Vol.7, No.

EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION

Earthq Eng & Eng Vib (2008) 7:415-426

December, 2008

DOI: 10.1007/s11803-008-1007-4

Seismic response of skewed RC box-girder bridges

Ahmed Abdel-Mohti and Gokhan Pekcan
Civil and Environmental Engineering Department, University of Nevada Reno

Abstract: It is critical to ensure the functionality of highway bridges after earthquakes to provide access to important
facilities. Since the 1971 San Fernando earthquake, there has been a better understanding of the seismic performance of
bridges. Nonetheless, there are no detailed guidelines addressing the performance of skewed highway bridges. Several
parameters affect the response of skewed highway bridges under both service and seismic loads which makes their behavior
complex. Therefore, there is a need for more research to study the effect of skew angle and other related factors on the
performance of highway bridges. This paper examines the seismic performance of a three-span continuous concrete box
girder bridge with skew angles from 0 to 60 degrees, analytically. Finite element (FE) and simplified beam-stick (BS) models
of the bridge were developed using SAP2000. Different types of analysis were considered on both models such as: nonlinear
static pushover, and linear and nonlinear time history analyses. A comparison was conducted between FE and BS, different
skew angles, abutment support conditions, and time history and pushover analysis. It is shown that the BS model has the
capability to capture the coupling due to skew and the significant modes for moderate skew angles. Boundary conditions and
pushover load profile are determined to have a major effect on pushover analysis. Pushover analysis may be used to predict
the maximum deformation and hinge formation adequately.
Keywords: skew bridge; seismic response; pushover analysis; time history analysis; modeling

1 Introduction
Many advances have been made in developing
design codes and guidelines for static and dynamic
analyses of regular or straight highway bridges.
However, there remains significant uncertainty with
regard to the structural system response of skewed
highway bridges as it is reflected by the lack of
detailed procedures in current guidelines. In fact, as
evidenced by past seismic events (i.e., 1994 Northridge
Gavin Canyon Undercrossing and 1971 San Fernando
Foothill Boulevard Undercrossing), skewed highway
bridges are particularly vulnerable to severe damage due
to seismic loads. Even though a number of studies have
been conducted over the last three decades to investigate
the response characteristics of skewed highway bridges
under static and dynamic loading, research findings
have not been sufficiently comprehensive to address
global system characteristics. Due to the fact that the
Correspondence to: Gokhan Pekcan, Dept. of Civil and
Environmental Engineering, University of Nevada Reno, NV
89557
Tel: (775) 784-4512; Fax: (775) 784-1390
E-mail: pekcan@unr.edu

Research Assistant; Assistant Professor

Supported by: In part by the California Department of
Transportation Under Caltrans Contract No. 59A0503, and the
Dept. of Civil and Environmental Engineering (UNR)
Received: October 3, 2008; Accepted: November 3, 2008

current seismic design guidelines do not provide explicit

procedures, a significantly large number of bridges are
at risk with consequential threat to loss of function, life
safety, and economy. Many of the existing bridges may
be prone to earthquake induced damage and may require
substantial retrofit measures to achieve the desired
seismic performance and post-earthquake serviceability.
Researchers and practicing design engineers need to fully
understand the overall system response characteristics
of skewed highway bridges for the proper detailing of
system components.
It is generally agreed that bridges with skew angles
greater than 30 degrees exhibit complex response
characteristics under seismic loads. Several studies have
investigated the effects of skew angle on the response of
highway bridges (Maleki, 2002; Bjornsson et al., 1997;
Saiidi and Orie, 1991; Maragakis, 1984). Saiidi and
Orie (1991) noted the skew effects and suggested that
simplified models and methods of analysis would result in
sufficiently accurate predictions of seismic response for
bridges with skew angles of less than 15 degrees. On the
other hand, Maleki (2002) concluded that slab-on-girder
bridges with skew angles up to 30 degrees and spans
up to 65 feet have comparable response characteristics
to straight bridges, and therefore, simplified modeling
techniques such as rigid deck modeling can be used in
many instances. Bjornsson et al. (1997) conducted an
extensive parametric study of two-span skew bridges
modeled with rigid deck assumption. In this study,

