EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)

DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL

COUPLING OF 3-D FOUNDATIONS AND STRUCTURES

JEFFREY S. MULLIKEN1 AND DIMITRIS L. KARABALIS2*

1 The LPA Group Inc., 2530 Devine Str., P.O. Box 5805, Columbia, SC 29250, U.S.A.
2 Department of Civil Engineering, University of Patras, 26500 Patras, Greece

SUMMARY
An efficient discrete model for predicting the dynamic through-the-soil interaction between adjacent rigid, surface
foundations supported by a homogeneous, isotropic and linear elastic half-space is presented. The model utilizes
frequency-independent springs and dashpots, and the foundation mass, for the consideration of soil—foundation
interaction. The through-the-soil coupling of the foundations is attained by frequency-independent stiffness and damping
functions, developed in this work, that interconnect the degrees of freedom of the entire system of foundations. The
dynamic analysis of the resulting coupled system is performed in the time domain and includes the time lagging effects of
coupled dynamic input due to wave propagation using an appropriate modification of the Wilson-h method. The basic
foundation interaction model is also extended to the evaluation of coupled building-foundation systems. ( 1998 John
Wiley & Sons, Ltd.

KEY WORDS: foundation—soil—foundation interaction; discrete model; soil—structure interaction

INTRODUCTION
It is broadly accepted nowadays, after several theoretical, experimental and field studies, that the dynamic
response of structural systems is significantly affected by the interaction with adjacent structures, in addition,
of course, to several other well-known factors. Even the dynamic response of the independent or structurally
interconnected adjacent foundations of a single structure is affected by through-the-soil dynamic coupling
among these foundations and, thus, the dynamic behaviour of the entire structure is influenced. This
phenomenon, rarely studied only during the last 30 years, is generally considered to fall within the
Soil—Structure Interaction (SSI) field although it clearly presents interactive characteristics beyond those of
the classic, and well understood, Foundation—Soil Interaction (FSI) problem.
The computation of dynamic interaction between adjacent foundations and structures requires the
solution of the FSI problem and the problem of foundation coupling through the supporting medium.
Solutions to either one of these problems can be obtained using rigorous analytical formulations, which
usually pertain to simple geometries and loading functions, or numerical procedures, usually FEM or BEM,
for more complicated configurations. In addition, a limited number of approximate discrete models with
relatively few Degrees Of Freedom (DOFs) have also been developed. These models provide practical and
efficient alternatives to computationally expensive numerical solutions or mathematically complicated
analytical formulations since their usage requires the application of basic concepts of structural dynamics
and access to a modest computer. While their order of accuracy is considered inferior to that of the more
rigorous analytic or numerical solutions it usually falls within the range of accuracy expected for

* Correspondence to: Dimitris L. Karabalis, Department of Civil Engineering, University of Patras, 26500 Patras, Greece. E-mail:
[email protected]

CCC 0098—8847/98/070687—24$17)50 Received 9 September 1997

( 1998 John Wiley & Sons, Ltd. Revised 19 January 1998
688 J. S. MULLIKEN AND D. L. KARABALIS

a soil—foundation system which by its nature cannot be accurately defined. The focus of this work is on the
development of a simple discrete model for dynamic 3-D Foundation—Soil—Foundation Interaction (FSFI)
analysis based on results obtained by more precise numerical investigations. Therefore, in the following,
specific reference is made to those published works which advanced the development of such discrete models.
The interested reader can find updated literature reviews on analytic and numerical solutions of various
FSFI problems in the recent articles of Qian and Beskos,1,2 and Mohammadi and Karabalis.3
One of the earliest efforts to develop practical discrete models for dynamic FSI was undertaken by Lysmer
and Richart.4 Their investigation resulted in a simple Single-Degree-Of-Freedom (SDOF) model for
computing the forced vertical response of foundations attached to an elastic half-space. This concept of
a lumped-parameter representation of foundations on a half-space is described in great detail and further
developed by Whitman,5 Richart et al.,6 Newmark and Rosenblueth,7 Gazetas8 and Wolf,9,10 among several
others. Meek and Veletsos11 extended these basic models by introducing an additional DOF to the dynamic
system. Generalizations of this method were developed by De Barros and Luco12 for surface and embedded
foundations and Wolf and Somaini13 for embedded foundations. Wolf and Paronesso14,15 used Multiple-
Degree-Of-Freedom (MDOF) lumped-parameter models for surface and embedded foundations on/in a soil
layer over bedrock. Wolf16,17 has also presented a set of ‘consistent’ lumped-parameter models for soil
media.
The through-the-soil coupling of foundations was first introduced by Whitman5 as an important problem
requiring study. Among the first publications in this area one should mention those of Warburton et al.,18,19
who performed studies of two masses with identical circular bases attached to an elastic half-space, and
MacCalden and Matthiesen,20 who developed a matrix formulation for the solution of the induced dynamic
displacement of a foundation near a harmonically loaded foundation attached to an elastic half-space.
However, comparison studies, presented in the latter publication, between theoretical and experimental
results showed significant discrepancies. A special mention should also be made to the mathematically
rigorous solutions presented by Triantafyllidis21 and Triantafyllidis and Prange22 which, however, are
unavoidably restricted to specific geometries. Among the latest numerical solutions one can mention the
time-domain BEM approach of Guan and Novak23 based on the half-space solutions provided by the same
authors,24 and the frequency-domain BEM formulation of Qian and Beskos.1,2 In both cases half-space
solutions have been used requiring the discrertization of only the contact surface between the soil medium
and the foundation. Huang25 and Karabalis and Huang26 have also reported on a time-domain solution of
the 3-D FSFI problem using the BEM in conjunction with the Stokes fundamental solutions. Soil layering
along with FSFI was studied by Karabalis and Mohammadi27,28 using a frequency-domain BEM and full-
space fundamental solutions. Thus, their solution required the discretization of the interlayer contact surface
in addition to that of the free surface. The dynamic structure—soil—structure interaction problem has been
studied analytically by Luco and Contesse,29 Wong and Trifunac30 and Murakami and Luco.31 In these
articles the loading functions were obliquely incident SH waves and the structures under consideration
consisted of two or more shear walls founded on semicylindrical rigid foundations. Their numerical studies
showed that groups of closely spaced buildings can result in interaction effects near the fundamental
frequencies of the buildings and at very low frequencies. Structural coupling, in addition to through-the-soil
coupling, between adjacent foundations has been considered by Mohammadi and Karabalis3 for the case of
a railway line using a frequency-domain BEM/FEM formulation. Their system of up to nine ties interconnec-
ted by rails showed significant sensitivity to both structural and through-the-soil coupling the influence of
which was studied independently. An approximate analytical—numerical approach was proposed by Lee and
Wesley32 for the solution of a 3-D through-the-soil interaction problem involving three rigid circular
foundations and spring-mass models for the superstructures attached to the foundations.
In the work described in this paper, coupling functions corresponding to discrete springs and dashpots
have been developed for the computation of the dynamic interaction of massive surface foundations
supported by a homogeneous, isotropic, linear elastic half-space. The time-lagging effects associated with
wave propagation are taken into consideration via a modification of the Wilson-h method33 developed for

