You are on page 1of 13

Computer-Aided Civil and Infrastructure Engineering 22 (2007) 293–305

Nonlinear Static Procedure for Seismic Vulnerability


Assessment of Bridges
S. Banerjee & M. Shinozuka∗
Department of Civil and Environmental Engineering, University of California, Irvine, Irvine, CA, 92697

Abstract: The impact of an earthquake event on the per- by means of graphical representation. Over the past few
formance of a highway transportation network depends years, CSM has been studied and applied for the seismic
on the extent of damage sustained by its individual com- evaluation of structures, mainly RC buildings. The Ap-
ponents, particularly bridges. Seismic damageability of plied Technology Council (ATC-40, 1996) documented
bridges expressed in the form of fragility curves can eas- nonlinear static analysis procedures including CSM for
ily be incorporated into the scheme of risk analysis of a seismic evaluation of concrete buildings. The same con-
highway network under the seismic hazard. In this con- cept is currently under investigation for use in bridge
text, this article focuses on a nonlinear static method of analysis, design, and seismic evaluation (Fajfar et al.,
developing fragility curves for a typical type of concrete 1997; Abeysinghe et al., 2002; and Zheng at al., 2003).
bridge in California. The method makes use of the capac- Although analytical methods are available to perform
ity spectrum method (CSM) for identification of spectral nonlinear time history analysis, this method, as more
displacement, which is converted to rotations at bridge practice oriented, attracted more interest and support
column ends. To check the reliability of this current ana- from the profession.
lytical procedure, developed fragility curves are compared Fishinger et al. (1997) performed seismic analysis of
with those obtained by nonlinear time history analysis. Re- viaducts with continuous deck using elastic and inelastic
sults indicate that analytically developed fragility curves methods and showed that elastic analysis failed to pre-
obtained from nonlinear static and time history analyses dict response in many cases. Shinozuka et al. (2000a) de-
are consistent. veloped fragility curves of Memphis bridges using CSM
following the procedure prescribed in ATC-40 (ATC,
1996). They compared these analytical fragility curves
1 INTRODUCTION with those computed by nonlinear time history analysis
and concluded that fragility curves obtained from these
Use of fragility curves is an increasingly more significant two analytical procedures were in a good agreement for
component of seismic risk assessment not only of individ- the state of minor damage but did not match very well
ual structures, but also of urban infrastructure systems for the state of major damage. Recently, pushover analy-
including highway networks in seismically active regions. sis is carried out for the evaluation of the Greveniotikos
This article develops a nonlinear static method, making bridge in Greece (Abeysinghe et al., 2002). All these re-
use of the capacity spectrum method (CSM), for the pur- cent studies indicate that the use of CSM for accurate
pose of constructing the fragility curves for a typical type estimation of bridge performance under seismic ground
of concrete bridge in California. motion is now a great interest to researchers.
CSM was first introduced in the 1970s for a rapid evalu- In order to establish a standardized procedure, this
ation of building performance under earthquake ground article focuses on nonlinear static methods to analyze
motion (Freeman, 1998). It correlates structural capac- a multispan reinforced concrete bridge under sixty (60)
ity with demand related to earthquake ground motion ground acceleration time histories in the Los Angeles
area developed originally for the Federal Emergency
∗ To whom correspondence should be addressed. E-mail: shino@ Management Agency (FEMA) SAC (SEAOC-ATC-
uci.edu. CUREe) steel project. A class of Caltrans’ bridges with


C 2007 Computer-Aided Civil and Infrastructure Engineering. Published by Blackwell Publishing, 350 Main Street, Malden, MA 02148, USA,
and 9600 Garsington Road, Oxford OX4 2DQ, UK.
294 Banerjee & Shinozuka

multiple span, skew angle 0◦ –20◦ , and supported on soil through the interaction of capacity and elastic (5%
type C (i.e., soft soil as per UBC 1993) is represented damped) demand spectrum. Fragility curves at different
by this example bridge. The analysis is performed with damage levels are generated by numerically simulating
aid of the finite element computer code “SAP2000 NL” performance displacements that are obtained from per-
(Computer and Structures, 2002). Two spectra are the formance points. To check the consistency of this cur-
key elements in CSM: “demand spectrum,” representing rent analytical procedure, developed fragility curves are
intensity of the seismic ground motion to which bridges compared with those obtained from time history analy-
are subjected, and “capacity spectrum,” representing the sis of the example bridge. Though one example bridge
bridges’ ability to resist the seismic demand. The “perfor- is analyzed here using SAP2000, the methodology and
mance point” indicates the intersection of demand and procedure are general enough to utilize in other com-
capacity spectra. Because of the hysteretic damping of puter codes and applicable for other reinforced concrete
structures during excitation, elastic response spectrum bridges.
(5% damped) should be reduced to inelastic response
spectrum by an appropriate spectral reduction factor. 2 NONLINEAR MODELING OF THE BRIDGE
Here one simplified approach (Reinhorn, 1997), alterna-
tive to the current professional design procedure given To demonstrate this methodology, a 242 m long, five span
in ATC-40 (1996) for building structure, is followed to Caltrans’ bridge (Sultan and Kawashima, 1994) with one
generate inelastic demand spectrum for which reduc- expansion joint is considered. Figure 1a shows geometric
tion factor due to structural nonlinearity is estimated configuration and boundary conditions of the example

