Вы находитесь на странице: 1из 14

EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. (2012)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.2197

Effect of the angle of seismic incidence on the fragility curves


of bridges

Marco Torbol*,† and Masanobu Shinozuka


Department of Civil and Environmental Engineering, University of California, Irvine, Irvine, CA 92697, USA

ABSTRACT
This study examines the effect of the angle of seismic incidence θ on the fragility curves of bridges.
Although currently, fragility curves of bridges are usually expressed only as a function of intensity measure
of ground motion (IM) such as peak ground acceleration, peak ground velocity, or Sa(o1), in this study they
are expressed as a function of IM with θ as a parameter. Lognormal distribution function is used for this
purpose with fragility parameters, median cm and standard deviation z to be estimated for each value of θ
chosen from 0 < θ < 360 . A nonlinear 3D finite element dynamic analysis is performed, and key response
values are calculated as demand on the bridge under a set of acceleration time histories with different IM
values representing the seismic hazard in Los Angeles area. This method is applied to typical straight
reinforced concrete bridges located in California. The results are validated with existing empirical damage
data from the 1994 Northridge earthquake. Even though the sample bridges are regular and symmetric with
respect to the longitudinal axis, the results indicate that the weakest direction is neither longitudinal nor
transverse. Therefore, if the angle of seismic incidence is not considered, the damageability of a bridge
can be underestimated depending on the incidence angle of seismic wave. Because a regional highway
transportation network is composed of hundreds or even thousands of bridges, its vulnerability can also
be underestimated. Hence, it is prudent to use fragility curves taking the incident angle of seismic waves
into consideration as developed here when the seismic performance of a highway network is to be analyzed.
Copyright © 2012 John Wiley & Sons, Ltd.

Received 5 June 2011; Revised 13 March 2012; Accepted 29 March 2012

KEY WORDS: angle of seismic incidence; seismic vulnerability; seismic risk assessment; fragility curves;
SRA of lifeline systems

1. INTRODUCTION

The seismic risk assessment (SRA) of highway network systems is needed for the estimation of their
resilience and sustainability. Bridges are prominently vulnerable components of these networks and
fragility curves represent the damageability of these bridges. References [1–3] calculated empirical
fragility curves from the damage caused by real earthquakes (1994 Northridge and 1995 Kobe), and
references [1, 4–6] calculated analytical fragility curves with finite element modeling and nonlinear
dynamic analyses. These fragility curves did not consider the angle of seismic incidence (θ) and are
only functions of an intensity measure of ground motion (IM), as seen in Equation 1.

Fragility ¼ PðDSjIM Þ (1)

where DS is the damage state and the fragility is defined as the probability that the bridge suffers from
this state of damage under the given IM.

*Correspondence to: Marco Torbol, Department of Civil and Environmental Engineering, University of California,
Irvine, Irvine, CA 92697, USA.

E-mail: mtorbol@uci.edu

Copyright © 2012 John Wiley & Sons, Ltd.


M. TORBOL AND M. SHINOZUKA

Any monotonic non-decreasing function is a fragility curve candidate, but it must be bounded between
0 and 1. The lognormal cumulative distribution function, which satisfies these two conditions, Equation 2,
is used here:
 
ln IM  ln cim
Pi ðIM Þ ¼ Φ (2)
zi

where Pi is the probability that the ith damage state will result under the ground motion IM and cim and zi
are the median and the standard deviation of the lognormal distribution function, respectively.
The effect of the angle of seismic incidence on a single structure was studied by references [7–9].
Up to now, fragility curves did not consider the angle of seismic incidence. They were expressed
only as a function of the IM without regards to the angle of the incoming seismic wave, Equation 1,
or only the transversal and longitudinal directions were considered, as shown in Equations 3 and 4.
!
ln IM  ln cim;long
P ðIM Þ ¼ Φ
i
(3)
zilong

