ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 27 (2007) 414–426

www.elsevier.com/locate/soildyn

Effect of isolation on fragility curves of highway bridges based on

simplified approach
Kazi Rezaul Karima,, Fumio Yamazakib
a
Department of Civil, Architectural and Environmental Engineering, University of Missouri-Rolla, 326 Butler-Carlton Hall, 1870 Miner Circle,
Rolla, MO 65409, USA
b
Department of Urban Environment System, Faculty of Engineering, Chiba University, 1-1-3 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan
Received 24 October 2005; received in revised form 8 October 2006; accepted 24 October 2006

Abstract

The trend of isolating highway bridges is on the rise after the recent large earthquakes in Japan, the United States, and other countries.
Recent investigation shows that isolated systems perform well against seismic forces as the substructures of such systems experience less
lateral forces due to energy dissipation of the isolation device. Hence, it is anticipated that there might be an effect on fragility curves of
highway bridges due to isolation. In this study, 30 isolated bridge models were considered (and they were designed according to the
seismic design code of highway bridges in Japan) to have a wider range of the variation of structural parameters, e.g. pier heights,
weights, and over-strength ratio of structures. Then, fragility curves were developed by following a simplified procedure using 250 strong
motion records, which were selected from 5 earthquake events that occurred in Japan, the USA, and Taiwan. It is observed that the level
of damage probability for the isolated system is less than that of the non-isolated one for a lower level of pier height. However, having the
same over-strength ratio of the structures, the level of damage probability for the isolated system is found to be higher for a higher level
of pier height compared to the one of the non-isolated system. The proposed simple approach may conveniently be used in constructing
fragility curves for a class of isolated bridge structures in Japan that have similar characteristics.
r 2006 Elsevier Ltd. All rights reserved.

Keywords: Strong motion records; Highway bridges; Dynamic analyses; Fragility curves

1. Introduction the empirical approach. It is assumed that structural

parameters and input motion characteristics (e.g. frequency
Fragility curves are regarded to be useful tools for contents, phase, and duration) have influence on the
estimating the extent of probable damages (slight, moder- damage of structures for which there will be an effect on
ate, extensive, and complete) of structures due to an fragility curves.
earthquake [1–4]. It shows the probability of structure The present authors [5] developed a set of analytical
damages as a function of ground motion indices, e.g. peak fragility curves for highway bridge piers based on
ground acceleration (PGA) and peak ground velocity numerical simulation considering the variation of input
(PGV). They allow estimating a damage level for a known ground motions. It was found that there is a significant
ground motion index. effect of earthquake ground motions on fragility curves.
Yamazaki et al. [1] developed a set of empirical fragility They also developed a simplified method [6] to construct
curves for highway bridges based on actual damage data the fragility curves of non-isolated highway bridges
from the 1995 Kobe earthquake. However, the type of considering the variation of both input ground motions
structure, structural performance (static and dynamic) and and structural parameters. It was also found that there is a
variation of input ground motions were not considered in significant effect of both earthquake ground motions and
structural parameters on fragility curves.
Corresponding author. Tel.: +1 573 341 4479; fax: +1 573 341 4729. The trend of isolating highway bridges is on the rise after
E-mail address: [email protected] (K.R. Karim). the recent damaging earthquakes in Japan, the United

0267-7261/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2006.10.006
 ARTICLE IN PRESS
K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426 415

States, and other countries. Recent investigation shows design code in Japan [11] that a pier should be divided at
that isolated systems perform well against seismic forces as least into 50 slices), and the force–displacement relation-
the substructures of such systems experience less lateral ship at the top of the bridge pier is obtained by using the
forces due to energy dissipation of the isolation device [7]. shear vs. strain and moment vs. curvature relationships of
Hence, it is anticipated that there might be an effect on all cross-sections. Using the elastic stiffness (obtained from
fragility curves of highway bridges due to isolation; in the force–displacement relationship), the non-linear dy-
other words, fragility curves for non-isolated bridges may namic response analyses [12] are performed for the selected
not be applicable to predict the extent of probable damages input ground motions, which are normalized to different
for isolated systems since the fragility curves of the two excitation levels.
systems might be different. The damage to the structure (pier) is then quantified by a
The purpose of this study is to develop fragility curves damage index (DI) that is obtained by using a damage
for isolated bridges by following a simplified procedure [6], model [13] and the number of occurrence of a particular
and to compare them with the ones of the non-isolated damage rank is counted by calibrating [14] the damage
systems. In this objective, 30 isolated bridge models are indices in different excitation levels, which is used to obtain
considered (and they to designed according were the the damage ratio of each damage rank in each excitation
seismic design code of highway bridges in Japan) to have level. The damage ratio is then plotted on a lognormal
a wider range of the variation of structural parameters, e.g. probability paper [1,5] from where the two parameters of
pier heights, weights, and over-strength ratio of structures. the fragility curves, i.e., mean and standard deviation are
A total of 250 strong motion records are considered as the obtained by performing a linear regression analysis.
input motions, which were selected from 5 earthquake Finally, fragility curves are constructed for each damage
events that occurred in Japan, the USA, and Taiwan. Then, rank with respect to the ground motion indices using the
using the selected input motions and isolated bridge obtained mean and standard deviation. The procedures
models, fragility curves are obtained with respect to ground adopted for constructing the analytical fragility curves can
motion parameters by following a simplified approach [6]. be summarized as follows:

2. Development of fragility curves 1. Selection of the earthquake ground motion records.

2. Normalization of PGA of the selected records to
2.1. Empirical fragility curves different excitation levels.
3. Making a physical model of the structure.
Yamazaki et al. [1] developed a set of empirical fragility 4. Performing a non-linear static pushover analysis and
curves based on actual damage data from the 1995 Kobe obtaining the elastic stiffness of the structure.
earthquake, and showed the relationship between the 5. Selection of a hysteretic model for the non-linear
damages that occurred to the expressway bridge structures dynamic response analysis.
and the ground motion indices. In this approach, the 6. Performing the non-linear dynamic response analysis
damage data of the expressway structures due to the Kobe using the elastic stiffness and the selected records.
earthquake were collected, and the ground motion indices 7. Obtaining the damage indices of the structure in each
along the expressways were estimated based on the excitation level using a damage model.
estimated strong motion distribution using Kriging tech- 8. Calibration of the damage indices for each damage
nique. The damage data and ground motion indices were rank to obtain the damage ratio in each excitation
related to each damage rank, and the damage ratio for each level.
damage rank was obtained. Finally, using the damage ratio 9. Plotting the damage ratio in each excitation level on a
for each damage rank, the empirical fragility curves for the lognormal probability paper and obtaining the mean
expressway bridge structures were constructed assuming a and standard deviation of the fragility curves for each
lognormal distribution [1,8]. damage rank by performing a linear regression
analysis.
2.2. Analytical fragility curves 10. Construction of fragility curves using the obtained
mean and standard deviation with respect to the
The present authors [5] developed a set of analytical ground motion indices for each damage rank assuming
fragility curves for highway bridge piers based on a lognormal distribution.
numerical simulation and considering the variation of
input ground motions. The procedures adopted to con-
struct the analytical fragility curves are briefly described 2.3. Simplified approach to develop fragility curves
below.
In this approach, first, the non-linear static pushover The present authors [6] also developed a simplified
analysis of the structure is performed [9,10], which includes method to construct fragility curves for non-isolated
the shear vs. strain and moment vs. curvature analyses of bridges based on the observed correlation between the
the cross-sections (it is recommended in the highway bridge fragility curve parameters and structural parameters. The
 ARTICLE IN PRESS
416 K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426

procedure adopted to develop the simplified expressions of Similar procedure has also been adopted to obtain the
fragility curves is briefly described below while the details expression for standard deviation x and the expression for
can be found somewhere else [6]. x is given as
In this approach, first, the fragility curve parameters
x ¼ ½b0 þ b1 h þ b2 h2 ½a0 þ a1 Dy. (5)
mean l and standard deviation x are obtained by
performing a series of both non-linear static pushover It should be noted that the regression coefficients of Eq.
and dynamic response analyses. Then, the relationships (4) are different than that of Eq. (5), however, same
between mean l and standard deviation x with the over- symbols are used for simplicity. It should also be noted that
strength ratio y [6,11] are obtained considering all the data the expressions of fragility curve parameters mean l and
points without making any subgroups. The relationships standard deviation x given in Eqs. (4) and (5), respectively,
are also obtained by making the data points into some hold true for all damage ranks, i.e. slight, moderate,
subgroups, for instance, data points for different codes, extensive, and complete with respect to both PGA and
pier heights, weights, etc. It is observed that l and y shows PGV. Another point also has to be noted that to perform
higher correlation for the data points of each level of pier regression analysis and to obtain regression coefficients b0,
height. Based on this observation, l for different levels of b1, b2, a0 and a1 of Eqs. (4) and (5), fragility curve
pier heights are obtained by fixing some y using the parameters mean l and standard deviation x are obtained
relationships between l and y that are obtained for by following the same procedures given in the preceding
different levels of pier heights. Then, the relationship section, i.e. ‘‘analytical fragility curves’’ section, which
between l and h is obtained using the following regression provides the foundation for developing the simplified
model: method to obtain the expressions of fragility curve
parameters [6].
lh ¼ b0 þ b1 h þ b2 h2 , (1)
3. Bridge models
where lh is the mean with respect to h, h the height of the
pier, and b0, b1 and b2 are the regression coefficients. Like
3.1. Description of bridge models
the data points for each level of pier height, it is also found
that there is a strong correlation between l and h for
In order to obtain simplified expressions of fragility curve
different y. It is also observed that the relationships
parameters for isolated bridges, a total of 30 bridge models
between l and h obtained for different y are quite parallel,
are considered to have a wider range of the variation of
which implies that knowing only one of the relationships
structural parameters, and they were designed [15] according
between l and h for a given y, the other relationships for
to the seismic design code of highway bridges in Japan [11].
different y can also be obtained knowing only some scale
For the selected bridge models, the piers are considered
factors for a change of y. In this objective, the scale factors
rectangular and fixed to the base [16], and a lead–rubber
are obtained for changing different y for different pier
bearing (LRB) is considered as the isolation device [7,16,17].
heights considering the relationship between l and h
The ground condition is considered as type II, the regional
obtained for a y equal to 1.0, and the scale factor Fy is
class is considered as A, and the standard lateral force
given as
coefficient khco is considered as type II [6,11].
F y ¼ a0 þ a1 Dy, (2) The bridge models are divided into three categories, viz.
bridges designed with different years’ seismic codes, bridges
where Fy is the scale factor with respect to the change of y, having different pier heights, and bridges having different
Dy the change of y given as (y1), and a0 and a1 are the span lengths or weights, however, the number of spans for
regression coefficients. Although, the scale factors for the all-bridge models is assumed to be four. The
different levels of pier heights are found to be very similar, substructures (piers) for any typical bridge model are
however, to minimize the error that might results for considered to be similar, in other words, one pier model
different levels of pier heights, the average scale factor can be considered as the representative of all other piers for
obtained for different pier heights is considered [6]. Hence, a particular bridge structure. This assumption is adopted to
the l value can readily be obtained using Eq. (1) for a avoid a rigorous computation necessary to perform non-
known h, and then simply multiplying it by the scale factor linear pushover analyses for the all piers of a particular
Fy of Eq. (2) that can be obtained for a known Dy . In other bridge model. It should be noted that the non-linear
words, the l value can be obtained by using the following pushover analysis is performed in order to obtain the
expression: elastic stiffness of the substructure. The physical model is
l ¼ lh F y . (3) considered as the one shown in Fig. 1 and the analytical
model is shown at the right side of the physical model. It
Substituting for lh and Fy from Eqs. (1) and (2) into Eq. (3) can be seen (Fig. 1) that the analytical model consists of a
gives portion of superstructure, the isolation device (LRB) and
the substructure (pier). Since the main concern is the
l ¼ ½b0 þ b1 h þ b2 h2 ½a0 þ a1 Dy. (4) damage of the pier, therefore, this simple analytical model
 ARTICLE IN PRESS
K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426 417

