Vous êtes sur la page 1sur 15

Minerals Engineering, Vol. 7, Nos 2/3, pp. 185-199, 1994 0892--6875/94 $6.00+0.

00
Printed in Great Britain 1993 Pergamon Pre~ Ltd

BREAKAGE PROPERTIES OF ORE BLENDS

D. YAN and R. EATON ~

WA School of Mines, Kalgoorlie, Western Australia


Dominion Mining Limited, Mt. Morgans, Western Australia
(Received 5 July 1993; accepted 16 August 1993)

ABSTRACT

The blending of different ore types is common practice to provide a consistent feed to a
process either in terms of uniform assay or rock hardness. When several different
deposits of varying grindabilities are blended prior to grinding, the hardness of the ore
blend is usually estimated as the average of the component grindabilities. It is often
suspected that this is not the real case. b~vestigations were therefore carried out to
determine the variation of the Bond Work Index (W~ of an ore blend as a function of
bletut composition.

Bomt Work Index measurements were carried out on a sample of 100% pure hard rock
(Wi = 14) and 100% soft rock ( ~ = 6) and three blends of the two components. The
work imtices were measured by the Bomt method and the Magdalinovic method. Both
methods gave comparable results for the pure components but differed markedly for the
blends. The work index, as measured by the Bond method, did not give a linear
relationship with blend composition but was weighted towards the harder component.
This wouM indicate that it is mainly the harder component which is comprising the
circulating load. In the Magdalinovic method, no recycle streams are employed and the
resultant work index measurement more closely follows a linear relationship with blend
composition.

As an aid to the understanding of the grinding mechanisms and interactions taking place
in the mill, analysis of the breakage properties (breakage distribution function and
breakage rate) of each pure component and the blends were also carried out. Qualitative
analysis of the breakage properties suggests there is an interaction between the
components of the blend which affects their it~lividual breakage rates. From these results,
breakage properties of the harder material appears to have a greater influence on the
overall breakage properties and the Bond Work Index of the blemt than the softer
material.

Attempts were made to use the measured breakage parameters and the batch grituting
equation to simulate the Bond Work Index test and hence calculate the Wi.

Keywords
Breakage, Blend, Bond Work Index
ME 7-2/3---~ 185
186 D. YAN and R. EATON

INTRODUCTION

The aim of this project is to investigate the grinding of ore blends, specifically their Bond Ball Mill Work
Index, and how it varies with differing blend compositions.

As an aid to understanding the grinding mechanisms and interactions taking place in the mill, analysis of
the breakage properties of each blend was also carried out.

Work concentrated on two gold ores from the Meekatharra district in Western Australia. The ores were
chosen because their grinding characteristics differed significantly. Miekey Doolan ore is a relatively
hard, competent, quartz-fuchsite-carbonate material, while Vivians ore is a soft, highly weathered
material. These and other ores from the mine's open pits are blended together prior to milling.

PREVIOUS W O R K ON BLENDS

Holmes and Patching [1], carried out grindability tests on an artificial mixture of quartz and limestone.
Each component was crushed to minus 3.35 mm and combined in a range of quartz to limestone ratios.
The Maxson, Cadena and Bond method was used to analyse the grindability of the mixtures. It is
essentially similar to the standard Bond method, and uses the same laboratory mill and conditions.

Grindability tests were carried out on the pure components as well as the mixtures. The ratio of each
component in each size fraction of the feed and product size distributions was determined by chemical
analysis for carbon dioxide.

It was assumed the size reduction was first order with respect to the volume of each component in the
mill. The first order rate constant, or the specific rate of reduction, k, was calculated for each component
from the feed and product data. The grinding performance was compared in terms of the ratios kL/kQ,
kL'/kQ', k L / k L' and kQ/kQ', where:

k,. kQ ,kL',kQ' specific reduction rate for the pure limestone, the pure quartz, the
limestone component and the quartz component of the mixture
respectively.

