Chemical Engineering Science, 1965, Vol. 20, pp. 953-963. Pergamon Press Ltd., Oxford. Printed in Great Britain.

A solution to the batch grinding equation

K. J. REID

Division of Chemical Engineering, C.S.I.R.O., Melbourne

(Received 25 January 1965;in revised form 10 March 1965)

Abstra~-A practical approximation to the fundamental integro-differential equation for batchgrinding

is presented and a solution developed. The solution is written in terms of two experimentally deter-
mined basic parameters of the system, (1) rates of breakage and (2) breakage function, and is in a
form suitable for simple extension to describe continuous grinding systems.

A different approach was reported in 1953 by

SEDLATSCBBKand B&s [4]. They proposed a set of
SIZE reduction, or comminution, is a unit operation differential equations to describe the batch grinding
of primary importance in modern industry and, process and a series of experiments to yield the
although it has been used for many centuries, it still unknown coefficients. These experiments were done
has no broad theoretical foundation. using aluminium oxide, and the technique was
In recent years there has been a wealth of litera- shown to be applicable to a system divided into four
ture on crushing and grinding [l], but this has been size intervals. Their approach has been criticized [5]
essentially of an empirical or descriptive nature, with since their experimental technique necessitates the
only a few papers contributing to the theoretical dubious assumption that the probability of breakage
background. The limited number of related publica- is independent of the proportions of any other sizes
tions in the chemical engineering journals is not a present, and the unknown coefficients have to
true reflection of the importance of this unit opera- be obtained by the simultaneous solution of the
tion in the chemical industries and it is expected that differential equations.
chemical engineers will show an increasing interest The following year BASS[6] presented a mathema-
in all aspects of solids processing in the near future. tical theory for the milling process and was the first
In 1948 EPSTEIN[2] established the statistical to derive the fundamental mass balance for batch
approach to the breakage of solids and, in particu- grinding as an integro-differential equation. Under
lar, introduced the two basic comminution func- certain restrictions he showed that the formal solu-
tions (a) the probability of breakage of material of a tion to the basic equation yields an expression
given size and, (b), the size distribution of the pro- equivalent to that of the previously semi-empirical
ducts of a breakage event. For particular cases of Rosin-Rammler size distribution. This theory was
these probability and distribution functions he was a formal derivation of the practical solution given
able to show that the products of a multiple break- in the earlier paper and presented no method of
age process would tend to a log-normal size distri- overcoming the limitations already discussed.
bution as is frequently observed in practice. A matrix representation of breakage, using the
GAUDIN et al. [3] used radioactive albite as a concepts of EPSTEIN, was published in 1956 by
tracer to follow the breakage of single screen sizes in BROADBENT and CALLCOT~[7]. This approach is of
a normal feed of albite but reported results at one great value for describing the breakage process, but
grinding time only. Although this paper showed a the assumptions introduced by the authors to relate
practical way of observing the behaviour of material their mathematical representation to experimental
originally contained in a single screen interval, there work are open to serious criticism [5].
was no attempt to explore a mathematical repre- In 1962 two computer solutions to the integro-
sentation of the system. differential equation for batch grinding were, pre-

953
 K. J. REID

sented. GARDNERand AUSIN [8], using a radio- THE BASIC EQUATION

active tracer technique, measured the basic com-
The fundamental equation of comminution may
minution functions for three coals in a Hardgrove
be written
grindability machine. An iterative solution to the
integro-differential equation for batch grinding was d2F(x, t) dF (x,t)
- = -k(x) 7
used to calculate product size distributions, using axat
the known feed size distribution and the experi-
mentally determined comminution functions, and
the calculated results were in agreement with their
experiments. This work established the feasibility
+
s xoaqx,u)
x
-.k(u).yy.
ax
aF(a, 0
da

