You are on page 1of 11

Engineering Structures 26 (2004) 15731583

www.elsevier.com/locate/engstruct

Comparison of random vibration methods for multi-support


seismic excitation analysis of long-span bridges
K. Soyluk 
Department of Civil Engineering, Gazi University, Maltepe, Ankara 06570, Turkey

Received 22 December 2003; received in revised form 27 May 2004; accepted 28 May 2004

Abstract

In this paper, the spatial variability eects of ground motions on the dynamic behaviour of long-span bridges are investigated
by a random vibration based spectral analysis approach and two response spectrum methods. The spatial variability of ground
motions between the support points is taken into account with the coherency function, which arises from three sources: incoher-
ence, wave-passage and site-response eects. Random vibration analyses are performed on two deck-type arch bridges and a
cable-stayed bridge model. Power spectral density function and response spectrum values used in random vibration analyses are
determined depending on the recordings of September 20, 1999, ChiChi, Taiwan earthquake. The results strongly imply that the
ltered white noise ground motion model can be accepted as a rather convenient model to represent actual earthquake ground
motions. It can be also observed that the structural responses for each random vibration analysis depend largely on the intensity
and frequency contents of power spectral density functions.
# 2004 Elsevier Ltd. All rights reserved.

Keywords: Random vibration analysis; Spatially varying ground motion; Spectral analysis; Multiple support response spectrum method; Deck
arch bridge; Cable-stayed bridge

1. Introduction Deterministic analyses of long-span bridges for


multi-support and travelling seismic wave eects are
One of the engineering challenges of long-span carried out by various researchers [13]. These studies
bridges is the determination of their dynamic behav- outline the requirement of the consideration of multi-
iour under the eect of dierential earthquake ground support and travelling seismic wave eects for the
motions. Since an earthquake excitation consists of the dynamic response analysis of long-span bridges. One of
superposition of a large number of waves with dierent the leading random vibration analyses of long-span
characteristics, the dierent positions along a long-span bridges to multi-support excitation is performed by
bridge generally are subject to dierent motions. In cal- Abdel-Ghaar and Rubin [4]. This study also under-
culating the dynamic response, the assumption of uni- lines the importance of the multi-support excitation.
form ground motion at the supports of these extended The eect of spatially varying ground motions on the
structures cannot be considered valid. Dierential random vibration response of bridges has been studied
ground motions may induce responses, commonly generally by spectral analysis approach in the last
known as pseudo-static component, which is absent in decade. For long multi-support structures, the spatially
structures uniformly excited at their base. The non- varying ground motion becomes a multi-support exci-
synchronous ground motion can induce a response tation. Zerva [5] evaluated the responses of continuous
very dierent from that calculated considering the same two- and three-span beams to spatially varying ground
ground motion at each support. motions and examined the validity of the commonly
used assumption of equal support motion. Hao [6] ana-

Tel.: +90-312-2317400/2231; fax: +90-312-2308434. lysed the in-plane stochastic responses of incompress-
E-mail address: ksoyluk@gazi.edu.tr (K. Soyluk). ible circular arches to spatially correlated multiple
0141-0296/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2004.05.016
1574 K. Soyluk / Engineering Structures 26 (2004) 15731583

