Engineering Structures, Vol. 19, No. 2, pp.

105-112, 1997

ELSEVIER

S0141-0296(96)00029-6

Copyright 1996 Elsevier Science Lid Printed in Great Britain. All rights reserved 0141~0296/97 $17.00 + 0.00

Torsional response of building structures to spatial random ground excitations

Hong Hao
School of Civil and Structural Engineering, Nanyang Technological University, Nanyang Ave, Singapore 639798 (Received July 1995; revised version accepted January 1996)

This paper applies the random vibration method to analyse the coupled lateral and torsional responses of symmetric and asymmetric building structures subjected to spatially varying horizontal multiple ground excitations. The responses of a single-storey building model with different eccentricities and varying numbers of supports are analysed. Ground excitations at the multiple structural supports are assumed to be zero mean spatial random processes. Their intensities and spatial variations are modelled by a spectral density function and an empirical coherency function. The standard deviations of various response quantities such as base shear, torque and relative column displacements are calculated. Numerical results are presented in terms of dimensionless parameters for their general application. Discussions on the torsional responses induced by both spatial ground excitations and building eccentricities are made. The adequacy of the torsional provisions of current seismic codes is also examined. Copyright 1996 Elsevier Science Ltd.

Keywords: multiple excitations, eccentricity, torsional response

1. Introduction
an asymmetric 7 rigid circular plate supported by columns on a rigid circular mat foundation that rests on soil. An empirical coherency function 1 was also used to model the ground motion spatial variations in torsional response analysis 8. In the later study 8, the structural model is an asymmetric rigid plate supported by four columns. The structural models in those studies 6-s are, however, very idealized and are the two extreme cases in terms of the multiple excitation effects. A rigid foundation directly resting on soil 6'7 is equivalent to a rigid plate supported by an infinite number of supports. Since the multiple excitation effects on structures will be reduced by increasing the number of supports due to the averaging effect on the spatially varying ground excitations, a structural model with infinite supports results in the minimum multiple excitation effects. On the other hand, a structural model with four columns 8 results in the maximum multiple excitation effects. In the present investigation, the responses of a singlestorey building model are analysed by the random vibration method. The same empirical coherency function 1 is used to model the ground motion spatial variations since it was based on the actual recorded earthquake motions. The standard deviations of the various response quantities such as base shear, torque and relative column displacements are

Torsional responses of building structures could be induced by structural eccentricities, and by spatially varying multiple ground excitations. Structural eccentricities-induced torsional responses have been intensively studied and are well understood, Torsional responses induced by torsional ground motions were first studied by Newmark ~. The work was then extended to include the combined horizontal, translational and torsional motions 2. In those two studies, the torsional ground motions were estimated by considering the phase shift of a propagating plane wave, but the loss of coherency between the spatial ground motions due to wave propagation was not considered. Recently, actual recorded spatial translational motions at the bases of over 30 buildings during recent earthquakes have been used to estimate torsional ground motions 3-s, the estimated torsional motions were then used to analyse the torsional responses of building structures. Torsional responses induced by spatially varying multiple support excitations have also been investigated 6-8. In the first two studies 6'7, the ground motion spatial variations were modelled by a theoretical coherency function 9, the structural models used in the analyses are a symmetric 6 or

105

106
calculated. The relative column displacements, which are related to the column shear forces, were not investigated in the previous studies 6 ~. The effect of the number and positions of the multiple supports on the torsional responses is also first investigated. Numerical results are presented and discussed in terms of the dimensionless parameters e/d and f~d/v,, for their general application, where e and d are, respectively, the stiffness eccentricity and building dimension in the direction normal to the multiple ground excitations; f, is the uncoupled lateral vibration frequency and v, is the apparent ground motion propagation velocity, d/v, is the time required for ground motion to propagate through the structure, and f,d/v, measures the phase difference between the dominant ground motion propagation phase shift and the uncoupled lateral vibration mode. Using the numerical results obtained in the study, the adequacy of the torsional provisions of the major current seismic codes~t ~4 are also examined. The objectives of the present study include: ( 1) to further understand the combined effect of structural eccentricity and spatial ground motions on the torsional responses of building structures; (2) to analyse the relative column displacements induced by spatial ground excitations; (3)to investigate the effect of the number and positions of the multiple supports on the torsional responses; and (4)to examine the adequacy of the torsional provisions of the major current seismic codes in accounting for the torsional responses of building structures subjected to spatial multiple support excitations. With these primary objectives, a few assumptions are made to simplify the analysis in the study. They are: ( 1) foundation soil is relatively firm so that the dynamic soil-structure interaction effect is neglected; (2) supporting columns rest directly on soil without any mat foundation; and (3)responses are small and remain in the linear elastic range.

