You are on page 1of 18

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS, VOL.

24, 917-934 (1995)

DYNAMIC ANALYSIS OF LARGE 3-D UNDERGROUND


STRUCTURES BY THE BEM

A. A. STAMOS AND D. E. BESKOS


Department of' Civil Engineering, Uniuersify o f Patras, GR-26500, Patras, Greece

SUMMARY
An up to date literature survey on the dynamics of underground structures is presented briefly. The dynamic response of
large three-dimensional underground structures to external or internal dynamic forces or to seismic waves is numerically
determined by the frequency domain boundary element method. This method is used to model both the structure and the
soil medium, which are assumed to behave as linear elastic or viscoelastic bodies. The full-space dynamic fundamental
solution is employed in the formulation and this requires a free soil surface discretization, confined to a finite portion
around the area of interest, in addition to soil-structure interface and free structural surface discretizations. The dynamic
disturbances can have a harmonic or a transient time variation. The transient case is treated with the aid of numerical
Laplace transforms with respect to time. Various numerical examples involving lined cavities and long lined tunnels
buried in the full- or the half-space subjected to harmonic or transient external forces or seismic waves are presented to
illustrate the method and demonstrate its advantages.

INTRODUCTION
Underground structures, such as tunnels for the improvement and extension of the road and railroad
network, various underground lifeline systems (sewerage, water, gas, electric and telecommunication pipe-
lines), large undergound storage facilities for solid or liquid fuels, undergound structures for the protection of
people (nuclear shelters) or sensitive equipment, as well as underground power plants play nowadays a very
important role in the development of a country. These structures should be properly designed to withstand
safely externally or internally applied dynamic loads or the action of seismic waves. The purpose of this paper
is twofold: first to present briefly an up to date literature review on the dynamics of underground structures
and second and most important to describe a general, accurate and efficient numerical methodology for
determining the dynamic response of large three-dimensional underground structures.
The analysis of underground structures is complicated due to their interaction with the surrounding soil,
especially under dynamic conditions. This probably explains why the existing literature on the dynamic
behaviour of underground structures is not as rich as it is for the case of above ground structures.
Historically, the first studies on the dynamic analysis and design of underground structures had to do with
underground nuclear However, a great interest was created very quickly for the dynamic
(especially seismic) analysis and design of other underground structures, such as tunnels and pipeline^,^-^
which, in general, are less affected by seismic motion than surface structures. More information on the
dynamic analysis and design of underground structures can be found in O k a m ~ t o , Owen ~ , ~ and S ~ h o l l , ~
Agrawal et aI.," Manolis," Manolis and Beskos" and B e ~ k o s . ' l4~ ,
In general, the existing methods for dynamic analysis of underground structures can be placed in the
following categories:

(1) The quasi-static method without soil-structure interaction, which assumes conservatively that the
structure is flexible enough to follow the deformation of the surrounding soil. Thus by using the

CCC 0098-8847/95/060917-18 Received 15 October 1994


0 1995 by John Wiley & Sons, Ltd. Revised 29 December 1994
918 A. A. STAMOS A N D D. E. BESKOS

maximum values of amplitude and wavelength of the soil seismic deformation, the maximum structural
strains can be determined. Kuese14 and Aisiks and Tarshansky” describe the application of this
method to the seismic design of the San Francisco subway, while Sakurai and Takahashi,I6 Newmark’
and Shah and Chu17 apply it to underground pipelines.
(2) The quasi-static method with soil-structure interaction, which can also be applied to stiff structures in
soft soil. Here use is made of the simple beam on elastic foundation model. Using this method Aoki and
Hayashi’’ studied the seismic response of long tunnels, while Wang and O’Rourke,’’ Nelson and
Weidlinger” O’Rourke and El Hmadi2’ and Singhal and Zuroff 2 2 of underground pipelines. More
refined models of this category are those of Constantopoulos et Navarro and Samartir~,’~
Navarro2’ and especially Penzien et which are used for design. Penzien et ~ l . , ’ ~also employ finite
elements under plane strain conditions for the design of the tunnel cross-section.
(3) Experimental dynamic methods, such as those of Young and M ~ r p h y , ’Bustamante,28 ~ Bulson,’’
Bal~ara,~’ Okamoto and Tamura,6 Goto et Howard and I b a n e ~Simonis
, ~ ~ and N a ~ hNash , ~ ~et
Hamada,35Tamura et and Chen et
(4)The dynamic method in conjunction with lumped masses, springs and dashpots for modelling the soil
and discrete or continuous systems for modelling the structure. Here one can mention Dawkins
et ~ l . , ~ ’Yuan and Walker,39 Hindy and N ~ v a k , ~Wong ’ and Weidlinger:’ Krauthammer et aL4” and
Weidlinger and H i r ~ m a in n ~connection
~ with discrete structural modelling and Costantino and Vey,44
Muleski et Muleski and Ariman,46 Takada and Tanabe47 and Manolis and Beskos4’ in
connection with continuous structural modelling.
( 5 ) Analytical dynamic methods, which are confined to problems of pipes or lined tunnels of circular
cross-sections buried in the infinite or semi-infinite elastic space. One can mention here the works of
Garnet and Crou~et-Pascal,4~ Lee and Trifunac,” El-Akily and Datta,” Datta and Shah,52 Wong
et Datta et ~ l . , ’ ~Balendra et OLeary and Datta,56 Datta et and Wong et al.58The
results of these analytical methods for two- and three-dimensional problems are very useful in
understanding the behaviour of underground structures to seismic waves and assessing the accuracy of
the various numerical methods.
(6) Numerical methods requiring an interior discretization of the body analysed, such as the finite
difference method (FDM) and the finite element method (FEM) in the frequency or time domain. These
methods are used both for modelling the structure and the soil in two- or three-dimensional problems
of underground structures. From the plethora of the relevant works, one can mention those of Ang and
N e ~ m a r k ,6o~ Wilkins
~, et and Robinson62 in connection with the FDM and Farhoomand and
Wilson,63 Okamoto and Tamura,6 Goto et ~ l . , ~ Yamada,64 ’ Howard and I b a n e ~ , ~El-Tahan
’ and
R e d d ~Pelz,66, ~ ~ Hwang and L y ~ m e rGomez-Masso
,~~ et a1.,68Nelson and Gong,69 Gomez-Masso
and Attala,70 Monsees and Merrit,71 Chen et Savidis et and N a ~ a r r oin, ~connection
~ with
the FEM.
(7) The boundary element method (BEM), a numerical method requiring, for linear problems, only
a surface discretization of the body. This method can be used for modelling both the soil and the
structure in a frequency or time domain formulation. Two- and three-dimensional problems involving
the dynamic analysis of underground structures have been solved by the BEM, as it is evident in the
works of Manolis and B e ~ k o sKobayashi ,~~ and N i ~ h i m u r a ,K~ ~~b, a~y~a s h i Vardoulakis
,~~ er ~ l . , ~ ’
Kitahara et ~ l . , ~ ’Wang and Banerjee,” Von Estorff et al.,” Stamos and Beskos,82 Luco and de
B a r r ~ s ~and ~ , ’Stamos
~ et aLa5
(8) Hybrid numerical schemes, which combine the FEM for the structure or the structure and a portion of
the surrounding soil (near field) with another numerical (FDM or mainly BEM) or analytical method
for the remaining soil (far field).86Hybrid schemes combining the FEM with an analytical-numerical
method (Gupta et ~ 1 . ’ ~or) an analytical method (Datta et and Wong et ~ 1 . ’ ~or) the FDM (Chen
and Kra~thamrner~’*~’) have been used for dynamic analysis of underground structures. The FEM
has been mostly combined with the BEM in the time domain as it is evident in Von Estorff and
Kausel,” Antes,92 Von Estorff and Antes,93 Von Estorff et aLal and Stamos et ai.,85who have used
this hybrid scheme for tunnels under plane strain conditions. Liu et have used a special
LARGE 3-D UNDERGROUND STRUCTURES 919

FEM/BEM hybrid scheme in the frequency domain for analysing pipes and tunnels under three-
dimensional conditions. Finally one should also mention the FEM/special BEM hybrid scheme of
Underwood and G e e r ~ , ~and
~ . ’Mathews
~ and Geersg7for dynamic analysis of underground cylin-
drical and spherical shells.