416

EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION

maximum relative abutment displacement (MRAD)

was considered as a critical quantity associated with
failures due to unseating. MRAD was found to be
influenced strongly by the impact between the deck
and the abutments. Critical skew angle as a function of
span length and width was introduced to maximize the
rotational impulse due to impact and was found to be
between 45 and 60 degrees.
In evaluating and comparing results across various
analytical studies, one must consider the underlying
assumptions and idealizations implemented in the
analytical treatment of the skew highway bridges. These
may pertain to material modeling, inelastic (hysteretic)
response characteristics of components, as well as
conditions at the boundaries, soil-structure interaction,
component geometry (i.e., idealized beam-stick versus
full finite element), superstructure (i.e., rigid versus
flexible), seismic mass (i.e., distributed versus lumped),
etc. For instance, Meng and Lui (2000) suggested that
the effects of modeling boundary conditions properly
may outweigh the effects of skew angle on the overall
dynamic response characteristics of a bridge. In fact,
differences in assumptions can create inconsistencies in
results as seen in the analysis of the Foothill Boulevard
Undercrossing, which sustained severe damage during
the San Fernando earthquake. A study conducted by
Wakefield et al. (1991) concluded that the failure was
controlled by rigid-body motion, which agreed with a
previous study conducted by Maragakis (1984). But, a
study conducted by Ghobarah and Tso (1974) explained
that the failure was induced by flexural and torsional
motion. Ghobarah and Tso (1974) assumed the deck
was fixed at the abutments, while Wakefield et al. (1991)
assumed free translation of the deck at the abutments.
For the present study, a benchmark bridge is selected and
the effect of various parameters interacting with the skew
angle is studied. Also, the adequacy of using simplified
beam-stick (BS) models and pushover analysis is tested
versus FE model and time history analysis.

2 Benchmark bridge
To facilitate the objective of this study, a
highway bridge was chosen from Federal Highway
Administrations (FHWA) Seismic Design of Bridges
Series (Design Example No.4). The bridge is a
continuous three-span box-girder bridge with 97.536 m
(320 ft) total length, spans of 30.48, 36.576, and 30.48
m (100, 120, and 100 ft), and 30o skew angle (Fig. 1;
FHWA, 1996). The superstructure is a cast-in-place
concrete box-girder with two interior webs and has a
width-to-span ratio (W/L) of 0.43 for the end spans and
0.358 for the middle span. The intermediate bents have a
cap beam integral with the box-girder and two reinforced
concrete circular columns. Reinforced concrete columns
of the bents are 1.219 m (4 ft) in diameter supported on
spread footings. The longitudinal reinforcing steel ratio
of the column is approximately 3% and the volumetric

Vol.7

steel ratio of the spirals is approximately 0.8%. Also, the

axial load ratio of the column is 14%. The abutments
are seat-type with elastomeric bearings under the
web of each box girder. In the longitudinal direction,
movement of the superstructure is limited by the gap
between the superstructure and the abutment back-wall.
In the transverse direction, interior shear keys prevent
the movement. This bridge was designed to be built
in the USA in a zone with an acceleration coefficient
of 0.3g following 1995 AASHTO guidelines. Seismic
performance of highway bridges is expected to be
affected by a combination of various parameters such
as: skew angle, boundary conditions, superstructure
flexibility, stiffness and mass eccentricity, in-span
hinges and restrainers, width-to-span ratio, and direction
of strong motion components with respect to orientation
of bridge bent and abutments. This study investigates
the effect of some of these parameters summarizes as:
skew angle, abutment gap, shear keys, and direction of
earthquake motion.

3 Modeling of bridges
To facilitate a comparative parametric study of the
seismic response of skewed highway bridges, a number
of detailed three-dimensional (3D) finite element (FE)
as well as beam-stick (BS) models were developed
(Fig. 2) using SAP2000 (2005). In all of the models,
the superstructure was assumed to be linear-elastic, and
all of the nonlinearity was assumed to take place in the
substructure elements including bents, internal shearkeys, bearings, and abutment gap (gap opening and
closing). The benchmark bridge was altered to produce
models with different skew angles, but with the same
overall dimensions. Skew angles of 0, 30, 45, and 60
degree were of interest.
For FE models, FE mesh was used to model deck,
soffit, girders, and diaphragms. On the other hand, bent
columns and footing were modeled using 3-D frame
elements. For the nonlinear analyses, nonlinearity is
assumed to take place in the form of localized plastic
hinges at the top and bottom of columns. The behavior
of nonlinear [uncoupled] axial and moment hinges
is characterized by the axial force-displacement and
moment-rotation relationship, respectively. It should
be noted that the coupling between moments developed
about the x-axis and y-axis was not considered in
modeling. Footing-soil interaction was modeled using
linear translational and rotational springs at base of the
footings (Fig. 2). Nonlinear link elements were used to
model bearings, abutment gap, and shear keys.
The bearings were designed based on a shear
modulus of elasticity (G) of 1.03 MPa (150 psi), and
bearing dimensions of 0.127 m (5 in.) height and 0.165
m2 (256 in.2) of area that supports the bridge. The initial
stiffness (ko) is calculated to be 1.33 kN/mm (7.68 kip/
in.). The failure limit state is assumed to occur at 100%
shear strain (127 mm (5.0 in.) displacement) and the

Ahmed Abdel-Mohti et al.: Seismic response of skewed RC box-girder bridges

corresponding lateral load is 169 kN (38.4 kips).

