Minerals Engineering 24 (2011) 15861594

Contents lists available at SciVerse ScienceDirect

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

Prediction of the disc wear in a model IsaMill and its effect on the ow
of grinding media
C.T. Jayasundara a, R.Y. Yang a, A.B. Yu a,, D. Curry b
a
b

Laboratory for Simulation and Modelling of Particulate Systems, School of Materials Science and Engineering, The University of New South Wales Sydney, NSW 2052, Australia
Xstrata Technology, Brisbane, Queensland 4000, Australia

a r t i c l e

i n f o

Article history:
Received 24 October 2010
Accepted 22 August 2011
Available online 19 September 2011
Keywords:
Wear
Stirred mills
Discrete element method
Granular ow
Mineral processing

a b s t r a c t
Mill wear is a critical issue in mineral industries. It affects mill performance and the cost of replacing
worn parts is high. Understanding wear and its effect would provide a useful insight for process optimisation. This paper combines the discrete element method (DEM) with a commonly used wear model to
predict the wear pattern of stirring discs in a model IsaMill. The results show that wear is more severe at
the outer face of discs and the lifting side of holes. The simulated wear pattern has been compared with
those observed in practice. The effect of disc wear on the ow of grinding media is also examined, showing that with the increasing wear, impact energy increases while power draw shows a mix of slight
increase and decrease. The ndings would be useful to the improvement in the design and control of
IsaMills.
2011 Elsevier Ltd. All rights reserved.

1. Introduction
IsaMill is a horizontal high speed stirred mill that operates with
very high power intensity. It has been increasingly used in the
mineral industry for ne and coarse grinding, because of its high
energy efciency compared with other conventional grinding techniques (Gao et al., 2001). Inside the shell, there are rotating grinding discs mounted on a shaft, which is coupled to a motor and
gearbox. The grinding discs agitate the media and ore particles in
slurry that is continuously fed into the feed port. A special product
separator is installed to allow the product to exit and generate high
centrifugal forces to retain the media inside the grinding chamber.
Despite becoming a popular ne grinding technique, IsaMill is
still a new technology, and its optimum control and scale-up rely
on experimental methods and industry experience, rather than
the detailed scientic principles. Gao and Forssberg (1993) and
Gao et al. (1996) conducted extensive laboratory tests to investigate the performance of IsaMills. Using plant data, mill scale-up
studies have also been done under certain operating conditions
based on the specic input energy (Gao et al., 2001). Investigations
on horizontal stirred mills which are similar to IsaMills have been
reported by Blecher et al. (1996) and Kwade (1999). They showed
that grinding performance of stirred mills can be represented by
the stress intensity and stress numbers, which are derived based
on empirical formulae. Their studies were focused on the ow phe-

Corresponding author. Tel.: + 61 2 9385 4429; fax: + 61 2 9385 5969.

E-mail address: [email protected] (A.B. Yu).
0892-6875/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mineng.2011.08.011

nomena of the slurry without considering the ow of grinding

media.
While these studies have been useful, particularly at the early
stage of development, they are largely based on the macroscopic
observations and the ndings are sometimes difcult to generalize.
To understand the underlying mechanisms of grinding, more detailed information at a microscopic scale is required. Numerical
simulation provides a useful research tool to achieve this goal. In
particular, numerical models based on the discrete element method (DEM) can produce information about the forces acting on and
the trajectory of individual particles (Cundall and Stack, 1979).
DEM was rst introduced to ball mill simulation by Mishra and
Rajamani (1994) and followed by others for different types of
grinding mills as a whole (Inoue and Okaya, 1996; Cleary, 1998;
Kano and Saito, 1998; Hoyer, 1999). It has been further developed
and widely used to study various particulate systems as reviewed
by Zhu et al. (2007 & 2008). DEM has not been applied to IsaMills
until recent. Yang et al. (2006) showed that DEM can be applied to
IsaMill process under dry conditions and later it was extended to
multiple discs for studying the microdynamic properties related
to the grinding performance (Jayasundara et al., 2006a). It was
shown that grinding process in stirred mill follows the rst-order
kinetics and the rate constant can be correlated to the impact energy (Jayasundara et al., 2010). While these studies lay foundation
to numerical modelling of IsaMill process, there are other issues
associated with the grinding process which have not been
addressed.
Wear is another important issue in grinding process. Discs in
IsaMills experience severe wear due to continuous interactions

