You are on page 1of 15

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS, VOL.

23,931-945 (1994)

2-D TRANSIENT SOIL-SURFACE FOUNDATION INTERACTION


AND WAVE PROPAGATION BY TIME DOMAIN BEM
HIROKAZU TAKEMIYA*, GUAN FEI' AND YOSHIKI SUKEYASU:
Department of Civil Engineering, Okayumu University, Tsushima-Naka,3-chome. Okuyama 700, Japan

SUMMARY
This paper shows an effective implementation of the half-plane Green function for surface strip impulses (Lamb's
problem), which was previously developed in a closed form by the authors, into the time-domain boundary element
method for the analysis of related initial boundary value problems. The time-stepping algorithm utilizing Heaviside step
function makes the solution process free from the Rayleigh wave front singularity. Illustrative analyses performed
include that: First, the response due to an impulsive uniform strip loading is dealt with in order to check the accuracy of
the present solution and to interpret the associated wave motion in the medium. Second, a rigid massless strip surface
foundation is analysed when subjected to various impulsive loadings in vertical, horizontal and rotational directions to
observe which wave is most concerned with the respective foundation motion. The field response is also of interest with
respect to distance attenuation.Third, the dynamic cross-interactionbetween active and passive foundationsthrough soil
is investigated when multiple strip foundations are placed separately on a half-space with a certain distance.

INTRODUCTION
The soilstructure interaction system is a mixed boundary value problem in the sense of elastodynamics. The
importance of this problem is first noticed in the analysis of machine foundations in terms of the soiI
impedance function, which induces the inertial coupling with a superstructure, and then the effective input
motion through wave diffraction for the analysis of soil-structure interaction due to incoming earthquake
motions. In view of these dynamic analyses, alternative approaches have so far been undertaken: (1) the
frequency-domain analysis for the steady-state response computation and its inverse Fourier transform by
the fast Fourier transform algorithm for a transient response evaluation; (2) the direct time-domain analysis
by step-by-step time integration. The frequency-domain analysis extracts the dynamic features stated above
as frequency-dependent nature. The time-domain analysis, on the other hand, can facilitate a clear-cut
interpretation of inertial and kinematic interactions with soil by tracing the diffracted waves.
This paper concerns the dynamics of a single or multiple foundations on an elastic half-space. A brief
review was first made of the related topic from the frequency-domain analysis since the inherent nature in the
elastic wave propagation in a half-space is identical regardless of the approach. As regards half-space/half-
plane elastodynamic problems, we must mention the classical paper of Lamb' and the later extended
applications by many researchers. Specifically, the soil compliance function (the inverse of the impedance
function) for a rigid foundation overlying an elastic half-space was formulated rigorously by Thomson and
Kobori2 by using the double Fourier transform method on a scattered wave field and then using a numerical
integration along a single-valued path for the inversion process in order to avoid the branch points and
Rayleigh singularities. Karasudhi et aL3 took the contohr integration approach on a complex plane in
executing such infinite integrals involved over the wave number. Wong and Luco4 employed the subelements
idea within which point loads were assumed in order to account for the traction distribution beneath a rigid

* Professor.
Formerly, visiting research fellow.
Graduate student.

CCC 0098-8847/94/09093 1- 15 Received 18 February 1993


0 1994 by John Wiley & Sons, Ltd. Revised 23 August 1993
932 H. TAKEMIYA, G . FEI AND Y. SUKEYASU

