Two-Dimensional Time Domain El

INFORMATION TO USERS
This manuscript has been reproduced from die microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may
be from any type of computer printer.
The quality of this reproduction is dependent upon the qnality of the

copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize m aterials (e.g., maps, drawings, charts) are reproduced by

sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in
reduced form at the back of the book.
Photographs included in the original manuscript have been reproduced

xerographically in this copy. Higher quality 6” x 9" black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly
to order.
University Microfilms International

A Bell & Howell Information C om pany
300 N orth Z ee b Road. Ann Arbor. Ml 48106-1346 USA
313/761-4700 800/521-0600
O rder N um ber 0230380
Two-dimensional tim e dom ain elastodynam ic displacem ent

discontinuity m ethod w ith m ining applications
Siebrits, Eduard, Ph.D.

University of Minnesota, 1092
C opyright ©1992 by S iebrits, E duard. All rig h ts reserved.
UMI
300 N. Zeeb Rd.
Ann Aibor, MI 48106
TWO-DIMENSIONAL TIME DOMAIN
ELASTODYNAMIC
DISPLACEMENT DISCONTINUITY METHOD
WITH
MINING APPLICATIONS
A THESIS
SUBMITTED TO THE FACULTY OF THE GRADUATE SCHOOL
OF THE UNIVERSITY OF MINNESOTA
BY
EDUARD SIEBRITS
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
FEBRUARY 1992
UNIVERSITY OF MINNESOTA
This is to certify that I have examined this bound copy of a doctoral

thesis by
Eduard Siebrits
and have found that it is complete and satisfactory in all respects, and
that any and all revisions by the final examining committee have been
made.
Steven L. Crouch
Name of Faculty Advisor
OhOi/C/li
Signature of Faculty Advisor
Ff/k. l2>. 1^92/

Date
GRADUATE SCHOOL
Copyright Eduard Siebrits 1992
"Shall 1 refuse my dinner because 1 do not fully
understand the process of digestion?”
0. Heaviside1
"Neither seeking nor avoiding mathematical exercltations

we enter into problems solely with a view to possible
usefulness for physical science."
Lord Kelvin and Peter Guthrie Talt1
Karman, T. and Blot, H.A. 1940. Mathematical Methods In Engineering.

New York: McGraw.
GO AHEAD,
MAKE ONE MORE CHANGE
Acknowledgements
The work described in this thesis forms part of the Chamber of Mines of
South Africa Research Organization's program into non-linear rock
deformation mechanisms, and was sponsored by the South African Gold
Mining Industry's Specialist Post-Graduate Training Scheme, during the
period 1988 to 1992.
I would also like to thank:
Prof. Steven Crouch, my advisor, for all the "big picture" talks, for
your excellent guidance, and for providing me with luxurious office
space and a 486 computer.
Dr. John Napier, for your enthusiastic support, encouragement, and

numerous suggestions during the course ofmy research.
Dr. You Tian, for helping me with numerous aspects of the two-
dimensional work.
Dr. Nick Gay, my mentor, for believing inmy abilities, andsupporting

me during the period 1981 to 1991.
Dr. John Ryder, for Introducing me to boundary element methods in 1985,

and for all your practical and theoretical insights over the years.
Dr. Anthony Peirce, for tirelessly helping me with some of the

mathematical details of the transform approaches.
Marc Loken and You Tian, for your friendship, and for the never-ending
discussions about all aspects of our research efforts.
Matthew Handley and Nic Lightfoot, for your friendship, and for taking
me underground to show me what I was supposed to be modelling in the
first place.
John Crouch, for providing all the excellent post-processing graphics

capabilities, except for the contour plots, which were provided by
Mining Stress Systems (Pty) Ltd and GeoLogic Research Inc.
Space Science Graphics at the Space Science Center, University of

Minnesota, for drawing all the figures.
(i)
Abatras.t
Two versions of a cine domain elastodynamic displacement

discontinuity method are developed for analyzing two-dimensional
problems in solid mechanics. Linear, continuous in time and piecewise
constant in space interpolations are assumed for one version, and
linear, continuous in time and piecewise linear in space interpolations
are assumed for the other. Analytical integrations in space and time are
used throughout. Both formulations are time-marching.
The displacement discontinuity methods are developed to model a
class of transient stress boundary value problems, where the unknowns
are transient discontinuities in displacement across known boundaries.
The boundaries typically represent the surface of a crack or an
extensive tabular excavation, in an infinite, isotropic, and homogeneous
medium.
The methods are demonstrated for several geomechanics problems of
practical interest in deep.underground mining. These mining problems
detail the interactions that occur between elastodynamic waves and
tabular mining excavations. In particular, eddy currents are generated
by both shearing and normal wave sources. The compressional and shear
wave fronts and the eddy currents cause rapid changes in stress, and are
associated with higher particle velocities than the surrounding areas.
This work is the first general elastodynamic displacement
discontinuity method developed for solving two-dimensional problems, and
lays the foundation for the development of special elements to model
frictional slip behavior and the automatic growth of cracks.
(ii)
TABLE OF CONTENTS
Acknowledgements ............................................ (1)

Abstract .......................... (li)
1 INTRODUCTION ............................................. 1
2 LITERATURE REVIEW ........................................ 5
3 THEORY: TWO-DIMENSIONAL FORMULATIONS ....................... 9

3.1 Introduction ..................................... 9
3.2 Mathematical Derivation........................... 10
3.2.1 Fundamental Solutions ...................... 10
3.2.2 Two-Dimensional Representation Theorem...... 13
3.2.3 Displacement Discontinuity Method .......... 14
3.3 Time and Space Integrations ........................ 17
3.3.1 Temporal Integrations .... 17
3.3.2 Spatial Integrations ....................... 19
3.3.2.1 Linear/Constant Formulation ......... 19
3.3.2.2 Linear/Linear Formulation ........ 22
3.4 Numerical Implementation ........................... 27
3.4.1 Linear/Constant Formulation................ 27
3.4.2 Linear/Linear Formulation.................. 32
3.5 Tangential Stress Calculation ...................... 36
3.5.2 Linear/Linear Formulation......... 38
3.6 Particle Velocity and Acceleration Terms ............ 39
3.6.2 Linear/Linear Formulation.................. 40
3.7 Symmetry Conditions ............................... 42
4 CAUSALITY ............................................... 43
5 WAVE MECHANICS THEORY.................................... 45
5.1 Basic Theory ..................................... 45
5.2 Mechanics of Crack-Wave Interactions ............... 49
6 SINGLE ELEMENT RESULTS .................................... 53
7 APPLICATIONS ............................................ 80
7.1 Solution of Mining Problems ........................ 80
7.2 Linear/Constant Formulation........................ 81
7.2.1 Selberg's PressurizedCylindrical Cavity ..... 81
7.2.2 Shear Fault Daylightlng into Tabular
Excavation ................................ 83
7.2.3 TWo Reef Problem .......................... 94
7.3 Linear/Linear Formulation.................. 102
7.3.1 Half-Plane Under Prescribed Time-Dependent
Normal Traction ........................... 102
7.3.2 Inclined Crack Interaction with Shear
Wave Source ............................... 105
8 SPECIAL ELEMENTS ......................................... 110

8.1 Elastic Joint Element ............................ 110
8.2 Cohesionless-Frictionless Fault Element ............. 120
8.3 Rapid-Yielding Hydraulic Prop Element .............. 120
9 STABILITY ................................................ 125
10 CONCLUSIONS ............................................. 135
11 RECOMMENDATIONS ........................................... 137
REFERENCES .................................................. 139

Appendix 1: Temporally and spatially Integrated constant
sub-functions ................................... 150
Appendix 2: Temporally and spatially Integrated linear
sub-functions .... 156
Appendix 3: Temporally and spatially integrated constant and
linear velocity sub-functions ..................... 165
1
1 INTRODUCTION
Rockbursts and rockfalls are two major problems affecting the

safety of miners In deep-level mining operations. The threat of
rockbursts Intensifies with depth because rock pressure Increases with
depth. Mining alters the stress distribution in the surrounding rock
mass, and seismic events or violent failures of the rock mass can result
so that the stresses can equilibrate. These events are often associated
with shear displacements along geologic discontinuities, which sometimes
daylight into the mining excavations.
Deep-level mining operations are often of an extensive tabular
nature because the ore is typically located in an extensive thin layer.
The elastostatlc displacement discontinuity method (Crouch and Starfield .
1983, Napier and Stephansen 1987) is routinely used by rock mechanics
engineers worldwide to model numerically the displacements and stresses
generated by mining under these conditions. The elastostatlc
displacement discontinuity method is a numerical technique based on the
solution to a discontinuity in displacement over a finite-sized crack in
an infinite elastic continuum. The elastodynamic displacement
discontinuity method is a natural extension of this technique which
allows the study of dynamic effects in a tabular mining environment.
There are essentially two types of problems that can be modelled
with the elastodynamic displacement discontinuity method. The first is
the simplest. Imagine a displacement discontinuity surface (e.g. fault
or tabular excavation) of arbitrary shape, in an infinite elastic
continuum, on which a known time history of shear and normal components
of displacement discontinuity exists. The elastodynamic displacement
discontinuity method can be used to calculate the time histories of
displacements, velocities, accelerations, and stresses anywhere in the
surrounding continuum, using the prescribed transient displacement
discontinuities directly in a discretized set of convolution integrals.
1
The second basic problem Is more complicated because a system of
equations must generally be solved for the transient displacement
discontinuities. Given a known time history of shear and normal
tractions on a displacement discontinuity surface, the corresponding
time history of displacement discontinuity on that surface can be
determined. Thereafter, the displacements, velocities, accelerations,
and stresses can be determined anywhere In the surrounding coptinuum, as
in the first type of problem mentioned above.
The assumptions made in the displacement discontinuity method are
that the rock mass is infinite, homogeneous, isotropic, and linearly
elastic. The displacement discontinuity surface can represent any linear
or non-linear material behavior that can be incorporated in the model as
a boundary condition (Crouch 1980). This is the simplest possible model
that can be postulated. Clearly, the rock mass does not match these
idealistic assumptions. However, elastodynamic boundary element methods
can be (and have been) extended to model non-linear, anisotropic, and
semi-infinite domains (see Ahmad and Banerjee 1990, for example). It is
also possible to include multiple material types. These refinements can,
in principle, be included in the models described in this work. It is
important to realize that the use of elastodynamic computer models in
mining applications is a very new field, and these models thus do not
necessarily need the added complications that would result from more
sophisticated assumptions. There are a large number of areas that can be
investigated with the relatively simple model used herein.
The original intent of this work was to develop a higher order
three-dimensional elastodynamic displacement discontinuity method,
thereby extending the work of Hack (1991). However, problems arose
concerning the numerical stability of the three-dimensional displacement
discontinuity method (and other boundary element methods), and
alternative approaches were investigated.
A three-dimensional displacement discontinuity method was
2
attempted where time was transformed into the Laplace domain In the hope
of avoiding problems of numerical Instability. There are a number of
undesirable aspects to this approach. The Integrals have to be
numerically Integrated using complex arithmetic, and the Inversion of
the displacement discontinuity solution back into the time domain can
only be done numerically. Dynamic fracture propagation (one of the long
term goals of this work) is not possible to Implement since the solution
is obtained over a predetermined range of time steps (or frequencies),
and hence it is impossible to check the rupture status of a particular
element after each time step. Notwithstanding all of the above, this
approach was implemented, but appeared to give poor results for cases
when the loading pattern was varied rapidly, spatially and temporally.
A two-dimensional time domain approach was then decided upon.
There were two major reasons for attempting a two-dimensional higher
order approach instead of a three-dimensional higher order approach.
First, it was unclear how much would be gained by a higher order
approach over the work of Hack (1991) in terms of alleviating the
numerical instability problem evident in his work. The higher order
approach posed a new challenge that other boundary element methods do
not have. The displacement discontinuity integrals are hypersingular at
the element edges, which means that internal collocation points are
required, and the resulting spatial shape functions are only piecewise
connected between adjacent elements. It was easier to develop the higher
order approach in two dimensions to investigate this problem.
The second reason for pursuing the two-dimensional approach was
that the modelling of problems where elements join each other at
arbitrary angles is comparatively easy to implement in two dimensions,
but requires the development of general quadrilateral or triangular
elements in three dimensions. This type of three-dimensional approach
has not yet been fully developed in elastostatics, and the insights
gained from a two-dimensional model can guide a future extension to
3
three dimensions.
The two-dimensional solutions do have less applicability than the
three-dimensional ones. However, they are very useful when performing
numerical experiments because physical interpretation is easier, and
because computer run times and storage requirements are reduced.
Furthermore, because deep-level mining excavations are often of an
extensive tabular nature, the assumption of plane strain is often
justified.
In what follows, Section 2 is a brief literature review of
elastodynamic methods. Section 3 contains mathematical details of the
two-dimensional elastodynamic displacement discontinuity method.
Sections 4 and 5 deal with causality and basic wave mechanics. Sections
6 and 7 are devoted to some example problems that demonstrate the
applicability of this work. Section 8 details special elements that can
be implemented. Section 9 is a discussion about issues of numerical
stability, and Sections 10 and 11 present the conclusions of this work,
and recommendations for future work in this field.
4
2 LITERATURE REVIEW OF BOUNDARY ELEMENT METHODS
This review will concentrate on elastodynamic boundary element

methods. There are a number of classical texts on elastodynamics and
integral techniques, viz. Love (1944), Wheeler and Sternberg (1968), Fao
and Mow (1973), Acheiibach (1973), Eringen and Suhubi (1975), Aki and
Richards (1980). General texts on dynamic boundary element methods
include those by Beskos (1987a) and Manolis and Beskos (1988). The
reader is referred to Beskos (1987b) for an exhaustive review of dynamic
boundary element methods.
Integral formulations for potential theory problems and elasticity
appeared in the literature in the 1800s. The first notable contributions
were those of Helmholtz (1860) and Kirchhoff (1883) on the wave
equation, Fredholm (1903), Muskhelishvlli (1953), and Mikhlln (1957) on
integral equations, Somigliana (1889-1890) and KupradZe (1963) on
elastostatics, and Kellogg (1929) on potential theory.
The first numerical applications appeared in the 1960s. For
example, Jaswon (1963) and Symm (1963) did work on potential theory,
Salomon (1964) on the displacement discontinuity method, Banaugh and
Goldsmith (1963a) in steady-state acoustics, Friedman and Shaw (1962) in
transient acoustics, Rizzo (1967) in elastostatics, Banaugh and
Goldsmith (1963b) in steady-state elastodynamics.
The first work on a transient elastodynamic integral formulation
was that of Cruse and Rizzo (1968) and Cruse (1968), who used a Laplace
transform boundary element approach to solve a wave propagation problem
in a half plane. Their approach was based on the work of Sternberg and
Eubanks (1955, 1957) and Doyle (1966). A numerical inversion algorithm
due to Fapoulis (1957) was used to obtain the time domain solutions, but
this was found to give good results only for early times.
Manolis (1980) and Manolis and Beskos (1981) extended Cruse's work
to investigate stress concentrations around arbitrarily shaped cavities
5
due to the scattering of elastic waves. They used a Fourier tranform
approach with an inversion algorithm due to Durbin (1974), which was
considerably more accurate than Papoulla's, even to late times, but was
more time consuming because of the need to carry out the computations in
complex arithmetic. Nlwa et al. (1975, 1976) and Kobayashi and Nlshimura
(1982) also used Fourier transform methods to solve transient
elastodynamic problems.
Time domain formulations were pioneered by Friedman and Shaw (1962)
in the field of acoustics. Cole et al. (1978) were the first to develop
an elastodynamic formulation in the time domain, but their method was
restricted to the anti-plane strain case, and was found to be unstable.
This was apparently due to an incorrect numerical treatment of the
convolutions involved (see Banerjee et al. 1987). Nlwa et al. (1980) and
Karabalis and Beskos (1984a, 1984b) first developed general two- and
three-dimensional elastodynamic methods in the time domain using the
direct integral approach.
Manolis (1983) compared three two-dimensional direct boundary
element approaches, viz. the Laplace transform, Fourier transform and
time domain. He found that the Laplace approach was superior to the
Fourier approach for problems with constant in time traction boundary
conditions. He also found that the time domain approach was slowest,
which is not surprising, because his two-dimensional time domain
formulation was obtained by numerically integrating the three-
dimensional solution. There was not very good agreement between the
different methods, but this could be due to an error in the time domain
kernels (see Nicholson and Mettu 1988), which also appears in the Nlwa
et al. (1980) formulation. The time domain approach was found to be more
accurate than the other two approaches for early times.
Banerjee et al. (1986) obtained a direct time formulation for
solving three-dimensional elastodynamics problems. Analytical in time
and numerical in space integrations were used with constant in time and
6
quadratic in apace elements. Ahmad and Banerjee (1988a) extended this
work to include analytical time integration, assuming a linear
interpolation.
Tian (1990) developed a two-dimensional direct and fictitious
stress boundary element method, using analytical Integrations in time
and space. The direct approach utilized a quadratic spatial variation
and a linear temporal variation, whereas- the indirect approach used a
constant spatial and linear temporal variation. The methods were
demonstrated for a variety of mining applications.
Hack (1991) developed a three-dimensional elastodynamic
displacement discontinuity method using constant in space and linear in
time elements with analytical integrations. The method was demonstrated
for a variety of mining applications, where the mining geometries were
of a tabular nature. Multiple surfaces of displacement discontinuity can
be modelled with this approach.
Other work on crack-like geometries in three dimensions under
dynamic loading conditions.includes the contributions of Haskell (1969),
Das (1980, 1981, 1985, 1988), Day (1982), Das and Kostrov (1987), and
Hirose and Achenbach (1989). The work of Haskell (1969) was limited to
field calculations due to prescribed dynamic displacement discontinuity
functions. The work of Das (1980, 1981, 1985, 1988) and Das and Kostrov
(1987) allowed the solution of traction boundary value problems, using a
direct boundary integral approach, but was only demonstrated for single
cracks. Day (1982) used a finite difference approach to study crack
propagation problems on single surfaces. Hirose and Achenbach (1989)
developed a time domain displacement discontinuity method which was
demonstrated for a penny-shaped crack.
There are other kinds of boundary methods such as the domain
boundary element method or mass-matrix approach (Nardini and Brebbia
1982), the T-matrlx approach (Waterman 1976) and the Born approximation
method (Chu et al. 1980). In the mass-matrix approach, the acceleration
7
terms are taken as equivalent body forces in a volume Integral. In the
T-matrix approach, the boundary values, the unknowns and the Green's
function of the problem are expanded in terms of a set of basis
functions. An Integral representation of the problem is used to
construct the transition or T-matrix, which connects the knowns and
unknowns. In the B o m approximation method, the problem is formulated in
the frequency domain as a direct boundary element method, and the
resulting integral equations are solved in an iterative fashion. The
final solution is obtained as a finite sum of successive approximate
solutions.
As for two-dimensional modelling of crack-like geometries, Das and
Aki (1977) developed a two-dimensional boundary integral method, based
on Lamb's (1904) half-plane fundamental solutions. The integrations were
performed numerically. The method was demonstrated for in-plane and
anti-plane problems, but was limited to a single fault surface.
Andrews (1985) presented a two-dimensional boundary integral
method for modelling plane-strain shear rupture, driven by a slip-
weakening friction law. The integrations were done part analytically,
and part numerically. The method was limited to a single pre-defined
fracture line.
Finite element methods (e.g. Johnson 1990) and semi-analytical
methods (Dragonl 1990) have also been used to investigate fault slip
behavior under various conditions. Finite difference methods (e.g.
Cundall 1990) have been used to model the dynamic interactions between
cracks and stress waves, but these methods are currently limited to
geometries that are grid-based because of the nature of finite
difference solutions.
8
3 THEORY: TWO-DIMENSIONAL FORMULATIONS
3.1 Introdustigp
Two elastodynamic displacement discontinuity methods are presented

for two-dimensional infinite domains under plane strain conditions. The
temporal variation of the displacement discontinuity function is assumed
to be linear and continuous. The spatial variation is assumed to be
constant in the first method (linear/constant), and piecewise linear in
the second method (linear/linear). The methods are developed to solve
mixed (traction and/or displacement) boundary value problems. The
algorithms are time-marching. All integrations are analytically
performed and are "causal," which means that only the dynamically active
parts of each element are integrated. Symmetry is Included in the
linear/constant algorithm to allow for reduced data manipulation in
problems exhibiting symmetries parallel to the xx and/or x2 axes. The
linear/linear algorithm can also make use of symmetry, but this has not
been included in the current work.
The numerical models are verified in four ways. First, the models
are checked against Selberg's pressurized cylindrical cavity problem
(1951), in order to check the solution algorithm, the co-ordinate
transformations, and the tangential stress calculations. Second,
verification of the complete solution and field point algorithms is
obtained by checking the final static results, obtained via the dynamic
formulations, against equivalent elastostatlc displacement discontinuity
methods. It should be noted that this check requires the use of the
complete dynamic formulation, and is thus a rigorous check. Third, a
dynamic analytical solution exists for a suddenly pressurized finite-
length crack. This is used to check some components of the solution
algorithm. Finally, a half-plane problem is compared with two other
boundary element methods.
The models are demonstrated for simple single and multiple reef
9
problems. This work can be combined with either a fictitious stress
method or a direct boundary element method, so that a general problem
involving large excavations and crack-like geometries in two-dimensional
space can be solved.
3.2 Mathematical Derivation

