Powder Technology.

38 (1984) 255 - 266 255

A Simulation Model for An Air-Swept Ball Mill Grinding Coal

L. G. AUSTIN, P_ T_ LUCKIE

Deportment of Mineral Engineering. The Pennsylvania State University. University Parh. PA (U_S._.l.)
K. SHOJI
h-ure Research Laboratory (formeriy h’ennedy Van Saun Corporation), Hiroshima (Japan)

R. S. C. ROGERS and K_ BRAILLE

Kennedy Van Saun Corporation. Danuille. PA (USA.)
(Received February 3.1983;in revised form May 6.1983)

SUMMARY industrial dry grinding of coal is always

carried out with a high rate of air-sweeping to
The conventional model for grinding in remove fines from the mill_ The inlet air is
tumbling ball milk was modified to allow for heated to give drying in the mill so that the
air-sweeping, for the case where all the product is fine dry coal- Surface moisture
material is carried out of the mill in the air prevents the fines from being swept out and
stream. It was shown that this type of mill greatly reduces grinding efficiency-
can be treated as a single fully mixed reactor. The simulation of tumbling ball mills wit.h
The values of the internal classification mathematical models has reached a fairly
numbers given by the air-sweeping were advanced stage 13, 4, 51 but the concepts
determined for a 1 m by l-5 m pilot-scale mill have not been applied to coal grinding be-
grinding coal. The resuits showed that only cause the models have not included the effect
1 to 270 of the mih charge are exposed to the of air-sweeping- In this paper, we will modify
sweeping action per mill revolution_ S and B the models to allow for air-sweeping, assum-
values determined in a laboratory mill were ing that the removal of fines allows the frac-
scaled-up for use in the continuous miiZ modei ture to proceed in a first-order manner [2].
and the simulations gave product size distri- The characteristic of an air-swept ball mill is
butions and mill capacities which agreed with that the entire flow of powdered coal from
the pilot-scale mill data within the experi- the mill is in the air stream- The model
mental accuracy of the pilot-scale data. incorporates closed circuit via an estemal
classifier and t.he simulation is validat.ed by
tests on a pilot-scale mill_

INTRODUCTION
THEORETICAL TREATMENT
The most widely used equipment for
grinding coal to the sizes prescribed for Consider the mill receiving a feed of known
pulverized coal firing of utihty boilers are the size distribution: as customary [3], the size
ball-race and roll-bowl types of mill. However, intervals of the size distribution are chosen in
there are some advantages to the use of \/2 geometric sequence corresponding to the
tumbling ball mills for coal grinding and standard sieve sequence, with the masimum
several manufacturers supply such systems. size interval denoted by 1, the second size
It has been known for many years that coals interval by 2, and so on_ The nth interval is
do not grind well if fines are allowed to buiid the ‘sink’ interval containing all material less
up in the grinding zones. This has been than the smallest size measurement, 2’70 mesh
recently confirmed [l, 21 by studies of the (53 pm) for the tests described here_ The feed
specific rates of breakage of coals in ball-race is denoted by the set of numbers fi, where f,
and tumbling ball mills. For this reason, is the weight fraction in size interval 1, etc_ It

@ Elsevier SequoialPrinted in The Netherlands

256

willbe assumed that breakage obeys the firs& action at the mill exit, pi = Wi and eqn. (2)
order hypothesis becomes
rate of breakage of size i = SiWiR’ (1) i-l
pi = + T C bi$ijp~
where W is the mass of coal acted on by the i=1
mill, Wi is the fraction of the mass of size i i>l
and Si is the specific rate of breakage of size i, l=Gi<n (W
time-‘_ It will also be assumed that primary
breakage of size j produces a suite of daughter where -r is the mean residence time defined by
fragments described by the set, of numbers b,, 7 = W/F. To modify this to allow for air-
where b, is the fraction of material which is sweeping, the left-hand side of the mass-rate
of smaller size i after primary breakage of balance becomes
Iarger size i (i > i); note that
rate of removal of size i in the air stream
+ rate of removal of size i in the powder
5 b, =1
i=n
flow