Torsional response of building structures: H. Hao

of the diaphragm, but mass is only symmetric about the x-axis with an eccentricity e between CM and CS in the x-direction. The ground motion is assumed to propagate in the xdirection, thus the multiple excitations at the columns along the x-direction experience both coherency loss and phase shift, while those at the columns symmetric about the x-axis experience coherency loss only. Since the current seismic codes~ ~4allow one to consider the seismic forces in different directions separately, only ground excitations in the ydirection are considered. Thus, the structure has twodegrees-of-freedom, namely one lateral in the y-direction, and one torsional about a vertical axis through CM. There are, in total, N~ x N, support excitations; where N~ is the number of supports along the x-direction, and N,. is that along the y-direction. The equilibrium equations for both dynamic and total responses can be established by standard structural dynamic method 8. For the structural model shown in Figure 1, the mass and stiffness matrices can be obtained as

A(~ Nv

N~ Nv

EEk ,
i=1 j = l

gExik,
i=1 j ~ l Nx N

K~s =

Nx N

EEx,k, EE( , +
i=t/=] i=t/=t

(2)

K~b =

2.

Structural model and equations

-kl

I k 2

"'" - k 6 k6xi

"'"

-kN~N" I

k]]xn k l 2x2

kN~NX~/

(3)

The structural model considered is a single-storey building consisting of a rigid roof diaphragm supported by columns. All the masses are lumped at the centre of mass (CM). The lateral resistance of the structure is provided by vertical columns along the resisting planes in the x- and y-directions as shown in Figure 1, where kij is the lateral stiffness of the ijth column, in which ij denotes the column being the ith in the x-direction and jth in the y-direction. The lateral stiffness of the structure is assumed to be symmetric with the centre of stiffness (CS) located at the geometric centre

where m is the lumped mass, I is the polar moment inertia of the structure about a vertical axis through CM, xi and yj are the co-ordinates of column ij with origin at CM as shown in Figure 1, K,, is the stiffness matrix corresponding to the two structural degrees-of-freedom, and K,, is that corresponding to the coupled structural degrees and the support displacements. A similar approach to that presented in the previous study 8 can be used to calculate the base shear and torque.

/"

,,/'y

/~/

>~ /

Z.'//

////

Va

1)/I /

2/1/l/

...

Figure 1 Single-storey building structural model with Nx N v columns

Torsional response of building structures: H. Hao

For the structural model considered in this study, they are derived as
N Ny N~ Ny

107

I .lexp

(-irrUPt
\ v,, /

V= 2Zki;u
i=1 j=l N~ N

+ ZZxikqO
i-I j=l N x Ny

(4)

r:

EEx,
i=l j = l

.u +
i--I j = l

2+

Y~)koO

+
(5)

))exp

(10)

where u and 0 are dynamic lateral and torsional responses, respectively. The relative column displacements of the first row columns at the two edges away from and close to CM are derived as
Ul = u' + ff \/(d12
i

where ]Yiil is the coherency loss function and e x p ( - i ~ d ~ / v , ) measures the phase shift between ground motions at points i and j; dlj and d~j are the projected distances between points i and j on ground surface in ground motion propagation direction and its transverse direction, respectively; and 2"n'an b , ~ ~ + 2Ir + c n ' n = l ' 2 ' (ll)
m

%=

e) 2 + b2/2

Vii

(6) (7)

0.314 rad/s -- to --< 62.83 rad/s

__ m

U ~ = u' - 0 ~ ,v:(d/2 - e) 2 + b2/2 - VuA

where u' and 0' are the total lateral and torsional responses, v~ and VNA are the ground displacements at supports 11 and N~I, respectively. As can be noted, V and T are functions of the dynamic responses u and 0, while U~ and U G are functions of the total responses u' and if, respectively, implying that the relative column displacements depend on both the dynamic and the quasistatic responses.

