International Journal of Mineral Processing 112113 (2012) 6370

Contents lists available at SciVerse ScienceDirect

International Journal of Mineral Processing

journal homepage: www.elsevier.com/locate/ijminpro

Modelling and simulation of rougher otation circuits

J. Yianatos a,, C. Carrasco a, L. Bergh a, L. Vinnett a, C. Torres b
a
b

Automation and Supervision Centre for the Mining Industry, CASIM, Chemical and Environmental Engineering Department, Federico Santa Mara Technical University, Valparaso, Chile
Divisin El Teniente, Codelco, Rancagua, Chile

a r t i c l e

i n f o

Article history:
Received 7 June 2011
Received in revised form 9 June 2012
Accepted 14 June 2012
Available online 21 June 2012
Keywords:
Flotation
Modelling
Simulation
Rougher cells

a b s t r a c t
At present, large mechanical otation cells of 100 to 300 m3 are used in rougher operation in different industrial otation plants around the world. However, in spite of the advances in fundamental research and the
notable growth in equipment size, there is still a lack of reliable data for industrial otation modelling and
simulation. In this work, a procedure for modelling and simulating rougher otation banks that is based on
operating variables and parameters tted by empirical data from plant measurements is presented.
Recently, a new methodology for describing industrial otation by separating the collection and froth zones
has been developed. This approach consists of using a new apparatus for direct bubble load measurement
below the pulpfroth interface in industrial cells. The procedure allows independently estimating the froth
recovery and the collection zone recovery. Metallurgical characterisation is developed by plant sampling,
mass balance adjustment and applying a short-cut method for kinetic characterisation of otation banks. In
addition, the effect of grinding is characterised using otation models with distributed otation rates (for
each particle size class) and a low number of parameters.
From the experimental data, a non-linear distributed model to simulate rougher otation banks was developed, which allows describing the industrial operation as a function of the main operating variables
(e.g., mass ow rate, solid percentage, grinding degree, pulp level and supercial air rate). The simulator
was calibrated and validated using experimental data from the rougher operation at El Teniente Division,
Codelco-Chile.
The new approach is a practical tool that is useful for otation operation diagnosis, for robust design and
analysis of new control and optimisation strategies and for evaluation of the effect that design and operating
variables have upon the collection and froth zone performance in large otation cells. This methodology can
be applied to other otation operations, such as cleaner and scavenger circuits that use mechanical otation
cells.
2012 Elsevier B.V. All rights reserved.

1. Introduction
Two types of mechanical otation cells are employed in industrial
applications: self-aspirated and forced-air otation cells. The main
differences between the cells are the type and location of the rotor
and the mechanism used for the air input. In self-aspirated cells, the
rotor is located near the top of the cell, whereas in forced-air cells,
the rotor is located near the bottom.
Mechanical cell designs have included some modications, such as
the use of froth-crowders and internal radial launders. These modications have been focused on improving the froth transport and consequently the mineral recovery.
The minimum number of cells per bank can be estimated from a
hydrodynamic analysis assuming perfectly mixed cells. For example,
it has been observed that when using four or more cells in series,

Corresponding author.
E-mail address: [email protected] (J. Yianatos).
0301-7516/$ see front matter 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.minpro.2012.06.005

the time scale-up factor required to reach the same recovery of a

plug ow operation (batch) with the same otation rate constant is
close to 1.15 (Arbiter, 2000; Yianatos et al., 2006). Thus, the mixing
effect can be compensated by increasing the bank mean residence
time, e.g., increasing the number of cells or the cell size.
Bourke (2002) noted an important aspect related to the minimum
number of control points for proper bank operation. For instance, in a
ten-cell arrangement, 2-2-3-3, only four level control points are available, whereas a ve-cell arrangement, 1-1-1-1-1, has ve level control points. In plant practice, a minimum of 46 cells in series,
including at least four control points, is recommended.
The behaviour of a otation bank has been typically modelled as a
set of tanks in series with the same properties, i.e., the same residence
time distribution, otation rate and froth characteristics in each cell.
At present, important efforts have been made to characterise otation
circuits as distributed systems in order to obtain a more realistic description of the process. Maldonado et al. (2007) discussed a control
strategy to improve the grade of the concentrate using the froth
depth along a rougher otation bank as the manipulated variable.

