design are typically 3.5 or higher over

the lifetime of the bridge. However, they permit lower reliability indices in the context of

specific evaluations of individual existing bridges, either by explicitly reducing the target

safety index in the calibration leading to the load factors for evaluation, which are lower

45
than those used in bridge design, or by directly scaling down the live loads used in the

assessment to reflect the lesser requirements for evaluation compared to the design level.

The ultimate limit states are typically required as the governing limit states for

safety checking for majority of the bridge types; serviceability and fatigue are not

regularly mandated unless signs of distress or fatigue related defects are observed. Rating

procedures and the assessment live load models vary the most from country to country,

but for the most part, a check on design load is typically performed prior to the capacity

estimation respect to actual vehicles; the latter, in general, is the basis for posting. The

view towards load testing is different from country to country, which leads to different

treatment of the provisions on this subject in different guidelines. Test protocols and

details that are critical for a load test to be successful and informative may not be

addressed. The AASHTO MBE (2008) has the most comprehensive provisions on load

testing of the condition assessment guidelines reviewed.

46
 CHAPTER 3
IMPROVED FRAMEWORK FOR BRIDGE RATING PRACTICE – A
PROTOCOL WITH THREE LEVELS OF ASSESSMENT

The LRFR option in the AASHTO Manual for Bridge Evaluation extends the limit

state design philosophy to the bridge evaluation process in an attempt to achieve a

uniform target level of safety for existing highway bridge systems. However, the current

capacity rating formulation in the MBE only supports capacity evaluation at one level of

sophistication (considering individual members, with system effects addressed indirectly

through a “girder distribution factor”) and treats existing bridges as “generic” structures

when, in fact, they have individualized features that contribute to capacity. The

uncertainty models of load and resistance embedded in the LRFR rating format represent

typical values for a large population of bridges involving different materials, construction

practices and site-specific traffic conditions. Although the LRFR live load model has

been modified for some specific cases (average daily traffic, redundancy), the bridge

resistance model can be better “customized” for an individual bridge by incorporating

available site-specific knowledge to reflect the fact that each bridge is unique in its as-

built condition. A rating procedure which does not incorporate in situ data properly may

result in inaccurate ratings (and consequent unnecessary rehabilitation or posting costs)

for otherwise well-maintained bridges, as indicated by many load tests (Nowak and

Tharmabala, 1988; Bakht and Jaeger, 1990; Moses, et al, 1994; Fu and Tang, 1995;

Faber, et al 2000; Barker, 2001; Bhattacharya, et al, 2005). Advancing the current bridge

evaluation practice requires better understanding of bridge system behavior, better

47
utilization of available in situ data as well as better modeling of the live load process and

other time-dependent factors such as fatigue, corrosion and concrete aging. This research

develops an improved bridge evaluation framework that directly addresses the

deficiencies in current practice noted in Chapter 2.

The improved practical rating framework illustrated in Figure 3.1 has three levels

of assessment of increasing complexity. In the first level, the deterministic format of the

AASHTO LRFR method is retained, but the correlation between visually determined

condition rating and the capacity evaluation is established, so that the bridge deterioration

can be taken into account quantitatively in the load rating process. If a bridge does not

pass the first level checking, the second level assessment is optional. At this level, site-

specific data about individual structural properties obtained from material tests,

diagnostic load test and from in-depth structural analysis can be incorporated in rating

calculations. Information needed for this second level assessment usually doesn’t require

significant cost investment and the checking format can still be presented in a

deterministic manner. Finally, if a bridge exhibits unsatisfactory capacity ratings after

the first two levels of assessment and if a further comprehensive evaluation is believed to

be warranted, the third level of assessment rates the bridge at system level by proof load

testing or by using information about its past performance. The proof load test can either

be conducted on the real structure in situ, or it can be done “virtually” through a finite

element analysis that is sufficient to describe the load-carrying mechanisms affecting the

bridge’s capacity and its load ratings. This third level of assessment involves a structural

system reliability analysis which inevitably requires some level of probabilistic modeling

and therefore is not given in a deterministic format.

48
Figure 3.1 Improved Bridge Rating Framework

49
 At the first level of the proposed rating framework, bridge condition rating as a

measure of bridge’s deterioration status is factored into bridge capacity rating metric.

Statistical data on bridge condition rating history taken from the National Bridge

Inventory (NBI) is analyzed, and quantitative models of material degradation and bridge

degradation developed in other research studies are incorporated, leading to a quantitative

correlation between the condition rating history of a large bridge population and the

bridge degradation modeled stochastically. Using statistics presented by this correlation

the bridge condition rating is incorporated in bridge capacity rating process through a

simple condition factor with values that are consistent with the reliability requirement

embedded in AASHTO LRFR.

The second level of assessment is designed to consider the uniqueness of each

existing bridge and to incorporate its in situ information at the component level in the

bridge rating framework. For example, material strengths in situ may be vastly different

from the standardized or nominal values assumed in design and current rating practices

due to strength gain of concrete on one hand and deterioration due to aggressive attack

from physical or chemical mechanisms on the other. In situ load distributions in the

bridge structure observed from diagnostic load test or a simplified FE analysis are often

significantly different from the girder distribution factors specified in the code.

Occasionally, certain bridge components might have a load carrying mechanism that can

not be appropriately modeled by traditional code provisions. Proper consideration of

these factors is likely to contribute to a more realistic capacity rating of existing bridges

and practical guidance for doing so is required. The second level of rating analysis

clearly highlights the learning process of a given bridge through field inspections and

50
provides incentives to obtain quantitative in situ measurements through modern non-

destructive evaluation (NDE) techniques, such as Carpenter hammer sounding, Schmidt

rebound hammer and ultrasonic pulse velocity.

In the third-level evaluation, a bridge is evaluated at its system level in situations

where the first two level of analysis produce low ratings, or analysis is difficult to

perform due to deterioration or lack of documentation. Bridge system reliability can be

estimated by incorporating results from a properly conducted proof load test or

information regarding performance collected in past routine periodic bridge inspections.

The feasibility of using finite element modeling, validated through either systematic field

inspection supported by NDE technologies or through diagnostic load tests, to conduct

“virtual” proof load tests of bridge systems, is investigated in this research. An

examination of the role and limitations of proof load testing is performed using FE

models that have been validated by diagnostic load tests on several bridges believed to be

representative of those bridges of the most concern of GDOT. Additionally, satisfactory

bridge performance history provides information comparable to what is learned from a

proof load test could also be favorable to the bridge capacity evaluation. This

information is currently neglected in the load rating process and will be incorporated in

the third-level assessment of the proposed evaluation framework.

51
 CHAPTER 4
SELECTION, TESTING AND FINITE ELEMENT MODELING OF
THE SAMPLE BRIDGES

In order to accomplish the study objectives stated in Chapters 1, four bridges that

are representative of older bridges typically of concern with regard to rating or posting

were identified from an examination of the Georgia bridge inventory. These four

bridges subsequently were used in an integrated program of analysis and load testing to

support the recommended improvements to the bridge rating process in the Chapter 3

framework.

4.1 SELECTED SAMPLE BRIDGES FOR TESTING AND ANALYSIS

The Georgia Bridge Management System database was examined carefully to

determine general characteristics of bridges that would be candidates for posting. Of the

approximately 2,000 Georgia bridges that require posting, 77% fall into one of three

categories:

• Reinforced concrete T-beam bridges, representing 21%;

• Steel girder bridges, representing 53%; and

• Pre-stressed concrete I-girder bridges, representing 3%.

Figure 4.1 shows the primary structural types of bridges constructed over each decade

from the 1940’s to the present. Figure 4.2 identifies the number of bridges from each

category that have been posted as unfit for some or all of the state legal load vehicles. Of

those posted pre-stressed bridges, 57% were constructed after 1980; in contrast, only 2%

52
Figure 4.1 Bridge Categories Identified by Decade of Construction

Figure 4.2 Posted Bridges Identified by Decade of Construction

53
of the posted reinforced concrete bridges and 10% of the posted steel girder bridges were

constructed after 1980. Seventy-six percent of the bridges that are candidates for

strengthening/posting in Georgia were constructed in the two decades following 1945,

were designed for H-15 loading and had simple spans, with lengths of 40 to 70 ft.

Following a review of the existing documentation on these bridges, site visits

aimed at determining their suitability and testing feasibility, and discussions with State

bridge maintenance engineering staff, four bridges were finally identified for diagnostic

load testing, further in-depth finite element analysis and reliability assessment [O’Malley,

et al, 2009].

• Reinforced concrete bridge – straight (Bridge ID: 129-0045)

• Reinforced concrete bridge – skewed (Bridge ID: 015-0108)

• Pre-stressed concrete girder bridge (Bridge ID: 223-0034), and

• Steel girder bridge (Bridge ID: 085-0018)

These bridges and their structural characteristics are summarized in the following

paragraphs.

4.1.1 Reinforced Concrete Bridge – Straight (ID: 129-0045)

This bridge carries SR 156 over Oothkalooga Creek, was designed using the

AASHTO 1953 specification for H-15 loading, and was built in 1957. It is located one

mile west of Calhoun, GA in Gordon County. SR 156 is a two-lane road. The bridge has

eight spans, seven of which are 40 ft (12.19 m) and one (over the channel) 45 ft (13.7 m).

The girders are 18 ½ in x 24 ¾ in (46.99 cm x 62.87 cm), except for the long span which

is 31 ¾ in, and are spaced 7.2 ft (2.19 m) apart. The bridge has a deck width of 32.3 ft

(9.85 m) and a road way width of 25.7 ft (7.83 m). The bridge carries an ADTT of 458.

54
The concrete deck has a condition rating of 5, the supporting reinforced concrete T-beam

superstructure is rated at 7, and the concrete bent and pier substructure are rated at 6. The

latest inspection report indicates that all caps have minor hairline cracking, and that

several areas of exposed cap reinforcement are present. All beams are reported to show

signs of typical flexural cracking. The entire deck has moderate surface deterioration,

scaling, and cracking. It has also been repaired in several notably bad sections. The

bridge has not been posted, but was scheduled for replacement in 20081.

Figure 4.3 Straight T-Beam Bridge (ID: 129-0045, Gordon County)

4.1.2 Reinforced Concrete Bridge – Skewed (ID: 015-0108)

This 12-span structure over a long flood plain and a creek carries Old Alabama

Rd. over Pumpkinvine Creek 3.7 miles south of Cartersville, GA in Bartow County. The

two-lane bridge structure has a skew of 30 degrees and an ADTT of 709, was designed

using the AASHTO 1977 specifications for HS-20 loading and dates to 1979. The eleven

1
This bridge was demolished later on. Concrete samples were obtained; in situ concrete strength was
measured. The in situ strength then is used to update the bridge rating in the second level evaluation
presented in Chapter 6

55
spans over the flood plain are carried by 40-ft (12.19 m) reinforced concrete T-beams.

The 70-ft (21.34 m) span over the channel is supported by AASHTO type II pre-stressed

concrete girders. The current bridge ratings for substructure, superstructure, and deck are

6, 6, and 7 respectively, and the bridge is posted for three truck loads: H (21 tons),

Tandem (19 tons), and Log (24 tons). There is minor cracking and spalling in a number

of the bents and abutments, as well as in the T-beams, but none is in need of immediate

repair.

