SEISMIC INTERACTION OF STRUCTURES AND S O I L S :

STOCHASTIC A P P R O A C H
By Anestis S. Veletsos, 1 M e m b e r , A S C E , a n d A i u m o l u M . P r a s a d , 2
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

Student M e m b e r , A S C E

ABSTRACT: A study of soil-structure interaction for seismically excited simple

structures is made considering both kinematic and inertial interaction effects. The
information and concepts presented elucidate the nature and relative importance of
the two effects and make it possible to assess readily the influences of the more
important parameters. The response quantities examined are the ensemble means
of the peak values of the lateral and torsional components of the foundation input
motion and of the associated structural deformations. The results are evaluated over
wide ranges of the parameters involved and are compared with those obtained for
no soil-structure interaction and for kinematic interaction only. Simple, physically
motivated interpretations are given for the observed differences. For the important
special case of vertically incident incoherent waves, simple closed-form approxi-
mate expressions are presented for the transfer functions of circular massless foun-
dations.

INTRODUCTION

In evaluating the response of structures to earthquakes, it is normally as-

sumed that all points of the ground surface beneath the foundation are ex-
cited synchronously and experience the same free-field motion (ATC 1978;
FEMA 1986; Veletsos 1977); the latter term refers to the motion which would
be induced at the foundation-soil interface if no structure were present. The
assumption of synchronous interface free-field ground motions is strictly valid
only for vertically propagating coherent wave fields; in reality, the motions
may vary from one point to the next (Abrahamson and Bolt 1985; Hari-
chandran and Vanmarcke 1986; Loh 1985; Yamahara 1970). Even when the
wave front is plane and propagates in a perfectly homogeneous medium, it
may impinge the foundation at a finite angle, leading to motions at neigh-
boring points which in the words of Kausel and Pais (1987) are "delayed
replicas" of each other. Known as the wave passage effect, the consequences
of such an action have been the subject of numerous previous studies (Bog-
danoff et al. 1965; Luco and Mita 1987a; Morgan et al. 1983; Newmark
1969; Roesset 1980; Scanlan 1976; Veletsos et al. 1976; Werner et al. 1979)
and are reasonably well understood.
Several additional factors contribute to the spatial variability of the free-
field ground motion. The individual wave trains may emanate from different
points of an extended source and may impinge the foundation at different
instants and with different angles of incidence, or they may propagate through
paths of different physical properties and may be affected differently in both
amplitude and phase by the characteristics of the travel paths and by reflec-
tions from, and diffractions around, the foundation. The spatial variability
of the ground motion due to these factors will be referred to as the ground
'Brown and Root Prof., Dept. of Civ. Engrg., Rice Univ., Houston, TX 77251.
2
Grad. Student, Dept. of Civ. Engrg., Rice Univ., Houston, TX.
Note. Discussion open until September 1, 1989. To extend the closing date one
month, a written request must be filed with the ASCE Manager of Journals. The
manuscript for this paper was submitted for review and possible publication on April
7, 1988. This paper is part of the Journal of Structural Engineering, Vol. 115, No.
4, April, 1989. ©ASCE, ISSN 0733-9445/89/0004-0935/S1.00 + $.15 per page.
Paper No. 23418.

935

J. Struct. Eng., 1989, 115(4): 935-956

 motion incoherence effect. This effect, which would exist even for horizon-
tally polarized vertically propagating shear waves, has been the subject of
only exploratory recent studies (Hoshiya and Ishii 1983; Luco and Mita 1987b;
Luco and Wong 1986; Matsushima 1977; Mita and Luco 1987; Novak and
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

Suen 1987; Pais and Kausel 1985).

The motion experienced by a rigid foundation is clearly different from the
free-field ground motion. The actual motion may conveniently be evaluated
in two steps. First, the so-called foundation input motion is computed; this
is defined as the motion which would be experienced by the foundation if
both it and the superimposed structure were massless. Computed with due
provision for the rigidity of the foundation, the foundation input motion in-
cludes both horizontal and torsional components even for a purely horizontal
free-field ground shaking. The difference in the responses of the structure
computed for the foundation input motion and the free-field motion at some
reference or control point of the ground surface is known as the kinematic
interaction effect. The greater the degree of ground motion incoherence or
the plan dimensions of the foundation in comparison to the length of the
dominant seismic waves, the more important this effect is likely to be.
The actual motion of the foundation is also influenced by its own inertia
and the inertia of the structure, and by the interaction or coupling between
them and the supporting soils. For a structure subjected to a purely horizontal
free-field ground shaking, not only are the horizontal and torsional com-
ponents of the actual foundation motion different from those of the corre-
sponding input motion, but the actual motion may also include rocking com-
ponents about horizontal axes. Contributed by the overturning tendency of
the superstructure, the latter components may be particularly prominent for
tall slender structures and for soft soils. These factors are provided for in
the second step of the evaluation process.
The term inertial interaction effect refers to the difference in structural
responses computed for the actual motion of the foundation and the foun-
dation input motion. The total soil structure interaction is clearly the sum of
the kinematic and inertial interaction effects.
Although the inertial interaction effects have been the subject of numerous
studies (FEMA 1986; Roesset 1980; Veletsos 1977, 1978; Veletsos and Meek
1976), they have generally been examined at the exclusion of the kinematic
interaction effects, and the interrelationship of the two effects has not been
adequately assessed. The objectives of this paper are: (1) To elucidate the
nature of both types of interaction for seismically excited simple structures;
(2) to assess the effects and relative importance of the numerous parameters
involved; and (3) to present information and concepts with which the effects
of the principal parameters may be evaluated readily. Primary emphasis is
placed on the kinematic interaction effects.
The structures investigated are presumed to have one lateral and one tor-
sional degree of freedom in their fixed-base condition and to be excited by
obliquely incident, horizontally polarized, incoherent shear waves. The tem-
poral variation of the free-field ground motion is expressed stochastically by
a local power spectral density (psd) function, and its spatial variability is
specified by a cross psd function. The response quantities examined include
the ensemble means of the peak values of the lateral and torsional compo-
nents of the foundation input motion and of the corresponding structural
deformations. These deformations are displayed in the form of pseudovel-
936

