SPECIAL SECTION: COMPUTATIONAL SCIENCE

RESEARCH ARTICLES

Numerical simulation of particulate materials

using discrete element modelling
Thallak G. Sitharam†
Department of Civil Engineering, Yamaguchi University, Tokiwadai 1-16-1, Ube 755 8611, Japan

cal solution techniques such as finite element,

Discrete element method (DEM) is a numerical technique finite difference and boundary element methods are
pioneered by Cundall1 for problems in rock
founded on the continuum assumption and have been
mechanics where the continuity between the separate ele-
proved as extremely useful tools for the engineer to solve
ments does not exist. Discrete numerical simulations with
particles are currently used in several different scientific many complex engineering problems. However, the contin-
disciplines. The DEM is capable of handling a wide range of uum mechanics models are phenomenological and are pri-
material behaviour, inter-body interaction force laws and marily concerned with mathematical modelling of the
arbitrary geometries. DEM models granular materials observed phenomenon without detailed attention to their
which can freely make and break contacts with their fundamental physical significance. In particulate materials
neighbours and is capable of analysing interacting deform- forces are transferred through contacts between the parti-
able bodies undergoing large absolute or relative motions. cles. Problems occur with the continuum assumption due to
DEM-related research has proved to be very meaningful for the particulate material’s inherent granular nature and the
enhancing the understanding of granular materials. This consequent deformation and failure modes. This discrete
tool has been used for developing and validating constitu- behaviour makes the constitutive relationship of particulate
tive relationships of particulate materials such as soil, media complex and many laboratory tests are necessary to
rock, grain or ceramic powder by using appropriate parti- understand the behaviour before the physical processes in
cle properties, sizes, shapes, and gradation. Moreover, these materials can be modelled. In materials dominated by
DEM is the best tool for explaining the experimentally ob-
particulate behaviour it is advantageous to treat the medium
served facts from a more fundamental approach. Perhaps
as an assemblage of particles rather than as a continuum as
one of the major advantages of the discrete particle model
approach is that what is being done during a simulation is it permits the exploration of actual mechanisms involved.
obvious at all the times. The discrete element model incor- Also, there are many situations especially in large deforma-
porates induced anisotropy, shear dilatancy and stress path tion problems and fracture in engineering materials where
dependency from more fundamental principles and has continuity cannot be assumed and other techniques must
proven its usefulness to generate insights into the proc- be utilized.
esses involved with particulate media. DEM is an explicit An alternative approach is discrete element model (DEM),
finite difference code and because of the explicit nature of which is capable of analysing multiple interacting deform-
the algorithm, it is necessary to use a very small time-step able continuous, discontinuous or fracturing bodies, un-
of simulation to guarantee numerical stability and accu- dergoing large displacements and rotations. The method is
racy. Many researchers are attempting to expose the ad- general in its ability to handle a wide range of material con-
vent of the effective use of the supercomputer and parallel stitutive behaviour, interaction laws and arbitrary geome-
computing for engineering problems by modifying the dis- tries. Currently, there are two-dimensional models available
crete element simulation codes. This will allow a major with particles as circular discs, ellipses and polygons and
step forward in handling a very large number of particles
three-dimensional models with particles as spheres,
for the more realistic simulations of particulate solids. In
blocks/bricks, and ellipsoids. Also DEM has progressed
this paper, an overview of the discrete element modelling is
presented along with some of its applications in engineer- from the early rigid and simply deformable elements to its
ing. present state of complex elements which place at its dis-
posal the full range and the power of the finite element
method. The method solves for the dynamic equilibrium
1. Introduction equations for each body subjected to body and boundary
forces. Highly dynamic effects can also be simulated in-
The continuum assumption is used as a basis for the idea- cluding stress wave propagation, vibration and damping2.
lization of many materials such as soils, rocks, rockfills, ag- DEM has evolved from various disciplines such as
gregates, concrete and other engineering materials. Numeri- physics of particles3,4, from geomechanics1,5,6 and from
structural engineering7,8. Theoretical background for the
†
On leave from Indian Institute of Science, Bangalore 560 012, method is derived from the fields of aero-elasticity9, the fi-
India. nite element method10 and the finite difference method11.
e-mail: [email protected]
876 CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000
 SPECIAL SECTION: COMPUTATIONAL SCIENCE

