26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

SIMULATION OF SINGLE PARTICLE BREAKAGE USING NON-ROUND

PARTICLES IN ROCKY DEM
Flávio P. André*, Alexander Potapov†, Clovis Maliska Jr.† and Luís Marcelo Tavares*
*
Department of Metallurgical and Materials Engineering, Universidade Federal do Rio de
Janeiro – COPPE/UFRJ, Cx. Postal 68505, CEP 21941-972, Rio de Janeiro, RJ, Brazil
e-mail: [email protected]

†
ESSS Rocky, Florianópolis, SC, Brazil
e-mail: [email protected]

ABSTRACT

Simulations using the discrete element method (DEM) have arisen as an effective tool to
understand the dynamics inside different comminution and classification devices. In most of
these simulations, breakage is often left to a post processing stage or is disregarded altogether.
In several types of crushers and mills this will result in unrealistic simulations, since particle flow
and particle size reduction cannot be decoupled. The present work analyzed results from DEM
simulations of single particle breakage tests adopting a suite of phenomenological and empirical
models, named the Tavares Breakage Model, to validate the implementation of the model on
ROCKY-DEM, a commercial DEM platform. Good correlation between previously validated
numerical and simulated results were found for different particle breakage properties studied
demonstrating the very good potential of the model to represent breakage inside a DEM
environment.

Keywords: Discrete element method, particle breakage, comminution, Rocky DEM

1. INTRODUCTION

The discrete element method (DEM) is a numerical method originally proposed by

Cundall & Strack (1979) which allows the simulation of individual particle motion in a granular
medium based on the application of Newton's second law of motion. DEM has intensively
contributed to the understanding of several equipment and processes in a wide range of
applications across different areas. In the mineral processing area, DEM has enabled the analysis
of internal conditions and dynamics inside different comminution and classification devices.

Crushing and handling are key processes in both aggregates and mining industries. Over
the years, several attempts to understand the breakage phenomena and describe them using
comminution models have been made (Napier-Munn et al., 1996; King, 2001) The latest
developments in the field are mechanistic models, which now allow the prediction of particle
size reduction in a post processing stage on the basis of the information of energy transfer
collected on DEM simulations of comminution devices (Weerasekara et al., 2013). The adoption
of the discrete element method has provided the necessary information to understand the
fundamentals of comminution devices, allowing the optimization of operation conditions of
crushers and mills. Although undesired, breakage is also important in a number of other unit
operations, namely handling, mixing, compaction and separation. Likewise, attempts have been
made to address breakage in some of these operations also as a post-processing stage (Tavares
& Carvalho, 2012).

1
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

In several instances in practice, however, simultaneous simulation of particle motion

and breakage in the DEM environment becomes necessary. This becomes critical in cases when
particle size reduction is necessary to ensure a proper description of particle flow through the
equipment. Examples of these are gyratory and cone crushers, as well as high-pressure grinding
rollers. Different approaches have been proposed to assess this limitation, such as the bonded-
particle model (BPM) and the particle replacement model (PRM). The former one uses a set of
spherical particles bonded together in each contact point using bonding beams (Potyondy &
Cundall, 2004), while the latter adopts the substitution of a particle by a set of progenies after a
breakage event (Cleary, 2001). Although these approaches have been demonstrated to be able
to represent breakage with some success, they contain severe limitations regarding the lack of
mass conservation due to the use of spherical particles, high computational demand and lack of
rock material breakage properties such as the distribution of particles fracture energy (Jiménez-
Herrera et al., 2018).

A model that was previously validated, and that encompasses descriptions of several
inherent rock material properties, namely, breakage probability, particle weakening due to
repeated stresses and energy-specific fragment size distribution (Tavares, 2004, 2009; Carvalho
& Tavares, 2013) has been implemented in ROCKY-DEM, being called Tavares Breakage Model.
Single particle impact tests are widely used to describe ore breakage behavior due to their ability
to replicate stress mechanisms that are present in several comminution equipment. As such, the
present work compares predictions using the continuous version of the breakage model to
results of simulation of single particle breakage by both impact tests and drop weight tests in
ROCKY-DEM.

