You are on page 1of 13

Powder Technology 356 (2019) 60–72

Contents lists available at ScienceDirect

Powder Technology

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

Laboratory-scale validation of a DEM model of a toothed double-roll


crusher and numerical studies
Yun-Wang Li a, La-La Zhao a,⁎, Er-Yi Hu a, Kang-Kang Yang a, Jing-Feng He b, Hai-Shen Jiang b, Qin-Fu Hou c
a
School of Mechatronic Engineering, China University of Mining and Technology, Xuzhou 221116, PR China
b
School of Chemical Engineering and Technology, China University of Mining and Technology, Xuzhou 221116, PR China
c
Department of Chemical Engineering, Monash University, Clayton, VIC 3800, Australia

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

Article history: A discrete element method (DEM) model of the crushing process of a toothed double-roll crusher (TDRC) is
Received 12 March 2019 established using the bonded particle model. DEM results and experimental data are compared quantitatively
Received in revised form 8 July 2019 and a relatively good agreement is observed. The effects of rotation speed and structure of crushing rolls on
Accepted 9 August 2019
the performance of TDRC are investigated numerically. The results show that when the rolls' speed is relatively
Available online 10 August 2019
high, the nipping condition would be improved, and more cracks could be created to release the increasing strain
Keywords:
energy, generating more fractions of small sizes in the products. But, when rolls' speed exceeds 150 rpm,
Crushing process the crushing performance would not be significantly improved. A reasonable working gap and better nipping be-
Toothed double-roll crusher haviour are obtained using the spiral-tooth-roll or the staggered-tooth-roll. The validated DEM model could be
Discrete element method applied to gain a fundamental understanding of the crushing mechanisms of TDRC.
Bonded particle model © 2019 Elsevier B.V. All rights reserved.
Laboratory-scale validation

1. Introduction based crushers, TDRC breaks particles by a combination of pressed,


split and shear forces. It should be noticed that the shear resistance of
A toothed double-roll crusher (referred to hereafter as TDRC) con- materials is generally much lower than the compression resistance.
sists of a pair of oppositely rotating rolls, which have a series of Thus, the particles would be crushed more easily resulting from the
intermeshing teeth protruding from the roll surfaces. The crushing ac- combined failure which makes TDRC be an efficient and energy-saving
tion of TDRC is a combination of compression and ripping, where the crushing machine [7].
teeth can dig into the rock and larger pieces than the roll diameter can Most of the current crushing models are based on experimental and
be handled. The TDRC is typically employed in the primary (coarse) industrial data or empirical equations, which lack insightful under-
crushing stage of soft materials such as ore coal, limestone, and soft or standing of the crushing process and fail to implement the effect of op-
sticky iron ores. The particle flow and breakage behaviour during the erating parameters on the particle flow and crushing behaviour. Thus, it
crushing process are important for the performance of TDRC. But, the is very helpful to understand the crushing phenomena at a particle
understanding of these fundamentals is limited at present. scale. However, it is difficult to study particle breakage with the current
The failure mechanisms of particles during the crushing process experimental techniques. To overcome the above limitation, numerical
with different types of crushers are generally different. Currently, the methods based on the discrete element method (DEM), originally pro-
crushers applied in the industries could be roughly divided into two cat- posed by Cundall and Strack [8], were developed based on Newton's
egories: compression and impact-based crushers [1,2]. In compressive motion laws. DEM has proved to be an effective numerical technique
crushing machines (e.g., jaw crusher [3], cone crusher [4]), the material for investigating the mechanical behaviour of discrete particle flows.
could be broken into coarse particles mainly resulting from the induced Successful applications of DEM models can be found in many fields in-
tensile failure, and fines mainly from the compressive failure near the cluding mineral processing, chemical engineering and agricultural engi-
points of loading. There are internal stresses in the particles broken by neering [9,10].
the pressure that can later cause cracking. However, the materials in im- To date, there is an increasing number of reports of DEM simulations
pact crushers (e.g., hammer mills [5], impact mills [6]) are crushed of particle breakage for both the crushable granular materials and the
mainly by tensile failure, which causes immediate fracture with no re- comminution machines (crushers and grinding mills) in the literature
sidual stresses. Compared with the above compression or impact- [11]. The bonded particle model (BPM) and the particle replacement
method (PRM) are two main particle breakage modelling approaches
⁎ Corresponding author. within the framework of DEM, which have been compared by simulat-
E-mail address: lala.zhao@cumt.edu.cn (L.-L. Zhao). ing the impact on particle beds [12]. BPM has been widely used to

https://doi.org/10.1016/j.powtec.2019.08.010
0032-5910/© 2019 Elsevier B.V. All rights reserved.
Y.-W. Li et al. / Powder Technology 356 (2019) 60–72 61

investigate the breakage behaviour of particles, especially in was presented, which took operational parameters as inputs and
geomechanical applications. For instance, the breakage of the sand par- could predict the output performance [36]. The breakage behaviour of
ticle was numerically investigated by adopting BPM, including the frac- coals in the industrial sieving crusher and ordinary teeth roll crusher
ture of individual sand particles [13], the high-pressure triaxial tests on was compared based on industrial scale tests, in which both the con-
crushable sand material [14] and the particle-scale mechanics of sand sumed energy and the product size distribution were analysed [7]. Be-
crushing in compression and ripping [15]. The evolution of cracks sides these empirical models and experimental studies, the particle
through the granular recycled construction and demolition materials flows and breakage in both single-roll and double-roll crushers were
was clarified and the breakage energy was accurately predicted by numerically studied based on DEM using a breakage model [1].
DEM-BPM [16]. The influence of the primary particle size distributions DEM is becoming a popular method for the study of crushing pro-
in core-shell agglomerates on the breakage behaviour under uniaxial cesses. However, DEM studies about the crushing mechanism of parti-
compressive load was simulated based on BPM [17]. The mechanical be- cles in different types of crushers were commonly conducted without
haviour of transversely isotropic rock was simulated using bonded- a detailed discussion of the reliability of the models [1,2,4]. There were
particle DEM model with embedded smooth-joint contacts [18]. Addi- only a small amount of validation studies of the DEM model of crushers,
tionally, DEM combining BMP was applied to simulate the impact where a direct comparison between DEM simulations and correspond-
breakage of the rice kernel and the breakage characteristic of the ellip- ing experimental results of the crushing process was made [23,37].
soidal agglomerate was revealed [19]. Besides the crushable granular Moreover, the crushing mechanism of the TDRC under different
materials, BPM was also used to model the crushing process for the operational conditions and structural parameters is still lack of
comminution machines. The fracture behaviour of a single spherical understanding.
and cubic rock in a laboratory jaw crusher was numerically studied In this work, the DEM model of the crushing process of TDRC is
based on BPM [20]. DEM simulation of the breakage of bonded agglom- established using the BPM. The corresponding laboratory-scale experi-
erates in a ball mill was investigated, where the agglomerates were con- ments with coal particles are conducted using a specially designed
structed with BPM [21,22]. Similarly, the breakage of rock particles in a TDRC model. DEM simulation results and experimental data are com-
cone crusher was simulated based on DEM-BPM [23]. pared quantitatively under a range of operating conditions. Overall
PRM, which is initially introduced by Cleary [24], is another alterna- good agreement is observed in the crushing performance quantities.
tive approach for simulating the breakage of particles. For example, a Then, comprehensive simulations of the crushing process are per-
DEM model of cone crusher was explored using PRM to predict the formed to provide a fundamental understanding of the effects of opera-
crushing performance and to represent the size reduction of rocks tional and structural parameters on the crushing performance. It is
[25]. Similarly, DEM simulation of particle flow and compression break- expected that this work could provide a concrete base for further devel-
age of non-spherical particles passing through an industrial scale cone opment and study of TDRC based on the numerical method.
crusher were studied in which PRM was used in the DEM particle break-
age model [4,26]. Moreover, numerical works presented in the refer- 2. Laboratory device
ences [1, 2] applied DEM with PRM to study a broad range of existing
compression and impact based crushers to understand machine level A specially designed laboratory-scale TDRC was constructed,
operation. From the simulations, the evaluation of powder, product allowing the observation of the crushing process and precise control
size, throughput rate and crusher wear were made. More recently, a of the rotation speed of rolls. Fig. 1 shows the schematic of the model
new DEM replacement model was proposed [27] based on a discrete used in DEM simulations and a photo of the laboratory-scale crusher
fragmentation multigenerational approach with a breakage criterion model used in the experiment. The laboratory crusher mainly consists
[28], which could simulate sequential breakage and better represent of a frame supporting the device, a reducer (with a motor) providing ro-
the breakage process. tation force, a coupling connecting with the driving shaft, two synchro-
Compared to the other type of crushers, the crushing mechanism of mesh gears passing rotation motion, two oppositely rotating toothed
toothed roll crusher is far less understood. There are only limited studies rolls crushing particulate materials, four bearing supports holding the
about the crushing process of roll crushers in the literature. For two rolls and a front baffle made of plexiglass. The rotation speed of
predicting the product size distribution after breakage, a breakage the rolls is controlled by a variable-frequency drive (VFD) connected
model of particles in a smooth double-roll crusher was developed, to the motor. A feeding bin placed on the top of the crushing chamber
based on a framework proposed decade ago. The breakage was pre- is used for storing coal ore particles. A collecting bin is placed under-
dicted based on an empirical relationship between particle properties neath the crushing chamber to collect the crushed particles. Four uni-
and breakage parameters [29]. After that, a combined breakage- versal wheels (with locks) are installed at the bottom of the fame for
agglomeration model was proposed to represent the particle behaviour the convenience of moving the crusher.
and to predict the product size distribution of a double-roll coal crusher. Fig. 2 presents the schematic of the crushing toothed-ring and driv-
The model parameters were estimated from experimental data of ing shaft of the TDRC. There are four teeth evenly distributed around the
single-size crushing tests [30]. Similarly, a mathematical model of an in-
dustrial double-roll crusher was provided and validated in terms of the
crushing performance, which was originally proposed for decades and
the breakage parameters estimation was based on industrial data [31].
In addition, for safe and reliable operation of the smooth double-roll
crusher, the influence of nip conditions, rotation speed and feed mate-
rial characteristics on the crushing and bearing force dynamics were in-
vestigated [32]. To overcome the disadvantage of low reduction ratios of
conventional roll crushers (typically limited to about 3:1), a non-
cylindrical roll crusher was developed in which the reduction ratio can
be over 10:1 from the tests with a model prototype [33]. For hard rock
conditions, the significant high energy efficiency of high pressure grind-
ing roll (HPGR) crusher was confirmed based on the industry applica-
tion data [34]. Furthermore, the specific energy requirement of an
HPGR was estimated from laboratory-derived ore characterisation Fig. 1. Configuration of the TDRC: (a) schematic of the model used in DEM simulations,
data [35]. Recently, a transient dynamic model of the HPGR crusher (b) a photo of the laboratory-scale TDRC used in experiments.
62 Y.-W. Li et al. / Powder Technology 356 (2019) 60–72

