See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/323934162

Discrete element modelling of vibrating screens

Article  in  Minerals Engineering · June 2018

DOI: 10.1016/j.mineng.2018.03.010

CITATIONS READS

0 30

3 authors, including:

Ahad Aghlmandi Eren Caner Orhan

RWTH Aachen University Hacettepe University
5 PUBLICATIONS   0 CITATIONS    12 PUBLICATIONS   32 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Numerical modeling of industrial screens View project

TKI Performance Tests View project

All content following this page was uploaded by Ahad Aghlmandi on 04 May 2018.

The user has requested enhancement of the downloaded file.

 Minerals Engineering 121 (2018) 107–121

Contents lists available at ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

Discrete element modelling of vibrating screens T

⁎
Ahad Aghlmandi Harzanagh , E. Caner Orhan, S. Levent Ergun
Hacettepe University, Department of Mining Engineering, 06800, Beytepe, Ankara, Turkey

A R T I C LE I N FO A B S T R A C T

Keywords: Screening is one of the most widely used unit operations in mineral processing plants. In crushing circuits, the
Discrete element method proper selection and sizing, optimization and their operation as efficient as possible are essential in terms of the
Simulation performance and profitability of the crushing circuit and the whole plant.
Vibrating screen There are various empirical models used for the sizing and the performance prediction of the screening
LIGGGHTS solver
operation. Due to the high margin of errors during calculations and assumptions based on empirical models and
Spherical particles
Irregularly shaped particles
owing to the continuous increase in the computational capacity of computers, the numerical methods such as
Validation discrete element method have been increasingly used for the simulation of screening. However, there is still a
Pilot scale deficiency in the literature in the area of the pilot or industrial scale validation of simulation results.
In this study, the effects of various design and operating variables (namely, feed rate, deck inclination, vi-
bration frequency, amplitude and direction, aperture size) on the efficiency of screening are investigated by
means of DEM simulation of industrial vibrating screens. Spherical and irregularly shaped particles (multi-
spheres) were used in the simulations. Additionally, the prediction capability of the simulator was revealed
based on the validation tests conducted on a pilot scale vibrating screen.

1. Introduction
Screening is one of the most common unit operations in mineral
processing plants. It is used for various purposes starting from the
production at the mine site until making it suitable for use as a definite
end-product. Classification, size limitation of crusher feed, dewatering,
slime cleaning, solid recovery, washing and etc. are some of the main
utilization purposes of screening in mineral processing plants (Mular
et al., 2002). Screening is widely applied in ore preparation and many
other areas (food, textile, etc.). The design, sizing and separation effi-
ciency of screens has a direct effect on the quantity and properties of
the target products, the total power consumption and the efficient op-
eration of the crushers and consequently the profitability of the mineral
processing plant.
The role and importance of screening in whole mineral processing
operations had been resulted in many modelling attempts. There are
different approaches to the modelling of the screening process.
Probabilistic and kinetic models were one of the first attempts in the
modelling of screening as a phenomenological process (Whiten, 1972)
and (Ferrara and Perti, 1975). On the other hand, empirical or capacity
models approach the screening process modelling by investigating the
effect of different variables on the screening performance experimen-
tally. Karra (1979)’s model is one of the most popular examples of these
models. These kinds of models are used mostly by screen manufacturers
and as they have been developed for a specific type of screen, they
should be used as a guide only (Wills and Napier-Munn, 2006).
Limitations of phenomenological and empirical models in re-
presenting realistic simulations and increasing computational capacity
of computers led to the popularity of particle-based simulation ap-
proaches such as the discrete element method (DEM) in mineral pro-
cessing which was first introduced by Cundall (1971) and the details of
which are given by Cundall and Strack (1979).
DEM is appropriate for screening operation because it is capable of
reflecting dynamic processes associated with screening such as segre-
gation, passage and transport by providing necessary parameters. As a
result, it is possible to have a deep insight through screening, as well as
optimizing operation and design parameters (Elskamp and Kruggel-
Emden, 2015).
One of the first applications of DEM on screening was the studies of
Shimosaka et al. (2000) where 3D batch simulations were performed
with a limited number of particles. The first attempts to simulate con-
tinuous screens were performed by Li et al. (2002, 2003). In these
studies, the effects of single particle size and near-mesh particles by the
look to the thickness of the bulk particle layer in screening performance
were investigated. Cleary investigated the performance of double deck
banana screen at different acceleration values (Cleary et al., 2009a,b).
Additionally, Dong and Brake (2009) simulated a multi-deck banana
screen by studying the effects of operational conditions and geometry

⁎
Corresponding author.
E-mail address: ahad@hacettepe.edu.tr (A. Aghlmandi Harzanagh).

https://doi.org/10.1016/j.mineng.2018.03.010
Received 28 July 2017; Received in revised form 20 February 2018; Accepted 13 March 2018
0892-6875/ © 2018 Elsevier Ltd. All rights reserved.
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