In this equation F(x, t) gives the weight fraction of

(1)

of measuring the fundamental properties of the

material finer than size x after grinding a given feed
system and using them to predict experimental
material with size distribution F(x, 0) and maximum
results, and also confirmed that the basic batch
size x0, for a time t, and B(x, a) is the breakage
grinding equation is a valid representation of the
function, giving the fraction of material finer than
physical system.
size x obtained by primary breakage of material of
GAUDIN and MELOY [9] in a purely theoretical
size a. Primary breakage refers to conditions under
paper used the matrix notation for the solution to
which the products of any breakage event undergo
the batch grinding equation and presented calcu-
no further fracture. The remaining term, k(x), is
lated results for several theoretical batch grinds
the fractional rate of breakage of material of size x
involving different probabilities of breakage and
and has been referred to by previous workers as
breakage distribution functions. Their calculated
either the probability function or the selection
size distributions showed the same general form
function. This approach leads to the assumption of
as experimental results reported in the litera-
a time independent rate of breakage which implicitly
ture.
defmes a first-order reaction, since if M(x, t) =
This work was extended by MELOY and BERG-
aF(x, t)/ax . dx gives the weight fraction of material
STROM[lo] in 1964, who used the matrix representa-
in the size range x to x + dx, then the fractional rate
tion for several hypothetical grinding circuits and
of breakage is
showed the relative importance of each of the con-
tributing parameters. dM(x, t)
All of the papers described have been directed = k(x)
- M(x, t) dt
towards obtaining a fuller understanding, and an
accurate mathematical representation, of the batch and by integration
grinding system and any attempt to extend this
M(x, t) = M(x, 0) exp[ - k(x)t] (2)
approach to a continuous grinding system has in-
volved the stated (or implicit) assumption of plug Equation (1) is an integro-differential equation
flow. and results from a mass balance on the weight frac-
An experimental study of breakage in a continu- tion of material M(x, t) in the differential size range
ous grinding system was reported by KELSALL [l l] from x to x + dx. The equation states that the net
in 1964 and led to a search for a workable mathem- rate of production of material of any size x equals
atical representation for continuous grinding. the rate of production of that size from breakage of
The batch grinding solution presented here is in a all material initially larger than size x minus the rate
form which enables the results for a continuous of breakage of material of size x.
mill, in which there is a distribution of residence Equation (1) as it stands has no simple analytical
time, to be calculated relatively easily. There are solution but under certain conditions may degener-
also certain advantages in using this solution for ate to an equation with known solution. If we
batch grinding calculations since neither an iterative consider the largest size of any feed, or start with
solution nor matrix multiplication, both of which feed of one size only, the decay of the largest size is
rely on the use of digital computers, are used. given by equation (2). A second degenerate case

954
 A solution to the batch grinding equation

arises when the kernel k(a) [aB(x, a)/ax] vanishes, meters which are defined below and which can be
i.e. when k(or) or aB(x, a)/ax = 0. If k(a) = 0 there directly determined by experiment. The term is
is no breakage at any size and consequently the used in this sense to differentiate the resulting equa-
weight at any size remains constant at its initial tion from the purely theoretical form of equation
value. If i%(x, cc)/dx = 0 then B (x, a) is constant (1).
and this corresponds to a system in which a particle Since size distributions are usually determined by
on breakage crumbles completely to a very fine sieving techniques using a constant screen ratio, and
powder. Under such conditions equation (2) again a solution continuous in time has advantages when
holds and all the feed material would decay with its extended to describe continuous grinding systems,
own rate constant. A more interesting analytical a finite difference approximation to equation (1) is
solution exists for the special case when the kernel sought which uses a geometric size interval and
k(a)[dB(x, a)/ax] takes on a constant value K, e.g. maintains a continuous time scale.
when Consider a set of 112screens numbered from the
coarsest downwards, with a screen ratio /3. The
k(a) = Ka and B(x, a) =z (3) general screen with number i, has an aperture of
size Xi and the size of the next coarser screen is
under which condition the equation degenerates to xi-l = /3xi Material which is retained on screen i
after passing through screen i - 1 has a size range
a’F(x, t) _Kx aF(x, t) + K Xo=(a, t)
-= -.da (4) from Xi to Xi-l, and has a fractional weight wi(t) at
dxat ax sx ax time t. For brevity this material, which contains a
This equation has the solution [9] single size fraction, will be referred to as being of
size Xi, i.e. the size of the screen on which it is
2(x, t) = z(x, 0) exp( - Kxt) (5) retained. If w,,,(t) represents the fractional weight
where z(x, t) is the fraction of material larger than passing the finest screen then
size x at time t, i.e. z(x, t) = 1 - F(x, t). Com-
parison of equations (5) and (2) shows that under + 2 wi(t>= l
WreS(t)
i=l
the special conditions of equation (3) the weight of
material above size x decays as if it were all exactly since the sum of alI fractional weights is unity.
of size x. Material originally of size xi (i.e. in the range xi to
In the absence of an analytical solution to equa- fixi) is defined as being unbroken if it is retained on
tion (1) a finite difference approximation must be the original screen i after exposure to any breakage
used in which either or both of the size and time system. Events which result in the breakage of the
variables are divided into small increments. The larger particles in the range in such a way that some
accuracy of the approximate solution depends on part of the product is retained in the same screen
the size of the increments used. The use of the finite interval cannot be measured experimentally, and all
difference representation requires careful redefini- particles which remain on the original screen are
tion of the contributing functions in terms of the considered unbroken, as defined above.
increments selected and the resulting solution must The breakage function modified to incorporate
be examined as a function of the increment size the screen ratio /I is written B(B, x, a) and gives the
since the exact solution results only in the limiting size distribution of the products of primary breakage
case of the differential increment. These precau- of material originally in the size range a to pa.
tions have not always been observed by previous Since any practical method aimed at determining
workers. the breakage function must involve the definition of
breakage and its consequences discussed above, it
follows that the breakage function will vary with the
THE PRACTKAL EQUATION
screen ratio employed.
The term “practical equation” is used here to The breakage function can also be represented by
describe an equation based on the practical para- the fractional weights retained on screens below the

