Corrosion and Seismic Analyses of Reinforced Concrete Bridges

by

Hanmin Wang

January 2022

A dissertation submitted to the

faculty of the Graduate School of
the University at Buffalo, The State University of New York
in partial fulfillment of the requirements for the
degree of

Doctor of Philosophy

Department of Civil, Structural and Environmental Engineering

 Copyright by

Hanmin Wang

2022

All Rights Reserved

ii
To my family.

iii
 Acknowledgments

First and foremost, I am extremely grateful to my supervisors, Dr. Ravi Ranade and Dr.

Pinar Okumus, for their guidance, support and patience over the course of my doctoral studies.

They inspired me to think independently in my research and encouraged me when I had a difficult

time in my life. It was my fortune to work with them over the past 3.5 years.

I would like to express gratitude to my Ph.D. committee members, Dr. Andrew Whittaker

and Dr. Negar Elhami Khorasani. Their valuable comments and review of my dissertation have

improved the quality of this research and enhanced my understanding of earthquake engineering

and probabilistic analysis.

I would like to extend my gratitude to all the faculty members and students of the Institute

of Bridge Engineering (IBE) at the University at Buffalo (UB). The monthly meetings of IBE

improved my skills in communication and presentation and provided me with opportunities to

learn many interesting research topics.

I appreciate all the faculty members in the Department of Civil, Structural and

Environmental Engineering at UB. Because of their interesting courses, I accumulated

fundamental knowledge in civil engineering, which further benefited my Ph.D. research. I would

also like to thank Professor Darrell Kaminski for his advice on my career development.

iv
 I am thankful to all my friends and colleagues at UB. I appreciate every discussion with

Dr. Shoma Kitayama on many interesting topics during my lunchtime. I would also like to thank

Dr. Cancan Yang, Dr. Xuan Gao, Dr. Nan Hua, Dr. Chingching Yu, Dr. Haifeng Wang, Dr. Yushan

Fu, Mr. Shaopeng Li, Mr. Rahul Raman, Mr. Amr Soliman, Mr. Hyunmyung Kim, Mr. Rodrigo

Castillo and Mr. Seyed Omid Sajedi for their help and friendship during my days at UB. I am

especially grateful to Mr. Yunpeng Shi for his generous help and support during my recovery phase

after surgery. My life at UB was memorable and fun because of them.

Finally, I would like to thank my parents, Ping Wang and Yuefeng Han, for their love,

support and encouragement throughout my entire life. This accomplishment would not be possible

without them. I dedicate this dissertation to them.

v
 Abstract

This dissertation focuses on the long-term behavior of RC highway bridge piers exposed

to corrosion and seismic events. The specific objectives are: (1) Present a systematic framework

for analyzing RC bridge piers subjected to corrosion deterioration and earthquake, and utilize the

framework to investigate the improvement in fragility functions of RC piers due to improved

durability of the cover material; (2) Incorporate local environmental conditions and vehicle spray

and splash mechanisms into a corrosion model for improved bridge durability assessment; and (3)

Calibrate the developed corrosion model using field-based bridge assessment data.

For objective (1), a systematic framework comprising of a corrosion model and a structural

model was presented for RC bridge piers. The corrosion model was used to estimate concrete

cracking and reinforcement area loss as functions of time, which were used in the structural model

for obtaining seismic fragility functions of RC bridge piers at discrete time points during their

service lives. The corrosion model accounted for pitting corrosion and the influence of cover

cracking on the corrosion rate. The influence of using a ductile fiber-reinforced concrete (FRC)

for cover of RC bridge piers on seismic fragility functions was investigated. Results showed that

corrosion could have a significant impact on the seismic fragility functions for RC bridge piers.

The improvement in durability of the RC bridge pier enabled by FRC cover translated into a lower

probability of seismic damage over time. A parametric study showed that the pre-crack corrosion

rate and longitudinal reinforcement ratio were the most influential parameters for the example pier

fragility functions, whereas corrosion initiation time and cover thickness had negligible effects

within the domains investigated in this study. The framework was further applied to evaluate eight

vi
bridges from Washington State Department of Transportation to demonstrate its use of the

framework on a bridge inventory. The results showed that although corrosion deterioration

affected all bridges in the group, some bridges could become more vulnerable to seismic damage

over time due to exposure conditions and pier configurations.

For objective (2), the developed corrosion model was improved by proposing a new

method for estimating chloride exposure (and thereby corrosion susceptibility) of bridges

individually based on unique local characteristics of the distance between roadside and structure,

snow precipitation, salt application, and traffic patterns. This method correctly differentiated

between bridges with high and low chloride exposure, and thus removed the need for assuming

the same chloride exposure for all bridges in the same region (the current practice).

To achieve objective (3), a novel method is proposed for linking the physics-based

corrosion model to field-based condition assessments. Corrosion-model predictions (i.e., surface

crack width and concrete spalling) and deterioration curves, based on bridge element condition

ratings, were correlated based on transportation agency guidelines. To demonstrate the process of

calibrating the input parameters of the corrosion model, crack widths and areas of spalling from

deterioration curves and the physics-based corrosion model were compared for an RC bridge. Input

parameter ranges that lead to a match between the corrosion model and field data predictions were

identified.

vii
 Table of Content

Dedication iii

Acknowledgments iv

Abstract vi

List of Figures xiii

List of Tables xv

Nomenclature xvi

Chapter 1: Introduction 1

1.1 Background and motivation 1

1.2 Scope of the research 2

1.3 Research goals and objectives 4

1.4 Dissertation organization 5

Chapter 2: Literature Review 7

2.1 Review of corrosion modeling 7

2.1.1 Corrosion initiation phase 8

2.1.2 Corrosion propagation phase 9

2.2 Review of seismic fragility analysis 10

2.3 Review of fragility analysis considering corrosion effects 11

Chapter 3: Combining Corrosion and Seismic Fragility Analyses to Investigate the Impact

of Rebar Cover Material 13

3.1 Introduction 13

3.2 Corrosion model 15

viii
 3.2.1 Corrosion initiation 16

3.2.2 Corrosion propagation 16

3.2.3 Corrosion model properties for the example bridge column 24

3.3 Seismic fragility of a corroded bridge column 27

3.3.1 Ground motion selection (step 1) 27

3.3.2 Structural model (step 2) 28

3.3.3 IDA (step 3) 29

3.3.4 Time-dependent damage states and damage index (step 4) 29

3.3.5 Estimation of the parameters of fragility curves (step 5) 32

3.4 Framework as applied to a bridge column 32

3.4.1 Finite element model 33

3.4.2 Corrosion and fragility results for the example column 37

3.5 Parametric study 44

3.5.1 Investigated parameters 44

3.5.2 Corrosion parameters 45

3.5.3 Sensitivity of the corrosion model to the corrosion parameters 46

3.5.4 Structural parameters 49

3.5.5 Sensitivity of the fragility curves to corrosion parameters 49

3.5.6 Sensitivity of the fragility curves to structural parameters 52

3.6 Conclusions 58

Chapter 4: Seismic Damage Risk Assessment of Aging Bridges 60

4.1 Introduction 60

4.2 Corrosion Modeling 60

ix
 4.2.1 Parameters used in corrosion modeling 61

4.2.2 Outcomes of corrosion modeling 62

4.3 Seismic Fragility Analysis 62

4.3.1 Development of fragility curves for various bridge ages (Step 6) 63

4.4 Integration of Seismic Hazard and Fragility Curves 64

4.4.1 Method to generate seismic hazard curves 65

4.4.2 Method to integrate seismic hazard curves and fragility curves 66

4.4.3 Outcomes of integration of seismic hazard and fragility curves 67

4.5 Bridge Details and Modeling Strategies 68

4.5.1 Bridge details 68

4.5.2 Modeling strategies 73

4.6 Results and discussion 78

4.6.1 Corrosion model results 78

4.6.2 Seismic damage probability 80

4.7 Conclusions 84

Chapter 5: Estimating Chloride Exposure of Reinforced Concrete Bridges Using Vehicle

Spray and Splash Mechanisms 86

5.1 Introduction 86

5.2 The proposed method 89

5.2.1 Deicing salts application rate 89

5.2.2 Water film thickness 90

5.2.3 Water sprayed and splashed by one heavy-duty vehicle 92

5.2.4 Chloride ions sprayed and splashed by one heavy-duty vehicle 96

x
 5.2.5 Chloride ions sprayed and splashed by all vehicles in one winter season 96

5.2.6 Chloride ions deposition on bridge substructure 98

5.2.7 Assumptions 99

5.2.8 Application 101

5.3 Demonstration of the Proposed Model with an Example 101

5.4 Comparison of model estimates with experimental results 104

5.4.1 Experimental data 104

5.4.2 Analytical model prediction 106

5.4.3 Model prediction versus test data 107

5.5 Effects of parameters on the surface chloride content 110

5.6 Conclusions 113

Chapter 6: Linking the Physics-Based Deterioration Model to Field-Based Condition

Assessments 116

6.1 Introduction 116

6.2 Corrosion modeling for predicting concrete spalling 119

6.2.1 Simplified concrete spalling geometry 119

6.2.2 Pressure for concrete spalling 120

6.3 Bridge inspection program and condition rating 122

6.4 Bridge deterioration curves 125

6.5 Deterioration interpretation from bridge element condition ratings 126

6.6 Corrosion model calibration 129

6.7 Demonstration of the Calibration Process with an Example 132

6.8 Effects of different corrosion input parameters on crack width 137

xi
 6.9 Summary and conclusions 143

Chapter 7: Concluding Remarks 145

7.1 Summary and conclusions 145

7.2 Scientific contributions and research impacts 148

7.3 Recommendations for future research 150

References 153

xii
 List of Figures

Figure 2.1: Reinforcement corrosion 7

Figure 2.2: Chemical reactions in the corrosion propagation phase 9

Figure 3.1: Pitting corrosion 18

Figure 3.2: Corrosion rate change with time 23

Figure 3.3: Flowchart of determining the slope of corrosion rate with time 23

Figure 3.4: Column model components 34

Figure 3.5: Material properties used in the finite element model 36

Figure 3.6: Rebar area loss and cover crack width with time 39

Figure 3.7: Fragility curves of the example bridge column with concrete cover 40

Figure 3.8: Fragility curves of the example bridge column with ECC cover 41

Figure 3.9: Comparison of fragility curves obtained using conventional concrete and ECC covers

at 0 years and 75 years 43

Figure 3.10: Effects of the corrosion parameters on remaining longitudinal reinforcement area

after 75 years 48

Figure 3.11: Influence of corrosion initiation time of longitudinal reinforcement on column

fragility over time 50

Figure 3.12: Influence of pre-crack corrosion rate on column fragility over time 51

Figure 3.13: Influence of K on column fragility over time 52

Figure 3.14: Influence of cover thickness on column fragility over time 53

Figure 3.15: Influence of concrete compressive strength on column fragility over time 55

Figure 3.16: Influence of longitudinal reinforcement ratio on column fragility over time 56

xiii
Figure 3.17: Influence of transverse rebar volumetric ratio on column fragility over time 57

Figure 4.1: Seismic hazard curve for an example bridge in Washington State 65

Figure 4.2: Seismic hazard function given for an example bridge 67

Figure 4.3: Layouts of WSDOT bridge substructures 73

Figure 4.4: Modeling simplification of the bridge pier 74

Figure 4.5: Model details and material models 77

Figure 4.6: Damage probability for bridges from WSDOT 83

Figure 5.1: Mechanisms of vehicle spray and splash 92

Figure 5.2: Location of specimens for field tests for bridges L1, L2 and L3 106

Figure 5.3: Comparison of 𝐶 from field observations and model estimates 108

Figure 5.4: Effects of parameters on surface chloride content 112

Figure 6.1: Simplified concrete spalling model shown in cross-section view 119

Figure 6.2: Calibration framework 131

Figure 6.3: Demonstration of the calibration process 132

Figure 6.4: Calibration of corrosion rate before cracking 133

Figure 6.5: Effect of varying 𝜆(𝑡) on crack width evolution 139

Figure 6.6: Effect of varying R on crack width evolution 140

Figure 6.7: Effect of varying K on crack width evolution 140

Figure 6.8: Effect of corrosion initiation time on crack width evolution 141

Figure 6.9: Effect of 𝑊 on crack width evolution 142

xiv
 List of Tables

Table 3.1: Parameters considered in the corrosion model of the example bridge column 26

Table 3.2: Repair strategies for each damage state in Caltrans Seismic Design Criteria 31

Table 3.3: Logarithmic median and standard deviation of ln(IM) of fragility functions for the

example bridge column with different materials 43

Table 3.4: Parameters considered in the parametric study 45

Table 4.1: Parameters used in corrosion model 62

Table 4.2: Summary of WSDOT bridges 69

Table 4.3: Model details of bridges 76

Table 4.4: Corrosion initiation time 79

Table 4.5: Remaining rebar area 80

Table 5.1: Model parameters for the piers of the example bridge 102

Table 5.2: Input parameters for different field tests 107

Table 6.1: Condition rating guidelines of NYSDOT bridge component evaluation 124

Table 6.2: AASHTO ECR guidelines 124

Table 6.3: Condition states of reinforced concrete elements based on surface crack widths 127

Table 6.4: Condition states of reinforced concrete elements based on areas of concrete spalling

127

Table 6.5: Relationship between NYSDOT ECR and crack width 129

Table 6.6: Model parameters of the bridge column 135

Table 6.7: Values of the corrosion model input parameters that led to a reasonable match with

field-based data 138

xv
 Nomenclature

𝐴 mm2 Original rebar area

Area parameter 1 depending on D0, 𝑃(𝑡), 𝜃 and 𝑏 in
𝐴 mm2
pitting corrosion model
Area parameter 2 depending on D0, 𝑃(𝑡), 𝜃 and 𝑏 in
𝐴 mm2
pitting corrosion model
𝐴(𝑡) mm2 Rebar area at time t
Ratio of exceedance frequency at a given PGA to the
𝐴 - exceedance frequency at a PGA that is 10-times
smaller than this PGA
ADT vehicles/day Total average daily traffic volume per day
ADTT vehicles/day Traffic volume of heavy-duty vehicles per day
b m Tire width
𝑏 mm Width of the pit area
𝐶 mm Thickness of concrete cover
𝐶 kg/m3 Critical chloride content
Chloride concentration at spatial coordinate x, and
𝐶 (𝑥, 𝑡) kg/m3
time t
𝐶 kg/m3 Chloride content at the concrete exterior surface
Amount of chloride ions sprayed by vehicles per
𝐶_ kg/m3/winter
winter season in kg/m3/winter
Amount of chloride ions splashed by vehicles per
𝐶 _ kg/m3/winter
winter season in kg/m3/winter
Amount of chloride ions generated by vehicle spray
𝐶_ kg/m3/winter
and splash per winter season
𝐷 mm2/s Effective chloride diffusion coefficient
D0 mm Original rebar diameter
Distance between the roadside and bridge pier or
𝑑 m
structure in m

xvi
 The limiting value of EDP used to define a damage
𝑑 -
level
𝑑 mm Aggregate size of concrete
𝑑 mm Reference aggregate size of cocnrete
𝐸 MPa Elastic modulus of concrete
𝐸 MPa Elastic modulus of rust
EDP - Engineering demand parameter
erf (. ) - Error function
𝑓 MPa Yield strength of non-corroded reinforcement
𝑓(𝑡) MPa Yield strength of corroded rebar at time t
𝑓 (𝑡) MPa Yield strength of transverse rebar at time t
fc MPa Compressive strength of unconfined concrete
𝑓 MPa Tensile strength of concrete
𝑓 (𝑡) MPa Compressive strength of confined concrete at time t
𝐻(𝑎) - Probability of exceeding of a PGA (𝑎)
ℎ m/day Water thickness on the road surface per day
ℎ m Depth of water film picked up on each rotation
ℎ m/winter Snowfall precipitation per winter season
𝐼𝑀 - Ground motion intensity measure
Factor that indicates the proportion of the tire width
𝑘 -
that is not a groove to the total tire width
Mass loss of steel per unit length consumed to
𝑀 kg
produce rust
𝑀 kg Mass of rust per unit length of one rebar
Total amount of deicing salts applied per winter
𝑀 kg/m2/winter
season
Amount of deicing salts applied per day during winter
𝑀 kg/m2/day
season
Mass flow rate of water generated by bow wave per
𝑀𝑅 kg/s
heavy-duty vehicle in kg/s

xvii
 Mass flow rate of water generated by capillary
𝑀𝑅 kg/s
adhesion per heavy-duty vehicle in kg/s
Mass flow rate of water generated by side wave per
𝑀𝑅 kg/s
heavy-duty vehicle in kg/s
Mass flow rate of water generated by tread pickup per
𝑀𝑅 kg/s
heavy-duty vehicle in kg/s
Percent of rebar mass loss (or area loss per unit
𝑚 -
length) due to corrosion
m - Number of ground motion intensity levels
𝑚 - Regression coefficient for considering the size effect
𝑁 - Number of traffic lanes
𝑛 - Volume expansive coefficient
Total ground motion records at jth ground motion
𝑛 -
intensity levels
𝑃(𝑡) mm Pit depth
𝑃 MPa Internal pressure caused by rust expansion
𝑄 - Percent weight loss (or area loss) of rebar
R - Pitting factor
𝑟 mm Radius of the thick-wall cylinder (concrete cover)
𝑟 mm Radius of rebar
Distance from the center of the thick-wall model to
𝑟 mm
the interface between the rust and concrete
Density of water in the air generated by bow wave per
𝑆𝐷 kg/m3/vehicle
heavy-duty vehicle in kg/m3/vehicle
Density of water in the air generated by capillary
𝑆𝐷 kg/m3/vehicle
adhesion per heavy-duty vehicle in kg/m3/vehicle
Density of water in the air generated by side wave per
𝑆𝐷 kg/m3/vehicle
heavy-duty vehicle in kg/m3/vehicle
Density of water in the air generated by tread pickup
𝑆𝐷 kg/m3/vehicle
per heavy-duty vehicle in kg/m3/vehicle

xviii
 Total density of water in the air generated by heavy-
𝑆𝐷 kg/m3/vehicle
duty vehicle spray and splash in kg/m 3/vehicle
Total density of deicing salt generated by heavy-duty
𝑆𝐷 _ kg/m3/vehicle
vehicle spray and splash in kg/m3/vehicle
Total density of chloride ions generated by heavy-
𝑆𝐷 _ kg/m3/vehicle
duty vehicle spray and splash in kg/m 3/vehicle
T0 years Corrosion initiation time
T1 years Time when the peak corrosion rate is reached
tcr_lon years Longitudinal rebar corrosion initiation time
𝑡 years Corrosion initiation time
𝑡 days/winter Number of days with snow per winter season

𝑉 mile/hour Heavy-duty vehicle speed in mile/hour

𝑉 km/hour Heavy-duty vehicle speed

kg/lane per unit
𝑉 Daily application rate of deicing salts
length/day
𝑊(𝑡) mm Crack width of concrete cover at time t
Average crack width derived from deterioration
𝑊 (𝑡) mm
curves at time t
Lower bound of crack width derived from
𝑊 (𝑡) mm
deterioration curves at time t
Upper bound of crack width derived from
𝑊 (𝑡) mm
deterioration curves at time t
𝑊(𝑡) mm Crack width of ECC cover at time t
𝑊 mm Critical crack width
𝑊, , , - Fitting coefficients
𝑊 m Road width where deicing salts were applied
Number of records which cause a particular damage
𝑧 -
state in the jth IM level
𝑖 Current density
𝛽 - Parameter of standard normal distribution

xix
 𝛽̅ - Estimated parameter of standard normal distribution
𝛿 mm Thickness of the porous zone
𝛿 mm Radial displacement caused by rust expansion
Salt mass applied per unit area of road / Water mass
𝛿 -
per unit area of road
The maximum compressive strain in the confined
𝜀 -
concrete
𝜀 - Strain in transverse rebar at ultimate strength
𝜀 - Tensile strain of ECC in percent
𝜀∗ - Tensile strain of ECC
𝜃 - Parameter of standard normal distribution
𝜃̅ - Estimated parameter of standard normal distribution
Angle parameter 1 depending on D0, 𝑃(𝑡), and b in
𝜃 -
pitting corrosion model
Angle parameter 2 depending on D0, 𝑃(𝑡), and b in
𝜃 -
pitting corrosion model
Amount of chloride ions as a proportion of the
𝜃 -
amount of sodium chloride
The angle between one of the outside cracks and the
𝜃 -
exterior surface of concrete cover
𝜆(𝑡) mm/year Corrosion rate
𝜆(𝑡) mm/year Corrosion rate at critical crack width
𝜆(𝑡) mm/year Corrosion rate before cracking
𝜌 - Longitudinal reinforcement ratio
𝜌 kg/m3 Mass density of rust
𝜌 kg/m3 Mass density of the original (non-corroded) steel
𝜌 (𝑡) - Volumetric ratio of transverse rebar at time t
𝜌 kg/m3 Density of water
∆𝐴(𝑡) mm2 Cross-sectional area loss of reinforcement at time t

xx
 Standard normal cumulative distribution function
Φ(. ) -
(CDF)
𝜈 - Poisson’s ratio of concrete
𝜈 - Poisson’s ratio of rust
𝜎 MPa Circumferential stress caused by rust expansion
Ratio of chloride ions sprayed and splashed by heavy-
∅ -
duty vehicles to light-duty vehicles.
∅ mm Reference rebar size
Fraction of heavy-duty vehicles as a proportion of
Θ -
total vehicles
Factor to account for the concrete brittleness in
𝜂 -
tension

xxi
 Chapter 1: Introduction

1.1 Background and motivation

Traditional infrastructure, such as roads, bridges, dams, and ports, is necessary for

transporting people and goods between and provides access to vital resources (e.g., water, minerals,

etc.). High-quality infrastructure is key for national and global economic growth (Heintz et al.,

2009). Highway bridges are a major part of the infrastructure, and they are in urgent need of repair

and maintenance in the US. According to the American Society of Civil Engineers (ASCE) 2021

infrastructure report card (ASCE, 2021), the current condition of bridges in the US was graded

“C”. Due to deterioration, it was estimated that the condition of bridges would be worse with time.

For example, the annual cost of bridge rehabilitation was estimated to increase from $14.4

billion/year to $22.7 billion/year from 2014 to 2034 (FHWA, 2017). Over 50% of bridges in US

are built with reinforced concrete (RC) (NACE, 2013). Therefore, the deterioration of RC bridges

significantly affects the health of the overall infrastructure in the US. Two major threats to the

health of RC bridges are corrosion and natural hazards.

Reinforced concrete bridges located in coastal areas and cold regions are subject to

corrosion. Bridges constructed over the sea are directly exposed to chloride ions due to waves and

tides (Bertolini et al., 2004). At the same time, the wind carries chloride ions from the sea to the

surfaces of piers and decks of RC bridges that are located several miles inland (Slamova et al.,

2012; Weyers et al., 1994). In cold regions, deicing salts are applied on roadways to prevent snow

and ice accumulations and improve driving conditions during the winter season. Although

chloride-based salts (e.g., sodium chloride, calcium chloride, and magnesium chloride) are

1
effective in deicing road surfaces, they also release a substantial amount of chloride ions (Cl -) to

the surrounding environment that can accumulate on the surfaces of RC bridge structures (Lindvall,

2001, 2003; Tang & Utgenannt, 2007; Xi et al., 2018). The released chloride ions are a major

source of reinforcement corrosion in RC bridges.

The RC bridges located in the seismic zone are susceptible to damage due to earthquakes.

Earthquakes can cause concrete cracking and spalling, and rebar yielding and fracture. Some of

these impacts could cause a sudden loss of functionality of a bridge, resulting in significant

downtime requiring repair/reconstruction and even loss of life (Basöz et al., 1999; Bommer et al.,

2002; Grossi, 2005). In the US, many aging RC bridges are located both in a corrosive environment

and in a seismic region, such as bridges located along the west coast (e.g., Washington State).

A systematic framework is needed to accurately evaluate the performance of RC bridges

subjected to the combined effects of corrosion and earthquake shaking. Such an evaluation would

allow engineers and policy makers to effectively assess risk and manage infrastructure assets with

limited resources. It also enables innovation in materials and structural design to improve the long-

term resilience and durability of RC bridges.

1.2 Scope of the research

There are many deterioration mechanisms that can reduce the capacity of an RC bridge,

including corrosion, alkali-aggregate reaction, and erosion of soil surrounding a bridge pier. This

study considers only corrosion as a deterioration mechanism. Moreover, corrosion is caused by

2
seawater for bridges in coastal regions and caused by deicing salts for bridges in cold regions. This

study focused on corrosion of RC bridges in cold regions caused by deicing salts.

Although considering the durability of all RC bridge components is important, this research

focused on the evaluation of RC bridge piers under corrosion and earthquake hazards. This is

because the bridge piers are critical structural components for resisting extreme loads, such as

earthquake, flooding, and vehicle impact, and their failure can cause the failure of the whole bridge.

For instance, according to Mitchell et al. (1995), most collapses of highway bridges subjected to

the Northridge 1994 earthquake were attributed to the bridge column failures. Bridge

superstructures such as girders and decks are important in maintaining the safety and serviceability

of bridges, however these structural components normally perform elastically under seismic events

(Ramanathan, 2012); therefore, superstructures were not considered in this research.

This study focused on RC bridge columns with flexural failure due to the combined effects

of corrosion and seismicity. Only slender bridge columns (cross-section diameter/ height ≤ 0.15)

with expected flexure failures are considered. Investigating shear failure was out of the scope of

this dissertation.

For seismic analyses, this study only considered the horizontal components of ground

motions, did not consider aftershocks or soil-structure interaction in order to reduce the

computational cost of the analyses. Finally, the scope of the work was limited to short and medium

span bridges with more than one span. Hazards other than seismicity (e.g., wind) have not been

considered.

3
1.3 Research goals and objectives

This doctoral research has three major objectives: (1) Develop a framework for seismic

damage risk analysis of corroding RC bridge piers, and utilize the framework to investigate the

reduction in seismic damage risk of RC bridge piers due to improved durability of the cover

material; (2) Incorporate local environmental conditions and vehicle spray and splash mechanisms

into the corrosion model for improved bridge pier durability assessment; and (3) Link the physics-

based corrosion model to the field-based condition ratings, and use the linkage for calibrating the

input parameters of the corrosion model. Specific research tasks to achieve the aforementioned

goals are as follows:

1. To identify the challenges in combining corrosion modeling and seismic fragility

analysis by reviewing existing frameworks.

2. To improve existing corrosion modeling methodologies by considering the effects of

concrete cracking on corrosion rate and develop a corrosion model for a ductile fiber

reinforced concrete, called herein Engineered Cementitious Composite (ECC).

3. To develop a systematic framework for performing the seismic fragility analysis of a

deteriorating RC bridge column, and demonstrate the framework on an example bridge

column.

4. To investigate the effects of replacing conventional concrete with ECC in the cover of

RC bridge columns on their durability and seismic performance.

4
 5. To demonstrate the framework on a group of existing RC highway bridges for

prioritization of intervening actions based on the evolution of damage risk with time.

6. To capture the impact of local environmental exposure conditions and vehicle spray

and splash mechanisms on surface chloride exposure of RC bridge piers.

7. To combine the physics-based corrosion model developed in this research with field-

based condition assessment data for calibrating corrosion model parameters.

8. To investigate the effects of various input parameters of the framework on corrosion

model results and fragility functions that incorporate corrosion effects.

1.4 Dissertation organization

This dissertation is organized into seven chapters. Chapter 2 presents a review of existing

studies on corrosion modeling, seismic fragility analysis, and methods for combining corrosion

modeling and seismic fragility analysis. In Chapter 3, a novel corrosion model that considers the

effects of concrete cracking on corrosion rate is presented. Additionally, a new method for

incorporating the unique cracking mechanism in ECC and similar fiber-reinforced concretes into

the corrosion model is developed. The method used for performing seismic fragility analysis of

RC bridges is also detailed in this chapter. A sensitivity study investigating the effects of different

parameters of the framework on fragility functions is presented at the end of Chapter 3 to determine

the most influential model parameters. Chapter 4 demonstrates the application of the framework

5
developed in Chapter 3 for analyzing the long-term performance of a group of aging bridges. The

advantages and disadvantages of the developed framework are discussed in this chapter. Chapter

5 introduces a new model for estimating the surface chloride exposure of RC bridges considering

vehicle spray and splash mechanisms and the local environmental exposure of highway bridges.

