You are on page 1of 12

Life-Cycle Resilience of Aging Bridges under Earthquakes

B Sharanbaswa Vishwanath1; and Swagata Benerjee, A.M.ASCE2

Abstract: A quantitative framework was developed to assess life-cycle resilience of deteriorating reinforced concrete (RC) bridges under
seismic ground motions (GMs). Deterioration of a three-span RC bridge due to chloride-induced corrosion was considered, and three-
dimensional (3D) finite-element (FE) models of the bridge were generated at pristine and four degraded conditions. Nonlinear time history
analyses of bridge models were performed to develop time-variant bridge fragility curves, which were utilized further to estimate post-
Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

earthquake direct and indirect losses arising from bridge damage and associated downtime. Sigmoidal functions were used to numerically
simulate realistic recovery paths depending on the extent of bridge damage after earthquakes. Finally, bridge resilience was estimated by
integrating the bridge’s seismic vulnerability, losses, and recovery functions. Obtained results revealed a degrading trend of resilience of the
bridge as it aged, signifying the importance of considering a life cycle–oriented framework for estimating time-variant resilience of aging
bridges. At a specific state of degradation, bridge resilience was observed to reduce linearly or nonlinearly with increasing GM intensity or peak
ground acceleration (PGA). Finally, an uncertainty analysis with key uncertain input parameters showed resilience of the pristine bridge at any
specific PGA to vary following a normal distribution. DOI: 10.1061/(ASCE)BE.1943-5592.0001491. Ó 2019 American Society of Civil
Engineers.
Author keywords: Bridge; Life-cycle resilience; Vulnerability; Corrosion; Loss; Earthquake.

Introduction Decò and Frangopol 2013; Ou et al. 2013; Akiyama and Frangopol
2014; Rokneddin et al. 2014; Biondini et al. 2015; Alipour and
Corrosion is a major factor of bridge aging and degradation with Shafei 2016; Ni Choine et al. 2016; Dong and Frangopol 2016;
time. Statistics indicate that about one-sixth of bridges in the United Kezhiyur et al. 2016; and Thanapol et al. 2016). The overall ob-
States are either in poor condition or structurally deficient, and about servation from all these past studies was that corrosion poses greater
40% of bridges have crossed 50 years in their service lives (FHWA threats to bridges due to higher bridge seismic vulnerability at cor-
2018). The scenario is not much different in other parts of the world. roded condition than that at pristine condition. This is an obvious
As the existing bridge infrastructure continues to age, it becomes outcome because corrosion results in reduction of steel area, deg-
increasingly more susceptible to damage under extreme conditions. radation of bond strength between concrete and steel bars (Lee et al.
Certainly, corrosion does not always cause failure of aging bridges, 2002), and concrete spalling, leading to further exposure of rein-
but it contributes to an accumulated risk, particularly when main- forcement bars in reinforced concrete (RC) members to the corro-
tenance is not done in a timely manner. The worldwide failure in- sive environment (Coronelli and Gambarova 2004). Apart from this
ventory of bridges provides abundant examples of collapse of in- direct threat of corrosion on overall health of bridges, corroded
service aging bridges under normal circumstances and due to natural bridge components take part in contaminating groundwater.
disasters such as earthquakes and floods. One such example is the The previously mentioned studies on seismic performance as-
failure of the old Surajbari Bridge in India during the Bhuj Earth- sessment of corroded bridges made meaningful advancements in
quake of 2001. The bridge, built in late 1960s, was found to be in this field of research; however, not much knowledge of seismic
bad condition prior to the earthquake due to significant corrosion of resilience of aging bridges is available now. With resilience rec-
rebar in an aggressive environment of salt water combined with high ognized as an important indicator of bridge performance that ac-
temperatures (Madabhushi and Haigh 2005). The bridge’s deterio- counts for vulnerability, losses, and postdisaster recovery measures,
rated condition accelerated its failure in 2001 during the earthquake. research is indeed needed to comprehend the impact of corrosion on
Extensive research has been performed over the last decade to life-cycle resilience of bridges under seismic hazard. Thus far, the
investigate the performance of corroded bridges and aging highway issue has broadly been addressed in Biondini et al. (2015) and Dong
transportation systems under seismic hazard (examples include and Frangopol (2016). Considering a representative corroded RC
Choe et al. 2009; Ghosh and Padgett 2010; Simon et al. 2010; bridge, Biondini et al. (2015) estimated seismic resilience of the
Alipour et al. 2011; Gardoni and Rosowsky 2011; Zhong et al. 2012; bridge for its 50-year life span in which the bridge’s time-variant
functionality indicator was expressed in terms of its base shear ca-
pacity and displacement ductility at various time instances. The
1
Doctoral Student, Dept. of Civil Engineering, Indian Institute of Tech- process, however, did not calculate any physical loss arising from
nology Bombay, Powai, Mumbai 400076, India. Email: sharan.baswa@iitb bridge damage and inaccessibility owing to seismic events. The
.ac.in analytical framework proposed by Dong and Frangopol (2016) for
2
Associate Professor, Dept. of Civil Engineering, Indian Institute of estimating seismic resilience of a corroding RC bridge incorporated
Technology Bombay, Powai, Mumbai 400076, India (corresponding au- seismic vulnerability (in the form of fragility curves), losses, and
thor). Email: swagata@civil.iitbac.in
Note. This manuscript was submitted on November 2, 2018; approved
recovery models. The study, however, did not perform any case-
on June 5, 2019; published online on September 6, 2019. Discussion period specific fragility analysis and considered fragility curves from
open until February 6, 2020; separate discussions must be submitted for HAZUS-MH MR4 (HAZUS) software developed by FEMA (2003)
individual papers. This paper is part of the Journal of Bridge Engineering, for the pristine condition of the bridge. Moreover, the use of a simple
Ó ASCE, ISSN 1084-0702. recovery model in that study signifies the need for a robust, ap-
© ASCE 04019106-1 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


propriate, and easy to apply recovery model for life cycle resilience t0EZþTLC
QðtÞ
assessment of aging bridges. R¼ dt ð1Þ
With an objective of developing a quantitative framework for TLC
t0E
estimating life-cycle resilience of deteriorating bridges under possi-
ble seismic hazard scenarios, the present study analyzed a three-span where Q(t) ¼ time-dependent functionality, which can be defined
RC bridge with corrosion deterioration over a period of 100 years. for a time instant t as follows:
Three major components of this framework—vulnerability, losses,
and recovery—are discussed. Seismic fragility analysis of the bridge QðtÞ ¼ 1  ½LðI; Trec Þ  fHðt  t0E Þ  Hðt  ðt0E þ Trec ÞÞg
was performed at both pristine and degraded conditions (i.e., at 25,
50, 75, and 100 years of deterioration) to generate a time-variant  frec ðt; t0E ; Trec Þ ð2Þ
vulnerability model of the bridge as it aged. Postearthquake losses
arising from bridge damage and associated downtime were estimated Here L(I,Trec) ¼ loss function; H() ¼ Heaviside step function;
in terms of direct and indirect losses incorporating probable pre- and and frec ¼ recovery function that depends on recovery time Trec,
postevent traffic information on the bridge. The recovery model the time required to recover functionality of a damaged bridge for
Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

