Minerals Engineering 105 (2017) 22–35

Contents lists available at ScienceDirect

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

Effect of lifters and mill speed on particle behaviour, torque, and power
consumption of a tumbling ball mill: Experimental study and DEM
simulation
Xiaolei Bian, Guoqiang Wang, Hongdi Wang, Shuai Wang, Weidong Lv ⇑
College of Mechanical Science and Engineering, Jilin University, Changchun 130025, China

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

Article history: Crushing and grinding consume most of the energy in mineral processing. Ball mill is an important kind
Received 2 April 2016 of grinding equipment used to decrease the size of ore particles. The power consumption of a ball mill
Revised 26 October 2016 is one of the most important parameters to consider in the design of a ball mill because it determines
Accepted 29 December 2016
its economic efficiency. The power consumption is usually determined by charge fill level, lifter height,
lifter number, and mill speed. However, almost all of the classical theories for calculating the power
consumption of ball mills disregard the effect of lifters and only focus on rotation rate, charge fill level,
Keywords:
as well as size and shape of grinding media, thereby may causing errors. Discrete element method
Experimental study
DEM simulation
(DEM) can simulate the motion and interaction of particle materials. Thus, this method is widely used
Particle behaviour to simulate the working process of ball mills, which yields many valuable research outcomes.
Ball mill Moreover, the results obtained from DEM usually should be validated with experiments. In this paper,
simulation results of particle behaviour, mill torque, and power consumption obtained from DEM sim-
ulation are compared with experimental results in detail to validate the correctness of the simulation
results. Especially, the particle behaviour is validated both qualitatively and quantitatively. The DEM
results are shown to be highly consistent with the experimental results. The torque of liners and baffles
are affected by lifter height, lifter number, and mill speed. Moreover, the changes in the torque and
power consumption of a ball mill can be effectively explained using two important factors: lifter and
particle area ratio.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Mill liners protect mill shells from abrasion and lift ore particles
and grinding media to a high position. Therefore, liners/lifters must
be able to bear high-impact loads during the grinding process. The
wear rate of these components is high, and these parts tend to
break or incur wear-out failure, which can seriously affect the pro-
duction efficiency of ball mills. Thus, the research on liner structure
and configuration of ball mills is crucial to improve the production
ability and economy of ball mills. The DEM is a powerful tool to
effectively explain the movement of ore particles and improve
the operation, production, and efficiency of ball mills by optimizing
their structure. Many researchers have used DEM to extensively
study ball mills, and they have obtained valuable results in the last
two decades (Rajamani and Mishra, 1996; Morrison and Cleary,
2004; Djordjevic, 2005; Powell and McBride, 2006; Powell et al.,
⇑ Corresponding author.
E-mail address: lvweidongchn@163.com (W. Lv).
2006; Cleary et al., 2008; Morrison and Cleary, 2008; Cleary,
2009a; Mulenga and Moys, 2014; Cleary, 2015; Geng et al., 2015;
Bbosa et al., 2016).
The structural characteristics of liners/lifters and mill speed
substantially influence particle motion in ball mills and dramati-
cally influence the grinding process and power draw of ball mills.
DEM was first used by Mishra and Rajamani (1992, 1994a,b) to
study the motion of ball charge and the power draw of tumbling
mills in two dimensions; in the study, the effect of liners was also
considered. Cleary (1998) then studied in great detail the effect of
mill speed on charge behaviour, torque, and power draw; although
they obtained instructive results, such results were still based on
two dimensions. Cleary (2001a,b) then made significant progress
by using three-dimensional models to predict the power draw of
semi-autogenous (SAG) mills. Generally, fill level, mill speed, shape
and number of lifters, and particle properties (friction coefficient as
well as shape and size of particles) substantially influence the par-
ticle behaviour and power draw of ball mills (Cleary, 2001c; Hong
and Kim, 2002; Djordjevic, 2003). Cleary et al. (2003) then

http://dx.doi.org/10.1016/j.mineng.2016.12.014
0892-6875/Ó 2016 Elsevier Ltd. All rights reserved.
 X. Bian et al. / Minerals Engineering 105 (2017) 22–35 23

provided an innovative approach to investigating ball mills by 2. Methodology

evaluating the accuracy of the results obtained from 2D and 3D
DEM simulations on the basis of a comparison of the simulation 2.1. Empirical formulas
and experimental results; the researchers also quantitatively stud-
ied how particle shape affects the power draw of ball mills. Many formulas can be used to calculate the net power of a mill.
Hlungwani et al. (2003) adopted the DEM simulator Millsoft (2D) These formulas come in two basic types: those used when a ball
to study the effect of lifter shape and mill speed on power draw mill works at a relatively low speed and those used when a ball
and load on the lifters as well as to verify the results obtained with mill works at a relatively high speed. The dividing line is shown
experiments. Powell and McBride (2004) illustrated media motion in Fig. 1 (Hou, 2015).
and grinding regions (head, departure shoulder, center of circula- When a ball mill works at a relatively low speed, its net power
tion, equilibrium surface, bulk toe, and impact toe). The wear rate can be obtained from Formula (1) as follows:
and service life of liners/lifters can significantly affect the beha-
3
viour of ball mills, whereas the load applied on lifters determines sin ðH=2Þ sin b
Ne ¼ mgrn  ; ð1Þ
the lifespan of lifters (Kalala et al., 2005, 2008). Makokha and 45u
Moys (2006, 2007) and Makokha et al., (2007) analyzed the
where m is the mass of the charge. A schematic of the parameters in
changes in grinding efficiency and productivity when cone lifters
Formula (1) is shown in Fig. 2 (Hou, 2015). Fig. 2 shows the horizon-
are used. Furthermore, a comminution model was built by
tal section of a ball mill, O is the center of the mill section, and OY is
Powell et al. (2007, 2008) to tackle the fundamental causes of rock
the vertical axis. According to Fig. 2, r u is the distance between O
breakage in various comminution devices. Cleary (2009b) made
and the centroid of the charge (M), r is the internal radius of the
major progress by studying a two-chamber cement mill in full
mill, H is the angle of charge (°), and n is the rotational speed (rpm).
scale. Generally, these researchers have performed a significant
When a ball mill works at a relatively high speed, its net power
amount of work and obtained interesting and important results.
consumption can be determined, as indicated in Table 1, where m
Nevertheless, progress can still be made, including in the qualita-
is the mass of the charge, u is the fill level, w is the rotation rate
tive validation of particle behaviour or charge motion.
(constant fraction of the critical speed), q is the density of the gran-
Major progress has also been achieved in the study of particle
ular pile, V is the internal volume of the ball mill, D is the internal
motion in ball mills and validation of the results obtained from
diameter of the ball mill, and dj is the maximum diameter of the
DEM (Powell and Nurick, 1996; Govender et al., 2001; Govender
steel balls.
and Powell, 2004; McBride et al., 2004). Recently, Rezaeizadeh
et al. (2010) conducted a series of experiments and concluded
that increasing mill speed and lifter number as well as reducing 2.2. DEM
fill level can raise the impact load and frequency of particles,
which can in turn improve the efficiency of ball mills. Powell DEM is a numerical method to deal with the kinematic and
et al. (2011) used EDEM software to predict the wear rate of lif- mechanical behaviour of complex granular systems. DEM regards
ters and built a model to study the relationship between liner the discrete system as many discrete units with certain shapes
shape and the efficiency of ball mills. Wang et al. (2012) then and masses. DEM is mainly used in kinematics and mechanics
examined the distribution of collision energy and maximum research on discrete particles, such as rocks, soils, and powders,
impact energy and analyzed how the size of steel particles affects to conduct a complex dynamic calculation of granules under com-
grinding efficiency. Maleki-Moghaddam et al. (2013) simulated plex physical conditions. DEM has been widelyused in mining,
the flow of particles inside a ball mill using GMT software, veri- chemical engineering, and biological medicine, particularly in the
fied the results through experiments, revised the particle trajec- research on ball mills. DEM is a highly convenient approach to
tory equation, and optimized the design of liners. Toor et al. determine the particle behaviour, torque, and power draw of ball
(2013) optimized the design of liners to prolong their life, mills in different conditions. Consequently, DEM is used in the pre-
increase mill production, and decrease power consumption. sent study as a tool to examine the effect of lifters and mill speed
Cleary and Morrison (2016) and Weerasekara et al. (2016) used on the particle behaviour, torque, and power draw of a ball mill.
DEM to study the comminution mechanisms and energy utiliza- The material and numerical parameters used in the DEM model
tion of ball mills. The use of DEM coupled with other numerical are shown in Table 2. The DEM code used in this study is EDEM.
methods to study ball mills has become a hot research topic
(Jonsén et al., 2011, 2012; Beinert et al., 2015; Mayank et al.,
2015).
None of the empirical formulas for computing the power con-
sumption of ball mills considers the effect of lifters (Bond, 1961;
Chen Bingchen, 1981). In the present study, the effect of lifter
height, lifter number, and mill speed on the particle behaviour, tor-
que, and power consumption of ball mills, respectively, are ana-
lyzed quantitatively by DEM and experiments. Specifically,
particle behaviour is validated both qualitatively and quantita-
tively in detail. The results of this study are expected to provide
insights into the motion of particles and may help optimize the
working process of ball mills on the basis of lifter height, lifter
number, and mill speed. The changes in the torque and power con-
sumption of ball mills can be effectively explained using two
important parameters: lifters and particle area ratio. The torque
of liners and baffles is also examined when the operating parame-
ters of ball mills are changed. The results indicate that lifter height,
lifter number, and mill speed affect the magnitude of torque on lin-
ers and baffles. Fig. 1. The dividing line of formulas.
24 X. Bian et al. / Minerals Engineering 105 (2017) 22–35

