Preprint

UCRL-JC-146404

The Distinct Element

Method - Application to
Structures in Jointed Rock
J. Morris, L. Glenn, S. Blair, F. Heuze

Thisarticle wassubmitted
to
InternationalWorkshop
Meshfree
Methods
of Partial Differential
Equations,Bonn,Germany,
September
11-14, 2001

November30, 2001
U.S.DerJartment
of Energy
Lawrence
Liverrnore
National
Laboratory

Approvedfor public release; further dissemination unlimited

DISCLAIMER
This document was prepared as an account of work sponsored by an agency of the United States
Government. Neither the United States Government nor the University of California nor any of their
employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for
the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or
represents that its use would not infringe privately owned rights. Reference herein to any specific
commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not
necessarily constitute
or imply its endorsement, recommendation, or favoring by the United States
Government or the University of California. The views and opinions of authors expressed herein do not
necessarily state or reflect those of the United States Governmentor the University of California, and
shall not be used for advertising or product endorsement purposes.
This is a preprint of a paper intended for publication in a journal or proceedings. Since changes may be
made before publication, this preprint is made available with the understanding that it will not be cited
or reproduced without the permission of the author.
This report has been reproduced directly from the best available copy.
Available electronically

at http: //www.dqe.gov/bridge

Available for a processing fee to U.S. Department of Energy

and its contractors in paper from
U.S. Department of Energy
Office of Scientific and Technical Information
P.O. Box 62
Oak Ridge, TN 37831-0062
Telephone: (865) 576-8401
Facsimile: (865) 576-5728
E-mail: [email protected]
Available for the sale to the public from
U.S. Department of Commerce
National Technical Information Service
5285 Port Royal Road
Springfield, VA22161
Telephone: (800) 553-6847
Facsimile: (703) 605-6900
E-mail: [email protected]:gov
Online ordering: http: / / www.ntis.gov / ordering.htm
OR
Lawrence Livermore National Laboratory
Technical Information Departments Digital Library
http: //www.llnl.gov/lid/Library.html

The Distinct Element Method - Application

Structures in Jointed Rock
Joseph Morris,

Lew Glenn, Steve Blair,

Geophysics and Global Security Division,

Laboratory, Livermore, U.S.A.

and Francois

to

Heuz@

Lawrence Livermore National

Abstract. The Distinct Element Method (DEM) is a meshfree method with applications to rock mechanics, mining sciences, simulations of nuclear repositories,
and the stability of underground structures.
Continuum mesh-based methods have been applied successfully to many problems in geophysics. Even if the geology includes fractures and faults, when sufficiently large length scales are considered a continuum approximation may be
sufficient. However, a large class of problems exist where individual rock joints
must be taken into account. This includes problems where the structures of interest have sizes comparable with the block size. In addition, it is possible that while
the structure may experience loads which do no measurable damage to individual
blocks, some joints may fail. This may launch smaller blocks as dangerous projectiles or even cause total failure of a tunnel. Traditional grid-based continuum
approaches are wholly unsuited to this class of problem. It is possible to introduce
discontinuities or slide lines into existing grid-based methods, however, such limited
approaches can break down when new contacts form between blocks.
The distinct element method (DEM)is an alternative,
meshfree approach. The
DEMcan directly approximate the block structure of the jointed rock using arbitrary polyhedra. Using this approach, preexisting joints are readily incorporated
into the DEMmodel. In addition, the method detects all new contacts between
blocks resulting from relative block motion.
We will describe the background of the DEMand review previous application
of the DEMto geophysical problems. Finally we present preliminary results from
a investigation into the stability of underground structures subjected to dynamic
loading.