constitutes a sigmoidal function that is capable of producing different earthquake event E. A detailed discussion of postevent loss as-
recovery patterns according to the extent of bridge damage and sessment and recovery in a distressed highway network is presented
associated idle time for postevent recovery actions. Finally, an un- in Banerjee et al. (2019). It is found that in the majority of cases,
certainty analysis was conducted for the as-built bridge to obtain postevent loss estimation requires information on bridge seismic
possible variability of seismic resilience due to uncertain parameters vulnerability. Therefore, time-variant seismic vulnerability of de-
in the constituent models (i.e., vulnerability, losses, and recovery). grading bridges is discussed first in this article, followed by dis-
Outcome of this research not only shows the importance of consid- cussions of loss and recovery functions for seismic resilience as-
ering a life cycle–oriented approach for estimating seismic resilience sessment of bridges.
of aging bridges but also identifies the need for bridge maintenance
and provides a platform for future research aiming at the development
of strategic plans for scheduling bridge maintenance interventions. Seismic Vulnerability of Degrading Bridges
Chloride-induced corrosion is one of the major sources of deterio-
ration of concrete bridges, mainly from the application of deicing
Seismic Resilience of Degrading Bridges salt and airborne chlorine in the marine environment (Neville 1995;
Enright and Frangopol 1998). It has two major phases: corrosion
The concept of resilience has evolved over the last four decades in
initiation and propagation (Stewart and Rosowsky 1998). For ini-
diverse fields of research and has been applied at individual, com-
tiation, the diffusion process of chloride ion penetration in an RC
munity, and organizational levels. For bridges and bridge networks
member can be modeled by Fick’s second law of diffusion and is
under single and multiple hazards, a systematic and comprehensive
empirically expressed as (Stewart and Rosowsky 1998)
review of resilience assessment is presented in Banerjee et al.
(2019). Resilience of a bridge under seismic excitations can be
vCðx; tÞ v2 Cðx; tÞ
calculated from the normalized area under the functionality curve ¼D ð3Þ
between time of occurrence of the seismic event (t0E) and a time vt vx2
instant, t0E þ TLC, as shown in Fig. 1. Here, TLC is called controlled In this, C(x,t) ¼ the chloride ion concentration up to a depth x from
time, a time set by stakeholders to recover the functionality of the the surface of a concrete member under corrosion at t years after
bridge, partially or fully, after the event. Previous studies observed construction, and D ¼ the apparent diffusion coefficient of the
that resilience significantly depends on residual functionality of concrete medium. Corrosion starts just after the corrosion initiation
bridges after an earthquake event and the process of postevent time, Ti (when chloride concentration on the surface of reinforce-
functional recovery over time. Mathematically, seismic resilience is ment bars exceeds the critical chloride content, Ccr). Parameters
expressed as (Cimellaro et al. 2010) to calculate Ti depend on exposure condition and location of the
bridge under consideration. The current study considers the bridge
exposure condition from Enright and Frangopol (1998), in which
the effect of corrosion due to deicing salt on flexural strength of
T-beams in a concrete bridge girder was studied. Considering the
uncertain nature of the corrosion initiation process, this previous
study used appropriate probability distributions of input variables
and performed Monte Carlo simulations for Ti. Simulation results
showed Ti to follow a lognormal distribution with mean value of
15.84 years and standard deviation of 8.23 years. Following this
observation, the current study uses the value of Ti equal to 15 years,
considering the same bridge exposure condition and depth of con-
crete cover. Assuming uniform corrosion of bridge piers, Ti is used
to calculate the area reduction of reinforcement bars due to corrosion
degradation as expressed in Eq. (4) (Thoft-Christensen et al. 1996)
8 p
>
> n d2 for 0  t  Ti
< 4 i
AðtÞ ¼ ð4Þ
Fig. 1. Schematic representation of seismic resilience of degrading > 2
bridges. : np½di  rcorr ðt  Ti Þ
>
for t > Ti
4
© ASCE 04019106-2 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


In this, A(t) ¼ area of rebars at time t years after construction; n ¼ as earthquake strikes. Due to damage, reduced functionality of
total number of reinforcing bars susceptible to corrosion; di ¼ initial bridges not only causes economic liability to the community
diameter in millimeters of these reinforcement bars; and rcorr ¼ rate for postdisaster repair and rehabilitation but also hinders mobility
of corrosion. A uniform rate of corrosion over the bridge life cycle is in everyday life. Owing to bridge damage, the total expected
considered here (Frangopol et al. 1997). loss is defined in the literature in two categories—direct and
Fragility curves are developed to capture the effect of corrosion indirect.
degradation on seismic performance of bridges. These curves pro- Direct loss (LD) defines the monetary value associated with
vide probabilistic estimation of bridge seismic damageability. Fol- postdisaster repair of damaged bridge components. It can be com-
lowing the traditional definition, fragility curves are expressed in the puted by multiplying the probability of bridge damage at a damage
form of two-parameter lognormal distribution functions (Banerjee state (obtainable from fragility curves) with a damage ratio corre-
and Shinozuka 2007) as follows: sponding to that damage state (FEMA 2003; Zhou et al. 2010). The
   damage ratio is taken as 0.03, 0.08, 0.25, and 0.67, respectively,
ln aj =ck for minor damage, moderate damage, major damage, and collapse
Pðaj ; ck ; fk Þ ¼ F ð5Þ
fk states of a bridge (FEMA 2003). This well-accepted approach
Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

for direct loss assessment has been used in a number of studies


Here P() ¼ probability of exceeding a damage state k; aj ¼ peak on seismic resilience assessment of bridges (refer to Banerjee
ground acceleration (PGA) of jth ground motion (GM), and ck and et al., 2019, for further information). Note that LD represents a
fk ¼ fragility parameters that represent median and log-standard percentage of bridge reconstruction cost and hence provides an
deviation at kth damage state of the bridge, respectively. The vari- immediate idea of losses incurred in comparison with reconstruc-
able k takes values equal to 1–4 based on four bridge damage states, tion cost.
namely, minor damage, moderate damage, major damage, and col- Besides direct loss, bridge damage results in traffic conges-
lapse, under seismic excitations. The current study developed fra- tion. Vehicle users need to take detours to complete trips, which
gility curves for all four damage states of a bridge under consider- leads to higher operation cost of vehicles. Thus, indirect loss (LI)
ation at its nondegraded (i.e., pristine) and degraded conditions. involves, for example, traffic delay, business interruption, income
These curves were compared to measure the variation in bridge loss, and relocation expense, and its estimation requires pre- and
fragility from nondegraded to degraded condition to identify the postevent traffic information on a bridge under consideration.
enhanced seismic vulnerability of the bridge due to corrosion. The current study considers indirect loss in two categories: op-
erating cost of vehicles on detour (Cop) and cost due to vehicle
time loss (CTL) owing to bridge damage (Dong and Frangopol
Loss Functions
2015; Zheng et al. 2018). Operating cost of vehicles (Cop) due to
Along service lives of bridges, economic losses are imminent detour is expressed as (Stein et al. 1999; Dong and Frangopol
due to bridge damage and downtime when a sudden disaster such 2015)

X  
TRD

TRD

Cop ¼ PE ðDS ¼ kÞ  Cop;car 1  þ Cop;truck  Dl  ADT ð6Þ
k
100 100

where Cop,car and Cop,truck ¼ average costs of operation of cars and trucks, respectively, in dollars per kilometer; Dl ¼ detour length in
kilometers; ADT ¼ average daily traffic; and TRD ¼ average daily truck traffic ratio as a percentage. Cost due to time loss (CTL) of vehicle
users and goods traveling through detours can be given as (Stein et al. 1999)

