You are on page 1of 9

ARTICLE IN PRESS

Minerals Engineering xxx (2005) xxx–xxx


This article is also available online at:
www.elsevier.com/locate/mineng

Some gas dispersion characteristics of mechanical flotation machines


Jan E. Nesset, Jose R. Hernandez-Aguilar, Claudio Acuna,
Cesar O. Gomez, James A. Finch *
Department of Mining, Metals and Materials Engineering, McGill University, The Wong Building, 3610 University Street, Montreal,
Quebec, Canada H3A 2B2

Received 12 July 2005; accepted 18 September 2005

Abstract

The gas dispersion properties that are the subjects of this communication include three measured parameters, gas superficial velocity
(gas rate), gas holdup and bubble size, plus the derived parameter, bubble surface area flux. Over the past 10 years sensors to measure
these parameters have been introduced and those originating from McGill University are briefly described.
The sensors have seen extensive use in plant campaigns and experiences on four cell types at three sites are discussed. The impact of
frother in controlling bubble size by retarding coalescence is illustrated, indicating the need to understand chemistry/machine interac-
tions in interpreting gas dispersion results. An increase in Sauter mean bubble size (D32) with superficial gas velocity (gas rate, Jg)
was found, which was fitted empirically by D32 = D0 + C, J ng where D0  0.5–0.6 mm and C and n are fitted parameters dependent
on the system (machine and chemistry). Following this, a consistent trend in bubble surface area flux (Sb) vs. Jg was also observed.
The relationship between Sauter mean and number mean (D10) is suggested to interpret the nature of the distribution. Analysing gas
holdup vs. gas rate shows both a dependence on location in the cell as well as cell position in a bank, factors which must be considered
when comparing data from different plants and circuits.
The instrumentation and methodologies being developed are presenting new opportunities for incorporating cell hydrodynamics into
metallurgical diagnostics. A case study at one plant serves as illustration.
Ó 2005 Elsevier Ltd. All rights reserved.

Keywords: Flotation machines; Flotation frothers; Flotation bubbles

1. Introduction The sensors and instruments employed here are those


described in Gomez and Finch (2002). Designed for plant
The last 10 years have seen significant advances in the work, they are robust and provide automated data acquisi-
measurement of so-called gas dispersion properties in flota- tion, and increasingly, on-line data analysis. There are other
tion machines, namely: gas superficial velocity (or simply devices at varying levels of sophistication (Powell et al.,
gas rate) (Jg), gas holdup (eg) and bubble size (diameter, 2000; Yianatos et al., 2001; Grau and Heiskanen, 2003).
Db). From these can be derived the bubble surface area flux As of writing, the team has conducted gas dispersion
(Sb = 6Jg/Db, where Db is usually the Sauter mean diame- campaigns at 12 sites on five continents. These have
ter, D32). Each paremeter carries information that charac- included: investigating summer/winter swings in metal-
terizes gas dispersion, the Sb being of prime interest lurgy (the case study selected for discussion); interpreting
because it is related to flotation rate (Gorain et al., 1997; differences between recovery in pilot and full-scale cells;
Hernandez et al., 2003). determining the impact of gas rate profiles on banks;
comparing the effect of different frothers and collector
types on gas dispersion; detailing the response of self-aer-
*
Corresponding author. Fax: +1 514 398 4492. ation machines to process variables; and comparing spar-
E-mail address: jim.finch@mcgill.ca (J.A. Finch). ger systems for flotation columns. This paper extracts cell

0892-6875/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mineng.2005.09.045
ARTICLE IN PRESS