Numerical Simulation

of Discontinua

A wide range of applications

involve materials or systems which are discontinuous at some level. While some systems may be intrinsically
discontinuous
(such as items of furniture
in a room) other discontinuous systems are well
approximated by a continuum. Soils, for example, are clearly discontinuous
at the microscale, however, in practice, the behavior of soils are well represented by continuum methods. This approximation
is possible because the
scale of the objects of interest (dams, walls, etc.) is large compared with
typical soil grain. There are also applications
where a discontinuous system
may be approximated better by a continuum or discontiuum depending upon

Joseph Morris et al.

the specific information sought. For example, continuum mesh-based methods have been applied successfully to many problems in geophysics. Even
if the geology includes fractures and faults, when sufficiently large length
scales are considered a continuum approximation maybe sufficient. However,
a large class of problems exist where individual rock joints must be taken
into account. This includes problems where the structures of interest have
sizes comparablewith the block size. In addition, it is possible that while the
structure may experience loads which do no measurable damageto individual
blocks, some joints may fail. A continuum, mesh-bashed treatment of such
systems is usually inappropriate.

Discrete

Element Methods

Cundall and Hart, 1992 review manyof numerical techniques that have been
developed to simulate the behavior of discontinuous systems. Cundall and Hart, 1992
define a discrete element method to have the following properties:
- Allows finite displacements and rotations of discrete bodies, including
complete detachment
- Recognizes new contacts automatically as the calculation progresses.
Cundall and Hart, 1992 further define four basic classes of discrete element
methods:
-

Distinct element methods (DEM)

Modal methods
Discontinuous deformation analysis
Momentum-exchange methods

Distinct element methods (DEM)use an explicit scheme to evolve the

equations of motion of discrete bodies directly. The bodies may be rigid
or deformable (by subdivision into elements). Early approaches employed
rigid disks or spheres with compliant contacts (Cundall and Strack, 1979,
Cleary, 1991). Cundall, 1980, Cundall and Hart, 1985 and Walton, 1980 developed 2-dimensional DEMswhich employed arbitrary polygons. More recently, fully three-dimensional DEMswith rigid or deformable polyhedral
blocks with compliant contacts were developed (Cundall, 1988, Hart el al., 1988),
When using the DEM,contacts are always deformable and can employ detailed joint constitutive models which mimic observed fracture properties
(normal stiffness, shear stiffness, dilation, cohesion, etc.). Quasi-static analysis can be performed by relaxation of the body motion. The DEMis considered in more detail in the following section.
Modal methods (Williams et al., 1985,Williams and Mustoe, 1987) are similar to the distinct element methodin the case of rigid blocks. However,for
deformable bodies, modal superposition is used. This method appears to be
better suited for loosely-packed discontinua. In dynamic simulation of dense

Distinct Element Method

packings, eigenmodes are not revised to account for additional contact constraints.
In contrast with the Distinct element method, Discontinuous deformation
analysis (Shi, 1988 and Shi and Goodman,1988) assumes contacts are rigid.
The bodies may be rigid or deformable and the condition of no-penetration
is achieved by an iterative scheme. Deformability is achieved through superposition of strain modes.
Momentum-Exchange methods (Hahn, 1988) assume both the contacts
and bodies to be rigid. Collisions are modeled by an instantaneous exchange
of momentumbetween contacting bodies. It is possible to include friction
sliding in the contact model.

Distinct

Element Method

The Distinct element method (DEM)can directly approximate the block

structure of the jointed rock using arbitrary polyhedra. Using this approach,
preexisting joints are readily incorporated into the DEMmodel. By nature,
the Distinct element method can readily handle large deformation on the
joints. In addition, the method detects all new contacts between blocks resulting from relative block motion. The Lagrangian nature of the DEMsimplifies tracking of material properties as blocks of material move. It is also
possible to guarantee exact conservation of linear and angular momentum.
~-hrthermore, by using an explicit integration scheme, the joint models can
be very flexible. In particular, the joint constitutive model can incorporate
experimentally observed effects such as, cohesion, joint dilation, and friction
angle.
The DEMhas been applied to a wide range of problems in geomechanics. For example, Antonellini and Pollard, 1995 simulated the formation of
shear bands in sandstone using the DEM.Morgan, 1999a,b applied the DEM
to the mechanics of granular shear zones. Heuz~ et al., 1993 used the DEM
to analyze explosions in hard rock. Sanderson and Zhang, 1998 used a twodimensional distinct element method to investigate the evolution of damage
zones and fluid flow in fracture networks and around faults. Cundall, 2001
reviews the application of the DEMto simulation of granular material and
rock.
The DEMhas also been coupled to other techniques in order to exploit
the relative advantages (accuracy vs. speed) of each in different regions
the computational domain. For example, Lorig et al., 1986 used a hybrid
model that incorporated both distinct element and boundary element methods (BEM)to analyze the behavior of jointed rock. The DEMwas used to simulate the near-field behavior in detail, while the BEMprovided boundary conditions that simulated the far-field rock mass. Pan and Reed, 1991 presented
details of a coupled distinct element-finite element methodthat was successfully applied to several field-scale test problems. Mori, Otsu, and Osaknda, 1997