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 689

the needs of this work. For the solution of the associated FSI problem, discrete models along the lines of
those proposed by Richart et al.,6 Gazetas8 and Wolf 9,10 have been adopted, while the development of the
coupling functions is based on the numerical results presented by Huang.25 The dynamic response of
a number of foundations due to external forces applied to one foundation is computed directly in the time
domain utilizing the coupling functions. These functions are also used, in the ‘numerical results’ section, to
compute through-the-soil coupling effects of closely spaced buildings subjected to ground acceleration. The
employment of such a direct time-domain analysis has the additional advantage of enabling the considera-
tion of non-linear soil and structures by simple modifications of the basic models described herein.
The original work of Mulliken34 is the basis of this article while later developments and results produced
by the authors have also been incorporated. Parts of this work have also been presented in Mulliken and
Karabalis.35

FOUNDATION—SOIL—FOUNDATION SYSTEM
The geometry and nomenclature for the foundation—soil—foundation system studied in this work are as
shown in Figure 1. This system can be viewed as an assemblage of individual unconstrained foundation—soil
systems coupled by the, common to all of them, soil medium. The problem is solved by first defining the
individual FSI systems with discrete masses, springs and dampers where each mode of vibration is considered
as an independent DOF, as shown in Figure 2. Subsequently, foundation coupling due to wave propagation
is introduced by an approximate method developed by Mulliken.34 This method introduces frequency-
independent coupling functions developed for each foundation DOF which result in springs and dashpots
that link the various DOFs of the foundations, as shown in Figure 3. The incorporation of these springs and

Figure 1. Foundation system geometry and nomenclature

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
690 J. S. MULLIKEN AND D. L. KARABALIS

Figure 2. Basic one-dimensional soil—foundation discrete model

Figure 3. Foundation coupling model

dashpots into the solution of a coupled dynamic system is achieved by a proposed modification of the
Wilson-h method so that time-lagging effects due to wave propagation are also taken into consideration.
The following three assumptions underline the development of the coupling model described in this work:
(a) coupling between foundations is determined for each mode of vibration independent of the other DOFs,
i.e. vertical motions at one foundation produce only vertical responses at the adjacent foundations, rocking
input produces only rocking responses, etc., (b) the static stiffness of each individual foundation—soil system is
not significantly effected by the close proximity of the adjacent foundations (an exception is made later for the
rocking motion), and (c) the coupling forces at one foundation occur at some time after the displacement of
the adjacent foundation, and the time lag is directly related to the wave speeds in the supporting medium.
Given these assumptions, the static stiffness and damping of individual foundation—soil systems are stored in
one set of matrices, while the coupling stiffness and damping are maintained as off-diagonal terms in separate
matrices. This is essential to the FSFI problem investigated in this work since, as it turns out, the presence of
the nearby foundations, modelled by the coupling terms, does not influence (except for the rocking motion in
the very near field) the individual stiffness and damping of single foundations. In addition, if constant stiffness
and damping matrices with off-diagonal terms were used, it would introduce direct coupling between DOFs,

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 691

and time-lagging effects between motions of adjacent foundations would not be possible. However, this work
evidences the fact that ignoring this time lag significantly alters the characteristics (frequency content and
amplitude) of motion in the time domain.
In the following, first the frequency-independent FSI models utilized in this work will be discussed. Next,
the time-domain solution of the forced dynamic response of single-degree-of-freedom FSI systems via the
Wilson-h method is presented for completeness. The proposed modifications to the Wilson-h method
required in order to compute coupling effects between multiple, adjacent foundations is then introduced.
Given these concepts the development of frequency-independent coupling functions is then described. As it
will be exhibited, these coupling functions enable the computation of dynamic through-the-soil coupling
between any number of square foundations for various time-dependent loading functions. Although the basic
concepts of the Wilson-h method are utilized in this work, any other direct integration method of preference
can be used, in concert with the described modifications, to evaluate foundation coupling.

Discrete FSI models

The relationship between an applied loading function on a rigid, massive foundation attached to an elastic
medium, and the corresponding displacement function of the foundation, can be determined for each
foundation DOF using a simple SDOF lumped parameter model, as shown in Figure 2. The mass, M,
represents the mass (or rotational inertia) of the foundation, m, plus a virtual soil mass (or rotational inertia),
m , attached to the foundation. This total mass is supported by a rigid support through a spring having
7
a stiffness, K, and a dashpot having a coefficient, C. These three parameters should be defined for each DOF.
Newmark and Rosenblueth7 describe the concept of virtual soil mass as a simple approach to the
consideration of the soil mass contribution to FSI problems. This method enables the use of frequency-
independent discrete models for the solution of a strongly frequency-dependent problem. After some
numerical experimentation with various combinations of available coefficients for M, K, and C the virtual
soil mass coefficients adopted in this work are those proposed by Richart et al.6 and Gazetas,8 while for the
stiffness those of the Wolf10 model are used. The damping coefficients while they are in the form proposed by
Wolf10 have been adapted so that in combination with the corresponding coefficients for mass and stiffness
produce the best fit to the various available data for the SDOF soil—foundation system. Table I below
reproduces the equations used in the computation of these coefficients with a being the half-width of a square
foundation and » "JG/o denoting the shear wave velocity, where G and o are the shear modulus and mass
4
density, respectively, of the soil medium.