3.

Spectral densities a n d analysis p r o c e d u r e

The cross-power spectral density matrix of the multiple ground excitations is ~,

&~( ito) = S~( to)

")#21

---

Y2(Nx+N,)

i
\'Y(N r + Nv )1

!
~I(N x + Ny) 2

i
...

] )
(8)

where G(to) =
( t o l2 __ ~ 2 ) 2 0)2 .4.

4~to2~2

1 + 4~ 2 2
to~ 1-toni + 4 g2~gz F (9)

when w > 62.83 rad/s, oq (o9) and O2(O)) become constants and equal their values at N = 62.83 rad/s, respectively. In this study, the apparent velocity v<, -- 500 m/s is assumed, and /3~ = 1.109 x 10-4, ~ 2 = 6.73 x 10-5, a~ = 3.853 x 10-3, bl = -1.811 x 10-5, cl = 1.177 x 10-~, a2 = 5.163 x 10-3, b2 = -7.583 x 10 -6, and c2 = -1.905 x 10-4 are used, which are values derived from processing the recorded motions during Event 45 at the SMART-1 array t6. In order to study the sensitivity of the coherency loss on torsional responses, the coherency loss corresponding to motions recorded during Event 30 at the SMART-1 array is also used j6. Figure 2 shows the coherency loss functions corresponding to dlj = d = 10m, 25 m and 5 0 m and d b - - 0 m. As can be seen, the motions recorded during Event 45 are more correlated than those recorded during Event 30. It should be noted that the two events chosen in the study represent the typical medium and large earthquakes in engineering design. The magnitudes for Events 30 and 45 are, respectively, ML = 6.3 and ML = 7.0, and the peak ground accelerations at the recording site are 0.084g and 0.25g. It should also be noted that the ground motion apparent velocity, which depends on the ground motion incident angle and the average shear wave velocity of the site, is usually unknown. The apparent velocity affects the spatial excitation phase shift. The assumption of v, = 500 m/s, however, does not affect the generality of the numerical results presented later since all the results are presented in terms of the dimensionless parameter f x d / v , . In this study, numerical results are calculated f o r f x d / v , ranging from 0.0 to 2.0, which covers most realistic engineering structures and site conditions, f~d/v,, = 0.0, 1.0 and 2.0 imply that the uncoupled lateral vibration mode and the dominant ground motion phase shift are in phase; and fxd/v,, = 0.5 and 1.5 imply that they are out-of-phase. In particular fxd/v,, = 0.0 implies uniform ground excitation since it is equivalent to / setting d = 0.0, thus dij __ d t0 __ 0.0, in equation (10). Using the random vibration method, the spectral densities of the base shear and torque are derived as

is the filtered spectral density of ground excitationsJS; in this study, tot = 1.636 rad/s, ~1 = 0.619, toe = 57r rad/s and ~u = 0.6 are used, these are typical values for a firm soil site; F is a scale factor depending on the ground motion intensity, and 70 is the coherency loss function given as ~

s~(g) =
i=t j=l /

108 1.0 0.8

>.

Torsional response of building structures: H. Hao

1.0 0.8
>.,

o0.6
L

o0.6
C @

=0.4 o ) 0.2 0.0 0

i

d=lOm d=25m d=50m Event 45

i i

~0.4 ...........
(D

0.2
Event 30
i

0.0 10

4 6 Freq(Hz)

4 6 Freq(mz)

10

Figure2

Typical coherency loss values between multiple excitations

2\/(d/2

e) 2 +

b2]2 real[S.,o,(iF)] b2/2 real[So,~N(iF)]

(15)

+ xikij
i=l j = l

2So(F)