64

J. Yianatos et al. / International Journal of Mineral Processing 112113 (2012) 6370

Tsatouhas et al. (2005) reported a decrease in the froth recovery

along a rougher bank in two industrial rougher circuits. Dobby and
Savassi (2005) reported a change of otation rate in the collection
zone along a bank of cells based on the results of batch tests.
The distributed character of a otation rougher bank requires
mass balances around each cell for solid and liquid phases in order
to describe the operating conditions downward the bank, i.e. mean
residence time per cell, froth depth, entrainment, and solid liberation,
among others. In addition, the performance of each otation cell in
terms of mineral recovery and concentrate grade depends on the otation rates in the collection zone and the cleaning effect controlled
by the froth depth. On this subject, Finch and Dobby (1990) described
industrial otation equipment separating the collection and froth
zone to identify the effects independently in order to estimate the
overall otation performance. Reviews on the collection and froth
zone modelling have been reported in literature by Finch and Dobby
(1990), Mathe et al. (1998) and Yianatos (2007), where detailed hydrodynamic characterisation of the collection zone and summaries of
the froth transport modelling are presented.
Several approaches for otation process simulation have been
proposed in literature which typically simplify or remove the froth
phase. For example, Sosa-Blanco et al. (1999) developed a simulator
to integrate the grinding and otation process, where the otation
cells were modelled as a perfect mixer with entrainment in the
froth phase; Ferreira and Loveday (2000) described the collection
zone as a perfect mixer and the froth phase in terms of the froth
recovery for a otation circuit with 3 nodes; Casali et al. (2002) proposed a dynamic rougher simulator without considering the froth
zone due to the close to plug ow behaviour; Maldonado et al.
(2007) considered that the volumetric concentrate owrate depends
on the froth depth linearly; among others.
Currently, several commercial simulators exist: the JK Simoat
mineral processing simulator (Schwarz and Alexander, 2009), which
includes operational variables such as supercial gas rate, bubble surface area ux, froth recovery, entrainment, water recovery, solids per
size class, mean residence times, overall otation rate constants and
mineral recovery per size class; the FloatStar mineral processing simulator (MinTek, 2012), which uses advanced process control to overcome design-related limitations and maximise the otation circuit
performance; the UsimPac mineral processing simulator (Metso,
2012), which considers the raw materials, products, water and
waste for equipment dimensioning, carrying out mass balances, energy consumption evaluation, among other results; and the MetSim
simulator (MetSim, 2009), which allows mass and energy balance development in complex otation plants while taking into account
chemical, control and equipment dimensioning aspects as well as
capital cost estimations, among others. In spite of the wide commercial application of the aforementioned simulators, in general the
software is closed for the user, with limited information about the
process modelling. An exception is the JK Simoat (Schwarz and
Alexander, 2009), in which the collection and froth zones are separately modelled. The collection zone hydrodynamic is modelled by a
perfect mixing regime whereas the froth zone is represented by a
probability of detachment as a function of the froth residence time.
On the other hand, the concept of using an overall otation rate, accounting for the collection and froth zones together, is still employed.
The overall rate constant depends on the bubble surface area ux, the
mineral oatability and the froth recovery.
The simulator reported in this paper is focused on providing a detailed phenomenological description of the otation process in which
a rougher bank has been modelled as a distributed system, i.e., each
cell is an independent unit that consists of two different zones: collection and froth, which are linked by mass balance and not hindered by
an overall rate constant. Moreover, the different liberation grades for
the mineralogical species were taken into account. In order to obtain
a representative description of the actual otation process, the

variations in the collection otation rates and residence time, as

well as the froth transport characteristics and the froth recovery
downstream from the bank of cells were considered. The simulator
has been successfully tested in the rougher otation SAG plant of El
Teniente, Codelco-Chile, which consists of four banks of seven selfaspirated Wemco cells of 130 m 3 (array 1-2-2-2).
2. Single-cell characterisation: pulp and froth model
For modelling, scale-up and analysis purposes, two zones are distinguished in the otation cell: (a) the collection zone, where the
particle-bubble aggregate is formed and carried to the pulpfroth interface (true otation) and (b) the cleaning zone (froth), located between the pulpfroth interface and the concentrate overow, where
entrained particles have the chance to drop back to the collection
zone. This zone is now described by the froth recovery. The global
cell recovery RG is related to the collection zone recovery RC and the
froth recovery RF, taking into account the overall mass balance around
the cell (Eq. (1)) (Finch and Dobby, 1990; Finch et al., 2008).
RG