foundation in an approximate manner. The main purpose of their work was to evaluate the frequency-
dependent soil impedance/compliancefunctions in accordance with the degrees of freedom of the addressed
foundations. In view of the reality in which multiple foundations are placed with close spacing, their
cross-interaction through soil becomes important. Among others, Aydinoglu and Cakiroglu' focused such
inertial interaction between two structures, pointing out the shift of fundamental natural frequency due to it.
Triantafyllidis,6 addressing kinematic cross-interaction of three strip foundations, evaluated the resultant
force to act on the respective foundations when a unit harmonic displacement is imposed on one of them. The
results indicate apparent fluctuations representing the diffracted wave field in the underlying half-space. Both
investigations concluded that the effect of the cross-interaction of closely spaced structures through soil
should be taken into account for better estimation of these dynamic behaviours. Wang et ~ l . from , ~ the
analysis of cross-interaction between flexible strip foundations, showed the dependent nature of impedance
on both frequency and their spacing.
Recently, as a methodology for analysis suited to an infinitely extending medium, the Boundary Element
Method (BEM) has been successfully applied to half-space/half-plane elastodynamic problems. Karabalis
and Beskos' were the first to apply the time-domain BEM for the analysis of a rigid surface foundation
placed on a homogeneous elastic half-space. Also, in the same approach, Spyrakos and Beskos' investigated
the response of a rigid/flexible strip foundation. Spyrakos and Antes" demonstrated their comparative BEM
solutions based on different formulations. Antes and von Estorff computed flexibility of the soil to follow
the degrees of freedom of a massless rigid foundation on a half-plane soil or on a uniform stratum. All of the
above BEM were based on the Stokes' solution for an impulsive concentrated (line) force application to
a half-space/half-plane. Takemiya et ul.,' noting the effective use of distributed force for elementwise
convolution integrals in the direct BEM formulation, developed an analytical solution for them from the
integral transform procedure.
In view of the semi-infinite boundary of a half-space /half-plane, developing the relevant Green function for
use in the boundary integral equation for related problems is preferable to employing the fundamental
solution of a full-space and adjusting the free surface condition by some discretization. Incorporating the
half-space/half-plane solution in the BEM has seemed to involve more complexity than utilizing the Stokes'
solution. The use of the latter fundamental solution was considered as less time consuming for the BEM
computation. Quite recently, however, Takemiya and Guan' have developed a closed-form representation
of the half-space displacement Green function for an impulse of uniformflinear intensity distribution over
a strip on the surface. This enables us to compute the stable response quite efficiently. Stepwise approxima-
tion of an arbitrary time-varying excitation enhances its wide application to the related initial boundary
value problems.
In what follows, the aforementioned closed-form Green function is implemented into the time-domain
BEM with stepping algorithm for the transient analysis of soil-structure interaction and illustrative analyses
are performed. First, the surface transient response due to an impulsive uniform strip surface traction on
a half-space is demonstrated in order to check the response feature with respect to its duration, and the
associated wave propagation in the medium is then interpreted. Second, a rigid massless strip surface
foundation is analysed when subjected to various impulsive loads in the vertical, horizontal and moment
directions to observe which wave is most concerned with the respective motion. The field response is also of
interest to find the vibration attenuation from the source location. Third, the dynamic cross-interaction
between active and passive foundations through half-space soil is investigated for the first time directly in the
transient response when multiple foundations are placed side by side with a specified distance.

HALF-SPACE GREEN'S FUNCTION FOR STRIP IMPULSE


Suppose a homogeneous, isotropic and elastic half-space subject to an impulsive strip loading of G(t)-time
variation on the surface, as illustrated in Figure 1. In such a medium, in addition to P and SV wave
propagation of velocity LY and p, respectively, the Rayleigh surface wave is induced whose velocity is denoted
by CR. Takemiya and Guan13 have shown an efficient straightforward inversion procedure for the
LaplaceFourier Transform method to obtain the relevant closed-form time-domain Green function at any
SOIL-SURFACE FOUNDATION INTERACTION 933

Figure 1. Impulsive strip traction on a half-space (Lamb's problem)

observation point in the medium. Surface response due to an arbitrarily distributed traction application can
be expressed as

Functions xl(x, t, q) (2 = 1,2) are related to the spatial distribution of the traction and to the way of
application over the width 2w. Specifically,in case of impulsive uniformly distributed horizontal and vertical
tractions of unit intensity with the centre being located at the origin, x's are expressed as

Equation (2a) indicates that the above Green function has a logarithmic (In)singularity at the Ray.-@ wave
front. Based on the impulsive loading, the response for an arbitrary time-varying surface traction defined as
T(t),the functions x,(x, t, q) ( I = 1,2) in equation (1) are replaced by the convolution integral of
934 H. TAKEMIYA, G. FEI A N D Y. SUKEYASU

The In-type singularity in equation (2a) can therefore be removed analytically or numerically through the
time-convolution integral of equation (3).
The computation time is split into increments of equal duration At. When traction variation is assumed to
be constant in each time interval [kAt, ( k + l)At], then
T(t) = T k ( H ( t- kAt) - H ( t - ( k + l)At)), kAt < t < (k + I)At (4)
in which T k is the intensity at the kth time step and H ( t - z) is the Heaviside step function defined by