C.T. Jayasundara et al. / Minerals Engineering 24 (2011) 15861594

with grinding media and slurry, causing changed stirrer geometry

and mill performance. The costs of replacing worn components and
production lost during the mill downtime are signicant (Rule
et al., 2008). Therefore, better understanding of wear phenomena
can improve mill performance by extending disc life span.
Different types of wear models and equations are reported in
literature to study the wear phenomena in industrial processes.
Most of them are derived based on empirical equations. Based on
contact-mechanics Archard (1953) showed that worn volume can
be represented by the sliding distance, applied load and the material hardness. This model can be applied to systems where there is
a constant contact between two surfaces. It does not account for
the wear due to striking nature of particles on the contact surface,
which is generally described as erosion. Wear due to erosion had
been extensively studied by Finnie (1960) and Bitter (1963). They
showed that the erosion can be described by the impact velocity
and angle of attack. Although there are different wear models
available, it was reported that no single predictive equation or
group of limited equations could be found for general and practical
use (Meng and Ludema, 1995).
On the other hand, a number of experimental methods have
been developed to examine the abrasive wear in grinding mills
(Spero et al., 1991). Among these tests, Yancey, Geer and Price
(YGP) abrasion test appears to be the most widely used and accepted test for mineral ores. Results obtained from these tests
are sometimes arguable because, it was reported that the results
obtained from the YGP abrasion test were not in good agreement
with those obtained in the CE-hammer mill abrasion test (Sligar,
1986). Extensive experiments have been reported by Powell
(1991) to examine the inuence of mill-liner wear on grinding performance of ball mills. A liner-wear monitor was developed that
could be used to take many readings of the wear prole of the
worn surfaces. Experiments were carried out over a year to examine the inuence of mill-liner wear on grinding performance of
grinding mills. Wear and performance characteristics of lifter bars
were evaluated in relation to their height, together with the costlife performance of the backing materials used for liners in rotary
mills. Although experiments and plant trials were used to select
the liner material and design, they do suffer from a number of
drawbacks such as downtime of monitoring the liners, taking a
long time, usually over a year to yield results, and often the results
can be inconclusive as plant conditions vary with time (Powell
et al., 2006).
Therefore, alternative techniques such as numerical modelling
are becoming more popular in nowadays to investigate the wear
phenomena in grinding mills. Using a simplied DEM model Radziszewski and Tarasiewicz (1993) showed that dissipated energy
can be used to predict the wear prole of ball mill lifters. Although
there are many wear mechanisms occurring in a grinding mill,
these investigators focused on adhesive and abrasion wear mechanisms. After a back-calculation from experimental data under dry
conditions, they obtained wear model parameters for abrasive
grinding wear. Kalala et al. (2005) used DEM to simulate the wear
prole of lifters in ball mills and compared with plant data. In their
model, both impact and abrasion damages were considered. The
wear surface was divided into small elements and the data was
accumulated in each element and the worn surface prole was
determined. They found that contribution of impact energy to
the wear of the lifter is more important than the contribution of
abrasive energy. Since this model is computationally intensive, at
present large scale 3D simulations are prohibitively costly (Kalala
et al., 2005). Based on DEM, Cleary et al. (2010) developed a 3D
wear model for the so called Hicom mill. In their wear model, both
impact and abrasion damages were considered. Impact damage
was predicted by two measures: (i) energy dissipated in the normal direction during collisions between the particles and the liner;

1587

(ii) excess kinetic energy of impact for collisions above a specied

threshold damage velocity. The abrasion damage was predicted by
another two measures: (i) shear work done by the contact model
for collisions between particles and the liner; (ii) kinetic energy
of each collision weighted by the collision angle which can be described by the Finnie wear model. It was observed that the abrasion wear is more representative than impact wear on the Hicom
mill.
These previous studies showed that depending on the mill type
and the grinding media ow pattern, dominant wear mechanism
could be different. Since numerical modelling of wear is computationally intensive, particularly for large systems such as IsaMills,
depending on the mill type and the scale of the system one has
to employ a feasible wear model in a simulation study. Up until
now no experimental or numerical studies have been reported
on the disc wearing in IsaMills and the consequent effect of wear
on the mill performance. The scope of this work is twofold: (i) to
develop a wear model based on DEM which can be used to explore
the disc hole wear of IsaMills; (ii) to investigate the effect of disc
wear on the ow of grinding media. The simulated results are compared with real wear prole to verify the proposed model.
2. Model development
2.1. DEM model
The DEM model employed in this work has been detailed elsewhere (Jayasundara et al., 2006a; Yang et al., 2006) and will be
briey described here. In a DEM simulation, each particle possess
translational motion and rotational motion, which can be described by Newtons second law of motion, given by

mi

dv i X n

Fij Fsij mi g
dt

and

Ii

dxi X

Ri  Fsij  lr Ri jFnij jxi

dt

where vi, xi and Ii are, respectively, the translational velocity, angular velocity and moment of inertia of particle i, Ri is a vector running
from the centre of the particle to the contact point with its magnitude equal to particle radius Ri. Fnij and Fsij are respectively, the normal and the tangential contact forces imposed on particle i by
particle j. For estimating the contact forces, the Hertz-Mindlin contact model was employed. The contact forces of normal and tangential directions, Fnij and Fsij , were calculated according to the following
equations:

Fnij


pp
3
2 p
 n2n  c E R
 nn v ij  n
^
E R

nij
ij
n
3

and

h

3=2 i
Fsij sgnns ljF nij j 1  1  minns ; ns;max =ns;max

e 2 , and Y and r
e are, respectively, Youngs modwhere E Y=1  r
ulus and Poisson ratio; nn is the overlap between particles i and j; nij
is a unit vector running from the centre of particle j to the centre of
 R=2 for mono-sized particle. The normal damping
particle i; R
constant, n, is the material property directly linked to the coefcient
of restitution e. ns and ns,max are, respectively, the total and maximum tangential displacements of particles during contact. The
present study used 3 mm glass beads with material properties of
density 2.5103 kg m3, Youngs modulus 1.0  107 N m2, Poisson
ratio 0.29, sliding friction coefcient 0.2 and restitution coefcient
0.68. Simulated material properties such as particle density and
Poisson ratio are the same as glass beads. The sliding friction

1588

C.T. Jayasundara et al. / Minerals Engineering 24 (2011) 15861594

coefcient of glass beads was measured by using pin-on-disc Tribometer (ASTM G99); this value is used for the particle-drum contact. Restitution coefcient is obtained by dropping a sample of
3 mm glass beads under gravity and measuring the rising height.
Youngs modulus used in the simulation is smaller than the real
glass beads (100 GPa) to reduce the simulation time. The current
value ensures that the maximum overlap is less than 3% of particle
diameter and has been demonstrated that it does not affect the nal
results much (Zhou et al., 1999; Yang et al., 2003b).
In industrial IsaMills, the main function of the product separator
is to retain the media in the mill. In addition, it will act as an impeller and pump the slurry back to the grinding chambers. The effect
of slurry is not considered in this work. Therefore, in this model, we
do not consider the effect of the product separator. The mill used in
this work consists of a xed chamber, a rotating shaft and three
stirrers which are rigidly attached to the shaft, as shown in
Fig. 1. A simulation starts from a packing process in which the shaft
and the discs are at rest and particles are fed into the mill to form a
stable packed bed. Then the shaft and stirrers start to rotate at a given speed to agitate the particles. For all simulations, the mill is
lled up to 80% by volume (41,000 particles) and mill speed is
set to 1000 rpm. All the results are analysed when the system
reaches the steady state. The steady state is determined by monitoring the torque on the mill shaft. Fig. 1 shows the sectional front
and end views of the mill.
2.2. Wear model
Wear of a solid surface due to particle erosion is widely encountered in industrial processes, such as erosion of turbine blades,
control of sand blasting, damage of helicopter propellers, and wear
of pneumatic pipelines (Fan et al., 1991). In these applications the
target surface is attacked by solid particles entrained in a uid
stream. In general, the extent of surface erosion by impingement
of abrasive particles depends on factors such as particle impinging
velocity, impact angle, properties of impacting particles and properties of target material. These concepts have been the foundation
for most of the wear models including Finnies wear model (Finnie,
1960) which has been widely used in many industrial processes. By
far, the majority of erosion models have been developed based on
Finnies model or its modied versions (Bhasker, 2010; Lester et al.,
2010).
In stirred mills, particles near the disc are subjected to intensive
collisions with the disc. These collisions occur at an angle with the
disc or they may slide on the disc which leads to abrasive wear.
Regardless of the nature of collision event, the rate of wear in

grinding mills is dependent in some way on the rate of energy

absorption of the boundary surfaces during collision with particles
(Cleary, 2001). The exact form of this dependence is currently
unknown Cleary et al. (2010) used two energy based methods
to predict wear arising from each of the impact and abrasion
mechanisms and found that wear due to abrasion is more representative than wear due to impact. They showed that the abrasive
wear can be predicted by two measures: (i) shear work done by the
contact force model; and (ii) kinetic energy weighted by the collision angle, which is based on the Finnie wear model. It was reported that these two measures equally give comparable results
with the real wear data obtained for the Hicom mill.
The Finnie model was developed considering a single rigid abrasive particle which strikes on the target surface in such a way as to
displace or cut away part of the surface. Removal of material is
somewhat similar to the tooth of a milling cutter or the grains
on a grinding wheel, as schematically shown in Fig. 2. The volume
of material Q, removed by a single abrasive particle of mass m and
velocity v is given by (Finnie, 1960):

