Analysis of ball mill grinding operation using mill power specific kinetic

parameters
V. K. Gupta*, Shivani Sharma
Department of Fuel and Mineral Engineering, Indian School of Mines, Dhanbad 826 004, India

ABSTRACT

With a view to developing a sound basis for the design and scale-up of ball mills, a large

amount of data available in the literature were analyzed for variation of the two key mill

performance parameters: power specific values of the absolute breakage rate of the coarsest

size fraction, S*, and absolute rate of production of fines, F*, with some of the important

operating and design variables such as the mill speed, ball load, particle load, ball diameter and

mill diameter. In general, values of both the mill performance parameters were found to vary

significantly with the mill operating conditions. The nature and relative magnitude of variation

for the two parameters also differed significantly. Moreover, the effect of any particular variable

on the S* and F* values was found to be significantly different for different sets of operating

conditions. It has been emphasized that, as the purpose of grinding is to produce fine particles,

the mill design and scale-up work should be based mainly on the F* parameters. Moreover, it is

not correct to regard the S* values to be independent of the mill design and operating variables

as a general rule, especially for a fine analysis of the performance of the grinding systems.

Keywords: Ball mills; Design and scale-up; Power specific parameters; Breakage rate; Fines

production rate

______________________________________________________________________________

*Corresponding author: Tel: +91 326 2235246; fax: +91 326 2296563

Email address: gupta.vinaig@gmail.com

1. Introduction

The world over, every year several billion tons of metallic ores, minerals, cement and various

other solids used in the ceramic and chemical industries are subjected to size reduction in ball

mills. The specific energy consumption value for grinding of these materials typically ranges

from 5 to 50 kWh/ton. Thus, a significantly large amount of electrical energy is consumed in the

ball mill grinding operation. It is, therefore, important to establish the optimum values of various

mill operating parameters, such as the mill speed, ball load, ball diameter and particle load, from

the energy consumption point of view.

Another important task associated with the ball mill grinding operation is to establish a sound

basis for carrying out scale-up of ball mills based on laboratory or pilot scale test work. This

requires studies related to the influence of mill diameter on production rate of particles of a

desired size distribution.

During the last four decades, considerable amount of work has been done pertaining to the

above mentioned two tasks using a phenomenological grinding kinetics mathematical model

derived from population balance considerations. This model is based on two sets of parameters:

specific breakage rate and breakage distribution parameters for various size fractions of particles

[1-4]. The equation used for describing batch grinding kinetics is

1
()
= () + , Sj Mj (t) (1)

=1

where Mi (t) is the mass fraction of the particulate solids in the sieve size interval i (bounded by

size of the aperture of upper sieve xi-1 and size of the aperture of lower sieve xi), t is grinding

time, Si is specific breakage rate for particles of size class i (fractional rate at which material
breaks out of size interval i), and bi,j is the weight fraction of the material breaking out of sieve

size interval j that reports to sieve size interval i.

Herbst and Fuerstenau [5], Kim [6], Malghan [7], Malghan and Fuerstenau [8], Siddique [9],

and Fuerstenau [10] have analyzed the variation of grinding rate of some selected sieve size

fractions (such as a 10/14 mesh size fraction) of quartz, dolomite and limestone with various

operating variables in batch ball mills of different diameters. They concluded that the absolute

grinding rate (the product of specific grinding rate and weight of the particulate contents of the

mill) per unit net power input to the mill does not vary with the mill operating conditions such as

the mill speed, ball load, particle load and mill diameter. It was mentioned that the breakage

distribution parameters were also to a good first approximation independent of these operating

variables within the normal operating range. Their proposition can be mathematically expressed

as [11,12]

Si = Si [PH] (2)

where the proportionality constants Si* are independent of mill design and operating variables, P

is net mill power input and H is the weight of particulate contents of the mill. Si* can also be

defined as absolute breakage rate per unit power input (Si *= Si H/ P). For a batch mill drawing

constant power, the specific energy input to the mill E* is given by

E = P t H (3)

Incorporation of Eqs. (2) and (3) in Eq. (1) gives

1
( )
= ( ) + , ( ) (4)

=1

This energy-size reduction relationship predicts that for a given material and feed size

distribution a necessary condition for identical product size distribution in different batch mills is
identical specific energy inputs into each mill independent of mill dimensions and mill

operating variables in the normal operating range [11,12].

Besides the grinding rate of the coarser size fractions, the rate of production of the desired

size fine product is also an equally important parameter for characterization of the mill

performance. Let the weight fraction of material finer than size xi at time t be denoted by Fi (t).