Y
n

ru O X
β

Θ M
Fig. 3. Experimental setup.

acquisition device (NI USB-6211) can deliver a data stream from

the signal converter to the laptop. Finally, the control module of
the device can be used to switch the device on and off and to adjust
Fig. 2. The parameters of Formula (1). its speed (0–60.87 rpm).
As shown in Fig. 4, the left side of the back end-wall is con-
nected to a flange, which is fixed connected with a transmission
Table 1
The empirical formulas to calculate the power of a ball mill. shaft. Thus, power is transmitted from the shaft to the drum. Inside

Authors Formulas
Chen Bingchen N e1 ¼ m  1 R0:5 w3 ð1 - K 4 Þ½2:25  w4 ð1 þ K 2 Þ
2102 u
Davis N e2 ¼ 2:31qVD0:5 w3 ½6ð1  K 4 Þ  8w4 ð1  K 6 Þ þ 3w8 ð1  K 8 Þ
Olevskiy N e3 ¼ 2:31qVD0:5 f0:94w7 ½16ð1  K 6 Þ  36w4 ð1  K 8 Þ
þ28:8w8 ð1  K 10 Þ  8w12 ð1  K 12 Þg
Levinson N e4 ¼ 7:808qVD0:5 uwð1 þ 0:274w2 Þ
Bond dj 12:5D
N e5 ¼ 1:08m½4:879D0:3 ð3:2  3:0uÞwð1  0:1
Þ þ 1:102
29 104 50:8
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pu
The K used in Table 1 isK ¼ 3 1  2:52w 2.

Table 2
Fig. 4. Drum of the experimental setup.
Parameters used for the DEM simulations.

Density of steel 7800 kg/m3 Table 3

Density of perspex Geometric details of the lifters used in this study.
1200 kg=m3
Poisson’s ratio (steel) 0.25 Number Lifter height (mm) Lifter top width (mm)
Poisson’s ratio (perspex) 0.3
Coefficient of restitution (steel–steel) 0.5 1 25 12
Coefficient of restitution (steel–perspex) 0.4 2 17 12
Static friction (steel–steel) 0.5 3 10 18
Static friction (steel–perspex) 0.5 4 5 18
Rolling friction (steel–steel) 0.01 5 3 18
Rolling friction (steel–perspex) 0.01

2.3. Experiment

A laboratory-scale ball mill with an internal diameter of

573 mm and axial length of 160 mm is used in this study. The com-
ponents of the experimental setup are shown in Fig. 3. The rated
power of the motor is 1.5 kW and the rated speed is 1400 rpm. A
reducer is used to decrease the speed of the shaft, and the reduc-
tion ratio is 23; thus, the rotation speed of the shaft is reduced to
60.87 rpm after the reducer. A torque and speed tester can be used
to indicate the torque and rotational speed signals in the form of
frequency signals. A signal converter can be used to transform
these signals from frequency to standard voltage. A laptop can
indicate and record the numerical information of torque and rota-
tional speed. The drum of the experiment setup is the most impor-
tant component of the device. The DC-regulated power supply
provides 24 V DC voltage for the torque and speed tester. The data Fig. 5. The structure of drum.
 X. Bian et al. / Minerals Engineering 105 (2017) 22–35 25

the cylinder shell are 24 rows of bolt holes that are spaced circum- and back). The front end-wall consists of two pieces of transparent
ferentially around the mill shell. The inner side of the cylinder shell semicircular perspex. The behaviour of the particles in the drum
and the lifters installed on it act as liners and lifters. For conve- can be observed clearly through the perspex. The material of the
nience during replacement, lifters are machined independently back end-wall is steel (same as the cylindrical shell). The two
and fixed with the cylinder shell in the form of bolt connections. end-walls are both flat and featureless. The structure of the end-
The parameters of the lifter bars are shown in Table 3. The length walls exerts a significant effect on the charge behaviour in grinding
of the lifter bars covers the back end-walls to the front end-walls. mills (Govender et al., 2013). Therefore, the torque of end-walls
The detailed structure of the drum is shown in Fig. 5. The drum independent from the drum is worth studying. The aspect ratio
is composed of a cylindrical shell, lifters, and two end-walls (front of a SAG mill is usually low (the aspect ratio of a product from
CITIC Heavy Industries Co. Ltd. is 0.49, £11 m  5.4 m), and the
Table 4
The diameter distribution of steel balls.
aspect ratio of the experimental mill used in this study is 0.28.
Although the aspect ratio of the experimental mill is lower than
Diameter (mm) Number Percent by volume (%) that of an industrial-scale SAG mill, studying the behaviour of ball
20 111 5.68 mills using a lab-scale ball mill remains interesting, and the results
15 795 17.16 obtained still have some instructive value in industry applications
13 3051 42.87
(McBride et al., 2004; Powell and McBride, 2004; Govender et al.,
8 10474 34.29
2013; Cleary and Morrison, 2016).