m2
k2, c2
m1
k1, c1

Fig. 1. Physical model of an isolated bridge system used in this study.

Table 1
Structural properties for the 30 isolated bridge models used in this study

Design code Span length, L ¼ 30 and 40 m (w ¼ 500 kN/m) Reinforcement

Pier height (m)

6 9 12 15 18 Long. Tie

Section Section Section Section Section rl (%) rt (%)

aa bb aa bb aa bb aa bb aa bb Area ratio Vol. ratio

1964 2.0 2.8 2.6 3.2 3.0 3.5 3.4 3.8 3.5 4.0 1.21 0.09
1980 2.1 3.0 2.8 3.2 3.2 3.8 3.8 4.0 3.8 4.2 1.25 0.32
1995 2.2 3.0 2.8 3.4 3.2 4.0 3.8 4.2 4.0 4.5 1.36 1.03

s0c (MPa) and ssy (MPa) are taken as the same for the all codes, and they are taken as 27 and 300, respectively.
a
Dimension in the longitudinal direction in m.
b
Dimension in the transverse direction in m.

is adopted in this study, which suffices to see how the mended that the yield force of the LRB can be taken as 5%
bridge pier behaves with an isolation device under seismic W, which provides a reasonable balance between reduced
loading. It should be noted that if one is interested to see forces in the piers and increased forces on the abutments.
the behaviour of several other components (e.g. failure of While several options may be considered, however, in this
deck, cap beam, piles, abutment, etc.) of a particular bridge study, the yield force and yield stiffness of the LRB are
structure, then a fancy model can be adopted based on any taken as 5% W and 5% W/mm, respectively. Given the
commercially available FEM based software for instance yield force level and the lead yield strength of 10–10.5 MPa
SAP2000 [10], however, that is not the purpose of this [15,16], the number and cross-sectional area of the lead
study. plugs can be designed (Fig. 2). The advantage of LRB is
Table 1 shows all the structural properties for different that it has low yield strength and sufficiently high initial
categories of bridges having span length of 30 and 40 m stiffness that results in higher energy dissipation [7,15–17].
with superstructure weight as 500 kN/m. Note that same
structural properties have been considered for the all-
3.3. Analytical model
bridge models having a span length of 40 m, in other
words, changing only the span length or weight of the
The analytical model consists of a portion of super-
superstructure while all other parameters being unchanged.
structure, the isolation device (LRB) and the substructure
It can be seen that the pier cross-section changes for
(pier) of the isolated bridge system, which is modeled as a
different seismic design codes even having the same height,
two-degree-of-freedom (2DOF) system [7,16,17], a bilinear
and it changes from smaller to larger from the 1964 code to
hysteretic model was considered for the both substructure
the 1995 code. It can also be seen that the pier cross-section
[18] and isolation device [7,15–17], the post-yield stiffness
also changes due to the changes of pier height even if it is
was taken as 10% of the initial stiffness for the both
designed with the same seismic code, and it changes from
substructure and isolation device [16,17], the damping
smaller to larger from pier height 6 to 18 m. One can also
matrix C is evaluated by using the Rayleigh damping
see that the longitudinal (area ratio) and tie (volumetric
[12,15], and the damping constant hi is found by using the
ratio) reinforcement also changes for different seismic
following expression [11]:
codes, and the value goes higher from the 1964 code to the
Pn T
1995 code. j¼1 hj Fij Kj Fij
hi ¼ , (6)
FTi KFi
3.2. Isolation device-LRB
where hj is the equivalent damping constant of element j,
Kawashima and Shoji [17] recommended that the yield Fij is the mode vector of element j of the ith vibration
force of the LRB can be taken as 10–20% weight of the mode, Kj is the equivalent stiffness matrix of element j, Fi is
superstructure (W), while Ghobarah and Ali [16] recom- the mode vector of the overall structure of the ith vibration
 ARTICLE IN PRESS
418 K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426