2 J.E. Nesset et al. / Minerals Engineering xxx (2005) xxx–xxx

characterization data from three concentrators: North (radial and depth variations in gas distribution in a cell
American Palladium Ltd.Õs Lac des Iles (LDI) (Ontario, can be expected).
Canada); Impala PlatinumÕs UG2 (Impala) (S. Africa);
and Noranda/FalconbridgeÕs Raglan (Raglan) (Quebec, 2. Instrumentation
Canada). The paper focuses on forced air mechanical
machines: at Raglan, (nominal) 28-m3 Outokumpu OK- The basic measurement is the gas superficial velocity
28 U-bottom cells; at LDI, 130-m3 Outokumpu OK-130 (Jg), namely the volumetric flow rate of gas (air) (Qg) per
tank cells; and at Impala, 50-m3 Bateman 50 and 30-m3 cross-sectional area of the cell (A). i.e., Jg = Qg/A with
Metso RCS 30 tank cells. Common gas dispersion fea- units typically cm/s. This is often termed simply Ôgas rateÕ.
tures are identified, for example the dependence of bubble The data reported here were measured using the so-called
size on gas rate. There is no intent to compare cell on/off Jg sensor (Fig. 1a) (Torrealba-Vargas et al., 2004).
designs as this would require a more complete under- It comprises a tube to collect bubbles by natural buoyancy
standing of the role of at least the following: pulp prop- and infers Jg from the rate of increase in pressure once a
erties (density, viscosity), mechanical factors (energy valve is closed (specimen Ôpressure-variationÕ curves are
input, mixing), solution chemistry (notably frother dosage included, Fig. 1b). The estimation of Jg requires the bulk
and sometimes salt concentration) and sampling location density (qbulk) and this is usually taken at the same time

Fig. 1a. Schematic of McGill on/off gas rate sensor (Note, level in tube descends as air accumulates).

Fig. 1b. Example pressure variation curves for the on/off sensor.
ARTICLE IN PRESS

J.E. Nesset et al. / Minerals Engineering xxx (2005) xxx–xxx 3

taking two pressure measurements a known depth apart.


(The latest sensor design incorporates both Jg and qbulk
measurements in one device.) The method has the advan-
tage of also providing a direct measure of real-time varia-
tion in cell pulp level, a variable required for correction
of Jg (as well as bubble size). The standard procedure is
to report Jg at the location (depth) sampled, typically
0.5–1.0 m below the pulp–froth interface. This is a region
away from the impeller where conditions are less turbulent
and bubble surface flux represents that leaving the pulp and
entering the froth.
Gas holdup (eg) is the volumetric fraction (or %) of gas
in the aerated slurry. The sensor is conductivity-based
(Tavera et al., 1996) comprising (Fig. 2) two flow cells,
an open cell to measure the conductivity of aerated slurry
(jslg) and a siphon cell to measure the conductivity of de-
aerated slurry (jsl). The siphon cell has a conical base with Fig. 3a. Schematic of McGill bubble size analyzer.
a small opening that acts to restrict bubble entry and thus
establish a bulk density difference between the fluid inside
and outside the cell. This density difference drives a flow effects on the bubble images. The assembly is first filled
