Influence of Traffic Loading on the Seismic Reliability

Assessment of Highway Bridge Structures

Jayadipta Ghosh, S.M.ASCE1; Colin C. Caprani2; and Jamie E. Padgett, A.M.ASCE3
Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Traditionally, the impacts of traffic and earthquake loading have been considered independently when assessing bridge reliability.
This paper presents a framework for joint seismic and live-load fragility assessment of highway bridges. Full probabilistic analyses accounting
for variation in bridge parameters, ground motion, and truck load and position are proposed to develop bridge system fragility curves and to
identify the critical truck position that renders the bridge most vulnerable to earthquakes. A fragility surface is derived for the critical truck
position at which the failure probability is conditioned on the governing vehicle weight in addition to ground motion intensity, thus depicting
the impact of truck load on bridge seismic fragility. This fragility surface is convolved with the governing vehicle weight distribution (obtained
from weigh-in-motion data) and probability of truck occurrence (function of truck flow rate) to determine traffic-informed conditional seismic
reliability estimates. The proposed methodology is demonstrated on a case study of a multispan continuous steel girder bridge in the central and
southeastern United States. The framework can find ready extensions to assess the joint impact of earthquake and live loads for other bridges and
hazard conditions and can offer a basis for deriving reliability-based load combinations consistent with emerging trends in bridge design. DOI:
10.1061/(ASCE)BE.1943-5592.0000535. © 2013 American Society of Civil Engineers.
Author keywords: Bridge engineering; Fragility analysis; Seismic reliability; Live-load reliability; Multihazard threat; Joint seismic and
live-load reliability.

Introduction goods (National Safety Commission 2007) and expectations of

future truck traffic that exceeds present standards. The impact of
such vehicles on bridge seismic performance has yet to be sys-
Motivation tematically explored.
The functionality of highway bridges is threatened by a range of
stressors including natural hazards, such as earthquakes, aging, Scientific Context
deterioration, and demands from heightened traffic loads. Whereas In recent decades, seismic fragility analysis of highway bridge
much literature exists solely on the live-load reliability of bridges structures has emerged as a powerful tool for vulnerability assess-
(Nowak et al. 2001; Du and Au 2005; Czarnecki and Nowak 2007) ment of bridges located in earthquake-prone regions. For predefined
or on their vulnerability to seismic hazards alone (Basoz and input parameters, such as the ground motion intensity, bridge fra-
Kiremidjian 1999; Choi et al. 2004; Gardoni et al. 2002; Nielson and gility models help to predict the probability of meeting or exceeding
DesRoches 2007a), a negligible amount of research focuses on the a particular damage state. Traditionally, seismic fragility models for
simultaneous consideration of live-load and earthquake threats. This highway bridges investigate the impact of the ground motion hazard
is particularly important, because a significant percentage of bridges alone on bridge-component and system-level vulnerability (Basoz
located in moderate to high seismic zones, characterized by typical and Mander 1999; Basoz and Kiremidjian 1999; Shinozuka et al.
nonseismic detailing (Choi et al. 2004; DesRoches et al. 2004; 2000; Mackie and Stojadinovic 2001; Nielson and DesRoches
Wright et al. 2011), have been labeled as functionally obsolete 2007a; Straub and Der Kiureghian 2008). Only recently have
(ASCE 2009), suggesting older design features and geometries and researchers extended this notion to vulnerability assessments of
an inability to accommodate current traffic volumes, vehicle sizes, bridges under joint threats or multiple hazards, such as the in-
and weights. The situation is further exacerbated by half a million corporation of aging and deterioration mechanisms (Choe et al.
overweight trucks being employed in the United States to transfer 2009; Ghosh and Padgett 2010) or impacts of scour (Alipour et al.
2013) on bridge seismic fragility. Given the unpredictable nature of
1 earthquakes, truck and traffic loads may be present during the
Graduate Research Assistant, Dept. of Civil and Environmental En-
gineering, Rice Univ., Houston, TX 77005 (corresponding author). E-mail:
seismic excitation as evidenced by past earthquake events, for in-
[email protected] stance, during the 1994 Northridge and 1995 Kobe earthquakes
2
Lecturer, Dept. of Civil Engineering, Monash Univ., Clayton, Victoria (Basoz and Kiremidjian 1998; Chang 2000). These instances
3800, Australia; formerly, Lecturer, Dept. of Civil and Structural Engineer- coupled with a lack of simultaneous consideration of earthquake
ing, Dublin Institute of Technology, Dublin 1, Ireland. E-mail: colin.caprani@ and traffic loads underlines the need to develop a joint fragility
monash.edu assessment framework.
3
Assistant Professor, Dept. of Civil and Environmental Engineering, Rice
Univ., Houston, TX 77005. Scientific Contribution
Note. This manuscript was submitted on February 14, 2013; approved on
July 15, 2013; published online on July 17, 2013. Discussion period open There is a pressing need for joint fragility assessment and risk
until May 5, 2014; separate discussions must be submitted for individual modeling to explore the role of traffic loading on the seismic per-
papers. This paper is part of the Journal of Bridge Engineering, © ASCE, formance of highway bridges. Further, the increasing availability of
ISSN 1084-0702/04013009(11)/$25.00. weigh-in-motion (WIM) data and the obvious variability of both

© ASCE 04013009-1 J. Bridge Eng.