Joseph Morris et al.

used a viscoplastic finite-element method to generate boundary conditions

for a DEMsimulation of grain alignment in the "mushy-state" forging of
magnets. In addition, the DEMhas been applied to simulate the transient
dynamics of fracturing solids. Munjiza, Owen, and Bicanic, 1995 developed a
coupled finite discrete-element technique where deformable discrete elements
were allowed to separate into two or more elements once a critical stress (or
strain) state was reached. Sawamotoet al., 1998 used a simple DEMwith
cylindrical elements to assess local damageto reinforced concrete.
The flexible nature of the DEMpresents a few challenges. The algorithm
for detecting an classifying contacts must be both robust and efficient. Realistic three-dimensional fracture networks require a great many DEMblocks
and concomitant computational expense. These two issues are discussed in
the following sections.

3.1

Detecting

Contacts

As the distinct element blocks move, initial contacts between blocks may disappear, and new contacts form. The DEMcode must rapidly determine if two
blocks are interacting. In addition, the code must be able to classify the type
of contact so an appropriate constitutive model may be applied. For example,
the behavior of a vertex pushing into a face maydiffer qualitatively from two
faces in contact. The methodused should be able to identify all conceivable
cases of two block interaction. In particular, near degenerate cases, such as
two faces barely overlapping, should not cause the algorithm to fail.
One approach is to exhaustively consider the arrangement of two blocks
by looking at each pair of vertices. If the blocks have m and n vertices
each, this direct approach requires order mn operations to determine if two
blocks interact. In addition, the algorithm in three dimensions is very complex
and requires manydifferent cases to be identified and treated appropriately.
Ghaboussi et al., 1993 proposed training neural networks to perform contact
detection. While neural networks are relatively simple to implement they
require extensive training.
The concept of a "Common-Plane"(Cundall, 1988) provides an alternative approach to reduce the complexity of the contact detection algorithm.
Intuitively, the approach can be thought of as introducing a rigid, massless plate between candidate interacting pairs of convex polyhedra (see Figure 1). The algorithm proceeds by iterating the position and orientation of
this "common-plane" to maximize the distance between it and each block.
If the blocks are overlapping, the common-planewill orient to minimize the
penetration of the blocks through the common-plane. Figure 2 shows some
examples of how the common-planeis oriented for two-dimensional polygonal
elements. Of particular interest is the treatment of non-convexblocks. In this
case the block must be subdivided into convex polyhedra and common-planes
are generated for each pair of convex polyhedra belonging to separate blocks.

Distinct Element Method

The common-planecan be used to simplify contact classification. Cundall, 1988

suggest classifying the contact type by simply counting up the numberof interacting vertices on each block. For example, two penetrating vertices from
one block and three from the other would indicate an edge interacting with
a face. In addition, the common-planesimplifies the application of joint constitutive models. For example, the orientation of the common-planecan be
used to determine the normal of the contact and dictates a co-ordinate system
which can be used by joint models.