Discrete numerical simulations with particles are currently interaction of large number of elements without excessive
used in several different scientific disciplines, to examine a memory requirements.
wide range of physical systems with inherent ‘granularity’ One of the important elements of the DEM model is the
(stellar dynamics, hydrodynamics, molecular dynamics, and explicit incorporation of Coulomb’s frictional behaviour at
electrostatic plasma; Hockney and Eastwood12). The DEM contacts between elements. Slippage between elements is
approach has been applied to problems on an atomic scale, permitted, when the tangential or shear force at a contact
max
to modelling of sedimentary rocks13,14, and to slope stability exceeds a critical value, F s defined
, by:
problems in which particle dimensions are up to many feet15.
The method is used in process engineering to study the Fsmax = c + tan φµ Fn , (1)
flow of granular media in hoppers or down-inclined
chutes 16,17; the development of stresses around tunnel where Fn is the normal force at the contact, and c and φµ are
openings in jointed rocks18; nonlinear static and dynamic the cohesion and the angle of internal friction at contacts,
soil structure interaction problems19; and the stress and respectively.
flow response of naturally fractured fluid reservoirs13. Cun-
dall and coworkers extended the DEM to soil using two-
dimensional discs20 and three-dimensional spheres21 and
2.1 Calculation cycle of DEM
subsequently these models were extended
A typical calculation cycle of DEM is shown in Figure 1.
to handle particles of any shape. Numerical simulations
Each DEM cycle involves application of Newton’s second
using DEM have been attempted by several workers1,14,20,22–
26 law of motion, the velocities x& and θ&which will be used in
to model granular media.
the force–displacement are obtained from the current resul-
An excellent source of references with respect to DEM is
tant force and the moment acting on the disc, given by:
contained in refs 27 and 28.
m &x& = ΣFi ,
(2)
Iθ&& = ΣM ,
2. Discrete element method
where Fi is the unbalanced force on each element, M is the
DEM is a numerical technique pioneered by Cundall1 for resultant moment on each element, m is the mass of the ele-
rock mechanics problems where the continuity between the ment, I is the moment of inertia of the element and
separate elements does not exist. In DEM, the equilibrium are the element accelerations. For each element, accelera-
contact forces and displacements of a stressed assembly of tions are integrated over a small time-step to give velocities
discs are found through the medium of disturbances origi- x& and θ& and displacements (x and θ). The time-step cho-
nating at the boundaries, which is a dynamic process. The sen is small enough that the velocities and accelerations are
(&x& and θ&&)
speed of propagation is a function of physical properties of constant over it. This is followed by a force displacement
the discrete medium. DEM is based on the idea that this law through which the unbalanced forces on each individ-
time-step may be chosen so small that during a single time- ual element are calculated. A simple interaction law at the
step disturbances cannot propagate from any disc further contact of the two interacting discs can be written as:
than its immediate neighbours. It is this key feature of DEM
which makes it possible to model the ( Fn ) new = ( Fn ) old + ∆Fn ; ( Fs ) new = ( Fs ) old + ∆Fs ,

Figure 1. A typical calculation cycle in DEM.

CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000 877

SPECIAL SECTION: COMPUTATIONAL SCIENCE

∆ Fn = K n ∆n; ∆Fs = K s ∆s, (3) to achieve equilibrium. Following the application

