Minerals Engineering 19 (2006) 13571361

This article is also available online at:

www.elsevier.com/locate/mineng

Eects of grinding media shapes on load behaviour

and mill power in a dry ball mill
N.S. Lameck *, K.K. Kiangi, M.H. Moys
School of Chemical and Metallurgical Engineering, University of the Witwatersrand, Private Bag 3, Johannesburg, WITS 2050, South Africa
Received 31 August 2005; accepted 26 January 2006
Available online 20 March 2006

Abstract
Little attention has been given to the study of grinding media shapes and how they aect load behaviour and power draw in grinding
mills. To a large extent the focus has been on how grinding media shapes aect milling kinetics in terms of specic rate of breakage and
breakage distribution parameterisation.
This study investigated the eects of three media shapes (cylpebs, spherical and worn balls) on load behaviour and mill power draw at
various mill speeds and load lling. An inductive proximity probe was used to determine the load orientation of the grinding media
charge while a load beam enabled measurement of power draw.
The variations in toe and shoulder positions among media shapes were observed. This was also reected in the power draw. The
power increases to a maximum with increasing mill speed for all media shapes. The maximum power draw was reached at dierent mill
speeds for the three studied media shapes.
 2006 Elsevier Ltd. All rights reserved.
Keywords: Grinding; Comminution; Power; Grinding media

1. Introduction
Studying the comminution processes and understanding
dierent parameters that aect it has increasingly became
of interest to many researchers in the eld of mineral processing. Milling kinetics (Austin et al., 1984; Herbst and
Lo, 1989), load behaviour (Liddell and Moys, 1988; Powell
and Nurick, 1996a,b,c; van Nierop and Moys, 1997; Dong
and Moys, 2003) and mill power (Yildirim et al., 1998;
Morrell, 1993) have been studied as functions of media
size, feed particle size distribution, fraction of mill lled
with balls and powder, mill diameter, mill speed and other
variables aecting them.
Load behaviour as an additional mill control parameter
is an important aspect because of the relation between the
variation of mill power with operating parameters (Austin

Corresponding author. Tel.: +27 11 717 7558; fax: +27 11 403 1471.
E-mail address: nlameck@prme.wits.ac.za (N.S. Lameck).

0892-6875/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mineng.2006.01.005

et al., 1984) and its impacts on mill control. This necessitates the need to study various parameters that aect the
load behaviour due to the benets that would come about
not only for control purposes but also for economical operation of the mill. More importantly the laws of motion that
govern the load behaviour do not change with size thus
scaling up from pilot to industrial scale mills does not aect
the load motion.
Grinding media directly aect the load behaviour and
consequently the operations of industrial mills in terms of
product size, energy consumption and grinding costs. Size
and shape distribution of the originally spherical balls
changes continuously as a result of impact breakage and
due to dierent wear mechanisms taking place inside the
ball mills. The wearing and breakage produce mostly
smooth, non-spherical ball shapes, some characterised by
exposed cavities or gross porosity holes (Vermeulen and
Howat, 1989). The movements of these irregularly shaped
components are believed to dier signicantly from those
of larger, more rounded components. Spherical media are

1358

N.S. Lameck et al. / Minerals Engineering 19 (2006) 13571361

also associated with high foundry production costs in comparison with other shapes that can be produced. An avenue
of using alternate media shapes (cylpebs, conipebs and
boulpebs) over spherical balls brings about various advantages such as lower foundry production costs, minimum
porosity produced on casting and an expected increase in
area, linear and point contact which the media make with
each other (Cloos, 1989).
It is long believed that a load comprised of a dierent
media shape other than spherical balls might inuence
the performance of a tumbling mill through variations in
load behaviour, charge segregation and power drawn by
the mill as well as the breakage kinetics. The extent of this
inuence has not yet been established. Even with the very
little done toward studying media shape eects (Herbst
and Lo, 1989; Yildirim et al., 1998; Shi, 2004), all eorts
have exclusively been focused on breakage rate, ignoring
other parameters dening mill performance such as load
behaviour and mill power. This paper investigates the inuence of dierent media shapes on the load behaviour and
power drawn by the mill as a function of mill lling and
speed.
The charge orientation is an important aspect of load
behaviour and is generally accepted to assume a quartermoon shaped prole (depending on load volume) with
toe and shoulder positions being key parameters. These
are dened, respectively, as the angular position where
the lifters/liner comes into contact with the charge and
where the charge departs from the liners, respectively.