H ( t - z) =
i 1 t>z
0 t d z
Substituting equations (2) and (4) into equation (3) leads to
j+ 1
jl(x,(k + l)At,q) = 2t Ck T'k-j' 1 ( - 1 ) .
-
j=O ;=j
J+1+1
'{ ( q t i + w +x)ln(qt; + w +x) -(qti - w -x)ln(qti - w -x)

+ (qti + w x) In (qti + w x) (qt; w + x)In (qti - w + x)}


- - - - (64

in which t k + = ( k + j)At. Note that equation (6a) excludes the wave front singularity.
For computational reasons, the semi-infinite integrals involved in equation (1) can be performed without
difficulty after converting them into the definite integrals by taking a new variable q' = l/q. The numerical
integration scheme proposed by Clenshaw and Curtis14 is employed in this paper, in which the integrand is
expanded into the Chebyshev series and the range of integration over each segment is adjusted to [ - 1,11.

TIME-DOMAIN BEM IMPLEMENTATION


Soil-structure interaction, which constitutes a mixed boundary value problem, is formulated by the direct
time-domain BEM. The substructure method is applied for the analysis. Here, the surface foundation and the
half-space beneath it are addressed for the analysis. The boundary integral equation in use of the Green
function, fulfilling the half-space condition, is given for surface problems by

in which x is a focus position and y is a field position, U,, denotes the displacement Green function for a unit
impulsive point force application to a half-space medium, u, and t, denote, respectively, the interface
displacement and traction and subscripts c1 and indicate the direction of the displacement response and the
traction exertion respectively. Equation (7) means that the discretization at the interface, bringing in nodes
only along the interface to account for the given interface condition, suffices the analysis unless free field
response is additionally expected.
Displacement and traction in the eth element are approximated by an introduced interpolation function
for the concerned nodal values. When isoparametric elements are used with interpolation function + ( 5 ) for
SOIL-SURFACE FOUNDATION INTERACTION 935

the space variable, these are expressed as


M+l
{ue(T,t),te(T,01 = 1 4,"<r,(U"*"(t),
m= 1
T'." (t>> (8)

in which each element of length 2ae has (M + 1) numbers of points placed within it and 5 is the local
co-ordinate attached to that element. Also, the time discretization is conducted, by introducing the
interpolation function $ ( t ) for displacement, as

{Ue*'"(t), = C {$;(t)Uesm,k,
Tepm(t)) $;(t)Te,m*k) (9)
k

These discretizationsimply that the displacement and the associated tractions are defined at all nodes both in
space and time.
= { u : , q , . . . ujXujz
{Uk}= k k . . . u ~ ~k u ,T,
,} (Tk}' = { tkl x tkl , . . . t kj x t kj , . . . t,,t,,}
k k T (10)
in which the quantities with superscript ( k ) indicate those values during the time interval [kAt,(k + l)At].
Substituting equations (8) and (9) into equation (7) yields

k JO e rn J-a,

Selecting position x at the respective nodal points (collocation method), we can evaluate a flexibility matrix
G(t) composed of the displacement Green function for a system of distributed loading.

G(tk) = Gk =

When the time-varying tractions are prescribed stepwise by $,(t)Tk = [ H ( t k ) - H ( t k - At)]Tk, the trans-
formation property of the convolution integral and decoupling of the preceding solution from future ones are
applicable. Subsequently, we get a time-stepping algorithm for the governing equation, then

k = 1 J -AI
K-1
=GOTK+ (GK-kTk)
k= 1

so that the associated discretized traction can be expressed as

In case of rigid foundations, the degrees of freedom can be expressed in terms of horizontal displacements
A;(n = 1,2, . . . , n), vertical displacements A:, and rotation angles 4'' at these centres (Figure 2), so that the
interface nodal displacements are
u:(x, t ) = A:(t) u,"(x,t) = A,"(t) + O"(t)x
or in a matrix form
936 H. TAKEMIYA, G . FEI A N D Y. SUKEYASU

Figure 2. Dynamic cross-interaction of multiple rigid strip foundations through a half-space soil