’ The size distributions produced from a Let the rate of removal of size i by air-
tumbling ball mill also depend on the resi- sweeping be ri, then the total rate of removal
dence time distribution (RTD) of material in of powder by air-sweeping is R = Zyri and
the mill. A particularly convenient method of eqn. (2) becomes
allowing for the RTD is to consider the mill
as a series of fully mixed stages. For the i-l

purpose of this paper, equal sizes of the stages ri +Pi(F- R) = fiF + W C bi$jWi -SiwiW
i=l
are assumed, but the equations developed can i>l
be readily extended to stages of unequal size_ (3)
Figure 1 illustrates the system and nomen-
clature. It is assumed that she powder acted on by
the grinding action is the flowing powder and
that the suspended powder is not subject to
breakage action; then, Wi =pi_ It is also
assumed that a fraction 77of the powder maSS
W is exposed to the air stream by tumbling,
per mill revolution. For a mill rotation speed
o, the amount of size i material in W exposed
Fig_ 1. Illustration of air-swept mill as a series of per unit time is pivOW_ It will be assumed
stages_
that the exposed material of size i has a
probability Ci of being retained in the hold-up
Austin and Gardner [6] have given the and hence the fraction swept up will be
mass-rate balance of a single fully mixed mill 1 -Ci or
operating at steady state as
ri = pi??0 W(1 - Ci) (4)
rate of removal of size i material in product
Inserting into eqn. (3) and rearranging using
= rate of addition of size i material in feed
7 = W/F,
+ rate of production of size i material by
breakage of all larger sizes - rate of i-l
breakage of size i material fi + TC b&jpi
j=1
or, using eqn. (l), l<i<n
pi = (1 - E) + Si7 + qO(l - Ci)7
i-l (5)
PiF = fiF + W C b,SjWi - SiWi W (2)
i=1 where
i>l

where F is the feed rate and pi the product E = R/F = n)w-&l - cj)pj (54
size distribution. If there is no classification 1
E is thus the fraction of the feed rate removed - - T/m = 7/m
in the air stream; if the RTD of the powder Is2- (1 _Ei)(l --El) --- (1 -Eg_,)
flow in the mill corresponded to one fully-
mixed reactor, E = 1_ l<k<m @a)
The treatment is extended to m fully
mixed stages in series as follows (see Fig_ l)_ il = 7/m

A mass-rate balance on the second stage is where -r is the overall mean residence time
identical except that the product piF(1 - E) is defined by IV/F and
the feed into the second stage. Since it is
known that the hold-up level along this type lGk<m (6b)
Ek = Tk,&l - ‘+)&.k
of mill does not change very much, it can be 1
assumed that the hold-up is equal in each
stage, at W/m_ For the relatively short mills There is no powder flow out of the mth
used in coal grinding, the balls are fully mixed reactor, so E, = l_ For a given set of 1 - ci
throughout, SO the same values of Si and b, vdue~, and given 77, W, w and fi, the value of
can be used in each section_ It will also be F must be such as to lead to E, = l_ The size
assumed that particles in the air stream are distribution of the air-swept product is
rapidly swept out and do not fall back into
the bed of powder. Then, for the first stage
or
i-l
fi * 71 C bi+jPi. 1
i=1 pi’ = wyCi) 2 pi_k
l=si<n (7)
Pi-l= t1 k=1
-El) + SiTI + qO(l -Ci)Tl

l<i=Gn If the values of n and Sj are assumed

constant along the mill, a mass balance on size
i over the total mill gives

E 1= r1T-J i: (1 --jlPi 1
gri_,=fiF + ‘i ‘g’ bijSjgpi_&
1 ?7lj=, 1
- si
W m
- x:Pi_ L_
1
m 1
i=1 i>l

For the second stage or

i-l i-l

Pi.1 +~2 C bijSjPj.2

pi = TJwFV(I - ci jwi = fiF + tV 3 bijSjU?j -
i=l j=1
i>l
piy2 = (1 - fZ*) + Sira + 7jW(l -Ci)T*
- Si WWi (8)
ldidn
where
r2 = (W/m)/(F -R,) = 711(1 - EI) m
wi=xPi.k m
1 I

E2 = T27& (1 - CilPi.2 isthe mean size distribution in the mill, ri is

1
the total flow of size i in the air stream and
IV is the total mass of powder in the mill_
In general [ 73, then,
Thus, since r = TV/F,
i-l
i-l
Pi. k - 1 -t Tk c b$$P~.k
i=1 fi -i- i C bi$TiWj
i=l
Pivk = (1 - ek) + Sirk -I- n~(l -Ci)Ti; Wi = l=Gi<n (9)
SiT ~ ‘I;lO(l -Cj)T
l<ibn
lCk=Gm (6) and
 258

