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

net/publication/248475855

Level control strategies for flotation cells

Article  in  Minerals Engineering · November 2003

DOI: 10.1016/j.mineng.2003.06.004

CITATIONS READS

27 1,603

2 authors:

Petteri Kämpjärvi Sirkka-Liisa Jamsa-Jounela

Neste Oil Aalto University
2 PUBLICATIONS   62 CITATIONS    186 PUBLICATIONS   1,842 CITATIONS   

SEE PROFILE 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 version, published in Minerals Engineering 16 (2003) 1061–1068

LEVEL CONTROL STRATEGIES FOR FLOTATION CELLS

P. Kämpjärvi, S-L. Jämsä-Jounela

Helsinki University of Technology

Laboratory of Process Control and Automation
Kemistintie 1, 02150 Espoo, FINLAND
E-mail: [email protected]

Abstract: Flotation is a difficult process to run efficiently. One way to make flotation
performance better is to improve cell level control. However, controlling pulp levels in flotation
cells is a complex control task because of strong interactions between the levels in flotation cells.
Therefore advanced controllers are needed to give good level control. This paper deals with a
model of six flotation cells in series. Simulations are performed to compare different control
strategies. Four control strategies are considered: one SISO controller and three different MIMO
controllers including a new multivariable controller. It is shown that level control performances
of the MIMO controllers are significantly better than that of the classical SISO controller.

Keywords: Flotation machines; Froth flotation; Process control; Mineral processing

1. Introduction feedback control and a Kalman filter for rougher

flotation control. Hammoude and Smith (1981) used a
Level control of flotation cells is a very complex task linear model to develop a minimum–variance
due to high interactions between the process variables. controller for recleaning. New advanced control
A control action implemented at any point in the method has been also recommended for the control of
flotation circuit tends to be transmitted to both flotation processes by (Stenlund and Medvedev,
upstream and downstream units, and sometimes with 2000).
amplification. Large variations in the flow rate to the
first cell and varying composition of the raw ore also The aim of this research is to study and compare
cause problems. different control strategies from the point of view of
cell level control. In addition, a new control strategy is
Flotation cells are conventionally controlled by presented and implemented. Its performance is
isolated PI-controllers. PI control works well when the compared to three different strategies: one traditional
cell being controlled is isolated. However, in a SISO control strategy, and two MIMO control
flotation circuit where interactions are strong, PI strategies. The strategies are compared by the means
control does not meet the requirements of high control of special performance indices.
performance. Hence a considerable amount of
research has been carried out over the last few years to In the following the mathematical model of a
develop better control techniques for flotation circuits flotation cell is first developed and a five cells in
(Jämsä-Jounela et al., 2001). series cell configuration is constructed. The
simulations are then performed in order to
Niemi et al. (1974), Koivo and Cojocariu (1977) used determine suitable control parameters for
a single cell model when developing an optimal controllers of the cell levels. Simulation was
control algorithm via applications of the maximum performed with Matlab 5.2.0 and its Simulink
principle. Andersen et al. (1981) and Zargoza and library.
Herbst (1987) reported an application of state
Fig. 1. Schematic diagram of flotation cells