Gap link elements were used to model the abutment
gap which its stiffness activates only when the gap closes.
On the other hand, the multi-linear plastic link elements
were used to simulate the response of internal shear
keys. Capacities of internal shear keys were estimated
based on comprehensive study done by Megally et al.
(2002). Three shear keys, one at center of each cell, were
aligned along the skew angle of the bridge (Fig. 3).
For simplified BS models, the same procedure of
modeling was followed. Nonetheless, superstructure

4'-0''

1'-0'' (TYP)

1'-6''

20'-0''

8''

4'-0''

12'-0''
6'-0''

43'-0''
5'-6''
5'-6''
Bent

12'-0''

4'-0'' Column (TYP)

3'-6''

2'-0''
(TYP)

Concrete hinge
(TYP)

14'-0''14'-0''
Square FTG (TYP)

417

was represented by a single beam element having the

equivalent properties of the entire deck (Table 1). Also,
the bent cap was modeled using a 3D frame element with
a high inertia to facilitate the force distribution to the
columns. Additional mass was assigned at abutments,
mid-spans, and bent caps to account for the additional
weight of the diaphragms. The abutments are modeled
by lumping the bearing, shear key, and gap link elements
from the FE model into individual nonlinear link
elements respectively used in BS Model (Fig. 3).

11'-3 1/8''

11'-3 1/8''
Column

Column

Section
(a) Bent elevation

Brg abut A

Brg abut B

320'-0''
Bent 1

Bent 2

120'-0''
30 Skew
(TYP)

100'-0''

''

'-0

26

43'-0''

100'-0''

Exp

Bent 2

Bent 1

Brg abut A

Plan
Brg abut B

No.4

Fix
Pin

Fix
Pin

Exp

Elevation
(Looking parallel to bents)
(b) Plan and elevation views
Fig. 1 Benchmark bridge geometry (*1ft = 304.8 mm and 1in. = 25.4 mm)

418

EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION

Vol.7

Column 3
(C3)

Skew
angle

Z
Y

Column 4
(C4)
Nonlinear hinges
(all columns)

Skew
angle

Column 1
(C1)

Translational and
rotational springs
(all columns)

X
Column 2
(C2)

Skew
angle

Z
Y

Skew
angle
X

(a) Finite element (FE) model

Nonlinear hinges
(all columns)
Footings

Rigid links
Translational and
rotational springs
(all columns)

(b) Beam-stick (BS) model

Fig. 2 FE and BS model details

Skew
angle

Skew
angle
B & SK

SK
B&G

SK
B&G

BBearing element
Y
GGap element
SKShear key element

SK
B&G

B&G

B
B&G

BBearing element
Y
GGap element
SKShear key element

(a) FE model

(b) BS model
Fig. 3 Abutment support detail of bridge models

Table 1 Section properties for beam-stick model

Superstructure

Bent cap beam

Area (m )

6.76

2.51

1.17

Ix - torsion (m4)

10.16

863

0.22

Iy (m4)

3.46

863

0.11

Iz (m4)

83.69

863

0.11

2915

2402

2402

Density (kg/m )
3

4 Analysis of skew highway bridges

As mentioned earlier, the main objective of the study
is to examine the effect of various modeling assumptions
on the seismic response of skewed highway bridges.
Table 2 presents the analytical matrix of the study.
The analysis includes modal, pushover, and linear and
nonlinear time history analyses. Also, a comparison was
conducted between the simplified BS and FE models to
assess the accuracy of BS models to capture response of
models with different skews. Another comparison was
done between results of nonlinear pushover analyses
versus linear and nonlinear time history analyses. Note
that analyses were conducted for two extreme cases of
boundary conditions Case I with shear keys and Case

Bent column

II with no shear keys. Both pushover analysis and time

history analyses results are discussed in more detail in
the following paragraphs.
4.1 Modal analysis
Modal analyses were conducted for each skew and for
two abutment support conditions (Case I and Case II) to
determine the vibration modes of the FE and BS models
discussed previously. Note that structural dynamic
characteristics of bridges are expected to be captured
more accurately by the FE models. Nonetheless, based
on the comparisons between the FE and BS models,
conclusions can be drawn with respect to the level of
accuracy of approximations due to simplified BS models