Figure 4.4 Skew T-Beam Bridge (ID: 015-0108, Bartow County)

4.1.3 Pre-stressed Concrete Girder Bridge (ID: 223-0034)

This bridge carries State Route 120 over Little Pumpkinvine Ceek approximately

5 miles south of Dallas in Paulding County GA. It was designed using the AASHTO

1989 for HS-20 loading specifications and was constructed in 1992. The main structural

system consists of pre-stressed concrete I-Beams arranged in four simply supported spans.

The bridge is 216 ft (65.8 m) long and is comprised of two 40-ft (12.2-m) Type II pre-

stressed I-girder spans and two 68-ft (20.7-m) Type III prestressed I-girder spans. The

56
centerline of the bridge is essentially perpendicular to the girder supports. The bridge has

a deck width of 43¼ ft (13.2 m) and a roadway width of 40 ft (12.2 m). The 68-ft (20.7-

m) spans are comprised of five type III I girders that are composite with the 9⅛ in (232

mm) thick slab (Figure 4.5). The bridge is in good condition, with substructure,

superstructure and deck condition numbers of 7, 8 and 7, respectively. It is not posted.

The ADT is 6550.

Figure 4.5 Pre-Stress Bridges (ID: 223-0034, Paulding County)

4.1.4 Steel Girder Bridge (ID: 085-0018)

This bridge carries SR 136 over the Etowah River 5.7 miles east of Dawsonville,

Georgia, in Dawson County. It was designed using the AASHTO 1961 specification,

with interim revisions through 1963 for H-15 loading, and was constructed in 1965. The

bridge is 196 ft ( 59.7 m) long and its four 49 ft (12.2 m) spans are supported by four

steel girders spaced at 8 ft on centers; the two facia girders are W33x118, while the two

interior girders are W33x130, with a full-depth diaphragm located at mid-span. The

57
two-lane bridge has a (non-composite) concrete deck, with overall width of 32 ft (9.75 m)

and a roadway width of 26 ft (7.92 m). The centerline of the bridge is perpendicular to

the girder supports. The bridge was last inspected on June 30, 2005, and at that time the

deck and substructure both were assigned a condition assessment rating of 6. That

inspection report indicates that there is spalling, aggregate exposure, and transverse

cracking in the deck in all spans. The bridge was determined to require posting, and has

been posted for a 21-ton H load, 25-ton HS load, 23-ton Tandem load, 32-ton 3-S-2 load,

and 27-ton Log load. The piles have minor pitting and the beams have minor deflections.

The bridge carries an ADTT of 280.

Figure 4.6 Steel Girder Bridge (ID: 085-0018, Dawson County)

58
4.2 DEVELOPMENT OF FINITE ELEMENT MODEL OF SELECTED BRIDGES

Three-dimensional (3D) nonlinear finite element models (FEMs) of the

superstructures of each of the four bridges were developed from design and construction

documents obtained from the Georgia Department of Transportation. The purpose of

these FE models was threefold. First, they provided a basis for comparison with the

simplified analytical evaluation procedure currently used by GDOT for bridge rating and

for identifying issues that might not be apparent with the existing component-based

deterministic rating format presented in bridge evaluation manuals. Improving bridge

rating guidelines requires an understanding of bridge system behavior subjected to

extreme load events which may well exceed the load level applied in the diagnostic tests.

Second, they were used to assist in designing the diagnostic load tests of the four bridges.

Finally, once validated through the diagnostic load tests (described subsequently), they

were used to conduct “virtual” proof load tests of other bridges of interest in the Georgia

Bridge Inventory. These ‘virtual” load tests, along with the system reliability analysis

described in Chapter 7, are an essential ingredient of the technical support of the

improved rating protocol introduced in Chapter 3.

All FE models of the sample bridges were developed using the ABAQUS

commercial FE package [Simulia, 2006]. Prior to the conduct of the load tests, these FE

analyses were performed using anticipated vehicle weights and arrangements to assist in

designing test instrumentation, to identify test vehicle locations, and to anticipate and

guard against potential bridge vulnerabilities that might become apparent during the

diagnostic load tests. Following the load tests, FE analyses again were performed using

the actual test vehicle weights and wheel locations measured during the tests, and

59
predicted responses were compared with test measurements to determine the accuracy

with which FEA can predict bridge behavior.

In the FE analyses of all four bridges, the failure mechanism in the concrete was

assumed to be either cracking in tension or crushing in compression. In the absence of

information on the in situ strengths of steel and concrete, the strengths specified on the

construction documents were assumed in bridge modeling and assessment. Of course,

this assumption introduces uncertainty in the comparison of model predicted structural

behavior and test measurements, as described subsequently. The stress-strain curve

proposed by Todeschini [1964] was utilized to model concrete behavior under

compression; in tension, the stiffness and strength reductions caused by cracking were

taken into account by the smeared crack technique [Kupfer, 1973; Hillerborg, 1976], in

which crack initiation is based on strength criteria and crack propagation is based on

fracture mechanics-based energy criteria. Deformed bar reinforcement and prestressing

strands both were assumed to have uniaxial elastic-plastic stress-strain behavior. All four

sample bridges were simply supported by the pier caps or abutments; these supports were

modeled by pin-roller boundary conditions.

For the reinforced and prestressed concrete bridges, the concrete deck, girders and

transverse diaphragms were modeled using 3D continuum solid elements. Steel

reinforcement was modeled using a distributed approach, in which the reinforcing bars

were smeared into membrane layers and embedded in the concrete at appropriate

locations. The pre-stressing strands in the girders were modeled individually using truss

elements embedded in the solid concrete elements. The pre-stress in strands was

replicated by applying an initial stress condition to the truss elements so that when the

60
bridge reaches self-equilibrium under such condition, the strands have the effective pre-

stress indicated in the design documents. The compatibilities between rebars/pre-stressing

strands and concrete were enforced. Shear reinforcement was ignored in the FE model of

the bridge superstructure. For the steel girder bridge (Bridge ID 085-0018), each girder is

modeled using beam elements for the flanges and shell elements for the web.

The FE model of the reinforced concrete T-beam span (bridge ID 129-0045) as

shown in Figure 4.7 had 420,928 degrees of freedom. The FE modeling of other bridges

was at a similar level of resolution. The research character of this investigation dictated

this level of resolution; such a level would not be required for routine bridge condition

assessment by analysis. More details of the FEMs can be found in O’Malley, et al [2009].

Figure 4.7: FEM of the Gordon County Bridge (ID: 129-0045)

61
4.3 VALIDATION OF FINITE ELEMENT MODELS OF THE SELECTED BRIDGES
THROUGH DIAGNOSTIC LOAD TESTS

The load tests were performed by GDOT employees, following the test protocols

instituted by the principal investigators and under their supervision. Each bridge was

tested using up to four DOT trucks; the truck wheel loads were measured prior to each

day’s testing. Details of the testing program, including the bridge selection,

instrumentation, testing process and the post-test assessment of the measurements, can be

found elsewhere [O’Malley, et al, 2009]. A summary of the results of the analysis and

load test conducted on the RC T-beam Bridge with straight approach (ID 129-0045) is

presented in Figures 4.8 and 4.9. Figure 4.8 shows the test arrangement of the trucks on

the bridge. The truck weight is summarized in Table 4.1. Figure 4.9 compares

displacements at midspan predicted by the FE analysis and the displacements measured

by potentiometers at the same locations.

When the RC T-beam Bridge with straight approach (ID 129-0045) was fully

loaded by four DOT trucks, totaling 223 kips (1,068 kN) as in Table 4.1, the bending

moment at mid-span of the bridge was 2.25 times the bending moment under the H-15

design load configuration. The maximum measured deflection of the beams at the mid-

span under such loading was 0.28 inches (7 mm). The span/800 deflection limit for

concrete T-beam bridge stipulated in AASHTO Standard Specifications [1992] (section

8.9.3.1) is 0.6 inches (15 mm). Clearly, the 1953 AASHTO specification that was in effect

at the time this bridge was designed incorporated a significant margin of safety.

Notwithstanding its age and design load, there is no evidence from this assessment that

this bridge was structurally deficient when evaluated according to modern bridge design

and rating criteria. Similar observations were made for the other bridges analyzed in this

62
study. The implications of these observations will be examined in detail later in Chapter

7.

The comparisons between predicted and observed maximum girder deflections

under load from four trucks for all four bridges tested are summarized in Table 4.2. The

results indelicate good agreement was achieved for all four bridges. The discrepancies

were invariably within 20% and, in the majority of cases, were substantially less. Such

differences can be attributed to various uncertainties associated with experimental data

collection under field conditions and the many assumptions made in the FE analyses,

including magnitude and homogeneity of in situ material properties and idealized

boundary conditions. In view of these factors, results of the FE analyses of the four test

bridges are considered sufficient to describe and quantify the load-carrying mechanisms

that affect the bridge capacity and its load ratings.

Figure 4.8: Schematic of Concrete Reinforced T-Beam Bridge

63
Table 4.1 Truck Weight (lb) Details for RC T-Beam Bridge (ID: 129-0045) Test

Load on Load on Load on Overall truck

axle 1 axle 2 axle 3 weight

Truck 1 18,400 19,100 19,000 56,500

Truck 2 19,100 17,400 17,100 53,600

Truck 3 19,500 19,300 19,000 57,800

Truck 4 17,800 18,700 18,600 55,100

Table 4.2 Comparison of the Maximum Deflections Measured in the Test and
Predicted by FE Analysis

Bridge ID Measurement FEM

RC bridge – straight
129-0045 0.28 0.26
approach
RC bridge – skewed
015-0108 0.14 0.16
approach
PC bridge 223-0034 0.20 0.22

Steel girder bridge 085-0018 0.36 0.43

4.4 CLOSURE

This Chapter has summarized the load testing and analysis phase of the study to

examine current bridge rating procedures and to improve them using reliability –based

64
methods. Four bridges that are typical of bridges of concern in rating and posting were

selected for load testing and analysis. Finite element models of these bridges were

developed to assist the design of the load tests and in the interpretation of the results. The

bridge test results, in turn, were used to validate and improve the finite element modeling.

The measured deflections in all cases were in good agreement with those predicted by the

FE model. All four bridges remained well within the elastic range when loaded to an

intensity that is well above their design load. The maximum deflections measured during

the load tests were on the order of 25% - 50% of the span/800 limit on deflection stipulated

in the AASHTO design specifications. Experience with these load tests suggests that

basing performance assessment of an existing bridge on global response measurements,

such as displacement, as opposed to local responses, such as strain, minimizes the

likelihood of errors in test interpretation and misjudgments of safety that may be prompted

by spurious local non-homogeneous or material behavior. It was also observed that

redundancy in measurements, through multiple gauges at a single location and gauges at

multiple locations in a single element, is essential to achieve accurate conclusions from the

condition assessment and should be utilized whenever practical.

In Chapters 6 and 7, this combined analytical and experimental approach will be

used to develop an improved bridge rating framework in which in situ material testing,

condition data, and history of successful service life performance are properly integrated

through system reliability analysis into an improved capacity rating metric.