From the above eight categories of methods seven of them do account for the important phenomenon of
soil-structure interaction. However, categories (2) and (4) account for soil-structure interaction in an
approximate manner through springs and dashpots, whose characteristics are usually difficult to determine
and primarily used for preliminary design, leaving only five categories which treat the phenomenon in an
exact manner. Out of these categories, experimeaal methods, although invaluable for validation of the
theory, are expensive and time consuming, while analytical methods are of very restricted application. Thus,
numerical methods (categories (6H8) represent the only accurate and inexpensive manner for analysing
realistic dynamic soil-structure interaction problems involving underground structures. Use of domain type
of numerical methods, such as the FDM or the FEM, creates artificial boundaries for the infinite or
semi-infinite soil medium and requires the use of either very large and uneconomical meshes or expensive
non-reflecting boundaries. The BEM, however, requires only a surface discretization, is more accurate, and
takes automatically into account the radiation conditions at infinity.12-14 Thus an efficient and accurate
treatment of dynamic analysis of underground structures can be accomplished by either using the BEM for
both the structure (s) and soil or by using the FEM for the structure and the surrounding soil and the BEM
for the far soil field.
In this paper large underground structures are dynamically analysed in a three-dimensional context
by assuming linear elastic or viscoelastic material behaviour and full contact between soil and structure at
their interface. The term ‘large’ is used here to denote either finite structures with all their dimensions
comparable or very long structures with substantial cross-sectional dimensions. The frequency domain BEM
is used for both the structure and the soil because of the linearity of the problem. Indeed when the
FEM/BEM scheme is used, the symmetry existing in the FEM is destroyed after the coupling, as it is the case
with the BEM used for both the soil and the structure. However, the final size of non-symmetric matrices in
the former scheme is larger than of those in the latter scheme. Besides by coupling the BEM for the structure
with the BEM for the soil, a better matching is achieved at the soil-structure interface than in the case of
using the FEM/BEM scheme. Of course, the latter appears to be better when material or geometric
non-linearities are present. From the extensive list of references just presented, only very few of them deal
with numerical dynamic analysis of three-dimensional linear underground structures, namely, References 67
and 71 employing the FEM and References 83,84 and 94 employing special analytical/BEM and FEM/BEM
schemes, respectively. Material non-linearities considered in References 26,41,65,66,69,73,86,89,90 and 96
have to do only with two-dimensional cases. Thus, in view of the above, this paper presents a numerical
method for linear dynamic analysis of general three-dimensional underground structures, which appears
to be more accurate and efficient than existing ones. The method works in the real frequency domain
when the dynamic disturbance, in the form of external forces or seismic waves, is harmonic or the
complex one (Laplace transformed with respect to time) when it is a transient one. In the latter case
a numerical inversion of the transformed solution is required to obtain the time domain response. For
linear problems, the frequency domain BEM is more advantageous than the time domain BEM, because
the static-like form of the governing equations leads to conceptual and programming simplicity, there
are no causality requirements to be satisfied during the time history and viscoelastic material behaviour can
be very easily incorporated in the formulation. The full-space dynamic fundamental solution is employed
in the present BEM and this requires a certain part of the free soil surface to be discretized in addition
to the inner and outer surfaces of the underground structure. Coupling between the BEM domain corres-
ponding to the soil and that of the structure is accomplished through equilibrium and compatibility at their
interfaces. Quadratic boundary elements, advanced integration schemes, efficient equation solvers, simple
and efficient treatment of corners and edges, symmetry and antisymmetry capabilities and an accurate
numerical Laplace transform inversion algorithm make the proposed method an accurate and efficient
computational tool.
920 A. A. STAMOS AND D. E. BESKOS

BEM FORMULATION AND SOLUTION


Consider an underground elastic structure of an arbitrary three-dimensional shape in an elastic half-space
subjected either to arbitrary external or internal dynamic loads F1 and F 3 , respectively or to plane SH, SV,
P and Rayleigh (R) seismic waves of arbitrary direction and temporal variation, as shown in Figure 1. For the
determination of the dynamic response of the soil-structure system of Figure 1 use is made of the frequency
domain BEM for both the soil and the structure and the resulting equations are coupled together through
equilibrium and compatibility at the soil-structure interface.
The equations of motion of a three-dimensional, homogeneous, isotropic and linear elastic body of volume
V and boundary surface S , on the assumption of zero body forces, have the form
(1+ p ) u i , i j+ pu' 1 . 1"
1 = piij (1)
where ui = ui ( x , t ) is the displacement vector at point x E Vand time t , A and p are the Lame elastic constants,
p is the mass density, commas and overdots denote differentiation with respect to space and time,
respectively, indices i , j take the values 1, 2 and 3 corresponding to the Cartesian directions xl,x2 arid x3,
respectively and repeated indices denote summation. The above governing equations are accompanied by the
initial
Ui(X,O) = uoi, tii(X,O) = uoi; xE v (2)
and the boundary
Ui(X, t ) = u:; x E S,
(3)
ti(X,t) = c'ijnj = t r ; x E s,
conditions, where S, u S, = S, ti is the traction vector, oijis the stress tensor, nj is the unit outward normal
vector on S and u? and t r denote prescribed values of displacements and tractions. Application of Laplace
transform with respect to time to equations (1H3)under zero initial conditions for reasons of simplicity
yields the complex frequency (Laplace transformed) equations
(1+ p ) u i , i j + p u j , i i = ps2uj (4)
G((X, s ) = ;
:il x E S,
- - (5)
ti(X, s) = c 7 i j n j = t?; x E s,
where overbars indicate transformed quantities and s is the, in general complex, Laplace transform
parameter.