of slurry out through the orifice in through the top of the with process water (to limit changes in chemistry experi-
cell (hence the description ÔsiphonÕ). This flow both com- enced by bubbles as they enter), then a valve near the base
pletes the elimination of air (bubbles cannot enter against of the tube is closed and the tube inserted at the desired
the outflow of slurry) and replenishes the slurry in the cell. location in the cell (below the froth and typically at the
With the two conductivities, MaxwellÕs model for the con- same location as for Jg). The valve is opened and bubbles
ductivity of a dispersion of spherical non-conducting rise by buoyancy into the viewing chamber. Here they
objects (bubbles in this case) in a continuous conducting encounter the sloped window and spread into an approxi-
medium (slurry) is solved (included in Fig. 2). (Note, the mate mono-layer, which reduces bubble overlap and pro-
model involves the ratio of conductivities, hence their vides an unambiguous plane of focus. For several
actual value, or whether they vary during measurement minutes the bubbles remain in clear view until particles
(e.g., due to reagent addition, change in percentage solids, (carried by the bubbles and released as they burst) accumu-
etc.) is not a factor.) Highly saline solutions require modi- late and cloud the chamber. Example images illustrating
fications to the sensor calibrations in order to maintain lin- the clarity are included (Fig. 3b). Typically over 10 000
earity and sufficient sensitivity. bubbles are counted after the distribution has reached
Bubble size distribution is the third gas dispersion steady state (the initial bubbles entering the viewer are
parameter. The technique employs a sampling-followed- the largest in the population as they rise up the sampling
by-imaging technique (Hernandez-Aguilar et al., 2004). tube faster than the others). The data are processed to yield
The set-up (Fig. 3a) comprises a sampling tube connected frequency distributions (number, surface area and volume)
to a closed, sloped viewing chamber with rear illumination and various statistical means, usually the number mean
through a diffuser (filter) to minimize light diffraction (D10) (sensitive to small bubbles) and volume-surface (Sau-
ter) mean (D32) (sensitive to larger bubbles and appropriate
metric to calculate bubble surface area flux).
Ideally, all gas dispersion measurements should be
made simultaneously at the same location with the pres-
sure and temperature recorded. In this way correction
to ÔstandardÕ conditions can be made, which is essential
to compare data from different cells. Practical consider-
ations often impose compromises. A discussion of the
plant sampling methodology used to investigate the rela-
tionships between metallurgical performance and the gas
dispersion measurements is given by Nesset et al. (2005).
The key is to conduct campaigns over a wide range of
the major variables (e.g., froth depth, air rate, reagent
addition level, Jg profile), preferably at the same time con-
ducting metallurgical sampling. Our standard practice is
Fig. 2. Schematic of McGill gas holdup sensor (including Maxwell to include repeat tests on separate days to measure
model). reproducibility.
ARTICLE IN PRESS