of Newton’s second law, a test for existence of physical
where Fn and Fs are the normal and shear contact forces, Kn contacts is performed. If the contact exists between two
and Ks are the element stiffness in normal and shear direc- elements, incremental normal and tangential contact dis-
tions, and ∆n and ∆s are the normal and shear relative dis- placements are calculated from rigid body movements of
placements at a contact. ∆Fn and ∆Fs are the each element, and the linear and angular velocities at each
incremental normal and shear contact forces respectively. element centre during a fixed time interval. It may be noted
The components ∆n and ∆s of the relative displacements at that the linear force displacement law discussed above is
a contact are obtained by integrating the relative velocity not inherent in the method. A nonlinear law such as that of
component with respect to time. The relative velocity com- Hertzian contact could equally well be employed and has
ponents are calculated by projecting relative velocity at the been used in many simulations presented in the next sec-
contact point in normal and tangential directions to the con- tion. Calculations are repeated till the unbalanced forces on
tact plane. the particles are very small or zero. A general purpose DEM
The calculations presented are repeated till the unbal- can treat nonlinearities which may arise from large dis-
anced forces on the particles are very small or zero for static placement, rotation, slip, separation and material behaviour.
problems. Interaction relationships for other types of con- The incremental motion of the elements may give rise to
tacts such as rock joints may include cohesion, dilation, interactions between elements and thus strain rates are
damage to asperites and stress-dependent friction. Also generated within the elements. From the constitutive be-
there is no restriction on where one element may make con- haviour of the material, the element will experience elastic or
tact with another. The connectivity or interaction of element viscoelastic behaviour or possibly fracturing. Fracturing of
to element is computed automatically in the algorithm, since a discrete element is relatively easy because all interactions
the elements can rapidly change neighbouring elements with other elements occur through boundary forces2. Thus
upon motion or fracturing. once the geometric bookkeeping of creating two elements
Rheological elements of the discrete element model for a from one has been accomplished, the standard discrete ele-
simple disc element are shown in Figure 2. In DEM artificial ment algorithms will take care of determining new contacts
damping is incorporated to dissipate the kinetic and forces. It must be mentioned here that the structure for
energy so as to bring the assembly to static equilibrium fracturing is relatively straightforward, however the rules for
under imposed boundary conditions. DEM includes two fracturing are not well developed29.
forms of viscous damping such as contact damping and
global damping. Contact damping (Cn) operates on relative
velocities at the contacts and is visualized as resulting from
dashpots acting at normal and shear directions at the con-
tact as shown in Figure 2. The contact damping coefficients
are taken to be proportional to contact stiffnesses. Global
damping (Cm and C1) operates on the absolute
velocities of discs and is visualized as the effect of dash-
pots connecting each particle to the ground. Coefficients of
global damping are taken proportional to absolute velocities
of both translational (Cm) and rotational movement (C1).
Friction damping (Cs) occurs during sliding when the abso-
lute value of the shear force at each contact is at the value
of F ms ax. The use of additional damping other than friction is
necessary in order that the assemblies reach a state of equi-
librium for all conditions. If neither contact nor global
damping were included, the assemblies could only reach
equilibrium if slip occurred.
Although the assembly is dynamic, static equilibrium is
achieved by applying the boundary stresses in terms of
small increments for all the static problems. Equilibrium con-
tact forces and displacements are found in an assembly of
particles, stressed by imposed boundary conditions
through a series of calculation cycles which trace the
movement of each particle. These movements are the
results of diffusion of the disturbance due to external app-
lication of forces and displacements. Because of the unbal- Figure 2. Rhelogical elements of DEM along with the damping
anced forces, the particles try to rearrange themselves so as mechanisms adopted in DEM.

878 CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000

 SPECIAL SECTION: COMPUTATIONAL SCIENCE

2.2 Method of analysis good to conclude that the discrete element model simulates
well the contact interaction and is a valid tool for fundamen-
The method of analysis consists of mainly three steps30: tal research in understanding
micromechanical behaviour of particulate media. Sitharam14
1. Contact processing and detection: The essence of con-
has presented several other validation tests using the DISC
tact detection is the determination of the surfaces and vol-
model.
umes of intersection between two bodies. Once the contact
The power of the technique can be appreciated from the
geometry has been determined, one can easily enforce the
idealized uniaxial compression test illustrated in Figure 4
contact conditions. In this case the program should deter-
(ref. 2). The left-hand plate moves towards the specimen
mine the contacts automatically within itself. This is done
against a second fixed plate. When the stress in the speci-
through a contact algorithm which checks every face/point
men exceeds a user-defined limit, fracture occurs. Further
contact of every element with every face/point of every
compression causes large motion and repeated fracturing,
other element.
with some fragments flying off with high velocity. It should
2. Contact and interaction forces: The numerical modelling
be noted that the fracturing proceeds without intervention.
of contact between bodies is a rather complex subject. It is
The fracture orientation, new element generation and new
however noted that any form of interaction law or numerical
interactions are automatically calculated within the program.
technique can be accommodated within the discrete element
This relatively straightforward discrete element analysis
framework. A most significant feature of the contact prob-
would be extremely difficult with conventional FEM analy-
lems involving friction is that deformations, in general, de-
sis.
pend on the loading history and thus the loading process is
followed with time. The earlier contact interaction law de-
4. Applications in engineering
scribed is a linear interaction model.
3. Element motion in which the dynamic equation of Many applications of DEM are reported in the literature. In
motion is applied for rigid body motion as applied to a sin- this section only selected applications of DEM in engi-
gle discrete element acted upon by an arbitrary set of point neering in general and structural engineering in particular,
forces. Adopting an incremental displacement geomechanics encompassing both soil and rock mechanics
approach linearizes these equations. applications, and granular flow problems have been pre-
sented. Especially in rock mechanics and soil mechanics
3. Validation tests applications, the joint/contact behaviour is the controlling
parameter and the intact rock/soil element deformations may
DEM consisting of 2-dimensional discs was developed and
have little effect. In such cases the rigid body DEM is an
validated by Cundall and Strack31. They have compared the
excellent tool. The following selected DEM applications
force vector plots obtained from the discrete element model
from the literature are presented to highlight the power of
consisting of 2-dimensional discs with the corresponding
discrete element modelling.
force vector plots obtained by photo-elastic experiments by
Figure 5 shows the three-point loading of a concrete
De Josselin de Jong and Verruijit32 as shown in Figure 3.
beam consisting of bonded discs33. Figure 5 a shows the
The correspondence between the plots was sufficiently

