Beruflich Dokumente
Kultur Dokumente
Technology,
34 (1983)
261
261 - 274
AUSTIN
and K. BRAME
Department
of dlineraZ Engineering.
PA I6802
(U_S_A_)
(Received
MineraZ
Processing
Section.
PcnnsyZrionia
State
Lbiuersity,
Cjniuersity
hrk.
March 11,1982)
SUMMARY
INTRODUCTION
The Bond method [I, 23 for sizing tumbling ball mills has been used successfully for
many years. On the other hand, a more
detaiied method of analysis of the grinding
process has been developed in recent years
[3 - 113, based on mathematical
models of
t.he process. This method of analysis has been
variously
called
the phenomenological
approach 131, the size-mass
balance 14, 121
method or the population
balance met.hod
[5]_ The objective of this paper is to analyze
the advantages and disadvantages of the two
methods, compare the predictions,
and show
how the advantageous
features of each
method can be combined.
The primary purpose of the Bond calculation is to predict the mill size and mill power
to give a desired capacity Q (t/h) from a wet
overflow ball mill in normal closed circuit
operated with a circulation ratio C = T,fQ of
2.5 (see Fig. 1). The input to t-he calculation
is the make-up feed size characterized by the
Make-Up
Feed
Mill Producr
Fig_ 1_ A grinding circuit with recycle of the *aarse stream to the mill feed. The symbols
of solid.
0032-5910/3310000-00001so3.00
0 Elsevier Sequoia/Printed
in
The Netherlands
262
THE
BOND
METHOD
We,.,
= (110)(44
.
(1)
where p1 is the opening of the classifying
screen used in the test; xgT is the 80%passing-size of the product in micrometers;
and #oT the SO%-size of the original feed,
which is near 2000 pm. The factor l-10
converts the original Bond Work Index in
kilowatt ho&s per short ton to kilowatt hours
per metric ton.
Bond correlated the results from the
laboratory mill with those from a wet overflow ball mill which was 8 ft (2-44 m) in
diameter, operated in closed circuit, with a
circulation ratio of 2.5. His tests showed that
the specific grinding energy E was empirically
related to the make-up feed and circuit
product sizes by
(-
- 1
E=WIy!&-g_
(2)
D < 3.81 m
2.44/D)-*
D 2 3.81 m >
WZ = (W&,)(0.914)
(3)
0.9375)
P~LD*-~
(4)
(5)
265
0.937J2)
Q = 6.13
SW&es% -go
(
eqns. (2),
O-10,
2x
& -
-2
(3),
-)
D d 361
)
(6)
D3-3(L/D)pb(J
0.937J2)
@, I.
Q = 8.00
SWltest go
(
TABLE
-go
2;:;$c
>
D 2 3231 m
Correcting
[$(pr
Fineness of grind XQ
Low reduction
(ml
=G
70
60
50
40
30
20
15
10
5
3
1.00
1.02
1.05
1.19
1.17
1.32
1.47
l-77
2.67
3.87
Over-sized
feed: .$a = 1 + ($
ratio
>6
6
5
4
3
2.5
2.25
2.00
l-75
1.50
-7)
(2)
Open circuit
a(p,)
(%I
IxQ
1.00
l-03
1.04
l-05
1.08
1.11
1.14
1-20
l-33
1.9
( 4ooo~~l_l~~13~~~~~~~~
50
60
70
80
90
92
95
9s
1.035
1.05
1.10
l-20
1.10
1.46
1.57
l-70
1)
factor
c is the product
of a series of
empirical correcting factors to WI for different conditions
(see Table 1): (i) fineness of
grind; (ii) low reduction ratio; (iii) over-sized
feed; (iv) open circuit_ Figure 2 gives a typical
result of the calculation_
The
ADVANTAGES
AND
DISADVANTAGES
OF THE
BOND METHOD
-I
f
IOD
IO
GO k-PASSING
1000
SIZE
OF
PRODUCT.
IOOOD
Q.
