Академический Документы
Профессиональный Документы
Культура Документы
14 (1976)
Technology,
@ Elsevier Sequoia
The Effect
Sk.
71 - 79
Lausanne - Printed in the Netherlands
71
L G. AUSTIN
Deporlment
Pa. 16802
of
Materials
Section,
7%~ Pennsylcnnia
State
Uniuersity.
University
Park.
(U.S.A.)
K. SHOJI
hisso Engineering
Compnrsy
Ltd..
Department
of Technical
Deuelopment.
l-4-3,
Einji-rho.
Chiyodn-1:~.
Toi:yo
(Japan)
P. T. LUCKIE
Kennedy
(Received
Van Saun
September
Corporation,
Danviile.
Pa. I7821
(bi_S_A.)
10. 1975)
SUMMARY
INTRODUCTION
(1)
ABNORMAL
BREAKAGE
of size i = SiWi
(2)
GRINDING
Fig. 1. Experimental
8 x 10 mesh cement
3
TIME.
MINUTES
dw (t)
-=
dt
dwl.a(t)
dt
+(1-o)
dwldt)
dt
where
dw,(t)
dt
dwdf)
dt
= -%=Lwl_A(t)
= -%w1l3
(f)
QX~eXp[-(WXf)"]
(4)
PARTICLE
SIEVE
SIZE.mm
EFFECT
OF BALL
DIAMETER
(5)
(6)
f
z
e*E*r
d(E/fl)
'73
_..
0.5
0 01
of esp(-z*/!?)Q(z)
VS. z -
2.5.
09
r
= s,foIp
Fig. 3. Log-probability
plot
different
ball diameters_
of y us. particle
size for
= Q
log xi log
log p
(3
OF BALL
d. mm
for
the
log u = exp(-z%/2)lQ(X,)
where z, = log(x,/p)/log
a; and x, is the
particle size for maximum S for the given baii
diameter. Figure 4 shows a plot of the function exp(-9/2)/Q(z)
WEUS z + 2.5, which
gives over the range of interest:
=na= 1.75(2.3ofi
log u)~~* -
2.5
or
log(X,/P)
DIAMETER
= log u [4.352(a
log a)452 -
2.51
(7)
PARTICLE
Fig. 6. Comparison
with experimental
(a
intervals)_
SIEVE
SIZE.
of the predictions
of eqn. (10)
resmelts, for data of Kelsall et al.
Knowing a and a, then, x,/cc can be determined. For the data of KeisalI et al_, x,/p = l-15_
Figure 5 shows the variation of x, with ball
7-I
TABLE
Material
Goid
(h)(lO
in
min-
Mill
diameter (mm)
k;,(lo3)
0.65
20
206
kz
ore
harder
mean
softer
-50
0.86
2.75
0.95
3.56
1.23
1.79
1.99
2.36.
1.69
1.91
0.59
0.68
1.66
1.73
0.86
157
195
l-25
O-40
1x-l
1.33
570
303
2.14
O-75
ND.
0.98
104
196
Cement clinker II
harder
meifn
Qi.KWtr
(Kelsall EL nf. )
mf=an
Cement clinker
mean
*It1
= 0.2
s,
(8)
(9)
P, a)
2-51)
WI 9
or
k, = k,,(D/I)
W)
(12)
where
D = k, d2/antilog(log
d2/antilog{log
oC4.352
(a log u)Q=2.51)
(13)
where nl = 0.1 to 0.2; further experimental
work will be needed to decide on the correct
value for n1 _
Table 1 summarizes the data values for k,,
kIO and u, determined by leas~squares regression analysis.
