You are on page 1of 8

Chemrcol En~mcerrng

Srience
Printed in Great Britain.

MIXING

Vol. 40. No. 7, pp. IO6 I-1068.

I985

ooos-2509/x5
23.00 + .Oo
Pcrgamon Press Ltd.

CHARACTERISTICS
OF INDUSTRIAL
COLUMNS
G. S. DOBBY*

FLOTATION

and J. A. FINCH

Department of Mining and Metallurgical Engineering, McGill University, 3450 University Street,
Montreal, Quebec, Canada H3A 2A7
(Received

12 December

1982; Revised

January 1984;

accepted

13 April

1984)

columns are counter-current bubble columns that have recentlybeen applied to froth
flotation separation of minerals. A unique feature of a flotation column is the water stream applied at
the top of the column for the purpose of washing entrained nonfloatable particles from the floatable
bubble-particle aggregate, consequently permitting high upgrading in a single separation stage. The

Abstract-Flotation

objective of our work is the development of a scale-up methodology for flotation columns, and as a
first step toward this objective we have measured the dispersion parameters of large industrial columns.
Pulsed tracer tests were performed with both liquid and solid tracers on 0.46 and 0.91 m square columns,
each 13 m high. The liquid axial dispersion coefficient appears to be linearly dependent upon column radial
dimension. As well, it is confirmed that the axial dispersion coefficient of mineral particles is the same as that
of the liquid, and that the residence time of solid particles is unaffected by a bubble swarm.
lNTRODUCHON

An inherent limitation with the flotation of fine particles in conventional froth flotation cells is recovery of
hydrophilic (gangue) particles by mechanical entrainment in the water reporting to the froth[ 11. The
method of minimizing entrainment is to create a 5530
cm thick froth at the slurry surface. The froth permits
the gangue to drain back to the pulp while retaining
the hydrophobic particles which are eventually discharged over the cell lip. This cleaning action is seldom
sufficient and sequential stage flotation is necessary.
The alternative approach to the cleaning of
entrained particles has been introduced with column
flotation. Industrial flotation columns are typically 13
m high by 0.45 to 0.91 m square. A schematic diagram
of a flotation column is shown in Fig. 1. Three distinct
zones are indicated. Feed slurry enters the column at
the top of the recovery zone, 9-l 1 m high, and flows
downward against a counter-current flow of l-2 mm
dia. air bubbles. After collision, particles that are
sufficiently hydrophobic will adhere to the bubbles and
the bubble-particle aggregate will rise into the cleaning zone. The cleaning zone incorporates a unique
feature; water is added from an array of perforated
pipes located about 5-10 below the discharge lip. This
washwater drains downward and gives the cleaning
zone the appearance of a packed bubble bed. The
downward velocity of the washwater is regulated by
the difference between the tailings flowrate and the
feed flowrate (or bias).
Gangue particles that are entrained in bubble
wakes or that shortcircuit from the feed point are
washed back from the cleaning zone to the recovery
*Now at Dept. Metallurgy and MaterialsScience, Univer-

sity of Toronto, Toronto, Canada.

zone. Cleaning by froth drainage is not required and


the sole purpose of a thin froth layer above the
washwater addition point is the transport of concentrate over the column lip. Because of this unique
cleaning action, a flotation column can upgrade a fine
sized concentrate in a single step, where conventional
flotation machines would require several sequential
stages[2,3].
The objective of our work is the development of a
scale-up methodology for flotation columns. An important scale-up factor is the mixing characteristics of
the recovery zone, because the particle collection
process that occurs in this zone is usually the rate
determining step in the overall flotation process.
Particle collection is considered to be first-order with
respect to solids concentration if the bubble surface
is clean. Levenspeil[4] has shown that for a first-order
rate process the conversion within a given sized
reactor can be predicted using the values of the rate
constant and the two reactor mixing parameters:
mean residence time and vessel dispersion number.
As a first step in
our scale-up study the axial
dispersion parameters of both liquid and solids in the
recovery zone of industrial flotation columns have
been determined.
DISPERSION