3.2.1 Fundamental Solutions
The basis of the two-dimensional formulation is Stokes's (1883)

fundamental solution, integrated from negative to positive infinity
along the x3 axis. This implies that the two-dimensional solution
differs from the three-dimensional one in two ways. First, the two-
dimensional solution is one order smoother (in space only) than the
three-dimensional one, which aids numerical implementation. Second, the
three-dimensional solution is due to a signal emitted at a single
instant in time and space, whereas the two-dimensional solution also
depends on signals emitted at all later times. Thus, the two-dimensional
solution has a distinctive "tail." This phenomenon is termed "diffusion"
(Eringen and Suhubi 1975), and is simply due to the integration from
negative to positive infinity of the three-dimensional solution. The
tail is due to the wave arrivals from all points along the integrated
direction. Figure 3.1 shows the fundamental displacements u generated by
impulsive point and line sources fc in three and two dimensions,
respectively, for the simplified case of the vector wave equation given
by
cVu+b-— = (3.1)
dt*
10
U y
t t
r/c p/c
r2 - x ^ , 1 - 1 to 3 pl - x ^ , a - 1 , 2
r2 - x32 + p2
Figure 3.1: Three- and two-dimensional displacements due to Impulsive

point and Hue sources, respectively, for vector wave equation.
The two-dimensional displacement fundamental solution is given by
V3o * 0
(3.2)
H(t-JL) o2t f(t--i)

(3.3)
33 f ------- ——dij
2xpC2 I (H2-r2)0-3
ot 1 e
2xp dtp cx ^ (ij2-r2)°*s fl Ci
t-_L
°2
H(t~— ) r f i/f<t-i-w)dw] +
c2 i («?2-r2)03 ij c2
o2t f(t-JL)
i22H(t-JL) r !i_d„] (3.4)
c,
2 °2 i (92-r2)0*5
11
where
a and j8 each range from 1 to 2,
* - la - £l
vij - vyte.tji.olfCt))
cx - compreaalonal (P) wave velocity
c2 - shear (S) wave velocity
£ - receiver or field position
£. - source position
This solution is Stokes's displacement aolution integrated along the x3

axis from negative to positive infinity (see Eringen and Suhtibi 1975, p
410). In (3.2) to (3.4), vAJ is symmetric, and represents the xA
component of displacement at point x and time t due to a force uniformly
distributed along a line at point £, time s - 0, parallel to the Xj
axis, and having a time dependent magnitude of f(t). The v33 term
provides the "anti-plane" motion, while the v^ term provides the
"plane" motion in two-dimensional elastodynamics. The fundamental
stresses are obtained directly by substitution into Hooke's law to get
9v._ 9v„ 3va,

If.» <3-5>
These fundamental displacements and stresses are the two-dimensional

equivalents of the Stokes state in three dimensions.
The displacements in the dynamic doublet state (needed for the
development of the displacement discontinuity method) are obtained from
the stresses in the two-dimensional Stokes state. The stresses in the
dynamic doublet state are obtained by substitution of the dynamic
doublet displacements into Hooke's law. It will be seen later that only
vtJ is actually needed to obtain the displacement discontinuity method
formulation, since v^ only needs to be discretized. The necessary
12
derivatives are then obtained directly from the discretized form of v4j,
The fundamental solution vtJ has two important properties:
causality, and translation. Causality implies that vAj - 0 if (t-r/cj-s)
2 0, and means that the receiver point cannot feel any dynamic effects
from the source point before the wave arrives at the receiver point.
Translation is a dual property. Time translation is a reciprocal
relation for source and receiver times given by vij(JS»t;£,s|f(t)) -
t-s;£,0|f(t)). Space translation is a reciprocal relation for
source and receiver positions given by v^ta.tif^slftt)) -
Vji(£,t;x,s|f(t)). Time and space translation can be combined. In this
work, time translation is used repeatedly during the temporal
discretization of the equations, and causality is used during
the spatial and temporal Integration processes. Space translation has
not been included in this work, because it is only useful in grid-based
systems where a repetitive mesh occurs.
3.2.2 Two-Dimensional Representation Theorem
The two-dimensional representation theorem, applicable when the

boundary and initial conditions are free from the x3 coordinate, can be
obtained by spatial integration of the three-dimensional result (e.g.
see Eringen and Suhubl 197S), and is given by
M i . t ) " | l vut<5.t;£|t(s)1(x,t)) - ^nHkte.tlllu^x.tJJJdStx)
+ pJvJ^x.tiilf^x.OJdVCx) +
r, aV^Cx.t;!)
pftVo^xJV^x.t ;£) + Uptte) ]dV(x)
(3.6)
13
where
vlk°(x,t;£|t(B)i(ji,t)) - Vik(&.t;£) * t<B)1(s,t)

<7(B)lk°<X.t;£|ui(x,t)) - rlflt(s,t;£t0|6(t)) nf * u^s.t)
Vlk(ai,t;£) - vtt(x,t;£,0|«(t))
Uoi(x) - u^x.O)
v01(jl) - duL(&,0)/8t
* - time convolution operator
and t(B)1 denotea the traction component tt on surface S with outward

unit normal n. This result is due to De Hoop (1958). This theorem
connects the boundary tractions and body forces with the boundary
displacements, and is the basis of the two-dimensional direct boundary
integral method. In order to obtain the two-dimensional displacement
discontinuity method, consider the simplified case when the initial
conditions and body forces are identically zero. In this case (3.6)
simplifies to
V i . t ) “ |[v^(x,t;£|t<s)1(x,t)) - ff“„)ik(x,t;£|ui(x,t))]dS(x) ^ ^
3.2.3 Displacement Discontinuity Method
The displacement discontinuity formulation can be obtained from

(3.7) using an approach originally due to Lamb (1932) and later Peirce
(1983) in elastostatics, but extended here to elastodynamlcs. Consider a
finite body V in an Infinite domain. Denote the exterior of the body as
Ve. The boundary S delimits the two regions. Let n{,)(jO and n(i)(x) be
the outward unit normals at s for regions V and V0, respectively, as
shown in Figure 3.2, and let point £ € V. Then (3.7) becomes
14
u^ti,t) - |[v^(x,ti£|t(
<nl(
>.>)i(x,t)) - aJn(.))tt(x,t;£|uil)(x.t))]«iS(x)
(3.8)
Figure 3.2: Unit normal directions on boundary S.
The contribution to the displacements at £ € V due to the stresses and

displacements in the exterior is zero (Eringen and Suhubi 1975), so that
0 " - ^ n{i))ik(x,t;£|u‘*)(x,t))]dS(x) ^ 9)
Adding (3.8) to (3.9), and noting that q (1> - gives
«^l)(£.t> - Jlvj^x.tjilT^x.t)) - ^ n(.))ik(x.t;£|D1(x,t))]dS(x) (3

10)
where
Dx - u ‘1} - 4
" (3.11)
15
In order to obtain the displacement discontinuity method, the constraint
that the tractions must be continuous across the boundary is imposed,
and (3,10) reduces to
(3.12)
Equation (3.12) gives the k-th displacement component at point £ due to

a discontinuity in the i-th displacement component at point g. The plane
motion equations will only be developed in this work since the anti-
plane motion equations are simply a subset of these, From (3.6), the
displacements in (3.12) and the Hookean stresses can then be written as
(3.13)
a
(3.14)
where Da(£,s) is the displacement discontinuity source function, and

where commas indicate spatial differentiation with respect to x1. Recall
that the stresses are continuous across S, and the displacements are
discontinuous across S. The integrals in (3.13) must thus be evaluated
in a limiting sense to obtain the displacement discontinuity formulation
so as to ensure that the correct jump conditions are imposed. One
procedure is to first integrate the effect on £ of (3.13) and (3.14)
over an element centered at £. The displacements and stresses are then
determined for £ -♦ £+ and £-*£'. The resulting displacement and stress
integral equations can then be used to obtain the displacement
16
discontinuity distribution on S in time, for any well-posed boundary
value problem. Once the distribution has been determined, field values
can be found using (3.13) and (3.14), hence the label of "indirect"
given to the displacement discontinuity method. Alternatively, a dynamic
displacement discontinuity distribution can be directly specified, and
field values directly calculated, if so desired.
3.3 Time and Space Integrations

3.3.1 Temporal Integrations
A piecewise linear temporal variation is assumed for the

displacement discontinuity function D(£,t), and will be denoted by f(t)
for simplicity. The simplest way to perform the integrations and develop
the time-marching algorithm is to integrate vtJ for the case f(t) - t.
The piecewise linear formulation is then obtained by combining three
such functions, staggered at successive time steps. Hence,
f(tk) - [H(t-V l > - ^ - ♦HU-t^)^]^ (3.15)
where tk - kAt, and rk - t - tk.

The above is permissible because of the linearity of the boundary
element method equations (principle of superposition applies), and
because of the time translation property of all the fundamental
solutions. Equation (3.15) forms a "hat" function at time tk
(Figure 3.3).
17
t
lk-1 ^ tk+1
Figure 3.3: Temporal Interpolation.
Combining a bat function at each time step results in a piecewise linear

temporal variation, as shown in Figure 3.3. Assuming a linear temporal
variation f(t) - t, the two sets of time integrals in (3.4) are (Tian
1990)
c2t t--!L
f ---- — — dn - tcosh-^f^) - -I(c|t2-r2)0-3
I (ij2-r2)0,3 r c2 (3.16)
t-2
J . d'\ 0 . f *(t-i-./)dv - -L [^ (c2t2+ 2 r 2)cosh"l(— ) -
* (ij2-r2)0,3 c c3 6 2 r
-Ir2(c2t2-r2)0-3 - ii(c2t2-r2)1,3]
12 36
(3.17)
where c - ca or c2. Substitution of (3.16) and (3.17) into (3.4) gives
18
3.3.2 Spatial Integrations
3.3.2.1 Linear/Constant Formulation
A piecewise constant spatial variation is assumed for the

displacement discontinuity function D(£,t). Without loss in generality,
assume that a displacement discontinuity element occupies the position
|x| S a, y - 0, as shown in Figure 3.4.
-a +a
Figure 3.4: Master element geometry.

Apply the following change of coordinates to a typical spatial integral:
u X-L
Jf(*-?iiy-f2 )df1 - f f(x,y)dx (3.19)
I x-u
where
r -x +
L - lower limit
U - upper limit.
The integral limits are variable since the spatial integrations obey
causality. There are essentially four different cases to consider when
specifying the Integral limits (Figure 3.5):
1) no wave arrival at receiver point from source element,

2) wave arrival at receiver point from all parts of source
element,
3) wave arrival at receiver point from points inside the source
element,
4) wave arrival at receiver point from one part of source element.
(1) I • (3)
(2) (4)
Figure 3.5: Different types of source element integration limits.
20
From <3.18) and (3.19), the Following spatial Integrals, due to Tlan
<1990), are then required:
fcosh*1<£^)dx - <x)cosh*1<— ) + (ct)sin"l( 2L_ )-

J r r (c2t2- y V 3
(y)tan_1(--------- ) <3.20)
rJ^cV-rV-’dx- <fH)tan-1< ) -sin*^----x „ )

J r2 y y<c2t2-r*)0 5 (c2t2-y^0,3
<3.21)
f J L ^ t 2-?2)0,3*!* - <c2t2-r2)0,3 - <ct)cosh"1<4^) <3.22)

J r2 r I
f(c2t2-?)°'s(fi - *<c2t2-r2)0-3 + <f2t-:f )sln-1<---- ) (3 2 3 )

J 2 2 ^z.^zjo.s
fJ^cV-r2)0 3d* - <£±f)tan-1<___ ^ ) ♦ ^ c V -^)03

J r* 2 cty3 y<c2 t2 -r2)0,3 2^ ?
(3.24)
f 4 ( c 2t2-?)03dx - <J_)cosh-2< ^ ) - ‘ (c2t2- ? ) » (3.25)

Jr* 2c t f 2t*
Notice that, at the wave front (r - ct), all terms In <3.20) to

<3.25) are zero except those involving an arctan or arcsln function.
These terms, and some of their first and second spatial derivatives
(needed In the displacement discontinuity method), are non-zero and
finite. The arctan term provides the displacement jump across the
element in the displacement discontinuity method. A complete list of all
21
the derivatives of the spatially integrated form of (3.18) is given in
Appendix 1. The arcsin term can be rewritten as
sin*1(. ) - tan_1( )
(c2^2 ir2\0.5
tz-y ) (c V - r 2)0-9
/„2«.2
,-i/ (c t -rr2\D.5
) ), X > 0
,2^2 _=2 \«.S

- - - - tan~1(
2 (3.26)
which allows easier evaluation when r - ct. It is important to note that

in order to obtain the spatial derivatives of the integrated form of
(3.18), Leibnitz's rule must be strictly applied, since the Integral
limits are functions of the spatial variables (r) in general. However,
it is fortuitous that (3.18) is zero at the wave front. This implies
that integration and differentiation orders can be Interchanged since
all terms in Leibnitz’s rule that involve the integral limits are zero.
Both types of arctan terms behave in a similar manner to the
arctan(y/x) term that arises in two-dimensional elastostatic boundary
element methods, the complication being the extra dynamic parts.
3.3.2.2 Linear/Linear Formulation
A piecewise linear spatial variation along each element is assumed

for the displacement discontinuity function D(£,t). Each element
contains two collocation points or nodes. The nodes are chosen to lie
within the element, since the spatially integrated solution is strongly
singular at the element edges, as is the case in statics. Without loss
in generality, assume that a displacement discontinuity element occupies
the position |x| £ a, y - 0, as shown in Figure 3.6.
22
Figure 3.6: Master element geometry
The collocation positions can be chosen to lie anywhere within the range
jx| < a. In statics, Crawford and Curren (1982) choose the collocation
positions to be located at x - ±a/2°*s. Napier (1990) has determined
that these positions are the optimum in terms of calculating the correct
volume of closure for a single element in statics. In dynamics, if the
nodes are too close to each other or to the edges of the element,
numerical instabilities will develop sooner because the kernel matrices
become ill-conditioned. In this work, the nodes have also been
positioned at x - ±b - ±a/2 0,3.
The linearly varying displacement discontinuity function can be
separated into a constant and a linear part. In the case of the constant
part, apply the following change of coordinates to a typical spatial
integral:
U X-L
(3.27)
where
r -x +y
L — lower limit
U - upper limit.
Similarly,applying achange of coordinates to the linearpart gives
U x-L x-L
JV1f(x-$1,y-€a)d{1 - x f f(x,y)dx - Jxf(x,y)dx (3.28)
I x-U x-U
The integral limits are variable since the spatial integrations obey
causality. As in the constant/linear formulation, there are essentially
four different cases to consider when specifying the integral limits
(Figure 3.7):
1) no wave arrival at receiver point from source element,

2) wave arrival at receiver point from all parts of source
element,
3) wave arrival at receiver point from points inside the source
element, j
4) wave arrival at receiver point from one part of source element, j
i t
(2) (4)
VI
Figure 3.7: Different types of source element Integration limits.
24
For cases 2 through 4, care must be taken to ensure that the appropriate
shape functions are included for each coefficient calculation. For
example, in the case of a wave arrival at a receiver (collocation or
field) point from one part of the source element (Figure 3.8), both the
left and right shape functions have non-zero contributions. In Figure
3.8, even though the left collocation point associated with the left
shape function is outside the radius of influence, an influence
coefficient is still generated, because part of the shape function is
within the radius of influence. Similarly, the right collocation node
generates an influence coefficient because part of the right shape
function is within the radius of influence. A shape function will only
generate no influence if the displacement discontinuity function is
identically zero at its associated source collocation point.
Figure 3.8: Detail of integration contributions for shape functions.
From (3.18), (3.27) and (3.28), the following spatial integrals,

due to Tian (1990), are then required:
[cosh‘1(4 r)dx - (x)cosh‘l(£H) + (ct)sin_1( ---- ) -
(3.29)
(y)tan*1(
25
fccosh-H— )dx - (^Jcosh-H— ) - ( ^ X c V - ? ) 0'5 (3.30)
J r 2 r 2
C - L ^ t 2-?)0-3^ - (£H)tairl< Ct\ ,..) - ain'H m ]

j r2 y y(c2 t2-r*)°-5 ( c ^ - y 2)0-3
(3.31)
f J L f c V - r V ^ d x - (cV-r 2)0,3 - (ct)cosh'1 (4r) (3.32)

J f
,2^2 - 2
J(c2t2-?V-3dS - ^ ( c ^ - r 2)0*3 + (i_LJL)sin-1( _ _ ^ - i l ) (3 .33 )
Jx(c«ta-r»)»-sfi - -I(cV-r*)‘-s (3>34)
Jfi?(cV-?)#^ - i(c2t2-yz)2sln*1(
8 (c2t2E_
_ ^ )0.3) - ^8( c V * ? ) 1'3 ♦
i ( c !t2-?)°’! (3.35)
J J* 2 c ty y(c2t2-r2)0-3 2 y2rz
(3.36)
f— (c^2-?2)0,5^ - (J_)cosh-l(4H) - -i-OsV-r*)0-9 (3 37)

J r* 2 ct r i?
Notice that, at the wave front (r - ct), all terms in (3.29) to

(3.37) are zero except those Involving an arctan or arcsln function.
These terms, and some of their first and second spatial derivatives
26
(needed In the displacement discontinuity method) are non-zero and
finite.
3.4 Numerical Implementation

3.4.1 Linear/Constant Formulation
The numerical implementation of the elastodynamic displacement

discontinuity method closely follows the methodology laid out in Crouch
and Starfield (1983) for the two-dimensional elastostatic method. The
only effective difference between the static and forward-marching time-
domain dynamic methods is the inclusion of source and receiver time
loops in the dynamic algorithm. Recall that the static solution
algorithm is given by (Crouch and Starfield 1983)
Hi ” (3.38)
(3.39)
where
Hi - K , uJi
S-i - [*.. ffJi
Bj - [D„ Dn]j
Aij " A,n;A,,,,
fiij ” [*■■• B »n! Bn»i BmJiJ
s - shear effect, n - normal effect.
For a displacement boundary value problem, i ranges from 1 to the number

of elements N in (3.38), and (3.39) is unused. For a traction boundary
value problem, 1 ranges from 1 to N in (3.39), and (3.38) is unused. In
a mixed boundary value problem, i ranges from 1 to N, and a combination
27
of (3,38) and (3,39) is used. The dynamic algorithm can similarly be
written as (Tian 1990)
h" - C1])" + S*-!,*-! Sr*1? (3.40)
where
m - current time step number

