Minerals Engineering 66–68 (2014) 77–83

Contents lists available at ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

Distributed flotation kinetics models – A new implementation approach

for coal flotation
Philip Ofori a,⇑, Graham O’Brien b, Priyanthi Hapugoda b, Bruce Firth a
a
CSIRO Energy Technology, 1 Technology Court, Pullenvale, QLD 4069, Australia
b
CSIRO Earth Science and Resource Engineering, 1 Technology Court, Pullenvale, QLD 4069, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Composition-dependent flotation kinetic models with distributed flotation rate constants for predicting
Received 29 December 2013 the flotation response of the various components of coal have been available for quite some time.
Revised 2 July 2014 However, the full potential of these models have not been realised because the floatability distribution
Accepted 16 July 2014
of particles, a key requirement for the utilisation of the models, has not been readily accessible experi-
Available online 21 August 2014
mentally to allow the validation of these models.
A new approach has been developed that allows the coal flotation feed to be fractionated into particles
Keywords:
classes of size and composition using a microscopic characterisation tool called Coal Grain Analysis.
Coal
Flotation kinetics
Previous work had determined the contact angles of discrete coal maceral groups and associated
Liberation minerals. Combining this information with the particle composition data from the Coal Grain Analysis
Modelling methodology provides a pathway for estimating the contact angles of heterogeneous coal particles via
Process optimisation Cassie’s Equation. The approach provides a unique opportunity to estimate the flotation rate constants
for the different particle classes.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction k is the flotation rate constant and R1 is the ultimate recovery.

It is recognised that coal particles in flotation slurry possess a
Flotation kinetics models that draw direct analogy with chemi- range of rate constants dependent on floatability (Imaizumi and
cal reaction kinetics, particularly the first order models, are the Inoue, 1965). There is a distribution of rate constants due to the
most widely used in the literature. The models are formulated in different particle size and composition classes present in the feed
terms of the rate of flotation and can be quantified in terms of coal. This may be taken into account in the flotation models as:
the physical, chemical and hydrodynamic parameters of the flota- X
n X
l
 
tion system. RðtÞ ¼ mij  R1ij 1  expðkij tÞ ð2aÞ
For instance, composition-dependent flotation models that are i¼1 j¼1

capable of predicting the flotation response of the various compo- or

nents of mineral ores and coal have been available for a long time.
Xn X l  
A simple example of these models is the first order kinetic flotation ðkij tÞ
model given in Eqs. (1a) and (1b). In this model the total recovery
RðtÞ ¼ mij  R1ij ð2bÞ
i¼1 j¼1
1 þ kij t
R(t) at flotation time t for a system in plug flow and for a system
exhibiting perfect mixing respectively are given by (Garcia where mij is the mass fraction of particles of size class i and compo-
Zúñiga, 1935; Kelsall, 1961): nent j in the slurry and the sum of the mass fractions is equal to 1, kij
represent the flotation rate constant of particles of size class i and
RðtÞ ¼ R1 ½1  expðktÞ ð1aÞ
component j, n is the number of size classes and l is the number
    of components.
1 kt
RðtÞ ¼ R1 1  ¼ R1 ð1bÞ The full potential of these models have not been realised for
ð1 þ ktÞ 1 þ kt
coal flotation because the floatability distributions of the particle
classes, a key requirement for the utilisation of these models have
not been readily accessible experimentally for the models to be
⇑ Corresponding author. properly validated. Current implementation approaches for coal
E-mail address: philip.ofori@csiro.au (P. Ofori). utilise only a two-component system with fast and slow flotation