For the rate of production of material finer than size xi we can write

1
()
= , () (5)

=1

where Bi,,j denotes fraction of broken product that is finer than lower size limit of the size

interval i when particles of size interval j undergo breakage. It is well known that in the batch

grinding operation, when a single size feed such as a 10/14 mesh size fraction is ground, the

initial rate of production of material finer than a given size remains constant for a short but

significant duration of time, depending on the fineness of the chosen size. This phenomenon is

known as zero order production of fines [13,14]. Let us now define a new parameter power

specific absolute rate of production of fines, Fi*, as

= ( ()/ )( ) (6)

Combining Eqs. (2), (5) and (6) we have

= , () (7)
=1

From Eq. (7) it follows that if both Sj* and Bi,j parameters were independent of the mill operating

conditions, then for a given size distribution of the particulate charge of the mill at time t, the

value of the parameter Fi* should also be independent of the mill operating conditions. In case of

a single size feed charge also the same should be true for the time domain corresponding to zero-
order production of fines. However, analysis of some of the published data and our own data

showed that this is not true. Significant variations were observed in Fi* with the mill operating

conditions.

As the main purpose of grinding operation is to produce fine particles of desired size and size

distribution, a study of variations in the Fi* parameters with the mill operating conditions should

be of greater concern and value for carrying out mill design and scale-up. Therefore, it was

decided to carry out a detailed analysis of the available experimental data with regard to the

variation of the two sets of mill power specific parameters, Si* and Fi*.

2. Approach to Analysis of Experimental Data

The technique commonly used for determination of the S value for the top size interval is

based on an assumption that the grinding rate is independent of grinding time [1-10,15-18].

Thus, the disappearance kinetics for the top size fraction in the particulate charge are first order,

as expressed below

dM1 dt = S1 M1 (t) (7)

From this equation we obtain:

ln M1(t) = ln M1(0) S1 t (8)

Generally, the product obtained from the first test is ground sequentially several times and, in

accordance with Eq. (8), the S value is obtained by fitting a straight line [1-10, 15-18]. It has

been claimed based on such graphical representation of experimental data that very frequently

the first order hypothesis is an excellent approximation to the truth [4]. But, this is not confirmed

by calculating the actual S values for different intervals of grinding time using the following

relationship
S1 = [ ln M1(t1) ln M1(t2)] / (t2 t1) (9)

Gupta [19] and Gupta et al. [20-22] have pointed out that when the crushed material obtained

from jaw and/or roll crushers is used as the feed to the batch ball mill, as is generally the case

with the data available in the literature, grinding rate of the topmost size interval is not time

independent as regarded by most of the researchers. Actually, it is found that the batch grinding

data is characterized by a high initial rate of breakage for a period of approximately 0.5 1 min.,

followed by significantly lower grinding rate values which vary with grinding time in a

somewhat random fashion[19,20]. There is no standard procedure for estimating a unique S

value from such data. The values of the grinding rate reported by researchers show a definite

personal bias [19,20]. Therefore, correlations developed by researchers are likely to be in error.

Gupta [19] and Gupta et al. [20--22] have pointed out that for a crushed particulate charge

large variations in the rate of breakage during the initial period of grinding are observed due to a

change in the particle strength and shape distributions. These distributions are characteristic of

the machine used for production of particles. With a view to avoid the above mentioned

problems, it has been recommended [20-22] that the crushed material should be pre-ground in

the test ball mill for a period of 1-2 min. Pre-grinding of the crushed material ensures that the

strength and shape distributions of particles do not undergo any significant changes during the

test work. Therefore, no significant variations in the grinding rate (as well as the breakage

distribution parameters) of material with grinding time are observed.

Further, a close examination of the published data on batch grinding of single size feeds

showed that during the initial 0.5 minute or so, the rate of production of fines is either

significantly lower or higher than the constant value exhibited during next 1-2 minutes [20].

Therefore, if rate of production of fines is estimated graphically by plotting Fi(t) versus t, the
estimated value is likely to get affected by the initial abnormal production rate value. It may be

mentioned that as the rate of production of fines is jointly determined by the S and B values,

which undergo significant change in the initial period due to change in the particle strength and

shape distributions, the initial variation in the value of the rate of production of fines too is to be

expected.

Estimation of the values of grinding rate parameters and rate of production of fines was a

straight forward job while analysing our own data as the test material was pre-ground in the ball

mill and, therefore, both the sets of parameters exhibited only very small and systematic

variation with grinding time. However, it became necessary to establish proper guidelines for

obtaining most likely estimates of the desired parameters from the data of other investigators.

Based on a comparative study of the results obtained by us for the crushed and pre-ground feed

charges, it was decided to ignore the initial abnormal values and take the average value for the

time interval during which the values were observed to be more or less constant. To illustrate the

proposed methodology of estimation of the values of model parameters, two examples are

presented in Tables 1 and 2. In these Tables, instead of the absolute values of the grinding rate

and rate of production of fines for different time intervals, the values relative to the value

observed for the first time interval, S and F, are shown for ease in visualising the nature of

variation with time.