!
ln IM  ln cim;tran
P ðIM Þ ¼ Φ
i
(4)
zitran

where cim;long and cim;tran indicate the median values of fragility when the bridge is subjected to a seismic
wave in the longitudinal and transverse directions, respectively. Similar definition applies to zilong and zitran.
In the SRA of the highway transportation network, the same fragility curve represents the
damageability of similar bridges. For example, in the Los Angeles area, bridges are reinforced concrete
bridges designed with the same code or at least the same basic principle and have similar behaviors.
However, these bridges are generally found in different locations and have different orientations. A
scenario earthquake strikes each bridge from a different angle. Therefore, fragility curves must be
developed as a function of an intensity measure of ground motion (IM) given the angle of seismic
incidence (θ) in order to enhance the accuracy of the SRA result, as shown in Equation 5.
!
ln IM  ln cim;θ
Fragility ¼ P DS; θjIM Þ ¼ Φ (5)
ziθ

2. DEVELOPMENT OF FRAGILITY CURVES

The following method is used to estimate the parameters of fragility curves as a function of ground
motion intensity measure given the angle of seismic incidence. The capacity of and the demand on
the structure are evaluated for each possible angle of seismic incidence, the capacity by nonlinear
static moment–curvature (M–c) analyses and the demand with nonlinear dynamic analyses.
In the SRA of a highway transportation network, a model represents an entire class of bridges with
similar characteristics. Within the same class, the geometry, material properties, natural frequencies,
and mode shapes are similar but not identical. However at each bridge site, the same scenario
earthquake generates a different acceleration time history with different frequency content.
Furthermore, the same scenario earthquake strikes each bridge from a different angle and excites
different natural frequencies of the bridge. Therefore, although spectral acceleration is considered to
be a more accurate intensity measure compared with peak ground acceleration (PGA), it is
unreliable in this case. However, if the fragility curves are calculated for a specific bridge under a
known acceleration time history and if the geometry, material properties, natural frequencies, and

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
THE ANGLE OF SEISMIC INCIDENCE ON THE FRAGILITY CURVES OF BRIDGES

mode shapes of the bridge are known, the spectral acceleration at the first natural frequency (Sa(o1)) is
the most appropriate intensity measure.
In this study, the fragility curves are developed for the use in the ensuing future study where the
SRA approach is used to predict the performance of highway transportation networks under future
seismic events, which are represented by a set of scenario earthquakes.

2.1. Model bridge


The method is applied to a model bridge. The detail of a typical straight reinforced concrete bridge
located in California is shown in Figures 1, 2, and 3. It has a length of 242 m with five spans and an
expansion joint. The bridge has four circular columns and is supported by seat type abutments. The
deck is a cast-in-place concrete box girder.
The finite element model of the bridge is developed following the guidelines for reinforced concrete
bridges in California [10]. Note that the deck is modeled with elastic beam elements because it is
expected to perform elastically during an earthquake.
Each column, which is the most important component of this type of bridges, is modeled with the
force beam element with localized plasticity as developed and implemented by [11]. It is stable and
computationally more efficient than the displacement based element. The plasticity is localized at the
top and the bottom of each column in the plastic hinge region. The length of plastic hinge (lp) is
calculated according to Equation 6 [12].

 
lp ¼ 0:08L þ 0:022fye dbl ≥ 0:044fye dbl fye in MPa (6)

where L is column length, fye is yield strength of longitudinal rebars, and dbl is diameter of
longitudinal rebars.
Each plastic hinge is modeled with a fiber section, shown in Figure 4. Therefore, the nonlinear
behavior of the confined concrete, the rebars, and the cover concrete are represented by the different

Figure 1. Bridge model, side view.

Figure 2. Bridge model, plan view.

Figure 3. Geometry and properties of the cross sections.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. TORBOL AND M. SHINOZUKA

Figure 4. Fiber section discretization at the plastic hinges.

fibers. All columns have the same cross section, material properties, and axial load of 9479 kN
(2131 kips) equivalent to the dead load of the deck.
The expansion joint is modeled according to the DOF. The two heads of the deck have a pin
connection during an out of plane displacement. During the opening, the expansion joint is modeled
by a gap (0.5 in) and a hook element in series. The hook represents the restrainers, which controls
excessive relative displacement between the two heads as well as avoiding unseating. During the
closure, the expansion joint is modeled by a gap (1.0 in) and a hysteretic element in series. The
hysteretic element represents the pounding effect of the two heads, and it was developed by [13].
At abutment locations, an initial gap of 0.0508 m (2 in) is provided between the bridge deck and the
abutment. The subsurface profile used in the study in [14] consists of highly plastic over consolidated
clay with a depth of 8.2 m (Figure 5). Layers of silt, silty sand, and fine to medium sand are present
below the top clay layer. Unit weight of soil varies from 1.84 to 2.00 t/m3, that is, 115 to 125 lb/ft3.