£ - & or or combination of A and £ - 2N x 2N coefficient matrix
fc - 11 or £ or combination of u and z - 2N x 1 vector of known
boundary displacements and/or tractions
Q - 2N x 1 vector of displacement discontinuities.
N - number of elements
The matrices in (3.40) are fully populated in general. It is clear that

the unknown displacement discontinuities If at the current time step m
are obtained via a convolution between the known coefficients and known
displacement discontinuities from all previous times.
The algorithm in (3.40) can be made implicit or explicit,
depending on the nature of matrix fi1. Define the dimenslonless factor Q1
as
where At is the tlme-step size. For most geometries, if Q1 £ 0.5 then

(3.40) is explicit, otherwise it is implicit. The method will be
explicit when the compressional (fastest) wave does not reach any of its
neighboring elements in one time step. For some geometries, where
elements are close to each other, in order for the solution algorithm to
remain explicit, a smaller critical Ql is often required.
In the explicit case, matrix Q1 contains self-effect information
28
only, and Is strictly diagonal. Algorithm (3.40) can then be marched in
time without the solution of an equation system. In the implicit case,
£* is banded about the diagonal, and a system of equations must be
solved at each time step, using either a direct or an iterative solver.
At each time, and for each source-receiver combination, the
coefficients A and A consist of four sub-terms (hence the size of
2Nx2N), just as in the static case. In order to determine general
expressions for the influence coefficients in A end A, consider
initially the simple case of the field displacements and stresses
generated by a master element (Figure 3.4) oriented in the global co
ordinate system (x,y). The stresses and displacements, from (3.40), are
given by
_ £ [B£ k.iD k + B--k.iD kj

kel
< ■i *iC X ]
k-1 (3.42)
- E [[(A+2/i)p(V1112 + V1 2 1 1 ) + A/i(V1 2 2 2 + V2212) ]"-k*lD* +

k-l
[(A+2/i)2 V1212 + A(A+2/i)(Vllll + V2222) + A2 V1212]B"k*1Dy ]

B
o r> rA®’k+1nk ^ A®-k4lT\ki
°xy - D* + A*y Dy]
k-l
B
m r' fa k , AB-k*lnki
‘' y y - E l ^ , Dx +Aw Dyl
k‘l (3.43)
where
A q - ji2 (V1122 + 2 V1212 + V2211)

Agy - (A+2p)/i(V1222 + V2212) + A/j(V1112 + V1211)
29
Ay, - (A+2jt)p(V1222 + V2212) +Am (V1112 +V1211)
A,, - (A+2M)a V2222 + 2A(A+2p)V1212 +A2 Vllll (3.44)
and
- /i(VH2 + V121)
- (A+2/OV122 + AV111
By, - m (V122 + V221)
B„ - (A+2^)V222 + AV121 (3.45)
where VIJK and VIJKL are given In Appendix 1. The above can be written
in terms of the local co-ordinate system (x,y), shown in Figure 3.9, as
x - (x-c,)cos£ + (y-cy)sin0
y - -(x-c3t)sin^ + (y-Cy)cosj9 (3.46)
by simply replacing (x,y) by (x,y) in (3.42) through (3.45).
Figure 3.9: Local and global co-ordinate systems.
The displacements and stresses in the global co-ordinate system (x,y)

due to the displacement discontinuities occurring on an element in the
local system (x,y) can be found by substituting (3.42) and (3.45) into
30
ux - u^cos/J - UySirv?
Uy - Ujsln^ + u^coafl (3.47)
and
#n - o— cos2/9 - 2a— aiiy?cosj8 + ff— sin2/}
&jry - (cr— -a— )ainficoafi + a— (cos2/9-sin20)
ffyy - <7— sin2/9 + 2 a— sin/3cos/9 + a^coazfi

(3.48)
The solution algorithm can be structured so that the displacement

discontinuities are calculated with respect to either the global or the
local co-ordinate system. In the latter case, which has been used in
this work, the methodology is as laid out in Crouch and Starfield
(1983). The procedure is as follows. Let a particular source and
receiver element be located at (x,y)j with orientation /3j and (x,y)i
with orientation fiL, respectively, as shown in Figure 3.10. Define the
source displacement discontinuities to be (Dj.Dy)j, in terms of a local
co-ordinate system at the source element.
Figure 3.10: Source and receiver element geometry.
31
The global displacements and stresses at the receiver point (x,y)1
are then given by (3.42), (3.43), (3.47), and (3.48). The displacements
and stresses at the receiver element relative to the local co-ordinate
system of the source element are then given by (3.42) and (3.43), but
with (x,y) simply replaced by (x,y) throughout. The displacements and
stresses In the local co-ordinate system of the receiverelement
relative to the local co-ordinate system of the sourceelement are then
given by
( u ,) i - (us ) i cos7 + (Uy)t s ln 7
(“nh - “(“gJiSiwr + (Uy)tcos7
" (tfs)iC08S + 2(ffSy>iflinT'COS7 + (o— )1sin27
" <ff35c)isln27 - 2(ff^)tsln7COS7 + (ojy)1cos27
(ff.Ji - - <«fr)1]sin7C0S7 + ( o ^ J c o s ^ - sin27]

QJ.4“)
where 7 - (^t - /9j). This completes the derivation. Equations (3.49)

provide the influence coefficients for matrices A and £.
The complete solution algorithm has been implemented in a FORTRAN-
77 computer program, called TW04Ds (Two-Dimensional Dynamic Displacement
Discontinuity Method solution). The field point algorithm has also been
implemented in a program called TW04Df. Both programs have been
structured in a manner similar to TWODD of Crouch and Starfield (1983),
and will jointly be referred to as TW04D.
3.4.2 Linear/Linear Formulation
The numerical Implementation of the higher order elastodynamic

displacement discontinuity method is similar to the equivalent higher
order elastostatic method. The only effective difference between the
32
static and forward-marching time-domain dynamic methods is the inclusion
of the source and receiver time loops in the dynamic algorithm. The
higher order static solution algorithm is given by
Hi “ 2j-i,2h fiij fij (3.50)
Hi “ Ay Dj (3.51)
where
Hi “ tu.. '‘nil
Hi “ [o'., "nh
fij - ED.. Dn]j
&1J “ [îii ^«n> ^nt» ^nnlij
lij " ' ®n«* ®mJiJ
s - shear effect, n - normal effect.
N — number of elements
For a displacement boundary value problem, i ranges from 1 to the number

of unknowns 2N in (3.50), and (3.51) is unused. For a traction boundary
value problem, i ranges from 1 to 2N in (3.51), and (3.50) is unused. In
a mixed boundary value problem, i ranges from 1 to 2 N, and a combination
of (3.50) and (3,51) is used. The dynamic algorithm can similarly be
written as
k" - fi1 £■ + £"'k+1 (3.52)
where
m - current time step number

& or £ or combination of & and £ - 4N x 4N coefficient matrix
£ - u or £ or combination of u and £ - 4N x 1 vector of known
33
boundary displacements and/or tractions
fl - 4N x 1 vector of displacement discontinuities.
N - number of elements
The matrices in (3.52) are fully populated In general. It Is clear that

the unknown displacement discontinuities at the current time step m
are obtained via a convolution between the known coefficients and known
displacement discontinuities from all previous times. Algorithm (3.52)
Is always Implicit, irrespective of the value of Ql, because matrix
is never strictly diagonal. "Cross-pollination" terms between
collocation points within the same element result in a system of
equations that must be solved to update the solution. For most
geometries, if Ql £ (1 - 1/20,3), then neighboring elements have no
effect on the current element until the next time step or later. In this
case, the solution algorithm could be interpreted as a mixed
implicit/explicit one.
At each time, and for each source-recelver combination, the
coefficients A end consist of sixteen sub-terms (hence the size of
4Nx4N), just as in the static case. The influence coefficients for the
master element in the global system are given by:
A,; - /iz(U1122 + 2 U1212 + U2211)

Agy - (A+2/i)p(U1222 + U2212) + A/i(U1112 + U1211)
A,, - (A+2/j)/i(U1222 + U2212) + Ap(U1112 + U1211)
A,, - (A+2/i)2 U2222+ 2A(A+2/i)U1212 + A2 Ullll (3.53)
and
figg - /i(U112 + U121)

Bgy - (A+2/OU122 + AU111
- ji(U122 + U221)
34
Byy - (X+2 f i ) V2 2 2 + AU121 (3.54)
and where UIJK and UIJKL are given by
Ulll - [-V11 + Will + (b-x)Vlll]/(2b)
U221 - [-V22 + W221 + <b-x)V221]/(2b)
U121 -[-V12 +W121 + (b-x)V121]/(2b)
U222 - [W222 + (b-x)V222]/(2b)
U112 - [W112 + (b-x)V112)/(2b)
U122 - [W122 + (b-x)V122]/(2b)
Ullll - [-2 Vlll + Wllll + (b-x)Vllll]/(2b)
U2211 - [-2 V221 + W2211 + (b-x)V2211]/(2b)
U1211 - [-2 V121 + W1211 + (b-x)V1211]/(2b)
U2222 - [W2222 + (b-x)V2222]/(2b)
U1122 - [W1122 + (b-x)V1122]/(2b)
U1222 - [W1222 + <b-x)V1222]/(2b)
U2212 - [-V222 ♦ W2212 + (b-x)V2212]/(2b)
U1112 - [-V112 + W1112 + (b-x)V1112]/(2b)
U1212 - J-V122 + W1212 + (b-x)V1212]/(2b)
for a left-hand side collocation point, and by
Ulll - [Vll - Will + (b+x)Vlll]/(2b)
U221 - [V22 - W221 + (b+x)V221]/(2b)
U121 - [V12 - W121 + <b+x)V121]/(2b)
U222 - [-W222 + <b+x)V222]/(2b)
U112 - [-W112 + (b+x)V112]/(2b)
U122 - [-W122 + (b+x)Vl22]/(2b)
35
Ullll - [2 Vlll - Wllll + (b+x)Vllll]/(2b)
U2211 - [2 V221 - W2211 + (b+x)V2211]/(2b)
U1211 - [2 V121 - W1211 + (b+x)V1211]/(2b)
U2222 - [-W2222 + (b+x)V2222]/(2b)
U1122 - [-W1122 + (b+x)V1122]/(2b)
U1222 - [-W1222 + (b+x)V1222]/(2b)
U2212 - [V222 - W2212 + (b+x)V2212]/(2b)
U1112 - [V112 - W1112 + (b*x)Vlll2]/(2b)
U1212 - IV122 - W1212 * (b+x)V1212]/(2b) (3.56)
for a right-hand side collocation point. Kernels VIJ, VIJK, VIJKL, and
WIJK, WIJKL are given in Appendices 1 and 2, respectively. The complete
solution algorithm, and the accompanying field point algorithm, has been
implemented in a FORTRAN-77 computer program,called TN04D(version 2).
3.5 Tangential Stress Calculation

3..J5.1 Llneflr/Cvnetant,Fvnnulotlgn
The computation of the tangential stress along the positive and

negative sides of the boundary mirrors the procedure outlined in Crouch
and Starfield (1983), except that all variables are functions of space
and time. The tangential stress can be written in terms of the
displacement discontinuity components. The tangential stress at the
center of the i-th element consists of the sum of a continuous and a
jump part:
36
1* 1 / 1- l*v 1 / 1- 1*\
fft " *°t > “ - at >
°t - * *t+> + (3.57)
The first terms on the right of (3.57) are the continuous parts and the
second terms on the right of (3.57) are the jump parts, and are given by
•|(oJ‘ + oj*) - [(A+2/t)p(V1112 + V1211) + A/<(V1222 + V2212)]D,i
+■ [(A+2/*)2 V1212 + A(A+2/i) (Vllll ♦ V2222) + A2 V1212JD*
(3.58)
1 (0^- - oj’) - JL-UD^costf1*1 - f)1) - D,l-1cos(/81 - -
[D^slntf1*1 - fil) ♦ D^sintf 1 - 0 1'1)]/^] (3.59)
where
d* - ai_lcob(/31 - /Ji_1) + 2a1 + a^cos^ 1*1 - 0l)
a1 - element 1 half-length
PL - angle of element 1 from x-axls, (3.60)
Note that the above tangential stress computation refers to the Induced
stresses. In order to compute the total or resultant tangential
stresses, the Initial stress state must be added to the Induced
tangential stresses.
37
In the linear/linear formulation, the computation of the

tangential stress is done slightly differently from the method used in
the linear/constant formulation. The computation of the tangential
stress along each side of the boundary is similar to another procedure
outlined in Crouch and Starfield (1983), except that all variables are
functions of space and time. The tangential stress is calculated on the
positive and negative sides of the i-th element, at each collocation
point, and consists of the sum of a continuous part, due to the normal
stress, and a jump part (constant along the element), due to the
displacements on each side of the element:
JLo*
1-./ ■
i-
i- - .2ft
7* r 8u;
.... _+--v-
l-i/ as i-i/ (3.61)
The first terms on the right of (3.61) are the jump parts and the second
terms on the right of (3.61) are the continuous parts, and are given by
K . <«,*>« - K h
da 2b
(3-62>
as " 2b
a - (ff„l)i or (ffnl)|i, if left or right collocation node,

respectively (3.63)
Note that the above tangential stress computation refers to the Induced
stresses. Inorder to compute thetotal or resultanttangential
stresses, the initial stressstate must be added tothe induced
38
tangential stresses.
3.6 Particle Velocity and Acceleration Terms

3.6.1 Linear/Constant Formulation
The particle velocities at arbitrary field points can b<#

calculated by using finite differences on the dynamic displacements.
Alternatively, the particle velocities can be determined analytically.
In the former case, the particle velocities at the m-th time step are
given by
at 2At
at 2At (3.64)
In the latter case, recall the following property of the convolution

operator (Eringen and Suhubi 1975):
8 -
at at
Since B^t-O) - 0, the velocities are given by
(3.65)
39
where
8B
” - p(V112T + V121T)
at
3B.„
“ - (A+2p)V122T + AV111T
at
aB
- p(V122T + V221T)
at
__2£ - <A+2m )V222T + AV121T

at (3.66)
and where VIJKT Is given in Appendix 3. The latter method is more

accurate, especially for computations close to the source elements, and
when the velocity pattern is changing rapidly with time relative to the
size of the time step.
If the velocity coefficients are temporally differentiated a
second time, in order to obtain particle accelerations, the
accelerations are found to contain a singular term l/(ct-r)°‘9, and
hence are infinite at the shear and compressional wave fronts, when ct -
r, and regular elsewhere. This is purely a mathematical phenomenon, and
is a function of the temporal discretization. The reader is also
referred to Aki and Richards (1980, p. 846) for further confirmation. A
smoother temporal discretization would result in non-singular particle
accelerations. In this work, the particle accelerations are calculated
using a central finite difference on the particle velocities. The
accelerations at the wave fronts will thus he non-singular.
The particle velocities at arbitrary field points can be

calculated by using finite differences on the dynamic displacements.
40
Alternatively, the particle velocities can be determined analytically.
In the former case, the particle velocities at the m-th time step are
given by
Hi
at
-< i
2At
Hi
at
C 1 - UT 1
2At (3.67)
In the latter case, the velocities are given by
- £ [ aB“ - D* + ,
1 ?" D*]
at £1 at ■ at "
^ -E [ "k - n.
k]
at h at • at nJ (3.68)
where
gB
— - - /i(U112T + U121T)
at
aB
— “ - (A+2p)U122T + AU111T
at
dB
— “ - p(U122T + U221T)
at
6B
— - - (A+2/OU222T + AU121T
and where U1JKT Is as defined In (3.55) and (3.56), but replace V1J,
VIJK, and WIJK by VIJT, VIJKT, and WIJKT, respectively. VIJT, VIJKT, and
UIJKT are given in Appendix 3. The accelerations are once again
mathematically singular at the wave fronts (ct - r), as is the case In
the linear/constant formulation, and are approximately calculated via a
41
finite difference on the particle velocities.
3.7 Symmetry Conditions
Symmetry conditions are incorporated in an identical manner to

that used in the static algorithms of Crouch and Starfield (1983) in the
case of the linear/constant formulation. The details will thus not be
repeated here. A line of symmetry with respect to a particular boundary
element is included by introducing an image element at the appropriate
location. Note that both the loading pattern in space and time as well
as the physical geometry of the problem must be symmetric in order to
use the symmetry option.
The implementation of the symmetry option in the case of the
linear/linear formulation is similar to the linear/constant case, but
more complicated. Care must be taken to Include the correct shape
functions during the imaging process. The symmetry option is easier to
implement in the linear/constant formulation since the displacement
discontinuity is assumed to be constant along each spatial element.
42
L CftttSALIH
Causality la defined simply as the relation between a cause and

its effect. In elastodynamics, the fundamental solutions define a causal
relationship between the source and receiver positions. A receiver
position cannot experience any dynamic motion until the fastest
(compressional) wave arrives from the nearest source position. Causality
is thus a time- and space-dependent property.
If an obstacle (e.g. a cavity) exists directly between a source
and receiver position, then the receiver position can first experience a
dynamic effect from the source position at a time corresponding to the
shortest route between source and receiver positions, taking into
account the presence of the obstacle. The receiver position is said to
be located in a "shadow zone" if it 1b on the opposite side of the
obstacle from the source position.
There has been much discussion in the boundary element literature
(e.g. Antes and von Estorff 1987, Triantafyllidis and Dasgupta 1987, and
others) about whether dynamic boundary element methods can successfully
capture the shadow zone behavior. It has been suggested that, in the
case of non-convex domains, these zones must be sub-divided into convex
domains with Interfaces to satisfy the transition conditions between
sub-domains. (A convex domain is one Where each point in the domain can
be connected to all other points in the domain by a straight line.)
The argument that dynamic boundary element methods are incapable
of modelling non-convex domains is incorrect. Boundary element methods
can model non-convex domains (e.g. see Tian 1990). Boundary element
methods that do not accurately model shadow zones do not have
sufficiently accurate integrations along the boundaries.
It should be clearly understood that boundary element methods are
derived from fundamental solutions that are valid only for infinite
(e.g. Stokes state (Stokes 1883)) or semi-infinite (e.g. Lamb's problem
43
(Lamb 1904)) convex domains. When a non-convex problem Is modelled In
statics or dynamics, the boundary element method solution is obtained by
specifying a stress or displacement field along a fictitious boundary,
which represents the real boundary in the problem, in such a way as to
obtain the correct boundary solution. Convexity is thus a non-issue
because the boundary element method always solves a convex domain
problem.
Causality is only important during the integration process. Care
must be taken to only integrate over the part of the element that is
dynamically active. Non-causal integrations introduce gross numerical
errors, and quickly lead to numerical instabilities.
44
5 WAVE MECHANICS THEORY
5.1 Basic Theory
There are analytical solutions available for stationary and