on the screening performance. The effects of screen motion parameters Table 1

like vibration amplitude, frequency and direction for linear and circu- The data used in DEM simulations.
larly vibrating screens investigated by Chen and Tong (2010) and Zhao
Particle size (mm) 26.6 20.6 15.7 12.1 9.5 6.7 4.7 2.8
et al. (2011) respectively. Validation of obtained data from DEM si- Feed size distribution (%) 10 10 20 8 14 14 14 10
mulations against experimental data without scaling up is rare in the Feeder height (mm) 275
literature and limited to the studies of Delaney et al. (2012) and Zhao Vibration type Linear (oscillatory motion)
et al. (2016). Also according to Elskamp and Kruggel-Emden (2015), Screen aperture (mm) 10.5
Particle density (kg/m3) 2700
Hilden (2007) validated DEM application to screening process suc- Young's modulus (N/m2) 5 × 107
cessfully. Fernandez et al. (2011) and Dong and Yu (2012) tried to Poisons ratio 0.45
couple DEM with SPH (smoothed particle hydrodynamics) and CFD Coefficient of restitution 0.3
(computational fluid dynamics) to simulate wet banana screens and Sliding friction coefficient 0.5
Rolling friction coefficient 0.01
sieve bends respectively. There are some studies utilizing non-spherical
Time Step (s) 5 × 10−6
particles to increase the reality of DEM simulations in recent years. Simulation Duration (s) In the range of 25–35 s
Kruggel-Emden and Elskamp (2014), Elskamp and Kruggel-Emden
(2015) and Zhao et al. (2017) used non-spherical particles and found
more realistic results compared to the experimental data. feeder, which can provide up to 15 tons per hour of dry feed below
This study will address the effects of some operational and design 30 mm. The data used for setting up the DEM simulations are sum-
parameters like feed flowrate, screen deck inclination, vibration am- marized in Table 1.
plitude, vibration frequency and vibration direction on the performance Several simulations were performed for the testing of the model
of an inclined vibrating pilot scale screen. Open source LIGGGHTS parameters (such as Young’s modulus, time-step, Poisson’s ratio, etc.) in
software was used for performing DEM simulations. During the simu- accordance with the former studies in the literature and also to main-
lations, both spherical and irregularly shaped particles were studied. tain a reasonable time-step based on Rayleigh time step in order to
The irregularly shaped particles were modelled using multi-spheres achieve reasonable simulation times.
(sphere clumps). The effects of different parameters on screening per- To investigate how different parameters affect screening perfor-
formance were tried to be explained with familiar terms for minerals mance, various simulations were carried out at various feed flowrate,
engineers (partition curve, screening efficiency, mean residence time of screen deck inclination, vibration amplitude, vibration frequency and
passing particles and etc.). The validity of the DEM simulations has vibration direction values while keeping the other parameters constant.
been tested by experimental data. The actual (experimental) data was In the experimental tests, −30 mm crushed aggregate samples were
compared to the simulation results using screening efficiencies and used. During the initial simulations, the particles were assumed as
partition coefficients. spherical particles. Table 2 shows the values of parameters used in si-
mulations.
A simulation begins by creating and discharging the particles from
2. Pilot-scale vibrating screen geometry and simulation
the feeder. The particles that reach the screen surface will either pass
conditions
the surface or flow along the screen towards the discharge end. It is
important to evaluate particles motion data after achieving macro-
In this study, a pilot-scale vibrating inclined screen, which is
scopic steady state at which the inlet flow rate is equal to the sum of the
available in the mineral processing laboratory of Hacettepe University,
flowrates of oversize and undersize streams for each size fraction. Fig. 2
was used to perform validation tests. The typical Hertz-Mindlin particle
shows the typical illustration of simulations.
contact model is used in this work as described in details in (Jahani
At the end of the simulations, precise information such as the co-
et al., 2015). Open source LIGGGHTS solver was used to perform DEM
ordinates of each particle at each time step, the residence time of a
simulations. In order to run the simulations under the same conditions,
particle or particle size class on the screen surface, whether the screen is
3D models of the screen was prepared. Fig. 1 shows real and CAD il-
at steady state, etc. are obtained.
lustrations of the screen and available feeding system. This poly-
urethane screen surface consists of six square panels in length and two
in width, which gives 90 × 30 cm dimensions. The opening of the 3. Results and discussion
screen is 10.5 mm and the inclination of the screen surface is 10° which
is adjustable. The feeding system consists of a bunker and vibrating The simulations produce huge amount of data such as location,

Fig. 1. Real (a) and CAD (b) versions of vibrating screen and feeder system.

108
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

Table 2 100 100

List of values of parameters used in simulations. 90 90

Screening efficiency (%)

Screening efficiency(%)

Mass of undersize (%)

80 80
Parameter Case 1 Case 2 Case 3 Case 4 Case 5 Case 6
70 Mass of undersize (%) 70
Feed Tonnage (t/h) 5 10 15 20 – – 60 60
Deck inclination (°) 5 10 15 20 – –
50 50
Vibration Amplitude (mm) 3 4 5 6 – –
Vibration Frequency (Hz) 15 20 25 30 – – 40 40
Angle of vibration direction 30 45 60 90 120 150 30 30
(°)a 20 20
a
Relative to the horizontal in the direction of the screen discharge.
10 10
0 0
velocity, angular velocity and torque of every particle at every time step 5 10 15 20
of simulation and the data can be visualized by means of appropriate Feed rate (t/h)
post-processing software. For extracting the required information from
Fig. 3. Effects of feed rate on screening efficiency and mass of undersize stream.
the bulk data, an in-house developed program was coded. This program
reads through the data generated by the simulation, and the informa-
tion such as whether the system reaches steady state, whether a particle specifications, screen geometry and screen motion, which are given in
reports to undersize or oversize stream, the size distributions of coarse Table 1, as well as the screen inclination (10°) were kept constant. The
and fine products and the residence time of each particle. amplitude, frequency and direction of vibration were 5 mm, 25 Hz and
The overall efficiency of screening process was calculated using Eq. 90° respectively.
(1). (Wills and Napier-Munn, 2006). Fig. 3 shows the changes of screening efficiency and mass of un-
dersize stream and Fig. 4 shows the corresponding partition curves at
c−f different feed rates.
E=
c (1−f ) (1) As can be seen in Figs. 3 and 4, screen efficiency and mass of un-
dersize stream reduce by increasing feed rate but there is no significant
where f is the fraction of material above the cut size in the feed and c is
difference between 5 and 10 t/h feed rates which probably is the effi-
the fraction of material above the cut size in the oversize stream. Also,
ciency limit at the current configuration. Partition curves also confirm
partition coefficients defined as the percentage of the feed reporting to
the mentioned effect and also show that the cut size is decreasing with
the oversize product were calculated and plotted for every size fraction.
increasing feed rate. The main differences between partition curves
Partition curve provides the cut size of the screening which always is
exist in three fine sizes (2.8, 4.7 and 6.7 mm) which means that the
smaller than screen aperture and sharpness of the curve represents the
recovery loss is more pronounced at finer fractions than the coarser
efficiency of the screening process (Wills and Napier-Munn, 2006).
(near-mesh) fractions (Fig. 4).
Having more detailed information, recovery of undersize particles at
Fig. 5 shows the cumulative recoveries of longitudinal panels of the
the undersize stream for every panel of the screen besides the mean
screen for different particle sizes (undersize particles). As it is apparent
residence time of undersize particles for different size fractions were
from Fig. 5, although all recovery curves reach the similar peak value at
calculated.
the end of the deck for fine particles (2.8 mm), it seems that at low feed
During the evaluation of the simulation results, the 9.5 mm-particles
rates, most of the fine particles pass through the apertures at first panels
are regarded as near mesh particles as their sizes are the closest to the
of the screen but at higher feed rates these particles evenly divide be-
aperture size which is 10.5 mm.
tween panels. The difference between the shapes of the recovery curves
is gradually diminishing with increasing particle size (4.7 mm) and al-
3.1. Effects of feed flow rate most disappears for larger particles (6.7 and 9.5 mm) which become
almost parallel.
Variation in feed flow rate can affect screening performance by Study of the mean residence time (MRT) of particles reporting to
changing the dynamics of interactions between particles, bed depth, undersize stream at different feed rates (Fig. 6) shows that MRT is in-
etc. In order to examine the effects of feed rate on screening perfor- creasing with increasing feed rate. As expected, the increase of feed
mance and related factors, simulations at 5, 10, 15 and 20 t/h feed rate flowrate reduces the probability of particles contact and pass through
have been performed. Other simulation parameters like ore

