ICR article draft

I Optimised ball size distribution

GRINDING MEDIA

by Harun Koak,
emas Dkm Sanayi AS,
Turkey,
Prof Hakan Benzer &
Dr Namk Aydogan
Hacettepe University, Turkey

ince the 1980s tumbling mills were

designed to be consistintly used to
supply fine grinding for the cement
industry. Due to their large and heavy
design, these mills evolved into ball mills
and tube mills (long tumbling) mills. After
this development, the cement-making
process was composed of a short mill
operated in closed circuit with an air
separator or long mill operated in open
circuit. Subsequently, the short mills
and tube mills were combined with a
division head, separating the mill into two
compartments. In the 1920s and 1930s,
many combinations were tried for multiple
compartment mills for cement grinding
(Lynch and Rowland, 2005).
In tube mills, about 1-2 per cent of
energy supplied is used in fracturing
particles and the remaining energy ends
up as heat, so internal fittings such as
media charge, the mill lining system
and diaphragms became important.
Studies have shown that the effect of
a good design versus a bad design is
considerable, with savings on the specific
energy consumption of 5-10 per cent
being achievable. Similar increases in mill
throughput were also possible (Duda,
1985).

Modern milling
Today, grinding is still an expensive
process in the manufacture of cement and
several variables can affect the efficiency
and productivity of a dry grinding line
such as operating conditions of the
separators, air flow through the mill and
ball sizes in the mill compartments.
Optimising variables in the grinding
lines is an important step in minimising
the cost of production of cement per

Ball size distribution is an important parameter in the application of ball

mills, grinding performance can be significantly affected depending on the
design of the grinding media. It is a difficult task to evaluate performance
and where the problematic points are within a finish mill circuit.
Simulation is the best tool to employ to improve the circuit performance if
the model structures are accurate. Trial and error is an expensive and timeconsuming approach. Industrial case results by Cemas indicate that a
15 -20 per cent gain can be achieved by optimising the ball size
distribution, using developed model structures.

tonne. Ball size distribution is also a

critical parameter for an efficient grinding
operation. A survey covering most of
cement plants in Turkey has shown
that an extremely large range of ball
size distribution is employed to grind
the same type of material with similar
grinding characteristics in different plants.
It also found that their energy
consumption (kWh/t) exhibits large
variations from plant to plant.
Conventional ball size selection is based
on Bonds equation which advocates
charging mills with the largest size of
balls that can be used. The parameters
affecting the choice are feed size (F80),
mill type, specific gravity, work index,
critical speed and mill diameter. From
the largest size, the distribution of ball
sizes are determined using an empirical
formula. This approach has been widely
accepted as an industrial standard and,
therefore, its validity has not been
challenged for years (Gupta and Yan,
2006).

Modelling and
distribution of ball size
Mathematical modelling requires
collecting extensive data around
grinding circuits operating at different
conditions so that the developed models
can successfully predict the full-scale
operation. Modelling and simulation
techniques have been successfully applied
to optimise wet grinding systems. These
models are capable of reflecting the effect
of design and operational characteristics
of wet grinding ball mills on breakage
rate/discharge rate (r/d) combined model
parameter defined on the basis of Perfect
Mixing Model (Whiten, 1976).

A focus on ball size

optimisation in a cement mill

Several investigators (Austin et

al, 1975; 1984; Zhang, 1988; 1992;
Benzer, 2000; 2001; Slanewski, 1985;
Viswanathan, 1988; Ozer, 2006) have
studied mathematical modelling of
cement mills on the basis of population
balance models (PBM) for simulation and
optimisation of cement grinding circuits.
However, to date there is no extensive
research on the analysis of effect of
design and operational characteristics of
dry grinding ball mills on the grinding
model parameters namely breakage and
discharge rate functions of particles, based
on the industrial scale data.