>'
in which {A} are the displacements of the total foundations {A} = {. . . {A:A;O} . . . and [D] is a trans-
formation matrix from the degrees of freedom of those rigid bodies to their interface nodal displacements
[D] = diag[D"]. The submatrix due to the nth foundation of rn-subelements is given by

(16)
0 x2 0 x,
Variable x, is measured as the horizontal distance in the local co-ordinate of the mth element with its origin
located at the centre of the foundation it belongs to.
The associated resultant horizontal force and the moment equilibrium are then obtained by summing up
the element-wise traction contribution of the respective foundations, together with the inertial forces of mass
M" and mass moment of inertia I ksuch that
N
p:(t) - M"A: = 1 wt::(t)
ne= 1

N
p;(t) - M"AZ = wt::(t)
ne= 1

p&(t) - I"$" = c
N

ne= 1
2wt::(t)xne

Substituting equation (13) into equation (17) and viewing equation (1 5 ) lead to the governing equation.
Under a certain incoming motion uf, it should be added to the foundation response of equation (9). Hence,
K-1
[MI {A"] + [KO] {A"} = {P"]+ [D]TIGO]-l
{I$} +C [G"-'] (Tk) (1 8)
k=l

in which [MI stands for the mass matrix, and [KO] = [DIT [ G o ] - ' [D] defines the instantaneous stiffness
matrix. Solution of equation (18) can be obtained by numerical integration, for instance, the Newmark
numerical integration algorithm.

NUMERICAL COMPUTATION
Addressing attention to initial boundary value half-space problems, the following illustrative analyses are
performed: (1) the response under the application of surface impulsive traction; (2) the response of a massless
rigid surface foundation under various loads; (3) the associated field surface response at different distances; (4)
the kinematic interaction response of multiple rigid surface foundations through soils.
SOIL-SURFACE FOUNDATION INTERACTION 937

Surface response of a half-space under impulsive uniform strip surface tractions


For the purpose of validation of the present solution method and investigation of the time-stepping width,
considered first are the responses of a half-space due to a vertical/horizontal uniform impulsive strip load
acting for a short duration. The medium has the properties of Young's modulus E = 1.724 x lo7 kN/m2,
mass density p = 30 890 kg/m3, and Poisson's ratio v = 0.25, giving the P wave velocity a = 3788 mfs and
S wave velocity /3 = 2187 m/s. Different normalized time steps AT = aAt/x, 5A7, 10A.rare employed in order
to observe its effect on transient response. The representation for the response, i.e., ( n p / B ) U i jvs. time T,
denoting the horizontal distance of the field point from the source by x and shear rigidity by p, follows the
normalization that Forrestal et al.' took for vertical line loading. Surface responses are depicted for various
distances against the strip width ratio x / w as a parameter. We notice that as the ratio x/w is increased, the
solution approaches that for a line loading. Solution for x / w = 100 results in almost the same responses as
those for line loading. In this case, when a very short impulse duration AT = 1.074 x lo-' is chosen, small
fluctuations appeared in the response due to the computation problem, although the actual values are very
small. For strip loading, body wave branch points and Rayleigh pole singularities are excluded so that
a stable solution is easily obtained. In contrast to the responses for 6(t)-typeimpulsive loading which has the
Rayleigh wave singularity," a drastic change can be observed in that the Rayleigh wave arrival shows
a bounded smoothly varying feature, especially for a large ratio of x/w. Regarding time step for response
computation in conventional BEM utilizing the so-called fundamental (full-space) solution, Rayleigh wave
field is not fulfilled originally in the kernel function but only after solving the discretized BEM model.
Therefore, the criteria on the node placing and suitable time stepping are claimed for better computation
from the wave propagation point of view. In the present method, however, the non-dimensional time step for
the response computation may be taken small enough to be able to satisfy causality and propagation for P,
S and Rayleigh waves within the expected accuracy. The features in Figures 3 are of help for substantiating
the discretization of arbitrary time variation of loading for response computation.