65
 (a) Loaded with Truck 1 (b) Loaded with Truck 1 & 2

(c) Loaded with Truck 1, 2 & 3 (d) Loaded with Truck 1, 2, 3 & 4

Figure 4.9: RC T-Beam Bridge Girder Displacements Due to Truck Loadings

66
 CHAPTER 5
LEVEL-ONE AASESSMENT: CORRELATION BETWEEN
CONDITION RATING AND CAPACITY RATING

5.1 INTRODUCTION

A bridge subjected to environmental attack and excessive traffic loads may

experience changes in resistance, and its capacity rating should reflect these changes.

Currently, in the AASHTO Manual for Bridge Evaluation, the physical condition of a

bridge is reflected in the capacity rating equation (eq. 6-1; eq 6A.4.2.1-1 in the MBE)

through a condition factor φc., where condition ratings from 1-9 that are assigned during

routine inspections according to National Bridge Inspection Standard (1996) (as

summarized in Table 5.1) are categorized into three qualitative descriptions of bridge

condition – good, fair and poor - and these qualitative descriptions are then connected to

the resistance calculation in the capacity rating through the condition factor, φc , as

summarized in Table 5.2 (AASHTO LRFR Table 6-2 ).

The current categorization of bridge condition above is purely empirical and the

development of the φc values is not theoretically supported by structural reliability

analysis. This chapter introduces a new method to correlate bridge condition ratings with

capacity ratings, taking into account both the underlying physics of bridge deterioration

phenomena and bridge conditional rating history data from routine inspections. A

revised set of values of φc that are tied to the rating equations have been developed to be

consistent with the structural reliability-based evaluation philosophy embodied in the

AASHTO MBE (LRFR option) and to incorporate recent developments in bridge

67
resistance degradation modeling and comprehensive databases of bridge condition rating

history.

Four steps are taken. First, a state-of-the-art bridge degradation model that

describes the physical process of bridge deterioration in a stochastic fashion is selected.

Second, the average bridge condition rating history model that illustrates the condition

rating as function of bridge service age is established. Third, the bridge condition rating

history is linked to the statistical models of bridge resistance by mapping the condition

rating history model in step two onto the bridge degradation model in step one. Finally,

a reliability-based optimization technique is used to identify a set of φc values that

express this correlation in a deterministic manner in the rating formula to satisfy the

reliability requirement embodied in AASHTO LRFR. These steps are illustrated in the

following sections, in which condition factors are developed for reinforced concrete

bridges. The methodology introduced can be applied to other types of bridges as well.

5.2 STOCHASTIC BRIDGE DETERIORATION PROCESS

Quantitative models of deterioration of reinforced concrete structures have been

developed in many research studies [Albrecht and Naeemi, 1984; Mori and Ellingwood,

1993; McCuen and Albrecht, 1995; Thoft-Christensen, 1998; Enright and Frangopol,

1998]. These models can be incorporated in a real-time bridge reliability assessment.

The uncertainties in resistance of an existing bridge are at least equal to those of a newly

designed bridge. Once the bridge begins to deteriorate, its mean resistance usually

decreases and the uncertainty in resistance generally increases. Time-dependent structural

resistance can be modeled as [Mori and Ellingwood, 1993]:

R (t ) = R0 g (t ) (5-1)

68
in which t is elapsed time, R0 is the resistance variable of a newly-constructed bridge and

g(t) is the degradation rate. The mean and COV of random variable g(t) can be expressed

for many common deterioration mechanisms [Enright and Frangopol, 1998], in first

approximation, as linear functions of time, as in eq (5-2):

1, t ≤ T0
Ε[ g (t )] = 
1 − k1 (t − T0 ) , t ≥ T0 (5-2)
V [ g (t )] = k 2 t

in which k1 and k2 are constants and T0 is the mean time required to initiate corrosion.

Bridges are exposed to many environmental stressors. The extent and the rate of strength

loss, g(t), depends on the aggressiveness of the environment and the properties of

construction materials. Chemicals, moisture, and cycle of extreme temperature are the

most common environmental factors that influence the strength of the structure.

For concrete bridges subjected to environmental attack, the strength degradation

mechanism can be classified as affecting either the concrete or the steel reinforcement, or

both. Concrete deteriorates because of internal pressures which are caused primarily by

chemical reactions in the cement (sulfate attack), by chemical reactions between the

cement and the aggregates (alkali-silica reaction), or by freeze-thaw cycles. Reinforcing

steel deteriorates primarily because of corrosion. In addition, the corrosion products

cause internal pressure that can lead to cracking and spalling of the concrete (Val and

Melchers, 1997). Enright and Frangopol [2000] found that most damage of RC bridges is

caused by water leakage through transverse joints in the deck and the corrosion is most

frequent damage mode. Corrosion of reinforced concrete is a two-stage process

consisting initiation (carbonation penetration or chloride ion ingress) and propagation

(metal loss).

69
 Table 5.1 NBIS Instruction for Superstructure Condition Rating

Condition
Description
Rating

9 EXCELLENT CONDITION

8 VERY GOOD CONDITION – No problems noted

7 GOOD CONDITION – Some minor problems

SATISFACTORY CONDITION – Structural elements show some minor

6
deterioration

FAIR CONDITION – All primary structural elements are sound but may have
5
minor Section loss, cracking, spalling or scour

4 POOR CONDITION – Advanced section loss, deterioration, spalling or scour

SERIOUS CONDITION – Loss of section, deterioration, spalling or scour have

3 seriously affected primary structural components. Local failures are
possible. Fatigue cracks in steel or shear cracks in concrete.

CRITICAL CONDITION – Advanced deterioration of primary structural

elements. Fatigue cracks in steel or shear cracks in concrete may be present or
2
scour may have removed substructure support. Unless closely monitored, it may
be necessary to close the bridge until corrective action is taken.

IMMINENT FAILURE CONDITION – Major deterioration or section loss

present in critical structural components or obvious vertical or horizontal
1
movement affection structure stability. Bridge is closed to traffic but corrective
action may put back in light service.

0 FAILED CONDITION – out of service. Beyond repair

Table 5.2 Condition Factor, φc (AASHTO LRFR, 2005, Table 6-2)

Structural Condition of Member (Condition Rating) φc

Good or satisfactory ( >= 6) 1.00
Fair (= 5) 0.95
Poor ( <= 4) 0.85

70
 Corrosion initiates after an initiation time (T0), at which time the steel

reinforcement becomes depassivated due to carbonation or chloride ion ingress

(Novokshchenov 1989; Whiting et al. 1993) and the chloride content in the concrete at

the depth of the steel reinforcement reaches a critical concentration. The corrosion

initiation time can be expressed as [Thoft-Christensen et al, 1977]

X2 −1 C 0 − C cr − 2
T0 = [erf ( )] (5-3)
4 Dc C0

where X = concrete cover (cm); Dc = chloride diffusion coefficient (cm2/year); Co

=equilibrium chloride concentration at the concrete surface ( in percent of weight of

concrete); and Ccr = critical chloride concentration at which corrosion begins (in percent

of weight of concrete). The corrosion initiation time is dependent on four random

variables(X, Dc, Co, Ccr). The main descriptors of these random variables can vary

considerably for different bridges.

Once corrosion has initiated, the cross-sectional area of reinforcement decreases

with time at a rate that is dependent on the number of reinforcement bars actively

corroding and the diameter of the individual bars. For the general case where the steel

reinforcement is composed of bars of various diameters, which begin corrosion at

different times, the time-variant area of steel, A(t), is [Mori and Ellingwood, 1994; Thoft-

Christensen et al. 1997; Enright and Frangopol 1998a]:

π n
At =
4
∑ [D ( t )]
j=1
j
2
(5-4)

D jo for t ≤ T1 j

Where D j ( t ) = D jo − rcorr ( t − T1 j ) for T1 j ≤ t ≤ T1 j + D jo / rcorr (5-5)

0 for t ≥ T1 j + D jo / rcorr

71
and Dj (t) = diameter of bar j at time t; n = number of bars; Djo = initial diameter of bar j;

rcorr = corrosion rate; t = elapsed time, and TIj = corrosion initiation time for bar j. For

rectangular nonprestressed members in which the strength of compression steel is

neglected, the nominal flexure resistance Mn of a concrete beam is given by [LRFD

1994]:

1 As f y
M n = At f y (d − ) (5-6)
2 0.85 f c' b

where fy = specified yield strength of reinforcing bars; d = distance from the extreme

compression fiber to the centroid of the non-prestressed tensile reinforcement; f’c =

specified compressive strength of concrete at 28 days; and b = width of the compression

face of the member.

To investigate the corrosion initiation time and strength degradation function for

typical RC bridges and to define the coefficients in eq 5-2, Enright and Frangopol [2000]

performed Monte Carlo simulations using eqs 5-3 through 5-6 and the statistics of key

initial resistance and corrosion random variables listed in Table 5.3. The mean and

coefficient of variation of the resistance random variables are based on the information

presented in MacGregor et al [1983]. Values for the corrosion random variables are based

on probabilistic corrosion studies summarized in Enright [1998]. Those studies found that

the corrosion initiation time T0 is lognormally distributed and is increasing with the depth

of the concrete cover. The mean of the degradation function, E[g(t)], decreased with

time while its cov V[g(t)] increased, as shown in Figure 5.1. For reinforced concrete

bridges subjected to environmental attack with medium rate of corrosion, the mean value

of T0 is approximately 10 years and coefficients k1 and k2 in eq (5-2) equal to 0.0031 and

0.0027, respectively; substituting these parameters into eqs 5-2 yields the time-dependent

72
relations presented in Figure 5.2. In other words, the mean of the resistance R(t) will

reduce to 80% of its original valve and the COV of g(t) alone will increase to 13%, after

a 75 year period of exposure. These statistics must be factored into the time-dependent

reliability analysis of the reinforced concrete bridge.

Figure 5.1 Mean and COV of g(t) of Time-variant Bending Resistance with
Different Corrosion Rate rcorr (Enright and Frangopol, 2000)

73
 Table 5.3 Random Variable for MC Simulation
(Enright and Frangopol, 2000)

Variable Coefficient
Description Mean
(units) of variation
fy (Mpa) Steel reinforcement yield strength 310.5 0.12
f’c (Mpa) Concrete compression strength 19.0 0.18
DM (mm) Initial diameter of flexure reinforcement 35.8 0.02
dM (mm) Initial depth of flexure reinforcement 68.73 0.03
XM1 (mm) Cove depth of flexure steel, layer 1 5.08 0.05
XM2 (mm) Cove depth of flexure steel, layer 2 12.70 0.05
Dc (cm2/yr) Diffusion coefficient 1.29 0.10
Co (wt % conc.) Surface chloride concentration 0.20 0.10
Ccr (wt % conc.) Critical chloride concentration 0.025 0.10
rcorr (mm/yr) Corrosion Rate 0.15 0.30

Figure 5.2 Time-dependent Mean and Coefficient of Variation of

Bridge Degradation Function g(t)

74
5.3 SYNTHESIS OF THE CONDITIONAL RATING DATA FROM NBI

Bridge condition ratings (see Table 5.1) for individual components are assigned

following inspections which usually are conducted once every two years. If such

condition ratings for a bridge are available for a relatively long period of time, the data

can be used to establish a deterioration model for that bridge. National Bridge Inspection

Standard (NBIS) coding guide identifies the specific bridge elements that must be

inspected and provides instructions on how to conduct an inspection. Although state

transportation agencies are not required to make their bridge inspection programs

identical to that describe in the coding guide, they are required to have databases that can

easily be converted to NBIS format for reporting to FHWA so that data from various

states can be combined to form the National Bridge Inventory (NBI).