Table I. Mass, stiffness and damping coefficients for one-dimensional discrete FSI model

Virtual soil
Mass (inertia) Equivalent mass (inertia), Static stiffness Damping
ratio, b radius, r m K C
0 7
(1!l) m 2a 0)27m 4)7Ga 0)8a
Vertical K
4 or3 Jn b 1!l »
0 s
(7!8l) m 2a 0)095m 9)2Ga 0)163a
Horizontal K
32(1!l) or3 Jn b 2!l »
0 s
3(1!l) m 2a 0)24m 4)0Ga3 0)6a
Rocking K
8 or5 J3n
4 b 1!l »
0 s
m 2a 0)045m 0)127a
Torsion 8)31Ga3 K
or5 J3n
4 b V
0 s

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
692 J. S. MULLIKEN AND D. L. KARABALIS

In view of the above definitions, the equation of motion for each DOF of a single foundation is

Mÿ(t)#CyR (t)#Ky(t)"F(t) (1)

where ÿ(t), yR (t) and y(t) are the acceleration, velocity and displacement, respectively, for the DOF under
consideration, and F(t) is the externally applied forcing function.

¹ime-domain formulation—SDOF system

The step-by-step linear acceleration method with the Wilson-h extended step is used for the integration of
the equation of motion (1). The basic assumption of the Wilson-h method, as shown in Figure 4, is that the
acceleration in the equation of motion can be represented by a linear function during the time step q"h*t.
By extending the time step, *t, by the factor h, the numerical stability of the solution is assured regardless of
the time step chosen as long as h*1)38. In the following, the Wilson-h method is outlined, for completeness
purposes, on the basis of the description provided by Paz.33
From Figure 4, it is seen that the area under the acceleration line over the extended step (t"t #q), i.e. the
i
incremental velocity, is evaluated by

*ª yR "ÿ q#1 *ª ÿ q (2)

i i 2 i
and, thus, the incremental displacement is

*ª y "yR q#1 ÿ q2#1 *ª ÿ q2 (3)

i i 2 i 6 i
Solving equation (3) for the incremental acceleration *ª ÿ and substituting in equation (2) yields
i
*ª ÿ "a *ª y !a yR !3ÿ (4)
i 4 i 2 i i
*ª yR "a *ª y !3yR !a ÿ (5)
i 1 i i 3 i

Figure 4. Linear acceleration assumption in the Wilson extended time interval

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 693

where
3 6 q 6
a " , a " , a " , a " (6)
1 q 2 q 3 2 4 q2

Taking the difference between the equilibrium conditions defined at t and t #q, the dynamic equation of
i i
motion (1) can be expressed in incremental form as
M*ª ÿ #C*ª yR #K*ª y "*ª F (7)
i i i i
where the incremental load *ª F is determined by linear interpolation as
i
*ª F "F #(F !F ) ) (h!1)!F (8)
i i`1 i`2 i`1 i
Direct substitution of equations (4) and (5) into this incremental equation of motion (7) results in an equation
for the incremental displacement which can be written conveniently as
KM *ª y "*ª FM (9)
i i
where the effective stiffness matrix is
KM "K#a M#a C (10)
4 1
and the effective incremental load is
*ª FM "*ª F #(a M#3C) yR #(3M#a C) ÿ (11)
i i 2 i 3 i
The incremental displacement for the extended time interval is obtained from equation (9) which upon
substitution into equation (4) yields the incremental acceleration for the extended step. Subsequently, the
incremental acceleration *ÿ for the normal time increment *t is determined by linear interpolation as
i
*ª ÿ
*ÿ " i (12)
i h
After computing the incremental acceleration, the displacement and velocity at the end of the time interval
t #*t can be computed by direct substitution into equations (2) and (3) using the normal time increment,
i
while the corrected acceleration is computed directly from the equation of motion (1) at t"t , i.e.,
i`1
1
yR "yR #ÿ *t# *ÿ *t (13)
i`1 i i 2 i
1 1
y "y #yR *t# ÿ *t2# *ÿ *t2 (14)
i`1 i i 2 i 6 i
ÿ "M~1 (F !CyR !Ky ) (15)
i`1 i`1 i`1 i`1
The Wilson-h procedure, outlined above, can, of course, be repeatedly used to compute the values of
acceleration, velocity and displacement at subsequent time steps until the desired final time of the analysis is
reached.

¹hrough-the-soil coupling of adjacent foundations

The through-the-soil coupling of adjacent foundations, as shown in Figure 1, is achieved via frequency-
independent stiffness and damping coefficients, as shown in Figure 3. As previously explained, the basis of the
required modification to Wilson-h method is that the static stiffness of the individual foundations is
essentially unchanged by the presence of other foundations, and that a finite time period exists between the
response of one foundation and the influence of that response on all other foundations adjacent to it.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
694 J. S. MULLIKEN AND D. L. KARABALIS

The formulation described above for a single FSI system can be expanded to include multiple FSI systems
by resorting to matrix notation. Thus, if no consideration is given to through-the-soil coupling effects the
stiffness, mass and damping matrices of a group of independent FSI systems can be written in the form of
diagonal matrices, with the subscripts indicating each foundation, as

K 0 0 2 M 0 0 0C 0 2
1 1 2 1
0 K 0 2 0 M 0 0 C 0 2
[K]" 2 , [M]" 2 2
, [C]" 2
0 0 K 2 0 0 M 0 0 C 2
3 3 2 3
F F F } F F F } F F F }
(16)
and, similarly, the displacement, velocity, acceleration, forcing function vectors as

GH GH GH GH
y yR ÿ F
1 1 1 1
y yR ÿ F
y" 2 , yR " 2 , ÿ" 2 , F" 2 , (17)
y yR ÿ F
3 3 3 3
F F F F
The inclusion of through-the-soil coupling is achieved by the introduction of stiffness and damping
coefficients that inter-connect the individual FSI systems, as shown in Figure 3. These coefficients are stored
as off-diagonal terms in a separate set of matrices, i.e. independent to those shown in equation (16), which are
of the form