growing cracks under certain types of loading conditions. These
solutions are often given In complex arithmetic, and are also usually
limited to only certain components of displacement, velocity, and/or
stress In certain field positions. The reader Is referred to Freund
(1990) for further details. It is informative to highlight certain
details from Freund (1990) In order to better understand more
complicated problems.
In what follows, the continuum surrounding the crack is assumed to
be stress-free and quiescent prior to loading. In the case of a suddenly
applied (step function) pressurization an to a crack of finite or semi-
infinite length under plane strain conditions, the envelope of the
leading wave-front moves outwards at the compressional wave speed. For
field points near the crack faces compared to the distance to the
nearest crack edge (A in Figure 5.la), the transient response is a plane
compressional wave.
(a) (b)
Pdi
Figure 5.1: Wave-fronts due to a (a) pressurized and ( b) shear crack.
45
No shear wave is induced in this region by the applied normal traction.
As the plane wave passes a field point, the vertical stress component
Uyy changes discontinuously from zero to -ern (i.e. compressive). The
particle velocity also changes discontinuously from zero to on/(pc1) in
the direction of the wave motion. In other words, for the above-
mentioned region only, the vertical stress at the wave front reaches the
same magnitude as the applied traction, irrespective of the distance
away from the crack face) Note that there will always be at least one
point at the compressional wave front in position A. For field points
under the influence of a crack edge, compressional, shear, and head
waves (see Section 5.2) are generated, and field quantities attenuate
with distance away from the crack.
In the case of a shear traction am applied suddenly to a crack, a
region can exist (A In Figure 5.1b) where only a plane shear wave is
generated. In this region, the shear stress at all field points
along the shear wave front jumps from zero to -a, and the particle
velocity jumps from zero to.os/(pc2) In the direction of the wave
motion. Near a crack edge, compressional, shear, and head waves are
generated, as before.
If, instead of a Heaviside traction input, a crack is loaded with
a time-varying input traction, the resulting behavior can be obtained
from the foregoing by superposition (Freund 1990). For example, if the
input traction varies linearly from zero to a finite value over a finite
rise time, the stress (vertical for normal input, or shear for shear
input) at the wave front in field positions A of Figure 5.1 will be
zero, and then build up as the diffused wave front arrives.
An analytical solution exists for the case of a stationary,
suddenly pressurized crack of length L, as depicted in Figure 5.2.
46
0.3—i t-*o
-* K
y' ky
1 o*H(t)
g Two^h ► X’ ♦ ♦ ♦ ♦ ♦ + ♦ *
S u .. > li 73
*§) S • at <P
| « 2-
1?
(2 m
s < r-o n
Ifc-C,
U.-Uy
/ c d<
Figure 5.2: Normalized normal displacement of lower crack face versus

normalized distance (after Freund (1990).
The vertical displacement along the lower crack face is given in complex
form (Freund 1990) by
t
. . olx'|F(0)i?T t
u^x'.t) - nl 1 *-- f I ml F.CD K-l- - f)df
c2/ (5.1)
- —ix i I* I
«i
where
*±<f>
F±(r) - —
(j l ± r>s±(r)
o± ( 0 - (_L ± f)o.s
S±(f) - exp[ln S±(?)]
1
<>2
V « 9 2- 4 ) ( 4 V ) ) 0-3
-i «=2 _dij_
In S±(f) - - 1 1 tan'H
(-1 - 2 n2)2 9±f
cf
x' - -(x + L)
47
and where the Rayleigh wave velocity cR « c2(0.862 + 1.14u)/(l+v).
Freund (1990) calculates (5.1) numerically, for v - 0.3, and plots a
normalized crack face displacement against normalized "distance" (Figure
5.2).
Freund (1990) claims that Figure 5.2 is valid for all points along
the crack face, which is clearly incorrect. First, the sign of the
vertical axis should be negative. Second, if one considers the two
extreme cases when x' - 0 and x' - -L, it is clear that Figure 5.2 is
incorrect for the latter case, since one expects Uy - 0 when x ‘ » 0 or
x* - -L. The static analytical solution for a pressurized crack can be
used to validate this. For very late times, near the origin of Figure
5.2, when static conditions have effectively been reached, the static
analytical solution can be used to determine the exact slope expected
near the origin of Figure 5.2 for a particular position along the crack
face. The procedure is as follows. Following Freund's (1990) notation,
the static analytical solution (Crouch and Starfield 1983) is given by
u^(x+L/2) - nL(l-* < » g 2>!)0-3 (5 .2)
For example, when x - -L/2, the normalized crack face displacement at

the center of the crack, by substitution of (5.2), becomes
K — (l-»/)L (5>3)
Vi*
and the normalized "distance" becomes
Cjt 2c1t
The slope of the analytical solution in Figure 5.2, as time t tends to

infinity, from (5.3) and (5.4), is thus given by
48
- -(1-v) (5.5)
Using the above procedure, It is now possible to infer which position

along the crack face was used to obtain Freund's (1990) result, by
estimating the initial slope. The vertical displacements at two
positions along the crack face, obtained from TW04D, are plotted in
Figure 5.2, corresponding to x' - -L/4 and x' - -L/2.
Furthermore, an observation by Freund (1990), inferred from Figure
5.2, that retardation of the crack surface motion begins at the Rayleigh
wave arrival is incorrect. The axes of Figure 5.2 have been normalized
with time, and it is thus not possible to deduce when retardation occurs
using this graphical format. Retardation of the crack motion actually
begins at a later time.
Before the arrival of the compressional waves from the crack
edges, all applicable positions along the crack faces are governed only
by the applied crack pressure. The vertical displacement and velocity
along the lower face is given by -onc2zt/(^c1) and -onc22/(/iCi) >
respectively (Freund 1990). This displacement is the horizontal dashed
line plotted in Figure 5.2. TW04D matches these results.
5.2 Mechanics of Crack-Wave Interactions
The interaction of an elastodynamic stress wave and a crack is a

complicated process. The reader is referred to Rossmanith (1983) for a
thorough discussion on such interactions and the use of photo-elasticity
to Investigate them. Given the simple case of a plane elastic
compressional (F-) wave interacting with a stationary plane linear crack
(Figure 5.3), Rossmanith (1983) Identifies nine different dynamic zones:
49
I. undisturbed zone,
II. incident P-wave zone,
III. P-wave scattering zone,
IV. shear (S-) wave scattering zone,
V. P-wave reflection zone,
VI. S-wave reflection zone,
VII. head wave zone,
VIII. P-wave diffraction zone,
IX. shadow zone.
VIII
PrP
II
Figure 5.3: Schematic diagram of planar P-wave source and crack

interaction.
SO
Most of Che defined zones are self-explanatory. The so-called scattering
zones are the regions under the Influence of the crack tip reflections
and refractions. The so-called reflection zones are the regions tinder
the influence of the crack side reflections. Rayleigh waves can generate
on both sides of the crack just behind the S-wave front In region IV.
The shadow zone IX Is the region on the leeward side of the crack that
has not yet experienced any dynamic action because it is protected by
the crack. The head wave delimits the region between zones VII and VIII,
and is a diffracted shear wave. The reader is also referred to De Hoop
(1958) for the mathematical derivation of these zones for a crack with
zero thickness.
Rossmanith also depicts the interaction between a point source and
a crack (Figure 5.4). Note that both ends of the crack will create the
so-called wave scattering zones.
An elastodynamic numerical model should be capable of defining the
wave fronts and the shadow zone. The secondary zones (IV, VI, VII) are
extremely difficult to isolate practically or numerically because they
are superimposed on the major zones (I, II, III, V, VIII, IX).
P P
Figure 5.4: Schematic diagram of P-wave point source and crack

interaction.
51
TW04D is capable of defining the najor zones, but does allow some
"leakage" into the shadow zone, the degree of which depends on the mesh
resolution and severity of loading. In the linear/constant formulation,
the displacement discontinuity function is piecewise constant in space,
and the linear/linear formulation assumes a piecewise linear variation
in space. The jumps in displacement discontinuity at element edges thus
can allow erroneous wave scattering zones (type III, IV) to develop, and
hence transmission of dynamic waves into the shadow zone. This numerical
problem can be alleviated by increasing the number of elements along
each crack.
52
6 SINGLE ELEMENT RESULTS
A series of single element results are presented in this section

for both the linear/constant and linear/linear formulations. A
horizontal element is activated by a prescribed Heaviside shear or
normal traction. Figure 6.1 provides all the input details. Snapshots,
at a particular time step, of displacement, velocity, acceleration, and
principal stress vectors are detailed in Figures 6.2 through 6.17.
Figures 6.2 through 6.5 are the results of the shear traction
loading for the linear/constant formulation, and Figures 6.6 through 6.9
are the results of the normal traction loading for the linear/constant
formulation. The reader is reminded that the particle accelerations are
singular at the compressional and shear wave fronts (Section 3,6), and
are only calculated approximately by applying a finite difference to the
particle velocities. Care should thus be taken when interpreting the
acceleration results. They should only be interpreted in a qualitative
sense. The particle accelerations, and the particle velocities to a
lesser extent, highlight the positions of the wave fronts.
Figures 6.10 through 6.13 are the results of the shear traction
loading for the linear/linear formulation, and Figures 6.14 through 6.17
are the results of the normal traction loading for the linear/linear
formulation. Note that the peak particle velocities occur along the
shear wave front for the shear traction loading, and along the
compressional wave front for the normal traction loading.
Time histories of all non-zero components of displacement,
velocity, and stress are given in Figures 6.18 through 6.45 for three
field points in and out of the plane of the displacement discontinuity
element, as shown in Figure 6.1. Horizontal and vertical displacements,
velocities, and stresses (u*, Uy, vIt vy, a^, aw ) and shear stresses
(o^) are presented for both normal and shear input tractions, but only
for the linear/linear formulation. Figures 6.18 through 6.30 are the
53
results of the shear traction loading at all positions, and Figures 6.31
through 6.45 are the results of the normal traction loading at all
positions.
Figures 6.9, 6.17, and 6.45 show the large normal stress at the
compressional wave front, of the same magnitude as the applied normal
traction, as discussed in Section 5.1. Similarly, Figure 6.30 shows a
large shear stress at the shear wave front, of the same order of
magnitude as the applied shear traction (Section 5.1).
The dynamic displacement vector snapshots (e.g. Figures 6,10 and
6.14) detail lobes or eddies that move away from the source element with
time. These lobes are associated with higher particle velocities
(Figures 6.11 and 6.15) than the surrounding areas. The peak particle
velocities are located perpendicular to the source elements, and at the
wave fronts. Figures 6.29 and 6.43 detail the velocity jumps at the
shear and compressional wave fronts, due to the applied shear and normal
tractions, respectively (Section 5.1). However, in a general problem the
input tractions are usually more distributed in space and time, and
higher particle velocities are therefore often associated with the eddy
currents, as will be seen in Section 7.
54
(75,75)
y*
10m
(35,0)
(25,-25)
(0, -35)
window
(-75, -75)
Q1 - 0.3 for snapshots a s ,crn(MPa)

- 0.1 for histories
p - 2700kg/m3 +100
E- 70GPa
v - 0.2
Ax - 10m
Figure 6.1: Single Element Geometry and Loading
55
TMOtt Dynolo tuplM W M i at HU) tlM « tw
j|Bnl
- I tt. «ictr
tV t f*
V.' ',*
atunMry
Minima: I . O I ? M > M
HtMflM OOOMISM
'— t i l - MfMiicatian: 10001
••«•* »*■'
Figure 6.2: L/C displacements due to single shear source
TMMO; vaiMitiM at aotft tin itap: Mvtfo.Mt
\ \•0 * *
^ I 4 ** *
k «dX+**
t *+ *
• * * \J ** K * ~ * d d £ L % * •
’ I * •*V i •* ■
■" I
7\ " • ' " V " -
- - *
MfniticltlM: IX
•r * • i ' P + • <t^ « 1« *
* * • »>»»
( t 4 »■• **
/ ( t * * * ' *?
a direction
Figure 6.3: L/C velocities due to single shear source.
56
ttMllPlttOM It
Minihntiin: tt
Figures 4: L/C accelerations due to single shear source.
TKMtt OynaaU p rln c lM l i itt M M tla> (UR m m . d i t
Itnnon
CO«pr#t«K
•XX'Xxs'x
Boundary
HiniM t.IMab MPa

NtHM X Hte MPa
~++*x HtoniricitiM o<ri
ss-xy.'y* * x
■ d lra c tla n
F{t,gv>re S 5: L/C principal stresses due to single shear source.
57
TMMIfc tlm
Figure 6.6: L/C displacements due to single normal source.
a t MM tla a a U K BavM.Mt
\W\it ttt
. t r/j>* «■((!
\V%«a <k ^ t * .»«♦.*>***
k^ J ^ ^ ^
%'
' '
U I1 <' r* •
*
* * ** *% ‘ *
• M ••»«••
' •*'* ' I* I’ ' *
S».' *Mi I^
•»’'' • J*y
’*4«•"*'* •
r • ' % *» ^ ^
//r^r # | ^
✓ tfV/ 1 * 1
x dirt t|m_______
\I7
Figure 6.7: L/C velocities due to single normal source
58
M n lm tu m at
Figure 6.8: L/C accelerations due to single normal source
-H-N++—
i+
V. ' ^
Figure 6.9: L/C principal stresses due to single normal source
59
UN
1 . . 1&-* ‘ * *
Figure 6.10: L/L displacements due to single shear source
Mk
Figure 6.11: L/L velocities due to single shear source
60
A c n ln tla m i t
Figure 6.12: L/L accelerations due to single shear source
tlM
* + *XJr+* * •/
♦+*X*+ ♦ *
Figure 6.13: L/L principal stresses due to single shear source
61
t!M
Figure 6.14: L/L displacements due to single normal source
i a t M U t l x ita a : BnvBL.MI
% . . . » tr
K » « «• |» • « • /
^ * * # #• •• • » » * ^
•f | i•
e & <* i i i i i
- /*
S
MtniM 0 OQMM*e
i MUM 713194
M g m f )t*tior> I*
'' '
1 9* • *
«e«a I •« ( •• ^
# •It I••«I
44 \\
\/
Figure 6.15: L/L velocities due to single normal source.
62
T M M M MMllPfltlNI It
Si
Figure 6.16: L/L accelerations due to single normal source
Figure 6.17: L/L principal stresses due to single normal source
63
TMHOvK uy < • t l a a H M M : iu y a t.a a t
Figure 6.18: L/L vertical displacement at 1. single shear source
IMHO** *y «• tla a a t M M : l« y it.< a t
Figure 6.19: L/L vertical velocity at 1. single shear source.
64
TMMOvK ■*» >1 t l M I t 0 0 ,0 ) : U « r t l - * t
EigM.Ee..6.20; L/L shear stress at 1. single shear source
65
TWHOvt: UK r a t l M i t 0 0 . - M : U n s t .M t
Figure 6.21: L/L horizontal displacement at 2. single shear source
TMOKOvI: vx VI tlM at US, -n): IvuS.Mt
tin W
Figure 6.22: L/L horizontal velocity at 2. single shear source
66
Taoaora: i n n t l a a a t ( M l- M l i w a t . a s t
tla a W
Figure 6.23: L/L vertical displacement at 2. single shear source.
MMM n vs tla a a t H * M s lvyaf.Sat
Figure. 6.24: L/L vertical velocity at 2. single shear source.
67
TM M M a n «a t l a a a t H 4 M : l a x n l . a a t
-i.i
tla a W
Figure 6.25: L/L horizontal stress at 2. single shear source
TMMO«E am va tta a a t (H .-O I: u y y a a.a at
-i.i
tla a W
Figure 6.26: L/L vertical stress at 2. single shear source
68
m u m ra y n tla a a t H - B i la m r tl.a a t
Figure 6.27: L/L shear stress at 2. single shear source
69
TKHOvE UKv* tla at M.'M: Ueol.M
rt8vr?. .§.,t,Z8j-l</L h9rlg9nt.fll.,4i8placCTgnt gt ?, single shear ggvtrce.
TMMMfc vx «■ tla at W.-1S: ivnt.Ott
Figure 6.28: L/L horizontal velocity at 3. single shear source.
70
TN04DVB m y v i fclM a t 10.- M : lu y m l.a a t
figure 6.30: L/L shear stress at 3. single shear source
71
TN040VC BynwtB iMUnnt n tlaa i t HI: U n i.M
tu» HI
Figure 6.31: L/L horizontal displacement at 1. single normal source
tnmdvs valoclty n bibb <t OB.01: Inl.et
Figure 6.32: L/L horizontal velocity at 1. single normal source
72
1MMM: But n tlaa at H D i iawil.<*t
Figure 6.33: L/L horizontal stress at 1. single normal source
TWHOvS: Sw v l tlM i t (M.OI: U y y l.d tt
4.11
Figure 6.34: L/L vertical stress at 1. single normal source
73
Figure 6.35: L/L horizontal displacement at 2. single normal source
TNMM: va VI tlM a t MS.-BB: t n l . M t
IMP
-1.1
t i m W
Figure 6.36: L/L horizontal velocity at 2. single normal source
74
tiaa M
Figure 6.37: L/L vertical diaolaeement at 2. single normal source
tm m m >y w tim a t U5.-B S : Iv y t.M t
Figure 6.38: L/L vertical velocity at 2. single normal source
75
TMHOvB B n * • t l a a M M . ' M : t t w a . M t
M.I
tla a w
t w m : tyy «• t i n a t (ta, -a ft: layya.aat
tl a a W
76
TMMOvfc tar n l i e at H H R i itart.Mt
Figure 6 41: L/L shear stress at 2. single normal source
77
TMMvfc uy va U b i t A - M : w y l.M t
Figure 6.42: L/L vertical displacement at 3. single normal source
IMMDvfe vy va tla a a t | 0.> W : ivya.aat
tla a M
Figure 6.A3: L/L vertical velocity at 3. single normal source
78
H M M t n m t i n it M . * H : l*ral.*t
TMMOvft ■nr va m m a t 10.- M : U y y i.aa t
-ft.
79
7 APPLICATIONS
7.1 Solution of Mining Problems
In underground mining problems, the rock mass is subject to an

initial state of stress (due to gravity and tectonic influences). When
excavation (mining) or fracturing of the rock mass takes place, the
initial stress state is disturbed. The total stresses trtl at any point
in the rock mass can then be represented as the sum of the initial
stresses (<7 ^ ) 0 and the stress changes (due to mining) o'l3 at that
point. In elastostatics, the stress changes, or Induced stresses, are
those that are required by the displacement discontinuity method in
order to calculate the displacement discontinuities induced by mining.
The total stresses are required to check for failure under Mohr-Coulomb
conditions, for example.
If more than one mining step takes place then, for each mining
step, the induced stresses are the stress changes from one mining step
to the next. The total stresses are the sum of the initial stresses and
the induced stresses from all mining steps.
Similarly, the total displacements are calculated as a sum of the
initial and induced displacements. However, only the induced
displacements are usually of interest.
In elastodynamics, the induced displacements and stresses are of a
temporal nature, and are used by the displacement discontinuity method
to obtain the transient displacement discontinuities. The particle
velocities and accelerations are always obtained from the induced
displacements. Transient total displacements or stresses, at a
particular time, are obtained by adding the static total displacements
or stresses (at time t - 0) to the transient induced displacements or
stresses at that time.
80
7.2 Linear/Constant Formulation
7.2.1 Selbera's Pressurized Cylindrical Cavltv
An analytical solution, due to Selberg (1951), exists for the

tangential stress along the wall of an Infinitely long cylindrical
cavity, which is subjected to a dynamic Heaviside pressure loading. This
solution is used to partly validate TV04D. Figure 7.1 is a normalized
tangential stress versus normalized time plot for the case where the
pressurized cylinder is modelled with eight displacement discontinuity
elements over one quarter of a circle. Symmetry about both axes is used.
The Selberg problem is a non-trivial problem for the displacement
discontinuity method to model. The closed contour of elements can
represent two problems, depending on the direction of boundary
traversal. The first is the Selberg problem (exterior problem), and the
second is an infinitely long circular rod subject to a dynamic
diametrical normal tensile traction (interior problem). Furthermore, if
symmetry is not used for an interior problem, a mixed boundary value
problem must be solved generally, because extra elements are required
with zero prescribed displacements to prevent rigid body motions. Of
course, this does not apply in the Selberg (exterior) problem.
The interior and exterior problems are inextricably linked with
each other in the displacement discontinuity method. The perturbations
at normalized times cxt/R - 2, 4, 6 and 8 in Figure 7.1 are waves from
the opposite side of the hole that have travelled directly across the
hole.
81
TN04D: Tang stress vs tise along hole boundary: ltslgal.dat
Figure 7.1: Tangential stress versus
O
time for Selberg
0.0
0.117824
7.2.2 Shear Fault Davliehtlng Into Tabular Excavation
A simple example problem Is given to demonstrate an application of