non-isolated ones result due to the energy dissipation of the

16−φ32 Bolts 5−φ110 Lead Plugs isolation device [7,16,17], and it implies that the isolated
system performs better against seismic forces than the non-
isolated system does.

C C
4. Simplified expressions of fragility curve parameters
1050

920
525
Fragility curve parameters l and x for the 30 isolated
bridge models are obtained by following the same
procedure given in the ‘‘analytical fragility curves’’ section
[5] using the selected 250 records as the input motions.
Then, simplified expressions for both the l and x of Eqs. (4)
850 425 and (5) are obtained by following the same procedure given
in the ‘‘Simplified approach to develop fragility curves’’
Plan view of LRB bearings assembly
section [6]. Fig. 5 shows the graphical representation to
obtain the simplified expression for l for a slight damage
Cover Plate with respect to PGA. Fig. 6(a) shows the relationships
110 Lead Plug 3 mm Steel Plate
between l and y obtained for different damage ranks with
201

respect to PGA for a y equal to 1.0, and the corresponding

average scale factors for l obtained for different damage
250 20 mm Rubber Layer
ranks are shown in Fig. 6(b). Finally, the regression
141

850 30 mm Steel Plate

coefficients of Eqs. (4) and (5) are obtained for all the
Sectional view at C-C
damage ranks with respect to both PGA and PGV by
Fig. 2. Details of a lead–rubber bearings (LRB) that can be used as an performing both linear and nonlinear regression analyses,
isolation device for the isolated bridge system. This is an example how the and the regression coefficients are shown in Table 2. Note
LRB can be designed given the lead yield strength and yield force level
that the corresponding R2 values are also shown in the
[15,16].
same table.

4.1. Numerical example

mode, and K is the equivalent stiffness matrix of the overall
structure. To see how the simplified expressions of fragility curve
parameters work, a different bridge structure is considered,
3.4. Displacement and energy demands which was not used to obtain the simplified expressions.
The bridge was designed according to the recent seismic
Recent investigation shows that isolated systems per- design code for highway bridges in Japan [11]. It is assumed
form well against seismic forces as the substructures of that only the number of spans, span length, superstructure
such systems experience less lateral forces due to energy weight, height and cross-section of the pier can be changed
dissipation of the isolation device [7]. Also, damage to the while other conditions being the same as that of the 30
structure for a given input motion is related to both bridge models that were used to develop the simplified
displacement and energy demands [13]. Hence, it is expressions. For the example bridge structure, the number
necessary to see how the displacement and energy demands of spans is assumed to be 5, the length of each span is taken
of isolated systems differ from that of the non-isolated as 50 m, the weight is taken as 320 kN/m, the height of each
ones. pier is taken as 8 m, and the cross-section of each pier is
Fig. 3 shows the plots of displacement and energy taken as 2.5 by 3 m. The over-strength ratio y is calculated
histories [19] for the substructure and bearing of an isolated [6,11] as 1.21. Now, knowing the height of the pier as 8 m
system obtained from the JMA Kobe NS record of the and y as 1.21, the fragility curve parameters l and x for
1995 Kobe earthquake. It can be seen that both the different damage ranks with respect to both PGA and PGV
displacement and energy demands of the substructure of are obtained using the simplified expressions given in Eqs.
the isolated system is less than that of the bearing. Fig. 4 (4) and (5), and using the regression coefficients given in
shows the plots of displacement and hysteretic energy Table 2. l and x are also obtained by performing a series of
demands of the substructures for an isolated and a non- both nonlinear static pushover and dynamic response
isolated system obtained from the JMA Kobe NS record. analyses.
One can see that both the displacement and energy Table 3 shows the list of the fragility curve parameters
demands of the substructure of the isolated system are less for the example bridge structure obtained from both
than that of the substructure of the non-isolated ones. The analytical and simplified methods, and the corresponding
lower level of both displacement and energy demands of errors e for both l and x with respect to the analytical ones
the substructure of the isolated system than that of the are also shown in the same table. Figs. 7 and 8 show the
 ARTICLE IN PRESS
K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426 419

a Displacement response hystory b x105 Energy of the substructure

30
substructure
bearing 2

Displacement [cm]
20

Energy [kN-cm]
1.5 input
10 hysteretic+damping
damping
0 1

-10 0.5

-20 0
0 20 40 60 80 0 20 40 60 80
Time [s] Time [s]

c x105 Energy of the bearing d x105 Energy of the system

8 10
Energy [kN-cm]