excitations. Harichandran et al. [7] studied the Golden spectrum method for the seismic analysis of cable-
Gate suspension bridge and two deck arch bridges to a stayed bridges subjected to correlated stationary ran-
general spatially varying earthquake ground motion dom ground motion. The method was based on the
which neglects the site-response eect. It was concluded relationship between the power spectral density func-
that the use of identical excitations is in general unac- tion and the response spectrum of the input ground
ceptable for those long-span bridges. Zembaty [8] pre- motion, and the fundamentals of the random vibration
sented a numerical sensitivity study of the local site theory. Although the analysis takes into account the
eects on a four-span bridge response. An analysis of a partial correlation of ground motions between the sup-
bridge response with supports founded on dierent ports, the nite propagation eect of the ground
soils was carried out. Allam and Datta [9] presented a motion and the site-response eect was ignored.
frequency domain spectral analysis for the seismic Dynamic properties of long-span bridges should also
analysis of cable-stayed bridges for the multi-compo- be experimentally tested. This verication can be
nent stationary random ground motion. The ground accomplished by the instrumentation of existing
motion was represented by its power spectral density bridges, full-scale testing of prototype bridges and lab-
function and a spatial correlation function. Zanardo oratory model testing. The literature about the experi-
et al. [10] carried out a parametrical study of the mental testing of long-span bridges is very limited.
pounding phenomenon associated with the seismic However, in a study performed by Garevski et al. [16],
response of multi-span simply supported bridges and a small-scale dynamic model with articial mass simu-
highlighted that multi-support analysis gives results lation of the Jindo Bridge (JB) was designed and built.
markedly dierent from the uniform dynamic analysis. Sinusoidal and random signal tests by using a shaking
Dumanoglu and Soyluk [11] investigated the stochastic table were used to determine the dynamic character-
response of a cable-stayed bridge subjected to spatially istics of the model. A comparison of the test results
varying ground motions based on a recently developed with those obtained from the mathematical model of
model. The importance of site-response eect was the prototype was made to verify the mathematical
investigated particularly. These studies performed by model verication. Good agreement between experi-
random vibration based spectral analysis method gen- mental and analytical results was obtained. Since the
erally underline the signicance of the spatial varia- main idea of this study is to perform a numerical study
bility of ground motions between support points. related to the spatial variability of ground motions and
The response spectrum method has proven to be an its eects on the response of arch and cable-stayed
accurate and practical method for seismic analysis. bridge models, the verication of the dynamic proper-
Most seismic codes specify the earthquake motion in ties of these bridges are not considered.
terms of the response spectrum and not the power The objective of this study is to compare the random
spectral density. Literature on the application of the vibration responses of deck-type arch bridges and
random vibration based response spectrum analysis for cable-stayed bridges to spatially varying ground
seismic analysis of lifeline structures is meagre. In motions including the wave-passage, incoherence and
recent years, Der Kiureghian and Neuenhofer [12] site-response eects. In this study, three dierent ran-
developed a response spectrum method for the seismic dom vibration methods are utilised to determine the
analysis of linear multi-degree-of-freedom, multiply- dynamic behaviour of these long-span bridges. As one
supported structures subjected to spatially varying of these methods is the spectral analysis approach of
ground motions. The rule provides the response of a random vibration theory based on the power spectral
linear system subjected to incoherent support excita- density specication of the ground motion, the other
tions directly in terms of the conventional response two methods are random vibration based response
spectra at the support degrees of freedom and a coher- spectrum methods. Although the spectral analysis of
ency function describing the spatial variability of the cable-stayed bridges subjected to spatially varying
ground motion. Nakamura et al. [13] applied this ground motions is studied by Dumanoglu and Soyluk
response spectrum method on the Golden Gate bridge [11], it has not been found any publication that
for spatially varying ground motions. It was found that includes the random vibration based response spectrum
the new response spectrum method oers a simple and analysis of cable-stayed bridges subjected to spatially
viable alternative for seismic analysis of multiply-sup- varying ground motions including the wave-passage,
ported structures. Der Kiureghian et al. [14] performed incoherence and site-response eects. As a new contri-
a comprehensive investigation of the multiple support bution to the literature the random vibration based
response spectrum method for seismic analysis of three spectral and response spectrum analyses of deck-type
to ve-span bridge structures. Special attention was arch bridges subjected to spatially varying ground
given to the eect of site response arising from vari- motions including all the spatial variability eects are
ation in the soil conditions at dierent supports of the also studied in this paper. Another objective of this
structure. Allam and Datta [15] presented a response study is to verify the ltered white noise (FWN)
K. Soyluk / Engineering Structures 26 (2004) 15731583 1575

ground motion model widely used in the literature, support degree of freedom k, Dk(wi, fi) denotes the dis-
before conducting extensive random vibration investi- placement response spectrum ordinate at support
gation of the bridges. This verication is performed by degree of freedom k for the frequency and mode of i,
comparing the structural responses for the FWN and quk ul , quk slj , qski slj are cross-correlation coecients
ground motion model and actual ground motions. between the support motions and the modes of the
structure.
2. Spectral analysis approach