8
< mv 2 sin 2a  3 sin2 a

a 6 18:5

a P 18:5

8p

cos2 a

Y`

X`
10
110

18

9
90
25

(a)

where p is the yield stress of the target material. In this work it is

assumed that the disc is made of high strength alloy steel (ASTM
A514) which has the yield stress of 690 MPa.
To examine the dominant wear mechanism, the spatial distribution of shear energy and kinetic energy on the disc surface were
accumulated for four revolutions of the disc (Fig. 3). Shear energy is
dened as the shear work done by the contact model for collisions
between particles and the disc (Cleary et al., 2010). The results
show that abrasion would likely to be the dominant wear mechanism for the outer surface and impact is most likely to be the main
mechanism for the holes. Since the focus of this work is to examine
the wear mechanism in the hole region, wear in the hole region can
be best captured by the Finnie wear model which was developed
based on kinetic energy.
From DEM simulations, information such as particle-disc impact velocity, angle and location can be obtained. The disc is divided into triangular surface mesh on which the worn volume
data are collected. When a particle is in contact with the disc, angle
a and the relative velocity v, is obtained from the simulation and
the corresponding worn volume is calculated according to Eq.
(5). When the system is at steady state, the simulation is carried
out for 1 s and the accumulated worn volume is recorded in the
corresponding triangular element. Note that the Finnie model only

39
Y

mv 2
24p

(b)

Fig. 1. Geometry of the model stirred mill: (a) sectional front elevation; and (b) sectional end elevation. All dimensions in mm.

1589

C.T. Jayasundara et al. / Minerals Engineering 24 (2011) 15861594

Fig. 2. Abrasive particle of velocity

v striking a surface and removing material.

captures the wear phenomena caused by a single particle attack

where the angle a and the relative velocity v are calculated just before the contact. Thus, the model cannot be applied to a situation
where a particle is hit by other particles mid-way during the scouring impact. In fact, if a particle is hit by other particles when it is in
contact with the surface, wear is caused by the abrasion. As we do
not consider the abrasion in this work, the effect of other particles
is not considered in this analysis.
In practice, it takes months before discs wear out and need to be
replaced. It is computationally expensive and takes months for
DEM models to simulate the whole wear process. Instead, the present study adopts the following procedure to model the wear process. Similar technique has been used by others to investigate
wear on ball mills (Radziszewski and Tarasiewicz, 1993; Kalala
et al., 2005).
The simulation is rstly performed on a new disc for 1 s and the
amount of material worn out can be recorded over the whole disc.
Fig. 4a shows the spatial distribution of wear rate on a new disc.
Signicant wear can be seen near the disc holes and outer face of
the disc (disc rotation anti-clockwise). For the outer edge, the
velocity of particles is relatively low, compared to that of the disc.

As a result, the particle-disc relative velocity increases, leading to

an increased wear rate. On the other hand, the disc holes act as lifters and the particles in this region collide frequently with the hole
surface, again leading to an increased wear rate. Change of outer
face prole has not been considered, because the focus of this work
is to investigate the effect on the hole shape on grinding performance. Fig. 4b shows the corresponding wear rate in the hole region where b is the clockwise angle from the line connecting the
disc and hole centres (Fig. 4b inset). From the simulations, it is observed that the hole has the maximum wear at b  40, which is
the lifting side of the hole, and the minimum wear at b = 225.
By plotting the wear rate from b = 225 (as shown in Fig. 4b), it is
shown that the wear pattern is almost symmetric around b = 40.
The wear rate is considered as a constant over an arbitrary time
period, e.g. one week, so that the total worn volume near the holes
within that time can be calculated. Depending on the mesh resolution and the time period, the wear prole has a tendency to be very
spiky. The spikes can be removed by appropriate smoothing technique or by selecting a proper smooth prole (Cleary, 1998; Kalala
et al., 2005). The extent of the smoothing is compromised by the
required accuracy and the simulation time. For instance, if the
worn area of interest is small, by using a very ne mesh, accurate
results can be obtained within a reasonable time. The same approach, however, could not be used when the worn surface is considerably large. Therefore, in order to achieve the results within a
reasonable time, the following technique has been adopted in the
hole region.
The worn volume in the hole region, which has to be removed in
order to obtain the worn prole, is calculated by considering the
total worn volume, accumulated on the mesh that covers the inner
surface of the hole. Note that the resolution of the mesh size is
approximately 1 mm (33% of the particle size). From the spatial

Fig. 3. Energy dissipation on the disc surface after four revolutions (anti-clockwise): (a) shear energy; and (b) kinetic energy.