2. TAVARES BREAKAGE MODEL

The proposed Tavares Breakage Model is a model that comprises a series of equations
proposed by Tavares (2004, 2009) and Tavares & King (1998, 2002) to address important
phenomena in particle breakage. The model decouples dominant mechanisms involved in the
comminution process and is based on the dissipated energy in each collision event. The model
has been successfully used, as a post processing stage, to predict particle breakage in ball mills
(Tavares & Carvalho, 2009), autogenous and semi-autogenous mills (Carvalho & Tavares, 2011),
impact crushers (Cunha et al., 2014), as well as degradation of lump ores during handling
(Tavares & Carvalho, 2012). It has now been adapted to discrete elements in ROCKY-DEM
version 4.2.

The adoption of non-round particles in Rocky DEM allows the model to overcome
limitations of other approaches regarding mass and volume conservation after breakage
(Jiménez-Herrera et al., 2018). Besides that, the model’s mechanistic approach accounts for
breakage properties that are not fully covered by other breakage models already available in
commercial DEM environments. Briefly, the main sub-models contained in the Tavares Breakage
Model are presented as follows.

2.1. Statistical distribution of fracture energy

The model considers that the driver for breakage of each individual particle is the energy
dissipated in each contact. When this energy dissipated surpasses the minimum energy required
to break the particle, called particle fracture energy, the particle will fail. The model also
recognizes that particles of the same size and composition do not break with the same amount

2
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

of applied energy due to uneven presence of internal flaws, besides differences in particle
shapes (Tavares & King, 1998). The distribution of particles fracture energy has been successfully
represented by the upper-truncated lognormal distribution (Tavares & King, 1998) consisting of
a cumulative probability distribution for a measured value of energy dissipated in the contact,
that is:
1 ln ∗ − (1)
( )= 1 + erf
2 √2
and

∗ (2)
=
−
where Emax is the upper truncation value of the distribution, E50 and σ² are the median and the
geometric variance of the distribution, respectively.

2.2. Effect of particle size on fracture energy

The fracture energy is strongly related to particle size. The size of flaws and defects
decrease as particles become finer, leading to an increase in the particle strength and fracture
energy. This relationship between particle size and median fracture energy may be well
described using the expression (Tavares, 2004)

(3)
= 1+

where E∞, do and are model parameters fitted to experimental data and dj is the
representative size of the size class of interest.

2.3. Damage accumulation and particle weakening

Whenever the collision energy is lower than the energy required to fracture a particle,
the particle will not break. However, as a result of the impact, the particle can become more
amenable to break on a subsequent stressing event due to the accrual of internal damage
(Tavares & King, 2002; Tavares, 2009). Damage sustained in the nth impact cycle is given by

2 (4)
∗ ,
= ∗
(2 − 5 + 5)
where γ is the damage accumulation coefficient, Ek,n, the contact specific energy and En-1, the
particle energy threshold. For values of Ek,n greater than En-1, the particle will break. If not,
damage will be accumulated, and the adjusted fracture energy of particles will be given by
∗
= (1 − ) (5)
where En is the particle fracture energy after sustaining damage.

2.4. Particle fragmentation

Progeny fineness will be related to the stressing energy based on the parameter t10,
corresponding to the percentage in weight of the original particle that passes through a sieve
with aperture equal to 1/10th of its original size. The expression for t10 is an adaptation of the
originally equation proposed by Narayanan & Whiten (1988), being given by (Tavares, 2009),

3
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

, (6)
= 1 − exp −

where A and b’ are model parameters that are fitted to experimental data.

The complete particle size distribution is estimated by the incomplete beta function
distribution according to the equation:
⁄ (7)
100
( )= (1 − )
∫ (1 − )

where the cumulative mass of the particles passing a screen (tn) with size x is defined by the
calculated t10 value and the original size of the stressed particle.