Figure 5. Soil strata from Sultan and Kawashima [14].

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
THE ANGLE OF SEISMIC INCIDENCE ON THE FRAGILITY CURVES OF BRIDGES

The bridge columns are supported on pile foundations with 18.3 m length, and each foundation consists
of a group of 40 piles, each with diameter of 0.38 m.
The soil structure interaction is modeled with P–y spring. An equivalent single pile is considered to
represent the pile group following [15]. The cross section of this single pile is equal to the total area of
the piles present in the pile group, and the soil surrounding the equivalent pile is modeled with P–y
spring (Figure 6). The properties of each P–y spring are related to the soil type at that particular
depth (Figure 5). The sand recommendations are values taken from [16], whereas the clay values
were retrieved from [17].

2.2. The capacity as a function of θ


For a typical reinforced concrete bridge located in California, the damage caused by an earthquake is
localized at the plastic hinge regions of the column in the form of flexural damage. The shear damage
in the column is prevented by the dense transversal reinforcement, while the unseating of the expansion
joint is prevented by the steel restrainers. Therefore, the rotational resistance of the plastic hinge
regions represented by the moment–curvature relationship is indicative of the bridge’s overall
resisting capacity.
In this study, the cross section of the column is descritized with a fiber model, and the moment–
curvature analysis is performed on the model for each selected angle of seismic incidence. For
example, a circular cross section with evenly distributed rebars has the M–c relationship independent
from the angle of seismic incidence (Figure 7), whereas an oval cross section, which is also used for
bridges in California (Figure 8), has a different M–c for different angles of seismic incidence (Figure 9).
Once the M–c relationship is known, the progressive damage in the plastic hinge is represented by
the progressive damage states. The threshold values of rotational ductility are taken from [18]. For this
specific circular cross section for each M–c, the ductility ratio 1.0 is the yield state and the ductility
ratio 9.15 is the ultimate state calculated according to [19]. The ultimate state is the collapse damage
state. The rotational ductility at the other damage states, minor, moderate, and major, are
proportional to the drift limit. For the circular column of the sample bridge, the ductility ratio values
are 3.39 for minor damage, 4.75 for moderate damage, 8.43 for major damage, and 9.15 for
collapse. This calibration of quantified bridge damage states is consistent with the physical damage
state descriptions given in HAZUS’99 [20]. The column under consideration is circular, and it has
the same M–c and the same rotational ductility thresholds for any angle of seismic incidence.
Instead, the column in Figure 8 is oval; therefore, not only does the M–c change with the angle of
seismic incidence but also the rotational ductility thresholds change with it.

2.3. The demand as a function of θ


A set of nonlinear dynamic analyses is performed on the bridge model for each selected angle of
seismic incidence. The data set includes 30 time histories of ground acceleration. It was originally

Figure 6. Nonlinear P–y springs element detail.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. TORBOL AND M. SHINOZUKA

MOMENT−CURVATURE
40,000

35,000

30,000

Moment [Kip*ft]
25,000

20,000

15,000

10,000

5,000

0
0 0.00005 0.0001 0.00015 0.0002 0.00025
Curvature [1/in]

Figure 7. Moment–curvature of the column for the interval 0 < θ < 90 .

Figure 8. Example of oval cross section commonly used in California bridges.

MOMENT−CURVATURE
30,000

25,000

20,000
Moment [kip*ft]

15,000

10,000
0 deg
15 deg
30 deg
5,000
45 deg
60 deg
75 deg
90 deg
0
0 0.00005 0.0001 0.00015 0.0002 0.00025
Curvature [1/in]

Figure 9. Moment–curvature of the oval cross section for the interval 0 < θ < 90 .

