Professional Documents
Culture Documents
Abstract: In seismic analysis of bridges, the designer chooses the direction of the applied earthquake forces arbitrarily. This paper
investigates the effects of seismic force direction on the responses of slab-girder skewed bridges in response spectrum and time history
linear dynamic analyses. The combination rules for orthogonal earthquake effects, such as the 100/ 30, 100/ 40 percentage rules and the
SRSS method are also examined. It is concluded that either the SRSS or the 100/ 40 percentage rule in the skew direction should be used
Downloaded from ascelibrary.org by University of Leeds on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.
in the response spectrum analysis of skewed bridges. For time history analysis none of the combination rules provide conservative results.
In this case, the application of paired acceleration time histories in several angular directions is recommended.
DOI: 10.1061/共ASCE兲1084-0702共2006兲11:1共122兲
CE Database subject headings: Bridges, skew; Seismic effects; Finite element method; Earthquakes.
and parallel to the abutments. Note that, due to symmetry of It is seen that the periods of vibration are coupled and depend on
support springs, the center of mass and stiffness coincide, and the stiffnesses Ke and Kc.
hence, translational inertia loads at the center of mass do not The translational mode shapes of a skewed bridge associated
cause any torsional loading. with periods derived above are in a direction that is skewed
A two-dimensional finite element model of the bridge and ␥ degrees, where
the three degrees of freedom at the center of mass are shown in
Fig. 4. The springs ke represent the stiffness of each elastomeric Kxx  + sin2 ␣
bearing in the x direction. Likewise, the springs kc represent the tan ␥ = = 共6兲
Kxy sin ␣ cos ␣
total lateral stiffness of the end-diaphragms in the t direction.
Let M and Io represent the total mass and the mass moment in which  = Ke / Kc.
of inertia of the bridge superstructure. These are assumed to be Note that, for  = 0, the second translational mode shape
concentrated at the center of mass. of a skewed bridge is in the t direction—that is, parallel to the
Assuming zero damping, the equation of motion for the model abutments 共␥ = ␣兲. This happens only when Kc is extremely large
bridge in Fig. 4, subjected to ground acceleration in the x and y compared to Ke. This is the case for skewed bridges supported on
directions 共ügx , ügy兲, can be written as follows elastomeric bearings on both sides. For very large  values, as for
a pin-supported bridge, ␥ approaches 90°; the motion will be
冤 冥冦 冧 冤 冥冦 冧 冦 冧
M 0 0 üx Kxx Kxy Kx ux Mügx along x axis.
The stiffness matrix in Eq. 共1兲 is diagonal for a straight bridge
0 M 0 üy + Kyx Kyy Ky uy = Mügy 共1兲 and all the nondiagonal terms are equal to zero. This means that a
0 0 Io ü Kx Ky K u 0 force applied in the x or y direction at the center of mass will not
rotate the deck and produces only translations in the same direc-
In the case of free vibration, the right hand side of the above
tion. As a result, a nonskewed bridge will have its fundamental
equation is set equal to zero.
mode of vibration in the x 共or y兲 direction only. On the contrary,
The stiffness term Kxx⫽summation of all stiffnesses in the
for an elastically supported skewed bridge, the stiffness matrix is
x direction. Likewise, the stiffness term Kyy⫽summation
of all stiffnesses in the y direction. Transforming the local
coupled and nondiagonal terms exist. This gives rise to motion in analyses are performed on short- to long-span slab-girder bridges.
both x and y directions for the translational modes of vibration. In The spans of 10, 20, and 30 m are chosen as representative spans
the case of a skewed bridge, the stiffness terms Kxy and Kyx, of such bridges and the properties are listed in Table 1. The cross-
Downloaded from ascelibrary.org by University of Leeds on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.
though equal to each other, are not equal to zero. However, due to frames lateral stiffness is obtained by assuming angle diagonals
symmetry, unless the base motion is torsional, the skewed bridge shown in Fig. 2 to be present in each 2 m panel. The top of the
does not experience torsional motion either. In other words,
cross frame is connected to a rigid deck.
Kx = Kx = Ky = Ky = 0. This is always the case as long as the two
The finite element analysis model of the skewed bridge
lines of support at the ends 共abutments兲 are parallel to each other.
Having only K, the torsional mode is totally uncoupled from is shown in Fig. 4. The SAP2000 finite element program, by
translational modes of vibration and can be treated independently. Computers and Structures, Inc. 共2000兲, is used for the analysis.
The period of vibration for the torsional mode is The model has two displacement degrees of freedom 共ux , uy兲
and one rotational 共u.兲 Since cross frames and lateral load-
冑
resisting elements of the superstructure are supposed to remain
Io elastic under earthquake forces, only linear dynamic analysis
T = 2 共7兲
K is performed on the models. This is consistent with AASHTO
Article 4.6.2.8.2.
