Article on optimizing cement milling process

1. CEMENT GRINDING OPTIMISATION
Dr Alex Jankovic, Metso Minerals Process Technology Asia-Pacific, Brisbane , Australia
e-mail: alex.jankovic@metso .com
Dr Walter Valery, Metso Minerals Process Technology Asia-Pacific, Brisbane, Australia
Eugene Davis, Metso Minerals Asia-Pacific, Perth, Australia
ABSTRACT
The current world consumption of cement is about 1.5 billion tonnes per annum and it is
increasing at about 1% per annum. The electrical energy consumed in cement production
is approximately 110 kWh/tonne, and around 40% of this energy is consumed for clinker
grinding. There is potential to optimise conventional cement clinker grinding circuits and
in the last decade significant progress has been achieved. The increasing demand for
“finer cement” products, and the need for reduction in energy consumption and green
house gas emissions, reinforces the need for grinding optimisation.
This paper describes the tools available for the analysis and optimisation of cement
grinding circuits. The application of the Bond based methodology as well as Population
Balance Models (PBM) is presented using a case study. The throughput of current
conventional closed grinding circuit can be increased by 10-20% by pre-crushing the
clinker using the Barmac crusher. A potential for application of stirred milling
technology for fine cement grinding was also discussed.
Key words: dry grinding, process optimisation, modelling
INTRODUCTION
For all dry grinding applications, cement production is certainly the most important. The
estimate for the world energy consumption for cement production is 18.7 TWh which is
approximately 0.02% of total world energy consumption per year. The world
consumption of cement was about 1.72 billion tones in 2002 and it is increasing at about
1% per annum.
Cement production process typically involves:
• grinding limestone (and other raw materials to achieve the right chemical
composition) to about 90% passing 90 microns in a dry circuit,
• making cement by the chemical reaction between the components of the ground
mixture. This chemical reaction occurs at high temperature in a rotary kiln,
• grinding the cement clinker nodules to 100% passing 90 microns in a dry circuit.

2. Grinding occurs at the beginning and the end of the cement making process.
Approximately 1.5 tonnes of raw materials are required to produce 1 tonne of finished
cement. The electrical energy consumed in the cement making process is in order of 110
kWh/tonne and about 30% of which is used for the raw materials preparation and about
40% for the final cement production by cement clinker grinding. Production costs and
environmental concerns are emphasizing the need to use less energy and therefore the
development of more energy efficient machines for grinding and classification.
The world energy consumed for cement production is similar to energy used for grinding
in US mining industry, which is a significant fraction of the world total. In the minerals
industry, research in modelling and simulation of the grinding process has a long and
successful history. However, in the cement industry, the grinding process is more of an
“art” than engineering and the equipment manufacturers exclusively hold the “know-
how”. The process is designed and operated using carefully guarded “recipes” and rules.
In such as environment there is little or no room for process understanding and
improvement. This paper discusses how techniques developed in minerals processing can
be applied in cement grinding optimisation. Only the cement clinker grinding is discussed
and area of raw material preparation is not covered.
CEMENT GRINDING
For most of the twentieth century, the dry grinding circuits for the production of finished
cement from cement clinker consist of two-compartment tube mills and the air separators.
It is not uncommon to produce the cement in an open circuit. Advances in cement
grinding technology is slow and these advances are limited to more developed countries.
Approximately 95% of the feed to the cement grinding circuit are clinker and the rest of
the feed are “additives” which includes grinding aids. The quality of cement is measured
by the surface area or the Blane index. The unit of the Blane index is m2
/kg, and this
index is determined by the Blane air permeability test. The surface area of the cement
powder depends on size distribution of cement particles; smaller particles have larger
surface area. If the particle size distribution is known, the Blane index can be successfully
predicted (Zhang et al, 1995)
The cement clinker grinding circuit reduces the feed from 80% passing size between 10
and 20 mm to 100% passing 90 microns. The size reduction takes place in a two-
compartment tube mill; the first compartment of the mill is shorter than the second
compartment. The coarse clinker is ground in the first compartment where larger balls
(80, 60, 50 mm) are used and the fine grinding is done in the second compartment where
smaller balls (below 25 mm) are used. A diaphragm (see Figure 1) separates the two
compartments and allows only particles below a certain size to pass to the second
compartment. Ground material exits the mill through the discharge grate which prevents
grinding balls from leaving the mill. A proportion of material, mostly fines, is “air-swept”
out of the mill. The final product is the fine fraction of the air classifier and the coarse
fraction returns to the mill.