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
THE ANGLE OF SEISMIC INCIDENCE ON THE FRAGILITY CURVES OF BRIDGES

developed for the Los Angeles area under the FEMA SAC steel project, found in [21]. The data set is
divided into three categories with the probability of exceedance of 2%, 10%, and 50% in 50 years.
Each ground motion has two horizontal components. One component is strike-parallel, and the other
component is strike-normal. This dataset includes a large variation in the return period and covers a
wide range of seismic hazard level. Therefore, it is appropriate for the development of analytical
fragility curves.
A larger size of the ground motion dataset can increase the accuracy by reducing the confidence
interval of the fragility parameters. However, availability of bi-directional ground motion dataset is
limited in number, especially at large values of the intensity measure. In this respect, a definitive
method to generate an artificial bi-component or tri-component ground motion is pursued by many
researchers, demonstrated by [22], among others. However, this can be achieved also by spectral
representation method [23].
The angle of seismic incidence θ represents the angle between the main component and the bridge
longitudinal axis. During each nonlinear dynamic analysis, the main component of the ground motion,
which is the strike-parallel, is always taken as ‘x-axis’ and the orthogonal component, which is strike-
normal, is always taken as ‘y-axis’, as shown in Figure 10. The coordinates of the nodes and the
orientations of the elements are a function of the angle of seismic incidence. At each new angle of
seismic incidence, the model is rotated, making it not necessary to decompose and recombine the
components of the ground motion.
Thus, under each angle of seismic incidence, 30 acceleration time histories (each with its PGA
value) are used to develop four fragility curves associated with four bridge damage states. The PGA
associated with each ground motion is the maximum value of the combination of the two
components. Each set of fragility parameters cm and z is evaluated by means of the maximum
likelihood method [3]. This will make it possible to develop fragility curves for each selected value
of θ with each predetermined bridge damage state classified in minor, moderate, major, and collapse
depending on the demand.

2.4. Parallel computing


The entire procedure can be accelerated through parallel processing. Every M–c analysis and every
nonlinear dynamic analysis are solved with OpenSeesMP, the parallel version of OpenSeesW [24].
Therefore, the overall computational time is divided by the number of core available, and it is
reduced to a reasonable amount. However, in many cases, the total number of analyses can be
reduced through symmetry. If a structure has no symmetry, such as a skew bridge, because of its
abutments’ geometry and location of columns, the fragility curve has to be calculated for the
interval 0 ≤ θ ≤ 360 . If the structure is symmetric with respect to the longitudinal axis, for example,
a multiple span straight bridge, the interval is reduced to 0 ≤ θ ≤ 180 . If the structure is symmetric
with respect to the transversal axis, for example a curved bridge with constant curvature, the interval
is reduced to 90 ≤ θ ≤ 90 . If the structure is symmetric with respect to both the longitudinal axis
and the transversal axis, such as the sample bridge, the interval is reduced to 0 ≤ θ ≤ 90 .

Figure 10. Orientation of orthogonal ground motion components.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. TORBOL AND M. SHINOZUKA

‘For each angle of seismic incidence’ indicates that a Δθ has to be chosen. The interval Δθ is
another critical number in this analysis. A smaller value of Δθ obviously requires a longer
computational time to cover the entire range of the angle to be considered. However, this process is
compensated by the better accuracy in the estimation of the overall seismic vulnerability of the
bridge for more precise results for ensuing SRA analyses.

3. RESULTS

For the sample bridge, the lognormal fragility curves are obtained by means of the maximum
likelihood method [1]. Parameters are listed in Table I and fragility curves in Figures 11–14.
The results indicate that the bridge is most vulnerable at a different angle of seismic incidence for each
damage state. For the minor damage state, the weakest direction is θ = 30 with 22.4% difference between
the strongest and the weakest direction. For the moderate damage state, the weakest direction is θ = 15
and θ = 60 with 22.0% difference. For the major damage state, the weakest direction is between
60 < θ < 75 with 48.2% difference. For the collapse damage state, the weakest direction is θ = 75
with a 62.9% difference. The percentage difference is calculated with respect to the mean value of that
specified damage state over the angle of seismic incidence. Therefore, fragility curves developed only
under the longitudinal direction and the transverse direction will underestimate the damageability of
the bridge.

Table I. Parameters of the analytical fragility curves of the sample bridge.