The specific rate of reduction of each component was observed to be lower when ground in a mixture.
There was no significant change in the product size distributions of the components in the sieve sizes as
their proportions in the feed changed.

They also concluded that the proportion of each component in the mixture had no effect on the rates of
breakage. The effects of enforced changes in the feed, such as the mixture composition, is buffered by
changes in the circulating load composition.

Somasundaran and Fuerstenau [2], carried out investigations of the grinding of mixtures, also using a
mixture of quartz and limestone. With the aim of determining the extent to which grinding energy is
consumed by one component more than another, they ground 1 kg samples of 1:1 mixtures of 4.75 x 2.36
ram material, and 1:1 mixtures of 4.75 x 2.36 mm quartz with 300 x 150 I.tm limestone, and 4.75 x 2.36
mm limestone with 300 x 150 I.tm quartz.

Batch tests were run in laboratory ball and rod mills and each size fraction of the product was analysed
to determine the relative proportions of each component. From this information, the energy consumed
in the reduction of each component was calculated. For the equal proportioned 4.75 x 2.36 mm material
only a very slight difference in the energy consumed by each component was observed. However, in the
milling of the different size fractions, the coarser component consumed a greater proportion of the
grinding energy. This phenomenon was more pronounced in the case of rod milling, and the harder
Breakage properties of ore blends 187

quartz consumed the greater proportion of energy during the longer batch tests. This is the result of the
harder, coarse component protecting the fines from impact by the grinding media, by virtue of their larger
size. While being a commonly recognised effect in rod mills, Somasundaran and Fuerstenau consider
this effect to be relatively small in ball milling, and the fraction of energy consumed by any material is
in proportion to the volume fraction of the material in the ball mill.

Rose and Matsumura [3], assumed that the size distribution of the material obtained by grinding a single
material under a specified set of conditions and for a specified time could be described by a matrix
equation relating the product size matrix to the feed size distribution matrix and a grinding matrix which
defines the grinding of the material in a specified size interval.

Knowing the grinding matrices of the pure components, Rose and Matsumura investigated the
predictability of the product size distribution for the grinding of the mixture. They derived an equation
which calculated the effective grind time required for the component in the mixture to produce the same
product size distribution as the pure component.

Fuerstenau and Venkataraman [4], approached the comminution of mixtures by applying the basic batch
grinding equation of the size-mass balance model.

The batch grinding experiments were carried out dry in a laboratory mill 25.4 cm in diameter and 27.9
cm in length. Initially, 1700 x 1180 I~m calcite and quartz feed material was ground separately for
varying grind times under standard mill operating conditions.

For the grinding of calcite-quartz mixtures, the two minerals were mixed in the ratio of 1:1 by volume
and ground for the desired time period under the same operating conditions. The size distribution of the
ground mixture and each component in the mixture was determined.

The results indicated that when minerals are broken in ball mills in the presence of other minerals, the
breakage rate functions are time-independent. For the grind times used in these experiments, the breakage
rate function of the softer mineral (calcite) increased when ground in the presence of the harder mineral
(quartz), compared to the pure component. However the breakage rate function of the quartz was reduced
when ground in the presence of the calcite. They observed that the breakage distribution function, Bil,
was normalisable in the case of mixture grinding, regardless of mill size, mill speed and ball charge, as
they are for the pure components.

Locked-cycle tests showed that in the grinding of mixtures, a relatively long time is required to attain
steady state, during which time the composition of the mill contents continuously change. During this
period, the breakage rate functions must also change continuously. The composition of the total charge
inside the mill at steady state is controlled by the relative grindabilities of the minerals in the mixture, that
is by the value of the breakage rate function of each mineral in the mixture. From this, they concluded
that in the grinding of multi-component feeds, the breakage rate function of the individual components
will vary with changes in the composition of the mill holdup, depending on the relative
hardness/grindability of the individual mineral components.

They also make the interesting point that, if each of the components of a mixture exhibit first order
grinding kinetics, the disappearance kinetics of the mixture as a whole must exhibit some deviation from
first order kinetics. This suggests the possibility of abnormal breakage conditions.