RC RF
1RC 1RF

This approach is used for both (a) kinetic parameter estimation in

the collection zone based on industrial mass balance measurements,
which allow obtaining RG and RF, and (b) evaluation of the overall
mineral recovery in each cell using the previously calibrated simulator, i.e., using hydrodynamic, otation rate, granulometric and froth
zone parameters as input.
2.1. Collection zone, RC
The mineral recovery RC in the collection zone can be described by
the general expression Eq. (2) (Polat and Chander, 2000; Yianatos,
2007).



kt
Et F k dt dk
RC RMAX 1e

0 0

where E(t) is the residence time distribution function for continuous

processes with different mixing characteristics, F(k) is the otation
rate distribution for mineral species with different otation rates
and RMAX represents the maximum otation recovery at innite
time. The term (1-e kt) represents the mineral recovery of a rstorder process with invariant kinetic constant k, as a time function.
2.1.1. Residence time distribution, E(t)
The hydrodynamic of otation cells can be characterised by means
of the residence time distribution function, E(t). The mixing regime of
mechanical cells arranged in banks has been typically represented by
a model of N perfect mixers in series (Yianatos et al., 2001). However,
for describing the E(t) of single large cells, the perfect mixing condition is not valid based on the empirical evidence. At present, better
ts for industrial data can be obtained using alternative approaches
(Lelinski et al., 2002; Yianatos et al., 2008a). For example, the hydrodynamic of a otation cell has been well represented by means of the large
and small (LSTS) tanks in series model given by Eq. (3) (Yianatos et al.,
2008a).
Et

exptP =S  exptP = L 
S P

where L is the mean residence time of the large tank, S is the mean
residence time of the small tank and p is the dead time. In addition,
the ratio S/L is usually close to 0.1 in mechanical cells (Yianatos et
al., 2005). Eq. (3) was considered for the simulator implementation.

J. Yianatos et al. / International Journal of Mineral Processing 112113 (2012) 6370

65

Fig. 1 shows good agreement between the LSTS model and the experimental data recorded in a Wemco cell of 130 m 3.
2.1.2. Flotation rate distribution, F(k)
Several models have been proposed in order to evaluate the
otation rate distribution in otation cells. The Dirac delta function, (k kB), and rectangular models correspond to single parameter models, where (k kB) represents a single otation
rate constant, kB.
The rectangular model for the i-th size class (+100#, 100+ 325#,
325#, where # means mesh) and for the n-th cell, given by Eq. (4), was
used for simulating the otation rate distribution in a distributed manner
n
because of its exibility and low number of parameters, kmax,i
.
n
F i k

8
<

1
knmax;i
:
0

; if

0bkbknmax;i

; if

k > knmax;i

4
Fig. 2. Recovery prediction model for batch tests.

Then, the mineral recovery in the collection zone of class i and cell
n
n,RCi
, was calculated by means of Eq. (5):
n

RC;i RC

MAX;i


kt
n
Et F i kdt dk
1e

0 0

where RnC MAX,i is the maximum collection zone recovery of the i-th class
in the n-th cell. Notice that because of the amount of valuable mineral
decreases along the otation bank, the maximum recovery value in
each size class (i.e., ne, intermediate and coarse) also decreases. If
this change is not taken into account, the otation rate distribution
in each cell will be hindered by the variation of RnC MAX,i.
Therefore, the valuable mineral was described by a fraction with a
probability (> 0) of being oated and a non-oatable fraction. Thus,
the amount of oatable, valuable and non-valuable material feeding
a single cell in a bank was obtained from a mass balance in the previous cell, where the recoveries were known. To estimate the maximum recovery of the raw material entering the rougher circuit,
batch tests for each size class were carried out using the same feed
samples from the mass balance measurements. Fig. 2 shows a batch
test result along with the RMAX estimation per size class.
The same residence time distribution was considered in Eq. (5) for
each size class because the empirical data showed no signicant differences in the mean residence time and the distribution shape.
Samples for the mass balances per size class were obtained for a
complete description of the rst 3 cells and the overall bank. The results for the global and froth recovery in each cell allowed estimating

Fig. 1. Model t to industrial data (non-oatable tracer), L = 5.5 min, S = 0.5 min,
p = 0.8 min.

the kinetic parameters in cells 1, 2 and 3 and to extrapolate the results

for the bank. Additionally, the air ow rate, the froth height and the
bubble load were measured during the same sampling campaign in
order to characterise the rougher circuit.
To characterise the decrease of the otation rate distribution of
mineral class i along the rougher bank, Eq. (6) was used. The parameters were tted by regression. The model variables ( and n) were
chosen considering the signicant dependence of the otation rate
on the particle size and the cell number.
kratio

knmax;i
k1max;i

 
 
n1
exp
i

Eq. (6) allows obtaining the rectangular model parameters for the
n-th cell as a function of the rst one. The i parameter depends on
the characteristic size of the i-th size class, and is a t parameter.
Fig. 3 shows a comparison between the proposed model and the
plant data in three size classes for a chalcopyrite mineral.
2.2. Cleaning zone, RF
The cleaning zone was characterised by means of the froth recovery, which can be estimated from bubble load measurement along

Fig. 3. Plant data and model describing changes in the collection rate along the otation
bank.

