Column Flotation Simulation and Control An Overview

Minerals Engineering 22 (2009) 519529
Contents lists available at ScienceDirect
Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng
Column otation simulation and control: An overview

Jocelyn Bouchard a,1,2,3, Andr Desbiens b,1, Ren del Villar c,*,1, Eduardo Nunez a,3
a
Xstrata Process Support, 6, Edison Road, Falconbridge, Ontario, Canada P0M 1S0
LOOP (Laboratoire dobservation et doptimisation des, procds Process Observation and Optimisation Laboratory), Universit Laval, Pavillon Adrien-Pouliot,
Dpartement de gnie lectrique et de gnie informatique, Qubec, Qubec, Canada G1V 0A6
c
LOOP (Laboratoire dobservation et doptimisation des procds Process Observation and Optimisation Laboratory), Universit Laval, Pavillon Adrien-Pouliot,
Dpartement de gnie des mines, de la mtallurgie et des matriaux, Qubec, Qubec, Canada G1V 0A6
b
a r t i c l e
i n f o
Article history:
Received 9 August 2008
Accepted 3 February 2009
Keywords:
Column otation
Modelling
Simulation
Process control
a b s t r a c t
Even after having been used for several years in mineral processing plants, the full potential of the column
otation process is still not fully exploited. There is obviously more than one reason explaining this, but one
important contributing factor is certainly the under usage of available control capabilities. Researchers and
practitioners have been interested in column otation simulation and control for more than two decades.
This paper provides an overview of the literature focused on these specic elds of research. It also discusses some future investigative issues and how the current industry may benet from past developments.
2009 Elsevier Ltd. All rights reserved.
Contents
1.
2.
3.
4.
5.
6.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Process description. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.
Prediction of recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.
Dynamic behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.
Soft sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Process control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.
Intermediate level control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.
Control strategies based on metallurgical objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Current practice, research trends & future applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.
Current practice and challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.
Steady-state simulation: metallurgical performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.
Sensor development and applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.
Dynamic modelling and simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.
Process control myths and reality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
....
....
....
....
....
....
....
....
....
....
....
....
....
....
....
....
....
....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
* Corresponding author. Tel.: +1 418 656 7487; fax: +1 418 656 5343.
E-mail addresses: jbouchard@xstrataprocesssupport.ca (J. Bouchard), desbiens@gel.ulaval.ca (A. Desbiens), rene.delvillar@gmn.ulaval.ca (R. del Villar).
1
Partially supported by NSERC (Canada).
2
Partially supported by FQRNT (Qubec).
3
Tel.: +705 693 2761x3427; fax: +705 699 3431.
0892-6875/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mineng.2009.02.004
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
520
520
521
521
522
522
523
524
524
524
524
524
525
525
526
526
527
527
520
J. Bouchard et al. / Minerals Engineering 22 (2009) 519529
1. Introduction
The introduction of otation columns in mineral processing
plants caught the attention of many researchers in the last two
decades of the twentieth century. Column otation simulation
and control progressively became prevailing elds of investigation.
Almost twenty-ve years after the installation of the rst commercial otation column in the Western world (Finch and Dobby, 1990),
it is worth examining where these studies have led, what are the
current research interests, and how the current mining industry
and practitioners may benet from past developments. This paper
aims at complementing the last published review (Bergh and
Yianatos, 2003). It provides an overview of the literature dealing
with column otation simulation and control, and discusses research trends and industrial application issues.
Emphasizing the operating variables, Section 2 gives a
description of the process. A summary of relevant with respect
to simulation and control publications dealing with modelling
considerations are presented in Section 3. The discussion is organized according to three research areas: the prediction of the
recovery, the analysis of the process dynamic behaviour and the
development of soft sensors. Section 4 focuses on process control
aspects and reviews most of the published applications. Finally,
Section 5 discusses industrial practice, research interests and future developments required to meet the present industrial needs.
2. Process description
A schematic representation of a otation column is given in
Fig. 1. A typical unit has three input streams (conditioned mineral
pulp feed, air, and wash water), and two output streams (the concentrate and tailings). Under normal operating conditions, the column volume is split into two distinct regions according to their air
content (volume fraction): a collection or pulp zone (less than 20%
of air) and a cleaning or froth zone (more than 70% of air).
To perform a otation separation, several reagents are generally required: collectors, frothers, and regulators (activators,
depressants and pH modiers). These chemicals are added in a

preliminary stage, i.e. the pulp conditioning, and/or directly to
the otation cell, and have obviously an important impact on
the metallurgical performance. However, as the type and concentration of most of the reagents may be determined off-line in laboratory and strongly depend on the duty, they are not considered
in this paper.
Specic variables characterize a otation column operation,
namely: the froth depth also called froth layer height and corresponding to the complement of the pulp level (or pulpfroth interface position) , bias, gas hold-up, and bubble surface area ux.
They are explained hereafter.
Froth depth. The froth depth (H) denes the relative height
of the cleaning and collection zones. Consequently, it determines
the mean solids residence time in the column, partially affecting
the recovery in both the pulp and the froth.
Bias. Dened as the the net downward ow of water through
the froth (Finch and Dobby, 1990), the bias may be qualitatively
interpreted as the fraction of the wash water ow used for froth
cleaning. It is worth noting that the on-line dynamic measurement
of the bias (J b , expressed as a supercial velocity) still presents a
difculty (see Section 3.3).
Air hold-up. It represents the gas volume fraction within the collection zone. The air is often considered as a otationreagent and
practitioners have been interested in monitoring the gas dispersion
within the column using the air hold-up (g ).
Bubble surface area ux. The collection of mineral particles by
bubbles greatly depends on the amount of bubble surface available. Therefore, a more adequate way of tracking the inuence
of gas to the otation process is to use the amount of bubble surface per unit time and unit of column cross sectional area instead
of the gas hold-up. This becomes obvious when comparing the
otation performance of similar volumes of air in the form of a
swarms of small bubbles (large specic area) or a swarm of fewer
larger bubbles (smaller specic area). The resulting bubble surface area ux, or Sb , may be evaluated from the bubble ow rate
(nb ), the bubble surface, and the cross-sectional area of the column (Ac ). Assuming a suitable mean bubble diameter (db ), it
can be estimated with
2
Sb
Fig. 1. Flotation column.
nb db p 6J g
Ac
db
where J g is the gas supercial velocity. Fig. 2 illustrates the bubble