242.0 m
41.18 m 53.38 m 53.38 m 53.38 m 41.18 m

10.7 m
21.0 m

(a)

Embankment Deck Deck

Linear
Column
Force

Abutment

Expansion Joint Potential Plastic


5.08 cm
Relative M Hinge
Force

Displacement My
kGap kHook
2.54 cm
1.27 cm Relative
Displacement
kGap
θy Rotation (θ)

(b)
Fig. 1. (a) Bridge model under consideration (not in scale); (b) Nonlinearities in bridge model.
Seismic vulnerability assessment of bridges 295

bridge under consideration, and Figure 1b represents the 50000


approximate nonlinearities in it. The bridge is supported
on four identical 21 m high circular RC columns of di- 40000
αPYk

Moment (kN-m)
ameter 2.4 m. The columns are of the same sectional and
material properties. The deck is made of a 3-cell con- 30000
My = 36493 kN-m
crete box girder with a cross-section of 13 × 2 m. Young’s θy = 3.73x10-3 rad
20000
modulus and mass density of concrete are taken as, re- k αPY = 0.01675
spectively, 27.79 GPa and 2.40 MN/m3 with 5% damping
10000
ratio at each mode of vibration. Computed Bi-linear
Finite element computer code SAP2000 NL is used in 0
the ensuing nonlinear analysis. The bridge deck is inte- 0.00 0.01 0.02 0.03 0.04
grated with the column bents, so full continuity is con-
Rotation (rad)
sidered at its girder-column joints. Bridge superstructure
is free to rotate at abutment locations for longitudinal Fig. 2. Bilinear moment–rotation curve.
excitation, although its translational motion is limited
to the initially provided gap between the bridge girder
and abutments. Axial force develops at the interface due tion in moment–rotation relationship of bridge columns
to the passive earth pressure of embankment soil if the under several earthquake ground motions. From this ex-
girder comes into contact with abutments. According to ercise, little difference in girder displacement and col-
Caltrans recommendations (California Department of umn rotation is observed when bridge deformation is in
Transportation, 2004), the longitudinal abutment stiff- moderate range (i.e., at low to moderate performance
ness is level of the bridge). Only when column rotation is close
  to its ultimate point, the degradation model develops
h
Kabut = Ki × w × in S.I. units (1) somewhat larger deformation (typically 8% more) than
1.7 the bilinear model.
where Ki is the initial embankment stiffness At the expansion joint, the example bridge is mod-
(=11.5 kN/mm/m), w is the width of the backfill eled such that the two ends of the expansion joint can
in m and h is the height of the backfill in m. During move independently in the longitudinal direction and
transverse out-of-plane motion, it can rotate freely rotate in longitudinal plane although they have no rel-
although movement is restrained by wingwalls and ative vertical movement. During out-of-plane motion,
concrete shear keys which are, in general, assumed to be they are assumed as pin connections as lateral transla-
rigidly connected with abutments and do not dissipate tions of two ends of the bridge deck at expansion joint
any energy through yielding (Federal Highway Admin- are the same. The opening and closure of the joint is
istration, 1996). Therefore, during transverse motion, modeled using hook and gap element. The gap element
the bridge deck and abutments move together as rigidly represents the effect of pounding between two adjacent
joined. To incorporate soil effects behind wingwalls at bridge decks at the expansion joint and abutment loca-
abutments, translational (linear) springs are attached tions during longitudinal movement of the bridge. At
at the end of the bridge girder. Spring stiffness is abutment locations, an initial gap of 0.0508 m (2 in) and
determined according to Caltrans recommendations for a linear spring with stiffness of 0.29 × 106 kN are pro-
abutment stiffness in transverse direction (California vided between the bridge deck and abutment (Figure
Department of Transportation, 2005). 3a) that jointly represent the gap element at this loca-
Due to seismic excitation, bending moment is gener- tion. The gap element at the expansion joint is modeled
ated in columns, which may lead to the formation of as a linear spring having stiffness 1.5 × 109 kN (Kim
plastic hinges at both ends of bridge columns. Moment– and Shinozuka, 2003). Here the initially provided gap
rotation relationship at plastic hinges generates complex is 0.0254 m (1 in) (Figure 3b) and pounding develops
hysteretic behavior. For the sake of simplicity, how-
ever, bilinear springs are used to represent this nonlin- 0.0508
0.0254m
m 0.0254
0.0254mm
k =k 0.29x106 kN
0.013 m
k =k1.5x106 kN
earity at both ends of bridge columns. The computed j i
k =k 1.5x109 kN
j i j
j i j i j
and approximated bilinear moment–rotation curves of
bridge columns are derived based on Priestly et al. (1996)
and showed in Figure 2. In this context, two separate (a) Gap Element for Pounding (b) Gap Element for Pounding (c) Hook Element for

sets of time domain analysis of a reinforced concrete at Abutment at Expansion Joint Restrainer at Expansion Joint
bridge have been performed considering bilinear hys-
teretic behavior, and stiffness and strength degrada- Fig. 3. Nonlinear modeling of gap and hook elements.
296 Banerjee & Shinozuka