Fig. 2. Typical illustration of simulations.

109
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

100 1.2
90
5 t/h 1
Partition coefficient (%)

80
70 10 t/h 0.8

MRT (s)
60
15 t/h 0.6
50 5 t/h
40 20 t/h 0.4
10 t/h
30 0.2 15 t/h
20 20 t/h
0
10
2 3 4 5 6 7 8 9 10
0
1 10 100 Particle size (mm)
Particle size (mm) Fig. 6. Mean residence time of particles at different feed rates.

Fig. 4. Partition curves for different feed rates.

100 100
Screening efficiency(%)

Mass flow of undersize (%)

the screen apertures. There is an increasing trend of MRT also for each 90 90

Screen efficiency (%)

80 Mass of undersize (%) 80
particle size, however at the coarsest size (9.5 mm), which are the near
mesh particles, has a similar or lower MRT value compared to the finer 70 70
size fraction (6.7 mm) which is an indication that at the tested feed rate 60 60
range, the effect of feed flow rate is less pronounced on the MRT and 50 50
hence the recovery of the near mesh particles. 40 40
30 30
3.2. Effects of screen inclination 20 20
10 10
For examining the effects of screen inclination on screening per- 0 0
formance and related factors, four simulations at screen inclinations of 5 10 15 20
5°, 10°, 15° and 20° were performed. Other simulation values like feed Screen inclination (°)
specifications, screening geometry, screening motion (Table 1) were
Fig. 7. Effects of screen inclination on screening efficiency and massflow of undersize
kept constant. Feed rate, vibration amplitude, vibration frequency and stream.
vibration direction were 10 t/h, 5 mm, 25hz and 90° respectively.
Fig. 7 shows the variation of screening efficiency and mass flow of
the undersize stream depending on the screen inclination and Fig. 8 mesh particles.
shows the related partition curves at different screen inclinations. Fig. 9 shows cumulative recoveries of longitudinal panels of the
As is apparent from Fig. 7, screening efficiency and mass of un- screen for different particle sizes. As is apparent from Fig. 9, the re-
dersize reduces with increasing deck inclination. According to partition covery of undersize particles to undersize stream is higher for lower
curves (Fig. 8), cut size is higher in lower screen inclinations and screen inclinations in every panel of the screen except in the first panel
screening becomes more efficient. At screen inclination of 5°, the re- at which the recovery at 5° is lower than the inclination of 10°. This
covery of near mesh particles (9.5 mm) to the oversize product also uneven behavior of lower recovery in the first panel is probably due to
decreases meaning that screen inclination is effective for treating near the excessive accumulation of particles at the feed end and preventing

2.8 mm 4.7 mm
Cum. Rec. to undersize (%)

Cum. Rec. to undersize (%)

100 100
80 80
60 60
40 5 t/h 10 t/h 40
20 20
15 t/h 20 t/h
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number

6.7 mm 9.5 mm
Cum. Rec. to undersize (%)

60 10
Cum. Rec. to undersize (%)

50 8
40
6
30
4
20
10 2
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number
Fig. 5. Cumulative recoveries of longitudinal panels of the screen at various feed rates.

110
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

100 2
90 5° 10°
Partition coefficient (%)

5°
80
10° 1.5 15° 20°
70
15°

MRT (s)
60
20°
50 1
40
30
0.5
20
10
0 0
1 10 100 2 3 4 5 6 7 8 9 10
Particle size (mm) Particle size (mm)
Fig. 8. Partition curves for different screen inclinations. Fig. 10. Mean residence time of particles for different screen deck inclinations.

the particles reach the screen surface, but at the next panel recovery 100 100
Screening efficiency
increases rapidly. Additionally, the recovery of near mesh particles is 90 90
(%)

Screening efficiency (%)

Mass of undersize (%)

considerably higher in low screen deck inclinations and final panels of 80 80
Mass of undersize (%)
the screen are responsible for this increase. 70 70
As expected, MRT of particles is higher in low screen deck inclina- 60 60
tions especially at the inclination of 5°, which means undersize parti- 50 50
cles, had more time and hence chance to contact the screen surface and 40 40
pass through the apertures (Fig. 10). High MRT of near mesh particles 30 30
at the inclination of 5° also proves the role of final panels of the screen
20 20
in the recovery of these particles.
10 10
0 0
3.3. Effects of screen vibration parameters 3 4 5 6

In this study, the effects of linear vibration parameters, including Vibration amplitude (mm)
vibration amplitude, vibration frequency and vibration direction, were Fig. 11. Effects of vibration amplitude on screening efficiency and mass of undersize
investigated separately. These parameters have significant importance stream.
in the design of vibrating screens as they determine the motion of the
screen surface and have a considerable influence on particle–particle amplitudes of 3, 4, 5 and 6 mm have been performed by keeping the
and particle–surface collisions. During the tests the screen inclination other parameters constant. Physical properties of the feed as well as the
was kept constant at 10°. screen motion are given in Table 1. In the simulations, the feed flow
rate, screen inclination, vibration frequency and vibration direction
3.3.1. Vibration amplitude values were 10 t/h, 10°, 25 Hz and 90° respectively. Fig. 11 shows the
In order to examine the effects of screen vibration amplitude on the changes of screening efficiency and mass of undersize stream and
screening performance and related factors, four simulations at vibration Fig. 12 shows the related partition curves for different vibration

2.8 mm 4.7 mm
Cum. Rec. to undersize (%)
Cum. Rec. to undersize (%)

100 100
80 80
60 5° 60
40 10° 40
20 15° 20
20°
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number

6.7 mm 9.5 mm
Cum. Rec. to undersize (%)

Cum. Rec. to undersize (%)

80 20

60 15

40 10

20 5

0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number
Fig. 9. Cumulative recoveries of longitudinal panels of the screen for various screen deck inclinations.