264
AND
MAKE-UP
FEED
ODUCT
closed
circuit
treated
as two
only be valid as a gross method for mills operating under normal closed circuit conditions.
Esperimental
batch grinding tests (see Fig.
4) show that the rate of breakage (under
normal conditions) of material sized within a
J?Z upper-to-lower
screen interval f0110Ws a
first-order breakage law:
broken, but if they are measured before rebreakage occurs, the mass fraction arriving in
size interval i from breakage of size interval j
is symbolized
by bi.i, where X$Z?+ 1bi.i = lThis set of numbers varies from one material
to anotherA balance of material being broken into and
being broken out of size interval i is: rate of
accumulation
of size i material = sum rate of
production
from breakage of all larger sizes
(j = 1 to i - 1) - rate of breakage of size i to
smaller sizes. In symbols, this is
dw,(r)W
o
-
EXPERIMENTAL
CALCULATED
GRiNDING
TIME
@WdUTESI
20
dt
ri=l
C
i=l
1 i>
bi_jSjWj(t)?V
1
-SjWi(t)W
I
or
i-
dw,(t)
=
df
-Siwi(t)
+ C
i=,
i>l
bi_jSiwi(t),
TZ2 i Z 1
('7)
This is the basic set of equations for firstorder batch grinding_ The set of numbers Si is
a precise index of the weakness or ease of
grindability of each size, and varies from one
material to another_ The solution of the equations with a known starting feed of w,(O),
wZ(0), etc, for time t, gives ml(t), wz(t), etc.,
thus giving the size distribution produced.
This set of equations would apply to a
continuous
mill operating at steady state if
the entering solid flowed through the mill as a
plug with all material staying in the mill for
the same residence time T_ However, ball mills
have a residence time distribut.ion (RTD) because forward and backward mixing occurs
along the mill axis. The model must therefore
be extended to allow for the actual RTD, as
first described by Reid [14]_ If a fraction
Q(t) dt of the feed stays in the mill for a time
t, and leaves between t and f + dt, this fraction will be broken as if it were batch g-round
for time t. Its size distribution upon leaving
will be the set of numbers wi(t), obtained by
the solution of eqn_ (7) for each value of i.
Simultaneously,
material that has been in the
mill for different lengths of time will also be
leaving; at steady state. the total product in
size interval i, Fpi, which leaves the mill will
be the weighted sum of all fractions of size i
product:
PiF=
J
0
w,(t)@(r)Fdt
266
or
pi = j Wi(t)@(t) dt
0
RELATIONS
(8)
(9)
( I(
xi
X0
l*
1
(xi/ilA
n>i>l
(11)
(10)
aa
1
l+66J2_3
(12)
=~(--1.32w
a a (l/cZ~)
(ii)
a=DNa
No=
1.0
(13)
(14)
(15)
1
1+ exp[15_7(+,
-0.94)]
C,=
c
1 + 6-6&Z-31 + 6.652-3
0.1 1 + exp[15_7(&
= 4, s
esp[---1_32(U-
@CT- o-1
1 + exp[ 15_7(& -
Ur)]
-
(17d)
O-94)]
O-94)]
(i7e)
where sj is the overall. specific rate of breakage of size i for amisture of m different ball
sizes. (Note that S, = 0, since material cannot
break out of the sink interval.)
If the values of bi_i are cumulated from i =
n, the cumulative values B&=
Zi=,, bk_j)
have the form shown in Fig_ 6_ This can be
expressed as the sum of two straight lines of
slope y and /3, with a fraction Q, of slope y and
a fraction 1 - @ of slope /3s
Biej
x22
aj;
i
Lri
t+
1
(1 -
X,-,.3
61,) Xi
( 1
n>i>j
(19)
(16)
aT
where
N2 = 0.1 to O-2
(17a)
C2 = (dT/d)No
ATo1
(17b)
c, = (D/DT)N,
N1 = O-5
(17~)
266
from
which biSj are readily calculated. It is
assumed that the biSj values do not change
with mill diameter or mill conditions, providing that rotational speeds are reasonably close
to the maximum power condition so that the
balls are cascading properly, and that the
slurry viscosity is low enough to avoid
changes in 7 [19]. However, the values change
somewhat for different ball diameters [ 191.