media, cylpebs and worn balls. The worn media were

obtained from an industrial dry ball mill used by ESKOM
(South Africa power utility) to produce pulverized fuel following an experiment done to characterise ball size and
shape distribution (Lameck, 2005).
Material properties such as coecient of friction, size
and shape (Zhou et al., 2002) aect media packing. With
this fact in consideration Eq. (1) was used to calculate
the spherical ball charge mass at 15%, 20% and 25% lling
levels. The voidage of 0.4 was used in calculating spherical
balls mass. The charge mass involving cylpebs and worn
balls were obtained as respective equivalent weight to that
of spherical balls. The respective percent lling levels for
cylpebs were 14.6, 19.5 and 24.3 at the voidage of 0.36
(0.02) and that of worn balls at 0.37 (0.01) voidage were
14.2, 18.9 and 23.6. The brackets show the standard deviations of the measured media voidage. This implies that
media shapes were compared based on the same mass criterion. Measurements were performed with speeds ranging
from 20% to 90% of critical speed. Balls between 22.4 and
26.5 mm size class for spheres and worn balls and
24 22 mm (l d) for cylpebs were used.

2. Experimental equipment and method

The patterns of the signals obtained over one mill revolution from the inductive probe are shown in Fig. 1 for the
three media shapes. The rst momentary contacts were
considered to be the result of balls bouncing around due
to the turbulent nature of the charge; the actual toe position was therefore taken as a point when the probe was
in persistent detection of media, while the shoulder position
was obtained when detection of media around sensing face
ceased due to media departure from the mill liners (Fig. 1).

2.1. Description of load position

Threshold
5

Inductive probe signal (Volts)

Experiments were conducted in 0.54 m-internal diameter

mill with 0.4 m length. The internal mill length was
adjusted to 100 mm using a movable diaphragm. The mill
is driven by a 2.5 kW variable speed motor mounted in a
mill rig, which has been described previously (Moys
et al., 1996).
Twelve 20 mm high trapezoidal lifters with 45 face
angle and 50 mm base width were inserted inside the mill.
An inductive probe inserted at 50 mm from the mill front
end through the mill shell between two lifters was used to
obtain the charge position (Kiangi and Moys, 2006). The
probe was able to detect any metal present within the range
of 38 mm from its sensing face. The inductive probe
installation has been described (Kiangi and Moys, 2006).
The inductive probe enabled the measurement of data
used to obtain the load orientation. A signal (produced
by a light emitting diode (LED) and a phototransistor)
when the mill passes a mirror xed at the 12 oclock position served as a reference point. All the signals were sent
through an interface to a data acquisition computer. A
video camera was also used to record tumbling charge
behaviour inside the mill through a transparent front end
plate.
Three types of media were available to study the eect of
media shape on load behaviour and power; spherical

0:6J pD2inside LqBall

1
4
where J is the load volume as a fraction of the mill volume,
and L is the internal mill length.
Charge mass kg

Shoulder

Toe

4
3

Spheres

12 oclock
(0, 360)

Cylpebs

1
0
-1
-2
-3
-4
110

Worn balls
155

200
245
Position (Degrees)

290

335

Fig. 1. Typical signals produced by inductive probe (one rev., N = 50%

Ncrit, J = 20%).