3. Figure 1. Diaphragm between the two compartments of the two-compartment mill
– view from first compartment
In the past 20 years, high pressure grinding roll (HPGR) technology has been used in pre-
crushing of clinker. Presently, many American and European cement grinding circuits
have HPGR which increases grinding capacity and energy efficiency.
CEMENT GRINDING SIMULATION
To optimise cement grinding, standard Bond grinding calculations can be used as well as
modelling and simulation techniques based on population balance model (PBM). Mill
power draw prediction can be carried out using Morrell’s power model for tumbling mills
(Morrell, 1998).
Bond method
The established technique for determining power requirements for ball mills is the Bond
method (Bond, 1961). This method also involves the application of some ‘efficiency
factors’ as described by Rowland (1975).
Bond’s equation describes the specific power required to reduce a feed from a specified
feed F80 to a product with a specified P80:

4. Wm = Wi (
80P
10
-
80F
10
) (1)
where: Wm - is mill specific motor output power (kWh/t),
Wi - is the Bond ball mill work index (kWh/t),
P80 - is sieve size passing 80% of the mill product (µm),
F80 - is sieve size passing 80% of the mill feed (µm),
The “efficiency factors” (Rowland, 1975) modify equation 1 so that it caters for circuit
conditions which are different from that Bond used to develop his original equation.
There are efficiency factors for dry grinding, open circuit ball milling, mill diameter,
oversize feed, grinding finer than 75 microns and too large or too small reduction ratios.
For cement application, the dry grinding (EF1), mill diameter (EF3), oversize feed (EF4)
and fine product (EF5) factors are relevant. Therefore, the equation for the specific power
requirement is:
Wm = EF1 EF3 EF4 EF5 Wi (
80P
10
-
80F
10
) (2)
According to Bond the specific power (calculated using Equation 1) should be multiplied
by EF3 where the mill diameter exceeds 8ft. Rowland (1975) modified the application of
EF3 and stated it should be used up to mills of 12ft. For mills larger than 12ft the value of
EF3 remains constant at the value for 12ft mills.
EF3 = (8/Dft)0.2
= ( ) .2 0 2.44
D
(3)
where: Dft - is the mill diameter inside liners (ft)
D - is the mill diameter inside liners (m)
The oversize feed factor (EF4) caters for situations where the feed size is coarser than a
specific size limit (F0), which is a function of ore hardness. Bond argued that if the feed
ore were coarser than this, bigger balls would be needed to break the coarser feed
particles at the expense of grinding of smaller particles. Conversely, if smaller balls were
used to grind the finer particles, the smaller balls would not break the coarser particles.
Either way a grinding inefficiency would result. The EF4 factor is applied only when the
F80 is greater than F0 and has a value greater than 1.
The correction factor, EF4, for the ore feed size is calculated as follows (Rowland, 1975):
EF4 = [Rr + (Wi - 7) (
0
080
F
FF −
)] / Rr (4)
P0 = 4000 (
Wi
13
)0.5
for ball mills (5)
Rr = F80/P80 (6)
where: Rr - size reduction ratio