J. Bridge Eng. 2014.19.

 2 3
truck flow and the distribution of truck weights across differing lnðSC Þ 2 lnðr1 Þ

  6 lnðPGAÞ 2 7
regions mean that good estimates of joint seismic and live-load 6 r2 7
fragility are now possible. Additionally, ongoing trends in bridge P DScomponent PGA ¼ F6 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 7 (2)
6 bDjIM þ bC
2 2 7
design readily apply LRFD; however, a reliability-based approach 4 5
has only been taken to derive the combination of dead load and r2
vehicular load in AASHTO (2012). To enable future opportu-
nities for more consistent reliability-based treatment of load com- where DScomponent 5 component damage state under consideration;
binations (Ghosn et al. 2003), such as earthquakes and vehicular loads, Fð × Þ 5 cumulative distribution function of a standard normal
a method for joint seismic and live-load fragility is first required. distribution; peak ground acceleration (PGA) is adopted as the in-
Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

This work proposes a general framework in which site-specific tensity measure IM; and SC and bC 5 median and dispersion
traffic characteristics can be coupled with conditional seismic re- estimates of the component capacity, respectively.
liability estimates to arrive at traffic-informed seismic fragility Bridge system-level fragility is derived assuming the bridge as
curves reflecting the likelihood of truck presence. This framework a series system wherein failure of a single component is represen-
should find ready application by practitioners for more refined tative of bridge failure (Nielson and DesRoches 2007b; Nowak and
seismic assessment of problem bridges, offer guidance for risk Cho 2007). Distribution parameters are estimated following a Monte
analysts regarding the relative importance of considering simulta- Carlo simulation from the PSDM and capacity estimates to arrive at
neous live loads and earthquakes, and even form a basis for deriv- the overall bridge system-level fragility
ing load factors for LRFD. "
#
Finally, the proposed framework is applied to an example bridge
  lnðPGAÞ 2 ln medsys
and traffic stream, both to illustrate its application and to discover 
P DSsystem PGA ¼ F (3)
dispsys
indicators of the general phenomenon. Whereas previous research
has thoroughly investigated the impact of earthquake load on the
response of critical bridge components, the proposed framework is where medsys and dispsys 5 median values (in units of g) and
used to assess if superimposed truck live-load contributes further to logarithmic standard deviations of the system fragilities; and
their seismic vulnerability. DSsystem 5 system-level damage state of interest.

Seismic Fragility Assessment of Highway Bridges

Traffic Load Modeling
A brief review of the steps involved in the state-of-the-art procedure
for seismic fragility assessment of highway bridges is presented first, Site-Specific Traffic Modeling
prior to developing a general framework for joint seismic and live-
load vulnerability assessment that integrates concepts from both Because most codes of practice for bridge loading apply to a wide
seismic fragility analysis and traffic load modeling. geographic area with many different levels of traffic loading se-
verity, they are necessarily conservative. Consequently, especially
Step 1: Probabilistic Seismic Demand Models for for the assessment of existing bridges, it is now well recognized that
a site-specific bridge loading model can offer substantial savings in
Bridge Components
rehabilitation costs. The better ability of site-specific traffic models
First, statistical relationships called probabilistic seismic demand to reflect realistic loading motivates their adoption for joint live-load
models (PSDMs) are derived to establish a relationship between the and seismic fragility modeling, averting the bias that may be in-
median of the peak seismic response of a particular bridge compo- troduced by adopting code-based load models. Although there are
nent and the intensity of the ground motion [Eq. (1)] (Mackie and many variations for site-specific traffic models in the literature
Stojadinovic 2001; Kunnath et al. 2006; Nielson and DesRoches (Caprani 2012), a reasonably simple yet representative model is
2007a; Ghosh and Padgett 2010) developed for ready inclusion in the joint seismic and live-load
fragility calculation framework. This model assumes that traffic is
lnðSD Þ ¼ lnðr1 Þ þ r2 lnðIMÞ þ ɛ (1) free-flowing and that the bridge length is sufficiently short so that the
presence of two or more trucks can be neglected.
where SD 5 median value of the component demand; r1 and r2
5 regression parameters; IM 5 ground motion intensity; and ɛ 5 zero-
Truck Presence Modeling
mean normally distributed model-fitting error, which also allows
estimation of the dispersion or lognormal standard deviation of the Because different sites may have different average traffic speeds,
seismic demand, denoted by bDjIM . Bridge-component PSDMs are even with the same flow, a useful traffic metric to consider for the
typically developed from statistical analysis of nonlinear time present purposes is the mean truck density (trucks per km). If the
history responses of bridge samples subjected to sample ground number of trucks in a single lane per hour is Q (i.e., the truck flow
motions representative of the bridge location. rate) and assuming a constant (or mean) speed for the traffic stream
of v ðkm=hÞ, the truck density (trucks=km) is
Step 2: Seismic Fragility Curves for Bridge Components
Q
and System r¼ (4)
v
In the next step, PSDMs of key components and correlations be-
tween component responses are used in conjunction with compo- Thus, for example, a single-lane site with 200 trucks=h and mean
nent capacity estimates to develop fragility curves. Under the speed of 88:5 km=h ð55 mi=hÞ has a truck density of r 5 200=88:5
assumption that the demand and component capacities are lognor- 5 2:26 trucks=km.
mally distributed, the seismic fragility of bridge components can be Haight (1963) describes a general approach for counting the
found in closed form (Melchers 1999) number of arrivals in a period of time t or a length L (assuming

© ASCE 04013009-2 J. Bridge Eng.

