See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245158515

Control of pulp levels in flotation cells

Article  in  Control Engineering Practice · January 2003

DOI: 10.1016/S0967-0661(02)00142-9

CITATIONS READS

12 174

4 authors, including:

Sirkka-Liisa Jamsa-Jounela
Aalto University
178 PUBLICATIONS   1,210 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Artificial Intelligence applied for industrial lime kiln control View project

PAPYRUS View project

All content following this page was uploaded by Sirkka-Liisa Jamsa-Jounela on 09 June 2019.

The user has requested enhancement of the downloaded file.

 Author's accepted manuscript, published in Control Engineering Practice 11 (2003) 73–81

Control of pulp levels in flotation cells

S.-L. J.ams.a-Jounela*, M. Dietrich, K. Halmevaara, O. Tiili
Laboratory of Process Control and Automation, Aalto University, P.O. Box 6100, Kemistintie 1, 02150 Espoo, Finland

Abstract

The control of pulp levels in flotation cells is a complex control task. Many different aspects such as cell structure and valve sizing,
and their effects on control, have to be taken into account. Control becomes more and more difficult as the complexity of the cell
configurations and the number of cells used in configuration increase. In this paper the mathematical model of a flotation cell is first
developed and several configurations of cells are constructed. Simulations are then performed in order to determine suitable control
parameters for PI and feed-forward controllers of the cell levels. Finally, the effects of valve sizing and cell structure are presented
and discussed.
r 2003 Elsevier Science Ltd. All rights reserved.

Keywords:
Level control; Control valves; Characteristic curves; Valve sizing; Flotation