(compressive force) at the interface of the two adjacent rule is applied to combine all modal parameters of these
bridge decks when relative displacement exhausts this two modes in order to determine “lateral” forces at each
initial gap width. The hook element represents the re- node of the bridge for the longitudinal pushover analysis.
straining cable that can be severed under excessive ten- Figures 5a and b show the capacity curves of the bridge
sile stress, or producing failure of anchorages through obtained from the longitudinal pushover analysis which
which restrainers are tied to decks. Figure 3c shows the represent, respectively, the overall base shear force as a
hook element as a linear string with stiffness 1.5 × 106 function of the bridge girder displacement and rotation
kN/m and initial slack of 0.013 m (0.5 in), and axial force at the plastic hinge region of the bridge column. In both
generates when the restrainer gets engaged by loosing cases (Figures 5a and b), yield states are obtained from
this initial slack. Also linear translational and rotational bilinear approximation.
springs are introduced at column bases to account for According to the general trend of CSM, it is required
soil effect. An effective moment of inertia is considered to convert the capacity curve to the capacity spectrum
to account for the cracked state of concrete columns. in acceleration–displacement response spectra (ADRS)
format. By definition, the acceleration–displacement re-
sponse spectrum represents the variation of spectral ac-
3 NONLINEAR STATIC ANALYSIS celeration (Sa ) with spectral displacement (Sd ), which
IN THE LONGITUDINAL DIRECTION are respectively, the maximum values of absolute ac-
celeration and relative displacement of a SDOF system
3.1 Development of capacity spectrum under a certain damping ratio. These measures can be
from capacity curve computed using the information obtained from capacity
analysis of the example bridge as given in Equations (3)
Capacity curve is given in the form of a force-
and (4):
displacement curve that represents the capacity of a
structure within and beyond elastic limit. Computation V/W
of total shear force generated at bridge supports as a Sa = (3)
α
function of displacement of the superstructure consti-
tutes the capacity curve by definition. This procedure is girder
also referred to as pushover analysis. To generate the ca- Sd = (4)
PFφgirder
pacity curve for building structures, horizontal displace-
ment at roof level is considered as a most critical dis- where W is overall dead weight of bridge, girder is hori-
placement component. However, for bridge structures zontal displacement of girder, φ girder is amplitude of the
displacement of the bridge girder in the longitudinal di- fundamental mode at girder, α and PF are modal mass
rection is used to develop the capacity curve. To develop coefficient and modal participation factor of the funda-
this plot, static nonlinear analysis (pushover analysis) is mental mode defined as follows
performed using SAP2000 NL computer code.
 2
As stated in ATC-40 (1996) the “lateral” forces (i.e., N
forces in a horizontal plane) are applied in proportion to (wi φi )/g
the fundamental mode shape, and are described as α=
i=1
  (5)
   
N 
N


N
wi /g wi φi2 g
Fi = wi φi wi φi V (2) i=1 i=1
i=1

where F i is the “lateral” force on node i (i = 1, 2, . . . , N),  


wi is the dead weight assigned to node i, φ i is the ampli- 
N

tude of the fundamental mode at node i, V is the base  (wi φi )/g 


 i=1 
shear and N is the total number of nodes. PF = 


 (6)
The example bridge has the 1st mode (T1 = 1.94 sec- 
N


wi φi g
2
onds) of vibration in the transverse direction and the i=1
next two modes (T2 = 1.56 seconds and T3 = 1.39 sec-
onds) are in the longitudinal direction as shown in Figure For the longitudinal pushover analysis of the exam-
4. This figure also schematically shows all nodal points ple bridge, the SRSS rule is applied to combine modal
where lateral loads are applied. In this study, these three amplitudes of modes 2 and 3 in order to compute α
modes are captured and incorporated into the lateral and PF. These values are estimated as 0.911 and 12.65,
load pattern for pushover analysis. As second (2nd) and respectively. Figure 6 shows the capacity spectrum of the
third (3rd) modes are very close to each other, SRSS bridge in the longitudinal direction.
Seismic vulnerability assessment of bridges 297

Fig. 4. (a) Schematic model and (b) mode shapes of the example bridge.