111
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

100 1.4
90
Partition coefficient (%)

1.2
80 3mm
70 1
4mm

MRT (s)
60 0.8
50 5mm 0.6
40 3mm 4mm
30 6mm 0.4
5mm 6mm
20 0.2
10
0
0 2 3 4 5 6 7 8 9 10
1 10 100
Particle size (mm)
Particle size (mm)
Fig. 14. Mean residence time of particles at various vibration amplitudes.
Fig. 12. Partition curves at different vibration amplitudes.

3.3.2. Vibration frequency

amplitudes. Screening efficiency and mass of undersize stream are re- For examining the effects of screen vibration frequency on the
ducing with the increase in vibration amplitude and it means that in screening performance and related factors, four simulations at vibration
high vibration amplitudes the chance of contact between particles and frequencies of 15, 20, 25 and 30 Hz were performed by keeping other
screen surface is decreasing. According to Fig. 12, cut size of screening parameters constant. Physical properties of the feed and the screen
is higher in low vibration amplitudes and the sharpness of performance motion are given in Table 1. The feed flow rate, screen inclination,
curve is higher in lower vibration amplitudes. Additionally, the re- vibration amplitude and vibration direction values were 10 t/h, 10°,
covery of near mesh particles in the oversize product is lower in low 5 mm and 90° respectively. Fig. 15 shows the dependence of screen
amplitudes so vibration amplitude seems to have a considerable influ- efficiency and mass of undersize stream on the frequency and Fig. 16
ence on the near mesh particles as well. Fig. 13 shows the recovery of shows the corresponding partition curves at different vibration fre-
undersize particles in the undersize stream in longitudinal panels of the quencies.
screen at various vibration amplitudes. As expected, the recoveries are Screening efficiency and mass of undersize reduces with increasing
higher at lower amplitude values as particles travel closer to the screen vibration frequency (Fig. 15) and as can be seen in Fig. 16, the cut size
surface. The difference between the recoveries of near mesh particles also decreases. It is possible to decrease the recovery of near mesh
(9.5 mm) is significant and this emphasizes the effect of vibration am- particles to oversize particles by reducing vibration frequency.
plitude in treating near mesh particles. Fig. 17 shows the cumulative recoveries of longitudinal panels of
The study of the mean residence time of passing particles (Fig. 14) the screen at various vibration frequencies and sizes. The effects of
shows that there is small difference between the mean residence time of vibration frequency on screening performance is more apparent for
particles reporting to undersize stream for finer particles but there is a near mesh particles.
clear difference for near mesh particles and MRT is higher in lower The study of the mean residence time of particles reporting to un-
amplitudes. This proves also that, compared to other size fractions, the dersize stream also emphasizes this fact as the change in the mean re-
near mesh particles have less probability of contacting the screen sur- sidence time with decreasing frequency is more pronounced for near
face at higher vibration amplitudes. mesh particles. Similar to the effects of vibration amplitude, the smaller

2.8 mm 4.7 mm
100 100
Cum. Rec. to undersize (%)

Cum. Rec. to undersize (%)

80 80
60 3mm 60
40 4mm 40
5mm
20 20
6mm
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number

6.7 mm 9.5 mm
100 30
Cum. Rec. to undersize (%)

Cum. Rec. to undersize (%)

80 25
20
60
15
40
10
20 5
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number
Fig. 13. Cumulative recoveries of longitudinal panels of the screen for various vibration amplitudes.

112
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

100 100 1.5

90 Screening efficiency(%) 90
Screening efficiency (%)

Mass of undersize (%)

Mass of undersize (%)

80 80
70 70 1

MRT (s)
60 60
50 50
40 40 15 Hz
0.5
30 30 20 Hz
20 20 25 Hz
10 10 30 Hz
0
0 0
2 3 4 5 6 7 8 9 10
15 20 25 30
Particle size (mm)
Vibration frequency (Hz)
Fig. 18. Mean residence time of particles at different vibration frequencies.
Fig. 15. Effects of vibration frequency on screening efficiency and mass of undersize
stream.
Vibration direction
100
90
Partition coefficient (%)

15 Hz Vibration
80 angle
70 20 Hz
60 25 Hz
50 Screen surface
30 Hz
40 Fig. 19. Schematic sketch of vibration direction.
30
20
3.3.3. Vibration direction
10
Vibration direction is an important parameter on designing vi-
0 brating screens as it controls the direction of forces applied to particles
1 10 100
and therefore the velocity and the residence time of particles. For ex-
Particle size (mm) amining the effects of vibration direction of the screen on the screening
Fig. 16. Partition curves for different vibration frequencies. efficiency, six simulations at vibration direction values of 30°, 45°, 60°,
90°, 120° and 150° relative to horizontal axis was performed (Fig. 19).
The vibration motion of the screen was selected to be linear in these
vibration frequency values cause the particles bounce less on the screen
simulations. The effects of the vibration motion with other shapes
surface which result in the increased probability of near mesh particles
(namely circular and elliptic) will be investigated elsewhere. Physical
contacting the screen surface and passing through the apertures
properties of the feed and screen motion are given in Table 1. The feed
(Fig. 18).
flow rate, screen inclination, vibration frequency and vibration fre-
quency values were 10 t/h, 10°, 5 mm and 25 Hz respectively.