J. Struct. Eng., 1989, 115(4): 935-956

 ocity response spectra and compared, over wide ranges of the parameters
involved, with those obtained for no soil-structure interaction and for ki-
nematic interaction only. Simple, physically motivated interpretations are
given for the observed differences.
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

A fundamental step in the analysis of a structure-foundation-soil system

is the evaluation of the transfer functions of the foundation. Defined for
harmonically excited massless foundations, these functions relate the am-
plitudes of the horizontal and torsional components of foundation input mo-
tion to the amplitude of the free-field ground motion. The relevant functions
are evaluated herein by a relatively simple, approximate procedure, and their
accuracy is assessed through comparisons with available exact solutions for
special cases. In addition, simple closed-form expressions are presented for
these functions for the important special case of vertically incident, incoh-
erent waves.

SYSTEM CONSIDERED

The system investigated is shown in Fig. 1. It is a linear structure of mass

m and height h, which is supported through a foundation of mass mf at the
surface of a homogeneous elastic half-space. The circular natural frequencies
of lateral and torsional modes of vibration for the structure when fixed at
its base are denoted by px = 2trfx and pe = 2irfe, respectively, in which fx
and / e are the associated frequencies in cycles per unit of time; and the cor-
responding percentages of critical damping are denoted by t,x and £9, re-
spectively. The foundation mat is idealized as a rigid circular plate of neg-
ligible thickness and radius R which is bonded to the half-space so that no
uplifting or sliding can occur, and the columns of the structure are presumed
to be massless and axially inextensible. Both m and mf are assumed to be
uniformly distributed over identical circular areas. The supporting medium
is characterized by its mass density, p, shear wave velocity, v,, and Poisson's

(a) (b)

FIG. 1. System Considered: (a) Three-Dimensional View; (b) Top View of Foun-
dation
937

J. Struct. Eng., 1989, 115(4): 935-956

 ratio, v. This structure may be viewed either as the direct model of a single-
story building frame or, more generally, as the model of a multistory, mul-
timode structure that responds as a system with one lateral and one torsional
degrees of freedom in its fixed-base condition.
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

The free-field ground motion for all points of the foundation-soil interface
is considered to be a unidirectional excitation directed parallel to the hori-
zontal x r a x i s , as shown in Fig. 1, with the detailed histories of the motions
varying from point to point. Such motions may be induced by horizontally
polarized, incoherent shear waves propagating either vertically or at an ar-
bitrary angle with the vertical, a v . The intense portions of the motions are
represented by a stationary random process of limited duration, t0, and a
space-invariant, local psd function, Sg = Sg((t>), in which w = the circular
frequency of the motions. The spatial variability of the motions is defined
by a cross psd function, S(r1,r2,oi), in which r, and r 2 are the position vec-
tors for two arbitrary points.
A decreasing function of the frequency co and of the distance between the
two points, |ri - r 2 |, the function S f a . ^ . w ) is taken in the form suggested
by Harichandran and Vanmarcke (1986) as

/ d\~ d2\
S(r,,r 2 ,w) = rflr, - r 2 | , c o ) e x p -m Sg(a>) (1)

in which T, referred to as the incoherence function, is a dimensionless, de-

creasing function of |r t — r 2 |; i = V ~ T ; d\ and d2 = the components of ri
and r 2 in the direction of propagation of the wave front [see Fig. 1(b)]; and
c = the apparent horizontal velocity of the front. The latter quantity is related
to the angle of incidence of the waves, a„, by

(2)
sin <xv
The product of the exponential term in Eq. 1 and Sg represents the wave
passage effect, whereas the product TSg represents the effect of ground mo-
tion incoherence. The peak value of T is unity and occurs at rt = r 2 .
Several different expressions have been suggested for the incoherence
function (e.g., Harichandran and Vanmarcke 1986; Hoshiya and Ishii 1983;
Loh 1985; Luco and Wong 1986; and Mita and Luco 1987), and there is no
general agreement at this time on the form that may be the most appropriate
for realistic earthquakes. In this study, the single-parameter, second-order
function recommended by Mita and Luco (1987) is used:

7<o|r, - r 2 |
r(|r, - r2|,a>) = exp (3)

in which 7 is a dimensionless factor, taken between zero and 0.5.

A different approach to the study of this problem has been taken by Pais
and Kausel (1985). They have attributed the ground motion incoherence to
arrays of uncorrelated, obliquely incident waves arriving from different di-
rections within a sector of the supporting medium. The kinematic interaction
effects in this approach are represented by weighted averages of the com-
ponent wave passage effects.

938

J. Struct. Eng., 1989, 115(4): 935-956

 KINEMATIC INTERACTION EFFECTS

Spectral Characterization of Foundation Input Motion

Let Sx be the psd function of the horizontal component of the foundation
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

input displacement and Sy be the corresponding function for the circumfer-

ential or tangential displacement component along the periphery of the foun-
dation. Further, let S^ be the cross spectral density function for the com-
ponent displacements. Whereas Sx and Sy are real-valued, 5^, is generally
complex-valued.
These functions were evaluated from the cross spectral density function,
S(rur2,a>), by application of the averaging technique employed by Iguchi
(1984) and Scanlan (1976) in their studies of wave propagation effects. This
approach leads to

Sx = ~i\ S{Yur2MdAldA2 (4a)