3. Results and Discussion

With a view to facilitating a direct visualization of the percentage change in the S* and F*

values with various operating conditions, in this section results are presented in various figures in

terms of the normalized S* and F* values, denoted by S** and F**, respectively. In each case,
normalization was done using the S* and F* values corresponding to a chosen value of the

operating condition. For example, while discussing the effect of mill speed, normalization was

done with respect to 50% critical speed.

The S* value was calculated for the top most size interval, designated by i=1. Depending on

the top size, which varied from 10 to 20 mesh (Tyler Series: 1700 and 850 microns, respectively)

in most cases, a suitable choice of upper particle size limit had to be made for the F* parameter.

The selected size had to be reasonably finer than the top size and also be representative of a

typical desired product. Taking in to account these two factors and the results of data analysis

pertaining to the trend of variation for products of different fineness, it was decided to select 100

mesh (150 micron) as the upper size limit of the ground product. Therefore, F* values were

calculated using data corresponding to weight percent of ground product passing 100 mesh sieve.

3.1 Effect of mill speed

3.1.1 Effect of mill speed in dry grinding

Kim [6] has presented batch grinding data for 10/14 mesh size fraction of limestone ground in

a 25.4 cm ball mill (mill speed, N: 0.5-0.8 (50-80 % critical); ball size, d: 25.4 cm; ball load, J:

0.5 (corresponding to 50% filling of mill volume by static ball charge; particle load, U: 1.0

(corresponding to 100% filling of interstitial volume of static ball charge). Using this data the S*

value for 10/14 mesh size fraction and the F* values for -100 and -200 mesh products were

calculated. Fig. 1 shows variation of S* and F* with mill speed in terms of their normalized

values, S** and F** (for normalization, their respective values at 50% critical speed were used).

It can be seen that both the S** and F** values increase significantly up to 70% critical speed,
followed by a sharp drop. However, the increase in the two F** values is much more than that

observed for S**. While the peak S** value corresponds to an 18% increase, the corresponding

F** values for 200 and 100 mesh size show an increase of 32 and 39%, respectively.

Fuerstenau [10] has given values of S* for batch grinding of a 7/9 mesh size fraction of

dolomite in a 25.4 cm ball mill (ball size, d: 2.54 cm; ball load, J: 0.5; particle load, U: 1.0) at

four different speeds: 53, 60, 70 and 90 % critical. This data is presented in Fig. 2. It can be seen

that as the mill speed increases from 53 to 70% critical, the S** value increases by about 21%. A

further increase in the mill speed leads to a sharp fall in the S** value. This trend of variation in

S** is similar to that observed in Fig.1 for Kims data on limestone.

Gupta [23] has presented results on variation of grinding rate of different size fractions of

quartz, limestone, and two types of cement clinkers with mill speed. A 29 cm diameter mill with

a built in wavy liner was used. The ball charge consisted of equal number of 1.27, 1.90 and 2.54

cm balls. The ball load corresponded to J=0.4 and the particle load corresponded to U=1. An

analysis of the published experimental data on variation of the mill power with mill speed shows

that in the range 55-70% critical speed the torque corresponding to the net power drawn by the

mill either remains practically constant [10,24] or it increases gradually with the mill speed by

about 8% [17]. These data can be utilized for evaluating the energy efficiency of the grinding

operation. For this purpose, let us define an energy efficiency parameter as the ratio of the

power specific grinding rate at 55 and 70 % critical speeds, [S*(55)/S*(70)]. Values of this

parameter were calculated for different size fractions of four materials studied by Gupta [23]

assuming the mill torque to be independent of the mill speed. These values are given in Table 3.

According to the proposition of Herbst and Fuerstenau [5] that the S* values are independent of

the mill operating conditions, the value of the parameter under consideration should be observed
to be 1 in all cases. However, as can be seen in Table 3, the value of this parameter is observed to

vary from 0.79 to 1.43, being greater than one in most cases. It can be readily seen that a value

greater than one implies that particles were ground more energy efficiently at 55% critical speed.

Further, a close inspection of the data presented in Table 3 would show that in general as the

particle size decreases, the energy efficiency at 55% critical speed increases. This fact is more

important for a relatively soft material like limestone because even for a coarse size fraction,

14/20 mesh, grinding at 55% critical speed is 23% more energy efficient. However, in case of

quartz, a hard material, the energy efficiency for 10/14 mesh size fraction is 21% lower at the

55% critical speed. Therefore, grinding a coarse feed of quartz at a lower speed is not desirable.