3.2 Development of demand spectrum Figure 7 shows the horizontal acceleration time history,
elastic acceleration response spectrum (5% damped)
Sixty (60) ground acceleration time histories in the Los
and elastic demand spectrum in ADRS format of ground
Angeles area developed originally for the Federal Emer-
motion from the 1940 El Centro Earthquake (LA01,
gency Management Agency (FEMA) SAC steel project
PGA: 452.03 cm/second2 ). It is important to note that
(http://nisee.berkeley.edu/data/strong motion/sacsteel/
each set of 20 ground motions is scaled linearly such that
ground motions.html) are utilized here. They consist
return periods are the same but not PGA.
of 3 sets of 20 actual earthquake records, each set
linearly scaled so as to have return periods of 2,500
years, 475 years, and 72 years and are representative of 3.3 Performance point
earthquakes with exceedance probabilities, respectively
of 2%, 10%, and 50% in 50 years. For the purpose Performance point is the intersecting point of capac-
of CSM, elastic acceleration response spectra (5% ity spectrum and demand spectrum. To incorporate the
damped) are generated from these acceleration time energy dissipation of structures through hysteretic be-
histories. Furthermore, these are converted to ADRS havior beyond yield, the elastic demand curve of an
format utilizing Equation (7) and referred to as elastic earthquake ground motion should be reduced with a
demand spectrum. proper reduction factor. ATC-40 (1996) suggests spec-
tral reduction factors for a range of constant peak spec-
tral acceleration and constant peak spectral velocity. But
T2 a demand spectrum of ground motion contains spectral
Sd = Sa g (7) acceleration and displacement. Therefore, it is necessary
4π 2
298 Banerjee & Shinozuka

20000 0.6

Acceleration (g)
0.3
Total Base Shear, V (kN)

15000 0
-0.3
-0.6
10000
0 10 20 30 40 50
Time (sec)
5000
(a)
2.5

Spectral acceleration, Sa (g)


0
0 0.2 0.4 0.6 0.8 1 2.0

Displacement of the Bridge Girder (m)


1.5
(a)
1.0
20000
Total Base Shear, V (kN)

0.5
15000

0.0
10000 0 1 2 3 4 5 6
Time, T (sec)
(b)
5000
2.5
Spectral acceleration, Sa (g)

0
2.0
0 0.01 0.02 0.03 0.04
Rotation at Plastic Hinge Region (rad)
1.5

(b)
1.0
Fig. 5. Capacity curves in the longitudinal direction
of the bridge. 0.5

that spectral reduction factors should be computed in 0.0


terms of spectral acceleration and displacement. Re- 0 0.2 0.4 0.6 0.8 1
cent studies developed a few methods to determine in-
Spectral displacement, Sd (m)
elastic demand spectra from elastic spectra (Krawinkler (c)

Fig. 7. The 1940 El Centro Earthquake (LA01); (a)


0.4 acceleration time history; (b) elastic acceleration response
spectrum (5% damped); and (c) elastic demand spectrum in
Spectral Acceleration, Sa (g)

ADRS format.
0.3

0.2
and Nassar, 1992; Miranda and Bertero, 1994; Reinhorn,
1997; Fajfar, 1999; and Rossetto and Elnashai, 2005).
0.1 Among these, the procedure given by Reinhorn (1997)
is used to produce the inelastic demand spectrum and,
hence the inelastic response. According to this pro-
0.0
cedure, the inelastic spectral displacement (Sin d ) and
0.0 0.2 0.4 0.6 0.8 1.0
inelastic spectral acceleration (Sin
a ) can be expressed as
Spectral Displacement, Sd (m)
 
Fig. 6. Capacity spectrum in the longitudinal direction Sde 1 Se
Sdin = 1 + (Rc − 1) ≥ d (8)
of the bridge. R c R
Seismic vulnerability assessment of bridges 299

2.5
Capacity Spectrum Capacity Spectrum

Spectral acceleration, Sa (g)


2.0 Elastic Demand Spectrum
Elastic Demand Spectrum
Q/W = Sa/g

Inelastic Demand Spectrum

1.5
Performance
Displacement
1.0

Qe/W
0.5
y
Q /W

0.0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Spectral displacement, Sd (m) Spectral displacement, Sd (m)

Fig. 8. Estimation of spectral reduction factor, R. Fig. 9. Calculation of performance displacement for LA01.

  
Sae 1 c pends on these two spectra and takes different values de-
Sain = 1 + αPY (R − 1) (9) pending on structures and earthquake ground motions.
R c
where Sde and Sae , respectively, are the elastic spectral dis-
placement and acceleration, R is the reduction factor, c 3.4 Development of fragility curves
is a constant and α PY is the post-yield hardening coeffi- For sixty earthquake ground motions, sixty performance
cient of the structure. The reduction factor R is defined displacements (Sd ) are computed separately from inter-
as the ratio of maximum elastic force (Qe /W) to yield action of capacity and inelastic demand spectra and con-
force (Qy /W) as shown in Figure 8. The constant c can verted to the displacement of the bridge girder with the
be computed as aid of following equation obtainable immediately from
Toa b Equation (4). By doing so, the spectral displacement
c= + (10) is converted to the physical displacement of the bridge
1 + Toa To
girder.
where T o is the initial time period of the structure, a and
b are factors as given in Table 1 (Krawinkler and Nassar, girder = Sd × PF × φgirder (11)
1992). Corresponding to these estimated displacements of
the bridge girder for all sixty ground motions, rota-
Following this procedure, 60 inelastic demand spectra tions at plastic hinge regions of bridge columns are com-
are generated from corresponding elastic demand spec- puted and converted to “rotational ductility demand.”
tra. Figure 9 shows the inelastic demand spectrum gen- The term “rotational ductility demand” is the signa-
erated from the elastic spectrum of the 1940 El Centro ture representing seismic performance levels and de-
earthquake (LA01). fined as the ratio of rotation (θ) of the plastic hinge
The intersecting point of capacity and inelastic de- region of the bridge to its yield rotation (θ y ). There-
mand spectra represents performance displacement (Sd ) fore, rotational ductility demands at yield point and ul-
of the example bridge under that particular ground mo- timate point of the example bridge are, respectively,
tion (Figure 9). It should be noted that this procedure is 1.0 and 9.16. The performance of this bridge under 60
interactive between the capacity spectrum and the elastic ground motions is categorized in five damage states
demand spectrum. Hence, the reduction factor (R) de- namely “almost no,” “minor,” “moderate,” “major”
and “collapse.” Threshold rotational ductility demands
Table 1 at these five damage states are estimated in propor-
Coefficients as given in Krawinkler and Nassar (1992) tion to the damage state definition given by Dutta and
Mander (1998) and presented in Table 2. In order to
α PY a b estimate the damage level of the bridge under one par-
ticular earthquake ground motion, the computed rota-
0% 1.0 0.42 tional ductility at the plastic hinge region is compared
2% 1.0 0.37
with the threshold rotational ductility demand at each
10% 0.80 0.29
damage state.
300 Banerjee & Shinozuka

