You are on page 1of 19

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2004; 33:375393 (DOI: 10.1002/eqe.357)

Seismic response of a cable-stayed bridge deck under


multi-component non-stationary random ground motion

Said M. Allam1 and T. K. Datta 2; ;


1 Department of Civil Engineering; Alexandria University; Egypt
2 Civil Engineering Department; Indian Institute of Technology; New Delhi 110016; India

SUMMARY
A Markov method of analysis is presented for obtaining the seismic response of cable-stayed bridges
to non-stationary random ground motion. A uniformly modulated non-stationary model of the random
ground motion is assumed which is specied by the evolutionary r.m.s. ground acceleration. Both
vertical and horizontal components of the motion are considered to act simultaneously at the bridge
supports. The analysis duly takes into account the angle of incidence of the earthquake, the spatial
correlation of ground motion and the quasi-static excitation. A cable-stayed bridge is analysed under a
set of parametric variations in order to study the non-stationary response of the bridge. The results of
the numerical study indicate that (i) frequency domain spectral analysis with peak r.m.s. acceleration as
input could provide more r.m.s. response than the peak r.m.s. response obtained by the non-stationary
analysis; (ii) the longitudinal component of the ground motion signicantly inuences the vertical
vibration of the bridge; and (iii) the angle of incidence of the earthquake has considerable inuence on
the deck response. Copyright ? 2003 John Wiley & Sons, Ltd.

KEY WORDS: Markov method of analysis; seismic response of cable-stayed bridge; non-stationary
seismic excitation; modulating function; quasi-static bridge response

INTRODUCTION

A number of studies have been reported in recent years on the seismic analysis of cable-
supported bridges [19]. Some of these studies investigated also the eect of soilstructure
interaction on the seismic response of the bridge and made attempts to control the seismic
response using passive and hybrid control strategies. However, most of these studies have
been made either for a specied earthquake record or for an earthquake assumed to be a sta-
tionary random process. Very few studies considered an earthquake input as a non-stationary
random process. Hyun et al. [10] developed a method for non-stationary analysis of
suspension bridges subjected to multi-support excitations which was found to be mainly

Correspondence to: T. K. Datta, Civil Engineering Department, Indian Institute of Technology, New Delhi 110016,
India.
E-mail: tkdatta@civil.iitd.ernet.in
Received 13 February 2002
Revised 25 March 2003
Copyright ? 2003 John Wiley & Sons, Ltd. Accepted 28 August 2003
376 S. M. ALLAM AND T. K. DATTA

dependent upon the enveloping function of the time history of ground motion. The non-
stationary responses were obtained in terms of time-dependent variance functions. There have
been many studies on simpler structures to obtain their response to non-stationary seismic
excitations [1117]. Lin [11] treated the non-stationary excitation as a sequence of random
pulses. By modelling the earthquake as a ltered Poisson process, Shinozuka et al. [16, 17]
developed a procedure to obtain the time-dependent variance of the response. Debchaudhury
et al. [12, 13] and Gasparini et al. [14, 15] developed a method for obtaining the response
of multi-degree of freedom systems to non-stationary seismic excitation using the Markov
approach. The advantage of the Markov approach for the non-stationary analysis of structures
for seismic excitation is that it does not require the derivation of the evolutionary frequency
response function H (w; t), which may be dicult to obtain in many complex structures such as
suspension bridges. Furthermore, the method directly obtains the evolutionary r.m.s. response
of the system. However, the application of the Markov approach for the response analysis
of cable-stayed bridges to non-stationary seismic excitation is not straightforward. It involves
some complexities because of (i) the horizontal component of ground motion contributing to
the vertical vibration of the deck and (ii) the contribution of the pseudo-static component of
vertical motion to the total vertical vibration of the bridge deck.
Herein, the vertical response of the bridge deck of cable-stayed bridges to multi-component
partially correlated non-stationary random ground motion is obtained using a Markov formu-
lation unlike that presented by Hyun et al. [10]. Furthermore, the formulation is fairly general
in the sense that it takes into account the eects of the angle of incidence of the earthquake,
the ratio between the horizontal and vertical components of ground motion, the quasi-static
component of the response, and the movements of the tower supporting the cables. The uni-
formly modulated non-stationary model of the ground motion is considered in the study. Using
the proposed method of analysis, a cable-stayed bridge is analysed under a set of parametric
variations in order to investigate the non-stationary characteristics of the bridge responses.