J. Bridge Eng. 2014.19.

 a constant speed for traffic). Denote the assumed headway (distance Joint Seismic and Live-Load Fragility Assessment
in time or space from front of leading vehicle to front of following
vehicle) distribution as gðxÞ. Then, the probability of up to n vehicles
arriving on a length of road L is Generalization of Fragility Concept
The conventional seismic fragility assessment procedure outlined
Pðnumber of vehicles on L # nÞ ¼ Pðsum of n þ 1 gaps $ LÞ (5) earlier results in a conditional probability of damage state exceed-
ance given a PGA, PðDSjPGAÞ. This probability can be seen as
The sum of n 1 1 random identically distributed variables with a special case of PðDSjPGA, qÞ with q 5 0, where q is a random
general probability density gðxÞ is given by the ðn 1 1Þth convo- variable representing live load with probability density fQ . The
Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

lution of the densities. Vehicles have a minimum length within distribution fQ will, in general, be a mixed distribution (Benjamin
which no arrival can occur, including the physical length and and Cornell 1970), including a discrete probability of there being no
a minimum gap that drivers will keep in jammed conditions, as- live load present and the spectrum of probabilities of the differing
sumed to be about 2 m. Denoting this length D, then the actual gap live-load magnitudes. Hence, for traffic-informed seismic fragility
distribution is assessment, the unconditional probability should ideally be com-
puted after accounting for live load as
gD ðxÞ ¼ gðx 2 DÞ (6) ð
PðDSjPGAÞ ¼ PðDSjPGA, qÞ × fQ ðqÞdq (10)
Haight (1963) provides the general result for n vehicles. For
simplicity, however, the presence of one truck at a time on the bridge
is slightly overestimated by neglecting multiple truck presences in
the same lane and writing Representation of Live Load
The live load of Eq. (10) should be representative of the actual traffic
Pðone truckÞ  1 2 Pðno trucksÞ (7) at the site. The live-load models of many codes of practice are
notional only and are commonly calibrated to reflect load effects and
For the lengths and densities considered in this paper, the prob- not the actual magnitude of the load itself. Finally, as noted earlier,
ability of occurrence of two or more trucks in the same lane are the increasingly wide availability of WIM data suggests that a single
successively orders of magnitude less than the probability of one truck, representative of the traffic at the site, may be easily derived
truck, and so Eq. (7) is a slight overestimation, because it includes for ease of applicability of the general framework. Therefore, it is
these probabilities in the one-truck probability. Consequently, from assumed for the present purposes that the live load at the site can be
Haight (1963) represented by a single truck. In making this representation, it is
assumed that there is negligible influence on the seismic fragility of
  the variability in truck geometry, namely, number of axles, axle
G½1, mðL 2 DÞ Gð0, mLÞ
Pðone truckÞ ¼ 1 2 2 (8) spacings, and other factors. Importantly though, the representative
Gð1Þ Gð0Þ truck can be assigned a gross vehicle weight (GVW) histogram
representative of all traffic at the site.
where Gðn, xÞ 5 upper incomplete gamma function; GðnÞ 5 complete Other representations of live load are clearly possible, such as a
gamma function; L 5 length of the bridge; and m 5 mean truck uniformly distributed force or mass of random magnitude. Indeed,
density in trucks=m ðr 3 1023 Þ. The assumptions underlying Eq. (8) the equivalent uniformly distributed loads of some codes of practice
have been tested using a Monte Carlo simulation of a heterogeneous may be attractive in this regard. However, preliminary investigations
traffic stream based on the model described in O’Brien and Caprani
(2005). It is found that Eq. (8) offers a slightly conservative esti-
mation of the probability of occurrence of a truck on a length of road
given a truck flow rate and velocity distribution. As such, this model
is deemed appropriate for modest truck flows and bridge lengths
under about 100 m.
This discussion relates to one traffic lane only. If it is assumed
that the truck arrivals in adjacent lanes are independent [and there is
some evidence to suggest that they are not, but the correlation is
generally weak (O’Brien and Enright 2011)], then the probability of
one truck being present on multiple lanes is found from Eq. (8) using
the combined truck density across the lanes.
Fig. 1 shows the variation of the probability of the presence of one
truck by flow rate for different bridge lengths. It can be seen that the
probability is more sensitive to bridge length than truck density
and may be reasonably well approximated by straight line fits enve-
loped by

Pðone truckÞ  ðL 2 18ÞQ  1025 (9)

where L 5 bridge length (25 # L # 100 m); Q 5 mean hourly truck

Fig. 1. Probabilities of observing one truck by truck density and bridge
flow (0 # Q # 600 trucks=h); and a mean traffic speed of 80e90 km=h
length, assuming D 5 22 m
is assumed.

© ASCE 04013009-3 J. Bridge Eng.

J. Bridge Eng. 2014.19.

 indicate that such globally applied live-load models may not be a geographical area subject to similar ground motions must be
suitable, because bridge-component seismic fragilities are sensitive identified.
to the more localized loading offered by actual trucks. Further, the
trucks of notional load models (such as the HS-20 truck) are cali-
brated to give similar load effects (such as bending moment and Example Application: Description
shear force) to real traffic and not to represent the actual traffic itself.
Therefore, the notional live-load models could underestimate the To illustrate application of the methodology, a regional joint seismic
true influence of live load on the system seismic fragility, and so, and live-load assessment is undertaken for the central and south-
a representative single truck is used instead. eastern United States (CSUS). A representative seismically vul-
Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