Figure 1. Three-dimensional underground structure under dynamic loads and seismic waves
LARGE 3-D UNDERGROUND STRUCTURES 92 1

Using the reciprocal theorem one can obtain an integral representation of the solution of equation (4) in
the form

c i j ( 5 )t7,((, s) = 1 [Uijc(x,
Js
s) - Tijiii(x,s)] dS(x)

where oij
is the fundamental solution tensor of equation (4) for the full-space and Tij is the corresponding
traction tensor obtained from uij through Hooke’s law, point x E S and the coefficient cij = hij if 5 E V or
cij = 0.5 hij if 5 E S and 5 is a smooth point with dij being the Kronecker’s delta. The fundamental solution
oij,
pair ( Tij)is a function of r = Ix - 51 and exhibits in three-dimensions singularities of the type O(l/r) and
O(l/v2) as r --t 0 for Uijand Tij, respectively. Explicit expressions for Uijand Fijcan be found, e.g. in Manolis
and Beskos.’
Solution of the boundary integral equation (6) for 5 E S and cij = 0.5 hij is accomplished numerically by the
BEM. Discretization of the surface S into a finite number of eight-noded quadratic quadrilateral boundary
elements leads to a matrix representation of equation (6) of the form

where {ti} and { t>are the Laplace transformed vectors of boundary nodal displacements and tractions,
respectively and [ 53 and [TI are influence matrices with entries integrals involving the components of the
fundamental solution pair over boundary elements. Details about the computation of these integrals are
given in the next section. The above BEM formulation is appropriate for bodies under external forces (and/or
internal forces if they are doubly connected) and thus equations (6)and (7) can only be used for the structure
of the soil-structure system of Figure 1. They can also be used for the soil of that system, only when subjected
to external forces. However, when there are seismic waves present, the above formulation can be used for the
soil, only in a modified form.
When the soil-structure system of Figure 1 is subjected to seismic waves, the boundary integral equation
for the soil body bounded by surfaces S1 and S2 has the form (Niwa et

where
A i j = hij - cij if 5 E S2
(9)
Aij = O if5ES1
and the total displacements and tractions Ui and ti,respectively, are related to their scattered ti; and $ and
free-field U fand counterparts by
- -
ui = @’+u:, ti = t ; + t-fi (10)
The free-field displacements u f are known for given seismic waves and they actually represent the sum of the
incident and the reflected waves on the free soil surface S , . Explicit expressions for them under harmonic
time variation can be found for SH, SV, P and R waves, e.g. in Reference 12, while the corresponding f f can be
obtained through Hooke’s law. When 5 is an interior soil point, equation (8) is also valid with A,, = 0 and
c.. = 8..
IJ I!’

Discretization of the surface S2 and a finite portion of the surface S1 (around the region of interest, as
shown in Figure 1) into eight-noded quadratic quadrilateral boundary elements and use of equations (10)
enables one to write equation (8) in the matrix form
[ P ] { P }= [ P ] { l P }+ { f }
922 A. A. STAMOS AND D. E. BESKOS

7
where vector { } contains known nodal boundary values associated with the free-field motion, superscript
g stands for ‘ground’ and the remaining matrices and vectors are analogous to those in equation (7) in the
sense that displacements and tractions correspond to the total field.
Having in mind Figure 1, one can rewrite equations (11) and (8) for the soil and the structure, respectively,
in the partitioned forms

where superscript t‘ stands for ‘liner’ and the subscripts 1,2 and 3 correspond to the free soil surface S1, the
ground-liner interface S2 and the inner surface of the liner S 3 , respectively. Coupling of equations (12) and
(13) is accomplished through equilibrium and compatibility conditions at the ground-liner interface, which
read
iC,= - i;, U C2 - u2
-g (14)
Thus equations (12) and (13) with the aid of equations (14), can be combined to yield

Employment of the boundary conditions of the problem, e.g. f! = 0 and ?


: = 0 when there are only seismic
waves present and rearrangement of equations (15) results in

In general, equations (15) after using the boundary conditions and rearranging take the form
CAI {$I = {V (17)
where { $} and { 6) are the vectors of unknown and known quantities and [A] is a full and non-symmetric
matrix but of much smaller size than corresponding ones in the FDM or the FEM for the same problem.
After solving equation (17) all boundary displacements and tractions become known. Thus use of equation
(6) with cij = 6, and equation (8) with cij = 6 , and A , = 0 determines displacements inside the structure and
the soil, respectively, and from there stresses via Hooke’s law. A numerical inversion of the transformed
solution can finally produce the time domain response. If the given loading or seismic waves have
a complicated time variation, their direct Laplace transform has also to be done numerically. More details
about all these computations can be found in the next section.
In case the dynamic disturbance varies harmonically with time, it is more convenient to solve the problem
in the real frequency o domain because no numerical inversion is required. With the simple replacement of
s by io, where i = 0, one can easily go from one domain to the other.I2 Finally, one should stress the
LARGE 3-D UNDERGROUND STRUCTURES 923

fact that the solution of the problem in the frequency (real or complex) domain permits an easy treatment of
viscoelastic material behaviour. Indeed one simply has to replace A and p by their complex values A ( l + 2ip)
and p(1 + 2iB), respectively, where /3 is the hysteretic damping coefficient."

COMPUTATIONAL DETAILS
On the basis of the methodology developed in the previous section, a general and user oriented computer
program in FORTRAN was developed for the dynamic analysis of large three-dimensional underground
structures. This section briefly describes the most important features of this program, such as the computa-
tion of regular and singular integrals, the treatment of corners, edges and discontinuous boundary condi-
tions, the consideration of multiple regions, the possibility of taking into account symmetry and antisym-
metry, the computation of internal displacements and stresses as well as stresses on the boundary, the
solution of a system of linear algebraic equations and the numerical evaluation of direct and inverse Laplace
transforms.
The boundary integrals in equations (6) and (8) are evaluated numerically with the aid of a discretization
involving eight-noded quadratic quadrilateral boundary elements. Thus these integrals have as integrands
products of the displacement U i j or traction Tij fundamental tensor times the shape functions and can be
regular if r # 0 or singular if r = 0 due to the singular behaviour of O(l/r) for Ojj and O(l/r2) for Tij,
respectively. Regular integrals are evaluated numerically by standard Gauss-Legendre quadrature. Singular
integrals are evaluated with the aid of triangle polar coordinates, which reduce the order of singularity by
one, and mappings of the boundary elements onto plane squares, as described in Li et Thus, singular
integrals involving Uij become regular and are evaluated numerically, while those involving Fijbecome of
O(l/r) and are evaluated with the aid of the rigid body motion concept." The presence of non-zero initial
conditions and body forces creates volume integrals in equations (6) and (8), which can be evaluated also
numerically by the method of Li et al.99However, the present program does not have as yet this capability.
When the geometry of the body presents corners or edges, one has three or two, respectively, traction
vectors at a nodal point, as shown in Figure 2, and assuming that only the displacement vector is prescribed
there, the number of unknowns exceeds the number of available equations at that node by six or three,
respectively. The additionally needed equations can be supplied by a method based on the use of tangential
derivatives and Hooke's law, originally proposed for two-dimensional elastostatics by Alarcon et a1.l" In
three-dimensions and with reference to Figures 2(a) and 2(b) one can write at a corner traction-stress

Figure 2. Corners and edges at the boundary with their traction and unit normal and tangential vectors
924 A. A. STAMOS A N D D. E. BESKOS

relations, Hooke's law and strain-displacement relations in matrix form as

where {ti> = ( t l , t 2 , t 3 , t ; , t ; , a22,c33,5 1 2 , a13,023}T,explicit expressions for ma.trices