2.8 mm 4.7 mm
Cum. Rec. to undersize (%)

Cum. Rec. to undersize (%)

100 100
80 80
60 60
40 15 Hz 20 Hz 40
20 25 Hz 30 Hz 20
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number

6.7 mm 9.5 mm
Cum. Rec. to undersize (%)
Cum. Rec. to undersize (%)

100 40
80 30
60
20
40
20 10

0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number
Fig. 17. Cumulative recoveries of longitudinal panels of the screen at various vibration frequencies.

113
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

100 100 sharply as particles have a tendency to flow in the reverse direction
90 90 (towards the feed end) and these give extra time to particles to stay on
Screening efficiency (%)

Mass of undersize (%)

80 80 the screen surface and find a chance to pass through apertures. Partition
70 70 curve of vibration inclination of 150° almost matches to partition curve
60 60 of ideal screening and approximately 80% of near mesh particles report
50 50 to undersize stream.
40 40 Fig. 22 shows the cumulative recoveries of longitudinal panels of
30 30 the screen for various vibration directions and particle sizes (under-
20 20 size). Extracted data proves the ineffectiveness of the vibration in-
Screening efficiency(%)
10 10 clination at low angles and emphasizes the effect of high vibration in-
Mass of undersize (%) clinations on screening performance especially for near mesh particles.
0 0
30 60 90 120 150 In the case of screen inclination of 150°, the recovery of undersize
particles is low at initial panels of the screen but increases rapidly at
Vibration ditection (°) following panels and reaches a maximum value. This phenomenon is
Fig. 20. Effects of vibration direction on screening efficiency and mass of undersize probably due to the aggregation of particles at first parts of the screen
stream. causing a larger bed depth. Fig. 23 shows the screen-shot of the ani-
mations at the 15th second of the simulations at various vibration di-
100 rection angles and the increase of the bed-depth could clearly be ob-
90 served.
30° Fig. 24 shows the mean residence time of particles reporting to
partition coefficient (%)

80
45° undersize stream at various vibration angles. According to Fig. 24, MRT
70 60° is almost constant at different vibration directions and the particle sizes
60 90° at lower angles but it follows an increasing trend in the case of larger
50 120° vibration direction angles. At 150°, MRT is very high even for finer
40 150° particles compared to the other vibration direction angles. Although
30 higher residence time leads to better performances, it also reduces the
capacity of the screen as well. Considering Figs. 23 and 24, it seems that
20
the mean residence time of particles reporting to undersize stream
10
could be a representative index to describe the bed depth and optimal
0 capacity at a certain efficiency of the screening operation.
1 10 10 0
Particle size (mm)
Fig. 21. Partition curves for different vibration directions. 4. Non-spherical particles

Fig. 20 shows the effects of various vibration directions on the In order to increase the reality of the simulations, non-spherical
screening efficiency and the mass flow of the undersize stream and particles were used and the results of the simulations were compared
Fig. 21 shows the corresponding partition curves. As is apparent from with earlier simulations. For the modelling of the irregularly shaped
these figures, the screening efficiency, the mass of undersize stream and particles, multi-spheres (sphere clumps) method was used which is
the cut size are not so much sensitive to vibration direction in the range supported by LIGGGHTS’ multi-sphere module. It uses the concept of
of 30–90° but at larger angles (120° and 150°), these values increase combining spheres and creating a new non-spherical particle template.
It is needed to import coordinates of the centers and also radiuses of the

2.8 mm 4.7 mm
Cum. Rec. to undersize (%)

Cum. Rec. to undersize (%)

100 100
80 80
60 60
40 40
30° 45°
20 60° 90° 20
120° 150°
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number

6.7 mm 9.5 mm
Cum. Rec. to undersize (%)

Cum. Rec. to undersize (%)

100 100
80 80
60 60
40 40
20 20
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Panel Number Panel Number
Fig. 22. Cumulative recoveries of longitudinal panels of the screen for various vibration directions.

114
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

26.6mm

30° 45° 60°

2.8mm

90° 120° 150°

Fig. 23. Illustration of simulations at various vibration directions at the 15th second.

10 100
9 30° 45° Screening efficiency (%)
90
8 60° 90° 80
7 120° 150° 70
MRT (s)

6 60
5 50
4 40
3 30 Non-spherical
2 20
Spherical
1 10
0 0
2 3 4 5 6 7 8 9 10 1 2 3 4
Particle size (mm) Amplitude (mm)
Fig. 24. Mean residence time of particles at different vibration directions. Fig. 26. Comparison of the screening efficiency at various vibration amplitudes for
spherical and non-spherical particles.

spheres as a text file to the solver. It is possible to use several particle

templates in a simulation. For the creation of desired particle templates
with an optimum number of spheres, an in-house program was coded
which takes a high-resolution particle shape, decrease the number of
vertices according to the desired resolution and then fill the solid with
spheres. With this methodology, one sphere is created at one vertex and
is expanded until the sphere touches another vertex and this action is
repeated for every vertex at the end of which a simplified sphere clump
is obtained. Fig. 25 shows the stages of creating a particle template.
It is apparent that the use of high-resolution particles gives more

High resolution 12 spheres 50 spheres

Fig. 25. The stages of creating a non-spherical particle template.

115
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

100

90 Amplitude 1mm (non-spherical)

80
Amplitude 1mm (spherical)

Partition coefficient (%)

70
Amplitude 2mm (non-spherical)
60
Amplitude 2mm (spherical)
50

40 Amplitude 3mm (non-spherical)

30 Amplitude 3mm (spherical)

20 Amplitude 4mm (non-spherical)

10
Amplitude 4mm (spherical)
0
1 10 100
Particle size (mm)
Fig. 27. Partition curves of simulations at various vibration amplitudes (spherical and non-spherical particles).