Thus from the limited work that has been done on the breakage of mixtures, there seems to be
contradictions as to the effects of composition on the breakage characteristics of the blends.
188 D. YAs and R. EATON

ASSESSING THE BREAKAGE CHARACTERISTICS OF ORE BLENDS

Bond m e t h o d

The mill is operated in locked cycle, grinding the charge for a number of revolutions, sieving on the
desired screen to remove undersize and then replacing the undersize with new feed. The grind time for
the new mixed feed is based on a calculated number of revolutions required to produce a circulating load
of 350%.

The procedure is repeated until a constant mass ratio of 2.5 for oversize to undersize is achieved for 3
consecutive cycles. Typically this may take 7 to 10 cycles. Size analysis of the sieve undersize to find
the 80% passing size of the product, allows the calculation of the Work Index, Wi,test using the empirical
equation:
44.5
Wl,~. t = 1 . 1 0 (1)
pc'z~.1,..I
~'~ blp
10.[1/P~ - 1/F~ ]

where Pc = test-sieve size in #m;


Gbp = mass of the test-sieve fresh under-size per mill revolution (g rain-l), and;
P80 = 80 % passing size of circuit product.
F80 = 80 % passing size of new feed.

The Magdalinovic test

The majority of attempts to simplify the standard Bond test have been aimed at shortening the procedure.
With such a long, complex procedure, the opportunity for systematic, or procedural error is considerable.
Also, the time required to complete a full Bond Work Index test can be anywhere upwards of 8 hours,
making it of little use in the optimisation or control of an operating industrial circuit.

The abbreviated procedure put forward by Magdalinovic [5] was developed from work by Smith and Lee
[6], comparing the Bond test results with batch grinding results, which showed there is an empirical
relationship between the locked cycle Bond test and batch grinding tests.

The Magdalinovic procedure reduces the standard Bond test to just two grinding cycle.s. The first grind
cycle is used to estimate the grinding rate to adjust the number of revolutions in the second cycle to obtain
the equivalent of 350 % recirculating load. It uses the standard Bond test mill and mill conditions.

Size-Mass balance model

In their favour, the Bond-type tests are comparatively simple. Their empirical development, based on a
large range of plant data, means that they do work well in many circumstances.

Circuit analysis in the Bond method is based on the 80 9; passing size of the whole circuit, hence it bears
no direct physical or theoretical correlation with what is actually occurring inside the mill. It neglects
the effect of the size distributions' shape on the comminution process, and hence does not consider many
second-order effects.

The Size-Mass balance model addresses the effect of the feed size distribution, and several other factors
not dealt with by the Bond Model.

The basic batch grinding equation for expressing the mass balance on material in size class i during the
course of comminution is given below [7]:

dwi(t)/dt = -Si.wi(t) + L"bij.Sj.wj(t) (2)


Breakage properties of ore blends 189

where wi(t) is the mass fraction of material in the ith size class at time t, S i is the breakage rate function
for material in the ith size class (which gives the fractional rate at which material is broken from the size
class at time t) and b~j is the breakage distribution function (which gives the fraction of material that
becomes particles in the ith size class when material in the jth size class is broken). The parameters Si
and Bij then describe the particle breakage and allow simulation of the milling process.

EXPERIMENTAL PROCEDURES

The Bond Work Index determination was carried out in a standard test mill (30 cm diameter), under
standard conditions [8].

The breakage parameters were determined experimentally using the "One-Size Fraction" technique,
outlined by Austin and Bhatia [9]. Tests were carried out on the two ores and one blend. The size
fractions chosen for the tests were:

-2360 + 1700 i.tm


-1700 + 1180 ~n,
-850 + 6001.tm
-300 + 212tim
-212 + 150 I~m

Since only a small size interval is used in these tests, it was no longer relevant to define the blend
composition in terms of volume, (since the bulk density of the material has changed with the size
distribution), hence mass fraction composition was adopted.