surface area ux concept. However, industrial bubblers usually produce a broad bubble size distribution, making the use of a single
mean value rather inacurate, since this same value could be obtained from quite different size distributions.
Finch et al. (2000) presented a discussion on gas dispersion
characterization. Using data from literature, they suggested a quasi-linear relationship restricted to the range of calibration between g and Sb . Heiskanen (2000) criticized the work of Gorain
et al. (1995a,b, 1996, 1997, 1999) on the gas dispersion in otation
machines, with respect to measuring methods for gas velocity and
bubble size. According to him, further studies on the linear relationship between the otation rate constant (k) and the bubble
surface area ux proposed by Gorain et al. (1997) are also required.
Deglon et al. (1999) were also critical about the linear kSb relationship. Based on the results of a simulation study, they claimed
that the near linear region corresponds to a transition from the predominance of the sub-process of particlebubble attachment to that of
particlebubble detachment in mechanical otation cells (Deglon
et al., 1999).
General instrumentation issues for on-line monitoring of column otation operation are discussed by Bergh and Yianatos
(2003) and Bouchard et al. (2005b).
Fig. 2. Sb concept.
3. Modelling
For process control purposes, research work dealing with column otation modelling can be organized in three categories
according to their aim: prediction of the recovery, analysis of dynamic behaviour and development of soft sensors.
3.1. Prediction of recovery
Besides works for scale-up purposes (e.g. Finch and Dobby
(1990), Rubinstein (1995), Alford (1992) and OConnor et al.
(1995)), other studies aimed at predicting the recovery of column
otation operations were presented but considering control and
optimisation applications.
Pioneering efforts started in the eighties. Luttrell et al. (1987),
proposed a static simulator based on a population mass balance
(air bubbles, unattached solid particles, and bubbleparticle aggregates). Mass transport was considered in the model using uid
ows and particle buoyancy, while the bubbleparticle rate attachment was evaluated using rst principles. Some processes, such as
the bubble loading and mixing properties, were explained under
pre-specied operating conditions, using a semi-fundamental approach requiring the calibration of two empirical coefcients. The
simulator attempted to predict the recovery of a specic column
otation operation for design, control, optimisation and scale-up
purposes.
At the same time, Sastry and Lofftus (1988) also developed a
simulator using a similar approach, but considering the dynamic
mass balance equations. The resulting tool opened the door to
time-dependent investigations, which are very useful to study process control strategies. The assumption of constant air and water
hold-up, along with the impossibility to analytically solve the general model, represented the greatest limitations of their work.
The addition of air and water mass balance equations to a ve
well-mixed zone approximation instead of the three zones model
of Sastry and Lofftus (1988) made the simulator proposed by Pate
and Herbst (1989) a more exible tool. Their approach also replaced the axially dispersed plug ow model with a distributed
volume mixers-in-series approximation to increase computational
efciency. The air mass balance was however considered on a static basis according to the assumption that air hold-up is subject to
very fast changes compared to water volume. Particles could be of
any size and were divided in three classes: free valuable mineral,
free gangue and locked. Similarly to above-mentioned simulators,
the proposed model exhibited certain empirical features regarding
the calibration of rate constants and the description of some phenomena using correlations (water entrainment, water drainage,
521
etc.). Later on, the same research team used this dynamic simulator
to design a methodology for selecting a control strategy for a column otation unit (Lee et al., 1991).
Cruz (1997) made a further step and proposed a fully dynamic
simulator of column otation metallurgical performance. Her
work, notably based on fundamental considerations, included a
comprehensive description of complex phenomena, such as bubble
coalescence in the froth and bubble loading, and considered particle and bubble size distributions as well as a particle composition
distribution. The design was based on the application of a population balance to a vertically distributed volume mixers-in-series
framework: a perfectly mixed aeration zone, a perfect-mixers-inseries lower collection zone, a single perfectly mixed feed zone, a
perfect-mixers-in-series upper collection zone, the interface, and
three plug ow volumes for the froth (stabilized froth, wash water
addition zone, and draining froth). Notwithstanding improvements
in column otation simulation, being impossible to carry out dynamic ow rate variations signicantly narrowed the extent of potential control applications. In fact, operating conditions were xed
off-line and stayed constant throughout the simulation. The process was then driven from an initial to a nal state as dened by
the simulated operating conditions. Gas hold-up and solids ow
rate changes were computed, but the dynamic variations of froth
depth were not considered. Despite this limitation, Cruz nevertheless achieved a major breakthrough, even though it seems that this
work remained unnoticed by the mineral processing community.
Recently, Bouchard et al. (2006) made a new attempt to develop
a dynamic fundamentally-based column otation simulation
framework, but only preliminary results were presented.
Using a more global approach, Kho and Sohn (1989) obtained a
predictive model for talc recovery based on empirically estimated
rate constants and residence time distributions for the liquid and
solids.
Luttrell and Yoon (1991), noticeably inspired by research work
supervised by Finch (Finch and Dobby, 1990), developed a static
simulator based on hydrodynamic principles, aiming at predicting
the recovery of a column otation operation. In 1993, they proposed a scale-up procedure (Luttrell et al., 1993). Besides vessel
geometrical characteristics considerations, their discussion
emphasized on air spargers and wash water distribution systems.
The column diameter was determined from the maximum froth
carrying capacity, while the rate constant and mean particle retention time were used to obtain its height. The effect of axial mixing,
air hold-up, and gas and liquid ow rates were also considered in
the procedure.
Following similar ideas, Alford (1992), from the JKMRC in Australia, gathered the results of many researchers to develop a column otation static simulator. The model was global (exhibiting
only a single zone) and was considered a useful tool to study ow
sheet congurations and scale-up of industrial units, assuming an
appropriate calibration.
Tuteja et al. (1994) published a review of the most relevant
models for the prediction of recovery under a clear classication:
kinetic models and non-kinetic models, i.e. completely empirical
regression models.
teyaka and Soto (1995) were interested in the modelling of the
recovery of negative bias column otation operation. Neglecting
the effect of the turbulence within the vessel, the model was based
on the probability of recovering a given particle to the concentrate.
Gupta et al. (1999) also worked on this topic, focusing on phosphate otation. In order to nd some relationships between rate
constants and operating variables, they suggested a hybrid model
combining rst principles and neural networks. Once calibrated,
the prediction of the effect of frother concentration, particle size,
air ow rate, and bubble diameter on phosphate recovery was
made possible.
522
It must be emphasized that for industrial applications, calibrating any model is a tedious process, and the difculty increases with
the number of parameters. The more the model relies on empirical
data the more its validity is compromised as process conditions
exit the range of calibration. However, requiring a calibration
should not preclude the use of a given model since many sensors
commonly utilized rely on calibration (e.g. on-stream analyzer,
density-meter, etc.).
Other researchers approached the issue of modelling of the
recovery from a more particular point of view. For instance, Yianatos et al. (1998) developed a semi-empirical model to predict the
froth recovery in an industrial otation column, as a function of
the main operating variables (air and water supercial velocities
and froth depth). Neethling and Cilliers (2001) suggested a comprehensive fundamental model to study the effect of wash water
on froth performance. Many complex phenomena were incorporated in their description: bubble coalescence, liquid drainage, particle settling and particle dispersion.
New developments and studies dealing with characterizing
hydrodynamic properties will help improving current models and
simulators. A revision of underlying assumptions commonly used
previously is to be expected as macroscopic behaviours are better
understood.
3.2. Dynamic behaviour
Despite the validity of the approach initiated by Sastry and Lofftus (1988) and pursued by Pate and Herbst (1989) and Cruz (1997),
the dynamic modelling of column otation based on rst principles
has not yet been successful to stand out as a solution to practical
problems. These attempts were eventually dropped to focus on
empirical techniques (transfer functions, state-space or fuzzy models) to model the behaviour of process variables such as froth
depth, air hold-up, and concentrate grade. Noteworthy results
were obtained with linear models for restricted ranges of operation
(Pal and Masliyah, 1990; Pu et al., 1991; Bergh and Yianatos, 1994;
Bergh et al., 1995; del Villar et al., 1999; Bouchard et al., 2005b).
However, some limitations were identied when considering
wider operating regions (Carvalho et al., 1999) as the linear behaviour assumption does not hold anymore. Desbiens et al. (1998) and
Milot et al. (2000) tested gain-scheduling and multi-model
schemes to overcome these problems and cope with nonlinearities
in a straightforward manner.
More recently, Vieira et al. (2005) proposed a fuzzy modelling
strategy to obtain a MIMO (multiple inputs/multiple outputs) representation of the behaviour of froth layer height, bias and air holdup. Despite a good agreement between the model and process,
their work illustrates fairly well the main difculty associated with
empirical-based representations attempting to stand as fully general tools: the empirical cost. In fact, the accuracy of black-box
models relies almost exclusively on the information contained
within the experimental data. The more complex is the behaviour
to be modeled as a result of the number of independent variables,
interactions and nonlinearities the more data are required for
calibration purposes. In practice, following the example presented
in their paper, empirical nonlinear models must be restricted to
relatively narrow range of operations. Other researchers followed
a hybrid path to overcome such a problem. For instance, Dumont
et al. (2001) used two models of the froth depth behaviour essentially based on rst principles to demonstrate how mixing the
empirical approach with the fundamental one, using basic physical
laws (Newtons second law, ideal-gas law, Archimedes principle,
etc.), leads to more general tools which require less experimental
data for calibration.
The development of dynamic relationships linking operating
variables (froth depth, bias, air hold-up, and bubble surface area
ux) to the metallurgical performance of otation columns (grade