(b) 7

(a)

6
5

Wear rate (mm /month)

Aw

wear rate
1 unit = 1 mm3/month

4
3
2
1
0
-135

-90

-45

45

90

(deg)

Fig. 4. Initial wear rate: (a) Spatial distribution of wear; and (b) wear rate of the hole surface.

135

180

225

1590

C.T. Jayasundara et al. / Minerals Engineering 24 (2011) 15861594

Fig. 5. Schematic representation of the worn area within a hole as determined by the proposed numerical scheme.

(b)

(a)

wear rate
1 unit = 1 mm3/month

(c)

(d)

Fig. 6. Evolution of disc hole prole with time: (a) new disc; (b) moderately worn disc after 4 months; (c) worn disc after 8 months; and (d) worn discs of industrial IsaMill
after 3000 h of operation (courtesy of Xstrata Technology, Australia).

wear distribution (Fig. 4a) it is shown that the worn prole should
represent the crescent shape worn area Aw (Fig. 4a). To represent
this area, a circle is used whose centre is determined by moving
it along the line which shows the angle of maximum wear rate
(Fig. 5). Intersection of this circle with the hole, closely matches
the high wear area (Aw). Now the shape of the hole is modied
by removing the part of the disc, overlapping the high wear circle.
The simulation is then performed again for another second, based
on the new design and the new wear pattern is obtained for the
next week. Thus, the third circular wear increment can be located
as shown in Fig. 5. Note that the angle of the line which corresponds to the maximum wear, increases with the subsequent hole.
Such process can be repeated to simulate the wear process over a
very long time.
The highest wear rates can be seen at the disc outer face and
near the holes regions. Disc outer face wear causes reduction in
disc radius. As a result, the gap between disc outer face and the

mill drum will be reduced. However, this may not affect the particle ow signicantly because the majority of the energy transfer
from disc to particles occurs through holes (Jayasundara et al.,
2006b). Therefore, the present study only considers the change of
hole prole and the effect of disc wear on the outer face has not
been considered.

3. Results and discussion

3.1. Wear pattern
Fig. 6 shows the wear pattern on the initial disc geometry.
Fig. 6b and c show the predicted geometry of the progressively
worn holes after 4 and 8 months, respectively. It is evident that
as wear increases, the hole geometry turns into an elongated shape
which moves towards the disc outer face as observed in practice

1591

C.T. Jayasundara et al. / Minerals Engineering 24 (2011) 15861594

outwards, reducing the gap between hole edge and the outer face, as
observed in Fig. 6. This will eventually lead to structural failure of the
disc, which is not desirable for effective grinding.

initial
2 months
4 months
6 months
8 months

10
8

Wear rate (mm /month)

12

3.2. Effect of disc wear on particle ow

6
4
2
0
-135

-90

-45

45

90

135

180

225

(deg)
Fig. 7. Wear rate on the hole surface as a function of b.

(Fig. 6d). The wear of the disc may lead to a structural failure of the
disc. Since the wear rate and pattern can be estimated, the life span
of the disc can be predicted.
To quantitatively describe the disc wear behaviour, Fig. 7 shows
the wear proles along the angular direction of a disc hole at different times. With time progressing, the wear rate of the disc increases,
while the angle at which the wear rate has the maximum value increases. As the holes wear out and become bigger, more particles
are captured, so there are stronger interactions between particles
and discs, resulting in an increased wear rate. While the wear pattern is still symmetric to a large degree, the peak shifts to the right.
The wear distribution becomes less symmetric with a higher wear
rate occurring in the region 40 < b < 225 and a lower rate occurring
in the region 225 < b < 360. This causes the hole boundary to move

(a)