A parametric study to determine the most influential parameters on surface chloride content of RC

piers is also presented. Chapter 6 presents a method to combine the physics-based corrosion model

(developed in Chapter 3) with field-based bridge condition rating data to calibrate corrosion model

input parameters and improve corrosion damage estimation for RC bridges. Finally, Chapter 7

summarizes the developed scientific impacts and original contributions of this research and

identifies future research opportunities for building upon this dissertation. A list of references

follows the body of this dissertation.

6
 Chapter 2: Literature Review

2.1 Review of corrosion modeling

In general, reinforcement corrosion in RC structural members occurs in two phases:

corrosion initiation and corrosion propagation. In the initiation phase, chloride ions from the

outside environment gradually ingress through the concrete cover to the steel reinforcement

surface. Once the chloride concentration at the reinforcement surface reaches a critical level, the

rebar corrosion initiates, which marks the beginning of the corrosion propagation phase. The cross-

sectional areas of the transverse and longitudinal rebars decrease during corrosion propagation.

The corrosion products occupy 3-6 times more volume than the uncorroded reinforcement, which

generates tensile stress in the surrounding concrete. The concrete cover cracks when the tensile

stress reaches the tensile strength of concrete. Figure 2.1 below illustrates the process.

Figure 2.1: Reinforcement corrosion (adapted from Tuutti (1982))

7
2.1.1 Corrosion initiation phase

The concrete cover usually protects the reinforcement against corrosion in RC structural

members. The pH of the concrete pore solution is usually greater than 12.5 due to the presence of

Na+ and K+ ions (in cement) and calcium hydroxide (Ca(OH)2). The Ca(OH)2 is a by-product of

cement hydration. Such high pH facilitates the formation of a thin protective layer of iron-oxide

around the steel reinforcement, which protects it from further corrosion. However, the protective

layer can be destroyed due to chloride ion accumulation at the steel-concrete interface. The loss of

the protective layer makes the steel reinforcement vulnerable to corrosion.

There are three transport mechanisms of chloride ions through the concrete cover:

permeation, absorption, and diffusion (Basheer et al., 2001). Diffusion is the most commonly

observed transport mechanism, and it is used in the majority of computational models. There are

both semi-empirical models (L.-O. Nilsson et al., 1997; Life-365™, 2020; Luping, 2008; Weyers

et al., 1994) and numerical models (Baroghel-Bouny et al., 2011; Johannesson et al., 2007; Samson

et al., 1999) presented in the literature. Semi-empirical models for corrosion initiation were

developed based on the statistical analysis of experimental data from field or laboratory tests,

combined with Fick’s second law of diffusion. Numerical models were developed based on

detailed coupled transport mechanism models and chemical equilibrium models. In this research,

the semi-empirical modeling approach was selected for simulating the corrosion initiation phase

due to the computational efficiency.

8
2.1.2 Corrosion propagation phase

As mentioned before, rebar area reduction, concrete cracking, and concrete spalling - all

happen in the corrosion propagation phase. There are two types of reactions that occur in this phase:

Anodic reaction and Cathodic reaction. Anodic reaction involves oxidation of iron to ferrous ions

(Fe2+) and releases electrons (e-). The Cathodic reaction absorbs the released electrons and reduces

oxygen, in presence of water, to hydroxyl ions (OH-). Finally, rust (Fe(OH)2) is produced by the

combination of Fe2+ and OH- ions. Figure 2.2 below illustrates both the reactions.

Figure 2.2: Chemical reactions in the corrosion propagation phase (adapted from Fakhri (2019))

Similar to the corrosion initiation, both empirical and numerical models have been

proposed in the literature for simulating corrosion propagation. Empirical models were developed

by performing statistical analysis of the data from field tests and laboratory experiments, whereas

the numerical models considered fundamental chemical reactions and modeled them using finite

element methods. A thorough summary of the empirical and numerical models for the corrosion

propagation phase can be found in Otieno et al. (2011).

9
2.2 Review of seismic fragility analysis

Seismic fragility curves are conditional statements of probability that a structure will meet

or exceed a specified level of damage for a given ground motion intensity measure. These curves

have found widespread use in probabilistic seismic risk assessment of highway bridges. The

ground motion intensity measures are typically single intensity measures such as peak ground

motion acceleration (PGA) or spectral acceleration at the geometric mean of the longitudinal and

transverse periods. Fragility curves are commonly generated from (1) empirical data based on

actual earthquake damage data (Basöz et al., 1999; Yamazaki et al., 2000), (2) expert opinion or

judgment (ATC, 1985) and (3) numerical simulation (Nielson & DesRoches, 2007; Ramanathan,

2012). Because actual earthquake damage data is limited and expert opinions can be subjective,

numerical simulations are commonly used for developing fragility curves.

Two common approaches for developing numerical seismic fragility curves are: (1)

Nonlinear response history analysis (NRHA), also called ‘Cloud analysis’, and (2) Incremented

dynamic analysis (IDA) (Muntasir & Shahria, 2015). The NRHA approach consists of four steps:

(1) selection of N ground motions to represent the earthquake scenarios at the structural location,

(2) development of Z numerical models of the structure utilizing statistical sampling methods (e.g.,

Monte Carlo simulations) and finite element methods, (3) pairing of the sampled models from step

2 with the selected ground motions to perform dynamic analysis, and (4) estimation of the

statistical parameters defining the fragility curves utilizing statistical modeling (e.g., linear

regression). In the IDA approach, the fragility curves are developed by scaling the ground motion

incrementally based on ground motion intensity measures (e.g., peak ground acceleration) and

10
performing dynamic analyses at each scaled intensity measure. More details about IDA are

discussed in Section 3.3.

The major difference between the NRHA and IDA approaches is the scaling of the ground

motions in IDA, which increases the computational cost of IDA compared to NRHA. However,

the incremental scaling of the ground motions in IDA allows tracking the performance of a

structure from elastic behavior through yielding to dynamic instability (structure failure), which

provides a more comprehensive evaluation of the structure under seismic loads. Therefore, IDA

was used to develop seismic fragility curves in this research on the corrosion effects on structure

performance under earthquake. More details about the fragility analysis can be found in Muntasir

& Shahria (2015).

2.3 Review of fragility analysis considering corrosion effects

There are several studies that have incorporated corrosion effects into seismic fragility

functions for RC bridges. The existing seismic risk analyses of corroding bridges can be broadly

classified into two categories, depending on the corrosion model: studies that considered unform

corrosion and studies that considered pitting corrosion. The majority of studies have utilized a

uniform corrosion model (Akiyama et al., 2011; Alipour et al., 2010; Biondini et al., 2014; Choe

et al., 2008, 2009; Ghosh & Padgett, 2010). Only a limited number of studies have used a pitting

corrosion model due to greater complexity of this model (Afsar Dizaj et al., 2018; Cui et al., 2018;

Ghosh & Sood, 2016; Yijian Zhang et al., 2019). In real RC bridges under chloride-rich

environments, pitting corrosion is the most commonly observed corrosion mechanism (Darmawan,

2010; Stewart & Al-Harthy, 2008).

11
 In addition, various failure types such as shear failure of RC bridge columns were

investigated (Choe et al., 2008, 2009; Yijian Zhang et al., 2019). Changes of thresholds for

defining the damage states in fragility analysis due to corrosion were considered by some studies

(Afsar Dizaj et al., 2018; Biondini et al., 2014; Ghosh & Sood, 2016). All these studies indicate

that corrosion significantly increases the seismic damage probability of RC bridges.

12
 Chapter 3: Combining Corrosion and Seismic Fragility Analyses to

Investigate the Impact of Rebar Cover Material

3.1 Introduction

Corrosion of steel reinforcement in reinforced concrete (RC) highway bridges is one of the

most common durability problems. In the United States, in 2013, the annual direct cost of repair

and maintenance of bridges that are deteriorating due to corrosion is estimated to be $13.6 billion

(NACE, 2013). In addition to its impact on the economy, rebar corrosion can also affect the safety

of a structure when a repair is overdue due to budget constraints. An extreme event such as an

earthquake can inflict more damage on a corroded bridge as compared to a sound bridge due to

the reduction of the buffer capacity. Therefore, a systematic framework to estimate the seismic

risk of corroded RC bridges is needed. As mentioned in Chapter 2, several studies have

investigated the seismic risk analyses of aging bridges. However, significant knowledge gaps

remain in the existing framework as discussed below.

First, the effects of concrete cracks on the corrosion rate have not been considered by the

aforementioned studies. Corrosion-induced cracks in concrete cover increase the corrosion rate

significantly (Cao et al., 2013; Yuan et al., 2010), and therefore ignoring their effects may cause

underestimation of corrosion effects on seismic fragility analysis of bridges (Cui et al., 2018).

Second, the effects of ductile fiber-reinforced concrete (FRC) covers on the long-term

seismic performance of RC bridge columns have not been investigated in the aforementioned

studies. Applying a cement-based jacket (cementitious mortar jacketing and UHPC) to a damaged

13
structure is a commonly-used rehabilitation method (Chalioris et al., 2019; Meda et al., 2016; Yang

Zhang et al., 2019; Zhu et al., 2020). However, only a few studies have investigated the benefits

of using FRC covers to improve long-term durability, and no studies investigated the combined

effects of corrosion and seismicity on columns with FRC covers.

Lastly, the unique multiple-cracking mechanism of ductile FRC (as opposed to localized

cracking in RC) and its influence on the corrosion rate, especially in the corrosion propagation

stage, have not been incorporated in the existing corrosion models. This chapter addresses the

above limitations of the existing studies.

The ductile FRC considered in this Chapter is called Engineered Cementitious Composite

(ECC), which has been shown to provide significantly better protection from corrosion than

conventional concrete (Li, 2019). The tensile strain capacity of an ECC is at least two orders of

magnitude greater than that of conventional concrete (Li, 2003). Instead of forming large cracks

similar to concrete, an ECC absorbs damage through the formation of micro-cracks of widths (or

openings) less than 100 μm even at large deformations well beyond its elastic limit. The

compressive strength of ECC is similar to that of a moderate strength concrete (40-50 MPa). ECC

has been used in structures to improve their shear capacity, energy dissipation, and ductility

(Fischer & Li, 2002; Fukuyama et al., 2000; Kanda et al., 1998). At the same time, previous

research on the transport properties of ECC suggests that this material, even when strained in

tension up to 3%, exhibits lower water permeability and effective chloride ion diffusivity

comparable to uncracked concrete, by virtue of its intrinsically tight crack width (Fakhri, 2019;

Lepech & Li, 2009; Mihashi et al., 2011; Sahmaran et al., 2007). The difference in performance

14
between ECC and concrete is more significant during the corrosion propagation stage, as ECC can

resist the tensile hoop stresses created around the rebars by the expansive corrosion products.

Recent research (Fakhri, 2019) has shown the effectiveness of using precast ECC covers for

reducing corrosion rates in chloride-rich environments.

The objective of this Chapter is to demonstrate a quantitative framework to determine the

influence of improved structural durability, enabled by the use of ECC cover, on the seismic

performance of an RC bridge pier. The computational framework consists of two parts: corrosion

modeling and seismic fragility analysis. In the subsequent sections of this Chapter, corrosion

initiation and propagation models are defined for concrete and ECC. The corrosion model is

different than existing models because it considers crack width effects and incorporates the

influence of the unique cracking behaviour of ECC on corrosion rate. Then, seismic fragility

functions, incorporating corrosion effects, are introduced. The framework is demonstrated using

an example bridge column that uses two alternative cover types: conventional concrete and ECC.

Long-term durability benefits of using ECC cover were investigated. Finally, a parametric study

is presented to identify the influence of critical parameters on the durability and fragility

assessments.

3.2 Corrosion model

Chloride-induced corrosion is considered in this study, which is the major cause of bridge

degradation in the U.S., particularly in the northern states. It happens in two phases: initiation and

propagation. In the initiation phase, chloride ions from the outside environment gradually ingress

15
through the concrete cover to the steel reinforcement surface. Once the chloride concentration at

the reinforcement surface reaches a critical level, the rebar corrosion initiates, which marks the

beginning of the corrosion propagation phase. The cross-sectional areas of the transverse and

longitudinal rebars decrease in the propagation phase due to corrosion. As the rust expands, it

generates tensile stress in the surrounding concrete. The concrete cover cracks when the tensile

stress reaches the tensile strength of concrete.

3.2.1 Corrosion initiation

The time to corrosion initiation is typically determined using the Fick’s second law of

diffusion. The time to corrosion initiation can be calculated using the Crank’s solution (Crank,

1979):

𝐶 (𝑥, 𝑡) = 𝐶 1 − erf (1)

A 2D corrosion diffusion model can provide a better estimate for corrosion initiation time.

However, for simplicity, a 1D model is used in this study (Equation 1).

3.2.2 Corrosion propagation

Active corrosion of steel reinforcement happens during the corrosion propagation phase.

Corrosion can either occur uniformly throughout rebar due to the formation of a corrosion macro-

cell or occur locally due to the formation of a corrosion micro-cell or a pit (Bertolini et al., 2004).

Chloride ion diffusion in RC bridges typically leads to local or pitting corrosion, which is

considered in our study.

16
 The corrosion propagation phase can be further divided into four stages. In the first stage,

the corrosion products fill the porous zone at the rebar-concrete interface without causing any

stress in the surrounding concrete. Once the pores are saturated, in the second stage, the corrosion

products start to apply outward radial pressure by expanding as their volume is 3-4 times that of

the original steel, causing tensile hoop stress in the surrounding concrete. This continues until the

tensile stress in concrete reaches its tensile strength, at which point the cover cracks, marking the

end of the second stage. In the third stage, the corrosion rate increases due to an increase in the

number of cracks as well as an increase in the widths (openings) of the existing cracks. In the

fourth stage, after reaching a certain crack width, the corrosion rate stabilizes as the built-up

corrosion products fill up the cracks and hinder the transportation of oxygen (Yuan et al., 2010).

All the above stages of corrosion are captured in our model.

3.2.2.1 Pitting corrosion

During pitting corrosion, a rebar’s cross-sectional area will continuously decrease over

time. As discussed above, pitting corrosion is typically observed in real RC structures during the

corrosion propagation stage. As the name suggests, pitting corrosion creates a pit starting at the

surface of the rebar. Unlike uniform corrosion that reduces rebar area throughout the length of the

rebar, in pitting corrosion, the rebar mass loss is local as the depth of the pit becomes larger with

time. Val & Melchers (1997) proposed a hemispherical model to simulate pitting corrosion, as

shown in Figure 3.1.

17
 Figure 3.1: Pitting corrosion (adapted from Val & Melchers (1997))

The rebar area, A(t), at time t after corrosion initiation and can be estimated using the

following equations (Val & Melchers, 1997):

√
⎧ −𝐴 −𝐴 , 𝑃(𝑡) ≤ 𝐷
⎪
𝐴(𝑡) = 𝐴 −𝐴 , 𝐷
√
< 𝑃(𝑡) ≤ 𝐷 (2)
⎨
⎪ 0, 𝑃(𝑡) > 𝐷
⎩

The parameters 𝐴 and 𝐴 can be estimated by the following equations (Val & Melchers, 1997):

( )
𝐴 = 0.5 𝜃 −𝑏 − (3)

( )
𝐴 = 0.5 𝜃 𝑃(𝑡) − 𝑏 (4)

𝜃 = 2arcsin( ), 𝜃 = 2arcsin( ) (5)

( )

( )
𝑏 = 2𝑃(𝑡) 1 − (6)

The pitting depth, 𝑃(𝑡), is be expressed by equation (Val & Melchers, 1997):

𝑃(𝑡) = R ∫ 𝜆(𝑡) 𝑑𝑡 (7)

18
where, 𝜆(𝑡) is the corrosion rate, expressed as:

𝜆(𝑡) = 0.0116 × 𝑖 (8)

In addition to the reduction in rebar area, pitting corrosion also causes a reduction in the

effective yield strength as expressed by Equation (9) below. This equation is from a study (Du et

al., 2005) which tested 87 rebar specimens that were representative in sizes of the ones used in

bridge columns. It should be noted that the nominal yield strength of the corroded rebar likely does

not change due to corrosion. However, pitting causes local stress concentration, and the residual

force capacity of the corroded rebar decreases more rapidly than the average cross-sectional area

of the corroded rebar (Du et al., 2005). Therefore, the effective yield strength must be used instead

of the nominal yield strength to account for the capacity loss of rebars due to corrosion.

𝑓 (𝑡) = (1.0 − 0.005𝑄 )𝑓 (9)

where, 𝑄 is the percent weight loss (or area loss) of rebar expressed as:

( )
𝑄 = ∗ 100 (10)

3.2.2.2 Crack initiation time

The build-up of rust causes tensile stress to increase with time in the concrete surrounding

the rebar. When this stress reaches the tensile strength of concrete, a crack is formed in the concrete

cover. Assuming that all the corrosion products remain inside, the rebar mass loss computed below

can be converted into internal radial pressure caused by rust expansion using Equations (11), (12),

and (13) (El Maaddawy & Soudki, 2007). The concrete around the rebar was modeled as a thick-

19
walled cylinder, in which the circumferential stress was calculated using Equation (14). For

simplifying the model, the entire cover was assumed to crack instantaneously when the tensile

stress in the circumferential direction at the interface between the rust and concrete reached the

tensile strength of concrete. As the rebar mass loss (𝑚 ) is a function of time, crack initiation

time was calculated using Equation (11) to Equation (14).

𝑚 =𝑄 × 𝑢𝑛𝑖𝑡 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 𝑜𝑓 𝑠𝑒𝑐𝑡𝑖𝑜𝑛 (1 𝑚𝑚) (11)

𝑃 = − (12)
. ( )( ) ( )( )

𝜓= (13)
( )

𝜎= (1 + ) (14)

3.2.2.3 Time-dependent crack width

After crack initiation, crack width gradually increases with time. Due to different crack

control abilities of concrete and ECC, two different time-dependent crack width models were

considered. For concrete, the equation below (Vidal et al., 2004) was used:

𝑊(𝑡) = 0.0575 (∆𝐴(𝑡) − ∆𝐴(𝑇 )) (15)

Similar to Equation (15), an empirical equation describing the time-dependent crack width

of ECC as a function of rebar area (or mass) loss was derived in our study. For this purpose, three

relationships were needed: (1) the relation between rebar area loss and outward radial displacement

caused by expansive corrosion products, (2) the relation between the outward radial displacement

and average hoop strain in the cover, and (3) the relation between average hoop strain and average

20
crack width of ECC. As mentioned above, unlike concrete, the crack width in ECC is controlled

by the bridging fibers. These three relationships were determined as follows.

Considering the densities of the original steel and corrosion products and the rebar

diameter, a model (Equation (16)) proposed by El Maaddawy & Soudki (2007) was used to

calculate the relationship between outward radial displacement (𝛿 ) and rebar mass loss. In our

study, the 𝜌 /𝜌 ratio was assumed to be 0.5 (Liu & Weyers, 1998; Molina et al., 1993) and the

𝑀 /𝑀 ratio was assumed to be 0.622 (El Maaddawy & Soudki, 2007).

− = 𝜋𝐷 (𝛿 + 𝛿 ) (16)

Fakhri (2019) performed experiments to investigate the relationship between outward

radial displacement and the average hoop strain of ECC by performing a linear regression on the

experimental data. The resulting expression is given:

𝜀 ∗ = 0.009𝛿 + 3.6 × 10 (17)

In another experimental study, Ranade et al. (2014) determined the relationship between

the tensile strain and the average crack width of ECC as given by Equation (18). In our study, the

mean crack width given in Ranade et al. (2014) was used. Using Equations (16), (17), and (18)

simultaneously, an empirical relation between rebar mass loss due to corrosion and average crack

width in ECC was developed.

𝑊(𝑡) = 36.1𝜀 − 101.0𝜀 + 105.8𝜀 + 23 (18)

where 𝜀 = 𝜀 ∗ ∗ 100

21
3.2.2.4 Corrosion rate

Previous studies (Cao et al., 2013; Fakhri, 2019; Yuan et al., 2010) have shown that the

existence of cracks in both ECC and concrete increases the corrosion rate significantly, depending

on the crack width. This is because the cover cracks provide an easier ingress path for water,

oxygen, and chloride ions. As explained above, in stage 3 of corrosion propagation, after cracking

on the concrete cover, the corrosion rate increases with an increase in the number and widths of

the cover cracks. Due to a lack of experimental data and ease in numerical analysis, the corrosion

rate was assumed to increase linearly with time in this study. In stage 4, the corrosion rate reaches

a peak value and remains constant thereafter, as the cover crack width reaches a critical value

depending on the material. The corrosion rate change within the corrosion propagation phase is

shown in Figure 3.2. The slope (dλ⁄dt) of corrosion rate with time was iteratively determined

using the steps as shown in Figure 3.3.

22
 Figure 3.2: Corrosion rate change with time

Figure 3.3: Flowchart of determining the slope of corrosion rate with time

23
3.2.3 Corrosion model properties for the example bridge column

A three-span highway bridge with a concrete deck and concrete tee girders supported by a

single column bridge pier was chosen to demonstrate the framework, additional details of the

bridge are given in Section 3.4. The example bridge column is located in an area with exposure to

de-icing salts and hence to a high corrosion environment.

3.2.3.1 Conventional concrete cover

Corrosion initiation phase: In this study, the example bridge was assumed to be a new bridge, and

therefore, the initial chloride concentration in the cover concrete was assumed zero. The surface

chloride concentration (Cs) was assumed to be 3.6 kg/m3, which represents the exposure

environment created by the de-icing salts, based on Weyers et al.(1994). The chloride diffusion

coefficient (𝐷 ) was assumed to be 7.4  10-12 m2/s for conventional concrete (Weyers et al.,

1994). The critical chloride content for corrosion initiation at the rebar-concrete interface was

assumed to be 0.72 kg/m3 (Life-365™, 2020).

Corrosion rate: A high corrosion rate of 0.0116 mm/Year suggested in the literature (Andrade &

Alonso, 2001) was assumed prior to crack initiation. Due to the lack of peak corrosion rate data in

real bridges, the peak corrosion rate after crack initiation was assumed two times (equal to 0.0232)

that of the corrosion rate before crack initiation. It is also assumed that no repair was performed

during the bridge’s service life. The critical crack width beyond which the corrosion rate becomes

constant was assumed to be 0.2 mm for concrete (Cui et al., 2018).

24
3.2.3.2 ECC cover

Corrosion initiation phase: The environmental exposure, therefore surface chloride concentration

(Cs) and initial chloride concentration, for the ECC cover, was the same as that for the concrete

cover. The effective chloride diffusion coefficient assumed for ECC was 0.64 times that of

conventional concrete due to different microstructures of concrete and ECC (Sahmaran et al.,

2007). Although the critical chloride content changes with the type of concrete, its value for ECC

was conservatively assumed to be the same as that for concrete due to limited experimental data.

Corrosion rate: Average corrosion rate of ECC assumed in this study was 0.72 times that of

conventional concrete based on the experimental data (Fakhri, 2019). Thus, the corrosion rates

before and after cracking of ECC were 0.0084 mm/Year and 0.0167 mm/Year, respectively. The

average crack widths in ECC are typically limited to 0.1 mm by the fibers. Some experimental

data suggests that the corrosion rate in ECC can become constant at crack widths less than 0.1 mm

(Fakhri, 2019). However, due to limited experimental data, the corrosion rate was assumed to

become constant after the crack width reached 0.1 mm. Table 3.1 summarizes all the parameters

used in the corrosion model of the example column. The justification for selecting these parameters

is provided in Section 3.2.3.1. All assumed values can be changed in the future based on the field

data for a specific column. The computational framework demonstrated in this Chapter will be

applicable regardless of the model parameters.

25
 Table 3.1: Parameters considered in the corrosion model of the example bridge column

Conventional
Parameters ECC Source
concrete
𝐶 (kg/m3) 3.6 3.6 Weyers et al. (1994)
Sahmaran et al.
𝐷 (m2/s) 7.4×10-12 4.7×10-12 (2007); Weyers et al.
(1994)
Critical chloride content at
0.72 0.72 Life-365™ (2020)
rebar surface (kg/m3)
C (mm) 50.8 50.8 -
Elastic modulus (GPa) 33 14 Wang & Li (2007)
Andrade & Alonso
𝜆(𝑡) (mm/Year) 0.0116 0.0084 (2001); Fakhri et al.
(2019)
Explained in Section
𝜆(𝑡) (mm/Year) 0.0232 0.0168
3.2.3.1
Critical crack width, Cui et al. (2018);
0.2 0.1
𝑊 (mm) Fakhri et al. (2019)
El Maaddawy &
𝜈 0.2 0.2 Soudki (2007),
AASHTO (2020)
El Maaddawy &
𝛿 (mm) 20×10-3 20×10-3
Soudki (2007)
Val & Melchers
R 6 6
(1997)

Corrosion-induced deterioration in bond strength between rebar and concrete was not

considered in this study because the influences of corrosion on bond strength are limited when the

structure has substantial confinement (Fang et al., 2004; Hanjari et al., 2011; Tondolo, 2015),

which is also the case in the example column investigated in this study. In addition, as the most

critical section of the column for both pitting corrosion and seismicity is the column base (Lindvall,

2003; Yuan et al., 2017), our model is governed by the pitting corrosion at the base of the column.

26
3.3 Seismic fragility of a corroded bridge column

A seismic fragility functions provide the probability of a structure exceeding a specified

level of damage as a function of a ground motion intensity measure (IM) such as PGA or spectral

acceleration at the geometric mean of the longitudinal and transverse periods. A seismic fragility

curve is commonly modeled by a lognormal cumulative distribution function, as shown in

Equation (19). In this study, the Engineering Demand Parameter (EDP) (e.g., beam plastic rotation,

derived damage index, etc.) and the distribution parameters (𝜃 and 𝛽) were estimated from an IDA

and statistical analysis of its results.

( )
𝑃(𝐸𝐷𝑃 ≥ 𝑑 |𝐼𝑀 = 𝐼𝑀 ) = Φ (19)

The fragility functions in this study were created following these steps: (1) Selection of

ground motions per FEMA P695 (ATC, 2009), (2) Development of a structural (finite element)

model, (3) Performing IDA (Vamvatsikos & Cornell, 2002) for the selected ground motions, (4)

Defining and selecting a suitable damage index and damage states, and (5) Estimation of

parameters of the fragility curves using statistics. It should be noted that there are multiple methods

to create fragility curves (Nielson & DesRoches, 2007), and some steps may vary between various

methods.

3.3.1 Ground motion selection (step 1)

A far-field ground motion record set, including 44 records recommended by FEMA P695

(ATC, 2009), was used. The PGA was used as the earthquake intensity measure. It should be noted

that PGA is not always suited to capture the effects of earthquake intensity, particularly in the case

27
of complex structures (e.g., a high-rise building), because larger PGA can sometimes cause less

severe damage to these structures. However, in this chapter, a bridge column was modeled as a

single degree freedom system, and the simulations showed that the seismic demand of this system

increased with an increase in the PGA. This finding was consistent with the findings of others in

the literature (Padgett et al., 2008; Ramanathan, 2012). Besides PGA, peak ground velocity (Karim

& Yamazaki, 2001) and spectral acceleration (Ramanathan, 2012) have been used as intensity

measures for seismic fragility analysis of bridges. All the ground motion records used in this study

were scaled linearly from 0.1g to 1.5g with increments of 0.1g and applied in the longitudinal

direction (along the traffic) to the bridge column. The column base was fixed and the column top

was free to translate in the longitudinal direction and rotate for this single degree freedom system.

Additional details of the model are provided in Section 3.4.1.

3.3.2 Structural model (step 2)

Finite element models of bridge substructure elements (i.e., columns) were created in an

open-source finite element software, OpenSEES (McKenna et al., 2000). The bridge columns were

simulated using a single degree of freedom, fiber-based, nonlinear, forced-based beam-column

elements. The superstructure and other bridge elements were represented as masses but were

otherwise excluded from the model for computational efficiency. Including all bridge elements in

the models (Ghosh & Padgett, 2010; Ghosh & Sood, 2016; Nielson & DesRoches, 2007) might

have an impact on the absolute probability of damage; however, for understanding the relative

effects of corrosion-related deterioration and cover material on the bridge fragility, a model with

the substructure element only was deemed acceptable. Additional details of the finite element

model are provided in Section 3.4.