0 !Kc(1,2)
!K 0 !C
!C
c(1,3) 2 c(1,2) c(1,3) 2
!K 0 !K !C 0 !C
[K ]" c(2,1) c(2,3) 2 , [C ]" c(2,1) c(2,3) 2
c !K !K 0 c !C !C 0
c(3,1) c(3,2) 2 c(3,1) c(3,2) 2
F F F } F F F }
(18)
The subscripts here indicate the foundations that the coupling terms connect. The assembled matrices should
be symmetrical since the reciprocity condition K "K and C "C holds for the coupling
#(i,j) #( j,i) c(i,j) c(j,i)
terms. The computation of these coupling terms is the subject of the next section.
The separation of the stiffness and damping matrices into non-coupling, equation (16), and coupling,
equation (18), coefficients becomes necessary for two reasons. The first reason is that the effective stiffness
matrix in equation (10) should not be altered by the presence of adjacent foundations. The second reason is
that the coupling forces occur at some time after each foundation responds, e.g. if foundation i undergoes
displacement, velocity and acceleration responses at time t, foundation j will be affected by those responses at
some time t#*¹, as depicted in Figure 5, *¹ being the time lag for a coupled response. Thus, at any time
during the analysis, the response of a foundation is the result of applied (external) and internal (SSI) forces at
the current time, and forces due to coupling between foundations that are caused by the response of adjacent
foundations at past time. By storing the coupling terms into separate matrices, it becomes possible to include
in the forces applied in the current time step at each individual foundation those coupling forces that are due
to previous responses at each adjacent foundation.
The actual time lag *¹ for the various coupled motions considered in this work are listed in Table II.
These formulae have been derived from observations on numerical results presented by Huang25 for a study
on two adjacent rigid, surface square foundations using a time-domain BEM. As it was expected *¹ depends
on the distance d between foundations, the foundation half-width a, and the corresponding wave velocity.
i,j
( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 695

Figure 5. Time lag in coupled response between loaded and unloaded foundation

Table II. Time lag for coupled dynamic response between two adjacent,
rigid, surface, square foundations

Vertical Horizontal Rocking Torsional

motion motion motion motion

3 (d#a/2) (d#a/2) (d#a/2) (d#a/2)

*¹ (sec)
4 » » » »
S p S S

» in Table II is the compression wave, or P-wave, velocity defined as

P

S
2!2l
» "» (19)
P S 1!2l

with l being the Poisson ratio.

The incorporation of coupling between foundations in the Wilson-h method involves modifications to the
effective incremental load in equation (11) and the computation of acceleration at the end of the normal time
increment in equation (15). The matrices [yJ ] , [yRI ] and [ÿI ] introduced here are a collection of displacement,
i i i
velocity and acceleration vectors, respectively, of each foundation at times that lag the current time t by *¹
i
and are of the form

G HG HG HG H
0 y (t !*¹ ) y (t !*¹ )
1 i 2,1 1 i 3,1 2
y (t !*¹ ) 0 y (t !*¹ )
[yJ ] " 2 i 1,2 2 i 3,2 2
(20)
i y (t !*¹ ) y (t !*¹ ) 0
3 i 1,3 3 i 2,3 2
F F F }

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
696 J. S. MULLIKEN AND D. L. KARABALIS

G HG HG HG H
0 yR (t !*¹ ) yR (t !*¹ )
1 i 2,1 1 i 3,1 2
yR (t !*¹ ) 0 yR (t !*¹ )
[yRI ] " 2 i 1,2 2 i 3,2 2
(21)
i yR (t !*¹ ) yR (t !*¹ ) 0
3 i 1,3 3 i 2,3 2
F F F }

G HG HG HG H
0 ÿ (t !*¹ ) ÿ (t !*¹ ) 2
1 i 2,1 1 i 3,1
ÿ (t !*¹ ) 0 ÿ (t !*¹ ) 2
[ÿI ]" 2 i 1,2 2 i 3,2 (22)
ÿ (t !*¹ ) ÿ (t !*¹ ) 0 2
3 i 1,3 3 i 2,3
F F F }

As an example, the first vector of the displacement matrix in equation (20) consists of the displacements of all
foundations, with the exception of foundation 1, which are lagged based on the distance between each
foundation and foundation 1, as noted in the subscripts referenced in the time lag, *¹ . These displace-
1,j
ments are required in order to determine the effect that prior displacements at all foundations have on
foundation 1 in the current time increment. The second vector provides the lagged displacement input for all
foundations that effect the response of foundation 2, and so on. The number of vectors is determined by the
number of foundations. The same notation is used in the assembly of the time-lagged velocity, equation (21),
and acceleration, equation (22), matrices.
The computation of coupling forces which is accomplished by multiplying the stiffness and damping
coupling matrices in equation (18) by the time-lagged displacement, velocity and acceleration matrices in
equations (20)—(22), requires special treatment. To produce vectors of force by the multiplication of two
matrices the following formula, shown symbolically, is used
n
MFI N " + a ) b (23)
i i,k k,i
k/1
where a represents the stiffness or damping coupling matrix, b represents the time-lagged displacement,
i,k k,i
velocity or acceleration matrix, and n is the number of foundations. Thus, an application of equation (23)
yields the forcing vector at a foundation based on the prior response of all foundations adjacent to it.
After the assemblage of the coupling matrices of equations (18), and the time-lagged matrices of equations
(20)—(22), the computation of the effective incremental load, including through-the-soil coupling forces, can
be achieved by modifying equation (11) as
M*ª FM N "M*ª FN #(a [M]#3[C])MyR N #3[C ][yRI ] #(3[M]#a [C])MÿN #a [C ] [ÿI ] (24)
i i 2 i c i 3 i 3 c i
Similarly, the acceleration at the end of the normal time increment, including the coupling forces, can be
computed by modifying equation (15) as
MÿN "[M]~1 (MFN ![C]MyR N ![C ][yRI ] ![K]MyN ![K ][yJ ] ) (25)
i`1 i`1 i`1 c i`1 i`1 c i`1
With these modifications to the well-known formulae of the Wilson-h method the total response of each
foundation including through-the-soil coupling can be computed. Apparently, the modifications introduced
by equations (24) and (25) do not alter the general flow of computations in the Wilson-h method and, thus,
existing computer codes require only minor changes.
As in any numerical method, the selection of the time step is critical to the accuracy of the solution of
a step-by-step integration method. It is well established, see, for example, Karabalis and Beskos,36 that using
the Wilson-h method accurate results can be obtained when the time interval is taken to be no longer than
one-tenth of the natural period of the system being analysed. In addition, a sufficiently small time interval
should be chosen such that the variation of the input load with time is properly represented. It was found in

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 697

the course of this work that a time interval of one-twentieth of the lowest period of the coupled system was
required to limit inaccuracies. Eigenvalue analyses using the Stodola method33 were performed to determine
the lowest period. The stiffness matrix necessary in these analyses was computed by a direct summation of the
stiffness matrices of equations (16) and (18).