Fig. 6. Effect of mill speed on particle behaviour for the experimental ball mill for different rotation rates: (a) w = 0%, (b) w = 60%, (c) w = 70%, (d) w = 75%, (e) w = 80%,
(f) w = 90%.
26 X. Bian et al. / Minerals Engineering 105 (2017) 22–35
In conducting the experiment, one of the semicircle perspex is
removed, and the steel balls are placed in the drum. The distribu-
tion of different steel balls is shown in Table 4, and the ball mixes
used for experiment and simulation are exactly same. Then, the
removed perspex is installed, and the control module is switched
on. The mill speed is adjusted through the control module, and
the rotational speed and torque data are recorded on a laptop. A
high-speed digital camera is used to record particle motion and
send pictures and videos to the computer for further analysis.
Finally, the device is switched off, and power outage is determined.
3. Experimental validation of empirical formulas and DEM
simulations
3.1. Particle behaviour
In this section, Hl (lifter height) = 10 mm, N l (lifter number)
= 24, and the fill level is 35%. Fig. 6 shows the comparison between
the DEM simulation and experiment results of particle behaviour
when the mill is operated in six typical rotation rates. The particles
are colored according to their speed, with red indicating high speed
and blue indicating low speed. The particle behaviours in the sim-
ulation and experiment are highly consistent when the mill speed
is changed. Consequently, the DEM simulation result for the parti-
cle behaviour of the ball mill is accurate and reliable. Cleary and
Hoyer (2000) studied mill charge motion by comparing the DEM
predictions with their experiment; focusing on fill level, they found
that the simulation result fits the experiment well when the fill
level is changed. A comparison of DEM and experiment results
for a scale model of a SAG mill was conducted by Cleary et al.
(2003), and the variables they studied were particle shape and
mode of 2D and 3D simulations. In the present study, we use mill
speed as the variable for validating particle behaviour.
Fig. 6 shows that when w ¼ 0% (Fig. 6a), the particles are static,
and when w ¼ 60% (Fig. 6b), the particle pile shows a relatively
clear free surface, similar to the figure in the paper by Govender
et al. (2013). When the ball mill is rotated, the particles move from
the bottom of the drum to the shoulder position on the upper left
part of the drum. The particles at the shoulder position begin to
slip and eventually slide across the free surface to the toe position
(defined by Owen and Cleary, 2015) at the lower right part because
of gravity, thereby completing the cycle movement. The particles
that slide downward across the free surface are called cascading
particles (Cleary, 2001c). At w ¼ 70% (Fig. 6c), w ¼ 75% (Fig. 6d),
and w ¼ 80% (Fig. 6e), the particle behaviour is substantially the
same as that when w ¼ 60%, although some of the particles at
Fig. 7. Ball mill sections with grids overlaid.
the shoulder position eventually fall on the toe position, and the
shoulder position becomes slightly high. The particles that are
thrown from the shoulder position are called cataracting particles
(Cleary, 2001c). The number of cataracting particles increases
when the mill speed increases gradually.
When the rotation rate of the ball mill increases to w ¼ 90%
(Fig. 6f), the quantity of the cataracting particles increases sharply,
covering almost half of its total quantity. Moreover, almost the
entire cross section of the drum is covered by high-speed particles.
The exact shoulder and toe positions become indistinct because
the particles cover almost the entire cross section of the drum.
Additionally, the color of the particles that are in contact with
the liners/lifters gradually becomes jacinth, which means that
the speed of these particles further increases. Large quantities of
particles collide on the liners/lifters directly when w ¼ 90%
(Fig. 6f). This collision leads to an excessive increase in the wear
rate of liners/lifters. Naturally, this condition should be avoided
in industries.
The ability of DEM to predict particle behaviour can be vali-
dated qualitatively in Fig. 6. However, a qualitative validation
alone is insufficient. Therefore, validating the particle behaviour
quantitatively is necessary. Inspired by the idea that grids can be
overlaid on the picture of a ball mill section (Govender and
Powell, 2004; McBride et al., 2004; Govender et al., 2013), we per-
form the same method in this study, as shown in Fig. 7. However,
our goal is to use the grids to determine la (particle area ratio),
which is given by
la ¼ Sp =Sm ; ð2Þ
where Sp is the area occupied by the particles in the section of the
ball mill and Sm is the area of the entire mill section.
Given that the parameters used in the simulation and experi-
ment are exactly same, we can confirm that the magnitudes of
la in the results of the simulation and experiment should be
almost the same if the simulation result is correct. Sp is obtained
through the number of grids occupied by the particles; the same
approach is used to determine Sm . Then, we can obtain a series of
magnitudes of la when the ball mill rotates at different speeds,
as shown in Fig. 8. To avoid random errors, we use Sp and Sm as
the average values of five magnitudes (when the ball mill rotates
at a special speed, five pictures are captured randomly; we can
then obtain Sp and Sm from each picture and thus determine five
magnitudes for Sp and Sm ).
According to the previous analysis, when the mill speed is
extremely low (w 6 60%), a large proportion of particles move at
a relatively low speed and conduct low-energy grinding but
 X. Bian et al. / Minerals Engineering 105 (2017) 22–35
Fig. 8. Validation of particle behaviour quantitatively.
100
Simulation result
Experimental result
80
60
Torque (Nm)
40
20
0
55 60 65 70 75 80 85 90
Rotation rate (%)
Fig. 9. Average torque as a function of mill speed for the experimental ball mill.
without effective high-energy impact, which reduces the produc-
tivity of the ball mill. However, when the mill speed is extremely
high (w P 90%), large quantities of particles move at high speed
and collide with the liners/lifters directly, thereby increasing the
wear rate of the liners/lifters. Thus, liners require frequent changes,
and considerable economic losses occur. The rotation rate should
be properly set for ball mills to work efficiently and economically.
Fig. 8 shows that the particle area ratios in the DEM simulation and
experiment are almost the same when the rotation rate changes
(the maximum difference is 4.8% when w ¼ 80%). Moreover, the
particle area ratio shows a degree of direct proportionality to the
rotation rate. All the particle area ratios in the experiment are
slightly lower than those in the DEM simulation because the end
of the lifter bars is not clear in the section of the experimental ball
mill and simplification was performed when determining Sm . Gen-
erally, the DEM simulation result for the particle behaviour of the
ball mill is highly accurate and reliable through qualitative and
quantitative validation. Furthermore, the method proposed in this
paper can be used to address the problem in selecting the speed of
ball mills.
3.2. Mill torque and power
Fig. 9 shows the change in mill torque with rotation rate when
the fill level is 35%, N l ¼ 24, and Hl ¼ 17 mm. Generally, the results
27
indicate that the DEM simulation and experimental results are
highly similar, except when w ¼ 90% (the difference is 10.1%,
which is attributed to the fact that when the mill works at a high
speed, it causes serious vibration, and the precision of the torque
and speed tester decreases). This characteristic proves that DEM
is a good method for predicting the torque of ball mills. When
the mill speed is raised, the torque increases until it reaches the
peak point of the curve at the rotation rate between 70% and
75%; it gradually decreases when w P 75%. These results are sim-
ilar to those obtained by Cleary (1998), that is, the mill torque
increases marginally until it reaches the peak when w ¼ 80%,
which is higher than the result reported in this paper.
Fig. 10 shows the power of the ball mill, which is determined by
using different methods with varying mill speeds. In this study, the
experiment result is selected as the standard power. When the fill
level is 35%, the ball mill works at a low-speed zone when
w 6 68%, and when w > 68%, the ball mill works at a high-speed
zone (Fig. 1). Thus, for a speed of w ¼ 60% (Fig. 10a), the speed of
the ball mill is in a low-speed zone, in which case formula (1)
should be used. The result of formula (1) is much higher than the
results of the simulation and experiment. This difference indicates
that formula (1) is not suitable for use in this condition, whereas the
results of the DEM simulation and experiment are quite similar.
For a mill speed of w ¼ 70% (Fig. 10b), the results obtained
using the Chen Bingchen formula, Levinson formula and DEM are
similar to the standard result, whereas the results acquired with
other formulas are markedly different from the standard result.
When w ¼ 75% (Fig. 10c), the conclusion is similar to that for
w ¼ 70%, except that the result obtained with the Olevskiy formula
is relatively close to the standard result (the difference changes
from 24.3% to 13.5%). For a speed of 80% (Fig. 10d), all the results
obtained with the aforementioned formulas are close to the stan-
dard result, except for the result obtained with the Bond.
formula (47.1% higher than the standard result). This result
demonstrates that the accuracy of empirical formulas is the high-
95 est when the rotation rate is approximately 80%. In industries, ball
mills usually operate at speeds near 80%, which indicates that the
given formulas are appropriate. For a speed of 90% (Fig. 10e), the
deviation of the empirical formulas is larger than that of the stan-
dard result, whereas the DEM simulation result is relatively close
to the standard result. Such an outcome demonstrates that
employing empirical formulas is not appropriate under this condi-
tion and should be replaced with DEM simulation.
The accuracy of different empirical formulas changes when the
rotation rate varies. The accuracy of empirical formulas is obvi-
ously lower than that of the DEM simulation when mill speed
changes. This difference can be mainly explained by the fact that
the effect of lifters is not considered in the empirical formulas.
When the rotation rate is 60%–90%, the DEM simulation results
and standard results are noticeably similar. This similarity demon-
strates that DEM is an accurate numerical calculation method that
can predict the power consumption of ball mills more accurately
than other methods could.
4. Results and discussion
4.1. Influence of lifter height and mill speed
Liners/lifters gradually experience wear and fatigue failure
because of the effect of impact and erosion. During the working
process, lifter height gradually declines and becomes too low to
bring the particles to a sufficiently high location. This condition
adversely affects the working efficiency of ball mills. Under this
condition, liners must be replaced to recover the working ability
of the ball mills. In this section, we employ DEM simulation to
28 X. Bian et al. / Minerals Engineering 105 (2017) 22–35
Fig. 10. The comparison of ball mill power.
study the effect of lifter height and mill speed on the particle beha-
viour, torque, and power of a ball mill. The parameters in this sec-
tion are as follows: fill level is 35%, w ¼ 75%, N l ¼ 24, and lifter
height changes from 0 mm to 25 mm.
4.1.1. Particle behaviour
The falling trajectory of the particles in the present study is sim-
ilar to that in the work of Cleary (2001c), and the particle streams
are clear and dense, as shown in Fig. 11a. As a result of abrasion,
the lifter height gradually decreases, the particle streams become
sparse, and the quantity of high-speed particles decreases
(Fig. 11b). For Hl ¼ 10 mm (Fig. 11c), the toe position moves
slightly lower, and the quantity of high-speed particles further
decreases. At Hl ¼ 5 mm (Fig. 11d) and Hl ¼ 3 mm (Fig. 11e), extre-
mely few cataracting particles exist, and most of the particles flow
in low-speed phase. Thus, productivity gradually decreases, and
the liners require replacement. Fig. 11f depicts the motion of par-
ticles when Hl ¼ 0 mm (no lifter). Compared with the particle
behaviour in Fig. 11a, that in Fig. 11f has remarkably changed, indi-
cating the significant effect of lifter height. Owen and Cleary (2015)
found that the strongest effect of lifter height lies on the extent and
location of the cataracting stream; these findings are generally
similar to ours, except that they set three different types of lifter
height and mainly used a streak pattern instead of particle visual-
ization in their study.
4.1.2. Variation of torque and power with lifter height and mill speed
All the magnitudes of torque decrease as the mill speed
increases, except for the torque corresponding to Hl ¼ 0 mm
(Fig. 12). When lifter height is high, the corresponding torque
begins to decline earlier (torque line begins to decline at a speed
of 60% when Hl ¼ 25 mm, at a speed of 70% when Hl ¼ 17 mm, at
a speed of 75% when Hl ¼ 5 mm, and at a speed of 80% when
Hl ¼ 3 mm; the torque curve increases continually as the mill
speed increases when Hl = 0 mm). In addition, at w P 80%, a large
torque is needed for the ball mill when the lifter height is low.
When the drum rotates, the particles are forced to flow by the
lifting force of the lifters, the friction of the liners, and the friction
of the front and back end-walls. The specific torque can be
obtained by multiplying each force with the corresponding dis-
tance to the center of the drum. The total torque provided by the
lifters and liners is called liner torque. In addition, the total torque
provided by the front and back end-walls is called baffle torque.
Fig. 13 shows the effect of lifter height on the liner torque and
baffle torque of the experimental ball mill at different mill speeds.
The upper six curves denote the liner torque, whereas the six lower
curves denote the baffle torque. For w ¼ 60%, the liner torque is
approximately directly proportional to lifter height. As the mill
speed increases to 70%, all the magnitudes of the liner torque
increase, except for the curve corresponding to Hl ¼ 25 mm. The
liner torque of Hl ¼ 17 mm is the maximum torque among the
six torque curves when w ¼ 70%. At w ¼ 75%, all the torque curves
 X. Bian et al. / Minerals Engineering 105 (2017) 22–35 29