THEORY

Seismic excitation
Seismic excitation is assumed to be a multi-component uniformly modulated non-stationary
random process. The three components of the ground motion are assumed to be dened in
the three principal directions of the earthquake and are assumed to be directed along the
principal axes of the bridge x, y, z or shifted with an angle  as shown in Figure 1. The evo-
lutionary r.m.s. acceleration for each component of the ground motion is specied. The spatial
correlation between the seismic excitations at two points is given by a correlation function.

ij = exp(arij ) cos(2Ko rij ) (1)

where a and Ko are parameters which depend on the direction of wave propagation as well
as wave type and earthquake type. The values of a and Ko are taken as 4.769 and 2.756,
respectively, as given by Loh [18], rij is the distance between the stations i and j measured
in the direction of wave propagation, which is assumed to coincide with the major principal
axis of the ground motion (u) as shown in Figure 1.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK 377

Figure 1. Layout of the principal axes of the bridge (x; y; z) and the principal
directions of the ground motion (u; v; w).

The non-stationary support excitations are considered as the outputs of lters excited by
the evolutionary white noise. For the formulation of the problem, the lters are augmented at
each support degree of freedom and are dened by

Xf i + 2f i !f i X f i + !f2i Xf i = Si + Wi
i = 1; 2; : : : ; ns (2)
Si + 2si !si S i + !si2 Si = Wi

where ns is the number of exciting degrees of freedom, i.e. the size of the excitation vector;
{Xf } is the vector of output of lters which is the input to the bridge supports at their degrees
of freedom; !f i and !si are the i-th lter parameters representing the predominant frequencies
and the other two parameters f i and si represent the damping ratios; {S } is the vector of
intermediate response and {W } is the vector of evolutionary white noise having a covariance
matrix as
ww (t; t + ) = E {[W (t)  w (t)] [W (t + )  w (t + )] T } = Q(t):() (3)
where  w is the mean vector of {W }, () is the Dirac delta function and Q(t) is known as
the matrix of white noise intensities. By integrating both sides of Equation (3)
 
ww (t; t + )= Q(t)() d 


(4)


ww (t; t + ) d =Q(t)

Thus, Q(t) is the integral of the covariance function of white noise excitation components. In
the present formulation, the elements of the intensity matrix are modelled as piecewise linear

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
378 S. M. ALLAM AND T. K. DATTA

functions of time, although they can take any shape. Typical elements of the covariance matrix
are

wiwi (t; t + ) = E {[Wi (t) Wi (t)][Wi (t + ) Wi (t + )] T }

= qii (t):() (5)

wiwj (t; t + ) = E {[Wi (t) Wi (t)][Wj (t + ) Wi (t + )] T }

= qij (t):()

= ij [qii (t)qjj (t)]1=2 () (6)

ij is the correlation function between excitations corresponding to the i-th and j-th d.o.f. and
is given by Equation (1), and Wi and Wj are the mean values of the i-th and j-th elements
of the vector {W }. In general, the ground motion is dened by its free-eld record which
is the output of the lter. The inputs to the lters represent the bed-rock excitation and the
lters reect the soil media. It will be subsequently seen that the formulation requires the
specication of the matrix Q(t) of the intensities of white noise. The elements of [Q(t)]
are determined with the help of the specied evolutionary r.m.s. acceleration of free-eld
ground motion and the characteristics of the lters as given by Soliman and Datta [19] and
Debchaudhary et al. [12, 13].

Modelling of the bridge deck


The bridge deck is idealized as a continuous beam over the outer abutments and the interior
towers as shown in Figures 2(a) and (b). The eect of the cables is taken as vertical springs
at the points of intersection between the cables and the bridge deck. The eect of the spring
stiness is taken as an additional vertical stiness to the exural stiness of the bridge. The
following assumptions are made for the formulation of the problem: (i) the bridge deck is
assumed to be a continuous beam (the girder and the tower are assumed to be axially rigid);
(ii) the beam does not transmit any moment to the tower through the girdertower connection;
(iii) the towers are assumed xed at the locations of the pier or well foundation; (iv) cables
are assumed straight under high initial tension due to dead load and are capable of supporting
negative force increment during vibration without losing the straight conguration; (v) an
appropriate portion of the mass of the cables is included in the dynamic analysis of the
bridge deck, and is assumed to be uniformly distributed over the idealized deck (in addition
to the deck mass); (vi) the beamcolumn eect in the stiness formulation of the beam is
considered for the constant axial force in the beam produced due to the initial tension in the
cable; and (vii) the cable dynamics are ignored for the bridge deck vibration, i.e., the tension
uctuations in the cable are assumed as quasi-static, and to not introduce any non-linearity
into the system.