3.1.2 Effect of mill speed in wet grinding

Kim [6] has presented a large amount of data on wet grinding of a 10/14 mesh size fraction of

limestone under different operating conditions. It is well known that the wet grinding kinetics are

strongly particle size distribution dependent. Therefore, a coarse (-14 mesh: 41-63%; -200 mesh

5.5-11%), a medium fine (-14 mesh: 63-86 %; -200 mesh: 11-21%) and a fine (-14 mesh: 87-97

%; -200 mesh 21-32 %) size distribution regimes were selected for studying the variation of S*

and F* parameters with mill speed in the range 50 to 80% critical. It may be mentioned that

though the feed and product size distributions for all tests were not identical, they were very

close (except for tests corresponding to 50% critical speed). The other mill operating conditions

were: % solids: 60; N: 0.6; U: 1.0, J: 0.5.

Results for the coarse size distribution regime are given in Fig. 3. It can be seen that as the

mill speed increases from 50 to 80 % critical, there is an overall decrease of 37 % in the S**

value for the 10/14 mesh size fraction, and 29 to 34 % decrease in the two F** values
corresponding to 200 and 100 mesh sizes, respectively. Moreover, in the range 50 to 60 %

critical speed, while the S** value shows a very sharp decline, the two F** values decline only

marginally.

Fig. 4 shows results obtained for variation in the S** and F** values in the medium fine size

distribution range (-14 mesh: 63-86%; -200 mesh: 11-21%). It can be seen that as the mill speed

increases from 50 to 70% critical, while the S** value increases by about 17%, the two F**

values do not increase significantly. Further increase in the mill speed results in a sharp drop in

the values of all three parameters.

For the fine size distribution regime, the trend of variation for the two F** values was

observed to be similar to that observed in case of the coarse and medium fine regimes. However,

the shape of the curve for the S** parameter was observed to be quite different from the shape of

the S** curves shown in Figs. 3 and 4. Based on these results it was concluded that in case of

wet grinding only the F* parameter can provide a reliable basis for the mill scale-up and design

work.

3.2 Effect of particle load in dry grinding

Berlioz (16) studied the effect of particle load on grinding rate of a 7/9 mesh size fraction of

natural, uncrushed dolomite in a 25.4 cm diameter laboratory ball mill (N: 0.6, J: 0.5, U: 0.2 to

1.6). Due to an unusually high initial rate of grinding, Berlioz defined sieve size interval 1 as

7/10 mesh instead of 7/9 mesh. The S* and F* values were calculated using the actual mill power

data reported by Berlioz. For calculation of the F* values 100 mesh was taken as the upper size

limit for the ground product. Further, the S* and F* values were normalized using the respective

values corresponding to U=1. Fig. 5 shows variation of the normalized values of the two energy
specific parameters with the particle load, U. As we move from right to left, the S** value

remains practically constant up to U=1 and the F** value remains constant up to U=0.8. Beyond

these limits, both the S** and F** values continuously decrease up to the U value of 0.2.

However, the overall decrease of 0.54 in the S** value is observed to be significantly higher than

the overall decrease of 0.38 in the F** value. These results show that as the U value decreases

the breakage distribution function becomes increasingly finer. Thus, the power specific absolute

rate of production of fines does not decrease as much as is the case with the power specific

absolute rate of breakage.

We carried out a detailed study on the effect of particle load and ball load on grinding kinetics

by carrying out experiments at two different mill speeds (55 and 70 % critical) and four levels of

ball load ( J: 0.2,0.3,0.4 and 0.5). The particle load U was varied in the range 0.25 to 2.0. As

mentioned earlier in section 3.1.1, a 29 cm diameter mill with a built in wavy liner was used

[22,23]. The ball charge consisted of equal number of 1.27, 1.90 and 2.54 cm balls. The feed

charge consisted of a well distributed -20 mesh feed of quartz that was prepared by grinding -8

mesh crushed quartz for two minutes in the test ball mill at the selected value of the mill speed.

For any given set of experiments, feed charges of almost identical size distributions were

prepared using a rotating table sampling device. Generally two short duration grinding tests were

carried out to obtain about 50% total reduction in the amount of the top size fraction, 20/28

mesh. The deviation in the S values calculated for two time intervals was found to be less than

2%. In almost all cases the second value was found to be higher than the first value. Further,

based on the correlations available in the literature, the duration of each grinding experiment was

calculated for obtaining almost identical product size distribution in all tests of a particular set of

experiments. In this context, it is important to point out that the phenomenon of zero-order
production of fines is not relevant to our experiments as a distributed feed was used instead of a

single-size feed. Therefore, a meaningful comparison of the rate of production of fines can be

carried out if only the feed and product size distributions are same. In about 40 % cases the

product size distributions were not close enough for comparing the rate of production of fines.