Energy [kN-cm]
8
6
6
4
4
2 input
2
input
hysteretic+damping hysteretic+damping
damping damping
0 0
0 20 40 60 80 0 20 40 60 80
Time [s] Time [s]

Fig. 3. Displacement and energy histories of an isolated bridge system obtained from the JMA Kobe NS record of the 1995 Kobe earthquake, (a)
displacement response histories of the substructure and bearing, (b) energy of the substructure, (c) energy of the bearing, and (d) energy of the system.

Displacement of Hysteretic energy

a the substructure b x105 of the substructure
12 5.5
isolated isolated
10 non-isolated
5 non-isolated
4.5
8
Displacement [cm]

4
Energy [kN-cm]

6 3.5
4 3
2 2.5
0 2
1.5
-2
1
-4 0.5
-6 0
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Time [s] Time [s]

Fig. 4. (a) Displacement response histories, and (b) hysteretic energy of the substructures of an isolated and a non-isolated bridge system obtained from
the JMA Kobe NS record of the 1995 Kobe earthquake.

fragility curves for all damage ranks with respect to PGA error for l with respect to both PGA and PGV is found to
and PGV, respectively, obtained from both analytical and be only 1.8%, and for x it is found as 19.8%.
simplified methods. It can be seen that the fragility curves It should be noted that l controls the amplitude and x
obtained by both analytical and simplified methods seem to controls the shape of the fragility curves. The 19.8% error
be very close with respect to PGV, however, a very small for x does not necessarily mean that it might result in a
difference is observed with respect to PGA for the all- significant effect on the fragility curves, and the evidence
damage ranks. Note that the maximum error with respect can be seen in the fragility curves (Figs. 7 and 8). Hence,
to both PGA and PGV for both l and x are shown in Table the error terms for both l and x given in Table 3 seem to be
3 with an underline mark. It can be seen that the maximum within an acceptable range, and the simplified method may
 ARTICLE IN PRESS
420 K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426

a b
7.5 7.5
h6, R2=0.985
λ=5.88+0.52θ, R2=0.788 2
h9, R =0.932
2
h12, R =0.964
Mean, λ 7 7 h15, R2=0.980

Mean, λ
h18, R2=0.994

6.5 6.5

6 6
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Overstrength Ratio, θ Overstrength Ratio, θ

c d
7.5 1.25
θ =1.0
θ =1.5

Scale Factor, Fθ
θ =2.0
1.2
7
Mean, λ

1.15
h6
6.5 h9
1.1 h12
h15
h18
6 1.05
5 10 15 20 0 0.5 1 1.5 2 2.5
Height, h Change of Overstrength Ratio, Δθ

Fig. 5. Relationship between (a) l and y obtained for the 30 isolated bridge models used in this study, (b) l and y for different pier heights, (c) l and h for
different y, and (d) Fy and Dy for different pier heights, all for a slight damage with respect to PGA.

a b
7.2 1.25
Slight Slight
Moderate
7 Extensive Moderate
Scale Factor, Fθ

Complete 1.2 Extensive

Mean,λ

6.8 Complete

6.6 1.15
6.4
1.1
6.2
6 1.05
5 10 15 20 0 0.5 1 1.5 2 2.5
Height, h Change of Overstrength Ratio,Δθ

Fig. 6. Relationship between (a) l and h for y equal to 1.0, and (b) average Fy and Dy obtained for different damage ranks with respect to PGA.

conveniently be used to construct the fragility curves for highway bridges. In this study, following the same
isolated bridge structures knowing the height h and over- procedure, simplified expressions are also developed to
strength ratio y only. It should be noted that the simplified construct the fragility curves for isolated highway bridges,
expressions of fragility curve parameters are obtained which are given in the preceding section. Since simplified
based on a set of isolated bridge systems, and these expressions show the correlation between the fragility curve
simplified expressions for fragility curve parameters may parameters and the structural parameters, they might
conveniently be used to construct the fragility curves of conveniently be used to construct the fragility curves for
similar kind of isolated bridge structures that fall within the the both isolated and non-isolated bridges. However, since
same group and have similar characteristics. the two systems are different, it is necessary to see how the
fragility curves of both the systems differ from each other
5. Fragility curves for both isolated and non-isolated bridges based on the simplified expressions.
Fig. 9 shows the plots of the relationship between
The present authors [6] also developed simplified fragility curve parameter mean l and pier height h for a y
expressions to construct fragility curves of non-isolated equal to 1.5 obtained from the simplified method for both
 ARTICLE IN PRESS
K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426 421

Table 2
List of the regression coefficients for the fragility curve parameters obtained from the simplified method