and recovery) is another interesting issue. Empirical graphical relationships were presented by Nesset et al. (2005), Hernandez-Aguilar et al. (2006) and Bartolacci et al. (2008). Work by Alexander
et al. (2005) and Schwarz et al. (2008) showed the use of Sb ; g ,
froth recovery (Rf ), entrainment and drainage parameters (n and
d), and the ore oatability parameter Pi to predict otation performance from simulations.
3.3. Soft sensors
Modelling efforts also targeted improving the instrumentation
for otation columns. The pulp level has attracted considerable
attention through experimental work to improve pressure- and
conductivity-based techniques (Finch and Dobby, 1990; Bergh
and Yianatos, 1993; del Villar et al., 1995a,b, 1999; Maldonado
et al., 2008a).
In the past few years, noteworthy prototypes have also been
presented for the air hold-up, even though no commercial applications seem to be reported in the literature. As for the froth depth,
sensors using pressure gauges were rst suggested (Finch and Dobby, 1990), but the more accurate conductivity probes constitute
the current research trend (Gomez et al., 1995; Tavera et al.,
1998; Prez-Garibay and del Villar, 1999; Tavera and Escudero,
2002; Nez et al., 2006a). In fact, as shown by Nez et al.
(2006a), the assumption of a constant slurry density may introduce
some inaccuracy problems with pressure-based sensors. A unique
solution to the gas holdup monitoring problem is also offered by
TM
CiDRA Corporation with the SONARtrac . OKeefe et al. (2007,
2008) reported successful applications of this multivariable uid
velocity/gas hold-up sensor, based on a passive sonar technology,
in a variety of mineral processing facilities. Despite a great deal
of work, on-line estimation techniques for the bubble surface area
ux and the bias have not reached the same degree of maturity,
although work is ongoing at Universit Laval and McGill
University.
Bubble surface area ux monitoring issues are related to bubble
size estimation. Off-line devices for bubble size estimation have
been developed by Randall et al. (1989), at McGill University (Finch
et al., 1999; Chen et al., 2001; Hernandez-Aguilar et al., 2002; Gomez and Finch, 2002, 2007), by Grau and Heiskanen (2002, 2003),
Schwarz and Alexander (2006a) and also by Rodrigues and Rubio
(2003). For fully on-line applications, only an indirect method
based on the so-called drift ux theory (Dobby et al., 1988; Yianatos
et al., 1988; Xu and Finch, 1990; Zhou et al., 1993; Banisi and Finch,
1994; Ityokumbul et al., 1995; Li, 2003) seems conceivable at the
moment. However, recent advances in bubble size distribution
estimation and modelling are promising for accurate on-line computation of mean bubble diameter (Maldonado et al., 2008b).
Soft sensors developed for bias estimation have always been
based either on conductivity or mass balance calculations, direct
or indirect (i.e. through calibrated empirical relationships such as
neural networks or regression models), and assumed a steady-state
operation (Uribe-Salas et al., 1991; Bergh et al., 1995; Prez et al.,
1996; Carvalho et al., 1999; Bouchard et al., 2005b). It is only recently that Maldonado et al. (2008c) proposed a way to monitor
the transient state of this variable at least in a two-phase (water
and air) laboratory application. The innovation comes from an estimation of the dynamics through a change of conductivity below
the interface. It is worth mentioning that this is the rst attempt
ever showing some potential for industrial applications. However,
the method is yet not adapted for slurries and the required calibration is likely to put a curb on sustainable use in a plant unless it
could be performed in a laboratory-scale column.
Persechini et al. (2000) also developed and tested three soft sensors to monitor froth depth, bias and air hold-up for a simplied
523
two-phase system. The latter process variables were inferred using

only two ow-meters (air and wash water) and two pressure
gauges mounted on the column side. The proposed approach is certainly interesting but would require further developments. A few
reservations that are worth mentioning are:
ing. An optimisation refers to the objective search for the

best solution, based on a process model, between all possible cases in a given range and considering input and output
constraints. An improvement consist in nding a better situation than the base case.
simplistic models were used to make up for the absence of pulp