Computation of coupling coefficients

Having defined the time domain analysis procedure for incorporating through-the-soil coupling of
foundations, the attention is now turned to the method used to determine the coupling functions for adjacent
foundations. These functions depend on the distance between the foundations and the foundation size, and
are necessary to compute the stiffness and damping coupling coefficients shown symbolically in equation (18).
The stiffness and damping coupling coefficients for each DOF are defined in their basic form in Table III.
The reader will recognize that the equations in Table III are the same as those reproduced in Table I from
Wolf 10 with the exception that the numerical constants are replaced by functions of the dimensionless
distance ratio d /a, ! for stiffness and ( for damping. These coupling functions were developed via an
i,j
iterative process using the direct integration scheme described above for a two-foundation system in which
one of the foundations was loaded and the responses of both foundations were computed. The off-diagonal
stiffness and damping coupling coefficients were adjusted until the loaded and unloaded foundation
responses fit existing time-domain solutions of the FSFI problem to within an average of 0)5 per cent error at
each time step. In this particular work the calibration of the stiffness and damping functions ! and ( was
based on Huang’s25 time-domain BEM solutions for two identical, square foundations attached to an elastic
half-space with Poisson ratio l"1 . The ! and ( functions produced by this iteration process correspond to
3
dimensionless distance ratios of 0)25, 0)50, 1)0, 2)0 and 3)0 only and were verified for various foundation
masses and soil material constants on the basis of the results provided by Huang.25
However, it would be useful, for apparent practical reasons, to develop continuous functions ! and ( valid
for any chosen distance ratio d/a. For this purpose standard regression analyses are used which, depending
on the form of the functions ! and (, are based on logarithmic or exponential curve fitting formulae, of the
form
Logarithmic: ½ "a * LOG (X ) (26)
i 10 i
Exponential: ½ "a * 10 (u * Xi ) (27)
i
where a and u are constants and X represents the distance ratio d/a at which coupling constants have been
i
computed from the above iterative process. These fitted curves allow the determination of coupling
coefficients for any value of the distance ratio d/a, and also make possible the extrapolation of data beyond
the distance ratios used in the development of the ! and ( functions via the specific iteration process on the
basis of the available numerical data.
The stiffness and damping coupling functions between any two foundations developed via the above
described process are shown graphically in Figure 6 and given in an explicit form in Table IV. To apply these

Table III. Coupling coefficients for foundation—soil—foundation interaction

Vertical Horizontal Rocking Torsional

motion motion motion motion

Ga Ga Ga3
Stiffness ! ] ! ] ! ] ! ]Ga3
3 (1!l) 1 (2!l) (2 (1!l) (3

Ga2 Ga2 Ga4 Ga4

Damping ( ] ( ] ( ] ( ]
3 » (1!l) 1 » (2!l) (2 » (1!l) (3 »
S S S S

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
698 J. S. MULLIKEN AND D. L. KARABALIS

Table IV. Explicit expressions of coupling functions for foundation—soil—foundation interaction

Vertical motion Horizontal motion Rocking motion Torsional motion

Stiffness, ! 1)614]10!0)16257(d/a) 3)7561]10!0)18995(d/a) !(0)04234!0)2396 0)05931

]LOG (d/a))
10
Damping, ( 8)504 13)2875 7)3823!6)775 4)4429—2)9125
]LOG (d/a ) ]LOG (d/a)
10 10

Figure 6. Coupling functions for: (a) vertical; (b) horizontal; (c) rocking and; (d) torsional motion versus distance ratio d/a

functions to multiple foundations, the stiffness and damping coupling coefficients between the first and
second foundation, as shown in Figure 3, are computed by using the ratio d /a in the coupling functions of
1,2
Tables III and IV. Accordingly, d /a is used for coupling between the second and third foundation, d /a
2,3 1,3
between the first and third foundation, etc. This procedure can be applied to any number of foundations in

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 699

a row. It should be noted at this point, that the accuracy of the computed coupling functions between any two
foundations is not, in general, affected appreciably by the existence of other foundations between them. This
observation is supported by the detailed numerical results presented by Mohammadi and Karabalis27,28 and
is shown to hold true in the verification studies presented in the ‘Numerical Results’ section.

Near-field effects
During the development of the coupling functions proposed in this work, the influence of the near-
field effect upon the accuracy of the results was also studied. It was determined, through extensive
numerical verification studies,34 that near-field effects are inconsequential for all but the rocking mode
of vibration. Further analysis of this vibration mode indicated that the individual rocking stiffness and
damping of each foundation was influenced by the close proximity of the adjacent foundations. In order
to incorporate this effect for distance ratios d/a)0)5, stiffness and damping matrices in equation (16) are
replaced by
K #K* 0 0
1 1 2
0 K #K* 0
[K]" 2 2 2
0 0 K #K* 2
3 3
F F F }
(28)
C #C* 0 0
1 1 2
0 C #C* 0
[C]" 2 2 2
0 0 C #C* 2
3 3
F F F }

where, upon computing the coupling coefficients, K and C ,

c(i,j) c(i,j)
d of foundations
K*" + Kc(i, j) (iOj ) (29a)
i
j/1
d of foundations
C*" + Cc(i, j ) (iOj ) (29b)
i
j/1
Equation (29) provides the means to incorporate on the independent stiffness and damping of each
foundation the cumulative influence of all foundations adjacent to it.
It has also been determined that, in addition to the above modification to the stiffness and damping
matrices, the stiffness and damping coupling functions for distance ratios d/a)0)5 are computed more
accurately using the following expressions
!"!(0)967!0)880 (d/a))
(30)
("0.204
However, it should be pointed out that the stiffness and damping functions for rocking vibrations shown in
Table IV are still applicable for distance ratios d/a'0)5.

NUMERICAL RESULTS
This section provides numerical examples utilizing the methodology and coupling functions described in the
previous sections of this work. First, results for a two-foundation system are presented for each DOF and

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
700 J. S. MULLIKEN AND D. L. KARABALIS

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 701

Figure 7. Loaded and unloaded foundation response versus time due to a rectangular impulse load

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
702 J. S. MULLIKEN AND D. L. KARABALIS

various distance ratios, and comparisons are made against those BEM solutions used in the development of
the coupling functions introduced in this work. Subsequently, in an attempt to establish the universal
applicability of the proposed methodology, comparison studies with references to a Green’s function solution
of a two-foundation system and a finite element solution of a three-foundation system are also shown. Lastly,
a study of the interaction of three adjacent structures undergoing earthquake excitation is presented by
appropriately extending the methodology introduced in this work.