outflow from the first cell, and Cv = valve

2. Mathematical modelling of flotation cells in coefficient.
series
The equations for cells 2, 3 and 4 are respectively (i
In the flotation process the pulp is fed into the first = 2, 3 and 4)
cell and the froth is collected in the launders. The
feed can be measured by a flow measurement, and Vi
 ( Fi 1  Fi )  KCv (ui 1 ) yi 1  yi  hi 1
the remaining pulp flows into the next cell. The t
magnitude of the flow depends on the pressure  KCv (ui ) yi  yi 1  hi
difference between two adjacent cells, the position
(2)
of control valves, and the viscosity and density of
the pulp. The magnitude of the pressure difference
The equation for the last cell (n = 5)
can be determined from the physical height
difference of the cells. The pulp level in a cell is
measured and controlled by adjusting the control Vn
 ( Fn 1  Fn )  KC v (u n 1 ) y n 1  y n  hn 1
valve. t
 KC v (u n ) y n  hn
In the following a flotation cell is considered as a (3)
tank of perfectly mixed pulp. Since the pulp is
perfectly mixed, the density is the same throughout In the case of an ideal tank, the cross-section of a
the cell, i.e. there are no spatial density gradients in cell is assumed to be constant. The pulp levels in
the cell. As the froth flow is small compared with the cells can therefore be written as:
the pulp flow, it is ignored in the outgoing flows.
The impact of the air feed on the pulp level is also y1 (q  F1 )
ignored.  
t A1 (4)
The cells under study do not have the properties of q  KC v (u1 ) y1  y 2  h1
ideal tanks because the cross-sectional area of the
A1
cell is not constant. A mathematical model for the
physical properties will be developed and discussed
next. y i K i 1C v (u i 1 ) y i 1  y i  hi 1

t Ai
2.1. Single cells in series
KC v (u i ) y i  y i 1  hi

In a flotation process several single cells are connected Ai
in series as shown in Fig. 1. (5)

For the first cell in the series: where i = 2, 3, and 4.

V1 y n KC v (u n 1 ) y n 1  y n  hn 1
 ( q  F1 )  q  KCv (u1 ) y1  y 2  h1 (1) 
t t An (6)
where q = feed rate to the first cell, y1 = pulp level KC v (u n ) y n  hn
in the first cell, y2 = pulp level in the second cell, hn

An
= physical difference in height between the cells, u1
= control signal, K = constant coefficient, F1 =
where n = 5.
 V2
2.2. Double cells in series  F 1 F  2 g ( y1  y 2 )ll hl
t (10)
Mathematical models for the double cells can be  K 2 Cv (u1 ) y 2  y 3  h1
derived in a similar manner (Jämsä-Jounela et al.,
2003). The principle differences compared with the
2.3. Modelling the Outokumpu flotation cells
mathematic models of single cells in series are that, in
a double cell, both pulp levels are controlled by
Fig. 3 shows the cross-section of the flotation cells
manipulating a control valve in the second cell
under study. The pulp level changes as a function of
outflow, as can be seen from Fig. 2. The cells are
the cell volume. The change is linear from the
physically on the same level.
bottom of the cell to the starting level of the launders
and boosters, denoted by Hlowerpart in Fig. 3. The
The pulp level in the second cell is measured and
cross-section of the pulp subsequently decreases on
compared with the set point. This signal is used as
moving upwards because the launders and
input to a PI-controller. The output signal of the
‘boosters’ reduce the volume of the cell.
controller is the desired valve position as denoted by
u1. The cells in a double cell are connected via a
flange. The flange limits the maximum flow between
Rr,out LC LB
the cells. The pressure difference in the cells is the
only driving force for the flow, and depends on the
density of the pulp. LR rC r2 rB
HC
HB

y1 y2 h a
y3
q
F1 F2
F3
u1
Hlowerpart
Fig. 2. Double flotation cell series

The velocity of outflow from the first cell can be

derived using Bernoull’s equation, resulting in
Rtotal
Fig. 3. Cross-section of a flotation cell of Outokumpu
v  2 g ( y1  y2 ) (7) Mintec

where g = gravity, y1 = pulp level in the first cell, and Calculation of the pulp volume can be divided into
y2 = pulp level in the second cell. two sections: the pulp level below and above the
level of Hlowerpart. When the volume is smaller than
The volume flow across the flange can be calculated the volume of the cylinder´s Hlowerpart, the pulp level
as is determined as
V . (11)
htot 
F1  vAflange  2 g ( y1  y2 )ll hl (8) Rtot
2