Figure 3. Comparison of force vector plots obtained from the discrete element model31 consisting of two-
dimensional discs with the corresponding force vector plots obtained by photo-elastic experiments by De
Josselin de Jong and Verruijit32 .

CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000 879

SPECIAL SECTION: COMPUTATIONAL SCIENCE

force distribution prior to cracking and Figure 5 b shows the

particle configuration after some bond breakage has oc-
curred. This test was simulated using particle flow code in
two dimension (PFC2D) of Itasca33. The particles may repre-
sent individual grains in a material such as sand, or they
may be bonded together to represent a solid material such
as concrete or rock. Fracturing was modelled when bonds
break under load. The example described here presents simple results of a
particulate mechanics approach to understand the constitu-
tive behaviour of granular materials to give a flavour of this
emerging branch of materials science. The essence of the
a modelling approach is illustrated in Figures 6 and 7 (ref. 35).
Figure 6 shows the computer-generated assembly of two-
dimensional disc particles at the start of the simulation and
several stages of compaction from initially generated no
touching discs. By applying isotropic stress on the bound-
ary particles of the initially generated assembly, compaction
was initiated. One can clearly see how contact forces gener-
b ate as the sample gets compacted and finally a homogene-
ous sample with uniform distribution of contact forces
between the discs is achieved using DEM (see Figure 6 d ).
Figure 7 a shows the stress–strain curves for a biaxial test
conducted on the compacted
assembly. Based on micromechanics, stress has been repre-
sented by the contribution from arrangement of particles
and their orientations (known as fabric), and contact forces
c
including normal and shear forces are presented in Figure
7 b. For the details one can refer to Sitharam35. The beauty
of particulate simulation using DEM is the availability of
micro-parameters at all stages of the simulation. Using the
micro-parameters such as average coordination number,
contact forces (both normal and shear) and their orienta-
tions and displacement trajectories, one can look for better
d understanding of the material behaviour and develop or
validate relationships between
micro- and macro-parameters (stress, strain, etc.). For these
simulations, we have used only linear contact law. However,
the stress–strain curve observed was nonlinear. Nonlinear
behaviour is attributed to rearrangement of particles
(change in fabric) and the macroscopic averaging of the
Figure 4. Idealized uniaxial compression test simulation-using movement of particles over a large assembly of particles. In
DEM (Mustoe et al.50 ). real granular systems, nonlinear contact interaction may be

a b

Figure 5. Three-point bending test of a concrete beam consisting of bonded discs33 .

880 CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000

 SPECIAL SECTION: COMPUTATIONAL SCIENCE

ascribed to surface asperities or slight analyses of rock masses.