In a review of the NBI database, Bolukbasi, et al (2004) found that the average

condition rating history of non-interstate RC bridges often can be modeled by a 3rd order

polynomial, as shown in Figure 5.3:

C (T) =8.662-0.146T + 0.003T2 - 3.09E5T3 (5-7)

where C(T) is the condition rating of the bridge at age, T, in years. This model yields a

condition number 4 in 70 years. Jiang and Sinha (1989) developed a similar polynomial

model with slightly different coefficients, which indicated 71 years to a condition state 4.

Weyers et al (1989) computed an average condition rate deterioration rate which

indicates 65 years to condition state 4 and 78 years to condition state 5. These models

are consistent with one another, and the model raised in the study by Bolukbasi, et al

(2004) is adopted in this study. The condition rating assigned during routine inspections

varies from inspector to inspector, and thus is a random variable that might affect time-

75
dependent reliability analysis. A study on the accuracy of inspection documentation

[Phares et al, 2004] has revealed that the distribution of condition ratings is normal.

Figure 5.3 Average Condition Rating as Function of Time for Non-interstate

Highway Bridges (Bolukbasi et al, 2004)

5.4 CORRELATION BETWEEN CONDITION RATING AND RESISTANCE MODEL

The correlation between condition rating C(T) and the statistical descriptors of

degradation g(t) is developed by mapping the average condition rating history of non-

interstate concrete bridges, as shown in Figure 5.3, onto the 75-year stochastic resistance

degradation model with medium degradation rate in Figure 5.2, leading to the rating-

dependent mean and COV of function g(t) shown in Figure 5.4 by the solid lines. When

76
flexural resistance is considered, Ro is described by a lognormal distribution with mean of

1.14 times the nominal flexural strength Rn and COV of 13%, respectively (Nowak, 1999).

The dashed lines in Figure 5.4 show the mean, E[R/Rn], and COV, V[R/Rn], of the

resistance modal as a function of condition rating. By mapping the resistance with an

average degradation rate on the mean condition rating history, the proposed statistical

descriptions of resistance as a function of condition rating are finally independent of

corrosion rate.

Figure 5.4 Time-dependent Mean and COV of Bridge Flexural Capacity

5.5 CALIBRATION FOR THE CONDITION FACTOR, φc

Using the statistics in Figure 5.4 along with the load models used in the AASHTO

LRFD (Nowak, 1999), the bridge condition rating values can be included in the

77
estimation of time-dependent failure probability and reliability index of a given bridge.

To further facilitate the bridge rating practices which utilize a deterministic format, a set

of φc-values necessary to achieve the target reliability requirements consistent with the

AASHTO LRFR method was obtained by minimizing the mean-square error between the

target βT and the reliability achieved by the use of specific values of φc, as illustrated in

Figure 5.5. The difference in target reliability (βT), 3.5 at the inventory level vs 2.5 at the

operating level, is reflected in difference in the live load factor used for these levels,

therefore does not affect the calibration of φc. The optimal condition factors determined

from this analysis are presented in Table 5.4.

Figure 5.5 Optimal Condition Factors for Different Condition Ratings

78
5.6 CLOSURE

This chapter illustrated a methodology for incorporating qualitative measures of

bridge condition into a quantitative reliability-based evaluation of bridge load rating.

Correlation between condition rating and capacity rating was established through a set of

condition factors with sound reliability basis. The development of φc was illustrated for

reinforced concrete bridges, but the methodology could easily be applied to other types of

bridges if supporting time-dependent deterioration statistics for those bridges similar to

those presented in Figure 5.3 become available. The AASHTO LRFR deterministic

rating format (eq 2-6), together with the revised values of condition factor φc developed

in this chapter, should be incorporated in the first-level assessment of the proposed rating

framework.

Table 5.4 Proposed Condition Factors

Structural Condition Rating (SI&A Item 59) φc

≥8 1.00
7 0.95
6 0.85
5 0.75
≤4 0.70

79
 CHAPTER 6
LEVEL-TWO ASSESSMENT: BRIDGE RATING USING
COMPONENT-LEVEL INFORMATION

The level-one assessment introduced in Chapter 5 utilizing the deterministic

format in AASHTO LRFR incorporates only limited quantitative in situ knowledge and

therefore could produce overly conservative ratings for many well-maintained older

bridges [Nowak and Tharmabala, 1988; Bakht and Jaeger, 1990; Moses, et al, 1994; Fu

and Tang, 1995; Faber, et al 2000; Barker, 2001; Bhattacharya, et al, 2005]. The models

of uncertainty in load and resistance (discussed in section 2.2.3) used in calibrating the

LRFR rating equation (eq. 2.6) represent typical values for a large bridge population

involving different materials, construction practices and site-specific environmental and

traffic conditions. Although AASHTO has modified the live load model in LRFR rating

to account for differences in site-specific average daily truck traffic, the resistance model

also should be “customized” for an individual bridge by incorporating available site-

specific knowledge to reflect the fact that each bridge is unique in its as-built condition.

The second level assessment in this Chapter provides bridge engineers with an

option to include additional site-specific information in the bridge rating process to

achieve a better evaluation of bridge capacity, when such effort is believed to be

warranted. Knowledge gained from diagnostic load tests, from validated finite element

analysis or from in-situ material tests can be used to revise the LRFR estimates (Chapter

5). An investment in the level-two assessment often can be justified, particularly for

bridges that carry heavy traffic or have level-one rating factors in the range of 0.7-1.0.

80
6.1 RATING USING IN SITU MATERIAL STRENGTH

Section 4.1.1 summarized the load test of Bridge ID 129-0045, a reinforced

concrete T-beam bridge that was designed according to the AASHTO 1953 Design

Specification for H-15 loading, and was constructed in 1957. The specified 28-day

compression strength of the concrete was 2,500 psi (17.2 MPa), while the yield strength

of the reinforcement was 40 ksi (276 MPa). This bridge was load-tested in September,

2006. Subsequent to the conduct of the load test, the bridge was demolished in May,

2008, providing an opportunity to secure drilled cores to determine the statistical

properties of the in situ strength of the 51-year old concrete in the bridge.

Four-inch diameter drilled cores were taken from the slab of the bridge prior to its

demolition. Seven (7) cores were taken from the slab at seven different locations along

both the length and width of the bridge. Cores also were taken from three of the girders

which were in good condition after demolition; these were cut into 8-in (203 mm) lengths

and the jagged ends were smoothed and capped, resulting in a total of fourteen (14) girder

test cylinders. Tests of these 4 x 8 in (102 x 203 mm) cylinders conformed to ASTM

Standard C42, and the results are presented in Table 6.1. An analysis of these data

indicated no statistically significant difference in the concrete compression strength in the

girders and slab, and the data were therefore combined for further analysis. The mean

(average) compression strength of the concrete is 4,820 psi (33 MPa) and the coefficient

of variation is 12%, which is representative of good-quality concrete [Bartlett and

MacGregor, 1996]. The mean strength is 1.93 times the specified compression strength

of the concrete. This increase in compression strength over a period of more than 50

81
years is typical of the increases found for good-quality concrete of this vintage by other

investigators [Washa and Wendt, 1975].

If these results are typical of well-maintained older concrete bridges, the in situ

concrete strength is likely to be substantially higher than the 28-day strength that is

customarily specified for bridge design and also is used in condition evaluation.

Accordingly, the rating criteria should provide the Bridge Engineer with incentives to use

the best possible information from in situ material strength testing whenever feasible

when performing a bridge rating [Ellingwood et al 2009]. It is customary to base the

specified compression strength of concrete on the 10-percent exclusion limit of a normal

distribution of cylinder strengths [ACI Standard 318-05]. Using Bridge ID 129-0045 as

an example, the 75% lower confidence interval on the 10-percent exclusion limit of

compression strength, fc, for the sample of 21 tests can be expressed as,

f c = (1 − kV) X (6-1)

in which X = sample mean, V = sample coefficient of variation, and k equals 1.520

(Ellingwood, et al, 2009) to obtain the 75% lower confidence interval on the 10th

percentile value of the distribution, based on 21 samples. Substituting the statistics in

Table 6.1, one would obtain fc = (1-1.520 × 0.12) × 4,820 = 3,941 psi (27.17 MPa), a

value that is 58% higher than the 2,500 psi (17.2 MPa) that otherwise would be used in

the rating calculations.

In the FE modeling of this bridge that preceded these strength tests, the concrete

compression strength was set at 2,500 psi (17.2 MPa), which was the only information

available before the material test. In order to determine the impact of using the actual

concrete strength on the rating process, the finite element model was revised to account

82
for the increased concrete compression strength (and the corresponding increase in

stiffness) into the analysis of the bridge. Only a modest enhancement in the estimated

bridge capacity in flexure was obtained, but a 34% increase was achieved in the shear

capacity ratings for the girders using the results of Table 6.1.

Table 6.1 Compression Tests of Cores from RC Bridge (ID: 129-0045)

Source Number Average Standard Coefficient of

(psi) deviation (psi) variation
Girder 14 4,880 603 0.12

Slab 7 4.698 573 0.12

Overall 21 4,820 586 0.12

6.2 RATING USING REALISTIC LOAD DISTRIBUTIONS

Girder distribution factors (GDFs) are used to distribute the traffic loads to the

individual girders so that bridge design and evaluation can be performed on an individual

member rather than a system basis. The GDFs are an important ingredient of bridge

capacity evaluation. The GDFs are different in the AASHTO ASD and LF rating

methods from those in the LRFR rating method; these differences are one reason for the

complaints received in the survey (discussed in section 2.3) regarding the inconsistency

in the bridge ratings obtained from these three methods. These differences were

examined using finite element models developed for the four typical Georgia bridges

83
(summarized Chapter 4). Bridge responses captured by FE analysis (validated by the

load tests, as discussed in section 4.3) were used to assess the in situ GDFs .

The GDFs for girder moment and girder shear are different. A comparison of the

moment distribution factors from the different methods for the four tested bridges is

tabulated in Table 6.2. The GDFs obtained from the FE models are substantially less

than those calculated using any of the existing methods; accordingly, the load ratings

calculated with the GDFs obtained by the FE models would substantially exceed the load

ratings that are obtained by the existing methods in the AASHTO MBE. For the straight-

approach reinforced concrete girder bridge, for example (first line in Table 6.2), the FE

analysis of this bridge indicates that when load is placed on the bridge to maximize the

moment in one of the interior girders, only about 41% of the applied load actually went

into that interior girder. The girder distribution factor in the current LFR method

however would require up to 60% of live load to be apportioned to that girder, while with

LRFR, the percentage would be 69%. While the LFR procedure results in a rating factor

of 0.75 for the HS-20 design load checking at Inventory level, the rating of this bridge

using load distribution factor obtained from FE analysis for the same vehicle is 1.10,

representing an increase of 47% [Wang et al, 2009; Ellingwood, et al, 2009]. Similar

results were observed for the other three bridges considered in this dissertation.