Pi ’ = mW(l - Ci)Wi ldi<n (IO) fixed, the coal feed rate follows from F = llr,-
Because the values of Ci vary with air flow
The conclusion is thus reached that for these rates, it is necessary in the simulation to: (a)
assumptions the mill behaves as if it were a specify the air-flow rate; (b) calculate Xi, 7,
single fully mixed mill containing a mean pi’; (c) calculate F for a correct value of W;
overall size distribution. Since the flow (d) repeat for different values of air flow rate
rate of material presented to the internal until the desired value of F is obtained by
air classification is VOW, and the rate of trial-anderror search or interpolation.
fine material leaving is QWWCi(l -Ci)UJi.
the equivalent internal circulating load is
l/Xi(l -Ci)Wi- DETERMINATION OF BREAKAGE PARAMETERS
4ND SCALE-UP RELATIONS

DISCUSSION OF MODEL Laboratory ball mill test

In addition to the air-sweeping action, ball
In this type of system, a large particle in mills grinding coal have another characteristic
the feed entering the mill must stay in the which is somewhat different from the usual
mill as long as necessary to break it down in a ball milling of mineral rocks. The coal feed to
z series of steps to sizes small enough to be the air-swept ball is typically a crushed feed
swept out. A large particle which moves to with a top size of about 1 in, as shown in
the discharge end of the mill will strike the Fig. 2. The presence of appreciable quantities
end wall and back-mis into the mill contents. of such large sizes means that care must be
On the other hand, a small feed particle can taken to describe the breakage of these large
soon be swept out. The concept of residence sizes with reasonable accuracy_ The descrip-
time distribution starts to lose meaning under tion of the breakage of small sizes which are
these circumstances_ The assumption that nipped and crushed in a normal breakage by
particles will not fall back into the bed once the balls has been well studied 15, S], but the
they are entiained is clearly an over-simpli- kinetics of breakage of larger sizes have not
fication which cannot apply to a long mill been so fully investigated [ 9] _
[ 73 _ However, the experimental measurement The normal values of S and B were mea-
of size distributions along the axis of a pilot- sured on a coal g-round dry in a cylindrical
scale mill (see later) showed that there was steel laboratory ball mill of 195 mm diameter
only a minor variation along the mill. In and 5.22 litre volume- The ball charge used
addition, the treatment of the closed cir- was 4810 g of 1 in dia. steel balls, correspond-
cuit system (Appendix 1) shows that the
effective overall classifier action is given by
(1 - Ci)(l -.Q) where 1 -si is the fraction w P/CIRCUIT PRODUCT
returned by the external classifier. Examina- N
L
tion of the classifier data show that terms in
r.0

Si dominate over those in Ciy SO that the final

result of closed-circuit operation is not highly
sensitive to the values of ci_
For these reasons, the simple model of a
fully mixed mill was used, with the values of
Ci being mean effective values over the mill as
a whole_ This gives a particularly simple
1290 ACFH
system for experimental analysis. The com- 970 IbS~h.
putational procedure is given in Appendix 1. c =,‘I8

For ~IIOWII VdUeS of Si, bif, Ci and T)O, there

is a unique value of 7. In practice, the air flow
rate is usually adjusted to match the coal
feed rate so that a correct filling level of Fig. 2. Experimental size distributions for coal
powder is obtained in the mill, using the ground in 0.98 rn I.D. pilot-scale air-swept ball mill
sound of the mill as a guide. Since IV is thus fitted with a twincone classifier_
 259

P
ing to a fractional filling of the mill volume
by the bed of balls of J = 0.2, using a formal (12)
porosity of 0.4 for the ball bed (290 lb/ft3).
The dry mass of coal used was 122 g; assum- It was found that the cumulative daughter
ing an apparent specific gravity of coal of fragment distribution Bii was normalized
1.3 and a formal porosity of O-4, the formal [12] for the coal (that is, 4:_ = a constant for
bulk density of coal is 770 kg/m3, giving different breaking size j), and ‘1~~= O-36(5),
a fractional filling of the mill volume of y = O-90, p = 2.8. The values of S and B are
f, = 0.03. The experimental methods have shown in Figs. 3 and 4_
been described in detail elsewhere 110 J_ The experimental values of the normal
For particles which are small with respect values of Si can be converted to other powder
to the ball size, it has been fGund that the
values of Si vary with particle size -Xi, ball
diameter d and mill diameter D according to
simple power functions
Si(d) a XiQDNIjdNo .W)
where Ni is close to O-5, and (Y is characteristic
of the material, 0.83 for this coal. The value
of iv0 is still in question due to conflicting
results of various tests but was taken [ll]
as l-O_
In the usual way [ 123, the primary daugh-
ter fragment distribution calculated using the
BII method was expressed in cumulative form
s oo~L_-__ ___._-_L;1__ _. _ .~_ ~. _.