66

J. Yianatos et al. / International Journal of Mineral Processing 112113 (2012) 6370

with a mass balance around the cell. Thus, the froth recovery, RF, was
calculated by Eq. (7) (Yianatos et al., 2008b):
C X C
RF
B J G AC X B

where is the maximum recovery in the froth, n is the froth stability

factor for the n-th cell and F, n is the gas mean residence time in the
froth of the n-th cell, which is given by Eq. (9) (Yianatos et al., 2008b):
F;n

H F;n G
JG

where HF,n is the froth height and G is the mean gas concentration
(hold-up) in the froth.
The empirical evidence showed that the froth stability of each cell
downstream from the bank is reduced, i.e., the froth selectivity
increases along the cells arrangement. Therefore, the n parameter
in Eq. (8) was modelled as follows:
n Mn RC;n :

The water recovery, RW, is dened as the fraction of feed water

reported to the concentrate stream and is given by Eq. (11):
RW

where B is the bubble load, JG is the supercial gas rate, AC the crosssectional area of the cell, C is the solid mass ow in the concentrate
and XC and XB are the concentrate and bubble load grade,
respectively.
It has been shown that Eq. (7) is useful for calculating the froth
recovery and the valuable minerals collected by true otation
from experimental data (Yianatos et al., 2008b). In addition, Eq. (8)
is usually employed for modelling the froth zone (Gorain et al.,
1998; Mathe et al., 1998; Zheng et al., 2004).



RF;n exp n F;n

2.3. Water recovery

10

where Mn is the amount of valuable mineral entering the n-th cell, RC,
n is the overall recovery of the collection zone and is a t parameter.
Notice that MnRC, n corresponds to the overall amount of valuable
mineral entering the froth zone by true otation.
Fig. 4 shows the Cu grade of the bubble load and at the top of froth
in a single cell, for different size classes. Based on these results, a
rather non-selective froth was observed, which is consistent with the results reported by Yianatos et al., 2008b. Therefore, the same froth
recovery per size class was assumed.

WC
WF

11

where WC corresponds to the water mass ow rate in the concentrate

and WF is the feed water mass ow rate to each cell. The calculation of
RW was carried out from solid percentage data and mass balance
results.
For simulation purposes, the water recovery was modelled as a
function of the froth height, the froth stability factor and the supercial gas rate for each cell (Eq. (12)):
RW;n exp

F;n n

JG

12

where is the maximum water recovery, and and are t parameters. Notice that Eq. (12) is similar to the froth recovery relationship
in Eq. (8). In addition, this equation considers the inuence of the
froth behaviour in the water recovery as a distributed system.
2.4. Gangue recovery
The gangue is recovered into the concentrate by entrainment caused by the water reported to the concentrate stream. The gangue ren
covered per size class in the n-th cell, RG,i
, was estimated by means of
Eq. (13) (Yianatos and Contreras, 2010).
n

RG;i EF i RW;n

13

where EFi is the entrainment factor per size class. The gangue recovery
per size class can then be obtained from RW measurements and
empirical EFi values.
3. Single-cell simulation
In order to evaluate the contribution of each individual cell in the
otation bank recovery, the collection zone recovery by true otation,
RC, the froth recovery of oatable minerals, RF, the water recovery, RW,
and the gangue recovery, RG, were calculated for each cell. In addition,
the mass ow rate of water and solids per size class in the tailings of
each cell were obtained by means of a mass balance.
The global recovery of mineral m, size class i, in cell n is given by
m

Ri;n

RC;i :RF;i
1RC;i RC;i :RF;i

14

Also, the concentrate grade X of mineral m, size class i, in cell n is

m
F i;n X F;m
i;n Ri;n

C;m

X i;n

2
P

3
P

m1 i1

Fig. 4. Comparison between Cu grade in the bubble load (CB) and at the top of froth (TOF).