4 Amplitude 1 mm 4 Amplitude 2 mm
Non-spherical
3 3 type1 (12 spheres)
MRT (s)
MRT (s)

Non-spherical
2 type2 (5 spheres)
2
Spherical

1 1

0 0
2 4 6 8 10 2 4 6 8 10
Particle size (mm) Particle size (mm)

4 Amplitude 3 mm 4 Amplitude 4 mm

3 3
MRT (s)

MRT (s)

2 2

1 1

0 0
2 4 6 8 10 2 4 6 8 10
Particle size (mm) Particle size (mm)
Fig. 28. Mean residence time of particles (reporting to under size stream) at various vibration amplitudes.

realistic results but for optimizing the performing time of the simula-
tion, two particle templates which consist of 12 and 5 spheres respec-
tively were used in non-spherical simulations. Determining the size of
the non-spherical particles is challenging as it differs in different di-
rections. To perform an effective sizing, the concept of lab scale sieving
was used in a simulation environment. For the determination of the
sieve size of a sphere clump, firstly the CAD models of sieves each with
10 apertures were created at aperture sizes of 22.6, 19.0, 13.0, 11.2, 8,
5.6, 4.0, 2.0 mm. In other words, the sphere clumps were sieved in the
DEM environment to precisely determine the corresponding sieve size
of each sphere clump. The motion of each sphere clump was simulated
using DEM at these sieves to determine whether the clump will report
to undersize. Repeating this procedure many times for each sphere
clump at various size scales, the range of size scale that corresponds to
each size fraction was determined.
As screening performance is very sensitive to vibration amplitude
(Section 3), simulations with vibration amplitudes of 1, 2, 3 and 4 mm
were performed for both spherical and non-spherical particles at the
same conditions. Fig. 26 shows the comparison of the screening effi-
ciency at various vibration amplitudes for spherical and non-spherical
simulations and Fig. 27 shows corresponding partition curves.
According to Fig. 26, screening efficiency reduces with increasing
vibration amplitude for both non-spherical and spherical particles but
efficiency values are higher for non-spherical particles.
Partition curves also verify the mentioned effect and show that the
cut size decreases with increasing vibration amplitude. Fig. 28 shows
the mean residence time of particles reporting to under size stream at
various vibration amplitudes for both non-spherical and spherical par-
ticles.
The mean residence time of non-spherical particles is apparently
higher than spherical particles. As less frictional force is applied on the
spherical particles, they readily roll towards the discharge end and
spend less time on the screen surface. This discloses the high efficiency
of screening process with non-spherical particles as high residence time
means a high probability of contact between particle and screen surface
and therefore high chance of passing through apertures. In the case of

116
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

Amplitude (1mm) Amplitude (2mm)

Amplitude (3mm) Amplitude (4mm)

Fig. 29. Illustration of simulations with non-spherical particles at various vibration amplitudes at the 11th second of simulations.

Amplitude (1mm) Amplitude (2mm)

Amplitude (3mm) Amplitude (4mm)

Fig. 30. Illustration of simulations with spherical particles at various vibration amplitudes at the 11th second of simulations.

vibration amplitude of 1 mm the difference between mean residence
time of spherical and non-spherical particles is the highest for all par-
ticle sizes but in other vibration amplitudes this difference is lower
especially for finer particles as these particles leave the screen surface
much easily, but for the coarser particles (towards near-mesh particles)
this difference is maximum.
Figs. 29 and 30 show the illustrations of simulations with non-
spherical and spherical particles at various vibration amplitudes at the
11th second of simulations respectively. According to these figures, the
accumulation of particles at the initial parts of the screen is very high in
the simulation performed by non-spherical particles and vibration
amplitude of 1 mm. This phenomenon explains the high mean residence
time of fine particles in this case.
Blinding of apertures with near mesh particles is inevitable in actual

117
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

Fig. 31. The stages of the blinding of screen apertures with non-spherical particles at various time-steps.

Table 3 to determine which results are closer to actual experimental data, va-
Feed flow rates of performed experimental tests. lidation tests were performed.
Test 1 Test 2 Test 3 Test 4

Feed flow rate (t/h) 0.24 3.99 5.33 8.96

5. Validation attempts

As mentioned above, a pilot scale inclined vibrating screen was used

100 for the validation experiments. Original deck inclination of the screen
90 was adjustable and was initially set to 10°. The feed rate and the vi-
Screening efficiency (%)

80 bration frequency of this vibrating screen could also be modified in

70 order to test various conditions. A calcite sample crushed to −30 mm
60 was classified to eight size fractions and combined at desired propor-
50 tions to generate the feed to the screen. In order to investigate the
Experimental test
40 validity of simulations and have a comparison between simulations
Simulation (non-spherical)
30 with spherical and non-spherical particles, an experimental test at four
Simulation (spherical)
20 different feed rates, two different aperture sizes and three different
10
vibration frequency values were performed. Particle size distribution
given in Table 1 was used in the tests. The vibration type of the pilot
0
0 2 4 6 8 10 screen was slightly elliptical but simplified to linear with 110° stroke
Feed rate (t/h) inclination according to horizontal axis, 25 Hz vibration frequency and
5 mm amplitude was identified as vibration frequency and amplitude of
Fig. 32. Calculated screening efficiencies for experimental tests and simulations per-
formed by spherical and non-spherical particles.
pilot scale screen.

screening operation and it is known that this phenomenon reduces the

efficiency of the screening by reducing the open area of the screen
surface. In the simulations with spherical particles, blinding was not
observed due to the high screen inclination (10°) and especially the
significant difference between the aperture size (10.5 mm) and the
diameter of near mesh particles (9.5 mm). However, in the case of si-
mulations performed with non-spherical particles, the blinding was
detected as particles do have different sizes in different directions.
Fig. 31 shows the stages of blinding of screen apertures with non-
spherical particles in a representative simulation which was performed
with near-mesh particles only.
Last of all, screening efficiency and sharpness of the separation is
higher for simulations performed with non-spherical particles. In order
• Feed flow rate:
Table 3 shows the values for feed flow rate during the experimental
tests.
After system reached steady state, undersize and oversize streams
were sampled at the same time and the size distributions of the samples
were measured. Using the data obtained, the size distribution of the
reconstituted feed, screening efficiency and partition coefficients were
calculated.
Besides the size distribution of reconstituted feed, other measured
specifications such as ore density, feed flow rate, screen deck inclina-
tion, vibration shape, vibration amplitude, vibration frequency and
direction were imported to the simulation code and the simulations
were performed using spherical and non-spherical particles separately.