4 J.E. Nesset et al. / Minerals Engineering xxx (2005) xxx–xxx

some residual frother), the Sauter mean (D32) decreases


to a constant value (ca. 0.6 mm) (note the number mean,
D10, is less sensitive). The inserts reveal how the distribu-
tions evolve. In particular, note the volume frequency
where at zero addition there are two modes, at ca. 1 mm
and ca. 5 mm, and as frother is added the peak at 1 mm
grows at the expense of the other.
One interpretation is that bubbles are produced at ca.
1 mm and some coalesce to form the larger bubbles (ca.
5 mm), a process progressively retarded by addition of
frother. Around 20 ppm (in this case), coalescence appears
to be completely prevented, which has been termed the crit-
ical coalescence concentration (CCC) (Cho and Laskowski,
2002). At concentrations above the CCC the volume fre-
quency approaches the number frequency indicating a dis-
tribution that is narrowing (the two means would be
coincident for a mono-sized dispersion). There are some
issues arising from this ÔcoalescenceÕ model which are dis-
cussed elsewhere (Finch et al., 2005).

3.2. Bubble size versus gas rate

Fig. 5 summarises the experience for the four machine


types showing the Sauter mean size increases with gas rate.
(This is among the first data showing this relationship in
operating cells.) The trend follows a power law,
D32 ¼ D0 þ CJ ng ; ð1Þ
where D0 is the bubble size (diameter) at Jg = 0 and C and
n are parameters dependent on the bubble production
Fig. 3b. Example bubble images: top left, lab, air–water, fine frit porous mechanism, system chemistry, and, possibly, slurry proper-
sparger; top right, same as left but coarse frit; mid left, industrial ties. The relationship is a modification of that proposed by
mechanical machine, base metal slurry; mid right, column flotation cell, jet Dobby and Finch (1986) for porous spargers in columns
type sparger, base metal slurry (large bubble isolated for demonstration (their equation did not include the D0 term). The Bateman
purposes); bottom left, industrial mechanical machine, PGM slurry; low
air rate; mid right, same as left but high air rate.
cell was deliberately run to approach zero gas rate and sug-
gests D0  0.5–0.6 mm, which is supported by extrapola-
tion of the other data. Given the machines all produce
3. Results and discussion bubbles by mechanical induced shear, we can hypothesize
that D0 is the (mean) starting size generated by that mech-
3.1. Bubble size and chemistry anism. Data gathered on a Denver laboratory cell mimics
the Db–Jg relationship (Sung-So Do, 2004), suggesting a
The bubble size observed in a flotation cell (of any vehicle for more controlled testing than generally allowed
design) is the result of the primary generation process in plant trials.
and secondary processes, notably coalescence. The slurry The increase in mean bubble size with gas rate may
properties (percent solids, viscosity, etc.) may influence result from an effect on the primary and/or secondary (coa-
either but certainly solution chemistry influences coales- lescence) processes. Fig. 6 shows the number frequency
cence. In most cases, frother is added to retard coalescence, response to gas rates in two cases. In Fig. 6a (Impala) there
which Harris (1976) noted was more important than the is a steady increase in modal size as gas rate increases, sug-
machine in controlling bubble size. Some collectors such gesting the primary process is affected. An effect on the pri-
as fatty acids and amines also affect coalescence, as can mary process may be related to the decrease in power draw
high concentration of some salts. Any discussion of bubble as gas rate is increased (Arbiter et al., 1976), i.e., there is
size in flotation machines must start by recognizing the less energy for bubble break-up, but is also seen with por-
chemistry/machine interaction (Cho and Laskowski, 2002). ous spargers (Dobby and Finch, 1986) where the same con-
The role of frother is illustrated in Fig. 4 which follows sideration apparently would not apply. In Fig. 6b (LDI), as
the impact of sequential additions to a laboratory flotation Jg increases the peak initially at ca. 1 mm shifts to a larger
cell containing process (Impala) water (no solids). With size (primary process again), while between the two highest
increasing dosage (the process water may have contained gas rates, the ÔprimaryÕ mode remains at the same size but
ARTICLE IN PRESS

J.E. Nesset et al. / Minerals Engineering xxx (2005) xxx–xxx 5

0.4

0.3
Fraction

0.4
0.2 Volume
Surface 0.3
Number
0.1

Fraction
0.2 Volume 0.4
0.0 Surface
0.1 1.0 10.0 0.1 Number 0.3
Db (mm)

Fraction
0.0 0.2 Volume
0.1 1.0 10.0 Surface
Number
Db (mm) 0.1

0.0
3.0 0.1 1.0 10.0
Db (mm)
2.5
0.4
D10 or D32 (mm)

2.0

1.5 0.3
D10

Fraction
1.0 D32 0.2 Volume
Surface
0.5
Number
0.1
0.0
0 20 40 60
0.0
Frother Addition (pm) 0.1 1.0 10.0
Db (mm)

Fig. 4. Impact of frother dosage on mean bubble size (D32 and D10) and on frequency distributions (number, surface area, volume) at selected frother
dosages.

3
40
Outokumpu TC 130 LDI
2.5 Jg=0.11
Jg=0.11
30
Frequency (%)

Jg=0.48
Jg=0.48
2
D32 (mm)

Bateman TC 50 Impala Jg=1.34


Jg=1.34
20
1.5 Metso RCS 30 Impala
Outokumpu 28U Raglan 10
1

0.5 0
0.1 1.0 10.0
0
0 0.5 1 1.5 2 2.5 3 Db (mm)

Jg (cm/s) Fig. 6a. Impact of gas rate on number frequency (Impala—Bateman).


Fig. 5. Bubble size (Sauter mean diameter) versus gas rate for the four cell
types at the three plants.