THEORY

Liquid

For bubble columns operating at relatively low air


flow rates and with small bubbles, which corresponds
to the recovery zone of flotation columns, the plug
flow dispersion model has been shown to provide a
good description of the axial mixing process[5, 61. if
a tracer is injected near the top of a recovery zone of
length L, the mass transport equation which describes

1061

G. S. DOBBY and J. A. FINCH

1062
r--Wash

Concentrate

YW

water

Feedy-1

Fig. 1. A flotation column, showing the three principle


zones: (1) recovery zone; (2) cleaning zone; and (3) thin
froth. Sample points used are indicated by 0.

its concentration at an axial distance from the injection point, x, and time, t, is given by:
a%

ac

ac

ax

at

E,----u---=0

ax2

(1)

where EI is the liquid axial dispersion coefficient and


u, is the interstitial liquid velocity. Radial dispersion
is assumed to be zero, so the mixing conditions can
be quantified by two parameters: the mean liquid
residence time, 7,. and the vessel dispersion number,
N, = E,/u,L.
For a system where end effects are
absent

the

Laplace

transform

of

eqn

(1)

WLI:
F(s)

= exp $
(

1 -

(1 + ~N,T,S))I*

is given

they found k = 1 and n = 0.23. No mention was


made of the mode of bubble generation. Towel1 and
Ackerman [ 111 operated in the concurrent mode, with
both gas and liquid injected through a sparger having
a small number of relatively large holes (0.64 cm), on
columns of 0.4 and 1.07 m dia. and 0 5 us s 0.15 m/s.
They found k = 1.5 and n = 0.5.
A noncircular shape of the column radial section is
a complicating
factor. Alexander and Shah[ 121
showed that a rectangular column 0.076 x 0.23 m in
cross-section
gave a considerably
larger axial
diffusion coefficient than a cylindrical column of
equivalent diameter (diameter of a circle having the
same area as the rectangle). Akita and Yoshida[l3],
working with a 0.15 m square section bubble column,
determined that the liquid phase mass transfer
coefficient was the same as that of a cylindrical
column having a diameter equal to a side of the
square. By extension, this would imply that the liquid
axial dispersion coefficients were also similar.

Solid9

In column flotation it is the dispersion parameters


of the solids, not the liquid, that is of prime importance. Several investigators[6,
14, 151 have suggested
that the axial dispersion coefficient of fine solids in a
bubble column is the same as that of the liquid, i.e.

Em = E,, where E,.,+,is the solids axial dispersion


coefficient in a slurry having a volume fraction solids
4. No
valid

II (2)

where s is the Laplace operator. The corresponding


time domain solution is a unique function of r, and

upper

particle

size, or mass,

for which

this is

has been suggested. The other solids dispersion

parameter, the particle mean residence time, rPpo,


should be a function of the terminal settling velocity of
the particle in the hindered settling up+. For the case of
a descending slurry which is not contacted with gas
bubbles, a measure of T,,+is:

NM:
C(r) =[&Jexp[&-k-q)].

(3)

The effect of various physical and operating parameters upon the liquid axial dispersion coefficient
in bubble columns has been reviewed by Shah, Stiegel
and Sharma[8]. They conclude that for cylindrical
columns E, is essentially independent of liquid velocity and liquid properties such as viscosity, surface
tension, density, etc. In a subsequent
review article,
Shah et al.[9] generalize the effect of column diameter, D, and superficial
gas velocity, vs, upon Et
with E, = Dkvgn, where 1 Sk < 1.5 and 0.3 5 n 5 2.
As the present concern is with flotation columns
having a large radial dimension it is worthwhile to
examine the scarce quantity of published data on
large diameter columns.
Magnussen
and Schumacher [ lo] experimented
with cylindrical columns having 0.04 < D 1. 1 m at
0.01 I vg 5 0.1 m/s while operating in the countercurrent mode. For columns with no internal baffling