75
0.1
00
0
a
0.2
0.3
f,.FRACTfON
MILL VOLUME
Fig_ 7. Variation
powder filling.
of absolute
MIXTURE
OF BALL
FILLED
SIZES
(14)
dj
of balls of
EFFECT
OF MILL
rate of breakage
from
bi.j.kSi.k
OTHER
F.z 0~ Do-6
(16)
k2 = (f&T)*
AND
ON k,
(l/0.159)
[2( 0_159/f,)Q
l] )
(17)
where f, is the fractional volume filling by
powder and (f,S)*
is the maximum value of
Sf,, in the almost horizontal region of the
plot. Assuming this relation can be used for
other values of J,
2 Si.kmk
k=l
DLXMETER
k=l
bi,j =
of k, this reduces
s = (f.S)X
bi.i.&i.kmk
Hence,
BY BALL BED
Si(d)dm(d)
06
to
CONDITIONS
&xix,
05
0.4
Assuming independent
be_havior of the
balls, the mean value of S,, Si say, is
j&Z
04
BY POWDER
03
wnSEEBACH
(13) 50% BALLS
AUSTIN et of (141 20% BALLS
UC/O 159~[2:2(0.159/1,1~ 35-q
0.1
0.2
Wz/S Sj.kmk
k=l
(15)
[2(0.1-59/f,)-55
l]
(17a)
i6
TABLE
I
L
I
3
3
0
-
,.I,,,
EXPERIMENTAL_
StMULAlEO
t
t I.,,.
1000
100
SIEVE
SIZE,pm
Fig. 9. Rosin-Rammler
plot of circuit simulation
vcrsws experimental results; assuming simple power
function for S.
f 4.55 J -
20.4J3 + Y_2.9J4
(16)
where
0 (J)
(f,S)*
6.75s -I-4.253
(16a)
(J + 1.5J-
l]
6.7 J3 + 4.25 J)
(19)
Then, for a fixed value off,, k2 is proportional to F(J), as shown in Fig. 8_ Thus for
low values of J the specific rate of breakage
increases almost linearly with the number of
balls, other factors held constant. This is in
reasonable agreement with experimental
results 1143 _ Equation (19) wiII probably be
modified for different mills as more information on the variation of S with f, and J
becomes available. Note that for filling conditions to just fill the ball interstices, that is,
f, = 0_4J, the value of k2 does not vary widely
with J due to compensating factors in eqn.
(191.
Rates of breakage do not vary much with
fraction of critical speed provided that the
mill is run close to the optimum value for
maximum power, which lies between 65 and
31.4
3
2%
2
1%
1%
1
718
69
76
63.5
51
38
31.6
25.4
22.3
First
compartment
Second
compartment
mk
mk
0.30
0.37
0.23
0.10
O-19
0.30
0.26
0.25
overall
WC
0.10
0.12
0.07
0.03
0.13
O-20
0.18
0.17
APPLICATIONS
(26)
77
10.0
0.1
102
103
SIEVE
IO4
SiZE.prn
SIEVE
SIZE.
in the production
mill, scaled
no maximum esisted, so that the match between the computed and the experimental
size distributions of the mill product and
recycle is considerably better for the upper
sizes. In addition, the values of k for the singlecompartment tersus the two-compartment
will show directly the quantitative advantage
of using a mix of larger ball diameters in the
first compartment and smaller in the second
instead of a complete mix through the whole
mill. The reasoning is as follows:
The real mill is more efficient than the first
simulation using a mix of balls because it
contacts the larger feed particles with a higher
proportion of larger balls_ Thus, to force the
false simulation to fit the real result, an
artificially higher conversion factor (of 0.76)
is required. On the other hand, the second,
real, simulation gives a truer estimate of O-71;
the comparative difference in efficiency is
approximately 0.05;0.71,
or 7%. Put another
way, if the S, values were corrected using the
true 0.71 factor and applied to a singlecompartment mill with a miu of all balls, the
flow rate through the mill (and hence circuit
output rate) would have to be decreased to
get the desired product size distribution (by
about 7%), or if the same flow rate were used
the product size distribution would be coarser.