irregularities in particle shape. Although a nonlinear contact Figure 8 shows a simulation of the progressive caving of
interaction model such as Hertzian contact theory charac- a mine using a two-dimensional rigid block code UDEC
terizes the relationship between contact force and inter- (universal discrete element code)34. This code can be used
particle displacement in a more realistic manner, we have
observed that the stress–force–fabric relationship for a ho-
mogeneous assembly is independent of the contact interac-
tion model adopted for two-dimensional systems14,22. a
However, it is still of interest to further explore the outcome
of such implementation for three-dimensional particulate
assembly.
DEM has been applied to dynamical problems, such as
granular flow problems36–38, impact problems39, and geody-
namical problems40,41. Examples of granular flows
include transportation of food grains, pharmaceuticals and
coal through hoppers, chutes, and conveyer belts and also
fluidized bed reactors. For these examples, a dynamic proc-
ess was simulated using DEM by not considering the global
and mass damping options.
Other applications include packing of granular materials,
particulate segregation and mixing, and particulate drying. It
has also been extensively applied to quasi-static problems,
both two- and three-dimensional, with discs and spheres,
respectively30,39,42–44. Ohnishi and Kono45, Hart, Cundall and b
Lemos15 and Harper and Last13 have applied DEM to the

a b

Figure 7. Stress–strain plots and their micro-parameter constitu-

tions in terms of anisotropy coefficients.

c d
to study two-dimensional analysis of jointed rock or granu-
lar media with different particle shapes and sizes. The pro-
gram can also be used for stability analysis of dams on
jointed rock foundations, design and construction of open
cuts and underground excavations and also for mine extrac-
tion sequences, support requirements and mining methods.

Figure 6. Compaction simulation of discs51 .

CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000 881

SPECIAL SECTION: COMPUTATIONAL SCIENCE

Figure 9 shows a simulation of failure of an unlined tunnel

using UDEC. Figure 10 shows an application of three- a
dimensional DEM using 3DEC of Itasca for modelling of an
excavation through jointed rock.
Figure 10 shows the perspective view of 3DEC model with
several foreground blocks hidden to expose the
tunnel. The model contains 616 deformable blocks (with

Figure 8. Progressive caving of mine34 .

Figure 11. a, Footing response on a granular soil; b, Trajectories

of disc particles below a footing47 .

56,700 degrees of freedom) and 33 joints have been

modelled. It requires approximately 7 h on an 80486 com-
puter to reach a solution for excavation-induced stress ef-
fects for each excavation step46.
To apply DEM to a full-scale situation in geotechnical
engineering, one could utilize real-sized particles for the
entire problem with suitable material properties and full-
Figure 9. Failure of an unlined tunnel using UDEC33 .
scale loads. However this is impossible considering the
computational intensity required for handling a very large
number of particles. Nevertheless, application of the DEM
to realistic geotechnical problems may be possible with ap-
propriate measures47. The key to applying the DEM to a full-
scale problem lies in ensuring proper nonlinear stress–
strain–strength soil behaviour and maintaining stress
similitude in the simulation. Ting et al.47 have used the
centrifuge modelling principle, in which a small-scale model
of the full-scale problem is constructed using the same ‘soil’
behaviour. To guarantee stress similitude, the gravitational
Figure 10. Modelling of a tunnel (excavation) through jointed acceleration is increased by a factor α equal to the length
rock using 3DEC46 . reduction. However in rock mechanics, full-scale simula-
tions are being carried out (as shown in Figure 10) in real
field scale problems46,48. A bearing capacity test conducted

882 CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000

 SPECIAL SECTION: COMPUTATIONAL SCIENCE

by Ting et al.47 on the 8500 disc bin by translating a 0.04 m place because of reduction in contact forces around the
wide (model unit) rigid ‘footing’ vertically at 0.5 m/s for fracture, which causes concentrated deformations domi-
0.0225 s has been reported here. After 0.0225 s (0.28 of the nated by plastic slip and parting. The grains on the wall of
footing width), the footing was stopped and the bin allowed the fracture are not free floating, they are pushed against
to equilibrate until 0.03 s. Figure 11 a shows the initial 8500 the walls of the fracture by the pressure drop across them.
disc bin and footing. Figure 11 b shows the actual dis- This is a demonstration of the stabilizing effect of fluid flow.
placement vectors of disc particles at the end of the test. It is also pointed out here that the particulate medium
The ultimate footing pressure applied compares favourably around the fracture is dilating and yielding as a conse-
with the ultimate bearing capacity from Terzaghi’s theory quence of the very low effective stress arising because of
for sand with angle of internal friction (φ) = 26° and density high pore pressure. In an isotropically compacted assembly,
(γ) = 2000 kg/m3. Also, the incremental normal stress con- there is no preferential direction for fracture propagation;
tours at the end of the bearing capacity test compare well the fracture, once initiated because of large deformation
with Boussinesq’s stress contours (not shown here). around the injection point, creates its own anisotropy and
Simulation of a injection test in a particulate medium has propagates in the path of least resistance.
been carried out to model hydrofracturing in granular media A simulation of multiple interacting fractures in a discrete
using a two-dimensional disc model coupled with fluid element particulate system is shown in Figure 13 a and b.
flow14. These tests are carried out to study the fracture ini- The figure also shows the interaction of multiple (two) frac-
tiation pressures and on fracture orientations in cohesion- tures in an isotropically compacted system. Figure 13 a
less granular media where conventional fracture mechanics shows the contact force distribution in a
concepts are not applicable. Figure 12 shows the results of discrete element system when the two fractures are inter-
injection-induced deformation in the two-dimensional as- acting with each other. Figure 13 b shows the displacement
sembly of discs. Figure 12 a indicates the fracture initiation trajectories corresponding to the stage of Figure 13 a from
in the isotropic disc assembly due to injection in a central the initially compacted system.
pore or a void. This fracture is stable and is slowly growing
as the injection pressure is gradually increased. Figure 12 b
and c show the fracture propagation due to further increase a
in injection pressures. Contact-force redistribution takes