The effect of a bubble swarm upon this relationship


is unknown and, to the authors knowledge, unmeasured. It is suggested here that this effect is
gas velocities
minimal
at the relatively low
(0.01-0.03 m/s) employed in a Rotation column, and
that equation 4 is suitable for scale-up purposes.
The particle terminal settling velocity, u,,,,,, will be
estimated using the method developed by Concha and
Almendra for spheres[ 161, subsequently corrected for
hindered settling and nonsphericity. The terminal velocity of a single sphere, u,, is given by:

u s

where

2!$%[{1 +0.0921(~y-2y2
1 I

(5)

Mixing characteristicsof industrialflotation columns

l/3
1

and
p=

w2

[ 4 (Pp -

P/)&g

Then, u, can be corrected to account for the effect of


volume fraction solids, #I, using the empirical relationship developed by Richardson and Zaki[ 171:
2

= (1 -

#)

(6)

where m = 4.65 for Re c 0.2, m = 4.35 ReV0.03 for


0.2r.Re<l.O,
m=4.45Re-O.
for l.O<Re<500
and m = 2.39 for Re > 500. Since the volume fraction
solids will decrease with increasing x, due to capture
onto bubbles of many of the floatable particles, u,+
will generally increase during the particle descent. An
average value of u,+ is obtained by using the average
of +(feed), at x = 0, and 4(tailings), at x = L, in eqn
(6). Finally, us+ is modified to account for nonsphericity of the mineral particles. The shape adjustment factor used in this work is the sphericity, $J, as
given by Govier and Aziz[ IS], where + is the ratio of
the surface area of a volume-equivalent sphere to the
surface area of the particle.
PARAMETER

DETERMINATION

Residence time distribution (RTD) was measured


using the unsteady state method of tracer impulse
injection to the top of the particle recovery zone of
the flotation column, followed by discrete sampling
of the tailings discharge and subsequent sample analysis for tracer content. The parameter estimation
method was similar to that employed by Rice et
a1.[5, 61 which consisted of equating experimentally
determined weighted moments to the theoretically
derived moments, which are functions of t and N.
The weighting factor, exp(-st), gives less importance
to the tail of the data. An optimum value of s was
selected by a search routine which minimized the root
mean square (RMS) error between the experimental
and theoretical curves, where RMS error =

1/c
[

ktcmi

cfi)2fi

i-l

a minerals system is excessive dye adsorption by the


solids. Prior to the RTD experiments, laboratory
adsorption tests were performed with Ruorescein in
slurries of molybdenum and zinc sulfide concentrates.
A reasonably small fraction (~20%) of the fluorescein
was adsorbed. (In contrast, tests with two other dyes,
Rhodamine WT and methylene blue, showed ~90%
dye adsorption).
As well, fluorescein was not adsorbed
by air bubbles in separate tests in which water and
fluorescein were sparged with air in a laboratory
flotation machine. Fluorescein is light sensitive[20], so
collected samples were stored in the dark.
RTDs of solid particles were determined
with
manganese

dioxide

as measured

by

atomic

absorp-

tion of manganese. Manganese dioxide was selected


because it satisfied the following requirements of a
solid tracer: (1) a low background concentration of
the tracer in the tailings material; (2) a low flotation
rate (i.e., it is hydrophilic); (3) a specific gravity in the
range of most sulfides (the MnO, was measured by
pycnometer to be 5.19 g/cm3); (4) ease of analysis; and
(5) reasonable cost. Laboratory grade MnO, was
ground and classified to an upper particle size of 150
pm prior to the RTD experiments. From microscope
examination the sphericity, #, was estimated to be
0.8.

Column

operation

The tracer experiments were performed on two


flotation columns operating at Mines Noranda Limitee, Division Gaspi, Quebec in a MO& concentrate
upgrading circuit. Operation and performance of the
columns have been previously described[2,3]. The
two columns were 0.46 m (18 in.) square and 0.91 m
(36 in.) square. They shall be referred to hereon as I8
and 36 inch columns.
Operating conditions for the two columns at the
time of the RTD tests are listed in Table 1. Air
holdup was estimated from direct measurement of
voidage on the 18 in. column when there was only
water and frotber in the column with no feed flow.