End cross frames provide an important load path for the
seismically induced loads. Seismic forces at the deck would
Analysis Model
have to pass through the cross frames to arrive at the top of the
In the dynamic analysis of skewed bridges the designer must bearings. Zahrai and Bruneau 共1998兲 have shown that the inter-
choose the earthquake input direction and a response combination mediate cross frames do not affect the seismic performance of
rule as discussed in the previous section. To compare the effects slab-girder bridges significantly. Hence, they are not considered
of these parameters on the response of skewed bridges, a series of in the models.
Fig. 5. Results the of response spectrum analysis for the 20 m span bridge: 共a兲 displacement in the x direction; 共b兲 displacement in the y direction;
共c兲 elastomer force; and 共d兲 cross-frame force
Fig. 6. Comparison between the application of paired time histories and the SRSS combination rule for the 10-m span bridge: 共a兲 San Fernando
time history and 共b兲 Northridge time history
The deck slab is modeled with rectangular shell elements. lateral stiffness. The modeling of the superstructure is consistent
The girders are modeled with frame elements connected to with the recommendations of Mabsout et al. 共1997兲.
shell elements at each joint. They are free to rotate but restrained Furthermore, for single-span bridges, the abutment stiffness is
vertically at each end. This will capture the contribution of the not considered in the analyses. The justification for this assump-
girders’ weak axis moment of inertia to the superstructure stiff- tion is described herein. In the transverse direction, abutments
ness for transverse loading. In the longitudinal direction, the ends are, in general, much stiffer than the cross frames. Since the two
of the girders are attached to a spring representing the elastomer’s are modeled as springs in series, the abutment will not contribute
Fig. 7. Comparison between the application of paired time histories and the SRSS combination rule for the 20-m span bridge: 共a兲 San Fernando
time history and 共b兲 Northridge time history
Fig. 8. Comparison between the application of paired time histories and the SRSS combination rule for the 30-m span bridge; 共a兲 San Fernando
time history and 共b兲 Northridge time history
to the overall stiffness and only the effect of cross-frame stiffness However, Maleki 共2004兲 has investigated the gap effect in seismic
is considered. In the longitudinal direction, the abutment and soil analysis of bridges separately. Furthermore, it is assumed that the
stiffness is much higher than the elastomer. Being connected bridge will not impact the abutment backwall in the longitudinal
in series again, an equivalent spring will only capture the direction. These parameters only complicate the analysis, but the
effect of the elastomers. The small gap 共about 3 mm兲 between conclusions of this paper will remain unaffected. Note that only
the retainer and elastomer is usually assumed to be closed by single-span bridges are considered in this study and therefore no
designers. This is also the assumption by the present authors. piers exist.
Fig. 9. Comparison between the application of the 100/ 30, 100/ 40, and the SRSS combination rules in the x,y and n,t directions for the 10-m
span bridge: 共a兲 San Fernando time history and 共b兲 Northridge time history
Fig. 10. Comparison between the application of the 100/ 30, 100/ 40, and the SRSS combination rules in the x,y and n,t directions for the 20-m
span bridge: 共a兲 San Fernando time history and 共b兲 Northridge time history
Response Spectrum Analysis bridges with pinned supports. The n,t axes define the principal
vibration modes for skewed bridges supported on elastomeric
The models described above are subjected to the AASHTO’s bearings on both sides. Therefore, these axes, if chosen as input
design RS with acceleration coefficient A = 0.4 and Soil Type II. seismic direction, can be critical and produce the maximum
Input directions for the ground acceleration are chosen as x, y, n, responses for the bridge.
and t axes 共see Fig. 4兲. The x,y axes are representative of The responses in the x and y directions are combined using
translational vibration modes for straight bridges or skewed the SRSS, 100/ 30, and 100/ 40 rules. These are denoted as
Fig. 11. Comparison between the application of the 100/ 30, 100/ 40, and the SRSS combination rules in the x,y and n,t directions for the 30-m
span bridge: 共a兲 San Fernando time history and 共b兲 Northridge time history
Fig. 12. Responses for the El Centro ground motion applied at various angles: 共a兲 displacement in the x direction; 共b兲 displacement in the y
direction; and 共c兲 cross-frame force
0.3-xy and 0.4-xy on the graphs. The same combination rules combined with 100/ 30 and 100/ 40 rules shows a dependency on
are also applied to the responses in the n and t directions and the skew angle and can be higher or lower than the SRSS values.
denoted by 0.3-nt and 0.4-nt on the graphs. Since the SRSS Close agreement is found for the y displacement of the bridge
combination is independent of the input direction, it is considered using all combination rules and for all skew angles, as shown
the most accurate method for the RS analysis. The responses Fig. 5共b兲.