TW04D. Two displacement discontinuity surfaces, representing an inclined
tabular excavation ("stope") and a frictionless "fault” surface, are
depicted In Figure 7.2, The entire fault surface, which daylights into
the stope, is activated at time t - 0 by a simple Heaviside shear
traction loading. This applied shear traction corresponds to the "static
excess shear stress” developed along the line of the fault due to an
initial compressive primitive stress field of 54 MPa vertically and 27
MPa horizontally, and assuming a Mohr-Coulomb failure criterion with
cohesion 10 MPa and friction angle 30°. Static excess shear stress, or
"stress drop," is the shear stress available to trigger a shear-based
event (Ryder 1987). The static excess shear stress is calculated, using
an adapted version of the elastostatlc displacement discontinuity method
program TWODD (Crouch and Starfleld 1983), when only the stope exists in
the rock mass.
The dynamic friction angle, dynamic cohesion, and fault
stiffnesses are assumed to be zero along the "fault," which is a worst
case scenario. The normal traction is assumed to remain unchanged during
the dynamic phase. The surrounding continuum is assumed to have elastic
constants of E - 70 GPa, v - 0.2, and p - 2700 kg/m3.
The shear loading causes the fault surface and the stope to move
interactively in a dynamic fashion. Figure 7.3 shows a time history of
dynamic particle displacements at two field points, one in the
hangingwall, and the other in the footwall. It is clear from Figure 7.3
that the peak dynamic movements are much larger than the static
movements that would arise if the same problem were loaded statically.
Note that the stope shields the hangingwall from much of the dynamic
action. Host of the dynamic behavior occurs in the footwall region
between the stope and the fault. Figure 7.4a shows the peak particle
83
velocities (with a maximum value of about 2 m/s), collected from all
time steps, on a vertically oriented grid surrounding the stope and
fault. Figure 7.4b depicts a "snapshot” in time of the particle velocity
vectors.
Figure 7.5 is a significant result. The dynamic (i.e. dynamic
phase only) displacement vectors in Figure 7.5a2 and Figure 7.5b detail
the presence of dynamic eddy currents that are set in motion by the
shearing of the fault surface. A series of snapshots in time reveals
that the eddy current pattern in the footwall of the stope moves quasi
parallel to the stope and away from the fault as time is marched. The
rock mass thus experiences "turbulence" during a shear-based seismic
event, causing rapid changes in stress with time. The local mining
geometry directs the motion of this "turbulence."
The eddy current along the footwall appears to be typical of a
Rayleigh wave. A similar phenomenon is visible below the fault in the
lower right corner of Figure 7.5bl, and this position is clearly remote
from any surface. The eddy currents and the apparent Rayleigh wave are
rotating shear lobes that are generated by the shearing source. A series
of snapshots of the dynamic (i.e. dynamic phase only) principal stresses
is shown in Figure 7.6.
Figure 7.7a shows the static excess shear stress around the stope,
before the introduction of the fault. Calculations based on the total
stresses (dynamic plus static) reveal that some parts of the rock mass
near to the stope and fault surface will fail (by exceeding the Mohr-
Coulomb shear strength with tension cut-off of the material) that would
not fail under static conditions. Hence, a slightly larger and more
severe failure envelope develops under dynamic conditions than under
static conditions (Figure 7.7b).
It is perhaps important to note that the dynamic excess shear
stress applied to the fault surface at time t - 0 in the above example
problem is almost certainly smaller than the real excess shear stresses
84
that would develop In the rock mass. This is because of the so-called
"ratcheting” effect (Kuijpers 1989), which is a build-up of stress in a
fractured rock mass as mining proceeds, due to non-linear movements on
joint planes and fractures. The stress build-up can exceed the primitive
stress conditions. A larger applied dynamic loading at time t - 0 would
increase the size and severity of the dynamic failure envelope.
85
150m
2000m
depth 1(150,75)
(75.40)
66*
150m
(0.0)
(90.48, -5.65)
window
Ax-5m
Q1-0.6
of*"
or(MPa)
-►t
Figure 7.2: Stooe-fault example problem geometry and dynamic loading
86
Figure 7.3: Displacement versus time for field points In footwall and
fafflgingWfllli
87
Figure 7.4a: Peak particle velocities for entire dynamic phase
TMMOt P art lc la v a la e ltlw a t
- *- . « u i u rr t
« 4 # # « , ^ ^ ^
******* * J* A/t* *
Figure 7.4b: Particle velocities at time step 50.
88
at tlM
iifie*tion booi
tin
figure 7.5a; Dynamic displacement vectors at time steps 10 and 30
89
n o n « m u diwin — t m e t m m m u tw» w n m , w
*444444444
**44444444
*44444444*
**444444*4
44444**4
4*441
44444
V»»*‘
M in i* * 9.000)01049
M lilM 0 03790N
m a g n ific a tio n 900*
*- t f
--- * * * * *
«•* ** P*
»**ffiPPPJPPP*
»»*ffP999ffi*
w M M M M i m
» » > f f fl t I II t 1
xdirxetlon
IW
Mft dyn—
indupuciwnt vcttr* tt ran til*fp; BwcTO.drt ^"|
M U M H M IM 1
11
4 * 4 4 4 4 4 4 4
M ***........
::
It
........ 9 I
4*
h :::::::
ii<«<........
5 1
i *. «........
i f f . ........
M a g n ific a tio n 9001
•. .•. "1' 11
M1itlt i t
........, .
. . . » 1 111
* titf...,.
fft
1 dlrwtlon
£lBW9 7 Dynamic displacement vectors at time steps 50 and 70
90
tmmq: iy w ic r i n s i n i t t r i M at totti t i a turn a m c H .H t
• i i• < ' • m«fliticatioA on
ft dtPMtlBH
1NQ40e rfymate prtnotpal i t n a m at 10th tin itip: Imeao.Mt
•**'// (UJ/"
\UJ'"
-x
'x
*
" *---
' x Ui/ /S//+
f / / x / /t-
- -“"Mr---- '■*
■»X 'V •
,s *• ■XX
(V— >*
~ Hx +
^Xx+
A\\xv-+ (111
\N-fcJf+ »rv,y\
1 K\ W
Figure 7.6a; Dynamic principal stresses at time steps 10 and 30
91
IMMPc Annate fPtnetiM >W a a t W t tla a I W ■tw aM .nt
U|»d
aa^^^M Ion*ion
« Coapnt s ai on
InM irr
Hmiaufe 0 .3 9 0 U B HP*
maiwa M MlOMB*
MtuhciliM 011
m VXt
/ ' / / / H iXlOlPMt
l lMI +
TMMOE dynaata p rin cip al atraaaaa a t n t h tta a atap: ■ avecn.dat
Ten*ion
Cenontatiat
i*
[H N im aua 0 8*1251 HP*
ii
N ia iau * M I k l HP*
Na0n i f i c * t i w t (I i>
]\\\%xx*
• • t I I\ | \ | \
IXVVxxtx
n u l l \\\ \yxxx
+?•• + » } ♦ } \ \ \ H >i\ X
‘'XXXxn a
idt
lraa
ltli
un i \ m
Figure 7 . L-Pynamlc principal stresses at time Steps 50 and 70
92
Figure 7.7a: Static excess shear stress around stopd
fMW »t .!>»-
Figure 7.7b: Mohr-Coulomb failure envelope for entire dynamic phase.
93
7.2.3 Two Reef Problem
A shear "fault" daylightlng Into two Inclined tabular excavations

Is shown In Figure 7.8. The initial and boundary conditions are the sane
as before: the applied dynamic shear traction loading Is the static
excess shear stress along the line of the fault (with only the two
stopes present), and the applied dynanic normal traction Is zero. Figure
7.9 reveals that the peak particle velocities are greater than for the
previous example (since the applied loading is more severe), with a
maximum value of about 8 m/s. Notice that the region experiencing the
largest particle velocities is orthogonal to the fault line. This is to
be expected since the applied loading is of a shearing nature. Figure
7.10 details a series of dynamic displacement vectors. Once again, the
circular lobes, which move away from the fault line with time, are
evident.
Figures 7.11 and 7.12 show the dynamic principal stresses and
total (i.e. static + dynamic) principal stresses, respectively, at a
particular time step. In this example, the dynamic action is strong
enough so that it is not swamped by the primitive stress field. Figure
7.13 depicts the static and dynamic Mohr-Coulomb failure envelopes. It
is clear that the dynamic envelope is significantly larger than the
static one.
The reader should bear in mind that the examples presented in
Sections 7.2,2 and 7.2.3 are not necessarily realistic, and merely
demonstrate the possibilities of the dynamic displacement discontinuity
method in investigating mining problems. In reality, the fault surface
would have some residual friction and cohesion during the dynamic phase.
Furthermore, the two sides of the fault surface should be restrained
from overlapping. Special elements can be included to model frictional
slip behavior and fault shear and normal stiffnesses, and to ensure that
the displacement discontinuity elements do not exceed a specified amount
94
of closure. In the examples in Sections 7.2.2 and 7.2.3, if the atope
height and fault thickness are assumed to be 1 m and 0.1 m,
respectively, then the two sides of the fault surface (and Btope) do not
overlap because the dynamic closures are small.
95
150 m
WKV\V<A-
2000m | 54 MPa | 1(175.125)
depth A
z
(190.12,93.307)
(9 6 .1 5 ,5 9 -1 1 )^ .—
i w T m*
vblosô 2 7 M Pa
^ ^ - 0 — (93-97, 34.20) 150m
(91.79,9.295)
w indow
(25, -25)
t t
Ax - 5m
Q1 - 0.6
<sT
-►t
Figure_7.8: Two reef problem geometry and loading
96
figure 7.9; Peak particle velocities for entire dynamic phase.
97
WMt mi ai»»iwtifi »t tew n» im a mw. *»>nw! ««»w.d»t |
at
Eigure 7.10a; Dynamic displacement vectors at time steps 10 and 30
98
INMOc iy n < u » l v o eto re a t H t l i t i n i t * mm i l . M y ltiM I : K n flO lt.M t
*t t k k k t 1 tl
'IHUUII
*'* i V U I I *
'k V \ U
‘M U
IV■ ■
»I1
♦*
Minimi* 0 OOOWJM9
f t )' N la im a 0 M 7 U I6
11 * *9A>'Italian iwi
11 •
S.1 t » I
»•5" "
w
| H U ^ \ U W
l U U U < v
x M pn U M
TMMft p n anal m u w it tw i t m «u» It reaw. — Tinatl: BwTOld.dit \

kk u u m m
“ u u U H U » * M
‘ u u u llllMi
‘ M UV IIMM
*»U *■ •‘ 'Vkk
M ill
M il r
e # ;;
tttrf*
\ttrt f moftiricatian loai
tttff
i 1 1 »♦
t i n t
t i n t
.linn
* llinn
t l l i m n
« direction
Figure 7.10b: Dynamic displacement vectors at time steps 50 and 70
99
•vn
L L
Elgure 7.11: Dynamic principal stresses at time step 30
UUl tu p a m f , tfwiftaht): ctvaoji.i
Figure 7.12; Total principal stresses at time step 30
100
N
I nrtfc TK»<i w witn i*w m ^ » r i m w w m w w m w w m j u w ii 4 ,5i21£J
LE8EWD
4S
■■m
Flewe 7.13; -Mohr-Coulomb failure envelopes for static and entire

dynamic phase.
101
7.3 Linear/Linear Formulation
7.3.1 Half-Plane Under Prescribed Time-Dependent Normal Traction
Ahmad and Banerjee (1988b) and Mansur and Brebbla (198S) have
modelled a half-plane subjected to a prescribed time-dependent normal
traction, as detailed in Figure 7.14. The linear/linear version of TV04D
has been used to simulate this half-plane problem by approximating the
surface as a row of 20 displacement discontinuity elements. This
approximate way of modelling a half-plane is correct until the
compressional waves first arrive at the nearest edge of the surface
elements. Thereafter, the solution approximates a half-plane only in the
region near to the center of the surface elements for early times,
before the arrival of the reflections from the edges of the row of
surface elements.
The material properties are E - 2x103 lb/in2 (1.379 GPa), v -
0.15, and mass density p - 1.9534x10'* lb-s2/in* (2089 kg/m3). The
vertical stress and displacement at three field points are shown in
Figure 8.15 (originally from Ahmad and Banerjee 1988b). All three
approaches give similar results. Ahmad and Banerjee (1988b) use a
Laplace domain formulation, whereas Mansur and Brebbia (1985) use a time
domain direct scheme. The fourth result quoted in Figure 7.15, by Tseng
and Robinson (1975), is from a finite difference formulation. These
results deviate from the boundary element results because the finite
difference results include contamination from an inadequate transmitting
boundary, which is needed to simulate the semi-infinite domain, It
should be noted that elastodynamic finite difference methods do exist
with effective transmitting boundaries (e.g. Cerjan et al. 1985).
102
£ *90oklt
\r* OiiC
*t=5*» Ctl'O.Hi) 1kil
«■ Tlaf
%
s /ys /y/ ,v* y
* 'X/ /'/ ; / /V/ /////// *
/// / ✓*✓y✓<.‘
.
'////,'////
"•“** ^rjCttWeSllM
♦y
p*qn
/IX ,11 a .
‘(150',-10’)
x*(80\ -60’)
(75’,-75')
Figure 7.14: Geometry and loading on a half plane

-15.2 •6
*1
- 10.2
t A
i "i •
o
A .vamp
-5.1 •2 nldf IW (men*
o Nmuf
X TwftAVX
0.0
— \0 6o e0 100 12
Fi|ur« 7(b). Tim* Muory of diipiteem enl* , i t the internal point F(KT, -tff)
- 10.2
-S.i
t
V)
s wovr arrives ™K
-a.s
p wove arrives — I
o.o
Tin* (noMi .
Figure I. Time hUlorjr of ditptactmenl H| at the intentel point Of I SO, -10*1
-+.L
- 2.8 K TW0^.1>V2.
* * * • I f o rr- «
o«
-14
sioric solution
0.0
tla*(ant,
Figure 9. S iren « „ * ' internet point B(7S', -1ST
Figure 7.15; Field poln^ results from different methods (after Ahmad and
Bartertee 1988b).
104
7.3.2 Inclined Crack Interaction with Shear Wave Source
The interaction between a shearing source and an inclined

traction-free crack is demonstrated here. Figure 7.16 shows the geometry
and loading pattern. A series of snapshots of the dynamic displacement
vectors is shown in Figure 7.17. Note the shearing lobes that
progressively move out from the source elements.
Figure 7.18 details particle velocity vectors. The compressional
and shear wave fronts are now clearly visible. A reflected wave from the
surface of the inclined crack is evident in Figure 7.18b.
The principal stress vectors in Figure 7.19 also highlight the
wave fronts and reflection. The reader should be aware that the wave
fronts in the principal stress snapshots may appear to be uneven only
because of the resolution of the field point positions. The stresses
vary rapidly in time, and field points can easily capture or miss a
stress peak. Figure 7.19a clearly highlights the shadow zone beneath the
inclined crack. The resolution of boundary elements along the Inclined
crack is sufficient so that no leakage of the stress waves through the
inclined crack is evident.
105
J i —
(75,0)
4 elements o8
(19.485,-33.75)
(-19.485, -56.25)
window
(-75, -150)
o s (MPa)
Q1 - 0.6
v -0 .2 +100
p - 2700 kg/m3
E - 70 GPa t
4At
Figure. 7.16: Geometry and loading of shear source and crack Interaction.
106
o u M M e iiapn tiaa Mt
ic a it e n io o o i
O y n m e diaplacwant vaetora at tiM
x dlrtctfton
Figure 7.17; Dvnamlc displacements at time steps 50 and 100
107
MiniM* ftUlMt-IB
N tg n tr ie a tiM ft'
v ilH ity w e t a n a t
Ni|Ai(uiiion ft*
Eieure 7.18: Particle velocities at time steps 50 and 100.
108
1*040*1! OimUG irtM lo o I « t r » i m a t OOM U N M * omoo.ttt |
u p d
Itnaien
tounflary
i
NiniM* 0 NI»>9 *»•
Hail MM »14H M
Ha*A»ueatian o.fti
x flirte tiM
tt
0b*
s d ic ta tio n
Figure 7 ! Dynamic principal stresses at time Steps 50 anA inn
109
8 SPECIAL ELEMENTS
The displacement discontinuity method has the extremely useful

feature that, even though the surrounding continuum Is limited to linear
elastic assumptions, the displacement discontinuity elements may exhibit
any desired internal linear or non-linear behavior. Other boundary
element methods do not have this advantage. The elastostatlc
displacement discontinuity methods routinely include special elements to
model non-linear material behavior of interest in geotechnical problems.
For example, the displacement discontinuity function can be governed by
a Mohr-Coulomb law to simulate frictional behavior between each side of
the element. Special elements have been designed to numerically model
linear and non-linear backfill material, and also to model seam-type
behavior. The reader is referred to Crouch (1980) for details.
The elastodynamlc displacement discontinuity methods can also
utilize special element behavior. Special elements have been included in
TW04D to simulate such behavior.
8.1 Elastic Joint Element
The simplest special element to include in the static and the

dynamic displacement discontinuity methods is the elastic joint or
stiffness element (Crouch and Starfield 1983). A joint element is an
element containing a compressible filling, and is modelled as an elastic
spring with shear and normal stiffnesses that connects each side of the
displacement discontinuity element. The spring stiffnesses represent the
joint's elastic properties (Figure 8.1).
110
Figure 8.1; Representation of lolnt element.
The element Is assumed to have a thickness (h). Following Crouch and

Starfield (1983), the Induced stresses can be written in terms of the
displacement discontinuities as
o. - -K. D,
- '*» J>„ (8.1)
where
K, - ft0/h
^ ~ E0/h
fiQ, E0 - joint shear and Young’s moduli, respectively
h - joint thickness.
In general, the stiffnesses In (8.1) can be made time'dependent, but

have been restricted to being constants in this work.
In order to demonstrate the above, Figure 8.2 depicts a traction-
free vertical crack that is activated by a nearby shearing Bource.
Figure 8.3 shows the vertical displacements generated at two field
points for a control problem, where the crack is not present, and Figure
111
8.4 shows the responses at the sane two positions when the unfilled
crack is included. The correct modelling of the shadow zone effect
(Section 5) is clearly evident in Figure 8.4. The small oscillations at
the late times in Figure 8.4 are real, and are the secondary reflections
from the source element due to the original reflections from the
unfilled crack.
If the crack is now assumed to be a joint with shear and normal
stiffness, then Figure 8.5 shows the responses at the two field points
when the crack properties are the same as the surrounding material. The
results are identical to the control problem in Figure 8.3. Figures 8.6
through 8.8 are the results due to progressively softer filling
material. Figure 8.9 is the special case when the filling material is
assumed to have zero stiffnesses, and the results match Figure 8.4.
It is evident from Figures 8.6 through 8.8 that the inclusion of a
filling material in a crack can in fact temporarily increase the dynamic
responses at the field points. It is likely that this occurs because the
diffracted wave arrivals from the source or receiver crack edges can
become superimposed with the wave arrivals due to transmission through
the filling material.
In the case of an infinitely long crack under the action of an
Incident plane compresslonal or shear wave, Hyer et al. (1990) note that
as the elastic stiffness (i.e. K„ only for compresslonal wave, and K,
only for shear wave) of the crack is decreased, the transmitted wave is
both slowed and attenuated. The attenuation is characterized by
decreasing amplitude and filtering of the high frequency components.
Careful study of the results of Hyer et al. (1990) reveals that some
parts of the transmitted wave actually temporarily increase in amplitude
with decreasing stiffness. The results of Figures 8.6 through 8.8 are
thus not unreasonable.
Figures 8.10 and 8.11 detail the responses for the cases when the
filling materials contain zero shear and normal stiffnesses,
112
respectively. It is clear that, since the source generates mainly shear
waves, the shear stiffness has the major effect in terms of allowing the
transmission of waves through the filling material.
113
y (m)
&
h = 0.1 m
2.5
0.5
-► x (m)
W 5 6
-o J
-2.5
c s (MPa)
Ax =0.5 m
+100 Q1 = 0.6
P - 2 7 0 0 kg/m3
v = 0 .2
t
5At E - 70 G P a
M> - 2 9 .1 7 G P a
Figure 8.2l_Geometrv_and loading of lolnt activated bv shearing source.
114
I
TM40S ttPMMl flWci—nt vs U w M,0)«ssiua M tt.0) d— sssr fttsst.dst
.HNS
t. S.
tl— M
I
l
I
Figure 8.3: Responses without joint present.
TMMOc Vsrtlsal dlsplsesssnt vs tlss: <4.0)«ssUd snd M,0)*dseh#* luy.dst
/"\
Figure 8.4: Responses with unfilled crack present
115
M t t Mrtlnl aminmiit »• tlw h H nfMMi lu»WOO,<*i
Figure 8.5: Responses with filled lolnt present: loint - rock

properties.
1MMB VMICIl tliplWHt II U K 14,0)•Mil*. ( M I M M * lUyMO.Mt
Figure 8.6: Responses with filled lolnt present; 1 ^ - 7 0 GPa/m.