Importance of the
grinding media quality
Grinding ball production is a metallurgical
operation and its quality is determined
by the quality of the production process.
Quality of one process has an indirect
effect of the quality of another process.
OCTOBER 2011 ICR XX

GRINDING MEDIA

The ball specifications used in the cement

industry are briefly outlined in Table 1.
The ball quality parameters can be
classified as the wear rate, dimensional
tolerance, ovality, mismatch, gating
knolls and cavity. The ball wear rate in
the cement industry is 40g/t for a closed
circuit application to produce a cement
fineness of 3700Blaine as the abrasive
content increases the ball consumption
increases, ie 150-200g/t for cement with
slag.
The dimensional tolerances of the
grinding media should be in the maximum
range of 5 per cent. The rest of the
quality figures are outlined in Table 2.
Once the quality of the balls used
is assured then the optimal ball size
distribution design has to be carried out
for a better performance of the circuit and
the balls.

Measuring ball perfomance

Before an optimal design, a complete
analysis should be conducted for
each milling duty. Factors such as mill
dimensions, mill speed, mill power,
ore type, feed top size, feed size
distribution, throughput, charge volume
and product size should be considered.
The measurement must take account of
several parameters, ie:
energy consumption
capacity
product size
liner and grinding media performance
and economy.
Complete analysis involves extensive
circuit sampling, sample analysis, mass
balancing and data analysis. Circuit
sampling is carried out under normal
operating conditions, while the circuit is
running at its steady-state condition when
the fluctuation of the critical parameters
are minimised. A typical steady-state
grinding circuit condition output is shown
in Figure 1.

Table 1: technical specifications of the balls used in the cement

industry
Diameter (mm)
15,17,20,25,30
35,40,50
60,70,80,90,100,110

C%
2.6-3.1
2.2-2.7
2.0-2.5

Cr %
12-14
12-14
17-19

Hardness HRC
60-66
59-65
58-64

Table 2: ball quality parameters summary used in the cement industry

Ball size Ovality
Ball size Mismatch Ball size
Gating knolls and
(mm)
(%)
(mm)
(mm)
(mm)
cavity (mm)
18-30
5-8
<25
0.25
<25
1
40-60
9-14
25, 30, 40
0.25 25, 30, 40
1.5
70-100 15-20 50, 60, 70
0.4
50, 60
2

80, 90, 100
0.4
70, 80 ,90
2.5
100
3

Figure 2: a typical grinding circuit and sampling points

Limestone
Gypsum

Air
Sampling Points

Clinker
lter

Cement silo

Class.
separ

Ball Mill

approximately
1m intervals.
The number of
sampling points
inside the mill is
important because
it should be
enough to follow
Figure 2 A typical grinding circuit and sampling points
Figure 3. Sampling points inside the ball mill
the variation
Operating conditions are recorded
in size reductions. Size distribution of
during the sampling period. To evaluate
samples are determined down to about
the performance of the existing circuits,
2m. Raw size distribution data then
for each circuit sampling surveys around
need to be mass balanced, and flow
the circuit are needed. A
rates around the circuit are calculated
simplified flowsheet of
on the basis of measured feed flow
a typical circuit with its.
rates. This first attempt identifies the
sampling points is shown
potential bottlenecks. A grindability test
in Figure 2.
is performed on the circuit feed sample
Sampling studies include
to assess performance of the actual mill.
all streams around the
Separator performance is evaluated by
circuit, as well as inside
examining the efficiency curve. In the
the mill samples after a
Figure 3, inside the ball mill sampling
crush stop. The samples
points are presented.
inside the mill are collected
Milling performance depends on
Figure 1: a typical steady state condition output of the control
along the central-axis with
the ball size distribution poor vs good
system around the circuit

XX ICR OCTOBER 2011

Figure 3: sampling points inside the ball mill

GRINDING MEDIA

is essential for the simulation study.

Laboratory scale tests are performed
to generate the breakage distribution
information (Genc, 2009). In ball
milling studies the models developed
by Hacettepe University were used
(Lynch 2000, Benzer, 2001). The model
parameters were determined by using the
non-linear regression technique.
The model structure uses perfect mixing
approach, providing that the mill can be
modelled in segments.