(z = 0.7) Difference
Sample bridge (%)
Mean
Angle of seismic incidence θ (degrees) 0 15 30 45 60 75 90 value

Minor damage state, cm 0.69 g 0.69 g 0.63 g 0.68 g 0.78 g 0.73 g 0.79 g 0.71 g 22.4
Moderate damage state, cm 1.04 g 0.84 g 0.84 g 0.90 g 0.84 g 0.98 g 0.91 g 0.91 g 22.0
Major damage state, cm 1.93 g 1.44 g 1.44 g 1.44 g 1.23 g 1.23 g 1.45 g 1.45 g 48.2
Collapse damage state, cm 2.20 g 1.58 g 1.44 g 1.44 g 1.33 g 1.23 g 1.57 g 1.54 g 62.9

Minor Damage State


1
0 deg
15 deg
0.9 30 deg
45 deg
60 deg
0.8 75 deg
Probability of Exceeding a Damage State

90 deg

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
PGA [g]

Figure 11. Minor damage fragility curves for the interval 0 < θ < 90 .

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
THE ANGLE OF SEISMIC INCIDENCE ON THE FRAGILITY CURVES OF BRIDGES

Moderate Damage State


1
0 deg
15 deg
0.9 30 deg
45 deg
60 deg
0.8 75 deg

Probability of Exceeding a Damage State


90 deg

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
PGA [g]

Figure 12. Moderate damage fragility curves for the interval 0 < θ < 90 .

Major Damage State


1
0 deg
15 deg
0.9 30 deg
45 deg
60 deg
0.8 75 deg
Probability of Exceeding a Damage State

90 deg

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
PGA [g]

Figure 13. Major damage fragility curves for the interval 0 < θ < 90 .

Fragility curves developed represent the damageability of the bridge considered in Figure 1. Bridges
with a significantly different geometry respond differently. However, the same method can be applied
for the estimation of the fragility parameters of these bridges. The fragility curves of other bridges with
significantly different geometry and other characteristics must be developed accordingly.

3.1. Validation with field data


The fragility curves developed in this study are validated with those derived from the field data. In
1994, Caltrans’ inspectors gathered the damage data from more than 2000 bridges in the Los

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. TORBOL AND M. SHINOZUKA

Collapse Damage State


1
0 deg
15 deg
0.9 30 deg
45 deg
60 deg
0.8 75 deg

Probability of Exceeding a Damage State


90 deg

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
PGA [g]

Figure 14. Collapse damage fragility curves for the interval 0 < θ < 90 .

Angeles area subjected to the Northridge earthquake. These damage data are used for the purpose of
this study. The data set is divided into 18 classes in accordance with the bridge characteristics as the
number of spans (two classes: one or multiple), angle of skewness (three classes: between 0 and
20 , 20 and 60 , and 60 and 90 ), and soil condition (three classes: hard A, medium B, and soft C
as in UBC’93) [25]. Hence, the bridges are grouped in 2  3  3 = 18 sets of classes. The number of
bridges reported in the inspection having specific levels of damage including no damage is listed in
Tables II and III.
The sample bridge used in this study, which is straight, multiple span, on soft soil, introduced earlier
(Figure 1) and used for the present study is indeed the prototype of class 16 in this classification. It is
important to point out that when the total sample is divided into 18 classes as carried out here, the
sample size becomes so small that fragility parameters cannot be estimated by means of maximum
likelihood method, and even if they could be, the confidence level is so low that the values are not
reliable. For this reason, the fragility parameters are estimated only for classes 16, 17, and 18, as
shown in Table IV. The parameters of the fragility curves are calculated using Method 2 by
Shinozuka [1]. Method 2 forces the standard deviation of the fragility curves, z, to be identical. The
fragility curves are lognormal distribution functions. When they are plotted on a lognormal probability
paper if they have identical z, they are parallel and any crossing is avoided. If the parameters of the
fragility curves are calculated separately, which is Method 1, the fragility curves may cross each other.

Table II. 1994 Northridge earthquake empirical fragility curves, classes 1 to 9.