where Aflange = cross-sectional area of the flange, ll = where Rtot is the radius of the cell.
length of the flange, and hl = height of the flange.
If the pulp volume is equal or greater than Hlowerpart,
The pressure drop due to flowing resistances is the volume is written as
assumed to be negligible and can be ignored. The
change of the pulp volume in the first cell with respect V  Vlower_ part  Vupper_ part(h)  VB (h)  VR (h) (12)
to time can be written as
V1 where Vlowerpart = volume of the cell below the level
 q  F  q  2 g ( y1  y 2 )ll hl (9) of Hlowerpart, Vupperpart = volume of the cylinder above
t Hlowerpart, VB = volume of the boosters, and VR =
volume of the launders.
The outflow of the second double cell is similar to that
of a single cell, and the flow into the second cell is the
A level variable, the zero point of which is equal to
outflow of the first. Therefore the mathematical model
Hlowerpart is denoted by h. The volumes of the launder
is
and booster are obtained by geometric relations, and
substituting them in Eq. 12 results in
 The valves in the models are sized for a flow which
 has a retention time of 1.5 min in one cell. The flow
h 3 ( K B2 )  h 2 (K B Rtotal ) (13)
3 rate to the first cell is also calculated using this
 h( ( K R  Rtotal
2
))  (V  Vlowerpart)  0 retention time value.

Characteristic curves of the control valves are

where KB = constant coefficient of the booster produced by Larox Flowsys and were used in the
dynamics and KR = constant coefficient of the Simulink models in order to cause realistic and
launder dynamics. nonlinear behaviour for the valves.

3. Control strategies

The different control strategies are discussed and

described in the following sections. These strategies
are selected because they can be used with basic PI-
controllers and without any additional
instrumentation. Traditionally in flotation cell series
there is only one flow measurement in the beginning
of the series and level PI-controllers in every cell.

3.1. Feed forward controller

Fig. 4. Pulp level as a function of pulp volume A flow feed-forward controller monitors disturbances
in the inflow to the first cell and uses proportional
The third order equation has a solution in [Hlowerpart action to close or open the valves of the cell in order
Htotal], where Htotal is the total height of the cell. KB, to compensate for disturbances. Compensation is
KR and Vlowerpart are constants and specific for each linearly dependent on the difference between the
cell size. The total level of pulp can be determined current inflow and the normal inflow. The
by summarizing h and Hlowerpart. measurement signal is filtered in order to prevent the
feed-forward control from reacting to random
The effect of pulp level non-linearity can be seen in variation in the flow. However, this kind of controller
Fig. 4, in which the pulp level is presented in ideal does not provide any extra performance improvement
tank conditions as a function of the pulp volume. in the event of disturbances occurring somewhere else
The cell type is TC-50, the maximum volume of in the cell series. The model of the feed-forward
which is 50 m3. controller is shown in Fig.5.

Fin

2.4. Valve sizing and characteristic curve of the Measure- Fin.m Feed-
ment forward
valves
yset + e + Flotation y
+ u F cell
Valve sizing is based on the Cv value, which is PI Valve
dynamics
controller
calculated according to the ISA standard as follows: - as

 (14)
ym
Measurement
C v  1.17Q 
p
Where Fig. 5. Control diagram of feed-forward controller.

Q is the flow rate, m3/h, Cv the valve capacity 3.2. Decoupling controller
coefficient,  the pulp density, kg/m3, and p the
pressure difference over the valve. A decoupling controller is based on differential
equations (1)–(3). The purpose of the decoupling
The flow rate (m3/h) through the cell is calculated as controller is to eliminate the crosswise effects of
control loops, and hence the stability of a single
Vcell control circuit depends only on its own stability
Q  1.2 (15) features. The basic model of the decoupling controller
 / 60 is shown in Fig. 6.
where Vcell is the cell volume (m3) and  is the pulp
retention time in the cell.
 + y1 ( s ) influenced not only by the difference between a set
m1 ( s )
H 11 ( s ) point and the measured level in the tank, but also the
+ differences between set points and the measured levels
H 21 ( s ) in all the tanks in upstream. These variables are
summed and fed to the PI-controller of the cell.
H 12 ( s ) Furthermore, the variables are scaled by a suitable
m2 ( s ) + y2 ( s) factor depending on the valve size, position and
H 22 ( s ) process.
+