owmeters (ow rate proportional to the voltage at the pump
terminals),
the instrumentation would need to be adapted for a slurry application, which represents the ultimate objective, and
reliability and validity of the bias and gas hold-up measurements were not assessed, even for the steady-state operation.
For industrial applications, advanced control strategies generally require data validationestimation stages, such as massbalTM
ance reconciliation (e.g. Bilmat Real-Time ), observers (e.g.
Kalman-like lters), and fault detection and diagnosis (Berton
and Hodouin, 2003), for more robustness and accuracy. It must also
be emphasized that the success of any advanced control strategy
strongly relies on the regulatory control layer. Upgrading from a
lower to a higher level should be considered only if the lower level
is fully and sustainably implemented. Robustness, reliability and
simplicity should be the primary focus regarding any choice of
instrumentation and control algorithm.
Finch and Dobby (1990), Bergh and Yianatos (1993) and Rubinstein (1995) have presented general discussions about column otation control. Process and instrumentation issues are covered, but
conclusions drawn about the controllability of the process and the
ability of conventional feedback control algorithms to handle it
(Bergh and Yianatos, 1993) are not in line with multivariable control practice, nor with other results found in the literature as explained below.
A physical analysis of the system shows that there are multiple interactions between the input (independent) and output
(dependent) variables. The computation of the RGA (relative gain
array) matrix allows to quantify this level of interaction for the
steady-state. Persechini et al. (2004) have thus shown that for
the froth depth bias gas hold-up system, respectively controlled using the wash water, tailings, and air ow rates, the
RGA matrix is almost an identity matrix. The process would thus
exhibit low interactions and therefore, be a good candidate for
decentralized control strategies. Similar observations were also
made by Maldonado et al. (2007a). Even if this result is obtained
for a simplied two-phase system, it is difcult to imagine that
the result for a slurry operation would be substantially different.
It should be emphasized that tuning decentralized PIDs is trickier
than tuning PIDs for completely independent processes. Good results are difcult to achieve without an understanding of multivariable feedback control theory (see Skogestad and
Postlethwaite, 2005). For TITO (two-inputtwo-output) decentralized control schemes, Desbiens et al. (1996) presented a PID tuning technique.
The critics of the control performance of standard feedback control techniques are difcult to understand since useful results were
obtained by some researchers at least for restricted ranges of operation using PI controllers alone (del Villar et al., 1999; Persechini
et al., 2004; Bouchard et al., 2005b, linear predictive controllers
Pu et al., 1991; Chuk et al., 2001; Maldonado et al., 2007b), and
multi-model schemes (Milot et al., 2000; Bouchard et al., 2005b).
It must be emphasized that the controllability of any process remains unaffected by uctuating operating conditions as long as
manipulated variables are unsaturated and the dynamic behaviour
does not vary signicantly (compared to the model used for the
controller design). However, uctuating operating conditions add
a challenge that cannot be neglected. This can even lead to the failure of a given control strategy if the limitations are not clearly
understood and properly dealt with, no matter what control algorithm is being used. In fact, the vast majority of industrial control
issues come from instrumentation problems (type, design, location
and/or condition), control strategy (pairing, objective, etc.) and
controller tuning. Very seldom they are inherent to the control
technique per se.
In a previous study, Hyma and Salama (1993) suggested a similar approach, but only preliminary results were presented.
4. Process control
Column otation process control objectives may be structured
according to a three-level hierarchy.
(1) This is the basic regulatory control layer. Above all, it aims at
a steady operation for the column inputs (slurry feed if
upstream uctuations are damped in the pump box, air,
wash water and reagents), but refers in general to all control
elements (valves and pumps, including that of the tailings).
From a process/design standpoint, the rst level also
includes damping feed ow-rate disturbances as much as
possible using the surge capacity of a pump box.
(2) The intermediate level aims at maintaining process variables
having a strong inuence on metallurgical results (grade and
recovery), namely froth depth, bias, froth parameters, air
hold-up or bubble surface area ux, in a bounded region,
i.e. an acceptable operating zone where it is possible to handle the process and reach production objectives. This level is
often called stabilizing control (Finch and Dobby, 1990;
Kosick et al., 1991; Hyma and Salama, 1993; Bergh and
Yianatos, 1993; Bergh and Yianatos, 1995; Rubinstein,
1995; Bergh et al., 1998; Bergh and Yianatos, 1999). Process
and security constraints should normally be included at this
level. This could include handling a circulating load in a
cleaning circuit for instance. A model-based predictive controller (MPC) can manage such constraints very easily. The
alternative when only PID controllers are available is to
use either an override or a pseudo-cascade scheme (Lestage
et al., 1999).
(3) The third layer involves the determination of metallurgical
targets according to an economic criterion. From these targets, operating set-points for the subordinate level are
found to drive the process or the circuit from its current
state to a new one in order to achieve the economic objective. Finch and Dobby (1990), Kosick et al. (1991), Hyma
and Salama (1993) and Rubinstein (1995), among others,
refer to this upper level as optimizing control. The term
supervisory control is also commonly used when other tasks
such as data validation, fault detection, security or limiting
conditions are also carried out (Bergh and Yianatos, 1993;
Bergh et al., 1998; Bergh and Yianatos, 1999). However,
when based on the rigorous resolution of a quantitative
optimisation problem, the control community rather refers
to real-time optimisation (RTO). Supervisory control structure
are often based on fuzzy rules, trying to emulate the best
possible operator. The main difference between the two
schemes is one aims at optimising, and the other, at improv-
524
4.1. Intermediate level control

SISO control strategies have been suggested for secondary control objectives. The most widely used in the industry is froth depth
control using the tailings ow rate. Mauro and Grundy (1984) already reported the early application of such a strategy at Lornex
Mining Corporation. A few years later, Nicol et al. (1988) tested it
with a pilot-scale unit, and Moys and Finch (1988b), using a laboratory-scale unit. Desbiens et al. (1998) proposed a gain-scheduled
technique to cope with nonlinearities. Barrire et al. (2001) tested
two nonlinear controllers based on semi-physical models developed by Dumont et al. (2001). Another version of pulp level control
is based on the wash water ow rate (Moys and Finch, 1988a;
Finch and Dobby, 1990). An application at Les Mines Gasp was reported by Cienski and Cofn (1981).
TITO strategies have also been tested to achieve intermediate level control objectives. They all involve the froth depth and another
variable, the bias or gas hold-up. Ameluxen et al. (1988) intended a
froth depth and bias control, using the wash water ow rate and the
tailings ow rate respectively, at Southern Peru Copper Corporation
(Cuajone Division). Using a two-phase laboratory-scale column, del
Villar et al. (1999) tested a decentralized control structure, but
using the wash water ow rate to control the bias, and the tailings
ow rate for the froth depth. Slurry operation pilot-scale results
were presented by Bouchard et al. (2005b) for the same TITO system. Milot et al. (2000) used a two-phase application to test a multivariable nonlinear predictive controller (GlobPC Global
Predictive Control) also to control froth depth and bias.
Pu et al. (1991) proposed a predictive controller (DMC Dynamic Matrix Control) for a slurry laboratory application involving
the froth depth and gas hold-up. Process variables were estimated
using three pressure-transducers. Using a similar approach, Chuk
et al. (2001) and Nez et al. (2006b) tested a GPC (Generalized
Predictive Controller) and a GlobPC (Global Predictive Control),
respectively.
Persechini et al. (2004) and Maldonado et al. (2007a) tested
decentralized strategies for bias, froth depth, and gas hold-up (simplied two-phase water-air system). Another strategy, combining a
PI controller for the froth depth and a predictive controller for the
bias and gas hold-up was also proposed for the same system by
Maldonado et al. (2007b).
Other researchers and practitioners have tried more qualitative
techniques. Kosick et al. (1991) reported the implementation of an
expert system at Doe Run Viburnum (Misouri, USA) and at Nerco
Con Mine (Yellowknife, Canada). Carvalho and Duro (2002) tested
a fuzzy logic controller for the froth depth, bias, and gas hold-up on
a two-phase laboratory-scale unit.
4.2. Control strategies based on metallurgical objectives
Examples of column otation control based on metallurgical
objectives are scarce in the technical literature. Generally, the proposed schemes try to improve the performance in terms of the
metallurgical efciency of an individual unit or a circuit and cope
with irregular or undesirable situations.
Expert systems were implemented by McKay and Ynchausti
(1996) to supervise column operation by manipulating froth depth,
air ow rate, and wash water ow rate set-points. Other applications have been presented by Bergh and Yianatos (1996) (El Teniente, Codelco-Chile), and Bergh and Yianatos (1999) (Salvador,
Codelco-Chile). Besides illustrating the benets of improving control strategies, the latter papers show the overall work associated
with industrial implementations, including a pre-diagnosis step
to detect and correct operation and maintenance problems. Vendors like Metso Minerals and SGS Minerals Services also propose
expert systems to handle column otation operations.
Concentrate grade control in a zinc cleaning column using fuzzy