Class 1 2 3 4 5 6 7 8 9
Number of span Single Single Single Single Single Single Single Single Single
Soil type A A A B B B C C C
Angle of skewness (degrees) a < 20 20 < a < 60 a > 60 a < 20 20 <a < 60 a > 60 a < 20 20 <a <60 a > 60
No damage 57 33 23 24 9 3 232 107 41
Minor damage 2 0 0 1 0 0 3 3 4
Moderate damage 0 0 0 0 0 0 2 7 2
Major damage 0 0 0 0 0 0 1 3 1
Collapse 0 0 0 0 0 0 0 0 0
Total number of bridges 59 33 23 25 9 3 238 120 48

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
THE ANGLE OF SEISMIC INCIDENCE ON THE FRAGILITY CURVES OF BRIDGES

Table III. 1994 Northridge earthquake empirical fragility curves, classes 10 to 18.
Class 10 11 12 13 14 15 16 17 18
Number of span Multiple Multiple Multiple Multiple Multiple Multiple Multiple Multiple Multiple
Soil type A A A B B B C C C
Angle of skewness
(degrees) a < 20 20 < a < 60 a > 60 a < 20 20 < a < 60 a > 60 a < 20 20 <a < 60 a > 60

No damage 112 76 33 41 17 8 635 321 223


Minor damage 4 1 3 0 2 2 29 18 13
Moderate damage 1 2 3 0 0 4 28 29 17
Major damage 0 3 2 0 2 1 11 12 11
Collapse 0 0 0 0 0 0 2 1 3
Total number 117 82 41 41 21 15 705 381 267
of bridges

Table IV. Parameters of the empirical fragility curves, class 16, 17, and 18.
Empirical fragility curves Class 16 (z = 0.70) Class 17 (z = 0.80) Class 18 (z = 0.56)

Minor damage 0.72 g 0.66 g 0.50 g


Moderate damage 0.95 g 0.82 g 0.60 g
Major damage 1.45 g 1.50 g 0.84 g
Collapse damage 2.82 g 3.20 g 1.40 g

If two different failure modes are considered, this is acceptable, for example, flexural failure and shear
failure. However, within the same ductile failure mode, this is unacceptable because the damage states
must always be reached in the same order. The flexural failure is ductile, and the minor, moderate,
major, and collapse damage state orders must be maintained [11].
The parameters of the empirical fragility curves and the parameters of the analytical fragility curves
are compared in Table V. Because empirical fragility curves could be computed only as a function of
the intensity measure, for each damage state, the comparison is made between the median of the
empirical fragility curve and the mean, over the angle of seismic incidence, of the median of the
analytical fragility curve. The comparison indicates an agreement of 5% in each damage state
except for the collapse, but the empirical data of the collapse is a poor statistical sample that
contains only two bridges.

4. APPLICATION TO THE SEISMIC RISK ASSESSMENT OF A SMALL NETWORK

The computed analytical fragility curves can be either used by an existing software, for example,
REDARS 2 [26], or implemented in a new framework. REDARS 2, which was developed under
FHWA supervision, assesses the damage and the social loss of highway networks due to earthquake.
REDARDS 2 uses the National Bridge Inventory database as a major source for fragility curves of
bridges, but it can also use user-defined fragility curves, that is, fragility curves developed in this study.
In this study, however, a framework was developed to integrate fragility curves as a function of IM
given θ; in the SRA of networks, the framework follow the flow chart in Figure 15 [25]. A small
highway transportation network, 5 links, 4 nodes, and 10 bridges (Figure 16), is used to demonstrate

Table V. Comparison between parameters of analytical and empirical fragility curves, class 16.
Class 16 Empirical fragility curves Analytical fragility curves Difference (%)
Minor damage 0.72 g 0.71 g 1.00
Moderate damage 0.95 g 0.91 g 4.72
Major damage 1.45 g 1.45 g 0.10
Collapse damage 2.82 g 1.54 g 82.95

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. TORBOL AND M. SHINOZUKA

Figure 15. Seismic risk assessment flow chart for the highway transportation example.