Equation of motion; free vertical vibration and quasi-static analysis


The equation of motion for the dynamic displacement, i.e., the relative vertical displacement
y(xr ; t) of any segment r of the idealized deck with constant axial force Nr , neglecting the

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK 379

Figure 2. Problem identication: (a) layout of the bridge under multi-component ground motion;
and (b) idealization of the bridge deck.

shear deformation and rotary inertia, is given by

@ 4 y(xr ; t) @ 2 y(xr ; t) @y(xr ; t) Wr @ 2 y(xr ; t)


Ed Ir 4
+ Nr 2
+ Cr + = P(xr ; t)
@xr @xr @t g @t 2

Wr 
12 12
= gjr (xr )Xf j (t) cr gjr X f j r = 1; : : : ; Nb (7a)
g j=1 j=1

in which Ed Ir , Nr , Wr , g, and Ed are the exural rigidity, the constant axial force in the
r-th segment of the bridge deck produced by cables due to the dead load, the load per unit
length of the bridge, the acceleration due to gravity, and the modulus of elasticity of the
deck material, respectively. P(xr ; t) is dened as the applied load due to seismic excitations
at dierent support degrees of freedom. Xf j (t). j = 1; 2; : : : ; 12 are the accelerations at dierent
support degrees of freedom (Figure 2(a)) and gjr (xr ) is the vertical displacement of the r-th
segment of the bridge deck due to a unit displacement given at the j-th degree of freedom of
the supports. gjr (xr ) is obtained by solving for the entire bridge (i.e. deck, towers and cables),
considering no moment transfer between the deck and the tower, by a separate analysis using
standard structural analysis procedures. For multi-support excitation, it is quite often assumed
that the term Cr gjr , i.e., the damping force produced in the dynamic degrees of freedom due
to the movement of the support, may be ignored for a reasonable damping ratio since they do
not signicantly inuence the response [20]. With this assumption, the damping term of the
right-hand side of Equation (7(a)) is ignored and the equation retains the term containing the

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
380 S. M. ALLAM AND T. K. DATTA

ground acceleration only. The mode shapes and frequencies of the bridge are obtained by
the undamped free-vibration analysis as given by Chatterjee et al. [21].

State space formulation using modal coordinate


The dynamic displacement y(xr ; t) at any point in the r-th segment of the bridge deck is
given by


y(xr ; t) = n (xr )n (t) r = 1; 2; : : : ; Nb (7b)
n=1

where Nb is the number of beam segments, n (xr ) is the n-th vertical mode shape of the r-th
beam segment of the bridge, and n (t) is the n-th generalized coordinate. Furthermore, the
generalized equation of motion for the relative vertical vibration of the bridge deck can be
written as

12
n (t) + 2n !n n (t) + !n2 n (t) = jn Xf j (t) (8)
j=1

where Xf j (t) is the support displacement corresponding to the j-th degree of freedom
(Figure 2(a)) which is the output of a set of lters excited by the evolutionary white noise
as given by Equation (2); X f j (t) and Xf j (t) are the derivatives of Xf j (t), and jn is the
modal participation factor dened as

  
Nb W Lr
gjr (xr ) n (xr ) d xr
r=1 g 0
jn =
  Lr
j = 1; 2; : : : ; 8
Nb W
n2 (xr ) d xr
r=1 g 0

jn = 0:0 j = 9; 10; 11; 12

Combining Equations (2) and (8), the equation of motion of the bridge can be written in the
state space with white noise as input in the following form:
{Z(t)} = [A]{Z(t)} + [B]{W (t)} (9)
where {Z(t)} is the vector of state variables given by
[Z1j ; Z2j ; Z3j ; Z4j ; Z5n ; Z6n ; ] j = 1; : : : ; 12 and n = 1; : : : ; M (10)

in which Z1j = Sj ; Z2j = S j ; Z3j = Xf j ; Z4j = X f j j = 1; : : : ; 12


Z5n =n ; Z6n = n n = 1; : : : ; M
Wj (t)=Wu cos  Ww sin  j = 5; 6; 7; 8 (x-direct)
Wj =Wu j = 1; 2; 3; 4 (y-direct)
Wj =Wu sin  + Ww cos  j = 9; 10; 11; 12 (z-direct)

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK 381

Wu , Wv , and Ww are the three components of ground motion and  is the angle between Wu
and the x-direction (Figure 1). Explicit expressions of matrices A and B are given by Allam
[22]. The solution of Equation (9) in the time domain is given by
 t
{Z(t)} = e[A](tt0 ) {Z(t0 )} + e[A](t) [B]{W ()} d  (11)
t0

The matrix exponential in Equation (11) is dened by the power series as



tk
e[A]t = [A]k (12)
k=0 k!
Equation (11) can be written as
 t
{Z(t)} = [ (t; t0 )] {Z(t0 )} + [ (t; )] [B]{W ()} d  (13)
t0

where
(t; t0 ) = e[A](t; t0 ) (14)
(t; t0 ) is the state transition matrix and may be calculated in dierent ways [2325]. Here,
the method of similarity transformation of matrix A is used for establishing (t; t0 ) [24]. The
covariance of the state variable

 t1  t2
T
zz (t1 ; t2 ) = (t1 ; t0 )zz (t0 ) (t2 ; t0 ) + (t1 ; 1 )BQ(1 )(1 2 )B T T (t2 ; 2 ) d 1 d 2
t0 t0
(15)