N.S. Lameck et al. / Minerals Engineering 19 (2006) 13571361

The threshold voltage of 3.65 V was used to mark the toe

and shoulder positions.
The power drawn by the mill was generally given as:
2pNT
60

where N is mill speed in revolutions per minute and T is the

torque exerted by the charge and was measured using a
technique established with an accuracy exceeding one percent (Moys, 1993).
3. Results

Position (Degrees)

Results presented in Figs. 2 and 3 shows the variation of

toe and shoulder positions for mill lling levels of 15% and
25% with mill speeds. The error bars show the standard
deviations (r) of the data at the respective mill speeds.
The mill speed points for spheres and cylpebs were shifted
for 1% and 1%, respectively to allow unambiguous pre-

350

350

305

305
Shoulder

260

Spheres (N+1%)
Worn balls (Unshifted)
cylpebs (N-1%)

215
170
125
80

260

350

350

215

305

305

170

Toe

125

20

40
60
Mill speed, N (% Critical)

80

sentation of variability. One can generally note that, at

lower speed (660% of critical speed), the toe positions
are similar for all media shapes. However at higher mill
speeds, lower toe positions were obtained for cylpebs than
the other two media shapes.
Toe position data were analysed using a t-test at 95%
condence limit for hypothetical testing of signicant differences among the shapes. Spheres and worn balls, spheres
and cylpebs as well as worn balls and cylpebs dierences
were tested. From the t-test analysis results it was found
that at all mill speeds studied, toe positions varied among
media shapes. However at mill speeds of 660% of critical,
variations in toe positions with media shape were less than
that at speeds greater than 60% of critical.
Observation of charge shoulder positions showed that
the cylpebs are raised to a higher position as compared
to worn and spherical media. This is clearly shown in Figs.
2 and 3. The higher shoulder positions of cylpebs than
other media also suggested early cataracting and premature
centrifuging for cylpebs as compared to the other two
shapes.
The variations of toe and shoulder positions with load
lling (J) are shown in Figs. 46. The shoulder positions

80
100

Position (Degrees)

1359

Shoulder

260

260
J15
J20
J25

215

215

170

170
Toe

125

Fig. 2. Variation of toe and shoulder positions with mill speed (J = 15%)
(data mean std, N values shifted to allow unambiguous presentation of
variability).

80

125

10

20

30

40
50
60
70
Mill speed (% Critical)

80

90

80
100

Fig. 4. Variation of toe and shoulder position with mill speed at dierent
charge lling levels (spherical media).
350

350

305

305

350

350

260

305

305

260
Spheres (N+1%)
215

215

Worn balls (Unshifted)

Cylpebs (N-1%)

170

170
Toe

125

Positions (Degrees)

Position (Degrees)

Shoulder

Shoulder

260

260
J15
J20
J25

215

215

170

170

125

Toe
125

80

20

40
60
Mill speed, N (% Critical)

80

80
100

Fig. 3. Variation of toe and shoulder positions with mill speed (J = 25%)
(data mean std, N values shifted to allow unambiguous presentation of
variability).

80

125

10

20

30

40
50
60
70
Mill speed (% Critical)

80

90

80
100

Fig. 5. Variation of toe and shoulder positions with mill speed at dierent
charge lling levels (worn balls).

N.S. Lameck et al. / Minerals Engineering 19 (2006) 13571361

350

Positions (Degrees)

305
Shoulder

260

160

160

305

140

140

120

120

100

100

80

80

260

J15
J20

215

350

215

J25
170

170

Toe

125

125

80
0

10

20

30

40
50
60
70
Mill speed (% Critical)

80

90

80
100

Fig. 6. Variation of toe and shoulder positions with mill speed at dierent
charge lling levels (cylpebs).

Power (watts)

1360

60
40
20
0
0

60
Worn balls
Cylpebs
Spheres

J25
J15

10

20

30

40
50
60
Mill speed (% Critical)

70

80

40
20
90

0
100

Fig. 8. Variation of mill power draw with mill speed (J15 & J25).

