Está en la página 1de 8

L G. AUSTIN, P. T.

LUCKIE
The Pennsylrunia State Univemity. Mineml Processing Section. Univwsity Park PA I6802 (USA-1

and K SHOJI
Kennedy Van Saun Corpomtion, DanviRe. PA 17821 (IISA.)

(Received October X3,1981)

SUMMARY

The scale-up of breakage parameters


measured in the Hardmve mill to large mill
diameters, using a simulation model for
continuous milling, is discussed_ Mill capa-
city for race diameter D in meters is

Q = (1.33 X 104)Dz5Qi

where QH is mill capacity determined by the Fig_ 1. Illustration of the characteristic dimensions
model for the Hardgrove mill. The total mill of the baIkace mill_
power at normal grinding conditions for coal
is
for the El-7 mill, providing the mill model
m, = (l/lbn)29_5Dz5, kW incorporated internal and extamal clessifica-
tion actions, and providing the variation of
where qm is motor efficiency_ The rvlriations specific breakage rates with mass throughput
in fineness of grind with mill capacity, Hard- was incorporated into the computations_
grove Grindability Index, classifier setting, In this final paper, the simulation model
and ai~oal ratio are described_ Results from is used to predict product size distributions
an El. 7 mill agree with model predictions_ for a range of coals, and the scale-up of mill
capacity to larger mill diameters is discussed.

‘INTRODUCTION
THEORY OF THE SCALE-UP OF BREAJSAGE
In the first paper
in this series [l] , results RATES
were given on the breakage parameters of
coals, and on mill power, as a function of It is assumed that the geometry of the
mill conditions for the laboratory Hard- mills is constant, so that all dimensions
grove mill In part II of the series [Z] , results increase in equal ratio_ In this case, the mean
from a 430 mm (17 in_) dia pilot-scale ball- race diameter D is an index to the size of
and-race mill (Babcock El.7 mill) were used the mill. Figure I shows an idealized picture
to obtain scale-up factors from the Hardgrove of the mill geometry_ Although the coal
results, using a simulation model for the accelerates as it moves from the center of the
mill. It was concluded that the simulation rotating table radially to the race, it can be
ba&d on Hardgrove breakage data gave assumed that the radial velocity through the
approximately the correct size distributions actual race, vo, is approximately constant,

0 Ekevier Sequoia/Priited in the Netherlands


128

with a mean bed depth of coal in the race of different sizes. If the material being acted
6. Let the number of balls be n and let the upon were all of a given size (defined as at/2
width of the race be L. Let the coal take a size interval, see previous papers [l, 23)
mean time of At to pass through the race, the rate of breakage is known as the abso-
At = L/v,_ The fraction of the race area cov- lute rate of breakage of this size. Thus,
ered by rolling balls per unit time is nw,
where w is the rotational speed, in rotations absolute rate of breakage of size i, Ai 0~MDw
per unit time. (2)
At low w, material on a rotating disc will The application of these concepts to batch
not move to the periphery because the centri- grinding of a fixed mass in the race gives Ai =
fugal force does not exceed the force of fric- (M/nLG)(nw)(lrDLG), that is,
tion between the material and the surface of
the disc, or the internal friction of powder on Ai m MUD- @a)
powder_ As speed is increased, the material
will start to slip toward tbe periphery- The results for the tests in the Hardgrove mill
Because the centrifugal force increases rapidly [l] did, indeed, show that Ai m M and Ai 0~
with r (force per unit mass = w?r), a rotation w, with considerable accuracy.
speed which moves material slowly at the The total effect of the individual rates of
central feed can cause high velocities at the breakage of each size interval are contained
periphery by acceleration as the material within the simulation model [2] _ The
moves along the radius. Consider particles methods of solution using Hardgro~e Ai
at the feed region moving along a radius with values have been given [Z] , and the result
velocity u(r). The accekration along r is is a Qn which is related to the real mill capa-
city Q by
o,+ = -=--
du(r) dW) dr
dt dr dt Q=KQa (3)
and since u(r) = dr/dt , w*r = v(r)dv/dr.
Separating and integrating, assuming u(r) where the scale-up factor K is proportional
is small at small values of r, gives to Ai_ Comb;king with eqn. (2) gives