3. METHODS

3.1. Material breakage parameters

Parameters from two distinct ores from Brazil where used in the simulations. The first
consisted of limestone and the second a copper ore. The full suite of breakage parameters have
been previously estimated on the basis of single-particle breakage data (Carvalho & Tavares,
2013). In addition, additional material parameters were obtained from previous studies in the
authors´ laboratory (Jiménez-Herrera et al., 2018). The reason behind the choice of these
materials was the significant difference in their breakage strengths, so as to demonstrate the
application of the modeling approach from very soft to very tough materials. The procedures
used to estimate the model parameters can be found elsewhere (Tavares 2004, 2009).
Parameters used in single-particle simulations of the two materials are summarized in Table 1.

Table 1. Input parameter of the Tavares Breakage Model for limestone and copper ore

Copper ore Limestone

γ 5 5.4
σ² 0.638 0.642
A 67.7 53.3
b´ 0.0294 0.033
E∞ (J/kg) 213.5 7
do (mm) 8.073 100
φ 1.219 0.8
Emax/E50 4 4
α1.2/β1.2 0.51/11.95 0.19/7.78
α1.5/β1.5 1.07/13.87 0.56/7.51
α2/β2 1.01/8.09 0.78/5.55
α4/β4 1.08/3.03 1.12/3.01
α25/β25 1.01/0.53 1.17/0.54
α50/β50 1.03/0.36 1.43/0.40
α7/β75 1.03/0.30 1.92/0.42
Density (kg/m³) 2930 2710
Young’s Modulus (N/m²) 5e+08 5e+08

4
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

3.2. Modeling

Simulations were performed in the commercial software Rocky DEM using the
Hysteretic Linear Spring model for the normal component of the force and the Linear Spring
Coulomb Limit for the tangential component of the force and different particle properties
required were input. Single impact tests were performed by propelling particles against an anvil
at different velocities to represent a wide range of collision energies. Double impact tests were
also simulated by using drop weight tests (DWT) with cylindrical weights, so as to simulate the
standard JKDWT (Napier-Munn et al., 1996) Different collision energies where simulated by
changing the drop height and the mass of the falling cylinder. The simulation setups of both tests
are presented in Fig. 1.

Fig. 1. Simulation setup for the drop weight test (left) and the single impact test (right). Ore particles are
shown in red

A generic irregular superquadric particle with 25 corners, with vertical aspect ratio of
0.80 and horizontal aspect ratio of 1.00 was adopted for both materials to avoid introducing
discrepancies beyond material parameters. Different particle sizes and minimum sizes of the
fragments generated were used depending on the purpose of the simulation.

Particles of 5.5 mm in size, comprising the size range of 4.75 to 6.30 mm were used for
the DWT and single impact tests. Particle size distribution of the broken fragments of 48 particles
for each energy and material using DWT simulations were post processed and compared with
model predictions. The minimum size fragment for these simulations was limited to 0.25 mm,
that is, about 1/20th of the original particle size. Both tests were used in checking the breakage
probability model. In these tests, due to the random generation of fracture energies of the
particles, much larger samples (288 and 576 particles) were selected for each energy and
material in the DWT and the single impact test, respectively. In order to reduce computational
effort in these last tests, the minimum size of the fragments was set to 0.55 mm.

Furthermore, particle weakening by damage accrual was assessed in the simulations by

counting the number of broken particles after each drop on simulations involving a total of 20
consecutive drops. A sample of 100 particles for each material with particle size of 37.5 mm and
minimum fragment size of 5.0 mm was used. Before each drop, all particles were resting on a
surface to avoid drop height discrepancies. In order to enable comparison with the model, any
contact besides the first contact between the particle and the anvil was considered negligible
and, therefore, was not considered in the model predictions.