X  
TRD

  TRD
 
Dl

l l

CTL ¼ PE ðDS ¼ kÞ  CAW Ocar 1  þ CATC Otruck þ Cgoods  ADT þ ADE  ð7Þ
k
100 100 S SD S0

where cost terms CAW, CATC, and Cgoods ¼ average wage in dollars loss function L(I,Trec) in Eq. (2) can be computed by adding direct
per hour, average total compensation in dollars per hour, and loss (LD) and indirect loss (LI).
monetary value of time taken to transport goods in cargo in dollars
per hour, respectively; Ocar and Otruck ¼ average vehicle occupan-
Recovery Functions
cies for cars and trucks, respectively; l ¼ length of link in kilo-
meters; S ¼ average speed on detour in kilometers per hour; SD and The recovery function represents the recovery path for restoring
S0 ¼ average speeds on the damaged and intact bridge in kilometers pre-event functionality of bridges. It is achieved, partially or fully,
per hour, respectively; and ADE ¼ average daily traffic remaining by measures such as removing debris, sealing minor cracks, patch-
on the bridge after the event. Values of these parameters are case ing concrete, and repairing or replacing damaged bridge compo-
specific; these are discussed subsequently in the discussion of a nents. The literature has considered continuous or discrete functions
bridge scenario. to realistically simulate the actual functional recovery of bridges
The cost terms are multiplied with bridge damage probabilities at after earthquakes. While a time-dependent continuous recovery
various damage states [per Eqs. (6) and (7)], and their summation function presents day-to-day progress of bridge functional recovery
represents total indirect cost. This cost term is then divided by after an earthquake (Venkittaraman and Banerjee 2014; Dong and
bridge replacement cost in order to obtain the normalized indirect Frangopol 2016; Zheng et al. 2018), a discrete function refers to
loss (LI) (i.e., percent value of bridge replacement cost). Finally, the important phases along bridge recovery paths (Bocchini et al. 2014;
© ASCE 04019106-3 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


Sharma et al. 2017). In both cases, it can also happen that the re- the bridge recovery function frec can be expressed in terms of a
covered bridge performs better than the predamaged bridge, par- sigmoidal recovery function as
ticularly when gradual degradation occurs due to aging, as in the
current study. 1
frec ðtÞ ¼ ð8Þ
Venkittaraman and Banerjee (2014) used standard mathematical 1 þ eaðtbÞ
expressions, such as linear, exponential, and trigonometric, as re-
covery functions to bridges, and they observed the linear recovery where t ¼ time instant and a and b ¼ parameters previously men-
function to be the most reasonable among the three when no infor- tioned. As shown in Fig. 2, the rapidity of a recovery curve is
mation related to postearthquake restoration techniques for bridges signified by a, and inflation point b controls the movement of the
is available. Based on expert opinion survey data from ATC (1985), curve along the x-axis. The current study evaluates these shape
FEMA (2003) developed recovery curves in the form of normal- parameters a and b in terms of recovery time Trec; this facilitates
distribution functions for bridges at various damage states. Even development of damage state–specific recovery curves and elimi-
though the distributions are associated with high standard deviations nates uncertainty related to the shape of these curves once recovery
that may eventually lead to high variability in estimated resilience, paths are decided. An additional advantage of using a sigmoidal
Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

the model has been used in past studies on seismic resilience as- function for the bridge recovery model is the consideration of idle
sessment of bridges (Dong and Frangopol 2016; Zheng et al. 2018). time in the recovery process as a part of the function. In a realistic
A continuous sinusoidal function, proposed by Decò et al. (2013) scenario, idle time may depend on the extent of bridge damage. In a
and later used by Dong and Frangopol (2015), was derived based on previous study, idle time was assumed to be uniformly distributed
six parameters, namely, residual functionality, idle time, recovery between 1 and 2 months (Decò et al. 2013). The proposed model in
duration, target functionality, and two shape parameters. Uncer- the current study considers idle time equal to one-tenth of Trec such
tainty involved with the recovery process is captured through this that only 10% of bridge functional recovery is achieved by then at
model by assigning probability distributions to model parameters. any bridge damage state except for minor damage. At minor dam-
Although this model is capable of producing common bridge re- age, idle time is ignored and a linear recovery function is consid-
covery patterns, values of model parameters are location specific and ered, as the extent of bridge damage is minimal at this stage and
require calibration for bridges under consideration. Hence, this recovery is expected to be quick.
model may also lead to high uncertainty in estimated resilience if For four bridge damage states, parameters a and b are calculated
required information is not available in a real-life scenario. according to respective values of Trec and realistic recovery curves
In the current study, a sigmoidal function (also called a logistic are obtained as shown in Fig. 3. Here, the expected value of Trec at a
function) is considered to represent the recovery path of damaged damage state is taken to be equal to the time at which 84% func-
bridges. Such a function essentially provides an S-shaped curve tional recovery is achieved at the same damage state from HAZUS-
(Fig. 2) and has been used widely in management to project per- specified restoration curves for highway bridges. In doing so, a full
formance growth of manufacturing and service processes (Jónás correspondence is established between mean and three standard
2007). Sigmoidal functions are also used as growth models in deviation values of Trec at a damage state with HAZUS-specified
various disciplines, such as biology, economics, and demography time for 100% functional recovery at the same damage state.
(Carrillo and González 2002), and as an activation function in ar- Therefore, expected values of Trec equal to 1.2, 5.2, 117, and
tificial neural networks (Chen and Cao 2009). The main advantage 340 days were obtained at minor damage, moderate damage, major
of such a function is that it can grow without bounds or be limited to damage, and collapse states of bridges, respectively, and used for
restrict growth. The same concept is used here to project postevent deriving recovery curves (Fig. 3). Corresponding standard deviation
recovery of bridges over recovery periods. The function is con- values are 0.35, 1.9, 24.2, and 63 days, respectively. These standard
trolled by two parameters (a and b) that influence its shape and deviations of Trec are used subsequently in this article in analyzing
make it capable of producing different recovery paths depending on uncertainty in estimated seismic resilience of bridges.
the nature of bridge damage, probable time to mobilize recovery
actions, and time to complete the recovery process. Mathematically,
Time-Variant Seismic Resilience
of a Corroded Bridge

Description and Seismic Response


Analysis of the Bridge
A three-span RC representative bridge, designed in accordance with
AASHTO’s (2014) LRFD bridge design specifications, was se-
lected to evaluate its life-cycle resilience under seismic hazard. The
bridge was assumed to be located in the state of Washington, United
States, in order to obtain location-specific information on seismic
hazard and traffic flow on the bridge. Bridge corrosion due to the
application of deicing salt is an important issue there because the
Washington State Department of Transportation (WSDOT) needs
to do winter maintenance of roadways and bridges every year
(WSDOT 2017). A recent study by Yilmaz et al. (2018) provided
statistics on geometric attributes of WSDOT bridges. Following that
closely, the three-span bridge considered here was assumed to be
86 m long with a 36-m-long intermediate span and two 25-m-long
Fig. 2. General representation of a sigmoidal curve as a postevent
end spans (Fig. 4). The bridge has a four-cell prestressed box girder
bridge recovery function.
with a 17  2-m cross-sectional dimension. Both ends of the girder
© ASCE 04019106-4 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

(c) (d)

Fig. 3. Recovery models at different bridge damage states: (a) minor damage; (b) moderate damage; (c) major damage; and (d) collapse.