Common-plane
Plane movesto maximize

Blocksbroughttogether
Fig. 1. The "common-plane"
can be thought of as a rigid, massless plate between
candidate interacting pairs of convexpolyhedra. The algorithm proceeds by iterating the position and orientation of this "common-plane"
to maximizethe distance
betweenit and each block.
The iterative procedure of the common-plane approach is very easy to
implement and can be very efficient for manyclasses of problem. Typically,
the common-planeorientation from the previous time-step provides a good
initial guess of the current orientation. Provided the appropriate normal of
the contact has not changed muchbetween time steps, the iterative procedure
converges rapidly.
3.2

Parallelization

of DEM

The number of distinct elements used in a single simulation is limited by

the available computational power. Current work stations can handle simulations involving of order one hundred thousand elements. Several authors have
extended the capability of the DEMby utilizing high performance computing (HPC) systems. Cleary and Sawley, 1999 and Sawley and Cleary, 1999
have developed a portable parallel implementation of the DEMusing standard message passing libraries. Horner et al., 1998 and Carrillo et al., 1999
recently implementeda fully interactive large-scale soil simulation involving
hundreds of thousands of elements. Dowdinget al., 1999 developed a parallel
DEMto calculate the dynamic response of a cavern in a jointed rock mass.

Joseph Morris et al.

c-p
(b)
:
c-p

c-p

c-p (a)

Fig. 2. Someexamples of howthe common-planeis oriented for two-dimensional

polygonal elements. The last exampleshowsthe treatment of non-convexblocks.
The block is subdivided into convex polyhedra and a common-planeis generated
for each pair of convexpolyhedrabelonging to separate blocks.

They compared SIMD(single instruction, multiple data) performance with

MIMD(multiple instruction,
multiple data) and found MIMDprocessing
to provide the best overall parallelization. Ghaboussi et al., 1993 proposed
an alternative parallelization scheme on SIMDarchitectures by using parallelized neural networks to perform the contact detection. However, to date
this approach does not appear to have been implemented.
Our approach was to parallelize through spatial domain decomposition.
The entire problem domain is divided into nearest neighbor cells which are
used to identify neighboring blocks which are potential contacts. Each processor is assigned a contiguous region of nearest neighbor cells. Communication
occurs via message passing (MPI) at the start of each time step. All blocks
within neighboring cells are copied between processors. Calculations are performed on blocks which do not directly interact with neighboring processors
while communication occurs. Figure 3 shows the performance of our parallel
code for increasing problem sizes and increasing numbers of processors. Duplicate calculations are performed on each processor in the region of overlap
where blocks are copied back and forth. Consequently, speedup is best for
larger problems where the region of overlap between processors is a smaller
fraction of the total work performed.

Distinct Element Method

linear--,
1000 ~

y
oo

,,
O

8000 --

I
2

I
4

t
6

I0
8

!
1

112

T
14

I
16

18

# Pr0cs
Fig. 3. Performanceof our parallel code for increasing problemsizes and increasing
numbers of processors. Speedupis best for larger problems where the region of
overlap betweenprocessors is a smaller fraction of the total workperformed.

Attack/Defense of Buried Facilities

The particular application of interest is the prediction of damage to hard

and deeply buried targets (HDBT), such as commandand control centers,
or chemical, biological, and nuclear weaponsfacilities. Figure 4 depicts three
scenarios for attacking an HDBT:
1. Direct blast into the entrance passageway.
2. Collapsing a critical access point with a single explosion.
3. Several blasts to a point well abovethe facility.
To predict damage inflicted upon hard and deeply buried targets, several
coupled regions must be modeled (see Figure 5). In the immediate vicinity
of the blast, the ground shock is sufficient to rubblize the rock, material
strength is irrelevant, and the material behavior is hydrodynamic. Deeper
into rock, material strength becomesimportant. Finally, in the vicinity of
the target area, the detailed structure of the rock mass and the facility itself
are important.
4.1

Continuum

Approach

An estimate of peak velocity, displacement, and stress experienced at a given

depth from an explosion can be used as an estimator of damage done to
deeply buried tunnels. Typically, the depth of the tunnel is large compared

Joseph Morris et al.

Surface Multi-burst

Bomb-in-Portal

Fig. 4. Three possible scenarios for attacking a hard and deeply buried target: (1)
direct blast into the entrance passageway, (2) collapsing a critical access point with
a single explosion, or (3) several blasts to a point well above the facility.

Peak.

Peak"

Fig. 5. To predict damage inflicted upon hard and deeply buried targets, several
coupled regions must be modeled. In the immediate vicinity of the blast, the ground
shock is sufficient to rubblize the rock, material strength is irrelevant, and the
material behavior is hydrodynamic. Deeper into rock, material strength becomes
important. Finally, in the vicinity of the target area, the detailed structure of the
rock mass and the facility itself are important.