15

m
F i;n X F;m
i;n Ri;n

where the supra-indices C and F refer to the concentrate and feed

streams.
Notice from Eq. (15) that three size classes, i, were considered:
coarse (+100#), intermediate (100 + 325#) and ne ( 325#) in
order to take into account different liberation grades. In addition,
two types of minerals were included for the simulation: valuable
mineral e.g., chalcopyrite and molybdenite and gangue.
The overall recovery and concentrate grade of the otation bank were
calculated by integrating the distributed result of each cell, which were
determined by Eqs. (14) and (15).

J. Yianatos et al. / International Journal of Mineral Processing 112113 (2012) 6370

Table 1
Simulation results for the rougher otation bank.
Feed
Flow rate
Froth depth
Solid percentage
Froth recovery, Cu
Collection recovery, Cu
Overall recovery, Cu

Cell 1 Cell 2 Cell 3 Cell 4 Cell 5 Cell 6 Cell 7

tph 900
803.9 790.4 782.1 775.7 770.7 765.6
cm
5
12
12
8
8
8
8
%
42.0
43.1
43.8
44.8
45.7
46.6
47.6
%
83.4
63.5
56.8
63.1
48.1
27.7
6.0
%
70.4
64.5
59.1
53.8
45.7
38.8
31.5
%
66.5
53.6
45.1
42.4
28.8
14.9
2.7

67

Therefore, the simulation requires a set of test and previously

measured variables to calibrate the otation models, including:
mass balances, feed particle size distribution, batch tests, bubble
load, supercial gas rate, residence time distribution and froth
recovery.
5. Results
5.1. Case of study

4. Summary
The general procedure for developing and calibration of the simulator includes the following steps:
i) Sampling the rst, second and third cells of the bank together
with sampling the otation bank as a whole.
ii) Characterise the streams in terms of solid percentage by
weight, and mass and grades (Cu, Mo and Fe) per size class.
iii) Carry out mass balance adjustment per size class and estimate
mineral recoveries.
iv) Using the feed sample, conduct batch test to determine the
RMAX values per size class.
v) Estimate froth recovery, Eq. (7), using the measured values of
XC, XB, B and JG. The adjusted value of concentrate C and the
cross sectional area is also required.
vi) Using results from steps iii) and v), obtain the collection zone
recovery per size class from Eq. (1).
vii) Estimate the otation rate parameters per size classes from
Eq. (5) based on the RTD model, the distributed RMAX,i n along
the bank per size class and the collection zone recovery results
from step vi).
viii) Characterise the decrease of the otation rate distribution per
size class along the rougher bank using Eq. (6).
ix) From Eqs. (8), (9) and (10) and results from step v), determine
the froth recovery parameter using HF, n, G and JG data along
the bank together with the amount of valuable mineral entering the froth by true otation. An value of 92% was assumed
in Eq. (8).
x) Estimate gangue recovery from water recovery. First, obtain
water recovery using the mass balance results together with
the solid percentage by weight. Second, estimate parameters
of Eq. (12) using the JG, and HF, n, and n values along the
bank. Determine gangue recovery from mass balances and
bubble load measurements (Yianatos and Contreras, 2010)
and estimate EFi factors per size class.
xi) Based on the previously adjusted parameters (by regression),
Eqs. (3) to (13), the overall mineral recovery and the concentrate
grade can be computed from Eqs. (14) and (15), respectively.

The rougher otation circuit (SAG plant) at El Teniente, CodelcoChile, was considered. The rougher circuit consists of four banks
with seven self-aspirated Wemco cells (130 m 3) in series, in a 1-22-2 array.
For the simulator calibration, measurements of feed, tail and concentrate grades per size class, supercial gas rate, bubble load, solid
percentage, gas hold-up, residence time distribution and froth depth
proles along the rougher bank were carried out in the rougher otation SAG plant (Carrasco, 2010). These results allowed obtaining the
simulator parameters as well as to validate them. Table 1 shows the
simulation results for Cu performance corresponding to a feed tonnage of 900 tph, a feed grade of 1%, a solid percentage by weight of
42% and 18% particle size (% + 100#) in a rougher bank.
Table 2 presents a comparison between the simulator outputs
based on operational parameters and plant results for two tests conducted in the rougher bank. The operational parameters of Test 1
were as follows: feed tonnage 790 tph, Cu feed grade 0.99%, Mo
feed grade 0.021%, Fe feed grade 5.0%, percentage + 100# 19.8%,
solids percentage by weight 41.5% and froth height (rst cell) 5 cm.
For Test 2, the operational parameters were: feed tonnage 650 tph,
Cu feed grade 0.95%, Mo feed grade 0.020%, Fe feed grade 5.5%, percentage +100# 22.0%, solids percentage by weight 40.0% and froth
height (rst cell) 3 cm. A reasonable agreement between the simulated data and the plant performance was observed.
5.2. Sensitivity analysis
For visualisation of the simulation responses, a sensitivity analysis
for different operating conditions was carried out. The main variables
for this analysis were particle size distribution (PSD), froth depth in
the rst cell of the bank, feed tonnage, feed grade and solid percentage. These variables are interesting for control purposes because of
the economic impact in terms of mineral recovery and concentrate
grade in industrial otation banks.
5.2.1. Effect of particle size distribution (PSD)
The effect of changes in the PSD on the cumulative Cu recovery
and the incremental grade along the otation bank was studied. The