28
3.3.3 IDA (step 3)

IDA proposed by Vamvatsikos & Cornell (2002) was performed in this study. It involved

a large number of NRHA using ground motions that were scaled systematically to increasing

earthquake intensities (PGA) until a given damage level was achieved. The IDA yielded the

structural response data that was statistically analyzed to determine the fragility curve for that

damage level.

3.3.4 Time-dependent damage states and damage index (step 4)

The level of damage is typically characterized by discrete damage states defined by

certain thresholds of a damage measure. In this study, only flexure failure is considered because

the example bridge column used in this study was slender (cross-section diameter/ height = 0.15)

and had a large amount of confinement reinforcement (51 mm spacing, No.13 (diameter = 13 mm))

transverse reinforcement], which resulted in ductile performance under seismic load (Ramanathan,

2012). Corrosion affects the structural capacity, and therefore, damage thresholds should change

with time. In this Chapter, displacement ductility was used as the damage measure, which is

defined as the ratio of the peak lateral displacement (obtained from the dynamic analysis) to the

yield displacement. The yield displacement, obtained from a push-over analysis, is defined as the

lateral displacement corresponding to the yielding of the outermost longitudinal rebar under

tension. The four damage states used in this study are defined as follows:

29
(1) Slight damage occurs when the peak lateral displacement is equal to the yield displacement

(i.e., displacement ductility = 1). At this level, the concrete cover is assumed to have visible

cracks under earthquake load.

(2) Moderate damage occurs when the maximum compressive strain in the concrete core (confined

by steel reinforcement) at the column base reaches 0.002 (Ghosh & Sood, 2016). At this strain,

the concrete cover is assumed to have minor spalling.

(3) Extensive damage level was defined as the displacement ductility equal to the geometric mean

of the displacement ductilities corresponding to the moderate and the complete damage states

(Ghosh & Sood, 2016). At this damage level, the concrete cover might have extensive cracks,

concrete spalls, and reinforcing bars are visible.

(4) Complete damage level was defined as the displacement at which the maximum compressive

strain in the confined concrete reaches 𝜀 calculated in Equation (20) (Paulay & Priestley,

1992). This value of strain in concrete corresponds to the fracture of the first transverse tie

(Afsar Dizaj et al., 2018; Ghosh & Sood, 2016).

. ( ) ( )
𝜀 = 0.004 + ( )
(20)

As described above, material stress and strain values (in either steel or concrete) were used

in this study to determine threshold displacement ductilities corresponding to the damage states.

This approach is different from the studies that utilized fixed values of displacement ductility

(Billah et al., 2013) or other macro damage measures such as curvature ductility (Ghosh & Padgett,

30
2010; Ramanathan, 2012) and drift ratio (Stefanidou & Kappos, 2017). This approach enables us

to account for the effects of corrosion on damage measures over time, which is crucial for the

combined corrosion and seismic analysis of RC bridges. The displacement ductility used as a

damage measure in this study has been used for specifying the performance criteria for highway

bridges in Caltrans Seismic Design Criteria (Caltrans, 2019). Displacement ductility is related to

force reduction which is easier to consider in nonlinear structural response analysis (Priestley,

2000).

The damage states used in this study could be related to repair strategies as shown in Table

3.2 according to the Caltrans Seismic Design Criteria (Caltrans, 2019).

Table 3.2: Repair strategies for each damage state in Caltrans Seismic Design Criteria

Damage states Damage states Repair strategies

defined in this Chapter (Caltrans, 2019) (Caltrans, 2019)
Slight damage DS-1: Flexural cracks
Bridge is essentially under
DS-2: Minor spalling and elastic performance
Moderate damage
possible shear cracks

DS-3: Extensive cracks and

spalling Bridge repair is likely, but
Extensive damage bridge replacement is
DS-4: Visible lateral and/or unlikely
longitudinal reinforcing bars

DS-5: Onset of compressive

Complete damage Bridge replacement is likely
failure of core concrete

31
3.3.5 Estimation of the parameters of fragility curves (step 5)

The parameters of the fragility curves can be obtained using several methods (Nielson &

DesRoches, 2007). In this study, the maximum likelihood method (Baker, 2015) was used to

estimate the parameters of the fragility curves, as shown below:

𝑛 ( ) ( )
𝜃̅ , 𝛽̅ = 𝑎𝑟𝑔𝑚𝑎𝑥 ∑ {𝑙𝑛 𝑧 + 𝑧 𝑙𝑛Φ + 𝑛 −𝑧 𝑙𝑛 1−Φ (21)

3.4 Framework as applied to a bridge column

A three-span highway bridge with a concrete deck and concrete tee girders supported by a

single column hammerhead bridge pier was used as an example bridge in this study to demonstrate

the framework described above. This bridge is representative of several highway bridges in

California (Ramanathan, 2012). The diameter and height of the pier column were 914 mm and 6

m (typical range of height is 4.9m to 6.8m (Ramanathan, 2012)), respectively. The details of the

superstructure of this bridge can be found in Ramanathan (2012). The longitudinal reinforcement

consisted of 18 No. 36 rebars (diameter = 35.8 mm), corresponding to a longitudinal reinforcement

ratio of 2.8% (typical range is 1.49% to 5.35% (Ramanathan, 2012)). The transverse reinforcement

consisted of No. 13 stirrups at 51 mm, corresponding to a volumetric ratio of 1.25% (typical range

is 0.31% to 1.61% (Ramanathan, 2012)). The design compressive strength of conventional

concrete was 41 MPa, and the yield strength of reinforcing steel was 414 MPa.

32
3.4.1 Finite element model

A finite element model of the column was created. The column was modeled as a single

degree freedom system with a fixed base by using a force-based “nonlinear beam-column” element

(Taucer et al., 1991) in OpenSEES. A fiber section model was used to discretize the cross-section

of the column into fibers. The column configuration and the finite element model are shown in

Figure 3.4. Unconfined concrete or ECC properties were assigned to the cover elements, whereas

confined concrete properties, determined by the model of Mander et al. (1988) and Chang &

Mander (1994), were assigned to the core elements of the column. The material properties of

concrete under tension were determined according to Chang & Mander (1994) and ACI (2014).

The material properties of ECC used in this research are representative of the mix number M45 in

Wang & Li (2007).

33
 (a) Column configuration

(b) Fiber section detail of cross-section A-A

Figure 3.4: Column model components

34
 The stress-strain behavior of concrete was modeled using the Concrete07 material model

in OpenSEES. It implements the concrete model by Chang & Mander (1994) with simplified

unloading and reloading curves and takes into account tensile strength, descending portion of the

compressive stress versus strain relation, and stiffness loss upon unloading of concrete. After

attaining a crack width of 1.0 mm in cover concrete, the cover was assumed to be ineffective for

carrying seismic loads and was therefore removed from the dynamic structural analysis (Cheng et

al., 2019). Thus, the upper bound of crack width was 1 mm. Due to lack of test data on ECC, the

upper bound of crack width for ECC cover was also assumed to be 1 mm. The stress-strain

behavior of ECC was modeled using the ECC01 material in OpenSEES. It implements Han et al.

(2003)’s ECC model and represents the tensile stress-strain behavior of ECC material by a tri-

linear curve. The material properties of ECC used in this study are obtained from Wang & Li

(2007).

Reinforcing steel stress-strain behavior was modeled using the Steel02 material model in

OpenSEES, which is based on the Giuffre-Menegotto-Pinto model (Filippou et al., 1983b)

including isotropic strain hardening. The yield strength of the rebar used in this study was 414

MPa (Grade 60) and the ultimate tensile strain of the rebar (Grade 60) was determined by

AASHTO (2020). The Grade 60 rebar has an ultimate tensile strength of 550 MPa based on ASTM

A706 (ASTM, 2020). However, the tensile strength of the rebar was assumed to be 530 MPa in

this study because the material model used in this study is a bi-linear approximation of steel

behavior. Strain and stress values used to define the constitutive relationships of concrete, ECC,

and reinforcing steel are shown in Figure 3.5.

35
(a) Compressive material properties of (b) Tensile material properties of
concrete concrete

(c) Compressive material properties (d) Tensile material properties

of ECC of ECC

(f) Material properties of steel reinforcement

Figure 3.5: Material properties used in the finite element model

36
3.4.2 Corrosion and fragility results for the example column

3.4.2.1 Corrosion initiation phase

For the example bridge column, the corrosion initiation time was computed by equating

the right-hand side of Equation (1) to the critical chloride content listed in Table 3.1. Based on

these calculations, corrosion initiation time for the transverse reinforcement (approximately 3

years in normal concrete and 5 years in ECC) was shorter than that for the longitudinal

reinforcement (approximately 5 years in normal concrete and 8 years in ECC), as the transverse

reinforcement is closer to the exterior surface of the column than the longitudinal reinforcement.

Corrosion initiation time for the column with ECC cover was 60%-66% longer than that for the

column with normal concrete cover for both transverse and longitudinal reinforcement.

3.4.2.2 Corrosion propagation phase

The computed loss in the reinforcement area (using the calculations described in Section

3.2) is plotted as a function of time for the example bridge column in Figure 3.6 (a). The area loss

of the transverse reinforcement due to corrosion was significantly greater than that of the

longitudinal reinforcement regardless of the cover material. The ECC cover reduced the area loss

rate for both longitudinal and transverse reinforcement compared to the concrete cover, and the

reduction is more significant for the transverse reinforcement. For instance, after 50 years, the

longitudinal reinforcement area left with ECC cover was about 3.1% larger than that left with the

concrete cover, whereas the transverse reinforcement area left with ECC cover was about 23.8%

37
larger than that left with the concrete cover. Thus, the column with ECC cover exhibited

significantly enhanced durability compared to the column with a concrete cover.

The computed longitudinal and transverse crack widths (caused by the corrosion of

longitudinal and transverse reinforcement, respectively) are plotted against time in Figure 3.6 (b).

The transverse cracks started earlier than the longitudinal cracks in the bridge column with either

of the two cover types. The computed crack widths (both longitudinal and transverse) increased at

a significantly faster rate in the column with concrete cover compared to the column with the ECC

cover, which is expected due to the crack bridging effect of the fibers in ECC. The small jump in

the crack width of ECC cover is caused due to the empirical relation between tensile strain and

crack width of ECC (Equation (18)) that estimates a crack width of 23 m for zero strain. As

mentioned in Section 3.4.1, cover elements were removed from the structural analysis after

reaching a crack width of 1 mm. In contrast, neither longitudinal nor transverse cracks in ECC

cover reached 1 mm even after 100 years.

38
 (a) Rebar area loss

(b) Longitudinal and transverse cover crack width variation with time

Figure 3.6: Rebar area loss and cover crack width with time

3.4.2.3 Time-dependent fragility curves of the example column

The fragility curves of the example bridge column with conventional concrete cover

computed at 0, 25, 50, and 75 years of service life (using the procedure described in Section 4) are

shown in Figure 3.7. The fragility curve at 25 years was similar to that at 0 years (initial condition);

39
however, the effects of corrosion became more pronounced with time, as the probabilities of

damage for a given PGA at 50 and 75 years were significantly greater than those at 0 and 25 years.

Furthermore, the corrosion effects became more significant for higher damage states. For instance,

the probability of complete damage of the column for PGA of 1.0 g at 75 years increased by about

54% relative to the initial probability at 0 years. In comparison, the probability of extensive damage

increased by about 18% relative to the initial probability at 0 years.

Figure 3.7: Fragility curves of the example bridge column with concrete cover (damage states
defined in Section 3.3.4)

Figure 3.8 shows the fragility curves of the same bridge column with an ECC cover at 0,

25, 50, and 75 years of service life. Similar to the observations for the column with conventional

40
concrete cover, the corrosion effects became more significant with time and at higher damage

levels. Despite this, the corrosion effects on the fragility curves were reduced with ECC cover

compared to conventional concrete cover. At slight and moderate damage levels (in Figure 3.8),

there was virtually no change in the fragility curves relative to the initial condition due to enhanced

protection against corrosion offered by the ECC cover.

Figure 3.8: Fragility curves of the example bridge column with ECC cover

Figure 3.9 shows the differences between the fragility curves of the example bridge column

obtained with ECC cover and conventional concrete cover at 0 years and 75 years. The columns

with ECC and concrete covers had similar fragilities at 0 years at all damage levels. It is because

41
the example bridge column used in this study was slender (cross-section diameter/ height = 0.15)

and had a large confinement reinforcement amount (51 mm spacing, No.13 (diameter = 13 mm)

transverse reinforcement), which resulted in a ductile performance under seismic load. Therefore,

ECC cover had a negligible effect on the seismic performance of the example bridge column

without corrosion (at 0 years). However, after 75 years, the ECC cover reduced the probability of

exceeding all damage states due to the reduction in corrosion of both longitudinal and transverse

reinforcement. For instance, at complete damage, the damage probability at PGA of 1.0 g with the

ECC cover was reduced by about 40% relative to that with the concrete cover. The estimated

logarithmic median (𝜃̅ ) and standard deviation (𝛽̅ ) of ln(IM) of the fragility functions for the

bridge column with concrete cover and ECC cover was listed in Table 3.3.

42
Figure 3.9: Comparison of fragility curves obtained using conventional concrete and ECC covers
at 0 years and 75 years

Table 3.3: Logarithmic median ( 𝜃̅) and standard deviation (𝛽̅) of ln(IM) of fragility functions for
the example bridge column with different materials
Damage state
Slight Moderate Extensive Complete
𝜃̅ 𝛽̅ 𝜃̅ 𝛽̅ 𝜃̅ 𝛽̅ 𝜃̅ 𝛽̅
0 years -1.184 0.3294 -0.838 0.340 -0.291 0.361 0.199 0.382
Concrete 25 years -1.184 0.329 -0.838 0.340 -0.316 0.357 0.163 0.366
cover 50 years -1.318 0.344 -0.990 0.302 -0.494 0.390 0.005 0.388
75 years -1.360 0.353 -1.019 0.313 -0.691 0.349 -0.379 0.383

0 years -1.226 0.345 -0.888 0.343 -0.303 0.370 0.179 0.370

ECC 25 years -1.226 0.345 -0.888 0.343 -0.311 0.377 0.167 0.361
cover 50 years -1.249 0.364 -0.893 0.345 -0.331 0.363 0.152 0.352
75 years -1.249 0.364 -0.899 0.346 -0.379 0.341 0.060 0.386

43
3.5 Parametric study

A detailed parametric study was performed to understand the sensitivity of the above

results to the input parameters used in the corrosion and structural models. For the parametric

study, the geometry (height and diameter) and mass of the example bridge column were kept

constant, and the cover material was conventional concrete. Other parameters that were varied are

described below.

3.5.1 Investigated parameters

The parameters investigated in this study were grouped into two types: (1) corrosion

parameters, (2) structural parameters – based on their primary use in either the corrosion model

(Section 3.2.3) or the structural model (Section 3.4.1) as summarized in Table 3.4. First, the effects

of corrosion parameters on the main result of the corrosion model, i.e., reinforcement area loss as

a function of time (which was an input to the structural model), were investigated. Then, only those

corrosion parameters that had significant effects on the corrosion model, along with all the

structural parameters, were investigated further to understand their effects on the fragility of the

example bridge column. Only one parameter was varied at a time. The baseline values of the

parameters, values that belonged to the example column, are marked in with an asterisk. The

rationale for the baseline corrosion parameters is provided in Section 3.2.3. The rationale for the

baseline structural parameters is provided in Section 3.4.1. The rationale behind choosing the

values of the investigated parameters is discussed in the following sections.

44
 Table 3.4: Parameters considered in the parametric study

Parameter
Parameter Values
type
Longitudinal rebar
corrosion initiation time, 2, 5*, 10
tcr_lon (Years)
5.8 1.16 5.8 1.16 1.74
𝜆(𝑡) (mm/Year) 10-4 10-3 10-3 10-2* 10-2
Corrosion Very
Negligible Low Moderate High
High
K 1.5, 2.0*, 2.5
Critical crack width,
0.1, 0.15, 0.2*, 0.25, 0.3
𝑊 (mm)
Cover thickness, 𝐶(mm) 51*, 76
Unconfined compressive
strength of concrete, 27.6, 41.4*, 55.2
fc (MPa)
Longitudinal reinforcement
Structural 1.4%, 2.8%*, 4.0%
ratio, 𝜌

Volumetric ratio of
transverse reinforcement, 0.83%, 1.25%*, 1.66%
1
𝜌

*Values used in the baseline model.

1
Ratio of the volume of transverse confining steel to the volume of confined concrete core.

3.5.2 Corrosion parameters

The investigated corrosion parameters included corrosion initiation time (tcr_lon) for

longitudinal reinforcement, corrosion rate before cracking ( 𝜆(𝑡) ), the ratio (K) of

𝜆(𝑡) to 𝜆(𝑡) , and the critical crack width (𝑊 ). According to Weyers et al. (1994),

corrosion initiation time for the bridge deck reinforcement is approximately 2 to 5 years. In the

parametric study, tcr_lon of 2, 5, and 10 years were used for longitudinal reinforcement; the value

of 10 years was used due to the lack of data about the corrosion initiation time in bridge columns

that are expected to be less exposed than bridge decks to corrosive agents. The corrosion initiation

45
time depends on the fundamental material properties (chloride diffusion coefficient (𝐷 ) and the

critical chloride content), cover depth, and exposure (surface chloride content). The corresponding

corrosion initiation times for the transverse reinforcement were estimated using Equation (1) by

accounting for the reduced distance of the transverse reinforcement from the external surface of

the column as compared to longitudinal reinforcement.

The value of 𝜆(𝑡) was varied according to Andrade & Alonso (2001), who

categorized the corrosion rates into four levels ranging from negligible to high, as shown in Table

3.4. Andrade & Alonso (2001) categorized the corrosion rate greater than or equal to 0.0116

mm/Year as “high”. In order to differentiate the 𝜆(𝑡) of 0.0116 mm/year with 0.0174

mm/Year in this study, the value of 𝜆(𝑡) of 0.0174 mm/Year was marked as “very high”.

There is a lack of data in the literature about the peak corrosion rate after cracking of the concrete

cover (𝜆(𝑡) ) and the corresponding critical crack width (𝑊 at which the corrosion rate

reaches 𝜆(𝑡) ). Therefore, reasonable values of K (𝜆(𝑡) /𝜆(𝑡) ) and 𝑊 were

assumed for these parameters as shown in Table 3.4.

3.5.3 Sensitivity of the corrosion model to the corrosion parameters

The effects of the corrosion parameters listed in Table 3.4 on rebar area loss are shown in

Figure 3.10. The purpose of this sensitivity analysis was to select the corrosion parameters with

significant effects on the rebar area loss over time, which were consequently important for the

fragility analyses. Therefore, the results of the rebar area loss in Figure 3.10 are shown only at 75

46
years, which is the last time point with the maximum possible effects of corrosion used in the

fragility analyses.

Figure 3.10 (a) shows the influence of corrosion initiation time of the longitudinal

reinforcement (tcr_lon) on the remaining longitudinal reinforcement area after 75 years for different

corrosion rates (𝜆(𝑡) ). The remaining rebar area increases slightly with an increase in tcr_lon.

For example, the remaining rebar area after 75 years increases by 5% when tcr_lon increases from 2

years to 5 years for 𝜆(𝑡) = 0.0174 mm/Year. This is because the rebar area loss happens

only in the corrosion propagation phase, and longer corrosion initiation time delays the start of the

corrosion propagation phase. Furthermore, as discussed below, greater corrosion rate results in

larger rebar area loss, and therefore, the influence of tcr_lon on the remaining rebar area increases

with an increase in 𝜆(𝑡) .

Figure 3.10 (b) shows that the remaining rebar area decreases with an increase in corrosion

rate before cracking in concrete cover (𝜆(𝑡) ). For example, rebar area remaining after 75

years decrease by 8% when 𝜆(𝑡) increases from 0.0058 mm/Year to 0.0116 mm/Year for

rebar diameter (Drebar) of 36 mm. It is further observed in Figure 3.10 (b) that corrosion rates less

than 0.0058 mm/Year do not influence the remaining rebar area significantly. Therefore, only the

higher corrosion rates of 0.0058 mm/Year, 0.0116 mm/Year, and 0.0174 mm/Year were considered

in the fragility analysis.

Figure 3.10 (c) shows the relationship between rebar area loss and K (the ratio of the post-

crack peak corrosion rate to 𝜆(𝑡) ). As increase in the value of K increases the corrosion

47
rate after cracking, the remaining rebar area decreases with an increase in K. For example, the

rebar area remaining after 75 years decreases by 9% when K is increased from 1.5 to 2 for

𝜆(𝑡) = 0.0174 mm/Year. Figure 3.10 (d) shows that Wcri (the critical crack width at which

the peak corrosion rate is achieved after cracking) has a negligible effect on the remaining rebar

area, regardless of the corrosion rate. Therefore, Wcri of 0.2 mm was used for conventional concrete

in the following fragility analysis.

(a) Effects of tcr_lon (b) Effects of 𝜆(𝑡)

(c) Effects of K (d) Effects of Wcri

Figure 3.10: Effects of the corrosion parameters on remaining longitudinal reinforcement area
after 75 years

48
3.5.4 Structural parameters

The investigated structural parameters included cover thickness (𝐶), concrete compressive

strength (fc), longitudinal reinforcement ratio (𝜌 ) and transverse reinforcement ratio (𝜌 ).

These parameters are typically selected based on structural demands. The cover thickness values

of 51 mm (for bridges without saltwater, deicing salt, and coastal exposure) and 76 mm (bridges

with coastal exposure) were considered as per AASHTO (2020) for bridges without saltwater

exposure and with coastal exposure, respectively. Three values of the unconfined compressive

strength of concrete were considered: 27.6 MPa (4 ksi), 41.4 MPa (6 ksi), and 55.2 MPa (8 ksi),

which are commonly used in the bridge substructures in the United States.

Three values of the nominal diameter of the longitudinal reinforcement were considered:

25.4, 35.8 mm, and 43.0 mm (number of rebars was kept constant), resulting in longitudinal

reinforcement ratios of 1.4%, 2.8%, and 4.0%, respectively. These longitudinal reinforcement

ratios were within the 1.0%-4.0% range required by AASHTO (2020) for bridges in seismic zones.

Finally, the volumetric ratio of transverse reinforcement was varied by changing only the spacing

of the stirrups at 38 mm, 51 mm, and 76 mm, leading to volumetric reinforcement ratios of 1.66%,

1.25%, and 0.83%, respectively. The lowest value (38 mm) is the minimum spacing allowed by

AASHTO (2020).

3.5.5 Sensitivity of the fragility curves to corrosion parameters

Figure 3.11 to Figure 3.13 show the effects of the corrosion parameters with significant

effects on the reinforcement area loss (Section 6.3) on the column fragility at various times. In all

49
the figures, the probability of damage (z-axis) is plotted at various PGAs (0.0g, 0.3g, 0.6g, 0.9g,

1.2g, and 1.5g) and at various times (0 years, 25 years, 50 years, and 75 years), for varying

parameters. In these figures, coincident or parallel lines joining the result points for a given PGA

indicate that the parameter had a negligible influence on the fragility curve for that damage state

with time. Figure 3.11 shows that the corrosion initiation time of the longitudinal reinforcement

(tcr_lon) had negligible effects on the fragility curves in all damage states. This is likely because the

corrosion initiation times (2 to 10 years) were small compared to the time scale considered (tens

of years), which resulted in insignificant effects on the reinforcement area loss.

Figure 3.11: Influence of corrosion initiation time of longitudinal reinforcement (tcr_lon) on

column fragility over time

50
 From Figure 3.12 and Figure 3.13, it can be observed that both pre-crack corrosion rate

(𝜆(𝑡) ) and the ratio K (𝜆(𝑡) /𝜆(𝑡) ) had significant effects on the column

fragility, and the effects increased with time and severity of the damage levels. Higher 𝜆(𝑡)

and K led to increased reinforcement area loss at a given time, which in turn increased the

probability of damage. Furthermore, the effects of K on the column fragility were magnified with

an increase in 𝜆(𝑡) due to the increase in the post-crack peak corrosion rate.

Figure 3.12: Influence of pre-crack corrosion rate (𝜆(𝑡) ) on column fragility over time

51
 Figure 3.13: Influence of K (= 𝜆(𝑡) /𝜆(𝑡) ) on column fragility over time

3.5.6 Sensitivity of the fragility curves to structural parameters

Figure 3.14 to Figure 3.17 demonstrates the influence of the structural parameters on

seismic fragility curves. Figure 3.14 shows the impact of cover thickness on the column fragility

over time. Larger cover thickness is typically recommended in design standards for durability. The

diameter of the example column considered in this study was more than an order of magnitude

greater than the cover thickness, and therefore, the cover thickness had a negligible influence on

the column capacity. The increase in cover thickness from 50.8 mm to 76.2 mm changed the

corrosion initiation time only by 6 years for conventional concrete, which was significantly smaller

52
than the time interval (25 years) used for computing the fragility curves. Additionally, as defined

in Section 5.1, when the crack width in the concrete cover reached 1 mm, the concrete cover was

rendered ineffective and was removed from the dynamic analysis. For the example column, the

cracks in the concrete cover due to reinforcement corrosion reached 1 mm at 50 and 75 years for

all the damage states. Therefore, the probabilities of damage at all the damage levels at 50 years

and 75 years were the same for columns with different cover thickness. Due to these reasons, the

overall influence of cover thickness on the fragility curves was negligible.

Figure 3.14: Influence of cover thickness (𝐶) on column fragility over time

Figure 3.15 shows that the compressive strength of concrete (fc) had a negligible effect on

the fragility curves overtime for all damage levels, except for slight sensitivity at the moderate and

53
complete damage levels. For moderate damage (when compressive strain in confined concrete

reaches 0.002), higher strength and higher stiffness reduces the displacement and, therefore, the

damage probability of a column. For complete damage, smaller ultimate compressive strain 𝜀

(damage threshold) of higher concrete strengths overcomes the higher stiffness of higher strengths

of concrete, leading to a higher probability of damage. Slight damage state is not affected by the

concrete strength because the compressive stress in concrete is very small at this damage state, i.e.,

at initial yielding of the longitudinal reinforcement. Finally, the impact of concrete compressive

strength on extensive damage is minimal because this state is defined as the geometric mean of

moderate and complete damage, whose individual influences by compressive strength of concrete

cancel one another.

54
 Figure 3.15: Influence of concrete compressive strength (fc) on column fragility over time

Figure 3.16 shows that the longitudinal reinforcement ratio had significant effects on the

fragility curves at all damage levels. This is because the damage levels are governed by the flexural

response of the column, which is significantly influenced by the longitudinal reinforcement ratio

(long). Greater long increases the displacement threshold for yielding of the longitudinal

reinforcement, which reduces the probability of slight damage at a given ground motion intensity.

Greater long also increases the stiffness of the column, which reduces the strain in confined

concrete. Therefore, the probability of moderate damage, complete damage, and extensive damage

for a given ground motion intensity decreases with an increase in long.

55
 Figure 3.16: Influence of longitudinal reinforcement ratio (𝜌 ) on column fragility over time

Figure 3.17 shows that the volumetric transverse reinforcement ratio (trans) had negligible

effects on the fragility curves at slight damage and moderate damage levels. For higher damage

levels, the effects of trans were significant, but they decreased with time. This is because larger

trans results in larger confined concrete compressive strength (fcc) as well as larger compressive

strain in confined concrete, which effectively increases the ductility of the column leading to a

lower damage probability. However, this effect reduces over time due to the definition of complete

damage in this framework (Equation (20)). As observed in Equation (20), the corrosion of

56
transverse reinforcement over time reduces its confinement effectiveness, which in turn reduces

the difference of 𝜀 between the columns with different trans over time.