Figure 16. Highway transportation network example.

the importance of the angle of seismic incidence in the SRA. The sample network is subject to an
earthquake scenario, and three different cases are considered. In case 1, no earthquake scenario
strikes the network. In case 2, the earthquake scenario strikes the network, but fragility curves are
either longitudinal or transversal. In case 3, the earthquake scenario strikes the network, and fragility
curves are a function of IM and θ. The angle of seismic incidence is calculated taking into
consideration the bridge location and orientations and the earthquake scenario location and rupture
orientation. The rupture orientation must be considered because it guarantees the consistency
between the fragility curves, which were calculated using two-component ground motions strike-
parallel and strike-normal, and the earthquake scenario. First, the network, its links, its node, and its
vulnerable components are identified. Then, an earthquake scenario is considered, and its IM is
computed using an attenuation relationship at the location of each bridge. For each bridge, the
probability of exceeding each damage states is computed from the IM and the fragility curves. On
the basis of these probabilities, n damaged networks are generated using Monte Carlo simulation.
The origin–destination matrix is loaded on the intact network and on the n damaged networks using
a user equilibrium analysis algorithm, and the total travel time of each network is computed.
The drivers’ delay (h/day) is the difference between the total travel time of the intact network (TTTI),
case 1, and the total travel time of the damaged network, cases 2 and case 3. The drivers’ delay

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
THE ANGLE OF SEISMIC INCIDENCE ON THE FRAGILITY CURVES OF BRIDGES

represents the damage index of the network due to earthquake, and it is the seismic risk. The greater the
number of Monte Carlo simulations performed, the more accurate is the damage index. Because this is
a small network, 10,000 Monte Carlo simulations could easily be performed. The framework takes
advantage of parallel processing. Therefore, at the same time, different cores are performing
different Monte Carlo simulations.
Results indicate that case 1, the intact network, has a total travel time of 2762.7 h/day (TTTI). Case 2,
where fragility curves are either longitudinal or transversal, has a mean total travel time of 3079.7 h/day
(TTTLT) and a mean drivers’ delay of 317.0 h/day. Case 3, where fragility curves are a function of IM
and θ, has a mean total travel time of 3134.1 h/day (TTTθ) and a mean drivers’ delay of 371.4 h/day.
There is a 17% difference in the driver’s delay between case 2 and case 3 due to the different
fragility curves. Therefore, if the angle of seismic incidence θ is disregarded, the predicted damage
caused by the earthquake is underestimated because fragility curves as a function only of IM do not
capture the weakest direction of the structure. Furthermore, the sample network has only straight
bridges. A greater variation in the drivers’ delay is expected for a real network, where the
damageability of skew and curved bridges is strongly related to the angle of seismic incidence.

5. SUMMARY AND CONCLUSIONS

The study demonstrates the importance of the angle of seismic incidence in the SRA of highway
transportation networks. A method is developed and is applied to a sample bridge. The results are
validated with field data and used in the SRA of a sample highway transportation network.
Results indicate that in any damage state, the median of the fragility curves has a 22–62% variation
between the strongest and the weakest direction. However, the sample bridge is straight and symmetric.
Therefore, a larger variation is expected for irregular structures such as skewed bridges and curved
bridges. Results also indicate that variation of the fragility curves due to the angle of seismic
incidence affects the SRA of networks. The SRA of a small network with straight bridges has a
17% difference in the drivers’ delay. But for a larger highway transportation network, which also
includes skewed bridges and curved bridges, a larger variation is expected.
The method calculates analytical fragility curves of a single structure in terms of PGA and the angle
of seismic incidence. However, the structure is the prototype of a class. Therefore, the results should be
used only to represent the damageability of an entire class while performing the SRA of networks. If
the method is used to calculate the damageability of a single isolated structure, Sa(o1) should be
used as the intensity measure instead of PGA because the geometry, natural frequencies, and mode
shapes of a single structure are known.
In the future, fragility curves should be estimated as a function of IM, the horizontal angle of seismic
incidence θ, and the vertical angle of seismic incidence to include the near-field effect. Before
performing the SRA of a large highway transportation network, fragility curves as a function of IM
given θ have to be calculated for single span bridges, skewed bridges, and curved bridges.

ACKNOWLEDGMENT
The authors would like to thank Caltrans for providing the design of the bridge and earthquake damage
information.

