Advanced Powder Technology xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Advanced Powder Technology

journal homepage: www.elsevier.com/locate/apt

Original Research Paper

A discrete element method-based approach to predict the breakage of

coal
Varun Gupta a,⇑, Xin Sun a,1, Wei Xu a,2, Hamid Sarv b, Hamid Farzan b
a
Advanced Computing, Mathematics and Data Division, Pacific Northwest National Laboratory, Richland, WA 99354, USA
b
The Babcock & Wilcox Company, Barberton, OH 44203, USA

a r t i c l e i n f o a b s t r a c t

Article history: Pulverization is an essential pre-combustion technique employed for solid fuels, such as coal, to reduce
Received 16 March 2017 particle sizes. Smaller particles ensure rapid and complete combustion, leading to low carbon emissions.
Received in revised form 15 July 2017 Traditionally, the resulting particle size distributions from pulverizers have been informed by empirical
Accepted 24 July 2017
or semi-empirical approaches that rely on extensive data gathered over several decades during opera-
Available online xxxx
tions or experiments, with limited predictive capabilities for new coals and processes. This work presents
a Discrete Element Method (DEM)-based computational approach to model coal particle breakage with
Keywords:
experimentally characterized coal physical properties. The effect of select operating parameters on the
Discrete element method
Size distribution
breakage behavior of coal particles is also examined.
Coal breakage Ó 2017 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder
Cohesion Technology Japan. All rights reserved.
Brazilian test

1. Introduction
The comminution process has wide applications in several
industries, including mining, power, cement, paint and pharma-
ceuticals. The design and operating conditions for comminution
devices traditionally have been informed by empirical or semi-
empirical approaches derived from the vast data gathered over
the years during operations or experiments. Most of these process
models are based on the Population Balance Method (PBM)
[5,6,19]. Some recent applications of PBM approaches in milling
processes are given in [9,32,43]. A comprehensive review on the
fundamental bases of the PBM and its use for varied applications
is provided in [33].
One of the major applications of comminution in the power
industry involves the pulverization of coal to attain rapid and com-
plete combustion. In a coal-fired power plant, raw coal is pulver-
ized and partially dried in a mill prior to combustion. The
pulverized coal is then carried to the combustion chamber by flu-
idizing the coal particles using pre-heated air. The size distribution
and shape of the coal particles have a significant impact on the effi-
ciency of the mill and combustion process (also noted in [54]).
⇑ Corresponding author.
E-mail address: varun.gupta@pnnl.gov (V. Gupta).
1
Current address: Oak Ridge National Laboratory, USA.
2
Current address: PPG Industries Inc., USA.
Computational models for coal combustion also rely on the input
coal particles distribution after the pulverization process [17,36].
The design of these mills traditionally has been based on the
process models [35,42], which rely on the breakage distribution
of coal derived from Population Balance approaches. The Popula-
tion Balance Models are typically developed for particular devices
operating in a specific scenario. However, the breakage behavior
depends on the intrinsic physical properties of the coal. The PBM
derived for one particular coal might not be suitable for a different
coal type. Also, if there are changes in the pulverizer design or
operating conditions, these phenomenological models will require
further tuning. With the ever-increasing need for accuracy and
sophistication in the predictive capability of comminution models,
the importance of mechanistically driven computational models is
growing.
The Discrete Element Method (DEM) [16] provides an alterna-
tive mechanism-based modeling capability in understanding the
breakage behavior of granular material. It has been extensively
used in several applications to model fragmentation and breakage
of granular materials. A comprehensive review regarding the con-
tributions of DEM for modeling comminution process is provided
in [52]. Today, there are several commercial and open-source
DEM software programs available. A summary of the capabilities
of some of these softwares, with an emphasis on modeling tech-
niques for fracturing processes, is provided in [22]. Many research-
ers [18,21,41,48] use hybrid techniques where DEM is used in
conjunction with PBM. Such techniques typically consist of either

http://dx.doi.org/10.1016/j.apt.2017.07.019
0921-8831/Ó 2017 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.

Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
2 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx
Nomenclature
Symbols
Fn normal interaction force between particles
Kn normal interaction stiffness
kn individual particle stiffness
E particle-scale modulus of elasticity
ueq equilibrium distance between particles
ks shear interaction tangent stiffness
m particle-scale Poisson’s ratio
using DEM simulations to inform certain components in the PBM
or using these two methods at different length scales with the
transfer of information across scales.
Broadly speaking, two approaches are typically employed in
modeling the dynamic breakage of granular samples using DEM.
The first approach is based on the Bonded Particle Model (BPM)
[30], where particles are bonded together using some mechanical
constitutive laws. The breakage in this approach is controlled by
the strength and mechanical properties of each bond. Examples
of similar approaches from the literature include [20,39,45–
47,51]. One of the biggest challenges of this approach is that it
leads to a large number of particles in the simulation, making it
especially difficult to use for large-scale applications. An alterna-
tive approach is the replacement strategy [11–13,15]. This strategy
involves a rule that is used to determine the expected conditions
for particle fracture and then replaces the parent particle with a
closely packed assembly of daughter particles. The choice of pack-
ing and sizes of these daughter particles usually is rather arbitrary.
The packing must be carefully chosen so as to not induce any arti-
ficial stresses, which leads to a loss of mass in the system. This
approach also requires particle crushing strength, which usually
depends on the size of particles with high variability. On the exper-
imental side, only limited size-dependent strength data are avail-
able in the literature. Bieniawski [8] reported compressive
strength of coal specimen for sizes ranging from 0.75 to 60 in. To
the best of the authors’ knowledge, strength data for much smaller
particles (few hundred microns to mm scale) are not available in
literature and are difficult to obtain due to a lack of appropriate
test procedures for such length scales. It is also common to assume
that particle strength follows some statistical law, e.g., Weibull dis-
tribution [53], with respect to particle size. Several researchers
have assumed these distributions or other empirical laws to char-
acterize the particle strength in their work [4,24–26,28,34,40]. The
use of Weibull distribution for characterizing strength of brittle
material as a function of particle size remains debatable [7].
In this study, the first approach based on the BPM is used to
simulate coal particle breakage because of its algorithmic simplic-
ity and the availability of experimental test data for calibrating
bond strength. The particle breakage is modeled by constructing
the macro-particles (or agglomerates) from the smaller sub-
particles (or simply particles). The smaller particles are bonded
together using a constitutive law, and the bond breakage between
these particles constitutes the micro-fractures in the agglomerates.
Several of these micro-fractures coalesce, eventually leading to
breakage of the agglomerate into smaller fragments. The main
objective of this work is to provide an integrated experimental
and DEM-based computational framework to model the coal
breakage and to examine effect of different operating conditions
on the breakage rate. The application of this framework is demon-
strated on a surrogate small-scale model with periodic boundary
conditions. In this work, the particle-based open-source code Yade
Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
kr rotational stiffness
rc normal cohesion parameter
/ internal friction angle
sc shear cohesion parameter
cint interaction ratio
DT cr critical time step size
q particle-scale density
[50] has been used for the numerical simulations. The develop-
ment of an accurate DEM to study the breakage of behavior of coal
is not the focus of this work.
Note that the actual coal grinding process in mills also results in
much finer particles or a powder-like substance, because of abra-
sion and chipping. The presence of such powdery material can have
significant impact on the breakage behavior. The powder acts as a
lubricant, leading to reduced intergranular force transmissions.
Cleary [14] proposed a multiscale method to incorporate the effect
of fine particles or interstitial powder in the DEM simulations. Any
bulk motion or the powder lubricating effect is not accounted for in
the current model.
The following section provides details on the constitutive laws
and the DEM model used in this work. Section 3 discusses various
parameters and material properties used to construct the model. It
also covers the procedure for calibrating the bond strength
between particles using the Brazilian Test. The model’s capability
in modeling breakage is then demonstrated on a small numerical
sample of coal in Section 4. Because of the small model size and
inherent assumptions in the boundary conditions, the model is
currently not intended to provide quantitative validation for the
coal particle breakage obtained through actual grindability opera-
tions. However, it provides a framework which can be expanded
and utilized to study the effect of certain operating parameters
on the mill efficiency. The main conclusions from this work are
summarized in Section 5.
2. Model description
In this study, the soft-sphere approach of DEM [16] is adopted
to capture the motion and interaction of particles. It consists of
modeling an assembly of rigid, locally deformable particles that
interact with each other by contact forces. Quite contrary to the
common definition of ‘‘particles” in the field of mechanics as a
point in the space, here it means a body occupying a finite volume
of space. Only spherical particles are considered in this work,
although other shapes are also possible. The classic DEM used in
this work is based on an explicit numerical scheme. The contact
forces and moments, governed by contact laws, are obtained at
each time step using relative displacements and rotations of the
particles. The updated positions of the particles are then obtained
by integrating Newton’s laws of motion.
In this study, the intact coal material is represented by a set of
spherical agglomerates. Each agglomerate is composed of rigid
spherical particles of radius 100 lm connected by cohesive bonds.
Fig. 1 shows the example of an agglomerate consisting of several
100 lm spherical particles. The dense packing of spherical particles
was obtained by isotropically compressing a much larger cubic
sample of randomly arranged non-overlapping particles using rigid
walls, to a resulting hydrostatic stress state of very small magni-
tude. This packing is then cropped to the desired spherical shape
 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx
Fig. 1. A spherical agglomerate composed of 100 lm particles connected by
cohesive bonds.
of agglomerate. In the context of DEM, each particle is a discrete
element. Although, a much smaller particle size would be ideal, this
smallest particle size is chosen based on the computational feasi-
bility to limit the problem size so that a solution can be obtained
within reasonable computational times. Also, a wide size distribu-
tion of particles within an agglomerate will lead to increased pack-
ing density. However, for simplicity, a single particle size is chosen
to form the agglomerate in this work.
Two types of contact laws are employed in the present model.
For the interactions of particles from within an agglomerate, a
cohesive-frictional contact law is used. The intergranular lock
between the coal particles is approximated by a notional cohesive
bond. The interaction forces are defined using elastic stiffnesses
corresponding to normal force, shear force, and the moment
between the discrete elements. The normal interaction force
between the two particles i and j is given by
F ijn ¼ K ijn ðu  ueq Þ;
where u is the distance between the centers of the two particles and
ueq is the pre-defined constant equilibrium distance between the
centers of two interacting particles at which the normal interaction
force is zero. K ijn is the normal interaction stiffness between the two
particles i and j, defined as the harmonic mean of individual particle
stiffnesses, kn ¼ E r; E being the particle-scale modulus of elasticity
and r is the radius of the spherical particle.
i j
2kn kn
K ijn ¼ i j
kn þ kn
This linear relationship for the interaction force with symmetry
in tension and compression is considered for simplicity. Other
more sophisticated models for contact forces, some of which are
outlined in [23], could also be used. The shear force between the
two particles is defined in an incremental form
ij
DF ijs ¼ ks Dus ;
where Dus is the relative incremental tangential displacement
ij
between the two particles. ks is the shear interaction tangent stiff-
ness defined similar to the normal stiffness in Eq. (2),
i j
ij 2ks ks
ks ¼ i j
ks þ ks
Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
3
with ks ¼ E r m, and m being the particle-scale Poisson’s ratio. The
interaction law also accounts for the moment that provides the
rotational resistance of the particles. Like shear force, it is computed
in an incremental form as
ij
DM ij ¼ kr Dh; ð5Þ
where h is the relative incremental rotation between the two parti-
ij
cles. kr is the rotational stiffness defined proportional to the shear
stiffness as a r i r j ks , where a is a dimensionless factor. An upper
ij
bound on the moment is introduced as the elastic limit and defined
as a function of normal interaction force as
Mijlimit ¼ gF ijn minðri ; r j Þ; ð6Þ
where g is a dimensionless factor.
The fracturing of the coal particle agglomerates is captured by
the rupture of bonds. These bonds fail when either the tensile nor-
mal force (Eq. (1)) or the shear force (Eq. (3)) exceed their respec-
tive strengths. Due to asymmetry in loading and particle packing, it
is assumed that before the compressive stresses in any of the indi-
vidual particles reach their compressive capacity, the bonds of that
particle with the surrounding particles in the agglomerate would
rupture in either tension or shear. The maximum acceptable nor-
mal force in tension scales with the square of particle dimension
and is given by

2
jF ijn j 6 rc minðri ; r j Þ ; ð7Þ
where rc is the normal cohesion parameter with the dimensions of
stress. The maximum acceptable shear force is given by a combina-
tion of the cohesive and frictional components

2
F ijs 6 F ijn tanð/Þ þ sc minðri ; r j Þ ; ð8Þ
where / is the internal friction angle and sc is the shear cohesion
parameter. This failure in shear is analogous to Mohr-Coulomb fail-
ure criterion. The sign convention is such that the compressive nor-
mal force is positive, whereas normal force in tension is negative. As
the cohesive bond breaks, the shear strength drops to the purely
frictional component. This similar cohesive-frictional contact law
also has been used in [10,37,38].
ð1Þ
Interactions between particles from two different agglomerates
and between a particle and the domain walls are modeled using a
frictional constitutive law given simply by Eqs. (1) and (3). The lim-
iting value for shear force in these interactions is given by the fric-
tional component, F ijn tanð/Þ, in Eq. (8). For interactions between
two different types of materials, the lower of the two internal fric-
tion angles is used in the constitutive law.
In the classical DEM approach, only the strictly contacting par-
ticles are considered for interactions. Using a parameter called
ð2Þ
ð3Þ
ð4Þ
Fig. 2. Image of the Brazilian Test conducted on a coal specimen.
4 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx

350

300
Average = 208 psi

Splitting Tensile Strength (psi)

250

200

150

100

50

0
1 2 3 4 5 6 7 8 9 10
Specimen Number

Fig. 3. Splitting tensile strength of 10 different coal specimens obtained through the Brazilian Test.