Figure 3.17: Influence of transverse rebar volumetric ratio (𝜌 ) on column fragility over time

Summarizing the results of the parametric studies, the pre-crack corrosion rate and

longitudinal reinforcement ratio had the maximum impact on the column fragility, whereas

corrosion initiation time and cover thickness had negligible effects on the example column and

parameter ranges examined in this study. As mentioned in Section 4.4, only flexure failure was

considered in this study because the bridge column was slender (cross-section diameter/height =

57
0.15) and had tight confinement (51 mm spacing, No.13 (diameter = 13 mm) transverse

reinforcement).

3.6 Conclusions

This Chapter presents a systematic framework for determining the seismic fragility of

bridge columns undergoing corrosion deterioration. The framework takes into account the cover

material and its cracking characteristics, the effects of pitting corrosion on both longitudinal and

transverse reinforcements, cracking in conventional concrete, and the time-dependent damage

thresholds in the fragility analysis. The framework is demonstrated using an example RC bridge

column to demonstrate the influence of changing the cover material from conventional concrete to

a ductile fiber-reinforced concrete (ECC) on the durability and seismic fragility over a period of

75 years. For this purpose, a novel method to incorporate the influence of crack width and the

influence of the unique cracking behavior of ECC on the corrosion rate was developed and

presented in this manuscript. Finally, the sensitivity of the corrosion model and fragility curves to

various framework parameters was determined to identify their relative significance. The main

conclusions of this study are below:

1. The estimated percentage of reinforcement area loss was larger for transverse reinforcement

than for longitudinal reinforcement regardless of the cover material. This was due to the

shorter distance of the transverse reinforcement to the column’s exterior surface than that of

the longitudinal reinforcement.

58
2. During the corrosion propagation phase and after cracking of the cover, the computed

reinforcement area loss was lower with ECC cover than that with conventional concrete cover

due to better crack width control with ECC.

3. The computed effects of corrosion on the seismic performance (fragility) of the example

column became more significant with time. These effects were significantly lower for the

column with ECC cover than with conventional concrete cover. Furthermore, the estimated

improvements in the seismic performance of the example column due to corrosion reduction

by ECC cover were more pronounced at higher damage levels than at lower damage levels.

Thus, an improvement in column durability is expected to improve its seismic performance as

well.

4. Among the various parameters investigated in this study, the pre-crack corrosion rate and

longitudinal reinforcement ratio had the maximum impact on the column fragility, whereas

corrosion initiation time and cover thickness had negligible effects.

5. There are opportunities to improve the current framework in the future, including the

experimental investigations of bridge column with ECC cover under the combined effects of

corrosion and earthquakes, the location of pitting corrosion, and concrete spalling effects on

the corrosion rate in simulations.

59
 Chapter 4: Seismic Damage Risk Assessment of Aging Bridges

4.1 Introduction

The objective of this chapter was to apply the framework developed in Chapter 3 to a group

of real RC bridges. The goal of this exercise is to demonstrate the application of the corrosion and

seismic evaluation framework to inventories of bridges (as opposed to individual bridges) and to

demonstrate how the framework can be used to identify bridges in an inventory with the highest

need of maintenance. The group of bridges were selected from the inventory of the Washington

State Department of Transportation (WSDOT). The computational framework presented in

Chapter 3 included corrosion modeling and seismic fragility analysis. In this Chapter, the

framework presented in Chapter 3 was expanded to include risk analysis, which involved

integrating fragility functions with the annual probability of a seismic hazard exceeding a certain

ground motion intensity. The corrosion model, seismic fragility analysis and seismic risk analysis

are described in the next three sections, followed by the results and discussion and the main

conclusions of this project.

4.2 Corrosion Modeling

As mentioned in Chapter 3 chloride-induced corrosion is the focus of this study, as it is the

most common cause of bridge degradation in the northern and coastal US. The mathematical

equation describing the corrosion initiation and propagation processes are given in Chapter 3

Section 3.2. Therefore, the corrosion modeling methodologies were not discussed in this section

60
again. This section focuses on presenting a summary of the parameter values used for modeling

corrosion in the groups of bridges investigated in this study.

4.2.1 Parameters used in corrosion modeling

Corrosion initiation phase: In this study, all the bridges were analyzed from the time of original

construction, and therefore, the initial unbound chloride concentration in the cover concrete of all

the bridges was assumed zero. The surface chloride concentration (C s) was assumed 3.6 kg/m3 for

bridges from WSDOT, which represents the exposure environment created by the de-icing salts,

based on Weyers et al. (1994). Due to the lack of reliable corrosion data for Washington State, the

chloride diffusion coefficient (𝐷 ) was assumed to be 4.3  10-12 m2/s, based on available field

tests of bridge beams in USA (Enright & Frangopol, 1998). The critical chloride content for

corrosion initiation at the rebar-concrete interface was assumed to be 1.2 kg/m 3 (Life-365™,

2020).

Corrosion rate: A corrosion rate of 0.0116 mm/year suggested in the literature (Andrade & Alonso,

2001) was assumed prior to crack initiation. Due to the lack of peak corrosion rate data in real

bridges, the peak corrosion rate after crack initiation was assumed two times (equal to 0.0232

mm/year) that of the corrosion rate before crack initiation. It was also assumed that no repair was

performed during the bridge’s service life. The critical crack width beyond which the corrosion

rate became constant was assumed to be 0.2 mm (Cui et al., 2018). Table 4.1 summarizes all the

parameters used in the corrosion model.

61
 Table 4.1: Parameters used in corrosion model

Parameters Values Source

Weyers et al.
𝐶 (kg/m3) 3.6
(1994)
Enright &
Frangopol (1998);
𝐷 (m2/s) 4.3 10-12
Weyers et al.
(1994)
Critical chloride content
1.2 Life-365™ (2020)
at rebar surface (kg/m3)
Andrade & Alonso
𝜆(𝑡) (mm/year) 0.0116
(2001)
As explained in
𝜆(𝑡) (mm/year) 0.0232
section 3.4.2.1
Critical crack width,
0.2 Cui et al. (2018)
𝑊 (mm)
𝜈 0.18 El Maaddawy &
𝛿 (mm) 20 × 10 -3 Soudki (2007)
Val & Melchers
Pitting factor, R 6
(1997)

4.2.2 Outcomes of corrosion modeling

The outcome of the corrosion model is the estimation of loss of reinforcement area due to

corrosion for bridge piers at 0 years, 25 years, 50 years and 75 years along their lifespan. Both

longitudinal and transverse reinforcement in piers were considered.

4.3 Seismic Fragility Analysis

A detailed procedure to develop seismic fragility function has been discussed in Chapter 3

Section 3.3. There are 5 steps used in generating the fragility functions: (1) Ground motion

62
selection, (2) Structural model, (3) IDA, (4) Time-dependent damage states and damage index and

(5) Estimation of the parameters of fragility curves. In order to consider the time effects on the

seismic fragility curves, Step 6 was introduced and added to the seismic fragility analysis as

discussed below. The earthquake intensity measure, EDP, damage index, damage state definitions,

and the methods used to develop fragility curves can be found in Chapter 3 Section 3.3.

4.3.1 Development of fragility curves for various bridge ages (Step 6)

Due to the high computational cost of IDA, the fragility curves were computed in this study

only for four time points along a bridge’s service life: 0 years, 25 years, 50 years, and 75 years

since the original construction of a given bridge. As explained above, each fragility curve is

described by the lognormal distribution parameters 𝜃 and 𝛽. To be able to integrate fragility

functions with the hazard curves (Section 4.4), the values of lognormal distribution parameters

between 0 years, 25 years, 50 years and 75 years are also needed. These parameters were obtained

through curve fitting. For each lognormal distribution parameter ( 𝜃 and 𝛽 ), a second-order

polynomial as expressed by Equation (22) was fitted using the known values of 𝜃 and 𝛽 at years

0, 25, 50 and 75. A second-order polynomial function was considered because 𝜃 and 𝛽 follow such

a trend with time as reported by others (Ghosh & Padgett, 2010). Fitting was performed using the

nonlinear least square method and the method for obtaining coefficients of the polynomial, 𝑊 , 𝑊

and 𝑊 , was similar to other studies (Ghosh & Padgett, 2010). Here, 𝑃𝑎𝑟𝑎(𝑡) is either 𝜃 or 𝛽 as a

function of time t.

𝑃𝑎𝑟𝑎(𝑡) = 𝑊 × 𝑡 + 𝑊 × 𝑡 + 𝑊 (22)

63
 The outcome of seismic fragility analysis is seismic fragility functions at 0 years, 25 years,

50 years and 75 years along the service life of a bridge. Outputs of the corrosion model (loss of

reinforcing bar area) are used as input to the seismic fragility analyses to understand seismic

damage probability considering corrosion.

4.4 Integration of Seismic Hazard and Fragility Curves

A fragility function quantifies the probability of damage to a structure exposed to a given

set of ground motions, it does not account for the probability of the hazard. Seismic hazard curves

are used for this purpose (USGS, 2021), which provide the annual probability of exceeding a

certain ground motion intensity. Obtaining a seismic hazard curve was particularly important for

accuratly estimating the seismic risk for structures located in a specific region. Figure 4.1 shows a

seismic hazard curve for a bridge in Washington State in which the annual probability of

exceedance and PGA (an intensity measure) are plotted on log scales.

64
 Figure 4.1: Seismic hazard curve for an example bridge in Washington State (Bridge WS-1 in
Table 4.2)

4.4.1 Method to generate seismic hazard curves

Seismic hazard curves can be generated by both deterministic and probabilistic methods.

In this study, probabilistic seismic hazard analysis (PSHA) (McGuire, 2008) is used, which

considers the uncertainty in the location, size, and resulting shaking intensity of earthquakes

incorporating the historical data of a region. Seismic hazard curves were generated using the

Unified Hazard Tool (USGS, 2021) provided by the United States Geological Survey (USGS).

The hazard curve provided by the tool shows the annual frequency of exceeding a given PGA (at

a given location) plotted against the PGA, which can be converted to the probability of exceeding

that PGA using the Poisson distribution.

65
4.4.2 Method to integrate seismic hazard curves and fragility curves

ASCE 43-19 (ASCE, 2019) provides a method to calculate the probability of exceeding a

given damage state considering both the seismic hazard curve and the fragility curve for that

damage state. In this approach, the seismic hazard curve is modeled by an approximate equation,

as shown below:

𝐻(𝑎) = 𝐾 × 𝑎 (23)

where 𝐻(𝑎) is the probability of exceeding of a PGA (𝑎). 𝐾 is a constant, and 𝐾 is a slope

parameter defined by :

𝐾 = (24)
( )

where 𝐴 is the ratio of exceedance frequency at a given PGA (or spectral acceleration) to the

exceedance frequency at a PGA (or spectral acceleration) that is 10-times smaller than this PGA

(or spectral acceleration). The probability of exceeding a given damage state is obtained by

convolution of the seismic hazard curve and fragility curve, which is expressed by:

𝑃 = ∫ 𝐻(𝑎) × 𝑑𝑎 (25)

where 𝑃 is the probability of exceeding a damage state i, is the log-normal probability density

function, which is the derivative of fragility curve (Fc) with respect to the PGA (a). Using Equation

(23) and Equation (24) along with Equation (25), 𝑃 can be expressed analytically as shown below,

where 𝜃 and 𝛽 are parameters of the lognormal distribution representing the fragility curve for a

given damage state.

𝑃 = 𝐻(𝜃) × exp[(𝐾 𝛽) /2] (26)

66
4.4.3 Outcomes of integration of seismic hazard and fragility curves

The outcome of this process is the probability of exceedance of a given damage state as a

function of PGA, considering both the hazard curve and the fragility curve. For each bridge, the

seismic hazard curve generated by USGS was first converted to the analytical form in Equation

(23) using the nonlinear least square fitting method, as shown for one of the bridges in Figure 4.2.

The fragility curve was developed for a given damage state using the steps given in Chapter 3.

Then, the damage probability for that damage state was obtained using Equation (26).

Figure 4.2: Seismic hazard function given for an example bridge (Bridge WS-1 in Table 4.2)

67
4.5 Bridge Details and Modeling Strategies

This section provides a detailed description of the analyzed bridges and the modeling

strategies adopted in this study. Eight bridges from WSDOT were selected in total. All the bridges

were built between 1960 and 1965 with RC substructures. In the following sections, the details of

the bridges from WSDOT are given. Then the modeling strategies are discussed, including: (1)

material model for concrete and rebar, (2) finite element model, including element choices and

boundary conditions.

4.5.1 Bridge details

Details for WSDOT bridges are summarized in Table 4.2 including year built, number of

spans, span lengths, types of superstructure and substructure, sizes of rebars, design compressive

strength of concrete, and yield strength of rebars. and their layouts are given in Figure 4.3. It should

be noted that some bridges were upgraded to carry more lanes after the original construction, but

only the original design information was available for this study. Therefore, the layouts of WSDOT

bridge substructures are based on the original designs.

68
 Table 4.2: Summary of WSDOT bridges
Span length (m) Superstructure Substructure

Bridge Built year Spans Rebar Diameter (mm) Concrete Rebar

Substructure
Side Middle Deck type Girder type strength strength
type Longitudinal Transverse
(MPa) (MPa)
Concrete
WS-1 1959 3 17.8 23.8 - Columns 25 13 25 276
box
Prestressed
WS-2 1965 3 15.4 16.4 Concrete Columns 29 10 28 276
concrete
Voided
WS-3 1960 3 14.6 19.5 - Columns 29 13 25 276
concrete
Prestressed
WS-4 1960 4 10.7 19.3 Concrete Columns 32 10 25 276
concrete
Prestressed
WS-5 1963 3 16.3 16.3 Concrete Columns 29 10 28 276
concrete
WS-6 1966 3 14.8 17.5 Concrete - Columns 29 13 28 276
Prestressed
WS-7 1961 3 15.5 15.7 Concrete Columns 32 10 25 276
concrete
Prestressed
WS-8 1961 3 14.6 14.6 Concrete Columns 32 10 25 276
concrete

69
(a) Bridge WS-1

(b) Bridge WS-2

70
(c) Bridge WS-3

(d) Bridge WS-4

71
(e) Bridge WS-5

(f) Bridge WS-6

72
 (g) Bridge WS-7

(h) Bridge WS-8

Figure 4.3: Layouts of WSDOT bridge substructures

4.5.2 Modeling strategies

The IDA used for the fragility analysis (Chapter 3) has a high computational cost, and it

has to be performed multiple times for each bridge at different times across their lifespan. In order

to improve the overall computational efficiency for a group of bridges, two assumptions were

made in this study: (1) earthquake load was applied in the longitudinal direction (parallel to traffic)

73
of a bridge, and (2) the seismic performance of a bridge was represented by the seismic

performance of its substructure. The behviour of the substructure is represented by the

performance of one interior column as illustrated in Figure 4.4. The effective mass of the

superstructure components carried by a single column was calcualted according to the tributary

area of the column. This simplication is justified because the design of each column of a pier was

the same, the arrangement of the columns in a pier was symetric and the seismic load was applied

in the bridge longitudinal direction (out of plane direction in Figure 4.4).

Figure 4.4: Modeling simplification of the bridge pier

4.5.2.1 Finite element model

An open-source software, OpenSEES (Mazzoni et al., 2006; McKenna et al., 2000; USGS,

2021) was used to create a finite element model of each bridge pier, which was a column (single

74
or multiple columns pier), to generate the seismic fragility curves for various damage states. Each

pier (column) was modeled as a four-degree freedom system with a fixed base using a

“displacement-based beam-column” element in OpenSEES as shown in Figure 4.5. The number

of beam-column elements and the type of element (displacement vs force based elements) used in

this Chapter are different than the ones used in Chapter 3 because the columns investigated in this

Chapter are taller and more slender than the ones in Chapter 3. In this Chapter, the column is

modeled using additional elements and using the displacement-based beam-column element in

order to have convergent nonlinear structural analyses. A fiber section was used to discretize the

cross-section of a pier (column). The important parts of the model for an example substructure are

shown in Figure 4.5 (a). The model details of all the bridges investigated in this study are

summarized in Table 4.3. In this table, the transverse reinforcement ratio is calculated as the ratio

of transverse rebar volume to confined concrete core volume, f’c is the compressive strength of

unconfined concrete and A is the gross area of the cross-section.

4.5.2.2 Material models

Unconfined concrete properties were assigned to the cover elements, whereas confined

concrete properties, calculated by the model of Mander et al. (1988), were assigned to the core

elements. The stress-strain behavior of concrete was modeled using the Concrete07 material model

in OpenSEES. This material model implements the concrete model by Chang and Mander (Chang

& Mander, 1994) with simplified unloading and reloading curves, taking into account the tensile

strength, descending portion of the compressive stress versus strain relation, and reduced stiffness

during the unloading of concrete. After cover concrete attained the crack width of 1.0 mm, it was

assumed to be ineffective for carrying mechanical loads and was therefore removed from the

75
dynamic structural analysis (Cheng et al., 2019). Reinforcing steel stress-strain behavior was

modeled using the Steel02 material model in OpenSEES, which is based on the Giuffre-

Menegotto-Pinto model (Filippou et al., 1983a), including isotropic strain hardening. The

constitutive relationships of concrete07 and steel02 are shown in Figure 4.5 (c).

Table 4.3: Model details of bridges

Long. rebar ratio, Superstructure

Bridge Height Trans. rebar ratio,
weight, P P/(fc’×A)
name (m) 𝜌 𝜌
(KN)
WS-1 9.00 1.0% 0.20% 3197.5 19%
WS-2 5.14 1.1% 0.12% 1419.0 7.8%
WS-3 6.26 1.3% 0.20% 1700.9 12.0%
WS-4 5.70 0.9% 0.12% 1062.5 5.8%
WS-5 6.00 1.1% 0.12% 908.5 5.0%
WS-6 8.17 1.4% 0.20% 1615.9 8.9%
WS-7 5.53 1.0% 0.12% 745.9 4.5%
WS-8 7.00 1.0% 0.12% 715.5 4.4%
fc’: Compressive strength of unconfined concrete
A: Gross area of the cross-section

76
 (a) Substructure model components

(b) Typical cross section

Concrete07 of OpenSEES material library

Steel02 of OpenSEES material library

(c) Material properties

Figure 4.5: Model details and material models

77
4.6 Results and discussion

4.6.1 Corrosion model results

4.6.1.1 Corrosion initiation phase

The computed times to corrosion initiation for all the bridge substructures (columns) are

shown in Table 4.4. It should be noted that the surface chloride concentration and chloride

diffusion coefficient for all the bridges in the same state (e.g., Washington) were assumed to be

the same, as given in Section 4.2.1. Therefore, the difference in corrosion initiation times between

the bridges of the same state was solely due to the difference in their concrete cover thicknesses.

Larger cover thickness increases the distance between the external surface of the column and the

steel reinforcement, which in turn increases the time taken by the chloride ions to diffuse through

the concrete cover. As observed in Table 4.4, the corrosion initiation time increases with an

increase in the concrete cover thickness. Additionally, the corrosion initiation time for the

transverse reinforcement is shorter than that for the longitudinal reinforcement because the

transverse reinforcement is closer to the exterior surface of a column/wall than the longitudinal

reinforcement. A more refined prediction can be made if the local exposure conditions of each

bridge, such as deicing salt amount on a specific road, are known and captured in corrosion

modeling.

Among the WSDOT bridges, WS-4 and WS-1 exhibited the longest and the shortest

corrosion initiation times, respectively. Bridges WS-5, WS-6 and WS-7 have the same corrosion

78
initiation time for both the longitudinal reinforcement and transverse. Bridges WS-3 and WS-6

exhibited the shortest corrosion initiation time due to the smallest cover thickness.

Table 4.4: Corrosion initiation time

Cover thickness Corrosion initiation time (years)

Bridge number
(mm) Longitudinal rebar Transverse rebar
WS-1 57 20.0 13.0
WS-2 46 12.5 8.5
WS-3 38 10.5 6.0
WS-4 59 19.0 14.0
WS-5 46 12.5 8.5
WS-6 38 10.5 6.0
WS-7 46 12.5 8.5
WS-8 46 12.5 8.5

4.6.1.2 Corrosion propagation phase

As explained in Section 3.2, the rebar area reduces during the corrosion propagation phase.

The calculated remaining rebar areas after 25, 50, and 75 years as percentages of the initial areas

for all the bridges are shown in Table 4.5. In these calculations, corrosion is assumed to equally

reduce the area of all reinforcing bars. The results indicate that rebars with larger initial areas tend

to exhibit slightly less percentage area reduction over time due to corrosion. For example, consider

the longitudinal reinforcement bars in bridges WS-2 and WS-7. Although the corrosion initiation

times for both the bridges are the same, the remaining rebar area in WS-7 (with No.32 longitudinal

rebars) after 75 years is 2.5% larger than that WS-2 (with No.29 longitudinal rebars). Pitting

corrosion considered in this study causes the same absolute rebar area reduction regardless of the

bar size, which in turn lowers the percentage area reduction for larger bar sizes. WSDOT bridges

WS-3 and WS-4 exhibit the least remaining longitudinal reinforcement areas after 75 years in their

79
respective groups, which makes them the most vulnerable to seismic damage as shown in the

following sections.

Table 4.5: Remaining rebar area

Remaining rebar area (%)

Bridge number After 25 years After 50 years After 75 years
Long. Trans. Long. Trans. Long. Trans.
WS-1 99.9 99.0 97.0 81.0 87.9 46.0
WS-2 99.8 95.6 96.1 58.4 87.5 8.9
WS-3 99.7 96.2 95.5 71.7 86.5 34.4
WS-4 99.9 98.5 98.1 68.8 92.2 18.6
WS-5 99.8 95.6 96.1 58.5 87.5 8.9
WS-6 99.7 96.2 95.5 71.7 86.5 34.4
WS-7 99.8 95.6 96.9 58.5 90.0 8.9
WS-8 99.8 95.6 96.9 58.5 90.0 8.9
Long.: Longitudinal rebar
Trans.: Transverse rebar

4.6.2 Seismic damage probability

4.6.2.1 Impact of bridge characteristics on seismic damage probability

Figure 4.6 shows the damage probability variation over 75 years (since the built year) for

WSDOT bridges at all the damage states. For the slight damage state, which is marked by the onset

of yielding in the outermost longitudinal rebar, the damage probability is mainly controlled by the

height and the axial load ratio (P/(fc×A)) (given in Table 4.3) of the bridge substructures. The

tallest WSDOT bridge substructures, WS-1, WS-6, and WS-8, exhibit the lowest probabilities of

slight damage. This is because the bridge substructures with larger heights have smaller stiffness,

which increases the displacement causing yielding in the outermost rebar. Although the low

stiffness also increases the dynamic responses of the bridge column, the yielding of rebar happens

80
under low seismic force. Therefore, the increase in the dynamic response is not as significant as

the increase in the displacement for the yielding of the rebar, which results in a low probability of

damage at slight damage states. For bridge WS-7, although it is not as tall as WS-1, WS-6, and

WS-8, the probability of slight damage is low because the axial load ratio of WS-7 is much lower

than other bridges, resulting in the smaller inertial force and smaller displacement response in

dynamic analysis.

The damage probability is mainly controlled by the axial load ratio (P/(fc×A)) at moderate

and extensive damage states. As a result, bridges WS-1, WS-2, WS-3, and WS-6, which have the

largest axial load ratios, exhibit the largest probabilities of moderate and extensive damage among

the WSDOT bridges. This is because the large axial load ratio results in the larger seismic force

(inertial force), which increases the dynamic responses.

For the complete damage state, which is marked by achieving the ultimate strain in

confined concrete (calculated by Equation (20) in Chapter 3), the damage probability is mainly

affected by the axial load ratio and transverse reinforcement ratio of the bridge substructures (given

in Table 4.3). The bridges that exhibited the highest damage probability are WS-1, WS-2, and WS-

3. These bridges have the highest axial load ratio. As mentioned before, a large axial load ratio

results in a large seismic force. The bridge WS-6 has a larger axial load ratio than WS-2. However,

its probability of damage is lower because the WS-6 has a larger transverse reinforcement ratio,

which increases the ultimate strain in confined concrete, resulting in a more ductile performance.

81
 It is interesting to note that at moderate and extensive damage states, the damage

probability of bridge WS-4 increases the fastest among all bridges over time. This is because WS-

4 has the smallest concrete core left after 50 years. All the concrete cover is assumed to have

spalled when the crack width (due to corrosion damage) reached 1 mm (Section 5.2.2). All the

WSDOT bridge columns have the same gross section area (including cover). Therefore, after cover

removal due to corrosion damage, WS-Bridge 4 has the smallest residual section area (as it has the

largest cover thickness, as shown in Table 5), which resulted in the largest decrease in stiffness

and the largest increase in dynamic responses. According to AASHTO (2020), the target level of

safety for bridges should have a failure probability of less than 2.0×10 -4. The failure probability

computed in this Chapter is higher because the failure probability provided by AASHTO (2020)

is for bridges designed for Strength I Limit States (basic load combination relating to the normal

vehicular use of the bridge without wind). The failure probability is expected to be higher for

bridges under corrosion and earthquake.

82
 Figure 4.6: Damage probability for bridges from WSDOT

4.6.2.2 Relative seismic damage probabilities of individual bridges in a group

When evaluating their inventory, departments of transportation are expected to assign a

higher priority of maintenance, repair and retrofit to bridges that have the highest expected damage

probability in Figure 4.6 for Washington State bridges, the order of damage probability (and

therefore priority of intervening actions) of bridges in the group changed over time for some

damage states, as suggested by damage probability curves crossing one another in Figure 4.6.This

indicates that departments of transportation that do not consider deterioration in their seismic

damage assessment may not be able to correctly identify bridges with the highest risk of seismic

damage. It should be noted that the bridges selected for this study for demonstration of the

83
assessment were built in similar years. In addition, the corrosion model captured the general

environmental conditions in a state, but did not consider local exposure conditions for each

individual bridge. For bridges with large differences in age and local exposure conditions, the order

of seismic damage risk is expected to change more drastically than the bridges considered in this

study for the demonstration of the assessment method.

4.7 Conclusions

This Chapter demonstrates the application of a corrosion and seismic evaluation framework

for determining the seismic vulnerability of individual bridges in a group to aid in implementation

of the framework by departments of transportation to larger bridge inventories. The framework is

used to quantify seismic damage probability across service life, to identify bridges in a network

with a higher damage risk, and therefore, to inform maintenance decisions for a group of bridges.

A group of bridges, owned by Washington State Department of Transportation (WSDOT), is

considered in this study. The framework consists of three parts: corrosion modeling, seismic

fragility analysis, and seismic risk analysis. The first two parts were developed in Chapter 3. The

latter is added in this Chapter. The framework takes into account the effects of several key

parameters on the corrosion and structural models. Most importantly, the effects of pitting

corrosion on both longitudinal and transverse reinforcement areas, the effects of cracking in

concrete cover on corrosion rates, and the effects of site-specific seismic hazard curves are

considered in this framework. The main conclusions of this research are given below:

1. All bridges become more vulnerable to seismicity due to corrosion deterioration. Therefore,

neglecting the effects of corrosion deterioration may lead to unconservative estimates of seismic

84
damage risk. Seismic evaluation of bridges should be performed considering the current state of

bridge substructures, as documented by inspections (if available) or as predicted by a corrosion

model such as the one presented in this Chapter. For planning repair and retrofit actions for the

future, the combined seismic and corrosion model presented in this Chapter can be used to predict

the performance of bridges in an inventory considering both hazards, allowing prioritization of

intervening actions as well as quantification of the impacts of these actions on bridge performance.

2. The demonstration of application of the seismic and corrosion assessment of bridges to a group

of bridges showed that that the order of seismic vulnerability of bridges in an inventory can change

over time due to corrosion. This implies that departments of transportation that do not consider

corrosion in seismic assessments may not be able to identify bridges with the highest need of

maintenance, repair or retrofit. The effect of corrosion on the order of seismic damage probability

of bridges in a group is expected to be more significant for inventories with a large variability in

bridge age, construction practices (e.g., cover distance) and exposure condition.

3. The order of vulnerability did not change for some bridges. This may be due to the assumption

of similar local conditions (such as salt exposure), and similar ages for some bridges. Incorporation

of local conditions to the corrosion model may refine the results presented in this Chapter. A

method that considers local conditions on chloride exposure is presented in Chapter 5.

