Beruflich Dokumente
Kultur Dokumente
SUMMARY
Two practical approaches, response spectrum and time-history methods, are developed to evaluate the response of
flexible multi-degree-of-freedom (MDF) systems, notably long-span bridges, to multiple-support seismic excitations. For
practical convenience, ground motions within a group of adjacent supports on continuous soil or rock are assumed to be
uniform and synchronized, while those of different groups are treated as non-uniform and uncorrelated. The response
spectrum analysis is extended to include the cross-correlation of modal responses, which prove important when closely
spaced modal frequencies exist. An example of the significance of multiple-support excitations is illustrated by
application to a suspension bridge. Qualitatively comparable effects can be expected for other bridges of similar type or
dimensions.
INTRODUCTION
For the seismic design of flexible and extended-in-plane structures like long-span bridges, a dynamic
response analysis involving multiple-support excitations is necessary, mainly owing to (i) phase-shifts in
ground motions along the spans and (ii) possibly different soil conditions at the supports which induce non-
uniform excitations. Also, the supports of a bridge across a deep gorge could be separately excited because of
the independent rocking motion of each cliff.
In past analyses,', the power spectral density functions of available strong-motion records were applied
to define the multiple-support inputs. However, such records taken with suitably spaced arrays are not
abundant. Therefore, ground-motion coherencies are usually unknown or unreliable for practical use. As a
more expedient approach, response spectrum methods for multiple-support uncorrelated excitation^^^^ have
been developed to estimate the maximum expected system responses without knowing ground-motion
correlations. The assumption of uncorrelated inputs is convenient indeed and possibly rather
conservative.'.' However, (i) the proposed methods are based on the simple SRSS (square-root-of-sum-of-
squares) rule of modal combinations, excluding sometimes important cross-correlations of modal responses
and (ii) their consistency with the time-history analysis has yet to be confirmed.
The present paper is intended to facilitate engineering applications by; (i) extending the response spectrum
methods for multiple-support inputs to include modal response cross-correlations?. and (ii) proposing a
time-history modal superposition analysis on the same basis as this extended spectral method. Here, ground
motions are assumed to be uniform as well as synchronized (i.e. no phase-shifts) within a group of closely
spaced supports on continuous soil or rock, such as those of a series of short side-spans in cable-stayed
bridges, for example. O n the other hand, ground motions of different groups are treated as non-uniform and
uncorrelated. In addition, a possible extension of the proposed methods to the problem of 3-dimensional
multiple-support excitations, which could be important in practice, is discussed. An illustrative example of a
suspension bridge shows the significance of multiple-support inputs and the effect of modal response cross-
correlations. The result also confirm the compatibility of the response spectral and time-history modal
superposition methods presented here.
EQUATION OF MOTION
The linearized equation of 3-dimensional motion of a multi-degree-of-freedom system is given as
[MI o(t)+ [C] O ( t ) + [K] U ( t ) = 0 (1)
where [MI is the mass matrix, [C] is the viscous damping and [K] is the stiffness matrix including
geometrical stiffness due to the initial (i.e. dead-load) axial forces, as is the case, for example, with suspension
bridges. U ( t ) is the 3-dimensional displacement vector, which is separated into two parts6v7
N
u(t)= { U j ( t ) } = C
n = l
4njqn(t) + CDjiI {zi(t)} (2)
where $nj is the nth mode shape, q n ( t )is the nth generalized coordinate and [Dji] is the matrix of quasi-static
functions, i.e. the matrix of the jth nodal displacement due to a unit ground displacement 6 = 1 (in the
direction of an excitation) at the ith support-group. (Here, a group is defined to be closely spaced supports on
continuous soil or rock, in which the excitation is assumed to be uniform.) zi(t) is the seismic ground
displacement at the ith support-group.
Pre-multiplying equation (1) by {4nj}T, a set of decoupled equations is obtained, as follows:
where Bin is the nth modal participation factor for the ground acceleration at the ith support-group. The
effective modal mass in the nth modal equation of motion, when the ith support-group is excited, is as follows:
Min Pin4,'jCMI {Dji }
= Bi',M,* = (10)
The ratio of the sum of !dinto the total mass of the system (effective mass ratio) is given by
and may provide a reasonable way of evaluating the sufficiency of the adopted number of modes N . For
example, the Honshu-Shikoku Bridge Authority specifies in its seismic design rules that R , for the case of
uniform excitation should be not less than 0.95.
MULTIPLE-SUPPORT SEISMIC EXCITATIONS 347
When the viscous damping in different parts of the system varies, as will be shown in the subsequent
example, the modal damping ratio h, of the nth mode in equation (8) can be obtained through a weighted
average, using the modal energy En, of the structural parts k (k = 1 , 2, . . . , K). Dropping the subscript j ,
where $im, t,hjn are the corresponding components of the mth and nth equivalent mode shapes and Gij(m)is the
cross-spectral density function of the input ground accelerations zi(t), z j ( t )at the ith and jth support-group.