+ 2\/(d/2 - e) 2 +

2(~/=~kij)(~;_j~x,kgj)real(S~o(iF))
(12)

where S.,(F), So'(-~) are the spectral densities of the tot_al l a t e r a l a n d torsional displacements, S.,o,(i-G), S.%,(ioJ), So% ~(iw), Sdvu.,( io), a n d Sotv%( io)) are the corresponding cross-spectral densities between the total responses and the ground displacements. The standard deviations of various response quantities can be obtained by integrating the corresponding spectral densities from ~ = 0 to ~ = ~. The closed form integration of the above functions is almost impossible, hence, numerical integration is carried out to calculate the various response standard deviations, which are the square root of the integrations. In numerical calculation, the structural dimensions in the x- and y-directions are assumed to be d = 5 0 m and b = 10 m, respectively; the damping ratios for both lateral and torsional modes are assumed to be 5%. Only two rows of columns in the y-direction (Nv = 2) are assumed, but two, three and four columns in the x-direction (N~ = 2, 3 and 4) are considered in order to study the effect of the number of columns on the torsional responses. The columns are arranged as follows: ( 1 ) four columns at the corners of the floor (N, = 2, N~ = 2); (2) four columns at the comers plus two columns located at the midspan in the x-direction (Nx= 3, Ny = 2 ) ; and ( 3 ) t w o rows of columns with four columns in each row uniformly spaced with a separation distance of 12.5 m (Nx = 4, Ny = 2). The base shear and displacement responses are normalized by the base shear Vo and displacement Uo, which are the responses of the corresponding symmetric structure (e = 0 . 0 ) produced by uniform excitation (Y0--1.0). The torque is normalized by To = Voea, where ea = ole + ~d is the design eccentricity specified by seismic codes; in which ae represents the stiffness eccentricity and /3d is a code specified accidental eccentricity. Different codes specify different a and/3 values, for example, Canadian ~ and Mexican t2 codes request c~ = 1.5, /3 = 0.1, but U B C t3 specifies a = 1.0 and/3 = 0.05, and the New Zealand code ~4 specifies c~ = 1.0 and /3---0.1. In this study, c~ = 1.5 and /3 = 0.1 is used.

and

( )2
N N

s(F)
i=l .i=l

Nx N v

X7 +

kis

5' o

IN. ,i/..
S u , ( F ) = S.,(F) + [(d/2 +

)]

real(S,0(iF))

where S,(F) andSo(F) are the spectral densities of dynamic lateral and torsional displacements, respectively, and S,o(ito) is their cross-spectral density. The spectral densities of the displacements Ut and UN~ are

e)2+ bZ/2]So,(F)

+ Sg(F) -

2real[Sj,,,,(iF)] +

2 1 ( d / 2 + e) 2 +

b212 real[S.,d(iF)] b2/2 real[S&,,(iF)] bz/2]Sd(F)

(14)

- 2 1 ( d / 2 + e) 2 +

S u ~ ( F ) = S . , ( F ) + [(d/2 - e) 2 + + Se(F) - 2 r e a l [ S . , ~ ( i F ) ] -

Torsional response of building structures: H. Hao

4. Reduction of the base shear

109

Base shear is reduced by torsional responses caused by either building eccentricity or multiple excitations. Figure 3 shows the normalized base shears for structures with N~ = 3, different eccentricities and different uncoupled torsional to lateral vibration frequency ratios, namely torsionally flexible with OJo/Wx=2/3, intermediate with Wo/O~ = 1.0, and torsionally stiff with Wo/Ogx= 3/2. As can be noted, when e/d = 0.0, the reduction of the base shear by multiple excitations is independent of the uncoupled frequency ratio. However, those of the asymmetric structure depend on the frequency ratio. The in-phase excitation effect is pronounced at f~d/v, = 0.0 and 2.0. It is not pronounced, however, at f , d / v , = 1.0. Similarly, the out-ofphase excitation effect is not pronounced atfxd/v, = 0.5 and 1.5. This is because of the two columns at the midspan of the structure. Since those midspan columns are d/2 away from the supports at both edges in the x-direction, when f~dlv, = 1.0, the excitations between the supports at the two edges in the x-direction are in-phase, but those between the midspan and the edge supports are out-of-phase. The inphase excitation effect is pronounced at f,d/v,,=2.0 because the excitations between any two columns in the xdirection are in-phase. The spatial multiple excitations generally reduce the base shear more than the structural eccentricity. The base shears are most significantly reduced when the structure is symmetric and the multiple excitations are out-of-phase. In other words, the multiple excitation effects are more pronounced for symmetric structures than for asymmetric structures.

5.