5
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

4. RESULTS

4.1. Fracture energy distribution and breakage probability

Each material was tested on five different impact energies for both double and single
impact tests. It is known that particles of a certain size range do not present the same fracture
energy (Tavares & King, 1998) and this is shown in Fig. 2, which compares the breakage
probability distribution given by equations 1 and 2 to the simulated results for both types of
tests. The figure also illustrates that that Rocky DEM prediction was able to reproduce very well
the numerical distribution, presenting a small underestimation of breakage probability.

100
90
80
Breakage Probability (%)

70
60
50
40
Copper ore - Tavares Model
30 Copper ore - DWT
20 Copper ore - Single impact
Limestone - Tavares Model
10 Limestone - DWT
Limestone - Single impact
0
10 100 1000 10000
Impact energy (J/kg)

Fig. 2. Comparison between the modeled (lines) and simulated (symbols) breakage probabilities for drop
weight tests and single impact tests of 5.5 mm particles for different impact energies

4.2. Progeny size distribution

Progeny size distributions of broken particles for both materials were measured after
each impact and compared to results obtained by the model equations. Although the fineness
of the distribution is primarily based on the value of t10 determined using equation 6, it is chosen
here to illustrate fineness of the progeny using the t4 and the t2 values, which required
estimation using the incomplete beta function (Equation 7). This was necessary, given the
minimum size selected and the corresponding challenges in representing the finer sizes in the
simulations in DEM. Fig. 3 and Fig. 4 compare simulations outcomes with the expected values.
Good agreement between the modeled values of t4 and t2 and predictions using the DEM
simulations, which demonstrates the capability of the model in predicting breakage
quantitatively.

6
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

100
Copper ore - Tavares Model
90
Copper ore - DWT
80 Limestone - Tavares Model
Limestone - DWT
70
60
t4 (%)

50
40
30
20
10
0
10 100 1000 10000 100000
Impact energy (J/kg)

Fig. 3. Comparison between the modeled and simulated t4 values for drop weight tests of 5.5 mm
particles for different impact energies. Lines represent the continuous model and symbols the
simulations

100
90
80
70
60
t2 (%)

50
40
30
Copper ore - Tavares Model
20 Copper ore - DWT
Limestone - Tavares Model
10
Limestone - DWT
0
10 100 1000 10000 100000
Impact energy (J/kg)

Fig. 4. Comparison between the modeled and simulated t2 values for drop weight tests of 5.5 mm
particles for different impact energies. Lines represent the continuous model and symbols the
simulations

Size distributions of copper ore and limestone particles when stressed by different
impact energies are presented in Fig. 5 and Fig. 6, respectively, for drop weight tests. The
simulated values reproduced correctly the expected material behavior, generating finer
fragments when subjecting ore particles to stresses of higher magnitude due to the re-breakage
of fragments. As expected, due to the minimum size selected (0.25 mm), the model
underestimates values below around 1/10th of the initial particle size (t10) for all impact energies.

7
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

100
Tavares Model (431 J/kg)
DWT (431 J/kg)
Tavares Model (2237 J/kg)
80
DWT (2237 J/kg)
Tavares Model (5817 J/kg)
DWT (5817 J/kg)
Passing (%)

60

40

20

0
0.1 1 10
Particle size (mm)

Fig. 5. Comparison between modeled (symbols) and simulated (lines) progeny size distributions for drop
weight tests on 5.5 mm copper ore particles for different impact energies

100
Tavares Model ( 67 J/kg)
DWT (67 J/kg)
Tavares Model (320 J/kg)
80 DWT (320 J/kg)
Tavares Model (3772 J/kg)
DWT (3772 J/kg)
Passing (%)

60

40

20

0
0.1 1 10
Particle size (mm)

Fig. 6. Comparison between the modeled (symbols) and simulated (lines) progeny size distributions for
drop weight tests on 5.5 mm limestone particles different impact energies