Fig. 11. Effect of lifter height on particle behaviour for the experimental ball mill: (a) Hl = 25 mm, (b) Hl = 17 mm, (c) Hl = 10 mm, (d) Hl = 5 mm, (e) Hl = 3 mm and (f)
Hl = 0 mm.

75 roughly reach the maximum, except for the curves corresponding

to Hl ¼ 25 mm and Hl ¼ 0 mm. At w ¼ 80%, only the liner torque
70
of Hl ¼ 3 mm and Hl ¼ 0 mm continues to increase, whereas the
65 other curves of the liner torque start to decline. The liner torque
that corresponds to Hl ¼ 3 mm becomes the maximum among all
60
the curves when w ¼ 90% and w ¼ 100%. Moreover, the mill works
Torque (Nm)

55 at a relatively high-speed state under this condition, and the space

between different curves enlarges. In this condition, the liner tor-
50 que also becomes approximately inversely proportional to the lif-
45
LH=25mm ter height (except Hl ¼ 0 mm). All the six baffle torque curves are
LH=17mm
LH=10mm
approximately inversely proportional to the mill speed. In addition,
40 when the ball mill rotates at a specific speed, the baffle torque is
LH=5mm
35 LH=3mm roughly inversely proportional to the lifter height. Furthermore,
LH=0mm the baffle torque curve corresponding to Hl ¼ 0 mm is substantially
30 larger than the other curves.
55 60 65 70 75 80 85 90 95 100 105
Generally, torque behaviour can be explained by two factors,
Rotation rate (%)
namely, lifter and particle area ratio. Baffle torque is approximately
Fig. 12. Effect of lifter height on the torque of the experimental ball mill at different inversely proportional to the particle area ratio. Liner torque is
rotation rate between 60% and 100%. determined by the effect of the particle area ratio and lifter.

70
Liner, LH=25mm
Liner, LH=17mm
60 Liner, LH=10mm
Liner, LH=5mm
50 Liner, LH=3mm
Liner, LH=0mm
Torque (Nm)

40 Baffles, LH=25mm
Baffles, LH=17mm
Baffles, LH=10mm
30 Baffles, LH=5mm
Baffles, LH=3mm
20 Baffles, LH=0mm

10

0
55 60 65 70 75 80 85 90 95 100 105
Rotation rate (%)