6.3 RATING OF DEEP REINFORCED CONCRETE BRIDGE COMPONENTS

Out of the 2000 Georgia bridges that require posting according to the AASHTO

MBE, more than 800 are governed by deficient shear ratings of their reinforced concrete

pier caps. The posting of the steel girder bridge illustrated in Figure 4.6 (Bridge ID 085-

0018) is based on the assessed shear capacity of its center pier cap. Accordingly, prior to

84
conducting the load tests, a series of independent FE analyses of the reinforced concrete

pier cap for this bridge was undertaken. This pier cap behaves as a “deep beam,” in that

its shear span is relatively short (its shear span/effective depth ratio, a/d, is approximately

1.0). In contrast, the shear capacity equations in the traditional bridge rating procedure

(similar to those in ACI Standard 318, 2005,) are known to be valid for beams in which

a/d is greater than approximately 3, but may underestimate the actual shear capacity of

deep reinforced concrete beams, in some cases significantly (Hawkins et al, 2005).

Table 6.2 Comparison of the Moment Distribution Factors for Interior Girders

Bridge Type LFR/ASR LRFR FEM

Concrete T 0.597 0.69 0.407

Concrete Skew T 0.757 0.73 0.482

Pre-stressed 0.818 0.85 0.521

Steel Girder 0.725 0.72 0.513

The Strut-and-Tie (S&T) model has been proposed recently as an alternative

method for evaluating the shear capacity of reinforced concrete beams with short shear

spans [e.g., Tang and Tan, 2004]. To determine whether an S&T analysis might enhance

the rating of Bridge ID 085-0018, an independent rating analysis was performed. First,

the current GDOT rating of the bridge and its current posted limits was confirmed by an

85
independent analysis. Next, the capacity of the pier cap in shear was assessed using both

S&T and FE analyses. In the FE model, the shear capacity was assumed to be reached

when yielding initiates in the steel reinforcement acting as the tie of the S&T mechanism

over the support (see Figure 6.1) or the concrete compression strut crushes. The stress

contours in Figure 6.1, obtained from the FE analysis of the pier cap, clearly show the

development of arch action, which the S & T model captures. Table 6.3 compares the

posting limits for this bridge for five GA/State Legal loads (cf Figure 2.3). The S & T

shear strength model leads to ratings that range from 24% (HS-20) to 30% (Tandem) over

the current method. In turn, the S & T shear strength estimates are less than the FE

results, suggesting that the current posting limits for this bridge, which are based on

traditional shear capacity calculations, are unduly conservative. Had the S & T model

been used to determine the shear strength of the pier cap, the posted limits would have

increased to 34 tons for HS20, 33 tons for Tandem and 44 tons for 3S2, as tabulated in

Table 6.3.

To determine whether or not this conservatism is unique to this particular bridge,

an analysis was performed of pile caps at two additional bridges identified through the

Georgia DOT database – Bridge ID 083-0016 (built in 1966 for H-15 load; ADTT: 130)

and ID 097-0032 (built in 1962 for HS-15; ADTT: 120) – where a/d is smaller than 1.5

and the shear capacity of the pile cap also governs posting. Table 6.4, developed using

an HS-20 rating vehicle, reveals that the current rating procedure appears to result in

excessively conservative posted loads for these two bridges as well. The ratings are

particularly conservative when the point on the pier cap that supports the girder is close

to the supporting piles, leading to a short shear span. Accordingly, the strut-and-tie

86
model appears to be more appropriate for assessing the shear capacity of existing bridge

sub-structures for rating purposes.

Figure 6.1 Development of Arch Action in Deep Beam

Table 6.3 Shear Ratings (tons) for RC Bridge (ID 085-0018)

Method H20 HS20 Tandem 3S2 Log

GDOT Method1 22 25 24 32 28

Strut and tie 30 34 33 44 38

Finite element 40 45 43 58 49

1
The posted loads reported in this line of the table are taken from the GDOT database.

87
 Table 6.4 Shear Ratings (tons) for HS20 Vehicle

Bridge ID
Method
085-0018 083-0016 097-0032

GDOT Method18 25 24 23

Strut and Tie 34 28 32

Finite Element 45 34 40

6.4 CLOSURE

This chapter explored the possibility for improving bridge load ratings by

factoring site-specific knowledge (directly or indirectly obtained from in situ testing or

inspection) into the bridge rating process. Three examples illustrated in the above

sections indicate a significant gain in load ratings in level-two assessment. This level of

analysis provides an incentive to obtain quantitative in situ measurements through

modern non-destructive evaluation (NDE) techniques.

Material strengths in situ may be vastly different from the standardized or

nominal values assumed in design. In situ material sample tests can be used to update

load rating calculations. Concrete compression strength gain in a typical straight T-beam

bridge leads to a 34% increase in the shear capacity rating of that bridge. For some other

bridges, however, material strength might decrease due to aggressive attack from

physical or chemical mechanisms. Material tests in these situations can help the bridge

engineer obtain more realistic evaluations and reduce the likelihood of an unsafe rating.

Decisions as to whether or not to conduct in situ material tests should be based on reliable

88
bridge inspection data, bridge economics and engineering judgment. Test procedure

should be consistent with ASTM standards, and in situ sampling and interpolation of test

results should have a reliability basis [Ellingwood et al, 2009].

The load tests and supporting analysis herein indicated that analytical approaches

to bridge evaluation utilizing current girder distribution factors yield a conservative

measure of actual load-carrying capacity. This conservatism is the result of assumptions

made in the analysis regarding load sharing, composite action, support conditions and

secondary member behavior. In view of the economic consequences of posting, it is

apparent that if the customary rating practice suggests that a bridge is a candidate for

posting, a more accurate structural analysis model should be employed to verify whether

more accurate GDFs might change that decision.

The load-carrying mechanism in pier caps and other structural components that

have short shear spans and behave as deep beams is better modeled by the strut and tie

method than by the traditional ACI Standard 318 model. The use of this method is

permitted by the LRFR option in the AASHTO MBE, and bridge engineers should be urged

to adopt it. The rated capacity of the pier cap increases by up to 59% in the typical steel

girder bridge examined when the S&T model is used. Investigations of similar bridge pier

caps indicated that the level of conservatism is dependent on the dimensions of individual

pier caps and the placement of the girders that they support.

89
 CHAPTER 7
LEVEL THREE ASSESSMENTS: BRIDGE SYSTEM RELIABILITY

As part of the effort to develop the AASHTO LRFD Bridge Design Specifications,

extensive databases were developed to describe the strength of individual bridge girders

and vehicle live loads probabilistically [Nowak, 1999; Moses, 2001]. As noted

previously, that research focused on the capacity of individual bridge girders; system

effects were included only indirectly and approximately through new girder distribution

factors that were developed in the course of the project. This approach was also adopted

by AASHTO in the bridge rating methods found in the Manual of Bridge Evaluation that

currently is in use. While component-based design of a new bridge provides adequate

safety at reasonable cost, component-based evaluation of an existing bridge for rating

purposes may be overly conservative and result in unnecessary repair or posting. In

particular, it is preferable to perform load capacity ratings through a system-level analysis

if decisions regarding bridge posting or road closure are at issue.

The level-three assessment introduced in this Chapter investigates bridge safety at

the system level and presents an additional perspective on the (unknown) level of

conservatism furnished by the current generation of reliability-based condition evaluation

and rating procedures. This level of analysis could also provide an understanding of

bridge system behavior when subjected to extreme load events which may have

implications for the use of such methods in permit ratings for extreme vehicle loads.

In this Chapter, bridge ratings are examined through a bridge system reliability

analysis, and the possibilities of incorporating proof load tests and successful service

90
performance history into the rating analysis are explored. In order to determine the

additional level of conservatism (if any) that arises from system behavior, finite element-

based system reliability analyses of all four study bridges were conducted. These

analyses allow the characteristic values of the variables used in the condition evaluation

to be changed on the basis of the outcomes of a proof load test or service-proving load.

7.1 VIRTUAL PROOF LOAD TESTS OF SAMPLE BRIDGES

A properly conducted proof load test can be an effective way to evaluate the

structural performance of a bridge as a system and to update the bridge load capacity

assessment in situations where the analytical approach at the first and second levels

produces low ratings, or structural analysis is difficult to perform due to deterioration or

lack of documentation (Saraf and Nowak 1998). However, a proof load test represents a

significant investment in terms of capital, time, and personnel, and the tradeoff between

the information gain and the risk of damaging the bridge during the test must be

considered. Therefore, proof tests are rarely conducted by the State DOTs for rating

purposes (Wang, 2009).

One of the key conclusions from the bridge modeling and tests in Chapter 4, in

which bridge response measurements obtained from the load tests of the four bridges

were compared with the results of finite element analyses of those bridges using

ABAQUS [Simulia, 2006], was that finite element modeling is sufficiently accurate that it

can be used to conduct “virtual” load tests of the majority of similar bridges. These

virtual load tests can provide the basis for developing recommendations for bridge ratings

using structural reliability principles. As noted in the introductory section above, such

recommendations require the bridge to be modeled as a structural system in order to

91
properly identify the performance limit states on which such recommendations are to be

based.

7.1.1 Push-down Analysis

To identify such performance limit states and to gain a realistic appraisal of the

conservatism inherent in current bridge design and condition rating procedures, a series

of static “pushdown” analyses of the four bridges identified in Chapter 4 was performed

to determine their actual structural behavior when loaded well beyond their design limit.

In a pushdown analysis, two rating vehicles are placed side by side on the bridge in a

position that maximizes the response quantity of interest in the evaluation (maximum

moment, shear, deflection, etc.). The loads are then scaled upward statically and the

performance of the bridge system is monitored. The dead weight of the bridge structure

is included in the analysis. The response initially is elastic. As the static load increases,

however, elements of the bridge structure begin to yield, crack or buckle, and the

generalized load-deflection behavior becomes nonlinear. If the bridge structure is

redundant and the structural element behaviors are ductile, substantial load redistribution

may occur. At some point, however, a small increment in static load leads to a large

increment in displacement. At that point, the bridge has reached its practical load-

carrying limit, and is at a state of incipient collapse.