http://dx.doi.org/10.1016/j.mineng.2014.07.013
0892-6875/Ó 2014 Elsevier Ltd. All rights reserved.
78 P. Ofori et al. / Minerals Engineering 66–68 (2014) 77–83
rate constants kf and ks following Kelsall (1961) or as in the mod-
ification to include a non-floating component (Jowett, 1974). At
the heart of the selective flotation process is the floatability distri-
bution of the particles in the flotation environment; therefore a
quantitative estimation of floatability distribution of coal is abso-
lutely critical in the understanding and modelling of the behaviour
of coal flotation systems. There has not been clear-cut and direct
way of estimating the floatability of coal particles until now.
In minerals flotation, the liberation data provided by QEMSCAN
and Mineral Liberation Analyser (MLA) has allowed size by libera-
tion studies to be performed, such as the work of Sutherland
(1989). Recent work in minerals flotation area includes those of
Runge et al. (2004) and Welsby et al. (2010).
2. Background
2.1. Particle size and maceral composition-based Approach
A new approach has been developed that allows the coal flotation
feed to be fractionated into particle classes of size and maceral com-
position using a semi-automated microscopic characterisation tool
called Coal Grain Analysis. Microscopic analyses are conducted on
polished grain-mounts of the samples using an oil immersion lens fit-
ted to a reflected-light microscope. Microscope images were cap-
tured using the MACE300™ system for coal petrography. The
images are then processed to obtain compositional information on
each grain in the images. The abundances of vitrinite, inertinite, lipt-
inite and minerals and their respective areas are determined. The full
details of this characterisation tool and the information it provided
can be found in Jenkins and Kwan (2003) and O’Brien et al. (2007).
The microscopic characterisation of the individual grains pres-
ent in coal samples provides a unique opportunity to estimate
the flotation rate constants for the different grain classes. This pro-
vides a new tool for more detailed flotation performance analysis
and modelling. The coal flotation feed is divided into particles clas-
ses of size and maceral composition, each class with a distinct
degree of floatability and therefore a distinct flotation rate con-
stant. The summation of the recovery of the particles classes pro-
vides an estimate of the overall recovery. This approach provides
a framework for an innovative and robust pathway for flotation
process analysis and prediction of coal flotation outcomes.
The grains within each size fraction are classified into eight
composition classes consisting of liberated (single component)
and non-liberated (composite) grains. Additional composition clas-
ses can be accommodated if warranted. Single component grains
comprise of >95% of a single phase and other grains are classified
as composites. Microlithotype nomenclature is used for the single
component grains with grains comprising of >95% vitrinite, inerti-
nite, liptinite and minerals classified respectively as vitrite, inertite,
liptite and minerite. The composite grains were subdivided based
on their maceral/mineral associations as being either vitrinite-rich
(VitRich), inertinite-rich (InertRich), liptinite-rich (LipRich) or min-
eral-rich composites (MinRich) depending on the dominant macer-
al type by volume in that class.
2.2. Hydrophobicity of heterogeneous coal particles
Estimation of the hydrophobicity/floatability of heterogeneous
particles is critical in the understanding and prediction of the behav-
iour of particles in flotation systems. Previous work (Ofori et al.,
2010; Arnold and Aplan, 1989) had determined the contact angles
of discrete coal maceral groups of vitrinite, liptinite and inertinite
and associated minerals. Combining this information with the parti-
cle composition data from the Coal Grain Analysis methodology pro-
vides a pathway for estimating the contact angle of heterogeneous
coal particles via Cassie’s Equation. Using the contact angle data of
the flotation feed components as input data in new or existing flota-
tion kinetic models provides a unique opportunity to estimate the
flotation rate constants for the different particle classes.
The Coal Grain Analysis technique for microscopic characterisa-
tion of coal samples has made composition and area information of
particle classes accessible (O’Brien et al., 2007; Ofori et al., 2004,
2006). For coal of composite grain classes made up of n maceral
components such as vitrinite, inertinite and liptinite and also
minerals, a contact angle hc can be estimated from the composition
and fractional areas using Cassie’s Equation (Cassie, 1948) which
generalises to (Hey and Kingston, 2007):
X
n
cos hc ¼ xi cos hi ð3Þ
i
where hi is contact angle of discrete maceral component i, xi is the
fractional area of component i at the surface. The contact angles
for the different size and composition classes may be used as input
in distributed flotation kinetic models.
2.3. Distributed flotation kinetics models
A number of studies have examined experimental and theoret-
ical modelling approaches that aim to predict the flotation rate
constants for minerals from first principles. One of the first of these
studies was that of Sutherland (1948). Other investigators in this
area include Jameson et al. (1977), Yoon and Luttrell (1989),
Yoon (2000); Dobby and Finch (1987); Dai et al. (1999); Bloom
and Heindel (2003); Nguyen and Schulze (2004); Ralston et al.
(1999) and Runge et al. (2004).
In most of these studies, the models developed included flota-
tion rate constants described by a relationship of the following
form:
ki ¼ f ðH; O; Pi Þ ð4Þ
where ki is the flotation rate constant of particle class i, H is an
expression for cell hydrodynamics, O represents operating condi-
tions and Pi is the probability of capture of particle class i, which
depends on the particle floatability.
The flotation rate constant described by the above relationship
may be distributed in nature due to the floatability distribution of
the particles. When applied to coal flotation, the Coal Grain Analy-
sis tool may be used to classify the flotation feed into particle clas-
ses of size and maceral composition, each class with a distinct
degree of floatability and hence a distinct flotation rate constant.
The particle capture process represented by probability P in Eq.
(4) may be viewed as consisting of a sequence of sub-process
events, each with an associated probability of success. If these
probabilities are assumed independent of each other, the overall
probability of particle capture is the product of these probabilities,
(Schuhmann, 1942; Tomlinson and Fleming, 1965; Nguyen et al.,
1997). Then
P ¼ Pc  Pa  ð1  Pd Þ ð5Þ
where Pc is the probability of particle–bubble collision, Pa is the
probability of particle–bubble attachment and (1Pd) or Ps is the
probability that the particle will not detach from bubble.
Following the approach taken by Abrahamson (1978); Schubert
and Bischofberger (1978); Schubert (1999); Yoon (2000) and Pyke
et al. (2003), Eq. (4) may be written as:
2 0 112 3
 2 " #4=3
6 dBp 0:33e 4=9
@d14=9
j qp  qf j 14=9 7
k ¼ 45nB  p þ dB A 5  Pc  Pa  P s
2 m1=3 qf
ð6Þ
 P. Ofori et al. / Minerals Engineering 66–68 (2014) 77–83 79