Indices DR Parameters

l x
2
lh ¼ b0+b1h+b2h Fy ¼ a0+a1Dy xh ¼ b0+b1h+b2h2 Fy ¼ a0+a1Dy

b0 b1 b2 s R2 a0 a1 s R2 b0 b1 b2 s R2 a0 a1 s R2

PGA S 6.30 0.03 0.0024 0.022 0.984 1.00 0.11 0.00 1.00 0.40 0.0006 0.0006 0.007 0.990 1.00 0.39 0.00 1.00
M 6.58 0.01 0.0005 0.031 0.812 1.00 0.10 0.00 1.00 0.41 0.0002 0.0001 0.012 0.981 1.00 0.38 0.00 1.00
E 6.67 0.02 0.0010 0.017 0.906 1.00 0.10 0.00 1.00 0.27 0.0008 0.0003 0.015 0.971 1.00 0.63 0.00 1.00
C 7.02 0.02 0.0003 0.047 0.770 1.00 0.10 0.00 1.00 0.38 0.0043 0.0002 0.002 0.997 1.00 0.70 0.00 1.00
PGV S 4.42 0.001 0.002 0.013 0.998 1.00 0.31 0.00 1.00 0.59 0.001 0.0002 0.007 0.967 1.00 0.24 0.00 1.00
M 4.25 0.091 0.004 0.036 0.933 1.00 0.26 0.00 1.00 0.59 0.005 0.0005 0.008 0.978 1.00 0.48 0.00 1.00
E 4.51 0.080 0.004 0.016 0.994 1.00 0.29 0.00 1.00 0.63 0.004 0.0005 0.010 0.969 1.00 0.47 0.00 1.00
C 489 0.068 0.003 0.015 0.973 1.00 0.18 0.00 1.00 0.87 0.034 0.0007 0.011 0.991 1.00 0.67 0.00 1.00

DR: damage rank, S: slight, M: moderate, E: extensive, C: complete.

Table 3
List of the fragility curve parameters for the example isolated bridge structure obtained from both analytical and simplified methods

Indices DR Parameters

l x

Analytical Simplified Error, e (%) Analytical Simplified Error, e (%)

PGA S 6.60 6.56 0.61 0.52 0.51 1.57

M 6.82 6.77 0.77 0.47 0.45 3.46
E 6.94 6.89 0.72 0.42 0.41 2.14
C 7.14 7.03 1.55 0.42 0.34 19.77
PGV S 4.60 4.57 0.61 0.59 0.60 1.04
M 4.96 4.96 0.12 0.63 0.66 3.52
E 5.26 5.16 1.80 0.75 0.69 7.86
C 5.46 5.43 0.53 0.75 0.73 3.05

DR: damage rank, S: slight, M: moderate, E: extensive, C: complete.

isolated and non-isolated systems for different damage very high, for instance, say more than 20 m, then the
ranks with respect to PGA. Note that the simplified isolated system may not be so effective. It should be noted
expressions for the non-isolated system were taken from that fragility curves are also a function of standard
the previous study [6]. It can be seen that the l for the deviation x, and both the mean l and standard deviation
isolated system is higher than that of the non-isolated one x are also functions of scale factor Fy that is obtained for a
for the all-damage ranks for a lower level of pier height, given over-strength ratio y. Hence, to see the effect of
which implies that the level of damage probability for the isolation on fragility curves, it is necessary to construct
isolated system is less than that of the non-isolated one them for the both isolated and non-isolated systems
when the level of pier height is not so large. However, one considering all these factors.
can see that as the pier height changes from lower to a Fig. 11 shows the plots of the fragility curves for the
higher level, the mean l of the isolated system seems to get isolated and non-isolated bridges for an extensive damage
closer to the non-isolated one, and eventually, in case of with respect to PGA obtained from the simplified expres-
extensive and complete damages, it is less than that of the sions for different level of pier heights with an over-
non-isolated one after a certain level of pier height. Similar strength ratio y equal to 1.5. It can be seen (Figs. 11(a) and
trend is also found with respect to PGV, and the plots are (b)) that the level of damage probability for the isolated
shown in Fig. 10. system is less than that of the non-isolated one for a pier
The trend of converging the mean l of the isolated height of 5 and 10 m, respectively, and its damage level
system with that of the non-isolated one for a higher level seems to be similar to that of the non-isolated one when the
of pier height implies that if the pier height of the bridge is pier height is 15 m (Fig. 11(c)). Now, if one looks at
 ARTICLE IN PRESS
422 K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426

a Slight b Moderate
1 1
analytical analytical
simplified simplified
0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
PGA [cm/s2] PGA [cm/s2]

c Extensive d Complete
1 1
analytical analytical
simplified simplified
0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
PGA [cm/s2] PGA [cm/s2]

Fig. 7. Comparison of the fragility curves obtained from both analytical and simplified methods for an isolated bridge system with respect to PGA.

a Slight b Moderate
1 1
analytical analytical
simplified simplified
0.8 0.8
Probability
Probability

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 50 100 150 0 50 100 150
PGV [cm/s] PGV [cm/s]

c Extensive d Complete
1 1
analytical analytical
simplified simplified
0.8 0.8
Probability

Probability

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 50 100 150 0 50 100 150
PGV [cm/s] PGV [cm/s]

Fig. 8. Comparison of the fragility curves obtained from both analytical and simplified methods for an isolated bridge system with respect to PGV.