REFERENCES
1. Shinozuka M, et al. Statistical analysis of fragility curves. Journal of Engineering Mechanics-ASCE 2000;
126(12):1224–1231.
2. Basoz NI, et al. Statistical analysis of bridge damage data from the 1994 Northridge, CA, earthquake. Earthquake
Spectra 1999; 15(1):25–54.
3. Yamazaki F, Hamada T, Motoyama H, Yamauchi H. Earthquake damage assessment of expressway bridges in Japan.
In Proc. 5th U.S. Conference on Lifeline Earthquake Engineering. 1999.
4. Karim KR, Yamazaki F. A simplified method of constructing fragility curves for highway bridges. Earthquake
Engineering and Structural Dynamics 2003; 32(10):1603–1626.
5. Elnashai AS, Borzi B, Vlachos S. Deformation-based vulnerability functions for RC bridges. Structural Engineering
and Mechanics 2004; 17(2):215–244.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. TORBOL AND M. SHINOZUKA

6. Choi ES, DesRoches R, Nielson B. Seismic fragility of typical bridges in moderate seismic zones. Engineering
Structures 2004; 26(2):187–199.
7. Rigato AB, Medina RA. Influence of angle of incidence on seismic demands for inelastic single-storey structures
subjected to bi-directional ground motions. Engineering Structures 2007; 29(10):2593–2601.
8. Lopez OA, Torres R. The critical angle of seismic incidence and the maximum structural response. Earthquake
Engineering and Structural Dynamics 1997; 26(9):881–894.
9. Lopez OA, Chopra AK, Hernandez JJ. Critical response of structures to multicomponent earthquake excitation.
Earthquake Engineering and Structural Dynamics 2000; 29(12):1759–1778.
10. Aviram, Mackie, Stojadinovic. Guidelines for nonlinear analysis of bridge structures in California. PEER, Editor. 2008.
11. Scott MH, Fenves GL. Plastic hinge integration methods for force-based beam-column elements. Journal of Structural
Engineering-ASCE 2006; 132(2):244–252.
12. Kowalsky MJ, Priestley MJN, Seible F. Shear and flexural behavior of lightweight concrete bridge columns in seismic
regions. ACI Structural Journal 1999; 96(1):136–148.
13. Muthukumar S, DesRoches R. A Hertz contact model with non-linear damping for pounding simulation. Earthquake
Engineering and Structural Dynamics 2006; 35(7):811–828.
14. Sultan M, Kawashima K. Comparison of the seismic design of highway bridges in California and in Japan. Recent
selected publications of Earthquake Engineering Div., 1994.
15. Yuanbiao Y, Kazuo K, Hikaru H. Single beam analogy for analyzing nonlinear soil–pile group interaction under lateral
loading. In 16th ASCE Engineering Mechanics Conference, University of Washington, Seattle. 2003.
16. API, Recommended practice for planning, design and constructing fixed offshore platforms. In API RP 2A-WSD 20th
Edn, American Petroleum Institute. 1993.
17. Matlock H. Correlations for design of laterally loaded piles in soft clay. In Proc. 2nd Annual Offshore Technology
Conference. 1970.
18. Banerjee S, Shinozuka M. Mechanistic quantification of RC bridge damage states under earthquake through fragility
analysis. Probabilistic Engineering Mechanics 2008; 23(1):12–22.
19. Priestley MJN, Seible F, Calvi GM. Seismic Design and Retrofit of Bridges, I (ed.). John Wiley and Sons: NY, 1996.
20. HAZUS MH Estimated Annualized Earthquake Losses for the United States. FEMA: Washington, D.C. 2008, p. viii, 53.
21. SAC-steel. http://www.nisee.berkeley.edu/data/strong_motion/sacsteel/ground_motions.html. 2002.
22. Rezaeian S, Kiureghian AD. A stochastic ground motion model with separable temporal and spectral nonstationarities.
Earthquake Engineering and Structural Dynamics 2008; 37(13):1565–1584.
23. Deodatis G. Non-stationary stochastic vector processes: seismic ground motion applications. Probabilistic Engineering
Mechanics 1996; 11(3):149–167.
24. McKenna F. The Regents of the University of California. 1999.
25. Shinozuka M, et al. Effect of seismic retrofit of bridges on transportation networks. Earthquake Engineering and
Engineering Vibration 2003; 2(2):169–179.
26. Werner SD, et al. REDARS 2: methodology and software for seismic risk analysis of highway systems. MCEER-06-
SP08, 2008.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe

Вам также может понравиться