ACKNOWLEDGEiUEPil-S
LIST
a
A
hi
Equations are given which enable the
conversion of specific rates of breakage
measured for one set of ball milling conditions
to values for other conditions, in particular
the conversion from one ball diameter to
another or a misture of balls. The position of
the maximum in the specific-rate-of-breakage
versus particle size plot is related to ball
diameter d by x, = h, d. This can be compared to the well-known criterion for ball size
selection, that is, x = Kd where d is the
best ball size to break particles of size x_
Clearly, this old criterion is crude, since the
particles do not stay at size x as they move
through the mill. We have demonstrated that
using the developed relations combined with
a mill model for a large cement mill enables
us to predict that a mix of larger balls in the
first compartment and smaller in the second
compartment is 7% more efficient than the
bi.i.k
J
k
OF SYMBOLS
79
k-1
k2
k
K1
mk
n
Q(z)
Si
i3f
minSi.k
t
wiA
tf)
wi(rl
W
_x
*In
G,(J)
log u
REFERENCES
1 P_ T_ Luckie and L. G. Austin, Review introduction
to the solution of the grinding equations by digital
computation,
Miner. Sci. Eng., i (1972) 2-2_
2 W. H. Coghill and F. D. Devaney, Ball mill grinding,
U. S. Bureau of Mines, T_ P. 581 (1937),
p_ 47.
3 F_ C. Bond, Grinding ball size selection. Min. Eng.,
10 (1958) 592.
4 L. G. Austin, Review introduction to the description of grinding as a rate process, Powder Technol.,
5 (1971172)
1.
5 L_ G. Austin, K. Shoji and M. D. Ever-all,
Explanation cf abnormal breakage of large
particle sizes in laboratory mills, Powder Technol.,
7 (lSi3)
3.
6 W. J. Taut&, P. Meyer and L. G. Austin. Comparison of breakage parameters in two tumbling ball
mills of different diameters, Proc. Symp.
Automatic Control in Mining. Mineral and hIeta
Processing, IFAC, Sydney, Australia, Aug. 1973,
p- 21.
7 D. J. Kelcall and K. J. Reid, The derivation of a
mathematical model for breakage in a small
continuous wet ball mill. Proc. AIChE/Chem.
E.
Joint hleeting. London. 1965, Sect. -l, p_ 11.
S R. Snow, Grinding mill simulation and scale-up
of ball mills, Proc. 1st Int. Conf. Particle Technol..
IITRI, Chicago, Aug. 1973, p_ 28.
9 V. K. Gupta and P. C. Kapur, Empirical
correlations for the effects of particulate mass
and ball size on the selection parameters in the
discretized batch grinding equation, Powder
Technol.. 3 (1971) 217.
10 D. F. Kelsall, K_ J. Reid and C. J. Restarick,
Powder Technol., 1 (1968) 291.
11 L. G. Austin, Understanding ball mill sizing, Ind.
Eng. Chem. Process Des. Dev., 12 (1973) 121.
13 I, G. Austin, P_ T_ Luckie and H. M_ van
Seebach. Optimisation of a cement milling circuit
with respect to particle size distribution and
strength development,
by simulation models, 4th
Eur. Symp. on Size Reduction, Nuremburg,
Sept. 1955.
13 H. M. van Seebach. Effect of vapours of organic
liquids in the comminution
of cement clinker in
tube mills, Research Institute of the Cement
Industry, Beton-Verlag GmbH, Dusseldorf, West
Germany, 1969_
1-I L. G. Austin, k N. Beattie and R_ R. Klimpel,
Proc. 2nd Eur. Symp. on Size Reduction,
Amsterdam,
Dechema Monogr., 57 (1967) 281.
15 H. E. Rose and R. M. E. Sullivan, Ball, Tube and
Rod Mills, Chem. Pub., New York, 1958.
16 L. G. Austin, P_ T. Luckie and D. Wightman,
Steady-state simulation of a cement milling
circuit, Int. J_ Miner. Process_, 2 (1975) 127.
17 L G. Austin, P. T. Luckie and B. G. Ateya,
Residence time distributions in mills, Cem.
Concr. Res., l(l971)
211.