Therefore, F* values could not be calculated for these tests.

Fig. 6 shows variation of the normalized values of S*(20/28 mesh) and F*(-100 mesh) with

the particle load at 70% critical speed and 40% filling of the mill volume by the ball charge.

Normalization was done using the maximum values of the parameters obtained at a U value of

0.75. A comparison of the two curves shown in Fig. 6 with the corresponding curves in Fig. 5

would show two main differences: (i) for both the parameters the maximum value is reached at a

lower U value (0.75 as compared to 1.0); and (ii) the parameter values do not remain constant as

we move to the right from the maxima in fact a sharp decline is observed. There is, however,

one similarity in respect of the variation in the breakage distribution function with the particle

load. As we move to the left from a high U value, up to the occurrence of the maxima the

breakage distribution function does not change significantly, resulting in nearly parallel curves

for the two parameters. However, as we move further to the left from the maxima, the breakage

distribution function becomes increasingly finer resulting in relatively less decrease in the F**

value.

At present, it is not possible to properly explain the different patterns of results obtained in

the two studies. Though, there are small differences in the mill speed (N: 0.6 versus 0.7) and ball

load (J: 0.5 versus 0.4), in our opinion the main factor is the ball size distribution: a single size

ball charge versus a ball charge made up of balls of three different sizes. The void space in the

ball charge, the relative movement of different layers of balls and the type of breakage events
generated are expected to vary with the ball size distribution. In practice, the production mills

have a ball mix of a wide size distribution. Therefore, the results obtained in our study are of a

greater relevance.

For 70% critical mill speed, Fig. 7 shows variation of S** for 20/28 mesh size fraction of

quartz with particle load at three levels of ball load, J: 0.2, 0.3 and 0.4. It can be seen that all

three curves exhibit a maxima. The U value corresponding to the maxima, U*, decreases as the

ball load increases. Moreover, an increase in the U value beyond U* leads to a sharp decline in

the S** value for J=0.3 and 0.4. Fig. 8 shows results of similar experiments carried out at 55%

critical speed for J values 0.3, 0.4 and 0.5. Though the broad trend of variation is same in respect

of the location of the maxima and the degree of decline in the S** value for different J values, in

general the magnitude of variation is comparatively very small.

Thus, a small ball load and a low mill speed are preferable from the point of view of variation

in the specific energy consumption with particle load. However, in doing so one has to take in to

account the mill production rate also. Reasonably high levels of ball load and mill speed are

required to maintain a high production rate. Therefore, a practical compromise has to be made

between the mill production rate and variation in the specific energy consumption with particle

load due to changes in the feed rate and hardness of the feed solids.

3.3 Effect of ball load in dry grinding

Several researchers [4,17,18,25] have studied the effect of ball load on grinding kinetics.

Relevant data presented by Smaila [17] for 20/28 mesh quartz ground in a 60 cm diameter ball

mill ( N: 0.7, U : 0.75, J: 0.2-0.5) were analyzed for variation in the S* and F* values with the

ball load . These results are presented in Fig. 9 in terms of the relative values S** and F**. It can
be seen that both the curves exhibit a maxima at J=0.3 and a minima at a J value of about 0.45.

Further, variation in the F** value is more marked than the variation in the S** value (33 versus

15%). This shows that the breakage distribution function changed with the ball load.

Figs. 10 and 11 results obtained by us on variation of S** with ball load for 70 and 55%

critical mill speed. It can be seen that at 70% critical speed the maximum S** value is observed

at a J value of about 0.3, while at 55% critical speed the maximum value is observed at a J value

of about 0.4. A maximum of about 20% variation is observed in the S** value in both the cases.

Further, the degree of variation is more at lower U values. At a U value of about 1.5 there is

minimum variation in the S** value. Also, at a given U value the degree of variation is nearly the

same for the two mill speeds under consideration.

Fig. 12 shows variation in the F** value with ball load at 55% critical speed for three values

of U: 0.50, 0.75 and 1.5. In contrast to the trend of variation in the S** value at the same mill

speed (Fig. 11), it is observed that in all three cases the F** value increases monotonically with

J, without exhibiting a maxima. Another interesting observation is that the rate of increase in the

values of both the parameters is highest for a U value of 0.75.

3.5 Effect of ball diameter

The effect of ball diameter on grinding kinetics has been studied by several researchers [7,

17,18, 22]. Batch grinding size distribution data of Malghan [7] on 28/35 mesh limestone and

that of Smaila [17] on the same size quartz were found to be most suitable for comparing the

effect of ball diameter on grinding behaviour of two materials of significantly different hardness.