Torque and code provisions

Figure 4 shows the normalized torques for the structures with Nx= 3, and different eccentricities and frequency
ratios. The normalized torques depend strongly on the uncoupled frequency ratios. For uniform excitation (fxd/v, = 0.0), the normalization is always less than one, implying that the code specifications of c~ = 1.5 and/3 = 0.1 are adequate. When the structure is symmetric (e/cI = 0.0) and subjected to the uniform excitation, it has T/To < 0.5, implying that the accidental eccentricity specification of /3 = 0.05 is also adequate. T/To, however, increases rapidly as fxd/v, > 0.0. For a symmetric structure, the normalized torque is as high as 8
e

for the torsionally flexible structure, and 1.2 for the torsionally stiff structure. The peak torques are produced by the out-of-phase nonuniform excitations. The normalized torques of the asymmetric structures are smaller than the corresponding values of the symmetric ones. As shown in Figure 5, the normalized torque in fact decreases with the increase of the eccentricities, implying that the code specified equivalent static analysis of torsional response underestimates the torques of symmetric structures more than those of asymmetric structures if spatially varying multiple excitations are present. The above observations indicate that the code torsional provisions are adequate to account for torques induced by building eccentricity, but inadequate to account for torques induced by multiple ground excitations. The code provisions greatly underestimate torques in symmetric and torsionally flexible structures. The underestimation is less significant if the structures are torsionally stiff and have large eccentricities. For structures with frequency ratios co0/oJx --> 3/2 and e/d >--0.05, the code torsional provisions of ce = 1.5 and /3 = 0.1 are adequate. It should be noted that the above discussions on code torsional provisions are based on the results obtained from analysing an idealized structural model. Based on these results and those in the previous studies 3 8, it is still far from realistic to draw any solid conclusions from the code torsional provisions. However, the results do indicate the importance of the spatially varying ground excitations on building structures. Further studies are necessary before any code modification can be done. Those should include analysing some more practical building structures. The results presented above are comparable with those obtained in previous studies 7,8, but are higher than those obtained by Hahn and Liu 7 and lower than those by Hao and Duan 8 due to the different structural models used in the analyses as discussed above. For example, the peak accidental eccentricity, that is T/To when e/d=O.O, obtained by Hahn and Liu 7 is 0.36 at a frequency ratio WohO~=0.71, whereas it is 0.8 at o90/~0~=2/3 and 0.3 at Wo/O~ = 1.0 in this study as shown in Figure 4. To investigate the sensitivities of the coherency loss on the torsional responses, the base shear and torque of the symmetric structure with Wo/Ogx= 1.0 are also calculated by using the coherency loss of Event 30. Figure 6 shows the comparison with those obtained by using the coherency loss of Event 45. As can be seen, the coherency loss effect is
e /d=O. 1 1.0

/d =0.0

1.0-

wo/w=
\
0.8 --

1.0
1, ',k

wo/w== 2/3 . . . . . . . . . .
0.8 > >
O

o > >
0.6 --

0.6

0.4 0.0

0.4 0.0

0.5

1.0

1.5

2.0

0.5

1.0

1.5

2.0

/=ah,,

Figure3 Normalized

base shear as a function of dimensionless parameter

fxd/va

110
e/d:0.0 8 7 6
r

Torsional response o f building structures: H. Hao

e/d=0.1 5
/ i /

.--\ \

< p--

2
/ O F-F-/ /

;
/

wo/w, = 1.0 wo/w~ = a/2 wo/w= -- 2/3 . . . . . . . . .

C)

0.0

0.5

1.0

1.5

2.0

0.0

0.5

1.0

1.5

2.0

f ,al~.
Figure4 Normalized torque as a function of dimensionless parameter f~d/vo

hal~,

/d=0.0 e/d=O.05
e/d=O. 1 e/d=O.2
4-

/--

"x

.........

wo/w~ = 1.0
5-

F-

/
1 Event 5 0 Event 4 5

1-

wo/w~ =
0 0.0 ~ 0.5 j 1.0 Ad/~, j 1.5 i 2.0
0 0.0 I 0.5

1.0 ]

] 1.0

1.5

I 2.0

f ~d/v. Figure6 Comparison of base shear and torque for different coherency functions
These observations indicate that the column shear forces could be underestimated if one considers only the dynamic response when multiple support excitations are nonuniform. The underestimation is more significant for stiff structures than for flexible ones. Since the relative column displacements are dominated by the quasistatic responses, they are insensitive to the torsional to lateral frequency ratios.