5. F0 - optimum mill feed size (µm),
F80 - actual mill feed size (µm),
P80 - mill product size (µm),
The correction factor EF5 for the products finer than 75 µm (Rowland, 1975) is
determined by:
EF5 =
80
80
P*1.145
3.10P +
(7)
It was found in the crushing area that there are significant differences between the real
plant data and the Bond calculations and therefore empirical corrections were introduced.
The following modified Bond equation was proposed for crushing (Magdalinovic, 1990):
Wc =
cP
A
Wi (
cP
10
-
cF
10
) (8)
where:
Wc - is energy consumed for clinker crushing (kWh/t),
Wi - is Bond ball mill work index (kWh/t),
Pc -is sieve size passing 80% of the clinker after crushing (µm),
Fc - is sieve size passing 80% of the clinker before crushing (µm),
A - is an empirical coefficient, dependant on clinker and crusher properties
Based on the above considerations for crushing and grinding, the energy consumption for
the clinker pre-crushing and ball milling can be estimated using the following Bond
based model:
W = Wc + Wm (9)
As pre-crushing product size Pc is equal to the mill feed size F80 then:
W =
80F
A
Wi (
80P
10
-
cF
10
) + 1.3 * ( ) .2 0 2.44
D
* {[Rr + (Wi - 7) (
0
080
F
FF −
)] / Rr} *
80
80
P*1.145
3.10P +
* Wi (
80P
10
-
80F
10
) (10)
PBM models
In the area of dry particle reduction, the population balance models (PBM) for crushers,
HPGR, ball mills, air-swept ball mills and air separators have been developed (Lynch et
all 1977, Austin et al, 1980; Zhang,Y et al, 1988, Morrell et al, 1997, Benzer et al, 2001,

6. 2003). These models can be used to simulate cement grinding circuits and to assist their
optimisation.
The “work horse” of the cement grinding plant is the two-compartment ball mill,
commonly called the tube mill. Significant advances in model development were
achieved in recent years (Benzer et al, 2001, 2003) through research on industrial scale.
The breakage and transport mechanisms are better understood as well as the classification
action of the diaphragm. Further advances are expected in modelling the effect of
diaphragm design on classification and powder transport.
The basis for modeling the two-compartment ball mill is the perfect mixing ball mill
model. It can be illustrated by the following equation (Lynch, 1997):
i
ii
i
j
jj
d
pr
p
d
pra
f
i
j
ij
i +=+ ∑=
][
1
(11)
where:
fi - feed rate of size fraction i (t/h)
pi - product flow of size fraction i (t/h)
aij - the mass fraction of size that appear at size i after breakage
ri - breakage rate of particle size i (h-1
)
si - amount of size particles inside the mill (tonnes)
di - the discharge rate of particle size (h-1
)
The model consist two important parameter, the breakage function (aij) that describe the
material characteristic and breakage/discharge rate function (ri/di) which defines the
machine characteristics and can be calculated when feed and product size distribution are
known and breakage function is available.
The air classifier controls the final product quality. Therefore, the air classifier has a
crucial role in the circuit and a lot of attention is paid on the design and operation of the
air classifier. The classification action is modeled using the efficiency curve approach
(Lynch, 1997). Effect of the classifier design and operational parameters on the efficiency
is complicated and work is in progress to improve the current models.

7. THE CASE STUDY
Potential benefits of using the Barmac crusher for clinker pre-crushing were studied for a
cement plant. Figure 2 shows the proposed cement grinding circuit.
Double-deck
screen
Barmac
crusher
Clinker
Storage
Storage
Two-compartment
dry ball mill
Air
classifier
Finished
cement
Figure 2. Simplified cement grinding circuit with precrushing stage
Barmac B-series VSI (vertical shaft impactor) crushers are applied to a broad range of
materials in minerals and aggregate industry. Due to the “autogenous grinding action” it
is especially efficient for high abrasive materials such as cement clinker. The crushing
action is schematically presented in Figure 3.
When the incoming rock passes over the distributing plate the rock is divided in three
separate streams and is forcefully impacted on the rock lining at the distributing plate.
The material is being rapidly accelerated by the centrifugal force of the rotor action and is
compressed against the rock lining which is formed in the crushing chamber. Multiple
events occur and a variety of forces act on the individual particles as they proceed. The
crushed rock is discharged through the clearance between the crusher chamber and the
rotor wall.