Table 3: the ball size distribution before and after the change
Before
After
Ball Size
Weight
Ball Weight

required
Size required
(mm) (%)
(t) (mm)
(%) (t)
40
22
28.16
25
29
37.12
30
30
38.4
20
21
26.88
25
26
33.28
17
50
64
20
22
28.16
Total
100 128
Total
100
128

XX ICR OCTOBER 2011

Figure 4: ball size distribution along the

second chamber

It includes two sets of model

parameters, ie the breakage function (aij)
and a combined breakage/discharge rate
(ri/di) function.
Ball size distribution is significantly
effective on breakage rates but in some
cases it may have an effect on the
discharge rate as well. Performance
Figure 5: change in retained % along the mill axis
evaluation or a sampling study is
Figure 4 Ball size distribution along the second chamber
essential to calculate the breakage
rate of a certain system. Additionally,
material characterisation in terms of
the breakage properties should be
completed. In Figure 8 the breakage
rate distribution of the two different
ball size distribution is illustrated.
These samples were taken from
industrial mills. The grinding rate
l axis
of Plant 1 is higher than Plant 2,
Figure 5 Change in retained % along the mil
study
is needed to establish an
Figure 6: ball size distribuion change within the mill
optimum ball size distribution
for a given ore and operating
conditions. As previously stated,
trial and error applications are
mostly expensive and timeconsuming.

Determination of
the suitable ball size
distribution-modelling
and simulation

Average Ball size (mm)

Average Ball size (mm)

performance. Ball mill performance is

affected by the ball size distribution. In
the following paragraphs good and bad
case studies will be presented as a part
of previous work done to identify how a
ball size configuration may affect the unit
process.
In the first case the data collection
indicated that good performance is
achieved along the mill axis by obtaining
the suitable ball size configuration at
different points along the mill axis by
classifying liner action (Slegten, 1973).
Figure 4 shows the ball size distribution
along the mill axis. As the material
becomes finer along the mill axis, finer
ball size distribution is required therefore
the grinding performance is increased
with a systematic size distribution (see
Figure 5).
The analysis shows that systematic
change in ball size distribution results a
systematic size reduction along the mill
axis. In this case liner configuration helps
the gradation of the balls. This systematic
change of the size reduction is an
indication of normal operating conditions
of a grinding circuit, but this does not
necessarily mean that the optimum ball
size distribution is being used in the
system. System performance may be even
better with another ball size configuration.
In the second case study, poor ball size
use deteriorates the mill performance
along the mill axis. Figure 6 and 7 indicate
ball size distribution change and size
reduction along the mill axis, respectively.
Figure 7 shows that random size
distribution changes along the mill axis.
This case study addresses the fact that
when a good ball size distribution is not
used in the system, milling conditions
may be improved by another ball size
distribution choice. These two practical
case studies suggested that a thorough

Simulation is a valuable tool in

process technology if the process models
therefore, it is possible to expect higher
are accurate and if model parameters can
grinding performance in Plant 1.
be determined in a laboratory or plant.
Ball size
distribution
is correlated with
Figure 6 Ball Size distribuion
change
within the mill
Mathematical modelling is now used
breakage rates so for different ball size
widely for the design and optimisation of
distributions, different breakage rates can
grinding circuits. The data obtained from
be achieved. With an optimum ball size
the mass balancing studies were used in
distribution, it is possible to minimise the
the model studies.
energy consumption while maximising
Apart from the circuit data generated
the milling rate. The aim is to increase the
by extensive sampling, determination of
maximum impact energy and the work
feed material breakage characteristics
done by the charge.

GRINDING MEDIA

Figure 7: change in retained % along the mill axis

Figure 8: breakage rates of the two

different ball size distribution
Figure 7 Change in retained % along the mill axis
100

-1

breakage rate (hr )

10

Plant I

0.1

Plant II
0.01
0.001

0.01

0.1

10

particle size (mm)

The ball size distribution

recommended for the second
compartment of the mill is shown
in Table 3.
By changing the ball size
distribution in the mill, the
breakage rate in the second
compartment is improved (see
Figure 9). The simulation studies
were conducted and the results
indicated a 15 per cent specific
energy consumption decrease in
the process.
After implementation of the
new ball size distribution the
energy consumption decrease was
recorded as 17 per cent. The new
ball size distribution did not change
the power draw of the mill, but
the mills production capacity was
increased from 115tph to 150tph
at the same quality figures.
The energy consumption fell
from 34.48 kWh/t to 26.66 kWh/t.