Table 1 where Fo is the average driving force for the film drainage and rup-
Experimental conditions. ture over the distance hc to ho the critical and initial film
Condition Value thicknesses.
Air flow rate 4 L/min (±0.4 L/min)
For a successful particle capture, this film must thin and rupture
%Solids 10% during the contact time. The critical film thickness depends on the
Impeller rotation speed 1500 rpm surface properties of the particle and film. It increases with an
Froth scrapping rate Once every 15 s increase in surface hydrophobicity. Sharma and Ruckenstein
Pulp volume 4L
(1990) provide an approximate implicit analytical relationship as
Collector Diesel oil, (1 mL/kg dry solids)
Frother MIBC (0.1 mL/kg dry solids) follows:
 
2hc ð1  cos hÞ2 2hc
¼1þ 2
 exp ð10Þ
r1 sin h sin h r 1 sin h
where dp is particle diameter, dB is bubble diameter, dBp sum of
particle and bubble radii, (dp + dB)/2, e is the turbulent energy dissi- where h is the contact angle, r1 is hole radius taken to be 719 nm as
pation rate, m is the kinematic viscosity, qp is particle density and qf suggested by Sharma and Ruckenstein (1990). This relationship
is fluid density. may be further simplified by a truncated Maclaurin series expan-
The number of bubbles in a given volume V of flotation mixture sion of the exponential term and neglecting all but the first two
is determined by the gas hold-up eg and bubble size db as follows: terms, resulting in the following expression:

6V eg r 1 sin h
nB ¼ and eg ¼ Jg =ug ð7Þ hc ¼ ð11Þ
p 3
dB 2
The approximation given by Eq. (11) may be used to estimate
where Jg is superficial gas velocity and ug is bubble rise velocity.
the critical film thickness for particles of different hydrophobicities
By estimating Pc the probability of particle–bubble collision, Pa
as indicated by contact angles and used to estimate induction
the probability of particle–bubble attachment and (1Pd) the prob-
times from Eq. (9).
ability that the particle will not detach from bubble from available
or derived models for each of the particle classes, the distribution
2.3.3. Probability of particle–bubble aggregate stability
of rate constants may be obtained.
As the particle–bubble aggregates formed after collision and
attachment move through the slurry towards the froth zone, they
2.3.1. Probability of particle–bubble collision
are subjected to a number of stresses. If the resultant detachment
The particle–bubble collision process is controlled by the
force is higher than the hydrophobic forces tending to keep the
hydrodynamic forces generated by the fluid and the relative
aggregate together, detachments occur. The more hydrophobic
motion between particle and bubble. The particle–bubble collision
the particles, the more stable the attachment to the bubble.
process has been modelled in many different ways. The most pop-
The relationship for the probability of the formation of a stable
ular of these approaches is the route used by Sutherland (1948) in
aggregate given by Nguyen and Schulze (2004) for a flow field
his pioneering work.
dominated by tensile stress may be used:
The general relationship for the probability of collision, Pc, ( )
obtained by this approach may be given as 3rð1  cos hÞ
 n Ps ¼ ð1  Pd Þ ¼ 1  exp 1  2
ð12Þ
Rp dp ðam þ gÞ  Dq
Pc ¼ A ð8Þ
Rb
where r is gas liquid surface tension, g is acceleration due to grav-
where Rp and Rb are particle and bubble radius, A and n are param- ity, and Dq is the difference between the particle and fluid densi-
eters that depend on the Reynolds number of bubbles and therefore ties. The turbulent machine acceleration am is given by:
on the flow fields around the bubbles. Values or expressions for
am ¼ ð1:9Þe2=3 r 1=3 ð13Þ
these parameters have been derived by Sutherland (1948) for
potential flow conditions, by Yoon and Luttrell (1989) and Weber with r approximated by the sum of bubble diameter and particle
and Paddock (1983) for intermediate flow conditions and by diameter 2(Rb + Rp) and e is the turbulent energy dissipation rate.
Gaudin (1957) for Stokes flow conditions.
Flotation kinetics of combustible matter
2.3.2. Probability of particle–bubble attachment 100
Of the particles that collide with bubbles, only those that are 90
hydrophobic enough attach to the bubbles. The particle–bubble
80
Combustibles Recovery (%)