logic was reported by Hirajima et al. (1991). The control rules were
obtained by interviewing skilled operators. The strategy was
mainly based on manipulating the air ow rate to reach concentrate grade objectives.
Karr (1996) discussed the possibility of using a column otation
neural network model within an adaptive control architecture.
Simulation studies have also been conducted. Bergh et al.
(1998) presented a hybrid system combining expert and fuzzy logic for the supervision of a decentralized PID control strategy (froth
depth, wash water, and air ow rates). Chuk et al. (2005) suggested
a supervising expert system to prevent froth collapse.
5. Current practice, research trends & future applications
Previous sections have shown that considerable effort has already gone into developing column otation models and designing
suitable process control strategies. That being said, one can now
ask:
How the current industry could, from a practical point of view,
benet from these developments?
What are the needs for further investigations?
This section examines these issues according to four topics,
namely
steady-state simulation of metallurgical performance of column
otation,
dynamic modelling,
sensor development and applications, and
process control.
5.1. Current practice and challenges
Performance of industrial column is still often limited by design
and maintenance issues. Air sparger or nozzles require recurrent
cleaning to cope with water accumulation in air lines (if the air
supply is not pre-dried) and plugging with slurry (from the column). Such problems must be diagnosed by monitoring the pressure at every air injection point and dealt with rapidly because
they prevent a steady gas ow-rate and proper aerodynamics conditions in the column.
Another common issue with column otation operation is the
uneven wash water distribution. Wrong design and fouling of the
wash water tray are the main causes. Correcting the design is generally relatively easy, by adding bafes and/or feeding the water to
the tray at more than one point for instance. Fouling issues are
more difcult to cope with because they require a lot of diligence
for periodic cleaning.
Pinch valves are commonly used for tailings ow control. Even
though they provide a satisfactory performance, they remain relatively high maintenance control elements. Recent experiences (e.g.
at Xstrata Nickel Raglan) show that ceramic ball valves can nowadays be considered for long term and low maintenance operation.
In large column circuits, splitting a feed evenly into several
gravity fed parallel units in order to fully utilize the processing
capacity is challenging. Unfortunately, designing suitable pulp
distributors does not seem to be straightforward and has to be
dealt with on a case-by-case basis.
5.2. Steady-state simulation: metallurgical performance
Based on recent experimental developments, a new Windowsbased version of JKSimFloat, a otation simulator commercialized
by JKTech, is available. Harris et al. (2002) described how the software incorporates new ideas and models, and discussed practical
and specic issues related to design and optimisation studies.
JKSimFloat V6.1 is the tangible result of a collaborative research
project involving the Julius Kruttschnitt Mineral Research Centre
(JKMRC) at the University of Queensland (Australia), the Mineral
Processing Research Unit at the University of Cape Town (South
Africa), and the Mineral Processing Group at McGill University
(Canada). Previous outcomes of this project were the development
of measuring devices and procedures for the diagnostic performance
evaluation of otation cells and circuits, a methodology for modeling
the performance of otation circuits for optimisation studies, and a
methodology for modeling the performance of otation pilot-plants
for design studies (Harris et al., 2002). Application case studies
were presented by Schwarz and Alexander (2006b).
The software could become a powerful tool for process engineers to choose operating points for secondary-objective variables
(e.g. froth depth, gas hold-up, etc.) in order to reach metallurgical
objectives. As part of an off-line or real-time optimisation strategy,
the determination of column operating conditions could then be
made on a quantitative basis. However, establishing explicit relationships linking the metallurgical performance to the pulp level,
bias, gas hold-up and/or bubble surface area ux is still an inexact
science. The software raises high expectations: applications in
industrial optimisation studies will show if they can be fullled.
As a part of a comprehensive design methodology, the software
also aspires to become a key scale-up tool for otation units and
circuits.
5.3. Sensor development and applications
Over the past two decades, new measuring devices have been
proposed although industrial applications remain scarce. Industrial-academic partnerships could be of great benet for better process supervision. In fact, equipment and service suppliers have the
technical and practical expertise to design robust and reliable
products, whereas the universities have the facilities and mandate
to develop theoretical concepts and explore new avenues. The joint
work of JKMRC and its commercial branch JKTech is one example
of a successful combination of fundamental and industrial and
R&D. By taking advantage of the best of both worlds, nearly mature
technologies like conductivity-based probes for froth depth and
gas hold-up could then rapidly become standard devices.
Froth depth determination based on conductivity exhibits a
very good accuracy in laboratory- and pilot-scale units. Unlike
oat- and pressure-based techniques, it is unaffected by any uctuations in the pulp density or by air hold-up. The evaluation of
the conductivity prole across the froth could also allow an on-line
monitoring of the bias. However, a comprehensive trial in a plant
environment is yet to be performed. It is only then that a valid
comparison with commonly used methods (i.e. oat- and pressure-based) will become possible.
The standard addition method developed by Prez-Garibay and
del Villar (1999) for gas hold-up monitoring has also been tested
by Arizmendi-Morquecho et al. (2002) for solids hold-up measurement. Such a technique could become a low-cost and safe substitute for nuclear densimeters.
Current industrial use of bubble surface area ux (local reference) is nowadays for operation diagnosis, but opportunities for
process control to achieve a target metallurgical performance
could originate from the development of accurate and robust
on-line estimation method. Unfortunately, important information
related to the shape of the bubble size distribution, such as multi-modal, narrowness and tail behaviour is completely lost when
using a mean bubble diameter as with Sb . Therefore, formal control
strategies should take into account the Sb value in conjunction with
525
the shape of the bubble size distribution, which must be estimated