P
9

and c,,,~ is the measured normalized concentration at

time 1, and cfi is the fitted concentration, equation 3.


Hopkins et al. [ 191 show that a suitable range for the
value rs is roughly 1.2-5.5; the values obtained in our
work all lie between 1.5 and 2.0.
Using this technique, r and N were estimated from
the experimental data. t could be compared with the
measured quantity L/u, and E was determined from
E = JVL2/60r m/s, where L is given in meters and T
in minutes.
EXPERIMENTAL

Tracer

1063

selection

Liquid RTD was determined using the dye fluorescein measured with a portable spectrophotometer. A
problem that is often encountered when using a dye in

Test procedure

For each of the two tests, fluorescein and manganese dioxide were mixed with water and added to
the column at a point close to the interface of zones
1 and 2 by pouring them, from the top of the column,
down a funnel and tube arrangement. Total injection
time was approximately 30 sec. for the 18 in. column
and 60 set for the 36 in. column. Samples of tailings,
concentrate and feed were taken over the next 60 min
at the sampling points indicated in Fig. 1.
To obtain liquid for fluorescein analysis the samples were settled and decanted. Manganese measurements were performed on the tailings form the
18 in. column test only. In preparation for manganese
analysis, each sample was filtered, dried, and screened
into four size classes: +400-270
mesh, +270-200
mesh, +200-150
mesh and + 150-100 mesh.

1064

and J. A. FINCH

G. S. DOBBY
Table

1. Column

operating conditions
18 inch

Feed

flowrate

Washwater

(L/min)

flowrate
flowrate

Tailings
Concentrate

wt%

Volume

liquid

ua(m/s)

solids,

150

530

40

gas
air

velocity,

holdup,

point

to

air

ug(m/s)

D
spargei

(ml

The measured tailings and concentrate RTD curves


for the 18 and 36 in columns are shown in Fig. 2.
Concentration
of dye in the feed is also shown, indicating that a small amount of dye from the concentrate
and/or
tailings
streams
was recirculated
from the
subsequent
flotation stages to the feed of the column
being tested. As shown in Fig. 2, the recycled quantity
is minima1 and can be safely ignored. The tailings data

0.006

0.004

0.014

0.018

0.055

RESULTS

. TAILS
a3NC
* FEED

DYE
CONC
@Pm)

0.011

0.013

+Caverage)

Liquid RTD

190

fraction

Estimated
feed

35

(L/min)
velocity,

inch
380

solids

Superficial

L,

IL/min)

flowrate

Interstitial
Feed

(L/min)

36

120

10.0

were normalized
and fitted to the dispersion
mode1
with the procedure
previously
described.
Figure
3
shows that for both columns the fit is good, evidence
that the Laplace solution to the open ended dispersion
mode1 is adequate to describe mixing in large flotation
columns. The fitted parameters
are given in Table 2.
The
flotation
column
dispersion
coefficients
are
compared
with previously
published
data on large
counter-current
bubble columns in Fig. 4. The radial
dimension of the square sectioned columns are characterized by equivalent diameters (0.52 and 1.03m). The
interval-bars
for the Magnussen
and Schumacher[
lo]
data reflect that their measured dispersion coefficients
were average

0.070

9.5

velocity

36

values obtained

(0.01-0.1

m/s).

over a wide range

Gas

velocities

DATA
MODEL :

in the

of gas
18 in.

2=1X
min
N4 = 0.476

36
0-04-

DYE
CONC
(Ppm)

TAILS
CONC

FEED

t
0.08

12

18

0.06

DATA
MODEL :

2= 12.6 min

NC= 0.276

18

004'

oI)2,I

TIME (mid
Fig. 2. Dye concentration vs time for the tailings, concentrate and feed stream of the 18 in. and 36 in. square
columns.

TIME

(min)

Fig. 3. Normalized liquid RTD obtained from the tailings


stream of the 18 in. and 36 in. columns.