002 or :j
Bii = 2 b,i i>j RELATIVE SlZE xi /xi
k=n
Fig. 4. Cumulative daughter fragment distribution of
and the results fitted to Belle Ayre coal (d2 inierval).

0=195mm/O
4
.I =0.20 P
I
f,= 0.03 i
d =25mm
3

2
A i
I I I1ll‘l 1 rrr,.,l I t I.,.

0.1 I IO loo

PARTICLE SIEVE SIZE, mm

Fig. 3. S values used in the simulations: Belle Ayre coal (d2 size interval, values plotted at upper size of intervai).
260

and ball filling conditions using the empirical 6 X 8 was prepared and 75 kg (dry basis,
equation developed by Shoji et OZ.[S]: moisture content 15%) was loaded into the
1 mill. The mill was run for 112 min, stopped. a
si = exp(-1.2U) (13) sample of about 500 g taken by random
1 -i 6.652.3 scoop sampling, and the mill run for addi-
where U is interstitial filling of balls defined tional grinding times, with sampling for each
by tl = f,/O.4J. time, to give grinding periods of l/Z, 1 l/2,
For larger sizes, the power function of 3,4, 5 and 16 min. The size distributions at
eqn. 11 is modified [5] by an empirical each time were determined by screening of a
function Q(w) which makes the values of Si measured amount of dried coal, followed by
pass through a maximum and decrease at large redrying and weighing: weight loss was
sizes : assumed to be material of minus 270 mesh.
D NIXiQ The results are shown in Fig. 5.
S,(d) a N, Q(x) (14) The first-order plot for the top size interval
d is shown in Fig. 6. The mean value of gI
where calculated from the slope of the curve was
1 4.0 min- ‘_ Figure 5 shows the results of
Q(x)= A> 0 (15) simulation of this batch grind using the
1 + (xi/P)* laboratory S parameters scaled to this mill
The value of ~1is the size at which Q(x) = 0.5 diameter and ball mixture and assuming the
and A is an index of how steeply the values same normalized B values, using exponents
fall with increasing particle size. The experi- N, = 0.5, N2 = 0.2. It was clear that the break-
mentally determined values of M and A for the age in the pilot mill gave the same basic shape
test coal were ~1= 3.7 mm, A = 2.75 for of the size distributions as those in the
1 in balls in the laboratory mill. It was 195 mm dia. laboratory mill and the simula-
assumed that u varies with ball diameter and tion was satisfactory.
mill diameter according to For grind times beyond 5 min, the slowing
p a d*Dh’z down of breakage rate observed with small
(16)
mills was also seen in the pilot mill results. In
where Nz is about 0.2. This espression allows addition, a similar batch test in the pilot mill
for the effect of ball diameter and mili on the same coal without prior air-drying to
diameter on the position of the maximum in 15% moisture gave a significantly different
S values, via P, assuming A to be constant_ result, especially for breakage of the larger
The overall values of S for the mixture of sizes. The coal used in these tests was a sub-
ball sizes in the mill were assumed to be the bituminous coal (Belle _4yre South) with an
weighted sums of the S values for each bail as-received moisture content of 30%. It is
size, clear that for this type of coal it is necessary
that the laboratory batch test be performed at
Z?i= 5 mhSi(d,) moisture conditions corresponding to those in
(17)
k=1 the continuous test. The mechar.isms by
which moisture content affects the breakage
where ml is the weight fraction of balls of of larger sizes remains to be investigated-
d, = 1 in diameter, m2 that of d2 = 1 l/4 in,
m3 that of d, = 1 l/2 in, etc., (m, = 0.25, Pilot mill tests: Continuous grinding
m2 = see
mill Fig m
O-375, 3; = 0.375 for the pilot-scale The test mill was a cylindrical air-swept
, - . mill of 0.98 m mean ID_ by 1.53 m long,
with inlet and discharge trunnions of approxi-
Pilot mill tests: Batch grinding mately 0.3 m dia. The ball load corresponded
Iu order to test the scale-up relations of to 24% mill filling (J = O-24), with a ball mix
eqn. (ll), a pilot-scale mill of 0.98 I.D. by of 25% 1 in diameter, 37.5% of 1 l/4 in, and
1.53 m long was run as a batch mill with the 37-5% of 1 l/2 in. The mill was run in closed
same filling conditions as for continuous circuit at steady state, at 80% of critical
operation (see next section). A feed of coal speed, with the feed rate adjusted by an
predominantly in the size intervals 4 X 6 and experienced operator to give the correct
 MINUTE