The static pushdown analyses initially were performed using the FE platform

ABAQUS, with random material properties determined by their respective mean values

to obtain a “best estimate” of bridge capacity. In all four cases, two HS-20 vehicles were

placed side by side on two adjacent lanes at a point so as to maximize the elastic moment

in the girders. These vehicle loads were scaled upward by a load factor until the point

92
was reached at which the load-deflection plot indicated the onset on nonlinearity in the

bridge structural system. The dead load was held constant throughout the pushdown

analyses. An illustration of the static pushdown analysis of the reinforced concrete T-

beam Bridge ID 129-0045 is presented in Figure 7.1. The point of initial flexural

yielding occurs at a moment corresponding to approximately 4.31 times the two HS-20

design loads, at a deflection of approximately 1.4 in (36 mm), which is equal to

approximately 1/345 times the span. The ultimate live load capacity of this bridge,

defined as the point at which a small increase in load causes a large increment of

displacement or the FE solution failed to converge, is approximately 4.8 times the applied

HS20 loads. It is interesting to note from Figure 7.1 that this 51-yr old bridge shows a

considerable degree of ductility in behavior. The level of load imposed during the

diagnostic load test by the four fully loaded trucks is also shown in Figure 7.1; the

diagnostic test load was approximately 1.3 times the two side-by-side HS20 loads (in

terms of maximum moment). Clearly, the capacity of this bridge is substantially in

excess of what a girder-based calculation would indicate.

Similar pushdown analyses performed on the other three bridges yielded the

results summarized in Table 7.1. The elastic ranges of all four bridges are in excess of

four times the design load level, indicating the level of conservatism associated with

traditional design and rating procedures. The ultimate capacity, as shown in the 5th

column of the table, is related to the number of the beams. The Skewed T-beam Bridge

and the Prestressed Concrete Bridge both have five beams and clearly have more reserve

capacity than the other two bridges which have four beams each.

93
 5.0
Opensees
Abaqus
4.0

Live Load (Norm aliz ed)

3.0

2.0

Diagnostic Test
Load Level
1.0

0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Mid-span Displacement in Beam 2 (in)

Figure 7.1 Push-down Analysis of RC Bridge (ID 129-0045)

7.1.2 Finite Element Analysis Based Bridge System Reliability

To accelerate the FE-based reliability analysis, FE models of the four sample

bridges were developed using the open-source platform, OpenSees (McKenna, et al,

2007). The more detailed ABAQUS models, which had been validated in the load tests,

were employed to confirm the bridge structural behavior predicted by the OpenSees

models as the system was loaded beyond its design limit. Using the RC T-beam Bridge

again as an example, Figure 7.1 illustrates the consistency achieved between the

ABAQUS and the OpenSees models through a complete push-down analysis, in which

the bridge is loaded well into the inelastic range. Following this validation, the system

performance of the sample bridges was characterized statistically by propagating the

94
uncertainties in material strengths, stiffnesses and geometry through the OpenSees FE

models using a Latin Hypercube Sampling technique (Imam and Conover, 1980) to

achieve efficient coverage of the sample space with a manageable number (40) of FE

analyses. The random variables involved in these FE analyses are described by the

statistics defined in the LRFD databases as mentioned previously (Chapter 2). The limit

state of performance was taken as the point at which the bridge system exits the elastic

range, as identified from its load-deflection curve (cf Figure 7.1); that definition is

believed to represent the limit of usability/repairability of the bridge, but does not

represent life-threatening behavior if the structure possesses even a moderate amount of

ductility (cf Figure 7.1).

The flexural capacities so determined from this system reliability analysis were

rank-ordered and plotted on lognormal probability paper, as illustrated in Figure 7.2 for

the straight approach RC bridge (ID 129-0045). The lognormal distribution provides a

good fit to these data. The mean and coefficient of variation in the system capacity of

this bridge (at first yield) are 4.311 times the applied two HS-20 loads and 15%,

respectively. The variability is of the same order as the individual girder capacities

(Nowak, 1999), but the larger mean is characteristic of the beneficial system effects in a

system reliability assessment. Bridge system resistance distributions for other sample

bridges are summarized in the last column of Table 7.1. When used in a reliability

assessment with the same statistical load models used to develop the LRFD Bridge

1
Note that the mean value of bridge capacity obtained from the probability distribution of system capacity
is virtually identical to the bridge system capacity estimated in 7.1.1 with a deterministic analysis with all
parameters set equal to their mean values. One of the tenets of modern fragility analysis holds that the
mean response of an engineered system can be approximated by the response estimated by setting all
parameters equal to their mean values. Similar results have been found in seismic and wind fragility
modeling of building structures. Mathematically, the observation is tantamount to expanding the limit state
in a Taylor Series about the means and truncating the terms of order higher than unity.

95
Specifications, one obtains a system reliability index of 3.51 for the RC T-beam Bridge,

which is comparable to the safety level stipulated for a new bridge in AASHTO LRFD.

The rating factor based on the system capacity [cf Eq (2.6)] for the HS-20 vehicle at

Operating level is 1.74, presenting a 87% increase in rated load capacity comparing to

that calculated at the component level as stipulated in the AASHTO MBE. The

comparison of the ratings at system level and at the component level for other sample

bridges is tabulated in Table 7.2. It may be appropriate to factor in this additional

conservatism in bridge evaluation on a case-by-case basis, depending on the

consequences of the rating exercise.

0.995
R es is t anc e R (F E A )
0.99
Log norm al f it

0.95
0.9

0.75
Probability

0.5

0.25

0.1
0.05

0.01
0.005

0.001
0.5 0. 6 0. 7
10 10 10
Data

Figure 7.2 Lognormal Fit of System Resistance of the RC Bridge (ID: 129-0045)

96
 Table 7.1 Analysis of Bridge Capacity, Determined as the Point of First Yield

Elastic Ultimate Elastic System

Design Load Load Resistance
Bridge ID Type
load factor on factor on Distributions
HS20 HS20 (mean, cov)
RC – T – straight –
129-0045-0 H-15 4.31 4.80 LN(4.31, 0.146)
not posted

RC – T – skewed -
015-0108-0 HS-15 4.50 5.34 LN (4.50, 0.150)
posted

Prestressed –
223-0034-0 HS-20 5.94 6.87 LN (5.94, 0.108)
straight – not posted

Steel girder –
085-0018-0 H-15 5.37 5.71 LN (5.37, 0.111)
straight - posted

Table 7.2 Load Rating at Component Level vs System Level

(HS20 Operating Rating)

Component System Percentage

Bridge ID Type
level rating Level Rating of increase

RC – T – straight –
129-0045-0 0.93 1.74 87%
not posted
RC – T – skewed -
015-0108-0 2.00 2.61 31%
posted
Prestressed –
223-0034-0 1.72 2.94 71%
straight – not posted
Steel girder –
085-0018-0 1.14 2.54 122%
straight - posted

97
7.2 RATING USING SERVICE-PROVEN LOAD HISTORY

Many older bridges have performed well in service without any indication of

damage, but yet have been rated as structurally deficient without considering their

satisfactory performance over the years under the ever-increasing traffic volume and

truck weights. These bridges generally were designed for lower loads but for higher

factors of safety and, if well maintained, may have reliability levels that are equal to or

higher than those in modern construction. Surviving an extended service load history

which is stochastic in nature provides evidence of structural reliability that may be

comparable to what might be learned from a proof load test (Ellingwood, 1996; Stewart

and Val., 1999). Satisfactory service history should be considered, especially for old

bridges, in designing in-service inspection programs and in making decisions for

updating analytical ratings and load postings. The AASHTO MBE does not provide a

mechanism for updating structural resistance for service-proven bridges.

A proof test of a bridge enables the lower tail of the resistance distribution to be

truncated at the level of the maximum load carried as shown in Figure 7.3. In contrast,

for a service-proven bridge, the magnitude of the maximum load carried by the bridge

during its service history is unknown; however, it can be determined statistically using

the weigh-in-motion data described earlier (e.g., Nowak, 1999). For a structure surviving

a sequence of random vehicle loads, the magnitude of which is described by the

probability distribution function FQ(r) determined using weigh-in-motion data, the

revised strength f R'' (r ) can be written as (Ellingwood, 1996):

98
 FQ* (r ) f R (r )
f R'' (r ) = ∞
(7.1)
∫F
−∞
Q*
(r ) f R (r )dr

where fR(r) and FQ*(r) are the prior probability density function of resistance and the

cumulative load distribution function of the maximum load to occur during the service

period of interest, respectively. This updated density can be used in a structural

reliability assessment to determine the beneficial effect of successful service

performance.

To illustrate the benefit of prior successful bridge performance on rating, consider

the concrete T beam bridge (ID: 129-0045), which gave 51 years of serviceable

performance. Prior to considering the benefit of successful bridge performance, the use

of the mean and COV of bridge capacity presented by Nowak [1999] for new bridges in

Eqs (2.1) and (2.2) leads to the prior safety index β = 2.54. The updated distribution of

resistance, as determined from Eq 7.1 by Monte Carlo simulation, is illustrated in Figure

7.4. As a result, with the updated resistance, the estimated bridge failure probability

decreases and reliability index increases as the successfully service life of the bridge

increases, as illustrated in Figure 7.5. This increase in reliability translates to an increase

in bridge capacity rating factors, as indicated in Figure 7.6. Rating factors for this bridge

with respect to HL-93 design loading at inventory level, prior to and after considering the

51-year successful service life of this bridge, are summarized in Table 7.3. These results

indicate a 16% increase in flexural ratings and a 40% increase in shear ratings by

considering the 51-year service load history. Note that the rating in shear increases more

than that in flexure. The COV of the prior shear resistance is much larger than the COV

99
of prior resistance in flexure; knowledge of successful performance causes the effect of

the larger COV on the updated reliability to be diminished.

7.3 CLOSURE

The level-three analysis discussed in this Chapter has presented a structural

system reliability approach for bridge evaluation by proof load testing, which can be the

basis for possible improvements to the component-based LRFR option in the AASHTO

Manual of Bridge Evaluation. By using a FE-based bridge system reliability analysis,

ratings for the RC bridge ID: 129-0045 has increase by 86% percent comparing with the

component level analysis. Even just by considering this bridge’s 51-year successful

service life alone, ratings have increased by 16% in flexural and 40% in shear.

FE models for “virtual” proof load testing possess great potential in modern load

ratings, particularly in investigating the conservatisms that appear to be inherent in

traditional girder-based rating calculations and in avoiding the extensive cost and risk

associated with “real” in situ proof load testing. The feasibility of using finite element

analysis, validated through either systematic field inspection or through diagnostic load

tests, to conduct “virtual” proof load tests of bridge systems and support the improvement

of bridge evaluation practices, has been demonstrated in this Chapter.

100
 fL(r), fR(r)

Live Load Effect fL(r)

Initial Resistance fR(r)

Updated Resistance f ’R(r)

LP r
Figure 7.3: Structural Reliability Models for Bridge Proof Load Test

Figure 7.4 Influence of Service Load on Updated Distribution of Structural

Resistance for the RC Bridge (ID: 129-0045)

101
 Figure 7.5 Updated Failure Probabilities and Reliability Indices
for RC Bridge (ID: 129-0045)

Figure 7.6 Updated Rating Factors Respect to HL-93 at Inventory Level

for the RC Bridge (ID: 129-0045)

102
Table 7.3 Comparison of Rating Factors Computed Before and After Considering
Service Load History for RC Bridge (ID: 129-0045)

Rating Factor
Flexure Shear
Interior Exterior Interior Exterior
girder girder girder girder
Before updated by
0.75 0.65 0.45 0.46
service load history
After updated by
0.87 0.81 0.63 0.64
service load history

103
 CHAPTER 8
SUMMARY, CONCLUSIONS AND FURTHER WORK

8.1 SUMMARY

The proposed bridge rating framework developed in this dissertation addresses

condition assessment and evaluation by analysis, load test, or a combination of the two

methods, depending on the circumstances. Consistent with the AASHTO LRFD Bridge

Design Specifications, they have a sound basis in structural engineering and structural

reliability principles, allowing ratings to be updated as changing circumstances (traffic

demands, additional data, material deterioration, and other factors) warrant. The research

included the following four major activities: review and critical appraisal of existing

bridge rating procedures; bridge load testing; advanced bridge performance analysis

using finite element modeling; and structural component and system reliability analysis

to more closely couple the bridge rating to the in situ performance objectives.