8. Figure 3. Schematic of Barmac crusher operation
0
10
20
30
40
50
60
70
80
90
100
0.01 0.1 1 10 100
size (mm)
cum%pass
Clinker
case A
case B
case C
Figure 4. Clinker and the crushed clinker size distribution (case A, B, C) using
Barmac crusher
Figure 4 shows the clinker size distribution and predicted Barmac crusher size
distributions. These data are based on Barmac pilot testing and the simulations. It can be
observed that the crushed clinker has a significant amount (10-20%) of -75 micron

9. material. As the cement size is in this size range, it could be concluded that the Barmac
crusher produces a significant amount of finely ground cement.
Table 1. Bond method power calculation for two-compartment mill
Mill feed F80 (mm) EF4 Power required (kW) Difference (%)
15.5 1.06 3564 0.0
4.5 1.01 3251 8.8
3.0 1 3133 12.1
1.8 1 3018 15.3
It can be seen that a reduction in two-compartment mill power in the order of 9–15% is
calculated for different crushed clinker feed F80 sizes. Corresponding to the change in
factor EF4, a 5-6% reduction comes from improving the milling efficiency with finer
feed. In order to get this improvement, the ball charge size distribution in the first
compartment needs to be adjusted for finer feed.
As energy is consumed in pre-crushing stage, the total power required will be sum of
two-compartment ball mill and Barmac crusher power. Figure 4 shows how the total
comminution energy consumption depends on the Barmac crusher product size. It can be
seen that the total energy consumption would reduce with reducing the product size of the
Barmac crusher. This indicates that the cement grinding efficiency may be improved up
to 10% compared to a conventional circuit without pre-crushing. The alternative benefit
of introducing the Barmac crusher in pre-crushing stage is increasing the circuit capacity,
as the capital investment is relatively low compared to HPGR.
25
26
27
28
29
30
31
32
33
34
35
0 2 4 6 8 10 12 14 16
Barmac product P80 (mm)
TotalgrindingkWh/t
Figure 4. Specific power requirement for cement grinding circuit including pre-
crushing stage with Barmac crusher

10. The PBM modelling of the clinker grinding was carried out using the principles described
earlier. It should be noted that in order to obtain the “site specific” model constants,
detailed surveys of the milling circuit are required: the size distribution of the material in
each stream as well as from different points inside the mill. In this study the model
constants published in the literature (Benzer et all, 2001, 2003) were used. The base case
flowsheet generated in the JKSimMet grinding simulation software is shown in Figure 5.
The first compartment was modeled as two ball mills in series, diaphragm between the
first and second compartments was represented as a screen and the second compartment
is represented as one ball mill. Information given in the flowsheet is solids throughput
(t/h), 80% passing size and % passing 0.01 mm.
Figure 5. Greens Island base case JKSimMet simulation flowsheet
Using the above base case model, simulations with different feed size distributions (raw
clinker, and pre-crushed clinker using the Barmac crusher) were carried out, keeping the
product size constant at P80=0.038 mm. The resulting increase in throughput is shown in
Table 2.

11. Table 2. Simulated increase in throughput
F80 (mm) Simulated throughput
(t/h)
Increase (%)
15.5 110 0.0
4.5 125 13.6
3.0 135 22.7
1.8 140 27.3
Table 3 summarises the potential increase in throughput from using a Barmac crusher to
pre-crush the cement clinker. It should be noted that this result does not consider any
circuit physical limitations such as conveying, aeration and air classifier capacity.
Table 3. Predicted increase in throughput using the Bond and grinding modelling
method – base case throughput 110 t/h
Feed
F80(mm)
minus 75 microns in
Barmac product (%)
% throughput increase,
Bond method
% throughput
increase, PBM
15.5 0 0 0
4.5 12.1 9.1 13.6
3.0 15.4 14.3 22.7
1.8 20.5 18.2 27.3
It can be observed that Bond calculations gave less throughput increase than PBM
simulations. This could be because the PBM simulation results may have been
overoptimistic as the separator performance was kept the same. A model capable of
simulating changes in performance with different air separator loads was not available.
STIRRED MILLING POTENTIAL
As the product size decreases the energy required for particle breakage increases rapidly.
The pre-crushing stage increases the milling efficiency in the first compartment of the
two-compartment mill where coarse milling takes place. Pre-crushing however does not
affect milling in the second compartment apart from producing significant amount of
final product size (10-15 %). In order to produce the final product size, the length of
second compartment is usually double that of the first compartment. Smaller grinding
balls are also used in the second compartment. The efficiency of fine grinding in the
second compartment is largely controlled by the size of grinding balls. Due to limitation
of the mill’s rotational speed, the smallest ball size is usually restricted to about 15 mm.
Cement grinding using stirred mills (Pilevneli and Azizli, 1999) indicates that using
smaller media (5-8 mm range) improves grinding energy efficiency up to 50% using
stirred mills. For specialized types of cement, which are finer than Portland, this figure
would be even higher. Significant benefits of using Tower mill were also reported
(Shibayama et all, 2000) as well as the industrial applications for production of fine
12000 Blane cement. The stirred mills can be used in different roles and it is expected
that their application in the future will be significant.