Responses of a rigid surface foundation under various loadings


Consider an isolated single rigid strip foundation, as illustrated in Figure 4, of width 2B = 2.0 m placed on
a surface of a homogeneous elastic half-space. Soil properties are assumed for shear modulus
I( = 1.0x lo5 kN/mZ, mass density p = 2000 kg/m3 and Poisson ratio v = 0.33. Shear wave velocity then
corresponds to /3 = 223.6 m/s and pressure wave velocity to a = 443.9 m/s. This model was used by Antes
and von Estorff." Following their discretization of a rigid foundation, eight elements are taken here also for
it. Perfect bond condition is assumed at the soil-foundation interface. Impulse load is applied to the
foundation in the directions of the vertical, tangential and rocking motions for the duration
At = 5-63x s. Intensities are taken as P , = P , = 1776 kN/m for the first two loadings and
M y= 1776 kN for the third loading.
Computation results are depicted in Figure 5 for the response quantitiesf, = Cu,,fi = Gu,, f+ = GE2u+ vs.
dimensionless time z = tB/B in order to compare these with results by the above authors, who utilized direct
BEM solutions with full-space Green's function. These two results compare well; however, computing time is
judged to be much less by the present procedure than by the latter.

Soil surjhce responses through a loaded rigid foundation


Soil surface responses associated with the above foundation loading are of interest in this investigation,
focusing wave attenuation with the distance from the source. Computation results are depicted in Figures
6(a) and (b) for those close to the foundation and for those far from it. Normalization is taken for the
presentation in the form of (zpB/B)Uii( i = x, z ) for the response, and z = Bt/x for the dimensionless time, so
that we can observe causality of the wave propagation and attenuation by distance. Loading on a rigid
massless foundation gives rise to a mixed boundary value problem, which modifies soil traction distribution
beneath it differently from that for a uniform loading directly on the soil surface. In Figures qa) and (b) we
can notice that the wave field close to the foundation is dominated by body waves; namely, in the vertical
response due to vertical loading the P wave contribution is significant, and in the horizontal response due to
0

8l*
II
e

I
C

1
09'E
00'9 ,
09'E-
02" 02"-
00.3-o
0
/
09'E 02'1-
!
09'E- 9
I I I I
02' I
-7-
00'

II
c1

0
04. I 06'0 OE'O OE'O- 06'0- OS'I-o
940 H. TAKEMIYA, G. FEI AND Y. SUKEYASU

t At = 5.63 x lo4 [s]

P,=Ps 1776 "/m]

Figure 4. Computational model for a single foundation on a half-space

8 Z= tP/B
qal , 1.p -
/ --- L __ 4.m 5.00

VERTICAL
'
L HORIZONTAL
ROTATIONAL
-
--- present solution
Antes and von Estorffll
8-

Figure 5. Transient responses of a massless rigid foundation due to vertical, horizontal and rocking loading independently

horizontal loading, the S wave is appreciable, whereas at the far field the response is dominated by the
Rayleigh wave. Hence, in the vicinity of the foundation, the response due to loading through the rigid
foundation has a significantly different feature from that for uniform traction loading (see Figure 3).
Attenuation by distance is clearly noted. However, at a far distance, the feature becomes quite similar to that
for a uniform loading and the Rayleigh wave propagates with no attenuation by distance.
SOIL-SURFACE FOUNDATION INTERACTION 941

\
01

?! I I 1 1 I I I 1 1 I I I I I I I I I I I I I I I 1
0.w 0.20 0.40 0.60 0.m 1.00 I.M 1.40 1.60 1.80 z.m 2.20 2.40

z= pt/x
Figure 6(a). Transient vertical surface responses due to vertical loading through a massless rigid foundation

t.w ' 0.L ' 0.h ' 0.M ' 0.m ' 1.L ' I.& I.iO ' 1.M I 1.m ' 2.b ' 2.A ' 2.b '
z = pt/x
Figure 6(b). Transient horizontal surface responses due to horizontal loading through a massless rigid foundation

In contrast, the above features of the wave motion help in understanding the kinematic interaction among
multiple foundations through soil. It is emphasized that the present Green function approach has an
advantage over the conventional BEM solution method in computing surface response at any distance
without significantly increasing computer storage for free field consideration and thus reducing computing
time.