Distinct Element Method

with the size of the blocks making up the rock, and continuum approaches
have been very successful in reproducing measured attenuation rates from
the source. Lomovet al., 2001 present an approach for accurately modeling
projectile penetration and explosions in rock media. Using an Eulerian code
(GEODYN)
Lomovet al., 2001 fit a constitutive model (Rubin et al., 2000)
to peak velocity and displacement attenuation data from tamped (buried)
nuclear explosions in hard rock conducted in the U. S. and in the Hoggar
massif (Algeria). The model was then validated with tamped data from nuclear explosions at Degelen Mountain. The predicted attenuation rates for
peak velocity and displacement were also in good agreement with those measured in 500-ton surface HE shots at Degelen Mountain. This continuum

ModelFit

100

1000

R (m/kt1/3
Fig. 6. Constitutive Modelfit to ensembleof nuclear tampedexplosion (peak velocity/displacement) data

treatment demonstrates that peak velocity and displacement from tamped

explosions in granitic rocks can be predicted to within a factor of two over
ten orders of magnitude in yield (see Fig. 6). However, this approach does
not provide insight into the details of the damagedone to underground structures. Traditionally a rock mass is deemedto fail when the strength of the
material is exceeded. Failed rock is no longer able to withstand load without
undergoing inelastic strains. However,hard rock strength increases markedly
with increased pressure and yet it has been observed that functional damage
or even complete tunnel collapse can occur at stress levels far below those

10

Joseph Morriset al.

previously thought to be required. The following section investigates

phenomenon.

4.2

Results

this

from Underground Explosions

Recent experiments have suggested that repeated loading of hard and deeply
buried targets by, for example, multiple bombing runs may cause functional
damageor even complete tunnel collapse at stress levels far below those previously thought to be required. Similar behavior was also observed at the
NevadaTest Site during nuclear tests. In this section we discuss someobservations of tunnels in Granite and Tuff from the Nevada Test Site, subjected
to various peak stresses. The Pile Driver and Midas Myth tests were performed in granite and tuff respectively. The tunnel sections varied widely in
their degree of rock mass reinforcement and tunnel lining or support. In the
cases presented here, it was reported that no damagehad taken place in the
granite material where the free field stress had not exceeded 700 MPa. Some
sections withstood free-field stresses in excess of 200 MPa,whereas some others were damaged at less than 100 MPa. Well engineered tunnel designs can
withstand muchhigher loads than unsupported or unreinforced sections.
Figure 7 comparestwo different sized tunnel sections within granite, reinforced with concrete. Under approximately the same loading conditions, the
larger construction failed. Figure 8 compares two tunnel segments in granite
with rock bolt and wiremesh reinforcement. Both were subjected to approximately 100 MPapeak stress and the tunnel segment orientated side-on to
the source failed. Figure 9 showsthe collapse of two excavations in tuff which
were subjected to loads well below the strength of tuff. Both structures failed,
and the discrete nature of the rock mass is evident.
In all these cases, the peak stress experienced by the structure was well below the strength of the intact rock, and yet manyof these structures failed.
The mechanism for this damage is thought to be key-block displacement.
To comprehend how this works it must first be understood that the structure surrounding an underground tunnel does not normally consist of homogeneousrock. Instead the rock is typically jointed via naturally occurring
faults or cracks. The crack density may be enhanced by the very process that
forms the tunnel, e.g., drill and blast. The resulting blocks are held in place
by frictional forces and/or cementation contacts. Whenthe ensemble block
structure is subject to intense impulsive loading it is possible that one or a
few (key) blocks on the tunnel wall can be displaced enough to release the
equilibrium state on a large section of blocks. The key block(s) then can fall
into the open tunnel and the space previously occupied by these blocks becomes available for occupation by adjacent blocks; this process can lead to
catastrophic collapse of the entire roof of the tunnel. Even if this does not
occur, the structure can be seriously weakened and subsequent loading may
cause collapse at muchlower loading.