Figure 5 Normalized torque corresponding to different eccentricities

prominent but insignificant compared to the phase shift effect. The high loss of coherency (Event 30) slightly reduces the phase shift effect. Nevertheless, the coherency loss between the multiple excitations should not be neglected. It could become the dominating multiple excitation effect on structural responses when the phase shift between the multiple excitations is small and when considering the total responses of stiff structures ~<~8.

7. 6. Displacements
The relative column displacements at the two edges of the structure behave very differently from the dynamic displacement induced by uniform excitation, Uo. As shown in Figure 7, the normalized displacements are close to unity at very smallf, d/v,,, but increase rapidly asfid/v, increases. This is because of the quasistatic displacements. As discussed above, the relative column displacements consist of dynamic and quasistatic displacements. At very small rid/v., implying very small f,, the total displacement is mainly produced by dynamic response. As f i increases, the dynamic response decreases due to the dynamic amplification effect, but the quasistatic response is independent of the structural vibration frequency and remains a constant for all f,d/v.. Thus, the relative column displacements are very close to unity at very smallf.d/v.; and increase rapidly withf~d/v, because the dynamic displacement U o decreases rapidly with the increase of f~.

Effect of the number of supports

The aforementioned results have demonstrated that torsional responses depend strongly on the phase differences between the dominating ground motion phase shift and the vibration mode. Thus, the number and positions of the structural supports affect the torsional responses since they change the phase differences of the multiple support excitations. To study the effect, structures with N~ = 2 and 4 are analysed. The normalized shear forces are shown in Figure 8. As can be seen, the phase shift effect of the multiple excitations is most pronounced in the structure with N~ = 2, where the normalized shear force oscillates asfid/v, increases, and reaches its minimum values when the multiple excitation phase shift and the lateral vibration mode are out-of-phase atfid/v, = 0.5 and 1.5, and the maximum values when they are in-phase at f,d/v, = 0.0, 1.0 and 2.0. The reduction in shear forces is most pronounced for a structure with N, = 4 since the dominating phase shift of the multiple ground motions at the four supports in the x-

Torsional response of building structures: H. Hao

10
si ,//

111
10=
/,/ //

e/d:O.O e/d=0.1

~/d=0.2
we/m= =
1.0
I llld I k I i i iiii[

10 -~ 0 -2

,,,,,i

,,,,,i

10 1

-~

1 0 -'

10 -2

1 0 -' /=d/v,

Figure 7 Normalized relative displacements at two edges of structure

e/d=O.O
1.0 --

e/d=0.2 Nx=2 Nx=5 Nx=4 1.0 ........... ..... 0.8

~,!

0.8

0.6

0.6

we~fox = 1 . 0

0.4 0.0 0.5 1.0 1.5 2.0

0.4

0.0

O.5

I .O

1.5

2.0

Aa/~a
Figure 8 Normalized base shear for structures with different numbers of supports

f=<llv.

direction and the lateral vibration mode are never completely in-phase in the range fxd/v,, = 0.0-2.0. Similarly, the normalized torques are smaller for structures with more supports due to the phase shift effect, as shown in Figure 9. As can be seen, the reduction in the torsional responses is substantial if more structural supports are present to prevent the complete out-of-phase excitations. For example, for the symmetric structure with wo/~ox= 1.0, the maximum values of the normalized torque are 4.3 and 2.6 for N~--2 and Nx = 4, respectively. By
e/d=O.O 5-

increasing the number of supports and by properly arranging the supports to prevent completely out-of-phase multiple excitations, torsional responses decrease.

8.

Conclusions

Torsional responses of symmetric and asymmetric structures subjected to spatially varying multiple ground excitations have been studied. It was found that torsional responses depend strongly on the uncoupled torsional to
e/d=0.2 2-

Nx=2 Nx=5

.........

0 32

o
,/

' "~ \ "

\'. '~,...

-"-"-

"-.