3
4

-I
-I’! 1-L_I-I_L_I-lI- -_LLl-d
100 1000 loooo

SIEVE SIZE, pm
Fig. 5. Rcsulls of batch grinding 75 kg Belle Ayrc coal (J = 0.25, fc = 0.083, moisture contrn~ of cod = I 5%) in
pilot-scale mill.

suddenly and simultaneously stopped, and the

mill came to rest in less than one-third of a
revolution_ The contents were tipped out.
weighed and mixed, and sampies taken for
size analysis. Four tests at various coal flow
rates gave mill hold-ups of 157, 160, 157 and
175 lb of coal (dry basis), that is, a value of
about 160 lb. This is a fractional formal mill
filling of f, = 0.081, which is a fractional
filling of the ball interstices of U = 0.83.
Knowing the steady mass flow rate to the
classifier, the mass flow rate of recycle, and
the recycle and product size distributions, the
size distribution of the mill discharge, pi’, is
readiIy calculated. r is calculated from the
mill feed rate and hold-up (both dry basis)_
Thus the values of qw(l -cci) in eqn. (10)
Fig. 6. First-order plot for the 4 x 6 mesh top size in (pi’/TUri = 7)0(1 - Ci)) GUI be calculated_ It
the batch pilot mill. was found that these values became constant
for sizes smaller than about 60 pm. Since the
air velocity in the open space in the mill is
sound level. The mass flow rate of recycle about 1.2 m/s, it can be assumed that ci is
from the classifier was measured by diverting zero for these small sizes. The constant value
the flow to a collecting bin for a short time- is thus VW and since w is known (34.6 rpm),
interval, and the make-up feed rate was the value of r/ for the test conditions follows_
measured by a calibrated, variable-speed belt The results are shown in Fig_ 7_ The effect
feeder_ The exit air temperature was close of air flow rate on 71is given by
to 150 OF, ensuring coal of low free-moisture
content. The test system is described in detail 7) = 6.2 X 104Vo-s (18)
in another place [13 J_ Samples were taken for that is
feed, recycle and product size analysis The
mill power, feed and air flow rate were then q = 1.2 X lo-‘(u, m/s)“-s (I8a)
262

UPPER SIEVE SIZE. ym

1
[
gi
gr : i
0 &
2% -! 22 20 IS 16 14 12 10 8 6 4
ILlI.
0001i 0 1 ,,:.,i 1 :
SIZE INTERVAL NUMBER
IO 50 loo 400
Fig_ 8. Smoothed values of internal c?assification
APPARENT SPACE VELOCITY. V miri’
numbers cc (for d2 intervals. plotted at upper size)_
Fig. 7_ Value of 7 for different fiow rates of air
through a 0.95 m 1-D. by l-5 m long airswept ball
mill. (see Appendix 2), giving the results shown as
broken lines in Fig. 8. _4lthough this treat-
where V is the apparent space velocity (vol- ment is not very sophisticated, it gives results
ume changes per minute, based on an empty which are consistent with the esperimental
mill fraction of 1 -J) and u is velocity data and which aBow interpolation and
through the mill based on a cross-section of extrapolation in a logical manner. In addition,
1 -J_ It is concluded that the fraction of the the external classification values dominate
mill powder which could be swept up in the over those in ci, so that the final result of
air stream per mill revolution is only 1 to 2% closed circuit operation is not highly sensitive
of the hold-up and that it increases with air to the values of ci_
velocity_ The actual amount swept out of the
mill depends, of course, on the values of
ci and Wi- Since ~wW is the mass iate of
material exposed to the internal air-sweeping COMPARISON OF SIMULATIONS AND PILOT
classification action, and F is the mill mass MILL DATA
flow rate, the air-sweeping action gives an
apparent internal circulating load of The model discussed above was pro-
grammed and simulations performed for coal
1 -L C’ = QoWJF = ~WT (19) and air flow rates corresponding to contin-
where C’ is the apparent internal circulation uous tests run in the pilot mill_ The samples
ratio- For a value of i- of 5 min and q = 1% of make-up feed and recycle in the contin-
per revolution, this gives C” = 0.75. uous tests enabled the size distribution of the
Since q is now known, the value of ci can mill feed to be calculated_ These were used as
be calculated from the values of qo(l - ci)- feed to the mill simulation (see Table l),
The smoothed set of cc values for an air flow giving the typical result shown in Fig. 9. The
rate of 1290 actual cubic feet per minute simulation correctly predicts the long plateau
(ACFM) at the exit temperature of 150 “F in the size distribution of the powder in the
are shown in Fig. 8. Particles larger than mill, which is due to the shape of the feed
about 2000 pm fall back into the bed and size distribution and due to the larger feed
cannot leave the mill- The individual values sizes being to the right of the maximum of S
of Ci for different air flow rates were too values, as shown in Fig_ 3. The mill products
scattered to give a definite relation to air flow shown are passed through a classifier to
rate- However, it. appeared that the curves prepare the standard pulverized coal grind, see
were of the same shape but shifted along the Fig. 2, and the mill feed shown is, of course,
size scale to larger sizes as flow rate increased. made up of fresh feed and recycle. Table 2
The change of ci with air velocity was esti- shows the comparison between simulated
mated by a simple settling velocity treatment 0erssu.sexperimental results for two air-coal
TABLE 1
Data used in simulation: air flow rate = 1290 ACFM
(0.61 ma/s. u = 1.04 m/s).