No.4

Ahmed Abdel-Mohti et al.: Seismic response of skewed RC box-girder bridges

in general.
Note that the comparisons for Case II only are
presented in Table 3. For Case II, it was observed that
the principal longitudinal, transverse, vertical, and
coupled vibration periods predicted by both models
are generally in good agreement. The largest percent
difference is approximately 11% if 60o skew bridge is
excluded. The modal periods start to deviate and the
largest percent difference approaches to approximately
35%. However, for Case I, it was found that the principal
longitudinal, transverse, vertical, and coupled vibration
periods predicted by both models are in good agreement
with 13% relative difference as well. The largest percent
difference is approximately 6% if 60o skew bridge is
excluded. Similarly, the modal periods start to deviate
and the largest percent of difference becomes 33% for
larger skew angles.
4.2 Time history analysis
Linear and nonlinear time history analysis were
conducted on FE models with different skew angles
using seven ground motions in order to investigate

419

nonlinear response; effect of skew angle on seismic

performance, and effect of direction of strong component
of ground motion with respect to longitudinal and
transverse directions of the bridge models. Also, time
history analysis results were used to measure the
accuracy of pushover analysis. Since the hysteretic
behavior and failure mode of the shear keys could not be
modeled in SAP2000, two separate cases were studied:
one with rigid restraints at the shear key locations
(Case I), and one without restraints at these locations
(Case II). Time history analyses were conducted for
these two extreme cases (Cases I and II) and different
direction of components of ground motions.
4.2.1. Selection of ground motions
Design acceleration response spectrum (ARS) is
assumed to be that of Caltrans with MW of 6.5, PGA
of 0.3g, and soil type B. Four ground motions were
selected from the PEER strong motion data base (http:
//peer.berkeley.edu/smcat) among the historically
recorded motions with MW of 6.0 or larger and epicenter
distance of 20 km or less. These four motions were
Kobe, Landers, Sylmar, and Rinaldi. All of the four
ground motions were scaled to have PGA of 0.3g. The

Table 2 Analytical matrix

Analysis type
Modal analysis

Analysis cases
Modal

Pushover analysis

Transverse mode,
Multimode,
Uniform load
Kobe
Landers
Sylmar
Rinaldi
Simqke 1
Simqke 2

Time history analysisNonlinear and linear

Parameters
With shear keys (Case I)
Skew: 0, 30, 45, 60
With shear keys (Case I)
Skew: 0, 30, 45, 60

Without shear keys (Case II)

Skew: 0, 30, 45, 60
Without shear keys (Case II)
Skew: 0, 30, 45, 60

Restrained (Case I)
Strong motion in transverse
direction / Weak motion in
longitudinal direction (ST)
Skew: 0, 30, 45, 60

Unrestrained (Case II)

Strong motion in Longitudinal
direction / Weak motion in
transverse direction (SL)
Skew: 0, 30, 45, 60

Table 3 Percent differences between FE and BS periods for case II

Mode desc.
Trans./Coup.
(1,1)*
Longitudinal
(2,2)
Transverse
(4,4)
Vertical
(6,6)

0 skew

Vertical (12,9)

4.67

0.24
3.36
3.11
5.31

Mode desc.
Coupled
(1,1)
Coupled
(2,2)
Transverse
(4,4)
Vertical
(6,6)
Vertical
(12,9)

30 skew
o

2.04
8.06
2.86
1.62
1.89

Mode desc.
Coupled
(1,1)
Coupled
(2,2)
Transverse
(3,4)
Vertical
(6,6)
Vertical
(8,9)
Vertical
(12,12)

* The numbers refer to the corresponding FE and BS mode numbers (FE, BS).

45 skew
o

5.03
10.6
5.26
2.68
34.7
34.3

Mode desc.
Trans./Coup.
(1,1)
Long./Coup.
(2,2)
Transverse
(3,3)
Vertical
(4,4)
Vertical
(6,6)
Vertical
(10,9)
Vertical
(12,12)

%
60o skew
3.47
0.58
9.19
22.8
26.1
26.1
23.5

420

EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION

1.2
Caltrans ARS (0.3g)
Kobe EW
Landers EW
Sylmar EW
Rinaldi EW
Simqke 1
Sinqke 2
Simqke 3
Average

PSA (g)