12. CONCLUSION
There is a scope for significant optimisation of the traditional cement grinding circuits.
Grinding process modelling and simulation methods can be used for optimisation. A case
study conducted using the data from an industrial cement plant indicates that:
• Pre-crushing of cement clinker using a Barmac crusher offers realistic benefits to a
cement plant in terms of process efficiency.
• The introduction of the Barmac crusher can increase the cement circuit throughput in
order of 10-20%, providing that there is no capacity limitation in other parts of the
circuit. This is an attractive option due to relatively low capital investment of the
Barmac crusher. The overall energy efficiency of the circuit can also be improved in
order of 5-10%.
The stirred milling technology could further improve energy efficiency of cement
grinding. Initial work indicates great potential and significant development in this
direction should be expected.
REFERENCES AND BIBLIOGRAPHY
Austin, L. G., Weymont, N.P., and Knobloch, O., 1980. The simulation of air-swept
cement mill. Proceeding of European Symposium., Particle Technology. Amsterdam
Bond, F.C., 1985. Testing and Calculations. SME Mineral Processing Handbook ,
Norman L. Weiss, Editor in Chief.
Bond, F.C., 1961. Crushing and Grinding Calculations Parts I and II, British Chemical
Engineering, Vol 6., No 6 and 8.
Bond, F.C., 1962. Crushing and Grinding Calculations - April 1962 Additions and
Revisions. Allis-Chalmers Manufacturing Co., Milwaukee, Wisconsin.
Brugan, M.J., 1991. Mills at grinding edge. Pit and Quarry. Pp 34-41
H. Benzer, L. Ergun, M. Oner and A.J. Lynch, 2001. Simulation of Open Circuit Clinker
Grinding. Minerals Engineering, vol 14. No 7, pp 701-710.
H. Benzer, L. Ergun, M. Oner and A.J. Lynch, 2003.Case Studies of Models of Tube Mill
and Air Separator Grinding Circuits, Proceedings: XXII International Mineral Processing
Congress, Chief Editors: L. Lorenzen and D.J. Bradshaw, pp 1524-1533.

13. Lynch, A. J. (1977). Mineral Crushing and Grinding Circuit – Their Simulation,
Optimisation, Design and Control. Amsterdam, Elsevier Scientific
Magdalinovic, N., 1990. Mathematical Model for Determination of an Optimal Crusher
Product size. Aufbereitungs-Technik 31, Nr 5,
Morrell,S. and Shi,F.,and Tondo,l997., Modelling and Scale-up of High Pressure
Grinding Rolls, In the proceedings of the XX International Mineral Processing
Congress(IMPC), Aachen, Germany, September 1997.
Pilevneli, C.C., Khairun Azizi Mohd Azizli, 1999. Semi-Batch Dry Grinding Tests of a
Pilot Scale Vertical Stirred Mill. Proceedings of VIII Balkan Mineral Processing
Conference, Belgrade, Yugoslavia.
Rowland, C.A., 1975. "The tools of power: How to evaluate grinding mill performance
using Bond Work Index to measure grinding efficiency" AIME Annual Meeting, Tucson,
Arizona.
Shibayama, A., Mori, S., Bissombolo, A., 2000. Studies on Comminution Mechanism of
the Dry Tower Mill KD-3. Proceedings of the XXI International Mineral Processing
Congress, Rome.
Zhang, Y. M., Napier-Munn, T. J., and Kavetsky, A., 1988. Application of comminution
and classification modelling to grinding of cement clinker. Transaction Institution of
Mining and Metallurgy (Section C: Mineral Processing and Extractive Metallurgy). C207
- 213
Zhang, Y. M., Napier-Munn, T. J., 1995. Effects of particle size distribution, surface area
and chemical composition on Portland cement strength. Powder Technology 83, pp 245-
252