Interval Upper Feed size cf Si

number i size of distribution (mm-‘)
interval (% < size)
Wm)

1 38100 100 1-o 0.77

2 26900 99.15 1.0 1.71
3 19000 9519 1-o 2.98
4 12500 92-56 1-c; 4.18
5 9510 85.77 1-o 4.86
6 6730 83.86 1.0 4.93
7 4760 SO-31 1.0 4.26
8 3360 75.37 1.0 3.36
9 2350 72.54 1-o 2.68
Fig. 9. Comparison of experimental results and
10 1680 69.72 0.99 2.03
computed results; mill hold-up 160 ib (73 kg), air
11 1190 67-00 0.96 l-52
flow rate 1290 ACFM (6.61 m3/s) at 150 ‘F (66 ‘C).
12 841 64.86 0.92 1.14
13 595 63-19 0.85 0.86
14 420 61.04 0.76 0.64
15 297 5756 O-66 O-48 by-pass fraction {O =Ga G LO), dgn is for the
16 210 52.32 O-54 0.36 classifier curve, and h is related to Sharpness
17 149 44.22 0.42 0.27 Index, SI, by In SI = -2_1972/h. The result-
18 105 32.75 O-30 O-20
ing values of si were used in t-he closed-circuit
19 74 22.32 0.17 0.15
20 53 16.01 0.0 0.11 simulation model to give product size distri-
21 38 11.00 0.0 O-09 butions and capacities for comparison with
22 26 7.80 o-0 0 the experimental results, as shown in Figs.
10 and 11. It was concluded that the simula-
tion model predicted the correct mill capacity
rates. The simulations are in reasonable and fineness of product wit.hin the scatter of
agreement with the experimental results. the experimental results.
To obtain an indication of the level of In t.his particular pilot mill, the drive and
experimental variability of the pilot-scaIe bearing friction losses were 5.9 kW and the
data, a series of closedcircuit tests on the net mill power was 7.5 kK;, giving a total
same coal was performed at varying flow motor power of 16.2 kW, as the motor
rates_ The size distributions around the twin- efficiency was 82% The fan power at 1500
cone classifier were used to ca.Iculate si values ACF’M was 2-6 kW and at 2000 ACF41 was
fitted by the form [14] 6 kW_ It is clear that fan power is significant
1 compared with net mill power. The vaiue of
&=a+(l--a) (191 J = O-24 is lower than in conventional non-
1 + (X</dsc)-” I swept ball milling because the trunnions must
where the classification action is defined by be large enough to pass air without excessive
the three parameters of a, dsO and A: a is the pressure drop and fan power.

TABLE 2
Comparison of simulated uerszzsexperiment for continuous pilotscale mill tests

Air flow Mill feed rate (d-b.) AIill product

(ACFhI) Ublh )
(% < 200 mesh) (‘% c 250 mesh)
Exp. Sim.
Exp_ Sim. Esp. Sim_

1 890 1680 1685 64-S 63.1 50.7 48.7

2 1290 2380 2385 55-4 52.9 41.6 39.6
264

given by eqn. (13), based on results from a

laboratory mill of 195 mm I.D. Scale-up of
experimental breakage rates from a 195 mm
dia. laboratory mill gave values which, when
used in the air-swept simulation model,
correctly predicted the size distributions and
mill capacity for the pilot-scale mill, within
the experimental scatter of the pilot-scale mill
results
i - SIMULATION

f5
iI
ACKNOWLEDGEMENTS

CIRCUIT PRODUCT RATE, Ibs/hr This work was supported by research

Fig. 10.Comparison of experimental and predicted
contract No_ EX-76-C-01-2475 from the
fineness of grinding for Belie Ayre cod ground
closedcircuit in the pilots&e mill. U-S. Department of Energy, contract monitor
Mr. T. K. Lau_ A number of people have been
involved in the test work on which the results
are based, but we wish to thank especially
Mr. Ken Gardner, Chief Engineer, and Mr-
James Wilver, Manager of Testing Services, the
Kennedy Van Saun Corporation, Danville,
Pa_, U.S.A.