{aij)= {all,
[ A ] , [ E l and [C] can be found in Starnos"' with i , j = 1 , 2 , 3 denoting the three Cartesian axes and where
Laplace transform overbars have been omitted for simplicity. Combination of equations (18H20) results in
{ti) = CAI CEI [CI {ui,j> (21)
Denoting with T ~7;, and T:' the unit tangential vectors at the corner (Figure 2(b)), one can write
( ~ i , r >= CDI { ui,j } (22)
where { u ~ ,=~ {} u ~ u, ~~, ,u ~~ , .u ~~ ,, ~u ,~, , ~u '~, , ~u '~, , ~u "~, , ~u"~, , , , ,and
} ~ an explicit expression for matrix [D]
can be found in Starnos."' Thus, using equations (21) and (22), the tractions at a corner can be expressed in
terms of tangential displacement derivatives as
{ti} = CAI CEI CCI CDI-'{ui,r} (23)
where the { u ~ ,is~computed
} in terms of the nodal displacements of the three boundary elements at the corner
by appropriate differentiation of the shape functions, as shown in Starnos.'" Thus equation (23) provides the
six additional equations needed at the corner.
For the case of an edge (Figure 2(c)), one can use only three equations from equation (23) with the two
tangential vectors (7' and T " ) to be on the same boundary element. The case of discontinuous traction
boundary conditions can be simply treated by defining two different nodes with the same co-ordinates and
displacements but different known tractions at the point of discontinuity.
The present program can accommodate problems with multiply connected domains, i.e. problems
involving more than one underground structure as well as problems involving a small number of soil layers
on the top of the half-space. Of course, a discretization of the layer interfaces is necessary because of the use of
the full-space fundamental solution in the formulation. This layer discretization as well as that of the free soil
surface is truncated without introducing appreciable errors, especially if a small amount of material damping
is introduced in the soil (Dominguez and Meiselo2).The case of corners or edges in a multiply connected
body is treated by the method previously described for a single body. Even though theoretically, the
additional equations can be derived on the basis of any one of the domains converging to the corner or edge,
use of the domain with an interior solid angle closer to the right one practically leads to better results.lO'
The analysis of symmetric bodies with symmetric or antisymmetric loading, which takes into account
symmetry or antisymmetry in the BEM formulation and solution procedure, can require only a fraction of
the computer memory and computational time needed for a conventional analysis. The basic relations are

for the case of antisymmetry with the superscripts sm and asm denoting symmetric and antisymmetric,
respectively. In three-dimensions one may have symmetry with respect to three planes or a combination of
symmetry and antisymmetry. Use of relations (24) and (25)in the present methodology is done as described in
Stamos.' O
LARGE 3-D UNDERGROUND STRUCTURES 925

Once boundary displacements and tractions have been computed, use of equation (6) in discretized form
with cij = hij and equation (8) in discretized form with cij = tiij and A i j = 0 enables one to compute
subsequently displacements inside the structure and the soil, respectively. Interior stresses in the structure
and the soil can be obtained from the discretized versions of the equations that are produced on the basis of
Hooke's law and the displacement gradients obtained by differentiating equations (6) and (8). These stress
integral expressions involve space derivatives of the fundamental tensors Uijand rij, which can be found in
explicit form in Ahmad.lo3Stresses at a boundary point are computed by the conventional method, which
combines the traction-stress relation, Hooke's law and the expressions for the displacement derivatives in
terms of nodal displacements of the element containing that point."
The system of linear algebraic equations (1 7) has a coefficient matrix [A] which is non-symmetric. This
system can be solved by Gauss elimination with partial pivoting without any problem in a large computer.
However, the size of this system usually exceeds the capacity of the central memory of small computers and
use has to be made of the additional disc memory of the machine. The present program uses an out of core
block equation solver, which employs the Gauss elimination method (forward reduction and back substitu-
tion) to system (17) whose matrix [A]is partitioned into blocks each containing a certain number of columns.
After each elimination of coefficients in a block has been done in the central memory, the resulting submatrix
is stored in the disc. This way the number of transfers between central and disc memories is minimized and
because every column j is wholly into the central memory during elimination, i.e. the coefficients of the
unknown qj of all the equations are into the central memory, the equation with the largest coefficient Aij can
be found to be used for eliminating $ j . This solver is similar to earlier analogous solvers by Lachat and
Watson,lo4Das'" and Crotty.lo6 However, instead of using blocks consisting of either a number of rows,1o4
square sub mat rice^'^^ or rectangular submatrices without zeros,'06 it employs blocks consisting of a number
of columns and thus the pivoting process is optimized and the results become more stable."'
When transient dynamic disturbances are considered, the direct and inverse numerical Laplace transform
has to be used in accordance with the present methodology. Laplace transform is used here rather than
Fourier transform because it can easily take into account initial conditions and eliminates the problem of
fictitious eigenfrequencies because it works in the complex frequency domain. A practical way to avoid the
effect of fictitious eigenfrequencies even in the real frequency domain when dealing with harmonic problems
is to introduce a small amount of viscoelastic damping. In this work the numerical Laplace transform
inversion algorithm of Durbin,lo7which is a combination of sine and cosine finite Fourier transforms, is
successfully employed. The high accuracy of this algorithm is well documented'083'* and its only disadvan-
tage is that it works with complex data that increases the computational cost of inversion. Indeed one has to
solve equation (17) for a sequence of values of s and then obtain the response in discrete form for a sequence
of values of time t . The present program takes advantage of the fact that both the real and imaginary parts of
the transformed solution {@} after about half the number of sampling points vary slowly with s and solves
equation (17) only for some values of s and determines { $} for the remaining ones in that region by linear or
spline interpolation. Because most of the transient loads are complicated functions of time, their direct
Laplace transform has also to be done numerically. This is accomplished again by the algorithm of
Durbin"' as explained in Narayanan and Beskos.'"

NUMERICAL EXAMPLES
Three numerical examples are presented in this section to illustrate the proposed method of dynamic analysis
of three-dimensional underground structures and demonstrate its merits.

Example I
Consider a spherical cavity at a depth of 8 ft (2.438m) in the half-space lined with a concrete shell
of thickness h = 5 ft (1.524 m) and having an inner radius r = 25 ft (7.620 m), as shown in Figure 3. The
soil material constants are modulus of elasticity E, = 4.5 x lo6 psf (215547 x lo6 N/mZ), Poisson's ratio
ug = 0.33 and mass density p, = 3.23 Ibs2/ft4 (1665.452 kg/m3) and the shell material constants
E, = 4.5 x 10' psf (215.547 x 10' N/m2), u, = 015 and pc = 4.66 Ibsz/ft4 (2402.788 kg/m3). The soil surface
926 A. A. STAMOS AND D. E. BESKOS

m,
I
<l-r/20>P(t)

8' (2,438 m>

(1,52411)

Figure 3. Underground spherical shell under transient dynamic load on the half-space surface

,
Uy * 10000 / r
-
~ - -- .

- I I
3( 1 -present -t SAP %- Wang and Baneriee
. _ _ ~ - ---J I

1- __ ---A_- -1 i- -1A--1 J
0 1 2 3 4 5 6 7 a
t /t,

Figure 4. Vertical deflection history of point A of the shell of Figure 3

above the buried shell is loaded axisymmetrically by a load having a spatial and temporal variation as shown
in Figure 3 with t o = 00016 sec. The dynamic response of this buried shell was computed by the present
method. Due to axisymmetry only 1/4 of the system was considered. Thus 1/4 of every shell surface was
discretized into 28 quadratic quadrilateral boundary elements and 1/4or the free soil surface into 56 elements
in a sectorial area with 427 ft (130.15 m) radius. The numerical Laplace transform inversion was done with 20
values of the transform parameter. Figure 4 depicts the history of the vertical displacement at point A of the
LARGE 3-D UNDERGROUND STRUCTURES 921

shell as obtained by the present method as well as by the axisymmetric time domain BEM of Wang and
Banerjee" and the axisymmetric option of the FEM of SAP IV (Bathe et The results show
a satisfactory agreement.