Table 2
Comparison between plant performance and simulation results.
Test 1

Rougher recovery %

Final concentrate

07/12/2009

Cu

Mo

Cu grade %

Mo grade %

Solid %

Cu/Fe ratio

% 325 #

Plant data
Simulation
% Error

90.8
91.0
0.2

80.00
83.76
4.7

5.5
5.3
3.6

0.09
0.084
6.7

27.2
27.1
0.4

0.63
0.64
1.6

69.8
69.2
0.9

Test 2

Rougher recovery %

Final concentrate

14/12/2009

Cu

Mo

Cu grade %

Mo grade %

Solid %

Cu/Fe ratio

% 325 #

Plant data
Simulation
Error (%)

91.2
91.6
0.4

81.0
83.4
3.0

5.00
5.03
0.6

0.071
0.070
1.4

24.5
25.3
3.3

0.61
0.60
1.6

67.4
66.9
0.7

68

J. Yianatos et al. / International Journal of Mineral Processing 112113 (2012) 6370

Fig. 5. Comparison of (a) the cumulative Cu recovery and (b) the incremental Cu grade
as a function of the coarser feed content.

following variables were assumed constant for the simulation: feed

tonnage 800 tph, feed copper grade 0.98%, solid percentage by weight
40% and 8 cm froth depth along the bank. The increment in particle
size was represented by increasing the +100# fraction in the mineral
feed by 18% to 25%.
Fig. 5(a) shows a 1% decrease in the cumulative mineral recovery
when the particle size was increased by increasing the coarse material in the otation bank. From laboratory and plant experiments,
the ner classes ( 100#) showed a higher collection probability;
therefore, a coarser feed decreases the global recovery of the otation
bank (Carrasco, 2010). In addition, an increase in the incremental
concentrate grade was observed for the coarser feed (Fig. 5(b))
because of the decrease in the liberated ne gangue entrainment to
the concentrate.

5.2.2. Effect of froth depth

A change in the froth height of the rst cell, from 5 to 25 cm, was
considered in order to evaluate its impact on the mineral recovery
and concentrate grade of the otation bank. The rest of the operating variables was kept constant, i.e., feed tonnage 800 tph, 0.98%
feed copper grade, 40% solid percentage by weight, 8 cm froth
depth in all of the other cells and 20% + 100# in the mineral feed
size.

Fig. 6. Effect of froth depth change in the rst cell on (a) the cumulative Cu recovery
and (b) the incremental Cu grade.

Fig. 6(a) shows a 3% decrease in the cumulative mineral recovery

when the froth depth was increased from 5 to 25 cm, while Fig. 6(b)
shows the signicant impact that the froth depth had on the concentrate grade obtained in the rst cell. Also, a strong decrease in the incremental grade was observed downstream the rougher bank when
the rst cell was operated with froth depth of 25 cm. Finally, an increase of 1% in the rougher concentrate grade was obtained. Adequate
level control in the rst cell of a rougher bank is necessary to control
the entrainment of non-oatable material and the overall concentrate
grade in self-aerated cells. Otherwise, the grade of the rst cell constrains the maximum grade achievable along the whole otation
bank.
5.2.3. Effect of feed tonnage
The feed tonnage is relatively constant in the rougher otation
SAG plant because the four rougher banks are fed by two SAG mills,
reducing the variations in the feed ow rate. An increase in feed tonnage from 600 to 800 tph was analysed. Again, the following variables
were kept constant: feed copper grade 0.98%, solid percentage by
weight 40%, 8 cm froth depth along the bank and 20% +100# mineral
feed size.
Fig. 7(a) shows a decrease of 1%2% in mineral recovery from increasing the feed tonnage. In addition, a negligible effect of this
change on the incremental concentrate grade was observed
(Fig. 7(b)). The effect shown in the cumulative recovery is mainly
due to variation of the mineral residence time in the bank of cells.