The integral term is obtained by assuming t1 t2 , t0 61 6t1 and t0 62 6t2 , and the integration
is rst performed with respect to 1 , then with respect to 2 . In this case,

 t2
zz (t1 ; t2 ) = (t1 ; t0 )zz (t0 ) T (t2 ; t0 ) + (t1 ; 2 )BQ(2 )B T T (t2 ; 2 ) d 2
t0
(16)

putting t1 = t2 = t, the covariance matrix of the state vector is given by


 t
zz (t) = (t; t0 )zz (t0 ) T (t; t0 ) + (t; )BQ()B T T (t; ) d  (17)
t0

If the mean of the excitation vector is assumed to be zero (i.e. W = 0), then Equation (17)
fully describes the state output vector {Z(t)}. Thus, the covariance matrix of response can
be calculated at any time t provided that the covariance matrix at any previous time t0 , and
the matrix of strengths of the excitations (i.e. intensity matrix [Q(t)] of {W (t)}) are known.
The method of calculation of the piecewise linear elements (Figure 3) of the {Q(t)} matrix
is given in detail by Debchaudhary et al. [12, 13] and Soliman and Datta [19].

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
382 S. M. ALLAM AND T. K. DATTA

Figure 3. Piecewise linear strength intensity function.

Evolutionary mean square displacement and bending moment


The expression for the total displacement Y (xr ; t) at any bridge deck segment r at any time
t is given as

Y (xr ; t) = [(xr )]{} + [G(xr )]{Xf } (18)

where

[(xr )] = [1 (xr ) M (xr )]

[ { } ] T = {  1  M }

[G(xr )] = [g1 (xr ) g8 (xr )]

[{X f }] T = [Xf 1 Xf 8 ]

Then the evolutionary mean square of the displacement can be written as


Y2 (xr ; t) = [(xr )][ (t)][(xr )] T + [G(xr )][X f X f (t)][G(xr )] T

+ [(xr )][X f (t)][G(xr )] T + [G(xr )][X f  (t)][(xr )] T (19)

where [ (t)], [X f X f (t)], [X f (t)], and [X f  (t)] can be easily assembled from the covari-
ance matrix of the state variable [zz (t)].
Similarly, the evolutionary mean square value of the bending moment can be obtained by
2 2
using E1 d d(x
x2
r)
and E1 d dG(x x2
r)
instead of (xr ) and G(xr ) respectively.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK 383

NUMERICAL STUDY

The example problem of the cable-stayed bridge, shown in Figure 4, is the same as that
considered by Morris [26]. Uniformly modulated non-stationary ground motion is expressed
in terms of the evolutionary r.m.s. ground acceleration. Three modulating functions, shown
in Figure 5, are considered in the study. The value of
fg (the peak value of the r.m.s.
ground acceleration) is taken as 0:61 m= s 2 . The ground motion is described along the principal

Figure 4. Cable-stayed bridge used in the parametric study.

Figure 5. Modulating functions used in the parametric study.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
384 S. M. ALLAM AND T. K. DATTA

Figure 6. Strength intensity functions corresponding to the modulating functions.

directions of the earthquake by specifying dierent proportions between Ru , Rv , and Rw . For


the angle of incidence  = 0:0 , the ground motions in the three directions refer to those
corresponding to the longitudinal, vertical and transverse directions of the bridge. Three sets
of lter parameters (!si ; !f i = 0:1!si ; si = f i ) are used to represent soft, rm and very rm
soil conditions as (!si = 6:28; si = 0:4), (!si = 15:708; si = 0:6), (!si = 31:416; si = 0:8),
respectively. The strength intensity functions for the set of lter parameters describing the rm
soil for the three modulating functions are shown in Figure 6. It is seen that the maximum
value of the strength intensity function occurs at the same time where the corresponding
modulating function attains its peak. However, the shapes of the strength intensity functions
are not exactly the same as those of the corresponding modulating functions. The evolutionary
r.m.s. responses are calculated with lt =ld = 4; Ru : Rv : Rw = 1:0 : 1:0 : 1:0;  = 0:0 and the set
of lter parameters corresponding to the rm soil condition, unless mentioned otherwise. Note
that the r.m.s. response denotes the r.m.s. value of the total response given by Equation (18).