K. - 29.17 GPa/m.
116
IWMtt Vtrttul I M U M M VI tlar (4.01-MU*. H W - I H M di I W 7.MI
10.11Uk
tlH tel
Figure 8.7: Responses with filled lolnt present; 1 ^ - 7 GPa/m

K. - 2.917 GPa/m.
fWHOt vortlcol d U M m unt v» tlac ROlMOitd *nd A d l - d M i M tuyKOT.Mt
Figure 8.8: Responses with filled lolnt present: K„ - 0.7 GPa/m

Ki-s.,,.0 ■29_gPfl/ffl,
117
im m b v a rtlc a l d li» U n —nt n t l v <4M «f»Iid and MLOIidddlndr luyko.dat
tf» M
Figure 8 9: Responses with filled lolnt present: K„ - Kf - 0
118
( l * tfcf* ,Jt*t•
tlw 111
/
Figure 8.10: Responses with filled lolnt present: Kn - 70 GPa/m. K, - 0
TMMOt Mitlnl M e l m M t n U K HMwIll M. 01 Iuyli70t.dat

( |> 4 ; fit• 3.4 f c f e
Figure 8.11; Responses with filled lolnt present: K_ - 0. Kt - 29.17 GPa/m
119
9,2 <?9h9gl9nlegg-FrictlonUgB Fault Element
A cohesionless-frlctlonless fault is a joint with a compressible

filling that obeys the Mohr-Coulomb law under conditions of zero
cohesion and zero angle of friction (Crouch 1980). The induced normal
stress is given by (8,1b). In order to satisfy the Mohr-Coulomb law
under these conditions, the total shear stress is zero along the fault.
The induced shear stress is the prescribed dynamic shear traction
loading. This element is demonstrated in Section 8.3.
8.3 Rapid-Yielding Hydraulic Prop Element
Rapid-yielding props are a form of support designed to stabilize

mining excavations on a local scale during a seismic event. A 400 kN
rapid-yielding prop element response curve is shown in Figure 8.12
(after Ryder 1988).
-F„(kN )
"* -50 50 *10?) 150 200 250 300 350 ' * D" <mm)
Eizure _B_.12: Loading curve of 400 kN rapid-vleldine hydraulic prop

fllwwnt*.
120
The prop is ideally installed "prestressed" at 200 kN, and behaves in a
linear elastic way until yield at 400 kN. At yield, the prop operates as
a one-way system that yields under compression only. If unloading occurs
during the yielding process, the prop is assumed to only recover the
elastic phase until loading resumes, as shown in Figure 8.12.
This element has been Included in TW04D. Logic has been included
to make the prop element fail if the ride (or shear component of
displacement discontinuity) is greater than a preset tolerance during an
unloading phase, after the entire elastic phase has been recovered (at 0
kN), if the closure velocity exceeds 3 m/s at any stage during loading,
or if the maximum stroke of 350 mm is reached. Any density of props per
element can be chosen. Note that the plane strain assumptions imply that
the prop elements should be interpreted "per unit distance” in the anti-
plane direction.
As a simple demonstration of the cohesionless-frictionless fault
and prop elements, the single excavation problem of Section 7.2.2, as
shown in Figure 8.13, is redone, with the fault modelled using
cohesionless-frictionless elements. Part of the stope is assumed to
contain prop elements, as shown in Figure 8.13. At time t - 0, it is
assumed that the props are installed such that each prop is prestressed
to 200 kN. The fault surface is assumed to have a Young's modulus of E0
- 50 GPa and thickness h - 0.1m. The dynamic loading applied to the
fault surface is the same as for the problem in Section 7.2.2.
The dynamic displacement vectors at the 30-th time step are shown
in Figure 8.14 (compare with Figure 7.5a2). Notice that the displacement
vectors near to the fault run more parallel with the line of the fault
than in the case of Figure 7.5a2 because the fault stiffness reduces the
normal displacement discontinuity component along the fault.
The props have a minor influence on the dynamic behavior of the
stope because they are only able to support relatively small loads
compared to the applied dynamic loading in this case. The props in
121
elements 20, 19, and 18 fall at time steps 3, 5, and 8 respectively,
because closure velocities exceed 3 m/s at these elements. Prop element
17 remains stable for the duration of the dynamic phase.
122
y>v>vx
54 MPa
2000m | (150,75)
depth
(93.97; 34.20)
C -0
27 MPa
(0,0)
(90.48; -5.65)
window
(0,-75) |
Ax -5 m
Q1 - 0.6
-►t
(Elements 1 7 ,18,19,20 contain 400 kN
props, with density of S/element)
(Ride tolerance ■ 0.010 m)
Elgure,-g,.l?;..St9P9 vith props f l i g h t i n g fault interaction.
123
TMMtt tlH
MilM O.OIHMf
Figure 8.14: Dynamic displacement vectors at time step 30
124
9 STABILITY
The algorithms presented In this work are technically numerically

unstable. If the algorithms are lurched for a sufficient number of time
steps, numerical noise will begin to swamp the real behavior of the
system. The visible Instabilities occur later in the linear/linear
formulation than in the linear/constant strength formulation. The small
oscillations barely noticeable at the late times in Figure 9.1 are the
start of a visible decay in the quality of results.
Figure 9.1 is the identical result to Figure 7.3, which was
obtained from the linear/constant formulation. Figure 9.1 depicts 150
time steps with Q1 - 0.6 (see 3.41) of that 28 element problem. Figure
9.2 is identical to Figure 9.1, except that the linear/linear
formulation is used instead. The reader is reminded that the
linear/linear formulation requires twice the number of equations that
the linear/constant onedoes for the same number of elements.
Figure 9.3 is theidentical result to Figure 9.1, but in this case
the linear/constant formulation influence coefficient calculations are
performed in single precision instead of the usual double precision.
Notice that the single precision results are worse than the double
precision ones.
Figure 9.4 shows 300 time steps with Q1 - 0.6 of a 56 element
version of the same problem as before, using the linear/constant
formulation with double precision again. Now the numerical instability
is clearly visible. Figure 9.5 is the same as Figure 9.4, but single
precision has been used.
Figure 9.6 is thesame as Figure 9.4, but now the 56 element
problem is computed for 360 time steps with Q1 - 0.5, and the
instability is clearly less severe. Reducing the time step in this case
reduces the severity of the emergent instability, even though more time
steps are needed to cover the same total time span.
125
Finally, Figure 9.7 is identical to Figure 9.2, except that the 28
element problem is now run for 500 time steps. The instability is now
evident.
It has been suggested (Ahmad and Banerjee 1988a, and others) that
the Implicit time-marching formulations in boundary element methods are
unconditionally stable. This is questionable, and has never been
mathematically or numerically justified to the best of the author's
knowledge. An implicit scheme does not guarantee stability.
In an attempt to quantify the numerical instability issue, the
spectral radii or eigenvalues of the Influence matrices in the boundary
element methods have been determined following a procedure outlined in
Smith (1985). From Smith (1985), given a system of equations, where the
solution at the (i+l)-th time step is related to the known solution at
the i-th time step as
Hi+i - A Mi + fei (9.1)
where
- vector of known boundary values

A “ "influence" matrix
then it can be shown that the equation system is stable if the spectral
radius
P(A) s 1 (9.2)
and convergence is guaranteed if
p (A) * II A II * 1. (9.3)
126
The time-narching solution algorithms (3.40) or (3.52) can he re
written In a form similar to (9.1) that is amenable to quantifying their
behavior. Following the notation of Section 3, define an operator H such
that
C - BV (9.4)
where
- (fi1)'1 ( I W Lk - Sk-l,rl f * 1 H*) (9.5)

fe* - I*
ba - nk.2>B Lk b1 (9.6)
Combining (9.4), (9.5), and (9.6) yields (3.40) or (3.52). Now define
another operator such that
C” - ]f1 (9.7)
From (9.4) it follows that
jje-i _ jja-i (9.8)
and hence
- (Jf1)"1 ff""1 (9.9)
Substituting (9.9) into (9.4) gives the required form of (9.7) where
E»-jf(Ifi)-i (9 10)
For numerical stability, the spectral radius of operator £" must be less
127
than or equal to unity. In other words,
Pip) s l r h i <9 .i d
The above algorithm has been Implemented numerically to test for

stability. Both the linear/constant and the linear/linear formulations
produce spectral radii greater than one. It should be noted that the
operator in (9.7) is time-step dependent, whereas the influence matrix
in (9.1) is constant. It is not clear whether this should change the
stability criterion (9,11). Note that algorithm (9.7) could also be
written differently so that the operator refers back to the first time
step instead of the previous one.
The reader is referred to Cole et al. (1978) for details of a more
sophisticated stability analysis on the direct boundary element method.
The reader is also referred to Peirce et al. (1990) for a detailed
analytical study on the convergence properties of certain boundary
element methods. Finite element methods (e.g. Smolinski et al. 1987) and
finite difference methods (e.g. Smith 1985) can also be checked for
accuracy and stability by examining the eigenvalues of the system.
Numerical instabilities have also been noted in the three-
dimensional elastodynamlc displacement discontinuity method (Mack 1991),
in the two-dimensional elastodynamlc direct integral method (Tian 1990),
in the two-dimensional fictitious stress method (Tian 1990), and in the
three-dimensional fictitious stress method (Loken 1991). The numerical
instability problem in boundary element methods is fundamental to the
wave equation. In the wave equation, any wave motion generated by a
source function will continue indefinitely. Wave motion is spatially
attenuated as waves move away from the source, but the total energy of
the particle motion is constant (Akl and Richards 1980). Any
discretization errors introduced into the system, because of the
polynomial approximations of the displacement discontinuity function,
128
cause the total energy to Increase. As time Is marched, the accumulation
of these errors eventually begins to swamp the solution. Hence, more
accurate integrations (causal as opposed to non-causal), higher order
formulations, and numerical averaging techniques will delay the onset of
noticeable numerical noise. Temporal and spatial averaging techniques,
and especially repetitive temporal averaging techniques (e.g. Manolls et
al. 1986), are not desirable because they effectively damp the dynamic
behavior.
The use of double precision instead of single precision in the
kernel calculations of the computer program further delays the onset of
these instabilities. Due to the nature of the two-dimensional
fundamental solution, dynamic effects do not cease after the shear wave
arrival, as is the case in three dimensions, since wave effects continue
to arrive from all points along the infinitely long out-of-plane
integrated direction. This implies that the computer program can
experience precision problems when time t is large. The three-
dimensional elastodynamlc displacement discontinuity formulation (Mack
1991) does not have this particular problem, since all dynamic effects
are identically zero after the arrival of the shear wave.
Finite difference formulations of the wave equation do not seem to
have a numerical Instability problem. These methods do, however, suffer
from a numerical "mesh dispersion” problem, in which the quality of the
solution becomes Increasingly worse as more and more grid points
separate source and receiver locations. A stage can be reached where
numerical noise again swamps the real behavior. Dispersion can be
alleviated by the introduction of a higher order difference formulation
(e.g. see Bayliss et al. 1986).
In the boundary element formulations, it is unclear whether a
higher order method will ever he sufficient to eliminate numerical
instabilities. Perhaps the only way to eliminate these instabilities is
to force material attenuation of the system by the introduction of some
129
form of damping.
'From a practical atandpolnt, however, the dynamic formulations
presented In this work are clearly uaeable, even for fairly
sophisticated and lengthy numerical simulations. The accurate analytical
integrations ensure that the numerical instabilities only become
physically evident after a large number of time steps.
130
TMMX X dlapl n tiaa at <7&40l<daaftad and (7H. I S ^ n ll± aM.dat
xl
Figure 9.1; L/C responses: Ql-0.6. tielem-28. nt-150. double prec
twuov Z: x d iap i >■ tta a a t ro a e i-d e e n e d and Its. 18) -w ild ; anvt.dat
Figure 9.2: L/L responses: 01-0.6. nelem-28. nt-150. double prec.
131
TMMtt x alapl n tlaa a t (m <M "*oaaad ana m n i ^ l l t oMta.aat
Figure 9.3: L/C responses: 01-0.6. nelem-28. nt-150. single prec
1*040: x Btapl n t i n a t ■ I la id pta: aratn«170,40). uni t a - 170. IB): an.Bat
Figure 9.4: L/C responses: 01-0.6. nelem-56. nt-300. double prec
132
T M M M i tflap l va t l a a a t (7B. 40) ■— h id M (7 B ,in « M ltd : fl.a a t
/**v
Figure 9.7: L/L responses: Ql-0.6. nelem-28. nt-500. double prec
134
10 CONCLUSIONS
Two versions of a two-dimensional time domain elastodynamlc

displacement discontinuity method have been developed and demonstrated
for a variety of geomechanics problems. This work presents the first
general elastodynamlc displacement discontinuity method for two-
dimensional problems. The resulting algorithms are designed sp that
mixed boundary value problems can be solved.
Section 6 details the dynamic patterns that develop around a
single displacement discontinuity element, loaded by either a Heaviside
shear or normal traction source. These results show that, for this
loading configuration, the most abrupt dynamic response occurs
perpendicular to the element at the leading wave front. This is
confirmed by available analytical solutions.
The results of Section 7 show that complicated dynamic patterns of
displacement, velocity, acceleration, and stress can develop in a
general problem. The compressional and shear wave fronts, and the eddies
that develop due to both compressional and shear traction sources, are
associated with rapid changes in stress and higher particle velocities.
Further experimentation is required to fully quantify these effects.
Special elements were demonstrated in Section 8. The joint element
problem of Section 8.1 indicates that temporary amplification of dynamic
waves can occur near a joint, depending on the shear and normal
stiffnesses of the joint relative to the host medium.
The linear/linear version has been shown to be more stable than
the linear/constant version (Section 9). The linear in space
approximations of the displacement discontinuity function in the
linear/linear method are more accurate than the constant in space
approximations in the linear/constant method, and therefore introduce
smaller discretization errors, thereby improving the stability of the
algorithm.
135
Another advantage of the higher order method ia that field points
can be located much closer to source elements in the linear/linear
formulation than in the linear/constant formulation without any loss in
accuracy, typically as close as one tenth of an element length, as is
the case in elastostatics. In the linear/constant formulation, a field
point should not be located within an element length of a source
element, as is the case in elastostatics. Therefore, elements,can be
oriented closer to each other in the linear/linear formulation,
resulting in a greater variety of problems that can be modelled.
In conclusion, this work provides the basic tools for studying the
interactions between elastodynamic waves and tabular excavations. The
development of further special elements to model frictional slip along
joints and the automatic growth of cracks will extend the use of this
method.
136
11 RECOMMENDATIONS
A number of research areas should still be investigated with

respect to the elastodynamic displacement discontinuity method. The use
of some form of damping should be investigated as a possible way to to
eliminate numerical instabilities. In order to reduce the calculation
times, an algorithm that automatically chooses the time-step size
according to the level of dynamic action at a particular time would be
of interest. Such algorithms are available in finite element methods.
The implementation of the space translation property of the influence
coefficients would considerably speed up calculations for problems where
a regular mesh occurs. Variable element sizes should be implemented in
the linear/linear formulation. The anti-plane components can also be
Included in both formulations.
It is possible to include a facility in the models to predict the
automatic growth of fractures. This could be designed so that fracture
propagation occurs along a prescribed surface, with the rate of
propagation determined by the models, or at a prescribed rate, with the
direction of fracturing automatically calculated, or so that both
direction and rate are automatically calculated according to a fracture
criterion.
Special elements that can still be implemented include Mohr-
Coulomb frictional elements, slip-weakening frictional elements (Andrews
1985), and velocity-dependent frictional elements to model fault slip,
crack-tip elements for fracture mechanics applications, and non-linear
backfill elements to model backfill under transient dynamic loading.
The displacement discontinuity method can also be combined with
either a fictitious stress method (Tian 1990) or a direct boundary
element method (Tian 1990) so that problems with both tabular and non-
tabular excavations can be modelled. Such a combined method could
include multiple material types so that an inhomogeneous material can be
137
modelled.
From the success of the two-dimensional formulations, It can be
Inferred that there is merit in developing a higher order three-
dimensional elastodynamic displacement discontinuity method, even though
this would entail a significant amount of work. Flat triangular or
general quadrilateral elements are recommended so that elements can join
each other at arbitrary angles.
138
REFERENCES
Achenbach, J.D. 1973. Wave Propagation In Elastic Solids. Amsterdam:

North-Holland.
Ahmad, S. and Banerjee, P.K. 1988a. Tlme-Domaln Transient Elastodynamlc

Analysis of 3-D Solids by BEM. Int. J. Num. Meth. Engng. 26. 1709-
1728.
Ahmad, S. and Banerjee, P.K. 1988b. Multi-Domain BEM for Two-Dimensional

Problems of Elastodynamlcs. Int. J. Num. Meth. Engng. 26. 891-911.
Ahmad, S. and Banerjee, P.K. 1990. Inelastic Transient Dynamic Analysis

of Three-Dimensional Problems by BEM. Int. J. Num. Meth. Engng.
29. 371-390.
Andrews, D.J. 1985, Dynamic Plane-Strain Shear Rupture with a Slip-

Weakening Friction Law Calculated by a Boundary Integral Method.
Bull. Seism. Soc. Am. 75. 1-21.
Antes, H. and von Estorff, 0. 1987. On Causality In Dynamic

Response Analysis by Time-Dependent Boundary Element Methods.
Earthquake Engng. and Struct. Dyn. 15. 865-870.
Aki, K. and Richards, P.G. 1980. Quantitative Seismology: Theory

and Methods. New York: Freeman and Company.
Banaugh, R.P. and Goldsmith, W. 1963a. Diffraction of Steady Acoustic

Paves by Surfaces of Arbitrary Shape. J. of Acoustic Soc.
of Am. 35. 1590-1601.
139
Banaugh, R.F. and Goldsmith, U. 1963b. Diffraction of Steady Elastic
Waves by Surfaces of Arbitrary Shape. J. of Applied
Mechanics. 30. 589-597.
Banerjee, P.K., Ahmad, S. and Manolis, G.D. 1986. rransient

Elastodynanlc Analysis of Three-Dimensional Problems by Boundary
Element Method, Earthquake Engng. and Struct. Dyn. 14. 933-949.
Banerjee, P.K., Ahmad, S. and Manolis, G.D. 1987. Advanced Elastodynamlc

Analysis, in: Boundary Element Methods In Mechanics, vol 3.
Computational Methods in Mechanics series, ed: D.E. Beskos. New
York: North-Holland.
Bayliss, A., Jordan, K.E., LeMesurier, B.J. and Turkel, E. 1986. A

Fourth-Order Accurate Finite Difference Scheme for the Computation
of Elastic Waves. Bull. Seism. Soc. Am. 76. 1115-1132.
Beskos, D.E. ed. 1987a. Boundary Element Methods In Mechanics, vol 3.