¹wo-foundation system
The dynamic response of two massive, rigid, square surface foundations on a homogeneous elastic
half-space is studied first. The soil medium constants are: mass density, o"10)368 lb-sec2/ft4 (5343)5 kg/m3),
shear modulus, G"9)71175]105 ksf (46)5 GPa), and Poisson’s ratio, l"1 . The foundations have a mass
3
M"10]or3 and a width 2a"5 ft (1)524 m). The external loads on one of the foundations are impulsive
0
forces or moments, depending on the loaded DOF, in the form of a rectangular impulse with a magnitude of
100 lb (444)822 N) and a duration of 0)00001821 sec. Figure 7 provides a comparison of vertical, horizontal,
torsional and rocking responses for various distance ratios of the results obtained using the proposed
methodology and coupling functions and the BEM solutions of Huang.25 The agreement between the two
solutions is apparent. These comparison studies, although indicative of the accuracy of the proposed
methodology, cannot be considered ‘objective’ since the results of Reference 25 have been used for the
development of the coupling functions introduced in this work. Further similar studies of various distance
and mass ratios using the proposed model can be found in Mulliken.34

Figure 8. Non-dimensional loaded and unloaded foundation response versus time due to a rectangular impulse load

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 703

The results of the two-foundation system analysed by Guan and Novak23 is used as an independent
comparison study. In their study one foundation is loaded with a vertical force in the form of a rectangular
impulse and the responses of both the loaded and unloaded foundations are computed in time domain on the
basis of a Green’s function formulation. A system of two rigid, square foundations each having mass
M"1]or3 , and distance ratio d/a"2 is used for the comparison study. Guan and Novak23 provide their
0
results in terms of dimensionless time q"» t/2a, and displacement d"aGº /P , where º is the foundation
s z z z
vertical displacement and P is the magnitude of the vertical impulse load as shown in Figure 8. The results
z
from the proposed model and those from Guan and Novak23 are shown in Figure 8 in dimensionless form.
The close agreement of the results produced by the two methodologies should be noted.

¹hree-foundation system
To further demonstrate the accuracy of the proposed model, a finite element solution of a FSFI problem
involving three foundations, geometrically arranged as shown is Figure 9, is performed with the computer
program SASSI37 and the results are compared to those obtained using the proposed model. The mechanical
constants of the system are: mass of each foundation 1663)0124 lb-sec2/ft (24269)84 kg), shear wave velocity of
soil 600 ft/sec (182)88 m/sec), unit weight of soil 0)120 kips/ft3 (18)85 kN/m3), Poisson ratio of soil 1/3.
A decaying sinusoidal forcing function is applied on foundation 1 only, as shown in Figure 9. Once again the
close agreement of the results obtained by the two models is apparent.

Figure 9. Dynamic response versus time of a set of three foundations due to an external impulsive load

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
704 J. S. MULLIKEN AND D. L. KARABALIS

¹hree-foundation—structure system
In this example the foundation coupling model developed in this work is extended to study the
through-the-soil coupling of adjacent structures under earthquake induced excitation. The basic lumped
parameter model of a single structure described by Wolf10 is shown for completeness in Figure 10(a). This
model considers the structure as a SDOF system that accounts for the fundamental vibration mode with the
appropriate structural mass M , stiffness K and damping C . Induced rocking motions are taken into
4 3 3
account by supporting the structural model at a height h above the foundation by a rigid bar. The rigid bar is
supported at the foundation level by translational and rocking springs and dashpots that are functions of the
foundation size and the soil medium constants, as shown in Table I. By condensing the rotational DOF at the
top of the rigid bar, a system of matrices for each structure is formulated as
K #K hK !K
1 3 3 3
[K**] " hK K #h2K !hK
i 3 2 3 3
!K !hK K
3 3 3
C #C hC !C
1 3 3 3
[C**] " hC C #h2C !hC (31)
i 3 2 3 3
!C !hC C
3 3 3
M 0 0
0
[M**] " 0 I 0
i 0
0 0 M
4
where the subscript ‘i’ indicates the foundation number, and [K**] , [C**] and [M**] operate on the
i i i
displacement, velocity and acceleration vectors, respectively,

GH G H G H
y (t) yR (t) ÿ (t)
1 1 1
M½**N " y (t) , M½Q **N " yR (t) and M½® **N " ÿ (t) (32)
i 2 i 2 i 2
y (t) yR (t) ÿ (t)
3 3 3

Figure 10. (a) Lumped parameter representation of a foundation-supported structure; (b) series of SDOF structural systems supported
on independent foundations

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 705

where the subscripts 1, 2 and 3 indicate the three DOFs at each foundation supported structure as shown in
Figure 10(a). The earthquake forces are computed by multiplying the structural and soil masses by the
earthquake acceleration time history, ä (t)
g

G HG H
F (t) !M ä (t)
1 0 g
MF**N " F (t) " 0 (33)
i 2
F (t) !M ä (t)
3 s g
The consideration of multiple structures, as shown in Figure 10(b), is achieved by coupling the foundations’
translational and rocking DOFs using the coupling functions and the direct integration procedure proposed
in this work. The procedure presented for multiple foundations is also utilized here with the following
changes to the stiffness, damping and mass matrices of equation (16) in order to incorporate the building
properties

[K**] 0 0
1 2
0 [K**] 0
K" 2 2
0 0 [K**] 2
3
F F F }

[C**] 0 0
1 2
0 [C**] 0
C" 2 2
(34)
0 0 [C**]
3 2
F F F }

[M**] 0 0
1 2
0 [M**] 0
M" 2 2
0 0 [M**]
3 2
F F F }

where the subscripts indicate the foundation number. The displacement, velocity and acceleration vectors of
equation (17) are also re-written to include the structural behaviour

G H G H G H G H
M½**N M½Q **N M½® **N MF**N
1 1 1 1
M½**N M½Q **N M½® **N MF**N
y" 2 , yR " 2 , ÿ" 2 , F" 2 (35)
M½**N M½Q **N M½® **N MF**N
3 3 3 3
F F F F