Fig. 13. Effect of lifter height on the liner torque and baffle torque of the experimental ball mill at different mill speed.
30
Moreover, the lifter is a decisive factor when a ball mill works at a
relatively low speed; a high lifter height causes a large liner torque.
By contrast, the particle area ratio is the main factor when a ball
mill works at a relatively high speed; an increased particle area
ratio causes a small liner torque.
When a ball mill rotates at a low speed, almost all the particles
move at a low speed, and the particle area ratio is low. However, an
increasing number of particles behave in a high-speed state as the
mill speed increases. The overall distribution of particles becomes
sparse, and many particles become cataracting particles. The parti-
cle area ratio then increases gradually. Fig. 13 shows that the liner
torque is approximately directly proportional to the lifter height
when the mill speed is low (w 6 70%), particularly because the lif-
ter is the decisive factor when the mill rotates at a low speed. A
large number of particles are lifted, and the impact between the
particles and liners intensifies when high lifters are used. Thus, a
ball mill with high lifters requires a large torque to run the mill.
As the mill speed continues to increase, all the curves of the liner
torque gradually decline (except Hl = 0 mm). Besides, lifter height
is higher, when w P 75%, the earlier liner torque curve begin to
decline. In addition, the liner torque is roughly inversely propor-
tional to the lifter height when w P 80%(except Hl ¼ 0 mm)
because the particle area ratio becomes the decisive factor with
the improvement of mill speed. Therefore, the corresponding par-
ticle area ratio is high when the lifter height is high; the liner tor-
que decreases as a result.
When a ball mill rotates at a specific speed, the particle area
ratio is known to increase as the lifter height increases, and the
baffle torque decreases when the particle area ratio increases.
Thus, baffle torque is approximately inversely proportional to lifter
height. Similarly, baffle torque is approximately inversely propor-
tional to mill speed. In addition, when Hl ¼ 0 mm (no lifter exists),
most of the particles are low-speed particles even whenw ¼ 100%.
Thus, baffle torque can remain almost unchanged when mill speed
increases.
Fig. 14 shows the change of mill power with mill speed when
the lifter height is changed. When the ball mill rotates at a rela-
tively low speed (w 6 70%), the power curves are quite similar
(except Hl = 0 mm). By contrast, the power curves corresponding
to different lifter heights are gradually and clearly separated when
the ball mill rotates at a relatively high speed (w P 75%). This rela-
tionship is attributed to the lifter height slightly affecting the
power draw of a mill when the mill works at a relatively low speed.
However, lifter height strongly affects the power draw of a mill
when the mill works at a relatively high speed. The power draw
is roughly inversely proportional to the lifter height when
380
LH=25mm
360
LH=17mm
340 LH=10mm
LH=5mm
320 LH=3mm
LH=0mm
300
Power (W)
280
260
240
220
200
180
55 60 65 70 75 80 85 90
Rotation rate (%)
Fig. 14. Effect of lifter height and mill speed on power.
X. Bian et al. / Minerals Engineering 105 (2017) 22–35
w P 75% (except Hl ¼ 0 mm) because the particle area ratio is
the decisive factor in this condition. The high speed can cause
the particle area ratio to increase, and both the liner torque and
the baffle torque decrease as the mill speed increases, thereby
reducing the mill power. Fig. 12 indicates that almost all the torque
curves reach their peak points near the speed of w ¼ 80%, whereas
the power curves reach the peak points near the speed of w ¼ 90%.
Moreover, the mill power is obtained from the torque multiplied
by the corresponding rotation speed. Thus, increasing the rotation
speed delays the decline of the power consumption of a mill at a
speed of w ¼ 90% although the mill torque begins to decline at a
speed of w ¼ 80%. The changes in the mill power with mill speed
when the lifter face angle changes were studied by Cleary
(2001c); the results showed that the power rises with the mill
speed and reaches the top when w ¼ 100% before it declines signif-
icantly. The difference is that the top point appears at w ¼ 100%,
whereas in our study, the power is maximum when w ¼ 90%.
4.2. Effect of lifter number and mill speed
Lifter number is an important parameter that needs to be care-
fully considered in designing a ball mill. However, no theory has
been developed to determine lifter number. Thus, lifter number
is generally decided according to empirical knowledge and anal-
ogy. The effect of lifter number on the particle behaviour, torque,
and power of a ball mill is examined in this study. In this section,
the fill level of the ball mill is 35%, w ¼ 80%, Hl ¼ 10 mm, and the
lifter number is changed from 0 rows to 24 rows.
4.2.1. Particle behaviour
Fig. 15 depicts the change in particle behaviour when the lifter
number changes from 0 to 24. Fig. 15a shows the particle beha-
viour when N l ¼ 0. A majority of the particles move at a low speed
and cannot collide with high energy. Thus, the ball mill works at a
relatively low efficiency. When the lifter number is increased to
three, the quantity of the high-speed particles relatively increases,
and a sparse particle stream is thrown from the shoulder position
back down to the toe position (Fig. 15b). Two particle streams exist
when the lifter number is increased to six. The quantity of cataract-
ing particles is excessively low, thereby causing a discontinuous
high-energy impact and low working efficiency (Fig. 15c).
Fig. 15d shows the motion of particles when N l ¼ 12. Evidently,
the quantity of high-speed particles increases, and three sparse
particle streams exist. However, the working efficiency is relatively
low because the quantity of cataracting particles is still relatively
small. At N l ¼ 24, as depicted in Fig.15e, the quantity of high-
speed particles increases, and many relatively coherent particle
streams exist. Furthermore, in this condition, the particles suffer
from both low-energy abrasion and continuously efficient high-
energy collision. Thus, the ball mill works at a relatively high effi-
ciency state. Consequently, the lifter number of the ball mill used
in this study should be approximately 24 for the mill to function
well. The work of Cleary (2001c) mainly explored the effect of lifter
pattern on charge shape (N l ¼ 12, 24 was studied).
In summary, the lifter number of ball mills crucially affects par-
ticle behaviour. The cyclic process of particles in which the parti-
cles move up from the toe position to the shoulder position and
are thrown from the shoulder back down to the toe position is clo-
sely related to the lifter number. The lifting ability of lifters
decreases with the reduction of the lifter number, thereby causing
the majority of particles to move in a low-speed condition. Thus,
95 100 105 the kinetic energy of particles is extremely low that high-energy
collisions cannot effectively occur. As a result, the breaking and
grinding ability of ball mills decreases. Therefore, the lifter number
should be selected reasonably in the practical design and usage of
 X. Bian et al. / Minerals Engineering 105 (2017) 22–35 31

Fig. 15. Effect of lifter number on particle behaviour: (a) N l = 0, (b) N l = 3, (c) N l = 6, (d) N l = 12, (e) N l = 24.

ball mills, and a relatively increasing lifter number is favorable, as
indicated by the results of this study.
4.2.2. Variation of torque and power with lifter number and mill speed
Fig. 16 shows the change in torque with the variation of lifter
number when the ball mill is operated in different rotation rates.
Generally, the torque tends to increase when the rotation rate
changes from 60% to 80%, and the slope becomes small when the
lifter number increases. When the ball mill rotates at a relatively
high speed (w P 90%), the torque increases first and then
decreases when N l P 12. Furthermore, when w ¼ 100%, the curve
of the torque declines rapidly when N l P 6 because a large quan-
tity of centrifugal particles exist in the mill under this condition.
Thus, the effects between particles and liners are weak, and the
particle area ratio increases. As a result, the torque of the mill
decreases.
Fig. 17 shows that when the rotation rate increases from 60% to
90%, the magnitude of the liner torque corresponding to a large lif-
70
68
66
64
Torque (Nm)
ter number is greater than that corresponding to a small lifter
number because the lifter is the decisive factor in this condition.
The interaction of lifters and particles is high when a large number
of lifters exist. Under this condition, a large liner torque is required
to run the mill. However, when the rotation rate is between 95%
and 100%, the magnitudes of the liner torque become approxi-
mately inversely proportional to the lifter number because the par-
ticle area ratio becomes the decisive factor. Numerous cataracting
particles exist when a large number of lifters exist, thereby increas-
ing the particle area ratio and decreasing the liner torque.
The character of the baffle torque shown in Fig. 17 is highly sim-
ilar to that in Fig. 13. The magnitudes of the baffle torque decrease
when the rotation rate increases because the particle area ratio
increases with the increase in mill speed, thereby reducing the baf-
fle torque. Additionally, the baffle torque is approximately inver-
sely proportional to the lifter number, which means that an
increase in the lifter number decreases the corresponding baffle
torque. This relation is attributed to the fact that when a mill works
at a specific speed, the particle area ratio is high when the lifter
number is high, in which case the baffle torque is reduced.
Fig. 18 shows the change in mill power with lifter number when
the mill rotates at a rotation rate between 60% and 100%. The mill
power increases with an increase in the lifter number when the
mill rotates at a rotation rate from 60% to 80%. When the mill
works at a relatively high speed (w P 90%), the mill power
increases first and then decreases gradually with an increase in
the lifter number. This outcome occurs because when the mill