UD=wR=wD/2 (1) K = MDwl(MD~)ardzrove (4)

The length of time spent in the crushing where MDw for the Hardgrove mill under
region is L/v,, so the number of times the standard test conditions = 434 N m (320 ftlbf-
material is rolled over is 2nL/D_ For the per min).
assumption of geometric similarity L = D, The determination of R is not quite as
so it is concluded that the number of crushing simple as eqn. (4) indicates, because the tests
actions as the material passes through the race on the pilot-scale mill [2] showed that the
is constant irrespective of w. power draw by the mill and the effectiveness
The severity of tbe crushing actions on of the crushing action in the race depend on
each particle as the line contact L of ball-on- the make-up feed rate to the mill, since they
race rolls over the particles is propor%onal to depend on the amount of larger sizes in the
MhLG, where M is the total crushing load on race. not on M and ti alone. It is, of course,
R balls and LS is the material acted upon not possible to use feed sizes in the small
under the line contact_ The rate of mate&l Hardgrove mill comparable with the largest
passing is ?rD&,, which equals 7rD26w/2 feed sizes to a production mill. Therefore,
when + is substituted from eqn_ (I)_. Then, eqn. (4) was modified to
multiplying the material acted upon per unit
time by the- severity of. action times the K = %MDw/434 (5)
number of actions per pass gives the total rate
-of breakage as proportional to (M/~LG)(TD *- for M in newtons, D in meters, o in r.p.m.,
6w)(2nL/D), that is, proportional to MDw; It k, being ti empirical factor to allow for the
is also known [l] -that tie rat&s of -breakage difference _in tbe action of the Hardgrove
undo- identical conditions are different for mill from that of larger mills at maximum
Fig_ 2. The crushing load, rotatiunal speed and mill
shaft power of the Babcock E-type series of ball-race
mills (3).
Fig_ 3_ The capacities of the Babcock E-type mills
for a coal of 50 HGI and a grind of 70% < 200 mesh
c31-

power draw_ The feed sizes to a large mill are


chosen to give good crushing action in the w = SO/;/O, r.p.m. (3)
race_ For coal they are normally a natural
crushed size distribution with a top size of Thus, for this series of mills,
about 30 nm. Thus it can be assumed that
k, is constant in scaling from the Hardgrove Q = (1.14 X 104)koDt5QH_ (9)
mill to all larger mills.
The second paper in this series [2] showed Figure 3 shows the suppliers relation of
how the value of Qn can be calculated for a capacity to mill diameters_ The Dz5 rela-
desired fineness of grind (with all other tion of eqn. (9) is validated.
factors held constant)_ Combining eqns_ (3) By comparing the result of Fig. 3 with the
and (5) gives Qn predicted by the mill simulation using
Ai values for a coal of 50 Hardgrove index,
Q = k,MDwQH/434. (6) and a product specification of 70% minus
200 mesh (75 pm), it was found that k. was
Figure 2 shows the variation of crushing 1.17. Thus, finally,
load M and rotational speed w for a series of
large ball-and-race mills [3] as reported by Q = (1.33 X 104)Dt5Qn (10)
the supplier. The maximum centrifugal force
per unit mass at the race rim, proportional to for this type of mill, D in meters.
&D/2, is held constant. In addition, the
specific crushing load, M/nLD, is held cons-
tant. For geometrically similar mills L = D, THEORY OF THE SCALE-UP OF POWER
so
.. The power required to drive the mill action
M=8W2,1cN (7) is

and m, = (60 X 10-3)@Z~Do, kW (11)


130

for o in r-pm. The coefficient cris an effec-


tive coefficient of friction (dimensionless)
between the balls and the bed, defined by:
force to rotate the bed = m = 2TD, where
‘X’ is the torque_ Because the effectiveness of
the crushing action depends on the make-
up feed entering the race, it is expected that
p will depend on the depth, size distribution,
and type of material in the bed, and this was
Air flow \\;!
demonstrated in tests on the Hardgrove mill ACFM
1500
Cl1 -
Figure 2 shows the shaft power for the 54 \T N
standard grind in the series of large mills 133,
which fits the relation 35a
---
250 .