Example 2
Consider a spherical cavity in the infinite elastic space lined by a shell of thickness h = 0 . 0 1 ~ and in-
ner radius a, as shown in Figure 5. The material constants of the soil are E,, u, = 0.25 and pe and those
of the shell E , = 2.5Eg, u, = 020 and pc = 1 . 5 6 ~ A ~ . plane compressional transient wave of the form
uf = (Vo/cl) (clt - x - a) H(clt - x - a) propagates in the soil along the x direction with speed c1 and is
diffracted by the shell as shown in Figure 5. In the above expression for the free field motion uf, Vo is the
velocity jump, H is the Heaviside function and - E defines the point on the x axis where the wave front is
located at time t = 0. Thus the wave reaches the shell at t = 0 and lasts for a time t o so that clto = 10 a. The

Figure 5. Underground spherical shell under transient compressional wave in the full-space

I ''
, I .
~
+ Mathews
F (present)
_ _ _ _
a Geers - 1 Mathews a
T (present)
Geers

~
\ Mathews
E (present)
a Geers 1
1

,
>'

L 1 -1 -.L --1 ---I _--i

0 2 4 6 8 10 12 14 16 18 20
t c, /a
Figure 6. Horizontal deflection history of points F, T and E of the shell of Figure 5
928 A. A. STAMOS AND D. E. BESKOS

above diffraction problem was analysed by the present method in association with a discretization involving
six quadratic quadrilateral boundary elements for every surface (interior and exterior) of 1/4 of the shell and
with a Laplace transform inversion involving 20 sample values. Figure 6 portrays the displacement histories
along the x direction of points F , T and E of the shell as obtained by the present method and by the numerical
method of Mathews and G e e r ~The . ~ ~very good agreement between the two methods is apparent.

Example 3
Consider an infinitely long circular cylindrical tunnel of a concrete liner of inner and outer radii ri and r,,,
respectively and thickness h buried at a depth H = 5 r i in a half-space of hard soil medium subjected to
non-vertically incident (0, = 30") P and SV harmonic waves impinging in the direction of the tunnel
(0, = OO),as shown in Figure 7. The concrete material has a mass density pc = 2.24 x lo3 kg/m3, a modulus

Figure 7. Infinitely long circular cylindrical lined tunnel under P or SV harmonic seismic waves
LARGE 3-D UNDERGROUND STRUCTURES 929

50 100 150 200 250 300 350


8'

ux I 1 Luco and Barros 1

\
1

50 100 150 200 250 300 350


!b)
8"

TII Luco and de Barros

(C)
0
0
,50 100
I

150
1

200
1

250
1

300
8'
350
,

Figure 8. Normalized shell response amplitude to P waves versus polar angle 8: (a) radial displacement U,;(b) axial displacement U,;
(c) hoop stress C~~
930 A. A. STAMOS AND D. E.BESKOS

Luco and de Barros

8 present method
1.2

0 50 100 150 200 250 300 350


(a) e*

0' 1 I I , I I
0 50 100 150 200 250 300 350
(b) 8'

Eee
12
I + Liu at SI
I I8
present method 1 1

I0l A
I
10

0' , , 1 , 1 I

0 50 100 150 200 250 300 350


(C) 8'

Figure 9. Normalized shell response amplitude to SV waves versus polar angle 8: (a) radial displacement U,;axial displacement U,;
(c) hoop stress Xee
LARGE 3-D UNDERGROUND STRUCTURES 931

of elasticity E , = 1.6 x 10" N/m2 and a Poisson's ratio u, = 0.2, while the hard soil medium has correspond-
ingly pe = 2.664 x lo3 kg/m3, E , = 7.567 x lo9 N/m2 and u = 0333. The liner thickness h = O.lri. The
dimensionless frequency q = oro/nc, = 0105, where c, = is the shear wave velocity with p, being the
shear modulus of the soil. Using these data and employing the proposed method with the free-field
displacements for P and SV waves taken from Achenbach,"' the normalized displacement amplitudes
Ur(ro)= lu,/A I for the radial and U x ( r o =) lux/Al for the axial direction on I = r0 and the normalized hoop
stress amplitudes X e e ( U ) = Iaoe/op,c,AI on the centerline r = c1 = 1.05ri of the liner have been computed. In
the above, A denotes the amplitude of the free-field displacement.' l o
Figures 8(ak8(c) depict U,, U , and &, respectively, versus the polar angle 0 for the case of P waves, while
Figures 9(ak9(c) the same things for the case of SV waves. The proposed method was used by taking into
account symmetry about the xz plane (Figure 5) and in conjunction with a discretization involving 32
quadratic boundary elements for half of the free soil surface and 32 such elements for half of each of the two
surfaces of the liner. The above figures also show results due to Liu et and Luco and de B a r r o ~ ~ ~ , ~
obtained by special techniques, which reduce this three-dimensional problem to a two-dimensional one.
Reference 94 employs a FEM/BEM hybrid scheme, while References 83 and 84 the BEM in its indirect form
for the soil and a semi-analytic method for the shell. It is apparent that the present results are in close
agreement with those of Luco and de B a r r o ~while ~ ~ they
, ~ ~disagree with those of Liu et al,94 which seem to
be in error. The accuracy of the results of References 83 and 84 has also been verified by another method due
to Luco and de B a r r ~ s . ~ ~ . ' ~

CONCLUSIONS
A general numerical method for determining the dynamic response of three-dimensional large underground
elastic structures has been presented. The method employs the BEM in conjunction with Laplace transform
for both the soil and the structure(s) and treats both external dynamic loads and seismic waves acting on the
soil-structure system. When the dynamic disturbances are transient, a numerical Laplace transform inver-
sion of the transformed solution is required for obtaining the time domain response. However, when the
disturbances are harmonic, the response is obtained directly in the frequency domain and no inversion is
required. The method is characterized by high accuracy and efficiency and can easily take into account initial
conditions and viscoelastic material behaviour in the soil and/or the structure. Thus it can be used as the
appropriate tool for studying the dynamic behaviour and seismic vulnerability of various underground
structures.

ACKNOWLEDGEMENTS
The authors are grateful to the Greek General Secretariat for Research and Technology (grant 7821/13.5.88)
and the Greek-German bilateral scientific cooperation program for supporting this work. They also
acknowledge the help of Mr. G. D. Pavlatos in connection with Example 3 and the constructive comments of
the two reviewers and Professor G. B. Warburton. Finally thanks are due to Mrs. E. Kefala for her excellent
typing of the manuscript.

REFERENCES
1. M. L. Baron, H. H. Bleich and P. Weidlinger, 'Theoretical studies on the ground shock phenomena', Report No SR-19, The
MITRE Corporation, 1960.
2. ASCE Manual, 'Design of structures to resist nuclear weapons effects', Manual of Engineering Practice No. 42, ASCE, New York,
1961.
3. J. M. Biggs, 'Introduction to Structuraf Dynamics', McGraw-Hill, New York, 1964.
4. T. R. Kuesel, 'Earthquake design criteria for subways', J . struct. diu. ASCE 95, 1213-1231 (1969).
5. N. M. Newmark, 'Earthquake response analysis of reactor structures', in Proc. 1st int. con$ on structural mechanics.in reactor
technology, Berlin, 1971.
6. S. Okamoto and C. Tamura, 'Behaviour of subaqueous tunnels duringearthquakes', Earthquake eng. struct. dyn. 1,253-266 (1973).
7. S. Okamoto, Introduction to Earthquake Engineering, Wiley, New York, 1st edn., 1973: 2nd edn, 1980.
8. S. Okamoto (ed.), 'Earthquake resistant design katures of submerged tunnels in Japan', in Earthquake Resistant Design for Civil
Engineering Structures, Earth Structures and Foundations in Japan, Japan SOC.Civil Engrs, Tokyo, 1980, pp. 189-247.
932 A. A. STAMOS AND D. E. BESKOS