Under the assumption that z j ( t )and zj(t)are uncorrelated, the matrix [ G i j ( w ) ]becomes diagonal and can be
written as
Fi(w) =
j: zi(t)exp ( - iwt)dt
where Tis defined approximately as the strong-motion duration of an earthquake.'.2 The mean-square value
of a response component is given, then, by the area under the spectral density function C,(w), as follows:
G,(w)dm =
m
CC
n i
$ i m $in 2% (21)
where I$,! is the mean-square contribution of the nth mode due to excitation at the ith support-group.
R 1c c
= [ i m n p m n ~ i m ~ i n ] 1 ’ 2= [ (Rim)’ [Pmnl {Rin} lv’ (29)
Equation (29) is an extension of the CQC method for the case of multiple-support excitations.
Another way of estimating the maximum expected response in engineering practice is the time-history
analysis, which is based on the above established principle that, according to equation (29), the maximum
expected response to the multiple-support-group (uncorrelated) inputs can be expressed as the square-root-
of-sum-of-squares of all the contributions from the ith (i = 1, 2, . . .) excitation. In that case, the maximum
expected response R(t),,, is simply given, as follows:
where Ri(t)maxis the maximum absolute response when the ith support-group is excited, Rin(t) = $i,,qin(t) is
the nth modal contribution to R i ( t ) and qin(t)is the nth generalized coordinate which can be obtained by
solving equation (8).
MULTIPLE-SUPPORT SEISMIC EXCITATIONS 349
MULTI-DIRECTIONAL EXCITATION
Three-dimensional seismic ground motions are known to have a set of orthogonal principal axes, two in the
horizontal directions and one in the vertical, along which the components are uncorrelated.* On this
assumption, Der Kiureghian and Smeby’ developed a response spectrum method for the case of uniform
excitation, and have shown that; (i) when the two horizontal components are not along the principal axes of a
structure (see Figure l), the effect of correlation between the components does appear, but is rather small, (ii)
if the two horizontal components have identical (or nearly identical) intensities, then the effect of correlation
disappears, and (iii) discarding this effect, the response to the 3-dimensional components can be simply
combined through the SRSS (square-root-of-sum-of-squares) rule.
The third conclusion above is most commonly applied in current practice. Based on this concept, the
analysis in this paper can be extended to the case of 3-dimensional multiple-support excitations, which is
important in engineering practice. First, the response spectrum approach is discussed, in which SL): (w,h),
S s ) (w,h) and S%)(w, h) are the maximum acceleration response spectra for a single-degree-of-freedom
oscillator for excitation in the X - , Y- and Z-directions at the ith support-group. Using equations (28) and (29),
i i
1
one obtains the system responses R,, R , and R , corresponding to S e ) (w,h), Sa:) (w, h) and S s ) (w, h),
respectively. Then, the total response is given by i
CR = [ R i + R: + R2]’”
(3 1)
Structure
2
kR1
- Rigid-member
In the same way, a time-history analysis is possible, where z$"'(t), Z j y ' ( t )and zi")(t)are the ground-motion
accelerations in X - , Y- and Z-directions at the ith support-group. By solving equation (8) and then using
1
equation (30), one obtains the responses Rx(t)maxrRY(t)maxand Rz(t),,, for zix)(t), z i Y ) ( t )and zi")(t),
i i
1
i
1
respectively. Then, the total response is
CR = CR,(t)iax + R y ( l ) i a x + Rz(t)iax11'2 (32)
. STRONG-MOTION OURATION
T(O.501 = 10.321 (SECI ,
0 5 10 15 20 25 30
T I M E (SEC)
Figure 3(a). Accelerogram for El Centro (1940) NS component (scaled to 180 gal peak acceleration)
MULTIPLE-SUPPORT SEISMIC EXCITATIONS 35 1
Figure 3(b). SDF response spectra of El Centro (1940) NS component ( A , = 180 gal)
The free vibration periods shown in Table I are rather closely spaced, so that the off-diagonal terms
pmn(rn # n) in equation (29) are surely important. Also, particular attention is needed for the 16th and 17th
modes which have close frequencies and similarly dominant tower motions (see Figure 5); hence the SRSS
rule of modal combination is apparently not reliable in this example. The effects of modal cross-correlations
are confirmed by the response moment diagrams in Figures 6 to 9, from which it is concluded that (i) the
differences in response moments estimated by equation (29) and by the SRSS rule are quite large for both the
stiffening truss and the tower, (ii) the moments predicted by equation (29) are fairly consistent with those
predicted by the time-history method of equation (30),even in the case of the impulsive San Fernando record.
Referring to Table I, it can be seen that the participation factors of the symmetric modes (n = 2, 4,6, . . . )
for the inputs at the support-groups (A) and (B) cancel out in the case of uniform ground motion (A + B).