25
diminishes in frequency with a second mode becoming evi- Jg =0.43
Jg=0.43
dent (ca. 2.5 mm), suggestive of an effect on the secondary 20
Frequency (%)

Jg =1.0
Jg=1.0
process. The top gas rate may be too high for the frother Jg =2.4
15 Jg=2.4
dosage to fully retard coalescence.
The availability of bubble size distribution data invites 10
exploration of the properties. Plotting D32 against D10 pro-
vides some insight into the effect of gas rate on the distribu- 5
tion. Fig. 7 shows the result for two of the cases in Fig. 6. 0
The LDI (Outokumpu) data closely follow the identity, 0.1 1.0 10.0
which means the distribution remains narrow over the Db (mm)
gas rate range, compared with the Impala (Bateman) data
which indicates a widening distribution with increasing gas Fig. 6b. Impact of gas rate on number frequency (LDI—Outokumpu).
ARTICLE IN PRESS

6 J.E. Nesset et al. / Minerals Engineering xxx (2005) xxx–xxx

rate. The different response may be related to the degree of 100 Impala Bateman
coalescence but more experience is required to pursue this

Minus 1 mm bubbles (%)


reasoning. 80 Impala Metso

Fig. 8a indicates the distributions often have a consis- LDI OK


60
tent shape: the percentage minus 1 mm versus D10 for three
of the full size machines and for pilot and lab scale data LDI Pilot
40
(LDI) yield a unique relationship. This is expected from LDI Lab
integration of the fine end of the frequency distribution, 20
which usually approximates a normal distribution. The 0
coarse end is often not as well behaved and so the plot of 0.5 1.0 1.5 2.0 2.5
percentage minus 1 mm versus D32 is not a consistent rela- D32 (mm)
tionship (Fig. 8b). Evidence was presented by Nesset et al.
(2005) that the proportion of minus 1 mm bubbles can Fig. 8b. Plot of percentage minus 1 mm versus D32 does not show the
strongly influence fine particle recovery (true flotation same consistency as for D10 in Fig. 8b.
effect) and concentrate grade (entrainment effect), and
hence D10 should be included alongside the more common
D32 for linking with metallurgical performance. 80
Outokumpu 28U Raglan
Much as the availability of particle size data stimulated 70
comminution research, the same can be expected in flotation Metso RCS 30 Impala
60
from bubble size data. The hypothesized advantage of Bateman TC 50 Impala
matching bubble and particle size distributions is closer to
Sb (1/s) 50
Outokumpu TC 130 LDI
being tested given the measurement abilities now available. 40
30
3.3. Bubble surface area flux versus gas rate 20
10
Fig. 9 is the data in Fig. 5 presented as bubble surface
0
area flux. The values extend up to 80 s1, which is towards
0 0.5 1 1.5 2 2.5 3
Jg (cm/s)
3
Fig. 9. Bubble surface area flux versus gas rate for data in Fig. 5.
2.5

2
D32 (mm)

LDI
the high end of the range in the literature (Gorain et al.,
1.5 Outokumpu 1997; Powell et al., 2000). Xu et al. (1991) suggest a maxi-
1 mum Sb in water–air systems of ca. 160 s1.
Impala
Bateman
Following from Fig. 5, the trends could be fitted by
0.5
6J g
0 Sb ¼ . ð2Þ
0 0.5 1 1.5 2 2.5 3 ðD0 þ CJ g Þ
D10 (mm) After calibration (to provide estimates of D0, C and n), Eq.
Fig. 7. Sauter mean (D32) versus number mean (D10). (2) could be integrated into flotation simulators.

3.4. Gas holdup versus gas rate


100 Impala Bateman
This relationship was the first explored to characterize
Minus 1 mm bubbles (%)

80 Impala Metso flotation machines. It was used, for instance, to establish


the operating gas rate range of a cell (Cooper et al.,
60 LDI OK
2004). Fig. 10 shows the relationship in two cells (both
40 LDI Pilot OK-28) at Raglan in the rougher/scavenger bank. The typ-
ical linear trend (which holds to some gas rate above which
LDI Lab
20 gas holdup varies erratically) is noted, but with a different
relationship for the two cells. The corresponding bubble
0
0.0 0.5 1.0 1.5 2.0
size versus gas rate data did not show such a distinction
(bubble size from both cells is included in Fig. 5). This sug-
D10 (mm)
gests slurry properties do not affect bubble generation as
Fig. 8a. Plot of percentage minus 1 mm versus D10: note consistent much as they do bubble dispersion: gas holdup is sensitive
relationship regardless of source. to bubble rise velocity, which may have been influenced by
ARTICLE IN PRESS