Since the formulation of the random vibration 4. Seismic excitation for random vibration analyses
theory for spatially varying ground motion is given
previously by many researchers [719], in this study Both in the spectral analysis approach and response
only the required nal equations will be considered. spectrum method the mean of maximum responses
Spectral analysis approach is based on principles of depend on the cross-power spectral density function of
stationary random vibration theory and provides an ground acceleration
approximate estimate of the mean of the absolute
maximum response of the structure in terms of the Guk ul w ckl wGuk uk w Gul ul w1=2 3
power spectral density function of ground motion and
where ckl(w) is the coherency function describing the
a coherency function. Any response quantity can be
decomposed into dynamic and pseudo-static compo- variability of the ground acceleration processes for sup-
nents, i.e., z zd zs when there is a dierential exci- port degrees of freedom k and l as a function of fre-
tation at the supports. The total mean-square responses quency w. Recently, Der Kiureghian [18] proposed a
can be obtained from general composite model of spatial seismic coherency
function in the following form:
r2z r2zd r2zs 2Covzs ;zd 1

where r2zd and r2zs are the dynamic and pseudo-static ckl w ckl wi expihkl ww hkl ws  4
variances, respectively, and Covzs ;zd is the covariance i w s
where ckl w , ckl w and ckl w characterise the inco-
between the dynamic and pseudo-static responses zd herence, the wave-passage and the site-response eects,
and zs [17]. The mean of the absolute maximum respectively. For the incoherence eect, the extensively
response can be obtained depending on the peak used model proposed by Harichandran and Vanmarcke
response and standard deviation (rz) of z(t).
[19] is considered.
For the spectral analysis approach, the power spec-
tral density function of the ground acceleration char-
3. Response spectrum method
acterising the earthquake process is assumed to be of
The multiple support response spectrum method was the form of FWN ground motion model modied by
developed by Der Kiureghian and Neuenhofer [12] Clough and Penzien [20]
based on fundamental principles of stationary random
w4g 4f2g w2g w2
vibration theory. This rule provides the response of a Guk uk w S0  2
linear system subjected to incoherent support excita- w2g w2 4f2g w2g w2
tions directly in terms of the conventional response
w4
spectra and a coherency function. The combination
  5
2
rule for the mean of absolute peak response is given in w2f w2 4f2f w2f w2
the form [12]
" where S0 is the amplitude of the white-noise bedrock
X
m Xm
acceleration, wg and fg are the resonant frequency and
E maxjztj ak al quk ul uk;max ul;max
k1 l1
damping ratio of the rst lter, and wf and ff are those
X
m X
m X
n of the second lter. In this study, rm, medium and
2 ak blj quk slj uk;max Dl wj ;fj soft soil types are used at the support points of the
k1 l1 j1 considered bridges and the lter parameters corre-
#1=2
m X
X m X
n X
n sponding to these soil conditions are obtained from
bki blj qski slj Dk wi ;fi Dl wj ;fj 2 Der Kiureghian and Neuenhofer [12].
k1 l1 i1 j1 Der Kiureghian and Neuenhofer [12] derived the fol-
where, ak and bki are the eective inuence coecients lowing relation between the power spectral density of
and eective modal participation factors, respectively, the acceleration and the displacement response spec-
uk,max denotes the mean peak ground displacement at trum associated with each support motion for the
1576 K. Soyluk / Engineering Structures 26 (2004) 15731583

multiple support response spectrum method The CSCB is a two-lane solid-ribbed steel deck arch