Fig. 2. Schematic of the main parts of TDRC: (a) crushing toothed-ring, (b) driving shaft.

crushing ring. These teeth have steep moderately flat leading faces, where Wa is the average diameter of coal particle, H is the average mea-
highly curved trailing faces and a relatively sharp tip at the outer end. sured height, L is the average measured length and D is the average
There is a convex structure like a flat key in the shaft hole of the crushing measured width. Due to the limitation of reduction ratio and size of
ring, while there are corresponding six evenly distributed concave feeding bin, the average diameters of selected feeding coal particles
structures like keyways on the shafts. The crushing toothed-rings can are around 120–150 mm (listed in Table 1). To reduce the error caused
be precisely installed on the shafts and fixed by using sleeves, plain by the particle geometry, the shape and size of the selected feeding coal
washers and round nuts along the axis. particles are as similar as possible. Fig. 5 presents the crushed coal par-
Table 1 gives a summary of the geometrical and operational param- ticles in an experiment, in which case the parallel-tooth-roll with a ro-
eters used in the experiments. There are four (three) crushing toothed- tation speed of 50 rpm was applied. The particle size of the crushed
rings installed on the driving (driven) shaft. The inner diameter (shaft coal with an average diameter over 15 mm was measured manually,
hole of the crushing ring) and outer diameter (top circle of the crushing while that of crushed particles smaller than 15 mm were sieved by
ring) are 50 mm and 160 mm, respectively. The thickness of the ring is using the standard screens.
20 mm and the distance between two rings along the axis is 30 mm. The
centre distance between two rolls is 185 mm, and the rotation speeds
are 50 rpm, 100 rpm, 150 rpm and 200 rpm in the experiments. 3. Numerical model and simulation conditions
Based on the structures of crushing toothed-ring and shafts, the
relative installation angle between two crushing toothed-ring can be 3.1. Discrete element method
adjusted conveniently. There are three different structures of toothed-
rolls used in this work, as shown in Fig. 3. Fig. 3a shows the structure The typical Hertz-Mindlin particle contact model is adopted in this
of parallel-tooth-roll, where the tips of crushing toothed rings on both work. As shown in Fig. 6, kn and dn are the normal stiffness and normal
the driving and driven shafts are aligned with each other. For the struc- damping, kt and dt are the tangential stiffness and tangential damping,
ture of spiral-tooth-roll, the relative installation angle between two ad- and kr and dr are the rolling stiffness and rolling damping. The contact
jacent crushing rings is the same (60°), as shown in Fig. 3b. For the model considers the force of gravity and the normal and tangential
structure of staggered-tooth-roll (Fig. 3c), the first and third crushing forces acting on a particle. In addition, two torques produced by a tan-
rings on the driven (left) shaft are aligned, while the second ring is gential force and a rolling friction force are also considered to act on
installed with a different installation angle with a value of 60°. Similarly, the particles. In the present model, the motion of individual particles
the first and third crushing rings on the driving (right) shaft are aligned, is calculated by Newton's laws of motion, the translational and rota-
while the second and fourth rings are installed with the same installa- tional motions of particle i are, respectively, determined by [38,39]:
tion angle of 60° relative to the other two rings.
Fig. 4 shows the coal ore particles used in the experiments and the
measurement of the particle size. For the large single feeding coal parti- dV i X i n

mi ¼ mi g þ F n;ij þ F t;ij ð2Þ
cle, an average diameter is used to define the particle size, which is de- dt j¼1
termined by the average values of height, length and width of the
particle:
dωi X i  
n
HþLþD Ii ¼ T t;ij þ T r;ij ð3Þ
Wa ¼ ð1Þ dt
3 j¼1

Table 1
The geometrical and operational parameters used in the experiments.
where m and I are the mass and moment of inertia of particle i, ni is the
total number of particles in contact with particle i, Vi is translational ve-
Parameter Value locity, ω is angular velocity, t is time, and g is the acceleration of gravity.
Number of crushing toothed-ring 4 (driving shaft), 3 (driven shaft) The interaction forces and torques acting on particle i include the nor-
Diameter of crushing toothed-ring (mm) 50 (inner diameter), 160 (outer mal contact force, the tangential contact force, the tangential torque
diameter)
and the rolling torque, which are defined as Fn,ij, Ft,ij, Tt,ij and Tr,ij respec-
Thickness of crushing toothed-ring (mm) 20
Distance between two crushing 30 (along the axis) tively. It should be noticed that both the normal and tangential force in-
toothed-ring (mm) clude elastic force and damping force, which is used for the elastic
Centre distance between two rolls (mm) 185 deformation and the viscous dissipation between two contacting parti-
Feeding particle size (mm) 120–150 cles, respectively. The detailed calculation can be found in references
Rotation speed of rolls n (rpm) 50, 100, 150, 200
[38, 40, 41].
Y.-W. Li et al. / Powder Technology 356 (2019) 60–72 63

Fig. 3. Three different structures of toothed rolls: (a) parallel-tooth-roll, (b) spiral-tooth-roll and (c) staggered-tooth-roll.