85
 Chapter 5: Estimating Chloride Exposure of Reinforced Concrete Bridges

Using Vehicle Spray and Splash Mechanisms

5.1 Introduction

Corrosion of reinforcement is a major durability problem for reinforced concrete (RC)

bridges located in cold regions across the world. In these regions, deicing salts are commonly

applied on roadways to prevent snow and ice accumulations and maintain safe driving conditions

during the winter season. Although the chloride-based salts (e.g., sodium chloride, calcium

chloride and magnesium chloride) are effective in deicing road surfaces, they also release a

substantial amount of chloride ions (Cl -) to the surrounding environment, including the external

surfaces of RC bridge structures (Lindvall, 2001, 2003; Tang & Utgenannt, 2007; Xi et al., 2018).

The released chloride ions are the major cause of rebar corrosion in RC bridges.

Steel reinforcement in RC members is typically protected from corrosion by a passive

oxide layer that forms on the rebar surface due to the high alkalinity of the concrete pore solution.

However, this passive layer breaks down when a critical content (Ccrit) of chloride ions (Cl-)

diffuses through the concrete cover and accumulates at the rebar surface (Bertolini et al., 2004).

According to Tuutti’s classical model (Tuutti, 1982), the time needed for this process to occur is

called the time-to-corrosion initiation.

In the past several decades, corrosion of RC highway bridges caused by deicing salts has

been investigated by many researchers. Depending on the research approach, these studies can be

broadly classified into two categories: studies based on field tests (Andersen, 1997; Keserle et al.,
86
2021; Lindvall, 2001, 2003; Prah-Ennin, 2013; Tang & Utgenannt, 2007; Weyers et al., 1994; Xi

et al., 2018) and studies based on analytical or numerical simulations (Conciatori et al., 2018;

Kirkpatrick et al., 2002; Petcherdchoo, 2017; Stewart & Rosowsky, 1998; Wolofsky, 2011). Field

tests commonly involve collecting several samples of concrete from different components (e.g.,

decks and columns) of existing bridges, followed by laboratory determination of chloride content

in these samples as a function of cross-section depth (chloride profiles). Some studies also perform

statistical regression on large datasets obtained from the field tests to develop probability density

functions for use in numerical analyses. The analytical or numerical simulations typically estimate

the chloride content in RC bridges by modeling the chloride transport in concrete using physics-

based or finite element methods.

Majority of the existing studies (Andersen, 1997; Kirkpatrick et al., 2002; Lindvall, 2001,

2003; Tang & Utgenannt, 2007; Weyers et al., 1994; Xi et al., 2018) show that corrosion in RC

bridges is highly affected by the application of deicing salts and the traffic patterns. Although the

existing studies on corrosion of RC members provide useful insights into chloride penetration and

corrosion in bridges caused by the deicing salts, there are several limitations or assumptions in

their analyses. First, the amount of deicing salts used was not incorporated in most of the studies.

Using larger amounts of deicing salts will likely increase the surface chloride concentration and

the risk of corrosion.

Second, none of the previous studies considered the chloride ion transport mechanisms

governed by vehicle spray and splash. Deicing salts applied on the road surface can be transported

to the surrounding environment by many mechanisms such as runoff, splash, spray and plowing

87
(Blomqvist & Johansson, 1999). For the piers of bridges crossing another roadway, vehicle spray

and splash are two major transport mechanisms (Lindvall, 2001).

Third, the majority of the existing models cannot be used to distinguish between bridges in

similar climates but with different traffic patterns (e.g., interstate versus less-used rural highway),

between bridges crossing different features (e.g., crossing a highway versus crossing a river) or

between bridge elements with different proximities to roadways (e.g., bridge column next to a road

versus away from a road). The above limitations must be addressed for accurate estimation of

chloride exposure of highway bridges and to determine time-to-corrosion initiation.

A systematic method addressing the aforementioned limitations and specifically

incorporating the vehicle spray and splash mechanisms in the estimation of chloride exposure of

RC highway bridges is proposed in this Chapter. In this method, first, the amount of deicing salts

applied on the road surface and the water film thickness on the road surface are determined. Then,

an analytical model based on computational fluid dynamics analysis (Flintsch et al., 2014) is used

to determine the amount of water and dissolved chloride ions sprayed and splashed in the

environment by a heavy-duty vehicle. The traffic data is used to compute the water amount and

dissolved chloride ions for all the vehicles near a bridge. Finally, the chloride ions transferred from

the environment to the exterior surface of the bridge substructure are calculated using an empirical

model (Blomqvist, 2001; Lundmark & Jansson, 2008). This provides the surface chloride content

(𝐶 ), which can be further used to determine the time to corrosion initiation based on the transport

properties of concrete and the cover thickness of an RC bridge member.

88
 These steps of the proposed method are described in detail (with mathematical equations)

in the next section, followed by a comparison with observed 𝐶 in existing RC bridges. Using the

validated model, the effects of water thickness on the road, vehicle speed, traffic volume, the

number of heavy-duty vehicles, distance of the bridge substructure to roadside, and the total

amount of deicing salts applied during the winter season on the estimation of 𝐶 were investigated.

5.2 The proposed method

The deposition of chloride ions on the external surface of RC highway bridges due to the

application of deicing salts occurs in three stages. In the first stage, deicing salts are applied to the

road surface to melt snow and ice. Next, the melting snow and ice lead to the formation of a film

of water containing chloride ions (salt solution) on the road surface. Finally, the passing vehicles

spray and splash the salt solution from the road surface to the exterior surfaces of nearby RC

highway bridge members. All the above stages are captured in the proposed model.

5.2.1 Deicing salts application rate

The amount of deicing salts applied on a roadway per day during the winter season is

calculated by Equation (27). In this equation, 𝑀 is the mass of deicing salts applied per unit

roadway area per day (kg/m2/day), 𝑉 is the daily deicing salt application rate (kg/lane

per unit length/day), 𝑁 is the number of road lanes in one traffic direction where deicing salts

are applied, and 𝑊 is the road width (m) where deicing salts are applied.

×
𝑀 = (27)

89
 Typically, 𝑉 is influenced by traffic, road conditions and weather (NYSDOT,

2006). The annual average daily traffic (ADT), rush hour peak traffic, and overall traffic pattern

on a given day affect the efficiency of the plow/salt trucks. For example, during rush hour, the rate

of salt application is limited by traffic congestion causing a reduction in 𝑉 .

𝑉 can be determined from highway maintenance guidelines of departments of

transportation (DOT) (NYSDOT, 2006).

In some countries (e.g., Sweden (Lindvall (2003)), instead of using deicing salts

application rate (𝑉 ), DOTs use the total amount of deicing salts (𝑀 ) applied on

the road surface during the winter season (kg/m 2/winter) to record the deicing operations in an area.

For this case, the amount of deicing salts applied per day during the winter season is calculated

using Equation (28):

𝑀 = (28)

where 𝑡 is the number of days with snow during the winter season, which is obtained from

historical weather data of the region.

5.2.2 Water film thickness

The daily water film thickness on the road (ℎ ) is calculated using Equation (29). In this

equation, ℎ is the total snowfall during a winter season (m), and 𝑡 is the number of days

with snow during a winter season. The parameters ℎ and 𝑡 are determined from the

weather history of the bridge location. The key assumption underlying Equation (29) is that all the

90
fallen snow is melted by the deicing salts, which in turn gets converted into the water film on the

road, and becomes available for vehicle spray and splash. This assumption may be invalid in some

scenarios, as some amount of snow is lost through displacement by vehicles, runoff, ground

percolation, evaporation, plowing and other mechanisms.

ℎ = (29)

The ambient temperature and the environmental conditions such as humidity at the time of

salt application can also affect the water film thickness on the road. Sometimes, the deicing salts

are applied in advance of the snow or during freezing rain to avoid slippery conditions.

Additionally, due to the different densities between snow and water, the melted water has lesser

volume compared to the snow. However, estimation of water film thickness considering all of the

above and other factors is extremely challenging and there is no existing mathematical model or

observations for modeling it accurately. Therefore, the proposed model assumed that all the

deicing salts were applied during the snow days. Accurate estimation of the water thickness on the

road considering relevant factors is one of the opportunities to improve the current model.

Furthermore, the time over which the snow melts could be different from the time assumed

above (𝑡 ), as influenced by environmental temperature and traffic volume (Lysbakken, 2013).

Sudden spike in temperatures and large traffic volume could accelerate the snow melting process.

However, in our study, snow on a road is assumed to be cleared through melting on the day of

precipitation. These assumptions were made due to the absence of data or a mathematical model

that captures the above complex mechanisms. To understand the significance of the assumptions,

91
the authors investigated the effects of the parameter happ on the model results, as discussed in

Section 5.5.

5.2.3 Water sprayed and splashed by one heavy-duty vehicle

Vehicle spray and splash refers to the displacement of water from the road surface to the

surrounding environment by the interaction of vehicle tires with the road surface. According to

Weir et al. (1978), there are four primary mechanisms of vehicle spray and splash: Capillary

adhesion, tread pickup, bow wave, and side wave, which are illustrated in Figure 5.1.

Figure 5.1: Mechanisms of vehicle spray and splash (adapted from Weir et al. (1978))

Capillary adhesion refers to the absorption of water (present on the road surface) by the

tires through surface tension. During tire rotation, water detaches from the tire surface due to the

centrifugal effect and gets sprayed and splashed to the environment. Tread pickup refers to water

92
within the grooves of a tire being sprayed and splashed behind the tire by turbulent flow in the

grooves. The bow and side waves are caused by the physical displacement of water from the road

surface due to the vehicle tires. While the side wave sends the water in the direction perpendicular

to the traffic, the bow wave sends the water towards the front of the tire.

Based on the above understanding of the mechanisms of vehicle spray and splash, Flintsch

et al. (2014) developed a numerical model using computational fluid dynamics simulations to

quantify vehicle spray and splash. The focus of Flintsch et al. (2014) was to understand the effect

of spray and splash on road visibility. Our study used the model by Flintsch et al. (2014) to estimate

the salt transport from the road surface to RC bridges by incorporating the amount of deicing salts

(𝑀 ) and water thickness (ℎ ) calculated in Sections 5.2.1 and 5.2.2, as described below.

The mass flow rate (𝑀𝑅 ) in kg/s is the amount of water displaced by a single tire due to

mechanism i (capillary adhesion, tread pickup, bow wave, of side wave ). Equations (30), (31) and

(32) (proposed by Flintsch et al. (2014)) model the mass flow rates (𝑀𝑅 ) for different vehicle

spray and splash mechanisms. In these equations, 𝑉 is the heavy-duty vehicle speed (km/h),

b is the tire width (m), 𝑘 is the ratio of the tire width that is not a groove to the tire width, ℎ is

the depth (m) of the water film picked up in each rotation, ℎ is the water film thickness (m/day)

on the road determined from Equation (29), and 𝜌 is the density of water (kg/m3).

93
Capillary adhesion (CA):

𝑀𝑅 =𝑉 ×𝑏×𝑘×ℎ ×𝜌 (30)

Tread pickup (TP):

𝑀𝑅 = 𝑉 × 𝑏 × (1 − 𝑘) × ℎ ×𝜌 (31)

Bow and side waves (BW or SW):

𝑀𝑅 = 𝑀𝑅 = 0.5 × 𝑉 ×𝑏× ℎ −𝑘×ℎ − (1 − 𝑘) × ℎ ×𝜌 (32)

Using the mass flow rates (𝑀𝑅 ) for various spray and splash mechanisms as input, Flintsch

et al. (2014) conducted several computational fluid dynamics simulations of vehicle spray and

splash to compute the concentration of water kicked up to the environment by each passing heavy-

duty vehicle (a dump truck). A heavy-duty vehicle was used to understand the most severe effect

on road visibility (Flintsch et al., 2014). In the simulations, the highway environment and the

heavy-duty vehicle were modeled within a rectangular space that was meshed with approximately

4.2 million tetrahedral cells. The modeled truck was held stationary, and the mass flow rates (𝑀𝑅 )

calculated by equations (30) to (32) were assigned to the mesh nodes on the surface of the truck’s

tires.

An airflow was introduced in the meshed domain around the truck from front to back,

simulating the turbulent airstream formed by a passing vehicle. The airflow was modeled as an

incompressible flow with the same speed as that of the truck. While prevailing winds might affect

the trajectories of the sprayed and splashed water droplets, it was assumed to be negligible

compared to the airflow created by a passing vehicle (Flintsch et al., 2014). The ground was

modeled as a wall with a translational velocity in the same direction as and equal to the air velocity.

94
Water droplets were released into the airflow model, and their trajectories were calculated within

the turbulent airstream formed by the truck.

The water concentration in the environment was called “spray density” (𝑆𝐷 ) in the ith

mechanism by Flintsch et al. (2014). For each spray and splash mechanism, the water

concentrations or spray densities (𝑆𝐷 ), normalized by the respective 𝑀𝑅 , in the space behind the

truck were calculated for different vehicle speeds (𝑉 ). Flintsch et al. (2014) applied regression

analysis to the simulation results to develop relationships between 𝑆𝐷 , 𝑀𝑅 , and vehicle speed,

𝑉 for each spray and splash mechanism as shown in Equations (33) to (36) below. Note that

Equations (33) to (36) were obtained by regression analyses and therefore are not dimensionally

homogeneous. For these equations, the unit for 𝑆𝐷 is in kg/m3/vehicle (kg of water in 1 m3 volume

of air) per heavy-duty vehicle. The unit for 𝑉 is miles/hour and the units for 𝑀𝑅 is kg/s.

Capillary adhesion:

𝑆𝐷 = (−2.69 × 10 × 𝑉 + 2.43 × 10 ) × 𝑀𝑅 (33)

Tread pickup:

𝑆𝐷 = (1.16 × 10 × 𝑉 − 5.25 × 10 ) × 𝑀𝑅 (34)

Bow wave:

𝑆𝐷 = (2.67 × 10 × 𝑉 − 4.71 × 10 ) × 𝑀𝑅 (35)

Side wave:

𝑆𝐷 = (2.67 × 10 × 𝑉 − 4.71 × 10 ) × 𝑀𝑅 (36)

95
Therefore, the mass of water per unit air volume (𝑆𝐷 ) kicked up by each passing heavy-duty

vehicle is:

𝑆𝐷 = 𝑆𝐷 + 𝑆𝐷 + 𝑆𝐷 + 𝑆𝐷 (37)

5.2.4 Chloride ions sprayed and splashed by one heavy-duty vehicle

Next, the mass of water per unit air volume due to heavy-duty vehicles per day, 𝑆𝐷 , is

converted into the mass of salt per unit air volume (𝑆𝐷 _ ) kicked up along with the water

using Equations (38) and (39).

𝑆𝐷 _ = 𝑆𝐷 ×𝛿 (38)

Where, 𝛿 = = (39)
×

The most commonly used salt for deicing operations in the US is sodium chloride (NaCl).

The molar mass ratio (𝜃 ) of chlorine (corrosive agent) to NaCl (deicing salt) is 0.61

(35.45g/58.44g). Therefore, the mass of chloride ions per unit air volume (𝑆𝐷 _ ) kicked up

by each heavy-duty vehicle is calculated using Equation (40). Other types of deicing agents can

be considered in the model by adjusting the molar mass ratio in this equation.

𝑆𝐷 _ = 𝑆𝐷 _ ×𝜃 (40)

5.2.5 Chloride ions sprayed and splashed by all vehicles in one winter season

In the previous sections, chloride ions were computed for heavy-duty vehicles. In this

section, chloride ions were calculated for all vehicles considering the types of vehicles, traffic

96
volume, and the number of snow days. Vehicles are classified into two categories: heavy-duty

vehicles (trucks) and light-duty vehicles. The ratio (∅) of chloride ions sprayed and splashed by

heavy-duty vehicles to light-duty vehicles was assumed to be 6 based on the study by Denby et al.

(Denby et al., 2013). The cumulative mass of chloride ions per unit air volume (𝐶 _ ) sprayed

and splashed by all the vehicles passing near a bridge component is calculated by Equation (41).

The average daily traffic (ADT) is used to account for the spray and splash of all the

vehicles on the road lane nearest to a bridge component, whose corrosion exposure is under

consideration. This is the reason why Equation (41) has number of lanes (Nlane) in its denominator.

In this study, the bridge components of interest are piers of road-crossing bridges. ‘ADTT’ can be

calculated as a product of Θ (the fractional number of heavy-duty vehicles out of all vehicles) and

ADT. If ‘ADTT’ is the average number of heavy-duty vehicles per day, then subtracting ADTT

from ADT leads to the number of light-duty vehicles per day. Computing the collective spray and

splash of all the vehicles by effectively multiplying the spray and splash of one vehicle by the

number of vehicles ignores the interaction effects of multiple vehicles on the road simultaneously,

which can potentially reduce the amount of salt sprayed and splashed on the bridge substructure.

Accounting for all the days with snow (𝑡 ) in a typical winter season, the cumulative

mass of chloride ions per unit air volume (𝐶 _ ) sprayed and splashed by all the vehicles passing

near the bridge pier every winter can be expressed as shown below.

𝐶 _ = (𝑆𝐷 _ × × + 𝑆𝐷 _ × )×𝑡 (41)

∅

97
5.2.6 Chloride ions deposition on bridge substructure

Finally, the deposition of deicing salts on the surface of the bridge substructure is computed

using the empirical Equations (42), (43), and (44) (proposed in Blomqvist (2001); Lundmark &

Jansson (2008)), which take into account the distance between the edge of the road near the bridge

substructure and the bridge substructure. Multiple studies have reported that the deposition of

deicing salts decreases exponentially with distance (Blomqvist, 2001; Blomqvist & Johansson,

1999; Denby et al., 2013; Lundmark & Olofsson, 2007) . In these dimensionally nonhomogeneous

equations, 𝐶 is the chloride ion concentration at the surface of the bridge substructure in kg/m 3

units, 𝐶 _ is the chloride ions sprayed by vehicles (Lundmark & Olofsson, 2007) in kg/m 3

units, 𝐶 _ is the chloride ions splashed by vehicles (Lundmark & Olofsson, 2007) in kg/m 3

units, d is the distance between the road edge and nearby bridge structure in meter units and e is

the base of natural logarithm.

. .
𝐶 = 𝐶 _ ×𝑒 +𝐶 _ ×𝑒 (42)

𝐶 _ = 𝐶 _ × 0.015 (43)

𝐶 _ = 𝐶 _ × 0.985 (44)

In Equation (42), the exponential coefficients of 0.05 and 0.5 were adapted from

Blomqvist (2001), who investigated the deposition of deicing salts in two sites in Sweden. These

coefficients were further verified by Lundmark & Olofsson (2007). The proportions of chloride

ions sprayed (0.015 or 1.5%) and splashed (0.985 or 98.5%) were adapted from Lundmark &

Olofsson (2007). In their study, they measured the deposition of deicing salts along a highway in

Sweden, used Equation (43) to define the relation between the mass of deicing salts per unit area

98
(collected from the sites) and distance from roadside, and used nonlinear fitting techniques to

determine the proportions of sprayed and splashed chloride ions. It should be noted that these

proportions can change with the site characteristics of a bridge, such as the roadside environment

(forested or urban), traffic characteristics (direction and volume), wind direction, and road surface

condition. However, there are no other studies reporting chloride ion deposition data considering

site characteristics. Thus, Equations (42) to (44) were used as such in demonstrating the proposed

method in this study, and modifications can be made in the future as site specific data becomes

available.

5.2.7 Assumptions

As described earlier, there are some key assumptions used in the analytical model presented

in this Chapter, which can be removed upon appropriate data collection. These assumptions are

summarized as follows:

1) Chloride ions are released from deicing salts after dissolution in water (from melting

snow). However, only a fraction of the applied deicing salts can be expected to dissolve

in water that is sprayed and splashed by the vehicles onto the bridge surface, and the

rest of the salt runs off. Due to the lack of data, all the applied deicing salts are assumed

to dissolve in the water on the road surface in this study. This assumption is expected

to result in underprediction of corrosion initiation time.

2) The chloride ions are assumed to be either washed out due to rain or blown away by

wind after each winter season, thus avoiding the buildup of chloride ions at the surface

99
 over multiple seasons. This assumption is supported by field observations (Lindvall,

2001). It should be noted that although there is no accumulation of the surface chloride

over multiple years, the interior chloride concentration does increase over time. The

chloride ions are sprayed and splashed by the vehicles to the RC column surface. As

the concentration of chloride ions at the external surface of concrete is larger than that

inside the concrete, the diffusion of chloride ions occurs.

3) For bridge substructures subjected to salt-induced corrosion, the most corroded part of

a bridge pier is typically located at the base (Lindvall, 2001). In addition, the base of a

pier typically has the highest structural demand across pier height. Therefore, while the

salt accumulation on the surface of a bridge pier due to vehicle spray and splash varies

with height above the ground, this variation was neglected, and only the pier section at

the substructure base was considered in this study.

4) The computational fluid dynamics simulation results used in this study is based on a

single heavy-duty vehicle causing the spray and splash. As many vehicles pass by a RC

bridge substructure, the amount of sprayed and splashed chloride ions from a single

vehicle may be affected by other vehicles. The proposed method calculates the 𝐶 _

by multiplying the traffic volume (ADT) in Equation (41) with 𝑆𝐷 _ . This

assumption may be an overestimation of 𝐶 . A better estimation may be provided by

performing computational fluids dynamics analyses with multiple vehicles in the future.

However, this is time-consuming and complex and there are no computational fluid

dynamics simulations available in the literature considering spray and splash by

multiply vehicles. Thus, this study used equations developed by Flintsch et al. (2014)

100
5.2.8 Application

As mentioned above, the proposed method is expected to provide a more accurate

determination of the surface chloride concentration (Cs), as it accounts for the unique

characteristics of each bridge. The 𝐶 is further used to determine the time to corrosion initiation

using Fick’s second law of diffusion, which is a complex partial differential equation. Crank’s

solution (Crank, 1979) to the Fick’s second law of diffusion given in Chapter 3 Equation (1) is

commonly used to solve for the unknown parameters, where 𝐶 is the chloride concentration at a

distance x from the exterior surface and at time t, 𝐷 is the diffusion coefficient of concrete, and

erf(.) is the error function. For the readers’ convenience, Equation (1) is listed again in this section.

Equating 𝐶 to Ccrit at x = cover depth; corrosion initiation time, t, can be obtained using Equation

(1). It should be noted that there are more advanced models for computing the corrosion initiation

time such as ClinConc model (Luping, 2008). However, the main focus of this study is to estimate

the surface chloride exposure for RC bridges in the cold region as opposed to estimating the

corrosion initiation time. Therefore, a 1-D chloride ion diffusion model was used in the

demonstration of the application.

𝐶 (𝑥, 𝑡) = 𝐶 1 − erf (1)

5.3 Demonstration of the Proposed Model with an Example

In this section, an RC bridge (Lindvall, 2001) located in Gothenburg, Sweden, is used as

an example to illustrate the proposed model. Detailed measurements of surface chloride

concentrations on the exterior surface of this bridge are available in the literature (Lindvall, 2001),

101
which facilitate comparative evaluation of the proposed method. The bridge has two spans and a

concrete deck supported by three RC bridge columns, exposed to a large amount of deicing salts

during the winter season. The values of all the parameters used in the proposed model are given in

Table 5.1, along with the data sources. The calculation procedure for surface chloride

concentration and time to corrosion initiation of the example bridge using the proposed method is

as follows:

Table 5.1: Model parameters for the piers of the example bridge

Parameter Value Source

𝑀 3.0 kg/m2 (Lindvall, 2001)
(Weather-Spark,
ℎ 0.645 m
2021)
(Weather-Spark,
𝑡 127 days/winter
2021)
𝜌 997 kg/m3 -
K 0.75 (Flintsch et al., 2014)
∅ 6 (Denby et al., 2013)
Spray and splash
ℎ 0.0001 m (Flintsch et al., 2014)
model
𝑏 0.56 m (Flintsch et al., 2014)
ADT 26,000 vehicles/day (Lindvall, 2001)
ADTT 2,704 vehicles/day -
𝑁 4 lanes (Lindvall, 2001)
𝑉 80 km/h -
𝑉′ 50 mile/hour -
𝑑 3.05 m (Lindvall, 2001)
Bridge pier Concrete cover
45 mm (Lindvall, 2001)
properties thickness
Corrosion initiation 𝐷 2.50 × 10-6 mm2/s (Life-365™, 2020)
model 𝐶 1.68 kg/m3 (Life-365™, 2020)

102
 Step 1: The 𝑀 was assumed equal to 3.0 kg/m2 given in Lindvall (2001). 𝑡 is

assumed to be 127 days based on the annual weather report for the bridge location (Gothenburg,

Sweden) (Weather-Spark, 2021). Using 𝑀 and 𝑡 in Equation (27), 𝑀 was calculated

as 0.0236 kg/m2 per day.

Step 2: Using ℎ of 0.645 mm (determined from the annual weather report for the

bridge location (Weather-Spark, 2021)) in Equation (29), the ℎ was calculated as 5.08 mm.

Step 3: The speed limit of heavy-duty vehicles on the road under the bridge is 80 km/h (50

mph), which was used as 𝑉 in Equations (30) to (32). The values of K, ℎ and b used in

Equations (30) to (32) are shown in Table 5.1, which were the characteristics of heavy-duty

vehicles used in the simulations by Flintsch et al. (2014). Equations (30) to (32) were used to

compute 𝑀𝑅 = 0.94 kg/s, 𝑀𝑅 = 15.88 kg/s, 𝑀𝑅 = 23.36 kg/s and 𝑀𝑅 = 23.36 kg/s.

Using 𝑀𝑅 calculated from above in Equation (33) to Equation (36), water concentrations

in air due to various vehicle spray and splash mechanisms were computed as 𝑆𝐷 = 0.0010

kg/m3, 𝑆𝐷 = 0.0084 kg/m3, 𝑆𝐷 = 0.0203 kg/m3, 𝑆𝐷 = 0.0097 kg/m3 and 𝑆𝐷 =

0.0394 kg/m3.

Step 4: Using Equation (39), 𝛿 was calculated as 0.0047, which in turn was used as

input in Equation (38) to compute 𝑆𝐷 _ = 0.0019 kg/m3. As mentioned in Section 5.2.4,

𝜃 is 0.61 for sodium chloride, which was used to obtain 𝑆𝐷 _ = 1.12 × 10 kg/m

using Equation (40).

103
 Step 5: Using the values of ADT, ADTT, d, ∅ , 𝑁 and 𝑡 shown in Table 5.1,

𝐶 _ = 23.39 kg/m3 was computed using Equation (41).

Step 6: The surface chloride content,𝐶 of 5.32 kg/m3 was computed using Equations (42)

to (44).

Step 7: Using 𝐶 , cover thickness, 𝐷 , and 𝐶 , the corrosion initiation time for the

example bridge calculated using Equation (1) as 12.8 years.

5.4 Comparison of model estimates with experimental results

5.4.1 Experimental data

The analytical model predictions were compared to experimental data from two studies: (1)

the study by Tang & Utgenannt (2007) that collected data from 34 cube-shaped specimens placed

along a highway, and (2) the study by Lindvall (2001) that collected data from a single bridge

column of four bridges. To the authors' best knowledge, these results were the only surface chloride

content profiles measured in the field and reported in the literature.

Tang & Utgenannt (2007) reported the surface chloride contents of 34 cube-shaped

concrete specimens after being placed along highway RV 40 in Sweden for 10 years. These

104
locations are typically exposed to a large amount of deicing salts during the winter season. These

specimens were denoted as TP in this Chapter.

Lindvall (2001) reported chloride content data for one RC highway bridge column in four

bridges exposed to deicing salts in Sweden. These bridges were built between 1968 and 1972.

Lindvall (2001) obtained concrete samples by drilling cores in bridge columns. For each bridge,

only one bridge column was drilled at one or more locations and led to one or more surface chloride

data points. The diameter of each core was 50 mm, and the core depth reached from the external

surface of bridge piers to the reinforcement surface (through the entire cover depth). Chloride

profiles (total chloride content versus depth from the surface) of the concrete samples were

determined by Lindvall (2001) using the experimental procedure described in AASHTO T260

(2009). By using the data provided by Lindvall (2001), we computed the surface chloride contents

(𝐶 ) of those bridges using ASTM C1556 (2016). These bridge columns are named as L1 to L4 in

this Chapter.

