Evaluation of porosity changes in a granular material undergoing particle

crushing using DEM

L.E. Vallejo & S. Lobo-Guerrero

Department of Civil and Environmental Engineering, University of Pittsburgh, Pittsburgh PA, USA

ABSTRACT: Granular materials forming part of natural slopes, embankments, subgrades of foundations
and pavement structures are subjected to both static and dynamic loads during their engineering lives. As a
result of these loads, particle crushing occurs. The present study uses the commercial code PFC2D based on
the Discrete Element method (DEM) to analyze how the porosity of a simulated granular material changes as
a result of particle fragmentation. Even though DEM does not normally consider particle breakage, the code
can be modified to allow particle crushing. This can be accomplished by replacing a particle that has failed in
tension with a combination of many particles of different sizes. DEM was used since it allowed to visualize
and to quantify the changes experienced by the granular material when undergoing granular fragmentation.
The simulated granular material was subjected to two different geotechnical tests: a uniaxial compression test
and a direct shear test. The results from the two simulated tests showed that granular crushing produced an
important change in the porosities of the samples. However, these porosity changes were not uniform throughout
the samples, being concentrated in certain regions. This was further studied by analyzing amplified snap shots
of these critical regions. Initially, the particles that crushed formed isolated dense zones, but as crushing evolved
these dense zones became interconnected. These results can be used to understand and explain the effect of
particle crushing on the hydraulic conductivity of a granular material such as a granular base in a pavement
structure. If this granular material reaches a similar state to one of those obtained at the end of the simulations
where the dense zones became interconnected, excess pore water pressure and permanent deformations could
be generated as a result of inadequate drainage.

1 INTRODUCTION 2 PREVIOUS WORK

Granular materials forming part of embankments,
foundations and pavement structures are subjected to
both static and dynamic loads. As a result total or par-
tial particle breakage occurs. Particle breakage causes
settlements and a reduction in hydraulic conductiv-
ity. Changes in the original properties of a granular
material affect the stability of engineering structures
supported on it. Therefore, granular crushing is a
detrimental phenomenon that deserves to be carefully
examined.
The main focus of this study is to understand and
produce a visualization of how the porosity of a simu-
lated granular material changes as a result of particle
fragmentation. For this purpose, a simulated granular
material was subjected to two different geotechnical
tests: a uniaxial compression test and a direct shear
test. Porosity changes due to crushing are important
for a granular material since they affect its drainage
capacity and deformation.
Previous research has shown that granular crushing is
influenced by grain angularity, grain size, uniformity
of gradation, particle strength, initial porosity, and
creep [Lee and Farhoomand, 1967], [Hagerty et al.,
1993], [Lade et al., 1996], [McDowell and Bolton,
1998]. The effects of these parameters were deter-
mined from triaxial or uniaxial compression tests.
Studying crushing of granular materials has some
how been limited by the machines needed to develop
the considerable loads that can lead to grain frag-
mentation. One way to sidestep this is to use stan-
dard geotechnical equipment with weak materials
[Mandl et al., 1977], [McDowell and Khan, 2003],
[Lobo-Guerrero and Vallejo, 2004]. However, the stan-
dard geotechnical equipment does not allow for a
continuous visualization of the evolution of crushing.
Computer simulations in the form of DEM present
the option to continuously visualize the evolution of
crushing in granular materials. Since the original DEM

573
Copyright 2005 Taylor & Francis Group plc, London, UK
developed by Cundall and Strack [1979] did not con- a) P2 P3
sider particle breakage, different solutions have been
proposed in order to overcome this constraint. One
solution to this problem is to treat each granular parti-
cle as a porous agglomerate built by bonding uniform
smaller particles [Jensen et al., 2001], [McDowell and
Harireche, 2002], [Cheng et al., 2003]. Another solu-
tion is to replace the particles that are fulfilling a
predefined failure criterion with an equivalent group
of smaller particles [Tsoungui et al., 1999]. This study P1 > P2 > P3 P1
uses the second solution and a new simplified ten-
sile failure criterion that can be easily implemented on
DEM in order to study the changes in porosity that b) P1 c)
a granular material experiences as a result of particle
crushing.