62 RR=60% rotates at a high speed and with a large number of lifters, the par-
RR=70% ticle area ratio increases and thus causes the mill power to gradu-
RR=75%
60 RR=80% ally decrease. In addition, when the lifter number increases from 0
RR=90% to 24, the mill power is approximately directly proportional to the
RR=100%
58 rotation rate (except when w P 90%, in which case the power
curve begins to decline when N l P 5).
56

54 4.3. Comprehensive effect of lifter height, lifter number, and mill speed
0 5 10 15 20 25
on torque and power
Lifters number (row)

Fig. 16. Effect of lifter number on the torque of the experimental ball mill at According to the preceding analysis, mill speed, lifter height,
different rotation rate between 60% and 100%. and lifter number significantly affect the particle behaviour,
32
60
50
40
Torque (Nm)
30
20
10
0
55 60 65
Fig. 17. Effect of lifter number on the liner torque and baffle torque of the experimental ball mill at different mill speed between 60% and 100%.
400
300
Power (W)
200
100
0
0 5 10 15
Lifternumber (row)
Fig. 18. Effect of lifter number on the power of the experimental ball mill at
different mill speed.
torque, and power of a ball mill. The comprehensive effect of mill
speed, lifter height, and lifter number on the torque and power
of a ball mill is discussed using DEM in the following section.
Fig. 19 shows the effect of rotation rate, lifter height, and lifter
number on the torque of the ball mill. For N l ¼ 24 (Fig. 19a), the
torque curve corresponding to Hl ¼ 25 mm decreases with an
increase in the rotation rate. The curve shape is a standard quadra-
tic parabola. The torque curves corresponding to Hl ¼ 10 mm and
Hl ¼ 3 mm increase first and then decrease with increase in the
rotation rate. Unlike in other conditions, the torque curve corre-
sponding to Hl ¼ 0 mm is approximately directly proportional to
the rotation rate. At N l ¼ 12 (Fig. 19b), the distance between each
torque curve becomes small and shows the same changing trend in
comparison to that at N l ¼ 24, which is roughly and directly pro-
portional to the rotation rate when the ball mill works at a low
speed and is inversely proportional to the rotation rate when the
ball mill works at a high speed. The difference is that the magni-
tude of the torque corresponding to Hl ¼ 0 mm becomes the sec-
ond largest value when w ¼ 100%, and this magnitude is close to
the maximum (Hl ¼ 3 mm) and is significantly greater than the
magnitudes of the torque corresponding to Hl ¼ 10 mm and
Hl ¼ 25 mm.
At N l ¼ 6 (Fig. 19c), the trends of the whole curves do not
change relative to those shown in Figs. 19a and 19b. However,
the sensitivity of the torque curves to the rotation rate decreases,
X. Bian et al. / Minerals Engineering 105 (2017) 22–35
Liner, LN=24
Liner, LN=12
Liner, LN=6
Liner, LN=3
Liner, LN=0
Baffles, LN=24
Baffles, LN=12
Baffles, LN=6
Baffles, LN=3
Baffles, LN=0
70 75 80 85 90 95 100 105
Rotation rate (%)
indicating that the magnitudes of torque vary slightly when the
rotation rate increases. At N l ¼ 3 (Fig. 19d), the tendency of the
curves changes relative to those shown in Figs. 19a–19c. The curve
of the torque corresponding to Hl ¼ 3 mm increases constantly
with the rotation rate. The curve corresponding to Hl ¼ 10 mm
only decreases when w ¼ 100%. When w ¼ 100%, the order of the
torque changes to Hl ¼ 10 mm, Hl ¼ 3 mm, and Hl ¼ 0 mm, and
the magnitude of the torque corresponding to Hl =25 mm is much
RR=60% smaller than the former magnitudes. Additionally, all the torque
RR=70% magnitudes are directly proportional to the rotation rate, except
RR=75% for the one corresponding to Hl = 25 mm.
RR=80% The result in Fig. 19 indicates that when the lifter number
RR=90% changes from 24 to 3, the torque of the ball mill and lifter height
RR=100% exhibit a degree of direct proportionality when the ball mill runs
at a relatively low speed (60% 6 w 6 70%). The reason is that the
20 25 lifter is the decisive factor when a mill runs at a relatively low
speed and that a high lifter height can increase mill torque. The tor-
que curves basically decrease when w P 80%. In addition, when
the lifter is high, the torque curve begins to decline early. Mean-
while, the torque is roughly inversely proportional to the lifter
height when w ¼ 100% (except for the curves of the torque corre-
sponding to Hl ¼ 0 mm and Hl ¼ 3 mm in Fig. 19d) because the
particle area ratio becomes the decisive factor when a ball mill
rotates at a high speed. The particle area ratio is high when the lif-
ter height is increased, and this condition decreases the corre-
sponding torque of the ball mill. For Hl ¼ 0 mm, the torque of the
mill is approximately directly proportional to the mill speed when
the rotation rate changes from 60% to 100%. Moreover, the slope is
significantly small because increasing the mill speed can
strengthen the impact between the liners and the particles and
can slightly affect the particle area ratio. Thus, the torque of the
mill gradually increases with the improvement of the mill speed.
Additionally, Fig. 19 shows that the difference between the torque
curves gradually decreases when the lifter number is reduced from
24 to 3. This result indicates that the effect of the lifter height on
the torque of the ball mill becomes small when the lifter number
decreases.
The power of a ball mill is obtained by multiplying the corre-
sponding torque with the mill speed. Thus, the torque curves and
power curves are similar. Fig. 20 shows 78 groups of DEM simula-
tion results for the effect of rotation rate, lifter height, and lifter
number on the power draw of the ball mill. Each curve represents
the variation of the mill power with the rotation rate under a speci-
fic lifter height and lifter number.
At N l ¼ 24 (Fig. 20a), the power curves corresponding to differ-
ent lifter heights increase first, reach the maximum when the rota-
tion rate is roughly 90%, and then gradually decrease with the
 X. Bian et al. / Minerals Engineering 105 (2017) 22–35 33

(a) LN=24 (b) LN=12

70 70

65 65

Torque (Nm)

Torque (Nm)
60 60

55 55

50 50
LH=25mm LH=25mm
LH=10mm LH=10mm
45 LH=3mm 45 LH=3mm
LH=0mm LH=0mm

40 40
55 60 65 70 75 80 85 90 95 100 55 60 65 70 75 80 85 90 95 100
Rotation rate (%) Rotation rate (%)
(c) LN=6 (d) LN=3
70 70

65 65

Torque (Nm)
Torque (Nm)

60 60

55 55

50 50
LH=25mm LH=25mm
LH=10mm LH=10mm
45 LH=3mm 45 LH=3mm
LH=0mm LH=0mm

40 40
55 60 65 70 75 80 85 90 95 100 55 60 65 70 75 80 85 90 95 100
Rotation rate (%) Rotation rate (%)

Fig. 19. Effect of lifter height and lifter number on the torque of the experimental ball mill at different mill speed between 60% and 100%.