bridge. While the total bridge span reaches 371 m, the
wp2 2fw 4 Dk w;f 2 total length of the arch span is 213 m. The deck is div-
Guk uk w p 6
w wpf p ps ps w0 ided into three continuous segments by hinged tower
In this expression, p and wf are parameters selected by connections at the ends of the main span. The two
adjusting the power spectral density for low fre- dimensional model of the CSCB is shown in Fig. 2.
quencies, s is the duration of strong motion phase of The equivalent beam elements of the deck and the arch
are numbered from 1 to 11. The two truss elements
the motion, f is a reference damping ratio, and ps(w)0 is
representing the cables which transfer longitudinal
the peak factor of white noise. In this study, the fol-
loads from the deck to the arch are labelled 1 and 2.
lowing parameters proposed by Der Kiureghian and
The deck moment releases resulting from the hinge
Neuenhofer [12] are used: p 3 and wf 0:705 rad=s.
connections at the towers are also shown in the gure.
In both bridge models the approach spans are repre-
sented by translation and rotation springs. In order to
5. Description of the bridge models reduce the number of degrees of freedom in the bridge
analyses, the full three-dimensional model of each
5.1. Deck arch bridge models
bridge was reduced to a two dimensional bridge model
In this study, two deck-type arch bridge models are for input motions in the longitudinal direction. Further
chosen: The Cold Spring Canyon Bridge (CSCB) in reductions in the number of degrees of freedom were
California and the New River Gorge Bridge (NRGB) made by assuming that the arch and the deck at each
in WV. panel point would have the same vertical translations.
The NRGB is a four lane, box truss, steel deck arch While NRGB is modelled with 81 degrees of freedom,
bridge. Both the deck and the arch are essentially box the CSCB is modelled with 66 degrees of freedom. The
two-dimensional modelling technique used in this study
trusses consisting of four box girder chords connected
proved to be a very eective modelling method. The
by lateral and vertical truss members. The NRGB has
modal analysis responses presented in Ref. [21] were
a north approach span of 175 m, a south approach of
generally within 10% of the results based on full three-
193 m and a main span of 554 m. While the total
dimensional modelling analysis. In this study a modal
bridge span reaches 924 m, the total length of the arch
damping ratio of 5% is used for both bridges.
span is 518 m. The approach span deck segments are
isolated from the main span deck segments by expan-
5.2. Cable-stayed bridge model
sion joints at the ends of the main span. At these
points, the bottom chords of the approach span deck In this study, the JB built in South Korea is chosen
are pinned to the top of the bents while the bottom as a cable-stayed bridge model. JB has three spans: the
chords of the main span deck are attached to the bents main span of 344 m and two side spans of 70 m. The
by rollers. The two dimensional model of the NRGB is stays are arranged in a fan conguration and converge
shown in Fig. 1. The equivalent beam elements of at the top of the A-frame towers. The stiening girder
the deck and the arch are numbered from 1 to 14. and the towers of the JB were made from steel [22].
The two truss elements representing the cables which A 2% of damping coecient is adopted for the
transfer longitudinal loads from the deck to the arch response calculations. To investigate the random
are labelled 1 and 2. The deck axial force and moment vibration response of the JB, two-dimensional math-
releases resulting from the expansion joints at the ends ematical model is used for calculations (Fig. 3). It has
of the main span are also shown. been shown that a two-dimensional analysis of the

Fig. 1. Two-dimensional mathematical model of New River Gorge Bridge.


K. Soyluk / Engineering Structures 26 (2004) 15731583 1577

Fig. 2. Two-dimensional mathematical model of Cold Spring Canyon Bridge.

Fig. 3. Two-dimensional mathematical model of Jindo Bridge.

cable-stayed bridge provides natural frequencies and


mode shapes which are in close agreement with those
obtained by three-dimensional analysis [2]. The chosen
nite element model is represented by 420 degrees of
freedom. As the stiening girder and towers are repre-
sented by 139 beam elements, the cable stays are mod-
elled with 28 truss elements.