have seat-type abutments that are supported on elastomeric bear- depths, soil resistance values for p-y and t-z springs were calculated
ings. Each pier bent consists of two 8-m-long, 1.2-m-diameter piers using the American Petroleum Institute (API 2003) guidelines and
connected monolithically with the girder. Bridge foundations at pier Mosher’s (1984) relation, respectively.
bents have pile groups, each having a pile cap. Each pile group Seat-type abutments at both ends of the bridge girder have two
consists of eight 0.9-m-diameter piles, as shown in Fig. 4. major components for modeling—abutment response in the longi-
A 3D finite-element (FE) model of the bridge was developed tudinal and transverse directions and response of elastomeric
using Open System for Earthquake Engineering Simulation (Open- bearing in all three translational directions. The modeling technique
Sees) software version 2.4.0. The modeling technique from Yilmaz for bidirectional abutment–soil interaction closely follows that
et al. (2016) was adopted with required changes to capture time- adopted by Yilmaz et al. (2016). The backwall-backfill interaction
dependent degradation of the bridge due to corrosion. The bridge of abutments in the longitudinal direction was modeled with elas-
girder was modeled using an elastic beam-column element with toplastic gap elements. Stiffness properties of these elements were
mass concentrated at nodes along the centroid (Fig. 4). This ele- calculated following Caltrans (2010) guidelines. A similar element,
ment was expected to have linear elastic behavior under earthquake except for the gap definition, was utilized to represent abutment
excitations. stiffness in the transverse direction. Elastic–perfectly plastic mate-
Bridge piers are critical vulnerable components of the bridge. rial models were also used to capture nonlinear response of abut-
These elements were modeled to capture nonlinear behavior using a ment bearings in two horizontal translational directions. Corre-
displacement beam-column element along with fiber sections. Cross sponding stiffness values were calculated following Nielson (2005).
section of a bridge pier and reinforcement detailing are shown in Bearing response in the vertical direction was modeled with linear
Fig. 4. OpenSees Concrete07 and Steel02 material models were used elastic elements, and corresponding stiffness values were calculated
to define properties of concrete and steel reinforcement, respectively. following Warn et al. (2007).
Foundation piles are designed as capacity-protected elements. The modeling technique was used to generate five bridge models
These elements were modeled using elastic beam-column elements at t ¼ 0 (i.e., pristine), 25, 50, 75, and 100 years after construction.
in OpenSees software. A series of p-y and t-z springs were assigned In the degraded models, gradual deterioration of the bridge due to
along the length of each pile to represent nonlinear interactions of corrosion was captured by changing the diameter of longitudinal
piles with the surrounding soil. While p-y springs are used to rep- reinforcements of bridge piers as discussed in Eq. (4). A previous
resent soil resistance in two horizontal directions, t-z springs capture study on time-dependent fragility assessment of bridges observed
vertical soil resistance (shaft resistance) along the pile length that corrosion of an RC bridge girder has little influence on seismic
(Fig. 4). These two springs were modeled with material models vulnerability of the bridge (Ghosh and Padgett 2010). Following
PySimple1 and TzSimple1 from the OpenSees material library and this observation, the current study does not incorporate deterioration
applied at uniform intervals of 0.4 m along pile lengths. At various of bridge components other than piers in the FE model. It is un-
© ASCE 04019106-5 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Schematic representation of a bridge and FE modeling of bridge components. (1) ¼ stiffness of abutment back wall with gap property; (2) ¼
bearing stiffness in the longitudinal direction; (3) ¼ abutment wing wall stiffness; (4) ¼ bearing stiffness in the transverse direction; (5) ¼ vertical
bearing stiffness; MN ¼ master node; SN ¼ slave node.

© ASCE 04019106-6 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


derstood, however, that cumulative deterioration of bridge com- threshold values, and damage states of bridge components were
ponents that are apparently noncritical for bridge seismic perfor- obtained. The process was repeated for all five bridge models at
mance assessment may result in localized damage leading to bridge various degraded and nondegraded conditions and for all seismic
failure during a strong seismic shaking. GMs. For all cases, threshold limits of EDPs as given in Table 1
Modal analysis of the bridge at pristine condition (i.e., at t ¼ 0) remained unchanged. Having same threshold values of EDPs helps
revealed that the translational mode of vibration in the longitudinal in observing increasing damage (and hence higher damage states) of
direction was the fundamental and governing mode of vibration of bridge components as the bridge deteriorates and becomes more
the bridge. Mode shapes with time periods in the first four modes of seismically vulnerable with the progression of corrosion deteriora-
the bridge are shown in Fig. S1. With t, modal time periods in- tion along the bridge life span.
creased as bridge corrosion progressed. This observation is in ac- Obtained damage state information of various bridge compo-
cordance with the literature (Alipour et al. 2011). Mode shapes nents and corresponding PGA values of GMs are used in Eq. (5) to
remained unchanged with progressing corrosion degradation. generate component-level fragility curves for piers, abutments, and
Nonlinear time history (NLTH) analysis of the bridge was per- bearings at pristine and four degraded conditions of the bridge.
formed for 50 GMs obtained from existing sources for the state of These component fragilities were combined to form system-level
Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

Washington. Among these, 40 GMs were originally generated for fragility curves of the bridge, as shown in Fig. 5. Corresponding
the Seattle region as a part of FEMA=SAC Steel Project (Somerville median values are listed in Table 2. A log-standard deviation of 0.5
et al. 1997). The other 10 GMs were obtained from the Center for was used to avoid intersection between fragility curves. As can be
Engineering Strong Motion Data (CESMD 2018) for the state of seen in Fig. 5, there was a small increase in bridge-failure proba-
Washington and neighboring regions. A summary of these motions bility in the first 25 years of corrosion, since corrosion initiates in
is given in Table S1. Each motion was composed of two horizontal the bridge after 15 years. Beyond 25 years, bridge seismic vulner-
orthogonal components, and their PGA values covered a wide ability increases rapidly at all damage states. Such an increase in a
range, which is desirable for fragility analysis. NLTH analyses were bridge’s seismic vulnerability over its lifetime signifies the impor-
performed by applying both components of a motion simulta- tance of estimating bridge seismic resilience in a life-cycle context.
neously, and these components were interchanged to avoid direc-
tional uncertainty in the analysis result. The bridge was analyzed at
pristine and four degraded conditions to generate seismic fragility Life-Cycle Seismic Resilience of the Bridge
curves at all conditions. Resilience of the studied bridge at different phases along its life span
was calculated for various seismic events having PGAs ranging over
0.2–1.0 g. For an earthquake with a known PGA, probabilities of
Time-Variant Fragility Curves of the Bridge
exceedance at all four damage states of the bridge were obtained
Bridge responses at critical components such as piers, abutments, from fragility curves (Fig. 5), and resulting direct and indirect losses
and bearings were recorded from each time history analysis. Extent were calculated as discussed in preceding paragraphs. For calcu-
of damage in these critical components under seismic excitation was lating indirect loss, expected values of traffic-related parameters and
measured in terms of curvature ductility of piers (lu) and longitu- relevant distributions are summarized in Table 3. Here, average
dinal and transverse deformations of abutments (Da,long and Da,trans) daily traffic remaining on the damaged bridge after the event (ADE)
and elastomeric bearings (Db,long and Db,trans). These response is calculated based on the concept presented in Mackie and
quantities are referred to as engineering demand parameters (EDPs). Stojadinovic (2006) and Dong and Frangopol (2015) related to the
Yilmaz et al. (2016) provided details on obtaining the threshold traffic capacity remaining on bridges after seismic events. The traffic
limits of various EDPs under seismic excitations to define seismic flow is uninterrupted (i.e., all lanes are open) or at 100% when there
damage states of critical bridge components. While threshold limits is no damage to the bridge. At minor damage, traffic flow is assumed
for lu were obtained following Ramanathan (2012), the same for to be reduced by 25% (through weight restriction) until the com-
Da,long and Da,trans were calculated based on observations from pletion of postevent inspection and removal of debris, if any
multiple studies (Shamsabadi et al. 2007; Aviram et al. 2008; (FHWA 2006). In the same way, 50% (one lane open only) and 75%
Caltrans 2010). For obtaining threshold limits of Db,long and Db,trans, (emergency access only) restrictions on traffic flow on the bridge are
engineering judgments were made on the basis of observed values considered, respectively, for moderate and major damage states of
from Konstantinidis et al. (2008). Table 1 lists these threshold EDP the bridge. The bridge is completely shut down to traffic when it
values, with respective references, as applicable to the bridge con- experiences collapse. Based on these traffic flow assumptions, ADE
sidered in the current study. Obtained maximum values of EDPs on the bridge at various bridge damage scenarios was estimated.
from each time history analysis were compared with respective Note that these assumed values of traffic flow and speed reduction