"• JA JA

Sy = — I d,d1S{vuv2MdAldA2 (4b)
h JA JA

S
*y = 7 7 d2S(rl,r2,m)dA1dA2 (4c)
I«A JA JA
in which dAx and dA2 are elemental areas of the foundation; A = TTR2 = the
area of the foundation; and 7e = AR2/2 = its polar moment of inertia about
a vertical centroidal axis.
As is true of the corresponding exact expressions presented by Luco and
Mita (1987b), Eqs. 4 represent weighted averages of S(rur2,oi). However,
whereas the weighting functions in the exact formulation are the complex
distributions of the actual tractions at the foundation-soil interface, in the
procedure employed herein they are taken as linear functions. This is tan-
tamount to representing the restraining action of the supporting medium by
a series of mutually independent springs of the Winkler type (Scanlan 1976).
There are two main advantages to the use of the approximate formulation
over the exact formulation: (1) It reduces the number of independent param-
eters that must be considered, thereby simplifying the interpretation of the
results; and (2) for important special cases, it leads to simple, closed-form
expressions for the desired quantities. Additionally, the results are generally
of good accuracy.
For the circular foundations examined herein, it is convenient to express
T[ and r 2 in Eqs. 1 and 3 in terms of polar coordinates. On substituting Eq.
1 into Eq. 4, and making use of the appropriate coordinate transformation,
one obtains
r 1 /> 1 /•2'TT / - 2 T T

7 =— && exp(-tf A,) cos (c0A2) d^d^d^d^ (5a)

Sg TT J o Jo Jo Jo

7 =4 && exp(-tf A,) cos (c0A2)

\ IT Jo Jo Jo Jo

cos 9, cos 02 dQidd2d£,idl;2 (5b)

939

J. Struct. Eng., 1989, 115(4): 935-956

 =
"T — ' t\& exp (-b20Ai) sin (c0A:
Jg T Jo Jo Jo Jo

cos 62 d^dQidtid^ (5c)

Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

in which
A, = g + g - 25,62 cos (6, - 92) (6a)
A2 = 61 COS 0! - 62 COS 0 2 (6fc)

& and 62 are the radial distances of the two points normalized with respect
to the radius, R; Ql and 02 are the corresponding angular coordinates mea-
sured from the direction of wave propagation, as shown in Fig. 1(b); and
b„ and c0 are dimensionless parameters related to the well-known frequency
parameter, a0 = co7?/vs, as follows:
bD = ya0 (7)
and

c0 = - a0 = (sin av)a0 (8)

c
In the exact formulation of the problems presented in Luco and Mita (1987a,
1987b), the quantities 7, a0, and vjc appear independently.

Integration of Equations
For vertically incident incoherent waves, c0 = 0 and the interrelationship
of the free-field ground motion and the foundation input motion is defined
by the single parameter b0. Eqs. 5 in this case can be integrated exactly to
yield

Sx = - { \ - exp(-2fcj) [/o(2#) + 7,(2^)]} Ss (9a)

Sy = - {1 - exp(-2#) U0(2b2o) + 27,(2^) + I2(2b20)]} Ss (9b)

Sxy = 0 (9c)
in which I0, /,, and 72 are modified Bessel functions of the first kind of the
order indicated by the subscript. Eq. 9c indicates that the horizontal and
torsional components of the foundation input motion are statistically uncor-
rected. The derivation of these expressions is given in Appendix I.
For obliquely incident coherent waves, for which 7 = b0 = 0, the inter-
relationship of the two motions is defined completely by c0, and Eqs. 5 can
again be integrated exactly to yield
J c
i( o)~\ „
Sx = 2- r (10a)
Co

Sy = 4. Hco) Sg (106)
Co

0/tO

J. Struct. Eng., 1989, 115(4): 935-956

 ./,(c0),/"2(c0)
Sxy = i 8 _ ^ L 2 : 5s (10c)

in which J{ and / 2 are Bessel functions of the first kind of order one and
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

two, respectively. The latter expressions have been presented previously in

Pais and Kausel (1985). Note that S^ is purely imaginary, indicating that
there is a 90° phase angle in this case between the horizontal and torsional
components of foundation input motion.
For the more general case involving combinations of wave passage and
incoherence effects, formal integration of Eqs. 5 has not proved possible,
and the relevant expressions were integrated numerically.

Presentation of Results
The quantities y$JSg and \/Sy/Sg represent the transfer functions for the
amplitudes of the horizontal and rotational components of the foundation
input motion, and the magnitude of S ^ / V s ^ is a measure of the degree of
correlation or coherence of the component motions. A numerical value of
unity for the latter quantity indicates that the component motions are fully
correlated, while a zero value indicates that they are uncorrelated. These
quantities are plotted conveniently in Figs. 2 and 3 as functions of the mod-
ified frequency parameter
a0 = \/b\ + c20 = V f + sin2av a0 (11)
and the modified incoherence parameter

c0 sin a v
For incoherence effects only, a v = 0, 7 = °° and d0 reduces to ya0 = b„.
Similarly, for wave passage effects only, 7 = 7 = 0 and a0 reduces to
a0 sin a,, = c 0 .
Note that whereas the transfer function for the lateral component of the
foundation input motion, \/Sx/Sg, decreases monotonically in Fig. 2 with
increasing d0, the corresponding function for the torsional component,
V'Sy/Sg, increases from zero to a peak and then decreases monotonically.

Accuracy of Solutions
As a measure of the accuracy of the reported data, the results computed
for incoherence effects only and for wave passage effects only are compared
in Fig. 4 with the corresponding exact solutions of Luco and Mita (1987a,
1987b). Since the factors 7 and sin a v = vjc appear independently in the
exact solutions, several different values are considered for these parameters.
No comparisons are made for combinations of incoherence and wave passage
as the exact solutions are not available in this case.
Considering the uncertainties that are inherent in the definition of the in-
coherence function and in the choice of the parameter 7, the degree of agree-
ment in the two sets of results displayed in Fig. 4 is deemed to be quite
satisfactory. Note should also be taken of the fact that, excepting the narrow
frequency ranges where the curves for wave passage only exhibit notch-like
trends, the approximate solutions overestimate the amplitudes of foundation
input motions.