Eect of the degree of non-stationarity (the nature of the modulating function)


The degree of non-stationarity is denoted by the sharpness of the modulating function. The
sharper the variation of the modulating function with time, the greater is the degree of non-
stationarity. Figures 7 and 8 show the evolutionary r.m.s. responses (standard deviation =
r.m.s. value since the random process is assumed with zero mean) for the points 3 and 9
representing the locations at the mid-points of the outer span and the inner span, respectively.
It is seen that the nature of the evolutionary response and its maximum value depend upon the
modulating function being used. The sharper the modulating function (modulating function
(1)), the smaller is the maximum value of the r.m.s. response. This is the case because
the energy contents of the excitation for the sharper (normalized) modulating function are

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK 385

Figure 7. Evolutionary r.m.s. of the responses at point (3):


(a) displacement response; and (b) bending moment.

less. Furthermore, the variation of the r.m.s. response with time is dierent from that of
the corresponding modulating function of excitation with time. This is expected because of
the combined eects of the lack of correlation between support excitations, the contribution
of the horizontal component of ground motion on the vertical deck displacement and the
contribution of the quasistatic component of the response on the total response. The eect of
the non-stationarity on the response is shown by the dierence between the maximum r.m.s.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
386 S. M. ALLAM AND T. K. DATTA

Figure 8. Evolutionary r.m.s. of the responses at point (9):


(a) displacement response; and (b) bending moment.

response and the r.m.s. response as obtained from the stationary analysis (frequency domain
spectral analysis). The latter is determined with input as double-ltered PSDF of ground
acceleration having
fg = 0:61 m= s 2 (i.e. the peak r.m.s. acceleration of the evolutionary input).
The dierence between the stationary r.m.s. responses and the maximum value of the non-
stationary r.m.s. responses is about 17% for the modulating function (2), about 25% for the
modulating function (3) and about 45% for the modulating function (1). For further parametric
studies, the modulating function (3) is used. It is apparent from Figure 5 that the dierence
between the two responses becomes less as the modulating function becomes wider. This
is the case because the degree of non-stationarity becomes more as the modulating function
becomes narrower.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
Table I. Eect of the ratio between the three components of ground motion on the r.m.s. responses.
1.0 : 0.4 : 0.6a 0.6 : 0.5 : 0.6a 0.8 : 0.5 : 0.6a

Copyright ? 2003 John Wiley & Sons, Ltd.


Stationary Non-stationaryb Stationary Non-stationaryb Stationary Non-stationaryb
Point D M D M D M D M D M D M
1 0.0204 0.0 0.0201 0.0 0.0255 0.0 0.0251 0.0 0.0255 0.0 0.0251 0.0
2 0.1289 1707 0.1053 1366 0.0876 1176 0.0718 948 0.1092 1456 0.0894 1170
3 0.1436 1658 0.1177 1332 0.0968 1153 0.0795 934 0.1212 1421 0.0992 1147
4 0.1201 1009 0.0987 826 0.0809 672 0.0667 551 0.1013 847 0.0835 694
5 0.0204 604 0.0201 533 0.0255 554 0.0251 490 0.0255 607 0.0251 537
6 0.1547 702 0.1276 601 0.1046 494 0.0868 425 0.1308 605 0.1082 520
7 0.2726 1247 0.2221 1012 0.1814 852 0.1482 696 0.2287 1059 1.0865 863
8 0.3099 745 0.2510 614 0.2061 534 0.1675 445 0.2599 649 0.2110 538
9 0.3047 1130 0.2438 882 0.2014 758 0.1612 604 0.2548 940 0.2040 747
D displacement (m), M bending moment (t.m.).
a
R u : R v : Rw .
b
Peak value of the evolutionary r.m.s. response (modulating function (3)).
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK

Firm soil; partially correlated;  = 0:0 .

Earthquake Engng Struct. Dyn. 2004; 33:375393


387
388 S. M. ALLAM AND T. K. DATTA

Figure 9. Strength intensity functions for dierent soil parameters (soil condition).

Eect of the ratio between the three components of ground motion


Table I shows the eect of the ratio between the three components of the ground motion on the
maximum r.m.s. responses obtained from the non-stationary analysis and the r.m.s. responses
obtained from the stationary analysis. Three dierent ratios (Ru : Rv : Rw ) between the three
components of the ground motion are considered for the study, namely, (i) 1:0 : 0:4 : 0:6;
(ii) 0:6 : 0:5 : 0:6; and (iii) 0:8 : 0:5 : 0:6. The angle of incidence of the earthquake is taken as
 = 0:0 , i.e. the three components of the ground motion coincide with the principal directions
of the bridge (x; y; z). It is seen that the relative magnitude of the x components of the ground
motion has a predominant eect on the responses at the mid-points of the outer and the inner
spans. This is expected since the horizontal movement of the towers and the abutments (in
the longitudinal direction) substantially inuences the vertical movement of the bridge deck
because of the uctuation in the cable tension. Furthermore, it is observed that as Ru (= Rx )
decreases, the response also decreases. It is seen that a relative reduction in the vertical
component of the ground motion with a relative increase in the longitudinal component of
the ground motion may also increase the responses.