In this strategy each control valve can be regarded as

Fig. 6. Control diagram of a basic decoupling controller. a sluice gate of a dam. When a damned inventory is
too high in upstream, the valves are opened more than
The mathematical criterion to be fulfilled for usual, even when there seems to be no need to take
decoupling a tank i will be such an action on the basis of the levels in the
neighbouring vessels. The control diagram is shown in
Fiin  Fiout  0 (4) Fig. 7.

e2
where Fiin is a change of inflow to tank i. Using the e1 P2 e3
valve functions from Eqs. (1)–(3) the equation can P1 + + P3

be written as follows +

yset + Flotation y
+ e u F cell
Fi  Ki Cvi ui ( hi )
PI Valve
(5) -
controller dynamics
as

where hi is the level difference over the valve. ym

Measurement

Substituting in Eq. (4), it becomes

Fig. 7. Control diagram of a multivariable controller similar
K i 1Ci 1 ui 1  ui 1  hi 1  hi 1  (6) to Floatstar™ (tank 4).

 K i Ci ui  ui  hi  hi   0

3.4. New multivariable controller
Solving this equation for the change in the control
signal gives When some disturbance occurs in tank, it has an
effect to level in the previous tank in upstream.
Previously described multivariable control does not
ui  K i 1Ci 1 K i Ci ui' 1 h  h   u
'
i 1 i
'
i
take disturbances of this kind into account. Also
strategy does not take into account disturbances
 f ui' 1 , hi'1 , hi'  (7) arising from pulp feed. Therefore a flow feed
forward has been added to the system and the
Eventually, the control signal for a tank i becomes difference between set point and measured level in
next tank is also added to the controller of previous
ui  uPI  f ui 1 , hi 1 , hi 
tank. The control diagram is shown in Fig 8.
(8)
where u PI is the control signal from a PI-controller.
e2

In order to handle the variations from the inflow, the e1 P2

P3
e3

e5

feed forward is attached to the first tank. P1

+ P5
+ + Filter
Fin
+ Feedforward
Fin,m Measurement

yset e u +
3.3. Multivariable controller similar to Floatstar™
+
F y
PI controller Valve Flotation cell dynamics

- +

ym Measurement

A multivariable controller (Schubert et al., 1995)

controls the total inventory of material in the upstream Fig. 8. The control diagram of MV controller (tank 4).
tanks. In this control strategy, controlling a valve is
Fig. 9. Simulink model of the flow dynamics in a flotation cell

4. Simulations
t2

In the simulations a configuration of six TC-50 cells

in series was studied in accordance with the ideal tank
IAE   y t   y t  dt
t t1
sp
(10)