941

J. Struct. Eng., 1989, 115(4): 935-956

Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

FIG. 2. Magnitudes of Transfer Functions between Free-Field Ground Motion and

Foundation Input Motions

Other Meanings for Results

Although defined specifically for the displacement histories of the foun-
dation input motion, the spectral density ratios Sx/Ss, Sy/Sg, and S^/Sg also
define the ratios s//Sg, S^/Sg, S$/Sg and S£g, $y/Sg, $$/$£> of the corre-
sponding velocity and acceleration histories. Recall that the psd function for
the first derivative of a process is given by the product of (2TT/) 2 and the
psd function of the original process.

Spectral Characterization of Structural Response

With the psd functions of the foundation input motion established, the
corresponding functions of the structural response can be obtained by well-
established procedures [e.g., Lin (1976)]. Let S„ be the psd function of the
structural defomiation, u, induced by the lateral component of the foundation
input motion; and let Sv be the corresponding function of the deformation,
0,An

J. Struct. Eng., 1989, 115(4): 935-956

Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

FIG. 3. Normalized Cross PSD Function for Horiiontal and Torsional Compo-
nents of Foundation Input Motion

Present Solution
Luco's Exact
Solution

£ = 0.3

0.5

-j i i

FIG. 4. Comparison of Approximate and Exact Magnitudes of Foundation Trans-

fer Functions: (a) Incoherence Only, vjc = 0; (b) Wave Passage Only 7 = 0

943

J. Struct. Eng., 1989, 115(4): 935-956

 v = fyR, induced along the perimeter of the structure by the torsional com-
ponent of response. The quantity vj; represents the angular deformation of
the structure. These functions are related to the psd functions of the foun-
dation input accelerations, Ss and Sy-, by
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

Su = fc|2^ (13)
and
Sv = \Hv\2Si: (14)
in which Hu = the transfer function for lateral response, given by
1 1
H. = —2 1 (15)
pi co co
Px Px
Hv = the corresponding function for torsional response, obtained from Eq.
15 by replacing/?, by pe and ^ by £e; and vertical bars indicate the modulus
of the enclosed quantity. Similarly, the psd function S„ for the total defor-
mation at the most highly stressed point on the periphery of the structure,
w = u + v, is given by (Lin 1976)
Sw = S„ + S„ + 2 |Re(tf„fl?%)| (16)
in which S^ = the cross psd function of the lateral and circumferential com-
ponents of the foundation input accelerations; Re denotes the real part of the
indicated quantity; and a star superscript denotes the complex conjugate of
the quantity to which it is attached.

Characterization of Free-Field Earthquake Ground Motions

The local psd function for the set of acceleration traces considered in the
remainder of this paper is taken in the form
v _ f [ / 7 ( 0 . 5 +/ 4 )][1 - (/V/o)]5 0 f o r / < / 0 n ~
Si (17)
~{ 0 for/>/0
in which S0 = a constant; / = CO/2IT = the exciting frequency in cycles/
sec; and f0 = the cut-off frequency, taken as 15 cycles/sec. Of the same
general form as that employed in a related study by Pais and Kausel (1985),
the function Sg, along with the associated functions for ground velocity and
ground displacement, are plotted in Fig. 5, with all peaks normalized to a
unit value. As would be expected, the psd function for velocity decays much
more rapidly with frequency than that for acceleration, and the correspond-
ing displacement function decays even faster.
Let Xg be the mean of the absolute maximum peaks of the acceleration
traces, and Xg and Xe be the corresponding means of the velocity and dis-
placement traces. These values were computed from Der Kieureghian's (1980)
empirical expressions summarized in Appendix II, considering the duration
of the intense portion of the excitation to be t0 = 20 sec. The resulting values
a r e i g = 26.17V^, Xe = 1.417V^, and Xg = 0.2468\/^.
Foundation Input Motion
Before examining the response of the structure, it is desirable to compute
the mean peak values of the acceleration, velocity, and displacement traces
944

J. Struct. Eng., 1989, 115(4): 935-956

Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

FIG. 5. Normalized PSD Functions for Free-Field Ground Motions Considered

of the horizontal and circumferential components of the foundation input

motion. The relevant values for the horizontal component of motion are de-
noted by X, X, and X, and those for the circumferential component along
the periphery of the foundation are denoted by Y, Y, and Y. Computed by
Der Kieureghian's approximation from the appropriate psd functions, these
values are plotted in Fig. 6 normalized with respect to the mean peak values
of the corresponding histories of the free-field ground motion.
For the multifrequency, transient excitation considered in this section, the
solution is controlled by the effective transit time,

T = V 7 2 + sin2av T (18)

in which T = R/vs = the time required for the shear wave to traverse the
radius of the foundation; and by the modified incoherence parameter, y,
defined by Eq. 12.
The following observations may be made and inferences drawn from the
data presented in Fig. 6:

1. The reduction in the horizontal component of the foundation input motion

and the corresponding increase in the rotational component are greatest for ac-
celeration, much smaller for velocity, and almost negligible for displacement.
Since the foundation filters the high-frequency wave components more effec-
tively than the low-frequency wave components, the acceleration traces of the
ground motion, which are richer in high-frequency content than the velocity and
displacement traces, are influenced more than the latter traces.
2. Considering that the response of high-frequency systems is acceleration-
sensitive, whereas that of low-frequency systems is displacement-sensitive, it
should be clear that the effects of kinematic interaction would be important for
high-frequency systems and inconsequential for low-frequency systems. Fur-
thermore, medium-frequency systems which are velocity-sensitive would be ex-

945

J. Struct. Eng., 1989, 115(4): 935-956

Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

0.02 0.04 0.06 0.08 0.1

Effective Transit Time, T , sec

FIG. 6. Normalized Mean Peak Values of Lateral and Torsional Components of

Foundation Input Motions

pected to be affected moderately. That this is indeed the case is confirmed by

the data presented in the following sections.