A 20 cm diameter test mill and 40% ball charge were used, at 60 rpm, wet grinding for all the tests.

The charge was ground for periods from 0.5 to 15 minutes and the mill emptied. The charge was sized
by wet and dry sieving. After weighing, the sample was re.combined and returned to the mill and ground
for a further period.

RESULTS AND DISCUSSION

Comparison of the Bond and Magdalinovic Work Indices for pure and blended materials

A Bond and Magdalinovic test was completed on each of the pure ore types, and on a 50:50 (vol.) blend
of Mickey Doolan and Vivians. The results are shown in Table 1.

TABLE 1 Comparison of Work Indices from Bond and Magdalinovic tests

Vivians Vivimm Test Sieve Work Index (kWh/t) Difference

vol./vol. wt.~vt. S~ (~m) Bond Test Magdalinovi (%)

1.00 1.00 106 6.6 5.5 +16.7

0.50 0.42 106 12.3 8.4 +31.7

0.00 0.00 106 14.0 14.8 - 5.0

It is evident from the table that the Magdalinovic test has slightly over-estimated the Work Index for the
hard Mickey Doolan material, while it slightly under-estimated the Index for the softer Vivians ore.
Although the real error for both tests are of similar magnitude, the percentage error of the Vivians ore
is considerable due to its low Work Index.
190 D. YAN and R. EATON

The percentage error in the Work Index for the Mickey Doolan ore is in line with the errors quoted by
Magdalinovic (+7 %) for materials of a similar Bond Work Index or higher.

The Magdalinovic test Work Index for the blend was considerably less than that determined by the full
Bond test. The large variation for the blend can be explained by the simulated locked--cyclenature of the
full Bond test. In the locked-cycle grinding of soft and hard materials simultaneously, one would expect
the harder component to predominate in the oversize and the recireulating load, and eventually in the mill
hold-up. By comparison, the Magdalinovie test is effectively open circuit grinding.

Variation of Bond Work Index with blend composition

In all, five full Bond tests were carried out, one on each of the pure ores and three blends. The results
from the tests are given in Table 2.

TABLE 2 Variation of Bond Work Index with blend composition.

ViviAn.~ Vivians Test sieve F80 P80 ~3rams per Bond Work

vol./vol. wt./wt. Otm) 0tm) (tun) revolution Index (kWh/t)

1.00 1.00 106 2100 73 3.31 6.6

0.75 0.69 106 2440 82 1.90 11.0

0.50 0.42 106 2621 80 1.60 12.3

0.25 0.20 106 2669 82 1.46 13.5 .q

0.00 0.00 106 2762 85 1.42 14.0

The trend of the Wi with blend composition is shown in Figure 1 in terms of volume fraction of Vivians.
16

14

X 12

: 10
{3

"0 6 Bond Method


0 Magdalinovic
rn 4
Method

0 0.2 0.4 0.6 0.8

Volume fraction of Vivians in blend


Fig. 1 Variation of the Bond Work Index with blend composition
Breakage properties of ore blends 191

The results show that the harder Mickey Doolan ore has the greatest influence on the value of the Bond
Work Index of the blend. This effect can be explained by considering the locked-cycle nature of the full
Bond test. With the locked-cycle grinding of a blend made up of hard and soft components, one would
expect the harder component, (in this case the Mickey Doolan ore), to grind at a slower rate and hence
comprise the greatest proportion of the test-sieve-oversize material. Consequently, the harder material
builds up in the recirculating load and hence the mill hold-up. This effect was evident during the testa,
as the colour of the recirculating coarse material became lighter with the increasing Mickey Doolan
content. Mickey Doolan material is a pale green-grey colour while Vivians is a dark red-brown colour.

Indeed, one could not expect the grams per revolution for a blend to become constant until the harder
material had built up to a constant level in the mill hold-up. Thus, it could be said that the Bond test
determines the Work Index only for the specific situation of locked--cycle grinding of the blend, since the
composition of the mill hold-up material is not representative of the composition of the starting blend.
This opens up the possibility that the Magdalinovic test may, in fact give a truer indication of the energy
required to grind a particular blend composition.