attachment process is strongly dependent on the chemistry and

surface physical chemistry of the particles and bubbles. Several 70
models have been proposed for the probability of attachment,
60
(Sutherland, 1948; Dobby and Finch, 1987; Yoon and Luttrell, Rec OrgTotal
1989 and Ralston et al., 1999). 50
Rec Org +0.25 mm
The degree of particle hydrophobicity is inherent in the induc- 40
tion time. Attachment efficiency is a function of bubble and parti- Rec Org 0.25x0.125
30
cle size and induction time. The induction time may be estimated Rec Org 0.125x0
using the model of Wang et al. (2005) which is based on the work 20
of Vinogradova (1995) on film drainage between hydrophobic sur-
10
faces. This is given by:
0
2 0 1 2 3
1:5pldb hc
ti ¼   2 ln ð9Þ Flotation Time (min)
db  ho
1 þ dp F o
Fig. 1. Flotation kinetics of combustible matter for different size classes.
 80 P. Ofori et al. / Minerals Engineering 66–68 (2014) 77–83

3. Experimental methods
Contact angle

Batch flotation experiments were conducted to examine the

kinetics of flotation of the different grain types of the selected coal
2.5
62.7
39.7

55.9
42.2
31.6
51.0
and to obtain flotation recovery data to validate the flotation
model. A laboratory Denver flotation cell set up according to the
Grain ash (%)

Australian Standard for coal flotation (Australian Standard, 2004)

was used. Total flotation time was 3 min. Timed increments of flo-
tation concentrates were taken at periods ending at 0.25, 0.5, 1, 2
1.2
3.9
69.9
16.4
17.4
67.5
23.6

and 3 min to provide flotation kinetics information. Diesel oil

collector and MIBC frother at the concentrations specified by the
Grain mass abundance

Australian Standard were used. A summary of the experimental

conditions are provided in Table 1.
The feed, tailings and each timed sample increment of the con-
centrate were divided into size classes (3 sizes) which were subse-
quently divided into composition classes (6 classes) using the grain
18.85

35.59
24.99
1.72
9.15

9.69
100.00

analysis technique. For these samples, liptinite was virtually

absent and hence liptinite and liptinite-rich components were
absent. The cumulative recovery of each of these classes could then
Grain density

be determined as a function of time.

1.24
1.32
2.66
1.38
1.43

1.48
2.20

4. Results
Grain volume abundance

4.1. Size partitioning

Partitioning of the 0.250  0.125 mm size fraction of flotation feed sample into different grain composition classes with estimated contact angles.

The flotation feed, concentrate and tailings were partitioned

into three size fractions; +0.250 mm, 0.250 + 0.125 mm, and
0.125 mm. The flotation rate data was examined as a function
1.93

6.51
22.51

38.11
25.85
5.09

100.00

of particle size and composition classes. The cumulative combusti-

bles recovery as a function of flotation time for the three size frac-
tions are shown in Fig. 1. The figure shows that the intermediate
Bright mineral

size class exhibited the highest flotation rate and ultimate recovery
and the finest size class had the lowest flotation rate and ultimate
recovery. It is widely known that the recovery of coal and minerals
0.02