Mixing characteristics of industrial flotation columns

1065

Table 2. Liquid dispersion parameters


18
NI1, fitted
El,

fitted

Cmz/sl

T1*

fitted

Cmin)

*l*

estimated

from

L/u,

o18'&36'Flotation
Cciumm

.
0.07-

datP

.Magmhsen
l

Rice

- Rice

dataC5)
data(=)

(m*/s>

Inch

0.476

0.033

0.060

12.6

13.2

12.5

14.7

005-

EI = 0.0630
where D is the equivalent

Soli&

OOI-

36

have been averaged over gas velocities from 0.01 to


0.033 m/s.
If the data at low superficial gas velocities are
considered, it appears that k = 1 for large columns,
neglecting the effect of ug- It is proposed here that for
flotation columns with D > 0.2 m and operating at gas
velocities between 0.01 and 0.03 m/s (as is typical) the
following relationship holds:

0.09
-

Cmin)

Inch

0.278

+*
b

02

0*

Q6

ae

la

D (ml
Fig. 4. Liquid axial dispersion coefficient vs column diameter for 18 in. and 36 in. flotation columns, and bubble
columns of Magnussen and Schumacher [lo] and Rice er
al. [5, 61.

and 36 in. flotation columns were 0.0 14 and 0.018 m/s


respectively, so it would be expected that the dispersion coefficients of the flotation columns would fall at
the lower end of the Magnussen data range. This is the
case. Included in Fig. 4 is the data of Rice er a1.[5, 61
for column diameters from 0.1 to 0.3 m. Their data

Interval
(mesh)

+400-270

dAVE(WI)
44

TPO
tmin)
9.7

RTD

The measured and fitted normalized data for Mn02


in the 18 in. column are given in Fig. 5. The error bars
represent 2 two standards deviations, which accounts
for both analytical and sampling error. Analytical
relative standard deviation was 2% and background
Mn content was approximately 10% of the average Mn
assay. Sampling error was estimated by the method of
Gy[21].
Table 3 summarizes the fitted parameters.
Note that E,, does not vary with particie size. The
average of the solids axial dispersion coefficients was
0.035 m/s, compared with E, = 0.033 m/s. It is
evident from this data that the axial dispersion coefficient of mineral particles in a flotation column is the
same as that of the liquid.
The predicted mean residence times of the four size
classes are calculated in Table 4. The predicted and
measured rPc vs d are plotted in Fig. 6. Agreement
between measured and predicted values is good.

N
PO

CL,

RMS Error
(x10=)

0.239

0.037

0.31
0.32

+270-200

63

9.0

0.174

0.029

+200-150

88

7.3

0.186

0.038

0.43

+lSO-100

125

6.0

0.143

0.036

0.38

0.035

0.36

12.6

0.278

0.033

0.13

Solids

Average

water
TGeometric
mean
of
because
means
are
screen
series).

mesh
then

interval
related

(7)

column diameter, in meters.

Table 3. Fitted dispersion parameters for solids in 18 in. column


Size

m/s

(typically
by
JT as

in

selected
original

G. S. DOBBY and J. A. FINCH

Predicted
- Measured

.
8-

*
\.

6-

d,

0.

(urn)

Fig. 6. Mean particle(Mn03 residencetime vs particlesize,


predicted and measured. pp= 5.19, p = 0.013, JI = 0.8 and
do= 0.006.

Flow split
A material balance of the fluorescein split and
hence water split between the tailings stream and the
concentrate stream can be obtained with the use of
the flowrates (Table 1) and RTD curves (Fig. 2). In
the 18 in. column, 99.8% of the measured fluorescein
reported to the tailings stream and in the 36 in.
column, 99.2% reported to the tailings. (In both
cases, approximately
18% of the dye was unaccounted for and is assumed to have been adsorbed
by solids.) From a flow balance it was determined
that the concentrate water from the 18 in. column
was composed of 5% original feed water and 95%
wash water, while concentrate water from the 36 in.
column was 8% original feed water and 92% wash
water. The washwater is evidently very effective in
preventing feed water, and thus gangue particles,
from reaching the concentrate.