352 N. YAMAMURA AND H. TANAKA
Participation factor
z
3
0
a
0
-180--
I 1 t
1 1
Figure qa). Accelerogram for San Fernando (1971) NS component (scaled to 180 gal peak acceleration)
MULTIPLE-SUPPORT SEISMIC EXCITATIONS 353
3000
2000
1000
500
40C
-
J
4
3oc
2 20c
4
v)
-
g 1oc
I-
<
a
w
_1
w 5c
0
2 4c
I
g 3t
z
0
0.
v) 2c
w
P
1(
.-I--
Figure 4(b). SDF response spectra of San Fernando (1971) NS component ( A G = 180 gal)
Thus, the contributions of symmetric modes to the responses almost disappear in the latter case, The effect is
shown in Figures 10 and 11, where the moments in the stiffening truss by multiple-support excitations are
much larger than those by uniform ground motion, notably in the case of the El Centro record; hence,
uniform excitation over the entire span is not a reasonable assumption. Meanwhile, the moments in the tower
(see Figures 12 and 13) by multiple-support and uniform excitations do not differ so significantly, indicating
that the tower moments are somehow less sensitive to symmetric vibrational modes. It should also be added
that the tower moments in Figures 12 and 13 are excited mainly by the input at support-group (A). Therefore,
independent excitations at all the supports, i.e. different inputs at Pts. 1 and 89, could have produced larger
differences.
354 N. YAMAMURA AND H. TANAKA
14 000 --
- --
eE 12 000
2
@@O
0
63
w @@@@ @@ @@@@@
Figure 6 . Response moments in the stiffening truss (El Centro 1940 NS component; multiple-support excitations)
Figure 7. Response moments in the stiffening truss (San Fernando 1971 NS component; multiple-support excitations)
MULTIPLE-SUPPORT SEISMIC EXCITATIONS 355
0 200 400
-
600
M (MN.rn)
800 1000 1200 1400
@
0 20 000 40 000 60 000 80 000 100 000 120 000 140 000
Figure 8. Response moments in the tower (El Centro 1940 NS component; multiple-support excitations)
- M (tf.rn1
Figure 9. Response moments in the tower (San Fernando 1971 NS component; multiple-support excitations)
356 N. YAMAMURA AND H. TANAKA
12 000- --120
i
Multi-support input
5 000 ____ Uniform excitation
1-11 1 lo
0 4 : I : I I I ; 1 I 1 I I I I ' 0
@@@@@@@@@@@@@@@
Figure 11. Response moments in the stiffening truss (San Fernando 1971 NS component; time-history method)
Multi-support input
Uniform excitation
0 10000 2OWO
- 30000
M (tf.m)
40oM) 50000
Figure 12. Response moments in the tower (El Centro 1940 NS component; time-history method)
MULTIPLE-SUPPORT SEISMIC EXCITATIONS 357
-Multi-support input
__-- Uniform excitation
Figure 13. Response moments in the tower (San Fernando 1971 NS component; time-history method)
CONCLUSIONS
A response spectrum method and a time-history approach for engineering practice are presented to facilitate
the analysis of flexible MDF systems subjected to multiple-support seismic excitations. The number of modes
necessary to represent reasonably the responses can be examined by equation (11). A system with non-
uniform damping ratios may be treated using equations (12) and (13). An example of a suspension bridge
confirmed that (i) the proposed response spectrum and time-history methods yield fairly consistent responses,
with 10-30 percent4.’ differences between them, (ii) the SRSS rule of modal combination discarding pmn
(m # n) in equation (29) can lead to gross errors and (iii) multiple-support excitations can play an important
role for the design of long-span bridges, notably in estimating the stresses in the stiffening girder or truss,
because they strongly excite both symmetric and antisymmetric vibration modes.
REFERENCES
I . A. M. Abdel-Ghaffar and Lawrence I. Rubin, ‘Suspension bridge response t o multiple-support excitations’, J . eng. mech. diu. ASCE
108, 419435 (1982).
2. A. M. Abdel-Ghaffar and Lawrence I. Rubin, ‘Vertical seismic behaviour of suspension bridges’, Earthquake eng. struct. dyn. I I, 1-19
(1983).
3. Shiro Aoyagi, ‘On the response of long-span suspension bridges subjected to earthquake different between substructures’, Proc.
J S C E . 190, 3 7 4 8 (1971) (in Japanese).
4. Armen Der Kiureghian, ‘Structural response to stationary excitation’, J . eng. mech. d i n ASCE. 106, 1195-1213 (1980).
5. Armen Der Kiureghian, ‘A response spectrum method for random vibration analysis of M D F systems’, Earthquake eng. struct. dyn.
9,419-435 (1981).
6. R. W. Clough and J. Penzien, Dpnumics qf Structures, McGraw-Hill, New York, 1975.
7. R. D. Mindlin and L. E. Goodman, ‘Beam vibrations with time-dependent boundary conditions’, J . uppl. mech. A S M E , 17,377-380
( 1950).
8. J. Penzien and M. Watabe, ‘Characteristics of 3-dimensional earthquake ground motion’, Earthquake eng. struct. dyn. 3, 365-374
(1975).
9. Armen Der Kiureghian and Wiggo Smeby, ‘Probabilistic response spectrum method for multidirectional seismic input’, Trans. 71h
SMIRT, section M , Aug. (1983).
10. K. Kawashima, K. Aizawa and K. Takahashi, ‘Duration of strong motion acceleration record’, Proc. struct. earthquake eng. diu.
J S C E , 2,161-168 (1985).