1. Introduction On an industrial scale, flotation is usually performed

in large cells connected in series. In addition to the tank
A flotation plant control system usually consists of itself, each flotation cell also contains a mechanism for
conventional single-loop controllers. Due to the physical injecting air into the bottom of the cell through the
characteristics of the flotation process, especially the rotors; the launders are located at the top of the cell, as
disturbances caused by varying composition of the raw well as occasionally structures to funnel the froth
ore, their efficiency has been found to be limited. Some towards the launders.
new advanced control methods have been recommended In the control of flotation circuits the froth height can
for the control of flotation processes (Stenlund & be used to control either the concentrate or the tailings
Medvedev, 2000). Niemi, Maijanen, and Nihtil.a grade from a specific flotation bank. Increasing the froth
(1974), Koivo and Cojocariu (1977) used a single cell height increases the particle residence time in the froth
model when developing an optimal control algorithm phase, thus increasing gangue drainage. This produces a
via applications of the maximum principle. Andersen higher concentrate grade, although often at the expense
et al. (1979) and Zargiza and Herbst (1987) reported an of a slight loss of recovery. Advanced control strategies
application of state feedback control and a Kalman often use cell levels as manipulated variables. Therefore
filter for rougher flotation control. Hammoude and properly tuned and well performing level controls are a
Smith (1981) used a linear model to develop a minimum- basis for successful flotation control.
variance controller for re-cleaning. However, the use A set of conventional single-loop controllers is
and maintenance of control systems based on modern commonly used, although many problems have been
control theory have proved to be problematic in the reported, e.g. cyclic or interactive disturbances that
industrial environment. As a result, expert control frequently occur in plants with connected stages and
strategies for the logic control of situations not governed recycled streams (Schubert, Henning, Hulbert, & Craig,
by continuous control have been introduced (J.ams.a & 1995). The major disturbances in the operation of a
Herbst, 1988). flotation circuit are caused by changes in the properties
of the ore, or by instabilities in the grinding. Large
variations in the flow rate of the flotation feed caused by
fluctuations in the performance of the grinding circuit
r 2003 Elsevier Science Ltd. All rights reserved.
are not uncommon in concentrators. Disturbances of
this kind are undesirable because they are transmitted The cells under study do not have the properties of
through the flotation circuit, causing instabilities in its ideal tanks because the cross-sectional area of the cell is
performance. not constant. A mathematical model for the physical
The aim of this research is to study flotation control properties will be developed and discussed next.
mainly from the point of view of cell level control by
developing a simulation platform for controller design. 2.1. Single cells in series
The first part of the research consisted of a detailed
study of the effects of cell configuration and valve sizing In flotation several single cells are connected in series
on the control of flotation cells in series. The final aim of as shown in Fig. 1.
the project is the development of software for the plant The change in volume of the pulp with respect to time
operators, which provides advice in the selection of can be denoted for each cell in series using continuity
tuning parameters for the cell level controls. equations.
In the following the mathematical model of a flotation For the first cell in the series:
cell is first developed and several cell configurations are qV1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
constructed. The simulations are then performed in ¼ ðq  F1 Þ ¼ q  KCv ðu1 Þ y1  y2 þ h1 ; ð1Þ
qt
order to determine suitable control parameters for PI
and feed-forward controllers of the cell levels. Further- where q is the feed rate to the first cell, y1 the pulp level
more, a study is carried out on the effect of valve sizing in the first cell, y2 the pulp level in the second cell, h1 the
on control by sizing the valves with equations obtained physical difference in height between the cells, u1 the
from the literature. The developed platform is utilised in control signal, K the constant coefficient, F1 the outflow
further research on control strategies (K.ampj.arvi & from the first cell and Cv the valve coefficient.
J.ams.a-Jounela, 2002). Simulation was performed with The equations for cells 2, 3 and 4 are, respectively
Matlab 5.2.0 and its Simulink library. (i ¼ 2; 3 and 4)
qVi
¼ ðFi1  Fi Þ
qt pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2. Mathematical modelling of flotation cells in series ¼ KCv ðui1 Þ yi1  yi þ hi1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 KCv ðui Þ yi  yiþ1 þ hi : ð2Þ
In the flotation process the pulp is fed into the first cell
and the froth is collected in the launders. The feed can
be measured, e.g. by a magnetic flow meter, and the The equation for the last cell (n ¼ 5) is
remaining pulp flows into the next cell. The magnitude qVn
¼ ðFn1  Fn Þ
of the flow depends on the pressure difference between qt pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
two adjacent cells, the position of control valves, and the ¼ KCv ðun1 Þ yn1  yn þ hn1
viscosity and density of the pulp. The pressure difference pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
can be determined from the physical height difference of  KCv ðun Þ yn þ hn : ð3Þ
the cells. The pulp level in a cell is measured and
controlled by adjusting the control valve. In the case of an ideal tank, the cross-section of a cell is
In the following, a flotation cell is considered as a assumed to be constant. The pulp levels in the cells can
tank of perfectly mixed pulp. Since the pulp is perfectly therefore be written as
mixed, the density is constant throughout the cell, i.e. qy1 ðq  F1 Þ
there are no spatial density gradients in the cell. As the ¼
qt A1
froth flow is small compared to the pulp flow, it is pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ignored in the outgoing flows. The impact of the air feed q  KCv ðu1 Þ y1  y2 þ h1
¼ ; ð4Þ
on the pulp level is also ignored. A1

y1
y2
y3
q
y4
y5
u1
u2
u3
u4
u5

Fig. 1. Schematic diagram of a flotation series.