rn#otal) = 29.50=, kW_ (12) a2

The no-load shaft mill power given by the --- Air/coal = 1.73 Wlb
0
manufacturers for the E-type mill series larger J 20 40 60 80 1
than D = l-5 m (5 ft) increases as Dz5, that %<ZOOM&

is, Fig_ 4. The simulated effect of air flc.w rate on the


capacity and grind of the El.7 pilot mill; Western
m&no-load) = 5_5Dz5, kW Kex:tucky coal, HGI = 55 (ACFM t actual cubic
(12a)
feet per minute); i?is overall circulation ratio.

actual mill power = 24Dz5, kW. U2b)

Combining eqns. (7), (8) and (ll), and


comparing with eqn. (12b) gives a constant
value of P = O-093. Defining a specific energy
of grinding by E = Q/m, eqns. (9) and (11)
show that the specific energy is constant with
mill diameter, for a given value of p, for a
given grind. The power to drive the Hardgrove
mill under the standard conditions for which
Ai values were measured was 2-O W [l] _
Since MDw for the standard Hardgrove mill
on coal is 434 N m/m& this gives a value of
p = 0.088. The close correspondence between
these two values of P is largely coincidental,
because the feed conditions and bed depths
in the two types of .mills are clearly not
comparable_ However, this coincidence does 10 50 100
mean that the specific energy of grinding is size.Sun
approximately the same for the Hardgrove Fig. 5. Variation of size distributions with air flow
miB as for an industrial mill [4] _ rate (see Fig_ 4).
The maximum power for the pilot-plant
E-type mill was9 kW minus a no_load correc-
tion of 2.5 kW_ At conditions corresponding efficient in transferring power to breakage
to the standard grind, with a standard air- action as the larger mills. The total mill power
coal ratio-of I.73 kg/kg, the mill power is is
somewhat less; 8.5 kW. Since MDo = 1.61 >c
-10’: N m/mm, for this-mill, p = 0.071. The. m, = (l/q,)(29.5Dz5), kW (13)
value of k, for this mill was O-46. Thus, the
small El.7 miXdoes not appear to b& as ~ where 71, is the motor efficiency_
as -

0.6 -

Fig.. 6_ Partition curves for a twin-cone classifier: air flow in actual cubic feet per minute (ACF) is: I 1500 ACFM,
q1000 ACFM. A, 500 ACFM_

THE SFFECT OF MILL CONDITIONS of the normal value, the air-coal ratio is 2.42
kg/kg.
ln the second paper in this series 121, the Mill control via pressure drop through the
simulation model was used to show the effect mill works on the principle that the pressure
of varying air flow rate. A typical result is drop is dominated by the weight of the
shown in Figs. 4 and 5, where these computa- column of coal+ above the annulus. Hold-
tions were performed by assuming that the ing pressure drop constant for varying coal
variation of classification action with air flow feed rate ensures that the concentration (kg/
rate [Z] could be extrapolated to lower and m3) of coal in the air above the race is approx-
higher air flow rates: cis the overall circula- imately constant_ However, the change in
tion ratio [2] _ The capacity is higher for classification action caused by lower through-
example 3 than for 1 or 2 (allbeing 70% put will produce a somewhat finer size distri-
minus 200 mesh) because the low feed air/ bution, as shown in Fig. 4. Reducing the coal
coal ratio, in principle, gives higher circulat- feed rate and holding the air flow rate cons-
ing loads and a steeper size distribution, tant produces a finer grind (defined by %
as shown in Fig. 5. However, in practice, it minus 200 mesh), but if the rate is reduced
is only possible to operate with the feed air- below the broken horizontal line in Fig. 4,
coal ratio varied over narrow limits, since the grind becomes coarser again because the
the pressure drop becomes excessive at low decrease in mill power and breakage rate over-
ratios d=le to the large maSS of circulating coal weighs the increase in residence time [ 21.
suspended in the mill. ln addition, the race The adjustment of product fineness is
may choke at high coal flow rates, giving better accomplished by variation of the classi-
unstable mill operation. Figure 4 shows the fier setting. Figure 6 shows the variation of the
locus of operating conditions for a feed air- action of an external separator (which domi-
coal ratio of 1.73 kg/kg, which corresponds nates over the internal classification action)
to normal practice_ Turn down of the mill to as the vane setting is changed_ At 100% open,
50% capacity would give a finer grind and the separator acts somewhat like a simple
higher circulating load if the air-coal ratio expansion separator, with changes in air flow
were kept constant, so the ratio is increased rate giving relatively minor changes in dsO_
to comperzate. If the air is reduced to 70% Reducing the vane opening from 100% gives
132