J. Yianatos et al. / International Journal of Mineral Processing 112113 (2012) 6370

Fig. 7. Effect of feed tonnage change on (a) the cumulative Cu recovery and (b) the
incremental Cu grade.

5.2.4. Effect of feed grade

The effect of increasing the copper feed grade from 0.9% to 1.1%
was evaluated. In this case, the following variables were kept constant: feed tonnage 800 tph, 40% solid percentage by weight, 8 cm
froth depth along the bank and 20% + 100# mineral feed size.
A negligible effect on the cumulative recovery was observed when
the feed grade increased from 0.9% to 1.1%, as shown in Fig. 8(a).
However, a signicant impact on the incremental concentrate grade
was observed (Fig. 8(b)) because of the higher quality of the fresh
material entering the rougher bank. Based on the models described
in the previous sections, the key parameter related with the feed
grade is the froth stability factor, , which affects the froth and
water recoveries.
5.2.5. Effect of solid percentage
The solid percentage determines the mean residence time in otation
cells (rheological effects were not considered). An increase in solid percentage by weight from 30% to 40% was analysed while keeping the
rest of the operating variables constant: feed tonnage 800 tph, 0.98%
feed copper grade, 8 cm froth depth along the bank, and 20% +100#
mineral feed size.
Fig. 9(a) shows an increase in the cumulative recovery by changing
the solid percentage from 30 to 40%, due to the longer residence time of
minerals in the otation cells. Also, a slight increase in the concentrate
grade was observed in Fig. 9(b), because the water recovery to the
concentrate was slightly lower which reduced the gangue entrainment
relative to the valuable solid recovered into the concentrate.

69

Fig. 8. Comparison of (a) the cumulative Cu recovery and (b) the incremental Cu grade
as a function of the feed grade.

The modelling of the otation cells by separating the collection

and froth zones, without using an apparent rate constant, allows for
a better identication of the true otation process (collection efciency) as well as the froth separation process (product quality). Changes
in the main operating variables yield to the expected mineral recovery and concentrate grade trends, while the effects on the mineral
collection and the froth separation were clearly identied. In addition,
the distributed approach for the rougher bank modelling involves the
decrease of mean residence time (from actual RTD), froth stability
and otation rates, together with the increase of particle size,
along the bank. All these information was obtained from plant
measurements.
The model does not consider reagents as manipulated variables
explicitly, but they are partially represented by the otation rate prole (per size class) along the bank, Fig. 3. Bigger efforts must be made
to describe the rougher circuit performance as a function of the collector and frother dosage downward the bank.
6. Conclusions
A procedure for modelling and simulating rougher otation banks
as a distributed process was developed. The model parameters were
calibrated from plant operating data using a otation rate evaluation
based on the short-cut method, adjusted mass balances and a new approach to evaluate bubble load and froth recovery.
The simulator was validated using experimental data from the
rougher operation at El Teniente concentrator, which consists of otation banks of seven 130 m 3 cells in series. The simulation allowed

70

J. Yianatos et al. / International Journal of Mineral Processing 112113 (2012) 6370

References

Fig. 9. Comparison of (a) the cumulative Cu recovery and (b) the incremental Cu grade
as a function of feed solid percentage.

the evaluation of the industrial rougher otation bank as a function of

the main operating variables, including particle size distribution,
mass ow rate, solid percentage, pulp level and feed grade.
The new approach is a practical tool that is useful for otation operation diagnosis, for robust design and analysis of new control and
optimisation strategies, and for evaluating the effect that design and
operating variables have upon the metallurgical performance in
large otation cells.
This methodology can be applied to other otation operations,
such as cleaner and scavenger circuits, that use mechanical otation
cells.
Acknowledgements
Funding for process modelling and control research was provided
by CONICYT, project Fondecyt 1100854, NEIM, Project P07-087-F,
ICM-MINECON, Santa Maria University, project 270522.