 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi The outflow of the second double cell is similar to that
qyi Ki1 Cv ðui1 Þ yi1  yi þ hi1
¼ of a single cell, and the flow into the second cell is the
qt Ai
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi outflow of the first. Therefore the mathematical model is
KCv ðui Þ yi  yiþ1 þ hi
 ; ð5Þ qV2
Ai ¼ F1  F
qt pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where i ¼ 2; 3; 4:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 2gðy1  y2 Þll hl
qyn KCv ðun1 Þ yn1  yn þ hn1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼  K2 Cv ðu1 Þ y2  y3 þ h1 : ð10Þ
qt An
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KCv ðun Þ yn þ hn
 ; ð6Þ
An
2.3. Modelling the non-linear flotation cells
where n ¼ 5:
Due to the froth collecting structures inside the
2.2. Double cells in series
modelled non-linear flotation cells (Fig. 3) the relation
between the pulp volume and level is not linear.
Mathematical models for the double cells can be
The effect of pulp level non-linearity can be seen in
derived in a similar manner. In a double cell, both pulp
Fig. 4, in which the pulp level is presented in ideal tank
levels are controlled by manipulating a control valve in
conditions as a function of the pulp volume. The cell
the second cell outflow, as can be seen from Fig. 2. The
type is TC-50, the maximum volume of which is 50 m3.
cells are physically on the same level.
The pulp level in the second cell is measured and
compared with the set point. This signal is used as input 2.4. Valve sizing and characteristic curve of the valves
to a digital PI controller. The output signal of the
controller is the desired valve position as denoted by u1 : Valve sizing is based on the Cv value, which is
The cells in a double cell are connected via a flange. The calculated according to the ISA standard as follows:
flange limits the maximum flow between the cells. The rffiffiffiffiffiffi
r
pressure difference in the cells is the only driving force Cv ¼ 1:17Q ; ð14Þ
Dp
for the flow, and depends on the density of the pulp.
The velocity of outflow from the first cell can be where Q is the flow rate (m3/h), Cv the valve capacity
derived using Bernoulli’s equation, resulting in coefficient, r the pulp density (kg/m3), and Dp the
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pressure difference over the valve.
v ¼ 2gðy1  y2 Þ; ð7Þ
where g is the gravity, y1 the pulp level in the first cell, y2 R r,out LC LB
the pulp level in the second cell.
The volume flow across the flange can be calculated as
LR rC r2 rB
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
HC
F1 ¼ vAflange ¼ 2gðy1  y2 Þll hl ; ð8Þ
where Aflange is the cross-sectional area of the flange, ll HB
the length of the flange, and hl the height of the flange.
The pressure drop due to flowing resistances is h a
assumed to be negligible and can be ignored. The
change of the pulp volume in the first cell with respect to
time can be written as
qV1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ q  F ¼ q  2gðy1  y2 Þll hl : ð9Þ
qt

y1 y2 Hlowe rpa rt
y3
q
F1 F2
F3
R tota l
Fig. 2. Double flotation cell series. Fig. 3. Cross-section of a non-linear flotation cell.
 TC50, linear and nonlinear pulp level 0.15
4.6

4.4 0.1

Change in valve position

4.2 0.05
y = 0.5628x
4
0
3.8 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
-0.05
3.6
level

-0.1
3.4

3.2 -0.15

3
The flow difference of normal flow (m3/s)

2.8
Fig. 5. Change in valve position as a function of the flow difference for
a 525 mm valve.
2.6
40 45 50 55 60
volume

Fig. 4. Pulp level as a function of pulp volume.

by measuring the flow rate, and then to open or close the
valves of the cell in order to compensate the disturbance
as fast and efficiently as possible.
When using feed-forward control, compensation is
The mean flow rate (m3/h) through the cell is
linearly dependent on the difference between the current
calculated as
inflow and the normal inflow. An example of the change
Vcell in valve position as a function of the difference in flow
Q ¼ 1:2 ; ð15Þ
t=60 compared to the normal flow conditions is presented in
where Vcell is the cell volume (m3) and t is the pulp Fig. 5. The correlation is determined by using the least-
retention time in the cell. square method. The achieved change in the valve
The valves in the models are sized for a flow having a position is added to the valve position signal coming
retention time of 1.5 min in one cell. Characteristic from the PI controller. The measurement signal is
curves of the control valves are produced by Larox filtered in order to prevent the feed-forward control
Flowsys and were used in the Simulink models in order from reacting to random variation in the flow.
to cause realistic and non-linear behaviour for the
valves.
4. Simulations