0.02
0.00
0.00
0.00
0.00

0.00

in general is highest for the intermediate size classes. The lower

flotation rate of the fine particles is generally ascribed to the lower
probability of collision between the bubbles and particles as a
Dark mineral

result of the smaller mass and inertia. On the other hand, the lower
flotation recovery of the coarse particles is attributed to the higher
15.97
3.65
2.73
4.36
0.11

5.08
0.04

probability of particle detachment from the bubble as a result of

the high inertia of the coarse particles.
Inertinite

1.88

6.65
0.19

0.67
18.92

28.33
0.01

4.2. Component partitioning and rate constants

The various size fractions were further fractionated into compo-

Vitrinite

sition classes using the Coal Grain Analysis methodology. Table 2

55.68
4.18
1.48
22.20

27.80
0.01
0.01

shows the different grain classes (grain mass abundance) and esti-
mates of contact angles for the different grain composition classes
Inertinite rich (I > V, L, M)

for the 0.250 + 0.125 mm size fraction of the flotation feed. Sim-
Vitrinite rich (V > I, L, M)
Inertite (>95% inertinite)

Mineral rich (M > V, I, L)

Minerite (>95% mineral)
Vitrite (>95% vitrinite)

ilar data was generated for each size fraction of the feed, tailings
and timed sample increments of the concentrate. The recoveries
of the grain classes were then determined.
Fig. 2 is a plot of the cumulative phase recovery versus flotation
Grain class

time for the +0.250 mm size fraction. The figure shows that for this
size fraction, the liberated vitrinite (vitrite) component grains class
has the highest flotation rate and ultimate recovery. This is fol-
lowed by the vitrinite-rich composite grain class which has a
Single Component Grains

higher flotation rate and ultimate recovery than the liberated

inertinite phase. The component with the next highest rate of
Composite Grains

recovery and ultimate recovery is the inertinite-rich composite

grains. The rejection of the liberated minerals (minerite) in this
size fraction is almost complete, with low flotation rate and ulti-
mate recovery. The mineral-rich composite grains are recovered
Total
Table 2

to a greater extent than the minerite. The curve for the combined
grain classes (all grains) is also shown.
 P. Ofori et al. / Minerals Engineering 66–68 (2014) 77–83 81

Flotation kinetics of Coal Components (+0.250 mm) Flotation kinetics of Coal Components (0.125x0 mm)
100 100

90
90
80

Component Recovery (%)

80
70
Component Recovery (%)

70
60

60 50

+250 All Grain 40

50
+250 Vit Rec
30 125x0 All Grain rec
+250 Inert Rec
40 125x0 Vit Rec
+250 Min Rec 20 125x0 Inert Rec
+250 VitRich Rec 125x0 Min Rec
30 10 125x0 VitRich Rec
+250 InertRich Rec
125x0 InertRich Rec
20 +250 MinRich Rec 0 125x0 MinRich Rec
0 0.5 1 1.5 2 2.5 3 3.5
Flotation Time (min)
10

Fig. 4. Flotation kinetics of the different grain classes of the coal for the
0
0.125  0 mm size fraction.
0 0.5 1 1.5 2 2.5 3 3.5
Flotation Time (mins) recoveries of mineral-rich composites grains and liberated mineral
Fig. 2. Flotation kinetics of the different grain classes of the coal for the +0.250 mm
grains are higher than was the case with the coarser size fractions
size fraction. due to entrainment.
The flotation rate constants and the ultimate recoveries for the
different grain classes of the three size fractions were extracted
from the kinetic data by fitting the experimental data to the first
Flotation kinetics of Coal Components (0.250x0.125 mm) order flotation model (Eq. (2a)) using non-linear least squares
100
techniques and are presented in Table 3. Also shown is the product
90 of the flotation rate constant and ultimate recovery for a single
parameter comparison. Chi-squared goodness-of-fit tests showed
80 that the first order kinetic models of Eq. (2a) provided adequate
fit to the data at the 95% confidence level.
Component Recovery (%)