Figure 7 shows the agreement betweeen
computed batch grinding results (using the
solution of eqn. (7)) and the experimental
data, for the characteristic breakage parameters given in Table 2, which were determined
in a small laboratory mill with 25.4 mm diameter steel balls.
TABLE
= 1.16 kg
= 2.65
Slurry density
= 64 wt.% soiid
= 40 vol.% solid
Ball charge:
d = 25.4
J = 0.35
Mill:
D=195mm
V = 5200 cm3
r.p.m. = 72 = 70%
Breakage parameters:
D = 0.35 min-
fY = 0.80
$1 ;y65 mm
mm
critical
y = 1.25
Q = 0.55
RTD, MASS
TRANSFER
AND
HOLD-UP
Measurements of residence time distribution in mills [20] indicate that the residence
time distribution in full-scale wet overflow
ball mills can be approsimated by the simple
equivalent system of one larger fully-mixed
reactor in series with two equal smaller fullymixed reactors_ However, when the measured
values of mean residence time r are used to
calculate mill hold-up W from r = W/F it is
also apparent that many mills are operated in
an over-filled condition. At high mass flow
rates through the mill, i-e_, short residence
times, the values of interstitial filling U
exceeds 1, and the values of a and W to be
&~Iiaed
in the simulation
are functions
of flow
rate.
It is convenient to perform the computations of the simulations using a constant a
value corresponding to U = 1, aI say. The
value of r thus obtained is a formal or false 7,
rf say. It is clearly related to the real -r by
used
Q = Wreall~rea~U+ Cl
= &W-JTf(l
+ C)
(20)
where
Equation (12) gives a,,&,
thus
as a function of U,
K. = U exp(-1_32U)/exp(-1.32)
SIZE.
PC=
(21)
(22)
269
TABLE
Slowing-do&x
U-05
FFOM
TO
1.0
80%passing
tvm)
Slowing-down
size
1.0
0.97
0.91
0.53
0.74
O-61
0.5-i
0.5-l
> 100
oi0 _
_.I._
.._
05
_.~ I _..___I
1.0
INTERSTITIAL
90
80
70
60
50
40
t30
_____I:.......~
x.5
20
FlLLlNG
25
as a function
of ball
u=
[(F/w,)T,J0-5
l)]
(23)
SLOWING-DOWN
EFFECT
Examination
of the long grinding times of
Fig. 7 shows that the simulation predicts a
finer grind than that obtained experimentallyThis results from a slowing down of breakage
rate as the charge becomes fine, due to the
development
of high effective viscosity in the
slurry [ 22,231.
-4ustin and Bagga [24] have
shown that the results can be simulated using
false first-order grinding times 8, giving a
slowing-down
factor K defined by
I
= Sg(t)/Si(O)
COMPARISON
BALANCE
OF BOND
_4XD SIZE-MASS
SIMULATIONS
(2M
If 71
gives
(23),
(21)
factor X
= de/dt
(24)
~(l-b)
1
1 -i- (xJdso)-h
(25)
270
loo0
, i I ,:*i
---7---z
i
l0l---!--LL_LliL_
10
PRODUCT
Fig- 9_ Comparison
simulation methods
100
GO X-PASSING
I : i LLLI/
,000
SIZE, a,
pm
= 19 kWh/metric
ton.
& 7
l-XT
: : 9.
*I
SIMULATIDN
!
4
1
1
1
7
3
2
2
1
,oi
: i , : ,,.i
10
t!
XI,.
100
PRODUCT
GO X-PASSING
IODO
SIZE.
x0.