(a) LN=24 (b) LN=12

390 390

360 360

330 330
Power (W)
Power (W)

300 300

270 270

240 LH=25mm 240 LH=25mm

LH=10mm LH=10mm
210 LH=3mm 210 LH=3mm
LH=0mm LH=0mm
180 180
55 60 65 70 75 80 85 90 95 100 55 60 65 70 75 80 85 90 95 100
Rotation rate (%) Rotation rate (%)
(c) LN=6 (d) LN=3
390 390

360 360

330 330
Power (W)

Power (W)

300 300

270 270

240 LH=25mm 240 LH=25mm

LH=10mm LH=10mm
210 LH=3mm 210 LH=3mm
LH=0mm LH=0mm
180 180
55 60 65 70 75 80 85 90 95 100 55 60 65 70 75 80 85 90 95 100
Rotation rate (%) Rotation rate (%)

Fig. 20. Effect of lifter height and lifter number on the power of the experimental ball mill at different mill speed between 60% and 100%.

further increase of rotation rate (except Hl ¼ 0 mm). At N l ¼ 12 At N l ¼ 6 (Fig. 20c), only the power curve corresponding to
(Fig. 20b), the power curves corresponding to Hl ¼ 0 mm and Hl ¼ 25 mm continues to show the trend of increasing first and
Hl ¼ 3 mm increase constantly with the increase of the mill speed. then decreasing. The other power curves exhibit a constant
34 X. Bian et al. / Minerals Engineering 105 (2017) 22–35
increase with the increase of the rotation rate. When the lifter
number is reduced to three (Fig. 20d), all the power curves are
approximately directly proportional to the rotation rate. Only the
power curve corresponding to Hl ¼ 25 mm begins to decrease
when w P 90%. In addition, when w ¼ 100%, the magnitudes of
power are quite close, except for the one corresponding to
Hl ¼ 25 mm.
Additionally, when Hl ¼ 0 mm, the power curves, which change
with the rotation rate, are the same in the four graphs in Fig. 20.
With this curve as the reference curve, the power curves corre-
sponding to the other lifter height approach this curve as the lifter
number decreases. The power curves corresponding to Hl of 25, 10,
and 3 mm decrease with the decrease of the lifter number when
w 6 90% and increase with the decrease of the lifter number when
w P 90%. These results illustrate that with a decrease in the lifter
number, the effect of lifter height on the power consumption of a
ball mill gradually weakens.
5. Conclusion
In this study, the effects of mill speed, lifter height, and lifter
number on the particle behaviour, torque, and power of a ball mill
are examined by combining DEM simulation and experiment vali-
dation. The conclusions based on the observations, calculations,
and analysis are as follows:
 Mill speed significantly affects particle behaviour. The DEM
simulation results for particle behaviour are highly consistent
with the experimental results when the mill rotates at various
speeds (from qualitative and quantitative validation). Particle
trajectory is reasonable near the rotation rate of 75%–80%.
 Under a specific mill speed, lifter height affects the falling tra-
jectory of particles. When the lifter height is too low, most of
the particles move at a relatively low speed and consequently
decrease the work efficiency of the ball mill. However, the fall-
ing trajectory of the particles becomes too high, and many par-
ticles directly hit the liners and lifters when the lifter height is
too high. This condition can accelerate the wear rate of liners
and lifters. The reasonable range of lifter height is 10–17 mm
in the ball mill used in this experiment. Additionally, the lifter
number mainly affects the number of cataracting particles and
the characteristic of particle streams. The number of cataracting
particles increases, and the particle streams become increas-
ingly dense and continuous when the lifter number increases.
In addition, the ball mill used in this study works effectively
when N l ¼ 24. Thus, increasing the lifter number appropriately
is a good approach to improve the working efficiency of a ball
mill.
 The simulation and experimental results show that the torque
of a ball mill increases first and then decreases during the pro-
cess of improving the rotation rate. Furthermore, when the lif-
ters are high, the corresponding curve of the torque begins to
decline early. Moreover, the torque of the ball mill and the lifter
height exhibit a degree of direct proportionality when the ball
mill runs at a relatively low speed (w 6 70%) and a degree of
inverse proportionality when the ball mill runs at a relatively
high speed (w P 90%). When other factors remain constant,
the torque of a ball mill usually increases with an increase in lif-
ter number and finally becomes stable.
 The relationship between baffle torque and particle area ratio
shows a degree of inverse proportionality. Thus, any measure
that can increase the particle area ratio decreases the baffle tor-
que; examples include increasing mill speed, increasing lifter
number and lifter height, and so on. The liner torque depends
on two factors, namely, lifter (including lifter number and lifter
height) and particle area ratio. The former one is the decisive
factor at a low mill speed, whereas the latter is the decisive fac-
tor at a high mill speed. Consequently, when a ball mill runs at a
relatively low speed, the liner torque becomes large when the
ball mill has a high lifter height and large lifter number. How-
ever, when a ball mill runs at a relatively high speed, the liner
torque becomes small when a ball mill has a high lifter height
and large lifter number.
 The computational accuracy of the empirical formulas changes
when the ball mill runs at different speeds. All of the five formu-
las achieve the highest accuracy near the mill speed of w ¼ 80%.
When the mill speed changes, only the power consumption of
the mill obtained with the DEM simulation remains consistent
with the experimental results. The accuracy of empirical formu-
las is obviously poorer than that of the DEM simulation when
the mill speed changes. The major explanation is that the effect
of lifters is not considered in the empirical formulas. Conse-
quentially, DEM is an accurate numerical calculation method
that can predict the power consumption of ball mills more accu-
rately than other methods could.
 The torque and power of a ball mill usually increase with an
increase in mill speed when no lifter exists in the mill (N l ¼ 0
or HL ¼ 0 mm). Usually, the slope is extremely small. The torque
and power curves become close to the curve corresponding to
no lifters (N l ¼ 0 or HL ¼ 0 mm) when the lifters are few and
low. This condition indicates that the effect of lifters on the tor-
que and power of a ball mill gradually weakens.
 Lifter height and lifter number exert a slight effect on the power
consumption of a mill at a relatively low speed (w 6 70%). The
effect of these factors on the power consumption of the mill
increases with an increase in mill speed (w P 80%). The power
consumption of the mill shows a degree of direct proportional-
ity to lifter height and lifter number at a low speed and a degree
of inverse proportionality at a high speed.
Acknowledgements
The authors acknowledge the financial support from the Collab-
orative Innovation Center of Major Machine Manufacturing in
Liaoning and the National High-tech Research and Development
Projects (863), (project No: 2012AA062002). The authors would
also like to thank Paul W. Cleary for citing many of his papers
and getting some instructive and good ideas from his papers.
References
Bond, F.C., 1961. Crushing and grinding calculations. Brit. Chem. Eng. Part I 6 (6),
378–385, Part II 6 (8) 543–548.
Beinert, S., Fragnière, G., Schilde, C., Kwade, A., 2015. Analysis and modelling of bead