J.E. Nesset et al. / Minerals Engineering xxx (2005) xxx–xxx 7

20 presented. Noranda/FalconbridgeÕs Matagami concentra-


tor in northern Quebec, Canada, has been processing mas-
16
sive sulphide copper–zinc ores from local mines for the past
Rougher 7 40 years. The deposits have been similar, containing a high
Scavenger 3 proportion of pyrite and talc as gangue minerals. A strong
εg @ 2.5m (%)

12 seasonal variation in final zinc concentrate grade has been


a feature for much of this period, giving as much as a 3%
8 minimum (winter) to maximum (summer) change. (Addi-
tional details are provided by Nesset et al. (2005)).
The results of a year-long investigation, which included
4
studying gas dispersion as well as chemical characteristics,
revealed much finer bubbles in winter than in summer, indi-
0 cated as March and August, respectively, in Fig. 12. Sur-
0.0 0.4 0.8 1.2 1.6 2.0
veys were conducted at low, medium and high Jg for the
Jg @ 2.5 m (cm/s) zinc rougher bank (all cells with the same targeted Jg) with
a repeat on a separate day. Comparing gas rate against zinc
Fig. 10. Gas holdup versus gas rate in two cells (Note, measurements are
concentrate grade (Fig. 13a) it is evident that the normal
at a specified depth which is now the practice).
operating Jg (indicated by arrows) results in the best zinc
grade in August and the lowest grade in March. When
differences in percent solids between the two cells (Banisi viewed as Sb against zinc grade (Fig. 13b) the trends now
et al., 1995). correspond with an off-set, demonstrating that the higher
The comparison is complicated, however, by the fact bubble surface area flux obtained in winter (due to smaller
that measurement is at a particular location in a cell, i.e., bubbles) is driving the zinc grade down.
it represents local gas holdup and gas rate values. Fig. 11
shows the g–Jg trend in the same cell for different positions
of the gas rate sensor. Experience with mechanical cells
100
generally supports a radial symmetry, gas rate, gas holdup
and bubble size decreasing with distance from the central
Frequency Distribution

80
March
gas injection point. This is seen here: reading across, the
(% passing)

Jg at the central position is higher than at the wall (laun- 60 August


der). The result is a different (but still linear) trend for
the two locations. Consequently, when comparing cells 40
(e.g., to examine the possible slurry effect on gas holdup Cell E2

noted above), it is necessary to allow for the sensor 20 Jg= 0.62

location.
0
0 1 2 3 4 5 6
3.5. Linking to metallurgy: an example Db (mm)

By way of an industrial case study, the investigation into Fig. 12. A comparison of bubble size distributions at the Matagami
seasonal variation in zinc concentrate grade at Matagami is concentrator for March and August at the same gas rate setting.

20
Concentrate Grade (%Zn)

55
Located at center Aug/02
Located at launder 53
16
Aug/02 Repeat
51
εg @ 2.5m (%)

12 March/02
49
March/02
8 OK28
OK28 47 Repeat

45
4 0 1 2 3 4 5

August March
0 Jg (cm/s)
0.0 0.4 0.8 1.2 1.6 2.0
Jg @ 2.5 m (cm/s) Fig. 13a. Zinc concentrate grade versus gas rate for winter (March) and
summer (August) zinc rougher surveys at Matagami. Note, arrows
Fig. 11. Gas holdup versus gas rate with Jg sensor at two locations. indicate plant operating settings.
ARTICLE IN PRESS