Table 2 Table 3
Drift ratios and threshold rotational ductility demands at five Fragility parameters in the longitudinal direction of the bridge
damage states of the bridge
Fragility parameters
Drift ratios
(Dutta and Threshold rotational Damage state c(g) ζ
Damage state Mander, 1998) ductility demand Almost no 0.2184
Minor 0.4020
Almost no Yield 1.00
Moderate 0.7122 1.249
Minor 1.0% 1.58
Major 1.1520
Moderate 2.5% 3.33
Collapse 1.4340
Major 5.0% 6.24
Collapse 7.5% 9.16

In the past few years many authors performed bridge Pi1 = F(ai , c1 , ζ ) − F(ai , c2 , ζ ) (15)
fragility analysis following different approaches (Hwang
and Huo, 1994; Fukushima et al., 1996; Singhal and Pi2 = F(ai , c2 , ζ ) − F(ai , c3 , ζ ) (16)
Kiremidjian, 1998; Mander and Basöz, 1999; Shinozuka
et al., 2000b; Karim and Yamazaki, 2001; Gardoni et al.,
2002; Shinozuka et al., 2003; Choi et al., 2004; and Pi3 = F(ai , c3 , ζ ) − F(ai , c4 , ζ ) (17)
Rossetto and Elnashai, 2005). In the present study, ana-
lytical fragility curve corresponding to each damage state
is expressed in the form of a two-parameter lognormal Pi4 = F(ai , c4 , ζ ) − F(ai , c5 , ζ ) (18)
distribution function as given in Shinozuka et al. (2003).
The fragility parameters (median, c, and log-standard de- Pi5 = F(ai , c5 , ζ ) (19)
viation, ζ ) are estimated through a maximum likelihood
method such that these fragility curves do not cross each The fragility parameters are obtained by solving
other. Therefore, an equal value of ζ is needed to satisfy Equation (20) and are listed in Table 3. Also, Figure 10
this condition. Under this lognormal assumption, the an- shows the fragility curves of the example bridge for all
alytical form of the fragility function F(•) for the state damage states.
of damage at least k is,
   ∂ ln L(c1 , c2 , c3 , c4 , c5, ζ )
ln caki ∂ck
F(ai , ck, ζ ) =    (12)
ζ ∂ ln L(c1 , c2 , c3 , c4 , c5, ζ )
= =0 ∀k (20)
∂ζ
where ck is the median of the fragility function associ-
ated with the damage state of at least k, ζ is the common
log-standard deviation, ai is the PGA value to which the
bridge is subjected and [•] is the standardized normal
Probability of Exceeding a Damage

1
distribution function. The fragility parameters are com- Almost No
Minor
puted by maximizing the likelihood function, L, which is 0.8
Moderate
given by, Major
0.6

5 
State

n Collapse
L(c1 , c2 , c3 , c4 , c5, ζ ) = [Pik]xik (13) 0.4
k=0 i=1

where xik is 1 or 0, depending on whether or not the 0.2


bridge sustains damage state k under ai , and n is the total
number of ground motions under which the analysis is 0
0 0.2 0.4 0.6 0.8 1
performed. Pik is the probability that the example bridge
will suffer from a damage state k when subjected to ai PGA (g)
and is expressed as
Fig. 10. Fragility curves in the longitudinal direction of the
Pi0 = 1 − F(ai , c1 , ζ ) (14) bridge for five damage states.
Seismic vulnerability assessment of bridges 301

16000 0.5

Spectral Acceleration, Sa (g)


Total Base Shear, V (kN)

0.4
12000

0.3
8000
0.2

4000
0.1

0 0.0
0 0.2 0.4 0.6 0.8 1 0.0 0.2 0.4 0.6 0.8 1.0
Displacement of the Bridge Girder (m) Spectral Displacement, Sd (m)

(a) Fig. 12. Capacity spectrum in the transverse direction of the


bridge.
16000
Total Base Shear, V (kN)

Probability of Exceeding a Damage


12000 Almost No
0.8 Minor
Moderate
8000 Major
0.6
Collapse
State
4000 0.4

0.2
0
0 0.01 0.02 0.03 0.04
0
Rotation at Plastic Hinge Region (rad) 0 0.2 0.4 0.6 0.8 1
PGA (g)
(b)
Fig. 13. Fragility curves in the transverse direction of the
Fig. 11. Capacity curves in the transverse direction of the
bridge for five damage states.
bridge.