a b

b
b c

Figure 12. Hydrofracture initiation and propagation simulation in Figure 13. Multiple fracture interactions in a discrete element
a two-dimensional disc assemblage14 . system 14 .

CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000 883

SPECIAL SECTION: COMPUTATIONAL SCIENCE

Figure 14. Simulations of a penetrating pile in sand40 .

884 CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000

 SPECIAL SECTION: COMPUTATIONAL SCIENCE

Figure 14 shows the dynamic nonlinear response of struc- gradation. Moreover, DEM is the best tool for explaining
tural foundations, in particular the penetrating pile in a sand the experimentally observed facts from a more fundamental
bed 40. In this figure, the first column clearly shows the approach.
stages of penetration of a pile in a two-dimensional circular The results of compaction of 1000 disc particles
particle system. The second and third columns highlight the (Figure 6 a; after Sitharam35) to a dense system takes about
disc trajectories and contact force in the system. This exam- 6 h to perform on an 80486 microcomputer. More than the
ple highlights the failure pattern formed below a penetrating discrete element calculation cycles, bookkeeping of the par-
pile and 2 : 1 slope, which is generally assumed in all geo- ticles and their contacts are very time consuming. The com-
technical designs. This clearly shows that the observed puting power of today’s state-of-the-art computer
failure mechanism is realistic and technology offers the possibility to consider even more
the failure surfaces formed with almost 2 : 1 slope from the realistic three-dimensional systems with spheres, ellipsoids
point above the pile tip (see bottom most picture in column and blocks to simulate particulate materials, which was pro-
3 of Figure 14). hibitively complex and expensive a few years ago. However,
due to the explicit nature of the algorithm, it is necessary to
use a very small time-step of simulation to guarantee nu-
5. Other practical applications
merical stability and accuracy. To obtain proper aggregate
behaviour, enough particles must be used to ensure suffi-
Other areas in which useful practical results may be
cient number of contacts along any shear surface or struc-
obtained using the DEM include:
tural interface. Many researchers are
1. Soil–structure interaction problems. attempting to expose the advent of the effective use of the
2. Dynamic pressures behind a flexible retaining wall. supercomputer and parallel computing by modifying the
3. Ice mechanics and other related applied mechanics discrete element simulation codes. Slowly consensus is
problems. emerging on the use of a parallel processing machine with
4. Fundamental aspects of particulate mechanics and con- good graphics to simulate particulate materials using DEM
stitutive behaviour of soils and rocks. rather than vector processing machines49. It may
5. Soil-reinforcement interaction in reinforced earth struc- be expected that these studies will allow a major step
tures, viz. reinforced retaining walls and soil beds. forward to the more realistic simulations of particulate solids
6. Design of stone or gravel column in soft clays. and a better understanding of the constitutive behaviour of
7. Analysis and design of earth anchors. particulate materials.
8. Fracture mechanics and large displacement nonlinear
problems. 1. Cundall, P. A., Proc. of the ISRM Symp., Nancy, France, 1971,
vol. 2, pp. 129–136.