3.2. Bonded particle model that act in the normal and in the shear direction vary due to the relatively
normal and shear motion (displacement and rotation) of the sphere.
The bonded particle model (BPM), originally proposed by Potyondy The criterion of rupture of the bond is defined by comparing the crit-
and Cundall [42], is used in this work to model the breakable coal parti- ical normal stress σc and the critical shear stress τc, which are model pa-
cles. The BPM is based on bonding or glueing a cluster of primary parti- rameters, to the maximum values reached during a collision,
cles together using imaginary parallel bonds between each particle in determined by
contact, in which the connections can break when stressed beyond a
critical level. The primary particles bonded together can be called frac- ! n;total
F 2M nb
tion particles and the cluster created is defined as a meta-particle. Be- σ¼ b
þ R bσ c
Ab J b
fore bond formation and after bond breakage the particles interact ! s;total
ð6Þ
according to the Hertz-Mindlin no-slip contact model. The bonds are F 2Msb
τ¼ b
þ R bτc
formed between particles in contact at a pre-set time. When the con- Ab J b
nections are created, the forces and torques between the fraction parti-
cles are calculated by expressions representing the bonds in BPM Either when σ is greater than σc or when τ is greater than τc, the par-
[12,23,37,42]. The forces Fb and torques Mb that act on a single bond allel contact breaks.
are presented in Fig. 7, given by
3.3. Simulation conditions
!n n
ΔF b ¼ −kb  Ab  vn  Δt
!s s A CAD model of the TDRC used in the simulations was constructed,
Δ F b ¼ −kb  Ab  vt  Δt
s ð4Þ which matches the main structure of the laboratory-scale device
ΔMnb ¼ −kb  J  ωn  Δt shown in Fig. 1. The properties and physical constants of particles and
n J
ΔMsb ¼ −kb   ωt  Δt crusher model used in the simulations are listed in Table 2 based on
2
the previous relevant works [38,39,43]. The density, Young's modulus
where, and Poisson's ratio of the particle model were 1300 kg/m3, 1.0 GPa
and 0.3, while that of the crusher model was 7861 kg/m3, 79.92 GPa
Ab ¼ π  R2b and 0.29, respectively. The coefficient of restitution and rolling friction
1 ð5Þ for both inter-particle and particle-crusher were 0.5 and 0.05, respec-
J ¼ π  R4b tively. The coefficient of static friction for inter-particle was 0.6, while
2
the value for particle-crusher was 0.4. Individual meta-particle
and knb, ksb are the normal and shear bond stiffness, vn, vt, ωn and ωt are the modelled in BPM is illustrated in Fig. 8. The shape of each modelled
normal and tangential velocities, and the normal and tangential angular meta-particle mimics the actual shape of a coal ore particle obtained
velocities, respectively. Ab and J are the cross-sectional areas of the bond by 3D laser scanning, generating an STL (stereolithography) file,
and the polar moment of inertia. Rb is the bond radius, which depends on which is used as a selection space as seen in Fig. 8a. The position and ra-
the diameter of the smallest sphere in contact. The forces and torques dius of each fraction particle within the selection space is exported to

Fig. 4. Coal particles used in the experiment: (a) a photo of the feeding coal particles, (b) measurement of the particle size.
64 Y.-W. Li et al. / Powder Technology 356 (2019) 60–72

Fig. 5. Photos of the crushed coal particles. The scale at the bottom is centimetre.

form a meta-particle cluster model that will be used in the simulation. calculated by EDEM. The simulations were conducted on a desktop
Fig. 8b shows the virtual coal ore particle modelled using 5697 fraction computer whose CPU, RAM and hard disk are 3.4 GHz, 8 GB and 2 TB,
spherical particles, whose sizes followed a lognormal distribution [12], respectively.
with a minimum diameter of 2.03 mm and a maximum diameter of
10.53 mm, with a mean size of 4.69 mm. The number of fraction parti- 4. Results and discussion
cles filled into the virtual particle template is directly related to the
size of the fraction particle. And, the DEM simulation time step is also The DEM model of the crushing process of TDRC is established and a
calculated based on the smallest size of the particle. A smaller simula- series of simulations are conducted with the same conditions to those of
tion time step and an increasing number of breakage generations laboratory experiments. The size fractions of the crushed particles are
could be obtained with a further reduced size of fraction particle, both compared between the results from experiments and DEM simulations,
of which would severely increase the computational cost. Thus, the in which a single-feeding particle is adopted. The effects of the rotation
proper size of the primary particle should be chosen for achieving a bal- speed of the rolls and structures of the toothed rolls on the crushing pro-
ance between accuracy and computational efficiency. cesses are discussed based on DEM simulations with multi-feeding
Table 2 also presents the micro-parameters of the BPM model used particles.
in the simulations, which are calculated or chosen based on the relevant
works [42,44–46]. In the present work, the normal stiffness and shear 4.1. DEM results of the crushing process of TDRC
stiffness could be calculated based on the principle of BPM and the ma-
terial properties of coal particles, whose values were 200 GN/m3 and Fig. 9 presents snapshots of the crushing process of coal particles in
100 GN/m3, respectively. Here, the stiffness ratio was set to 2.0 accord- DEM simulation, in which case the parallel-tooth-roll with a rotation
ing to the results presented in the literature [44,47]. In addition, accord- speed of 50 rpm is adopted and particles are displayed as clustered
ing to the mechanical properties of coal, the compressive strength of bonds and coloured by the normal force of the bonding beam. In
coal particles is much larger than the tensile strength, and therefore Fig. 9a, virtual coal particles are fed into the feeding bin of TDRC. Then,
the fragmentation is usually considered to be initiated by tensile failure the particles fall freely down into the crushing chamber to be crushed,
[46,48]. Thus, both values of the critical normal stress and the critical which mainly consists of the two oppositely rotating toothed rolls.
shear stress were set to 3.0 MPa based on the experimental data of the Figs. 9b-d show the crushing process of a virtual coal particle at the
coal tensile strength [45,46]. The radius of Bonding was set to 2.5 mm time of 1.76 s to 1.78 s. When the particle is gripped by the teeth on
which was closed to the mean size of the fraction spherical particles. the rolls of each side, the teeth will dig into it and a combination
The above BPM parameters were calibrated by comparing the simula- crushing effect of compression and ripping will act on the particle.
tion results with the corresponding experiments, which should be rea- Once the normal stress or the shear stress acting on the bonding
sonable for simulating the crushing process of TDRC. In this work, the beams between the fraction particles exceeds the critical values, the
commercial software package EDEM (version 2.7) was employed to bonding beams will break and the virtual coal particle will be broken.
perform numerical simulations using the BPM. The simulation time Then, the broken pieces drop into the collecting bin. It should be noticed
step was set to be 1.0 × 10−6 s smaller than the Rayleigh time step that the broken pieces collide with the bottom of the collection bin with
a relatively high velocity because of the acceleration of the rotation rolls,
which may cause a further secondary crushing. During the crushing pro-
cess, the number of broken bonds is recorded, which can be used as a

Fig. 6. Contact forces between two contacting particles [38]. Fig. 7. Schematic representation of two particles bonded together with a cylindrical beam.
Y.-W. Li et al. / Powder Technology 356 (2019) 60–72 65