6. Numerical results

6.1. Verication of FWN ground motion model

In the rst part of this study the widely used ground


motion model, the FWN dened by the Eq. (5) is veri-
ed. For this purpose the recordings of 20 September
1999, ChiChi, Taiwan earthquake, namely the
ILA027E, HWA033N and TAP051W components
recorded at rm, medium and soft soil conditions,
respectively, are used. Acceleration power spectral den-
sity functions for these earthquake ground motion
components are given in Fig. 4. The spectral density
functions for the FWN ground motion model are
also given in these gures for each soil type. As indi-
cated, the lter parameters corresponding for each soil
condition are obtained from Der Kiureghian and
Neuenhofer [12]. To provide a fair comparison, the
amplitude of white-noise bedrock acceleration S0 is
obtained for each soil type by equating the variance of
the ground acceleration (Eq. (5)) to the variance of the
components of the ChiChi earthquake acceleration
recorded at rm, medium and soft soil conditions. The
calculated intensity parameter values for each soil
type are, S0 firm 4:85
10 5 m2 =s3 , S 0 medium Fig. 4. Power spectral density functions at (a) soft soil; (b) medium
1:28
10 4 m2 =s3 , S0 soft 9:33
10 5 m2 =s3 . As soil; (c) rm soil.
1578 K. Soyluk / Engineering Structures 26 (2004) 15731583

can be observed in Fig. 4, the acceleration spectral den- pared in Fig. 5 for both cases. It is obvious from these
sity functions obtained for the ChiChi earthquake gures that the results obtained for the actual ground
ground motion and the FWN model coincide with each accelerations are very close to those of the FWN
other. ground motion model response values. The dierence
For each component of the ChiChi earthquake of order up to 10% between these results outlines that
recorded at rm, medium and soft soil conditions, the FWN ground motion model can be used as ground
spectral analysis of the cable-stayed bridge model is motion model with an acceptable accuracy. After the
performed for uniform ground motion case with STO- verication of the FWN ground motion, this model
CAL [23] computer program. The same bridge model is will be used throughout the study as ground motion
also analysed for the ltered white noise ground for spectral analysis.
motion model based on the calculated intensity para-
meters. The vertical displacements of the cable-stayed 6.2. Random vibration analyses
bridge model calculated for each soil type are com-
In the second part of this study, deck-type arch
bridge models and a cable-stayed bridge model sub-
jected to spatially varying ground motions will be stud-
ied for dierent random vibration methods. In this part
of this study, the goal is to establish the random
vibration method that will generate the highest struc-
tural response. The random vibration methods con-
sidered in this study are as follows:

1. Spectral analysis approach of random vibration


theory which considers the FWN ground motion
model as an earthquake ground motion model
(Fig. 4).
2. Response spectrum method based on the relation-
ship between the power spectral density function
and the response spectrum of the input ground
motion. The response spectrum displacement is
derived from the power spectral density function of
ground acceleration by using Eq. (6). This response
spectrum method uses the displacement response
spectrum shown in Fig. 6 based on the components
of the ChiChi earthquake and the power spectral
density function of FWN of rm, medium and soft
soil types. This method will be named as PSDF
based response spectrum method throughout the
study.

Fig. 5. Vertical deck displacements of JB for (a) soft soil; (b) Fig. 6. Displacement response spectra obtained from Eq. (6) for dif-
medium soil; (c) rm soil. ferent soil conditions for ps w0 3:2497.
K. Soyluk / Engineering Structures 26 (2004) 15731583 1579

propagates from the Jindo Island site to the mainland


site for the considered cable-stayed bridge with an nite
wave velocity of 500 m/s. The considered earthquake
ground motions lasting up to 70.0 s are applied to the
bridges in the longitudinal directions, either in power
spectral density shape or in response spectrum form.