Table 1. Threshold limits of different damage states


Component EDP Minor damage Moderate damage Major damage Collapse Reference
Piers Curvature ductility 1  lu < 4 4  lu < 8 8  lu < 12 lu  12 Ramanathan (2012)
Abutment Longitudinal 50  Da,long < 100 Da,long  100 — — Shamsabadi et al. (2007);
deformation (mm) Aviram et al. (2008);
Transverse 50  Da,trans < 150 Da,trans  150 — — Caltrans (2010)
deformation (mm)
Elastomeric Longitudinal 40  Db,long < 86 86  Db,long < 310 310  Db,long < 533 Db,long  533 a

bearing deformation (mm)


Transverse 40  Db,trans < 86 86  Db,trans < 310 — — a

deformation (mm)
a
Based on engineering judgments and observations from Konstantinidis et al. (2008).

© ASCE 04019106-7 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

(c) (d)

Fig. 5. Fragility curves of the aging bridge at different damage states: (a) minor damage; (b) moderate damage; (c) major damage; and (d) collapse.

Table 2. Median fragility parameters (in g) at various damage states and on the damaged bridge enabled us to study the influence of flow
degraded conditions of the bridge reduction on postevent indirect loss and to calculate life-cycle
Years after Minor Moderate Major seismic resilience of the bridge. Accuracy of these values is not
construction damage damage damage Collapse examined here and is beyond the scope of this study.
0 0.29 0.43 1.06 1.51 Direct and indirect losses are combined to generate the loss ratio
25 0.25 0.39 0.90 1.29 L(I,Trec) of the bridge. Fig. 6 shows the variation of loss ratio of the
50 0.20 0.31 0.71 0.95 bridge for a wide range of seismic hazard and for various degraded
75 0.16 0.27 0.54 0.75 conditions of the bridge. Variations may follow nearly linear or
100 0.12 0.22 0.42 0.63 nonlinear patterns, depending on the bridge’s degraded condition.

Table 3. Parameters associated with indirect loss assessment of the bridge


Random variables
Parameter Mean COV Distributiona Deterministic Reference
Operation cost of car (Cop,car) ($=km) 0.4 0.2 LN — AASHTO (2003)
Operation cost of truck (Cop,truck) ($=km) 0.57 0.2 LN — AASHTO (2003)
Average wage per hour (CAW) ($=h) 13.12 0.3 LN — US Dept. of Labor (for WA) (2018)
Average total compensation per hour (CATC) ($=h) 22.21 0.3 LN — US Dept. of Labor (for WA) (2018)
Time value of goods transported in cargo (Cgoods) ($=h) 3.81 0.2 LN — AASHTO (2003)
Average detour speed (S) (km=h) 50 0.2 LN — Dong and Frangopol (2015)
Average speed on damaged link (SD) (km=h) 20 0.2 LN — Assumed
Average speed on intact link (S0) (km=h) 80 0.2 LN — Dong and Frangopol (2015)
ADT — — — 29,000 WSDOT (2016)
Daily truck ratio (TRD) (ADTT=ADT) — — — 11% WSDOT (2016)
Average vehicle occupancy for cars (Ocar) — — — 1.5 AASHTO (2003)
Average vehicle occupancy for trucks (Otruck) — — — 1.05 AASHTO (2003)
Length of detour (Dl) (km) — — — 12.5 Google Maps (2017)
Segment (link) containing damaged bridge (l) (km) — — — 7.5 Google Maps (2017)
Note: COV ¼ coefficient of variation; LN ¼ lognormal; ADTT ¼ average daily truck traffic; ADT ¼ average daily traffic; and WA ¼ Washington.
a
Based on Dong and Frangopol (2015).

© ASCE 04019106-8 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Variation of loss ratio with PGA. Fig. 7. Time-dependent seismic resilience of the bridge for various
earthquake loadings.

For segments where the trend is nearly linear, increase rate of loss
ratio with PGA is more for higher degraded conditions of the
bridge. This is an obvious outcome. Further, analysis shows that the
loss ratio increases linearly up to an approximate value of 0.6,
beyond which the increase rate of loss ratio with PGA decreases,
making the trend follow a nonlinear pattern. This is primarily be-
cause of the extent of damage the bridge suffers at various PGAs.
The trend follows a linear pattern when the bridge incurs minor to
moderate earthquake damage. The nonlinear pattern prevails when
higher damage states (i.e., major damage and collapse) contribute
more to the loss ratio. Hence, the trend quickly becomes nonlinear
with increasing PGA as bridge deterioration progresses.
To apply a damage-specific recovery function (frec) in Eq. (2), the
first task is to define the probable damage state of the bridge under
earthquakes of varied intensity. This was decided based on the
outcome of a previous study related to the development of state-
dependent fragility curves of bridges based on experimental ob-
Fig. 8. Life-cycle seismic resilience of the bridge.
servations (Banerjee and Chi 2013). The bridge in the current study
was said to have a damage state DSE under earthquake E if it suf-
fered from 30% or higher damage exceedance probability in m (> 0)
number of damage states and DSE was the highest among these. For bridge at PGA ¼ 0.68 g due to corrosion degradation of the bridge
example, the pristine bridge has 74.2%, 44.2%, 2.5%, and 0.4% over 100 years.
probabilities of exceedance in minor damage, moderate damage,
major damage, and collapse states, respectively, under a 0.4 g PGA
earthquake (from Fig. 5). Hence, the pristine bridge was considered Uncertainty Quantification of Resilience
to suffer from moderate damage under this earthquake, and the
recovery model corresponding to the moderate damage state was The discussed framework for life-cycle resilience assessment of
applied for assessing its resilience under that earthquake. A similar bridges may be associated with uncertainty due to uncertainties from
approach was taken in Karamlou and Bocchini (2016) to determine its key modules—vulnerability, loss, and recovery functions. A
bridge seismic damage states from fragility curves. This procedure number of previous studies estimated uncertainty in seismic vul-
for resilience assessment was applied for a number of earthquakes nerability models of bridges due to inherent uncertainties from
(having PGAs ranging over 0.2–1.0 g) and for all degraded condi- structural, geotechnical, and hazard parameters (Nielson 2005;
tions of the bridge. For each case, resilience was computed using Yilmaz et al. 2018; Cui et al. 2018). Uncertainty related to traffic
Eq. (1) and control times up to 170 days for minor, moderate, and parameters was considered by Decò et al. (2013) and Dong and
major damage states and 500 days for the collapse state. Fig. 7 Frangopol (2015) to study the variability in resilience, although
shows obtained resilience values of the bridge at pristine and de- uncertainty in the vulnerability model was not considered there.
graded conditions with varied PGAs. As expected, resilience de- Venkittaraman and Banerjee (2014) showed that uncertainties in
creased with increasing seismic intensity, following the reverse recovery time and control time may have great influence on esti-
pattern of that observed in Fig. 6 for the variation of loss ratio with mated seismic resilience of bridges. Based on these previous ob-
PGA. This trend is quite understandable because the increase in loss servations, the current study estimated uncertainty in bridge resil-
ratio results in a decrease in resilience. Moreover, resilience of the ience due to uncertainties from the bridge vulnerability model, the
bridge (at any particular PGA) degrades as deterioration progresses. indirect loss estimation model, and the recovery model of the bridge
Such degrading trends of resilience over the life cycle of the bridge when one or more model parameters were uncertain. The direct loss
are shown in Fig. 8 for three specific values of PGA. Analysis assessment model may produce uncertain direct losses primarily due
showed a maximum reduction of 38% in seismic resilience of the to uncertainty in the bridge vulnerability model when damage ratios
© ASCE 04019106-9 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