Chloride profiles for bridge L1, L2, L3 and L4 were determined at 2 locations, 4 locations,

4 locations and 1 location on a column, respectively. These locations are shown in Figure 5.2 for

bridges L1, L2 and L3. The location for the sample collected from bridge L4 was not reported. In

this figure, F locations face oncoming traffic and M locations face away from the oncoming traffic.

105
 Figure 5.2: Location of specimens for field tests for bridges L1, L2 and L3 (adapted from
Lindvall (2001))

5.4.2 Analytical model prediction

For comparing the proposed method for determining 𝐶 with the field tests (Lindvall, 2001;

Tang & Utgenannt, 2007), the parameters listed in Table 5.2 were used. Although the absolute

traffic volume of heavy-duty vehicles (ADTT) was not given by Lindvall (2001), the number of

heavy-duty vehicles as a fraction of total vehicles (Θ) was given in Tang & Utgenannt (2007) as

0.104. Considering that Tang & Utgenannt (2007) and Lindvall (2001) obtained their observations

from the highways in the same region (Gothenburg, Sweden), the same fraction was assumed for

simulating the cases in Lindvall (2001). Snowfall precipitation (ℎ ) and days with snow during

winter (𝑡 ) were determined from the past weather reports of Gothenburg, Sweden (Weather-

Spark, 2021) for simulating the environmental exposure in all the cases.

106
 Table 5.2: Input parameters for different field tests
ADT ADTT
Specimen/ 𝑉 𝑀 ℎ 𝑑
Study Bridge (vehicles/ (vehicles 𝑁 𝑡
name (km/h) (kg/m2) (m) (m)
day) /day)
Tang &
Utgenannt TP 80 1.54 12000 1250 0.645 0.70 2 127
(2007)
L1 80 2.80 27000 2808 0.645 3.00 4 127

L2 80 3.00 26000 2704 0.645 3.05 4 127

Lindvall
(2001)
L3 80 2.80 34000 3536 0.645 2.00 4 127

L4 40 2.30 2400 250 0.645 2.30 2 127

5.4.3 Model prediction versus test data

The values of 𝐶 obtained from the field specimens (Lindvall, 2001; Tang & Utgenannt,

2007) and computed using the proposed analytical model are compared in Figure 5.3. In this figure,

the ‘TP’ label is used to refer to the field 𝐶 values from Tang & Utgenannt (2007) (34 specimens

in total) and ‘L1’ to ‘L4’ labels are used for the bridges from Lindvall (2001). The averages of the

observed 𝐶 values are also marked on the figure. For bridges L1, L2 and L3, the ‘F’ position faces

the oncoming traffic, whereas the ‘M’ position faces the back of the traffic (Figure 5.2).

107
Figure 5.3: Comparison of 𝐶 from field observations (Lindvall, 2001; Tang & Utgenannt, 2007)
and model estimates

The chloride profiles of the field specimens taken from bridges L1, L2 and L3 varied

depending on the position of the sampled locations on the bridge columns relative to the traffic

direction, whereas the chloride profiles of the field specimens taken from bridge TP varied due to

difference in concrete quality, as explained below. As expected, the 𝐶 values observed at ‘F’

positions were greater than those at ‘M’ positions for bridges L1 and L3. However, for the bridge

L2, the largest 𝐶 was observed at ‘M2’ position because of the wind effects (that were neglected

in the analytical model calculations). For bridge L2, the wind effects were reported to have moved

the chloride ions accumulated on the ‘F2’ position to ‘M2’ position (Lindvall, 2001).

The difference in the Cs observations in the case TP was due to the difference in the concrete

quality (Tang & Utgenannt, 2007). Each concrete specimen (34 in total) used in TP was made with

different binders and water-cement ratios, which resulted in different chloride binding capacities.

The maximum 𝐶 observed in this case was from the specimen with the highest volume of blast-

108
furnace slag, which has a strong chloride binding capacity (Tang & Utgenannt, 2007). From Figure

5.3, it can be observed that for TP specimens, the 𝐶 predicted by the proposed model was close

to the maximum observed value (within 8%). The predicted 𝐶 was 51% higher than the average

observed value. The analytical model does not capture the impact of concrete quality on surface

chloride content due to the absence of data. Concrete moisture conditions and concrete resistance

to chloride ion ingression can also affect the estimation of surface chloride content. However,

limited data and models have been published on analyzing this complex mechanism. Concrete

quality, including chloride binding capacities, moisture condition and chloride ion diffusivity, can

be considered in future studies for improving the current model by collecting additional field data

and performing laboratory experiments.

For bridges L1, L2 and L3, the model estimates of 𝐶 were close to the values observed in

the field specimens taken from the ‘F’ positions (the ‘M’ position for L2), where the corrosion is

most likely to initiate first. The largest difference between the observed value at an ‘F’ position

and the model estimate was 1.71 kg/m3 (approximately 13% of the observed value).

The estimated 𝐶 for the bridge L4 using the proposed method was significantly less than

the only observed value for a few possible reasons. First, unlike L1, L2 and L3, there was only one

observation of Cs available for L4, which does not capture the inherent variability among

specimens. Second, the L4 bridge is located on a local road with a low vehicle speed limit of 30 to

40 km/h, whereas the proposed model utilizes Equations (33) to (36) that were developed based

on regression analysis on data with vehicle speeds ranging from 50 km/h to 96 km/h. Therefore,

the regression model used in the model might not be applicable to speeds outside this range. The

109
vehicle speed has a significant effect on the amount of water and chloride ions that can be sprayed

and splashed in the air as shown in Equations (33) to (36). This shows that there are limits to the

applicability of the proposed model, but such limitations can be removed through appropriate data

collection and corresponding updates to the model. Examples of valuable data include data for

local roads as opposed to major highways and a greater number of samples instead of only one.

Despite the differences between the observed and computed estimates of 𝐶 , the proposed

method can differentiate between the given specimens or bridges with different surface chloride

contents. For example, L3 and L4 bridges have the highest and the lowest surface chloride contents,

respectively, which is consistent with the model estimates. This information can be critical in

identifying bridges that should be prioritized for maintenance.

5.5 Effects of parameters on the surface chloride content

The effects of the following parameters on the surface chloride content ( 𝐶 ) were

investigated by varying these parameters within reasonable ranges: average daily water thickness

during a winter season (ℎ ), vehicle speed (𝑉 ), traffic volume (ADT), fraction of heavy-

duty vehicles as a portion of total vehicles (Θ), distance of the bridge substructure to roadside (𝑑),

and total deicing salts applied per winter (𝑀 ). Figure 5.4 shows 𝐶 as a function of these

variables for the example bridge substructure discussed in Section 5.3, which was the bridge

labeled as ‘L2’ in Section 5.4.

110
 Figure 5.4 (a) shows that ℎ has negligible effects on 𝐶 . Although a thicker water film

on the road surface would lead to a greater amount of water sprayed and splashed by passing

vehicles, the larger amount of water also reduces the concentration of deicing salts dissolved in

the melting snow, thereby nullifying the effect on 𝐶 .

Figure 5.4 (b) shows that 𝐶 increases with an increase in vehicle speed (𝑉 ). This is

because, as shown in Equations (33) to (36), the “spray density” (𝑆𝐷 ) has a positive correlation

with vehicle speed. Similar trends are observed for the variations of 𝐶 with 𝐴𝐷𝑇 and Θ, as shown

in Figure 5.4 (c) and (d), respectively. This is because 𝐶 is linearly dependent on ADT and Θ, as

shown in Equation (41).

Figure 5.4 (e) shows the variation of 𝐶 with 𝑑. Vehicle spray and splash as well as surface

chloride content decrease with increase in the physical distance (d) between the bridge substructure

and the nearest road lane, as shown in Equation (42). Figure 5.4 (f) shows that 𝐶 increases with

an increase in 𝑀 . This is because 𝑀 is used to determine the ratio of amount of deicing

salts on the road to the amount of water on the road as shown in Equation (39). The more deicing

salts applied per day results in the larger ratio ( 𝛿 ) , which further increase 𝜃 and

𝑆𝐷 _ as shown in Equation (40).

111
 (a) (b)

(c) (d)

(e) (f)

Figure 5.4: Effects of parameters on surface chloride content

112
5.6 Conclusions

A systematic method for incorporating the vehicle spray and splash mechanisms in the

estimation of chloride exposure of highway RC bridge piers is proposed in this chapter. This

method systematically accounts for the salts applied and snow precipitation to determine the water

film thickness on the road surface. Then, a regression model based on the results of computational

fluid dynamics analyses (Flintsch et al., 2014) is utilized to determine the amount of water and

dissolved chloride ions kicked up to the environment around a bridge by the passing vehicles

through various spray and splash mechanisms. Finally, the chloride ions transferred from the

environment to the exterior surface of a bridge substructure are calculated using an empirical

model. Thus, the proposed method allows evaluation of chloride exposure (and thereby corrosion

susceptibility) of bridges individually based on the unique characteristics of the crossed features,

snow precipitation, salt application, and traffic patterns, which is not possible with the existing

methods. The method can also be used to estimate corrosion initiation time considering local

conditions of an exposed bridge element. Application of the model was demonstrated on an

example bridge.

Concrete specimens placed on the side of a highway (denoted as TP) and RC bridge

columns (denoted as L1 to L4) with experimentally determined chloride profiles from the literature

were used to evaluate the predictions of the model proposed in this study. For specimens TP and

bridges L1, L2 and L3, the surface chloride content predicted by the proposed model was close to

the maximum observed value. For two (L1 and L3) out of the four bridges, the surface chloride

content values computed with the proposed method were close to the values observed in the field

specimens taken from the ‘F’ positions (facing the oncoming traffic). The maximum vehicle spray

113
and splash occur at these locations, and therefore, corrosion is most likely to initiate at these

positions. For the bridge L2, the largest surface chloride content was observed at the ‘M’ position

(facing away from the oncoming traffic) because of the reported wind effects that were neglected

in the model calculations. For bridge L4, the estimated surface chloride content using the proposed

method was significantly less than the only observed value, likely due to different speed limits and

traffic volume compared to the conditions assumed to derive the model equations and the small

number of data samples.

Overall, despite some discrepancies, the proposed method seems to satisfactorily

incorporate the vehicle spray and splash mechanisms in the estimation of chloride exposure on

highway RC bridges. For the most part, the predictions of the surface chloride content are closer

to the upper bound of field data that exhibit scatter due to variations in concrete quality (TP

specimens) and location of surface chloride content measurements (L1 to L4 bridge columns). As

described in the chapter, there are some limitations of the model, which can be addressed with

further data collection and recalibration of the empirical equations. Additionally, some amount of

snow is lost through displacement by vehicles, runoff, ground percolation, evaporation, plowing

and other mechanisms, which were not included in the current model due to a lack of reported

studies. Estimating the chloride exposure of other bridge components such as decks and curbs can

be considered as an extension of the current model. These mechanisms are the opportunities for

improving the current model in the future.

The effects of various model parameters, including water thickness on the road surface,

vehicle speed, traffic volume, the number of heavy-duty vehicles, distance of the bridge

114
substructure to roadside and the total amount of deicing salts applied during a winter season on

surface chloride content, were investigated. The results show that water film thickness on the road

surface has a negligible effect on the surface chloride content of nearby bridge elements. The

surface chloride content decreases with the increase in the distance between the nearby bridge

substructure and the roadside. The surface chloride content of nearby bridge elements increases

with increases in vehicle speed, traffic volume, the number of heavy-duty vehicles on roads and

the total amount of deicing salts applied during winter.

115
 Chapter 6: Linking the Physics-Based Deterioration Model to Field-Based

Condition Assessments

6.1 Introduction

The goal of infrastructure asset management is to preserve and manage the infrastructure

sustainably, while ensuring safety and functionality during its service life (AASHTO, 2013;

Dominguez, 2019). For reinforced concrete (RC) bridges, deterioration mechanisms such as

corrosion can gradually affect safety and functionality. If left unaddressed, managing corrosion

can be economically unsustainable. An effective asset management program must therefore be

able to predict deterioration and optimize the timing of repair activities for maximizing safety and

functionality under given budget and time constraints.

Field inspections are typically used to monitor the condition of bridges, and the resulting

condition ratings are used as input in asset management tools and software for determining the

repair priority of bridges, considering budget and time limitations. However, field inspections are

labor intensive and can be subjective. Furthermore, visual inspections rely on visible signs of

corrosion on the exterior surface of a bridge (e.g., corrosion stains and cracks), and may miss

severe cases of localized deterioration until dangerously severe reinforcement corrosion occurs

(Moore et al., 2001). To address these challenges, a physics-based method is needed for

supplementing field inspections improving the reliability of condition assessment and management

of RC bridges.

116
 Several researchers have developed models for estimating the condition of RC bridges

using field-based condition assessments (Agrawal et al., 2008, 2010; Morcous, 2011) or physics-

based deterioration models (Stewart, 2004; Val & Melchers, 1997; Vidal et al., 2004; Vu & Stewart,

2000; Wang et al., 2020). Field-based condition assessment studies commonly develop

deterioration models for bridges by collecting condition rating data from periodic bridge

inspections and by processing this data using deterministic or probabilistic regression/stochastic

methods. These deterioration models are referred to as “deterioration curves”. Since corrosion is

the main deterioration mechanism in RC bridges, studies based on physics-based deterioration

models typically utilize corrosion models that are developed using fundamental equations of

diffusion, supported by experimental data, and/or numerical simulations. The physics-based

corrosion models estimate deterioration indicators, such as the rebar area loss, concrete crack width,

concrete spalling, and the stiffness reduction of bridges as a function of time.

The existing studies (Agrawal et al., 2008, 2010; Bu et al., 2014; Morcous, 2011; Stewart,

2004; Val & Melchers, 1997; Vidal et al., 2004; Vu & Stewart, 2000; Wang et al., 2020) on

condition assessment of RC bridges have several limitations or rely on several assumptions. Field-

based condition ratings used to generate deterioration curves may have significant variability. This

is partly due to the subjective nature of visual inspections (Moore et al., 2001). Also, at a given

bridge age, there is a wide scatter in the condition ratings due to large differences in the progression

of deterioration in various bridges. These differences are caused by the unique characteristics of

bridges in terms of their location, environmental exposure, traffic, etc. Therefore, the resulting

average deterioration curve can be significantly different from the actual condition rating of an

individual bridge at a given bridge age.

117
 The physics-based models require input parameters, such as corrosion rate, relation

between crack width and corrosion rate, which are difficult to measure in the field. These models

estimate bridge deterioration in terms of indicators such as reinforcement area loss, concrete crack

width, and concrete spalling as a function of time. However, the asset management systems used

by various state departments of transportation (DOT) in the US rely on condition ratings from field

inspections. A quantitative link between the indicators from the physics-based corrosion model

and the field condition ratings is missing. Such a link is needed for an objective assessment of

structural health and for prioritizing maintenance and repair actions, given limited resources.

A systematic method linking the physics-based corrosion model to bridge condition ratings

is proposed in this chapter to address some of the aforementioned limitations of existing models.

The method consists of two parts: 1) a corrosion model, and 2) linking the corrosion indicators to

bridge condition ratings. In the subsequent sections of this chapter, an extension of the corrosion

model introduced in Chapter 3 to capture spalling is presented. This extension is needed as spalling

is used as a criterion in bridge condition rating. Next, bridge condition rating systems and

deterioration curves currently used by transportation agencies are reviewed. Then, the method to

link the corrosion model output (crack width and spalling) to the bridge deterioration curves is

introduced. The application of the method is demonstrated using an example bridge. Finally, a

parametric study is presented to identify the influence of critical parameters on the proposed

method.

118
6.2 Corrosion modeling for predicting concrete spalling

The methods for estimating the corrosion initiation time as well as modeling crack width

as a function of time and corrosion rate are given in Chapter 3. The focus of this section is to

extend the corrosion model discussed in Chapter 3 to predict concrete spalling as concrete spalling

is considered in condition rating of bridges.

6.2.1 Simplified concrete spalling geometry

Many studies (Moccia et al., 2021; Su & Zhang, 2019; Zhang et al., 2021) assumed a

simplified triangular concrete spalling area on a cross-section view as shown in Figure 6.1. In this

triangular spalling model, three cracks are assumed to form in the concrete cover. The angle

between one of the outside cracks and the exterior surface of concrete cover is denoted by 𝜃 .

Figure 6.1: Simplified concrete spalling model shown in cross-section view

119
6.2.2 Pressure for concrete spalling

Moccia et al. (2021) performed several experiments to investigate the relationship between

the internal pressure caused by rust expansion and concrete spalling. In their experiments, a

hydraulic device was inserted into a circular hole created in a concrete sample to apply radial

pressure, simulating rust expansion. When concrete spalling was observed during the test, the

corresponding pressure was recorded. By analyzing the test data and following the simplified

model described in Section 6.6.1, a semi-empirical model considering the tensile strength of

concrete, cover thickness and rebar size was developed to calculate the pressure for concrete

spalling. The spalling pressure is calculated by Equation (45) using this model:

/
× ∅
𝑝 = ×2 + × (45)
( ) ∅ ∅ ∅

In the above equation, 𝜂 is the factor to account for the concrete brittleness in tension, 𝑓 is the

concrete tensile strength (MPa), ∅ is the reference rebar size which is 20 mm and 𝑑 is the

reference aggregate size which is 32 mm (Moccia et al., 2021), ∅ is the rebar size (mm), 𝑑

is the average roughness (mm), which is related to the maximum aggregate size (mm), 𝑐 is the

concrete cover thickness (mm) and 𝑚 is the empirical coefficient from regression. This semi-

empirical study used Imperial Units.

As mentioned in Chapter 3, over time, the build-up of rust increases tensile stress in the

concrete surrounding the rebar. When this stress reaches the tensile strength of concrete, a crack

is formed in the concrete cover. Rebar mass loss is predicted by Equation (11) shown below (also

defined in Chapter 3). In this equation, 𝑚 is the percent rebar mass loss as a function of time,

120
and 𝑄 is the percent area loss of rebar as a function of time. Lu et al.(2011) proposed an

equation for internal radial pressure (𝑃 ) due to rust expansion, given by Equation (46). Here,

𝛿 is the thickness of the porous zone around the rebar, 𝑛 is the volume expansion coefficient (due

to greater volume of corrosion products compared to the reactants), ∅ is the rebar diameter,

𝐸 is the elastic modulus of concrete, 𝐸 is the elastic modulus of rust, 𝑣 and 𝜈 are the Poisson’s

ratios of concrete and rust, respectively. Assuming that all the corrosion products remain inside,

the internal radial pressure, corresponding to the rebar mass loss computed by Equation (11), can

be found using Equations (46) (Lu et al., 2011).

𝑚 =𝑄 × 𝑢𝑛𝑖𝑡 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 𝑜𝑓 𝑠𝑒𝑐𝑡𝑖𝑜𝑛 (1 𝑚𝑚) (11)

( ) ⁄ /∅
𝑃 = ( ) ⁄ ( ) ( ) /
( ) [( ) ]( ⁄ ) (46)
∅
𝑤ℎ𝑒𝑟𝑒 𝑟 = +𝛿

The time at which concrete spalls is determined by checking whether the radial pressure

due to rust expansion calculated by Equation (46) reaches the concrete spalling pressure

determined by Equation (45). Later in this chapter, the concrete spalling pressure is calibrated by

changing the parameter 𝜃 in Equation (45) as will be discussed in Section 6.6.

It should be noted that the pressure due to rust expansion shown in Equation (46) presented

in this Chapter is not the same as the one presented in Chapter 3. Equation (46) is preferred in this

Chapter because it considers the deformation of rust layer, unlike Equation (12) presented in

Chapter 3. The model presented in Chapter 3 is intended for predicting the surface cracking time,

until which the rust layer is thin and has negligible effects on the stress distribution around the

121
rebar. However, at the time of spalling, the rust layer is thicker. Based on previous studies, when

rust layer becomes thick (between crack initiation time and time of concrete spalling), the rust

layer will be restrained and compressed by the surrounding concrete (Caré et al., 2008; Liu & Su,

2018). and therefore, the deformation of the rust layer should be considered. Additional details of

the rust expansion function can be found in Lu et al. (2011).

6.3 Bridge inspection program and condition rating

The bridge inspection program in the US was established in 1968 in the aftermath of the

collapse of the 2,235-foot Silver Bridge, at Point Pleasant, West Virginia. After the failure of the

bridge, the USDOT decided to establish a national bridge inspection standard (Thomas et al., 2012)

to keep bridges safe and serviceable. After several decades of development, the bridge inspection

program became more standardized, more sophisticated and broader in scope compared to the first

version. For example, different procedures have been developed for different types of bridge

inspections (AASHTO, 2019) and new inspection technologies have been developed, such as sonic

testing and spectrum analysis (Thomas et al., 2012). One of the most common inspection types is

routine inspection, which is a two-year periodic inspection during the service life of a bridge.

Routine inspection uses measurements and observations to evaluate the condition of bridges and

documents changes in condition since the last inspection.

Visual inspection (observations) is the primary method used in routine inspection. It is

commonly performed by inspection teams consisting of several trained bridge inspection engineers.

During each inspection, the inspection engineers examine all bridge components, such as bridge

deck, girders, bearing, and bridge columns/bridge pier walls, and document the location, type, size,
122
quantity, and description of damage severity such as area of concrete spalling and surface crack

widths. The inspection engineers assign a number (0~7 or 0~9) to the examined bridge elements,

which is called bridge element condition rating. Similarly, bridge component ratings are used to

evaluate the overall condition of the superstructure and substructure of a bridge. The method

presented in this Chapter uses bridge element condition rating data. Additional details about the

rating systems can be found in the AASHTO Manual for Bridge Inspection (AASHTO, 2019).

The New York Department of Transportation (NYSDOT) used two different rating systems.

Before 2016, NYSDOT’s bridge inspection system had a scale of 1 (failure) to 7 (new) to evaluate

the condition of a bridge component. The description of each condition rating is given in Table

6.1. After 2016, NYSDOT transitioned to AASHTO’s element-based condition rating system

(AASHTO ECR). This system has rating scales ranging from CS1-good to CS4-severe, following

AASHTO Manual for Bridge Inspection (AASHTO, 2019). Table 6.2 shows the AASHTO ECR

guidelines. The method presented in this Chapter utilizes both scales due to their unique

advantages. AASHTO ECR scale (CS1 to CS4) is used because it provides more detailed

information on elements. However, since this system is newer (since 2016), data collected using

this scale are limited to a relatively short time period. The older NYSDOT scale (1 to 7) is less

detailed but have been used over a longer period of time leading to a larger condition rating data

set.

123
Table 6.1: Condition rating guidelines of NYSDOT bridge component evaluation (adapted from
NYSDOT (2014))

Condition rating Description

7 New condition, no deterioration.
6 Used to shade between rating s of 5 and 7.
Minor deterioration, but functioning as
5
originally designed.
4 Used to shade between ratings of 3 and 5.
Serious deterioration, or not functioning as
3
originally designed.
2 Used to shade between ratings of 1 and 3.
1 Totally deteriorated, or in failed condition.

Table 6.2: AASHTO ECR guidelines (adapted from NYSDOT (2017))

Condition Condition
General Condition Guideline
State Type
That portion of the element that has either no deterioration or the
deterioration is insignificant to the management of the element, meaning
that portion of the element has no condition based preventive
CS-1 Good maintenance needs or repairs. Areas of an element that have received
long lasting structural repairs that restore the full capacity of the element
with an expected life equal to the original element may be coded as good
condition.
That portion of the element that has minor deficiencies that signify a
progression of the deterioration process. This portion of the element may
CS-2 Fair need condition based preventive maintenance. Areas of the element that
have received repairs that improve the element, but the repair is not
considered equal to the original member may be coded as fair.

That portion of the element that has advanced deterioration but does not
CS-3 Poor warrant structural review. This portion of the element may need
condition based preventative maintenance or other remedial action.

That portion of the element that warrants a structural review to determine

the effect on strength or serviceability of the element or bridge; OR a
CS-4 Severe structural review has been completed and the defects impact strength or
serviceability of the element or bridge; OR a condition where that portion
of the element is no longer effective for its intended purpose.
CS-5 Unknown That portion of the element not assessable due to lack of access.

124
6.4 Bridge deterioration curves

Bridge element deterioration curves are developed using historical element condition rating

data of many bridges in a certain region. These curves are used to estimate the future condition of

bridges in the region and to plan intervening actions. Because the curves are developed based on

inventory data, deterioration curves represent average performance of bridges in the inventory

rather than the performance of individual bridges.

As mentioned in Section 6.1, deterioration curves are commonly developed by two

methods: deterministic methods and stochastic methods. Deterministic methods (Morcous, 2011;

Veshosky et al., 1994; Yanev & Chen, 1993) utilize statistical regression for determining the

relationship between the bridge condition rating and time. The deterministic deterioration curves

can be developed using straight-line, extrapolation, regression, and curve-fitting methods

(Morcous, 2011).

The stochastic methods (Agrawal et al., 2008; Mauch & Madanat, 2001; Morcous, 2011)

capture the uncertainty and randomness in bridge deterioration. The models developed by

stochastic methods can be classified as either state-based or time-based (Mauch & Madanat, 2001).

In state-based models, deterioration models are commonly modeled by Markov chains, which

represent the change of bridge condition ratings within a certain time (e.g., 2 years between

successive inspections) as a probability matrix. The probability matrix can be developed using the

bridge condition rating data and optimization theories. Then, the relationship between the

estimated bridge condition rating and time can be developed based on the probability matrix. In

time-based models, the duration of a bridge component at a particular bridge condition rating is
125
modeled as a random variable using either a Weibull-based probability distribution or a lognormal

distribution. The probability distribution is developed by fitting the time-based model with the

bridge condition rating data.

6.5 Deterioration interpretation from bridge element condition ratings

The goal of this Chapter is to establish a link between the physics-based corrosion

indicators (e.g., crack width) and the field-based bridge element condition ratings. In this section,

the bridge element condition ratings are correlated with crack width and spalled area, which are

later used as the basis of the linkage to the physics-based indicators. For this purpose, two

relationships are needed: (1) the relationship between the corrosion indicators and the AASHTO

ECR, and (2) the relationship between the AASHTO ECR and the NYSDOT element condition

ratings (NYSDOT ECR). It should be noted that element condition ratings from other states can

also be used for this practice. NYSDOT ECR are used in this study because of their availability to

the author.

To establish the first relationship mentioned above, Table 6.4 presents a correlation

between crack widths and condition ratings of RC bridge columns, as described in AASHTO ECR

(AASHTO, 2019). Similarly, using the same reference (AASHTO, 2019), Table 6.5 presents a

correlation between area of concrete spalling and condition ratings of RC bridge columns. In this

study, crack width and concrete spalling are used as corrosion indicators.

126
 Table 6.3: Condition states of reinforced concrete elements based on surface crack widths
(adapted from AASHTO (2019))
Condition
Defect Description
State
Cracks less than 0.012’’(0.3 mm) in width can be
Insignificant cracks or moderate
CS-1 considered “insignificant” and a defect is not
width cracks that have been sealed
warranted under the Element.
Unsealed moderate width cracks or
Cracks ranging from 0.012’’(0.3 mm) up to
CS-2 unsealed moderate pattern (map)
0.05’’ (1.27 mm) can be considered “moderate” .
cracking.
Cracks equal to or greater than 0.05’’(1.27 mm)
Wide crack or heavy pattern (map)
CS-3 can be considered “wide”. Extent and severity
cracking.
are not excessive and/or widespread,

The condition is beyond the limits established in

Wide crack or heavy pattern (map) condition state three (3) and /or warrants a
CS-4
cracking including ASR. structural review to determine the strength or
serviceability of the element or bridge.

Table 6.4: Condition states of reinforced concrete elements based on areas of concrete spalling
(adapted from AASHTO (2019))
Defect CS-1 - Good CS-2 - Fair CS-3 - Poor CS-4 - Severe
Delamination None Spall less than Spall greater than 1 The condition is beyond the
/Spall 1 inch (25 inch (25 mm) deep limits established in
mm) deep or or greater than 6 condition state three (3) and
less than 6 inches (152.4 mm) /or warrants a structural
inches in in diameter or review to determine the
diameter exposed rebar strength or serviceability of
the element or bridge.