1.0
0.8
0.6
0.4
0.2
0.0

Period (s)
Fig. 4 ARS spectra scaled to 0.3 PGA

remaining three ground motions were artificial ground

motions generated using SIMQKE (1999) to match
the target 5%-damped design ARS. Figure 4 presents
a comparison of the scaled ground motions along with
artificially generated motions and Caltrans ARS. The
average of the selected seven motions matches the
demand curve, especially within 0.4 to 0.7 seconds,
which is of major interest as periods of models fall
within this range. All four ground motions consisted
of two components with different intensities, therefore
two analysis cases were considered; the two cases
are designated as ST and SL. For the ST, the stronger
component was applied in the transverse direction while
the opposite was true for the SL case. On the other hand,
SIMQKE motions were applied equally in both directions.
4.3 Pushover analysis
Nonlinear static pushover analyses are performed on
BS and FE models developed with various skew angles
under consideration (0, 30, 45, and 60). The objectives
were to study: (1) the accuracy of BS models versus FE
models in capturing the overall nonlinear behavior of the
bridges through pushover analysis, (2) the applicability
of pushover analyses for relatively large skew angles,
and (3) the effect of the pushover load profile.
Pushover analysis is anticipated to be sensitive to the
orientation of shear keys and the pushover load profile.
Therefore, three pushover load profiles are considered.
These profiles are: (1) first dominant transverse
mode (Mode 1); (2) combination of transverse and a
longitudinal modes (Mode 1+2); and (3) uniform load.
Pushover capacities in terms of load and displacement
are reported for all of the examined parameters as well as
hinge formation sequence in columns (C1, C2, C3, and
C4), which are shown in Fig. 2(a).
4.3.1. FE versus BS models results
For straight bridges, FE and BS models predicted
comparable load and displacement capacities when Mode
1 or Mode 1+2 were applied to Case I. A slight difference
was observed between the results of both models when

Vol.7

uniform load profile was used for the pushover analysis.

For both load profiles Mode 1 and Mode 1+2, sequence
of hinge formation varied between FE and BS models.
Yielding of columns initiated at the bottom then at the
top in the FE model, and the opposite was the case
for the BS model. This was due to the difference in
modeling of the deck diaphragm between the models,
which is expected to affect the load distribution and path
from deck to columns. However, the same sequence of
hinge formation of columns was observed between the
models when the uniform load profile was applied,
although the FE model predicted larger pushover
capacities and displacement. On the other hand, a good
agreement was observed between pushover capacities
predicted by FE and BS models for all load profiles
for Case II. The same sequence of hinge formation
was observed in Mode 1 and the uniform load profiles.
For 30o skew, pushover loads predicted by both
models were in a good agreement in the first transverse
mode and uniform load profile for Case I. The sequence
of hinge formations was predicted consistently with both
modes and with similar values of pushover capacities
and displacements regardless of the load profile.
Furthermore, application of the uniform load profile
led to consistent results between the FE and BS models
to predict pushover curve and sequence of yielding for
Case II. This shows the adequacy of the uniform load
profile to predict pushover capacity of bridges relying
on simplified BS models for highway bridges with
moderate skew angles.
For 45o skew, similar behavior was observed as
compared to the 30o skew. However, yielding load
and displacement values started to deviate between
the FE and BS models when subjected to the first
transverse mode for Case I. For Case II, application
of the uniform load profile remains the first choice to
achieve consistency between the models (BS and FE).
The transverse load profile led to the prediction of larger
pushover capacities and displacements, whereas smaller
capacities and displacements were predicted with the
multi-mode profile in FE models as compared to the BS
models.
For 60o skew, when FE and BS models were
subjected to transverse mode and uniform load profiles,
similar pushover capacities were predicted for Case I.
Application of the multi-mode profile predicted smaller
capacity in the FE model. For Case II, the uniform load
profile still resulted in a reasonable agreement between
capacities between the FE and BS models, although the
BS model predicted larger displacements. Application of
the first transverse mode profile led to similar pushover
curves with larger pushover capacities predicted by the FE
model. A fair agreement was noticed between the predict
sequence of hinge formation between the models when
subjected to transverse mode and uniform load profiles.
4.3.2. Effect of skew angle
By considering each of the three load profiles,
different conclusions can be drawn. Application of the

No.4

Ahmed Abdel-Mohti et al.: Seismic response of skewed RC box-girder bridges

transverse mode leads to comparable hinge formation

stages and first yield displacement as skew varies, except
for 45o skew where larger first yield displacement was
predicted for Case I. Also, major variation in predicted
pushover capacity was observed as a function of skew.
For Case II, no certain yield displacement pattern was
followed. Also, similar pushover capacity was observed
due to the application of the transverse mode for all skews
except 45o skew, which predicted slightly larger capacity.
Application of the combined mode led to an
almost identical pattern for Case I. Pushover load and
displacement are similar regardless of the skew angle,
while capacity of the bridge with no zero skew was
predicted to be smaller. For Case II, pushover capacity
decreases with increasing skew angle. It was observed
that 45o has the smallest pushover capacity but it is close
to that of 60o skew model.
Application of the uniform load profile led to lower
pushover capacity and displacement as skew increases,

contradicting with results of the other load profiles for

Case I (Fig. 5). For Case II, it predicted smaller pushover
capacities than in the case of the transverse mode profile
(Fig. 6).
4.3.3. Effect of shear key
Figure 7 summarizes the effectiveness of shear key
for all load profiles. Application of the transverse mode
shows that effectiveness of the shear keys contribution
to pushover capacity is independent of skew angle. As
expected, pushover capacity of Case I is larger than
Case II due to the absence of the shear key, which
essentially resists lateral loads. Nonetheless, the location
of the first yield and ultimate capacity remain the same
among models with different skews. On the other hand,
application of the combined mode suggested that the
shear key does not provide a major additional capacity
as the skew changes. The application of uniform load
shows that the effectiveness of the shear key is dependent
on skew angles; it reduces as skew angle increases. For