Angle of repose (Degrees)

increased with increase in mill speed and charge lling and

are in agreement with previous ndings (Liddell and Moys,
1988; Moys and Skorupa, 1993). The toe positions were
steady with mill speed up to about 60% critical. Beyond
this speed, the spherical ball toe positions were almost constant (Fig. 4) while that of cylpebs shown a distinct
decrease in toe position with increase in mill speed (Fig. 6).
It was further observed that charge toe and shoulder
positions were aected by charge lling (Figs. 46). The
shoulder positions dierences for cylpebs (Fig. 6) between
dierent lling levels were not as signicant as for spherical
media.
The experimental results obtained at mill speeds 660
were used to estimate the dynamic angle of repose of the
charge for the three media shape. Fig. 7 compares the angle
of repose among the three media shapes at various mill
speeds for 20% mill lling. It was observed that cylpebs
have higher angle of repose than worn and spherical media.
Trends were very similar for 15% and 25% mill ling levels.
Although the real operating speeds are greater than 60%, it
is long established that at lower mill speeds, the angle of
repose of the charge can be used to estimate the charge volume; this assumes that, a simple chord joining the toe and
shoulder position can be used to represent the charge prole. This also can be used in estimating the dynamic coef-

60

60

50

50

40

40

30

30

Spheres
Worn balls

20

20

Cylpebs

10
0

10

10

20

30
40
50
Mill speed (% Critical)

60

0
70

Fig. 7. Comparison of angle of repose at various mill speeds (J = 20%).

cient of friction of the charge cascading down the toe

position.
It was further found that angles of repose of the charge
increased with mill speeds for the studied mill speeds.
3.1. Mill power
Sensitivity of power draw to media shape and operating
parameters in terms of speed and charge lling level was
analysed and is shown in Fig. 8. The power drawn was
found to be sensitive to media shape at all charge lling levels studied. At a mill speed of about 72%, which is about
the speed most mills are operated; power drawn by the cylpebs and spherical media is similar at all charge levels studied, however beyond this speed, power drawn by cylpebs
starts to decrease but that drawn by spheres is still increasing. Dierent media shapes reached the maximum power
draw at dierent mill speeds. The maximum power for
spheres occurs at a speed greater than 90% of critical
(i.e., higher than the highest speed tested).
4. Discussion
The load behaviour inside tumbling mills depends on
collective interaction and interlocking between individual
media and between media and lifters. These interactions
and interlocking are greatly inuenced by point or surface
contact among media to each other and/or with the lifters.
For media-liner contacts, the nature of contact is characterised by mainly two types of motions, rolling and sliding
on the faces of the lifter. These are signicantly aected by
media shape. Generally, spherical media is likely to be subjected to all two motions while irregular and cylpeb shapes
experience restricted rolling as well as sliding motions due
to surface contact. This may result in a delayed departure
from the mill surface at the shoulder and hence these media
may be raised to a higher shoulder position as compared to
spherical media.
Various power predicting models developed had lumped
the eect of grinding media properties in terms of size and

N.S. Lameck et al. / Minerals Engineering 19 (2006) 13571361

shape into a constant. This suggests that for any media

shape used, the maximum power would occur at a xed
speed. This is not the case as the experimental results show.
Cylpebs power draw reached its maximum at a lower speed
compared to worn and spherical media. On the other hand,
spherical balls reached maximum power draw at speed
greater than 90% which is contrary to ball mill data, this
is probably because of the higher lifter/media ratio used
such that more power was needed to provide lift to balls
which were still locked between lifters.
This observation is supported by shoulder position data
observed in Figs. 2 and 3. This may possibly be explained
by looking at shoulder positions among media shapes. As
load shoulder position increases, power also increases.
However, continuous increases in shoulder results in more
cataracting, increasing the number of balls in ight as well
as impacting at the toe of the mill. Energy is imparted back
to the mill by the impacting balls at the toe, leading to a
loss in power as the amount of centrifuging increases.
5. Conclusion
Information has been gained on the eects that media
shape have on load behaviour and mill power. Load positions that were quantied in terms of toe and shoulder
positions are dierent for dierent media shapes. Cylpebs
have higher shoulder positions at all the speeds studied
and spherical media have the lowest values. The trends
are such that for all three media shapes, shoulder positions
increase with mill speed. While the shoulder position for
spheres increases with mill charge lling, there is little variation in the shoulder positions with mill lling for cylpebs
media.
It has been found that at speed less than 60%, the toe
positions vary little with mill speed but above this speed,
cylpebs media have lowest angular toe positions of the
three. The toe positions for spheres are inherently constant
above mill speed of 60% and higher than worn and cylpebs
media while that of cylpebs media decreases with increase
in mill speed.
It was also observed that mill power was sensitive to
grinding media shape used. Power increases to the maximum with increasing mill speed for all media shapes studied. The maximum power draw for cylpebs media occurs at
a relatively lower speed in comparison to the other media
shapes studied. At lower speeds, cylpebs media draw more
power followed by worn balls and lastly spherical balls.
The load positions observations show that cylpebs are
more dispersed (i.e., spread out over the liner surface) than
the other two media shapes.
While power models are useful in evaluating dierent
control strategies in the light of impact that load behaviour
has on milling, information gained can be much useful in
improving these models and selection of the right media