Table 2 simulations and tests are calculated by the ratio of the difference be-
Model parameters used in the DEM simulations. tween simulations and tests to the value of the tests, as shown in
Parameters Value Fig. 10b. Generally, the trends of the results from simulations and
DEM material properties
the corresponding experiments are very similar. For the results of
Particle Crusher the experiment with the rotation speed of 50 rpm, the proportion of
Density (kg/m3) 1300 7861 each size of crushed particles shows an increasing trend (except for
Young's modulus (GPa) 1.0 79.92 the largest size fraction). The proportions of first two size fractions
Poisson's ratio 0.3 0.29
of particles are very small, whose values are 4% (b10 mm) and 6%
Inter-particle Particle-crusher
Coefficient of restitution 0.5 0.5 (10–15 mm), respectively. For the intermediate two size fractions,
Coefficient of static friction 0.6 0.4 the mass proportions increase respectively to 11% (15–25 mm) and
Coefficient of rolling friction 0.05 0.05 17% (25–35 mm). The last two size fractions of particles are the
BPM parameters main crushing products obtained in the test, whose values are 25%
Normal stiffness (GN/m3) 200 (35–45 mm) and 33% (45–55 mm), respectively. The relative error be-
Shear stiffness (GN/m3) 100 tween the simulation and test for size fraction of b10 mm is −25%,
Critical normal stress (MPa) 3.0 which means the mass proportion of this size fraction is under-
Critical shear stress (MPa) 3.0
predicted by simulations. For the size fraction of 10–15 mm, the rela-
Radius of Bonding (mm) 2.5
tive error is about 17%, which means the result is over-predicted. The
relative errors of the middle two size fractions of particles are both
basis for judging the crushing performance. The size fraction of the under-predicted, whose values are all around −18%. The largest rela-
crushed particles can be analysed based on the crushed pieces stored tive error is observed for the size fraction of 35–45 mm with the value
in the collecting bin, and the detailed methodology for calculating the being 40%. But, for the last size fraction (45–55 mm), the relative error
size distribution of discharged clusters can be found in the reference is only about −9%. In addition to human errors during the crushing
[23]. It should be noticed that particle shape is one of the governing fac- experiments, the main reasons causing the mismatch between the
tors affecting the fracture pattern and the probability of particle break- simulations and the corresponding tests should be related to the dif-
age in a crusher [16]. The particle tensile strength and breakage ference of geometric and mechanical properties of the particle model
energy, micro- and macro-mechanical behaviour of the particle during compared to the realistic coal particle. For instance, the shape of the
the breakage of realistic particles should be different from idealised feeding coal particles in the experiments was complicated and differ-
ones [16,49]. In the present work, the virtual particle was modelled ent to each other, while that of the particle models used in the simu-
based on the realistic shape of a coal particle, but it was still a simplified lations were the same and modelled using spherical fraction particles.
particle model formed by the spherical meta-particle cluster. Therefore, In addition, the real coal particles might be weakened by internal
this could be expected to generate significant quantitative errors in the micro-fractures, while the mechanical properties of the particle
DEM prediction of the product size distributions. model were homogeneous. Furthermore, the crushing behaviour of
particles in the simulations and the tests should be also different, dur-
4.2. Validation in terms of the size fraction of crushing product ing which the contacts between the particle and crushing teeth were
different causing the product size distribution different as well. Thus,
In the present work, three different structures of toothed rolls are it could be expected that the relative errors between the simulations
adopted to carry out the validation of the DEM model of TDRC. During and the experiments were generally random and difficult to control.
the simulations and the corresponding experiments, only one large For the experiment case with a rotation speed of 100 rpm, the mass
coal particle was fed into the crusher for each case (repeated three proportions of the first two size fractions are still small, whose values
times in experiments), due to the difficulty of manual measurement of are 6% (b10 mm) and 5% (10–15 mm), respectively. The mass propor-
the size of crushed particles in the experiments. Fig. 10 shows the prod- tion increases to 12% for the size fraction of 15–25 mm. The mass pro-
uct size distribution and relative error between the simulations and portions of the last three are all above 20%, in which the maximum
experiments for TDRC with three different roll structures, in which value is 28% for the size fraction of 35–45 mm. The corresponding values
the rotation speeds of the rolls are 50 rpm, 100 rpm, 150 rpm and of relative error for both the size fractions of b10 mm and 15–25 mm are
200 rpm, respectively. about 17%, while the former is under-predicted and the latter is over-
Figs. 10a and b show the validation results for the TDRC with predicted. For the size fraction of 10–15 mm, the largest relative error
parallel-tooth-roll. The proportion of each size of particles (Fig. 10a) is is observed with the value of 60%. But, it should be noticed that the
calculated by the ratio of the mass of one size fraction of crushed par- mass proportion of this size fraction is very small both for the experi-
ticles to the total mass of the product. And, the relative errors of the ment (5%) and simulation (8%). Hence, a small difference between ex-
mass proportion of each size fraction of product between the periment and simulation might cause a large relative error. The

Fig. 8. Illustration of creating meta-particle clusters of virtual coal ore particle with a realistic shape.
66 Y.-W. Li et al. / Powder Technology 356 (2019) 60–72

Fig. 9. Snapshots of the crushing process of virtual coal particles in DEM simulation (particles are displayed as clustered bonds and coloured by the normal force of the bonding beam).

relative errors for the last three size fractions are very small ranging Then, for the intermediate products (size fraction of 15–35 mm), the
from around −4% to −7%, which are under-predicted. mass proportion increases from 14% to 19%. The proportions of the
When the rotation speed of the rolls increases to 150 rpm, there are last two large products (size fraction of 35–55 mm) are all above 20%,
no large particles with the size fraction of 45–55 mm obtained both in whose values are 21% and 26%, respectively. The relative errors of the
the experiments and simulations. The mass proportions of the small mass proportion of each size fraction of product between the simula-
particles with the size fractions of b10 mm and 10–15 mm are 9% and tions and tests also fluctuate for different products. The mass proportion
10%, respectively. The crushing product mainly consists of the last of the size fraction b10 mm is under-predicted by about 29%, while that
three size fractions of particles, whose mass proportions are 25%, 21% of the size fraction of 10–15 mm is the highest and over-predicted by
and 31%, respectively. The relative error of the size fraction of about 33%. For the middle two size fractions, the relative errors are
10–15 mm is the largest with the value being 20%, and the rest values very small, which values are just about −7% and − 5%, respectively. In
are under 11%, in which the value of the size fraction of 35–45 mm is the large products, the relative error for the size fraction of 35–45 mm
just about 3%. Similar results are observed when the rotation speed of is obviously higher than that of the size fraction of 45–55 mm, which
the rolls further increases to 200 rpm. There are no large particles values are about 24% and − 8%, respectively.
(45–55 mm) in the crushing product and the size fraction of the main When the rotation speed of the rolls is 100 rpm, only the mass pro-
product is between 15 and 45 mm. The mass proportions of the first portion of size fraction b10 mm is under 10%, while the values of the rest
two small size of particles are 10% and 12%, and that of the last three size fractions are all above 10% ranging from 13% to 23%. The largest rel-
size fractions of particles are ranging from 21% to 31%. The relative er- ative error between the simulations and tests is about 23% for the size
rors of all the size fraction of particles are around 10%, in which the larg- fraction of 15–25 mm, and the smallest error is only about −6% for
est value is only about 14% for the size fraction of 15–25 mm. the size fraction of 45–55 mm. The values of relative error for the rest
The crushing results of the TDRC with spiral-tooth-roll are shown in size fractions are around 10%. When the rotation speed of the rolls in-
Figs. 10c and d. Again, the overall trends of the results from simulations creases to 150 rpm, there are no large particles with the size fraction
and the corresponding experiments are very similar. When the rotation of 45–55 mm in the crushing products. The mass proportions of all the
speed of the rolls is 50 rpm in the experiment, the mass proportion of rest size fractions are above 10%, in which the proportions of the first
each size fraction of particles shows a linear increasing trend. For the two size fractions are 11% and 18%, while that of the last three size frac-
first two fine products (size fractions under 15 mm), the mass propor- tions are ranging from 21% to 26%. The largest relative error is about 22%
tions are relatively low, which values are 7% and 9%, respectively. observed for the size fraction of 10–15 mm, while the smallest error is
Y.-W. Li et al. / Powder Technology 356 (2019) 60–72 67

(a) 40 (b)
Simulations Experiments
60 n=50 rpm

Proportion of each size particles/%


n=50 rpm n=50 rpm
35 n=100 rpm
n=100 rpm n=100 rpm 50
n=150 rpm
n=150 rpm n=150 rpm
30 40 n=200 rpm
n=200 rpm n=200 rpm

Relative error/%
30
25
20
20
10
15 0
-10
10
-20
5 -30
0 -40
<10 10-15 15-25 25-35 35-45 45-55 <10 10-15 15-25 25-35 35-45 45-55
Size fraction of crushed particles/mm Size fraction of crushed particles/mm

(c) (d) 60
40
Simulations Experiments
Propotion of each size particles/%

n=50 rpm n=50 rpm n=50 rpm


35 50
n=100 rpm n=100 rpm n=100 rpm
n=150 rpm n=150 rpm 40 n=150 rpm
30 n=200 rpm
n=200 rpm n=200 rpm

Relative error/%
30
25
20
20 10
15 0
-10
10
-20
5
-30
0 -40
<10 10-15 15-25 25-35 35-45 45-55 <10 10-15 15-25 25-35 35-45 45-55
Size fraction of crushed particles/mm Size fraction of crushed particles/mm