Calculation of the Breakage Distribution Parameters

The Cumulative Breakage Distribution, Bij, defined by Austin and Luckie [I0] as "The weight fraction
of material broken from size j which falls less than the upper size of size interval i." is commonly used
to characterise the size distribution resulting from breakage of material from a particular size interval to
a smaller size.

The Bij values were calculated using the BII method described by Austin and Luckie [10]. The values
of B i.~J for .some
.
size
.
fractions
.
and each material, are plotted in Figures 2 and 3, versus the dimensionless
relahve size interval, to determine if the values of Bij are normalisable. The Bij values are said to be
normalisable if the fraction which appears at sizes less than the starting size is independent of the starting
size. In terms of the plots, the curves should super-imposed upon each other if the Bij values are
normalisable.

iT/

r ~ j
m 0.1 v

-2360+1700
m

-1700+1180

' -850+600 um

-300+212 um

-21~2+1S0 u m

O.Ol I I I I I I
0.1 I

Relative Size x(i-1)Ixj


Fig.2 Variation of Bij with size for Hard Ore (Mickey Doolan)

From Figures 2 and 3, it can be seen that the pure Mickey Doolan and Vivians ore could be assumed to
have normalised Bij values. Plots of this nature allow the following general equation to be fitted:
192 D. Y A N and R. EATON

Bit = .(xi.t/xj)Y + (1 - O).(Xi.l/Xj~8 (3)

The parameter is the intercept at (xi_]/xj) ffi 1, y is the slope of the lower section of the Bil curve and
13 is the slope of the steeper section of the Bij curve. Determination of these parameters allows direct
comparisons to be drawn between different materials and blends.

The values of , y and 13 were evaluated by curve-fitting the measured data to eq. (3). Typically, y
values should be in the range of 0.5 to 1.5, while 13 values should be between 2.5 and 5.0 [11].

4....- I

im,a t
mlm

m -2360+1700

o -1700+1180

-850+600 um

o -300+212 um
*/

-212+150 um
0.1
0.1

Relative Size x(i-1)/xj


Fig.3 Variation of Bij with size for SoR Ore (Vivians)

Figures 4 and 5 show the variation of the Bij parameters with the top size of the size interval, for each
pure material. 13, and y appear on the plot to be approximately linear.

y tends to decrease and 13 increase with increasing particle size for the Mickey Doolan sample. These
changes in y and 13 indicate a change in the breakage mechanism with increasing particle size, from
shattering to more chipping and abrasion [12]. This reflects the increasing hardness of the large Mickey
Doolan particles, such that they resist crushing impacts from the grinding media. Such a change in
breakage mechanism suggest abnormal breakage of the larger particles.

For the Vivians sample, the tend in the breakage distribution parameters y and 13 is less pronounced and
could be taken as either invariant with particle size or slightly contrary to that of the Mickey Doolan ore.
The softer Vivians ore still undergoes the shatter type breakage even for the larger particle sizes.

At the present time, the breakage parameters of only one size fraction (-1700+ 1180 pm) of one blend
(50:50 by mass) of the two ore types has been carried out. The values of the breakage distribution
parameters for this size interval is compared with the pure components in Figure 6. The value of does
not vary greatly with the composition of the blend whereas both y and 13decrease with increasing % of
the softer component. The decrease in parameter value with composition is not linear. Work is
continuing on other size intervals and ore blend compositions to quantify the variation in breakage
properties with component composition.
Breakage properties of ore blends 193

10
,JII

7" " ---r--


0

IIII

0.1
I I
10 100 1000 10000
Size Xi (microns)
Fig.4 Breakage Distribution Function Parameters , y, 13 with particle size
(Mickey Doolan Ore)

10 i

4, I! !
-~---- ~_ ~-- ~ . . . . . . .