6.2.1. Arch bridge responses


For the NRGB it is assumed that the south supports
are founded on medium soil and the north supports are
founded on soft soil conditions. The mean of absolute
maximum vertical displacements of NRGB along the
deck are compared in Fig. 8 for the considered random
vibration methods. This gure shows that the
maximum vertical displacement occurs at the middle of
the bridge deck for the spectral analysis and PSDF
based response spectrum methods. Meanwhile for the
response spectrum method, maximum displacement
occurs close to the north supports. While the displace-
ments calculated by the PSDF based response spec-
trum are the smallest, the response spectrum method
yields the largest displacements. The spectral analysis
method induces 12% larger displacements at the middle
of the deck than those of the PSDF based response
spectrum method and 10% smaller displacement than
those of the response spectrum method. However,
because the maximum displacement obtained by the
spectral analysis method is at the middle of the bridge
and the response spectrum method causes maximum
displacement close to the north supports, the maximum
dierence between these methods is within 65%.
The mean of absolute maximum bending moments
at the deck members are compared in Fig. 9. This
gure indicates that the bending moments from the

Fig. 7. Displacement response spectra (f 0:02) of ChiChi earth-


quake recorded at the soil conditions of (a) soft soil; (b) medium soil;
(c) rm soil.

3. Multiple support response spectrum method based


on the response spectrum specication of the input
motion. This method will be called as response spec-
trum method in this study. This model uses the dis-
placement response spectrum curves of the three
components of the ChiChi earthquake as ground
motions (Fig. 7).

In this study, it is assumed that while the ground


motion is propagating across the deck arch bridges Fig. 8. Mean of absolute maximum vertical deck displacements of
from the south supports to the north supports, it NRGB.
1580 K. Soyluk / Engineering Structures 26 (2004) 15731583

approach and by Eq. (6) for multiple support response


spectrum method. Although the same earthquake
ground motion is used for both methods, the power
spectral density functions of ground accelerations char-
acterising the ground motions are dierent in character.
For the CSCB it is assumed that the south supports
of the approaching span are founded on rm soil, the
south support of the main span is founded on medium
soil, and the north supports are founded on soft soil
conditions. The calculated mean of absolute maximum
vertical displacements of CSCB along the deck are
compared in Fig. 11. Although this gure indicates that
the harmony between spectral analysis and PSDF
based response spectrum is rather close to the variation
obtained for the NRGB, the response spectrum
method yields somehow dierent variation than those
of the NRGB. In this case, the response spectrum
method causes even smaller responses than those of the
other two analyses. At the middle of the deck, spectral
Fig. 9. Mean of absolute maximum deck bending moments of analysis results 12% and 20% larger response values
NRGB.
than those of the PSDF based response spectrum
method and spectrum analysis, respectively.
spectral analysis and PSDF based response spectrum Mean of absolute maximum bending moments at the
methods are quite close to each other, namely of order arch members are compared in Fig. 12. This gure also
up to 10%, and are much smaller than the bending shows the agreement between the spectral analysis and
moments from the response spectrum method. The PSDF based response spectrum methods. Although
maximum dierence between the spectral analysis these two methods are dierent in character, the agree-
approach and the response spectrum method is 110%. ment of the results is caused by equal power spectral
This trend shows itself again in the comparison of density functions for both methods. The response spec-
bending moments of arch members (Fig. 10). The dis- trum method causes much larger bending moments
crepancy between the spectral analysis approach and than those of the former two methods. As the
the response spectrum method is most probably caused maximum dierence between the spectral analysis and
by the dierence of the variances of ground accelera- PSDF based response spectrum method is 15%, the
tions described by Eq. (5) for spectral analysis

Fig. 10. Mean of absolute maximum arch bending moments of Fig. 11. Mean of absolute maximum vertical deck displacements of
NRGB. CSCB.
K. Soyluk / Engineering Structures 26 (2004) 15731583 1581