The second relationship mentioned above, between the AASHTO ECR and the NYSDOT

ECR, is needed because (1) the AASHTO ECR system has been effective in New York State only

after 2016. Therefore, there is limited historical data for developing deterioration curves using

AASHTO ECR, and (2) existing deterioration curves in the literature for bridge elements in New

127
York State are generated based on the NYSDOT ECR, but the relationships between NYSDOT

ECR and crack width are not well-quantified.

The relationship between AASHTO ECR (CS-1 to CS-5) and NYSDOT ECR (1 to 7) is

established based on several assumptions and is shown in Table 6.5. AASHTO ECR of CS-4 is

assumed to correspond to a NYSDOT ECR of 3 because they are both used to indicate loss of

functionality of a bridge element. AASHTO ECR of CS-2 is assumed to correspond to a NYSDOT

ECR of 5 because both represent minor deterioration in a bridge element. AASHTO RCR of CS-

1 is assumed to correspond to the NYSDOT ECR of 7 and 6 because both are assigned to elements

with no deterioration or insignificant deterioration. Based on the aforementioned definitions, the

remaining AASHTO ECR of CS-3 is assumed to correspond to NYSDOT ECR of 4. The condition

states CS-4 and CS-3 have no differences in terms of crack width (see Table 6.3). Therefore, only

CS-3 was considered in the relationship between AASHTO ECR and NYSDOT ECR.

Due to the simplified model used in this study (Section 6.2.1), after concrete spalling, the

depth of the spalled area is assumed to be equal to the thickness of the concrete cover. For most

concrete transportation structures, the cover thickness is at least 1 inch. Therefore, when the time

for concrete spalling is reached, the condition rating is assumed to drop down to CS-3.

Using the two relationships described above (between crack width/spalling and AASHTO

ECR, and between AASHTO ECR and NYSDOT ECR), NYSDOT ECR is defined in terms of

crack width and concrete spalled area as shown in Table 6.5. It should be noted that the

assumptions made to connect the AASHTO ECR to NYSDOT ECR can be modified and refined

128
in the future, but these assumptions do not change the procedures for the general method of linking

physics-based and field-based methods described in the subsequent sections.

Table 6.5: Relationship between NYSDOT ECR and crack width

NYSDOT bridge element

7 6 5 4
condition rating

AASHTO bridge element

CS-1 CS-1 CS-2 CS-3
condition rating

Crack width (𝑊),

𝑊=0 0 < 𝑊 ≤ 0.012 0.012 < 𝑊 ≤ 0.05 𝑊 > 0.05
inch

Spall
Spall less than 1 greater than
inch deep or less 1 inch deep
Spalled area N/A N/A
than 6 inch or greater
diameter* than 6 inch
diameter*
* Since the depth of the spalled area is assumed to be the cover thickness and since cover thickness
is larger than 1 inch for most bridges, spalling is assumed at NYSDOT ECR of 4 or AASHTO
ECR of CS-3.

6.6 Corrosion model calibration

The corrosion model parameters, corrosion rate (𝜆(𝑡) ), pitting factor (R), corrosion

initiation time, the ratio (K) of 𝜆(𝑡) to 𝜆(𝑡) and critical crack width (𝑊 ), can be

calibrated using the deterioration curves introduced in Section 6.4. These curves represent the

average deterioration rate of bridges in a state or a region. Although the bridge inspection data of

a particular bridge can also be used for calibration, this data is limited, especially for bridges built

within the last 10 years. Using a limited data set to calibrate the corrosion model would cause a

large bias. Moreover, bridge deterioration curves are the most commonly used tools by state DOTs

129
for asset management. Therefore, using these curves for model calibration is relevant and is

expected to accelerate adoption by practicing engineers.

The method for calibrating the inputs of the corrosion model is illustrated in Figure 6.2 and

Figure 6.3. In the first calibration step, crack width is used as the criterion for calibration. The

corrosion model calculates the crack width as a function of bridge age (Chapter 3). Deterioration

curves provide the bridge ECR as a function of bridge age, but they can be converted to equivalent

crack widths as a function of bridge age using the interpretation described in Table 6.5. The output

of this step is the lower and upper bounds of crack widths and average of the bounds corresponding

to a condition rating. The calibration process involves using the least sum of squares method to

minimize the difference between crack width calculated by the corrosion model (𝑊(𝑡 )) and the

average crack width (𝑊 (𝑡 )) calculated as the average of the upper (𝑊 (𝑡 )) and lower bound

(𝑊 (𝑡 )) of crack widths derived from the deterioration curves as shown in Equation (47a). If the

crack width range determined from the deterioration curves does not have an upper bound value

(e.g., crack width > 0.05 inch for NYSDOT ECR of 4 and AASHTO ECR of CS-3 in Table 6.5),

the crack width calculated by the corrosion model should be larger than the lower bound crack

width determined from the deterioration curves (Equation (47b). The inputs of the corrosion model,

such as 𝜆(𝑡) , are changed iteratively until the conditions given in Equation (47) are satisfied.

( ) ( )
(𝑎) 𝑊 (𝑡 ) =

(𝑏) 𝑖𝑓 𝑊 (𝑡 ) 𝑖𝑠 𝑑𝑒𝑓𝑖𝑛𝑒𝑑, ∑ [𝑊(𝑡 ) − 𝑊 (𝑡 )] ≅ 0 (47)

(𝑐) 𝑖𝑓 𝑊 (𝑡 ) 𝑖𝑠 𝑛𝑜𝑡 𝑑𝑒𝑓𝑖𝑛𝑒𝑑, 𝑊(𝑡) ≥ 𝑊 (𝑡)

130
 Figure 6.2: Calibration framework

In the second calibration step, concrete spalling criterion is checked by iteratively changing

the cracking angle 𝜃 . Physically, the cracking angle is governed by the aggregate size,

aggregate location, and the aggregate-cement paste interfacial characteristics. Due to the

complexity of the concrete spalling process, it is difficult to accurately estimate 𝜃 from

fundamental parameters. Therefore, this variable was calibrated using field-data. In this study,

𝜃 in Equation (45) is calibrated in the range of 15° to 75° (Su & Zhang, 2019; Zhang et al.,

131
2021) to ensure that the criteria shown in Table 6.5 for both crack width and spalling are satisfied

simultaneously. Figure 6.3 demonstrates the process of calibration.

Figure 6.3: Demonstration of the calibration process

6.7 Demonstration of the Calibration Process with an Example

An example RC bridge located in Erie County, NY, crossing over a 2-lane state route, is

used to demonstrate the calibration process. The bridge has two spans and a concrete deck

supported by two RC bridge columns, exposed to a large amount of deicing salts during the winter

season. In this study, the corrosion rate of one bridge column was calibrated based on the

deterioration curves. The deterioration curves used in this study were developed by Agrawal et al.

132
(Agrawal et al., 2008), which represent the deterioration of concrete bridge columns in New York

State. The deterioration function is expressed as shown in Equation (48), where 𝑇 is the time in

years since bridge construction. By using Equation (48) and Table 6.5, the criteria for the corrosion

rate calibration can be developed as shown in Figure 6.4. It should be noted that the specific

deterioration curve shown in Equation (48) appears as a “straight line” in Figure 6.4. Typically,

the rate of deterioration (or decrease in condition ratings) reduces with time.

Figure 6.4: Calibration of corrosion rate before cracking (𝜆(𝑡) )

In Figure 6.4, the blue line is the condition rating expressed by Equation (48). The pink

lines were developed from the deterioration curve. They indicate the upper, lower bounds and

average of crack width at a given condition rating and the time when concrete spalls based on the

condition rating. The crack width (𝑊(𝑡)) curve computed by the corrosion model (the black line

in Figure 6.4) should satisfy Equation (47).

133
 𝑁𝑌𝑆𝐷𝑂𝑇 𝐸𝐶𝑅 = 7 − 0.0486218𝑇 − 0.00001326𝑇 − 0.0000012𝑇 (48)

Before calibration, the corrosion initiation time of the bridge column is calculated by

inputting the parameters shown in Table 6.6 into Equation (1) of Chapter 3. These input parameters

are within reasonable range, as supported by the references provided in Table 6.6. The corrosion

initiation time is calculated as 3.3 years using this input. Chapter 3 section 3.5.2 showed that the

corrosion rate before cracking (𝜆(𝑡) ) has the most significant effect on the corrosion model

results. Therefore, the calibration of this parameter is presented in this example. The effects of

calibrating other corrosion model inputs are discussed in Section 6.8.

134
 Table 6.6: Model parameters of the bridge column

Corrosion
Values before
model Units Source
calibration
parameters
𝐶 lbs/yard3 0.062 Weyers et al. (1994)
𝐷 inch2/year 0.13 Weyers et al. (1994)
Critical
chloride
lbs/yard3 2.02 Life-365™ (2020)
content at rebar
surface
Corrosion
years 3.3 -
initiation time
C inch 1.4 -
𝐸 ksi 3704 -
4.0×10-4 Andrade & Alonso
𝜆(𝑡) inch/year
(before calibration) (2001)
K - 2 -
Critical crack
width, inch 0.008 Cui et al. (2018)
(𝑊 )
Val & Melchers
R - 6
(1997)
𝐸 ksi 3 Liu & Su (2018)
El Maaddawy &
𝜈 - 0.18
Soudki (2007)
𝜈 - 0.5 Lu et al. (2011)
Liu & Su (2018); Lu
n - 4
et al. (2011)
35°
𝜃 - Su & Zhang (2019)
(before calibration)

135
 The procedure for calibrating the corrosion model for the example bridge is as follows:

Step 1: By inputting the value of 𝜆(𝑡) ,𝑊 , R, K and 𝐸 in the corrosion model as

illustrated in Chapter 3 and Chapter 6, the crack width as a function of time and time for concrete

spalling is calculated (dashed line in Figure 6.4).

Step 2: The lower bound, upper bound and average of lower and upper bounds of crack

width (pink dots) corresponding to bridge ages at which condition ratings (blue line) change are

obtained from deterioration curves. At 𝑡 = 19.7 years, 𝑊 (𝑡 ) = 0 inches, 𝑊 (𝑡 ) = 0.012

inches and 𝑊 (𝑡 ) = 0.006 inches. At 𝑡 = 38.5 years 𝑊 (𝑡 ) = 0.012 inches, 𝑊 (𝑡 ) =

0.05 inches and 𝑊 (𝑡 ) = 0.031inches. At 𝑡 = 57.4 years, 𝑊 (𝑡 ) = 0.05 inches.

Step 3: The crack widths calculated by the corrosion model (Step 1) are checked to

determine whether they satisfy the conditions of Equation (47). If they do not, the corrosion rate

(𝜆(𝑡) ) is changed and Step 1 is repeated iteratively until the conditions of Equation (47)

are satisfied. For this example bridge, 𝜆(𝑡) that satisfied the conditions given in Equation

(47) was 3.8×10-4 inch/year.

The comparison between the predicted crack width before and after calibration is presented

in Figure 6.4. The blue line is the deterioration curve developed by Agrawal et al. (Agrawal et al.,

2008) for NYSDOT bridges. As shown in the figure, the calibrated value of 𝜆(𝑡) is 3.8×10-

4
inch/year. Before calibration, 𝜆(𝑡) was 4.0×10-4 inch/year. The value before calibration

136
resulted in crack widths within the range of crack widths obtained based on the deterioration curve

but closer to the upper bound crack width.

Step 4: Using 𝐸 , 𝜐 , n, and 𝜐 shown in Table 6.6, the pressure due to rust expansion

(Equation (46)) is calculated as a function of time. Before calibration, concrete spalled at 38 years,

when the condition rating drops to NYSDOT ECR of 4, based on the assumption in Chapter 6.5.

The parameter 𝜃 is calibrated iteratively until the concrete spalling criterion shown in

Figure 6.4 is satisfied. The value of 𝜃 thus obtained from calibration is 33.5°.

6.8 Effects of different corrosion input parameters on crack width

This section serves two purposes: (1) to identify input parameters that have negligible

effects on the results so that calibration efficiency can be improved by excluding these parameters

from the calibration process, (2) to identify the input parameter combinations that lead to a match

between the corrosion model results and field-data.

In this section, the effects of different corrosion input parameters of the corrosion model

were investigated by varying these parameters one at a time. The input parameters varied were the

corrosion rate before cracking (𝜆(𝑡) ), pitting factor ( R), the ratio (K) of 𝜆(𝑡) to

𝜆(𝑡) , corrosion initiation time, and critical crack width ( 𝑊 ). The example bridge

presented in Section 6.7 was used as a baseline for this exercise. The values of the input parameters

that led to a match between corrosion model and field-based data are summarized in Table 6.7 and

discussed in this section.

137
 Table 6.7: Values of the corrosion model input parameters that led to a reasonable match with
field-based data

Corrosion model Source for

Units Values Typical range
parameters typical range
2.9×10-4, 3.3×10-4, Andrade & Alonso
𝜆(𝑡) inch/year 2.0×10-4 ~ 4.0×10-4
3.8×10-4*, 4.0×10-4 (2001)

Val & Melchers

R - 4.8, 6.0*,7.3 4.0 ~ 8.0
(1997)
K - 1.4, 2.0*, 2.6 - -
Corrosion initiation
years 3.3*, 5.0, 6.0, 7.0 - -
time
Critical crack width, 0.004, 0.008*,
inch - -
(𝑊 ) 0.012
* Baseline values that were kept constant when other parameters changed one at a time.

Figure 6.5 shows that the crack width increases with an increase in 𝜆(𝑡) . This is

because 𝜆(𝑡) is the major factor determining the rebar area loss as shown in Equation (7)

of Chapter 3, which in turn influences the crack width (Equation (15)). For the baseline values of

R, K, corrosion initiation time, and critical crack width shown in Table 6.7, the minimum and

maximum values that 𝜆(𝑡) can take while keeping the predicted crack width within the

range determined from deterioration curves are 2.9×10-4 inch/year and 4.0×10-4 inch/year,

respectively, which are within the range reported by Andrade & Alonso (2001).

138
 𝜆(𝑡)

Figure 6.5: Effect of varying 𝜆(𝑡) on crack width evolution

Figure 6.6 and Figure 6.7 show that the effect of the parameters R and K, respectively, on

crack width are significant. This is attributed to the fact that, similar to 𝜆(𝑡) , R and K are

major factors determining the rebar area loss in Equation (7). The minimum and maximum values

of R that maintain the crack width consistent with those determined from deterioration curves are

4.8 and 7.3, respectively, when other parameters are assigned the baseline values shown in Table

6.7. The minimum and maximum values of K that are within the calibration range are 1.4 and 2.6,

respectively, when other parameters are assigned the baseline values.

139
 R

Figure 6.6: Effect of varying R on crack width evolution

Figure 6.7: Effect of varying K on crack width evolution

140
 Figure 6.8 shows that corrosion initiation time has negligible effects on crack width. This

is likely because the corrosion initiation times (3 to 9 years) were small compared to the time scale

considered (tens of years), which resulted in insignificant effects on the reinforcement area loss. It

should be noted that 𝜆(𝑡) also covers a short time period before cracking (6-9 years).

However, 𝜆(𝑡) also influences the corrosion rate after cracking and the maximum corrosion

rate and has a significant effect on the crack width. All reasonable values of corrosion initiation

time led to crack widths that were within the calibration range, when other parameters were the

baseline values shown in Table 6.7.

Corrosion initiation time

Figure 6.8: Effect of corrosion initiation time on crack width evolution

Figure 6.9 shows that crack width is not sensitive to 𝑊 , i.e., the critical crack width at

which the corrosion rate reaches the maximum value. For the same

𝜆(𝑡) , 𝑅, and corrosion initiation time, the time at which concrete cracks computed by the

141
corrosion model is the same. Therefore, 𝑊 only affects the time during which the corrosion rate

increases, which is small (less than 10 years) compared to the service life of the bridges (75 years).

Therefore, 𝑊 has a limited effect on crack width. When all other parameters are assigned the

baseline values shown in Table 6.7, all reasonable values of 𝑊 lead to crack widths that are

within the range determined from the deterioration curves.

Figure 6.9: Effect of 𝑊 on crack width evolution

Based on the above results, the corrosion initiation time and 𝑊 do not need to be

considered as parameters to be calibrated. Although the effects of R and K are significant, it is hard

to collect data from real bridges or laboratory experiments for determining the values of these

empirical parameters. In the literature (discussed in Chapter 3), R ranges from 4 to 8 ((Val &

Melchers, 1997). It quantifies the distribution of pitting locations along a corroded rebar (Stewart,

2004). In real practice, it is difficult to observe all the corroded locations of a rebar embedded in a

142
bridge column. For K, which is the ratio of corrosion rate after cracking to that before cracking,

there are limited laboratory tests (Otieno et al., 2016; Yuan et al., 2010) and no field observations.

Thus, the most suitable parameter for calibration is determined to be the corrosion rate before

cracking (𝜆(𝑡) ). 𝜆(𝑡) determined from the calibration is within the range reported

by Andrade & Alonso (2001).

6.9 Summary and conclusions

A systematic method for linking the physics-based corrosion model and the field-based

condition assessments is proposed in this chapter. In this method, the physics-based corrosion

model introduced in Chapter 3 was expanded by incorporating concrete spalling because spalling

affects bridge condition ratings. Then, links between the surface crack width and bridge element

condition rating, as well as concrete spalling and bridge element condition rating were established

based on the guidelines of NYSDOT (2017) and AASHTO (2019). Finally, based on the developed

link, a calibration method for the inputs of the corrosion model was proposed by matching the

crack widths and spalling interpreted from deterioration curves with the outputs of the physics-

based corrosion model.

The use of the calibration method was demonstrated on an example bridge in New York

State. The calibration identified a range for corrosion rates before cracking that leads to a match

between the corrosion model and deterioration curve predictions. The upper bound of this range

corresponded to the rate reported in the literature by others (Andrade & Alonso, 2001).

143
 The effects of various model parameters, including the corrosion rate before cracking

(𝜆(𝑡) ), pitting factor (R), the ratio (K) of 𝜆(𝑡) to 𝜆(𝑡) , corrosion initiation

time and critical crack width (𝑊 ), were investigated. The results show that crack width increases

with the increase of 𝜆(𝑡) , 𝑅, and K. Corrosion initiation time and 𝑊 have negligible

effects on the crack width. 𝜆(𝑡) is recommended as the most appropriate parameter to be

calibrated in the proposed method because it is easier to be measured in real practice than other

parameters and has a significant effect on the crack width. 𝜆(𝑡) determined from the

proposed method is within the range reported by Andrade & Alonso (2001).

144
 Chapter 7: Concluding Remarks

7.1 Summary and conclusions

Brief summaries and the major conclusions of each chapter are given below. Chapter 1

provided the background and the goals and scope of the research. A literature review of the major

topics was presented in Chapter 2. Additionally, each chapter contained a review of the specific

topics covered in that chapter.

In Chapter 3, a systematic framework for determining the seismic fragility of bridge

columns undergoing corrosion was presented. The framework considered the type of cover

material and its cracking characteristics, the effects of pitting corrosion on both longitudinal and

transverse reinforcement, effect of cracks on corrosion rate, and the time-dependency of damage

thresholds used in the fragility analysis. The framework was demonstrated using an RC bridge

column with two different cover materials - conventional concrete and ECC. Overall, the computed

effects of corrosion on the seismic fragility of the column with either cover material became more

significant with time. Notably, the computed transverse and longitudinal reinforcement area losses

in the column with ECC cover were 63% lower and 75% lower than that with conventional

concrete cover due to better crack width control with ECC. As a result, the column performed

better with ECC cover than with conventional concrete cover under simulated seismic loads. For

example, the column with ECC cover exhibited 40% lower probability of extensive damage at

PGA of 1.0g compared to the concrete cover. Furthermore, these improvements were more

pronounced at higher damage levels (extensive damage and complete damage) than at lower

145
damage levels (slight damage and moderate damage). Thus, the more durable cover material was

shown to improve the seismic performance of the example RC bridge column.

In Chapter 4, the framework developed in Chapter 3 was used to study the seismic

vulnerability of a group of bridges undergoing corrosion. This exercise was performed to mimic

the application of the framework to an inventory of bridges. Eight bridges owned by the

Washington State Department of Transportation (WSDOT) were used for this purpose. The

framework incorporated site-specific seismic hazard curves. The analyses of a small group of 8

bridges showed that: (1) all bridges became more vulnerable to seismicity due to corrosion

deterioration; (2) the order of seismic vulnerability of bridges changed over time due to corrosion.

Therefore, corrosion must be considered in seismic assessments to prioritize bridges maintenance.

This is expected to be even more significant for state bridge inventories comprising of hundreds

of bridges with large variabilities in age and exposure condition.

Chapter 5 presented a systematic method for incorporating vehicle spray and splash

mechanisms in the estimation of chloride exposure of RC highway bridge piers, thus capturing

local exposure conditions of a bridge. In this method, the salt application and snow precipitation

during a winter season were used to determine the water film thickness on the road surface. A

regression model based on the results of computational fluid dynamics analyses from the literature

was utilized to determine the amount of water and dissolved chloride ions kicked up to the

environment around a bridge by the vehicles passing through various spray and splash mechanisms.

Finally, the chloride ions transferred from the environment to the surface of a bridge substructure

were calculated using an empirical model. Although the observed surface chloride concentration

146
differed from the computed values for some of the bridges considered in this study, the proposed

method correctly differentiated between the bridges with high and low chloride exposure. This in

turn can be used for improving the estimates of corrosion deterioration for a specific bridge.

Chapter 6 described a new method for linking the physics-based corrosion model to field-

based condition assessments used by transportation agencies for asset management. In this chapter,

the physics-based corrosion model developed in Chapter 3 was first expanded by incorporating

concrete spalling. Then, a link between model predictions (i.e., surface crack width and concrete

spalling) and deterioration curves, based on bridge element condition ratings, was then established

based on the guidelines of NYSDOT (2014) and AASHTO (2019). Finally, the inputs of the

corrosion model, such as corrosion rates before and after cracking, pitting factor, were calibrated

for an example RC bridge in New York State by comparing the crack widths and spalling from

deterioration curves and the physics-based corrosion model. The corrosion rate before cracking

(𝜆(𝑡) ) influenced the corrosion model more than other input parameters such as corrosion

initiation time and critical crack width. After calibration, it was determined that, if the condition

ratings of the example bridge follow the state-wide deterioration curve, corrosion rate before

cracking (𝜆(𝑡) ) must be between 2.9×10-4 inch/year and 4.0×10-4 inch/year for the example

bridge.

147
7.2 Scientific contributions and research impacts

The major scientific contributions and broader impacts of the research presented in this

dissertation are:

1. Improved corrosion model for RC bridge piers incorporating pitting corrosion, cracking,

and a new cover material: The existing computational models for combined corrosion and

seismic analyses of RC bridges assume uniform (global) corrosion of steel reinforcement,

leading to unconservative estimates of rebar area loss with time. The computational model

presented in Chapter 3 accounts for pitting (local) corrosion, which, although more

complex, is more representative of corrosion in real RC bridges. Furthermore, unlike

existing models, the corrosion model presented in this dissertation accounts for the effect

of cracking on the corrosion rates of conventional and fiber-reinforced concretes. This

enables the investigations of the effects of innovative materials, such as ECC, on the

durability and seismic fragility of RC bridges, which is not possible with the existing

models.

2. Developed a novel method for estimating chloride exposure of RC bridges considering

local exposure parameters: The existing corrosion or service-life estimation models (e.g.,

Life 365) determine the surface chloride exposure based only on the region (e.g., Western

New York) where the structure is located, which leads to the assumption of the same

chloride exposure of bridges located within the same region. In contrast, the method

developed in Chapter 5 allows evaluation of chloride exposure (and thereby corrosion

148
 susceptibility) of bridges individually based on the unique characteristics of the crossed

features, snow precipitation, salt application, and traffic patterns. Thus, the new method

for surface chloride estimation can differentiate between RC bridges in the same region but

with different traffic patterns and local conditions. The developed method can be used to

improve the estimation of corrosion initiation time for different RC bridges in a state bridge

inventory and prioritize maintenance accordingly.

3. Developed a link between the physics-based corrosion model and the field-based condition

ratings: There are several physics-based corrosion models presented in the literature, which

quantify deterioration in terms of parameters such as crack width and rebar area loss.

However, such physics-based models are seldom utilized by the transportation agencies for

determining the maintenance needs of RC bridges. Instead, the transportation agencies rely

on periodic field inspections that are time and labor intensive, subjective, and largely based

on visible surface damage (assisted with limited non-destructive testing). A new method

for linking the physics-based corrosion models to deterioration curves (obtained from field

inspections) is developed, which allows for a quantitative assessment of bridge

deterioration. At the same time, the deterioration curves can be used to calibrate and

improve the physics-based corrosion models, as demonstrated in Chapter 6. Overall, a state

bridge management program could benefit from a hybrid approach of utilizing field-based

inspections complemented by calibrated physics-based corrosion models.

149
7.3 Recommendations for future research

There are several possibilities for building upon this research in the future, as listed below:

1. The corrosion model in Chapter 3 can be further developed by considering the location of

pitting corrosion along the rebar, concrete spalling effects on corrosion rate, and using a

more advanced approach, such as two-dimensional (2D) diffusion of chloride ions (as

opposed to 1D diffusion assumed in this research), for calculating the corrosion initiation

time.

2. The application of ECC cover for reducing corrosion and improving seismic performance

of RC bridge columns was investigated only theoretically in Chapter 3. Laboratory or field

experiments should be performed to confirm the advantages of ECC cover. Moreover,

construction methods need to be investigated for a full-scale structure with concrete core

and ECC cover.

3. This study only considered far-field ground motions in dynamic analysis. Future studies

could consider near-field ground motions in the analysis for bridges.

4. Several assumptions are made in Chapter 5 for estimating the chloride exposure of RC

highway bridges. For example, all deicing salts are assumed to be dissolved in water after

deicing operations; the effects of runoff, ground percolation, evaporation, plowing, and

other mechanisms are ignored. Field experiments and data collection are needed to validate

or update these assumptions.

150
5. The vehicle spray and splash model presented in Chapter 5 is based on assumed scaling of

a single truck to account for vehicles of other sizes and shapes. Future research should

consider the interaction effects of multiple vehicles on spray and splash mechanisms, which

can further improve the accuracy of the proposed model.

6. Only limited surface chloride content data was available for the validation of the chloride

exposure model in Chapter 5. Future studies could collect surface chloride content data

from existing aging bridges to verify and refine the proposed model.

7. The corrosion assessment models proposed in this research (Chapter 5 and Chapter 6) focus

on bridge substructures such as bridge columns and bridge pier walls. Future studies could

expand the proposed model to estimate the corrosion of other bridge components such as

decks and girders.

8. The proposed corrosion model calibration method (Chapter 6) could be further improved

by collecting additional inspection data for specific bridges. The inspection data includes

photographs, bridge condition ratings, measurements of crack width, records of concrete

spalled area and rebar area loss.

9. There are other deterioration mechanisms that can reduce the capacity or functionality of

RC bridge components, such as freeze-thaw, shrinkage, thermal effects, alkali-aggregate

reaction, erosion, etc. Similarly, there are other extreme events (other than earthquakes),

such as vehicle impacts, floods, hurricanes, terrorist attacks, etc. that can lead to sudden

151
 loss of functionality of RC bridge components. Future studies could incorporate these

deterioration mechanisms and extreme events into the proposed framework.

10. The proposed framework did not consider uncertainties in corrosion parameters (e.g.,

corrosion rate before cracking) and structural parameters (e.g., longitudinal reinforcement

ratio). Future studies could incorporate uncertainties of corrosion and structural parameters

in the fragility analysis.

152
 References

American Association of State Highway and Transportation Officials (AASHTO) (2009).

Standard method of test for sampling and testing for chloride ion in concrete and concrete

raw materials, T260, AASHTO, Washington D.C., United States.

American Association of State Highway and Transportation Officials (AASHTO) (2013).

AASHTO Transportation asset management guide - a focus on implementation, AASHTO,

Washington D.C., United States.

American Association of State Highway and Transportation Officials (AASHTO) (2019).