Bridges that are typical those of current concern in rating and posting were selected

for load testing and analysis. Finite element models of these bridges were developed to

assist the design of the load tests and in the interpretation of the results. The bridge load test

results, in turn, were used to validate and improve the finite element modeling. The

measured bridge girder deflections in all cases were in good agreement with those

predicted by the FE model. This validated analytical and experimental combined approach

was then used to develop reliability-based framework to improve the current bridge rating

process.

104
 A three-level bridge rating framework was developed to provide bridge engineers

with several rating options. At the simplest level, the results of visual inspection are

incorporated in a format similar to that in the AASHTO Manual of Bridge Evaluation. The

higher levels involve more in-depth rating analyses that facilitate a customized rating of an

individual bridge by integrating in situ information into the capacity rating metric. This

study provides reliability-based methodologies and technical tools for performing ratings

systematically. Such in situ information can be obtained from material testing, condition

rating records, load tests, successful service life performance, and system reliability

analysis. The higher level ratings emphasize the importance of learning from inspections

in rating a given bridge and provide clear incentives to obtain quantitative in situ

measurements through field inspections, load tests and other modern non-destructive

evaluation technology.

8.2 RESEARCH CONCLUSIONS

Application of the proposed evaluation framework to existing steel, reinforced

concrete or pre-stressed concrete bridges indicated that posting requirements based on

current bridge evaluation practices, which do not incorporate available site-specific

knowledge in any detail, can be unduly conservative from a structural reliability

viewpoint. The proposed improvements recognize the uniqueness of an individual bridge

and take advantage of accessible in situ information to the extent feasible to produce

bridge ratings that provide for public safety without undue economic impact on the

community served.

The load tests and supporting analysis indicated that utilizing current girder

distribution factors can yield conservative measures of actual load-carrying capacity. This

105
conservatism is the result of assumptions made in the analysis regarding load sharing,

composite action, support conditions and nonlinear behavior, as well as the differences in

material strengths. In view of the economic consequences of posting, it is apparent that if

the customary rating practice suggests that a bridge is a candidate for posting, a more

accurate structural analysis model should be employed to verify whether more accurate

GDFs might change that decision.

The load-carrying mechanism in reinforced concrete pier caps that have short shear

spans and behave as deep beams is better described by the strut and tie model than by the

traditional ACI Standard 318 model. The use of this new capacity calculation method is

permitted by the LRFR option in the AASHTO MBE. A preliminary investigation of

similar bridge pier caps indicated that the level of conservatism is dependent on the

dimensions of the individual pier caps and the placement of the girders that they support.

Routine bi-annual inspections play an important role in bridge condition

assessment by providing in situ data to support the real-time bridge reliability estimate

and to assist the decision making regarding suitable maintenance. These inspections are

mostly completed using visual inspection techniques, do not result in quantitative

estimates of deterioration, and therefore are difficult to incorporate in a strength updating

process based on modern structural reliability assessment methods. This study provides a

method for linking qualitative visual inspections to quantitative reliability-based load

rating. A new set of condition factors was developed to couple the rating procedure more

closely to the results of bridge inspections

A properly conducted proof load test as well as bridge’s successful service history

can be effective way to assess the structural behavior of a bridge as a system and to

106
update the bridge load capacity at the system level, especially in situations where the

analytical approach at the first and second levels produces questionable ratings, or

structural analysis is difficult to perform due to deterioration or lack of documentation.

FE analyses conducted as “virtual” proof load tests possess great potential in modern load

rating, particularly in eliminating the conservatisms that appear to be inherent in

traditional girder-based rating calculations and in avoiding the risk and cost associated to

a “real” in situ proof load test. The feasibility of using finite element modeling, validated

through either systematic field inspection or through diagnostic load tests, to conduct

“virtual” proof load tests of bridge systems and support the improvement of bridge

evaluation practices, has been demonstrated in the study.

Experience in conducting the load tests suggested that basing the performance

assessment of an existing bridge on global response measurements, such as displacement,

as opposed to local responses, such as strain, minimizes the likelihood that spurious local

non-homogeneous or material behavior may lead to false conclusions regarding structural

system safety. It was also observed that redundancy in measurements, through multiple

gauges at a single location and gauges at multiple locations in a single element, is essential

to accurate condition assessment and should be utilized whenever practical.

8.3 RECOMMENDED FUTURE WORK

In the course of the research conducted in this dissertation, several topics worth

further investigation have been identified:

In bridge rating by the LRFR method, the performance based evaluation

philosophy is embedded in the load and resistance factors. Each load or resistance factor

should have a sound reliability basis and should enable performance objectives that are

107
consistent with current practice to be achieved. The system factor appearing in the LRFR

rating equation (Eq 2-6) should account for the reserve capacity and redundancy of the

bridge in the rating calculation, but is based on a component level rather than system

level of analysis. The system reliability analysis conducted herein demonstrates the

inadequacies of this component-based approach. The development and basis of the

system factors in LRFR should be more closely examined.

The proposed multi-level rating approach clearly highlights the learning process

of a given bridge through field inspections. Current bridge inspection programs seldom

do an adequate job of reflecting in situ condition quantitatively, making it difficult to

incorporate the results into the bridge capacity rating process. The use of modern non-

destructive evaluation (NDE) techniques, such as Carpenter hammer sounding, Schmidt

rebound hammer and ultrasonic pulse velocity, should be encouraged in bridge inspection

activities, as they may permit a revised estimate of in situ strength to be used in the

bridge rating calculations. The benefit and the cost of introducing these technologies

should be considered in designing in-service inspection programs and in maintenance

decisions. Inspection guidelines corresponding to these NDE techniques should be

established to coordinate with the updating of rating calculations.

While each existing bridge structure is unique, many bridge types may share

common failure mechanisms and may be governed by a relatively few parameters,

especially when a group of bridges is constructed from a limited numbers of standard

designs. It is therefore possible to classify bridges by identifying these governing

parameters and their variation within the population and to develop bridge type-specific

strategies for load rating and condition assessment. The type-specific strategy could

108
provide significant advantage for inspection and load rating of bridges shearing common

material, similar geometry and detailing, and the same critical behavior mechanisms.

Much of the development of the reliability assessment of bridges has focused on

the performance of the bridge superstructure. Less effort has been spent on studying

reliability of bridge substructures (piercap, piers and columns) and connections. Bridge

deficiencies may become apparent under seismic, hurricane and other extreme events, in

addition to traffic demands. Efforts should be increased to collect data on the bridge

substructure/connection behavior under extreme events in a time-dependent manner.

These results should be included in condition assessment framework expressed in terms

of reliability.

A decision engine to supplement the tiered rating framework is necessary for

authorities to maintain the functionality of the bridge system in a cost-effective manner.

If a bridge is rated below the minimum acceptance level, the assessing engineer should

examine the options of possible maintenance strategies on a future cost (or life-cycle

cost) basis. The bridge authority must have an internally consistent decision-making

system which interprets the rating results, estimates the cost of each alternative strategy,

examines the overall objective of bridge management in terms of risk, and assigns

priorities for bridge maintenance. The bridge condition assessment process will be able

to determine not only the structural adequacy of the bridge at the present time, but will

provide technical support for financial risk management strategies and future

maintenance options.

109
 REFERENCES

ACI 318-05 (2005), Building Code Requirements for Structural Concrete. American
Concrete Institute

Ang, A. H.-S. and Cornell, C.A. (1974) Reliability bases of structural Safety and Design,
Journal of Structural Division, ASCE, 100(3): 1755-1769.

Ang, A. H.-S. and Tang, W. H. (2007) Probability Concepts in Engineering Planning

and Design, Vol. I, Basic Principles, John Wiley

Ariyaratne, W. (1984) Case history of bridge tested in NSW since 1995. In: Chirgwin GJ,
editor. Proceedings of the AUSTROADS 1997 Bridge Conference ‘Bridging the
Millennia’, Sydney: AUSTROAD Inc., 1997; 3:283-90

Albrecht, P. and Naeemi, A. H. (1984) “Performance of weathering steel in bridges.”

NCHPR Rep. 272, Transportation Research Board, Washington D.C.

American Association of State Highway and Transportation Officials (2007). LRFD

Bridge Design Specifications, 4th Edition (including 2008 and 2009 interim revisions).
AASHTO, Washington D.C.

American Association of State Highway and Transportation Officials (1994). Manual for
Condition Evaluation of Bridges, 2nd Edition (including 1995, 1996, 1998 and 2000
interim revisions). Washington, DC: AASHTO, 2000.

American Association of State Highway and Transportation Officials (2003). Guide

Manual for Condition Evaluation and Load and Resistance Factor Rating (LRFR) of
Highway Bridges, 1st Edition (including 2005 interim revisions). AASHTO,
Washington D.C.

American Association of State Highway and Transportation Officials AASHTO (2008).

Manual for Bridge Evaluation, 1st Edition. American Association of State Highway
and Transportation Officials, Washington, DC.

Bakht, B. and Jaeger, L.G. (1990) “Bridge testing – a surprise every time.” J. Struct.
Engrg. ASCE 116(5):1370-1383.

Barker, M. G. (2001) “Quantifying Field-Test Behavior for Rating Steel Girder Bridges.”
Journal of Bridge Engineering, 6(4): 254-261.

110
Bartlett, M. F. and Sexsmith, R. G. (1991) “Bayesian technique for evaluation of material
strengths in existing bridges.” ACI Materials Journal, 88(2): 164-169.

Bartlett, M. F. and MacGregor, J. G. (1996). “Statistical Analysis of the Compressive

Strength of Concrete in Structures.” ACI Materials Journal, 93(2): 158-168.

Benjamin, J.R. and Cornell, C.A. (1970) Probability, Statistics, and Decision for Civil
Engineers, MrGraw- Hill Book Company

Bhattacharya, B., Li, D., Chajes, M., and Hastings, J. (2005). “Reliability-Based Load
and Resistance Factor Rating Using In-Service Data.” Journal of Bridge Engineering,
10(5): 530-543

Bolukbasi, M., Mohammadi, J and Arditi, D (2004). “Estimating the Future Condition of
Highway Bridge Components Using National Bridge Inventory Data.” Practical
Periodical on Structural Design and Construction, ASCE, Vol. 9, No. 1: 16-24

Chen, X. and Lind, N. C. (1983) Fast probability integration by three parameter normal
tail approximation, Structural Safety, 1(4): 269-276

Ditlevsen, O. (1981) Principle of normal tail approximation, Journal Mechanics Division,

ASCE, 107 (EM6), 1191-1208

Ellingwood, B., Galambos, T.V., MacGregor, J. C., and Cornell, C. A. (1980)

Development of a probability based load criteria for American National Standard A58,
NBS Special Publication 577, U.S. Department of Commerce, National Bureau of
Standard, Washington, D.C.