that were calculated include the displacements in the x and y Figs. 5共c and d兲 shows the elastomeric bearing and the end
directions, end cross-frame shear force, elastomeric bearing shear cross-frame shear forces. The elastomer forces are proportional to
force and base shear of the structure for all skew angles and the x displacement of the bridge and the curve shows similar
spans. Only partial results are shown here for the sake of brevity. variation. The cross-frame forces are reasonably predicted by all
Fig. 5共a兲 shows the bridge displacement in the x direction for combination rules except for 100/ 30 rule with input in the
the 20-m span using the combination rules described. The 10-m x,y directions.
and 30-m spans are not shown and behave similarly. It is seen In general, the 100/ 40 percentage rule with the RS input in the
that the 100/ 30, 100/ 40, and SRSS rules yield similar x displace- n,t directions has yielded the closest responses to the SRS method
ments for the response spectrum input in the n,t directions. in all cases and can be chosen as the preferred method in the RS
However, x displacement for the input in the x,y directions analysis.
Fig. 13. Responses for the San Fernando ground motion applied at various angles: 共a兲 displacement in the x direction; 共b兲 displacement in the y
direction; and 共c兲 cross-frame force
Fig. 14. Responses for the Northridge ground motion applied at various angles: 共a兲 displacement in the x direction; 共b兲 displacement in the y
direction; and 共c兲 cross-frame force
Time History Analysis responses for this bridge yields a value of 115 mm for the same
displacement. The same bridge behaves differently under the
In this section, the effect of input motion direction on the Northridge earthquake. The x displacement is equal to 147 and
response of skewed bridges in time history analysis is investi- 124 mm for the x,y and n,t input directions and 124 mm for the
gated. The bridge model shown in Fig. 4 with spans of 10, 20, SRSS method. This case clearly shows that none of the above
and 30 m are considered with skews ranging from 0 to 60°. Paired methods can yield the maximum response for a skewed bridge.
time histories for the 1940 El Centro earthquake with peak Therefore, in general, the x,y and n,t input directions could not
ground acceleration of 0.34g, the 1994 Northridge 共Sylmar be designated as the critical directions for all skew angles, and the
County兲 earthquake with peak ground acceleration of 0.84g, SRSS method does not yield a conservative response in all cases.
and the 1971 San Fernando 共Pacoima Dam兲 earthquake with
peak ground acceleration of 1.2g are chosen as input motions.
Damping is assumed to be 2% of critical. Accuracy of the Combination Rules in TH Analysis
To examine the adequacy of the combination rules, the 100/ 30,
Application of the Paired Time Histories 100/ 40, and the SRSS method that are commonly used in the RS
analysis are applied to the TH analysis of bridges. The input
The El Centro, Northridge, and San Fernando histories are
motions are not applied in pairs; rather, the strong component is
applied in the x,y and n,t directions. In each case, the pair of
applied separately in two orthogonal directions and the combina-
histories is also switched in direction and the maximum responses
tion rule is applied to responses. This is more conservative than
are noted. These responses are denoted as rគxy and rគnt on the
combining the responses due to a paired component of ground
graphs. The maximum displacement, elastomer force, and cross-
motion. Figs. 9–11 show the responses due to San Fernando and
frame force are recorded for each direction. However, the
Northridge ground motions based on the x,y and n,t input direc-
elastomer force is similar to the x displacement and is not shown
tions combined with the above rules. It is observed that the SRSS
on the graphs. In addition, the SRSS of responses when the
rule produces a more conservative response in the majority of
strong component of ground motion is applied in the x,y direction
is obtained. This is achieved by applying a single TH in the x and cases. However, it was shown previously that the SRSS method
y directions and finding the SRSS value of the two responses. itself is not conservative when a paired TH is applied simulta-
Paired time histories are not used for the SRSS value. The SRSS neously to bridges in the x,y- or n,t directions. Hence, the 100/ 30
value is independent of the input direction and is valid for the and 100/ 40 rules cannot be safely recommended for TH analysis
n,t axes as well. either.
Figs. 6–8 show the responses of bridges with skew angles
ranging from 0 to 60° due to San Fernando and Northridge Critical Input Motion Direction in TH Analysis
ground motions for spans of 10, 20, and 30 m, respectively. The
results due to El Centro are not shown due to size limitation on It was pointed out that the common input motion directions
the paper. It can be seen that the input directions considered do and combination rules are not always yielding the maximum
not uniquely define the maximum responses. The SRSS method responses for skewed bridges. In this section, the input motion
does not produce the maximum response for all skew angles direction is treated as a variable in an attempt to find the
either. For example, for a 10-m span bridge with 30° skew under critical direction by examining the responses. The three pairs of
San Fernando ground motion, the displacement in the x direction El Centro, Northridge, and San Fernando ground motions are
is equal to 92 and 168 mm when the motion is applied in the x,y applied at various angles to the bridge. The angle of input is
and n,t directions, respectively. The SRSS combination of varied in 10° increments, starting from the x,y axes and turning