I
I
m

0.1
10 100 1000 10000

Size Xi (microns)
Fig.5 Breakage Distribution Function Parameters ~ , y, 13 with particle size
(Vivians Ore)

Calculation of the Breakage Rates

The specific rates of breakage, S 1, were calculated using the data from the breakage tests.

In the analysis of the breakage of materials, it is useful to make the initial assumption that the breakage
of each size fraction is first order in nature. That is, the rate of disappearance of size 1 due to breakage,
is proportional to the amount of size 1 material in the mill holdup.
194 D. YAN and R. EATON

-d[wl(t).W]
*, wj(t).W (4)
dt

Since the mill holdup, W, is constant, eq. (4) becomes:

dw l(t)/dt = -SlWl(t) (5)


If S 1 does not vary with time:

Wl(t) = Wl(0).exp(-Sl.t ) (6)

3.5

2.5

~ 2 : <b

e 1.5 [] y

13

0.5

0
0 20 40 60 80 100

% Vivians
Fig.6 Breakage Distribution Function Parameters , y, 13 with Blend
Composition for -1700+ 1180 ~m Size Interval

If breakage is first-order, a plot of log {[wl(t)]/[wl(0)] } versus time,t, should produce a straight line with
slope, -Sl/2.303. This allows the value of S l to be determined. Figures 7 and 8 plot wt(t ) as a function
of grind time on log-linear scales. The values of S 1 were determined by linear regression of log
{[wl(t)]/[Wl(0)] } with respect to time t. Figure 9 shows the variation of S 1 with the top sieve size of each
size fraction. These plots indicate there is abnormal breakage of the larger sizes. Typically, plots of this
type show a maximum, where the breakage rates of the larger particles decrease with increasing abnormal
breakage effects, and the breakage rates of the smaller particle sizes decrease in accordance with the first-
order breakage assumption. The maximum occurs because the particles become large enough to
withstand the impacts of the tumbling charge. Interestingly, for both the hard and soft components, the
maximum particle size is approximately the same.

Figure 10 shows the plot of log {[Wl(t)]/[wl(0)] } versus time for the 50:50 (by mass) mix of Mickey
Doolan and Vivians ore. The plot is nolonger a single straight line but becomes steeper with increasing
grind time. There is thus some variation in the breakage rate with time, and hence deviation from first-
order breakage in at least one of the size fractions. Deviation from first-order grinding kinetics is termed
"abnormal" breakage.
Breakage properties of ore blends 195

A
\
O -2360+1700 um
IE
A
0.1 ~ -1700+1180 um
V
N\ ~.\
IE " -300+212 um

\ o -212+150 um
\
\
0.01
0 5 10 15

Grind Time (minutes)


Fig.7 Mass Fraction Remaining in the Top Size versus Grind Time for Mickey Doolan Ore

1.00

A
O
=E
A 0.10 o -2360+1700 um

=E -1700+1180 um

-300+212 um

0.01
0 2 4 6 8 10

Grind Time (minutes)


Fig.8 Mass Fraction Remaining in the Top Size versus Grind Time for Vivians Ore

One would expect a slight variation in the breakage rates with time for the blends. Two types of
abnormal breakage are possible. The breakage rate can either decrease or increase with time. For the
first case, in the early stages of the batch grind, breakage rates would be expected to be greater with the
grinding of the softer component. This grinding rate would slow with time as the harder material begins
to predominate in the top size interval.

Alternatively, the grinding rate may increase with time as the harder component, which was shielding the
softer component, is removed by breakage. If the harder component is shielding the softer component
in the early stages of breakage then the initial breakage rate could be expected to be similar to that of the
harder component while the later rate of breakage would be some mean value for the two components.
If the curve in Figure 10 is broken into two straight line sections indicating the initial rate of breakage
196 D. Y A N and R. EATON

and the rate of breakage at extended grind times then the breakage rates obtained from the slopes are
0.308 and 0.623 rain-] for the initial rate and the extended rate respectively. The initial rate is close to
the breakage rate for the same size fraction of Mickey Doolan, 0.286 min"t, while the extended rate is
considerably higher than the breakage rate for Vivians, 0.468 rain"l. This higher rate, at least, was also
observed by Fuerstenau and Venkataraman [4] in the grinding of their quartz/calcite mixtures.