Effects of Ground Motion Incoherence on Structural Response

Let Ux = the mean of the maximum values of the structural deformations
induced by the ensemble of lateral components of the foundation input mo-
tions, and Uy = the corresponding mean of the deformations induced at the
periphery of the deck by the torsional components. These quantities have
been evaluated for vertically propagating incoherent shear waves (7 = »),
and the results are displayed in Fig. 7 in the form of tripartite response
spectra. The solid curves in the upper part of the figure refer to lateral re-
sponse, and the lower curves refer to torsional response. Several values of
946

J. Struct. Eng., 1989, 115(4): 935-956

 5 r
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

An

Dashed lines are for

''h^V combination of horizontal
and torsional responses

0*02 -LLLUJ 1 i_j_Luul i i i I mil i ' ' I'fiiil

I * 10
f x , cps

T = 0.02 sec

0.02 0.1

FIG. 7. Effects of Ground Motion Incoherence on Maximum Deformations of

Structures with J, = £e = 0.02

the effective transit time parameter, f, are considered, including the limiting
value of f = 0 for which there is no kinematic interaction. The damping
factors for both modes of response are taken as £, = £e = 0.02.
The left-hand diagonal scale in the upper part of Fig. 7 represents Ux
normalized with respect to the mean peak value of the free-field displace-
ment, Xg; the vertical scale represents the corresponding pseudovelocity, Vx
= pxUx, normalized with respect to Xg; and the right-hand diagonal scale
represents the corresponding pseudoacceleration, Ax = pxVx, normalized with
respect to Xg. In an analogous manner, the three scales in the lower part of
the figure represent the deformation ratio, Uy/Xg; the pseudovelocity ratio,
Vy/Xg, in which Vy = peUy; and the pseudoacceleration ratio, Ay/Xg, in which
A
y = PoVy = plUy.
As anticipated from examination of the peak values of the foundation mo-
tions, the lateral component of the response of high-frequency systems in
Fig. 7 is affected materially by ground incoherence, and this effect is par-
ticularly large in the practically important region of the response spectrum
947

J. Struct. Eng., 1989, 115(4): 935-956

 within which the pseudoacceleration attains its maximum value. For 7 = 0.2
and f = 0.02 sec (a value corresponding to, say, R = 100 ft and vs = 1,000
ft/sec), the maximum value of Ax is 78% of that obtained for a fully co-
herent, uniform free-field ground motion; for f = 0.05 sec, the correspond-
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

ing ratio is 55%. The reductions are significantly less pronounced for me-
dium-frequency systems and practically negligible for low-frequency systems.
For systems of very high frequency, for which Ax may be considered to be
equal to the mean peak value of the foundation input acceleration, the per-
centage reductions are, of course, identical to those indicated in Fig. 6 for
the foundation input acceleration.
The general trends of the response spectra for the torsional deformation
in Fig. 7 are consistent with those of the corresponding curves for the foun-
dation input motion presented in Fig. 6. Specifically, in the low-frequency,
displacement-sensitive region, the response increases with increasing values
of the effective transit time, f, whereas in the high-frequency, acceleration-
sensitive region, the response values for f = 0.02 sec are higher than those
for the higher values of f considered. Furthermore, the percentage changes
in response are comparable to those for the controlling values of the foun-
dation input motion.
The component of the response contributed by the rotation of the foun-
dation is generally small, and the combined effect of lateral and torsional
responses is generally only slightly greater than that due solely to lateral
response. The mean maximum values of the total deformation for the most
highly stressed column along the periphery of the structure were evaluated
considering p9/px = 1.5, and the results are shown by the dashed lines in
Fig. 7. These results were computed by Der Kieureghian's approximation
making use of Eq. 16 for the psd function of the combined motion.

Comparison of Incoherence and Wave Passage Effects

Some of the response spectra for the incoherent ground motions presented
in Fig. 7 are compared in Fig. 8 with those computed considering: (1) Wave
passage effects only; and (2) combinations of wave passage and incoherence
effects represented by a value of 7 = 1. It should be clear that the results
are not particularly sensitive to the choice of the parameter 7, and that this
insensitivity is fully compatible with that observed in Fig. 6 for the peak
values of the foundation input motions. Indeed, the ratio of the low-fre-
quency limiting values of Ux for 7 = 0 and 7 = 00 i n Fig. 8 is almost
identical to that of the peak values of the lateral component of the foundation
input displacements in Fig. 6, and the ratio of the corresponding values of
Vy is almost identical to the displacement ratio of the torsional component
of the foundation input motion. Similarly, the ratios of the high-frequency
limits of Ax and Ay in Fig. 8 are identical to those obtained from Fig. 6 for
the mean peak values of the lateral and torsional components of the foun-
dation input accelerations. It follows that, to the degree of approximation
represented by the differences in the results displayed in Fig. 8, the effects
of ground motion incoherence may be replaced by those of wave passage,
and vice versa. This possibility has also been suggested by Luco and Wong
(1986) from examination of the relevant foundation transfer functions. In
implementing this replacement, it is important that the value of f be the same
in the two cases.

048

J. Struct. Eng., 1989, 115(4): 935-956

 INERTIAL INTERACTION EFFECTS

The inertial interaction effects are now evaluated by a simple modification

of the procedure used in previous studies in which the effects of kinematic
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

interaction were neglected [e.g., Veletsos (1977); Veletsos and Meek (1976)].
For each mode of excitation, it is only necessary to replace the free-field
motion by the appropriate component of the foundation input motion.
The following steps are involved in the analysis. First, the harmonic re-
sponse of the system is evaluated, making use of the appropriate complex-
valued foundation impedance functions. Next, the psd functions of the tor-
sional and lateral components of structural response are determined. The
desired mean peak values of the responses are then computed from Der Kiur-
eghian's approximation.