Finally, some further simplifications are made for the live-load nerable multispan continuous (MSC) steel-girder bridge is chosen,
model. It is assumed that the superimposed truck is stationary at constituting 11.6% of bridges in the region (Wright et al. 2011).
a random location on the bridge during the earthquake. Future re- Regional traffic characteristics are taken from 11 WIM sites in
search can investigate the impact of the moving vehicle or bridge- Alabama. The following sections will describe the finite-element
vehicle interaction on seismic bridge fragility. Although convolution bridge-modeling approach and the site-specific traffic model adopted
on truck position is possible, the truck is conservatively assumed for this study.
in this paper to be positioned to give the critical fragility of interest
(for either a single component or the system). As will be shown Seismic Hazard [Input A (Fig. 2)]
later, however, the seismic fragility is more sensitive to the truck
presence than the location of the truck, so this conservative location A total of 96 two-component ground motions are used from the
is not a gross simplification. Wen and Wu (2001) and Rix and Fernandez (2004) synthetic
ground motion suites in this study for the fragility analysis and
impact assessment of truck load on bridge seismic vulnerability.
Convolution for Traffic-Informed Seismic Fragility These ground motions are representative of a range of potential
ground motions for the CSUS region and have been extensively
The assumption of a single representative truck for the site-specific
used in the past for fragility analysis of pristine, retrofitted, or
live-load model is reasonable for lengths less than about 100 m and
aging highway bridges located in these regions (Nielson and
thus covers many bridges. Eq. (10) can therefore be adapted to read
DesRoches 2007a; Padgett and DesRoches 2009; Ghosh and
PðDSjPGAÞ ¼ PðDSjPGA, truckÞ × PðtruckÞ Padgett 2010).

þ PðDSjPGA, no truckÞ × Pðno truckÞ (11)

Bridge Data [Inputs B and C (Fig. 2)]
The truck presence model described earlier gives the comple- Fig. 3 shows the representative MSC steel bridge characterized by
mentary probabilities of truck presence [PðtruckÞ] or no presence the continuity of the steel girders over the interior bents and
[Pðno truckÞ], and the conventional seismic fragility assessment showing the superimposed truck atop the bridge deck. Typical
yields PðDSjPGA, no truckÞ. It therefore remains to evaluate the design details of this bridge type include multiple components that
probability of a damage state being exceeded conditional both on are susceptible to damage during earthquake events (Choi et al.
PGA and on the truck being present, PðDSjPGA, truckÞ. However, 2004; Nielson and DesRoches 2007a), such as (1) inadequately
the truck live load is itself conditional on the GVW distribution, reinforced concrete columns with widely placed transverse ties
fW ð×Þ, at the site and is found from and approximately 1% longitudinal reinforcement ratio, (2) vul-
ð nerable high-type steel fixed and rocker bearings with short seat
PðDSjPGA, truckÞ ¼ PðDSjPGA, truck, wÞ × fW ðwÞdw (12) widths, and (3) inadequately reinforced pile caps. Additionally,
large inertial loads attributable to the considerable deck mass and
the superimposed truck load may result in high demands on the
Thus, the proposed joint seismic and live-load fragility framework underreinforced columns, expansion bearings, and abutments
requires three inputs. The first two inputs consist of a site-specific during seismic events.
GVW histogram representative of truck traffic fW ð×Þ and the prob- The three-dimensional finite-element model for the case-study
ability of the occurrence of a truck, which is a function of the truck bridge configuration is developed using the finite-element plat-
density at the site, which is in turn a function of the flow rate and form OpenSees (Mazzoni et al. 2009) following the recommen-
mean speed at the site. The truck GVW distribution is readily dations by Nielson and DesRoches (2007a). For the superstructure
obtained from WIM data at the site, and the other information is modeling, the composite action of the steel girders and bridge deck is
obtained from standard traffic counters. The third input is the taken into account and modeled with linear elastic beam column
probability of damage state exceedance given the earthquake in- elements, because damage is not expected in the superstructure.
tensity for a stand-alone bridge with no superimposed truck load The weight of the superimposed truck axles at any position on
[PðDSjPGA, no truckÞ], as well as for a bridge with a truck of given the bridge deck is reflected via increased modal masses of the
GVW atop the bridge deck [PðDSjPGA, truck, wÞ]. pertinent deck nodes, additional vertical forces on steel fixed and
expansion bearings, and increased axial forces on bridge columns.
Analytical modeling of the bridge bearings is based on experimental
Overall Seismic Fragility Assessment Procedure
tests by Mander et al. (1996). Subsequently, bilinear truss elements
The proposed framework for the joint seismic and live-load fra- and link elements in OpenSees are used to model the prying (of the
gility assessment procedure is illustrated in Fig. 2. It can be masonry plate), rocking, and stiffness degradation of the bearing
implemented for a particular bridge or for a representative class in assembly along both longitudinal and transverse directions. Non-
a wider geographical area. In both cases, the live-load model must linear beam column elements with fiber-defined cross sections are
be representative of the amalgamated traffic data, and for the used to model columns and bent beams, whereas nonlinear springs
second case, a suite of bridges representative of the bridge stock in represent the abutments with lumped translational and rotational

© ASCE 04013009-4 J. Bridge Eng.

J. Bridge Eng. 2014.19.

Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. General framework for joint seismic and live-load fragility assessment: (a) site data and derived inputs; (b) calculation procedure for joint
seismic and live-load fragility assessment

Fig. 3. Case-study MSC steel girder bridge adopted in this study; the truck image is an adaptation of the conventional 18-wheeler truck diagram by
Padleckas (2011; used under a Creative Commons Attribution-Share Alike Unported 3.0 license: http://creativecommons.org/licenses/by-sa/3.0/
deed.en)

© ASCE 04013009-5 J. Bridge Eng.