70
250x125 All Grain Rec

60 250x125 Vit Rec

50
40
30
20
10
0
0 0.5 1 1.5 2 2.5
Flotation Time (mins)
Fig. 3. Flotation kinetics of the different grain classes of the coal for the
0.250  0.125 mm size fraction.
Fig. 3 shows the cumulative phase recovery as a function of flo-
tation time for the intermediate size fraction of 0.250  0.125 mm.
In this size fraction the rate of recovery and ultimate recovery of
vitrite are still the highest, however unlike the coarser size fraction
they are only slightly higher than those of the inertite and the vitr-
inite-rich composite grains.
The rate of recovery and ultimate recovery of the inertinite-rich
composite grains are significantly lower than inertite and vitrinite-
rich composite grains. For this intermediate size fraction, signifi-
cantly higher mineral-rich composite grains and liberated minerals
are recovered than for the coarser +0.250 mm size fraction.
Fig. 4 shows the cumulative phase recovery as a function of flo-
tation time for the finest size fraction of 0.125  0 mm. For this size
fraction, vitrite has the highest ultimate recovery as well as rate of
recovery but not much higher than those of inertite, vitrinite-rich
composite grains and inertinite-rich composite grains. Also
250x125 Inert Rec
250x125 Min Rec
250x125 VitRich Rec
250x125 InertRich Rec
250x125 MinRich Rec
3 3.5
4.3. Comparison of experimental and calculated recoveries and rate
constants
Flotation rate constants were calculated using Eq. (6) in con-
junction with expressions for the probabilities of collision, stability
of the particle–bubble aggregates given by Eqs. (8) and (12) respec-
tively. The expression for the probability of attachment derived by
Yoon and Luttrell (1989) were used and the parameters used for
Eq. (8) were also those of Yoon and Luttrell (1989) for intermediate
flow regime. Estimates of contact angles were made from the pro-
portion of macerals in the grain classes and the contact angle of
individual macerals using the generalised Cassie’s Equation (Eq.
(3)). The calculations were done for the three size fractions that
experimental data was available for. The calculated rate constants
for the organic grain classes have been plotted as a function of size
(see Fig. 5) to give an indication of the trend of the rate constants of
the components with particle size. This figure shows the familiar
trend in coal flotation whereby the flotation rate and recovery at
the fine size fraction is lower than the intermediate size fraction
due to the lower probability of collision for the fine particle. It also
shows the lower flotation rate of the coarse particles due to the
higher probability of detachment of coarse particles. It can also
be seen from the figure that the calculated rate constants of the
components capture the previously observed tendency of coarse
inertinite and inertinite-rich composite grains to be poorly recov-
ered compared with vitrinite and vitrinite-rich composite particles.
The calculated flotation rate constants of the components of the
coal from the probabilistic model of Eq. (6) as well as those
obtained by fitting the plug flow model to the experimental data
are provided in Table 4 for the +0.250 mm and 0.250 +
0.125 mm size fractions. A Chi-squared goodness-of fit test of the
two sets of flotation rate constants show satisfactory agreement
82 P. Ofori et al. / Minerals Engineering 66–68 (2014) 77–83

Table 3
Experimental flotation rate constants for the different grain classes for three size fractions.

Size fractions Grain classes Flotation rate constant (k) Ultimate recovery (R1) R1*k
+0.250 mm Vitrite 6.43 95.2 611.8
Inertite 3.98 81.3 323.6
Mineral 0.78 1.2 0.84
Vit-Rich 4.78 84.4 403.4
In-Rich 3.99 68.6 273.7
Min-Rich 1.88 14.0 26.3
Total 4.80 65.2 313.4
0.250  0.125 mm Vitrite 6.11 96.1 586.8
Inertite 4.14 98.9 410.4
Mineral 1.52 12.4 31.3
Vit-Rich 4.28 95.0 407.6
In-Rich 2.88 86.1 247.9
Min-Rich 1.93 25.8 49.8
Total 4.79 68.0 325.7
0.125  0 mm Vitrite 2.63 98.2 258.4
Inertite 1.90 98.1 186.3
Mineral 0.77 29.9 23.2
Vit-Rich 2.28 96.7 220.2
In-Rich 1.93 91.3 176.6
Min-Rich 1.23 36.4 45.1
Total 1.87 62.0 115.8

the probabilistic model equations that needed to be estimated.

Areas needing more work include the estimation of induction time
6
from contact angles of composite particles required in estimating
the probability of particle–bubble attachment. This requires the
Flotation rate constant (min-1)

5
estimation of the average driving force for film drainage and the
initial and critical film thicknesses as required in Eq. (9) which is
4
not straightforward.