Computational Methods in Mechanics series. New York: North-
Holland.
Beskos, D.E. 1987b. Boundary Element Methods in Dynamic Analysis.

ASME Applied Mechanics Reviews. 40. no. 1. 1-23.
Cerjan, C., Kosloff, D., Kosloff, R. and Reshef, M. 1985. A

Nonreflecting Boundary Condition for Discrete Acoustic and
Elastic Wave Equations. Geophysics. 50. 705-708.
Chu, L.L., Cakmak, A.K. and Askar, A. 1980. B o m Approximation for Wave
Scattering in Elastodynamlcs. in: Century 2 Pressure Vessels and
Piping Conf. ASME. paper no. 80-C2/FVP 72. New York: ASME Fubl.
140
Cole, D.M., Kosloff, D.D. and Minster, J.B. 1978. A Numerical Boundary
Integral Equation Method for Elaatodynaalcs. I. Bull. Seism. Soc.
Am. 68. 1331-1357.
Crawford, A.M. and Curran, J.H. 1982. Hlgher-Order Functional Variation

Displacement Discontinuity Elements. Int. J. Mum. Meth. Engng. 19.
143-148.
Crouch, S.L. 1980. Analysis of Stresses and Displacements around

Underground Excavations: An Application of the Displacement
Discontinuity Method. Department of Civil and Mineral
Engineering. University of Minnesota. Geomechanics Report.
Crouch, S.L. and Starfield, A.M. 1983. Boundary Element Methods In Solid
Mechanics. London: George Allen and Unwin.
Cruse, T.A. and Rizzo, F.J. 1968. A Direct Formulation and Numerical
Solution of the General Transient Elastodynamlc Problem. I.
J. Math. Anal. Appl. 22. 244-259.
Cruse, T.A. 1968. A Direct Formulation and Numerical Solution of the

General Transient Elastodynamlc Problem. II. J. Math. Anal. Appl.
22. 341-355.
Cundall, P.A. 1990. personal communication.
Das, S. and Aki, K. 1977. A Numerical Study of Two-Dimensional

Spontaneous Rupture Propagation. Geophys. J. R. astro. Soc. 50.
643-668.
Das, S. 1980. A Numerical Method for Determination of Source Time
141
Functions for General Three-Dimensional Rupture Propagation.
Geophys. J. R. astro. Soc. 62, 591*604.
Das, S. 1981. Three-Dimensional Spontaneous Rupture Propagation

and Implications for the Earthquake Source Mechanism. Geophys. J.
R. astro'. Soc. 67. 375-393.
Das, S. 1985. Application of Dynamic Shear Crack Models to the

Study of the Earthquake Faulting Process. Int. J. of Fracture.
27. 263-276.
Das, S. 1988. Relation between Average Slip and Average Stress Drop for
Rectangular Faults with Multiple Discontinuities. Bull. Seism.
Soc. Am. 78. 924-930.
Das, S. and Kostrov, B.V. 1987. On the Numerical Boundary Integral

Equation Method for Three-Dimensional Dynamic Shear Crack
Problems. J. of Appl. Mech. 54. 99-104.
Day, S.M. 1982. Three-Dimensional Simulation of Spontaneous

Rupture: The Effect of Nonuniform Prestress. Bull. Seism. Soc. Am.
72. 1881-1902.
De Hoop, A.T. 1958. Representation Theorems for the Displacements In an

Elastic Solid and their Application to Elastodynamlc Diffraction
Theory. Ph.D. dissertation. Tech. Univ. Delft.
Doyle, J.M. 1966. Integration of the Laplace Transformed Equations of

Classical Elastoklnetlcs. J. Hath. Anal. Appl. 13. 118-131.
Dragonl, M. 1990. A Model of Interseismlc Fault Slip in the Presence of
142
Asperities. Geophys. J. Int. 101. 147-156.
Durbin, F. 1974. Numerical Inversion of Laplace Transforms by Relating

them to the Finite Cosine Transform. J. Assoc. Comp. Mech. 15.
115-123.
Erlngen, A.C. and Suhubi, E.S. 1975. Elastodynamics II: Linear

Theory. New York: Academic Press.
Fredholm, I. 1903. Sur Une Classe D'Equations Foctlonelles. Acta. Hath.

(Sweden). 27. 365-390.
Freund, L.B. 1990. Dynamic Fracture Mechanics. Mew York: Cambridge

Press.
Friedman, H.B. and Shaw, R.P. 1962. Diffraction of Pulses by Cylindrical

Obstacles of Arbitrary Cross-Section. J. of Appl. Mech. 29. 40-46.
Haskell, N.A. 1969. Elastic Displacements in the Near-Field of a

Propagating Fault. Bull. Seism. Soc. Am. 59. 865-908.
Helmholtz, H. 1860. Theorie der Luftschwlngungen In Rohren mlt Offenen

Enden, J. Relne Angew, Hath. 57. 1-72.
Hlrose, S. and Acheribach, J.D. 1989. Tlme-Domain Boundary Element

Analysis of Elastic Wave Interaction with a Crack. Int. J. Num.
Heth. Engng. 28. 629-644.
Jaswon, M.A. 1963. Integral Equation Methods In Potential Theory. I.

Proc. of the Royal Society of London. A. 275. 23-32.
143
Johnson, E. 1990. On the Initiation of Unidirectional Slip. Geophys. J.
Int. 101. 125-132.
Karabalis, D.L. and Beskos, D.E. 1984a. Dynamic Response of 3-D Rigid
Surface Foundations by Time Domain Boundary Element Method.
Earthquake Engng. Struct. Dyn. 12. 73-93.
Karabalis, D.L. and Beskos, D.E. 1984b. Earthquake Response of 3-D

Foundations by the Boundary Element Method. In: Proc. 8th Uorld
Conf. Earthquake Engng. San Francisco. Englewood Cliffs: Prentice
Hall. 769-776.
Kellogg, O.D. 1929. Foundations of Potential Theory. Berlin: Springer.
Kobayashl, S. and Nishimura, N, 1982. Transient Stress Analysis of

Tunnels and Cavities of Arbitrary Shape due to Travelling Waves.
in: Developments in Boundary Element Methods. 2. London: Appl.
Sci. Publ. 177-210.
Klrchhoff, G. 1883. Zur Theorle der Llchstrahlen. Ann. Physik. 18. 663-
695.
Kuijpers, J.S. 1989. Generation of High Horizontal Stresses In Stope

Hanglngwalls. Chamber of Mines Research Organization. Internal
note R02/89.
Kupradze, V.D. 1963. Dynamical Problems of Elasticity, in: Progress In

Solid Mechanics, vol 3. eds: I.N. Sneddon and R. Hill. Amsterdam:
North-Holland.
Lamb, H. 1904. On the Propagation of Tremors over the Surface of an
144
Elastic Solid. Phil. Trans, of the Royal Society of London,
series A. 203. 1-42.
Lamb, H. 1932. Hydrodynamics. 6th ed. New York: Dover.
Loken, K.C. in prep. Three-Dimensional Elastodynamlc Fictitious Stress

Boundary Element Method. Ph.D. dissertation. University of
Minnesota.
Love, A.E.H. 1944. A Treatise on the Mathematical Theory of Elasticity.

4 edn. New York: Dover.
Mack, M.G. 1991. A Three-Dimensional Boundary Element Method for

Elastodynamlcs. Ph.D. dissertation. University of Minnesota.
Manolis, G.D. 1980. Dynamic Responses of Underground Structures. Ph.D.

dissertation. University of Minnesota.
Manolis, G.D. 1983. A Comparative Study on Three Boundary Element

Method Approaches to Problems In Elastodynamlcs. Int. J. Num.
Meth. Engng. 19. 73-91.
Manolis, G.D., Ahmad, S. and Banerjee, P.K. 1986. Boundary Element

Method Implementation for Three-Dimensional Transient
Elastodynamlcs. in: Developments in Boundary Elements-4. eds: P.K.
Banerjee and J.O. Watson. New York: Elsevier Applied Science. 29-
65.
Manolis, G.D. and Beskos, D.E. 1981. Dynamic Stress Concentration

Studies by Boundary Integrals and Laplace Transform. Int. J. Num.
Meth. Engng. 17. 573-599.
145
Manolis, G.D. and Baskos, D.E. 1988. Boundary Element Methods In
Elastodynamlcs. London: Unwin Hyman.
Mansur, V.J. and Brebbia, C.A. 1985. Transient Elastodynamlcs. in:

Topics in Boundary Element Research. 2. Berlin: Springer-Verlag.
Kikhlin, S.G. 1957. Integral Equations. Oxford: Pergamon Press.
Muskhelishvili, N.I. 1953. Singular Integral Equations. Groningen:

Noordhoff.
Myer, L.R., Pyrak-Nolte, L.J. and Cook, N.G.V. 1990. Effects of Single
Fractures on Seismic Wave Propagation, in: Rock Joints, eds:
Barton and Stephanson. Rotterdam: Balkema. 467-473.
Nardini, D. and Brebbia, C.A. 1982. A Mew Approach to Free Vibration

Analysis using Boundary Elements, in: Boundary Elements in
Engineering, ed: C.A. Brebbia. Berlin: Springer-Verlag. 312-326,
Napier, J.A.L. 1990. Modelling of Fracturing near Deep Level Gold Mine
Excavations using a Displacement Discontinuity Approach, in:
Mechanics of Jointed and Faulted Rock, ed: H.P. Rossmanith.
Rotterdam: Balkema.
Napier, J.A.L. and Stephansen, S.J. 1987. Analysis of Deep-level Mine

Design Problems Using the MIMSIM-D Boundary Element Program.
Mining. South African Inst, of Mining and Metall. Johannesburg. 1.
3-19.
Nicholson, J.tf. and Mettu, S.R. 1988. Computation of Dynamic Stress

Intensity Factors by the Time Domain Boundary Integral Equation
146
Method. I. Engng. Fracture Mech. 31. 759*767.
Niwa, Y., Kobayashi, S. and Azuma, N. 1975. An Analysis of Transient

Stresses produced around Cavities of Arbitrary Shape during the
Passage of Travelling Waves. Memoirs Fac. Engng. Kyoto University.
37. 28-46.
Niwa, Y., Kobayashi, S. and Fukul, T. 1976. Applications of Integral

Equation Methods to Some Geomechanlcal Problems, in: Numerical
Methods in Geomechanics, ed: C.S. Desai. New York: ASCE. 120-131.
Niwa, Y., Fukui, T., Sato, S. and Fujiki, L. 1980. An Application of the
Integral Equation Method to Two-Dimensional Elastodynamlcs.
Theoretical and Appl. Mech. 28, 281-290.
Fao, Y.H. and Mow, C.C. 1973. Diffraction of Elastic Waves and Dynamic
Stress Concentrations. New York: Crane Russak.
Fapoulis, A. 1957. A Mew Method of Inversion of the Laplace Transform.

Quarterly of Appl. Math. 14. 405-414.
Peirce, A.F. 1983. The Applicability of the Non-Linear

Boundary Element Method in the Modelling of Mining Excavations.
M.Sc. dissertation. University of Witwatersrand.
Peirce, A.F., Askar, A. and Rabitz, H. 1990. Convergence Properties of a

Class of Boundary Element Approximations to Linear Diffusion
Problems with Localized Nonlinear Reactions. Num. Meth. for
Partial Differential Equations. 6. 75-108.
Rizzo, F.J. 1967. An Integral Equation Approach to Boundary Value
147
Problem of Classical Elastostatlcs. Quarterly of Appl.
Hath. 25. 83-95.
Rosamanith, H.P. 1983. Dynamic Crack Analysis and the Interaction

between Cracks and Waves. In: Rock Fracture Mechanics, ed: H.F.
Rossmanith. CZSM courses and lectures, no. 275. International
Centre for Mechanical Sciences. New York: Springer-Verlag. 271-
300.
Ryder, J.A. 1987. Excess Shear Stress (ESS): An Engineering Criterion

for Assessing Unstable Slip and Associated Rockburst Hazards, eds:
G. Herget and S. Vongpaisal. Proc. 6th Int. Congress on Rock
Mechanics. ISRM. Rotterdam: Balkema.
Ryder, J.A. 1988. An Industry Guide to Methods of Ameliorating the

Hazards of Rockfalls and Rockbursts - 1988 Edition. Chamber of
Mines Research Organization.
Salamon, M.D.G. 1964. Elastic Analysis of Displacements and Stresses

Induced by Mining of Seam or Reef Deposits, part 4. J. of
South African Inst, of Mining and Metall. 65, 319-338.
Selberg, H.L. 1951. Transient Compression Waves from Spherical and

Cylindrical Cavities. Arkiv fdr Fysik. Band 5. nr 7. 97-108.
Somigliana, C. 1889-1890. Sulle Equazionl dell' Elasticlta. Ann. Mat.

II. 17. 37-64.
Smith, G.D. 1985. Numerical Solution of Partial Differential Equations:

Finite Difference Methods. Oxford: Clarendon Press.
148
Smolinski, P., Belytschko, T. and Liu, W.K. 1987. Stability of
Multi-Time Step Partioned Transient Analysis for First-Order
Systems. Computer Meth. In Appl. Mech. and Engng. 65. 115-125.
Sternberg, E. and Eubanks, R.A. 1955. On the Concept of Concentrated

Loads and an Extension of the Uniqueness Theorem In the Linear
Theory of Elasticity. J. of Rat. Mech. 4. 135-145.
Sternberg, E. and Eubanka, R.A. 1957. On Stress Functions for

Elastoklnetlcs and the Integration of the Repeated Wave Equation.
Quarterly of Appl. Math. 15.
Stokes, G.G. 1683. On the Dynamical Theory of Diffraction.

In: Mathematical and Physical Papers. II. ed. G.B. Stokes.
Cambridge: University Press. 243-290.
Symm, G.T. 1963. Integral Equations In Potential Theory. II. Proc.

of the Royal Society of London. A. 275. 33-46.
Tlan, Y. 1990. Boundary Element Methods In Elastodynamlcs. Ph.D.

dissertation. University of Minnesota.
Triantafyllidis, T. and Dasgupta, B. 1987. The Causality of the Boundary

Element Method In Elastodynamlcs. Supercomputer Institute.
University of Minnesota. UMSI 87/112.
Waterman, P.C. 1976. Matrix Theory of Elastic Wave Scattering. J.

Acoust. Soc. Am. 60. 567-580.
Wheeler, L.T. and Sternberg, E. 1968. Some Theorems In Classical

Elastodynamlcs. Arch, for Rat. Mech. and Anal. 31, 51-90.
149
onendlx 1; Temporally and spatially integrated constant sub-functions
„ 1 . y \ (cjt'-r*)*1 "?tV
7“ - 2 ^ ' — ^ - |(T * T > ? ------ — -------- ? -------
3 1 2 r
J»
H(t" ^ > r/ca

V ^ y2 v (c*t»-r*)«‘» e J t V (cft2-r2)0,5 ^
3 6 3 r2 3 r*
2
(Al.l)
H(t--L)
V12 - V21 - _ L [ ---- fi_[fiLi2(cit2-r2)0'3 - l2^(cft2-r2)0-3]
12 21 2k p c3 3 r* 3 r2
H(t-JL) 2 2
c2 eft2
(cft2-r2)0,5 - l H ( c 2t2-r2)0-3]] (A1.2)
r* 3 r2
H(t-— ) 2 2 2
C, eft2 V a 9 . - - . w2 eft2
V* - _ L [ ------------------------ - ( J L . -
2irp c3 3 r« 3 r2 6r 2
l/_2».2 _2v0.3 . , C lt x_._i.-1/ C lt >
H(t-JL)
~i t f «i !- -!- - l- - !- - -(- -c- - -f- -t- - 2-
* --- r
1 2)
/ 0-3 - \<J r
l_ T♦ ftl)(cit*-r2)0-
_ J V ^ 2 5♦
cf 3 r* 3r2 6r2
■j(cftz-r2)0,3 - (^ ) c o s h - 2( ^ ) J ] (A1.3)
150
In what follows, VIJ - Jv^dx, VIJK - d/dxk /VtJ dx, VIJKL - dVdxudxj/V^dx:
H(t-JL) 2
r 1 c. c.tx , c.t x c.t .......
K i* - -ji-t - ^ t < - ^ - ) c o s h - l ( _ i . ) - i ( 2 + —l y ) (c 2t 2- r 2) 0,3] -
J 2itp 2 r 6 rz
H(t-~) 2
C» C,tx . C.t x Cat 2 9 in*
>coali-l ( _ i . ) ♦ | ( 1 — L - X c ^ - r 2)0 5 -
cI 2 r 6 r2
i<c 2 t2+y2)sin l< _ _ 5 )♦

2 (cft2-y2)0'5
(A1.4)
(0 ,1 7 )tan'1(--- y X, -■■-)] ]
y(c2 t2-r2)0-3
H(t-JL) 2
Jv12dx - J L [ i i - K ^ Z j e o a h - ^ ^ ) - |(2+fil)(cft2-r2)0-5]
-i
H(t-JL)
2_ .rfZ^êosh”1^ __ - y(2+Z 2!!)<c22t2-r2)0-*]]
[ ( ^ ) c o s h - 1(Z2!)
r 6 r2
A(1.6)
H(t-— )
t
fv2zdx - -i-{ Zi-t-CCityJtan'^ ? + {ZÔcosh*1^ "> +
J 2np y^Cit2 _r2 j0 .s 2 r
■i(yz+cJt2)sin"1(— — J L _ _ ) - *(i-Zl!!)(c*t2-r»)0•’ ]
2 (cjt2-y2) • 6 r2
H(t-— ) 2 2
Ba C*tX * C*t v Cat a . _ . _
— [-(-i-Jcosh-H-2-) + 4(2+_i_)(c2t2-r2)0-5]
c3 ■ ' 2 r 6 r2
(A1.7)
151
CZ C 2 t X X V c 2 t l 1 S
* f ‘^ > “
•<,),In',(7 3 ^ >n d.10)

H(t-~) a .
c.t c.t r c,t
_ifv12dx - J _ [ — ii_[<fi!)cosh-i(^:) -
dyJ 12 2*p cl 2 r 2 r2
(ra-2ya) , cita_i\i.3i
6r ra
C * C n ( a C £ 1*
.*-2
C j t
a > ■
_£_[(_£_)cosh-1(-2-) - £<-i— -1)0,5
a " 2 I 2 r2
cz
- <ra-2ya) ( clt2.t î.31] (Al.12)