30
0 Skew
30 Skew
45 Skew
60 Skew

Base shear (103 kN)

25
20
15
10
5
0
0

30

60
90
Displacement (mm)

120

150

Uniform load (finite element)

Skew angle
0
30

45

60

Stage

Displ. (mm)

Base shear (kN)

421

Description

17

10399

Yield at bottom of C1, C2, C3, C4

II

61

22163

Yield at top of C1, C2, C3, C4

16

8646

Yield at bottom of C3

II

30

13552

Yield at bottom of C1, C2, C4 and at top of C2, C3, C4

III

48

16102

Yield at top of C1

IV

95

16905

Failure at top of C3

14

6063

Yield at bottom of C3

II

16

6801

Yield at bottom of C1, C2, C4

III

31

9694

Yield at top of C1, C2, C3, C4

IV

114

11471

13

5131

Yield at bottom of C3 and C4

II

15

5568

Yield at bottom of C1 and C2

III

29

7756

Yield at top of C1, C3, C4

IV

52

8998

Yield at top of C2

130

8922

Failure at top of C3 and C4

Failure at top of C3

Fig. 5 Uniform load pushover curves at various skew angles: Case I shear keys effective

422

EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION

Vol.7

12

Base shear (103 kN)

10
8
6
4

0 Skew
30 Skew
45 Skew
60 Skew

2
0
0

Skew angle
0

30

45

60

Stage
I

Displ. (mm)
13

30

60
90
Displacement (mm)

120

150

Uniform load (finite element)

Base shear (kN)
Description
5241
Yield at bottom of C1, C2, C3, C4

II

23

6886

Yield at top of C1, C2, C3, C4

III

144

8674

Failure at top of C1, C2, C3, C4

IV

144

3647

Failure at bottom of C1, C2, C3, C4

16

5987

Yield at bottom of C1 and C2

Yield at bottom of C3 and C4 and at top of C1, C2, C3, C4

II

26

7944

III

135

10344

21

5847

Yield at bottom of C1 and C2

II

47

9065

Yield at bottom of C3 and C4 and at top of C1, C3, C4

III

75

10664

14

4777

Yield at bottom of C2

II

17

5612

Yield at bottom of C1, C3, C4

III

30

7510

Yield at top of C1, C3, C4

IV

35

7947

Yield at top of C2

132

8883

Failure at top of C4

Failure at top of C1, C2, C3, C4

Yield at top of C2

Fig. 6 Uniform load pushover curves at various skew angles: Case II shear keys failed

moderate skew (< 30o), larger pushover capacity and

smaller displacement is predicted by Case I; while for
larger skew (> 45o), it suggests that the shear keys do
not have a significant effect on pushover capacity and
displacement.
4.4. Comparison between time history and pushover
analyses
In order to investigate the accuracy of pushover
analyses in predicting the lateral capacity of skewed
highway bridges, individual maximum responses
from nonlinear time history analyses are plotted on
the pushover curves. These data are plotted for the FE
models only (Figs. 8 and 9). The maximum resultant
base shear and maximum displacement at the center of
the midspan (control node) in the transverse direction
are plotted for each of the seven ground motions. Two
abutment support conditions are considered: Case I

(shear key/restrained) and Case II (without shear key/

not restrained) with two cases of ground motions. The
first case, ST, applies the stronger component in the
transverse direction of the bridge while the second case,
SL, is the opposite. It should be noted that additional
time history analysis was done using Kobe and Sylmar
motions scaled to 0.5g for Case I and ST case and
compared against pushover curves.
For Case I, the time history results compare well
for the 0 degree skew. Both ST and SL cases follow
the Mode 1+2 and Uniform Load pushover curves very
well; although, the base shear capacity is slightly over
predicted by the Uniform Load pushover curve. The
time history results tend to cluster around the first hinge
formation on the pushover curves. The first transverse
mode pushover curve predicts the closest response for
0.5g ground motions. For the 30 degree skew, all the
predicted pushover curves compare well to the time
history results for the ST and SL cases; although, the

No.4

Ahmed Abdel-Mohti et al.: Seismic response of skewed RC box-girder bridges

Base shear (103 kN)

Transverse mode

Mode 1+2

Case I (w/shear keys)

Case II (w/o shear keys)

30

423

Uniform load

0 Skew

20

10

Base shear (103 kN)