are considered to be deterministic at all damage states. This is why Therefore, 10 random parameters were considered in uncertainty
uncertainty in the direct loss model is not considered separately. analysis for resilience assessment of the pristine bridge. Among
Yilmaz et al. (2018) performed a rigorous uncertainty analysis these, the fragility parameter and recovery time take different values
for vulnerability assessment of a bridge considering eight uncertain of mean and standard deviation, depending on the damage state of
parameters related to the structure and underlying soil. That study the bridge. One hundred random samples for each input random
estimated 90% confidence intervals of seismic fragility curves at all variable were generated and random combinations across 10 ran-
damage states to represent possible variations of these curves. dom variables were made using the Latin hypercube sampling
Based on these observed variations, coefficients of variation of technique. For each random combination, resilience of the pristine
fragility curves are calculated to be 9%, 10%, 6%, and 6%, re- bridge was computed following the framework discussed in this
spectively, for minor damage, moderate damage, major damage, study. Thus, 100 values of resilience were obtained for each selected
and collapse states. The present study considered these coefficients value of PGA; their variability is shown in Fig. 9 with a gray band.
of variation of fragility curves from Yilmaz et al. (2018) and used This figure also shows mean values of resilience for varied PGAs.
these along with median fragility parameters of the pristine bridge For specific PGA values, resilience was observed to follow a normal
(from Table 2) to represent probable variability in fragility char- distribution. Fig. 10 shows histograms, normal probability density
Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

acteristics of the studied bridge at its pristine condition. For un- functions (PDFs), and normal probability plots of 100 realizations of
certainty quantification in indirect loss, probabilistic characteristics resilience obtained for 0.5 g and 0.8 g PGA earthquakes. These
of associated random parameters are summarized in Table 3. Re- normal probability plots signify that a normal distribution cannot be
covery time Trec is also taken to be an uncertain parameter; its mean rejected to define the probabilistic variation of resilience at the
values and standard deviations at different damage states are given pristine condition of the bridge. The same observations were made
previously in this article where the recovery function is discussed. for other GMs. The mean and standard deviation values of resilience
Previous studies by Decò et al. (2013) and Dong and Frangopol at 0.5 g PGA were found to be 89.8% and 1.30%, respectively.
(2015) considered recovery time Trec to follow a triangular distri- These values were 81.03% and 1.46%, respectively, at 0.8 g PGA.
bution. The current study considered Trec to follow normal distri- Note that the standard deviation of bridge seismic resilience in-
bution because no specific information on its probabilistic nature creases with increasing PGA, while the mean value gradually de-
was known from a real-life scenario. creases. This outcome indicates that bridge seismic resilience be-
comes more uncertain as intensity of GM increases.
The uncertainty analysis, as discussed here, did not include life
cycle study of the bridge. Nevertheless, the framework is general
and can be used to explore uncertainty in bridge resilience at any
degraded phase along its life span. A similar variation of bridge
resilience, as observed for bridge pristine condition, is expected for
any degraded condition of the bridge.

Conclusions

This research contributes to new knowledge of the impact of


chloride-induced corrosion on life-cycle seismic resilience of RC
bridges. Time-dependent fragility assessment of a three-span aging
RC bridge was performed at both pristine and degraded conditions
of the bridge (i.e., at 25, 50, 75, and 100 years of deterioration)
under earthquakes having varied levels of hazard. Obtained resil-
ience was observed to decrease following a linear (nearly) or non-
Fig. 9. Uncertainty in estimated resilience and its variation with PGA. linear path as GM intensity (i.e., PGA) increased. A linear decrease

(a) (b)

Fig. 10. Resilience of the pristine bridge at 0.8 g PGA and 0.5 g PGA: (a) normal probability papers; and (b) normal probability distributions with
distribution parameters.

© ASCE 04019106-10 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


in resilience with PGA was observed when the bridge was at pris- Aviram, A., K. R. Mackie, and B. Stojadinovic. 2008. “Effect of abutment
tine condition or had slight corrosion degradation. The nonlinear modeling on the seismic response of bridge structures.” Earthquake Eng.
nature became prevalent as the bridge degraded further with time Eng. Vib. 7 (4): 395–402. https://doi.org/10.1007/s11803-008-1008-3.
and was expected to suffer from higher damage states (such as Banerjee, S., and C. Chi. 2013. “State-dependent fragility curves of bridges
based on vibration measurements.” Probab. Eng. Mech. 33: 116–125.
major damage and collapse) under earthquakes. For any specific
https://doi.org/10.1016/j.probengmech.2013.03.007.
PGA, reduction of resilience as the bridge aged portrayed the time- Banerjee, S., and M. Shinozuka. 2007. “Nonlinear static procedure for seismic
dependent trend of life-cycle resilience of bridges under seismic vulnerability assessment of bridges.” Comput.-Aided Civ. Infrastruct.
shaking. Note that the study did not consider pitting corrosion, Eng. 22 (4): 293–305. https://doi.org/10.1111/j.1467-8667.2007.00486.x.
which may result in a faster rate of bridge degradation than that due Banerjee, S., B. S. Vishwanath, and D. K. Devendiran. 2019. “Multihazard
to uniform corrosion. Also, the impact of corrosion on bond resilience of highway bridges and bridge networks: A review.” Struct.
strength degradation in RC members and post-rust crack progres- Infrastruct. Eng. https://doi.org/10.1080/15732479.2019.1648526.
sion was not studied. Research is under way to explore these effects. Biondini, F., E. Camnasio, and A. Titi. 2015. “Seismic resilience of con-
An uncertainty using 10 input random variables from vulnera- crete structures under corrosion.” Earthquake Eng. Struct. Dyn. 44 (14):
2445–2466. https://doi.org/10.1002/eqe.2591.
bility, indirect loss, and recovery models of the bridge at its pristine
Bocchini, P., D. M. Frangopol, T. Ummenhofer, and T. Zinke. 2014.
Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