1361

shape to achieve specic requirement such as ne grinding

or mill power optimisation.
Acknowledgements
The authors acknowledge the nancial support from Eskom without which this work would not have been possible. The research group is also acknowledged for their
cooperation in conducting this work.
References
Austin, L.G., Klimpel, R.R., Luckie, P.T., 1984. Process Engineering of
Size Reduction: Ball Milling. AIME-SME, New York, USA.
Cloos, Ulrich, 1989. Cylpebs: An alternative to balls as grinding media
(Doering GmbH, West Germany). World Mining 36, 59.
Dong, H., Moys, M.H., 2003. Load behaviour and mill power. International Journal of Mineral Processing 69, 1128.
Herbst, J.A., Lo, Y.C., 1989. Grinding eciency with balls or cones as
media. International Journal of Mineral Processing 26, 141151.
Kiangi, K.K., Moys, M.H., 2006. Measurement of load behaviour in a dry
pilot mill using an inductive proximity probe. Minerals Engineering,
doi:10.1016/j.mineng.2006.01.007.
Lameck, N.S., 2005. Eects of grinding media shapes on ball mill
performance. M.Sc. Dissertation, University of Witwatersrand,
Johannesburg.
Liddell, K.S., Moys, M.H., 1988. Eects of mill speed and lling on the
behaviour of the load in a rotary grinding mill. Journal of The South
African Institute of Mining and Metallurgy 88, 4957.
Morrell, S., 1993. The prediction of power draw in wet tumbling mills,
Ph.D. Thesis, University of Brisbane.
Moys, M.H., 1993. A model of mill power as aected by mill speed, load
volume, and liner design. Journal of the South African Institute of
Mining and Metallurgy 93 (6), 135141.
Moys, M.H., Skorupa, J., 1993. Measurement of the radial and tangential
forces exerted by the load on a liner in a ball mill as a function of load
volume and mill speed. International Journal of Mineral Processing 37,
239256.
Moys, M.H., van Nierop, M.A., Smit, I., 1996. Progress in measuring and
modelling Load behaviour in pilot industrial mills. Minerals Engineering 9 (12), 12011214.
Powell, M.S., Nurick, G.N., 1996a. A study of charge motion in rotary
mills Part 1Extension of the theories. Minerals Engineering 9 (2),
259268.
Powell, M.S., Nurick, G.N., 1996b. A study of charge motion in rotary
mills Part 2Experimental work. Minerals Engineering 9 (3), 343
350.
Powell, M.S., Nurick, G.N., 1996c. A study of charge motion in rotary
mills Part 3Analysis of results. Minerals Engineering 9 (4), 399418.
Shi, F., 2004. Comparison of grinding media-Cylpebs versus balls.
Minerals Engineering 17, 12591268.
van Nierop, M.A., Moys, M.H., 1997. Measurement of load behaviour in
an industrial grinding mill. Control Engineering Practice 5, 257262.
Vermeulen, L.A., Howat, D.D., 1989. A sampling procedure validated.
Journal of The South African Institute of Mining and Metallurgy 89,
365370.
Yildirim, K., Austin, L.G., Cho, H., 1998. Mill power for smooth-lined
mills as a function of media type and shape. Fine Powder Processing
Technology, 6979.
Zhou, Y.C., Xu, B.H., Yu, A.B., Zulli, P., 2002. An experimental and
numerical study of the angle of repose of coarse spheres. Powder
Technology 125, 4554.