(e) (f)
40 60
Simulations Experiments n=50 rpm
Propotion of each size particles/%

n=50 rpm n=50 rpm 50


35 n=100 rpm
n=100 rpm n=100 rpm
40 n=150 rpm
30 n=150 rpm n=150 rpm
n=200 rpm
n=200 rpm n=200 rpm
Relative error/%

30
25 20
20 10

15 0
-10
10
-20
5 -30
0 -40
<10 10-15 15-25 25-35 35-45 45-55 <10 10-15 15-25 25-35 35-45 45-55
Size fraction of crushed particles/mm Size fraction of crushed particles/mm

Fig. 10. The product size distribution (left) and relative error between the simulations and experiments (right): the TDRC with parallel-tooth-roll (a and b), spiral-tooth-roll (c and d) and
staggered-tooth-roll (e and f).

only about −4% for the size fraction of 25–35 mm. The values of the rel- The relative errors of the rest size fractions are under 10%,
ative errors of the rest size fractions are ranging from 9% to 14%. Simi- among which the smallest value is just about 5% for the size fraction
larly, when the rotation speed of rolls further increases to 200 rpm, of 25–35 mm.
there is also no products with the size fraction of 45–55 mm and the The validation results for the TDRC with staggered-tooth-roll are
mass proportion of the size fraction of b10 mm is 10%. The proportions shown in Figs. 10e and f. When the rotation speed of the rolls is
of the rest size fractions evenly distribute between 19%–24%. The pro- 50 rpm in the experiment, the mass proportions of the first five size
portion of the size fraction of b10 mm is over-predicted by 20% and fractions of products show an increasing trend, whose values increase
that of the size fraction of 35–45 mm is under-predicted by about 16%. from 5% for the size fraction of b10 mm to 29% for the size fraction of
68 Y.-W. Li et al. / Powder Technology 356 (2019) 60–72

35–45 mm. The proportion of the largest size of the product is also 29%. rolls is 50 rpm (case 1), only a small part of fine products whose size
The mass proportion of the fine product that b10 mm is under- fraction is under 15 mm is obtained. The mass proportions of the prod-
predicted, whose relative error is about −20%. For the product between ucts with the size fraction of b10 mm and 10–15 mm are b4% and 8%, re-
10 and 15 mm, the largest error is observed, and the proportion is over- spectively. There are total about 24% of the intermediate size of particles
predicted by about 29%. The relative error decreases to about −17% for in the products (15–35 mm), in which about 10% is the particles with
the size fraction of 15–25 mm. For the rest three size fractions, the rela- size fraction of 15–25 mm and about 14% is the particles with size frac-
tive errors are small, which values are about −6%, 3% and 7%, respec- tion of 25–35 mm. The last two large size fractions of particles are the
tively. When the rotation speed of the rolls is 100 rpm, as the size main products, whose mass proportions are almost 65% in total. The
fraction increases the mass proportions of each size fraction of products maximum mass proportion is about 35% for the size fraction of 35–
gradually increase from 6% to 26%. The relative error for the first size 45 mm, and the mass proportion for the last size fraction (45–55 mm)
fractions is the largest, which value is about −33.3%. The mass propor- is above 30%. When the rotation speed of rolls is 100 rpm (case 2), the
tions for the size fractions of 35–45 mm is over-predicted by about 19%. mass proportions of the first two size fractions of products (b10 mm
For the rest size fractions, the values of relative error are under 10%, in and 10–15 mm) increase to about 5% and 9%, respectively. And, the in-
which the error is only about −4% for the size fraction of 45–55 mm. termediate two size fractions of products also increase, which values are
There are no products with the size fraction of 45–55 mm when the ro- about 14% and 22%, respectively. In contrast, the mass proportions of the
tation speed of the rolls is higher than 150 rpm. In the experiment with last two size fractions of products show a decreasing trend, whose
the rotation speed of 150 rpm, the mass proportion for the size fraction values drop to about 26% and 24%, respectively. When the rotation
of b10 mm is still low, whose value is 9%. The proportion increases to speed increases to 150 rpm (case 3), the mass proportion of each size
15% for the size fraction of 10–15 mm and that of the rest three size frac- fraction of products shows a similar trend with the simulation case
tions are all above 20%, whose values are 25%, 20% and 28%, respectively. with the rotation speed of 100 rpm. The mass proportions of the first
The relative errors of the first two size fractions are above 20%, in which two size fractions further rise to about 9% and 12%, respectively. For
the error for the size fraction of b10 mm is about −22%. The smallest the intermediate two size fractions, the mass proportions go up to
error is only about 4% for the size fraction of 35–45 mm, and the values above 23% for both the size fractions of 15–25 mm and 25–35 mm. Sim-
of the middle two size fractions are −12% and 15%, respectively. In the ilarly, the mass proportions for the last two size fractions further drop to
case with the rotation speed of 200 rpm, there are more products of about 16% for size fraction of 35–45 mm and 17% for size fraction of 45–
small sizes in the experiment, whose mass proportions are 12% and 55 mm. When the rotation speed further increases to 200 rpm (case
19% for the first two size fractions. The mass proportions of the rest 4) or 250 rpm (case 5), the overall mass proportions of each size frac-
three size fractions are all above 20%. The mass proportion of the first tions are similar to that of case 3. Compared to case 3 only slight changes
size fraction is under-predicted by 25%, which is the largest relative in the mass proportions of each size fraction can be found. For case 4, the
error in the case. The relative errors for the rest size fractions are proportions of both the first two size fractions are only about 1% higher
under 10%, in which the minimum is only about −4% for the size frac- than that of the case 3. The proportion of the size fraction of 15–25 mm
tion of 35–45 mm. increases by only about 3%, while the proportion of the size fraction of
As noted above, the size fractions of crushing products predicted by 25–35 mm is also about 23%. For the last two size fractions, the mass
the simulations are generally consistent with the corresponding exper- proportions are both about 2% lower than that of the case 3. For the
imental results. DEM model of the crushing process of the TDRC based last case (250 rpm), the proportions for the three groups of size frac-
on the BPM can be potentially used to study the crushing process of par- tions are about 8% and 14%, 24% and 24%, 15% and 15%, respectively.
ticulate materials. The mismatch between simulations and tests mainly All the results are similar to those of cases 3 and 4.
comes from the errors of physical parameters of particles, the random The above results obtained by the simulations could be preliminarily
collision behaviours between the teeth and particles, the difference of explained based on the crushing principle of TDRC. It should be noticed
geometric and mechanical properties of the real coal particles. More- that each toothed ring of the TDRC only has four crushing teeth evenly
over, the validation results also illustrated that the breakage behaviour distributed on the crushing ring. When the rolls' speed was relatively
of coal particle during the crushing should be mainly induced by tensile low, the contact between the feeding particle and the toothed ring
strength, which was in agreement with the micromechanical properties could be reduced during this short time event. Moreover, a working
of coal presented in references [44–46]. gap would be formed between the two parallel-tooth-rolls at a certain
time during the crushing process, into which the particle or the large
4.3. The effect of rotation speed and structure of rolls on the crushing size of fractions could be trapped but without any effective contact
process with the crushing teeth. Then, the feeding particle would not be broken
completely causing more large size fractions and less small size fractions
The rotation speed and the structure of crushing rolls of the TDRC are in the product. On the contrary, when the rolls' speed increased, the nip-
two critical factors for the crushing performance. Generally, it is known ping condition between the particle and the rolls could be improved
that the rolls' speed has a direct influence on the throughput, and the [32] causing relatively less large size fractions and more small size frac-
structure of the roll has a significant effect on crushing behaviour. It tions observed in the simulation.
would be helpful for understanding the fundamental mechanism to Fig. 11b shows the total number of broken bonds during the
study the crushing process at particle-scale. So, numerical studies crushing process of the TDRC with parallel-tooth-roll. The feeding parti-
based on the above validated DEM model of the TDRC are conducted cles modelled by bonded fraction particles are crushed into pieces under
to investigate the effect of the two factors on the crushing process of the crushing forces acting by the teeth. The number of broken bonds is
coal particles. In the simulation works, all the three different structures recorded during the crushing process until the simulation is finished.
of crushing rolls are adopted and ten virtual coal particles fed sequen- Generally, the total number of broken bonds increases gradually during
tially into the crusher in each simulation case. Fig. 11 presents the size the crushing process, and the final values show an increasing trend as
fraction of crushing product and the number of broken bonds during the rotation speed increases. All the crushing processes of the five
three groups of crushing processes at different rotation speeds of rolls cases with different rotation speed are ended at around 20s, which
for the TDRC with parallel-tooth-roll, spiral-tooth-roll and staggered- may be slightly different due to the random motion of feeding particles
tooth-roll. in the crusher and the complicated collision behaviour between parti-
The simulation results shown in Fig. 11a is the first group of the cles and crushing teeth. For case 1 (the rotation speed of 50 rpm), the
product size distribution for the TDRC with the parallel-tooth-roll final total number of broken bonds is 21,185. When the speed increases
under different rotation speeds of the roll. When the rotation speed of to 100 rpm (case 2), the increasing speed of the number is faster than in
Y.-W. Li et al. / Powder Technology 356 (2019) 60–72 69