Fig. 4. The model used for Brazilian Test simulation: (a) Intact coal sample and (b) Broken sample (beyond peak load).

interaction ratio (cint ) in the model allows for increasing the range
Table 1
Model and material parameters for the coal sample used in this study.
of interaction distance between the particles. Therefore, cohesive
Parameter interaction can exist between the particles separated by a distance,
Young’s modulus (E) 4 GPa deq , given by
Poisson’s ratio (m) 0.25
Density (q) 1400 kg/m3

Friction angle (/) 30 deq 6 cint r i þ r j : ð9Þ
a (Eq. (5)) 2.0
g (Eq. (6)) 0.15 This feature provides the capability to have an initial denser
Interaction ratio (cint ) 1.1 network of cohesive bonds between the particles of a granular
Cohesive strength parameter (rc ¼ sc ) TBD sample. Notably, the interaction ratio is set back to its default value
of 1.0 after the first simulation time step, so new contacts only
occur between strictly touching particles. Also, the equilibrium dis-
tance ueq (Eq. (1)) between each set of particles is set equal to the
Table 2 distance between the particles in their initial configuration. This
Material properties for the loading plates.
ensures that the cohesive interaction bonds initially are force-free.
Parameter The constitutive laws described herein do not include velocity-
Young’s modulus (E0 ) 200 GPa based damping. Therefore, an artificial numerical damping is used
Poisson’s ratio (m0 ) 0.25 to dissipate the kinetic energy of particles in the simulations. To
Density (q0 ) 8000 kg/m3 approximate the quasi-static conditions, a high value of 0.8 is used
Friction angle (/0 ) 30
for the damping coefficient. This ensures that the energy released

Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx 5

40

35

30

Force (N) 25

20

15

10

0
0 0.5 1 1.5 2 2.5 3 3.5 4
-4
Displacement (m) ×10

Fig. 5. The typical force-displacement response obtained in the Brazilian Test simulation.

40

2
35 5 10 % per minute
4
5 10 % per minute

30

25
Force (N)

20

15

10

0
0 0.5 1 1.5 2 2.5 3
-4
Displacement (m) × 10

Fig. 6. Effect of different strain rates on the force-displacement response in the Brazilian Test simulation.

by breakage of a bond does not have significant influence on the
sample’s mechanical behavior and breakage of subsequent bonds.
To ensure stability of the explicit integration scheme used for
the DEM simulations, the time step size must be smaller than
the critical time step [49] given by
rffiffiffiffiffi
qi
DT cr ¼ min : ð10Þ
ri Ei
A value of 0:5DT cr has been used for the simulations in this
work.
3. Model parameters and material properties
In addition to the physical properties of coal, the described
model also requires several critical material parameters governing
particle separation. Due to the lack of suitable laboratory charac-
terization tests and experimental data, determination of all the
parameters through calibration is not possible. One of the most
important properties to determine breakage behavior of any gran-
ular sample is the cohesive bond strength: rc and sc in Eqs. (7) and
(8). In this work, these strength parameters are calibrated based on

Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
6 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx

600

500
Splitting Tensile Strength (psi)

400

300

200

100

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
2 7
Cohesive Strength Parameter (N/m ) × 10

Fig. 7. Calibration of cohesive strength parameter based on the simulated Brazilian Test.

Fig. 8. Normalized particle size distribution obtained at different stages in the HGI
test.
the experimentally determined tensile strength of the actual coal
sample. Similar to the approach in [30], the cohesive bond strength
parameters in normal and shear are assumed to be the same, i.e.,
rc ¼ sc . Through some rigorous statistical analysis, it has been
shown in [31] that shear strength is the most influential parameter
among different calibrated properties for the BPM model of rock
particles.
The Brazilian test [2] is the most commonly used laboratory test
to indirectly determine the tensile strength of rocks and concrete,
and it has been used in this work to determine the splitting tensile
strength of the coal sample. In this test, a disc-shaped specimen is
placed between two rigid platens (shown in Fig. 2) and compressed
Fig. 9. Numerical sample of coal placed between the two loading plates.
continuously at a constant rate until the sample failure occurs. The
specimen’s tensile strength obtained by this test is given by [1]
2P
Rt ¼ ; ð11Þ
ptd
where P is the compressive load at failure point and t and d are the
thickness and diameter of the specimen, respectively. The test was
conducted on 10 different carefully prepared disc-shaped speci-
mens of 200 diameter coal with a thickness-to-diameter ratio, t=d,
in the range of 0.3–0.6. Loading was applied at a constant strain rate
of 2.5% per minute. The splitting tensile strengths obtained for the
10 specimens are reported in Fig. 3. The variability of strengths from
different specimens is attributed to the local inhomogeneities pre-
sent in the sample. The average value of the splitting tensile

Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx 7

100
Initial
90 25 cycles
31 cycles
80

Particle Fineness, % smaller

70

60

50

40

30

20

10

0
100 200 300 400 500 600 700 800
Particle Size (microns)

Fig. 10. Particle size distribution obtained with a compressive force value of F 0 .

strength from the 10 samples was 208 psi. In this work, this single
value for the splitting tensile strength is used in calibrating the
cohesive strength parameter for the coal sample. However, it is pos-
sible to define the cohesive strength as a probability distribution
corresponding to the variance observed in several Brazilian tests.
Next, a numerical Brazilian sample is constructed to calibrate
the cohesive strength. A disc-shaped numerical sample, 5-mm
diameter and 2.5-mm thickness, is generated by a random dense
packing consisting of 5402 spherical particles of radius 100 lm.
To limit the problem size, a much smaller sample is selected for
numerical simulations compared to the actual experimental spec-
imen size. It is assumed that the tensile splitting strength, which is
a macro-scale property, remains the same across this range of
specimen sizes. The sample is placed between two plates as shown
in Fig. 4(a). The bottom plate is fixed, and the top plate moves
down with a constant velocity corresponding to a strain rate of
4
5  10 % per minute. This higher strain rate is used to speed up
the computations, and a numerical damping ratio of 0.8 is used
to limit the effects of strain rate. In addition, it is numerically ver-
ified that the loading strain rate does not significantly impact the
results. Fig. 4(b) shows the predicted broken state of the sample
beyond the post-peak load.
The cohesive strength of bonds is the critical mechanical prop-
erty for the breakage simulations and is obtained through calibra-
tion in this study. Other model parameters are chosen to be typical
material properties and from literature. Properties including mod-
ulus and Poisson’s ratio are selected from the material properties
for coal reported in [27]. Although the particle-scale properties,
in general, differ from the macroscale mechanical properties, this
difference should not have any distinguishable impact on the coal
sample’s breakage behavior. The dimensionless coefficients for
moment equations are chosen from [38]. Table 1 summarizes the
model and coal material parameters used in this study. Also,
Table 2 lists the material parameters for the loading platens used
in this model.
Fig. 5 shows the typical force-displacement response obtained
through the simulated Brazilian Test. A value of
rc ¼ sc ¼ 3  107 N=m2 for the cohesive strength parameter is
used to obtain this plot. The abscissa is the vertical displacement
of the top plate, and the ordinate is the vertical component of
the reaction force exerted by the coal specimen on the top plate.
The peak load (highest ordinate) in this graph corresponds to the
specimen’s failure point and is used to compute the value of tensile
splitting strength from Eq. (11). Fig. 6 compares the force-
displacement response obtained with two different loading strain
rates. It is evident from this graph that the responses (including
the peak load value) are very similar in the two cases.
Simulations with different values of cohesive strength parame-
ter were conducted. Fig. 7 shows the tensile splitting strength
obtained for different values of cohesive strength parameter. Then,
a linear curve fit is used to obtain the cohesive strength corre-
sponding to the experimentally determined splitting strength
value of 208 psi. This calibration procedure provides a value of
2:39  107 N=m2 for the cohesive bond strength parameter, which
is used for subsequent simulations featured in this paper. Notably,
this cohesive strength is a function of the particle size, as also indi-
cated in [29,44], and needs to be recalibrated if different-sized par-
ticles are used in the DEM simulation.
4. Model predictions and discussion
The grinding section of a coal pulverizer generally consists of
compressing rollers (or tires) and a rotating table on which the
raw coal is fed. Each passage of the rotating table against the rollers
breaks the coal particles into progressively smaller sizes. The ease
of pulverization, or grindability, of coal is measured using the Hard-
grove Grindability Index (HGI) defined by a standard test method
[3]. Fig. 8 provides the particle size distribution obtained from
the HGI test conducted on the coal specimen. The test was con-
ducted for six minutes on 50-gm coal sample. The y-axis in the plot
represents the cumulative percentage of particles by mass smaller
than a particular size. The x-axis features the normalized particle
size. In a sieve analysis, these quantities are more commonly
referred to as particle fineness and particle size, respectively. The
solid line curve represents the initial size distribution of the raw

Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
8 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx

100 100
Initial Initial
90 25 cycles 90 3.33 cycles
31 cycles 6 cycles
Particle Fineness, % smaller

80 80 11 cycles

Particle Fineness, % smaller

70 70

60 60

50 50

40 40

30 30

20 20

10 10

0 0
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800
Particle Size (microns) Particle Size (microns)
(a) (b)

100 100
Initial Initial
90 2.5 cycles 90 1.66 cycles
3.33 cycles 2.5 cycles
80 6 cycles 80 3.33 cycles
Particle Fineness, % smaller

Particle Fineness, % smaller

70 70

60 60

50 50

40 40

30 30

20 20

10 10

0 0
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800
Particle Size (microns) Particle Size (microns)
(c) (d)
Fig. 11. Particle size distribution obtained with a compressive force value of: (a) F 0 , (b) 1:5F 0 , (c) 2F 0 , and (d) 2:5F 0 .

coal sample, while the other curves provide the size distribution
35 evolution at different revolution counts during the HGI test.
It would have been desirable to model the HGI test and numer-
ically predict the particle size distribution evolution (as in Fig. 8)
30
for different coal types based on their mechanical properties. How-
ever, the numerical simulation of such a large-scale system is not
25 yet feasible due to the limitations of currently available computa-
Number of cycles

tional capabilities. A full-scale numerical model to predict the par-

20 ticle breakage in a real HGI system would require simulating
millions of particles. These simulations can be prohibitively expen-
sive even with state-of-the-art computational resources because of
15
the extremely small time stepping required for numerical stability
of the explicit integration scheme used in the DEM. The main
10 objective of this paper is to provide a framework for modeling
the breakage of coal particles under realistic mechanical loadings.
5
Therefore, a much simpler model that does not introduce the geo-
metrical and operational complexities of a full-scale pulverizer is
used to demonstrate this computational framework.
0 The model used in this study consists of a small cuboidal coal
1 1.5 2 2.5
sample, placed between two plates which provide mechanical
Normalized Force
loading on to the sample. This initial mechanically stable DEM
Fig. 12. Number of cycles required to reduce 90% of the sample to 100-lm
particles.

Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx 9

100 100
Initial Initial
90 14 cycles 90 5 cycles
20 cycles 6.66 cycles
80 80
Particle Fineness, % smaller

Particle Fineness, % smaller

70 cycles 11 cycles

70 70

60 60

50 50

40 40

30 30

20 20

10 10

0 0
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800
Particle Size (microns) Particle Size (microns)
(a) (b)
100
Initial
90 1.66 cycles
2.5 cycles
Particle Fineness, % smaller

80 3.33 cycles

70

60

50

40

30

20

10

0
100 200 300 400 500 600 700 800
Particle Size (microns)
(c)
Fig. 13. Particle size distribution obtained with a friction angle value of: (a) 24 , (b) 27 , and (c) 30 .

sample configuration is prepared using the procedure described as  This sample is isotropically compressed using rigid walls to a
follows: resulting hydrostatic stress state of 1:0  105 N=m2 and sub-
Sample preparation: sequently uncompressed to relieve the stress. In this process, no
gravitational forces are considered, and zero friction angle is
 First, a sample of spherical balls is generated based on the given used between the particles to generate a dense sample. Also,
initial size distribution of the coal sample. This sample is cre- each agglomerate in itself is assumed to be rigid, and the
ated such that the spherical balls are in a gas-like state without arrangement of particles within each agglomerate does not
any contact. The particle size distribution and corresponding change during this process.
cumulative mass fractions are specified, leading to spherical  This dynamically compressed sample is still not mechanically
balls of various sizes. stable under gravity. Hence, the sample is now placed on a plate
 The volume of each spherical ball is then replaced with an at the bottom and with periodic boundary conditions in x and z
agglomerate of densely packed spherical particles of radius directions. The sample is allowed to settle under gravity with-
100 lm such that the entire volume of these particles exists out any other external forces. During this stage, the particles
within the volume of the spherical ball. The resulting total num- within each agglomerate are connected by cohesive bonds,
ber of particles in this numerical sample is 2226. Because of the and the bond strength is set to a sufficiently large value to avoid
various sizes of spherical volumes used to fit the agglomerates any bond breakage. Before the main breakage simulation com-
of particles, all of the agglomerates do not need to appear spher- mences, the bond strength is recovered to the required value
ical in shape. The spherical ball only acts as a restricting volume obtained by calibration in Section 3.
for agglomerate particles to lie within-the larger the spherical
ball, the more spherical the shape of an agglomerate. Fig. 9 depicts the configuration of the generated coal sample.
Each agglomerate is represented by a different color. The dimen-
sions of the granular sample are 3:6 mm  2:6 mm  3:6 mm. In

Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
10 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx
this periodic simulation, the plates are generated using the rigidly
attached clumps of nine smaller segments to avoid some numerical
issues that would arise by using one single plate. Periodic bound-
ary conditions are assumed in the x and z directions to approxi-
mate the boundary behavior of a much bigger sample. Of note, in
Fig. 9, some of the agglomerates cross the periodic boundary and
appear on the other side of the periodic domain.
The progressive breakup of the interparticle bonds leads to the
breakage of an agglomerate, resulting in individual particles and
granular fragments of different sizes and shapes. A granular frag-
ment here is defined as a group of particles connected by at least
one continuous cohesive bond chain. The individual 100-lm-
sized particles are assumed to be non-breakable. Therefore, the
crushing of these particles that leads to even smaller-sized parti-
cles, which should happen in an actual scenario, is not accounted
for. There is no unique way to define the size of granular frag-
ments. For example, it can be defined by the size of the equivalent
solid sphere, the size of the sphere circumscribing the fragment, or
simply by the number of particles in each fragment. In this paper,
the fragment size is characterized by the radius of gyration, Rg ,
given by,
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 XN
Rg ¼ r þ kxk  xmean Þk2 ;
N k¼1
where N is the total number of particles in a fragment. xk is the posi-
tion of center of the kth particle, xmean is the mean position of cen-
ters of all particles in a fragment, and r is the radius of a single
spherical particle. This measure is commonly used in polymer phy-
sics to describe the dimension of polymer chains. For consistency
purposes, the size of an intact spherical agglomerate is also defined
by the radius of gyration, which differs from the radius of the orig-
inal spherical ball. The mass of each fragment is determined by
counting the number of particles and then multiplying the volume
of these particles with the solid density.
There are several factors that affect coal’s grindability, including
some of the major operating parameters for a given mill hardware
configuration, such as the roller wheel compression against the
table and the rotational speed of the table (grinding ring). The pul-
verization process mainly involves two broad categories of
mechanical loading on the coal sample: 1. Compression force on
the coal sample and 2. Transverse shear induced by the rotation.
Other factors are raw coal type and their physical properties, which
can be accounted for in the DEM model parameters. The effect of
some operating parameters on the coal sample’s breakage behavior
is examined via numerical experiments. The model parameters
used in these numerical simulations are the same as those in
Tables 1 and 2. The boundary conditions for this problem include
a constant vertical downward force acting on the top plate. The
top plate is free to translate in the y-direction and is subjected to
a constant horizontal velocity of V 0 ¼ 0:1 m=s in the x-direction.
All other translational and rotational degrees of freedom are con-
strained. The bottom plate is assumed to be fixed.
4.1. Effect of top plate compressive force
In a pulverizer, the roller wheel compression against the table is
an important mill operating parameter, governing the breakage
behavior of the coal sample. The desired small coal particle sizes
are difficult to achieve with insufficient compression. Conversely,
excess compression can lead to accelerated wear and tear on the
equipment. Therefore, informed control of the compression is nec-
essary for optimal operation of the pulverizer. In the context of the
current model, shown in Fig. 9, the vertical downward force on the
top plate leads to the desired compression. The effect of the mag-
nitude of that vertical force is studied in this section. Fig. 10 shows
Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
70
60
50
Number of cycles
40
30
20
10
0
24 25 26 27 28 29 30
Friction angle (in degrees)
Fig. 14. Number of cycles required to reduce 90% of the sample to 100-lm
particles.
ð12Þ
the size distribution obtained at different stages of simulation for
an arbitrary compressive force value of F 0 ¼ 1:45 N. The y-axis in
the plot represents the cumulative percentage of fragments by
mass finer than a particular size. The x-axis depicts the fragment
size (Eq. (12)). The solid line curve represents the initial size distri-
bution of the coal sample, while the other two curves provide the
size distribution at two random cycle counts in the course of the
simulation. A cycle is defined as one complete movement of the
top plate in the periodic cell, i.e., when the left edge of the top
plates moves from the left to the right edge of the periodic cell.
The increased breakage of agglomerates into smaller fragments
with time can be clearly seen in the graph. The dash-dot curve cor-
responding to the instance of 31 cycle counts indicates that
P  90% of the coal sample by mass has reduced to single particles
of radius 100 lm. Further particle breakage is not allowed, and
100 lm is the smallest resolution of particle size considered in this
study.
Although the numerical model used in this study is simplistic
and quantitative validation with test data from a real system com-
parable to the one shown in Fig. 8 is not expected, the model
demonstrates the mechanistic capability to simulate the effect of
operating parameters on breakage behavior. Nevertheless, the
model offers qualitative similarity with the experimental results
(Fig. 8) in capturing the progressive breakage of the coal particles.
Noticeably, the timescale required for observing significant break-
age in the numerical sample is orders of magnitude smaller com-
pared to that observed in the actual HGI test. This primarily
stems from using a compact numerical sample, which results in
rapid transmission of stresses that lead to faster breakage. One of
the main reasons for carefully preparing a compact and mechani-
cally stable sample is to reduce the simulation time.
For comparison purposes, Fig. 11 provides particle size distribu-
tion results obtained with different compressive force magnitudes.
All other model parameters were kept the same. These graphs
clearly show that granular sample breakage becomes more rapid
as the compressive force increases, which is consistent with phys-
ical intuition. For instance, the reduction of more than 90% of the
coal sample to individual particles took much fewer cycles with
higher compression forces, clearly demonstrated by Fig. 12. The
graph shows the number of cycles required to reduce 90% of the
 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx 11

100 100
Initial Initial
90 13 cycles 90 25 cycles
16 cycles 31 cycles
80 80

Particle Fineness, % smaller

Particle Fineness, % smaller

70 70

60 60

50 50

40 40

30 30

20 20

10 10

0 0
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800
Particle Size (microns) Particle Size (microns)
(a) (b)

100
Initial
90 41 cycles
56 cycles
80
Particle Fineness, % smaller

70

60

50

40

30

20

10

0
100 200 300 400 500 600 700 800
Particle Size (microns)
(c)
Fig. 15. Particle size distribution obtained with a plate velocity value of: (a) 0:5V 0 , (b) V 0 , and (c) 2V 0 .

granular sample by mass to 100-lm particles. Meanwhile, the x- Fig. 13 shows the particle size distribution obtained by using
axis features the normalized values of compressive force. three different friction angle values. It is evident that the breakage
becomes considerably slower as the friction angle is reduced. The
reduction in friction angle leads to increased slippage between
4.2. Effect of friction angle
the plate and coal particles, thereby resulting in reduced force
transmission from the plate movement. The friction angle’s effect
In the simulation results presented thus far, a constant friction
on the breakage rate appears to be much more drastic compared
angle with an arbitrary, but reasonable, value of 30° has been used
to the compression force. The number of cycles required for reduc-
for interparticle interactions or interactions between the plates
tion of more than 90% of the coal sample to 100-lm particles is
and particles. Although the internal friction angle for interparticle
plotted against the friction angle in Fig. 14.
interactions is an inherent property of the coal material sample,
the frictional interaction between the coal sample and plates can
be altered by introducing some surface profile on the grinding balls 4.3. Effect of plate velocity
or table. In this section, the effect of varying the friction angle for
interactions between the plate and coal particles is studied. As The rotational velocity of the table in a coal pulverizer is
mentioned previously in Section 2, for interactions between dis- another important operating parameter for mill operations. For
crete elements of two different materials, the lower of the two fric- the model considered in this work, the top plate velocity repre-
tion angles is used in the constitutive law. Therefore, using a value sents the rotational velocity in an actual mill. The effect of varia-
of /0 (for plates) lower than the one used for coal particles, implies tion in the magnitude of velocity on the coal sample’s breakage
that /0 is employed for frictional constitutive law between plates behavior is considered in this section. Three different velocity val-
and coal particles. The compressive force magnitude of 2:5F 0 is ues are considered: 0:5V 0 ; V 0 , and 2V 0 . The compression force is
used, keeping all other model parameters the same as in the previ- taken as F 0 and friction angle as /0 ¼ 30 . Fig. 15 illustrates the
ous section. results that correspond to these plate velocities. Increasing the

Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
12 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx
plate velocity does not seem to increase the breakage rate. The
number of cycles required to achieve similar breakage in the case
of higher plate velocity also seems to increase in the same propor-
tion as velocity. This could possibly be due to much smaller time-
scales required for breakage in this model which, in turn, is
because of the use of a compact sample. This result indicates that
the additional energy input to the grinding system in the form of
increased rotational velocity might not contribute to the granular
assembly breakage.
5. Concluding remarks
Coal particle size distribution plays an important role in the effi-
ciency of combustion in coal-fired power plants. Traditionally, the
breakage distribution from the pulverization process is obtained
through some empirical or semi-empirical approaches. These
approaches do not provide accurate predictions for varied coal
types because the breakage behavior inherently depends on the
coal’s physical properties. The main contribution of this work is
to provide an integrated experimental and DEM based computa-
tional framework to model the coal particle breakage. The model
requires several input parameters, with cohesive strength between
the particles deemed the most important of all that affect the
breakage behavior of coal. A procedure to calibrate the cohesive
strength parameter using the Brazilian Test is presented. The mod-
el’s details and capabilities are demonstrated with the help of a
surrogate system subjected to periodic boundary conditions. The
effect of some of the operating parameters on the breakage behav-
ior is then studied. Although the numerical problem demonstrated
in this work is simplistic compared to a real pulverizer, it still pro-
vides the complete framework required to build a full-scale predic-
tive model for the pulverizer. As computational capabilities
advance, this model can be extended to simulate the actual particle
size distributions in a pulverizer. Meanwhile, the model predic-
tions provided in this work can provide some quantitative insights
on the coal grindability. Also, this modeling approach, upon valida-
tion, can help in accelerate the design of pulverizers and selecting
operating conditions for new types of coal.
Because of computational limitations, the smallest discrete ele-
ment resolution of size 100-lm is used in the current study. Much
smaller particle sizes are desirable for more accurate predictions.
In addition, using a wide size distribution of coal particles to form
the sample agglomerates will lead to a denser, more realistic
numerical representation of the coal sample. The approach used
in this work is based on Bonded Particle Method, which is compu-
tationally intensive due to the large number of particles in the sim-
ulation. An alternative and computationally efficient replacement
strategy requires the crushing strength of different-sized particles.
The determination of such size-dependent strength for much
smaller-sized particles (few microns to mm) via experimental pro-
cedures is a subject of future research.
Acknowledgments
The Pacific Northwest National Laboratory is operated by the
Battelle Memorial Institute for the U.S. Department of Energy
under Contract No. of DE-AC05-76RL01830. This work was funded
by the U.S. Department of Energy, Office of Fossil Energy’s Carbon
Capture Simulation Initiative (CCSI) through the National Energy
Technology Laboratory.
Disclaimer
This manuscript was prepared as an account of work sponsored by
an agency of the United States Government. Neither the United
Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
States Government nor any agency thereof, 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 endorse-
ment, recommendation, or favoring by the United States Govern-
ment or any agency thereof. The views and opinions of authors
expressed herein do not necessarily state or reflect those of the
United States Government or any agency thereof.
References
[1] Suggested methods for determining tensile strength of rock materials. Int. J.
Rock Mech. Min. Sci. Geomech. Abstr. 15(2) (1978) 47–51. ISSN 0148-9062.
http://dx.doi.org/10.1016/0148-9062(78)91677-7. <http://
www.sciencedirect.com/science/article/pii/0148906278916777>.
[2] Standard test method for splitting tensile strength of intact rock core
specimens, in: ASTM D3967-08. ASTM International, 2008. <www.astm.org>.
[3] Standard test method for grindability of coal by the hardgrove-machine
method, in: ASTM D409. ASTM International, 2012. <www.astm.org>.
[4] S. Antonyuk, J. Tomas, S. Heinrich, L. Morl, Breakage behaviour of spherical
granulates by compression, Chem. Eng. Sci. 60 (14) (2005) 4031–4044, http://
dx.doi.org/10.1016/j.ces.2005.02.038, ISSN 0009-2509 Granulation across the
length scales – 2nd International Workshop on Granulation <http://
www.sciencedirect.com/science/article/pii/S0009250905001442> .
[5] L.G. Austin, P.T. Luckie, Methods for determination of breakage distribution
parameters, Powder Technol. 5 (4) (1972) 215–222, http://dx.doi.org/10.1016/
0032-5910(72)80022-6, ISSN 0032-5910 <http://
www.sciencedirect.com/science/article/pii/0032591072800226> .
[6] L.G. Austin, R.R. Klimpel, P.T. Luckie, Process engineering of size reduction: ball
milling, Soc. Min. Eng. AIME (1984), ISBN 9780895204219
<https://books.google.co.in/books?id=7uScQgAACAAJ> .
[7] B. Basu, D. Tiwari, D. Kundu, R. Prasad, Is Weibull distribution the most
appropriate statistical strength distribution for brittle materials?, Ceram Int.
35 (1) (2009) 237–246, http://dx.doi.org/10.1016/j.ceramint.2007.10.003, ISSN
0272-8842 <http://www.sciencedirect.com/science/article/pii/
S0272884207003665> .
[8] Z.T. Bieniawski, The effect of specimen size on compressive strength of coal,
Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 5 (4) (1968) 325–335, http://dx.
doi.org/10.1016/0148-9062(68)90004-1, ISSN 0148-9062 <http://
www.sciencedirect.com/science/article/pii/0148906268900041> .
[9] E. Bilgili, B. Scarlett, Population balance modeling of non-linear effects in