3 BASES OF THE COMPUTER SIMULATION t= 2P1/LD

The PFC2D program produced by Itasca was used

to visualize the evolution of crushing in a granu-
lar material subjected to a uniaxial compression test
and a direct shear test. In the PFC2D program, par-
ticles are idealized as disks that interact with each
other at their contacts. This interaction is mainly gov-
erned by three models: the stiffness model, the slip
model, and the bonding model [Itasca, Theory and
Background, 2002]. Only the first two models were
used in this simulation. Since the PFC2D program does
not allow particle breakage, a new subroutine using
the FISH language [Itasca, FISH in PFC, 2002] was
programmed in order to allow this to happen. The sim-
plified failure criterion adopted during the simulation
considers:
Only particles with a coordination number equal to
or smaller than 3 are able to be broken. Particles
break more easily when their coordination number
is low [Tsoungui et al., 1999], [Lade et al., 1996].
For those particles having a coordination number
smaller than or equal to 3, the real loading con-
figuration such as the one presented in Fig. 1(a) is
assumed to be equivalent to the one obtained in a
diametrical compression test such as the Brazilian
test, as shown in Fig. 1(b). By using this sim-
plification the induced tensile stress, t , can be
approximated with the expression presented on
Fig. 1(b), where P1 is the value of the highest con-
tact force acting on the particle, L is the thickness
of the disk (unit thickness for the simulated case),
and D is the diameter of the disk. The induced ten-
sile stress in Fig. 1(b) could be more severe than the
stress induced in Fig. 1(a), but the authors believe
that the implemented simplification is justified due
to the simplicity in the calculation of the induced
tensile stress.
The tensile strength of a particle having a radius
of 1mm is predefined as max1mm = 3 106 Pa.
P1
Figure 1. Idealization of the induced tensile stress and
arrangement of the produced fragments.
This value was arbitrarily chosen by the authors.
It represents the strength of the virtual material.
By changing this value, the necessary stresses to
produce crushing will also change. It is assumed
that the tensile strength of a particle with a radius
r, max (r), is related to max1mm according to the
following relationship (where r is expressed in mm):
In this way, particles with a radius greater than
1mm have a tensile strength smaller than max1mm ,
and particles with a radius smaller than 1mm
have a tensile strength greater than max1mm . Other
researchers have reported experimental results in
order to describe the change in tensile strength
as a function of size for different materials such
as quartz fibers and sand grains [Billam, 1971],
[Nakata et al., 2001]. They found that the experi-
mental results could be described using an equation
similar in form to Equation (1). As it is expected, the
values of the two constants in Equation (1) depend
on the analyzed material.
Every particle with a coordination number smaller
than or equal to 3 is allowed to break if t > max (r).
It has been found that the distribution of fragments
produced by a tensile failure depends on the nature
of the tested material [Nakata et al., 1999], [Takei
et al., 2001]. For the purpose of this simulation, it was
assumed that if a particle is fulfilling the previously
established failure criterion, it is allowed to break into
a group of 8 particles having 3 different sizes, as shown

574
Copyright 2005 Taylor & Francis Group plc, London, UK
 Original particles

First generation of
crushing
Second generation
of crushing

Third generation
of crushing

0.24

0.22

0.20
Porosity, n

0.18

0.16

0.14 lower part n(1)

upper part n(2)
0.12

0.10
1E+03 1E+04 1E+05 1E+06
Applied force (N)

Figure 2. Porosity and crushing evolution during the uniaxial compression test.