3
In-Rich 5. Conclusions
Vit-Rich
2 Inertite
Vitrite The Coal Grain Analysis tool has provided data that has enabled
1
the partitioning of the coal particles to product or tailings to be
studied in greater detail than ever before possible. With the aid
of this characterisation technique, for the first time there is the
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 potential to implement a property-based coal floatability model
Particle Size (mm) in which the floatabilities can now be easily linked to particle size
and composition. Based on this capability, a framework has been
Fig. 5. Flotation rate constant versus particles size for coal component grain classes. developed for extracting flotation rate constants for particle classes
based on size and composition. A particle size and composition
Table 4 dependent kinetic flotation model is used to generate data for
Comparison of calculated and experimental flotation rate constants for the different varying feed particle composition and size distribution. The model
grain classes. is assessed for its ability to predict the effects due to changes in
Size fractions Grain classes Flotation rate constant particle properties. Distributed flotation rate constants are used
Experimental Model
for fitting the effects of particle heterogeneity, accounting explic-
itly for particle composition.
+0.250 mm Vitrite 6.43 5.25
Inertite 3.98 2.05
Floatability distributions of different particle classes as indi-
Mineral 0.78 0.44 cated by contact angles are able to be estimated and used as input
Vit-Rich 4.78 3.77 data for distributed flotation kinetic models for predicting compo-
In-Rich 3.99 2.30 nent flotation rate constants. The model predictions have been
Min-Rich 1.88 0.45
compared with laboratory experimental data. The calculated com-
Total 4.80 4.1
0.250  0.125 Vitrite 6.11 5.2 ponent rate constants compared satisfactorily with the experimen-
Inertite 4.14 5.1 tal data for the coarser particle size fractions but more work is
Mineral 1.52 1.1 required to improve the outcomes for the very fine size fraction.
Vit-Rich 4.28 5.2 The model predictions were able to replicate the characteristic
In-Rich 2.88 1.87
Min-Rich 1.93 0.77
maximum in rate constant and recovery for the intermediate size
Total 4.79 5.1 classes as well as mimic previous experimental observation that
coarse inertite and inertinite-rich composite grains have lower rate
for the intermediate size fraction, marginal agreement for the constants than vitrite and vitrinite-rich composite grains.
coarse size fraction and poor agreement for the finest size fraction
at the 95% confidence level. More work is still needed to improve Acknowledgement
this outcome, particularly for the fine size fraction. The significant
differences between the two sets of flotation rate constants are not The funding provided by the Australian Coal Association
surprising when you consider the many secondary parameters of Research Program (ACARP) in Australia is gratefully acknowledged.
 P. Ofori et al. / Minerals Engineering 66–68 (2014) 77–83 83