milling processes, Powder Technol. 153 (1) (2005) 59–71, http://dx.doi.org/
10.1016/j.powtec.2005.02.005, ISSN 0032-5910 <http://
www.sciencedirect.com/science/article/pii/S0032591005000598> .
[10] F. Bourrier, F. Kneib, B. Chareyre, T. Fourcaud, Discrete modeling of granular
soils reinforcement by plant roots, Ecol. Eng. 61 (Part C) (2013) 646–657,
http://dx.doi.org/10.1016/j.ecoleng.2013.05.002, ISSN 0925-8574 Soil Bio- and
Eco-Engineering: The Use of Vegetation to Improve Slope Stability <http://
www.sciencedirect.com/science/article/pii/S0925857413001936> .
[11] M.O. Ciantia, M. Arroyo, A. Gens, F.Calvetti, A new strategy to simulate particle
crushing in DEM analysis, in: Proceedings of 11th World Congress on
Computational Mechanics, 2014.
[12] M.O. Ciantia, M. Arroyo, F. Calvetti, A. Gens, An approach to enhance efficiency
of DEM modelling of soils with crushable grains, Géotechnique 65 (2) (2015)
91–110, http://dx.doi.org/10.1680/geot.13.P.218.
[13] P.W. Cleary, Recent advances in DEM modelling of tumbling mills, Miner. Eng.
14 (10) (2001) 1295–1319, http://dx.doi.org/10.1016/S0892-6875(01)00145-
5, ISSN 0892-6875 <http://www.sciencedirect.com/science/article/pii/
S0892687501001455> .
[14] P.W. Cleary, A multiscale method for including fine particle effects in DEM
models of grinding mills, Miner. Eng. 84 (2015) 88–99, http://dx.doi.org/
10.1016/j.mineng.2015.10.008, ISSN 0892-6875 <http://
www.sciencedirect.com/science/article/pii/S0892687515301047> .
[15] P.W. Cleary, M.D. Sinnott, Simulation of particle flows and breakage in
crushers using DEM: Part 1 compression crushers, Miner. Eng. 74 (2015) 178–
197, http://dx.doi.org/10.1016/j.mineng.2014.10.021, ISSN 0892-6875
<http://www.sciencedirect.com/science/article/pii/S0892687514003616> .
[16] P.A. Cundall, O.D.L. Strack, A discrete numerical model for granular assemblies,
Géotechnique 29 (1) (1979) 47–65, http://dx.doi.org/
10.1680/geot.1979.29.1.47.
[17] J. Fan, L. Qian, Y. Ma, P. Sun, K. Cen, Computational modeling of pulverized coal
combustion processes in tangentially fired furnaces, Chem. Eng. J. 81 (13)
(2001) 261–269, http://dx.doi.org/10.1016/S1385-8947(00)00212-6, ISSN
1385-8947 <http://www.sciencedirect.com/science/article/pii/
S1385894700002126> .
[18] B. Freireich, J. Li, J. Litster, C. Wassgren, Incorporating particle flow information
from discrete element simulations in population balance models of mixer-
 V. Gupta et al. / Advanced Powder Technology xxx (2017) xxx–xxx
coaters, Chem. Eng. Sci. 66 (16) (2011) 3592–3604, http://dx.doi.org/10.1016/j.
ces.2011.04.015, ISSN 0009-2509 <http://www.sciencedirect.com/science/
article/pii/S0009250911002570> .
[19] J. A Herbst, D. W Fuerstenau, Scale-up procedure for continuous grinding mill
design using population balance models, Int. J. Miner. Process. 7 (1) (1980) 1–
31, http://dx.doi.org/10.1016/0301-7516(80)90034-4, ISSN 0301-7516
<http://www.sciencedirect.com/science/article/pii/0301751680900344> .
[20] J. Huang, S. Xu, S. Hu, Influence of particle breakage on the dynamic
compression responses of brittle granular materials, Mech. Mater. 68 (2014)
15–28, http://dx.doi.org/10.1016/j.mechmat.2013.08.002, ISSN 0167-6636
<http://www.sciencedirect.com/science/article/pii/S0167663613001567> .
[21] R. Kumar, B. Freireich, C. Wassgren, DEM–compartment–population balance
model for particle coating in a horizontal rotating drum, Chem. Eng. Sci. 125
(2015) 144–157, http://dx.doi.org/10.1016/j.ces.2014.06.046, ISSN 0009-2509
Pharmaceutical Particles and Processing <http://
www.sciencedirect.com/science/article/pii/S0009250914003339> .
[22] A. Lisjak, G. Grasselli, A review of discrete modeling techniques for fracturing
processes in discontinuous rock masses, J. Rock Mech. Geotech. Eng. 6 (4)
(2014) 301–314, http://dx.doi.org/10.1016/j.jrmge.2013.12.007, ISSN 1674-
7755 <http://www.sciencedirect.com/science/article/pii/
S1674775514000213> .
[23] Stefan Luding, Introduction to discrete element methods: basic of contact force
models and how to perform the micro-macro transition to continuum theory,
Eur. J. Environ. Civ. Eng. 12 (7-8) (2008) 785–826.
[24] G.R. McDowell, Statistics of soil particle strength, Géotechnique 51 (10) (2001)
897–900, http://dx.doi.org/10.1680/geot.2001.51.10.897.
[25] G.R. McDowell, M.D. Bolton, On the micromechanics of crushable aggregates,
Géotechnique 48 (5) (1998) 667–679, http://dx.doi.org/
10.1680/geot.1998.48.5.667.
[26] G.R. McDowell, J.P. de Bono, On the micromechanics of one-dimensional
normal compression, Géotechnique 63 (11) (2013) 895–908.
[27] T.P. Medhurst, E.T. Brown, A study of the mechanical behaviour of coal for
pillar design, Int. J. Rock Mech. Min. Sci. 35 (8) (1998) 1087–1105, http://dx.
doi.org/10.1016/S0148-9062(98)00168-5, ISSN 1365-1609 <http://
www.sciencedirect.com/science/article/pii/S0148906298001685> .
[28] Y. Nakata, A.F.L. Hyde, M. Hyodo, H. Murata, A probabilistic approach to sand
particle crushing in the triaxial test, Géotechnique 49 (5) (1999) 567–583,
http://dx.doi.org/10.1680/geot.1999.49.5.567.
[29] D.H. Nguyen, P. Sornay, E. Azéma, Evolution of particle size distributions in
crushable granular materials, Geomech. Micro Macro (2014) 275–280.
[30] D.O. Potyondy, P.A. Cundall, A bonded-particle model for rock, Int. J. Rock
Mech. Min. Sci. 41 (8) (2004) 1329–1364, http://dx.doi.org/10.1016/j.
ijrmms.2004.09.011, ISSN 1365-1609 Rock Mechanics Results from the
Underground Research Laboratory, Canada <http://
www.sciencedirect.com/science/article/pii/S1365160904002874> .
[31] J. Quist, C.M. Evertsson, Cone crusher modelling and simulation using {DEM},
Miner. Eng. 85 (2016) 92–105, http://dx.doi.org/10.1016/j.
mineng.2015.11.004, ISSN 0892-6875 <http://
www.sciencedirect.com/science/article/pii/S0892687515301230> .
[32] R. Ramachandran, C.D. Immanuel, F. Stepanek, J.D. Litster, F.J. Doyle III, A
mechanistic model for breakage in population balances of granulation:
theoretical kernel development and experimental validation, Chem. Eng. Res.
Des. 87 (4) (2009) 598–614, http://dx.doi.org/10.1016/j.cherd.2008.11.007,
ISSN 0263-8762. 13th European Conference on Mixing: New developments
towards more efficient and sustainable operations <http://