Figs. 13 and 14 show the variation of S** (28/35 mesh) and F** (-100 mesh) values for quartz

and limestone, respectively. It can be seen that in both the cases the F**and S** values decrease
with increase in ball diameter. However, the decrease in the F** value is relatively less than that

observed for S**. This shows that an increase in the ball diameter results in a finer breakage

distribution function which compensates for decrease in the breakage rate to some extent.

Further, it can be seen that the decrease in F** is relatively more pronounced in case of

limestone, and the S** and F** values are quite close. Thus, we see that the effect of ball

diameter is not same for two materials of different hardness. Increase in ball diameter gives rise

to relatively less change in the breakage distribution function for the softer material. In case of

quartz, which is relatively harder than limestone, increase in ball size results in a greater change

in the breakage distribution function. Therefore, the rate of production of fines does not decrease

as much as it does in the case of limestone.

3.6 Effect of mill diameter

Malghan [7] has presented data on dry grinding of 8/10 mesh limestone in 12.7, 25.4 and 38.1

cm diameter ball mills. As shown in Figs. 15 and 16, data for three sets of operating conditions

were analyzed. These three sets included two values each of mill speed (N: 0.6 and 0.7), ball

load (J: 0.3 and 0.5) and ball diameter (d: 25.4 and 50.8 mm). Particle load was kept constant at

U=1. It can be seen in Fig.15 and Fig.16 that both the S** and F** values exhibit a maxima at

about 33 cm mill diameter. However, the nature and magnitude of variation for these two

parameters is different for different sets of operating conditions.

Siddique [9] has presented data on dry grinding of 10/14 mesh limestone in 25.4, 38.1 and

76.2 cm diameter ball mills. The mill speed, ball load and particle load values corresponded to:

(N: 0.6; J: 0.5; U: 1.0). Balls of four different diameters, 12.7, 19.05, 25.4 and 38.1 mm, were

used for preparing the ball charge. The proportion of balls by number was: 36, 24, 26 and 14 %,
respectively. Fig. 17 shows the variation of S** and F** with mill diameter for this set of

operating conditions. It can be seen that while F** increases steadily by 16% over the mill

diameter range under consideration, S** does not exhibit any definite trend of variation of a

significant magnitude.

In view of the fact that Siddique used a mixture of balls of different size and mills of

relatively bigger diameter, his results should be considered to be more relevant to the actual

operating conditions. However, it is felt that the available data is not adequate for arriving at any

definite conclusions. A 12.5 cm diameter mill is too small for the purpose of scale-up. A

preferable range of mill diameter for the scale-up study would be: 30 to 120 cm.

3.7 General discussion

It has been demonstrated above that the basic assumptions made by Herbst and Fuerstenau [5]

in proposing Eqs. (2) and (4) are not valid in general. These equations were originally proposed

based on the results of only eight dry batch grinding experiments carried out in a 25.4 cm

diameter mill, corresponding to the operating conditions: (N: 0.53, 0.6, 0.7,0.9; J: 0.5; U: 1.0) , (

N: 0.6; J: 0.35; U:1.0) and ( N: 0.6; J: 0.5; U: 0.8, 1.0, 1.6). The value of S* was found to vary

from 1.74 to 2.18 kWh/ton a variation of 12.5% around the mean value of 1.96 kWh/ton [10].

Thus, these equations are only approximately valid over a narrow range of operating conditions,

as mentioned by these authors themselves. However, this fact has been overlooked by several

other researchers [26-29] who have used these equations in their analysis of experimental data as

universal correlations. In view of the results presented above in sections 3.1 to 3.6, these

correlations should be used with a good degree of caution.

 An important result obtained is that the effect of any particular operating variable on the

parameters, S* and F*, is not same, i.e. the F* values do not vary in the same proportion as that

observed for S*. This is because of the fact that the B values also vary with the operating

conditions. It may be recalled that the F* values are determined by S* and B values, both.

Further, the effect of any particular variable on the S* and F* parameters is observed to be

different under different set of operating conditions. Therefore, the effect of any particular

operating variable, such as the mill speed, on S* or F* cannot be described in terms of mill speed

alone. Actually, the relationship is expected to include other variables (such as the ball load, ball

size distribution and particle load) also.

Attention needs to be drawn to the fact that no significant data is available in the literature

on the wet mode of operation. In the mineral processing plants, most of the grinding is carried

out in the wet mode. And, as shown above in Section 3.1, the effect of speed on S* and F* values

is quite different for the dry and wet modes of operation. It is quite likely that same is the case

with other variables. Therefore, it is important that adequate data is generated for the wet mode

of grinding operation covering the range of operating conditions of practical interest. Three

variables must be particularly emphasized: particle size distribution of the mill contents, ball size

distribution and ore hardness.