Many new applications using DEM are being carried out, 2. Pande, G. N., Beer, G., Williams, J. R., Numerical Methods in
e.g. simulation of comminution in ball mills, slurry flow, flow Rock Mechanics, John Wiley, New York, 1990, pp. 229–312.
of powders and grains, and physical animation for computer 3. Ashley, H., A.I.A.A. J., 5 March 1967, pp. 460–469.
graphics. Materials other than those discussed earlier, such 4. Huston, R. L. and Passerello, C. E., Comput. Struct., 1980, 12,
249–276.
as ice plates, masonry and jointed rocks are also ideal can- 5. Burman, B. C., Ph D thesis, James Cook University of North
didates for modelling using DEM. Queensland, Townsville, Australia, 1971.
6. Hocking, G., Ph D thesis, Imperial College, 1977.
7. Trolleppe, D. H., Proc. 4th Int. Conf. S.M. and F.E., London,
6. General remarks 1957, vol. 2, pp. 383–388.
8. Kawai, T., IABSE Colloquium, Copenhagen, 1979.
One can envisage many advantages and uses for the DEM. 9. Bisplinghoff, R. L. and Ashley, H., Principles of Aeroelasticity,
Dover, New York, 1962.
The power of DEM in simulating the real processes and in 10. Zienkiewicz, O. C., The Finite Element Method, McGraw-Hill,
identifying the mechanisms is evident. Thus it is an useful UK, 1977, 3rd edition.
tool in understanding the physics of the processes and 11. Wilkins, M. L., Calculation of Elastic–Plastic Flow, Methods in
problems. DEM clearly simulates deformation mechanisms Computational Physics, Academic Press, New York, 1964, vol.
3.
in jointed and particulate media in a more realistic way than 12. Hockney, R. W. and Eastwood, J. W., Computer Simulation us-
any of the continuum-based models. Because it can monitor ing Particles, McGraw-Hill, New York, 1981.
internal stresses and contact behaviour unobtrusively, 13. Harper, T. R. and Last, N. C., Geotechnique, 1989, 39, 1–11.
sample reproducibility can be guaranteed, and it is ideal for 14. Sitharam T. G., Ph D thesis, University of Waterloo, Waterloo,
Ontario, Canada, 1991, p. 303.
use in understanding fundamental particulate material be-
15. Hart, R., Cundall, P. A. and Lemos, J., Int. J. Rock Mech. Min.
haviour. In addition, since external stresses and stress Sci. Geomech. Abstr., 1988, 25, 117–125.
paths can be controlled precisely it can be used for devel- 16. Campbell, C. S. and Brennen, C. E., in Mechanics of Granular
oping and validating constitutive relationships of any par- Materials: New Models and Constitutive Relations (eds Jenkins,
ticulate materials such as soil, rock, grain or ceramic powder J. T. and Satake, M.), Elsevier, Amsterdam, 1983, pp. 326–
373.
by using appropriate particle properties, sizes, shapes and
CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000 885
SPECIAL SECTION: COMPUTATIONAL SCIENCE
17. Hawkins, G. W., in Mechanics of Granular Materials: New Mod- 33. Particle flow code in 2-dimensions (PFC2D), ITASCA, 1995.
els and Constitutive Relations (eds Jenkins, J. T. and Satake, M.), 34. Universal Distinct Element Code (UDEC), ITASCA, 1993.
Elsevier, Amsterdam, 1983, p. 305. 35. Sitharam, T. G., in Distinct Element Modeling in Geomechanics
18. Lorig, L. J. and Brady, B. H. G., in Design and Performance of (eds Sharma, V. M., Saxena, K. R. and Woods, R. D.), Oxford &
Underground Excavations ISRM/BGS, Cambridge, England, IBH, New Delhi, 1999, Chapter 3, pp. 47–88.
1984, pp. 105–112. 36. Walton, O. R., in Mechanics of Granular Materials: New Models