0.05tJ-u-iJ 1—i—i I i n i l i i i l 11 H I i '• i

0 . 0 5 0.1 I io 50
fx, cps
21 ,

0.02 ' ' ' ' , l ' ' I I l l ill i i ' I i i ' il ' i i
0 . 0 5 0.1 I 10 50
fg. cps

FIG. 8. Effects of Ground Motion Incoherence and Wave Passage on Maximum

Deformations of Structures with L = £« = 0.02

949

J. Struct. Eng., 1989, 115(4): 935-956

 The foundation impedances for the torsional mode of vibration were com-
puted from the approximate closed-form expressions of Veletsos and Nair
(1974), and those for the horizontal and rocking motions were computed
from the corresponding expressions of Veletsos and Verbic (1973). The cross
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

coupling terms between horizontal and rocking motions were presumed to

be negligible.
The principal parameters that influence the response of the system are the
characteristics of the free-field ground motion; the fixed-base natural fre-
quencies of the structure, fx and/ B , and the associated damping factors, ^
and £e; the height to base radius ratio, h/R; the mass density ratio for the
structure, defined conveniently as 8 = m/(TTpR2h), in which the denominator
represents the total mass of the structure when filled with the supporting
soil; and the wave transit times, T and f. It is important to note that whereas
the kinematic interaction effects are defined completely by f, the evaluation
of the inertial interaction effects requires the separate specification of the
parameters 7 and T. Other parameters affecting the response of the system
are Poisson's ratio for the supporting medium, v; the mass ratio of the foun-
dation and superstructure, ms/rn; the ratio //// of the mass moments of inertia
of the foundation and structure about horizontal centroidal axes; and the ratio
Jf/J of the corresponding polar moments of inertia. For the solutions pre-
sented herein, L = £* = 0.02, 8 = 0.15, v = 1/3, and mf (and hence If
and Jf) are considered to be negligible.

Results for Vertically Propagating Incoherent Waves

Fig. 9 shows response spectra for lateral and torsional responses obtained
for vertically propagating incoherent waves, taking 7 = 0.4 and T = R/vs
= 0.05 sec. Three sets of solutions are presented: (1) Making no provision
for soil-structure interaction, i.e., considering the foundation motion to be
equal to the free-field ground motion; (2) providing only for the kinematic
interaction effects, i.e., using as base excitation the foundation input mo-
tions; and (3) providing for both kinematic and inertial interaction effects,
i.e., analyzing the structure-foundation-soil system exactly as a coupled sys-
tem. In the analysis of the inertial interaction effects, two values of h/R are
used: a unit value, corresponding to short stubby structures, and a value of
3, corresponding to taller, more slender structures. Solutions (1) and (2) are
independent of h/R, and solutions (2) are valid for all combinations of 7
and T for which 7T = f = 0.02 sec.
Previous studies of soil-structure interaction involving only inertial inter-
action effects (ATC 1978; FEMA 1986; Veletsos 1977, 1978) have shown
that these effects may be evaluated to a high degree of approximation using
the free-field ground motion as the foundation input motion and merely mod-
ifying the relevant natural frequency and damping of the structure. The mod-
ified frequency and damping are taken such that, for each mode of vibration,
the magnitude and location of the resonant peak of the relevant harmonic
response are identical for the actual and replacement systems. For structures
for which the kinematic interaction effects are important, this approach would
require that the response of the structure be evaluated for the horizontal and
torsional components of the foundation input motion rather than for the free-
field ground motion.
The mean maximum values of the responses obtained by this approximate
procedure are shown by the dashed lines in Fig. 9, and the values of the
950

J. Struct. Eng., 1989, 115(4): 935-956

 7=0.4
T = 0 . 0 5 sec
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

I-
Xg

Approximate Solution
0.1
for Total SSI

0.05 XJLLlJL J 1 Liiml. J L-LLUJ-L

0.05 0.1 10 50
f x , cps

I
V,

*3

o.i

0.021 I I I I ! J ' I I I III

0.05 0.1

FIG. 9. Comparison of Effects of Kinematic and inertia! interaction on Maximum

Deformations of Structures with £, = £e = 0.02

modified natural frequencies and damping factors are identified in Fig. 10.
Denoted with a tilda superscript, the modified frequencies are, of course,
lower than the corresponding fixed-base frequencies, and the modified damping
factors are higher than the value of t,x = £e = 0.02 assumed for the fixed-
base structure.
The following trends should be observed in these figures:
1. Like kinematic interaction (KI), inertial interaction (II) may affect signif-
icantly the responses of systems in the medium- and high-frequency spectral
regions.
2. The II effects are generally more important than the KI effects.
3. Unlike kinematic interaction, which generally reduces the lateral response,
951

J. Struct. Eng., 1989, 115(4): 935-956

Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

ffl. cps

FIG. 10. Natural Frequencies and Damping of Modified Systems in Approximate

Analysis of Inertial Interaction Effects

inertial interaction may increase the corresponding response of tall, slender struc-
tures in the high-frequency region of the response spectrum. Such structures,
however, typically fall in the middle-frequency region of the spectrum, for which
the interaction effects are relatively small.
4. The II effects for low-frequency, highly compliant structures are negligible
because such systems "see" the half-space as a very stiff, effectively rigid me-
dium.
5. Provided the base excitation for the structure is taken equal to the foun-
dation input motion rather than the free-field ground motion, the concept of mod-
ifying the fixed-base natural frequencies and associated damping values of the
system provides a simple and highly reliable practical means for assessing the
II effects.
ORO

J. Struct. Eng., 1989, 115(4): 935-956

 It may be surprising that the values of/ e and £e in Fig. 10 are functions
of the ratio h/R. This is due to the fact that with the value of the mass ratio,
8, fixed in these solutions, the polar mass moment of inertia of the system,
/ , is different for different values of h/R.
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