References Ofori, P.K., O’Brien, G., Firth, B.A., Jenkins, B., 2004. Flotation process diagnostics and
modelling by coal grain analysis. In: Membrey, W., (Ed.), Proceedings of the
Tenth Australian Coal Preparation Conference, Australian Coal Preparation
Abrahamson, J., 1978. Collision rates of small particles in vigorously turbulent fluid.
Society, p. 102.
Chem. Eng. Sci. 30, 1371–1379.
Pyke, B., Fornasiero, D., Ralston, J., 2003. Bubble particle heterocoagulation under
Arnold, B.J., Aplan, F., 1989. The hydrophobicity of coal macerals. Fuel 68, 651–658.
turbulent conditions. J. Colloid Interface Sci. 265, 141–151.
Australian Standard AS 4156.2.1, 2004. Coal preparation part 2. 2: Froth flotation-
Ralston, J., Fornasiero, D., Hayes, R., 1999. Bubble-particle attachment in flotation.
sequential procedure. (Standards Australia).
Int. J. Miner. Process. 56, 133–164.
Bloom, F., Heindel, T., 2003. Modeling flotation separation in a semi-batch process.
Runge, K.C., Franzidis, J.P., Manlapig, E.V., 2004. A study of the flotation
Chem. Eng. Sci. 58, 353–365.
characteristics of different mineralogical classes in different streams of an
Dobby, G.S., Finch, J.A., 1987. Particle size dependence in flotation derived from a
industrial circuit. In: Lorenzen, L., Bradshaw, D.W. (Eds.), Proceedings of XXII
fundamental model of the capture process. Int. J. Miner. Process. 21, 241.
International Mineral Processing Congress, Cape Town, pp. 962–972.
Cassie, A.B.D., 1948. Contact angles. Discuss. Faraday Soc. 3, 11–16.
Schubert, H., 1999. On the turbulence-controlled microprocesses in flotation
Dai, Z., Fornasiero, D., Ralston, J., 1999. Particle–bubble attachment in mineral
machines. Int. J. Miner. Process. 56, 257–276.
flotation. J. Colloid Interface Sci. 217, 70–76.
Schubert, H., Bischofberger, C., 1978. On the hydrodynamics of flotation machines.
Garcia Zúñiga, H., 1935. La recuperación por flotación es una función exponencial
Int. J. Miner. Process. 5, 131–142.
del tiempo, (flotation recovery is an exponential function of time). Boln. Soc.
Schuhmann, R., 1942. Methods for steady state study of flotation problems. J. Phys.
Nac. Min. Santiago 417, 83–86.
Chem. 46, 891–902.
Gaudin, A.M., 1957. Flotation, second ed. McGraw-Hill, New York.
Sharma, A., Ruckenstein, E., 1990. Energetic criteria for the breakup of liquid films
Hey, M.J., Kingston, J.G., 2007. The apparent contact angle for a nearly spherical drop
on nonwetting solids surfaces. J. Colloid Interface Sci. 137 (2), 433–455.
on a heterogeneous surface. Chem. Phys. Lett. 447, 44–48.
Sutherland, K.L., 1948. Physical chemistry of flotation. XI: kinetics of the flotation
Imaizumi, T., Inoue, T., 1965. Kinetic consideration of froth flotation. Proceedings of
process. J. Phys. Chem. 52, 394–424.
the 6th International Mineral Processing Congress, Cannes (1963), pp. 563–579.
Sutherland, D.N., 1989. Batch flotation behaviour of composite particles. Int. J.
Jameson, G.J., Nam, S., Moo Young, M., 1977. Physical factors affecting recovery
Miner. Process. 2 (3), 351–367.
rates in flotation. Min. Sci. Eng. 9 (3), 103–177.
Tomlinson, H.S., Fleming, M.G., 1965. Flotation Rate Studies Mineral processing. In:
Jenkins, B.M., Kwan, H.A., 2003. Calibration of the MACE300™ system. In: 55th
Roberts, A. (Ed.), 6th International Minerals Processing Congress, Cannes, pp.
Meeting of the international committee for coal and organic petrology, Utrecht,
563–579.
The Netherlands 2003 B.
Vinogradova, O.I., 1995. Drainage of a liquid film confined between hydrophobic
Jowett, A., 1974. Resolution of flotation recovery curves by a differential plot
surface. Langmuir 11, 2213–2220.
method. Trans. Am. Soc. Mining Metall. Eng. 85, C263–C266.
Wang, W., Zhou, Z., Nandakumah, K., Masliyah, J.K., Xu, Z., 2005. An induction time
Kelsall, D.F., 1961. Application of probability in the assessment of flotation systems.
model for the attachment of an air bubble to a hydrophobic sphere in aqueous
Bull. Inst. Min. Metall. 70, 191–204.
solutions. Int. J. Miner. Process. 75, 69–82.
Nguyen, A.V., Schulze, H.J., Ralston, J., 1997. Elementary steps in particle–bubble
Weber, M.E., Paddock, D.J., 1983. Colloid and Interface Science, vol. 94(2), pp. 328.
attachment. Int. J. Miner. Process. 51, 183–195.
Welsby, S.D.D., Vianna, S.M.S.M., Franzidis, J.P., 2010. Assigning physical
Nguyen, A.V., Schulze, H.J., 2004. Colloidal Science of Flotation. Marcel Dekker Inc.,
significance to floatability components. Int. J. Miner. Process. 97, 59–67.
New York, pp. 640–645.
Yoon, R.-H., 2000. The role of hydrodynamic and surface forces in bubble-particle
O’Brien, G., Ofori, P., Jenkins, B., 2007. Semi-automated petrographic assessment of
interaction. Int. J. Miner. Process. 58, 129–143.
coal by coal grain analysis. Miner. Eng. 20, 428–434.
Yoon, R.H., Luttrell, G.H., 1989. The effect of bubble size on fine particle flotation.
Ofori, P., Firth, B., O’Brien, G., McNally, C., Nguyen, A, 2010. Assessing the
Miner. Process. Extr. Metall. Rev. 5, 101–122.
hydrophobicity of petrographically heterogeneous coal surfaces. Energy Fuels
24 (11), 5965–5971.
Ofori, P.K., O’Brien, G., Firth, B.A., Jenkins, B., 2006. Flotation process diagnostics and
modelling by coal grain analysis. Miner. Eng. 10, 102–113.