575
Copyright 2005 Taylor & Francis Group plc, London, UK
in Fig. 1(c). This distribution of sizes resembles the one
observed by the authors when crushing gravels in point
load tests.
In order to implement the failure criterion, a new
subroutine was programmed using the FISH language.
This subroutine automatically checks if a given particle
is fulfilling the failure criterion. If it does, the simu-
lation stops, and the particle that broke is deleted and
replaced by the set of particles shown in Fig. 1(c).
The subroutine does not restrict smaller particles from
continuing to break. Thus, different generations of
crushing can coexist inside the sample. The different
generations of crushing can be distinguishing by the
color of the particles as shown later.
4 UNIAXIAL COMPRESION TEST
4.1 Preparation of the sample
The first step was the construction of the rectangu-
lar container. This had a width of 5 cm and a height
of 10 cm. It was composed by 4 different walls hav-
ing all a friction coefficient of 0.7. The normal and
shear stiffness of these walls was set to 1 109 N/m.
After the construction of the container, 120 circular
disks with unit thickness and a uniform radius of 3 mm
were generated inside. Their positions were randomly
chosen by the program, having the limitation of no
overlaps between particles. The density of the particles
was set to 2500 kg/m3 , their normal and shear stfiness
were set to 1 108 N/m, and their friction coefficient
was set to 0.7.
After the particles were generated, they were
allowed to settle and reorganize under the action of
a normal gravity field (9.8 m/s2 ). In this way, the
sample had a loose structure at the beginning of the
test. After this, the upper wall of the container started
to move with a constant velocity of 1 107 m/step.
After some displacement, it made contact with the par-
ticles initiating the uniaxial compression conditions.
The velocity of the upper wall was constant during the
test. This slow vertical velocity was chosen in order to
guarantee that particles did not crush as a result of a
rapid movement of the wall.
4.2 Evolution of the crushing and porosity changes
After the loading stage started, the particles reorga-
nized and started to crush in response to the applied
load. Fig. 2 shows the applied force vs porosity curves
for the upper and lower regions of the sample. Also,
some snap shots of the sample are shown at differ-
ent values of applied load. The first two particles that
crushed were located in the middle and upper regions
of the sample, and in general crushing concentrated in
the upper region during the first stages of the simula-
tion. It can be seen that almost every particle from
576
Copyright 2005 Taylor & Francis Group plc, London, UK
a) 7.5x104 N
b) 1x105 N
Figure 3. Details of the pore structure at different values of
vertical force.
the upper row was broken. After these first stages,
crushing started to spread in to the middle region
and ultimately into the lower region. This behavior is
explained by the fact that the upper particles had to
bear large loads and their coordination numbers were
low since they were closer to the walls.
The porosity curves present two different regions.
The first one is characterized by a small decrement
in porosity while the second one is associated with
an important decrement. In other words, small val-
ues of deformation are associated with the first region
and large values are associated with the second region.
The transition between these two regions corresponds
to an applied vertical force between 2 104 N and
3 104 N, and is related to the starting of the crushing.
In this way, the change in porosity in the first region can
be explained by the rearrangement of the original par-
ticles while the change in porosity in the second region
is the product of particle breakage and the produced
rearrangement of the original particles and fragments.
As shown in Fig. 2, the values of porosity in the
upper part of the sample were always smaller than the
values in the lower part. This is consistent with the fact
that most of the crushing and consequently the reduc-
tion in volume of voids took place in the upper part.
Engineering properties such as the hydraulic conduc-
tivity of a crushable granular material will be affected
differently depending on the analyzed region of the
sample.
Fig. 3. shows some amplified details from the sam-
ple at different values of applied vertical force. It
can be observed how the produced fragments reor-
ganized forming a dense structure. If several of these
dense zones interconnect with each other, the drainage
capacity of the granular material will be severely
 0.22
0.20
0.18
Porosity
0.16
0.14
0.12
0.10
0 1
Horizontal deformation (nm)
Figure 4. Porosity and crushing evolution during the direct shear test.
affected. If a granular material in the field such as
a granular base or a sub base forming part of a pave-
ment structure reaches the condition shown in those
snapshots, excess pore water pressure and perma-
nent deformations could be generated as a result of
inadequate drainage.
5 DIRECT SHEAR TEST
5.1 Preparation of the sample
Porosity changes due to granular crushing in a direct
shear test were also evaluated using DEM. The same
virtual material used on the uniaxial test was used in
this test. The first step was the construction of a virtual
shear box. The top part of the box was made of two
vertical walls that were kept separated by a distance
equal to 6 cm and a movable horizontal wall. The lower
part of the box was fixed. The box had a width of 6 cm
and a height of 2.86 cm before shearing took place. The
shear and normal stiffnesses of the walls forming the
box were set to 1 109 N/m. The walls were assumed
to be frictionless.
Copyright 2005 Taylor & Francis Group plc, London, UK
Original particles
First generation of
crushing
Second generation
of crushing
porosity (left)
Third generation
porosity (right) of crushing
average
2 3 4 5 6
Only 48 disks were used in this simulation. The