CONCLUSIONS

1. The information and concepts presented herein provide valuable insight into
the nature of kinematic and inertial interaction effects for simple surface-sup-
ported structures subjected to earthquakes, and into the effects and relative im-
portance of the numerous parameters involved.
2. In the approximate method of analysis employed, the kinematic interaction
effects are defined completely by the effective transit time, f, and the modified
incoherence parameter, 7.
3. Even for vertically propagating waves, kinematic interaction may reduce
significantly the critical responses of high-frequency systems. These reductions
are generally smaller than those due to inertial interaction.
4. Reliable estimates of the effects of kinematic interaction on the peak values
of structural response may be obtained from knowledge of the corresponding
values of the acceleration, velocity, and displacement traces of the foundation
input motion. The latter quantities may be computed from analyses of the re-
sponse of the massless foundation to the free-field ground motion.
5. Insofar as the mean maximum values of the responses are concerned, the
kinematic interaction effects due to ground motion incoherence are similar to
those due to wave passage, and the two effects may be interrelated.
6. An excellent approximation to the inertial interaction effects may be ob-
tained by a previously recommended simple procedure (ATC 1978; FEMA 1986;
Veletsos 1977, 1978) using as base excitation the foundation input motion rather
than the free-field motion. The inertial interaction effects in this approach are
expressed by changes in the natural frequency of vibration and the associated
damping of the structure for the mode of vibration considered.

ACKNOWLEDGMENT

This study was supported by Grants 86-2034 and 87-1314 awarded to Rice
University by the National Center for Earthquake Engineering Research, State
University of New York at Buffalo. This support is appreciated greatly. Ap- •
preciation also is expressed to Yu Tang, Post-Doctoral Research Associate
at Rice, and K. Dotson and P. Malhotra, graduate students at Rice, for read-
ing an earlier draft of the manuscript and offering valuable comments.

APPENDIX I. DERIVATION OF EQS. 9

For incoherence effects only, the integrands in Eqs. 5a and 5b are sym-
metric about 4i = iz- This symmetry may be provided for by multiplying
these expressions by 2 and changing the upper limit of integration of £j2 from
unity to £,. On using the identity

953

J. Struct. Eng., 1989, 115(4): 935-956

 /„(z) = "(cos nQ)dd (19)
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

given as equation 9.6.19 in Abromowitz and Stegun (1970), the specialized

forms of Eqs. 5a and 5b are integrated with respect to the circumferential
coordinates to yield

g,& exp[-tf(g + m /o(2fc&&) «&<*& (20a)

o Jo
1
rei
7 = 16 I I (€i&)2 exp[-^(£ + g)] /I(2fcgg1g2) «&«& (206)
\ Jo Jo
The dummy variable £2 in these equations is then expressed as £j2 = s(ji. and
the resulting expressions are integrated with respect to s by making use of
the identity [see equation 6.631.8 in Gradshteyn and Ryzhik (1980)]

s"+l exp(-as 2 ) I„(2as)ds = — 2 #2«) (21)

4a
to yield

j = ^ | 2g, [1 - exp(-2#£) 70(26^?)] rf|, (22a)

7 =^1 £ {1 - exp(-2#£) [7o(2^£?) + 2/,(2tf £)]} dg, (226)

Finally, on letting a = 2b20£{ and making use of the identities

e-"a"In{a)da = — — [7„(2) + /„+1(z)] (23)

o 2n+l

a"I„^(a)da = z"I„(z) (24)

70(z) - h(z) = - 7,(7). (25)

given as equations 11.3.12, 11.3.25, and 9.6.26 in Abromowitz and Stegun

(1970), Eqs. 22a and 22b are integrated to yield Eqs. 9a and 9b. Eq. 9c
follows from the fact that the integrand in Eq. 5c is antisymmetric in this
case.
For small values of ba, application of Taylor's series expansion to Eqs.
9a and 9b yields

Sx = I 1 - bl + - b\ + (26a)

1 2
5V = | - bl - - bl + • • • (266)
V2 3

J. Struct. Eng., 1989, 115(4): 935-956

 APPENDIX ii. EVALUATION OF PEAK VALUES OF INPUT
AND RESPONSE

Let z(t) be a stationary, ergodic random Gaussian process with zero mean
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

and limited duration, t0, and let Z be the ensemble mean of its peak values.
Further, let G(u>) be the one-sided power spectral density of the process, and
X0, ^-l. and X2 be its first three moments, defined by

'G((a)d<a n = 0, 1, 2 (27)
Jo
The value of Z in Der Kiureghian's approach is evaluated conservatively
from
0.5772
Z = V 2 In (|xef„) + X0 (28)
V2 In 0 I A ) J

in which
2.1 or 2q\u0 if greater than 2.1 for q £ 0.1
|V 0 = \ (1.63 9 0 - 45 - 0.38)|xf„ for 0.1 s q < 0.69 (29)
\xt0 for q > 0.69
(x = ( I / T T Q V ^ / X Q = the mean zero-crossing rate of the process; and q =
V l ~ X*/(X0X2) = Vanmarcke's (1975) bandwidth parameter.