1400 Iscm laoo 2000 REFERENCES

AIR FLOW RATE. ACFM

Fig. 11. Circuit capacity versus air flow rate for Belle L. G. Austin and P_ Bagga, Powder TechnoL. 28
Ayre coal ground in the pilot-scale mill. (1981) 83.
L. G. Austin, J_ Shah. J. Wang, E. Gallagher and
P. T_ Luskie, Powder TechnoL. 29 (1981) 263.
Appendix -42 gives a method of scaling the L. G. Austin, Powder TechnoL. 5 (1971/72) 1.
internal classification to larger mills_ At the J. A_ Herbst and D_ W_ Fuerstenau, Inf. J_ of Min.
moment, however, it is not possible to scale- Proc. 7 (1980) 1.
up the total model to larger mills because the L. G. Austin, R R Khmpel, P. T_ Luckie and
variation of 71 with mill diameter is not R. S. C. Rogers, Symp. on MiZZing, AIME, Hawaii,
Sept_ 1982. pp_ 301 - 324_
kIlOWIl.
6 L. G. Austin and R. P. Gardner, in H. Rumpf and
D. Behrens (Eds.), Proc 1st European Symposium
in Size Reduction, Verlag Chemie, Weinheim.
CONCLUSIONS 1962, pp_ 232 - 248_
7 L. G. Austin, N. P_ Weymont and 0. Knobloch,
in K_ Schonert, W. Gregor and F. Hofmann
If it is assumed that the grinding conditions (Eds_), Preprints. Particle Technology 1980,
are constant along the mill, an air-swept mill Dechema, Vol_ B (1980) 640 - 655_
in which all of the mill product leaves in the 8 K_ Shoji, L. G_ Austin, F_ Smaila, K_ Brame and
air stream can be treated as a single fully P_ T. Luckie, Powder TechnoL. 31 (1982) 121.
9 T. Miles, I. Shah and L. G. Austin, Breakage Rates
mixed mill. The results from a 0.98 m diam-
and Breakage Dish-butions of Large Particles in
eter by 1.53 m long pilot-scale mill showed Ball Mills. in preparation.
that the percentage of mill hold-up exposed 10 L. G. Austin and V_ K. Bhatia, Powder TechnoL.
to the air-sweeping action per revolution of 5 (1972) 261.
the mill was about 1 to 2%, at a ball fiiing of 11 F. Smaila, K_ Brame, L. G. Austin and P_ T. Luckie,
24%, an optimum powder filling of 8%, and The Effect of Ball Diameter on Breakage Rates
and Breakage Distributions in a Laboratory Mill,
80% of critical speed. The optimum powder
in preparation_
filling represents a value of U = 0.83. This can 12 L. G. Austin and P_ T. Luckie, Powder TechnoL.
be compared with the optimum value of 0.84 5 (1972) 215
 . 265

13 P. T. Luckie and L. G. Austin, Coal Grinding size distribution to the mill is related to the
Technology: A Monuol for Process Engineers. make-up and recycle by fiF = giQ + tiT where
U.S. Department of Energy, FE-2475-25, avail- F = Q + T, Q being the make-up feed rate and
able National Technical Information Service,
Springfield, Va_. (1979) p_ 92.
T the recycle rate_ Because the circulation
14 L. G. Austin and R. R. Klimpel, Powder TechnoL. ratio is defined as C = T/Q, then 1 + C = F/Q_
29 (1981) 277. Since si is defined as the fraction of Fpi’
which is recycled, that is, tiT = FSipi’, then
fi(l + C) =gi + (1 + C)sipi’- Substituting into
APPENDIX 1 I COMPUTATIONAL PROCEDURE eqn. (Al-l) to eliminate fi gives
i--l
As discussed previously, the internal
gi + jTl bijYi*
sweeping action can be treated as an apparent
yi* = lGi<n
classification action- However, the programs
1 + l?O(l -si)(l - cil
we have used previously for mill circuits
[Al] choose a value of mean residence time si (X1-3)
r = F’/W and calculate C’ and, hence, F/W- where yi* = (1 + C)WiSir = WiSi7*- AS before,
For the air-swept system, the values of ci and the y* values are computed sequentially and
q depend on the air flow, so it is more con-
n-l
venient to set the air flow, which then defines G -r- C %Y,*
r_ The computation is performed as follows. n--l i- 1
Equation (9) can be put as 7* = c (Yj*/S,) -I-
i=l r?w(l -s,)(l -Z)
i-1 (Xl_‘Z)
fi + z biirj
i= 1 Then wi = yi*/Sir*, from eqn. (10) pi* =
i>l T*vc;(~ - Ci)Wi, (where pi* = p;‘(l + C)) and
Yi = n> iz= 1 (Al-l)
Bw(l --Ci)
1+
Si l-?-C= gpi*=
i=1
7*wJ
i=1
2 Wj(1 - Cj) (X1-5)
where Yi = WiSir_ The Y ~23~2s are calculated
sequentially starting at i = I, for known values Knowing C, the values of pi’ follow. The size
of b,, Si, Ci and VW_ (For example, the value distribution of the circuit product is given by
of Ci will probably be 1 for size 1, SO y1 = fI_)
4i = (1 + C)pi’(l -St) (_41_6)
Then, since CiWi = 1,
The circuit capacity Q is W/r*-
n-1
It is useful to note that the closed circuit
behaves like the open-circuit model but with
the value of 1 - ci replaced with an effective
Note that Y,, = 0 because S, = 0, but w,r is value of (1 - ci)(l -si). The computation
not zero_ It is given by program for closed circuit can thus be used
for open circuit by setting Si = 0 and gi = fi-