AASHTO guide manual for bridge element inspection, second edition, AASHTO,

Washington D.C., United States.

American Association of State Highway and Transportation Officials (AASHTO) (2020). LRFD

bridge design specifications, ninth edition, AASHTO, Washington D.C., United States.

American Concrete Institute (ACI) (2014). Building code requirements for structural concrete

and commentary, ACI 318-14, ACI, Farmington Hills, MI, United States.

Afsar Dizaj, E., Madandoust, R., & Kashani, M. M. (2018). Exploring the impact of chloride-

induced corrosion on seismic damage limit states and residual capacity of reinforced

concrete structures. Structure and Infrastructure Engineering, 14(6), 714-729.

Agrawal, A. K., Kawaguchi, A., & Chen, Z. (2008). Bridge element deterioration rates, C-01-51,

The City College of New York, New York, NY, United States.

Agrawal, A. K., Kawaguchi, A., & Chen, Z. (2010). Deterioration rates of typical bridge elements

in New York. Journal of Bridge Engineering, 15(4), 419-429.

153
Akiyama, M., Frangopol, D. M., & Matsuzaki, H. (2011). Life-cycle reliability of RC bridge piers

under seismic and airborne chloride hazards. Earthquake Engineering and Structural

Dynamics, 40(15), 1671-1687.

Alipour, A., Shafei, B., & Shinozuka, M. (2010). Performance evaluation of deteriorating highway

bridges located in high seismic areas. Journal of Bridge Engineering, 16(5), 597-611.

Andersen, A. (1997). Investigation of chloride penetration into bridge columns exposed to de-

icing salt, No. 82, Danish Road Directorate, Region Zealand, Denmark.

Andrade, C., & Alonso, C. (2001). On-site measurements of corrosion rate of reinforcements.

Construction and Building Materials, 15(2-3), 141-145.

American Society of Civil Engineers (ASCE) (2019). Seismic design criteria for structures, systms,

and components in nuclear facilities, ASCE/SEI 43-05, ASCE, Reston, VA, United States.

American Society of Civil Engineers (ASCE) (2021). 2021 Report card for America's

infrastructure, ASCE, Reston, VA, United States.

American Society for Testing and Materials (ASTM) (2016). Standard test method for determining

the apparent chloride diffusion coefficient of cementitious mixtures by bulk diffusion,

ASTM C1556, ASTM, West Conshohocken, PA, United States.

American Society for Testing and Materials (ASTM) (2020). Standard specification for deformed

and plain low-alloy steel bars for concrete reinforcement, ASTM A706, ASTM, West

Conshohocken, PA, United States.

Applied Technology Council (ATC) (2009). Quantification of building seismic performance

factors, FEMA P695, Federal Emergency Management Agency (FEMA), Washington

D.C., United States.

154
Applied Technology Council (ATC) (1985). Earthquake damage evaluation data for California,

ATC-13, Federal Emergency Management Agency (FEMA), Washington D.C., United

States.

Baker, J. W. (2015). Efficient analytical fragility function fitting using dynamic structural analysis.

Earthquake Spectra, 31(1), 579-599.

Baroghel-Bouny, V., Thiéry, M., & Wang, X. (2011). Modelling of isothermal coupled moisture–

ion transport in cementitious materials. Cement and Concrete Research, 41(8), 828-841.

Basheer, L., Kropp, J., & Cleland, D. J. (2001). Assessment of the durability of concrete from its

permeation properties: a review. Construction and Building Materials, 15(2-3), 93-103.

Basöz, N. I., Kiremidjian, A. S., King, S. A., & Law, K. H. (1999). Statistical analysis of bridge

damage data from the 1994 Northridge, CA, earthquake. Earthquake Spectra, 15(1), 25-

54.

Bertolini, L., Elsener, B., Pedeferri, P., Redaelli, E., & Polder, R. (2004). Corrosion of steel in

concrete: prevention, diagnosis, repair, WILEY-VCH Verlag GmbH and Co. KGaA,

Weinheim, Germany.

Billah, A. M., Alam, M. S., & Bhuiyan, M. R. (2013). Fragility analysis of retrofitted multicolumn

bridge bent subjected to near-fault and far-field ground motion. Journal of Bridge

Engineering, 18(10), 992-1004.

Biondini, F., Camnasio, E., & Palermo, A. (2014). Lifetime seismic performance of concrete

bridges exposed to corrosion. Structure and Infrastructure Engineering, 10(7), 880-900.

Blomqvist, G. (2001). De-icing salt and the roadside environment: Air-borne exposure, damage

to Norway spruce and system monitoring, PhD dissertation, KTH Royal Institute of

Technology, Stockholm, Sweden.

155
Blomqvist, G., & Johansson, E.-L. (1999). Airborne spreading and deposition of de-icing salt - a

case study. Science of the Total Environment, 235(1-3), 161-168.

Bommer, J., Spence, R., Erdik, M., Tabuchi, S., Aydinoglu, N., Booth, E., Del Re, D., & Peterken,

O. (2002). Development of an earthquake loss model for Turkish catastrophe insurance.

Journal of Seismology, 6(3), 431-446.

Bu, G., Lee, J., Guan, H., Blumenstein, M., & Loo, Y.-C. (2014). Development of an integrated

method for probabilistic bridge-deterioration modeling. Journal of Performance of

Constructed Facilities, 28(2), 330-340.

California Department of Transportation (Caltrans) (2019). Seismic design criteria, version 2.0,

Caltrans, Sacramento, CA, United States.

Cao, C., Cheung, M. M., & Chan, B. Y. (2013). Modelling of interaction between corrosion-

induced concrete cover crack and steel corrosion rate. Corrosion Science, 69, 97-109.

Caré, S., Nguyen, Q. T., l'Hostis, V., & Berthaud, Y. (2008). Mechanical properties of the rust

layer induced by impressed current method in reinforced mortar. Cement and Concrete

Research, 38(8-9), 1079-1091.

Chalioris, C. E., Kytinou, V. K., Voutetaki, M. E., & Papadopoulos, N. A. (2019). Repair of

heavily damaged RC beams failing in shear using U-shaped mortar jackets. Buildings, 9(6),

146.

Chang, G., & Mander, J. B. (1994). Seismic energy based fatigue damage analysis of bridge

columns: part 1-evaluation of seismic capacity, NCEER-94-0006, National Center for

Earthquake Engineering Research, University at Buffalo, Buffalo, NY, United States.

156
Cheng, H., Li, H.-N., Yang, Y., & Wang, D.-S. (2019). Seismic fragility analysis of deteriorating

RC bridge columns with time-variant capacity index. Bulletin of Earthquake Engineering,

17(7), 4247-4267.

Choe, D.-E., Gardoni, P., Rosowsky, D., & Haukaas, T. (2008). Probabilistic capacity models and

seismic fragility estimates for RC columns subject to corrosion. Reliability Engineering

and System Safety, 93(3), 383-393.

Choe, D.-E., Gardoni, P., Rosowsky, D., & Haukaas, T. (2009). Seismic fragility estimates for

reinforced concrete bridges subject to corrosion. Structural Safety, 31(4), 275-283.

Conciatori, D., Brühwiler, E., & Linden, C. (2018). Numerical simulation of the probability of

corrosion initiation of RC elements made of reinforcing steel with improved corrosion

performance. Structure and Infrastructure Engineering, 14(11), 1446-1454.

Crank, J. (1979). The mathematics of diffusion, Oxford University Press Inc. Oxford, United

Kingdom.

Cui, F., Zhang, H., Ghosn, M., & Xu, Y. (2018). Seismic fragility analysis of deteriorating RC

bridge substructures subject to marine chloride-induced corrosion. Engineering Structures,

155, 61-72.

Darmawan, M. (2010). Pitting corrosion model for reinforced concrete structures in a chloride

environment. Magazine of Concrete Research, 62(2), 91.

Denby, B. R., Sundvor, I., Johansson, C., Pirjola, L., Ketzel, M., Norman, M., Kupiainen, K.,

Gustafsson, M., Blomqvist, G., & Kauhaniemi, M. (2013). A coupled road dust and surface

moisture model to predict non-exhaust road traffic induced particle emissions (NORTRIP)

Part 2: Surface moisture and salt impact modelling. Atmospheric Environment, 81, 485-

503.

157
Dominguez, M. T. (2019). Transportation Asset Management Plan, New York State Department

of Transportation, Albany, NY, United States.

Du, Y., Clark, L., & Chan, A. (2005). Residual capacity of corroded reinforcing bars. Magazine of

Concrete Research, 57(3), 135-147.

El Maaddawy, T., & Soudki, K. (2007). A model for prediction of time from corrosion initiation

to corrosion cracking. Cement and Concrete Composites, 29(3), 168-175.

Enright, M. P., & Frangopol, D. M. (1998). Probabilistic analysis of resistance degradation of

reinforced concrete bridge beams under corrosion. Engineering Structures, 20(11), 960-

971.

Fakhri, H. (2019). Corrosion mitigation in reinforced concrete structures using engineered

cementitious composites, PhD dissertation, University at Buffalo, NY, United States.

Fakhri, H., Ragalwar, K. A., & Ranade, R. (2019). On the use of strain-hardening cementitious

composite covers to mitigate corrosion in reinforced concrete structures. Construction and

Building Materials, 224, 850-862.

Fang, C., Lundgren, K., Chen, L., & Zhu, C. (2004). Corrosion influence on bond in reinforced

concrete. Cement and Concrete Research, 34(11), 2159-2167.

Federal Highway Administration (FHWA) (2017). 2015 Status of the nation's highways, bridges,

and transit conditions and performance report to congress, FHWA, Washington D.C.,

United States.

Filippou, F. C., Bertero, V. V., & Popov, E. P. (1983). Effects of bond deterioration on hysteretic

behavior of reinforced concrete joints, UCB/EERC-83, Earthquake Engineering Research

Center, University of California, Berkeley, CA, United States.

158
Fischer, G., & Li, V. C. (2002). Effect of matrix ductility on deformation behavior of steel-

reinforced ECC flexural members under reversed cyclic loading conditions. Structural

Journal, 99(6), 781-790.

Flintsch, G. W., Tang, L., Katicha, S. W., de León Izeppi, E., Viner, H., Dunford, A., Nesnas, K.,

Coyle, F., Sanders, P., & Gibbons, R. B. (2014). Splash and spray assessment tool

development program, DTFH61-08-C-00030, Transportation Institute of Virginia Tech.,

Virginia Tech., VA, United States.

Fukuyama, H., Sato, Y., Li, V. C., Matsuzaki, Y., & Mihashi, H. (2000). Ductile engineered

cementitious composite elements for seismic structural applications. Proceedings of 12th

World Conference on Earthquake Engineering, Auckland, New Zealand.

Ghosh, J., & Padgett, J. E. (2010). Aging considerations in the development of time-dependent

seismic fragility curves. Journal of Structural Engineering, 136(12), 1497-1511.

Ghosh, J., & Sood, P. (2016). Consideration of time-evolving capacity distributions and improved

degradation models for seismic fragility assessment of aging highway bridges. Reliability

Engineering and System Safety, 154, 197-218.

Grossi, P. (2005). Catastrophe modeling: a new approach to managing risk, Vol. 25, Springer

Science and Business Media, Berlin, Germany.

Han, T.-S., Feenstra, P. H., & Billington, S. L. (2003). Simulation of highly ductile fiber-reinforced

cement-based composite components under cyclic loading. Structural Journal, 100(6),

749-757.

Hanjari, K. Z., Coronelli, D., & Lundgren, K. (2011). Bond capacity of severely corroded bars

with corroded stirrups. Magazine of Concrete Research, 63(12), 953-968.

159
Heintz, J., Pollin, R., & Garrett-Peltier, H. (2009). How infrastructure investments support the US

economy: employment, productivity and growth, Political Economy Research Institute

(PERI), University of Massachussetts, Amherst, MA, United States.

Hosseinpour, F., & Abdelnaby, A. (2017). Fragility curves for RC frames under multiple

earthquakes. Soil Dynamics and Earthquake Engineering, 98, 222-234.

Johannesson, B., Yamada, K., Nilsson, L.-O., & Hosokawa, Y. (2007). Multi-species ionic

diffusion in concrete with account to interaction between ions in the pore solution and the

cement hydrates. Materials and Structures, 40(7), 651-665.

Kanda, T., Lin, Z., & Li, V. C. (1998). Application of pseudo strain-hardening cementitious

composites to shear resistant structural elements. Proceedings of FraMCoS-3, Gifu, Japan.

Keserle, G. C., Sanchez, T., Conciatori, D., & Chouinard, L. (2021). Monitoring environmental

and climatic exposure conditions for structures in cold regions. Journal of Cold Regions

Engineering, 35(2), 04021007.

Kirkpatrick, T. J., Weyers, R. E., Anderson-Cook, C. M., & Sprinkel, M. M. (2002). Probabilistic

model for the chloride-induced corrosion service life of bridge decks. Cement and Concrete

Research, 32(12), 1943-1960.

L.-O. Nilsson, P. Sandberg, E. Poulsen, L. Tang, Andersen, A., & Frederiksen, J. M. (1997). A

system for estimation of chloride ingress into concrete, theoretical background, No. 83,

Ministry of Transport, Copenhagen, Denmark.

Lepech, M. D., & Li, V. C. (2009). Water permeability of engineered cementitious composites.

Cement and Concrete Composites, 31(10), 744-753.

Li, V. C. (2003). On engineered cementitious composites: A review of the material and its

applications. Journal of Advanced Concrete Technology, 1(3), 215-230.

160
Li, V. C. (2019). High-performance and multifunctional cement-based composite material.

Engineering, 5(2), 250-260.

Life-365™. (2020). Life-365 service life predication modelTM, Version 2.2.3, Life-365 Consortium.

Retrieved on 09/20 from http://www.life-365.org.

Lindvall, A. (2001). Environmental actions and response – reinforced concrete structures exposed

in road and marine environments, Master thesis, Chalmers University of Technology,

Gothenburg, Sweden.

Lindvall, A. (2003). Environmental actions on concrete exposed in marine and road environments

and its response - consequences for the initiation of chloride induced reinforcement

corrosion, PhD dissertation, Chalmers University of Technology, Gothenburg, Sweden.

Liu, Q., & Su, R. K. L. (2018). A displacement-based inverse analysis method to estimate in-situ

Young’s modulus of steel rust in reinforced concrete. Engineering Fracture Mechanics,

192, 114-128.

Liu, T., & Weyers, R. (1998). Modeling the dynamic corrosion process in chloride contaminated

concrete structures. Cement and Concrete Research, 28(3), 365-379.

Lu, C., Jin, W., & Liu, R. (2011). Reinforcement corrosion-induced cover cracking and its time

prediction for reinforced concrete structures. Corrosion Science, 53(4), 1337-1347.

Lundmark, A., & Jansson, P.-E. (2008). Estimating the fate of de-icing salt in a roadside

environment by combining modelling and field observations. Water, Air, and Soil

Pollution, 195(1), 215-232.

Lundmark, A., & Olofsson, B. (2007). Chloride deposition and distribution in soils along a deiced

highway–assessment using different methods of measurement. Water, Air, and Soil

Pollution, 182(1), 173-185.

161
Lysbakken, K. R. (2013). Salting of winter roads: the quantity of salt on road surfaces after

application, PhD dissertation, Norwegian University of Science and Technology,

Trondheim, Norway.

Mackie, K., & Stojadinović, B. (2001). Probabilistic seismic demand model for California

highway bridges. Journal of Bridge Engineering, 6(6), 468-481.

Mander, J. B., Priestley, M. J., & Park, R. (1988). Theoretical stress-strain model for confined

concrete. Journal of Structural Engineering, 114(8), 1804-1826.

Mauch, M., & Madanat, S. (2001). Semiparametric hazard rate models of reinforced concrete

bridge deck deterioration. Journal of Infrastructure Systems, 7(2), 49-57.

Mazzoni, S., McKenna, F., Scott, M. H., & Fenves, G. L. (2006). OpenSees command language

manual. Pacific Earthquake Engineering Research (PEER) Center, Berkeley, CA, United

States.

McGuire, R. K. (2008). Probabilistic seismic hazard analysis: early history. Earthquake

Engineering and Structural Dynamics, 37(3), 329-338.

McKenna, F., Fenves, G. L., & Scott, M. H. (2000). Open system for earthquake engineering

simulation. University of California, Berkeley, CA, United States. Retrieved on 09/20 from

http://opensees.berkeley.edu.

Meda, A., Mostosi, S., Rinaldi, Z., & Riva, P. (2016). Corroded RC columns repair and

strengthening with high performance fiber reinforced concrete jacket. Materials and

Structures, 49(5), 1967-1978.

Mihashi, H., Ahmed, S. F. U., & Kobayakawa, A. (2011). Corrosion of reinforcing steel in fiber

reinforced cementitious composites. Journal of Advanced Concrete Technology, 9(2), 159-

167.

162
Mitchell, D., Bruneau, M., Saatcioglu, M., Williams, M., Anderson, D., & Sexsmith, R. (1995).

Performance of bridges in the 1994 Northridge earthquake. Canadian Journal of Civil

Engineering, 22(2), 415-427.

Moccia, F., Ruiz, M. F., & Muttoni, A. (2021). Spalling of concrete cover induced by

reinforcement. Engineering Structures, 237, 112188.

Molina, F., Alonso, C., & Andrade, C. (1993). Cover cracking as a function of rebar corrosion:

Part 2 - numerical model. Materials and Structures, 26(9), 532-548.

Moore, M., Phares, B. M., Graybeal, B., Rolander, D., Washer, G., & Wiss, J. (2001). Reliability

of visual inspection for highway bridges, Vol. I, FHWA-RD-01-020, Federal Highway

Administration (FHWA), Washington D.C., United States.

Morcous, G. (2011). Developing deterioration models for Nebraska bridges, SPR-P1(11) M302,

Nebraska Transportation Center, Lincoln, NE, United States.

Muntasir Billah, A., & Shahria Alam, M. (2015). Seismic fragility assessment of highway bridges:

a state-of-the-art review. Structure and Infrastructure Engineering, 11(6), 804-832.

National Association of Corrosion Engineers (NACE) (2013). Highways and bridges. Retrieved

on 01/16 from https://www.nace.org/resources/industries-nace-serves/highways-bridges

Nielson, B. G., & DesRoches, R. (2007). Seismic fragility methodology for highway bridges using

a component level approach. Earthquake Engineering and Structural Dynamics, 36(6),

823-839.

New York Department of Transportation (NYSDOT) (2006). Highway maintenance guidelines:

snow and ice control, NYSDOT, Albany, NY, United States.

New York Department of Transportation (NYSDOT) (2014). NYSDOT Bridge Inspection Manual,

NYSDOT, Albany, NY, United States.

163
New York Department of Transportation (NYSDOT) (2017). NYSDOT Bridge Inspection Manual,

NYSDOT, NY, United States.

Otieno, M., Beushausen, H., & Alexander, M. (2016). Chloride-induced corrosion of steel in

cracked concrete - part I: experimental studies under accelerated and natural marine

environments. Cement and Concrete Research, 79, 373-385.

Otieno, M. B., Beushausen, H. D., & Alexander, M. G. (2011). Modelling corrosion propagation

in reinforced concrete structures - a critical review. Cement and Concrete Composites,

33(2), 240-245.

Padgett, J. E., Nielson, B. G., & DesRoches, R. (2008). Selection of optimal intensity measures in

probabilistic seismic demand models of highway bridge portfolios. Earthquake

Engineering and Structural Dynamics, 37(5), 711-725.

Paulay, T., & Priestley, M. N. (1992). Seismic design of reinforced concrete and masonry buildings.

John Wiley and Sons, Inc., Hoboken, NJ, United States.

Petcherdchoo, A. (2017). Closed-form solutions for bilinear surface chloride functions applied to

concrete exposed to deicing salts. Cement and Concrete Research, 102, 136-148.

Prah-Ennin, P. K. (2013). Quantifying corrosive potential of de-icing and anti-icing solutions to

steel and concrete bridge components, University of North Carolina, Charlotte, Charlotte,

NC, United States.

Priestley, M., (2000). Performance based seismic design. Bulletin of the New Zealand Society for

Earthquake Engineering, 33(3), 325-346.

Ramanathan, K. N. (2012). Next generation seismic fragility curves for California bridges

incorporating the evolution in seismic design philosophy, PhD dissertation, Georgia

Institute of Technology, Atlanta, GA, United States.

164
Ranade, R., Zhang, J., Lynch, J. P., & Li, V. C. (2014). Influence of micro-cracking on the

composite resistivity of engineered cementitious composites. Cement and Concrete

Research, 58, 1-12.

Sahmaran, M., Li, M., & Li, V. C. (2007). Transport properties of engineered cementitious

composites under chloride exposure. ACI Materials Journal, 104(6), 604.

Samson, E., Marchand, J., Robert, J. L., & Bournazel, J. P. (1999). Modelling ion diffusion

mechanisms in porous media. International Journal for Numerical Methods in Engineering,

46(12), 2043-2060.

Slamova, K., Schill, C., Wiesmeier, S., Köhl, M., & Glaser, R. (2012). Mapping atmospheric

corrosion in coastal regions: methods and results. Journal of Photonics for Energy, 2(1),

1-11.

Stefanidou, S. P., & Kappos, A. J. (2017). Methodology for the development of bridge‐specific

fragility curves. Earthquake Engineering and Structural Dynamics, 46(1), 73-93.

Stewart, M. G. (2004). Spatial variability of pitting corrosion and its influence on structural

fragility and reliability of RC beams in flexure. Structural Safety, 26(4), 453-470.

Stewart, M. G., & Al-Harthy, A. (2008). Pitting corrosion and structural reliability of corroding

RC structures: experimental data and probabilistic analysis. Reliability Engineering and

System Safety, 93(3), 373-382.

Stewart, M. G., & Rosowsky, D. V. (1998). Structural safety and serviceability of concrete bridges

subject to corrosion. Journal of Infrastructure Systems, 4(4), 146-155.

Su, R. K. L., & Zhang, Y. (2019). A novel elastic-body-rotation model for concrete cover spalling

caused by non-uniform corrosion of reinforcement. Construction and Building Materials,

213, 549-560.

165
Tang, L. (2008). Engineering expression of the ClinConc model for prediction of free and total

chloride ingress in submerged marine concrete. Cement and Concrete Research, 38(8-9),

1092-1097.

Tang, L., & Utgenannt, P. (2007). Chloride ingress and reinforcement corrosion in concrete under

de-Icing highway environment - a study after 10 years' field exposure, Statens

Provningsanstalt Technical Research Institute of Sweden, Borås, Sweden.

Taucer, F., Spacone, E., & Filippou, F. C. (1991). A fiber beam-column element for seismic

response analysis of reinforced concrete structures. Earthquake Engineering Research

Center, University of California, Berkeley, CA, United States.

Thomas, R., J. Eric Mann, & Chill, Z. M. (2012). Bridge inspector's reference manual, FHWA

NHI 12-049, Federal Highway Administration (FHWA), Washington D.C., United States.

Tondolo, F. (2015). Bond behaviour with reinforcement corrosion. Construction and Building

Materials, 93, 926-932.

Tuutti, K. (1982). Corrosion of steel in concrete, Swedish Cement and Concrete Research Institute,

Stockholm, Sweden.

United States Geological Survey (USGS) (2021). Earthquake hazards. Retrieved on 09/21 from

https://www.usgs.gov/natural-hazards/earthquake-hazards/earthquakes

Val, D. V., & Melchers, R. E. (1997). Reliability of deteriorating RC slab bridges. Journal of

Structural Engineering, 123(12), 1638-1644.

Vamvatsikos, D., & Cornell, C. A. (2002). Incremental dynamic analysis. Earthquake Engineering

and Structural Dynamics, 31(3), 491-514.

Veshosky, D., Beidleman, C. R., Buetow, G. W., & Demir, M. (1994). Comparative analysis of

bridge superstructure deterioration. Journal of Structural Engineering, 120(7), 2123-2136.

166
Vidal, T., Castel, A., & Francois, R. (2004). Analyzing crack width to predict corrosion in

reinforced concrete. Cement and Concrete Research, 34(1), 165-174.

Vu, K. A. T., & Stewart, M. G. (2000). Structural reliability of concrete bridges including

improved chloride-induced corrosion models. Structural Safety, 22(4), 313-333.

Wang, S., & Li, V. C. (2007). Engineered cementitious composites with high-volume fly ash. ACI

Materials Journal, 104(3), 233.

Wang, Y., Liu, C., Wang, Y., Li, Q., & Yan, B. (2020). Semi-empirical prediction model of

chloride-induced corrosion rate in uncracked reinforced concrete exposed to a marine

environment. Electrochimica Acta, 331, 135376.

Weather-Spark. (2021). Average weather in Göteborg. Retrieved on 02/21 from

https://weatherspark.com/y/71566/Average-Weather-in-G%C3%B6teborg-Sweden-Year-

Round

Weir, D. H., Strange, J. F., & Heffley, R. K. (1978). Reduction of adverse aerodynamic effects of

large trucks, Vol. I, FHWA-RD-79-84, Federal Highway Administration (FHWA),

Washington D.C., United States.

Weyers, R. E., Fitch, M., Larsen, E., Al-Qadi, I., Chamberlin, W., & Hoffman, P. (1994). Concrete

bridge protection and rehabilitation: chemical and physical techniques - service life

estimates, SHRP-S-668, Strategic Highway Research Program, National Research Council,

Washington D.C., United States.

Wolofsky, R. (2011). Corrosion initiation of concrete bridge elements exposed to de-icing salts

Master thesis, McGill University, Montreal, Canada.

Xi, Y., Jing, Y., & Railsback, R. (2018). Surface chloride levels in Colorado structural concrete ,

CDOT-2018-05, Colorado Department of Transportation, Denver, CO, United States.

167
Yamazaki, F., Hamada, T., Motoyama, H., & Yamauchi, H. (2000). Earthquake damage

assessment of expressway bridges in Japan. Proceedings of 12th World Conference on

Earthquake Engineering, Auckland, New Zealand.

Yanev, B., & Chen, X. (1993). Life-cycle performance of New York City bridges. Transportation

Research Record, 1389, 17.

Yuan, Y., Jiang, J., & Peng, T. (2010). Corrosion process of steel bar in concrete in full lifetime.

ACI Materials Journal, 107(6), 562.

Yuan, Z., Fang, C., Parsaeimaram, M., & Yang, S. (2017). Cyclic behavior of corroded reinforced

concrete bridge piers. Journal of Bridge Engineering, 22(7), 04017020.

Zhang, X., Zhang, Y., Liu, B., Liu, B., Wu, W., & Yang, C. (2021). Corrosion-induced spalling of

concrete cover and its effects on shear strength of RC beams. Engineering Failure Analysis,

127, 105538.

Zhang, Y., DesRoches, R., & Tien, I. (2019). Impact of corrosion on risk assessment of shear-

critical and short lap-spliced bridges. Engineering Structures, 189, 260-271.

Zhang, Y., Zhu, Y., Yeseta, M., Meng, D., Shao, X., Dang, Q., & Chen, G. (2019). Flexural

behaviors and capacity prediction on damaged reinforcement concrete (RC) bridge deck

strengthened by ultra-high performance concrete (UHPC) layer. Construction and Building

Materials, 215, 347-359.

Zhu, Y., Zhang, Y., Hussein, H. H., & Chen, G. (2020). Numerical modeling for damaged

reinforced concrete slab strengthened by ultra-high performance concrete (UHPC) layer.

Engineering Structures, 209, 110031.

168
 ProQuest Number: 28963803

INFORMATION TO ALL USERS

The quality and completeness of this reproduction is dependent on the quality
and completeness of the copy made available to ProQuest.

Distributed by ProQuest LLC ( 2022 ).

Copyright of the Dissertation is held by the Author unless otherwise noted.

This work may be used in accordance with the terms of the Creative Commons license
or other rights statement, as indicated in the copyright statement or in the metadata
associated with this work. Unless otherwise specified in the copyright statement
or the metadata, all rights are reserved by the copyright holder.

This work is protected against unauthorized copying under Title 17,

United States Code and other applicable copyright laws.

Microform Edition where available © ProQuest LLC. No reproduction or digitization

of the Microform Edition is authorized without permission of ProQuest LLC.

ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346 USA