on-line from sequential bubble size data points (Maldonado et al.,
2008b).
Supervision methods based on multivariate image analysis
(MIA) are promising options for froth characteristic monitoring.
For certain applications, MIA could complement on-stream analyzers (OSA) and help to overcome their limitations (e.g. time delays,
need of sampling systems, maintenance, calibration, etc.). For instance, Duchesne et al. (2003) proposed an application of MIA for
concentrate grade prediction. In this case, the grade predictions
could either be used to ll in gaps between OSA analyses or to
monitor extra concentrates without upgrading the sampling and
multiplexing systems. Liu et al. (2005) presented a novel method
to extract textural and color information related to the bubble size
distribution, and the presence and amount of clear windows (or
black holes) on otation froth bubbles. Liu and MacGregor (2008)
presented how the scores of the MR-MIA (multiresolutional multivariate image analysis) features could be directly used in froth control. Froth image analysis based on MIA could become the
alternative to some instrumentation problems of otation columns, for instance, to monitor water entrainment. If froth color
and/or texture is related to water content, it is possible to build a
regression model linking froth image characteristics to the water
hold-up of the concentrate. As in any image analysis application,
conclusions would be drawn using only surface features. Therefore,
even if the technique may provide a valuable information, the local
nature of the results should never be forgotten (the state of the
bubbles even one layer below the surface remains unknown). A
more detailed discussion on numerical image analysis potential
for the monitoring of froth characteristics was presented by Bartolacci et al. (2006).
Commercial froth imaging systems are available to monitor different froth parameters such as velocity, bubble size, stability and
colour intensity (e.g. Metso Minerals and SGS). They are widely
used in otation plants and there is a growing demand both for
new and existing column installations.
Another challenging area for future developments is the on-line
evaluation/modelling of the bubble size distribution (BSD). The use
of the BSD for control purposes could lead to great metallurgical
benets, particularly if it can be matched to the prevailing particle
size distribution feeding the otation unit. Work is presently
underway regarding this matter at Universit Laval.
5.4. Dynamic modelling and simulation
Above all, dynamic models are required for process control purposes. Laboratory- and pilot-scale studies have shown a high potential for the use of linear empirical models (transfer functions
and state-space models). However, such a black-box approach
has a limitation: it is restricted to conditions found in the data used
for calibration (operating points, ore characteristics, conditioning,
etc.) and, therefore, may require frequent recalibration. To overcome this drawback, to better understand the interacting process
involved and to obtain more versatile controllers, phenomenological model development should be the focus of researchers. After
the rst attempts in the late eighties (Sastry and Lofftus, 1988,
1989), Cruz (1997) made signicant progress in column otation
dynamic simulation. Combining the latest work on froth modelling
(Neethling and Cilliers, 2001; Neethling and Cilliers, 2002a; Neethling and Cilliers, 2002b; Neethling, 2008; Stevenson et al., 2003;
Stevenson, 2007; Nguyen et al., 2003) with the approaches proposed by Cruz (1997) and Bouchard et al. (2006) will probably provide the next generation of column otation dynamic simulators.
Fundamental dynamic models may be used to build robust simulators. For process control engineers, dynamic simulators are useful tools since:
526
(1) they allow the development and study of control strategies

without upsetting production,
(2) they allow pre-tuning before the actual implementation
hence reducing the commissioning period, and
(3) they can be used directly as a process model in predictive
controllers as suggested by Henson (1998).
Desbiens and Bouchard (2004) and Bouchard et al. (2005a) presented novel predictive control formulations to make use of this
latter concept.
It must be emphasized that simulator calibration and recalibration requirements can become overwhelming and therefore,
the number of empirical parameters should remain reasonable.
The range of validity and robustness can be extended typically
by including fundamental knowledge (i.e. physical laws, conservation principles, etc.) into the models, hence minimizing the dependence over empirical parameters and the data used for
calibration to explain the behaviour. Moreover, general trends
are more important than the actual accuracy for advanced process
control applications, particularly when the aim is to control low
frequencies or steady-states (e.g. grade and recovery).
5.5. Process control myths and reality
Column otation control is often seen as a problem that cannot
be handled by standard feedback control techniques. Such a vision
certainly put a curb on a major breakthrough in column control.
Most of column otation control issues are instrumentation
matters, related to the lack of commercial sensors to monitor key
process variables. Obviously, any linear control law is restricted
to a certain range of operation, but it does not automatically preclude their use for processes exhibiting some nonlinear behaviour.
In fact, column otation should be classied under the category of
quasi-linear systems with different operating points. Therefore, it
is not different from any other industrial process where standard
feedback control techniques have been successfully implemented.
Moreover, if a wide operating region must be considered, using
gain-scheduling and multi-model schemes may enlarge the range
of applicability of the controllers (Bouchard et al., 2005b).
To solve many column otation control problems, simple and
effective control techniques are already available as discussed by
Bouchard et al. (2005b). At the very least, pulp level control should
always be implemented. PID controllers are well suited for such an
operation and inexpensive commercial on-line pulp level measuring devices are already available (pressure gauges or oat coupled
with ultrasounds). Even though there is still room for improvement, plant practice shows that this is actually no longer a critical
issue since most of industrial applications involves froth depth
control. If, in addition, air ow-meters are used, the effect of the
air ow rate can easily be implemented to ensure a wider range
of operation (Desbiens et al., 1998). A tight froth depth control provides several possibilities to improve column operation. For instance, it greatly helps operators to reach production objectives.
It also provides a better environment to conduct experiments for
studying the effect of operating conditions on metallurgical results,
since steady-state operation is more easily reached and maintained. From a plant practice standpoint, a steady pulp level operation also relates to a great extent to using surge capacity to
attenuate high frequency throughput uctuations wherever
possible.
Bias and gas hold-up are generally not supervised with great
accuracy. The passive sonar technology recently introduced in
mineral processing plants (OKeefe et al., 2007; OKeefe et al.,
2008) seems to be mature enough now to solve the gas hold-up
monitoring issue. The technology also offers a great potential for
bias steady-state estimation since it can be adapted as well to
monitor concentrate ow-rates if the velocity requirements are