118
A. Aghlmandi Harzanagh et al. Minerals Engineering 121 (2018) 107–121

Feedrate 0.25 (t/h) Feedrate 3.99 (t/h)

100 100
Partition coefficient(%)

Partition coefficient(%)
80 80

60 60

40 40
Experimental
20 Simulation (non-spherical) 20
Simulation (spherical)
0 0
1 10 100 1 10 100
Paricle size (mm) Paricle size (mm)

Feedrate 5.33 (t/h) Feedrate 3.99 (t/h)

100 100
Partition coefficient(%)

Partition coefficient(%)
80 80

60 60

40 40

20 20

0 0
1 10 100 1 10 100
Paricle size (mm) Paricle size (mm)
Fig. 33. Partition curves of tested and simulated processes for various feed flow rates.

100 100
90
Experimental 90
Partition coefficient (%)

80
13mm
70 80
Screning efficiency (%)

60 Experimental
10.5mm
70
50
Simulation 60
40
30 13mm 50
20 Simulation 40
10 10.5mm Screening efficiency (%) Experimental
0 30
Mass of undersize (%) Experimental
1 10 100 20
Screening efficiency (%) Simulation
Particle size (mm) 10 Mass of undersize (%) Simulation
Fig. 34. Partition curves of actual and simulation results at different screen apertures.
0
15 20 25 30
The data generated as a result of simulations were used for calculating Vibration frequency (Hz)
screen efficiencies and partition coefficients. Fig. 35. Calculated screening efficiencies and mass of undersize in pilot scale tests and
Fig. 32 shows the calculated screen efficiencies for experimental simulations at various vibration frequencies with 13 mm aperture size.
tests and simulations performed with spherical and non-spherical par-
ticles. According to the Fig. 32, the efficiency of the pilot screening is
trends proving the fact that DEM simulation is capable to reflect dy-
reasonably close to the results of both simulations performed with
namic processes associated with screening process like segregation,
spherical and non-spherical particles. All results show a similar trend
transport and passage.
and that the efficiency is the lowest at the lowest feed rate (0.24 t/h)
Fig. 33 shows the partition curves of performed experimental test
and reaches a peak value at around 4 t/h and then decreases again with
and simulations. Like efficiency data, there is a good agreement be-
increasing feed flow rate. On the other hand, the simulations with non-
tween tested and simulated partition curves and the results obtained
spherical particles are much closer to the results of pilot scale screening.
with non-spherical particles are in very good agreement with pilot scale
These proves the fact that the representing real particles as spheres in
validation tests.
DEM simulations reduces the reliability of the results.
No bed was observed at lowest feed flow rate (0.24 t/h) and the bed
depth was higher with increasing feed flow rates. This phenomenon • Screen aperture:
shows the importance of optimal bed depth which provides best
An experimental test with the pilot scale screen was performed
transport and segregation of particles along the screen surface.
using screens with 10.5 and 13 mm apertures. Other parameters like
Screening efficiencies of simulated case results also follow similar