B 955
 K. J. REID

initial size and these will be denoted by h. If b&wAO)

k, _ kz x
material of original size Xj, i.e. retained on screenj, b(f) = ~~(0)exp(-k,t) +
undergoes primary breakage, then b,,j gives the
x [exp(-kzt) - exp(-kg)] +
weight fraction of the broken material (i.e. finer
than Xi>,which is retained on screen i where Xi < Xi + bwklwdo)
and consequently i > j. k, _ kl Cexpt-kit>- expt-Ml +
The fractional rate of breakage
w(--k,t)
ki= --
dwitt) (k3 _ k,)(k, _ k,) -

Wi(t) dt
expt
- kd) expt-WI
also incorporates the above definition of breakage - (4 - k,)(k, - k,) + tks - kd& - kd 1
and is assumed to be independent of time. This
assumption will be fully discussed later. (SC)
TO obtain a mass balance for size Xi, consider a etc.
small increment of time dt during which the amount
broken from size xi is kiwi(t)dt. The amount pro- The significance of each term in the equations (9)
duced from size xj, j < i, is bi, jkjwj(t)dt and the can be clearly seen e.g. in (SC).
amount produced from all sizes above xi is (a) The first term represents the breakage of
material originally of size 3.
i-l
(b) The second term represents material formed
-C b,jkjwj(t) dt
j=l by breakage of material originally size 2 to size 3.
(c) The third term represents material formed by
Hence, dividing through by dt, the rate of accumu- direct breakage of material from size 1 to size 3.
lation of material of size Xi equals the rate of pro- (d) The fourth term represents material broken
duction from all larger sizes minus the rate of from size 1 to size 2 and then rebroken from size 2 to
breakage size 3.
The number of terms involved in the ith size is
~ = ~~: bi,jkjwj(t) - kiw,(t) (7) 2(‘-l) and the algebra involved in writing solu-
tions for i > 4 becomes cumbersome. To over-
This equation is a particular finite difference form of come this problem terms in exp (- k,t) are collected
equation (1) and becomes identical to equation (1) and the solution written in the form
whenp + 1.
Wi(t) = i a,,i exp( - k,t) (10)
n=l

SOLCJTION
TO THE PRACTICALEQUATION with the coefficients a,,i to be determined.
The solution wi(t) to equation (7) may be obtained Substitute equation (10) into (8) and rearrange to
by writing the equation in the form obtain

i-l j

and integrating cjgl nzlkjbi,jan.jexp[th - k&l (11)

w,(t) = w,(O) exp(-k,t) (9a) Integrating

bz.lkiw,(O) x Wi(t) CXp(kit) - Wi(O)

wdt>= ~~(0)exp( - k,t) +
kz - kl
{expC(ki- k&l - 11
x Cexp( - kl t) - exp( - k, t)] (9b)