assumption. Therefore the effects of boosters and

launders were not considered. The valves were 100%
oversized according to the ISA standard, and the 5. Simulation results
retention time in each cell was 1.5 min. Control
strategies included conventional PI-controllers with The simulations of the configurations of six TC-50
feed-forward control, decoupling controller, a cells in series resulted in parameters for the PI-
multivariable controller similar to Floatstar™ and a controllers. Integration times in the traditional system
feed-forward multivariable controller. The simulation with a feed-forward controller were between 15 and
results of a +3 cm change in the set points of the cell 50 s and proportional gains between 0.8 and 1.2.
levels at times 100, 150, 200, 250, 300 and 350 s are Because MIMO control strategies respond better to
presented in the following. Making ±20% change in disturbances, the PI parameters were set faster.
the feed to the first flotation cell was also simulated Integration times in all the PI-controllers were set to
with different strategies. The set point of the cell level 15 s and gain to 1. In the decoupling controller the PI-
is lowest in the first cell, and the set point values parameters were between 15-50 s and 1-1.4 s,
increase on moving towards the last cell in the series, correspondingly.
where the operating range of the level controller is
smaller. The simulation schemes were constructed The responses of the feed-forward controller to
with Matlab 6.0.0 Simulink software. The Simulink disturbances in pulp feed and to set point changes are
model of the flow dynamics in the flotation cell is presented in Fig. 10. As can be seen, the –20% change
presented in Fig.9. in pulp feed is affecting all the cell levels in the series.
The set point changes in a cell also have undesirable
The controllers were tuned and compared using the effects on the adjacent cells. There is always a
following indices. The IAE index (integral of the considerably large perturbation in the level of the next
absolute value of the error) integrates the absolute cell every time a set point change is made in the
value of errors, and even-handedly weights all the system.
deviations. ISE (integral of the square error) gives
more weight to big deviations from the set point. The responses of the decoupling controller are
illustrated in Fig 11. The decoupling controller is a
MIMO controller, and it also takes into account the
t2
interactions between cells. As can been seen from the
 y t   y t  dt
2
ISE  sp
(9) graphs, the decoupling controller effectively
t t1 eliminates disturbances arising from changes in the
pulp feed. Furthermore, set point changes in the cells
do not affect to the other cells. The IAE and ISE indices, which depict the
performance of controllers, are shown in Tables 1–
The responses of configurations in which a controller 4. As was to be expected, the traditional SISO
similar to Floatstar™ and the new multivariable control with flow feed-forward had the poorest
controller are used are shown in Figs. 12 and 13. The figures in all cases.
new multivariable controller seems to be more robust
than the other controller, especially during changes in
pulp feed.

4.25 4.25

4.2 4.2

4.15 4.15

4.1 4.1

4.05 4.05

4 4

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

Fig. 10. Feed-forward controller. On the left the response to a -20% change in pulp feed, and on the right the response to set
point changes.

4.25 4.25

4.2 4.2

4.15 4.15

4.1 4.1

4.05 4.05

4 4

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

Fig. 11. Decoupling controller. On the left the response to a -20% change in pulp feed, and on the right the response to set
point changes.

4.25 4.25

4.2 4.2

4.15 4.15

4.1 4.1

4.05 4.05

4 4

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

Fig. 12. Multivariable controller similar to Floatstar™. On the left the response to a -20% change in pulp feed, and on the right
the response to set point changes.
Fig. 13. New multivariable controller. On the left the response to a -20% change in pulp feed, and on the right the response
to set point changes.

Table 1. The performance indices for feed-forward controller

Feed-forward ISE(+20%) IAE(+20%) ISE(-20%) IAE(-20%) ISE(s.p.c.) IAE(s.p.c.)
controller
1 3.10 348 3.7 325 2.7 298
2 2.9 361 3.1 360 4.0 520
3 2.7 371 3.3 391 5.0 712
4 2.2 341 2.7 371 5.33 818
5 2.0 322 2.4 361 6.5 970
6 0.5 159 0.4 147 3.7 604

Table 2. The performance indices for the decoupling controller

decoupling ISE(+20%) IAE(+20%) ISE(-20%) IAE(-20%) ISE(s.p.c.) IAE(s.p.c.)
controller
1 1.10 327 0.7 263 2 220
2 0.0008 10.2 0.0003 8.0 2.2 242
3 0.0000008 0.3 0.00000004 0.1 2.2 241
4 0.00000005 0.05 0.000000001 0.01 2.2 241
5 0.0 1.2 0.000004 0.4 2.1 228
6 0.04 29.5 0.019 19.5 2.1 179