119
A. Aghlmandi Harzanagh et al.
100
90
80
Partition coefficient (%)
70
60
50
40
30
20
10
0
1.00
Particle size (mm)
Fig. 36. Partition curves of pilot scale tests and simulations at various vibration frequency with 13 mm aperture size.
size distribution of the feed, feed flow rate and vibration frequency was
kept constant. After the evaluation of the test results, simulations were
performed using non-spherical particles as described above.
According to the results of the experimental tests and simulations,
the screening efficiency increased from 68.13% to 80.50% with in-
creasing the aperture size from 10.5 mm to 13 mm as expected. This
trend was also estimated in the simulations in which the values of the
screen efficiency were calculated as 67.11% and 78.38% respectively.
Fig. 34 shows the partition curves of the actual and simulated results at
different screen apertures. According to Fig. 34, there is a very good
agreement between simulated and actual screening with irregularly
shaped particles.
As expected, increasing the aperture size would increase the cut
size. The results show that increasing the screen aperture from 10.5 to
13 mm, the cut size is increased from 7.65 to 10.70 mm in the pilot test
and from 7.50 to 10.60 mm in the simulations.
• Vibration frequency:
In order to test the validity of simulations, further pilot scale tests
were performed at 17.87, 21.32 and 25 Hz vibration frequencies using a
frequency converter. Parameters like particle size distribution, feed
flowrate and screen aperture (13 mm) were kept constant and the si-
mulations were performed at the same conditions.
Results of the experimental tests and the simulations show a de-
creasing trend in screening efficiency and the mass of undersize with
increasing vibration frequency in both experimental and simulation
results (Fig. 35) and there is a very good agreement between the results
of the tests and the simulations.
Fig. 36 shows the partition curves of pilot scale tests and simulations
at various vibration frequencies. According to Fig. 36, the sharpness of
the curves is decreasing with increasing vibration frequency as ex-
pected and there is a very good agreement between the results of the
tests and simulations.
The results of the pilot scale validation tests employed at various
feed flow rate, aperture size and vibration frequency values show that
the DEM is capable of very closely estimating the actual screening op-
eration. Although the simulations with spherical particles are in good
agreement with actual results, the simulations with non-spherical par-
ticles are even better and provide much closer results to actual opera-
tion.
120
Minerals Engineering 121 (2018) 107–121
Simulation 25 Hz
Simulation 21.32 Hz
Simulation 17.87 Hz
Experimental 25 Hz
Experimental 21.32 Hz
Experimental 17.87 Hz
10.00 100.00
6. Conclusions
Particle flow in a typical pilot scale vibrating screen were numeri-
cally simulated by a three-dimensional DEM model using spherical and
non-spherical particles. The effects of major operational and design
parameters, namely feed rate, screen deck inclination, vibration am-
plitude, vibration frequency and vibration direction, on the screening
performance were investigated. The performance of the screening was
investigated in terms of screening efficiency, the mass flow of the un-
dersize stream, partition curve and mean residence time of particles
reporting to undersize stream.
For validating the results of the simulations, experimental tests were
performed using pilot scale vibrating screen. The results of the valida-
tion tests performed with various feed flowrates, aperture size and vi-
bration frequency, show that the DEM is capable of very closely esti-
mating the results of actual screening operation.
Besides the powerful predictions of overall efficiency, partition
curves and the size distribution of products, DEM also enables the
quantification of important parameters such as the residence time of
any particle and/or size fraction, the tracking of particles for the de-
termination of the load on any part of the screen, observation of bed
material and stratification, etc. which would provide better under-
standing of the effects of various design and operating variables and
would also be beneficial in the optimization of the screening operation.
Although reasonable estimations could be achieved using spherical
particles, the use of irregularly shaped particles provides much more
superior predictions enabling the reflection of actual phenomena such
as stratification of bed material, blinding of apertures and segregation
of particles, which the spherical particles fail.
The simulation environment developed in this study would be a
valuable engineering tool which can be used in design and optimization
of the screens and the screening operations.
Acknowledgement
The authors gratefully acknowledge the financial support of The
Scientific and Technological Research Council of Turkey (TUBITAK) to
the project entitled “Numerical Modelling of Industrial Screening”
(215M368).
References
Chen, Y., Tong, X., 2010. Modeling screening efficiency with vibrational parameters
based on DEM 3D simulation. Min. Sci. Technol. 20, 615–620. http://dx.doi.org/10.
A. Aghlmandi Harzanagh et al.
1016/S1674-5264(09)60254-4.
Cleary, P.W., Sinnott, M.D., Morrison, R.D., 2009a. Separation performance of double
deck banana screens – Part 1: Flow and separation for different accelerations. Miner.
Eng. 22, 1218–1229. http://dx.doi.org/10.1016/j.mineng.2009.07.002.
Cleary, P.W., Sinnott, M.D., Morrison, R.D., 2009b. Separation performance of double
deck banana screens – Part 2: Quantitative predictions. Miner. Eng. 22, 1230–1244.
http://dx.doi.org/10.1016/j.mineng.2009.07.001.
Cundall, P.A., 1971. A computer model for simulating progressive large-scale movements
in blocky rock systems. Nancy, p. Symp. Int. Soc. Rock Mech.
Cundall, P.A., Strack, O.D.L., 1979. A discrete numerical model for granular assemblies.
Géotechnique 29, 47–65. http://dx.doi.org/10.1680/geot.1979.29.1.47.
Delaney, G.W., Cleary, P.W., Hilden, M., Morrison, R.D., 2012. Testing the validity of the
spherical DEM model in simulating real granular screening processes. Chem. Eng. Sci.
68, 215–226. http://dx.doi.org/10.1016/j.ces.2011.09.029.
Dong, K.J., Brake, I., 2009. DEM simulation of particle flow on a multi-deck banana
screen. Miner. Eng. 22, 910–920. http://dx.doi.org/10.1016/j.mineng.2009.03.021.
Dong, K.J., Yu, A.B., 2012. Numerical simulation of the particle flow and sieving beha-
viour on sieve bend/low head screen combination. Miner. Eng. 31, 2–9. http://dx.
doi.org/10.1016/j.mineng.2011.10.020.
Elskamp, F., Kruggel-Emden, H., 2015. Review and benchmarking of process models for
batch screening based on discrete element simulations. Adv. Powder Technol. 26,
679–697. http://dx.doi.org/10.1016/j.apt.2014.11.001.
Fernandez, J.W., Cleary, P.W., Sinnott, M.D., Morrison, R.D., 2011. Using SPH one-way
coupled to DEM to model wet industrial banana screens. Miner. Eng. 24, 741–753.
http://dx.doi.org/10.1016/j.mineng.2011.01.004.
Ferrara, G., Perti, U., 1975. A contribution to screening kinetics. In: L Lth Int. Min. Proc.
Cong. Cagliari.
Hilden, M.M., 2007. dimensional analysis approach to the scale-up and modelling of
industrial screens. University of Queensland, PhD-thesis.
Jahani, M., Farzanegan, A., Noaparast, M., 2015. Investigation of screening performance
of banana screens using LIGGGHTS DEM solver. Powder Technol. 283, 32–47. http://
121
View publication stats
Minerals Engineering 121 (2018) 107–121
dx.doi.org/10.1016/j.powtec.2015.05.016.
Karra, V., 1979. Development of a model for predicting the screening performance of a
vibrating screen. CIM Bull.
Kruggel-Emden, H., Elskamp, F., 2014. Modeling of Screening Processes with the Discrete
Element Method Involving Non-Spherical Particles. Chem. Eng. Technol. 37,
847–856. http://dx.doi.org/10.1002/ceat.201300649.
Li, J., Webb, C., Pandiella, S.S., Campbell, G.M., 2002. A numerical simulation of se-
paration of crop seeds by screening—effect of particle bed depth. Food Bioprod.
Process. 80, 109–117. http://dx.doi.org/10.1205/09603080252938744.
Li, J., Webb, C., Pandiella, S.S., Campbell, G.M., 2003. Discrete particle motion onsieves -
a numerical study using the DEM simulation. Powder Technol. 133, 190–202. http://
dx.doi.org/10.1016/S0032-5910(03)00092-5.
Mular, A.L., Halbe, D.N., Barratt, D.J., 2002. Mineral Processing Plant Design, Practice,
and Control. SME, Vancouver, British Columbia.
Shimosaka, A., Higashihara, S., Hidaka, J., 2000. Estimation of the sieving rate of pow-
ders using computer simulation. Adv. Powder Technol. 11, 487–502. http://dx.doi.
org/10.1163/156855200750172088.
Whiten, W.J., 1972. The simulation of crushing plants with models developed using
multiple spline regression. J. South. African Inst. Min. Metall.
Wills, B.A., Napier-Munn, T.J., 2006. Mineral Processing Technology. Elsevier Science &
Technology Books.
Zhao, L., Zhao, Y., Bao, C., Hou, Q., Yu, A., 2017. Optimisation of a circularly vibrating
screen based on DEM simulation and Taguchi orthogonal experimental design.
Powder Technol. 310, 307–317. http://dx.doi.org/10.1016/j.powtec.2017.01.049.
Zhao, L., Zhao, Y., Bao, C., Hou, Q., Yu, A., 2016. Laboratory-scale validation of a DEM
model of screening processes with circular vibration. Powder Technol. 303, 269–277.
http://dx.doi.org/10.1016/j.powtec.2016.09.034.
Zhao, L., Zhao, Y., Liu, C., Li, J., Dong, H., 2011. Simulation of the screening process on a
circularly vibrating screen using 3D-DEM. Min. Sci. Technol. 21, 677–680. http://dx.
doi.org/10.1016/j.mstc.2011.03.010.