5 TIME HISTORY ANALYSIS OF BRIDGE


4 NONLINEAR STATIC ANALYSIS
IN THE TRANSVERSE DIRECTION The example bridge model with all nonlinear proper-
ties and boundary conditions mentioned earlier is ana-
As previously mentioned, the example bridge has its fun- lyzed in time domain for 60 earthquake ground motions
damental mode of vibration (T1 = 1.95 seconds) in the in both longitudinal and transverse directions. SAP2000
transverse direction. In this mode, observed α and PF NL computer code is again utilized in this analysis. The
values are 0.70 and 15.39, respectively. Following the pro- time history of structural responses are computed in
cedure described earlier in this article, the capacity curve
of the bridge in the transverse direction is developed and
plotted in Figure 11 and afterward, converted to the ca- Table 4
pacity spectrum as shown in Figure 12. Fragility parameters in the transverse direction of the bridge
Fragility analysis in the transverse direction of the
bridge is performed as described before. Analytical Fragility parameters
fragility curves of the example bridge for five damage Damage state c(g) ζ
states are shown in Figure 13 and fragility parameters
are listed in Table 4. Result indicates that for all damage Almost no 0.2680
states except for “Almost No,” the bridge is more frag- Minor 0.3880
Moderate 0.6408 0.96
ile in the transverse direction. This phenomenon could
Major 0.8929
be different for other bridges with no in-span expansion Collapse 1.0745
joint.
302 Banerjee & Shinozuka

terms of the ductility demand at top and bottom of each Table 5


column, the tensile force developed in restrainer and Fragility parameters obtained from time history analysis
the pounding force developed at the interface of adja-
cent bridge decks at the expansion joint and abutment Fragility parameters Fragility parameters
locations due to face-to-face impact during oscillation. in longitudinal in transverse
For the evaluation of seismic performance, possible fail- direction direction
ure modes and mechanisms of the example bridge under Damage state c(g) ζ c(g) ζ
seismic excitation are examined independently for each
time history through a rigorous computation (Banerjee Almost no 0.245 0.212
and Shinozuka, 2004). The result indicated that depend- Minor 0.365 0.295
Moderate 0.638 0.932 0.617 0.968
ing on the values of restrainer capacity and pounding
Major 1.060 0.856
force, failure at expansion joint is preceded by the for- Collapse 1.610 1.738
mation of plastic hinges at the column ends. Hence, the
expansion joint does not influence seismic performance
analysis of the example bridge.
Following the procedure given by Shinozuka et al.
(2003), analytical fragility curves are developed as two- should be noted that fragility curves are represented by
parameter lognormal distribution functions. To be con- lognormal distribution functions and constructed inde-
sistent with the fragility curves developed by nonlinear pendently. Therefore, to avoid the intersection of two
static procedure, the definition of damage states is kept fragility curves it is assumed that their log-standard
unaltered (Table 2). The fragility curves in the longitudi- deviations are identical. Figures 15 and 16 show the
nal and transverse directions of the example bridge are comparison in “almost no,” “minor,” “moderate,” and
shown in Figures 14 and fragility parameters are given “major” damage states of the example bridge in the
in Table 5. longitudinal and transverse direction, respectively. The
comparison is made in terms of median values of fragility
curves and common log-standard deviations of 1.03 and
6 COMPARISON OF ANALYTICAL 0.96 are used to construct all of these curves, respectively,
FRAGILITY CURVES in the longitudinal and transverse direction. The result
indicates, in both directions analytical fragility curves
To check the consistency of this current nonlinear static derived from pushover analysis are well in accordance
analytical procedure, developed fragility curves are com- with those from time history analysis for all four damage
pared with those obtained from nonlinear time history states. Therefore this nonlinear static analysis method
analysis in both longitudinal and transverse directions. It can be used as an alternative to the nonlinear time his-
tory analysis in assessing the seismic performance of
bridges. Also this provides confidence to use this method
1 to conduct seismic performance analysis of bridges when
Probability of Exceeding a Damage

ground motion comes from any arbitrary direction to


0.8 the bridge. A recent study by authors (will be published)
showed that the analytical fragility curves developed in
0.6 this study are consistent with the fragility curves given
State

in HAZUS (HAZUS, 1999).