www.sciencedirect.com/science/article/pii/S0263876208003225> .
[33] D. Ramkrishna, M.R. Singh, Population balance modeling: current status and
future prospects, Ann. Rev. Chem. Biomol. Eng. 5 (1) (2014) 123–146, http://
dx.doi.org/10.1146/annurev-chembioeng-060713-040241, PMID: 24606333
<http://dx.doi.org/10.1146/annurev-chembioeng-060713-040241> .
[34] J.A. Mendoza Rizo, Modeling rock cutting using DEM with crushable particles,
June 2010. <http://d-scholarship.pitt.edu/6570/>.
[35] K. Sato, N. Meguri, K. Shoji, H. Kanemoto, T. Hasegawa, T. Maruyama, Breakage
of coals in ring-roller mills part i. The breakage properties of various coals and
simulation model to predict steady-state mill performance, Powder Technol.
86 (3) (1996) 275–283, http://dx.doi.org/10.1016/0032-5910(95)03061-1,
ISSN 0032-5910 <http://www.sciencedirect.com/science/article/pii/
0032591095030611> .
[36] N. Schaffel, M. Mancini, A. Szlek, R. Weber, Mathematical modeling of {MILD}
combustion of pulverized coal, Combust. Flame 156 (9) (2009) 1771–1784,
http://dx.doi.org/10.1016/j.combustflame.2009.04.008, ISSN 0010-2180
<http://www.sciencedirect.com/science/article/pii/S0010218009001023> .
Please cite this article in press as: V. Gupta et al., A discrete element method-based approach to predict the breakage of coal, Advanced Powder Technology
(2017), http://dx.doi.org/10.1016/j.apt.2017.07.019
13
[37] L. Scholtés, F.V. Donzé, A {DEM} model for soft and hard rocks: role of grain
interlocking on strength, J. Mech. Phys. Solids 61 (2) (2013) 352–369, http://
dx.doi.org/10.1016/j.jmps.2012.10.005, ISSN 0022-5096 <http://
www.sciencedirect.com/science/article/pii/S0022509612002268> .
[38] L. Scholtés, F.V. Donzé, M. Khanal, Scale effects on strength of geomaterials,
case study: coal, J. Mech. Phys. Solids 59 (5) (2011) 1131–1146, http://dx.doi.
org/10.1016/j.jmps.2011.01.009, ISSN 0022-5096 <http://
www.sciencedirect.com/science/article/pii/S002250961100010X> .
[39] W. Schubert, M. Khanal, J. Tomas, Impact crushing of particle–particle
compounds–experiment and simulation, Int. J. Miner. Process. 75 (12)
(2005) 41–52, http://dx.doi.org/10.1016/j.minpro.2004.01.006, ISSN 0301-
7516 <http://www.sciencedirect.com/science/article/pii/
S0301751604000092> .
[40] A.P.S. Selvadurai, K. Sepehr, Discrete element modelling of fragmentable
geomaterials with size dependent strength, Eng. Geol. 53 (34) (1999) 235–241,
http://dx.doi.org/10.1016/S0013-7952(99)00039-3, ISSN 0013-7952 <http://
www.sciencedirect.com/science/article/pii/S0013795299000393> .
[41] M. Sen, D. Barrasso, R. Singh, R. Ramachandran, A multi-scale hybrid CFD-
DEM-PBM description of a fluid-bed granulation process, Processes 2 (1)
(2014) 89, http://dx.doi.org/10.3390/pr2010089, ISSN 2227-9717 <http://
www.mdpi.com/2227-9717/2/1/89> .
[42] K. Shoji, N. Meguri, K. Sato, H. Kanemoto, T. Hasegawa, T. Maruyama, Breakage
of coals in ring-roller mills part 2. An unsteady-state simulation model,
Powder Technol. 99 (1) (1998) 46–52, http://dx.doi.org/10.1016/S0032-5910
(98)00089-8, ISSN 0032-5910 <http://www.sciencedirect.com/science/
article/pii/S0032591098000898> .
[43] L.M. Tavares, R.M. de Carvalho, Modeling breakage rates of coarse particles in
ball mills, Miner. Eng. 22 (78) (2009) 650–659, http://dx.doi.org/10.1016/j.
mineng.2009.03.015, ISSN 0892-6875 Special issue: Comminution Special
issue: Comminution <http://www.sciencedirect.com/science/article/pii/
S0892687509000971> .
[44] S.C. Thakur, J.Y. Ooi, H. Ahmadian, Scaling of discrete element model
parameters for cohesionless and cohesive solid, Powder Technol. 293 (2016)
130–137, http://dx.doi.org/10.1016/j.powtec.2015.05.051, ISSN 0032-5910
Particle Modelling with the Discrete Element Method A success story of
{PARDEM} (www.pardem.eu) <http://www.sciencedirect.com/science/
article/pii/S0032591015004817> .
[45] C. Thornton, K.K. Yin, M.J. Adams, Numerical simulation of the impact fracture
and fragmentation of agglomerates, J. Phys. D: Appl. Phys. 29 (2) (1996) 424.
<http://stacks.iop.org/0022-3727/29/i=2/a=021> .
[46] C. Thornton, M. T Ciomocos, M.J. Adams, Numerical simulations of
agglomerate impact breakage, Powder Technol. 105 (13) (1999) 74–82,
http://dx.doi.org/10.1016/S0032-5910(99)00120-5, ISSN 0032-5910 <http://
www.sciencedirect.com/science/article/pii/S0032591099001205> .
[47] J. Tomas, M. Schreier, T. Groger, S. Ehlers, Impact crushing of concrete for
liberation and recycling, Powder Technol. 105 (13) (1999) 39–51, http://dx.
doi.org/10.1016/S0032-5910(99)00116-3, ISSN 0032-5910 <http://
www.sciencedirect.com/science/article/pii/S0032591099001163> .
[48] E.T. Tuzcu, R.K. Rajamani, Modeling breakage rates in mills with impact energy
spectra and ultra fast load cell data, Miner. Eng. 24 (34) (2011) 252–260,
http://dx.doi.org/10.1016/j.mineng.2010.08.017, ISSN 0892-6875 Special
issue: Comminution <http://www.sciencedirect.com/science/article/pii/
S0892687510002165> .
[49] V. Šmilauer et al, Yade Documentation, second ed., The Yade Project, 2015a.
http://dx.doi.org/10.5281/zenodo.34073. URL <http://yade-dem.org/doc/>.
[50] V. Šmilauer et al, Using and programming, in: Yade Documentation, second
ed., The Yade Project, 2015b, http://dx.doi.org/10.5281/zenodo.34043, URL
<http://yade-dem.org/doc/>.
[51] J. Wang, H. Yan, On the role of particle breakage in the shear failure behavior of
granular soils by DEM, Int. J. Numer. Anal. Meth. Geomech. 37 (8) (2013) 832–
854, http://dx.doi.org/10.1002/nag.1124, ISSN 1096-9853.
[52] N.S. Weerasekara, M.S. Powell, P.W. Cleary, L.M. Tavares, M. Evertsson, R.D.
Morrison, J. Quist, R.M. Carvalho, The contribution of {DEM} to the science of
comminution, Powder Technol. 248 (2013) 3–24, http://dx.doi.org/10.1016/j.
powtec.2013.05.032, ISSN 0032-5910 Discrete Element Modelling <http://
www.sciencedirect.com/science/article/pii/S0032591013003902> .
[53] W. Weibull, A statistical distribution function of wide applicability, J. Appl.
Mech. Trans. ASME 18 (3) (1951) 293–297.
[54] X. Xue, Y. Wang, Particle size distribution as a nonindependent variable
affecting pulverized-coal burnout in coal-fired power-plant boilers, Energy
Fuels 27 (8) (2013) 4930–4934, http://dx.doi.org/10.1021/ef400406w.