met. It must be emphasized that the steady-state assumption can
become acceptable when considering a sufciently long time operating window. Moreover, mass-balancing technique using mean or
TM
reconciled (using Bilmat Real-Time for instance) values over the
moving window would greatly improve accuracy.
On-line stream analysers are widely used to monitor the metallurgical performance. However, only seldom an explicit grade or
recovery closed-loop control is achieved. Off-the-shelf expert systems are available to do so, but a common plant practice is to manually adjust reagent addition when not ratio controlled from the
circuit head grade and/or manipulate gas rate and froth depth
(for instance at Xstrata Nickel Raglan and Strathcona concentrators). The effect of wash water does not seem to be as clearly understood even though the wash water ow-rate is often monitored.
The use of simple correlations obtained from in-plant empirical
studies (using design of experiment approach) is not a panacea, but
it may partly make up for the lack of measurements and provides a
rst quantitative basis to guide process operation improvements.
The same idea may also be used to link critical operating variables
to the metallurgical performance. If the limitations of such an approach are well understood, for instance regarding the validity
range and the necessity of recalibration, it may be used in off-line
or even real-time optimisation strategies. Lestage et al. (2002) presented a RTO application for a grinding circuit based on linear
empirical models. Such a successful methodology could be transposed to otation columns.
Implementing a rigorous optimisation plan requires investments involving nancial (purchase of a simulator, hiring of a consulting rm, etc.), human and time resources (for implementation
and the strict minimum maintenance). However, the reward, assessed in terms of
lower production costs (reagent consumption, etc.),
better metallurgical performance (grade and recovery),
lower product quality variability,
benets on personnel (availability, training on new technologies, etc.)
fewer environmental impacts,
etc.