It is proposed that the results presented in this paper should be verified through simulation of

the ball mill grinding operation under different sets of operating conditions using the discrete

element method [30-34]. This method enables the calculation of collisions and the resulting

motion of each and every ball in the charge, and it takes into account the internal geometry of the

mill as well as the size distribution of the ball charge. Such studies will enable us properly

account for the observed variations in the mill performance under different operating conditions.
 4. Conclusions

Analysis of a large amount of batch ball mill grinding data available in the literature has

shown that the values of the power specific absolute rate of grinding of the topmost size fraction,

S*, and the power specific absolute rate of production of -100 mesh fines, F*, are not

independent of the mill operating conditions and mill diameter. In general, both the parameters

exhibit a significantly large degree of variation, typically 20 to 40%. Further, the nature and

degree of variation for these two parameters often differ significantly.

The effect of any particular operating variable on the S* and F* values is significantly

different for different combinations of the values of the other operating variables. The breakage

characteristics (hardness of material) and the mode of grinding operation (dry/ wet) are also

important factors. It is, therefore, proposed that depending on the likely range of operating

conditions for the production mill, the optimum values of various operating variables should be

established by carrying out some well designed grinding experiments in a ball mill of at least 60

cm diameter.

Finally, as the purpose of grinding is to produce fine particles, it is recommended that the

mill design and scale-up work should be based mainly on the F* parameters.

Nomenclature

bi,j weight fraction of the material breaking out of sieve size interval j that reports to sieve

size interval i.

Bi,,j fraction of broken product that is finer than lower size limit of the size interval i when

particles of size interval j undergo breakage

E* energy input to the mill per unit mass of the particulate charge

F rate of production of fines relative to the value for the first time interval

F* absolute rate of production of fine material per unit power input

F** power specific absolute rate of production of fine material relative to the corresponding

value for some chosen value of the operating condition under consideration

Fi mass fraction of particulate solids finer than sieve i

H total mass of the particulate charge in the mill

J fraction of mill volume filled by static ball charge

Mi mass fraction of the particulate solids in the sieve size interval i

N fraction of critical speed of mill

P net power drawn by mill

S grinding rate relative to the grinding rate for the first time interval

S* absolute rate of breakage per unit power input

S** power specific absolute grinding rate relative to the power specific absolute grinding rate

for some chosen value of the operating condition under consideration

Si specific breakage rate for particles of size class i

t grinding time

U fraction of void volume of the static ball charge occupied by the particulate solids

5. References

[1] K. J. Reid, A solution to the batch grinding equation. Chemical Engineering Science 20

(1965) 953-963.
[2] L.G. Austin, A review introduction to the description of grinding as a rate process. Powder

Technology 5 (1971-72) 1-17.

[3] V.K. Gupta, P.C. Kapur, A critical appraisal of the discrete size models of grinding

kinetics. in: H. Rumf, K. Schonert, (Eds.), Proceedings, Fourth European Symposium

Zerkleinern, Dechema Monographien 79, Nr. 1576-1599, Verlag Chemie, Weinheim, 1976,

pp. 447-465.

[4] L.G. Austin, R.R. Klimpel, P.T. Luckie, The Process Engineering of Size Reduction:

Ball Milling, SME/AIME, New York, 1984, pp. 84-86.

[5] J.A. Herbst, D.W. Fuerstenau, Mathematical simulation of dry ball milling using specific

power information. Transactions SME-AIME 254 (1973) 343-348.

[6] J.H. Kim, Normalized model for wet batch ball milling. Ph.D. Dissertation, University of

Utah, Utah, USA, 1974.

[7] S.G. Malghan, The Scale-up of Ball Mills Using Population Balance Models. D.Eng.

Dissertation, University of California, Berkeley, California, 1975, pp.271.

[8] S.G. Malghan, D.W. Fuerstenau, The scale up of ball mills using population models and

specific power input. in: H. Rumf, K. Schonert, (Eds.), Proceedings, Fourth European

Symposium Zerkleinern, Dechema Monographien 79, Nr. 1576-1599, Verlag Chemie,

Weinheim, 1976, pp. 613-630.

[9] M. Siddique, A kinetic approach to ball mill scale-up for dry and wet systems. M.S.

Thesis, University of Utah, USA, 1977.

[10] D. W. Fuerstenau, Research on comminution process and simulation. Report No. BuMines

OFR 39-79, National Technical Information Service, U.S. Department of Commerce,

Springfield, VA. 22161, December 1978.

[11] J.A. Herbst, D.W. Fuerstenau, Scale-up procedure for continuous grinding mill design

using population balance models. International Journal of Mineral Processing 7 (1980) 1-

31.