Eect of the nature of the lter coecients


The dierent lter coecients denoting three dierent soil conditions have been used in
this study. The evolutionary strength functions for the three soil conditions are shown in
Figure 9. Although the evolutionary free-eld r.m.s. ground acceleration is the same for all
three soil conditions, the shapes of the evolutionary strength functions are dierent and hence
the evolutionary r.m.s. responses would be dierent for dierent soil conditions. This is the
case because the lter coecients (Equation (2)) modify the frequency contents of the free-
eld ground motion. Table II compares the maximum r.m.s. responses obtained from the

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
Table II. Eect of the nature of the evolutionary excitation due to soil conditions on the r.m.s. responses.
Soft soil Firm soil Very rm soil
(6.2832, 0.62832, 0.4, 0.4) (15.708, 1.5708, 0.6, 0.6)a (31.416, 3.1416, 0.8, 0.8)a

Copyright ? 2003 John Wiley & Sons, Ltd.


Stationary Non-stationaryb Stationary Non-stationaryb Stationary Non-stationaryb
Point D M D M D M D M D M D M
1 0.3794 0.0 0.3698 0.0 0.0509 0.0 0.0503 0.0 0.0104 0.0 0.0103 0.0
2 0.5124 5294 0.4736 4650 0.1549 2090 0.1272 1690 0.0613 885 0.0505 730
3 0.5813 5417 0.5391 4800 0.1707 2058 0.1403 1674 0.0670 888 0.0552 736
4 0.5319 4437 0.4981 4155 0.1427 1178 0.1178 968 0.0546 479 0.0449 397
5 0.3787 3001 0.3691 2782 0.0509 1065 0.0503 939 0.0104 552 0.0103 493
6 0.7816 3334 0.7418 3154 0.1847 886 0.1536 762 0.0638 396 0.0517 345
7 1.2069 3666 1.1332 3206 0.3181 1509 0.2600 1236 0.1109 681 0.0892 567
8 1.4097 2511 1.3268 2256 0.3615 964 0.2940 806 0.1203 464 0.0958 396
9 1.2211 2504 1.1330 2027 0.3522 1347 0.2850 1077 0.1186 553 0.0949 452
D displacement (m), M bending moment (t.m.).
Partially correlated: Ru : Rv : Rw = 1:0 : 1:0 : 1:0;  = 0:0 .
a
Filter coecients !s , !f , s , f .
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK

b
Peak value of the evolutionary r.m.s. response (modulating function (3)).

Earthquake Engng Struct. Dyn. 2004; 33:375393


389
390 S. M. ALLAM AND T. K. DATTA

Table III. Eect of damping on the r.m.s. responses (non-stationary response).


Soft soil Firm soil Very rm soil

Point % Da Mb D M D M
1 2 0.3698 0.0 0.0503 0.0 0.0103 0.0
4 0.2432 0.0 0.0352 0.0 0.0062 0.0
3 2 0.5813 5417 0.1403 1674 0.0552 736
4 0.3456 3642 0.1246 1232 0.0346 580
5 2 0.3691 2782 0.0503 939 0.0103 493
4 0.2426 1792 0.0352 682 0.0062 345
9 2 1.133 2027 0.2850 1077 0.0949 452
1.221c 2504c 0.3522c 1347c 0.1186c 553c
4 0.7853 1424 0.2046 725 0.0625 361
0.8924c 1872c 0.2614c 1085c 0.0942c 428c
a
D = Displacement (m).
b
M = Bending moment (t.m.).
c
Stationary r.m.s. response.

non-stationary analysis with the r.m.s. responses as obtained from the stationary analysis. It
is seen from the table that the responses are more for the lter coecients corresponding to
the soft-soil condition. The reason for this is attributed to the fact that the frequency contents
of the excitations for the soft-soil condition are centred around a narrow band of frequencies
which are close to the fundamental frequency of the bridge. The dierence between the
responses obtained by the stationary and non-stationary analyses remains nearly the same for
the three lter coecients.
In order to study the eect of damping on the non-stationary response for dierent types of
excitations due to soil condition, the damping of the bridge is changed to 4% (in the original
data it is taken as 2%). The results for the two cases are compared in Table III. It is seen
that the peak value of non-stationary r.m.s. response is signicantly reduced. The decrease is
about 33%. The corresponding decrease is found to be about 25% for the stationary r.m.s.
response. Thus, the eect of damping is greater for non-stationary response analysis.