3. Cell level control 4.1. Cell configurations

In industrial applications each flotation cell has In the simulations the configurations of three and six
separate PI controller for the level control via the TC-50 cells in series were studied (1) with the ideal tank
control valve of the cell outflow. In the simulation study assumption, and (2) with the non-linear cell model. The
the following PI algorithm is used: simulations were performed by using the valves sized
  according to the ISA standard and with 100% over-
T sizing. Control strategies with conventional PI control-
Dui ¼ K ðei  ei1 Þ þ ei ; ð16Þ
I lers and with feed-forward control were simulated. The
where Du is a calculated change in the control output, K simulated configurations are presented in Fig. 6.
the proportional gain, I the integration time, ei a current
error term (set point—process variable), ei1 a previous 4.2. Performance criteria
error term (T seconds ago), and T is the sampling
interval (seconds). The performance criteria set for the comparison of
In industrial process control applications (Niemi et al., alternative designs are overshoot, rise time (i.e. time
1974) feed-forward compensation of measurable dis- for the response to reach the desired value for the
turbances is commonly and successfully used. Available first time) and settling time (time for the response
feed flow measurement makes the feed-forward level to settle within 75% of the desired value). The closed
control possible. loop responses are used for calculating the dynamic
The main purpose of the feed-forward level control is performance criterion of integral of the square
to observe the disturbance in the inflow of the first cell error (ISE) and integral of the absolute value of
 Fig. 6. The simulation configurations of conventional PI controllers and PI controllers with feed-forward control.

the error (IAE): compensate in configurations based on the ideal tank

Z t2 assumption.
ISE ¼ ðyðtÞ  ysp ðtÞÞ2 dt; ð17Þ The set point changes could be performed fluently in
t¼t1 the case of the ideal tank assumption, as can be seen
Z from Fig. 11. There were considerable differences in
t2   oscillation and settling time between the ideal tanks
IAE ¼  yðtÞ  ysp ðtÞ dt: ð18Þ
t¼t1 and non-linear cells. The new set point was reached
rapidly and with only small oscillation in the case of
ideal tanks.
4.3. Simulation schemes Feed-forward control applied to double cell config-
urations improved control performance in a similar
The simulation schemes were constructed with Ma- manner as in the single cell configurations, as can be
tlab 5.2.0 Simulink software. A simulation scheme of seen in Fig. 12 and Table 4.
non-linear cell is presented in Fig. 7.

5. Simulation results 6. Pilot plant tests

The simulation results for systems with the ideal tanks The influence of valve sizing and additional feed-
and non-linear cells show that the cell levels can be forward control were also studied in a pilot plant
controlled more accurately using valves sized according (J.ams.a-Jounela, Laurila, Karesvuori, & Timperi, 2001).
to the ISA standard (Fig. 8 and Table 1). There were The test configuration consisted of a flotation cell, pulp
only minor differences in overshoots and in the settling circulation cell, pulp pump, air feed device and two
times. different control valves. The inflow was measured with a
A somewhat larger improvement in cell level control magnetic flow meter and the pulp level with a float. The
was accomplished when feed-forward control was tests were performed with a flotation cell with similar
implemented. In the case of disturbances in the feed, geometries as described above. The results achieved
the overshoots performed by the last cells are signifi- were analogous to those obtained in the simulations
cantly smaller (Fig. 9 and Table 2). (Fig. 13). The control performance was noticeably
The results in Fig. 10 and Table 3 also demon- improved by adding the feed-forward control. It was
strate that there can be differences in control also found that an adaptive feed-forward control gain
performance between ideal tanks and non-linear could result in even greater improvements in control
cells. The disturbances in feed were easier to performance.
 2
Outflow of the
1st tank
Volume
H(2) MATLAB
Step H Function 1 MATLAB
1
h2+H-h3 Sqrt s Function
1 Tank level
dV V Level
Level in the Flow out of
calculation
following tank 2nd tank

Hlip 2

lip height h-hlip Outflow

e x
-0.12 x2 Cv
e2 Discrete PI1
Set point for level xPI+xFF Larox 350 Dynamics
in the tank sp+nsp 3
(from lip height) xFF
Error2
1 1
To Workspace x2
Saturation1 Rate Limiter 350mm
Cv

New set point error Out1

0

ISE Display1

error Out1
0 Valve 2

IAE Display2

Fig. 7. Simulink model of the flow dynamics in a flotation cell.