8 J.E. Nesset et al. / Minerals Engineering xxx (2005) xxx–xxx

55 1. Coalescence is suggested by a bi-modal size distribution,


Concentrate Grade (%Zn)

where the peak at the larger size shrinks with increasing


53 August/02
frother dosage.
51 2. The Sauter mean bubble size increases with gas rate fol-
August/02
Repeat lowing the trend D32 ¼ D0 þ CJ ng where C and n depend
49
March/02
on chemistry/machine system and D0 is the extrapolated
size when Jg = 0.
47
3. D0 may represent the inherent (mean) size produced by
45 the mechanism; the data suggest D0  0.5–0.6 mm.
0 10 20 30 40 50 4. Following from 2, a consistent trend in bubble surface
area flux versus gas rate was found; the maximum Sb
August March was ca. 80 s1.
Sb (1/s)
5. The trend in Sauter mean vs. number mean is used to
Fig. 13b. The same data as Fig. 13a but plotted against bubble surface probe the nature of the distribution; as the two means
area flux. approach it indicates a narrowing distribution. The
D10 is an important measure of the fine end of the distri-
bution being a direct measure of the proportion of
minus 1 mm bubbles.
56
6. To compare gas dispersion data from different cam-
Zn Concentrate Grade

54 paigns must consider slurry and chemistry effects and


@ 70 %Recovery

August
R2 = 0.9166 the influence of sampling location.
52 7. Gas dispersion measurement offers unique new insights
March
50
into metallurgical behaviour with opportunities for pro-
Linear fit cess improvements, as illustrated by a case study.
48

46 Acknowledgements
0 10 20 30 40 50 60
Bubble Surface Area in Fraction <1 mm (%)
The authors wish to thank the site personnel for their
Fig. 14. The concentrate grade data for March and August correlate when enthusiastic assistance during the campaigns, and SGS
plotted against the proportion of the surface area flux within the minus Lakefield Research for facilitating two of these initiatives
1 mm bubble fraction.
(Impala and LDI). The authors represent the various teams
assembled to conduct the work and wish to acknowledge
Various relationships were investigated to try to explain the significant contribution of the following: M. Bailey, J.
this dependency and, ultimately, the one demonstrated in Torrealba-Vargas, L. Calzado and R. Dahlke. Funding
Fig. 14, showing the concentrate grade is inversely-linearly was from two NSERC (Natural Sciences and Engineering
related to the proportion of bubble surface area flux con- Research Council of Canada) CRD grants, one sponsored
tained in the minus 1 mm bubbles proved the most con- by INCO, Noranda, Falconbridge, Teck-Cominco, COR-
vincing (R2 = 0.92). At this juncture it is not known if EM and SGS Lakefield Research, and the other by
the role of the fine bubbles is to enhance collection of gan- AMIRA International under the P9 project.
gue particles by true flotation or entrainment mechanisms
(For further discussion see Nesset et al. (2005)). While References
not completely resolved, the example does show the value
of gas dispersion measurement in understanding metallur- Arbiter, N., Harris, C.C., Yap, R.F., 1976. The air flow number in
flotation machine scale-up. Int. J. Miner. Process. 8, 257–280.
gical behaviour. Banisi, S., Laplante, A.R., Finch, J.A., Weber, M., 1995. Effect of solid
particles on gas holdup in flotation columns.—11. Investigation of
4. Conclusions mechanisms of gas holdup reduction in presence of solids. Chem. Eng.
Sci. 50 (4), 2335–2342.
Gas dispersion measurements—gas rate (Jg), gas holdup Cho, Y.S., Laskowski, J.S., 2002. Effect of flotation frothers on bubble
size and foam stability. Int. J. Miner. Process. 64, 69–80.
(g), bubble size (Db) and bubble surface area flux (Sb)—are Cooper, M., Scott, D., Dahlke, R., Finch, J.A., Gomez, C.O., 2004.
reported on commercial flotation cells from three sites, Impact of air distribution profile on banks in a Zn cleaning circuit. In:
North American PalladiumÕs Lac des Iles concentrator Proceedings 36th Annual Meeting of the Canadian Mineral Processors
(nominal 130-m3 Outokumpu OK-130 tank cells), Impala of CIM, pp. 525–539.
PlatinumÕs UG2 concentrator (50-m3 Bateman, and 30- Dobby, G.S., Finch, J.A., 1986. Particle collection in columns—gas rate
and bubble size effects. Can. Metall. Quart. 25 (1), 9–13.
m3 Metso RCS tank cells) and Noranda/FalconbridgeÕs Finch, J.A., Moyo, P., Gelinas, S., 2005. Frother-related research at
Raglan concentrator (28-m3 Outokumpu OK-28U cells). McGill University, Miner. Eng., in press, doi:10.1016/
The following observations were made: j.mineng.2005.09.027.
ARTICLE IN PRESS