12 r-

1.0

as

B
g 0.6
B
Iif

0.4

02

0 I 1 I 1 1 I
0 20 40 60 80 100

Fig. 7_ Simulated capadty us. fineness of grin-d for Fig. 8. Variation of standard mill capacity (dry coal
fiied air-coal ratio (1.73 kg/kg), El-7 mill, Westim basis) with coal type and product fiiemssr 50% vane
Kentucky coaI HGI = 55_ setting; air-coal ratio = 1-73 by weight_

more swirl to the air and, hence, the separator THE EFFECT OF COAL TYPE
tends to catch lkger sizes more efficiently
and the dSo of the classification action is In the previous reports [l, 23 it was shown
reduced. However, this also increases the that the characteristic breakage parameters
apparent by-pass of fine material back to the intheHardgrovemillforanumberofcoals
mill contents_ At 50% vane setting, increased varied in a consistent manner with the
air flow rate sweeps out larger particIes. dsO standard Hardgrovegrindabilityindex(HGI),
increases, and the better sweeping action althoughtherewassubstantialscatterinthe
reduces the fraction of trapped fines returned data_ It was also shown that the primary
to the mill_ Thus the effect of air flow on daughterfragmentdistributions denoted by
dsO andon theappzuentby-passfractionact qi werenormalized(thatis, Bki =Bij) and
to compensate oneanotherto somedegree. these values werethe sameinthe E1.7 type
At25%vanesetting,thedifferenceinair velo- millforananthracite.
city from 5OOto1500 ACFMgivesavery Thevalues oftheabsoluterates ofbreak-
pronouncedeffectandthesteepness ofthe agevariedwithsizeintheform
classifiercurveisreduced,showingpoor
classification action_ Figure 7 shows the Ai = aW~Qi (14)
effectofvanesetting onthecapacity-tie where Qi was 1 for the smaller sizes, but fell
distributioncurves. It is clearthatthere is to less than 1 for sizes too large to be
an optimum vanesettingfor ma2dmurn mill nipped by the balls. In the simulations report-
capacity, which corresponds to the ed here it was assumed that the results are
production ofthesteepestproductsizedis- to be used for scaling to large mills with large
tributionforadesired%minus200mesh. balls, with a natural crushed coal feed having
Theeffectofvariationinfeedsizedistri- a.topsizeofabout30mm.Thusthevalues
bution can bestudiedusingthesimulation ofQiweretakenaslforallsises,since the
model, buttheresultsare possibly notcor- large ballsin industrial mills will nip all the
rect because the-efficiencyof the Franz- coal sizes. It was also assumed that E&j
fer of crushing action to the race contents values for all coals remain_the same for dif-
willchangev+ithchangeinmake-up feed ferent sizes of mill and that (Y remains the
size_ same_
HGI 6mczoDm%<2oo
s 40 Q5D a26
60 a91 Q56

t so 1.29 cs3
- SIMULATED
--_ ?AX.C.;K AND

50 I I I
40 50 60 70 80 90 100

HiSI

Fig. 9. Size distributions as a function of coal type Fig_ 10. Comparison of simulation sodel and sup-
(Hardgrove Index) corresponding to conditions of pliers recommended capacity m_ HGI and fineness
Fig_ 8. scaled to the El_7 mill_ of grind. l , 4 I, measured El.7 results.

