DEM Simulation of Comminution: Fragmentation and Size Distribution

Yucang Wang and Fernando Alonso‐Marroquín

Citation: AIP Conference Proceedings 1145, 863 (2009); doi: 10.1063/1.3180066

View online: http://dx.doi.org/10.1063/1.3180066
View Table of Contents: http://scitation.aip.org/content/aip/proceeding/aipcp/1145?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

DEM study on the interaction between wet cohesive granular materials and tools
AIP Conf. Proc. 1542, 947 (2013); 10.1063/1.4812089

Numerical simulation study of quasistatic grinding process on a model granular material

AIP Conf. Proc. 1542, 927 (2013); 10.1063/1.4812084

A DEM—FEM two scale approach of the behaviour of granular materials

AIP Conf. Proc. 1145, 443 (2009); 10.1063/1.3179957

DEM investigation of particle size distribution effect on direct shear behaviour of granular agglomerates
AIP Conf. Proc. 1145, 401 (2009); 10.1063/1.3179946

SIMULATION OF PARTICLE SIZE EFFECT ON DYNAMIC PROPERTIES AND FRACTURE OF PTFE‐W‐Al

COMPOSITES
AIP Conf. Proc. 955, 785 (2007); 10.1063/1.2833241

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.255.6.125 On: Tue, 11 Aug 2015 08:22:40
 DEM Simulation of Comminution: Fragmentation and Size
Distribution
Yucang Wang+ and Fernando Alonso-Marroquín*
*
ESSCC, the University of Queensland, Brisbane QLD 4067, Australia
+
MoSCoS, School of Mathematics and Physics, the University of Queensland, Brisbane QLD 4067 Australia

Abstract. ESyS_Particle, the 3-D Discrete Element Model, is used to model the fracture process and the size
distribution of rock fragmentation under two different loading conditions: 1) Slow uni-axial compression of a ball
aggregate which consists of small bonded particles; 2) Impact of a ball consisting of bonded particles to a stiff wall. We
reproduce the realistic fracture patterns, such as meridional cracks under slow loading and secondary cracks under fast
loading. We observe a transition from damage to fragmentation as impact velocity increases. The latter one is
characterized by generation of fines and a power law distribution of fragment sizes.
Keywords: Discrete Element model, fragmentation, size distribution.
PACS: 45.70

INTRODUCTION the realistic fracture patterns and fragment size

distribution.
Rock fracture and fragmentation play very
importance roles in mining industry. In process of ESYS_PARTICLE
comminution, an important issue is to determine the
relation between the input energy and the fragment ESyS Particle is a 3-D parallel DEM software
size distribution. Laboratory tests are essential to developed in ESSCC, the University of Queensland. It
investigate the fragmentation patterns and the energy is suitable to model brittle rock fractures and
budget associated to these processes. However, in earthquake processm[1]. There are several major
experiments it is difficult to quantify how the input differences between the ESyS_Particle and the other
energy is transformed in dissipation, acoustic waves existing DEMs. First, in the ESyS_Particle, the unit
and surface energy. Numerical models, such as quaternion q = q0 + q1i + q2j + q3k is used to describe
Discrete Element Method (DEM), do not have this
limitation because it is easy to access to the explicitly the orientations of particles. The quaternion
micromechanical data associated to fracture. With the represents a one-step rotation around the vector
recent advances of large scale simulations, numerical q1i + q2j + q3k with a rotation angle of 2arccos(q 0 ).
simulations with large number of particles can be At each time step the dynamic Euler equations and
achieved, (up to a million of particles in a single- quaternion equations are integrated such that the
processor, and billions of particles in supercomputers). orientation of each particle is uniquely known [2, 3].
Therefore DEM simulations of aggregates of bonded The second important difference is that a full set of
particles represent an attractive alternative to model interactions are included. There are three kinds of
fracture processes, as they provide detailed interactions in ESyS_Particle: bonded interaction,
information on breaking of bonds, energy budget and solely normal repulsive interaction and cohesionless
size distribution of the fragments that are not easily frictional interaction [3]. A special emphasis is on the
accessible in experiments. bonded interaction in which all three interactions
As a first step to understand fragmentation (normal, shearing forces and bending moment) in 2D
phenomenon, we use our Discrete Element software, and six (normal, shearing forces, bending and twisting
ESyS_Particle, to model fracture of a sphere sample moment) in 3D are transmitted between each bonded
under two different loading conditions. We reproduce particle pair [2, 3].