0
30 Skew

30

20

10

Base shear (103 kN)

0
45 Skew

30

20

10

Base shear (103 kN)

0
60 Skew

30

20

10

0
0

35

70

105

140 0

35

70

105

140 0

35

Displacement (mm)
Fig. 7 Effect of shear keys on pushover curves at various skew angles

70

105

140

EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION

Mode 2 Transverse mode

20

10

10

I
II
I
II
III
I
II
III
IV

Mode 1+2

Uniform load

# Pushover stages

10

40
80
120
Displacement (mm)

30 skew stages
Finite element
Description

Stage
Mode 2

20

40
80
120
Displacement (mm)

Vol.7

Uniform load

30

20

40
80
120
Displacement (mm)

Kobe
Kobe (0.5g)
Landers
Sylmar
Sylmar (0.5g)
Rinaldi
Simqke 1
Simqke 2
Simqke 3

Mode 1+2

30
Base shear (103 kN)

Base shear (103 kN)

30

Base shear (103 kN)

424

Yield at top of C1 and C2

Yield at top of C3 and C4 and at bottom of C1, C2, C3, C4
Yield at top of C2
Yield at top of C1, C3, C4 and at bottom of C1, C2, C3, C4
Failure at bottom of C2 and C3
Yield at bottom of C3
Yield at bottom of C1, C2, C4 and at top of C2, C3, C4
Yield at top of C1
Failure at top of C3

Fig. 8 Comparison of time history results and pushover capacities for 30 skew, ST direction, Case I - shear keys effective

20

10

30

Kobe
Kobe (0.5g)
Landers
Sylmar
Sylmar (0.5g)
Rinaldi
Simqke 1
Simqke 2
Simqke 3
# Pushover stages

30

20

10

00

40
80
120
Displacement (mm)

Mode 1+2
Base shear (103 kN)

Mode 2 Transverse mode

Base shear (103 kN)

Base shear (103 kN)

30

40
80
120
Displacement (mm)

Stage
Mode 1
Mode 1+2

Uniform load

I
II
I
II
III
IV
I
II
III
IV
V

Uniform load

20

10

00

40
80
120
Displacement (mm)

60 skew stages
Finite element
Description
Yield at bottom of C1, C2, C3, C4
Yield at top of C1, C2, C3, C4
Yield at bottom of C3
Yield at bottom of C1, C2, C4 and at top of C2 and C3
Yield at top of C1 and C4
Failure at top of C3
Yield at bottom of C3 and C4
Yield at bottom of C1 and C2
Yield at top of C1, C3, C4
Yield at top of C2
Failure at top of C3 and C4

Fig. 9 Comparison of time history results and pushover capacities for 60 skew, ST direction, Case I - shear keys effective

No.4

Ahmed Abdel-Mohti et al.: Seismic response of skewed RC box-girder bridges

Mode 1+2 pushover curve underpredicts the response

from the Kobe and Sylmar motions for the ST case
(Fig. 8). The Kobe, Landers, and Sylmar motions are
closer to the fully yielded mechanism predicted by
the pushover curves while the Rinaldi and SIMQKE
motions are closer to the first hinge formation. Once
again, the first transverse mode profile gives the better
prediction for ground motions beyond the design level,
but the Uniform Load also gives satisfactory results. For
the 45 degree skew and ST case, the three SIMQKE and
Landers motions fall below the pushover curves while
the Kobe, Rinaldi, and Sylmar motions suggest that the
pushover curves predicts smaller maximum base shear,
except for the first transverse mode profile. The time
history responses are clustered around the fully yielded
mechanism for Mode 1+2 and the Uniform Load. For
0.5g ground motions, the first transverse mode pushover
curve accurately predicts the maximum response from
the time history analyses. For the SL case, the pushover
curves over predict the maximum response due to
the three SIMQKE motions. In both Mode 1+2 and
Uniform Load, the maximum responses from the time
history analyses are clustered around the fully yielded
mechanism. For the 60 degree skew shown in Fig. 9,
the time history results compare well with the Mode 1+2
pushover curve for both cases. The first transverse mode
predicts a larger capacity and the 0.5g ground motion
results do not compare well with any of the pushover
curves; although, the results are conservative with respect
to the first transverse mode pushover curve. For the ST
and SL cases, the maximum responses are clustered
in between the partial and fully yielded mechanisms.
For Case II, for the ST and SL cases, the first
transverse mode and Uniform Load pushover curves
underpredict the response from all seven ground
motions for the 0 degree skew. The Mode 1+2 pushover
curve compares well with the time history results. All
the motions cluster near the fully yielded mechanism
for both cases. For the 30 degree skew, the pushover
curves and time history results compare well for the ST
and SL cases; although the first transverse mode and
Uniform Load pushover curves slightly underpredict the
time history responses, especially for the SL case. Once
again, the motions are clustered around the fully yielded
mechanism. For the 45 degree skew, the seven ground
motions are more scattered, especially the Kobe and
Landers motions. The scatter for the ST case follows
the first transverse mode and Uniform Load pushover
curves exactly. The Kobe and Landers motions cluster
around the fully yielded mechanism while the maximum
responses due to the rest of the motions fall near the
partially yielded system. For the SL case, all three
pushover curves slightly underpredict the time history
responses. For the 60 degree skew, the three pushover
curves once again slightly underpredict each ground
motion response for the ST and SL cases. All of the
ground motions are clustered near the fully yielded
mechanism for both cases.