956
 A solution to the batch grinding equation

Hence Consider material in the size range u to /?a. The

material which is broken is, by definition, that
Wi(t) = Wi(O) exp( - kit) +
which passes size u, and from equation (5) is given by
1 - exp( - KS). The material passing any size x,
[exp( - k,t) - exp( - kit)] smaller than a, is similarly given by 1 - exp( - Kxt)
(12) which, as a fraction of the broken material is
[l - exp( - Kxt)]/[ 1 - exp(K&)]. Expanding the
Equating coefficients in equations (12) and (10) exponential terms and letting t + 0, the practical
(i) for n # i breakage function is obtained as follows :
1 - exp( -Kxt)
(13) w, x, 4 =
1 - exp( -Kat) t+o
= 2 = B(x, a)
~1

(note that the summation runs from rz to i - 1, This result shows that under the conditions of equa-
since a,,jOforallj < n).
=
tion (3) the practical breakage function is identical
to the theoretical breakage function and is inde-
(ii) for II = i pendent of the screen ratio.
i-l

04) Using these practical functions it is possible to

n=l examine the solution given by equation (10). Con-
The advantages of the form of the solution given by sider the top two screens ; then
equation (10) are easily demonstrated by using i = 1; a,,, = ~~(0)
equations (13) and (14) to determine the coefficients
for the third screen and comparing the result with klb,lal,,
i = 2; a,,, =
equation (SC): kz - kl
and
i=3:n=l:al,=
klb3,1al,l W3,2a1,2
a2,2 = w2(0) - al,2
k, - kl + k3 - kl
The rates of breakage are
k2b3,2a2,2
n=2:a2,=
k, - kz kl = Kxl and k2 = Kx2 = 7
Kx,
n=3:a,,,= w3(0) - al.3 - a2,3
and the practical breakage function becomes
and

w3(0= q3 exp( - k,t) + a2,3 exp( - k,t) + b2,1 = 1 - z= 1 1 -

B
+ a3,3 exp(-k30 (15) Hence
al,2 = --a,,, = -w(O)
The practical solution given by equation (10) with
giving
equations (13) and (14) can be examined under the
conditions of equation (3) for which the analytical Z2(0 = w(t) + w20)

solution to the basic equation is known and given

by equation (5). Under these conditions the rate of
= CM9 + WZ@)Ied- W
breakage through any size x is independent of the = ~~(0) exp( - k,t)
size distribution above that size and it follows that
which is identical to equation (5).
the rate of breakage of a practical size fraction with
lower limit x will be the same as the rate of break- The analytical solution is obtained similarly for all
age of material of exact size X. The relationship other screens.
between the practical breakage function B(P, x, a)
The fraction passing any screen i is
and the theoretical breakage function B(x, a) can be
established for the same conditions as follows : Fi(t) = 1 - Zi(t) (16)

957
 K. J. RED

Table 1. The coefficients c”,i for coal A

Screen
size 840 590 420 297 210 149 105 74
P

\ i 1 2 3 4 5 6 7 8

1 0.72500 oa4304 0.08983 004524 0.03327 001926 0.01246 0*00716

2 0.95696 oa4153 0.13116 0.05091 oa4055 002168 0.01418
3 O-86863 -1.6112 l-0632 -0do735 0.11094 -0.03348
4 2.4348 - 4aO97 2.6147 -0.94711 0.25413
5 3.8623 - 67977 44422 - 1.5342
6 5.5306 -9.3163 5.5335
3 6.6762 - 6.9141
IO.155

Table 2. The coefficients c,,i for coal B

Screen
size 840 590 420 297 210 149 105 74 53
CL

i
1 2 3 4 5 6 7 8 9
\ II
0.69000 021022 0.13772 008547 0.05133 @03100 0.01909 001135 0.00663
: 0.78978 0.22499 O-12528 0.06810 0.03508 0.01851 0.01027 0~00516
0.63729 - 7.5395 34.961 - 104.14 214.56 - 349.48 394.98
8.3288 -71.120 316.57 - 884.76 1871.4 -2634.7
37.040 -398.65 1925.8 - 6293.0 12433
187.16 -1851.8 9798.7 -26911
597.21 - 77075 33548
2680.9 - 19575
2745.8