0.4

0.2
7 CONCLUSION
0
0 0.2 0.4 0.6 0.8 1 This study performs seismic vulnerability assessment of
PGA (g) one reinforced concrete bridge by making use of CSM
Almost No_Longitudinal Almost No_Transverse through a considerable amount of computational ef-
Minor_Longitudinal Minor_Transverse fort. Fragility curves are developed for different dam-
Moderate_Longitudinal Moderate_Transverse age states to give a physical interpretation of the failure
Major_Longitudinal Major_Transverse
Collapse_Longitudinal Collapse_Transverse probability of the bridge. A straightforward approach al-
ternative to the suggested procedure in ATC-40 (1996)
Fig. 14. Fragility curves of the bridge for five damage states is used to compute the reduction in the elastic response
developed from time history analysis. spectrum under 5% damping due to hysteretic energy
Seismic vulnerability assessment of bridges 303

1 1

Probability of Exceeding a Damage

Probability of Exceeding a Damage


0.8 0.8

0.6 0.6

State

State
0.4 0.4

0.2 Pushover Analysis (median = 0.218g) 0.2 Pushover Analysis (median = 0.402g)
Time History Analysis (median = 0.245g)
Time History Analysis (median = 0.365g)
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
PGA (g) PGA (g)

(a) Almost No Damage (b ) M i no r D a m a ge

1 1
Probability of Exceeding a Damage

Probability of Exceeding a Damage


Pushover Analysis (median = 0.712g) Pushover Analysis (median = 1.152g)

0.8 Time History Analysis (median = 0.638g) 0.8 Time History Analysis (median = 1.060g)

0.6 0.6
State

State
0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
PGA (g) PGA (g)

(c) Moderate Damage (d) Major Damage

Fig. 15. Comparison of fragility curves in the longitudinal direction of the bridge.

1 1
Probability of Exceeding a Damage

Probability of Exceeding a Damage

0.8 0.8

0.6 0.6
State

State

0.4 0.4

0.2 Pushover Analysis (median = 0.268g) 0.2 Pushover Analysis (median = 0.388g)
Time History Analysis (median = 0.212g)
Time History Analysis (median = 0.295g)
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
PGA (g) PGA (g)

(a) Almost No Damage (b ) M i no r D a m a ge

1 1
Probability of Exceeding a Damage

Probability of Exceeding a Damage

Pushover Analysis (median = 0.641g) Pushover Analysis (median = 0.893g)

0.8 Time History Analysis (median = 0.617g) 0.8 Time History Analysis (median = 0.856g)

0.6 0.6
State

State

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
PGA (g) PGA (g)

(c) Moderate Damage (d) Major Damage

Fig. 16. Comparison of fragility curves in the transverse direction of the bridge.
304 Banerjee & Shinozuka