Kinematic interaction among multiple foundations


An isolated structural model is an idealization of reality. Usually buildings are constructed at a site
separately but in close proximity to each other. Under such situations, coupling among buildings may exist
through soil. Here, in order to study such interactions, two or three foundation systems are investigated
under a unit impulse on one of them. Relative distance is specified by parameter D = 4 2 4 normalizing the
distance d between two adjacent foundations by the foundation width 2B. (See the illustration in Figures
7 and 8). Half-space soil is characterized by Young's modulus ,u = 1.0x lo6 kN/m2, density p = lo00 kg/m3
942 H. TAKEMIYA, G. FEI AND Y. SUKEYASU

ni D -0.5
1' 1

Y.

'7
0
2 s
I0
0
0 N
0 0
0.00 0.40 0.80 I .20 I .60 2. 0.00 0.40 0.90 I .20 I .GO 2
I(sec.x5xlO *) (SPC ~5x10
')

(a) Horizontal response of left foundation (b) Horizontal response of right foundation

c_J I.,

W
0 0
01
N
0 o I D-0.5 I I' ,I At=5X lo4

W
0

-F
v
0

0"
Y f p
9
0

N
0
0

8
0.
".-,,
-
0.00 0.40
'
0.80
~-
I .20
(Jec.xSx 10')
I .60 2.00
1 ;# 9
0.00
1 - 7
0.40
I
0.80
1
I .2u
r (.rec.x5x lo.')
I l
I .60
I
2.1
1

(c) Vertical response of left foundation (d) Vertical response of right foundation

Figure 7. Transient interaction responses of a massless two-foundation system through a half-space soil

and Poisson's ratio v = 1/4. A massless foundation of width 2B = 5 m is used. In the following numerical
computations, each foundation is discretized into eight subelements and time step is chosen as
At = 5 x s.
Numerical results are depicted in Figures 7(at7(d) for a two-foundation system and in Figures S(a)-S(f) for
a three-foundation system. In these figures, it is very clear that the coupling through soils depends on the
SOIL-SURFACE FOUNDATION INTERACTION 943

dimensionless distance D.This is interpreted in the sense of the propagation of P, SV and Rayleigh waves.
The response of a loaded (active) foundation, which is indicated by an arrow, shows some small fluctuations
caused by the waves emitted from it and reflected back to it from the adjacent foundation, which is called
herein the 'return wave motion'. This is noted to be appreciable for horizontal loading in the two-foundation
system, but it is less discernible for other cases. The response of a non-loaded (passive) foundation, indicated

u -0.5

1 1 1 1 1 1 1 1 1
00 0.40 0.e0 1.20 I .60 2.
r (SCC.XS~IU~*) I (sec.x5x10-*)

(a) Horizontal response of left foundation (b) Horizontal response of middle foundation

1 (.ccc.x5x1U-*)

(c) Horizontal response of right foundation (d) Vertical response of left foundation
Figure
944 H. TAKEMIYA, G. FEI AND Y. SUKEYASU

J h

N
0

m
I
' - I I I I
o-I--, , I
I
I
I 1 __
10 0.40 0.80 I .20 1 .60 2. 1 0.00 0.40 0.80 I .20 I 60 2.
r (5ec.x5x10-2) I (sec.xSx I o *)
(e) Vertical response of middle foundation (0 Vertical response of right foundation
Figure 8. Transient interaction response of a massless three-foundation system through a half-space soil

by a symbol 7 , shows the effect of radiating wave motion, which is called direct contribution. For vertical
loading situation of the two-foundation system, the response of a passive foundation in the proximity of the
active foundation, say D < 2, is noticeable due to the shear wave propagation. However, as the distance of
the focused passive foundation increases away from the loaded foundation, say D > 5, the response feature is
due mostly to the Rayleigh wave. For horizontal loading situation of the same foundation system, the
response of a nearby passive foundation, say D < 2, is dominated by the P wave and also for the distance, up
to I) = 5, and the Rayleigh wave contribution becomes comparable with the P wave at D = f 0. It can also be
found that the Rayleigh wave, which is induced by horizontal loading, indicates different features than those
induced by vertical loading. In view of the inset figures in Figures 7 and 8, the correspondence between the
passive foundation of the two-foundation system and the farthest passive foundation of the three-foundation
system is given by (20 + 1) for the same distance from the active foundation. The response comparison
between these passive foundations suggests that in the case of three foundations, the presence of the middle
passive foundation has almost no effect on the response of the farthest passive foundation. It also implies that
the response of the middle foundation is not affected by the passive foundation located on the opposite side of
the active foundation.