Table 3. The coefficients c,,i for coal C

Screen
size 840 590 420 297 210 149 105 74 53
P

i
1 2 3 4 5 6 7 8 9
\ n

1 0.60000 0.29000 0.18039 0.11427 007308 oa4790 0.03088 0.01988 0~01300

2 @71000 0.33308 0.16729 0.08955 om993 0.02989 0.01748 0~01010
3 048653 -0.99252 0.63743 -0~10445 -0.01229 -ONJO -0GO355
4 1.7110 -3.7641 2.6312 -044560 -0.07586 --001044
5 3.9640 - 8.9746 5.9966 -0.31354 -0.07709
6 7.3500 - 15.295 6.2482 -0.19764
7 10.695 - 14.202 5.7325
9.3064 - 19.141
14674

958
 A solution to the batch grinding equation

and can be obtained from a summation of weights and the batch grinding results calculated. The
retained on the individual screens since coefficients c, ,i are listed in Tables l-3, and the
calculated results compared with their experimental
zi(t) = jilwj(t) = i C,,i exp( - k,t) (17) and computed results in Figs. l-3.
II=1 Feed size distributions used for the calculation
where the coefficients C,,i are given by were obtained from the original graphs [5] since the
data were not tabulated. The figures used for frac-
tion passing 590 p were 0*275,0*3 1 and 0.4 for coals
A, B and C, respectively.
Figs. l-3 show that the results calculated by this
method are virtually identical to the computations
COMPARISON
WITH RESULTSOF GARDNER et al.
of GARDNERet al. The small differences are prob-
A complete set of experimental data for the ably due to minor errors in estimating the feed
breakage functions, rates of breakage and batch sizes.
grind products for three different coals has been
given by GARDNER et al. [5]. Their results are
EXAMINATION
OF ASSUMPTIONS
reported in terms of average sizes and they use a
,/2 screen ratio, consequently the screen size as Having established the validity of the practical
used in this work is 2/(1 + J2) = 0.83 times the solution by comparison both with experimental
average size. Using this factor their experimental results and a different solution to the basic equation,
results for breakage function and rate of breakage, it is now necessary to examine in detail the assump-
as defined in this paper, have been substituted tions on which the solution is based.
directly into the practical solution proposed here, The major assumption is that the rate of breakage

X Calculated from
X Calculated
from
equation (16)

I I I I I
60 100 200 400 600 loo0 60 100 200 400 600

Sire p Size p

FIG. 1. Comparison of calculated size distributions with the FIG. 2. Comparison of calculated size distributions with the
results of GARDNERet al. results of GARDNERet al.

959
 K. J. REID

(ii) k,, = Kx, and B(/?, xi, xi) = z(cf. equation 3)

under which condition

cR,p= 0 for n #g
and
cn,g= 1 for n=g
giving
-=dw(0 _k
9
w(t) dt
(iii) One k controls, i.e.
k,, > k for n# i
X Calculated from - 0°4 ki < k
equation (16)
Then
dw(t)
-= _k
I
w(t) dt
Consideration of equation (19) leads to the conclu-
1 I , I I 1 I sion that in general the fractional rate of breakage
60 100 200 400 600 1000

Size p
of material in a finite size range is not independent
of time and consequently the assumption of time
FIG. 3. Comparison of calculated size distributions with the
results of GARDNERet al. independence is, in general, theoretically invalid.
of a finite size range is time independent. Consider a Under certain circumstances the assumption is
screen ratio B close enough to unity to define a dif- justified but the probability of an experimental
ferential increment; then if g increments are re- system exactly satisfying the restraining conditions
quired to achieve a practical screen interval the is small. There is a wide range of possible rates of
weight w(t) on the screen as a function of time is breakage and breakage functions, and the extent of
the time dependence of the rate of breakage of a
w(t) = z,(t) = f c~,~exp(-k,t) single screen size will depend on the combined
II=1 effects of both parameters.
Therefore Experimental work on aluminium oxide [4],
coal [8] and quartz [l 11,in ball mills has shown that
dw(O
-=.f the decay of a single size is closely represented by a
dt n=l
- ~,,~kexp(- kt) straight line on semilogarithmic paper, and con-
sequently the time dependence of the rates of break-
and the fractional rate of breakage is
age of these materials is negligible under the circum-
stances of the particular experimental system and
dw(t)
-=-
n&n,gk.
exp(- k&) over the time range investigated. The validity of the
w(t) dt
(1% assumption for these three widely different materials
n=1
t c”,~ ew( - k0 suggests that it may be valid in general for ball mills.
However, in any particular system the question can
Equation (19) is time-independent when only be resolved by experiment.
(i) k,, = k for all n, giving The breakage function used in the practical solu-
tion is defined in terms of experimental measure-
-=Wt) _k ments and contains only the inherent assumptions
w(t) dt that the size distribution within a single screen