The values of breakage parameters [2]


were estimated from the relations with HGL
It was assumed that the transfer of crushing
action in huge mihs as a function of mill
capacity follows the same relation as that
measured in the El.7 pilot miII and that the
air-sweeping prevents any cushioning action
due to accumulation of fines [l] . Finally,
it was assumed that the vahres of cIas.sifier
selectivity s would be the same for all clas-
sifier-mill sizes (providing the air-coal ratios C.?pan’ty.
rph

are the same) as those measured for a twin- HGI W%C200 8O%CZCXJ
40 0.51 0.36
cone classifier. With these assumptions, the 60 1.00 0.86
simulation model was used to produce Qn 80 1.40 1.28
values, which can then be scaled to larger
miIi diameters using eqn_ (lo)_
The variation of mili capacity (dry coal
50 ma 5m 1000
basis) with product size and coal HGI is
*i*e pm
shown in Fig. 8, for 50% vane setting and a
Fig. 11. Ideal limiting size distributions.
fixed air-coal ratio. The corresponding size
distributions are shown in Fig. 9. These
results were put in a relative capacity form
for comparison with the suppiiers r&zom- appear to be based on early work by Hard-
mendations [S], as shown in Fig_ 10 where grove as reported by Frisch and Foster [5].
coal of 50 HGI and 70% minus 200 mesh is Figure 11 shows the limiting size distribu-
taken as 100. It is clear that the simulation tions [2], which are the size distributions
model predicts higher capacities for softer predicted if ail material less than a desired
coals than that recommended_ However, the top (cut) size were removed from the crush-
pilot-mill data (El.7 mih) fit the simuIation ing action without further breakage. Compa-
modd, not the recommended vaIues, which rison with the r-fits of Fig. 9 shows that the
134

ideal curves contain substantially less fines quantitative description of the classification
for. a given top size than the real mill condi- actions of the classifier on large mills is not
tions. However, the El.7 pilot mill uses 8.5 available. The classification actions have been
kW of grindirig power at standard‘grinding assumed to be dependent on air flow rate
conditions and the fan power for the air alone, whereas the coal* concentration into
sweeping action is about 1.5 kW, so the air- the classifier may also affect the action. Espe-
sweeping is already co nsuming a significant cially, the variation of breakage rates (of all
fraction of the totsl power requirement_ sixes) with the mixture of feed size (being
Increasing the rate of removal of fines by acted upon) requires investigation to deduce
bigher air-sweeping rates would increase quantitative corrections for this effect.
mill capacity at the expense of higher fan
power.
ACKNOWLEDGEMENTS

CONCLUSIONS This work was supported by research con-


tract No. EX-76-C-01-2475 from the US-
The simulation model is based on breakage Department of Energy, contract monitor Mr.
data measured in the Hardgrove mill, measure- T. K. Lau. Grateful acknowledgement is
ments of the variation of scale-up factor (effi- made to A. Birks, Babcock Product Engineer-
ciency of transfer of the crushing actionj ing, Ltd_ for very helpful discussions.
with make-up feed rate in the El.7 mill for a
natural crushed coal feed, and internal classifi-
cation actions measured on the El.7 mill_ REFERENCES
The model gives valuable insight into the
I, G_ Austin, J_ Shah, J. Wang, E_ Gallagher and
mode of operation of this type of mill and
P. T_ Luck%, Powder TechnoL, 2-9 (1981) 263 -
enables predictions to be made of the effect 275.
of operating variables_ However, work remains I._ G_ Austin, K Shoji and P_ T_ Luckie. Powder
to be done to refine and validate these predic- TechnoL, 33 (1982) 113 - 125.
tions_ It is known that the moisture content Babcock Product Engineering, Ltd-, Crawley, U-K
1976.
of the coal affects mill capacity, but the
H. M. van Seebach, Zem-RaZfz-Gips, 25 (1972).
variation of breakage parameters with 71_
moisture content in the Hardgrove mill has M. Frisch and A. C. Foster, Am. Sot. Test_ Mater..
not been investigated in this study. The PIWC.. 37 (1937) 441- 466.

También podría gustarte