J.E. Nesset et al. / Minerals Engineering xxx (2005) xxx–xxx 9

Gomez, C.O., Finch, J.A., 2002. Gas dispersion measurements in flotation Powell, A., Franzidis, J.-P., Manlapig, E.V., 2000. The characterization of
machines. CIM Bull. 95 (1066), 73–78. hydrodynamics conditions in industrial flotation cells. In: 7th Mill
Gorain, B.K., Franzidis, J-P., Manlapig, E.V., 1997. Studies on impeller OperatorsÕ Conference, Australasian Institute of Mining and Metal-
yype, impeller speed, and air flow rate in an industrial scale flotation lurgy, pp. 243–255.
cell—Part 4: Effects of bubble surface area flux on flotation kinetics. Sung-So Do, 2004. Hydrodynamic characterization of a Denver labora-
Miner. Eng. 10 (4), 367–379. tory flotation cell, M.Eng. Thesis, McGill University.
Grau, R.A., Heiskanen, K., 2003. Gas dispersion measurements in a Tavera, F.J., Gomez, C.O., Finch, J.A., 1996. Novel gas hold-up probe
flotation cell. Miner. Eng. 16, 1081–1089. and application in flotation columns. Trans. Inst. Min. Metall. (Sec. C:
Harris, C.C., 1976. Flotation machines. In: Fuerstenau, M.C. (Ed.), A.M. Miner. Process. Extr. Metall.) 105, C99–C104.
Gaudin Memorial Volume, SME of AIME, vol. 2, pp. 753–815. Torrealba-Vargas, J., Gomez, C.O., Finch, J.A., 2004. Model for the JK
Hernandez-Aguilar, J.R., Coleman, R.G., Gomez, C.O., Finch, J.A., and on–off McGill gas velocity sensors, Amira P9M Final Report,
2004. A comparison between capillary and imaging techniques for Flotation Module, vol. II, pp. 47–70.
sizing bubbles in flotation systems. Miner. Eng. 17 (1), 53–61. Xu, M., Finch, J.A., Uribe-Salas, A., 1991. Maximum gas and bubble
Hernandez, H., Gomez, C.O., Finch, J.A., 2003. Gas dispersion and de- surface area rates in flotation columns. Int. J. Miner. Process. 32, 233–
inking in a flotation column. Miner. Eng. 16 (6), 739–744. 250.
Nesset, J.E., Gomez, C.O., Finch, J.A., Hernandez-Aguilar, J., DeFeo, A., Yianatos, J., Bergh, L., Condori, P., Aguilera, J., 2001. Hydrodynamics
2005. The use of gas dispersion measurements to improve flotation and metallurgical characterization of industrial flotation banks for
performance. In: 37th Annual Meeting of the Canadian Mineral control purposes. Miner. Eng. 14 (9), 1033–1046.
Processors of CIM, 18–20 January 2005, Ottawa, pp. 401–422.

You might also like