Distinct Element Method

11

These results also indicate that engineered structures play a key role.
Furthermore, the orientation, spacing, and shear strength of geologic discontinuities (joints) control the behavior of a tunnel of a given diameter. Hard
rock joints dilate strongly before reaching peak strength, after which the
strength drops rapidly with increased loading. As a result of the controlling
effects of the joints it is not possible to estimation tunnel response via continuumbased analysis. While it is possible to augment continuum approaches
to handle fractures and faults by introducing discontinuities or slide lines into
existing grid-based methods, this approach becomesincreasingly difficult if
the connectivity of the block structures changes.

~ ~i~2.~,~.... , .........

(b)

(a)

Fig. 7. (a) A 7 ft inner diameter tunnel in concrete inside a backpacked13

diameter excavation within granite. Subjected to over 200 MPa,the tunnel suffered minimal damage.(b) A similar, but larger construction (12x7 backpacked,
reinforced concrete intersecting with a 26.7x16rock excavation) failed under comparable load.

5
5.1

Application

of DEMto Structures

in Hard Rock

Wave Propagation

In this section we demonstrate the ability of the DEMto solve wave propagation problems before using the DEMto simulate the response of realistic
tunnel geometries in jointed rock. Figure 10 shows idealized pulse, similar to
that experienced at some depth below a surface explosion. This pulse was
driven into a regular lattice of cubic DEMblocks of side length 1 m, with
normal stiffness of 3 101 Pa/m and density of 2 103 kg/m3. The pulse
is also shown after it has been transmitted through 200 1 m DEMblocks. In
this case, the joint modelwas elastic, so no energy is lost and the pulse peak

12

Joseph Morris et al.

(a)

(b)

Fig. 8. (a) A 16-ft diameter tunnel with no liner, with rock bolts and wiremesh
reinforcing within granite. Subjected to over 100 MPaend-on loading, the tunnel
suffered essentially no damage. (b) The same basic design, oriented with side-on
loading failed under the same load. The failed block structure can be discerned
through the wire mesh.

(a)

(b)

Fig. 9. (a) An excavation, reinforced with wire within tuff, collapsed at low stress.
(b) An excavation, reinforced with rockbolts within tuff, collapsed at low stress.

Distinct Element Method

is propagated with negligible attenuation.
by:

13

The expected wave speed is given

c=~

(1)

where I is the length of each block in the propagation direction, p is the block
density, and K,~ is the normal stiffness of the contacts. The simulation exhibits
a wave speed of 3820 m/s which compares well with the analytic result of
3873 m/s. Note that while this test assumed elastic joint models, hysteretic
contact forces have been considered by other authors (Heuz@et al., 1993) and
will be incorporated into our code in the near future.

0i

~_1.~

/
,

.,ll

/ /

t//i/

-3

-351VI/

~/V

-4

-- depth Om
--- depth199m

005

0.1

0.15

0.2

0,25

time(s)
Fig. 10. A idealized typical velocity profile is well propagatedthrough a stack of
200 1 m DEMblocks.

5.2

Simulations

of Tunnels

Finally we present preliminary results from a investigation into the stability of underground structures subjected to dynamic loading. The effects of
explosions on underground structures were studied in two dimensions by
Heuz@et al., 1993. Figure 11 shows a tunnel in jointed rock. The average
block size is approximately 1 m. The jointed rock mass is confined by 7.5 MPa
of lateral and vertical stress. The upper boundary was movedwith the driving
pulse shownin Fig. 10.
The simulation predicts that a substantial portion of the tunnel roof will
collapse, makingthe tunnel unusable. Althoughthis simulation is preliminary,

14

Joseph Morris et al.

it demonstrates the capabilities of the DEMwith this kind of problem. The

joint structure is realistic, with non-orthogonal joint planes. Blocks are free
to move and make new contacts with other blocks in the simulation. After
further validation, this simulation capability will be applied to a parameter
study of the stability of undergroundstructures.

Fig. 11. A preliminary simulation of a tunnel in jointed rock subjected to the pulse
shownin Fig. 10. The simulation predicts that a substantial portion of the tunnel
roof will collapse, makingthe tunnel unusable.