0200 -150 mesh

DISSCUSSION

TIME (min)

Fig. 5. Normalized MnO, RTD of the 18 in. column. (Not


all of the tailings samples taken were analyzed for manganese.)

Mixing parameters of large flotation columns have


been determined under industrial operating conditions. Results on these large columns show that the
liquid axial dispersion coefficient follows the dependance upon column diameter as suggested by
Magnussen and Schumacher, El a D. The square
cross-sectioned flotation columns (using the equivalent diameter) gave the same dispersion coefficients
as those determined by Magnussen and Schumacher
for similar diameter cylindrical columns. This simi-

Table 4. Predictionof rpr for 18 in. column. Conditions: pMn0,


(9T

f 2O); & = 0.006 (-2.5

= 5.19 g/cm; $ - 0.8; or = 0.013 pp


wt% solids); pr = 0.0127 m/s; T, = 12.6 min

44

-0033

.0032

.82

-0026

-83

10.5

9.7

63

-0064

.0062

-77

-0048

-73

9.1

9.0

88

-0116

.0113

.76

.0086

.60

7.5

7.3

125

-0210

.0205

.70

-0144

.47

5.9

6.0

Mixing characteristics

of industrial

flotation columns

1067

larity between square and cylindrical columns also


agrees
with
the
observation
of
Akita
and
Yoshida[l3].
These experiments have also verified the observation by other researchers that axial dispersion
coefficient of solids is equal to that of liquid. Then,
for large columns at low air rates,
E,, = E, = 0.0630

m/s.

(8)

For the purposes of scale-up and design of flotation


columns equation 8 is quite adequate, even though
it does not take into account the effect of gas velocity.
Gas velocity in column flotation will probably vary
from 0.01 to 0.03 m/s so its influence on eqn (8) should
be small.
The solids RTD tests have shown that the bubble
swarm has little effect on the mean residence time of
a mineral particle as determined simply by combining
the relative particle-fluid settling velocity with the
liquid velocity. The weight fraction solids at the time
of the test (3%) was lower than in normal operation
(~5-15~~). As well, the air holdup was smaller than
would be expected if bubbles approximately 1 mm in
diameter were being generated[6]. Nevertheless, the
outcome of the solids tests show that the RTD of
solids in a large flotation column can be predicted by
application of eqns (4) and (8); they provide good
scale-up criteria.
As a consequence of being able to estimate rP+,.we
can examine the effect that liquid velocity has upon
the ratio rpr/7,. This is presented in Fig. 7 for the
hypothetical case of pP = 4.0, p = 0.014, $ = 0.8 and
C$= 0.059 (20 wt/, solids). The diagram illustrates the
sensitivity of rpq/rr to the mean liquid velocity, especially for particles larger than 40 em. This could pose a
problem in using laboratory flotation columns that are
short, say 2-3 m, where u, must necessarily be kept low
in order to attain the same mean residence time of an
industrial, 10 m column.
CONCLUSIONS

The results of tracer tests performed on industrial


columns have led to the following conclusions:
(1) For flotation columns with D > 0.2 m and
operating at gas velocities between 0.01 and 0.03 m/s,
a reasonable expression for both liquid and solid axial
dispersion coefficients is given by Epe = El = 0.0630
where D is the equivalent diameter of the
m/s,
column, in meters.
(2) A good estimate of the mean residence time of
particles in a bubble swarm (ug -Z 0.03 m/s) can be
made from the expression TV = T~(u,/(u, + up,))_
(3) The equations stated in 1 and 2 above provide
a good criteria for scale-up of industrial sized flotation
columns.
(4) Washwater operation on the 18 in. and 36 in.
flotation columns is very effective, allowing less than
1% of the feed water to reach the concentrate.
Acknowhdgemenzs-The

authors wish to thank Centre de

RechercheNoranda, Mines Noranda Ltee, for providing a

Fig. 7. The ratio of mean particle residence time to mean


liquid residence time vs liquid velocity, assuming: 4 = 4.0,
p = 0.014, I& = 0.8 and $J = 0.059 (20 wt% solids).