condition resulted in an uncertain band of resilience estimated for “Resilience and sustainability of civil infrastructure: Toward a unified
earthquakes with PGAs ranging over 0.2–1.0 g. For any specific approach.” J. Infrastruct. Syst. 20 (2): 04014004. https://doi.org/10
PGA, resilience was observed to vary following a normal distri- .1061/(ASCE)IS.1943-555X.0000177.
bution. The mean value of the distribution decreased with in- Caltrans. 2010. Caltrans seismic design criteria. Version 1.6. Sacramento,
creasing PGA, while the standard deviation increased. This out- CA: California Dept. of Transportation.
come indicates higher uncertainty in seismic resilience of the Carrillo, M., and J. M. González. 2002. “A new approach to modelling
pristine bridge as intensity of GM increases. sigmoidal curves.” Technol. Forecasting Social Change 69 (3): 233–
Overall, the research outcome demonstrates the importance of 241. https://doi.org/10.1016/S0040-1625(01)00150-0.
CESMD (Center for Engineering Strong Motion Data). Accessed October
assessing seismic resilience of aging bridges in a life-cycle context.
19, 2018. https://strongmotioncenter.org.
Results obtained from this study facilitate future research aimed at Chen, Z., and F. Cao. 2009. “The approximation operators with sigmoidal
the development of strategic plans for retrofit and maintenance functions.” Comput. Math. Appl. 58 (4): 758–765. https://doi.org/10
prioritization in order to keep bridges safe and functional under .1016/j.camwa.2009.05.001.
distressed conditions. Choe, D.-E., P. Gardoni, D. Rosowsky, and T. Haukaas. 2009. “Seismic
fragility estimates for reinforced concrete bridges subject to corrosion.”
Struct. Saf. 31 (4): 275–283. https://doi.org/10.1016/j.strusafe.2008.10
Acknowledgments .001.
Cimellaro, G. P., A. M. Reinhorn, and M. Bruneau. 2010. “Framework for
analytical quantification of disaster resilience.” Eng. Struct. 32 (11):
This study was supported by the Industrial Research and Consul-
3639–3649. https://doi.org/10.1016/j.engstruct.2010.08.008.
tancy Centre (IRCC) at IIT Bombay, India, through Grant No. Coronelli, D., and P. Gambarova. 2004. “Structural assessment of corroded
15IRCCSG023 and by the Science and Engineering Research reinforced concrete beams: Modeling guidelines.” J. Struct. Eng.
Board (SERB) in India through Grant No. ECR=2017=000849. 130 (8): 1214–1224. https://doi.org/10.1061/(ASCE)0733-9445(2004)
Their support is gratefully acknowledged. 130:8(1214).
Cui, F., H. Zhang, M. Ghosn, and Y. Xu. 2018. “Seismic fragility analysis
of deteriorating RC bridge substructures subject to marine chloride-
Supplemental Data induced corrosion.” Eng. Struct. 155: 61–72. https://doi.org/10.1016/j
.engstruct.2017.10.067.
Fig. S1 and Table S1 are available online in the ASCE Library Decò, A., P. Bocchini, and D. M. Frangopol. 2013. “A probabilistic ap-
(www.ascelibrary.org). proach for the prediction of seismic resilience of bridges.” Earthquake
Eng. Struct. Dyn. 42 (10): 1469–1487. https://doi.org/10.1002/eqe.2282.
Decò, A., and D. M. Frangopol. 2013. “Life-cycle risk assessment of spa-
tially distributed aging bridges under seismic and traffic hazards.”
References Earthquake Spectra 29 (1): 127–153.
Dong, Y., and D. M. Frangopol. 2015. “Risk and resilience assessment of
AASHTO. 2003. A manual of user benefit analysis for highways. 2nd ed. bridges under mainshock and aftershocks incorporating uncertainties.”
Washington, DC: AASHTO. Eng. Struct. 83: 198–208. https://doi.org/10.1016/j.engstruct.2014.10.050.
AASHTO. 2014. AASHTO LRFD bridge design specifications. 7th ed. Dong, Y., and D. M. Frangopol. 2016. “Probabilistic time-dependent
Washington, DC: AASHTO. multihazard life-cycle assessment and resilience of bridges considering
Akiyama, M., and D. M. Frangopol. 2014. “Long-term seismic performance climate change.” J. Perform. Constr. Facil. 30 (5): 04016034. https://doi
of RC structures in an aggressive environment: Emphasis on bridge .org/10.1061/(ASCE)CF.1943-5509.0000883.
piers.” Struct. Infrastruct. Eng. 10 (7): 865–879. https://doi.org/10.1080 Enright, M. P., and D. M. Frangopol. 1998. “Probabilistic analysis of re-
/15732479.2012.761246. sistance degradation of reinforced concrete bridge beams under corro-
Alipour, A., and B. Shafei. 2016. “Seismic resilience of transportation sion.” Eng. Struct. 20 (11): 960–971. https://doi.org/10.1016/S0141
networks with deteriorating components.” J. Struct. Eng. 142 (8): -0296(97)00190-9.
C4015015. https://doi.org/10.1061/(ASCE)ST.1943-541X.0001399. FEMA. 2003. Multi-hazard loss estimation methodology, earthquake
Alipour, A., B. Shafei, and M. Shinozuka. 2011. “Performance evaluation model: HAZUS-MH MR4 technical manual. Washington, DC: FEMA.
of deteriorating highway bridges located in high seismic areas.” J. FHWA (Federal Highway Administration). 2006. Seismic retrofitting
Bridge Eng. 16 (5): 597–611. https://doi.org/10.1061/(ASCE)BE.1943 manual for highway structures. Part 1: Bridges. Publication No.
-5592.0000197. FHWA-HRT-06-032. McLean, VA: FHWA. Accessed September 25,
API (American Petroleum Institute). 2003. Recommended practice for 2018. https://www.fhwa.dot.gov/publications/research/infrastructure
planning, designing and constructing fixed offshore platforms: Working /bridge/06032/06032.pdf.
stress design. Washington, DC: API. FHWA (Federal Highway Administration). 2018. “Bridge condition by
ATC (Applied Technology Council). 1985. “Earthquake damage evaluation highway system.” Accessed October 26, 2018. https://www.fhwa.dot
data for California.” Rep. ATC-13. Redwood City, CA: ATC. .gov/bridge/britab.cfm.

© ASCE 04019106-11 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106


Frangopol, D. M., K.-Y. Lin, and A. C. Estes. 1997. “Reliability of re- Shamsabadi, A., K. M. Rollins, and M. Kapuskar. 2007. “Nonlinear soil–
inforced concrete girders under corrosion attack.” J. Struct. Eng. abutment–bridge structure interaction for seismic performance-based
123 (3): 286–297. https://doi.org/10.1061/(ASCE)0733-9445(1997) design.” J. Geotech. Geoenviron. Eng. 133 (6): 707–720. https://doi.org
123:3(286). /10.1061/(ASCE)1090-0241(2007)133:6(707).
Gardoni, P., and D. Rosowsky. 2011. “Seismic fragility increment func- Sharma, N., A. Tabandeh, and P. Gardoni. 2017. “Resilience analysis: A
tions for deteriorating reinforced concrete bridges.” Struct. Infrastruct. mathematical formulation to model resilience of engineering systems.”
Eng. 7 (11): 869–879. https://doi.org/10.1080/15732470903071338. Sustainable Resilient Infrastruct. 3 (2): 49–67. https://doi.org/10.1080
Ghosh, J., and J. E. Padgett. 2010. “Aging considerations in the development /23789689.2017.1345257.
of time-dependent seismic fragility curves.” J. Struct. Eng. 136 (12): Simon, J., J. M. Bracci, and P. Gardoni. 2010. “Seismic response and fragility
1497–1511. https://doi.org/10.1061/(ASCE)ST.1943-541X.0000260. of deteriorated reinforced concrete bridges.” J. Struct. Eng. 136 (10):
Google Maps. 2017. Accessed December 20, 2017. http://www.google 1273–1281. https://doi.org/10.1061/(ASCE)ST.1943-541X.0000220.
.com/maps. Somerville, P., N. Smith, S. Punyamurthula, and J. Sun. 1997. “Develop-
Jónás, T. 2007. “Sigmoid functions in reliability based management.” ment of ground motion time histories for Phase 2 of the FEMA=SAC
Periodica Polytech. Social Manage. Sci. 15 (2): 67–72. https://doi.org Steel Project.” SAC Joint Venture Project Rep. No. SAC=BD-97=04.
/10.3311/pp.so.2007-2.04. Richmond, CA: SAC Steel Project.
Karamlou, A., and P. Bocchini. 2016. “Sequencing algorithm with multiple- Stein, S. M., G. K. Young, R. E. Trent, and D. R. Pearson. 1999. “Priori-
Downloaded from ascelibrary.org by Carleton University on 09/07/19. Copyright ASCE. For personal use only; all rights reserved.