9. G . N. Owen and R. E. Scholl, ‘Earthquake engineering of large underground structures’, Report N o FHWA/RD-80/195, Federal
Highway Administration, Washington, DC, 1981.
10. P. K. Agrawal et al., ‘Seismic response of buried pipes and structural components’, ASCE, New York, 1983.
11. G. D. Manolis, ‘Dynamic behaviour of underground structures’, Shock uibr. digest 15(11), 7-18 (1983).
12. G . D. Manolis and D. E. Beskos, Boundary Element Methods in Elastodynamics, Unwin-Hyman, London, 1988.
13. D. E. Beskos, ‘Boundary element methods in dynamic analysis’, Appl. mech. rev. 40, 1-23 (1987).
14. D. E. Beskos, ‘Wave propagation through ground, in G. D. Manolis and T. G. Davies (eds), Boundary Element Techniques in
Geornechanics, Elsevier Applied Science, London, 1993, pp. 259406.
15. E. G . Aisiks and I. W. Tarshansky, ‘Soil studies for seismic design of San Francisco transbay tube’, in Vibration Eflects of
Earthquakes on Soils and Foundations, STP 450, ASTM, Philadelphia, 1969, pp. 138-166.
16. A. Sakurai and T. Takahashi, ‘Dynamic stress of underground pipelines during earthquakes’, in Proc. 4th world conf: on earthquake
eng., Vol . 2(2), B-4, Santiago, Chile, 1969.
17. H. H. Shah and S. L. Chu, ‘Seismic analysis of underground structural elements’, J. power diu. ASCE 100, 53-62 (1974).
18. Y. Aoki and S. Hayashi, ‘Spectra for earthquake resistive design of underground long structures’, in Proc. 5th world con$ earthquake
eng., Rome, Italy, 1973, Paper No 61.
19. L. R. L. Wang and M. ORourke ‘State of the art of buried lifeline earthquake engineering’, in Proc. current state of lifeline
earthquake engineering, ASCE, New York, 1977, pp. 252-266.
20. I. Nelson and P. Weidlinger, ‘Dynamic seismic analysis of long segmented lifelines’, J. pressure uessel technol. A S M E 101. 1G19
(1979).
21. M. J. O’Rourke and K. El Hmadi, ‘Analysis of continuous buried pipelines for seismic wave effects’, Earthquake eng. struct. dyn. 16,
917-929 (1988).
22. A. C. Singhal and M. S. Zuroff, Analysis of underground and underwater space frames with slip joints’, Comput. struct. 35,227-237
(1990).
23. I. V. Constantopoulos, J. T. Motherwell and J. R. Hall, ‘Dynamic analysis of tunnels’, in W. Wittke (ed.), Numerical Methods in
Geomechanics-Aachen, A. A. Balkema, Rotterdam, 1979, pp. 841-848.
24. C. Navarro and A. Samartin, ‘Simplified longitudinal seismic analysis of buried tunnels’, SoJware eng. workstations 4, 3-10
(1988).
25. C. Navarro, ‘Seismic analysis of underground structures’, in Proc. lOrh world con/: on earthquake eng., Madrid, A. A. Balkema,
Rotterdam, 1992, pp. 1939-1944.
26. J. Penzien, Y. J. Lee and W. Y. Jean, ‘Seismic analysis of rectangular tunnels in soft ground’, in Proc. IOth world con$ on earthquake
eng., Madrid, A. A. Balkema, Rotterdam, 1992, pp. 1619-1624.
27. D. F. Young and G . Murphy, ‘Dynamic similitude of underground structures’, J. eng. mech. dzu. ASCE 90, No. EM3, 1 1 1-1 33
( 1964).
28. J. I. Bustamante, ‘Model study of the dynamic response of the San Francisco rapid transit tube and ventilation caisson’,
Unpublished Report to Parson-Brinckerhofj:Tudor-Bechtel, Inst. of Eng., National Univ. of Mexico, Mexico City, 1965.
29. P. S. Bulson, ‘Blast loading of buried square tubes’, in B. 0.Skipp (ed.), Vibration in Ciuil Engineering, Butterworths, London, 1966,
pp. 224-234.
30. J. P. Balsara, ‘Blast loaded buried arches’, J. eng. mech. diu. ASCE 96, 1-16 (1970).
31. Y. Goto, J. Ota and T. Sato, ‘On the earthquake response of submerged tunnels’, in Proc. 5th world conf: on earthquake eng. Rome,
Italy, 1973, Paper No. 63.
32. G. E. Howard and P. Ibanez, ‘Analytical and experimental investigation of the dynamic response of underground nuclear power
plants’, in T. A. Jaeger and B. A. Boley (eds), Trans. 5th int. con/: on structural mechanics in reactor technology, Berlin, 1979,
North-Holland, Amsterdam, 1979, K 7/5, pp. 1-8.
33. J. C. Simonis and P. T. Nash, ‘Seismic testing of modeled buried pipelines’, in S. K. Datta (ed.), Earthquake Source Modelling,
Ground Motion and Structural Response, ASME, New York, 1984, pp. 175-190.
34. P. T. Nash, R. L. McGuire and A. B. Wenzel, ‘Blast damage of underground piping’, in S. K. Datta (ed.), Earthquake Source
Modelling, Ground Motion and Structural Response, ASME, New York, 1984, pp. 163-173.
35. M. Hamada, ‘Earthquake observation on two submerged tunnels and numerical analysis’, in Proc. 8th world con/: earthquake eng.,
San Francisco, Prentice-Hall, Englewood Cliffs, NJ, 1984, pp. 673-680.
36. C. Tamura, S. Okamoto, K. Kato and Y.Kito, ‘Deformation of tunnel in soft ground during earthquake’, in Proc. 8th world con$
earthquake eng., San Francisco, Prentice-Hall, Englewood Cliffs, NJ, 1984, pp. 71 1-718.
37. H. L. Chen, S. P. Shah and L. M. Keer, ‘Dynamic response of shallow-buried cylindrical structures’, J. eng. mech. ASCE, 116;
152-171 (1990).
38. W. P. Dawkins, ‘Analysis of tunnel liner-packing systems’, J. eng. mech. diu. A S C E 95, 679-693 (1969).
39. H. R. Yuan and R. E. Walker, ‘The investigation of a simple soil-structure interaction model’, in D. A. Howells, I. P. Haigh and
C. Taylor (eds), Dynamic Waves in Ciuil Engineering, Wiley, New York, 1970, pp. 247-266.
40. A. Hindy and M. Novak, ‘Earthquake response of underground pipelines’, earthquake eng. struct. dyn. 7 , 4 5 1 4 7 6 (1979).
41. F. S. Wong and P. Weidlinger, ‘Design of underground protective structures’, J . Struct. eng. ASCE 109, 1972-1979 (1983).
42. T. Krauthammer, N. Bazeos and T. J. Holmquist, ‘Modified SDOF analysis of RC box-type structures’, J. eng. mech. A S C E 112,
726744 (1986).
43. P. Weidlinger and E. Hinman, ‘Analysis of underground protective structures’, J. struct. eng. ASCE 114, 1658-1673 (1988).
44. C. J. Costantino and E. Vey, ‘Response of buried cylinders encased in foam’, J. soil mech.found. diu. ASCE 95, 1159-1 179 (1969).
45. G. E. Muleski, T. Ariman and C. E. Aumen, ‘A shell model of buried pipe in a seismic environment’, J. pressure uessel technol.
A S M E 101,4&50 (1979).
46. G. E. Muleski and T . Ariman, ‘A shell model for buried pipes in earthquakes’, Soil dyn. earthquake eng. 4, 43-51 (1985).
47. S. Takada and K. Tanabe, ‘Three-dimensional seismic response analysis of buried continuous or jointed pipelnes’, J. pressure uessel
technol. A S M E 109, 8G87 (1987).
LARGE 3-D UNDERGROUND STRUCTURES 933