4.3. Damage accumulation

Particle weakening was assessed on the basis of the cumulative number particles broken
in consecutive impacts. Due to the small deviation on the breakage probability estimation,
evident in Fig. 2, numerical analyses were conducted with the set impact energy, but also with
a reduced impact energy, so as to match exactly the breakage probability in the first impact. In
simulations with copper ore particles, this resulted in a reduction from the set impact energy of
94.7 J/kg to 85.6 J/kg, whilst for the limestone particles, the considered impact energy was

8
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

reduced from 10.5 J/kg to 9.4 J/kg. Fig. 7 compares both curves to the simulated results, which
demonstrate excellent agreement.

100 100
90 Copper ore 90 Limestone
80

Cumulative broken (%)

80
Cumulative broken (%)

70 70
60 60
50 50
40 40
30 30
Tavares Model (94.7 J/kg) 20 Tavares Model (10.5 J/kg)
20
Tavares Model (85.6 J/kg) Tavares Model (9.4 J/kg)
10 10
Single impact Single impact
0 0
0 5 10 15 20 0 5 10 15 20
Number of impacts Number of impacts
Fig. 7. Comparison between the modeled and simulated values of cumulative broken for consecutives
single impacts. The continuous lines are the expected values for cumulative broken and the dashed line
the values of cumulative broken for a reduced impact energy

5. DISCUSSION

As mentioned, the literature generated over more than a century of research on single
particle breakage is vast and enlightening. While attempts still remain being made to connect
microscale phenomena involved in fracture and fragmentation of particles, namely, crack
propagation and branching, are evolving, models that capture the phenomenology of the most
important aspects of particle breakage and that are amenable to be fitted from single-particle
breakage data are available. Among the various models available in this area, the so-called
Tavares Model is believed to be able to combine the level of detail and flexibility required for
describing the observed phenomena and the mathematical simplicity through explicit
expressions.

Adapting a continuous model to discrete elements in not a straightforward task and the
work presented herein demonstrates that the implementation in Rocky DEM captures, within
the limitations of the discretization, all key aspects of the model developed by Tavares and co-
workers. Indeed, the very good agreement between model equations and the simulation model
clearly shows that it represents a significant advance in comparison to the earlier model
available in Rocky DEM, whose limitations have been the object of earlier work (Jiménez-Herrera
et al., 2018).

In its present form the model will be particularly useful to describe breakage in a number
of systems, including those involved in mechanical degradation of materials, crushing and
grinding, bearing in mind the limitation of the minimum size and the corresponding
computational effort. One particular innovation in the present model in comparison to those
previously available in commercial platforms is its ability to describe weakening due to
unsuccessful stressing events. Indeed, amenability to particle weakening is an important
material property to be considered when simulating a number of comminution equipment such
as ball mills, semi autogenous and autogenous mills, due to the elevated number of collisions
that do not lead particles to fail.

9
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

6. CONCLUSIONS

Simulations of breakage of different materials on single-particle impact tests using the

Tavares Breakage Model available on Rocky DEM were compared in regard to their response to
breakage probability, progeny particle size distribution and particle weakening due to damage
accumulation.

It became evident that the model implemented in the simulation software described
particle fracture energy distribution with accuracy, presenting only a modest underestimation
in comparison to the model results for single impact tests for both materials and for double
impact tests involving only the tougher material studied (copper ore). The adoption of materials
with large differences in mean fracture energy indicates that the model is able to predict
breakage probability for a wide range of materials. Regarding the progeny size distribution, the
simulation model was generally capable of predicting the fineness of the distribution on the
basis of the energy applied, presenting good correspondence with the numerical solution of the
model. While overestimating breakage for weak materials under high impact energies, such as
for the weaker material studied (limestone), and presenting the opposite behavior concerning
the tougher material (copper ore), the re-breakage of progeny particles allowed the model to
correctly perform particle fragmentation with low discrepancy with modeled results. The model
also demonstrated the ability to predict particle weakening, showing good agreement between
the cumulative number of broken particles and the experimental data.