APPENDIX 2: VARIATION OF INTERNAL CLAS-

SIFIC-TION WITH AIR R_TE
where C, is the mean value of c for the sink
interval (fortunately, z = 0 in most cases)_ The concept used is that a particle of a
Thus, there is a unique value of T_ The values given size (and density) ent.rained in the air
of wi follow from Wi = Yi/SiT and eqn. (10) stream will follow a falling path determined
gives the product size distribution. by its starting position, settling velocity and
When the mill circuit is closed via an the time for the air to reach the end of the
external classifier, it can be assumed that the mill. If the path does not reach the bed before
external classification action can be described the particle is swept to the end of the mill, it
by a set of classifier numbers Si- The make-up will leave in the air stream- When the velocity
feed size distribution is gi and the actual feed of air flow through the mill is increased, an
266 .

entrained particle wilI follow the same path The equation is arranged in this form so that
and have the same probability of leaving if its setting a value for r’ enables the calculation
settling velocity is higher than the original of r for a given u’/u. The value of c for this
particle, so that its time for settling is de- value 0% r can be interpolated from an esperi-
creased in the exact ratio that the time for mental Ci versus ri CUNe.
the air to reach the end of the mill is de- In the case given here, u a 0.61 m3/s for
creased. Thus, as air velocity is increased the the ci curve of Fig. 8 (at 1290 ACFM). This
size of particle which leaves the mill by a was fitted by the empirical curve
particular path also increases.
In quantitative form, the treatment is as (A2.3)
follows. For spheres of 40 to 6000 pm, the
variation of settling velocity us in air with where x was in ym, that is, _r = 2r X 104. Thus,
particle radius r given by Fuchs [A23 was for LX’ in the desired t/2 sequence, eqn.
converted to the air viscosity and density at (A2.2) allows the calculation of x, and
150 ?F, corresponding to typical air tem- eqn. (A2.3) then gives ci-
peratures in the mill_ The result was fitted by With respect to mill diameter D, the same
the empirical expression set of geometrically similar settling paths
would be espected for the same L/D at the
log u, = 4-352 - 0.796f”-*” (A2_1)
same gas flow velocity. The ratio of LL’to u is
where r is in cm, ZJ,in cm/s. At a given air- calculated from volume flow rates V by
sweeping veiocity u in the mill, the experi- V-‘/(D’)7
mental value of ci represents the mean effect (A2_4)
of all possible paths for powder of radius ri l’r’i’
= VW*
with a flow time for air in the mill of t. At If the c relation of eqn. (A2.3) is to be used,
anjr other flow rate L(‘, t.he time is L-’= h/u’. V = 0.61 m3/s and D = 0.98 m, for L/D = 1.5,
In order for particles of radius r’ to go that is, u = 1.04 m/s.
through the same set of paths to give the
same ci, the settling velocity must increase REFERENCES
proportionally to l/C. Thus, u,‘t’ = u,t and
Al P. T. Luckie and L. G. Austin, Min. Sci. and
v,‘Iv, = u’ju. Using eqn. (A2.1), Engng.. 4 (1972) 24.
A2 N. A. Fuchs, The Mechanics of Aerosols, Mac-
J-= l/[ ($)-O-24 + 1.256 log(u’/u)]‘-” (A2.2) millan, New York, 1961.