The coupling matrices of equation (18) are also expanded to handle both translational and rocking vibrations
at each foundation
[K**]
0 [K**] 0 [C**] [C**]
c 1,2 c 1,3 2 c 1,2 c 1,3 2
[K**] 0 [K**] [C**] 0 [C**]
K" c 2,1 c 2,3 2 , C " c 2,1 c 2,3 2 (36)
c [K**] [K**] 0 c [C**] [C**] 0
c 3,1 c 3,2 2 c 3,1 c 3,2 2
F F F } F F F }

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
706 J. S. MULLIKEN AND D. L. KARABALIS

where
!Khc(i, j) 0 0
[K**] "[K**] " 0 !Krc(i, j) 0
c i,j c j,i
0 0 0
(37)
!Chc(i, j) 0 0
[C**] "[C**] " 0 !Crc(i, j) 0
c i,j c j,i
0 0 0

with Kh and Ch being the horizontal coupling coefficients and Kr and Cr the rocking
c(i,j) c(i,j) c(i,j) c(i, j)
coupling coefficients.
Finally, the time lagging of coupled foundation input is incorporated by the modification of the matrices in
equations (20)—(22). As previously discussed, the first vector in each matrix contains the displacements at all
foundations adjacent to foundation 1 that are lagged by the distances between these foundations and
foundation 1, the second vector provides these values in reference to foundation 2, etc. Similar, consider-
ations should be made for the multiple structure system in order to appropriately include through-the-soil
coupling. Since there are now three DOFs at each foundation supported structure, as identified in Figure
10(a), the following substitutions are made in equations (20)—(22):
y (t !*¹ ) 0 0
1,j k i,j
y (t !*¹ )" 0 y (t !*¹ ) 0
j k i,j 2,j k i,j
0 0 0

yR
(t !*¹ ) 0 0
1,j k i,j
yR (t !*¹ )" 0 yR (t !*¹ ) 0 (38)
j k i,j 2,j k i,j
0 0 0

ÿ
(t !*¹ ) 0 0
1,j k i,j
ÿ (t !*¹ )" 0 ÿ (t !*¹ ) 0
j k i,j 2,j k i,j
0 0 0

where the subscript ‘k’ indicates the current time increment, ‘i’ is the foundation at which we are computing
coupling forces and ‘j’ represents the foundations adjacent to foundation ‘i’. The subscripts ‘1, j’ and ‘2, j’
indicate the horizontal and rocking DOFs of foundation ‘j’.
Next, a numerical example utilizing the above procedure is presented. Three structures, each having the
same stiffness of 66 000 kip/ft (963)2 MN/m), mass of 26 kip-sec2/ft (379441.7 kg), 7 per cent of critical
damping and height above the foundation of 35 ft (10)668 m), are supported on independent rigid surface
foundations with half-widths a"20 ft (6)096 m). The separation between foundations is 10 ft (3)048 m). The
soil medium constants are: shear modulus"360 ksf (17)237 MPa), mass density"3)996 lb-sec2/ft4
(2059)45 kg/m3) and Poisson ratio"1 . To begin the study, sinusoidal acceleration time histories of unit
3
magnitude and frequency range 0—16 Hz, are input to a single foundation—structure system and a three
foundation—structure system, as described above. The peak response at each frequency for both systems is
shown in Figure 11. It is apparent that each of the structures in the three foundation system experiences

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 707

Figure 11. Structural response versus frequency for single- and multiple-building systems

Figure 12(a). Seismic response versus time for a single-building and the end building of a three-building system

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
708 J. S. MULLIKEN AND D. L. KARABALIS

Figure 12(b). Seismic response versus time for a single building and the middle building of a three-building system

motions and ‘resonance’ peaks that are significantly altered by the presence of the adjacent structures.
Although for this particular example this observation appears to be partially explained by the fact that the
coupled dynamic system frequency is near the fixed base natural frequency of the single structure, it also
demonstrates the importance of the FSFI phenomenon.
The time-domain responses of the single- and three-building systems are also computed given an artificial
earthquake input motion with a peak ground acceleration of 0)2g. The response of the single structure is
compared to the end structure of the three-building system in Figure 12(a) and the middle structure in Figure
12(b). It is evident again that large differences in response can be produced due to the presence of adjacent
structures.

CONCLUSIONS
Combining classical foundation—soil interaction models and the coupling models developed in this work, the
foundation—soil—foundation interaction of a number of adjacent rigid surface foundations can be computed.
This approach takes advantage of available data for foundation—soil—foundation dynamic interaction and
includes familiar terminology and methodologies of dynamic lumped parameter systems for incorporating
the dynamic through-the-soil coupling of adjacent foundations and structures. The numerical integration
technique utilized can also be modified to include the effects of non-linear soil and structures. Through
a variety of comparison studies performed in this work the proposed methodology for dynamic founda-
tion—soil—foundation interaction is proven accurate, efficient and applicable to the evaluation of linear and
non-linear systems. A modification of the above basic foundation—soil—foundation model is also used to

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
 DISCRETE MODEL FOR DYNAMIC THROUGH-THE-SOIL COUPLING 709

study dynamic structure-soil—structure interaction phenomena. These studies have demonstrated the direct
applicability of the basic model to more complicated soil-structure systems and the possible importance of
the dynamic through-the-soil interaction phenomenon in structural design.