19. Ting, J. M. and Corkum, B. T., Proc. CSCE Annual Conference, and Constitutive Relations, Elsevier, Amsterdam, 1983, pp.
Calgary, Canada, 1988, vol. 3, pp. 196–215. 327–338.
20. Strack, O. D. L. and Cundall, P. A., Report to NSF concerning 37. Walton, O. R., Braun, R. L., Mallon, R. G. and Cervelli, D. M.,
Grant Eng. 7620711, Part I, Dept of Civil and Mineral Engi- in Micromechanics of Granular Material, Elsevier, Amsterdam,
neering, Univ. of Minnesota, USA, 1978. 1988, pp. 153–161.
21. Cundall, P. A. and Strack, O. D. L., Mechanics of Granular 38. Kiyama, H. and Fujimara, H., Proc. JSCE, 1983, 333, 137–146.
Materials: New Model and Constitutive Relations (eds Jenkins, 39. Thornton, C. and Randall, C. W., Micromechanics of Granular
J. T. and Satake, M.), Elsevier, Amsterdam, 1983, pp. 137–149. Materials (eds Satake, M. and Jenkins, J. T.), 1988, pp. 133–
22. Bathurst, R. J., Ph D Dissertation, Queen’s University, King- 142.
ston, Canada, 1985. 40. Uemura, D. and Hakuno, M., Struct. Eng./Earthquake Eng.,
23. Rothenburg, L. and Bathurst, R. J., Geotechnique, 1989, 39, 4, 1987, 4, 1, 155s–164s.
601–614. 41. Casaverde, M. L., Iwashita, K., Tarumi, Y. and Hakuno, M.,
24. Ting, J. M. and Corkum, B. T., 6th Int. Conf. on Numerical Struct. Eng./Earthquake Eng., 1989, 6, 1, 131s–140s.
Methods in Geomechanics, Innsbruck, Austria, 1988. 42. Cundall, P. A., Drescher, A. and Strack, O. D. L., Proc. IUTAM
25. Nimbkar, M. S., M.Sc. (Engg.) thesis, Indian Institute of Science, Conf. on Deformation and Failure of Granular Materials, Delft,
Bangalore, India, 1996, p. 128. Balkema, 1982, pp. 355–370.
26. Nagaraja Rao, P. V., M. Sc. (Engg.) thesis, Indian Institute of 43. Cundall, P. A. and Strack, O. D. L., Geotechnique, 1979, 29,
Science, Bangalore, India, 1998, p. 154. 47–65.
27. Mustoe, G. G., Henriksen, M. and Huttelmaier, H-P. (eds), Pro- 44. Thornton, C. and Barnes, D. J., Acta Mech., 64, 45–61.
ceedings of the 1st U.S. Conference on Discrete Element 45. Ohnishi, Y. and Kono, M., Proc. JSCE, 1986, 376, 231–239 (in
Method (DEM), Colorado School of Mines, Golden, CO, 1989. Japanese).
28. John R. Williams and Graham G. W. Mustoe (eds), Proceedings 46. Hart, R. D., Proc. of 7th International Congress on Rock
of the 2nd International Conference on Discrete Element Mecha-
Method (DEM), Dept of Civil and Environmental Engineering, nics, Aachen, Germany, 1991, pp. 245–261.
MIT, IESL publications, 1993. 47. Ting, J. M., Corkum, B. T., Cauffman, C. R. and Carlo Greco,
29. Williams, J. R., COMPLAS 1987, Barcelona, Spain, 6–10 April Report to University of Toronto, Department of Civil Engi-
1987, Pineridge Press, UK. neering, ISBN 0-7727–7090-5 Publication No. 87–03, 1987.
30. Cundall, P. A., in Micromechanics of Granular Materials (eds 48. Dasgupta, B., Dham, R. and Lorig, L. J., Proc. 8th Int. Conf.
Satake, M. and Jenkins, J. T.), Elsevier, Amsterdam, 1988, pp. on Rock Mechanics, Tokyo, Japan, 1995.
113–123. 49. Cundall, P. A. and Hart, R. D., Eng. Comput., 1992, 9, 101.
31. Cundall, P. A. and Strack, O. D. L., Report to NSF concerning 50. Mustoe, G. G. W., Williams, J. R. and Hocking, G., The Discrete
Grant Eng. 7620711, Dept of Civ. & Min. Engineering, Univ. of Element Method in Geotechnical Engineering, Elsevier, Bark-
Minnesota, USA, 1979. ing, Essex, UK, 1986.
32. De Josselin de Jong, G. and Veruijit, A., Cahiers Groupe Fran- 51. Sitharam, T. G. and Nimbkar, Indian Geotech. J., 1996, 26,
cais Rheol., 1969, 2, 73–86. 243–268.

886 CURRENT SCIENCE, VOL. 78, NO. 7, 10 APRIL 2000