960
 A solution to the batch grindingequation

interval remains constant. A standard method of use of a finite time interval eliminates the advan-
presenting size distribution is the log-log plot of tages of the time-continuous solution. The compu-
fraction passing against size, and a “natural” system ter solutions of GARDNER and AUSTIN and of
is characterized by a smooth continuous curve. If GAUDINand MELOYboth use this approach, which
p is small (,/2 or less), natural size distributions can neglects the important factor that the basic func-
be well approximated by straight lines over single tions must be determined experimentally, and as a
screen intervals, and the assumption of constant consequence experimental results have to be cor-
size distribution within the interval is valid. For rected back to the differential size increment. Cor-
larger screen ratios or systems with highly artificial rections of this nature, which are neither simple to
feed size distributions, the assumption may break to use nor easy to justify, were used by GARDNER
down with the possibility of invalidating the analy- etal. [5], and have been fully discussed by them.
sis. The second approach is that adopted in this paper
As a result of these considerations screen ratios and involves the re-definition of rate of breakage
greater than ,/2 are not recommended for practical and breakage function in terms of the experimental
work in the field and due care should be taken to methods used for their determination. This ap-
preserve the natural size distribution during the pre- proach removes the necessity to correct the experi-
paration of material in a single screen interval. mentally determined functions and avoids any dis-
cussion of the validity of the corrections employed.
Comparison with previous work has also shown
DISCUSSION
that the present solution produces equivalent results
Of the three previously proposed methods of cal- although requiring a less involved mathematical
culating batch grind results, only that of Buss uses a procedure.
continuous time scale. His solution can be written Some workers have postulated [2, 71, and
in the form of equation (10) and his proposed others have shown experimentally for the systems
method for evaluating the coefficients U,,i involves studied [5, 121, that the breakage function, expres-
the assumption that the experimental system is des- sed on a relative size basis, is independent of initial
cribed by the equations and then, using the experi- size. Under this condition the present solution can
mental results, solving for the coefficients. However, be used, in principle, to compute rates of breakage
the accuracy of any method which solves a set of and breakage function given size distributions
equations from experimental data must depend on after batch grinding for two different times.
unavoidable experimental scatter and consequently A general purpose iterative computer programme
the results are of doubtful value without further was written to do this, and although the correct
direct experimental justification. functions could be obtained from computed
In a system where the coefficients are not specified results, the presence of very small experimental
in terms of the fundamental properties of the errors resulted in unrealistic values for rates of
material and equipment, it is difficult to visualise breakage or breakage function. As a result of the
any method for measuring these coefficients directly. high sensitivity to minor experimental errors it is
Furthermore, they do not provide a sound basis for considered that this approach is of little practical
the correlation of different materials, especially value. However the less ambitious approach of
since they are dependent on the feed size distribu- solving for either rate of breakage or breakage
tion. function, having determined the other by a tracer
There are essentially two approaches to the solu- technique, and, given one grinding result, proved
tion of the basic equation. The first accepts the successful even in the more complicated case of a
definitions of the rate of breakage and breakage continuous grinding system [12]. This approach
function based on a differential increment and uses should be of considerable value in comminution
standard mathematical procedures to handle an studies since the painstaking experimental work
arbitrary finite difference approximation to the required to measure both of the basic parameters
basic integro-differential equation. However, the can be considerably reduced.