[12] J.A. Herbst, Y.C. Lo, B. Flintoff, Size reduction and liberation, in (Eds. M.C. Fuerstenau,

K.N. Han) Principles of Mineral Processing, SME, Littleton, Colorado, 2003

[13] J.A. Herbst, D.W. Fuerstenau, The zero order production of fine sizes in comminution and

its implications in simulation, Transactions AIME 241 (1968) 538-546.

[14] V.K. Gupta, Zero order production of fines in ball and rod mill grinding: an explanation.

SME-AIME Annual Meeting, New York, N.Y., February 24-28, 1985. Preprint

Number 85-139.

[15] T.S. Mika, L.M. Berlioz, D. W. Fuerstenau, An approach to the kinetics of dry batch ball

milling. 2nd European Symposium on Comminution, Amsterdam, 1966. Dechema-

Monographein 57, Nr. 993-1026, Verlag Chemie, Weinheim, 1967, pp. 205-240.

[16] L.M. Berlioz, Kinetics of dry grinding, M.S. Thesis, University of California, Berkeley,

Calif., USA, 1966.

[17] F.M. Smaila, An investigation of the kinetics of dry batch ball milling. M.S. Thesis, The

Pennsylvania State University, Pennsylvania, USA, 1982.

[18] V. Deniz, The effects of ball filling and ball diameter on kinetic breakage parameters of

barite powder, Advanced Powder Technology 23 (2012) 640-646.

[19] V.K. Gupta, An appraisal of the linear first-order kinetic model based ball mill design

correlations, 1.World Congress Particle Technology, Part II: Comminution (6. European

Symposium Comminution), Nurnberg, April 16-18, 1986, pp. 605-620.

[20] V.K. Gupta, H. Zouit, D. Hodouin, Determination of the breakage rate and distribution
 parameters: Preparation of the test samples and design of the grinding experiments. Report

No. 82-19, GRAAIM, Laval University, Quebec, Canada, 1982.

[21] V.K. Gupta, D. Hodouin, R. Spring, SPOC Manual, Supplement 7.2: Kinetic ball mill

model parameter estimation, Division Report MRP/MSL 83-88(IR), CANMET, Energy

Mines and Resources Canada, 1983.

[22] V. K.Gupta, H.Zouit, D.Hodouin, The effect of ball and mill diameters on grinding

rate parameters in dry grinding operation. Powder Technology 42 (1985) 199-208.

[23] V.K. Gupta, Mill speed control for more efficient grinding. Proceedings 17th Canadian

Mineral Processors Annual Conference, Ottawa, January 22-24, 1985, pp. 639-646.

[24] M.H. Moys, A model of mill power as affected by mill speed, load volume, and liner

design. Journal of the South African Institute of Mining and Metallurgy 6 (1993), 135-141.

[25] K. Shoji, L.G. Austin, F. Smaila, K. Brame, P.T. Luckie , Further studies of ball and

powder filling effects in ball milling. Powder Technology 31 (1982) 121-126.

[26] K.S. Venkataraman, D.W. Fuerstenau, Application of the population balance model to the

grinding of mixtures of minerals. Powder Technology 39 (1984) 133-142.

[27] P.C. Kapur, D.W. Fuerstenau, Energy split in multicomponent grinding. International

Journal of Mineral Processing 24 (1988) 125-142.

[28] N. Khumalo, D. Glasser, D. Hildebrandt, B. Hausberger, An experimental validation of a

specific energy-based approach for Comminution. Chemical Engineering Science 62 (2007)

2765-2776.

[29] D. Touil, S. Belaadi, C. Frances, The specific selection function effect on clinker grinding

efficiency in a dry batch ball mill. International Journal of Mineral Processing 87 (2008)

141-145.
[30] B. K. Mishra, R.K. Rajamani, The discrete element method for the simulation of ball mills.

Applied Mathematical Modelling 16 (1992) 598-604.

[31] B.K. Mishra, R.K. Rajamani, Simulation of charge motion in ball mills. Part 1:

experimental verifications. International Journal of Mineral Processing 40 (1994) 171-186.

[32] B.K. Mishra, R.K. Rajamani, Simulation of charge motion in ball mills. Part 2:

Numerical simulations. International Journal of Mineral Processing 40 (1994) 187-197.

[33] A. Datta, R.K. Rajamani, A direct approach of modelling batch grinding in

ball mills using population balance principles and impact energy distribution. International

Journal of Mineral Processing 64 (2002) 181-200.

[34] R.M. de Carvalho, L.M. Tavares, Predicting the effect of operating and design variables on

breakage rates using the mechanistic ball mill model. Minerals Engineering 43-44 (2013)

91-101.