4.25 4.25

4.20 4.20

4.15 4.15

4.10 4.10

4.05 4.05

4.00 4.00

0 100 200 300 400 500 0 100 200 300 400 500

Fig. 8. Response of the levels to a +20% change in pulp feed. On the left for 100% valve oversizing, and on the right for valve sizing according to
the ISA standard.

Table 1
The influence of valve sizing: response of the pulp level in the last cell to a +20% change in the feed

Configuration Overshoot (cm) Overshoot (%) Rise time (s) Settling time (s) Dead time (s) ISE IAE

TC50 111111 525 3.2 0.762 2.5 55 7 0.01248 0.6838

TC50 111111 350 2.8 0.667 2.8 49 8 0.01221 0.6815
 4.25 4.25

4.20 4.20

4.15 4.15

4.10 4.10

4.05 4.05

4.00 4.00

0 100 200 300 400 500 0 100 200 300 400 500

Fig. 9. Response of the levels to a +20% change in pulp feed. 100% valve oversizing is used in both figures. On the left PI controller, and on the
right PI controller and feedforward controller.

Table 2
The influence of feed-forward control: response of the pulp level in the last cell to a +20% change in the feed

Configuration Overshoot (cm) Overshoot (%) Rise time (s) Settling time (s) Dead time (s) ISE IAE

TC50 111111 525 3.2 0.762 2.5 55 7 0.01248 0.6838

TC50 111111 525 FF 1.84 0.438 2.4 55 0.5 0.00216 0.2686

4.25 4.25

4.20 4.20

4.15 4.15

4.10 4.10

4.05 4.05

4.00 4.00

0 100 200 300 400 500 0 100 200 300 400 500

Fig. 10. Response of the levels to a +20% change in pulp feed. 100% valve oversizing is used in both figures. On the left are non-linear cells, and on
the right model based on the ideal tank assumption.

Table 3
The influence of cell geometry: response of the pulp level in the last cell to a +20% change in the feed

Configuration Overshoot (cm) Overshoot (%) Rise time (s) Settling time (s) Dead time (s) ISE IAE

TC50 111111 525 nonlinear 3.2 0.762 2.5 55 7 0.01248 0.6838

TC50 111111 525 linear 1.35 0.321 3 42 10 0.003334 0.3956

7. Conclusions the case of ideal tanks than with cells. Higher propor-
tional gains reduce the settling time of the system in PI
All the simulated configurations were successfully control. The opening speeds (from 0% to 100%) of the
tuned. Different valve sizes and different cell types had valves have been estimated to be 30 s regardless of the
an effect on the proportional gain of the PI control: the valve size, i.e. the bigger the valve, the shorter the rise
proportional gains were larger when using valves sized time.
according to the ISA standard than with oversized The valves were sized according to the ISA standard,
valves. Similarly, the proportional gains were larger in and the retention time in each cell was 1.5 min. The
 4.25 4.25

4.20 4.20

4.15 4.15

4.10 4.10

4.05 4.05

4.00 4.00

0 100 200 300 400 500 0 100 200 300 400 500

Fig. 11. Response of the levels to a +3 cm change in the set points. 100% valve oversizing is used in both figures. On the left are non-linear cells, and
on the right is the model based on the ideal tank assumption.

5.00 5.00

4.95 4.95

4.90 4.90

4.85 4.85

4.80 4.80

4.75 4.75

4.70 4.70
0 100 200 300 400 500 0 100 200 300 400 500

Fig. 12. Response of the levels to a +20% change in pulp feed. 100% valve oversizing is used in both figures. On the left is the PI controller, and on
the right, PI controller and feedforward controller.