(a) (b)
40 55

Propotion of each size particles/%

Total number of broken bonds/×103


n=50 rpm n=200 rpm 50 n=50 rpm
35 n=100 rpm n=250 rpm n=100 rpm
45
n=150 rpm n=150 rpm
30 40 n=200 rpm
25 35 n=250 rpm
30
20
25
15 20
10 15
10
5
5
0 0
<10 10-15 15-25 25-35 35-45 45-55 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Size fraction of crushed particles/mm t/s

(c) (d)
40 55

Total number of broken bonds/×103


Propotion of each size particles/%

n=50 rpm n=200 rpm 50


35
n=100 rpm n=250 rpm 45
30 n=150 rpm
40
25 35
30
20
25
15 20 n=50 rpm
10 15 n=100 rpm
n=150 rpm
10 n=200 rpm
5
5 n=250 rpm
0 0
<10 10-15 15-25 25-35 35-45 45-55 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Size fraction of crushed particles/mm t/s

(e) (f)
40 55
Total number of broken bonds/×103

n=50 rpm
Propotion of each size particles/%

n=50 rpm 50
35 n=100 rpm n=100 rpm
45
n=150 rpm n=150 rpm
30 40 n=200 rpm
n=200 rpm
n=250 rpm 35 n=250 rpm
25
30
20
25
15 20
10 15
10
5
5
0 0
<10 10-15 15-25 25-35 35-45 45-55 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Size fraction of crushed particles/mm t/s

Fig. 11. The size distribution of the crushing product (left) and the total number of broken bonds during crushing processes (right) at different rotation speeds of rolls for the TDRC with
parallel-tooth-roll (a and b), spiral-tooth-roll (c and d) and staggered-tooth-roll (e and f).

case 1, whose final number is 24719 (3534 higher than that of case 1). certain type of loading [42,50]. According to the crack propagation the-
Similarly, the final number of case 3 (rotation speed is 150 rpm) in- ory (Griffith theory), brittle materials release the strain energy mainly
crease to 28,355, which is 3636 higher than case 2. The total number by crack propagation and the ultimate strength of the material was pro-
of broken bonds increases significantly for the first three cases, resulting portional to the loading rate (equivalent to the rolls' speed here). When
from the increase of the fine and intermediate products and the de- the loading rate was high, more cracks were created to release the in-
crease of the large products with the increase of the rotation speed of creasing strain energy causing more fractions of small size in the prod-
rolls (see Fig. 11a). However, when the rotation speed is further in- ucts. Furthermore, it seemed that there should be a critical rolls' speed
creased to 200 rpm or 250 rpm, relatively small increases of the final (150 rpm) beyond which both the nipping behaviour and the fracture
numbers can be obtained. The final value of case 4 is 29,219, which is mechanics of materials would not be significantly improved.
only 367 higher than case 3. For case 5, the final number is 30219, Fig. 11c presents the second group of the size distribution of
which is 1000 higher than case 4. Consequently, when the rotation crushing product for the TDRC with spiral-tooth-roll at different rota-
speed is higher than 150 rpm, a further increase of the speed does not tion speeds and Fig. 11d is the corresponding total number of broken
significantly increase the number of broken bonds. The breakage of bonds during the crushing processes. One of the main difference be-
bonds could be related to the fracture dynamics of particles under a tween the parallel-tooth-roll and the spiral-tooth-roll is that the
70 Y.-W. Li et al. / Powder Technology 356 (2019) 60–72

working gap between the rolls is reduced, which means that more effec- 15–25 mm is almost the same, but the proportion of the size fraction
tive contacts between the particle or the large size of fractions and the of 25–35 mm is about 4% less than that of the case in the second
crushing teeth could be formed. Thus, it is expected that the nipping group. But for the large products, the mass proportions are about 4%
condition of particle and crushing performance with the spiral-tooth- and 8% higher than the values of the corresponding case of the second
roll could be improved. group. As the rotation speed increases to 150 rpm, the mass propor-
When the rotation speed of crushing rolls is 50 rpm, the mass pro- tions of the fine and intermediate products rise to about 7% and
portions of the first two fine fractions of particles are about 5% and 18%, 22% and 23%, respectively. The proportions of the large products
12%, respectively, which are higher than that of the cases with further drop to about 15% and 14%, respectively. Similarly, the mass
parallel-tooth-roll (about 4% and 8%). The mass proportions for the in- proportions of the fine and intermediate products are slightly less,
termediate two size fractions are also higher than that of the cases while the proportions of large products are a little higher than the
with parallel-tooth-roll (about 10% and 14%), whose values are about values of the case in the second group. When the rotation speeds fur-
14% and 18%, respectively. The last two large size particles are still the ther increase to 200 rpm and 250 rpm, there are also no obvious
main crushing products, which mass proportions are about 26% and changes in the mass proportions of each size fraction of products.
24%, respectively. But, the large size of products is relatively less than Moreover, similar results can be obtained compared to that of the
the first group of cases with parallel-tooth-roll (about 35% and 30%). cases in the second group.
When the rotation speed increases to 100 rpm, both the mass propor- The total numbers of broken bonds during the crushing process of
tions of the first two (fine product) and the middle two (intermediate the TDRC with staggered-tooth-roll at different rotation speeds
product) size fractions increase, while that of the last two size fractions displayed in Fig. 11f shows a generally similar trend to Figs. 11b and d,
(large product) decrease to about 21% and 17%, respectively. As the ro- which values are stable at about 23 s. When the rotation speed increases
tation speed further increases to 150 rpm, the mass proportions of the from 50 rpm to 100 rpm, the final total number of broken bonds in-
fine and intermediate products increase continuously, while that of crease from 26,056 to 30,277. As the rotation speed increase to
the large products further fall to about 9% and 8%, respectively. Again, 150 rpm, the final total number of broken bonds goes up to 34,600.
when the rotation speed further increases to 200 rpm or 250 rpm, But, when the rotation speed further increases to 200 rpm or 250 rpm,
only minor changes of the mass proportions for each size fractions are the final numbers slowly rise to 27,114 or 38,312, respectively. The
observed. Overall, the changes in the mass proportion for each size frac- final numbers are significantly smaller than the results shown in
tion of the product are similar to that of the last group of cases with Fig. 11d, as well as the increase between two adjacent cases. But, the
parallel-tooth-roll. However, there are generally more fine and interme- values are comparatively larger than that of the results shown in
diate products and less large products obtained in each case. Fig. 11b. Based on the above discussion, it could be seen that the
With the increase of rotation speed of the rolls, the total number of crushing performance was also improved compared to the parallel-
broken bonds for each case also shows an increasing trend, as shown tooth-roll because of a reasonable working gap between the rolls and
in Fig. 11d. The crushing processes finish at around 23 s being different a better nipping behaviour of particles in TDRC with a staggered-
from the cases with parallel-tooth-roll, related to the different structure tooth-roll.
of the crushing rolls and the different crushing acting between the tooth In summary, both the rolls' speed and the structure of crushing rolls
and particles. Generally, the final numbers of the broken bonds are of the TDRC have a significant effect on the crushing process. When the
clearly higher than that of the last group of cases. When the rotation rotation speed of the crushing rolls is under 150 rpm, the mass propor-
speed is 50 rpm, the final number is 31403, which is 10,218 higher tions of fine and intermediate products show an overall increasing trend
than the case with parallel-tooth-roll. When the rotation speed in- with the increase of the rotation speed. However, the proportions of the
creases to 100 rpm, the final number jumps to 41,406, which is 10,003 large products show an opposite trend, whose values decrease gradu-
higher than case 1 (50 rpm). The increase between the two cases is sig- ally with the increase of the rotation speed. Moreover, when the rota-
nificantly higher than that of the first two cases in the last group. As the tion speeds further rise to 200 rpm and 250 rpm, the mass proportion
rotation speed increases to 150 rpm, the final number further increases of each size fraction will fluctuate slightly for the cases with different
by 10,022. Similarly, when the rotation speed further increases to structures of crushing rolls. Generally, there are more products of fine
200 rpm or 250 rpm, the final total number of broken bonds only in- and intermediate sizes, but fewer products of large sizes obtained
creases by 856 or 1397, respectively. So, when the rotation speed is when the structure of spiral-tooth-roll is adopted compared to that of
higher than 150 rpm, a further increase of the speed does not signifi- the other two structures. The changes in the total numbers of broken
cantly increase the number of broken bonds. Furthermore, it is clear bonds during the crushing processes are overall consistent with the
that the structure of the rolls should have a significant effect on crushing product size distributions of the three groups of cases with different
behaviour. And, compared to the parallel-tooth-roll, the feeding particle structures of rolls. The increase of the broken number means that
would tend to more likely be crushed into relatively small size fractions. there are more small pieces of crushed particles produced during the
Fig. 11e presents the third group of the product size distribution for crushing processes. When the rotation speed of crushing rolls is under
the TDRC with staggered-tooth-roll at different rotation speeds. When 150 rpm, the final number of broken bonds increase rapidly. But the
the rotation speed of rolls is 50 rpm, the mass proportions of the fine growth will slow down when the rotation speed is higher than
products (first two size fractions) are also small being about 4% and 150 rpm. For the three groups of cases with different structures of
10%, respectively. For the intermediate products, the mass proportions rolls, both the total number of broken bonds and the increase between
are about 10% and 15%. The last two size fractions of large particles are two adjacent cases are the largest for the crushing processes with
also the main products, whose mass proportions are all about 30%. spiral-tooth-roll, while the values of the crushing processes
Overall, the product size distribution is close to that of the case with with staggered-tooth-roll are relatively higher than that of cases with
parallel-tooth-roll. And, the mass proportions of the fine and interme- parallel-tooth-roll.
diate products are lower than that of the case with spiral-tooth-roll, In addition, the numerical results obtained in the present work could
but it is opposite for the large products. When the rotation speed in- be preliminarily explained based on both the crushing principle of TDRC
creases to 100 rpm, the mass fractions of the fine products slightly in- and the fracture dynamics of particles. The effects of rolls' speed on the
crease above 4% and 13%. For the intermediate products, the mass crushing mechanism agreed with the results reported in the works
fractions are all about 16%. But, the mass fractions drop to about 25% [32,42,50]. And, similar conclusions about the effect of rolls structure
for the large products. Similar to the first case (50 rpm), the mass pro- on the size fraction of the product could be found in reference [7]. The
portions of fine products are less than that of the corresponding fundamental findings of the present work could also be further con-
case with spiral-tooth-roll. The proportion of the size fraction of firmed by these relevant studies.
Y.-W. Li et al. / Powder Technology 356 (2019) 60–72 71