J. Bridge Eng. 2014.19.

 springs at the base of the columns to model the pile foundations. The investigation of the force-displacement behavior of critical bridge
probabilistic models for the random variables corresponding to components during seismic shaking is presented only for the load
bridge structure include those previously identified for fragility case that most adversely affects bridge reliability. The development
analysis of MSC steel bridges in the CSUS (Nielson and of fragility surfaces corresponding to different truck GVWs for the
DesRoches 2007a), namely, concrete compressive strength, steel most unfavorable truck location is followed by the convolution with
strength, stiffness of foundation piles, damping ratios, and the gap the GVW occurrence probabilities.
between the deck and abutment.
Critical Truck Location and Component
Traffic Data and Model [Inputs D, E, and F (Fig. 2)] Response Analysis
Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

Traffic data typical of the CSUS is obtained from a regional network The variation in bridge system fragility with truck location is
of 11 WIM stations across northern Alabama, with 21.4 million assessed for 18 locations of 2.90 m increments: Load Cases 1–18
trucks recorded. In this database, 45% of all trucks are the standard [Fig. 6(a)]. The symmetry of the bridge about the centerline averts
American 18-wheeler, making it the most common type of truck by the need to redo the computationally expensive bridge fragility
far. It is termed a WB-20 by Transportation Research Board (2003) analysis for the WB-20 moving in the opposite direction. The fully
and is a Federal Highway Administration (FHWA) Class 9 vehicle, loaded truck GVW mode of 35.4 t [Fig. 5(b)] is used in this stage of
subtype S3 (FHWA 2011b). The geometric configuration of the the analysis [Fig. 6(a)], but the impact of different truck GVWs will
truck does not have much variability, and therefore, truck traffic is be found in the next section. Corresponding to each truck position, as
represented by a WB-20 truck with an associated GVW distribution.
The configuration of this truck is arrived at through a study of the
modes of the distributions of the relevant parameters and is given in
Fig. 4. For example, Fig. 5(a) shows the histogram of the percentage
of the GVW carried by the rear axle tandem, from which the mode
gives 42.5%, which is similar to that found from the legal limits
(FHWA 2011a). Fig. 5(b) gives the histogram of GVW for all WB-
20 trucks recorded. It shows two modes at 17 and 35 t, representing
empty and fully loaded trucks (partial loads are inefficient and so are
fewer), with a GVW range of approximately 10–55 t.
The CSUS representative bridge length is 66.9 m, and a 20 m
length for the WB-20 truck plus 2-m full-stop bumper-to-bumper
minimum gap (Caprani 2012) gives D 5 22 m. Therefore, for a free-
flow speed of 80 km=h (about 50 mi=h), the probabilities of one
truck occurring, from Eq. (8), are 4.4% for 80 trucks=h (truck density
of 1 truck=km) increasing to 16.4% for 320 trucks=h ð4 trucks=kmÞ.
These can be estimated from Fig. 1, in which the case-study bridge
length is given along with alternative lengths.

Example Application: Results

The location of the WB-20 truck and its GVW are random param-
eters, and the impact of these factors on the bridge fragility is
assessed. Whereas system-level fragility estimates are developed for
various locations of the WB-20 truck across the bridge deck, the

Fig. 4. Adopted configuration of representative truck (note that tandem

loads are evenly distributed between the axles); the truck image is an
adaptation of the conventional 18-wheeler truck diagram by Padleckas
(2011; used under a Creative Commons Attribution-Share Alike
Unported 3.0 license: http://creativecommons.org/licenses/by-sa/3.0/ Fig. 5. Histograms of WIM-measured truck parameters: (a) percentage
deed.en) of GVW carried by rear axle tandem; (b) GVW for all WB-20 type trucks

© ASCE 04013009-6 J. Bridge Eng.

J. Bridge Eng. 2014.19.

Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. (a) Different WB-20 truck locations investigated in this study (ranging from Load Case 1 to 18) to assess the impact on bridge system fragility
[the truck image is an adaptation of the conventional 18-wheeler truck diagram by Padleckas (2011), used under a Creative Commons Attribution-Share
Alike Unported 3.0 license: http://creativecommons.org/licenses/by-sa/3.0/deed.en]; (b) median values of bridge system fragility for the stand-alone
bridge with no superimposed truck and Load Cases 1–18; note that Load Case 5 leads to the highest change in median values and hence corresponds to
the most unfavorable truck position

Table 1. Bridge Component Response Parameters Used in This Study

well as for the no-truck case, PSDMs are developed for different Component Response parameter
bridge components (Table 1) followed by fragility analyses to de-
Column response Curvature ductility
termine the median and dispersion values of bridge system fragility.
Fixed bearing longitudinal response Deformation
Fig. 6(b) depicts the median values of bridge system fragility for
Fixed bearing transverse response Deformation
the extensive damage state (DS 5 extensive) for the different load
Expansion bearing longitudinal response Deformation
cases in addition to median value for the stand-alone bridge without
Expansion bearing transverse response Deformation
any superimposed truck load [PðDSjPGA, no truckÞ]. The extensive
Abutment active response Deformation
damage limit state is characterized by the following: the damage to
Abutment passive response Deformation
bridge components is visible; it requires repair and results in closure
Abutment transverse response Deformation
of the bridge for at least a week following a seismic event; and it often
leads to reduced traffic-carrying capacity for up to a month (Padgett
and DesRoches 2007). Because the median value of fragility cor- load on a bridge leads to a decrease in the median value of fragility
responds to the PGA intensity, which indicates a 50% probability of and hence an increase in bridge vulnerability. Load Case 5 is found
damage state exceedance, a reduction in median value signifies to be the most unfavorable to the case-study bridge fragility, leading
a more seismically vulnerable bridge. It is evident from Fig. 6(b) to approximately a 10% reduction in median values as compared with
that, regardless of the location, the presence of a superimposed truck the pristine bridge structure. Finally, the variation in the dispersion

© ASCE 04013009-7 J. Bridge Eng.