6r r*
H(t-— ) aa 2 2 2 2
d e 1 c» x 2 c.t2 1a 1
_ fv„dx - _ L [ ----- i-t-iLc-L— -!)1-3 +_(-4r+—i— ){_!__-I)0,5
cit cit .0 a
dxJ 22 2*/ c3 3r r2 6 r r2
c.t . c.t
( 1 Jcosh’^-l-)]
2 r
H(t-JL)
x2 eft2 . . 1 C?t2 C?t2
2— -1) * i(2r+— 5— )(_£_-l) 0.3
3r tz 6 r r2
H(t--L) 2 2
_ l r v 22dx - _ L [ ____!i_[-(c1t)tan’1( -- JH
C - - ) - 22<.!i!L-l)l s +
dyJ 22 2np cl lVl/ y(c2t2-r2)0, 3r r2
(y)sin"l(— -— ----- )]
(cft2-y2)°‘s
H(t-J-> 2
- C2 [_xy^c2
3 1 3r r* (A1.14)
153
H(t-JL) 2
1.3
J L f v u dx - _ L [ t (r*-2x2)(
dx * cf 3r8 r2
eft2* C?t2
1)0-5]
3rs
H(t S ) 2-2
2_[ j L ( r 2-2x2)(.°2t --1'15
I )1
3r3
(A1.15)
r3 3r3 3r r2
mt- — / 22 2 2
_ d!_rvu dx - _i_[_____________________ + y(F2-f!x2> < ^ l - i ) 23
dxdy
dyJ 2*p c3 r3 r2 3r3 r2
r3 r2 3r
(A1.16)
n^t“ ) a« 22
_£ifvu dx - J _ [ _________________(fil-i)2-3 -
dy*J 11 2xp c* 3r3 r3 r2 r2
eft2
(c22t2-y2)0S r3 r
x(rz-4y2) czt2 !.5

(Al.17)
3r3 r2
154
SSI
(OZ’IV)
e*£ z3 „ e3
g3 . e3S z3 „ e3 e . ^ *Z ZI r z P
"s3*3 (^ " * J£)iC +S‘° Z3*
z*zX° X° 1 I _XP A-fzP
'z z (T ' ^ H
(61 IV)
z3 ..e3 e3€
t t s o<I" ^z0 \£x. “ s t ( 1 “ a?o
„g„> (z^V_z3)x zo
z*z (T -5)H
g3 63 z3 c3 E , e° ,* » 2 „ r Xpxp
! o(X",5lo)iXx
zz " S l(T“^ 0 > (g^"g3>x ] To 1 I _XP A-f ?P
** < T _:1)H
(81'IV)
z3 „ e3
t l s o( l - , ) ^ x - e-x<1 - g0 ) ( x i , - 3 ) ^ J z0
z z 1 z < T " 3>H
. z3 e3 z3 e3€ ca <**Z fg*P

n .5'o V ’n ,lja To I ’ J zP
g z < T - :° H
H(t--I)
eft2
dx2Jp
—
3ra
x 3 eft2* Ov eft2 „«
r3 3r3 3r r*
3r
v3 cft^ v eft2 „-
(A1.21)
r 3r3 3r r2
H(t--l)
mt" — / 2 9 2 9
dxdyJ 2n p Ci3 r r 3 r r r
H(t--L)
y . 1 4x2x, c2 t2
r-(— ----1) - 1)13]]
r2 3 r r3 r2
(Al.22)
H(t-— )
eft2
-£lfv„dx - J_[
dy2J 2 itp (eft2—y2)0-3
3 r r3 r2
H(t-— ) 2«-2
\ cli!-n»>]j
3 r r3 r2
(A1.23)
156
Appendix 2: Temporally and spatially Integrated linear sub-functions
In what follows, WIJ - J xV1j dx, VIJK - d/dxk / xV4j dx, and
WIJKL - d2/dxkdxL J xVAJ dx.
H(t--I)
-£-fxVu dx - - L [ p L U — ^coih-1^ ) _ *<cat2-r2)0-3 *
dxJ 2*p 2 r 2
J L < r 1-2y1)<c*t*-r#)2-3]
6r'
H(t-— )
^ [ - ( f i l ^ J c o s h - ^ ^ ) + i(cft2-r2)0*3 +
J L ( r 2-2y2)(c|t2-r2)ls]] (A2.4)
fir*
H(t-JL)
-dyJ
i f x V n d x - _2xp
L [ - - -3^ - ( f ^ c2 o s h - 1^ r) * I2 ^ V - r 2)0 5 +
-L. (3r2-2y2)(cjt2-r2)13]
6r*
^L.[-3(f£)corii-1<^!) * ^ ( c 2t2-r2)0-3 +
es 2 r 2
-2-(3r2-2y2)(c|t2-r2)1-3]] (A2.5)
6r*
2 H(t--l)
-l_fxVn dx - _ L [ --- [J^(4y2-3r2)(c2t2-r2)1,3 - 2LZ(c2t2-r2)0-3]
dxdyJ 2np CJ 3r6 r*
H(t-Jl)
- --- ^2_[ JS^(4y*-3r2)(c*t2-rV-9 ♦ *£(cft2-r2)03]]
Cjj 3r r |
(A2.6)
159
H(t-JL)
_L.fxVu dx - JL[ ♦ -l-(t*-2xV)(cft*-r»)»-» ♦
dy • 2"P c{ 2 r 2r*
— 1- (3r*+4y2(2ya-3r2))(c?t2-r2)1,3]
6r®
H<t_— >
- _________________________ ♦ _i_(3r*+2y*)(cft2-r2)°‘s -
4 2 r 2r*
JL (3r*+4y2(2y2-3r2))(c2t2-r2)13]] (A2.7)
H(t-JL)
_lfxV12dx- — [---- lL.[i^(cJt2-r2)1-5]
dxJ 12 2np „3 3r*
H(t-— )
[if£(c2t2-rV-3)] (A2.8)
cj 3r*
H(t-— )
citx
— |xV12dx - _ L [ ---- [(City)tan_1( ) - * (cft2-r2)0-3 -
dyj 12 2xpl c3 'y(c2 t2.r2)0.5 2 1
_iL(cJt2-r2)13 - 0.5(y2+cJt2)sin"1(— -— ^---- )J

3r 1 (cft2-y2)0,3
°2 [(c2ty)tan~1( .-) - i(c2

2t2-r2)°-3
y(cft2-r2)C-S 2
-“ (c2t2-r2)1,3 - 0.5(y2+c|t2)sln-1( X ■ )]]

3r* (c22t2-y2)0-3
(A2.9)
160
H(t-— )
JxV12dx-JL[ p-[^2(r2-2xa)<cfta-r*)ls- 2LZ(cJta-ra)°‘3]
dx
: ca^[i^<r2
3re
-2x2)<c“2
V-rV s- i2(
r4
c2ta-r2)°-3]]
(A2.10)
H(t--l)
lxV12dx - _ L [ --- (r2-4y2)(c2ta-r2)1,3 - i ^ ( c 2t2-ra)“-3]
dxdyJ 2 up c® 3r8 r*
H(t— I)
Ca_[ Jil(r2-4y2)(c2t2-r2)1-5 - 2fy?<c2t2-r2)«-3]]
c3 3r®
(A2.ll)
2 H(t--L)
JL_ fxV12dx - _!_[ fi-KcitOtan"^ ^-1! ) - ysin_1(---------- ) +
d y 2J 14 2*/ c3 (cjt2-y2)0,5
^ ( c f t ^ - r 2)1-3 - 22ll(cftz-r2)0,5]
3r8 r*
H(t-— )
- ----25_[(c2t)tan"l(----- ) - ysln"l(---- 2 )+
cl y(c2t2-r2)0,3 (c|t2-y2)0,3
J - i2!(c|t!-r»)* s]]
3t' r‘ (*2 .12)
161
y tit
> "IT
a-
TOI
rt ft
«r
Mo to
rt
i
rt
a i o |h
to to
|x H X - ° h
H ov >IH°
162
to
I
to O I ft
TO o \
TO o
*< o In TO o
Id as
sr to sr ro H i to o
O W
O o
NM u X
ft
is X 5* m
i
I
n
H i to
I
*
I it
to >u
+
ro | X
ro|x ■i<
0
0to
•» to ft 0»
ft to IO X
to 1 V«J ♦
to
H
1 H
+
to K X
w to *r
Hm X to
I
it
to
I
H
I 0
H to io
to rt
to
1 0
> ► Hto to to 1
ro IO ft
1
to
*
H(t--L)
JlfxVjjdx - _ L [ ---J2_[3(f££)coBh■*(.!£> - ^ ( c f t 2-r2)l s -
dyJ “ 2wp CJ 2 r 2
-JL(3r2-2y2)(cft2-r2)15]
6r*
H(t-JL)
- --- ^ [ ( ^ J c o s h - V ^ ! ) - !<c*t2-r2)0-5 -
-3L(3rz-2y2)(c2t2-r2)15]] (A2.15)
H(t-— )
^LfxV^dx - J _ [ --- -2i_[-_22L(4y2-3r2)(cjt2-r2)1,5 - i ^ ( Cl2t2-r2)0 S]
dxdyJ 2xp c3 3r« r* 1
-^2— [-J2 L(4y 2-3r2)(Cjt2-r2)1,5 + Ji-Z(c?t2-r2)0S]]

2 3r® r*
(A2.16)
2 H(t~— )
î-fx^dx - _ L [ p - W - f i ^ c o s t r 1^ ) - — i-(3r*+2y*)(c2t2-r2)0,s -
dy J 2up cI 2 r 2r4
_ L < 3r* + 4y 2 (2y 2-3r2))(c|t2-r2)15]

6r6
H(t--I)
- --------[(.!£) cosh"1(f2!) - — i- (r*-2x 2y2)(c?t2-r2)0,s -
ca 2 r 2r4
J _ ( 3 r* + 4 y 2(2y 2-3r2))(o|t2-r2)1-3]] <A2 .17 )

6r°
163
2 H(t-JL)
-i-fxVjjdx - _ L [ --- ^ [ ( ^ c o s h - 1^ ) - - L ^ x V - r ^ c f t 2-!2)15 -
dx2J 2 np c3 2 r 6r*
_ L ( 2 y 2(3x2-r2) + r 2(2y2+3r2)) <cjt2-r2)0 5 ]
6r*
H(t-— )
- i--- ^ [ - ( ^ J c o s h - ^ ^ 2!) - _JL(8xaya_r*)(c2t2-r*)1,9 -
c* 2 r 6r8
_ L ( 2 y 2(3x2-r2) - r 2(8x2+r2))(c*t2-r2)0-5] ]
6r*
(A2.18)
164
Appendix 3: Temporally and spatially Integrated constant and linear velocity
I
Isub-functions
In what follows, VIJT - d/dt / VtJ dx, VIJKT - d2/dtdxk J V4j dx, and
WIJKT - d2/dtdxk / xVAJ dx.
H(t--l)
[c2tan'1(--- °ftX ) ♦ iSZc|t(c|t2-r2)0-3]]
y(c?t2-r2)0,3 r* ,j
(A3.1)
H(t-JL) H(t-— )
V12dx li— [i^cfrcjt2-,:2)0-3] - ----^ [ 2 2 c 2t(c2t2-r2)°-3]]
dtdxJ’ 2xp r* c. r*
(A3.2)
H(t--l)
i!_fv22dx - _ L (--- il_[(J±^!)cJt(o*t2-r2)0-3 ♦ (^)cosh-1(î!)]
itdxJ “ 2icp -3 2r* 2 r
H«:-— )
i 5_ I ( £ ^ L ) c 22t(c22t2-r2)0-3 - (^Jc o s h - ^ f 2!)]]
cl 2r* 2 r
(A3.3)
165
H(t-JL)
— !lfv22dx - - L [ ------- [-c1tan*1( > - S c “t<c*t*-r*>«]>
dtdyJ “ 2wfi c3 1 y(e‘tx-rs)°-9 r4
H(t--L)
M - ^ c 2t(c22t2-r2)0-5]] (A3.4)
H(t-— )
-*l|v12dx - _ L [ --- lL.[(^)cosh-1( ^ ) - (l!^z!)cft(cjt2-r2)0-3]
dtdyJ 12 2np c® 2 r 2r4
H(t-JL)
!2-[(5)cosh-l(
ff— ,___,
^ ) - ,____
( r2"2y2)c2t(c2t2-r2)0 5U
2 r 2r4
(A3.5)
H(t-— >
H(t-— )
^ 2-[<ii-^!)c|t(c2t2-r2)o s - (^Jcosh'^f2!)]]
c® 2r4 2 r
(A3.6)
166
H(t-— ) 2
J.fVlldx - - L [ Jl.[(V)co*-»(i!) - (^2)(c*t*-r*)«-*l

d tJ 11 2*p c* 2 r 2r*
H(t-JL) ,
CaX < Cat CatX 2 9 a f% a
— I - ( - i- ) c o s h _1( — ) - ( _ i - ) ( c * t * - r 2) 0-3 -
c2
3 2r2
(c2t)sln_1(— -— - ) + (cjy)tan_1<--- ------ )]]

(c2t2-y2)0,3 y(c2
2t-r2)0'3
(A3.7)
H(t-JL) 2
i2_[(^)cosh-l(— ) - (î2)(o|t2-r2)0-9]]
c? 2 r 2r2
(A3.8)
H(t~— ) 2
dr 1 c. c.x , c.t c.tx , , ...
Tth* • 2^'— * ("^r><c‘ 1 ‘
(c2t)sin*1(— -— ) - ( c ^ t a n * ^ --- -------- )]

(cjt2-y2)0’3 y(cjt2-r )
H(t-— )
c, c,x , c.t 2
c.tx 2 . ... .
2 [-(_2_)cosh-1(JL) ♦ (-±_)<c2t2-r2)0'3]]
c3
z 2 r 2r2
(A3.9)
167
H(t--i)
£_fxVu dx - - L [ Jl.K.^Jcoah-1^ ) + x<ra-2ya)Cl2t(c1
V - r 2)0-3]
dtdxJ 11 2 xp c3 2 r 2r*
H(t-— )
^[-(f^cosh'H— ) + x(r2~2y2) cft(c2t2-r2)05]]
s 2 r 2r«
2r4 2
(A3.10)
H(t--l)
L r x V u dx - _ L [ --- * [-(^)cortr2(*i) + y(3r2-2y2)e2t(e2t2.r2)0.3]
dyJ 11
dtdy 2 xp CJ 2 r 2r*
W -— > , ,
- ^ ( ^ ( f ^ O c o s h ' 1^ ) + Z 2 l ^ L l c | t ( c | t 2- r 2) 0-5]]
c? 2 r 2r*
(A3.11)
H(t-JL) _ H(t-JL)
*L[xV1
fxV,.dx _ L [r ----2l_[i^cft(c?t2-r2)o s] -
2dx - — f2_[i^c?t(c2t2-r2)0-5]]
tdxJ 12
dtdx 2xP c3 r* 1 1 3 r* * *
~1 ~2
(A3.12)
168
H(t--l)
. dZ fxV„dx - J _ [ ^L_[(c-yjtan-^ ) - (c?t)«in“1<---- ------ )
-
dtdyJ 2itp c* y(cJta-ra)0-5 (cft2-y2)0S
r*
H<t-— > ^
°2 •[(c2y)tan'1(--- *2 ) - (c’tOsln^. 51
c| w ^ « a f- 2 _ * » 2 ^ 0 .5
y(c2 2t2-r2)0S (c*t*-y*)°-s
.3
2Lc?t(c*t2-r2)0-3]]
r*
(A3.13)
H(t-— )
— ~— [xVjjdx - _ L [ ----[(fill)cosh'1(.lill) - H r-2~2y2),c?t(c?t2-r2)0-3)
dtdxJ 22 2np c3 2 r 2r*
H(t--L)
!i.[- ( ^ ) c o s h - ( ^ ) - i < £!^! >c2
2t(c2
2t2-r2)-]]
c.3 2 r 2r*
(A3.14)
H(t-— )
^-JxV^dx - =1
H(t--l)
5!2_E(^)c osh* 1(^2!) - Z-(l r2~?y2)c|t(c2t2-r2)0,3]]
c2
3 2 r 2r*
(A3.15)
169

Two-Dimensional Time Domain El

Uploaded by

Copyright:

Available Formats

Two-Dimensional Time Domain El

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Two-Dimensional Time Domain El

Uploaded by

Copyright:

Available Formats

INFORMATION TO USERS

The quality of this reproduction is dependent upon the qnality of the

Oversize m aterials (e.g., maps, drawings, charts) are reproduced by

Photographs included in the original manuscript have been reproduced

University Microfilms International

Two-dimensional tim e dom ain elastodynam ic displacem ent

Siebrits, Eduard, Ph.D.

C opyright ©1992 by S iebrits, E duard. All rig h ts reserved.

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

This is to certify that I have examined this bound copy of a doctoral

Ff/k. l2>. 1^92/

"Neither seeking nor avoiding mathematical exercltations

Lord Kelvin and Peter Guthrie Talt1

Karman, T. and Blot, H.A. 1940. Mathematical Methods In Engineering.

I would also like to thank:

Dr. John Napier, for your enthusiastic support, encouragement, and

Dr. Nick Gay, my mentor, for believing inmy abilities, andsupporting

Dr. John Ryder, for Introducing me to boundary element methods in 1985,

Dr. Anthony Peirce, for tirelessly helping me with some of the

John Crouch, for providing all the excellent post-processing graphics

Space Science Graphics at the Space Science Center, University of

Two versions of a cine domain elastodynamic displacement

Acknowledgements ............................................ (1)

2 LITERATURE REVIEW ........................................ 5

3 THEORY: TWO-DIMENSIONAL FORMULATIONS ....................... 9

6 SINGLE ELEMENT RESULTS .................................... 53

8 SPECIAL ELEMENTS ......................................... 110

9 STABILITY ................................................ 125

10 CONCLUSIONS ............................................. 135

11 RECOMMENDATIONS ........................................... 137

REFERENCES .................................................. 139

Rockbursts and rockfalls are two major problems affecting the

This review will concentrate on elastodynamic boundary element

Two elastodynamic displacement discontinuity methods are presented

3.2 Mathematical Derivation

The basis of the two-dimensional formulation is Stokes's (1883)

Figure 3.1: Three- and two-dimensional displacements due to Impulsive

The two-dimensional displacement fundamental solution is given by

H(t-JL) o2t f(t--i)

2xp dtp cx ^ (ij2-r2)°*s fl Ci

a and j8 each range from 1 to 2,

This solution is Stokes's displacement aolution integrated along the x3

9v._ 9v„ 3va,

These fundamental displacements and stresses are the two-dimensional

3.2.2 Two-Dimensional Representation Theorem

The two-dimensional representation theorem, applicable when the

M i . t ) " | l vut<5.t;£|t(s)1(x,t)) - ^nHkte.tlllu^x.tJJJdStx)

vlk°(x,t;£|t(B)i(ji,t)) - Vik(&.t;£) * t<B)1(s,t)

and t(B)1 denotea the traction component tt on surface S with outward

3.2.3 Displacement Discontinuity Method

The displacement discontinuity formulation can be obtained from

Figure 3.2: Unit normal directions on boundary S.

The contribution to the displacements at £ € V due to the stresses and

Adding (3.8) to (3.9), and noting that q (1> - gives

«^l)(£.t> - Jlvj^x.tjilT^x.t)) - ^ n(.))ik(x.t;£|D1(x,t))]dS(x) (3

Equation (3.12) gives the k-th displacement component at point £ due to

where Da(£,s) is the displacement discontinuity source function, and

3.3 Time and Space Integrations

A piecewise linear temporal variation is assumed for the

fcosh1<£^)dx - <x)cosh1<— ) + (ct)sin"l( 2L_ )-

f J L ^ t 2-?2)0,3! - <c2t2-r2)0,3 - <ct)cosh"1<4^) <3.22)

h" - C1])" + S-!,-! Sr*1? (3.40)

- E [[(A+2/i)p(V1112 + V1 2 1 1 ) + A/i(V1 2 2 2 + V2212) ]"-klD +