961
 K. J. &In

The advantage of a time-continuous system can lated data, both for steady state and transient
be seen by considering the expression for the steady results, and established the validity of this approach
state product from a continuous mill. The flow when applied to the particular system and conditions
characteristics of a continuous mill are given by the studied.
residence time distribution function 4(t), where
4(t)& gives the fraction of material entering at time
Acknowledgement-The author would like to thank Dr. J. A.
t = 0 which leaves in the time interval t to t + dt. BARKER for his valuable comments and discussion of the
Material with residence time t has a fraction retained problem.
above size Xi of Zi(t) hence zi(t)&t)dt gives the frac-
tion of material coarser than size xi with residence
time t. The total fraction coarser than xi is the sum
NOTATION
of these terms over all residence time :
an,r Coefficient of exp-(k,t) in equation (10) for
m fractional weight on screen i
Zi(cO)= Zi(t)&(t)
dt (20) B(x, 4 Theoretical breakage function for differential
s0 size increment, giving weight fraction passing
size x from primary breakage of material of
exact size CL
where Z,(W) is the steady state fraction retained
B@, x, 4 Practical breakage function for a single screen
above size Xi in the product from a continuous size increment, giving weight fraction passing
grinding system. Substituting for Zi(t), equation size x from primary breakage of material in
size range LYto /3a
(20) can be written Practical breakage function as weight fraction
retained on screen i from primary breakage of
materiaf on screen j
Zi(a) = i Cn,i mexp( - l&j(t) dt (21) Coefficient of exp-(k&) in equation (17) for frac-
II=1 s0 tional weight above screen i
W, t) Weight fraction of material finer than size x at
For many systems 4(t) can be represented time t
by a simple exponential function [13] and under In The value of the integral exp (--k&#(t)dt
OD
these circumstances the value I, of the integral I
k(x) Fractional rate of breakaLe of material of size x
j: exp (-k,t)~$(t) dt can be obtained analytically, min-I
leading to the algebraic solution kc Fractional rate of breakage of material in size
range x6 to /3x$ min-l
K Rate of breakage proportionality constant
rnin--lEl;l
zi(co) = t: G,Jn (22) M(x, t) Weight fraction of material in size range x to
PI=1
x+dx
m Total number of screens used
Even if the integral can only be calculated using a t Time min
digital computer, equation (20) is more rapidly w(t) Fractional weight on screen i at time t
wr,s(t) Fractional weight passing finest screen at time t
evaluated when zi(t) is available in the form of x Size p
equation (17) rather than being generated from an x0 Maximum size in feed CL
iterative or matrix solution. X6 Size of aperture of screen i
Zi(W) Steady state fractional weight grefter than size
Equation (21), and the corresponding form for XLin the product from a continuous system
the dynamic response to a step change in feed pro- 4x9 t) Fractional weight above size x at time I
perties, have been used to calculate results in a z*(t) Fractional weight above size xf at time t
Size p
small scale continuous ball mill [12]. Close agree- Screen ratio.
;
ment was obtained between experimental and calcu- 4(t) Residence time distribution function

REFERENCES
[l] D.S.Z.R. Crushing andGrinding. (Bibliography) Her Majesty’s Stationery Office, London 1958.
[2] EPSTEINB.,Zndustr. Engng. Chem. 194840,2289.

962
 A solution to the batch grinding equation

[3] GAUDM A. M., SPEDDENH. R. and KAUFMAND. F., Min. Engng. 19513,969.
[4] SEDLATSCHECK K. and Bass L., Powder Metall. Bull. 1953 6,148.
[5] GARDNERR. P., AUSTINL. G. and WALKERP. L., Special Report to the Coal Research Board of the Commonwealth
Pennsylvania, Number SR-23. January 1961,.
[6] BASSL., Z. angew. Math. Phys. 1954 5,283.
[7] BROADBENT S. R. and CALLCOT~T. G., J. Inst. Fuel 1956 29,191.
[8] GARDNER R. P. and AUSTINL. G., Symposium Zerkleinern, 1962 217. (Dusseldorf Verlag Chemie, V.D.I. Verlag).
[9] GAUDINA. M. and MELOYT. P., Trans. Am. Inst. Min. Engrs. 1962 223,401.
[lo] MELOYT. P. and BERGSTROM B. H., Proc. 7th Znt. Mineral Proc. Congr. New York, September 1964 p. 19.
[ll] KELsALLD. F. Proc. 7th Znt. MineralProc. Congr. New York, September 1964, p. 33.
[12] KEL~ALLD. F. and REID K. J., Joint Amer. Inst. Chem. Engrs./Znst. Chem. Engrs. Meeting, London, June 1965.
[13] WOLF D., and F~ESNICKW. Zndustr. Engng. Chem. Fundamentals 1963 2,287.

R&um&L’auteur presente une approximation pratique de l’equation fondamentale integro-

differentielle du broyage discontinu et en developpe une solution.
La solution est resolue en fonction de 2 parametres fondamentaux du systeme determines
experimentalement: vitesses de broyage et fonction broyage. Elle est facilement applicable g la
resolution des systemes de broyage continu.

963