contacts in wet-operating stirred media and planetary ball mills with cfd–dem
simulations. Chem. Eng. Sci. 134, 648–662.
Bbosa, L.S., Govender, I., Mainza, A., 2016. Development of a novel methodology to
determine mill power draw. Int. J. Miner. Process. 149, 94–103.
Bingchen, Chen., 1981. Power consumption calculation of ball mills. Metal Mine 7,
38–40 (in Chinese).
Cleary, P.W., 1998. Predicting charge motion, power draw, segregation, wear and
particle breakage in ball mills using discrete element methods. Miner. Eng. 11,
1061–1080.
Cleary, P.W., 2001a. Recent advances in DEM modelling of tumbling mills. Miner.
Eng. 14, 1295–1319.
Cleary, P.W., 2001b. Modelling comminution devices using DEM. Int. J. Num. Anal.
Meth. Geomech. 25, 83–105.
Cleary, P.W., 2001c. Charge behaviour and power consumption in ball mills:
sensitivity to mill operating conditions, liner geometry and charge composition.
Int. J. Miner. Process. 63, 79–114.
Cleary, P.W., Morrisson, R., Morrell, S., 2003. Comparison of DEM and experiment
for a scale model SAG mill. Int. J. Miner. Process. 68, 129–165.
Cleary, P.W., Sinnott, M., Morrison, R., 2008. DEM prediction of particle flows in
grinding processes. Int. J. Num. Meth. Fluids 58, 319–353.
Cleary, P.W., 2009a. Industrial particle flow modelling using discrete element
method. Eng Comput 26, 698–743.
 X. Bian et al. / Minerals Engineering 105 (2017) 22–35
Cleary, P.W., 2009b. Ball motion, axial segregation and power consumption in a full
scale two chamber cement mill. Miner. Eng. 22, 809–820.
Cleary, P.W., 2015. A multiscale method for including fine particle effects in DEM
models of grinding mills. Miner. Eng. 84, 88–99.
Cleary, P.W., Morrison, R.D., 2016. Comminution mechanisms, particle shape
evolution and collision energy partitioning in tumbling mills. Miner. Eng. 86,
75–95.
Djordjevic, N., 2003. Discrete element modelling of the influence of lifters on power
draw of tumbling mills. Miner. Eng. 16, 331–336.
Djordjevic, N., 2005. Influence of charge size distribution on net-power draw of
tumbling mill based on DEM modelling. Miner. Eng. 18, 375–378.
Geng, F., Gang, L., Wang, Y., Li, Y., Yuan, Z., 2015. Numerical investigation on particle
mixing in a ball mill. Powder Technol. 292, 64–73.
Govender, I., Powell, M.S., Nurick, G.N., 2001. 3D particle tracking: a rigorous
technique for verifying DEM. Miner. Eng. 14, 1329–1340.
Govender, I., Powell, M.S., 2004. Improved experimental tracking techniques for
validating discrete element method simulations of tumbling mills. J. Exp. Mech.
44, 593–607.
Govender, I., Cleary, P.W., Mainza, A.N., 2013. Comparisons of PEPT derived charge
features in wet milling environments with a friction-adjusted DEM model.
Chem. Eng. Sci. 97, 162–175.
Hlungwani, O., Rikhotso, J., Dong, H., Moys, M.H., 2003. Further validation of DEM
modeling of milling: effects of liner profile and mill speed. Miner. Eng. 16, 993–
998.
Hong, S.H., Kim, B.K., 2002. Effects of lifter bars on the ball motion and aluminum
foil milling in tumbler ball mill. Mater. Lett. 57, 275–279.
Hou, Y.J., 2015. Research on performance of large tumbling mill liners based on
DEM-FEM coupling approach. Masters dissertation, Jilin University (May 2015).
Jonsén, P., Pålsson, B.I., Tano, K., Berggren, A., 2011. Prediction of mill structure
behaviour in a tumbling mill. Miner. Eng. 24, 236–244.
Jonsén, P., Pålsson, B.I., Häggblad, H.Å., 2012. A novel method for full-body
modelling of grinding charges in tumbling mills. Miner. Eng. 33, 2–12.
Kalala, J.T., Bwalya, M.M., Moys, M.H., 2005. Discrete element method (DEM)
modelling of evolving mill liner profiles due to wear Part I. DEM validation.
Miner. Eng. 18, 1386–1391.
Kalala, J.T., Breetzke, M., Moys, M.H., 2008. Study of the influence of liner wear on
the load behaviour of an industrial dry tumbling mill using the Discrete
Element Method (DEM). Int. J. Miner. Process. 86, 33–39.
Makokha, A.B., Moys, M.H., 2006. Towards optimising ball-milling capacity: effect of
lifter design. Miner. Eng. 19, 1439–1445.
Makokha, A.B., Moys, M.H., Bwalya, M.M., Kimera, K., 2007. A new approach to
optimising the life and performance of worn liners in ball mills: experimental
study and DEM simulation. Int. J. Miner. Process 84, 221–227.
Makokha, A.B., Moys, M.H., 2007. Effect of cone-lifters on the discharge capacity of
the mill product: case study of a dry laboratory scale air-swept ball mill. Miner.
Eng. 20, 124–131.
35
Maleki-Moghaddam, M., Yahyaei, M., Banisi, S., 2013. A method to predict shape
and trajectory of charge in industrial mills. Miner. Eng. 46, 157–166.
Mayank, K., Malahe, M., Govender, I., Mangadoddy, N., 2015. Coupled dem-cfd
model to predict the tumbling mill dynamics. Proc. Iutam 15, 139–149.
McBride, A., Govender, I., Powell, M.S., Cloete, T., 2004. Contributions to the
experimental validation of the discrete element method applied to tumbling
mills. Eng. Comput.: Int. J. Comput.-Aided Eng. 21, 119–136.
Mishra, B.K., Rajamani, R.J., 1992. The discrete element method for the simulation of
ball mills. Appl. Math. Model. 16, 598–604.
Mishra, B.K., Rajamani, R.K., 1994a. Simulation of charge motion in ball mills. Part 1:
experimental verifications. Int. J. Miner. Process. 40, 171–186.
Mishra, B.K., Rajamani, R.K., 1994b. Simulation of charge motion in ball mills. Part 2:
numerical simulations. Int. J. Miner. Process. 40, 187–197.
Morrison, R., Cleary, P.W., 2004. Using DEM to model ore breakage within a pilot
scale sag mill. Miner. Eng. 17, 1117–1124.
Morrison, R.D., Cleary, P.W., 2008. Towards a virtual comminution machine. Miner.
Eng. 21, 770–781.
Mulenga, F.K., Moys, M.H., 2014. Effects of slurry filling and mill speed on the net
power draw of a tumbling ball mill. Miner. Eng. 56, 45–56.
Powell, M.S., Nurick, G.N., 1996. A study of charge motion in rotary mills part 2—
experimental work. Miner. Eng. 9, 343–350.
Powell, M.S., McBride, A.T., 2004. A three-dimensional analysis of media motion and
grinding regions in mills. Miner. Eng. 17, 1099–1109.
Powell, M.S., McBride, A.T., 2006. What is required from DEM simulation to model
breakage in mills? Miner. Eng. 19, 1013–1021.
Powell, M.S., Smit, I., Radziszewski, P., Cleary, P., Rattray, B., Eriksson, K., Schaeffer,
L., 2006. The selection and design of mill liners. Adv. Comm. 43, 331–376.
Powell, M.S., Govender, I., Kulya, C., McBride, A.T., 2007. Applying DEM outputs to
the unified comminution model – the SAG mill. Proc. Discreet Elem. Meth. 7,
27–29.
Powell, M.S., Govender, I., McBride, A.T., 2008. Applying DEM outputs to the unified
comminution model. Miner. Eng. 21, 744–750.
Powell, M.S., Weerasekara, N.S., Cole, S., LaRoche, R.D., Favier, J., 2011. DEM
modelling of liner evolution and its influence on grinding rate in ball mills.
Miner. Eng. 24, 341–351.
Rajamani, R. K., Mishra, B. K., (1996, October). Dynamics of ball and rock charge in
sag mills. In Proc. SAG (Vol. 19).
Rezaeizadeh, M., Fooladi, M., Powell, M.S., Mansouri, S.H., Weerasekara, N.S., 2010. A
new predictive model of lifter bar wear in mills. Miner. Eng. 23, 1174–1181.
Toor, P., Franke, J., Powell, M., Bird, M., Waters, T., 2013. Designing liners for
performance not life. Miner. Eng. 43, 22–28.
Wang, M.H., Yang, R.Y., Yu, A.B., 2012. DEM investigation of energy distribution and
particle breakage in tumbling ball mills. Powder Technol. 223, 83–91.
Weerasekara, N.S., Liu, L.X., Powell, M.S., 2016. Estimating energy in grinding using
dem modelling. Miner. Eng. 85, 23–33.