Fig. 11(d), then it can be seen that the level of damage trend is also observed on the fragility curves obtained for
probability for the isolated system is higher than that of the both isolated and non-isolated systems with respect to
non-isolated one where the pier height is 20 m. Similar PGV, and the plots are shown in Fig. 12.
 ARTICLE IN PRESS
K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426 423

a Slight b Moderate
7.5 7.5
non-isolated
isolated
7 7
Mean, λ

Mean, λ
6.5 6.5

6 6

5.5 5.5 non-isolated

isolated
5 5
0 5 10 15 20 25 0 5 10 15 20 25
Height, h Height, h

c Extensive d Complete
7.5 7.5

7 7
Mean, λ

Mean, λ
6.5 6.5

6 6

5.5 non-isolated 5.5 non-isolated

isolated isolated
5 5
0 5 10 15 20 25 0 5 10 15 20 25
Height, h Height, h

Fig. 9. Comparison of the relationship between l and h for y equal to 1.5 obtained from the simplified method for the isolated and non-isolated bridge
systems for different damage ranks with respect to PGA.

Fig. 10. Comparison of the relationship between l and h for y equal to 1.5 obtained from the simplified method for the isolated and non-isolated bridge
systems for different damage ranks with respect to PGV.
 ARTICLE IN PRESS
424 K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426

a h=5m b h=10m
1 1
non-isolated non-isolated
isolated isolated
0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
PGA [cm/s2] PGA [cm/s2]

c h=15m d h=20m
1 1
non-isolated non-isolated
isolated isolated
0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
PGA [cm/s2] PGA [cm/s2]

Fig. 11. Comparison of the fragility curves for the isolated and non-isolated bridge systems with respect to PGA obtained from the simplified method for
different pier heights, all for an extensive damage with y equal to 1.5.

a h=5m b h=10m
1 1
non-isolated non-isolated
0.8 isolated 0.8 isolated
Probability

Probability

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 50 100 150 0 50 100 150
PGV [cm/s] PGV [cm/s]

c h=15m d h=20m
1 1
non-isolated non-isolated
0.8 isolated 0.8 isolated
Probability

Probability

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 50 100 150 0 50 100 150
PGV [cm/s] PGV [cm/s]

Fig. 12. Comparison of the fragility curves for the isolated and non-isolated bridge systems with respect to PGV obtained from the simplified method for
different pier heights, all for an extensive damage with a y equal to 1.5.
 ARTICLE IN PRESS
K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426 425