CONCLUSION
In this paper, the closed form Green function of a half-space (Lamb's solution) for surface strip loads, which
was previously developed by the authors, was implemented into the direct time-domain boundary element
method for the analysis of related initial boundary value problems to demonstrate its effective application.
Illustrative computations for a half-space under an impulsive exertion of surface traction and for a rigid
overlying massless foundation under such various loading revealed the features of the associated wave
motion and the detailed interpretation was made for them. Further, this application to the kinematic
interaction analysis of multiple foundations through soil indicates some useful engineering findings. The
return contribution from the wave radiated from the active foundation and hit back to it from the passive
SOIL-SURFACE FOUNDATION INTERACTION 945

foundations is negligibly small, while the direct radiating effect from the active to the passive foundations is
appreciable when they are placed close to each other. Indirect contribution from the first passive to the
second passive foundation is also negligibly small. Hence, we only have to take into account the two
foundations interaction through soil, even for multiple foundations.
Since the present procedure evaluates the transient response directly, it may be extended straightforwardly
to the non-linear soil-structure interaction of interface slip or separation. Although the surface response is
only addressed herein, an extension to the internal response problems is very possible.

REFERENCES
1. H. Lamb, ‘On the propagation of tremors on the surface of an elastic solid‘, Phifos. trans royal SOC., London ser. A 203, 1-42 (1904).
2. W. T. Thomson and T. Kobori, ‘Dynamic compliance of rectangular foundations on an elastic half-space’, J . appl. mech. ASME 30,
579-584 (1963).
3. P. Karasudhi, L. M. Keer and S. L. Lee, ‘Vibratory motion of a body on an elastic half plane’, J. appf. mech. ASME 35, 697-705
(1968).
4. H. L. Wong and J. E. Luco, ‘Dynamic response of rigid foundations of arbitrary shape’, Earthquake eng. struct. dyn. 4, 579-587
(1976).
5. M. N. Aydinoglu and A. Cakiroglu, ‘Dynamic interaction between soil and a group of buildings, Proc. 6th world conJ on earthquake
eng., New Delhi, 15961601 (1976).
6. Th. Triantafyllidis, ‘An analytical method to determine the subsoil-coupling between rigid strip foundations bonded to the surface of
a half-space subjected to harmonic excitation’, Ingenieur-Archiu, 52, 145-157 (1982).
7. Y. Wang, R. K. N. D. Rajapakse and H. Shah, ‘Dynamic interaction between flexible strip foundations’, Earthquake eng. struct. dyn.
20, 44-454 (1991).
8. D. L. Karabalis and D. E. Beskos, ‘Dynamic response of 3-D rigid surface foundations by time domain boundary element method,
Earthquake eng. struct. dyn. 12, 73-93 (1984).
9. C. C. Spyrakos and D. E. Beskos, ‘Dynamic response of rigid-strip foundations by a time domain boundary element method’, 1nt.j.
numer. methods engng 23, 1547-1565 (1986).
10. C. C. Spyrakos and H. Antes, ‘Time domain boundary element method approaches in elastodynamics: A comparative study’,
Comput. struct. 24, 529-535 (1986).
11. H. Antes and 0. von Estorff, ‘Dynamic response analysis of rigid foundations and of elastic structures by boundary element
procedures’, Soil dyn. earthquake eng. 8, 68-74 (1989).
12. H. Takemiya, C. Y.Wang and A. Fujiwara, ‘2-D elastodynamic fundamental solution for distributed loads and BEM transient
response analysis of halfplane problems’, Struct. eng./earthquake eng. JSCE 10, 23s-33s (1993).
13. H. Takemiya and F. Guan, ‘Transient Lamb’s solution for surface strip impulses’, J. eng. mech. diu. ASCE (1993) (in print).
14. C. W. Clenshaw and A. R. Curtis, ‘A method for numerical integration on an automatic computer’, Nwner. math. 2,197-205 (1960).
15. M. J. Forrestal, L. E. Fugelso, G. L. Neihardt and R. A. Felder, Proc. eng. mech. din spec. con$, ASCE, 719 (1966).

You might also like