As the disc gradually wears out, it is expected that it will cause

different ow patterns and energy consumptions, hence different
milling performances. In this section, we investigate the effect of
disc wear on the ow properties such as ow velocity, porosity,
collision energy Ce, collision frequency Cf, impact energy intensity
Ei and power draw Pin. These properties have been found to be useful to describe the ow of grinding media in relation to grinding
performance (Jayasundara et al., 2006a, 2010). Collision energy is
dened as 1=2mv 2ij , where m is the mass of particles and
v ij j~
vi  ~
v j j the magnitude of the relative velocity of two colliding particles. The magnitude of the velocity was determined when
the particle contact is rst made. Collision frequency is the number
of collisions per particle per second. By denition, a collision must
come from two particles with a certain distance. Otherwise, they
are simply in enduring contact. If the collision gap is not dened
or set to very close to zero, there may be a large number of collisions with very low collision energies. In order to lter out small
collisions, it is necessary to dene the collision gap in the DEM simulation. Here consistent with our previous studies (Jayasundara
et al., 2006a, 2010), a collision is counted only when two particles
come to contact from a gap of at least 1% of particle diameter and
an existing contact between particles is not counted as a collision.
Impact energy intensity is dened as Ei = NCfCe, where N is the total
number of particles.
Fig. 8 shows the effect of wear after 8 months on the ow properties in terms of particle ow pattern, velocity eld and local
porosity (disc rotating clockwise). Different ow patterns are

(b)

velocity (m/s)

porosity

Fig. 8. Snapshot of particle ow pattern (top), and porosity and velocity distribution (bottom) on the radial plane YY0 for: (a) new disc; and (b) worn disc after 8 months.

1592

C.T. Jayasundara et al. / Minerals Engineering 24 (2011) 15861594

Fig. 9. Spatial distribution of the local impact energy on the radial plane YY0 (top), and the axial plane XX0 (bottom) for: (a) new disc; and (b) worn disc after 8 months.

(b) 1.2

0.003
new
after 8 months

0.0025
0.002
0.0015
0.001
0.0005
0

200

400

600

800

C (Hz)
f

Probability distribution

Probability distribution

(a)

new

after 8 months

0.8
0.6
0.4
0.2
0

10

15

20

25

-6

C (x10 J)
e

Fig. 10. For new and worn discs: (a) collision frequency distribution; and (b) collision energy distribution.

observed as a result of different hole geometries. With the new disc,

particles rotate from the lower region (about 4 oclock position) to
the upper region (around 11 oclock). Those particles in the upper
region with low centrifugal forces form the cascading ow regime
around 2 oclock position. When the discs are worn, energy transfer
from discs to particles is higher due to large holes. Particles are
more densely packed and form a solid centrifugally driven layer
around the entire perimeter of the mill, creating a large void in
the centre. Porosity can be used to quantify the ow structure of
particles (Yang et al., 2003a; Zhou et al., 2003). The so-called local
porosity is used in this work, which is achieved by dividing the calculation domain into a series of spherical cells of two particle diameters in size and the porosity is calculated for each cell. The results
at different times are collected and averaged to obtain the timeaveraged value. The spatial distribution of porosity also indicates
that when increasing the hole size, high porosity regions are developed in the mill centre. This is because, with large holes, as more
particles are captured by the rotating holes, energy transfer from

disc to particles is high. As a result, particles gain more kinetic energy and centrifuge towards the mill drum.
Simulation and experimental studies have shown that the impact energy among particles is related to the grinding rate (Kano
and Saito, 1998; Kano et al., 2001; Mori et al., 2004). In fact, our recent study showed that, dry grinding in a stirred mill studied in
this paper is a rst-order kinetic process with particle sizes, decaying exponentially with time. The grinding rate is correlated with
the total impact energy (Jayasundara et al., 2010). The above studies indicate the importance of impact energy in prediction of grinding rate. Fig. 9 shows the spatial distributions of impact energy in
the radial and axial planes. In the radial plane, both systems show a
ring of high impact energy region near the holes. The worn disc
shows a wider energy ring near the holes, compared to that of
the new disc due to more particles interact with the disc. Therefore, higher grinding rates can be expected with the worn disc. In
the axial direction, both systems show high impact energies close
to discs near the holes.

C.T. Jayasundara et al. / Minerals Engineering 24 (2011) 15861594

80

Acknowledgments

P and E (W)

Finnie model is a promising technique to predict the wear where

impact damage is dominant. Further studies are necessary to study
wear due to abrasion.

100

1593

60

The authors would like to thank the Australian Research Council

and Xstrata Technology for providing the nancial support for this
work. The permission granted by Xstrata Technology to publish
this paper is gratefully acknowledged.

40

20

Time (month)
Fig. 11. The effect of disc wear on the impact energy Ei and power draw P.