/
/
. m
/
/ /
/
I"--i
/
t~
S /
/
0.1 ,/
t~
v l/
(g

I11
' Mickey Doolan "

c--- Vivians

0.01
100 1000 10000

Top Size, Xi (microns)


Fig.9 Specific Rates of Breakage versus Size for Mickey Doolan and Vivians Ore

1.00

A
0
5
A
0.10

5
data point

- - Initial rate

...... e x t e n d e d rate

0.01
0 1 2 3 4 5 6

Grind Time (minutes)


Fig. I0 Mass Fraction Remaining in the Top Size versus Grind Time for a
50:50 Blend of Hard and Soft Ore (-1700+ 1180 ~m size interval)
Breakage properties of ore blends 197

It is expected then that some evidence of abnormal breakage in the grinding of the blends will occur due
to the difference in the breakage rates of the individual components. The question becomes one of h o w
the breakage parameters change with blends of two or more different components and how might the
change be predicted.

Simulation of the Bond w o r k Index test

If the breakage rate and breakage distribution function together allow simulation of batch grinding then
it should be possible to simulate the Bond Work Index test if the Si and Bij values are known.

A BASIC program was written to evaluate the Reid Solution of the batch/plug-flow grinding equation [13].

Values for the specific rate of breakage, SI, for each size fraction were extrapolated from the plots of S 1
as a function of interval top size, Figure 9.

A complete set of Bij values were calculated from the values of ~, y and 13 obtained from Figures 4 and
5. The calculated Bij values were close and for ease of calculation were considered to be normalised.

Since a complete analysis of the 50:50 blend of Mickey Doolan and Vivians has not yet been completed,
an estimate was made of the Si and Bij values based on the comparison between the breakage
characteristics of the -1700 + 1180 I.tm interval of the pure components and the blend. Since the Bond test
cycles are usually less than 3 minutes, the initial breakage rate of the blend is valid and since this rate is
very close to the rate for pure Mickey Doolan they were taken to be the same. Similarly, the breakage
distribution parameters, , , and 13 are increasing with increasing % Miekey Doolan in the blend but
because of the shape of the non-linear relationship, the parameters for the blend are still very close to the
pure Vivians values. Therefore the breakage distribution function for the 50:50 blend was taken to be
the same as the pure Vivians component.

Now the mill conditions used to measure the breakage parameters was different from the standard Bond
test mill conditions. As a result the values of the breakage rates will be different. The effect of mill
diameter on the breakage distribution function is small and hence the Bij values can be considered
unchanged [14]. However the breakage rates will need to be scaled up to the larger mill, and changed
milling conditions. The relationship between mill diameter and Si is given by ; [11]

Si ,c D0.5 (7)

This gives a scale factor of 1.22. This scaling of the Si values does not take into account the difference
in ball charge, wet grinding and different mill speed. To obtain a reasonable scaleup factor, a simulation
was performed to backcalculate the scale factor for the first grind cycle of the Bond test on the Mickey
Doolan ore. This produced a scale factor of 4.36. This scaleup factor was then used in the remaining
cycle calculations for Mickey Doolan and for the scaleup in all subsequent Bond Work Index simulations
on the other sample.

The results of the simulated Bond Work Index for the Mickey Doolan and Vivians material is given in
Table 3.

TABLE 3 Experimental and computer simulated Bond Work Indices

Material
E ~ Simulated

MiekeyDoolan 14.0 13.1

50:50 blend 12.0 11.1


Vivians 6.6 5.6
198 D. YAN and R. EATON

The accuracy of these simulations suggests it maybe possible to predict the Bond Work Index with more
experimentally-derived input.