5. Conclusions [2] M.D. Sinnott, P.W. Cleary, Simulation of particle flows and breakage in crushers
using DEM: part 2 – impact crushers, Miner. Eng. 74 (2015) 163–177.
[3] M. Johansson, M. Bengtsson, M. Evertsson, E. Hulthen, A fundamental model of an
The crushing processes of the TDRC were modelled based on the industrial-scale jaw crusher, Miner. Eng. 105 (2017) 69–78.
DEM-BPM model. The corresponding laboratory-scale experiments [4] P.W. Cleary, M.D. Sinnott, R.D. Morrison, S. Cummins, G.W. Delaney, Analysis of cone
crusher performance with changes in material properties and operating conditions
with coal ore particles were conducted. The simulation and experiment using DEM, Miner. Eng. 100 (2017) 49–70.
results were compared quantitatively under a range of operating condi- [5] G.Z. Li, R. Roufail, B. Klein, L. Nordell, A. Kumar, C.B. Sun, J. Koua, Experimental eval-
tions. The effect of the rotation speed and the structure of toothed rolls uation of the conjugate anvil hammer mill - comparison of semi-confined to con-
fined particle breakage, Miner. Eng. 137 (2019) 34–42.
on the crushing performance were discussed based on the simulation [6] E.G. Ardi, K.J. Dong, A.B. Yu, R.Y. Yang, A combined experimental and DEM approach
results. The following conclusions can be drawn from the present study: to determine the breakage of particles in an impact mill, Powder Technol. 318
(2017) 543–548.
(1) The general good agreements observed in the results of simula- [7] B.Q. Wang, G.F. Wang, H. Wang, Y. Ding, T. Chen, Comparison of breakage behavior
tions and tests in terms of the product size distribution are evi- between the sieving crusher and ordinary teeth roll crushers, Energy Sources A 41
(2019) 252–259.
dence of the validity of the DEM model of the crushing process [8] P.A. Cundall, O.D.L. Strack, Discrete numerical model for granular assemblies,
of TDRC based on BPM. The validated DEM model demonstrates Geotechnique 29 (1979) 47–65.
that it can be potentially used for prediction of the crushing be- [9] H.P. Zhu, Z.Y. Zhou, R.Y. Yang, A.B. Yu, Discrete particle simulation of particulate sys-
tems: a review of major applications and findings, Chem. Eng. Sci. 63 (2008)
haviour of the crushable materials and probing optimal perfor-
5728–5770.
mance and better operation of crushers. [10] H.P. Zhu, Z.Y. Zhou, R.Y. Yang, A.B. Yu, Discrete particle simulation of particulate sys-
(2) The rolls' speed is a key operational parameter of TDRC. The effect tems: theoretical developments, Chem. Eng. Sci. 62 (2007) 3378–3396.
[11] N.S. Weerasekara, M.S. Powell, P.W. Cleary, L.M. Tavares, M. Evertsson, R.D.
of rolls' speed on the crushing performance could be preliminar-
Morrison, J. Quist, R.M. Carvalho, The contribution of DEM to the science of commi-
ily explained based on both the crushing principle of TDRC and nution, Powder Technol. 248 (2013) 3–24.
the fracture dynamics of particles. When the rolls' speed (equiv- [12] N. Jiménez-Herrera, G.K.P. Barrios, L.M. Tavares, Comparison of breakage models in
alent to the loading rate) is relatively high, the nipping condition DEM in simulating impact on particle beds, Adv. Powder Technol. 29 (2018) 692–706.
[13] M.B. Cil, K.A. Alshibli, 3D assessment of fracture of sand particles using discrete ele-
would be improved and more cracks could be created to release ment method, Geotechnique Lett. 2 (2012) 161–166.
the increasing strain energy causing more fractions of small size [14] J.P. de Bono, G.R. McDowell, DEM of triaxial tests on crushable sand, Granul. Matter
in the products. Furthermore, a critical rolls' speed is obtained, 16 (2014) 551–562.
[15] K.J. Hanley, C. O'Sullivan, X. Huang, Particle-scale mechanics of sand crushing in
which value is 150 rpm in the present work. With a further in- compression and shearing using DEM, Soils Found. 55 (2015) 1100–1112.
crease of the rolls' speed, both the nipping behaviour and the [16] T. Afshar, M.M. Disfani, A. Arulrajah, G.A. Narsilio, S. Emam, Impact of particle shape
fracture mechanics of materials would not be significantly im- on breakage of recycled construction and demolition aggregates, Powder Technol.
308 (2017) 1–12.
proved. [17] M. Weber, A. Spettl, M. Dosta, S. Heinrich, V. Schmidt, Simulation-based investiga-
(3) The structure of the crushing rolls has a significant influence on tion of core-shell agglomerates: influence of spatial heterogeneity in particle sizes
the quality of the crushing product. A reasonable working gap on breakage characteristics, Comput. Mater. Sci. 137 (2017) 100–106.
[18] B. Park, K.B. Min, N. Thompson, P. Horsrud, Three-dimensional bonded-particle dis-
between the rolls and a better nipping behaviour of particles in
crete element modeling of mechanical behavior of transversely isotropic rock, Int. J.
TDRC with the spiral-tooth-roll or the staggered-tooth-roll are Rock Mech. Min. Sci. 110 (2018) 120–132.
helpful for improving the crushing performance. For the crushing [19] Y. Zeng, F.G. Jia, Y.W. Xiao, Y.L. Han, X.Y. Meng, Discrete element method modelling
of impact breakage of ellipsoidal agglomerate, Powder Technol. 346 (2019) 57–69.
processes of the TDRC with spiral-tooth-roll, both the total num-
[20] A. Refahi, J. Aghazadeh Mohandesi, B. Rezai, Discrete element modeling for
ber of broken bonds and the increase between two adjacent predicting breakage behavior and fracture energy of a single particle in a jaw
cases are the largest for the three structures. The values of the crusher, Int. J. Miner. Process. 94 (2010) 83–91.
crushing processes with staggered-tooth-roll are relatively [21] M.J. Metzger, B.J. Glasser, Numerical investigation of the breakage of bonded ag-
glomerates during impact, Powder Technol. 217 (2012) 304–314.
higher than that of cases with parallel-tooth-roll. Compared to [22] M.J. Metzger, B.J. Glasser, Simulation of the breakage of bonded agglomerates in a
the other two structures, there are more products of fine and in- ball mill, Powder Technol. 237 (2013) 286–302.
termediate sizes but fewer products with large sizes obtained for [23] J. Quist, C.M. Evertsson, Cone crusher modelling and simulation using DEM, Miner.
Eng. 85 (2016) 92–105.
the TDRC with spiral-tooth-roll. [24] P.W. Cleary, Recent advances in DEM modelling of tumbling mills, Miner. Eng. 14
(2001) 1295–1319.
The fundamental knowledge about breakage behaviour during the [25] H. Li, G. McDowell, I. Lowndes, Discrete element modelling of a rock cone crusher,
Powder Technol. 263 (2014) 151–158.
crushing process of TDRC is still limited at present. In future works, [26] G.W. Delaney, R.D. Morrison, M.D. Sinnott, S. Cummins, P.W. Cleary, DEM modelling
the crushing mechanism of TDRC should be further analysed based on of non-spherical particle breakage and flow in an industrial scale cone crusher,
the fracture dynamics of particles. The investigation about the energy Miner. Eng. 74 (2015) 112–122.
[27] P. Wang, Z. Karatza, C. Arson, DEM modelling of sequential fragmentation of zeolite
efficiency of TDRC should be conducted both experimentally and nu- granules under oedometric compression based on XCT observations, Powder
merically aiming to reduce the energy and costs consumed by the Technol. 347 (2019) 66–75.
crusher. In addition, the effect of particle shape on the quantitative pre- [28] M.O. Ciantia, M. Arroyo, F. Calvetti, A. Gens, An approach to enhance efficiency of
DEM modelling of soils with crushable grains, Geotechnique 65 (2015) 91–110.
diction of the product size distribution should be further examined. [29] S.K. Soni, S.C. Shukla, G. Kundu, Modeling of particle breakage in a smooth double
Moreover, the optimisation study of the structure of the roll and the op- roll crusher, Int. J. Miner. Process. 90 (2009) 97–100.
erational parameters also need extended research. [30] J. Kwon, H. Cho, M. Mun, K. Kim, Modeling of coal breakage in a double-roll crusher
considering the reagglomeration phenomena, Powder Technol. 232 (2012)
113–123.
Acknowledgement [31] I. Cotabarren, P.G. Schulz, V. Bucalá, J. Piña, Modeling of an industrial double-roll
crusher of a urea granulation circuit, Powder Technol. 183 (2008) 224–230.
[32] H. Lieberwirth, P. Hillmann, M. Hesse, Dynamics in double roll crushers, Miner. Eng.
This work is financially supported by the National Key Research and
103-104 (2017) 60–66.
Development Program of China (Grant No. 2018YFC0604503), the Na- [33] P. Velletri, D.M. Weedon, Comminution in a non-cylindrical roll crusher, Miner. Eng.
tional Natural Science Foundation of China (Grant Nos. 51675518, 14 (2001) 1459–1468.
51704293, 51775544, 51875568), the Priority Academic Program De- [34] D. Maxton, C. Morley, R. Bearman, A quantification of the benefits of high pressure
rolls crushing in an operating environment, Miner. Eng. 16 (2003) 827–838.
velopment of Jiangsu Higher Education Institutions and the China Schol- [35] S. Morrell, Predicting the overall specific energy requirement of crushing, high pres-
arship Council. sure grinding roll and tumbling mill circuits, Miner. Eng. 22 (2009) 544–549.
[36] M. Johansson, M. Evertsson, A time dynamic model of a high pressure grinding rolls
crusher, Miner. Eng. 132 (2019) 27–38.
References [37] M. Johansson, J. Quist, M. Evertsson, E. Hulthén, Cone crusher performance evalua-
tion using DEM simulations and laboratory experiments for model validation,
[1] P.W. Cleary, M.D. Sinnott, Simulation of particle flows and breakage in crushers Miner. Eng. 103-104 (2017) 93–101.
using DEM: part 1 – compression crushers, Miner. Eng. 74 (2015) 178–197.
72 Y.-W. Li et al. / Powder Technology 356 (2019) 60–72