Discussion

We have seen that continuum approaches can provide some details of potential damageto underground structures. Peak velocities and displacements
are well predicted to within a factor of two over ten orders of magnitude
in yield. However, this information along cannot provide complete damage
estimates. In particular, it has been observed that substantial damage, or
total collapse of a tunnel can occur at stresses well below the strength of the
rock. The mechanism for this damage is thought to be key-block displacement. Realistic simulation of failure Distinct element methods can simulate
collapse of tunnels for realistic fracture set geometryincluding the effects of
block motion.
In practice, however, we only have limited knowledgeof local fault zones.
To provide bounds on the response, we must study a range of probable
fault geometries. That is, we must perform stochastic analysis with many
realizations required to obtain adequate statistics to bound results. Current
three-dimensional scalar codes take days to simulate the dynamic loading Of
structures containing ~- 104 - 105 distinct elements. Wehave presented our
approach to parallelization
of the DEMby employing domain decomposition.
The common-plane approach to contact detection is both efficient and
easy to implement. However, under some circumstances the method of Cundall, 1988

Distinct

Element Method

15

gives misleading contact classifications.

The contact type is based upon counting the number of penetrating
vertices.
Thus, it is possible that a contact
which would be better approximated by two interpenetrating
faces will be
misclassified
as an edge to face contact. We will be investigating
more accurate contact detection
algorithms which build on the common-plane approach. Other future development of our DEMcode will include coupling
with other codes and the inclusion of structural
elements. In addition, more
advanced joint models including hysteresis (Heuz~ et al., 1993) will be incorporated into our code in the near future.
Acknowledgement This work was performed under the auspices of the U.S.
Department of Energy by the University of California,
Lawrence Livermore
National Laboratory under Contract No. W-7405-Eng-48.

References
Antonellini M.A., Pollard D.D. (1995) Distinct element modeling of deformation
bands in sandstone. J. Struct. Geol., 17:1165-1182
Carrillo A.R., West J.E., Horner D.A., Peters J.F. (1999) Interactive large-scale soil
modeling using distributed high performance computing environments. Int. J.
High Perf. Comput. Appl., 13:1:33-48
Cleary P.W. (1991) Extensions of the hybrid method for granular flows. In: Proc.
5th International Computational Techniques and Applications Conference, Adelaide, Australia
Cleary P.W., M.L. Sawley (1999) Three-dimensional modelling of industrial granular flows. In: Second International
Conference on CFDin the Minerals and
Process Industries, CSIRO, Melbourne, Australia, 95-100
Cundall P.A. (1980) UDEC-Ageneralized distinct element program for modelling
jointed rock, Final Tech. Rep. Eur. Res. Office (US Army Contract DAJA3779-C-0548); NTIS order No. AD-A087610/2
Cundall P.A. (1988) Formulation of a three-dimensional distinct element modelPart I. A scheme to detect and represent contacts in a system composed of many
polyhedral blocks. Int. J. Rock Mech. Min. Sci. & Geomech.Abstr. 25:107-116
Cundall P.A. (2001) A Discontinuous Future for Numerical Modelling in Geomechanics? Geotech. Eng., 149:1:41-47
Cundall P.A., Hart R.D. (1985) Development of generalized 2-D and 3-D disinct
element programs for modeling jointed rock, Misc. Paper SL-85-1, US Army
Corps of Engineers
Cundall P.A., Hart D.H. (1992) Numerical Modelling of Discontinua. Eng. Comput.,
9:101-113
Cundall P.A., Strack O.D.L. (1979) A discrete numerical model for granular assemblies. G~otechnique, 29:47-65
Dowding C.H., Dmytryshyn O., Belytschko T.B. (1999) Parallel processing for
discrete element program Comput. Geotech. 25:4:281-285
Ghaboussi J., Basole M.M., Ranjithan S. (1993) Three-dimensional discrete element
analysis on massively parallel computers. In: Second International Conference
on Discrete Element Methods, MIT, Cambridge, MA

16

Joseph Morris et al.