--~.. 1
b--

y
,'/
~o/~= = 1.0
0 I I I F
0 I I I I

0.(

0.5

1.0

1.5

2.0

0.0

0.5

1.0

1.5

2.0

f,d/~.
Figure9 Normalized torque for structures with different numbers of supports

f=d/v~

112 lateral vibration frequency ratios and the phase differences between the multiple ground motion phase shift and the vibration mode. Both structural asymmetry and multiple excitation reduce base shear and induce torque. The outof-phase multiple excitations significantly reduce base shear and induce torque. The multiple excitations produce more torsional responses than structural eccentricity if the structure is symmetric or asymmetric with small eccentricities. Relative column displacements depend on both dynamic and quasistatic responses, considering only dynamic responses greatly underestimates the relative column displacements of stiff structures. The number and positions of the multiple supports affect the phase difference between the multiple support motions. Properly arranging the structural supports to prevent completely out-of-phase excitations will substantially reduce the torsional responses. Based on the numerical results obtained in this study, it was found that the torsional provisions of a = 1.5 and 13 = 0.1 in current seismic codes are adequate for torsionally stiff and large eccentric asymmetric structures. They are, nevertheless, not adequate to account for the torsional responses of torsionally flexible structures induced by spatially varying ground excitations. The code specified equivalent static analysis of torsional responses underestimates the torques of torsionally flexible and symmetric or small eccentric asymmetric structures. A thorough investigation of the adequacy of the code torsional provisions for various practical structures and multiple excitations is needed.

Torsional response of building structures: H. Hao

torsion provisions from building records', J. Struet. Engng. ASCE 1994, 120(2), 597-616 4 De La Llera, J. C. and Chopra, A. K. 'Using accidental eccentricity in code-specified static and dynamic analyses of buildings', J. Earthquake Engng Struct. Dyn. 1994, 23, 947-967 5 De La Llera, J. C. and Chopra, A. K. 'Estimation of accidental torsion effects for seismic design of buildings'. J. Struct. Engng, ASCE 1995, 121(1 ), 102-114 6 Veletsos, A. S. and Prasad, A. M. 'Seismic interaction of structures and soils: stochastic approach', J. Struct. Engn~, ASCE 1989, 115(4), 935-956 7 Hahn, G. D. and Liu X. ~Torsional response of unsymmetric buildings to incoherent ground motions', J. Struet. Engng, ASCE 1994, 120(4), 1158-1181 8 Hao, H. and Duan, X. N. 'Seismic response of asymmetric structures to multiple ground motions'~ J. Struct. Engng. ASCE 1995, 121(11), 1557 1564 9 Uscinski, B. J. The elements ~ wave propagation in random media, McGraw-Hill, 1977, New York 10 Hao, H., Oliveira, C. S. and Penzien, J. 'Multiple-station ground motion processing and simulation based on SMART-1 array data', Nucl. Engng Des. 1989, 111, 293 310 11 Associate Committee on the National Building Code, NBCC, 'National building code of Canada 1990', National Research Council of Canada, 1990, Ottawa, Ontario 12 Gomez, R. and Garcia-Ranz, F. 'The Mexico earthquake of September 19, 1985 - complementary technical norms for earthquake resistant design, 1987 edition', Earthquake Spectra 1988, 4, 441-459 13 International Conference of Building Officials, 'Uniform building code', 199 I, Whittier, CA 14 Standards Association of New Zealand 'Code of practice for general structural design and design loadings for buildings: NZS 4203', 1992, Wellington, New Zealand 15 Ruiz, R. and Penzien, J. 'Probabilistic study of the behaviour of structures during earthquakes', Report EERC 69-03, 1969, Earthquake Engineering Research Center, University of California, Berkeley, CA 16 Hao, H. 'Effects of spatial variation of ground motions on large multiply-supported structures', Report UCB/EERC-89-07, 1989, Earthquake Engineering Research Center, University of California at Berkeley, CA 17 Zerva, A. 'Response of multi-span beams to spatially incoherent seismic ground motions', Earthquake Engng Struct. Dyn. 1990, 19, 819-832 18 Hao, H. 'Ground-motion spatial variation effects on circular arch responses', J. Engng Mech., ASCE 1994, 120(11 ), 2326-2341

References
1 Newmark, N. M. "Torsion in symmetrical buildings', Proc. 4th WCEE, Santiago, Chile, 1969, Vol. 2, pp. 19-32 2 Morgan, J. R., Hall, W. J. and Newmark, N. M. 'Seismic response arising from travelling waves', J. Struct. Engng, ASCE 1983, 109(4), 1010-1027 3 De La Llera, J. C. and Cfiopra, A. K. 'Evaluation of code accidental-