Eect of the angle of incidence of earthquake ()


The eect of the angle of incidence () on the displacement response is shown in Table IV.
 = 0:0 corresponds to the case when the major principal component of the earthquake
is along the longitudinal direction of the bridge and  = 90 indicates the case where the
moderate principal component of the earthquake is along the longitudinal axis of the bridge
(Figure 1). The minor principal component of the earthquake is always in the vertical direc-
tion. Furthermore,  = 90 denotes the case of fully correlated excitations. As the ratio between
the three components of the earthquake is taken as the same, the change in  predominantly
eects the correlation between excitations at any two points by modifying the separation length
rij = Lij cos  (Figure 1).
The table shows that the maximum response at any section of the bridge deck does not
necessarily occur for  = 90 ; it may occur for an angle of incidence  between 0 and 90 .
The responses are a minimum for  = 90 , i.e. for fully correlated ground motion. The critical

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
Table IV. Eect of the angle of incidence of the earthquake () on the r.m.s. displacement.
 = 0:0  = 30  = 45  = 70  = 90

Copyright ? 2003 John Wiley & Sons, Ltd.


Point Stationary Non-stationarya Stationary Non-stationarya Stationary Non-stationarya Stationary Non-stationarya Stationary Non-stationarya

1 0.0509 0.0503 0.0509 0.0503 0.0509 0.0503 0.0509 0.0503 0.0509 0.0503
2 0.1549 0.1272 0.1641 0.1338 0.1685 0.1375 0.1322 0.1110 0.0823 0.0727
3 0.1707 0.1403 0.1811 0.1477 0.1870 0.1526 0.1442 0.1209 0.0855 0.0749
4 0.1427 0.1178 0.1512 0.1239 0.1570 0.1287 0.1203 0.1014 0.0740 0.0661
5 0.0509 0.0503 0.0509 0.0503 0.0509 0.0503 0.0509 0.0503 0.0509 0.0503
6 0.1847 0.1536 0.2009 0.1665 0.2089 0.1735 0.1441 0.1217 0.0739 0.0662
7 0.3181 0.2600 0.3557 0.2898 0.3561 0.2914 0.2382 0.1962 0.0955 0.0802
8 0.3615 0.2940 0.4181 0.3387 0.3961 0.3229 0.2599 0.2124 0.1157 0.0945
9 0.3522 0.2850 0.4164 0.3333 0.3820 0.3068 0.2435 0.1948 0.1277 0.1046
Firm soil; partially correlated; Ru : Rv : Rw = 1:0 : 1:0 : 1:0.
a
Peak value of the evolutionary r.m.s. response (modulating function (3)).
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK

Earthquake Engng Struct. Dyn. 2004; 33:375393


391
392 S. M. ALLAM AND T. K. DATTA

value of  depends upon the section at which the response is obtained. This is the case
because the angle of incidence of earthquake inuences the response in a complex manner by
modifying the separation length being considered in the correlation function. Furthermore, the
dierence between the maximum r.m.s. response and the r.m.s. response as obtained by the
stationary and non-stationary analyses varies with . The value of  for which this dierence
becomes a maximum depends upon the section at which the response is desired.

CONCLUSIONS

The seismic response of cable-stayed bridges to multi-component non-stationary partially cor-


related random ground motion is obtained using a Markov approach. A uniformly modulated
non-stationary model of the random ground motion is assumed and is specied by the evolu-
tionary r.m.s. ground acceleration. The analysis duly takes into account the spatial correlation
of ground motion, the cabletowerdeck interaction and the quasi-static excitation. Using the
proposed method of analysis, a cable-stayed bridge is analysed under a set of parametric vari-
ations in order to investigate the non-stationary characteristics of the bridge responses. The
results of the parametric study lead to the following conclusions.
1. The shape of the modulating function depicting the degree of non-stationarity signicantly
inuences the evolutionary r.m.s. displacement response. The eect of the non-stationarity
is to decrease the r.m.s. response.
2. Frequency domain spectral analysis (stationary analysis) provides higher r.m.s. responses
(both displacement and moment) compared to the maximum r.m.s. responses obtained
by the non-stationary analysis; the dierence could be as much as 45%.
3. The sharper the modulating function, the greater is the dierence between the maximum
displacement r.m.s. response (of the non-stationary analysis) and the stationary r.m.s.
displacement response.
4. Responses are amplied for the lter coecients corresponding to soft-soil conditions.
However, the dierence between the maximum r.m.s. responses (both displacement and
moment) (of the non-stationary analysis) and the stationary r.m.s. responses remains
nearly the same for all soil conditions.
5. Fully correlated ground motion provides a smaller value of the displacement response.
6. The maximum displacement response does not occur for zero angle of incidence of earth-
quake (i.e., the major component coinciding with the longitudinal axis of the bridge). The
critical angle of incidence depends upon the response quantity of interest. Furthermore,
the dierence between the maximum r.m.s. response (of the non-stationary analysis) and
the stationary r.m.s. response diers with the angle of incidence of earthquake.
7. The longitudinal component of ground motion considerably inuences the vertical
vibration of the bridge deck.