Arbiter, N., 2000. Development and scale-up of large otation cells. Miner. Eng. 52 (3),
2833.
Bourke, P., 2002. Selecting otation cells: how many and what size? MEI on-line, www.
min-eng.com2002.
Carrasco, C., 2010. Modelacin Zona de Coleccin y Espuma para Supervisin y Control
Industrial, Master Thesis, Chemical Engineering Department, Santa Mara University,
Chile, 2010.
Casali, A., Gonzalez, G., Agusto, H., Vallebuona, G., 2002. Dynamic simulator of a rougher
otation circuit for a copper sulphide ore. Miner. Eng. 15 (4), 253262.
Dobby, G., Savassi, O., 2005. An advanced modeling technique for scale-up of batch otation results to plant metallurgical performance. Proceedings of the Centenary of
Flotation Symposium, Brisbane, Australia (June), pp. 99103.
Ferreira, J.P., Loveday, B.K., 2000. An improved model for simulation of otation circuits.
Miner. Eng. 13 (1415), 14411453.
Finch, J.A., Dobby, G.S., 1990. Column Flotation. Pergamon Press.
Finch, J., Nesset, J., Acua, C., 2008. Role of frother on bubble production and behaviour
in otation. Miner. Eng. 21 (1214), 949957 (November).
Gorain, B., Harris, M., Franzidis, J.-P., Manlapig, E., 1998. The effect of froth residence
time on kinetics of otation. Miner. Eng. 11 (7), 627638.
Lelinski, D., Allen, J., Redden, L., Weber, A., 2002. Analysis of the residence time distribution in large otation machines. Miner. Eng. 15 (7), 499505.
Maldonado, M., Sbarbaro, D., Lizama, E., 2007. Optimal control of a rougher otation
process based on dynamic programming. Miner. Eng. 20, 221232.
Mathe, Z.T., Harris, M.C., O'Connor, C.T., Franzidis, J.-P., 1998. Review of froth modelling
in steady state otation system. Miner. Eng. 11 (5), 397421.
MetSim, 2009. The Premier Process Simulator for WindowsRetrieved September 2009
from , http://www.metsim.com/2009.
Metso, 2012. USIM PAC simulatorRetrieved March 2012 from , http://www.metso.
com/2012.
MinTek, 2012. FloatStar Dynamic SimulatorRetrieved March 2012 from , http://www.
mintek.com.au/2012.
Polat, J., Chander, S., 2000. First order otation kinetics models and methods for estimation of the true distribution of otation rate constants. Int. J. Miner. Process.
58, 145166.
Schwarz, S., Alexander, D., 2009. JKSimFloat v6.1plus: improving otation circuit performance by simulationRetrieved September 2009 from , http://www.jktech.com.au/
2009.
Sosa-Blanco, C., Hodouin, D., Bazin, C., Lara-Valenzuela, C., Salazar, J., 1999. Integrated
simulation of grinding and otation application to a leadsilver ore. Miner. Eng.
12 (8), 949967.
Tsatouhas, G., Grano, S.R., Vera, M., 2005. Case studies on the performance and characterization of the froth phase in industrial otation circuits. Miner. Eng. 19,
774783.
Yianatos, J., 2007. Fluid ow and kinetic modelling in otation related processes: columns and mechanically agitated cells a review. Chem. Eng. Res. Des. 85
(2007), 15911603.
Yianatos, J., Contreras, F., 2010. Particle entrainment model for industrial otation cells.
Powder Technol. 197 (3), 260267.
Yianatos, J.B., Bergh, L., Condori, P., Aguilera, J., 2001. Hydrodynamic and metallurgical
characterization of industrial otation banks for control purposes. Miner. Eng. 14
(9), 10331046.
Yianatos, J.B., Bergh, L.G., Daz, F., Rodrguez, J., 2005. Mixing characteristics of industrial
otation equipment. Chem. Eng. Sci. 60, 22732282.
Yianatos, J., Henriquez, F., Oroz, A., 2006. Characterization of large size otation cells.
Miner. Eng. 19, 531538.
Yianatos, J., Bergh, L., Tello, K., Daz, F., Villanueva, A., 2008a. Residence time distribution in single big industrial otation cells. Miner. Metall. Process. 25 (1), 4652.
Yianatos, J.B., Moys, M.H., Contreras, F., Villanueva, A., 2008b. Froth recovery of industrial otation cells. Miner. Eng. 21 (1214), 817825 (Selected Papers from Flotation '07).
Zheng, X., Franzidis, J.-P., Manlapig, E., 2004. Modelling of froth transportation in
industrial otation cells, part I. Development of froth transportation models for
attached particles. Miner. Eng. 17, 981988.