Fig. 13. This gure reveals that the displacement values


calculated by spectral analysis are slightly larger than
the displacements from the PSDF based response spec-
trum method, of the order of 7% at the middle of the
deck where the displacements are the largest. The
closeness of the results of both methods can be attrib-
uted to the fact that the PSDF based response spec-
trum method depends heavily on the power spectral
density function of the FWN ground motion model.
The displacements obtained from the response spec-
trum method are generally much larger than those of
the other two methods especially at the main span, of
the order of 84% at the middle of the deck. It should
be also outlined that such a comparison is not entirely
fair, because structural responses can show important
variations for ground motions even with similar
response spectrum. The mean of absolute maximum
longitudinal displacements of the tower located at the
Jindo Island, are also compared for the random
Fig. 12. Mean of absolute maximum arch bending moments of vibration methods (Fig. 14). This gure also outlines
CSCB. the fact that the spectral analysis and PSDF based
response spectrum methods give close response values
dierence between the spectral analysis and response and the response spectrum method induces much larger
spectrum method is 125%. So, the most evident dier- responses. While the spectral analysis method causes
ence in the responses of both bridges is the response of 10% larger displacement value at the top of the tower
bending moments. compared to those of the PSDF based response spec-
trum method, the response spectrum method induces
6.2.2. Cable-stayed bridge responses 65% larger displacement value than those of the spec-
For the JB, it is assumed that the island abutment is tral analysis method. This trend observed for the deck
founded on rm soil, the island pier is founded on and tower displacements shows itself again in the
medium soil and the supports at the mainland site are comparison of deck bending moments (Fig. 15).
founded on soft soil conditions. The mean of absolute These results indicate that for the arch deck-type and
maximum vertical deck displacements are compared in cable-stayed bridge models the worst case of random

Fig. 13. Mean of absolute maximum vertical deck displacements Fig. 14. Mean of absolute maximum longitudinal tower displace-
of JB. ments of JB.
1582 K. Soyluk / Engineering Structures 26 (2004) 15731583

function of the FWN ground motion model and the


response spectrum of the earthquake ground motion.
Although these two methods are dierent in character,
the agreement of the results can attributed to the fact
that the power spectral density functions for both
methods are equal.
The multiple support response spectrum method esti-
mates based on the response spectrum specication of
the earthquake motion show trends which are consist-
ent with the trends observed in the other two random
vibration methods. However, there are signicant dif-
ferences between the results. The discrepancy between
the spectral analysis approach and multiple support
response spectrum method is mostly caused by the dif-
ference of the variances of ground accelerations
described by Eq. (5) for spectral analysis approach and
by Eq. (6) for multiple support response spectrum
method. So, even for the same ground motion dierent
random vibration methods may cause very dierent
Fig. 15. Mean of absolute maximum deck bending moments of JB. results depending on the intensity and frequency con-
tents of the power spectral density functions.
vibration analyses is the multiple support response
spectrum method which results in higher responses
depending on the intensity and frequency contents of References
the power spectral density function, especially for
member forces. [1] Nazmy AS, Abdel-Ghaar AM. Seismic response analysis of
cable stayed bridges subjected to uniform and multiple-support
excitations. Report no. 87-SM-1. Princeton (NJ): Department of
Civil Engineering, Princeton University; 1987.
7. Conclusions [2] Garevski M, Dumanoglu AA, Severn RT. Dynamic character-
istics and seismic behaviour of Jindo bridge, South Korea. Struc-
The rst part of this study involves the verication of tural Engineering Review 1988;1:1419.
the most widely recognised ground motion model, [3] Dusseau RA, Wen RK. Seismic response of deck type arch
bridges. Earthquake Engineering and Structural Dynamics
namely the FWN ground motion model. This part is
1989;18:70115.
followed by a comparative analysis which is performed [4] Abdel-Ghaar AM, Rubin LI. Vertical seismic behaviour of sus-
by a frequency domain spectral analysis approach, a pension bridges. Earthquake Engineering and Structural
power spectral density function based response spec- Dynamics 1989;11:119.
trum method and a multiple support response spectrum [5] Zerva A. Eect of spatial variability and propagation of seismic
method on deck-type arch bridges and a cable-stayed ground motions on the response of multiply supported struc-
tures. Probabilistic Engineering Mechanics 1991;6:21221.
bridge model subjected to spatially varying ground [6] Hao H. Arch responses to correlated multiple excitations. Earth-
motions. The spatial variability of ground motions quake Engineering and Structural Dynamics 1993;22:389404.
between the support points is taken into account with [7] Harichandran RS, Hawwari A, Sweiden BN. Response of long-
the incoherence, wave-passage and site-response eects. span bridges to spatially varying ground motion. Journal of
As earthquake ground motions, three dierent compo- Structural Engineering 1996;122(5):47684.
[8] Zembaty Z, Rutenberg A. On the sensitivity of bridge seismic
nents of the ChiChi earthquake recorded at rm, response with local soil amplication. Earthquake Engineering
medium and soft soil conditions, are employed in this and Structural Dynamics 1998;27(10):10959.
study. [9] Allam SM, Datta TK. Seismic behaviour of cable-stayed bridges
When compared with the actual earthquake ground under multi-component random ground motion. Engineering
motions, it can be observed that the results obtained Structures 1999;22:6274.
[10] Zanardo G, Hao H, Modena C. Seismic response of multi-span
for the ltered white noise ground motion model are simply supported bridges to a spatially varying earthquake
quite satisfactory. Therefore, from this stand point this ground motion. Earthquake Engineering and Structural Dynam-
model can be accepted as a rather convenient model ics 2002;31:132545.
for actual ground motions. [11] Dumanoglu AA, Soyluk K. A stochastic analysis of long span
The responses obtained from the frequency domain structures subjected to spatially varying ground motions includ-
ing the site-response eect. Engineering Structures 2003;25:
spectral analysis are in close agreement with results 130110.
obtained from the response spectrum method based on [12] Der Kiureghian A, Neuenhofer A. A response spectrum method
the relationship between the power spectral density for multiple-support seismic excitations. Report no. UCB/
K. Soyluk / Engineering Structures 26 (2004) 15731583 1583