scholarship to one of us (G. Dobby), and the Department


of Energy, Mines and Resources, Canada, for equipment
funding. Valuable experimental asssitance was provided by
M. Leroux, University, and Noranda plant personnel. Helpful discussions with M. Weber and A. R. Laplante are
appreciated.
NOTATIONS

tracer concentration
measured concentration, normalized
fitted concentration, normalized
column diameter, 111(for a square column, D is
the diameter of a circle with equivalent area)
particle diameter, cm (unless otherwise noted)
liquid axial dispersion coefficient, m2/s
solids axial dispersion coefficient, m*/s
acceleration due to gravity, cm/s*
column length, from tracer injection point to
the bubbler at the bottom of the column, m
liquid vessel dispersion number, E,/uJ
solids vessel dispersion number, E,,/u,,+L
particle Reynolds number, ap~,d/p
maximum t of RTD
interstitial liquid velocity, m/s
relative velocity of particles in a slurry with
volume fraction solids c$, m/s
velocity of single spherical particle, m/s
velocity of spherical particles in a slurry with
volume fraction solids C#I,m/s
superficial liquid velocity, m/s
superficial gas velocity, m/s
axial distance from tracer injection point ( + ve
downward)
fractional air holdup
liquid viscosity, p
liquid specific gravity, g/cm3
particle specific gravity, g/cm3
liquid mean residence time, min
mean residence time of solids in a slurry with
volume fraction solids $I, min
volume fraction solids in a slurry
particle sphericity =
surface area of volume equivalent sphere
surface area of particle

1068

G. S. DOBBY and J. A. FINCH


REFERENCES

[II Trahar W. J., Int. J. Min. Proc. 1981 8 289.


r21 Cienski T. and Coffin V., Proc. 13 th Canadian Mineral
Processors Meeting, p. 240 Ottawa 198 1.
[31 Coffin V. and Miszczak J., 14th Int. Mineral Processing Cong., paper IV-21. Toronto 1982.
I41 Levenspeil O., Chemical Reaction Engineering, Chap. 9.
Wiley, New York 1972.
I51 Rice R. G., Tupperainen J. M. I. and Hedge R. M.,
Gun. J. Chem. Engng 1981 59 677.
WI Rice R. G., Oliver A. D., Newman J. P. and Wiles R.
J., Powder Tech., 1974 10 201.
PI Ostergaard K. and Michelsen M. L., Can. J. Chem.
Engng 1969 47 107.
181 Shah Y. T., Stiegel G. J. and Sharma M. M., A.Z.Ch.E.
J. 1978 &I(3) 369.
[91 Shah Y. T., Kelkar B. G., Godbole S. P. and Deckwer
W. D., A.Z.Ch.E. J. 1982 28(3) 353.
1121 Alexander B. F. and Shah Y. T., Chem. Engng Jour.,
1976 11 153.

[ 131 Akita K. and Yoshida F., Ind. Engng Chem. Proc. Des.
Dev. 1973 12(l) 76.
[14] Imafuku K., Wang T. Y., Koide K. and Kubota H., J.
Chem. Engng Japan 1968 l(2) 153.
[ 151 Argo W. B. and Cova D. R., Znd. Engng Chem. Proc.
Des. Dcv., 1965 4 352.
[16] Concha F. and Almendra E. R., Znt. J. Mineral Proc.,
1979 5 349.
[17l Richardson J. F. and Zaki W. N.. Trans. Inst. Chem.
Engrs 1954 32 35.
[18] Govier G. W. and Aziz K., The Flow of Complex
Mixtures in Pipes. p. 12. Van Nostrand Reinhold, New
York 1972.
[19] Hopkins M. J., Sheppard A. J. and Eisenklam P.,
Chem. Engng Sci. 1969 24 113 1.
[20] Smart P. L. and Laidlaw I. M. S., Water Resources
Research 1977 13(l) 15.
[21] Wills, B. A., Mineral Processing Technology, 2nd Edn.,
p. 42 Pergamon Press Oxford 198 1.

You might also like