disks were generated inside an area that measured
6 cm in width and about 3.5 cm in height. This area
was slightly bigger than the final dimensions of the
shear box before shearing (6 cm in width by 2.86 cm
in height).The positions of the particles were randomly
chosen by the program. By using a gravity field
(g = 9.8 m/s2 ) the particles were allowed to settle and
rearrange. After this, the horizontal wall of the shear
box was moved downward until it made contact with
the particles and then a normal load equal to 3 104 N
was applied to the sample. After this, shearing was
induced in the sample. The shearing stage started by
moving the upper section of the shear box with a con-
stant velocity of 5 105 mm/step, and it was ended
when its horizontal displacement was equal to 5 mm.
This slow horizontal velocity was chosen in order to
guarantee that particles did not crush as a result of a
rapid movement of the walls.
5.2 Evolution of the crushing and porosity changes
Fig. 4 shows the porosities of different regions of the
sample at different values of horizontal deformation.
577
Also, Fig. 4 shows snap shots of the sample during
the shearing. The first particles that crushed were
in contact with the horizontal walls of the box. This
behavior is explained by the fact that the particles in
contact with the walls had to bear large loads and their
coordination numbers were low. The other particles
that were crushed during the simulation were located
in the middle of the shearing zone.
Fig. 4 shows that the porosities of the left part and
the right part of the sample followed a similar trend
during the simulation. The porosity of the left part of
the sample was always higher than the porosity of the
right side because of the initial arrangement of the par-
ticles. It can be observed that the overall porosity of
the sample did not significantly change during shear. It
seems that contractive changes in volume due to crush-
ing were offset by dilatancy during the shear process.
6 CONCLUSIONS
The porosity changes that a granular material experi-
ences as a result of particle crushing during uniaxial
compression and direct shear conditions were analyzed
using DEM. In order to allow particle breakage, a sim-
plified failure criterion was implemented in a FISH
subroutine that interacted with the course of the simu-
lations. For the uniaxial compression test, it was found
that particle crushing and the associated particle rear-
rangement caused a significant reduction on porosity.
Also, crushing did not uniformly develop through out
the sample, being concentrated in certain regions. The
obtained results can be used to understand and explain
the effect of particle crushing on the hydraulic conduc-
tivity of a granular material such as a granular base
forming part in a pavement structure. If this gran-
ular material reaches a similar state to one of those
obtained at the end of the simulation where the dense
zones became interconnected, excess pore water pres-
sure and permanent deformations could be generated
as a result of inadequate drainage.
Contrary to the results obtained on the uniaxial
compression test, the results of the direct shear test
showed that the average porosity of the sample was
not affected by the particle crushing. The contrac-
tive changes in volume due to crushing were offset
by dilatancy during the shear process.
REFERENCES
Billam, J. (1971). Some aspects of the behaviour of granular
materials at high pressures, Stress-strain Behaviour of
Soils: Proceedings of the Roscoe Memorial Symposium,
Cambridge University, 6980.
578
Copyright 2005 Taylor & Francis Group plc, London, UK
Cheng, Y.P., Nakata, Y., Bolton, M.D. (2003). Discrete ele-
ment simulation of crushable soil, Geotechnique, 53(7),
633641.
Cundall, P.A., Strack, O.D.L. (1979). A discrete numeri-
cal model for granular assemblies, Geotechnique, 29(1),
4765.
Hagerty, M.M., Hite, D.R., Ulrich, C.R., Hagerty, D.J. (1993).
One dimensional high pressure compression of granu-
lar media, ASCEs Journal of Geotechnical Engineering,
119(1), 118.
Itasca Consulting Group, Inc. (2002). PFC 2D (Particle Flow
Code in Two Dimensions) version 3.0: Theory and Back-
ground, and FISH in PFC.
Jensen, R.P., Plesha, M.E., Edil, T.B., Bosscher, P.J.,
Kahla, N.B. (2001). DEM simulation of particle damage
in granular media structure interfaces, The Interna-
tional Journal of Geomechanics, 1(1), 2139.
Lade, P.V.,Yamamuro, J.A., Bopp, P.A. (1996). Significance
of particle crushing in granular materials, ASCEJournal
of Geotechnical Engineering, 122(4), 309316.
Lee, K.L., Farhoomand, I. (1967). Compressibility and
crushing of granular soil in anisotropic triaxial compres-
sion, Canadian Geotechnical Journal, 4(1), 6886.
Lobo-Guerrero, S., Vallejo, L.E. (2004). Modeling of Mate-
rial Crushing in Granular Road Bases, Proceedings
for the UNBAR VI Conference, Pavements Unbound,
University of Nottingham, 3341.
Mandl, G., Jong, L.N.J., Maltha, A. (1977). Shear zones in
granular material, Rock Mechanics, 9, 95144.
McDowell, G.R., Bolton, M.D. (1998). On the microme-
chanics of crushable aggregates, Geotechnique, 48(5),
667679.
McDowell, G.R., Harireche, O. (2002). Discrete element
modelling of yielding and normal compression of sand,
Geotechnique, 52(4), 299304.
McDowell, G.R., Khan, J.J. (2003). Creep of granular
materials, Granular Matter, 5, 115120.
Nakata, Y., Hyde, A.F.L., Hyodo, M., Murata, H. (1999). A
probabilistic approach to sand particle crushing in the
triaxial test, Geotechnique, 49(5), 567583.
Nakata, Y., Kato, Y., Hyodo, M., Hyde, A.F.L., Murata, H.
(2001). One-dimensional compression behaviour of uni-
formly graded sand related to single particle crushing
strength, Soils and Foundations, 41(2), 3951.
Takei, M., Kusakabe, O., Hayashi, T. (2001). Time depen-
dent behavior of crushable materials in one-dimensional
compression tests, Soils and Foundations, 41(1), 97121.
Tsoungui, O., Vallet, D., Charmet, J.C. (1999). Numeri-
cal model of crushing of grains inside two-dimensional
granular materials, Powder Technology, 105, 190198.