Table 3. The performance indices for the multivariable controller similar to FloatstarTM
MV controller ISE(+20%) IAE(+20%) ISE(-20%) IAE(-20%) ISE(s.p.c.) IAE(s.p.c.)
1 7.81 499 9.02 527 2.2 202
2 0.1 65.0 0.07 58.6 2.3 243
3 0.03 25.2 0.03 26.5 2.3 248
4 0.0001 2.2 0.00006 1.4 2.3 249
5 0.0001 2.7 0.00008 1.5 2.3 245
6 0.01 18.5 0.002 7.8 1.7 151

Table 4. The performance indices for the new multivariable controller

new MV controller ISE(+20%) IAE(+20%) ISE(-20%) IAE(-20%) ISE(s.p.c.) IAE(s.p.c.)
1 1.79 216 2.1 231 2.2 204
2 0.03 30.7 0.02 27.4 2.2 202
3 0.007 12.3 0.007 13.2 2.2 211
4 0.0001 1.1 0.00009 1.7 2.2 211
5 0.0009 4.3 0.0005 3.5 2.2 212
6 0.005 8.1 0.003 6.9 1.9 149
 Zargiza, R., Herbst, J.A., 1987. A model based feed
6. Conclusions forward control scheme for flotation plants. In:
116th AIME annual meeting. Denver, CO, USA,
All the simulated configurations were successfully pp. 23-27.
tuned. It is noticeable that the classical SISO strategy
with feed-forward controller cannot even approach the
performances of the MIMO controllers. This is due to
high interactions between the control loops, which
SISO systems cannot take into account.

The differences between different MIMO systems are

somewhat smaller. All the controllers performed
robustly to disturbances in pulp feed and to set point
changes. The decoupling controller had the best IAE
and IDE indices. However, the decoupling controller
is sensitive to model uncertainties (Skogestad and
Postelwaite, 1996). This also means that process
changes can strongly degrade the control performance.

References

Andersen, R., GRonli, B., Olsen, T., Kaggernd, I.,

RAmslo, K., Sanvik, K., 1979. An optimal control
system of the rougher flotation at the Folldal Verk
concentrator, Norway. In: Proceedings of the 13th
International Mineral Processing Congress. New
York, USA, pp. 1517-1540.
Hammoude, A., Smith, H., 1981. Experiments with
self-tuning control of flotation. In: Proceedings of
the 3rd IFAC Symposium on Automation in
Mining, Mineral and Metal Processing. Oxford,
UK, pp. 213-218.
Jämsä-Jounela, S-L., Laurila, H., Karesvuori, J.,
Timperi, J., 2001. Evaluation of the future
automation trends in control and fault diagnostics
– a case study in flotation plant. In: 10th IFAC
Symposium on Automation in Mining, Mineral
and Metal Processing.
Jämsä-Jounela, S.-L., Dietrich, M., Halmevaara, K.,
Tiili, O., 2003. Control of pupl level in flotation
cells. Control Engineering Practise, 73-81.
Koivo, H., Cojocariu, R., 1997. An optimal control for
a flotation circuit. Automatica 13 (1), 37-45.
Niemi, A., Maijanen, J., Nihtilä, M., 1974. Singular
optimal feed forward control of flotation. In:
IFAC/IFORS Symposium on Optimization
Methods – Applied Aspects. Varna, Bulgaria, pp-
277-283.
Schubert, J.H., Henning, R.G.D., Hulbert, D., Craig,
I.K., 1995. Flotation control – a multivariable
stabilizer. In: XIXth IMPC, San Fransisco, vol. 3,
pp.237-241.
Skogestad, S., Postelwaite, I., 1996. Multivariable
Feedback control: Analysis and design. John
Wiley & Sons.
Stenlund, B., Medvedev, A., 2000. Level control of
cascade coupled flotation tanks. Future trends in
Automation in Mineral and Metal Processing. In:
Jämsä-Jounela, S.-L., Vapaavuori, E. (Eds.),
IFAC Workshop 2000, Helsinki, Finland, pp.
194-199.

View publication stats