J. Bridge Eng. 2014.19.

 of fragility estimates with truck location is negligible compared with Influence of Truck Weight on Seismic Fragility
the stand-alone bridge and approximately equals 0.50.
For the most unfavorable position of the WB-20 truck load (Load
A closer inspection of the seismic response for the critical truck
Case 5), fragility surfaces are derived for different truck GVWs.
position (Load Case 5) reveals an increase in the seismic demand of
Fragility analyses are conducted, successively increasing the truck
critical bridge components compared with that for the stand-alone
GVW from 10 to 60 t in increments of 10 t to cover the entire range of
bridge. For instance, Fig. 7(a) shows an increased peak column
GVW in the WIM data, shown in Fig. 5(b). Fig. 8(a) shows the
curvature in the first bent when the bridge is subjected to a ground
fragility curves for each truck GVW and the interpolated fragility
motion record with a PGA of 0.47 g. Additionally shown in Fig. 7(b)
surface. The steady increase in bridge fragility underlines the en-
is the increase in peak expansion bearing deformation in the second
hanced susceptibility of bridge damage state exceedance to in-
Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

bent along the longitudinal direction compared with that of the stand-
creased superimposed truck load. The smoothness of the fragility
alone bridge. Although Load Case 5 results in the maximum increase
surface and lack of discontinuities impart confidence in the 10-t
in the seismic demand of critical bridge components, leading to the
increments of truck GVW adopted in this study to develop the
highest reduction in the median value of bridge fragilities [Fig. 6(b)],
fragility surface.
a steady increase in seismic demand of these components attrib-
utable to superimposed truck load is observed for all load cases
(Load Cases 1–18).
Joint Seismic and Live-Load Fragility Assessment
Under the methodology explained earlier, a joint seismic and live-
load assessment is made for a one-lane bridge assuming a truck
density of 4 trucks=km, for which Eq. (8) gives the probability
of occurrence of one truck to be 16.4%. Fig. 9(a) shows the fragi-
lity curves for the cases: (1) no truck is present [PðDSjPGA, no truckÞ],
representing current practice; (2) one truck is present [PðDSjPGA,
truckÞ] with GVW histogram given by Fig. 5(b), found using
Eq. (12); and (3) the probabilistic presence of a random truck,
PðDSjPGAÞ, which uses the fragility curve from Eq. (11) [see also
Eq. (12)], as well as uncertainty in truck GVW. The median
fragilities for these cases are: (1) 0.514 g, (2) 0.481 g, and (3)
0.508 g, showing reductions of 6.33 and 1.07% for Cases 2 and 3
from the no-truck case.
Finally, to assess the influence of traffic flow variation across the
region, the change in fragility percentiles with truck density can be
ascertained, and the results are shown in Fig. 9(b). In this figure, the
reductions in the median and 95th-percentile PGAs are shown to
increase with increasing truck density. They are also quite similar,
suggesting a shift in the fragility curve, rather than a change in its
dispersion. However, the reductions are quite small. Further, it must
be recalled that it is assumed that the truck is located at the most
unfavorable position, and so, the real reduction could be smaller.
Although the previous sections demonstrated that the presence of
a truck atop the bridge deck may cause appreciable changes in
fragility relative to the no-truck case, after accounting for the GVW
distribution and truck density, the impact is found to be negligible for
the present investigation.

Conclusions

This work proposes an approach for joint seismic and live-load

fragility modeling and a method to quantify the impact of consid-
ering truck load presence within a seismic fragility analysis. In
recognition of the sensitivity of component seismic fragilities (and
hence system-level fragility) to truck location and configuration, it is
proposed that a representative single truck be defined for a site or
region under consideration. The critical representative truck location
Fig. 7. Increase in peak component responses of (a) RC columns and
on the component or system fragility is then identified. Because
(b) expansion bearings in the longitudinal direction when the case-study
truck GVW for a site or region is random, a convolution with the
bridge with WB-20 truck in Load Case 5 position is subjected to
GVW histogram is used so that the seismic fragility conditional on
a deterministic ground motion with PGA of 0.47 g; comparisons are
truck presence is found, resulting in a bridge fragility surface. Fi-
shown with respect to the component behaviors of the stand-alone
nally, a reasonably simple traffic model is presented, so that the
bridge with no superimposed truck load subjected to the same ground
probability of a truck presence can be found given the truck density
motion
at the site, and consequently, a joint seismic and live-load fragility
assessment can be carried out.

© ASCE 04013009-8 J. Bridge Eng.

J. Bridge Eng. 2014.19.

Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. (a) Case-study bridge system-level fragility curves for different GVWs of the WB-20 truck positioned at Load Case 5 location; (b) interpolated
three-dimensional fragility surface depicting the joint impact of the PGA of seismic shaking and truck GVW on bridge system fragility