[38] L. Zhao, Y. Zhao, C. Bao, Q. Hou, A. Yu, Laboratory-scale validation of a DEM model of [45] Y. Zhao, S. Liu, Y. Jiang, K. Wang, Y. Huang, Dynamic tensile strength of coal under
screening processes with circular vibration, Powder Technol. 303 (2016) 269–277. dry and saturated conditions, Rock Mech. Rock. Eng. 49 (2016) 1709–1720.
[39] L. Zhao, Y. Zhao, C. Bao, Q. Hou, A. Yu, Optimisation of a circularly vibrating screen [46] S. Zhong, F. Baitalow, M. Reinmöller, B. Meyer, Relationship between the tensile
based on DEM simulation and Taguchi orthogonal experimental design, Powder strength of irregularly shaped coal particles and various fuel properties, Fuel 236
Technol. 310 (2017) 307–317. (2019) 92–99.
[40] Q.F. Hou, Z.Y. Zhou, A.B. Yu, Micromechanical modeling and analysis of different [47] J.W. Zhou, Y. Liu, C.L. Du, S.Y. Liu, Effect of the particle shape and swirling intensity
flow regimes in gas fluidization, Chem. Eng. Sci. 84 (2012) 449–468. on the breakage of lump coal particle in pneumatic conveying, Powder Technol.
[41] Q.F. Hou, K.J. Dong, A.B. Yu, DEM study of the flow of cohesive particles in a screw 317 (2017) 438–448.
feeder, Powder Technol. 256 (2014) 529–539. [48] Y.X. Zhao, S.M. Liu, G.F. Zhao, D. Elsworth, Y.D. Jiang, J.L. Han, Failure mechanisms in
[42] D.O. Potyondy, P.A. Cundall, A bonded-particle model for rock, Int. J. Rock Mech. coal: dependence on strain rate and microstructure, J. Geophys. Res.-Solid Earth 119
Min. Sci. 41 (2004) 1329–1364. (2014) 6924–6935.
[43] L.L. Zhao, Y.W. Li, X.D. Yang, Y. Jiao, Q.F. Hou, DEM study of size segregation of wet [49] R. Fu, X.L. Hu, B. Zhou, Discrete element modeling of crushable sands considering re-
particles under vertical vibration, Adv. Powder Technol. 30 (2019) 1386–1399. alistic particle shape effect, Comput. Geotech. 91 (2017) 179–191.
[44] B.A. Poulsen, D.P. Adhikary, A numerical study of the scale effect in coal strength, Int. [50] J. Bruchmuller, B.G.M. van Wachem, S. Gu, K.H. Luo, Modelling discrete fragmenta-
J. Rock Mech. Min. Sci. 63 (2013) 62–71. tion of brittle particles, Powder Technol. 208 (2011) 731–739.

You might also like