Breakage in DEM is useful to provide fundamental understanding for the optimization

and operation of numerous industrial processes. Good results in single particle impacts indicates
that the model can be extended to simulation of comminution processes. The model adopted in
Rocky DEM offers several advantages in particle breakage simulations. The adoption of
polyhedral particle shapes conserves mass and volume after a breakage event and the progeny
presents realistic shape and size distribution. Moreover, the results shown great fidelity in
describing several aspects of particle breakage while its parameter fitting can be performed with
quite simplicity.

REFERENCES

CARVALHO, R.M., TAVARES, L.M., 2011, "Leaping forward in SAG and AG mill simulation using
a mechanistic model framework". In: Proceedings of SAG 2011, pp. 1–26, Vancouver.
CARVALHO, R.M., TAVARES, L.M., 2013, "Predicting the effect of operating and design
variables on breakage rates using the mechanistic ball mill model". Minerals
Engineering, v. 43–44, pp. 91–101.
CLEARY, P.W., 2001, "Recent advances in DEM modelling of tumbling mills". Minerals
Engineering, v. 14, pp. 1295–1319.
CUNDALL, P.A., STRACK, O.D.L., 1979, "A discrete numerical model for granular assemblies".
Géotechnique, v. 29, pp. 47–65.
CUNHA, E.R., CARVALHO, R.M., TAVARES, L.M., 2014, "A predictive model of the vertical shaft
impact (VSI) crusher". In: Proceedings of Comminution 2014, Cape Town.
JIMÉNEZ-HERRERA, N., BARRIOS, G.K.P., TAVARES, L.M., 2018, "Comparison of breakage
models in DEM in simulating impact on particle beds". Advanced Powder Technology, v.
29, pp. 692–706.

10
26th International Mining Congress and Mining Exhibition of Turkey, April 16-19th, 2019

KING, R.P., 2001, Mineral Processing Design and Operations.

NAPIER-MUNN, T.J., MORRELL, S., MORRISON, R.D., KOJOVIC, T., 1996, Mineral Comminution
Circuits - Their Operation and Optimisation.
NARAYANAN, S.S., WHITEN, W.J., 1988, "Determination of comminution characteristics from
single particle breakage tests and its application to ball mill scale-up". Transactions of
the Institution of Mining and Metallurgy, v. 97, pp. 115–124.
POTYONDY, D.O., CUNDALL, P.A., 2004, "A bonded-particle model for rock". International
Journal of Rock Mechanics and Mining Sciences, v. 41, pp. 1329–1364.
TAVARES, L.M., 2004, "Optimum routes for particle breakage by impact". Powder Technology,
v. 142, pp. 81–91.
TAVARES, L.M., 2009, "Analysis of particle fracture by repeated stressing as damage
accumulation". Powder Technology, v. 190, pp. 327–339.
TAVARES, L.M., CARVALHO, R.M., 2009, "Modeling breakage rates of coarse particles in ball
mills". Minerals Engineering, v. 22, pp. 650–659.
TAVARES, L.M., CARVALHO, R.M., 2012, "Modeling ore degradation during handling using
continuum damage mechanics". International Journal of Mineral Processing, v. 112–
113, pp. 1–6.
TAVARES, L.M., KING, R.P., 1998, "Single-particle fracture under impact loading". International
Journal of Mineral Processing, v. 54, pp. 1–28.
TAVARES, L.M., KING, R.P., 2002, "Modeling of particle fracture by repeated impacts using
continuum damage mechanics". Powder Technology, v. 123, pp. 138–146.
WEERASEKARA, N. S., POWELL, M.S., CLEARY, P.W., TAVARES, L.M., EVERTSSON, M.,
MORRISON, R.D., QUIST, J., CARVALHO, R.M., 2013, "The contribution of DEM to the
science of comminution". Powder Technology, v. 248, pp. 3–24.

11