425

5 Conclusions
As mentioned earlier, the behavior of skewed
highway bridges is complex and modeling assumptions
affect the predicted seismic performance. In this study,
various parameters were studied such as: skew angle,
shear key effect, direction of components of ground
motions, as well as the adequacy of the pushover method
to estimate the lateral capacity of skew bridges. The
following conclusions are drawn:
(1) Predicted modal properties with the FE and BS
models were comparable; the BS model was successful
in capturing the modal coupling due to the skew and
the significant modes needed for further analysis.
Nonetheless, FE models should be considered when
dealing with very large skew angles (> 30) in order to
capture the higher mode effects.
(2) Boundary conditions (hence modeling
assumptions) have a significant effect on pushover
analyses and choice of pushover load profile.
(3) For Case I and II, the Uniform Load profile was
the most consistent in predicting similar sequences
of hinge formation between the FE and BS models.
Ultimately, the BS model, when compared to the
FE model, accurately captured the overall nonlinear
behavior of the bridge when using the Uniform Load
profile.
(4) Based on the pushover analyses, each load profile
considered suggested a different conclusion regarding
the effectiveness of the shear keys; while the Uniform
Load profile results suggested that the effectiveness of
the shear keys reduces with skew angle, the Multimode
profile results suggested that the shear keys do not
necessarily improved the pushover capacity, and the
Transverse Mode profile results suggested that shear
keys are effective in providing additional capacity that is
fully effective despite the skew angle.
(5) The direction of the two horizontal components
of the strong motions relative to the longitudinal and
transverse directions did not have any significant effect
on the overall response.
(6) The pushover curves considered in this study
predicted the maximum deformation response and the
sequence of hinge formations reasonably accurately
when compared to response predictions from individual
nonlinear time history analyses.
(7) For larger skew angles, the results from the
nonlinear time history analyses agree with the observed
yield mechanism in the columns from the pushover
analyses.

References
American Association of State Highway and
Transportation Officials (AASHTO) (1998), LRFD
Bridge Design Specifications, 2nd Edition, Washington
D.C.

426

EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION

Bentz EC and Collins MP (2000), Response-2000,

Version 1.0.5, University of Toronto, Canada.
Bjornsson S, Stanton J and Eberhard M (1997),
Seismic Response of Skew Bridges, 6th U.S. National
Conference on Earthquake Engineering, pp.112.
Computers and Structures, Inc. (2005), SAP2000,
Version 10.0.5, Integrated Structural Analysis and
Design Software, Berkeley, CA.
Federal Highway Administration (FHWA) (1996),
Seismic Design of Bridges, Design Example No.
4 Three-span Continuous CIP Concrete Bridge,
Publication No. FHWA-SA-97-009, October 1996.
Ghobarah AA and Tso, WK (1974), Seismic Analysis of
Skewed Highway Bridges with Intermediate Supports,
Earthquake Engineering and Structural Dynamics, 2:
235248.
Maleki S (2002), Deck Modeling for Seismic Analysis
of Skewed Slab-girder Bridges, Engineering Structures,
24: 13151326.
Maragakis E (1984), A Model for the Rigid Body

Vol.7

Motions of Skew Bridges, Doctoral Thesis, California

Institute of Technology, Pasadena, CA.
Megally SH, Silva PF and Seible F (2002), Seismic
Response of Sacrificial Shear Keys in Bridge
Abutments, Report No. SSRP-2001/23, Department
of Structural Engineering, University of California, San
Diego.
Meng, JY and Lui EM (2000), Seismic Analysis and
Assessment of a Skew Highway Bridge, Engineering
Structures, 22: 14331452.
PEER Strong Motion Database (2000), Pacific
Earthquake Engineering Research Center, Berkeley,
CA.
Saiidi M and Orie D (1992), Earthquake Design Forces
in Regular Highway Bridges, Computers & Structures,
44(5): 10471054.
Wakefield RR, Nazmy AS and Billington DP (1991),
Analysis of Seismic Failure in Skew RC Bridge,
Journal of Structural Engineering, ASCE, 117(3): 972986.