Table 4
The influence of feed-forward control on the configuration with double cells: response of the pulp level in the last cell to a +20% change in the feed

Configuration Overshoot (cm) Overshoot (%) Rise time (s) Settling time (s) Dead time (s) ISE IAE

TC130 11222 525 0.92 0.184 3 47 8 0.00175 0.2884

TC130 11222 525 FF 0.4 0.08 4 30 0.5 0.00021 0.1024

oversized valves were considerably larger than those urations with ideal cells and with different geometry
sized using the ISA standard. This leads to the situation cells were compared. In every case the results were
where the valve is almost closed during normal clearly better.
conditions, and the valve is operating in the non-linear Feed-forward control implemented with PI control
area. The effect of valve size on the quality of control improved the compensation of the disturbances in every
was minor. Overshooting due to flow disturbances was case. An adaptive feed-forward control gain could result
compensated slightly better when the valves were sized in even bigger improvements in control performance.
according to the ISA standard. In contrast, changing the
set points of the pulp level with oversized valves often
even resulted in better performance indices. 8. Further studies
The geometry of the cell had the greatest effect on the
performance of the level control. A difference in The simulation platform introduced in this paper has
performance was clearly evident when the same config- been used in further studies to develop and compare
-300 References

-310 Andersen, R., Gronli, B., Olsen, T., Kaggerud, I., Ramslo, K., &
(a) Sandvik, K. (1979). An optimal control system of the Rougher
Flotation at the Folldal Verk Concentrator, Norway. Proceedings
-320 of the 13th international mineral processing congress, Vol. II (pp.
1517–1540). New York, USA.
-330 (b)
Hammoude, A., & Smith, H. (1981). Experiments with self-tuning
control of flotation. Proceedings of the 3rd IFAC symposium on
automation in mining, mineral and metal processing (pp. 213–218).
-340 Oxford, UK.
J.ams.a, S.-L., & Herbst, J. A. (1988). A simulation study of expert
(c) control system for flotation. Software for Computer Control 1988
-350 (pp. 67–75). New York, USA.
J.ams.a-Jounela, S.-L., Laurila, H., Karesvuori, J., & Timperi, J. (2001).
-360
Evaluation of the future automation trends in control and fault
0 50 100 150 200 diagnostics—a case study in flotation plant. Proceedings of the 10th
Fig. 13. Responses of the level to a change in pulp feed (0.35-1.1 m3/
View publication stats IFAC symposium on automation in mining, mineral and metal
processing. Tokyo, Japan.
min). The biggest overshoot (a) is obtained using a PI controller, and K.ampj.arvi, P., & J.ams.a-Jounela, S.-L. (2002). Level control strategies
noticeable improvements in control performance are achieved by for flotation cells. Proceedings of the 15th IFAC world congress on
implementing additional feed-forward controllers (b and c). automatic control. Barcelona, Spain.
Koivo, H., & Cojocariu (1977). An optimal control for a flotation
advanced multivariable control strategies for flotation circuit. Automatica 13, 37–45.
cells. Control performance of feed-forward control, Niemi, A., Maijanen, J., & Nihtil.a, M. (1974). Singular optimal
feed forward control of flotation. IFAC/IFORS symposium on
decoupling and different multivariable control methods
optimization methods—applied aspects (pp. 277–283). Varna,
have been studied, discussed and published by K.amp- Bulgaria.
j.arvi and J.ams.a-Jounela (2002). Schubert, J. H., Henning, R. G. D., Hulbert, D., & Craig, I. K. (1995).
Flotation control—a multivariable stabilizer. XIXth IMPC, Vol. 3
(pp. 237–241). San Francisco, USA.
Stenlund, B., & Medvedev, A. (2000). Level control of cascade
Acknowledgements
coupled flotation tanks. Future trends in automation in mineral and
metal processing, IFAC workshop 2000 (pp. 194–199). Helsinki,
The authors would like to thank Outokumpu Mintec. Finland.
In particular, special thanks to Heikki Laurila who Zargiza, R., & Herbst, J. A. (1987). A model based feed forward
helped with the practical details. Financial support of control scheme for flotation plants. 116th AIME annual meeting
(pp. 23–27). Denver, CO, USA.
TEKES is also gratefully acknowledged.