An example bridge and a large database of real traffic are used to loading is considered. However, the increased fragility (reduced
demonstrate the approach for the CSUS region. A WB-20 truck is median PGA) can be small once the probabilities of truck pre-
found to be representative of the truck traffic and is used to assess the sences and GVWs are taken into account. Indeed, unless the site-
influence of the presence of a single truck on the seismic fragility. A specific GVW distribution is significantly different from the
representative seismically vulnerable MSC steel-girder bridge from present case study, even for high flow rates, the truck traffic does
the region is also used. It is found that the median value of bridge not have a significant effect on the seismic fragility. It should be
seismic fragility reduces by between 5 and 10% when a truck with noted that only one-truck presences are considered in this work,
a modal GVW of 35.4 t is present at any random location on the and the truck is considered to be positioned at the determined
bridge. This median PGA depends on the location of the truck on the critical location for the duration of the earthquake. In reality, the
bridge, and a critical position of the truck partially on the exterior probability of the truck being at this location could also be con-
and interior spans is identified. For this location, it is found that the sidered, as well as other variables such as the duration of its tra-
median PGA reduces approximately linearly with increasing truck verse, the position of the truck at the time of arrival of the
GVW, yielding a reduction of about 14% for a truck of 60 t GVW earthquake, the dynamic interaction of the moving truck with the
relative to the conventional case where truck loads are neglected. bridge, and finally the height of the truck GVW above the bridge
Consideration of the regional GVW histogram, given the truck deck. Notwithstanding these possible refinements, the presented
presence, leads to a reduction in median PGA of about 6% compared methodology provides a basis for incorporating the considera-
with that of the bridge alone. Finally, for the bridge and traffic tion of truck loads in the seismic fragility estimation of bridges.
studied using the proposed methodology, even for heavy truck Further studies are required to explore the generalization of these
flows, the median PGA is reduced by less than 2%. conclusions to other bridge geometries and traffic loading con-
It is clear from the results presented in this work that it is like- ditions. Additionally, this framework can offer the basis for future
ly that bridges are more fragile during a seismic event if traffic investigation of reliability-based load factors for the case of

© ASCE 04013009-9 J. Bridge Eng.

J. Bridge Eng. 2014.19.

 multiple extreme events: Scour and earthquake.” J. Bridge Eng., 10.1061/
(ASCE)BE.1943-5592.0000369, 362–371.
ASCE. (2009). “ASCE: Infrastructure fact sheet.” Æhttp://www
.infrastructurereportcard.org/sites/default/files/RC2009_bridges.pdfæ
(Mar. 21, 2012).
Basoz, N., and Kiremidjian, A. S. (1998). “Evaluation of bridge damage data
from the Loma Prieta and Northridge, California earthquakes.” Tech-
nical Rep. MCEER-98-0004, Multidisciplinary Center for Earthquake
Engineering Research, Buffalo, NY.
Basoz, N., and Kiremidjian, A. S. (1999). “Development of empirical
Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

fragility curves for bridges.” Optimizing post-earthquake lifeline system

reliability, ASCE, Reston, VA, 693–702.
Basoz, N., and Mander, J. B. (1999). “Enhancement of the highway
transportation module in HAZUS.” Rep. to National Institute of Building
Sciences, National Institute of Building Sciences, Washington, DC.
Benjamin, J. R., and Cornell, C. A. (1970). Probability, statistics, and
decision for civil engineers, McGraw Hill, New York.
Caprani, C. C. (2012). “Calibration of a congestion load model for highway
bridges using traffic microsimulation.” Struct. Eng. Int., 22(3), 342–348.
Chang, S. E. (2000). “Disasters and transport systems: loss, recovery and
competition at the Port of Kobe after the 1995 earthquake.” J. Transp.
Geogr., 8(1), 53–65.
Choe, D.-E., Gardoni, P., Rosowsky, D., and Haukaas, T. (2009). “Seismic
fragility estimates for reinforced concrete bridges subject to corrosion.”
Struct. Saf., 31(4), 275–283.
Choi, E., DesRoches, R., and Nielson, B. G. (2004). “Seismic fragility of
typical bridges in moderate seismic zones.” Eng. Struct., 26(2), 187–199.
Czarnecki, A. A., and Nowak, A. S. (2007). “Reliability-based evaluation of
steel girder bridges.” Proc. Inst. Civ. Eng. Bridge Eng., 160(1), 9–15.
DesRoches, R., Choi, E., Leon, R. T., Dyke, S. J., and Aschheim, M. (2004).
“Seismic response of multiple span steel bridges in central and south-
eastern United States. I: As built.” J. Bridge Eng., 10.1061/(ASCE)1084-
0702(2004)9:5(464), 464–472.
Du, J. S., and Au, F. T. K. (2005). “Deterministic and reliability analysis of
prestressed concrete bridge girders: Comparison of the Chinese, Hong
Kong and AASHTO LRFD codes.” Struct. Saf., 27(3), 230–245.
Federal Highway Administration (FHWA). (2011a). “Federal Highway
Administration bridge formula weights.” Æhttp://ops.fhwa.dot.gov/
freight/publications/brdg_frm_wghts/index.htmæ (Jan. 29, 2013).
Federal Highway Administration (FHWA). (2011b). “Federal Highway
Fig. 9. Influence of traffic loading on seismic fragility: (a) change in Administration vehicle classification.” Æhttp://www.fhwa.dot.gov/policy/
fragility curves for three pertinent cases; (b) change in fragility per- ohpi/vehclass.htmæ (Jan. 29, 2013).
centiles with increasing truck flow Gardoni, P., Der Kiureghian, A., and Mosalam, K. M. (2002). “Probabilistic
capacity models and fragility estimates for reinforced concrete columns
based on experimental observations.” J. Eng. Mech., 10.1061/(ASCE)
0733-9399(2002)128:10(1024), 1024–1038.
combined earthquake and live loads by posing a method to com- Ghosh, J., and Padgett, J. E. (2010). “Aging considerations in the de-
pute the joint fragility for these event occurrences. velopment of time-dependent seismic fragility curves.” J. Struct. Eng.,
10.1061/(ASCE)ST.1943-541X.0000260, 1497–1511.
Ghosn, M., Moses, F., and Wang, J. (2003). Design of highway bridges for
Acknowledgments extreme events, National Academy, Washington, DC.
Haight, F. A. (1963). Mathematical theories of traffic flow, Academic, New
The authors thank the University of Alabama at Birmingham and the York.
Alabama DOT for their generous distribution of the weigh-in- Kunnath, S. K., Larson, L., and Miranda, E. (2006). “Modelling consid-
erations in probabilistic performance-based seismic evaluation: Case
motion data. Additionally, the support provided by the National Sci-
study of the I-880 viaduct.” Earthquake Eng. Struct. Dynam., 35(1),
ence Foundation under Grant No. CMMI-1234690 is gratefully
57–75.
acknowledged. Any opinions, findings, and conclusions or recom- Mackie, K., and Stojadinovic, B. (2001). “Probabilistic seismic demand
mendations expressed in this material are those of the authors and model for California highway bridges.” J. Bridge Eng., 10.1061/
do not necessarily reflect the views of the National Science Founda- (ASCE)1084-0702(2001)6:6(468), 468–481.
tion. Finally, the authors are indebted to the reviewers for construc- Mander, J. B., Kim, D. K., Chen, S. S., and Premus, G. J. (1996). “Response
tive commentary that helped improve the presentation. of steel bridge bearings to reversed cyclic loading.” Æhttps://ir.lib.buffalo
.edu/xmlui/handle/10465/5686æ (Jan. 16, 2012).
Mazzoni, S., McKenna, F., Scott, M. H., and Fenves, G. L. (2009). OpenSees
References command language manual, Univ. of California, Berkeley, CA.
Melchers, R. E. (1999). Structural reliability analysis and prediction, Wiley,
AASHTO. (2012). LRFD bridge design specifications, Washington, DC. New York.
Alipour, A., Shafei, B., and Shinozuka, M. (2013). “Reliability-based cali- National Safety Commission. (2007). “The National Safety Commission
bration of load and resistance factors for design of RC bridges under alerts: Nation’s road infrastructure under great stress.” Æhttp://alerts