Ellingwood, B., Galambos, T.V., MacGregor, J. C., and Cornell, C. A. (1982).

“Probability based load criteria: load factors and load combinations.” Journal of
Structural Division, ASCE, 108(5): 978-997.

Ellingwood, B. (1996). “Reliability-Based Condition Assessment and LRFD for Existing

Structures.” Structural Safety, 18(2): 67-80.

Ellingwood, B. (2000). “LRFD: implementing structural reliability in professional

practice.” Engineering Structures, 22: 106-115

Ellingwood, B.R., Zureick, A., Wang, N. and O’Malley, C. (2009). Condition assessment
of existing bridge structures, Task 4 - State of the art of bridge condition assessment.
Report of GDOT Project RP 05-01. Georgia Department of Transportation, Atlanta,
GA (ftp://ftp.dot.state.ga.us/DOTFTP/Anonymous -Public/Research_Projects/) 05/10.

Enright, M. P. and Frangopol, D. M. (1998a) Service life prediction of deteriorating

concrete bridges, J. Structural Engineering, 124 (3)

111
Enright, M. P. and Frangopol, D. M. (1998b) “Probabilistic Analysis of Resistance
Degradation of Reinforced Bridge Bean under Corrosion.” Engineering Structure, 20
(11) 960-971

Estes, A.C. and Frangopol, D.M. (1999) Repair optimization of highway bridges using
system reliability approach, J. Structural Engineering, ASCE, 125(7), 766-775

Faber, M. H., Val, D. V. and Stewart, M.G. (2000). “Proof load testing for bridge
assessment and upgrading.” Engineering Structures, 22(12): 1677-1689.

Fu, G. and J. Tang (1995) “Risk-based proof-load requirements for bridge evaluation.”
Journal of Structural Engineering, 121(3): 542-556.

Ghosn, M. (2000) Development of truck weight regulations using bridges reliability

models, J. bridge Engineering, ASCE, 5 (4), 293-303

Galambos, T. V., Ellingwood, B., MacGregor, J. C., and Cornell, C. A., "Probability
Based Load Criteria: Assessment of Current Design Practice," Journal of the Structural
Division, ASCE, 108 (5) 959-977.

Hall, W. B. (1988) Reliability of service-proven structures, J. Structural Engineering,

ASCE, 114 (3), 608-624

Hasofer, A. and N.C. Lint (1974) Exact and Invariant Second-moment Code Format,
Journal of Engineering Mechanical Division, ASCE, 100(1): 111-121.

Hawkins, N. M., Kuchma, D. A., Mast, R. F., Marsh, M. L., Reineck, K.-H., (2005).
NCHRP Report 549: Simplified Shear Design of Structural Concrete Members,
Transportation Research Board, Washington, D.C.

Hillerborg, A., M. Modeer, and P. E. Petersson (1976). “Analysis of Crack Formation

and Crack Growth in Concrete by Means of Fracture Mechanics and Finite Elements.”
Cement and Concrete Research 6: 773–782.

Iman, R. L. and Conover, W. J. (1980) “Small sample sensitivity analysis techniques for
computer models with an application to risk assessment.” Communi Statist. Theor.
Meth., A9 (17):1947-1842

Jiang, M. and Sinha, K.C. (1989) “Bridge service prediction model using the Markov
chain.” Transp. Res. Rec. 1223, Transportation Research Board, Washington, D.C.

Kim S. and Nowak A. S. (1997). “Load distribution and impact factors for I-Girder
bridges.” J of Bridge .Engineering, Vol. 2, No.3

Kupfer, H. B., and K. H. Gerstle (1973). “Behavior of Concrete under Biaxial Stresses.”
Journal of Engineering Mechanics Division, ASCE 99: 853-864.

112
Lind, N.C. (1977) Formulation of probabilistic design, J.Eng Mech. Div, ASCE, 103, No.
EM2.

McCuen, R. H. and Albrecht, P. (1995) “Composite modeling of atmosphere corrosion

penetration data.” Application of Accelerated Corrosion Tests to Service Life
Prediction of Materials, ASTM STP 1194, Philadelphia, 65-102

McKenna, F. T. (1997). “Object-oriented finite element programming: frameworks for

analysis, algorithms and parallel computing.” Ph.D. Thesis, University of California,
Berkeley, CA, United States.

Melchers, R., E. (1994) Structural system reliability assessment using directional

simulation, Structural Safety, 16(1, 2), 23-38

Melchers, R., E. (1989) Importance sampling in structural systems, Structural Safety, 6 3-

10

Melchers, R., E.(1999). Structural Reliability Analysis and Prediction, 2nd Ed., Wiley,
Chichester, England

Melchers, R., E. (1999) Reliability of deteriorating RC slab bridges, Journal of Structural

Engineering, 123(12), 1638-1644

Melchers, R., E. (2001). “Assessment of existing structures - approaches and research

needs.” Journal of Structural Engineering , 127(4), 406-411

Minervino, C., Sivakumar, B., Moses, M., Merza, D., and Edberg, W. (2004) New
AASHTO guide manual for load and resistance factor rating of highway bridges,
Journal of Bridge Engineering, 9(1): 43-54.

Mori, Y. and Ellingwood, B. (1993). “Reliability-Based Service Life Assessment of

Aging Concrete Structure.” Journal of Structural Engineering, 119(5): 1600-1621.

Mori, Y. and Ellingwood, B. (1994a) “Maintaining reliability of concrete structures I:

role of inspection/ repair.” Journal of Structural Engineering, 120(3): 824-845.

Mori, Y. and Ellingwood, B. (1994b) “Maintaining reliability of concrete structures II:

optimum inspection/repair.” Journal of Structural Engineering, 120(3): 824-845.

Moses, F. and D. Verma (1987). “Load capacity evaluation of existing bridges,” NCHRP
Report 301, National Research Council TRB, National Academy Press, Washington,
DC.

Moses, F., Lebet, J. P. and Bez, R. (1994). “Applications of field testing to bridge
evaluation.” Journal of Structural Engineering, 120(6): 1745-1762.

113
Moses, F., (2001) “Calibration of Load Factors for LRFR Bridge Evaluation.” NCHRP
Report 405, Transportation Research Board, National Research Council, Washington
D.C.
Madsen, H.O. (1987) Model updating in reliability analysis, Proc. ICASP5, Vancouver,
BC, 564-577.

National Bridge Inspection Standard (1996) Code of federal regulations, No. 23CFR650,
US government Printing Office, Washington, D.C., 238-240

National Cooperative Highway Research Program (NCHRP Web Document 28) (2001).
"Manual for Condition Evaluation and Load Rating of Highway Bridges Using Load
and Resistance Factor Philosophy" NCHRP 12-46. Lichtenstein Consulting Engineers,
Inc. Paramus, NJ

Nowak, A. S. and Tharmabala, T. (1988) “Bridge reliability evaluation using load test.”
Journal of Structural Engineering, 114(10): 2268-2279.

Nowak, A. S. (1993) “Live Load Model for Highway Bridges.” Structural Safety,
13(1993): 53-66.

Nowak, A.S. (1999) “Calibration of LRFD Bridge Design Code.” NCHRP Report 368,
Transportation Research Board, National Research Council, Washington D.C.

O’Malley, C., Wang, N., Ellingwood, B.R. and Zureick, A. (2009). “Condition
assessment of existing bridge structures: Report of Tasks 2 and 3 - Bridge Load
Testing Program.” Report of GDOT Project RP 05-01. Georgia Department of
Transportation, Atlanta, GA (ftp://ftp.dot.state.ga.us/DOTFTP/Anonymous-Public/
Research Projects/). 05/10

Phares, A. M., Washer, G. A., Rolander, D. D., Graybeal, B. A. and Moore, M. (2004)
Routine highway bridge inspection condition documentation accuracy and reliability, J.
of Bridge Engineering, ASCE, 9 (4) 403-413

Saraf, V. and Nowak, A. S. (1998), Proof Load Testing of Deteriorated Steel Girder
Bridges, Journal of Engineering Mechanics, ASCE 3(2): 82-89.

Shinozuka, M. (1983) Basic Analysis of Structural Safety, Journal of Structural

Engineering, ASCE 109(1): 721-740.

Simulia (2006), Abaqus Analysis User's Manual, Version 6.8, Dassault Systèmes Simulia
Corp. Providence RI, USA.

Standard Test Method for Obtaining and Testing Drilled Cores and Sawed Beams of
Concrete, (ASTM C 42-94) (1995) Annual Book of ASTM Standards, American
Society for Testing and Materials, Philadelphia, pp. 24-27

114
Stewart, M. G. and Val, D. V. (1999) Role of load history in reliability-based decision
analysis of aging bridges, Journal of Structural Engineering, ASCE 109(7)

Tabsh Sami, W. and Nowak A.S. (1991) Reliability of highway girder bridges, Journal of
Structural Engineering, ASCE 117(8), 2372-2388

Tang, C. Y., Tan, K. H. (2004). “Interactive mechanical model for shear strength of deep
beams,” J. Struct. Engrg., ASCE 130(10), 1534-1544.

Thoft-Christensen P. and M. J. Baker (1982) Structural Reliability and Its Application,

Springer-Verlag, Berlin

Thoft-Christensen P. (1995) Advanced bridge management systems, Structural

Engineering Review, 7(3), 151-163

Thoft-Christensen P. (1998) Life time reliability assessment of concrete slab bridges,

Proc., Optimal Performance of Civil Infrastructure systems, D. M. Frangopol, ed.,
ASCE, Reston, Va., 181-193

Todeschini, C., A. Bianchini, and C. Kesler (1964). “Behavior of concrete columns

reinforced with high strength steel”. ACI Materials Journal 61(6): 701-716

Val, D. V. and Melchers, R. E.(1997). “Reliability of deteriorating RC slab bridges”

ASCE, J. of Structural. Engineering, Vol 123, No.12 1638-1644

Wang, N., Ellingwood, B.R., Zureick, A. and O’Malley, C. (2009). “Condition

assessment of existing bridge structures: Report of Task 1 – Appraisal of state-of-the-
art of bridge condition assessment.” Report of Project GDOT No. RP05-01, Georgia
Department of Transportation, Atlanta, GA (ftp://ftp.dot.state.ga.us/DOTFTP/
Anonymous-Public/Research_Projects/) 05/2010

Wang, N., Ellingwood, B.R. and Zureick, A.(2010). Development of bridge rating
protocols using system reliability assessment, Part II: Reliability-based bridge rating. J.
Bridge Engrg., ASCE (under review).

Wang, N., Ellingwood, B. R., Zureick, A.-H. and O’Malley, C. (2010). “Development of
bridge rating protocols using system reliability assessment Part I: Analysis and
Verification by Load Testing.” J. Bridge Engrg., ASCE (under review).

Washa, G. W. and Wendt, K.F (1975) “Fifty Year Properties of Concrete.” ACI Journal
Proceedings, Vol 72, No.1 20-28

115