III
by
varying
the
classification
effi-
TABLE
SI = 0.5,
Product
Q 0: D=(L]D)
Simulation:
Q = Wa a D3*x-
(L/D)
D<
b = 0.3
by-pass
xs,,
By-p&
S-I_
Capacity
factor
Q/Q-
(pm)
420 to 38
0.6
0.6
0.6
O-5
OS
0.5
o-4
o-4
0.4
l-10
1.04
0.97
l-05
1.00
0.94
l-00
0.95
0.89
0.2
0.3
o-4
0.2
o-3
0.4
o-2
0.3
0.4
i
.
. . . . .i
.:
. . . . . 1
PRODUCT
GO X-PASSING
of Bond
Fig. 9).
ID00
100
SIZE.
x0.
ym
and simulation
scale in the
Since IV, = 0.5, both methods
same way with respect to mill diameter and
length and the conclusions from Fig_ 9 are unchanged, providing that the mass transfer/
over-filling relation scales in the same way_
operation,
IO
3-81
DISCUSSION
10 j
D < 3.81 m
Bond:
results
the
concepts
behind
the
method
(26)
His argument
can be reduced to this:
eliarnination of the variation of 80%passing
size of the product (xx) in a batch grinding
test shows the E a lO&&relation
to apply,
and the Griffit.h crack theory of fracture [27]
contains a l/(crack
length)*
term, therefore
eqn. (26) must be a fundamental
law_ Such
reasoning is clearly inadequate, but it is often
repeated without question; for esample [ 283 :
The Bond theory states that work input is
proportional
to the new crack tip length
produced in particle breakage___., which is an
essentially meaningless statement.
Even if
eqn. (26) were generally true, it should
logically be applied to the feed to the mill and
the product from the mill not the circuit feed
and circuit product_ Since it. was deduced for
batch grinding, it should apply to a mill in
plug flow, whereas RTD esperiments
show
mills to be closer to fully-mized.
Again, if it
were a fundamental
law, why should it be
necessary to use fineness-of-grind
and low
reduction ratio corrections to the law?
It must be concluded that the so-called
Third Law of Comminution
is not a fundamental law; and therefore it does not have to
hold under all circumstances as it is entirely
empirical_ The results presented in Fig. 7 do
indeed fit eqn. (26) (assuming E grinding
time, ie_ constant mill power), but they are
272
TABLE
Effec:
5
of Bond
Test Work
(kwhit)
Index
oversized
feed
Feed
s G
correction
Product
factor
Oversized
feed factor
Low reduction
ratio factor
-&-&
Capacity
(t/h)
(mm)
X8
(mm)
14
10
10
1
2
1.92
2.83
1.0
1.04
0.216
0.124
216
247
19
10
10
1
2
2.96
4.90
1.0
1.04
0.216
0.124
104
105
24
10
10
1
2
4.33
7.64
1.0
1.04
0.216
0.124
56
53
factor
(C =
80% <
CONCLUSIONS
Bond method will over-predict capacity, unless the slurry density is reduced and appropriate additives used.
ACKNOWLEDGEhIENTS
4
.5
10
11
12
13
1-f
15
16
17
1s
19
20
274
hfineral and hfetal Processing.
Pergamon Press (1980) 303 - 309_
J. C. Marchand, D. Hodouin and M. D. Evereli,
Proc_ 3rd IFAC Symp.. Automation
in hfining.
Mineral and Metal Processing, Pergamon Press
(1980) 295 - 302.
R. R. Klimpel, Powder TechnoL
(in press).
R. R. Klimpel and L. G. Austin, Powder Technor..
32 (1982) 239 - 253.
L. G. Austin and P_ Bagga, Powder TechnoL.
28
in hfining.
21
22
23
24
(1981) 83 - 90.
L. G. Austin and R. R. Kiimpel, Powder Technol.,
29 (1981) 277 - 281_
26 A. F_ Taggart, Eahdbooh
of Mineral Dressing,
Ores and Industrici: hfinerals. Wiley, New York,
25
1945.
27
p_ 100_