EERC-91/08. Berkeley (CA): Earthquake Engineering Research National Science Foundation MSU-ENGR-88-002. Michigan
Center, College of Engineering, University of California; 1991. (MI): Department of Civil and Environmental Engineering,
[13] Nakamura Y, Der Kiureghian A, Liu D. Multiple-support College of Engineering, Michigan State University; 1988.
response spectrum analysis of the Golden Gate bridge. Report [18] Der Kiureghian A. A coherency model for spatially varying
No. UCB/EERC-93/05. Berkeley (CA): Earthquake Engineer- ground motions. Earthquake Engineering and Structural
ing Research Center, College of Engineering, University of Dynamics 1996;25:99111.
California; 1993. [19] Harichandran RS, Vanmarcke EH. Stochastic variation of earth-
[14] Der Kiureghian A, Keshishian P, Hakobian A. Multiple support quake ground motion in space and time. Journal of Engineering
response spectrum analysis of bridges including the site-response Mechanics 1986;112(2):15474.
eect and MSRS code. Report No. UCB/EERC-97/02. Berkeley [20] Clough RW, Penzien J. Dynamics of structures, 2nd ed.
(CA): Earthquake Engineering Research Center, College of Singapore: McGraw Hill, Inc; 1993.
Engineering, University of California; 1997. [21] Dusseau RA. Unequal seismic support motions of steel deck
[15] Allam SM, Datta TK. Analysis of cable-stayed bridges under arch bridges. PhD thesis, Michigan State University, Michigan;
multi-component random ground motion by response spectrum 1985.
method. Engineering Structures 2000;22:136777. [22] Tappin RGR, Clark PJ. Jindo and Dolsan bridges: design. Pro-
[16] Garevski MA, Brownjohn JMW, Blakeborough A, Severn RT. ceedings of the Institution of Civil Engineers 1985;78:1281300.
Resonance-search tests on a small-scale model of a cable-stayed [23] Button MR, Der Kiureghian A, Wilson EL. STOCAL-User infor-
bridge. Engineering Structures 1991;13:59-66. mation manual. Report No. UCB/SEMM-81/2, Earthquake
[17] Harichandran RS, Wang W. Response of one- and two-span Engineering Research Center, College of Engineering, University
beams to spatially varying seismic excitation. Report to the of California; 1981.

You might also like