REFERENCES
1. Abdel-Ghaar AM, Rubin L. Suspension bridges response to multi support excitations. Journal of the
Engineering Mechanics Division (ASCE) 1982; 108:419 435.
2. Abdel-Ghaar AM, Rubin LI. Vertical seismic behaviour of suspension bridges. Earthquake Engineering and
Structural Dynamics 1983; 11:119.
3. Ali HM, Abdel-Ghaar AM. Modelling the nonlinear seismic behaviour of cable-stayed bridges with passive
control bearings. Computers and Structures 1995; 54:461 492.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393
SEISMIC RESPONSE OF A CABLE-STAYED BRIDGE DECK 393

4. Elassaly M, Ghali A, Elbadry MM. Inuence of soil conditions on the seismic behaviour of two cable-stayed
bridges. International Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts 1996;
33(5):231A.
5. Zheng J, Takeda T. Eects of soilstructure interaction on seismic response of PC cable-stayed bridge.
International Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts 1996; 33:233A.
6. Zheng J, Takeda T. Eects of soilstructure interaction on seismic response of PC cable-stayed bridge. Soil
Dynamics and Earthquake Engineering 1995; 14(6):427 437.
7. Villaverde R, Martin SC. Passive seismic control of cable-stayed bridges with damped resonant appendages.
International Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts 1995; 32(6):232A.
8. Allam SM, Datta TK. Seismic behaviour of cable-stayed bridges under multi-component random ground motion.
Engineering Structures 1999; 21(1):6274.
9. Park K-S, Jung H-J, Lee I-W. Hybrid control strategy for seismic protection of a benchmark cable-stayed bridge.
Engineering Structures 2003; 25(4):405 417.
10. Hyun CH, Yun CB, Lee DG. Nonstationary response analysis of a suspension bridge for multiple support
excitations. Probabilistic Engineering Mechanics 1992; 7:2735.
11. Lin YK. Nonstationary excitation and response in linear systems treated as sequences of random pulses. Journal
of the Acoustical Society of America 1965; 38(3):453 460.
12. Debchaudhury A, Gasparini DA. Response of MDOF systems to vector random excitation. Journal of the
Engineering Mechanics Division (ASCE) 1982; 108(4):367385.
13. Debchaudhury A, Gazis GD. Response of MDOF systems to multiple support seismic excitation. Journal of
the Engineering Mechanics Division (ASCE) 1988; 114(4):583 603.
14. Gasparini DA. Response of MDOF systems to nonstationary random excitation. Journal of the Engineering
Mechanics Division (ASCE) 1979; 105(2):13 27.
15. Gasparini DA, Shah A, Tsiatas G, Shein S, Sun W. Random vibration of cascaded secondary systems. Research
Report 8301, Department of Civil Engineering, Case Institute of Technology, Cleveland, Ohio, 1983.
16. Shinozuka M, Sato Y. Simulation of nonstationary random process. Journal of the Engineering Mechanics
Division (ASCE) 1967; 94(1):11 40.
17. Shinozuka M, Yang JN. Peak structural response to nonstationary random excitations. Journal of Sound and
Vibration 1971; 16:505 517.
18. Loh CH. Analysis of the spatial variation of seismic waves and ground movements from SMART-1 array data.
Earthquake Engineering and Structural Dynamics 1985; 13:561581.
19. Soliman HO, Datta TK. Seismic response of piping system to nonstationary random ground motion. Journal of
Sound and Vibration 1995; 180(3):463 473.
20. Clough RW, Penzien J. Dynamics of Structures, 2nd Edition. McGraw-Hill: Singapore, 1993.
21. Chatterjee PK, Datta TK, Surana CS. Vibration of cable stayed bridges under moving vehicles. Structural
Engineering International (IABSE) 1994; 4(2):116 122.
22. Allam SM. Seismic response of cable supported bridges. Ph.D. Thesis, submitted to Indian Institute of
Technology, New Delhi, 1998.
23. Nigam NC. Introduction to Random Vibrations. MIT Press: Cambridge, U.K., 1983.
24. Sinha PK. Multi-variable Control; an Introduction. Marcel Dekker Inc: New York, 1984.
25. Lin YK. Probabilistic Theory of Structural Dynamics. McGraw-Hill: New York, 1967.
26. Morris NF. Dynamic analysis of cable-stiened structures. Journal of the Structural Division (ASCE) 1974;
100:971981.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:375393