Fig. 10 shows the distributions of collision frequency and collision energy for the new and worn discs. Both collision frequency
distributions show two peaks at low and high frequencies
(Fig. 10a). Our previous study indicated that collision frequency
of particles between discs is higher than those near discs (Jayasundara et al., 2006a), so the rst peak corresponds to the collisions
close to the discs and the second peak corresponds to the collisions
between the discs. With disc wearing, the distribution is slightly
shifted to the right, suggesting a higher Cf than that of the new
one. As explained before, when the discs wear out, particles obtain
a higher kinetic energy. As a result, they move towards the mill
drum and packed closely. When particles are closely packed, the
chances to collide with other particles increase which leads to increase in Cf. Fig. 10b shows the distribution of the collision energy
of the worn disc slightly shifts to the right, suggesting an increase
in average collision energy. This is because, with larger holes, more
particles are captured by the rotating discs, and hence obtain higher kinetic energy, which in turn results in an increased Ce.
The results showed that when the holes wear out, more energy
is transferred to the particles. Therefore, the power draw is increased as shown in Fig. 11. The impact energy intensity Ei also increases with time at a faster rate than power draw. As Ei is related
to the grinding performance, the results suggest that the grinding
performance of mills may increase when discs wear out. However,
mill operation with worn discs is not preferable because it may
cause other adverse effects such as structural failure of the disc.
4. Conclusions
Wear is important to grinding processes and the accurate prediction of the wear pattern can help decision making in mill operation. By incorporating a wear model into our previous developed
DEM model, we have studied the wear pattern of the disc holes in a
model IsaMill. Although this approach is not perfect, it can be used
in large scale IsaMills to examine the wear in the disc hole region.
The simulation results show that, the wear of the disc holes is
caused by impact damage rather than abrasion. Most of the impact
wear takes place on the lifting side of the disc holes. When discs
gradually wear out, more particles are captured by the disc holes
and obtain high velocities. As a result, the collision energy, collision
frequency and impact energy increase, whereas the power draw
shows a mix of slight increase and decrease. The result suggest
that, grinding performance of mills may be improved when discs
wear out, but attention should also be given to possible adverse effects such as structural failure of the discs.
The present study further conrms that numerical modelling is
a cost-effective technique to study the wear of a solids handling
system. In particular, it demonstrated that DEM coupled with the

References
Archard, J.F., 1953. Contact and rubbing of at surfaces. Journal of Applied Physics
24, 981988.
Bhasker, C., 2010. Flow simulation in industrial cyclone separator. Advances in
Engineering Software 41, 220228.
Bitter, J.G.A., 1963. A study of erosion phenomena. Wear 6, 521.
Blecher, L., Kwade, A., Schwedes, J., 1996. Motion and stress intensity of grinding
beads in a stirred media mill 1. Energy density distribution and motion of single
grinding beads. Powder Technology 86 (1), 5968.
Cleary, P.W., 1998. Predicting charge motion, power draw, segregation and wear in
ball mills using discrete element methods. Minerals Engineering 11, 1061
1080.
Cleary, P.W., 2001. Recent advance in DEM modelling of tumbling mills. Minerals
Engineering 14, 12951319.
Cleary, P.W., Owen, P., Hoyer, D.I., Marshall, S., 2010. Prediction of mill liner shape
evolution and changing operational performance during the liner life cycle: case
study of Hicom mill. International Journal for Numerical Methods in
Engineering 81, 11571179.
Cundall, P.A., Stack, O.D.L., 1979. A discrete numerical model for granular
assemblies. Geotechnique 29, 47.
Fan, J., Zhou, D., Jin, J., Cen, K., 1991. Numerical simulation of tube erosion by
particle impaction. Wear 142, 171184.
Finnie, I., 1960. Erosion of surfaces by solid particles. Wear 3, 87103.
Gao, M., Forssberg, E., 1993. A study of the effect of parameters in stirred ball
milling. International Journal of Mineral Processing 37, 4549.
Gao, M., Forssberg, K.S.E., Weller, K.R., 1996. Power prediction for a pilot scale
stirred ball mill. International Journal of Mineral Processing 4445, 641652.
Gao, M., Young, M.F., Cronin, B., Harbort, G., 2001. IsaMill medium competency and
its effect on milling performance. Minerals and Metallurgical Processing 18 (2),
117121.
Hoyer, D.I., 1999. The discrete element method for ne grinding scale-up in Hicom
mills. Powder Technology 105 (13), 250256.
Inoue, T., Okaya, K., 1996. Grinding mechanism of centrifugal mills a simulation
study based on the discrete element method. International Journal of Mineral
Processing 4445, 425435.
Jayasundara, C.T., Yang, R.Y., Yu, A.B., Curry, D., 2006a. Discrete particle simulation
of particle ow in IsaMill. Industrial and Engineering Chemistry Research 19,
63496359.
Jayasundara, C.T., Yang, R.Y., Yu, A.B., Curry, D., 2006b. Discrete particle simulation
of particle ow in IsaMill Effect of mill properties, In: Comminution 06.
Minerals Engineering, Perth, Australia.
Jayasundara, C.T., Yang, R.Y., Yu, A.B., Rubenstein, J., 2010. Effects of disc rotation
speed and media loading on particle ow and grinding performance in a
horizontal stirred mill. International Journal of Mineral Processing 96, 2735.
Kalala, J.T., Bwalya, M., Moys, M.H., 2005. Discrete element method (DEM)
modelling of evolving mill liner proles due to wear Part I. DEM validation.
Minerals Engineering 18, 13861391.
Kano, J., Saito, F., 1998. Correlation of powder characteristics of talc during
planetary ball milling with the impact energy of the balls simulated by the
particle element method. Powder Technology 98, 166170.
Kano, J., Mio, H., Saito, F., Miyazaki, M., 2001. Correlation of grinding rate of gibbsite
with impact energy in tumbling mill with mono-size balls. Minerals
Engineering 14, 12131223.
Kwade, A., 1999. Wet comminution in stirred media mills research and its
practical application. Powder Technology 105 (13), 1420.
Lester, D.R., Graham, L.A., Wu, J., 2010. High precision suspension erosion
modelling. Wear 269, 449457.
Meng, H.C., Ludema, K.C., 1995. Wear models and predictive equations: their form
and content. Wear 181, 443457.
Mishra, B.K., Rajamani, R.K., 1994. Simulation of charge motion in ball mills. Part 1:
experimental verications. International Journal of Mineral Processing 40, 171
186.
Mori, H., Mio, H., Kano, J., Saito, F., 2004. Ball mill simulation in wet grinding using a
tumbling mill and its correlation to grinding rate. Powder Technology 143, 230
239.
Powell, M.S., 1991. The design of rotary-mill liners, and their backing materials.
Journal of the South African Institute of Mining and Metallurgy 91, 6375.
Powell, M.S., Smith, I., Radziszewski, P., Cleary, P., Rattray, B., Eriksson, K., Schaeffer,
L., 2006. The selection and design of mill liners, advances in comminution. In:

1594

C.T. Jayasundara et al. / Minerals Engineering 24 (2011) 15861594

Kawatra (Ed.), Society of Mining Metallurgy and Exploration. Colarado, USA, pp.
331376.
Radziszewski, P., Tarasiewicz, S., 1993. Simulation of ball charge and liner wear.
Wear 169, 7785.
Rule, C.M., Knopjes, L., Clermont, B., Philippe, C., 2008. Fine grinding
Developments in ceramic media technology and resulting improved plant
performance at Anglo Platinum. In: Third International Platinum Conference,
Sun City, South Africa.
Sligar, J., 1986. The abrasiveness of coal. In: NERDDP Report EG 87/652, Canbera.
Spero, C., Hargreaves, D.J., Kirkcakdie, R.K., Flitt, H.J., 1991. Review of test methods
for abrasive wear in ore grinding. Wear 146, 389408.
Yang, R.Y., Zou, R.P., Yu, A.B., 2003a. Microdynamic analysis of particle ow in a
horizontal rotating drum. Powder Technology 130, 138146.
Yang, R.Y., Zou, R.P., Yu, A.B., 2003b. Numerical study of the packing of wet coarse
uniform spheres. AIChE Journal 49 (7), 16561666.

Yang, R.Y., Jayasundara, C.T., Yu, A.B., Curry, D., 2006. DEM simulation of the ow of
grinding medium in IsaMill. Minerals Engineering 19, 984994.
Zhou, Y.C., Wright, B.D., Yang, R.Y., Xu, B.H., Yu, A.B., 1999. Rolling friction in the
dynamic simulation of sandpile formation. Physical Review A 269, 536.
Zhou, Y.C., Yu, A.B., Stewart, R.L., Bridgwater, J., 2003. Microdynamic analysis of the
particle ow in a cylindrical bladed mixer. Chemical Engineering Science 59,
13431364.
Zhu, H.P., Zhou, Z.Y., Yang, R.Y., Yu, A.B., 2007. Discrete particle simulation of
particulate systems: theoretical developments. Chemical Engineering Science
62, 33783392.
Zhu, H.P., Zhou, Z.Y., Yang, R.Y., Yu, A.B., 2008. Discrete particle simulation of
particulate systems: a review of major applications and ndings. Chemical
Engineering Science 63, 57285770.