48. G . D. Manolis and D. E. Beskos, ‘Dynamic response of framed underground structures’, Comput. struct. 15, 521-531 (1982).
49. H. Garnet and J. Grouzet-Pascal, ‘Transient response of a circular cylinder of arbitrary thickness in an elastic medium to a plane
dilatational wave’, J . appl. mech. A S M E 33, 521-531 (1966).
50. V. W. Lee and M. D. Trifunac, ‘Response of tunnels t o incident SH waves’, J . eng. mech. div. ASCE 105, 643-659 (1979).
51. N. El-Akily and S. K. Datta, ‘Response of a circular cylindrical shell to disturbances in a half-space: Part I, Theory; Part 11,
Numerical results’, Earthquake eng. struct. dyn. 8, 469477 (1980); 9, 4 7 7 4 8 9 (1981).
52. S. K. Datta and A. H. Shah, ‘Dynamic response of buried pipelines and tunnels’, in S. K. Datta (ed.), Earthquake Ground Motion and
its Eflects on Structures, ASME, New York, 1982, pp. 181-197.
53. K. C. Wong, A. H. Shah, S. K. Datta and P. M. O’Leary, ‘Dynamic amplification of displacements and stresses around buried
pipelines and tunnels’, in S. K. Datta (ed.), Earthquake Source Modelling, Ground Motion and Structural Response, ASME, New
York, 1984, pp. 147-161.
54. S. K. Datta, A. H. Shah and K. C. Wong, ‘Dynamic stresses and displacements in buried pipe’, J . eng. mech. ASCE 110, 1451-1466
( 1984).
55. T. Balendra, D. P. Thambiratnman, C. G. Coh and S. L. Lee, ’Dynamic interaction of twin tunnels’, in A. P. Boresi and K. P. Chong
(eds), Engineering Mechanics in Civil Engineering, ASCE. New York, 1984, pp. 368-371.
56. P. M. OLeary and S. K. Datta, ‘Dynamics of buried pipelines’, Soil dyn. earthquake eng. 4, 151-159 (1985).
57. S. K. Datta, P. M. OLeary and A. H. Shah, ‘Dynamic response of buried pipelines to incident longitudinal and shear waves’, J .
appl. mech. A S M E 52, 919-926 (1985).
58. K. C. Wong, S. K. Datta and A. H. Shah, ‘Three-dimensional motion of buried pipeline I: Analysis, 11: Numerical results’, J . eng.
mech. ASCE 112, 1319-1337; 1338--1345 (1986).
59. A. H. S. Ang and N. M. Newmark, ‘Computation of underground structural response’, D A S A Report N o . 1386. University of
Illinois, Urbana, 1963.
60. A. H. S. Ang and N. M. Newmark, ‘Development of a transmitting boundary for numerical wave motion calculations’, DASA
Report No. 2631, University of Illinois, Urbana, 1971.
61. M. L. Wilkins er al., ‘A method for computer simulation of problems in solid mechanics and gas dynamics in three-dimensions and
time’, Report UCRL-51574, Lawrence Livermore Laboratory, University of California, Pasadena, 1974.
62. A. R. Robinson, ‘The transmitting boundary-again’, in: W. J. Hall (ed.), Structural and Geotechnical Mechanics, Prentice-Hall,
Englewood Cliffs, NJ, 1970, pp. 163-177.
63. I . Farhoomand and E. L. Wilson, ‘Nonlinear finite element code for analyzing the blast response of underground structures’,
Report No. DACA 3 6 5 6 7 4 0 2 0 , Structural Engineering Laboratory, University of California, Berkeley, 1.970.
64. Y. Yamada, ‘Dynamic analysis of structures’, in J. T. Oden, R. W. Clough and Y. Yamamoto (eds), Advances in Computational
Methods in Structural Mechanics and Design, The University of Alabama Press, Huntsville, 1972, pp. 181-199.
65. H. El-Tahan and D. V. Reddy, ‘Seismic response of the cut-and-cover type reactor containments considering nonlinear soil
behaviour’, in T. A. Jaeger and B. A. Boley (eds), Trans. 5th int. con/: on structural mechanics in reactor technology, Berlin, 1979,
North-Holland, Amsterdam, 1979, K7/9 pp. 1-9.
66. R. Pelz, ’A fintie element study on earth covered structures subjected to impact loading’, in G. N . Pande and 0. C. Zienkiewicz
(eds), Soils Under Cyclic and Transient Loading, A. A. Balkema, Rotterdam, 1980, pp. 673-680.
67. R. N. Hwang and J . Lysmer, ‘Response of buried structures t o travelling waves’, J . geotech. eng. div. ASCE 107, 183-200 (1981).
68. A. Gomez-Masso, J. Lysmer, J. C. Chen and H. B. Seed, ‘Soil-structure interaction with Rayleigh waves’, Earthquake eng. struct.
dyn. 11, 567-583 (1983).
69. I. Nelson and C. Gong, ‘Dynamic response of buried concrete arch to blast loading’, in W. F. Chen and A. D. M. Lewis (eds), Recent
Advances in Engineering Mechanics and Their Impact on Civil Engineering Practice, ASCE, New York, 1983, pp. 989-992.
70. A. Gomez-Masso and I. Attala, ‘Finite element versus simplified methods in the seismic analysis of underground structures’,
Earthquake eng. struct. dyn. 12, 347-367 (1984).
71. J. E. Monsees and J. L. Merrit, ‘Seismic modelling and design of underground structures’, in G . Swoboda (ed.), Numerical Methods
in Geomechanics-lnnsbruck 1988, A. A. Balkema, Rotterdam, 1988, pp. 1833-1842.
72. S. A. Savidis, M. Ulrich and H. Klapperich, ‘Seismic loading of underground structures’, in A. S. Cakmak and I. Herrera (eds),
Srructural Dynamics and Soil-Structure Interaction, Comphtational Mechanics Publications, Southampton, 1989, pp. 4 0 3 4 1 3.
73. C. Navarro, ‘Effect of adjoining structures o n seismic response of tunnels’, Int. j . numer. anal. methods geomech. 16,797-814 (1992).
74. G. D. Manolis and D. E. Beskos, ‘Dynamic response of lined tunnels by an isoparametric boundary element method‘, Comput.
merhods appl. mech. eng. 36, 291-307 (1983).
75. S. Kobayashi and N. Nishimura, ‘Transient stress analysis of tunnels and caverns of arbitrary shape due to travelling waves’, in
P. K. Banerjee and R. P. Shaw (eds), Developments in Boundary Element Methods-2, Applied Science, London, 1982, pp. 177-210.
76. S. Kobayashi and N. Nishimura, ‘Analysis of soil-structure interaction by boundary integral equation method‘, in P. Lascaux (ed.),
Numerical Methods in Engineering, Pluralis, Paris, 1983, pp. 353-362.
77. S. Kobayashi, ‘Applications of boundary integrals equation method to geomechanics’, in T. Kawamoto and Y. Ichikawa (eds),
Numerical Methods in Geomechanics-Nagoya 1985, A. A. Balkema, Rotterdam, 1985, pp. 83-92.
78. 1. G . Vardoulakis, D. E. Beskos, K. L. Leung, B. Dasgupta and R. L. Sterling, ‘Computation of vibration levels in underground
space’, Int. j . rock mech. min. sci. geomech. abstrs, 24, 291-298 (1987).
79. M. Kitahara, T. Ando and Y. Muranishi, ‘On the effect of viscosity for the dynamic strain of underground structures’, in M. Tanaka
and Q. H. Du (eds), Theory and Applications of Boundary Element Methods, Pergamon Press, Oxford, 1987, pp. 75-84.
80. H. C. Wang and P. K. Banerjee, ‘Axisymmetric transient elastodynamic analysis by boundary element method’, Int. j . solids struct.
26, 4 0 1 4 1 5 (1990).
81. 0. Von Estroff, A. A. Stamos, D. E. Beskos and H. Antes, ‘Dynamic interaction effects in underground traffic systems’, Eng. anal.
bound. elem. 8, 167-175 (1991).
82. A. A. Stamos and D. E. Beskos, ‘Dynamic soil-structure interaction in 3-D underground structures’, in C. A. Brebbia,
J. Dominguez and F. Paris (eds), Boundary Elements XIV, Vol. 2, Elsevier Applied Science, London, 1992, pp. 169-178.
934 A. A. STAMOS AND D. E. BESKOS