CONCLUSIONS

Qualitative analysis of the breakage properties seems to suggest there is an interaction between the
components of the blend which affects their individual breakage rates. In this particular case, breakage
properties of the harder material appears to have a greater influence on the overall breakage properties
and Bond Work Index of the mixture.

In a more detailed analysis of blends, the ability to determine the composition of each size fraction of the
product would greatly improve understanding of how each material may affect the grinding rate of the
other. Chemical analysis for umque trace dements in the fractions is being investigated and would have
the following advantages:

(i) analysis of the composition of the recirculating load in the Bond test.

(ii) the breakage rates of each component in the blend could be compared with those of the blend
overall, and with those of the components when ground separately.

The Bond Work Index of a mixture of ore types is not simply the weighted mean of the Work Indices of
the pure components. Where ores of widely varying grindabilities are blended prior to milling the power
requirements and final grind cannot be known with great accuracy. With the determination of the
breakage parameters of the pure components and how these change with ore blend composition, it should
be possible to predict the Bond Work Index of a blend of any composition.

ACKNOWLEDGEMENTS

The authors would like to thank Dominion Mining Limited, Paddy's Flat Operations, for their assistance
in the experimental programme. In particular, the assistance of Mr. Harry Helle, Chief Area Metallurgist
and Mr. Gary Johnson, Group Technologist of Dominion Mining Limited and Mr Ben Grummels, Senior
Technician of the WA School of Mines is greatfully acknowledged.

REFERENCES

1. Holmes, J.A. & Patching, S.W.F., A preliminary investigation of differential grinding - grinding
of quartz-limestone mixtures. Transactions of the Institute of Chemical Engineers, 35, 111-124
(1957).
. Somasundaran, P. & Fuerstenau, D.W., Preferential energy consumption in tumbling mills.
Transactions of the American Institute of Mining, Metallurgy and Petroleum Engineers. 226, 132-
134 (1963).
. Rose, H.E. & Matsumura, S., Simultaneous grinding of two materials in a batch ball-mill.
Transactions of the Institution of Mining and Metallurgy, 94, C40-C46 (1985).
4. Fuerstenau, D.W. & Venkataraman, K.S., The comminution of multi-component feeds under
batch and locked-cycle conditions: Kinetics, simulation and energy distribution. International
Journal of Mineral Processing, 22, 105-118 (1988).
. Magdalinovic, N., A procedure for rapid determination of the Bond Work Index. International
Journal of Mineral Processing, 27, 125-132 (1989).
6. Smith, R.W., & Lee, K.H., A comparison of data from Bond type simulated closed-circuit and
batch-type grindability tests. Transactions of the American Institute of Mining, Metallurgy and
Petroleum Engineers, 241, 91-99 (1968).
. Austin, L.G., A review introduction to the mathematical description of grinding as a rate process.
Powder Technology, 5, 1-17 (1971/72).
Breakage properties of ore blends 199

. Bond, F.C., Crushing and Grinding Calculations, British Chemical Engineering, 6, 378-391,
543-548 (1960).
9. Austin, L.G. & Bhatia, V.K., Experimental methods for grinding studies in laboratory mills.
Powder Technology, 5, 261-266 (1971172).
10. Austin, L.G. & Luckie, P.T., The estimation of non-normalised breakage distribution parameters
from batch grinding tests. Powder Technology, 5, 267-271 (1971172).
11. Austin, L.G., Klimpel, R.R. & Luckie, P.T., Process Engineering of Size Reduction: Ball
Milling. American Institute of Mining, Metallurgy and Petroleum Engineers, pp 1-291, New
York, (1984).
12. Kelly, E.G. & Spottiswood, D.J., The breakage function: What is it really? Minerals
Engineering. 3, 405-414 (1990).
13. Reid, K.J., A solution to the batch grinding equation. Minerals Science and Engineering, 20, 953-
963 (1965).
14. Austin, L.G., Understanding ballmill sizing. Ind. Eng. Chem. Proc. Des. Develop., 12, 121-129
(1973).

Vous aimerez peut-être aussi