REFERENCES
1. J. Qian and D. E. Beskos, ‘Dynamic interaction between 3-D rigid surface foundations and comparison with the ATC-3 provisions’,
Earth. Engng. Struct. Dyn. 24, 419—437 (1995).
2. J. Qian and D. E. Beskos, ‘Harmonic wave response of two 3-D rigid surface foundations’, Soil Dyn. Earth. Engng. 15, 95—110
(1996).
3. M. Mohammadi and D. L. Karabalis, ‘Dynamic 3-D soil-railway track interaction by BEM-FEM’, Earth. Engng. Struct. Dyn. 24,
1177—1193 (1995).
4. J. Lysmer and F. E. Richart, ‘Dynamic response of footings to vertical loading’, J. Soil Mech. Found. Div. ASCE 92 (SM 1), 65—91
(1966).
5. R. V. Whitman, ‘The current status of soil dynamics’, Appl. Mech. Rev. 22, 1—8 (1969).
6. F. E. Richart, J. R. Hall and R. D. Woods, »ibrations of Soils and Foundations, Prentice-Hall, Englewood Cliffs, N.J., 1970.
7. N. M. Newmark and E. Rosenblueth, Fundamentals of Earthquake Engineering, Prentice-Hall, Englewood Cliffs, N.J., 1971.
8. G. Gazetas, ‘Analysis of machine foundation vibrations: state of the art review’, Soil Dyn. Earth. Engng. 2, 2—42 (1983).
9. J. P. Wolf, Dynamic Soil—Structure Interaction, Prentice-Hall, Englewood Cliffs, N.J., 1985.
10. J. P. Wolf, Soil—Structure-Interaction Analysis in ¹ime Domain, Prentice-Hall, Englewood Cliffs, N.J., 1988.
11. J. W. Meek and A. S. Veletsos, ‘Simple models for foundations in lateral and rocking motion’, Proc. 5th ¼orld Conf. on Earth. Engng.
vol. 2, 1973, pp. 2610—2613.
12. F. C. P. De Barros and J. E. Luco, ‘Discrete models for vertical vibrations of surface and embedded foundations’, Earth. Engng.
Struct. Dyn. 19, 289—303 (1990).
13. J. P. Wolf and D. R. Somaini, ‘Approximate dynamic model of embedded foundation in time domain’, Earth. Engng. Struct. Dyn. 14,
683—703 (1986).
14. J. P. Wolf and A. Paronesso, ‘Lumped-parameter model for foundation on layer’, Proc. 10th European Conf. on Soil Mech. Engng.,
vol. 1, 1991, pp. 287—290.
15. J. P. Wolf and A. Paronesso, ‘Lumped-parameter model for a rigid cylindrical foundation embedded in a soil layer on rigid rock’,
Earth. Engng. Struct. Dyn. 21, 1021—1038 (1992).
16. J. P. Wolf, ‘Consistent lumped-parameter models for unbounded soil: physical representation’, Earth. Engng. Struct. Dyn. 20, 11—32
(1991).
17. J. P. Wolf, ‘Consistent lumped-parameter models for unbounded soil: frequency-independent stiffness, damping and mass matrices’,
Earth. Engng. Struct. Dyn. 20, 33—41 (1991).
18. G. B. Warburton, J. D. Richardson and J. J. Webster, ‘Forced vibrations of two masses on an elastic half space’, J. Appl. Mech.
ASME 38, 148—156 (1971).
19. G. B. Warburton, J. D. Richardson and J. J. Webster, ‘Harmonic responses of masses on an elastic half space’, J. Engng. Ind. ASME
194, 193—200 (1972).
20. P. B. MacCalden and R. B. Matthiesen, ‘Coupled response of two foundations’, Proc. 5th ¼orld Conf. on Earthquake Engng., vol. 2,
1973, pp. 1913—1922.
21. T. Triantafyllidis, ‘Dynamic stiffness of rigid rectangular foundations on the half-space’, Earth. Engng. Struct. Des. 14, 391—441
(1986).
22. T. Triantafyllidis and B. Prange, ‘Dynamic subsoil coupling between rigid, rectangular foundations’, Soil Dyn. Earth. Engng. 6,
164—179 (1987).
23. F. Guan and M. Novak, ‘Transient response of multiple rigid foundations on an elastic, homogeneous half-space’, ¹rans. ASME 61,
656—663 (1994).
24. F. Guan and M. Novak, ‘Transient Response of an elastic homogeneous half-space to suddenly applied rectangular loading’, ¹rans.
ASME 61, 256—263, (1994).
25. C.-F. D. Huang, ‘Dynamic soil—foundation and foundation—soil—foundation interaction in 3-D’, Ph.D. Dissertation, Dept. Civil
Engng., Univ. South Carolina, Columbia, 1993.
26. D. L. Karabalis and C.-F. D. Huang, ‘3-D foundation—soil—foundation interaction’, in C. A. Brebbia and A. J. Kassab (eds),
Boundary Element ¹echnology IX (BETECH 94), CMP, Southampton, 1994, pp. 197—209.
27. D. L. Karabalis and M. Mohammadi, ‘3-D dynamic foundation—soil—foundation interaction on a layered soil medium’, in B. H. V.
Topping (ed.), Advances in Boundary Element Methods, Civil-Comp Press, Edinburgh, 1996, pp. 73—80.
28. M. Mohammadi and D. L. Karabalis, ‘3-D dynamic foundation—soil foundation interaction on layered soil’, Soil Dyn. Earth. Engng.,
in print.
29. J. Luco and L. Contesse, ‘Dynamic structure—soil—structure interaction’, Bull. Seism. Soc. Am. 63, 1289—1303 (1973).
30. H. L. Wong and M. D. Trifunac, ‘Two-dimensional, antiplane, building—soil—building interaction for two or more buildings and for
incident plane SH waves’, Bull. Seism. Soc. Am. 65, 1863—1885 (1975).
31. H. Murakami and J. Luco, ‘Seismic response of a periodic array of structures’, J. Engng. Mech. Div. ASCE 103, 965—977
(1977).
32. T. H. Lee and D. A. Wesley, ‘Soil—structure interaction of nuclear reactor structures considering through-soil coupling between
adjacent structures’, Nucl. Engng. Des. 24, 374—387 (1973).
33. M. Paz, Structural Dynamics ¹heory and Computation, 3rd edn, Van Nostrand Reinhold, New York, 1991.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)
710 J. S. MULLIKEN AND D. L. KARABALIS

34. J. S. Mulliken, ‘Discrete models for foundation—soil—foundation interaction in time domain’, M.S. ¹hesis, Dept. of Civil Engineering,
Univ. of South Carolina, 1994.
35. J. S. Mulliken and D. L. Karabalis, ‘Discrete model for foundation—soil—foundation interaction’, in A. S. Cakmak and C. A. Brebbia
(eds), Soil Dynamics and Earthquake Engineering »II, CMP, Southampton, 1995, pp. 501—508.
36. D. L. Karabalis and D. E. Beskos, ‘Numerical methods in earthquake engineering’, in D. E. Beskos and S. A. Anagnostopoulos
(eds), Computer Analysis and Design in Earthquake Resistant Structures: A Handbook, Chapter 1, CMP, Southampton, 1997,
pp. 1—102.
37. J. Lysmer, Tabatabaie-Raissi, M. Tajirian, F. Vahdani and F. Ostadan, ‘SASSI—a system for analysis of soil—structure interaction’,
Report UCB/GT/81-02, Geot. Engng, Univ. of California, Berkeley, 1981.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 687—710 (1998)