83. J. E. Luco and F. C. P. de Barros, ‘On the three-dimensional seismic response of a class of cylindrical inclusions embedded in
layered media’, in A. S. Cakmak and C. A. Brebbia (eds.), Soil Dynamics andEarthquake Engineering VI, Computational Mechanics
Publications, Southampton, 1993, pp. 565-580.
84. I. E. Luco and F. C. P. de Barros, ‘Seismic response of a cylindrical shell embedded in a layered viscoelastic half-space. I:
Formulation; 11: Validation and numerical results’, Earthquake eng. struct. dyn. 23, 553-580 (1994).
85. A. A. Stamos, 0.Von Estorff, H. Antes and D. E. Beskos, ‘Vibration isolation in road-tunnel traffic systems’, Int. j . eng. anal. des. 1,
111-121 (1994).
86. I. S. Sandler, ‘A method of successive approximations for structure-medium interaction problems’, in: A. J. Kalinowski (ed.),
Computational Methods for Injnite Domain Media-Structure Interaction, ASME, New York, 1981, pp. 67-82.
87. S. GuDta. J. Penzien, T. Lin and C. Yeh, ‘Three-dimensional hybrid modelling- of soil-structure interaction’, Earthquake eng. struct.
dyn. i0, 68-89 (1982).
88. K. C. Wong, A. H. Shah and S. K. Datta, ‘Dynamic stresses and displacements in a buried tunnel’, J . eng. mech. ASCE 111, ;!18-234
(1985).
89. Y. Chen and T. Krauthammer, ‘A combined ADINA-finite difference approach with substructuring for solving seismically induced
nonlinear soil-structure interaction problems’, Comput. struct. 32, 779-785 (1989).
90. Y. Chen and T. Krauthammer, ‘Seismic effects on large reinforced concrete lifelines-I Theory; I1 Implementation’, Compur . struct.
42, 129-135 (1992).
91. 0. Von Estorff and E. Kausel, ‘Coupling of boundary and finite elements for soil-structure interaction’, Erthquake eng. struct. dyn.
18, 1065-1075 (1989).
92. H. Antes, ‘Dynamic interaction analysis in wave propagation problems by a time-dependent boundary element method, in
G. Kuhn and H. Mang (eds), Discretization Methods in Structural Mechanics, Springer, Berlin, 1990, pp. 105-1 14.
93. 0. Von Estorff and H. Antes, ‘Dynamic response of tunnel structures’, in W. B. Kratzig et al. (eds), Structural Dynamics, A. A.
Balkema, Rotterdam, 1991, pp. 803-810.
94. S. W. Liu, S. K. Datta, K. R. Khair and A. H. Shah, ‘Three-dimensional dynamics of pipelines buried in backfilled trencheis due to
oblique incidence of body waves’, Soil dyn. earthquake eng. 10, 182-191 (1991).
95. P. Underwood and T. L. Geers, ‘Doubly asymptotic boundary element analysis of dynamic soil-structure interactioin’, Int. j . solids
struct. 17, 687-697 (1981).
96. P. Underwood and T. L. Geers, ‘Doubly asymptotic boundary element analysis of nonlinear soil-structure interaction’, in R. P.
Shaw et al. (eds), Innooatioe Numerical Analysisfor the Applied Engineering Sciences, University Press of Virginia, Charlot tesville,
VA, 1980, pp. 413-422.
91. I. C. Mathews and T. L. Geers, ‘A doubly asymptotic nonreflecting boundary for ground-shock analysis’, J . appl. mech. A S M E , 54,
489-497 (1987).
98. Y. Niwa, S. Hirose and M. Kitahara, ‘Application of the Boundary Integral Equation (BIE) method to transient response :analysis
of inclusions in a half-space’, Waoe motion 8, 77-91 (1986).
99. H. B. Li, G. M. Han and H. A. Mang, ‘A new method for evaluating singular integrals in stress analysis of solids by the direct
boundary element method, Int. j . numer. methods eng. 21, 2071-2098 (1985).
100. E. Alarcon, A. Martin and F. Paris, ‘Boundary elements in potential and elasticity theory’, Comput. struct. 10, 341-362 (1979).
101. A. A. Stamos, ‘Dynamic response of underground structures’, Ph.D. Thesis, University of Patras, Patras, Greece, 1994 (in Greek).
102. J. Dominguez and T. Meise, ‘On the use of the BEM for wave propagation in infinite domains’, Eng. anal. bound. elem. 8, i32-138
(1991).
103 S. Ahmad, ‘Linear and nonlinear dynamic analysis by boundary element method’, Ph.D. Thesis, State University of New York at
Buffalo, Buffalo, New York, 1986.
104. J. C. Lachat and J. 0. Watson, ‘A second generation boundary integral equation program for three-dimensional elastic analysis’, in
T. A. Cruse and F. J. Rizzo (eds), Boundary Integral Equation Method: Computational Applications in Applied Mechanics, ASME,
New York, 1975, pp. 85-100.
105 P. C. Das, ‘A disc based block elimination technique used for the solution of non-symmetrical fully populated matrix systems
encountered in the boundary element method‘, in C. A. Brebbia (ed.), Recent Advances in Boundary Element Methods, Pentech
.. 391-404.
Press. London. 1978. DD.
,

106. J. M.’Crotty, ‘Ablock equation solver for large unsymmetric matrices arising in the boundary integral equation method’, Int. j .
numer. methods eng. 18, 997-1017 (1982).
107. F. Durbin, ‘Numerical inversion of Laplace transforms: an efficient improvement to Dubner and Abate’s method‘, Comput. j . 17,
371-376 (1974).
108. G. V. Narayanan and D. E. Beskos, ‘Numerical operational methods for time-dependent linear problems’, Int. j. numer. methods
eng. 18, 1829-1854 (1982).
109. K. J. Bathe, E. L. Wilson and F. E. Peterson, ‘SAP IV, a Structural analysis program for static and dynamic response of linear
systems’, Report N o EERC 73-11, University of California. Berkeley, 1973.
110. J. D. Achenbach, Waoe Propagation in Elastic Solids, North Holland, Amsterdam, 1973.

You might also like