Hahn J.K. (1988) Realistic animation of rigid bodies. Comp. Graph. 22:299-308
Hart R., Cundall P.A., Lemos J. (1988) Formulation of a Three-dimensional Distinct
Element Model- Part II. Mechanical Calculations for Motion and Interaction of
a System Composed of Many Polyhedral Blocks. Int. J. Rock Mech. Min. Sci.
& Geomech. Abstr. 25:117-125
Heuz~ F.E., Walton O.R., Maddix D.M., Shaffer R.J., Butkovich T.R. (1993) Analysis of Explosions in Hard Rocks: The Power of Discrete Element Modeling. In:
Hudson J.A., Brown E.T., Fairhurst C., Hoek E. (Eds.) Comprehensive Rock
Engineering, Vol. 2, Analysis and Design Methods, Pergamon Press, 387-413
HomerD.A., Carrillo A.R., Peters J.F., West J.E. (1998) High resolution soil vehicle
interaction modeling. Mech. Struct. & Mach., 26:3:305-318
LomovI., Antoun T., Glenn L. (2001) Explosion in the granite field: Hardening and
softening behavior in rocks. In: Proceedings of 12th APSTopical Conference,
Shock Compression of Condensed Matter, June 24-29, Atlanta, Georgia
Lorig L.J., Brady B.H.G., Cundall P.A. (1986) Hybrid distinct element-boundary
element analysis of jointed rock. Int. J. Rock. Mech. Min. Sci. & Geomech.
Abstr., 23:4:303-312
MorganJ.K. (1999) Numerical simulations of granular shear zones using the distinct
element method 1. Shear zone kinematics and the micromechanics of localization. J. Geophys. Res., 104:B2:2703-2719
MorganJ.K. (1999) Numerical simulations of granular shear zones using the distinct
element method 2. Effects of particle size distribution and interparticle friction
on mechanical behavior. J. Geophys. Res., 104:B2:2721-2732
Mori K., Otsu M., Osakada K. (1997) Distinct element simulation of grain alignment
in mushy-state forg!ng of magnets. Int. J. Mech. Sci., 39:7:771-780
Munjiza A., Owen D.R.J., Bicanic N. (1995) A combined finite-discrete
element
method in transient dynamics of fracturing solids. Eng. Comput. 12:145-174
Pan X.D., Reed, M.B. (1991) A coupled distinct element-finite
element method
for large deformation analysis of rock masses. Int. J. Rock Mech. Min. Sci. &
Geomech. Abstr. 28:1:93-99
Rubin M.B., Vorobiev O.Y., Glenn L.A. (2000) Mechanical and numerical modeling of a porous elastic-viscoplastic
material with tensile failure. International
Journal of Solids and Structures 37:1841-1871
Sanderson D.J., Zhang X. (1998) Deformation, damage and fluid flow in fracture
networks and around faults. Fall Meeting of the American Geophysical Union.
Sawamoto Y., Tsubota H., Kasai Y., Koshika N., Morikawa H. (1998) Analytical
studies on local damage to reinforced concrete structures under impact loading
by discrete element method. Nucl. Eng. Des., 179:157-177
Sawley M.L., Cleary P.W. (1999) A parallel discrete element method for industrial
granular flow simulations. EPFL Supercomputing Review 11:23-29
Shi Gen-Hua, (1988) Discontinuous deformation analysis- a new numerical model
for the statics and dynamics of block systems. PhDThesis, University of California, Berkeley
Shi Gen-Hua, Goodman R.E. (1988) Discontinuous deformation analysis- a new
method for computing stress, strain and sliding of block systems. In: Cundall P.A. et al. (Eds.) Key Questions in Rock Mechanics, Balkema, Rotterdam,
381-383
Walton O.R. (1980) Particle dynamics modeling of geological materials, Rep.
UCRL-52915, Lawrence Livermore National Laboratory

Distinct

Element Method

17

Williams J.R., Hocking G., Mustoe G.G.W. (1985) The theoretical

basis of the
discrete element method. In: NUMETA
85, Balkema, Rotterdam
Williams J.R., Mustoe G.G.W. (1987) Modal methods for the analysis of discrete
systems, Comp. Geotechnics 4:1-19