Note that the over-strength ratio y is considered being a higher level of pier height compared to the one of the
the same for all levels of pier heights, and it is taken as 1.5. non-isolated system. It implies that the level of damage
It means that when the pier height is to be higher then even probability for the isolated systems tends to be higher for a
having the same over-strength ratio, the level of damage higher level of pier height. This might be due to fact that
probability for the isolated system goes higher compared to the failure of the isolation device was not considered in the
the one of the non-isolated system. Several factors may be present study as well as the strength of the isolation device
addressed in this regard. Firstly, simple expressions are was considered as the same for the all level of pier heights.
obtained for a limited range of pier heights, i.e. 6 to 18 m, In other words, the fragility curves may differ from the one
which may imply that the simple expressions are valid only where the failure of the isolation device as well as different
for the said range of pier heights. Secondly, the strength for level of strengths of the isolation device for different level
the isolation device is considered as the same (5% W) for of pier heights would have been considered.
the all level of pier heights, which may underestimate the The simple expressions may be a very useful tool, and
energy dissipation of the isolation device for a higher level may conveniently be used to construct the fragility curves
of pier heights. Thirdly, due to the elongation of the for isolated bridges in Japan that fall within the same
natural period, the pier with more height become less stiff group and have similar characteristics. However, it is
(i.e. more flexible) compared to the one with less height. anticipated that the simple expressions of fragility curves
This observation is supported with the fact that an developed in this study may not be applicable for the
isolation device is more effective for a stiff structure and isolated systems that have SSI effect, and a further research
it turns out to be very less effective for a flexible structure. is recommended in this regard. Also, the same simple
Finally, the failure of the isolation device is not considered expressions may not be applicable for other countries since
in this study, which may have some effect on the fragility it is based on only Japanese seismic design code, however,
curves for a higher level of pier height. the same simple approach may be adopted in constructing
Although, the soil–structure interaction (SSI) effect is the fragility curves using different seismic codes of other
generally not so severe for non-isolated bridge structures countries.
except for the case with strong soil non-linearity, isolated
bridges are regarded to be more susceptible to the effect of
SSI during an earthquake [20]. Thus, it is anticipated that Acknowledgements
there might be an effect on the fragility curves of isolated
bridges due to SSI; hence, a further study is recommended The authors would like to thank the editor and the
in this regard. Also, several other factors should be kept in anonymous reviewers who spent their valuable time in
mind, for instance, different design codes in other reviewing this paper. Their valuable comments in modify-
countries, several other components of bridge structures ing this paper is highly appreciated.
(e.g. failure of deck, cap beam, piles, abutment, etc.),
different seismic zones, different soil conditions, etc. While References
the present study considers only a few of them (e.g. failure
of bridge pier, seismic zone A, soil type II, lateral force [1] Yamazaki F, Motomura H, Hamada T. Damage assessment of
coefficient of type II, Japanese seismic design code for expressway networks in Japan based on seismic monitoring. In:
highway bridges, etc.), however, same simple approach can Proceedings of the 12th world conference on earthquake engineering.
be adopted in deriving the simple expressions for fragility 2000, Paper no 0551 [CD-ROM].
[2] Basoz N, Kiremidjian AS. Evaluation of bridge damage data from
curve parameters of isolated bridge structures considering the Loma Prieta and Northridge, CA earthquakes. Report no 127,
all the parameters or a combination of several parameters. The John A. Blume Earthquake Engineering Center, Department of
Civil Engineering, Stanford University, 1997.
6. Conclusions [3] Mander JB, Basoz N. Seismic fragility curves theory for highway
bridges. In: Proceedings of the fifth US conference on lifeline
earthquake engineering. TCLEE no 16, ASCE, 1999. pp. 31–40.
Simple expressions of fragility curve parameters for [4] Kircher CA, Nassar AA, Kustu O, Holmes WT. Development of
isolated highway bridge structures were obtained based on building damage functions for earthquake loss estimation. Earth-
numerical simulation with respect to the ground motion quake Spectra 1997;13(4):663–82.
parameters using 250 strong motion records. Fragility [5] Karim KR, Yamazaki F. Effect of earthquake ground motions on
curves for the both isolated and non-isolated systems were fragility curves of highway bridge piers based on numerical
simulation. Earthquake Eng Struct Dyn 2001;30(12):1839–56.
also constructed based on the obtained simplified expres- [6] Karim KR, Yamazaki F. A simplified method of constructing
sions and using the seismic design code of highway bridges fragility curves for highway bridges. Earthquake Eng Struct Dyn
in Japan. 2003;32(10):1603–26.
It was observed that the level of damage probability for [7] Chaudhary MTA, Abe M, Fujino Y, Yoshida J. System identification
the isolated system is less than that of the non-isolated one and performance evaluation of two base-isolated bridges using
seismic data. Journal of Structural Eng ASCE 2000;126(10):1187–96.
for a lower level of pier height. However, having the same [8] Sucuoglu H, Yucemen S, Gezer A, Erberik A. Statistical evaluation
over-strength ratio of the bridges, the level of damage of the damage potential of earthquake ground motions. Struct Saf
probability for the isolated system is found to be higher for 1999;20(4):357–78.
 ARTICLE IN PRESS
426 K.R. Karim, F. Yamazaki / Soil Dynamics and Earthquake Engineering 27 (2007) 414–426

[9] Bentz EC, Collins MP. Response-2000. Software program for [15] Priestley MJN, Seible F, Calvi GM. Seismic design and retrofit of
load–deformation response of reinforced concrete section. /http:// bridges. New York: Wiley; 1996.
www.ecf.utoronto.ca/bentz/inter4/inter4.shtmlS, 2000. [16] Ghobarah A, Ali HM. Seismic performance of highway bridges. Eng
[10] SAP2000. Integrated structural analysis and design software. Struct 1988;10:157–66.
Computers and Structures Inc.; 2000. [17] Kawashima K, Shoji G. Interaction of hysteretic behavior between
[11] EED. Design specifications of highway bridges. Part V: seismic isolator/damper and pier in an isolated bridge. J Struct Eng-JSCE
design. Technical memorandum of EED, PWRI, no 9801, 1998. 1998;44A:213–21.
[12] Chopra AK. Dynamics of structures: theory and application to [18] Kawashima K, Macrae GA. The seismic response of bilinear
earthquake engineering. Upper Saddle River, NJ: Prentice-Hall; 1995. oscillators using Japanese earthquake records. J Res, PWRI, Ministry
[13] Park YJ, Ang AH-S. Seismic damage analysis of reinforced concrete of Construction, Japan 1993;30:7–146.
buildings. J Struct Eng ASCE 1985;111(4):740–57. [19] Uang C-M, Bertero VV. Evaluation of seismic energy in structures.
[14] Ghobarah A, Aly NM, El-Attar M. Performance level criteria and Earthquake Eng Struct Dyn 1990;19:77–90.
evaluation. In: Proceedings of the international workshop on seismic [20] Chaudhary MTA, Abe M, Fujino Y. Identification of soil–structure
design methodologies for the next generation of codes. Balkema, interaction effect in base-isolated bridges from earthquake records.
Rotterdam, 1997. p. 207–15. Soil Dyn Earthquake Eng 2001;21(8):713–25.