© ASCE 04013009-10 J. Bridge Eng.

J. Bridge Eng. 2014.19.

 .nationalsafetycommission.com/2007/09/nations-road-infrastructure-under- Padgett, J. E., and DesRoches, R. (2009). “Retrofitted bridge fragility
great.phpæ (Mar. 25, 2012). analysis for typical classes of multispan bridges.” Earthq. Spectra, 25(1),
Nielson, B. G., and DesRoches, R. (2007a). “Analytical seismic fragility 117–141.
curves for typical bridges in the central and southeastern United States.” Padleckas, H. (2011). “Conventional 18-wheeler truck diagram.” Æhttp://
Earthq. Spectra, 23(3), 615–633. commons.wikimedia.org/wiki/File:Conventional_18-wheeler_truck_diagram
Nielson, B. G., and DesRoches, R. (2007b). “Seismic fragility methodo- .svgæ (Jul. 25, 2012).
logy for highway bridges using a component level approach.” Earth- Rix, G. J., and Fernandez, J. A. (2004). “Earthquake ground motion sim-
quake Eng. Struct. Dynam., 36(6), 823–839. ulation.” Æwww.ce.gatech.edu/research/mae_ground_motion/æ (Oct. 19,
Nowak, A. S., and Cho, T. (2007). “Prediction of the combination of failure 2009).
modes for an arch bridge system.” J. Constr. Steel Res., 63(12), 1561–1569. Shinozuka, M., Feng, M. Q., Kim, H.-K., and Kim, S.-H. (2000). “Nonlinear
Downloaded from ascelibrary.org by INDIAN INST OF TECHNOLOGY- Mumbai (IITM) on 02/16/15. Copyright ASCE. For personal use only; all rights reserved.

Nowak, A. S., Park, C. H., and Casas, J. R. (2001). “Reliability analysis of static procedure for fragility curve development.” J. Eng. Mech.,
prestressed concrete bridge girders: Comparison of Eurocode, Spanish 10.1061/(ASCE)0733-9399(2000)126:12(1287), 1287–1295.
Norma IAP and AASHTO LRFD.” Struct. Saf., 23(4), 331–344. Straub, D., and Der Kiureghian, A. (2008). “Improved seismic fragility
O’Brien, E. J., and Caprani, C. C. (2005). “Headway modelling for traffic load modeling from empirical data.” Struct. Saf., 30(4), 320–336.
assessment of short to medium span bridges.” Struct. Eng., 83(16), 33–36. Transportation Research Board. (2003). Review of truck characteristics as
O’Brien, E. J., and Enright, B. (2011). “Modeling same-direction two-lane factors in roadway design, National Cooperative Highway Research
traffic for bridge loading.” Struct. Saf., 33(4-5), 296–304. Program, Washington, DC.
OpenSees 2.4.3 [Computer software]. Berkeley, CA, Pacific Earthquake Wen, Y. K., and Wu, C. L. (2001). “Uniform hazard ground motions for
Engineering Research Center. Mid-America cities.” Earthq. Spectra, 17(2), 359–384.
Padgett, J. E., and DesRoches, R. (2007). “Bridge functionality relation- Wright, T., DesRoches, R., and Padgett, J. E. (2011). “Bridge seismic
ships for improved seismic risk assessment of transportation networks.” retrofitting practices in the central and southeastern United States.”
Earthq. Spectra, 23(1), 115–130. J. Bridge Eng., 10.1061/(ASCE)BE.1943-5592.0000128, 82–92.

© ASCE 04013009-11 J. Bridge Eng.

J. Bridge Eng. 2014.19.