CPl 145, Powders and Grains 2009, Proceedings of the 6 International Conference on
Micromechanics of Granular Media edited by M. Nakagawa and S. Luding
© 2009 American Institute of Physics 978-0-7354-0682-7/09/S25.00

863
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.255.6.125 On: Tue, 11 Aug 2015 08:22:40
 The third one is the way of updating forces and particles are bonded, they belong to the same
torques caused by the relative movements between two fragment. If another particle is bonded to any of the
particles. In other existing DEMs, the incremental particles in the group, it joins the group. This process
method is used to update the interactions between repeats until a particle is not bonded to any of the
particles. Instead of the incremental method, particles in the existing groups, then it starts a new
ESyS_Particle uses the Finite Deformation Method, in group. If two particles from two different fragments
which the total relative (translational and rotational) are found to be linked, then the two fragments are
displacements are calculated at each time step [5]. combined into one. In this way all particles can be
Using quaternion algebra, we proved that [3] an grouped to corresponding groups. By calculating the
arbitrary rotation between two rigid bodies or two volume of each fragment, the size distribution can be
coordinate systems cannot be decomposed into three obtained.
order independent rotations around three orthogonal
axes. However it can be decomposed into two Fragmentation size distribution under
rotations, which correspond to the relative twisting and
slow uni-axial compression
bending between two bodies in our model. The two
rotations are sequence-independent. The beauty of this
Figure 2 shows fracture patterns of a sphere
sequence-independent decomposition is that it respects
consisting of 29117 bonded particles under a slow uni-
the physical law and it is guaranteed that forces and
axial compression driven by two rigid walls. The
torques decided by such a two-step rotation are unique.
colors represent horizontal displacements, and
Numerical results show that when dealing with finite
discontinuities in colors show the formation of
rotations of particles, the incremental method is not as
fractures. Close to two contact zones with the rigid
stable and accurate as the method used in our model
walls, two circular, flatten areas are observed. Detailed
[3]. We also use a different criterion to judge the
studies show that they are two cone shaped fragile
breakage of bonds [2] and study how the spring
regions, formed mainly by shear fractures. Four
stiffnesses should be chosen [4]. Realistic fracture
meridional cracks are clearly seen, which are caused
patterns are reproduced using the ESyS_Particle [5].
by tensile fractures. This kind of pattern is also
Figure 1 shows the 3D fracture pattern of a brittle
observed in laboratory tests [6].
rock-like material under uni-axial compression.
Figure 3 shows the fragmentation size distribution
of Figure 2 at four different loading stages. It plots the
number of fragments within each volume size bin. As
time increases, the size of the largest fragments
decreases and percentages of medium fragments
increase, indicating continuing fracture. We do not
observe a good power law distribution since there are
relatively more large and small pieces. In some
literatures the percentage of particles bigger (or
smaller) than certain volume is plotted against the
volume sizes [7], which can be obtained by integration
from Figure 3.

FIGURE 1. Fracture pattern of a 3-D brittle rock-like

material under uni-axial compression.

FRACTURE PATTERNS AND SIZE

DISTRIBUTION
In ESyS_Particle, the breakage of a bond is an FIGURE 2. Fracture pattern of a sphere under slow loading,
explicit representation of a small fracture event. Under side view (left), top view (right). Colors represent
loading some bonds begin to break, while others are displacement, so that discontinuities of the color show the
still linked. At certain time intervals we output fractures.
information of the particle pairs which are still bonded.
An algorithm is developed to identify which particles
belong to a group (fragment). The idea is that if two

864
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.255.6.125 On: Tue, 11 Aug 2015 08:22:40
 1 o * V = 7.5
^ • V = 10
* Time step = 200000
* Time step = 280000 "V3 . • V = 20
•Time step = 360000
\ ^VV ^ o V = 30
0.1
0.1 oTime step = 440000
""0, ' '.TJ n
s 1 *
*
0.01 0.01 -Q ' . t--' s.