APPENDIX III. REFERENCES

Abrahamson, N. A., and Bolt, B. A. (1985). "The spatial variation of the phasing
of seismic strong ground motion." Bull. Seismological Soc. of Am. 75(5), 1247-
1264.
Abromowitz, M., and Stegun, I. A. (1970). Handbook of mathematical functions.
Dover Publications, Inc., New York, N.Y.
Appl. Tech. Council (ATC). (1978). "Tentative provisions for the development of
seismic regulations for buildings." ATC-3-06, Palo Alto, Calif.
Bogdanoff, J. L., Goldberg, J. E., and Schiff, A. J. (1965). "The effect of ground
transmission time on the response of long structures." Bull. Seismologic Soc. of
Am. 55, 627-640.
Der Kiureghian, A. (1980). "Structural response to stationary excitation." J. Engrg.
Mech., ASCE, 106(6), 1195-1213.
Fed. Emergency Mgmt. Agency (FEMA). (1986). NEHRP recommended provisions
for the development of seismic regulations for new buildings. Building Seismic
Safety Council, 1015 15th Street, N.W., Suite 700, Washington, D.C.
Gradshteyn, I. S., and Ryzhik, I. M. (1980). Table of integrals, series and products.
Academic Press, Inc., Orlando, Fla.
Harichandran, R. S., and Vanmarcke, E. H. (1986). "Stochastic variation of earth-
quake ground motion in space and time." / . Engrg. Mech., ASCE, 112(2), 154-
174.
Hoshiya, M., and Ishii, K. (1983). "Evaluation of kinematic interaction of soil foun-
dation systems by a stochastic model." / . of Soil Dynamics and Earthquake Engrg.,
2(3), 128-134.
Iguchi, M. (1984). "Earthquake response of embedded cylindrical foundation to SH
and SV waves." Proc. Eighth World Conf. on Earthquake Engrg., San Francisco,
Calif., 1081-1088.
Kausel, E., and Pais, A. (1987). "Stochastic deconvolution of earthquake motions."

955

J. Struct. Eng., 1989, 115(4): 935-956

 J. Engrg. Mech., ASCE, 113(2), 266-277.
Lin, Y. K. (1976). Probabilistic theory of structural dynamics. Robert E. Krieger
Publishing Co., Huntington, N.Y., 155-202.
Loh, C-H. (1985). "Analysis of the spatial variation of seismic waves and ground
Downloaded from ascelibrary.org by UNIV OF STELLENBOSCH-PERIOD on 09/04/20. Copyright ASCE. For personal use only; all rights reserved.

movements from smart-1 array data." J. Earthquake Engrg. and Struct. Dynamics
13, 561-581.
Luco, J. E., and Mita, A. (1987a). "Response of a circular foundation on a uniform
half-space to elastic waves." / . Earthquake Engrg. and Struct. Dynamics, 15,
105-118.
Luco, J. E., and Mita, A. (1987b). "Response of a circular foundation to spatially
random ground motion." J. Engrg. Mech., ASCE, 113(1), 1-15.
Luco, J. E., and Wong, H. L. (1986). "Response of a rigid foundation to a spatially
random ground motion." J. Earthquake Engrg. and Struct. Dynamics 14, 8 9 1 -
908.
Matsushima, Y. (1977). "Stochastic response of structure due to spatial variant earth-
quake excitations." Sixth World Conf. on Earthquake Engrg., New Delhi, India,
1077-1082.
Mita, A., and Luco, J. E. (1987). "Response of structures to spatially random ground
motion." Proc. Third U.S. Nat. Conf. on Earthquake Engrg., Charleston, S.C.,
907-918.
Morgan, J. R., Hall, W. J., and Newmark, N. M. (1983). "Seismic response arising
from traveling waves." J. Struct. Engrg., ASCE, 109(4), 1010-1027.
Newmark, N. (1969). "Torsion of symmetrical buildings." Proc. 4th World Conf.
on Earthquake Engrg., Santiago, Chile, A-3, 19-32.
Novak, M., and Suen, E. (1987). "Dam-foundation interaction under spatially cor-
related random ground motion." Proc. 3rd Int. Conf. on Soil Dynamics and Earth-
quake Engrg., Princeton Univ., Princeton, N.J.
Pais, A., and Kausel, E. (1985). "Stochastic response of foundations." Rep. No.
R85-6, MIT Dept. of Civ. Engrg., Cambridge, Mass.
Roesset, J. M. (1980). "A review of soil-structure interaction." UCRL-15262, Law-
rence Livermore Nat. Lab., Livermore, Calif.
Scanlan, R. H. (1976). "Seismic wave effects on soil-structure interaction." Earth-
quake Engrg. and Struct. Dynamics 4, 379-388.
Vanmarcke, E. H. (1975). "On distribution of the first-passage time for normal sta-
tionary random processes." / . Appl. Mech. 42, 215-220.
Veletsos, A. S. (1977). "Dynamics of structure-foundation systems." Structural and
geotechnical mechanics, W. J. Hall, ed., Prentice-Hall, Inc., Englewood Cliffs,
N.J., 333-361.
Veletsos, A. S. (1978). "Soil-structure interaction for buildings during earthquakes."
Proc. Second Int. Conf. on Microzonation, San Francisco, Calif., 1, 111-133.
Veletsos, A. S., and Meek, J. W. (1976). "Dynamic behavior of building-foundation
systems." J. Earthquake and Struct. Dynamics 3, 121-138.
Veletsos, A. S., and Nair, V. V. D. (1974). "Torsional vibration of viscoelastic
foundations." J. Geotech. Engrg., ASCE, 100(3), 225-246.
Veletsos, A. S., and Verbic, B. (1973). "Vibration of viscoelastic foundations." J.
Earthquake Engrg. and Struct. Dynamics 2, 87-102.
Veletsos, A. S., Erdik, M. O., and Kuo, P. T. (1976). "Structural response to trav-
eling seismic motions." Proc. 5th European Conf. on Earthquake Engrg., Istan-
bul, Turkey, Ch. 4, 63-1 to 63-14.
Werner, S. D., Lee, L. C , Wong, H. L., and Trifunac, M. D. (1979). "Structural
response to traveling seismic waves." / . Struct. Engrg., ASCE, 105(12), 2547-
2564.
Yamahara, H. (1970). "Ground motions during earthquakes and the input loss of
earthquake power to an excitation of building." Soils and foundations, Japan Soc.
of Soil-Mech. and Foundation Engrg. 10(2), 145-161.

956

J. Struct. Eng., 1989, 115(4): 935-956