input genetic operators: Application to disaster resilience.” Eng. Struct. tizing scour vulnerable bridges using risk.” J. Infrastruct. Syst. 5 (3):
117: 591–602. https://doi.org/10.1016/j.engstruct.2016.03.038. 95–101. https://doi.org/10.1061/(ASCE)1076-0342(1999)5:3(95).
Kezhiyur, A. J., S. Banerjee, V. Shankar, and P. Basu. 2016. “Age- Stewart, M. G., and D. V. Rosowsky. 1998. “Structural safety and service-
dependent seismic resilience of an aging bridge network.” In Proc., ability of concrete bridges subject to corrosion.” J. Infrastruct. Syst. 4 (4):
Transportation Research Board 95th Annual Meeting. Paper No. 16- 146–155. https://doi.org/10.1061/(ASCE)1076-0342(1998)4:4(146).
0632. Washington, DC: Transportation Research Board. Thanapol, Y., M. Akiyama, and D. M. Frangopol. 2016. “Updating the
Konstantinidis, D., J. M. Kelly, and N. Makris. 2008. “Experimental in- seismic reliability of existing RC structures in a marine environment by
vestigation on the seismic response of bridge bearings.” Rep. No. EERC incorporating the spatial steel corrosion distribution: Application to
2008-02. Berkeley, CA: Earthquake Engineering Research Center, bridge piers.” J. Bridge Eng. 21 (7): 04016031. https://doi.org/10.1061
Univ. of California, Berkeley. /(ASCE)BE.1943-5592.0000889.
Lee, H.-S., T. Noguchi, and F. Tomosawa. 2002. “Evaluation of the bond Thoft-Christensen, P., F. M. Jensen, C. R. Middleton, and A. Blackmore.
properties between concrete and reinforcement as a function of the de- 1996. “Assessment of the reliability of concrete slab bridges.” Structural
gree of reinforcement corrosion.” Cem. Concr. Res. 32 (8): 1313–1318. Reliability Theory Paper No. 157, Vol. R9616. Presented at the 7th IFIP
https://doi.org/10.1016/S0008-8846(02)00783-4. WG7.5 Working Conference on Reliability and Optimization of
Mackie, K. R., and B. Stojadinovic. 2006. “Post-earthquake functionality of Structural Systems. Aalborg, Denmark: Dept. of Building Technology
highway overpass bridges.” Earthquake Eng. Struct. Dyn. 35 (1): 77– and Structural Engineering, Aalborg Univ.
93. https://doi.org/10.1002/eqe.534. US Dept. of Labor. 2018. Accessed October 26, 2018. https://www.dol.gov.
Madabhushi, S. P. G., and S. K. Haigh, eds. 2005. The Bhuj, India Earth- Venkittaraman, A., and S. Banerjee. 2014. “Enhancing resilience of high-
quake of 26th January 2001: A field report by EEFIT. London: Earth- way bridges through seismic retrofit.” Earthquake Eng. Struct. Dyn.
quake Engineering Field Investigation Team, Institution of Structural 43 (8): 1173–1191. https://doi.org/10.1002/eqe.2392.
Engineers. Warn, G. P., A. S. Whittaker, and M. C. Constantinou. 2007. “Vertical
Mosher, R. L. 1984. Load-transfer criteria for numerical analysis of axially stiffness of elastomeric and lead–rubber seismic isolation bearings.” J.
loaded piles in sand. Part II: Load pile capacity curves for steel and Struct. Eng. 133 (9): 1227–1236. https://doi.org/10.1061/(ASCE)0733
concrete piles. Technical Rep. K-84-1. Vicksburg, MS: US Army En- -9445(2007)133:9(1227).
gineer Waterways Experiment Station, US Army Corps of Engineers. WSDOT (Washington State Department of Transportation). 2016. “2016
Neville, A. 1995. “Chloride attack of reinforced concrete: An overview.” annual traffic report.” Accessed October 19, 2018. http://www.wsdot.wa
Mater. Struct. 28 (2): 63–70. https://doi.org/10.1007/BF02473172. .gov/mapsdata/travel/pdf/Annual_Traffic_Report_2016.pdf.
Ni Choine, M., M. Kashani, L. N. Lowes, A. O’Connor, A. J. Crewe, N. A. WSDOT (Washington State Department of Transportation). 2017. “2017–
Alexander, and J. E. Padgett. 2016. “Nonlinear dynamic analysis and 2018 snow and ice plan.” Accessed October 19, 2018. http://www
seismic fragility assessment of a corrosion damaged integral bridge.” .wsdot.com/winter/snowiceplan.htm.
Int. J. Struct. Integ. 7 (2): 227–239. https://doi.org/10.1108/IJSI-09 Yilmaz, T., S. Banerjee, and P. A. Johnson. 2016. “Performance of two real-life
-2014-0045. California bridges under regional natural hazards.” J. Bridge Eng. 21 (3):
Nielson, B. G. 2005. “Analytical fragility curves for highway bridges in 04015063. https://doi.org/10.1061/(ASCE)BE.1943-5592.0000827.
moderate seismic zones.” Ph.D. thesis, School of Civil and Environ- Yilmaz, T., S. Banerjee, and P. A. Johnson. 2018. “Uncertainty in risk of
mental Engineering, Georgia Institute of Technology. highway bridges assessed for integrated seismic and flood hazards.”
Ou, Y.-C., H.-D. Fan, and N. D. Nguyen. 2013. “Long-term seismic per- Struct. Infrastruct. Eng. 14 (9): 1182–1196. https://doi.org/10.1080
formance of reinforced concrete bridges under steel reinforcement /15732479.2017.1402065.
corrosion due to chloride attack.” Earthquake Eng. Struct. Dyn. 42 (14): Zheng, Y., Y. Dong, and Y. Li. 2018. “Resilience and life-cycle perfor-
2113–2127. https://doi.org/10.1002/eqe.2316. mance of smart bridges with shape memory alloy (SMA)-cable-based
Ramanathan, K. N. 2012. “Next generation seismic fragility curves for bearings.” Constr. Build. Mater. 158: 389–400. https://doi.org/10.1016
California bridges incorporating the evolution in seismic design phi- /j.conbuildmat.2017.10.031.
losophy.” Ph.D. thesis, School of Civil and Environmental Engineering, Zhong, J., P. Gardoni, and D. Rosowsky. 2012. “Seismic fragility estimates
Georgia Institute of Technology. for corroding reinforced concrete bridges.” Struct. Infrastruct. Eng.
Rokneddin, K., J. Ghosh, L. Dueñas-Osorio, and J. E. Padgett. 2014. 8 (1): 55–69. https://doi.org/10.1080/15732470903241881.
“Seismic reliability assessment of aging highway bridge networks with Zhou, Y., S. Banerjee, and M. Shinozuka. 2010. “Socio-economic effect of
field instrumentation data and correlated failures. Part II: Application.” seismic retrofit of bridges for highway transportation networks: A pilot
Earthquake Spectra 30 (2): 819–843. https://doi.org/10.1193 study.” Struct. Infrastruct. Eng. 6 (1=2): 145–157. https://doi.org/10
/040612EQS160M. .1080/15732470802663862.

© ASCE 04019106-12 J. Bridge Eng.

J. Bridge Eng., 2019, 24(11): 04019106

You might also like