V '*••.
1 E-3 •cr t 1E-3
\
• 4 - 4 - 4

1E-4 b
1 E-4

10 100 1000 10000 100000 10 100 1000 10000 100000

Fragment volume Fragment volume

FIGURE 3. Fragment size distribution of Figure 2. FIGURE 5. Fragment size distribution of Figure 4.

Generally it is observed that higher impacting

Size distribution under fast loading: velocity means more fragments, higher percentage of
impact fracture of a ball to a rigid wall small pieces, lower percentage of large pieces and
reduced sizes of the largest pieces, indicating the
Figure 4 shows four snapshots of fracture patterns important role of input energy.
of a ball after impacting a rigid wall at different
velocities. When the velocity is small (v = 7.5 units),
only slight damage occurs with several meridional
cracks, and the ball mainly keeps its integrity. But with SUMMARY AND DISCUSSIONS
increase of velocities, the specimen breaks into more,
smaller pieces, and latitudinal secondary cracks We presented simulation results of brittle fracture
develop. of rock-like materials under two types of loading
conditions: the fracture of a sphere under slow uni-
axial compression and dynamic fracture of a sphere
colliding with a rigid wall. The realistic fracture
patterns, such as meridional cracks and secondary
cracks, are well reproduced.
A numerical algorithm was developed to calculate
the size of fragments. We do not obtain power law size
distribution in case of slow uni-axial loading, but good
power law distributions are observed for nearly four
orders of magnitude in volume in case of impact.
The origin of the power laws distributions of size is
still not clear. Previous studies based on percolation
clusters of broken bonds were successful to explain the
power laws, but they failed in the prediction of the
exponents [8]. The larger deviations for large
fragments were also reported by other studies [6].
Some even suggested that coarse and fine fragments
obey two different power law distributions [6]. Based
on numerical experiments [7] it is suggested that the
FIGURE 4. Fracture pattern of a sphere under impact,
V=7.5 units (top left), 10 (top right), 20 (bottom left), and 30
coarser fragments result from cracks propagated from
(bottom right). the loading/impact zone, whereas the finer fragments
result from secondary fragments created perpendicular
The fragmentation size distribution can be found in to the fractured surfaces.
Figure 5. Good power law distributions are observed Others questions need to be answered in the future
for nearly four orders of magnitude in volume. studies: Are the deviations from the power law for
large and small fragments caused partly by the size

865
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.255.6.125 On: Tue, 11 Aug 2015 08:22:40
 effects, or is it caused by different physical
mechanisms? How do initial conditions, i.e. input of
the contact parameters, the geometrical packing of the
sample, initial size distribution of particles, effect the
calculated fragmentation size distribution? Besides,
the studies in this paper are still rudimentary and
qualitative. The model is still not an accurate
representation of a rock sample yet since some
parameters, such as the size, porosity, grain size and
macroscopic elasticity etc, are not strictly chosen
according to a specific type of rock. This is why in
Figure 4 and 5 there is no detailed units for velocities
and volumes. Also there are no pre-existed flaws in the
rock at this stage. Accurate calibration and large scale
simulations are required to investigate these problems
in the future.

ACKNOWLEDGMENTS
We thank for the support of the ACcESS and
AuScope project. FAM is supported by the Australian
Research Council (project number DP0772409) and
the University of Queensland Early Carrier Grant.

REFERENCES
1. P. Mora and D. Place, International Journal Modern
Physics C4, 1059-1074 (1993).
2. Y. C. Wang, S. Abe, S., Latham and P. Mora, Pure and
Applied Geophysics 163, 1769-1785 (2006).
3. Y. C. Wang, Acta Geotechnica, in print (2008).
4. Y. C. Wang and P. Mora, Journal of Mechanics and
Physics of Solids 56, 3459-3474 (2008) .
5. Y. C. Wang, F. Alonso-Marroquin and P. Mora,
Granular matter, accepted (2009).
6. S. Z. Wu, K. T. Chau and T. X. Yu, Powder Technology
143, 41-55 (2004).
7. C. Thornton, M. T. Ciomocos and M. J. Adams, Powder
Technology 140, 258-267 (2004).
8. F. Kun, and H. J. Herrmann, Phys. Rev. E 59, 2623-2632
(1999).

866
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.255.6.125 On: Tue, 11 Aug 2015 08:22:40