Circuit simulations are carried out

to quantify the effects of potential
disturbances on the circuit performance.
Figure Breakage rates of the two dierent ball size distribution
Conclusion
For a predetermined breakage rate, it is
Ball mill operation is a high cost unit
possible to run the simulations to find
operation and is affected by several
out the product size distribution and the
parameters. Ball size distribution is an
production tonnage. Different ball size
important parameter to achieve a better
distributions provide different production
grinding performance. For an existing
rates at a given product fineness.
operation, measuring the performance
Therefore, it is possible to find out the
by circuit sampling is the starting point
best ball size distribution by running the
to evaluate where the bottlenecks are
simulation studies.
located and how they are related to the
In the case study given below, the
performance of the circuit. Once the
performance evaluation studies in
bottlenecks are defined, generally a better
a cement circuit addressed a poor
performance can be achieved by a better
performance. Therefore, an optimisation
ball size distribution configuration. After
study was performed by running the
completing the material characterisaiton,
simulation studies. As a result of such
modelling and simulation is the best
studies it was possible to decrease the
tool to achieve this goal. The aim is to
specific energy consumption of the
increase the breakage rate of the mill.
mill motor by changing the ball size
Ball size distribution is correlated with
distribution in the mill. In addition to this
breakage rates so for different ball size
change it was necessary to optimise the
distributions it is possible to determine the
operating parameters such as separator
breakage rates. With the correct ball size
parameters around the circuit according to
distribution, energy consumption can be
new grinding conditions.
minimised while maximising the
Figure 9: the breakage rate change between the existing
milling rate.
ball size distribution in the mill and the new ball size
distribution offered
The aim is to increase the
maximum impact energy and
the work done by the charge.
In this article it is shown that
by using this technique it is
possible to obtain reduced cost,
due to less energy consumption
per tonne of cement produced.
________________________ I
Breakage rate (t/h broken per ton in the mill)

1000

100

10

present condition
new condition

0.1
0.01

0.1

Size (mm)

10

References
AJ LYNCH, CA ROWLAND, 2005: The
History of Grinding, SME
AUSTIN LG, LUCKIC PT and WIGHTMAN
D, 1975: Steady-state simulation of a
cement-milling circuit,
International Journal of Mineral
Processing, 2, p127-150.
AUSTIN LG, LUCKIE PT, SHOJI K,
ROGERS RSC and BRAME K, 1984: A
simulation model for an air-swept
ball mill grinding coal, Powder
Technology, 38, p255-266.
A GUPTA, DS YAN, 2006: Mineral
Processing Design and Operation, The
Netherlands: Elsevier
BENZER H, ERGN L, LYNCH AJ, NER
M, GNL A, elik I, AYDOGAN N,
2001, Modelling Cement Grinding
Circuits Minerals Engineering, Nr14/11
pp1469-1482
BEBZER H, ERGN L, NER M, LYNCH
AJ, 2001: Simulation of Open Circuit
Clinker Grinding Minerals
Engineering, Nr.14/7 pp701-710
LYNCH AJ, ner M, BENZER H, 2000,
Simulation of Closed Cement Grinding
Circuit, ZKG, No. 53/10, pp 560-564.
GENC O; BENZER H, 2009: Single
particle impact breakage characteristics
of clinkers related to mineral composition
and grindability, Minerals Engineering
Volume: 22 Issue: 13 Pages: 1160-1165
ZER CE, Ergn L, BENZER H, 2006:
Modeling of the classification behaviour
of the diaphragms used in multi-chamber
cement mills, International Journal of
Mineral Processing, Vol.80 No.1
H DUDA W, 1985, Cement Data Book-1.
Berlin.
SLANEWSKI, A, 1985, Optimization of
tube mill power consumption with VoestAlpine Grooved Liner Plates,
ZKG International, (12).
SLEGTEN PM, 1973, Cement mill liners
segregate grinding media, Rock Products,
76 (3) .
WHITEN, 1976, Ball mill simulation using
small calculators, AusIMM Proc. 258 ,
47 53.
VISWANATHAN, KN, 1988, Computer
simulation and optimization of ball mills/
circuit, WC, 19 (4) , 143-148.
ZHANG, YM, NAPIER-MUNN,
TJ, KAVETSKY A, Application of
comminution and classification
modeling to grinding of cement clinker,
Transactions of Institution of Mining and
Metallurgy, Sect C 1988, 97, 207-213.
OCTOBER 2011 ICR XX