dissipation. Obviously the bridge model can be improved Federal Highway Administration. (1996), Seismic design of
and different set of failure modes can be added further bridges, Design Example No. 1 FHWA-SA-97-006.
although the basic methodology will remain, in principle, Freeman, S. A. (1998), Development and use of the capacity
spectrum method, in Proceedings of the 6th U.S. National
the same. Conference on Earthquake Engineering, EERI, Seattle,
Developed analytical fragility curves in both lon- Washington, Paper No. 269, on CD-ROM.
gitudinal and transverse directions of the bridge are Fishinger, M., Isakovic, T. & Fajfar, P. (1997), Seismic analysis
compared with those obtained from the nonlinear time of viaduct structures—Which method to choose? in P. Fajfar
history analysis. Comparison indicates that the agree- and H. Krawinkler (eds.), Seismic Design Methodologies for
the Next Generation of Codes, A. A. Balkema Publishers,
ments of analytical fragility curves are excellent for the Rotterdam, pp. 347–58.
states of “almost no,” “minor,” “moderate,” and “ma- Fukushima, S., Kai, Y. & Yashiro, K. (1996), Study on the
jor” damages. Therefore this nonlinear static analysis fragility of system—Part 1: Structure with brittle elements
method provides a more practice-oriented alternative in its stories, in Proceedings of the 11th World Conference on
to the nonlinear time history analysis in assessing the Earthquake Engineering, Pergamon, Elsevier Science Ltd.,
Oxford, UK, Paper No. 333.
seismic performance of bridges. Gardoni, P., Der Kiureghian, A. & Mosalam, K. M. (2002),
Probabilistic models and fragility estimates for bridge com-
ponents and systems, Technical Report PEER 2002/13, Pa-
cific Earthquake Engineering Research Center, University
ACKNOWLEDGMENTS of California, Berkeley, CA.
HAZUS (1999), Earthquake loss estimation methodology,
This study was supported by the Federal Highway Technical Manual (SR2), Federal Emergency Management
Agency through agreements with National Institute of
Administration under contract DTFH61-98-C-00094 Building Science, Washington, DC.
through the Multidisciplinary Center for Earthquake Hwang, H. H. M. & Huo, J.-R. (1994), Generation of hazard-
Engineering Research (MCEER) in Buffalo, NY. consistent fragility curves for seismic loss estimation stud-
ies, Technical Report NCEER-94-0015, National Center for
Earthquake Engineering Research (NCEER), State Uni-
versity of New York at Buffalo, NY.
Karim, K. R. & Yamazaki, F. (2001), Effect of earthquake
REFERENCES ground motions on fragility curves of highway bridge piers
based on numerical simulation, Earthquake Engineering &
Abeysinghe, R. S., Gavaise, E., Rosignoli, M. & Tzaveas, Structural Dynamics, 30(12), 1839–56.
T. (2002), Pushover analysis of inelastic seismic behavior Kim, S-H. & Shinozuka, M. (2003), Effects of seismically in-
of Greveniotikos bridge, Journal of Bridge Engineering, duced pounding at expansion joints of concrete bridges,
ASCE, 7(2), 115–26. Journal of Engineering Mechanics, ASCE, 129(11), 1225–34.
Applied Technology Council [ATC] (1996), Seismic Evaluation Krawinkler, H. & Nassar, A. A. (1992), Seismic design based on
and Retrofit of Concrete Buildings, ATC-40, Redwood City, ductility and cumulative damage demands and capacities, in
CA. P. Fajfar and H. Krawinkler (eds.), Nonlinear Seismic Anal-
Banerjee, S. & Shinozuka, M. (2004), Dynamic progressive fail- ysis and Design of Reinforced Concrete Buildings, Elsevier
ure of bridges, in Proceedings of ASCE Joint Specialty Con- Science Publisher Ltd., NY, pp. 23–40.
ference on Probabilistic Mechanics and Structural Reliability, Mander, J. B. & Basöz, N. (1999), Seismic fragility curve theory
Albuquerque, NM. for highway bridges, in Proceedings of the 5th U.S. Confer-
California Department of Transportation (2004), Seismic De- ence on Lifeline Earthquake Engineering, Reston, VA, 31–
sign Criteria, V. 1.3, Sacramento, CA. 40.
California Department of Transportation (2005), Bridge De- Miranda, E. & Bertero, V. V. (1994), Evaluation of strength re-
sign Aids, Sacramento, CA. duction factors for earthquake-resistant design, Earthquake
Choi, E., DesRoches, R. & Nielson, B. (2004), Seismic fragility Spectra, 10(2), 357–79.
of typical bridges in moderate seismic zones, Engineering Priestley, M. J. N., Seible, F. & Calvi, G. M. (1996), Seismic
Structures, 26, 187–99. Design and Retrofit of Bridges, John Wiley and Sons, Inc.,
Computer and Structures, Inc. (2002), SAP2000 Nonlinear NY.
Users Manual, V. 8, Berkeley, CA. Reinhorn, A. M. (1997), Inelastic analysis techniques in seismic
Dutta, A. & Mander, J. B. (1998), Seismic fragility analysis evaluations, in P. Fajfar and H. Krawinkler (eds.), Seismic
of highway bridges, in Proceedings of INCEDE-MCEER Design Methodologies for the Next Generation of Codes, A.
Center-to-Center Workshop on Earthquake Engineering A. Balkema Publishers, Rotterdam, pp. 277–87.
Frontiers in Transportation Systems, Tokyo, Japan, 311– Rossetto, T. & Elnashai, A. (2005), A new analytical proce-
25. dure for the derivation of displacement-based vulnerability
Fajfar, P., Gaspersic, P. & Drobnic, D. (1997), A simplified non- curves for populations of RC structures, Engineering Struc-
linear method for seismic damage analysis of structures, in tures, 27(3), 397–409.
P. Fajfar and H. Krawinkler, (eds.), Seismic Design Method- Shinozuka, M., Feng, M. Q., Kim, H.-K. & Kim, S.-H. (2000a),
ologies for the Next Generation of Codes, A. A. Balkema Nonlinear static procedure for fragility curve development,
Publishers, Rotterdam, pp. 183–94. Journal of Engineering Mechanics, ASCE, 126(12), 1287–95.
Fajfar, P. (1999), Capacity spectrum method based on inelas- Shinozuka, M., Feng, M. Q., Lee, J. & Naganuma, T. (2000b),
tic demand spectra, Earthquake Engineering and Structural Statistical analysis of fragility curves, Journal of Engineering
Dynamics, 28, 979–93. Mechanics, ASCE, 126(12), 1224–31.
Seismic vulnerability assessment of bridges 305

Shinozuka, M., Feng, M. Q., Kim, H., Uzawa, T. & Ueda, T. Sultan, M. & Kawashima, K. (1994), Comparison of the seismic
(2003), Statistical analysis of fragility curves, Technical Re- design of highway bridges in California and in Japan, Recent
port MCEER-03-0002, Multidisciplinary Center for Earth- selected publications of Earthquake Engineering Div., Public
quake Engineering Research (MCEER), The State Univer- Works Research Institute (PWRI), Japan (Technical Mem-
sity of New York at Buffalo, NY. orandum of PWRI No. 3276).
Singhal, A. & Kiremidjian, A. (1998), Bayesian up- Zheng, Y., Usami, T. & Ge, H. (2003), Seismic response predic-
dating of fragilities with application to RC frames, tions of multi-span steel bridges through pushover analysis,
Journal of Structural Engineering, ASCE, 124(8), 922– Earthquake Engineering and Structural Dynamics, 32, 1259–
9. 74.

You might also like