may exceed original expectations and the payback period tends to

be much faster than in the case of major capital investments (few
weeks to few months). This last point can partly be explained by
the fact that tens and even hundreds of thousands of dollars of
computer equipment, controllers, DCSs, PLCs, and sensors are
already available in mineral processing plants and, in most cases,
only their simplest features are being used and the most powerful
ones are often left on the shelf, mainly due to the lack of human
resources. In other words, many control wares have already been
paid off, but their potential has not been fully exploited.
More advanced control and supervision techniques may eventually be used to improve otation column operation. For instance,
nonlinear controllers based on phenomenological models could
enhance the range of validity, performance, robustness, and perhaps reduce the recalibration needs of control systems. Multivariate statistics could also be used to enhance monitoring and
diagnosis capabilities for otation column operations as recently
suggested by Bergh et al. (2005) and Bergh and Acosta (2008).
6. Conclusion
The modelling and control of column otation has received particular attention from the mineral processing community since almost three decades. An overview of the literature pertaining to this
eld shows that much has been achieved, but also that some technologies still remain to be transferred to industry. A signicant effort to bring modelling capabilities to process engineers is the
development of a commercial Windows-based simulator. Following the example of JKMRC and its commercial branch JKTech, fruitful collaborations between academics and practitioners should
help speed up developments in other areas such as on-line sensors
and process control. For many plants, important gains can be made
with the simple application of standard control techniques for
intermediate process variables (froth depth, bias, ow rates, etc.).
Finally, the use of multivariate statistics for on-line monitoring
and phenomenological dynamic modelling are current research
interests showing great potential. New technologies and revived
works will probably guide future investigative efforts.
Acknowledgements
The authors wish to thank Simon Garipy (Algosys) and Pr. Carl
Duchesne (Universit Laval) for fruitful discussions and exchanges
on plant practice and MIA, respectively.
References
Alexander, D., Bilney, T., Schwartz, S., 2005. Flotation performance improvement at
Placer Dome Kanowna Belle Gold Mine. In: Proceedings 2005 37th Annual
Meeting of the Canadian Mineral Processors. CIM, Ottawa, Canada, pp. 171201.
Alford, R.A., 1992. Modelling of single otation column stages and column circuits.
International Journal of Mineral Processing 36, 155174.
Ameluxen, R., Llerena, R., Dustan, P., Huls, B., 1988. Mechanics of column otation
operation. In: Sastry, K. (Ed.), SME Annual Meeting Column Flotation88.
Society of Mining, Metallurgy and Exploration, Phoenix, Arizona, pp. 149155.
Arizmendi-Morquecho, A., Garibay, P., Uribe, A.S., Nava, F.A., 2002. On-line solids
hold-up measurement in mineral slurries by the standard addition method.
Minerals Engineering 15 (6164).
Banisi, S., Finch, J., 1994. Reconciliation of bubble size estimation methods using
drift ux analysis. Minerals Engineering 7 (12), 15551559.
Barrire, P.-A., Dumont, F., Desbiens, A., 2001. Using semi-physical models for better
control. Part 2: nonlinear control of a pilot otation column. In: International
Federation of Automatic Control (IFAC) Symposium on Automation in Mining,
Mineral and Metal Processing. IFAC, Tokyo, Japan, pp. 131136.
Bartolacci, G., Pelletier, P., Tessier Jr., J., Duchesne, C., Boss, P.-A., Fournier, J., 2006.
Application of numerical image analysis to process diagnosis and physical
parameter measurement in mineral processes part I: otation control based
on froth textural characteristics. Minerals Engineering 19, 734747.
Bartolacci, G., Ourriban, M., Knuutila, D., Lockhart, A., Finch, J.A., Lau, C.H., Fortin, A.,
Goyette, G., 2008. The link between pulp zone and hydrodynamic
characteristics and froth stability in a 125 m3 otation column. In: 40th
Annual Meeting of the Canadian Mineral Processors, Ottawa, Ontario, Canada,
pp. 311332.
Bergh, L., Acosta, S., 2008. On-line fault detection on a pilot otation column using
linear PCA models. In: Automining 2008, Santiago, Chile, pp. 5766.
Bergh, L., Yianatos, J., 1993. Control alternatives for otation columns. Minerals
Engineering 6 (6), 631642.
Bergh, L., Yianatos, J., 1994. Experimental studies on otation column dynamics.
Minerals Engineering 7 (2/3), 345355.
Bergh, L., Yianatos, J., 1995. Dynamic simulation of operating variables in otation
columns. Minerals Engineering 8 (6), 603613.
Bergh, L., Yianatos, J., 1996. Hierarchical control strategy in columns at El Teniente.
In: Gomez, C.O., Finch, J.A. (Eds.), Column96 Proceedings of the International
Symposium on Column Flotation. Mineral Sciences and Engineering Section of
the Metallurgical Society of CIM and Canadian Mineral Processors Division of
CIM, Montral, Qubec, pp. 369380.
Bergh, L., Yianatos, J., 1999. Supervisory control experience on large industrial
otation columns. In: Hodouin, D., Bazin, C., Desbiens, A. (Eds.), Proceedings of
the International Symposium on Control and Optimization in Minerals, Metals
and Materials Processing. MetSoc, Qubec, Canada, pp. 299310.
Bergh, L., Yianatos, J., 2003. Flotation column automation: state of the art. Control
Engineering Practice 11, 6772.
Bergh, L., Yianatos, J., Acua, C., 1995. Hierarchical control strategy for otation
columns. Minerals Engineering 8 (12), 15831591.
Bergh, L., Yianatos, J., Leiva, C., 1998. Fuzzy supervisory control of otation columns.
Minerals Engineering 11 (8), 739748.
Bergh, L., Yianatos, J., Len, A., 2005. Multivariate methods applied to otation
columns. Minerals Engineering 18, 721723.
Berton, A., Hodouin, D., 2003. Linear and bilinear fault detection and diagnosis
based on energy and mass balance equations. Control Engineering Practice 11,
103113.
Bouchard, J., Desbiens, A., del Villar, R., 2005a. Optimization-free constrained
nonlinear control mineral processing applications. In: 16th IFAC World
Congress, Praha, Czech Republic, We-A22-T0/1.
527
Bouchard, J., Desbiens, A., del Villar, R., 2005b. Recent advances in bias and froth
depth control in otation columns. Minerals Engineering 18, 709720.
Bouchard, J., Desbiens, A., del Villar, R., Chirinos, J., 2006. Column otation
simulation: a dynamic framework. In: International Conference on Mineral
Process Modelling, Simulation and Control, Sudbury, Ontario, pp. 259275.
Carvalho, M.T., Duro, F., 2002. Control of a otation column using fuzzy logic
inference. Fuzzy Sets and Systems 125, 121133.
Carvalho, T., Duro, F., Fernandes, C., 1999. Dynamic characterisation of column
otation process laboratory case study. Minerals Engineering 12 (11), 1339
1346.
Chen, F., Gomez, C., Finch, J., 2001. Bubble size measurement in otation machines.
Chuk, O.D., Mut, V., Nez, E., Gutierrez, L., 2001. Multivariable predictive control of
froth depth and gas holdup in column otation. In: International Federation of
Automatic Control (IFAC) Symposium on Automation in Mining, Mineral and
Metal Processing. IFAC, Tokyo, Japan, pp. 8791.
Chuk, O., Ciribeni, V., Gutierrez, L., 2005. Froth collapse in otation columns: a
prevention method using froth density estimation and fuzzy expert systems.
Minerals Engineering 18, 495504.
Cienski, T., Cofn, V., 1981. Column otation operation at Mines Gaspe molibdenum
circuit. In: 13th Annual Meeting of the Canadian Mineral Processors. Canadian
Mineral Processors Division of the Canadian Institute of Mining, Metallurgy and
Petroleum, Ottawa, Ontario, pp. 240262.
Cruz, E.B., 1997. A comprehensive dynamic model of the column otation unit
operation. Ph.D. Thesis, Department of mining and minerals engineering,
Virginia Polytechnique Institute and State University, Blacksburg, Virginia.
Deglon, D., Sawyerr, F., OConnor, C., 1999. A model to relate the otation rate
constant and the bubble surface area ux in mechanical otation cells. Minerals
Engineering 12 (3), 291308.
del Villar, R., Prez, R., Diaz, G., 1995a. Improving pulp level detection in a otation
column using a neural network algorithm. In: Proceedings 27th Annual
Meeting of the Canadian Mineral Processors. Canadian Mineral Processors
Division of the Canadian Institute of Mining, Metallurgy and Petroleum, Ottawa,
Ontario, pp. 83100.
del Villar, R., Prez, R., Diaz, G., 1995b. Improving the three-pressure transducer
method of level detection in otation columns. In: Casali, A., Dobby, G., Molina,
M., Thoburn, W. (Eds.), Proceedings of Copper 95 Cobre 95 International
Conference, vol. 2. IIMCH, MetSoc, TMS, Santiago, Chile, pp. 247260.
del Villar, R., Grgoire, M., Pomerleau, A., 1999. Automatic control of a laboratory
otation column. Minerals Engineering 12 (3), 291308.
Desbiens, A., Bouchard, J., 2004. Constrained nonlinear predictive control based on
IMC-optimization. In: 11th International Federation of Automatic Control (IFAC)
Symposium on Automation in Mining, Mineral and Metal Processing, Nancy,
France, Thu-C1.
Desbiens, A., Pomerleau, A., Hodouin, D., 1996. Frequency-based tuning of SISO
controllers for two-by-two processes. IEE Proceedings on control theory and
applications 143 (1), 4956.
Desbiens, A., del Villar, R., Milot, M., 1998. Identication and gain-scheduled control
of a pilot otation column. In: International Federation of Automatic Control
(IFAC) Symposium on Automation in Mining, Mineral and Metal Processing.
IFAC, Cologne, Germany, pp. 337342.
Dobby, G.S., Yianatos, J.B., Finch, J.A., 1988. Estimation of bubble diameter in
otation columns from drift ux analysis. Canadian Metallurgical Quarterly 27
(2), 8590.
Duchesne, C., Bouajila, A., Bartolacci, G., Pelletier, P., Breau, Y., Fournier, J., Girard, D.,
2003. Application of multivariate image analysis (MIA) to predict concentrate
grade in froth otation processes. In: Proceedings 2003 35th Annual Meeting
of the Canadian Mineral Processors. CIM, Ottawa, Canada, pp. 511526.
Dumont, F., Barrire, P.-A., Desbiens, A., del Villar, R., 2001. Using semi-physical
models for better control. Part 1: modelling of a pilot otation column. In:
International Federation of Automatic Control (IFAC) Symposium on Automation
in Mining, Mineral and Metal Processing. IFAC, Tokyo, Japan, pp. 137142.
Finch, J.A., Dobby, G.S., 1990. Column Flotation. Pergamon Press, Oxford, UK.
Finch, J., Gomez, C., Hardie, C., Leichtle, G., Filippone, R., Leroux, D., 1999. Bubble
surface area ux: a parameter to characterize otation cells. In: 31th Annual
Meeting of the Canadian Mineral Processors, Ottawa, Canada, pp. 199209.
Finch, J., Xiao, J., Hardie, C., Gomez, C., 2000. Gas dispersion properties: bubble
surface area ux and gas holdup. Minerals Engineering 13 (4), 365372.
Gomez, C., Finch, J., 2002. Gas dispersion measurements in otation machines. CIM
Bulletin 95 (1066), 7378.
Gomez, C.O., Finch, J.A., 2007. Gas dispersion measurements in otation cells.
Gomez, C.O., Tavera, F., Finch, J.A., Prez, R., del Villar, R., 1995. Continuous gas
holdup measurement in otation columns. In: Proceedings 27th Annual
Meeting of the Canadian Mineral Processors. Canadian Mineral Processors
Division of the Canadian Institute of Mining, Metallurgy and Petroleum, Ottawa,
Ontario, pp. 202216.
Gorain, B., Franzidis, J., Malanpig, E., 1995a. Studies on impeller type, impeller speed
and air ow rate in an industrial scale otation cell part I: effect on bubble
size distribution. Minerals Engineering 8 (6), 615635.
Gorain, B., Franzidis, J., Malanpig, E., 1995b. Studies on impeller type, impeller speed
and air ow rate in an industrial scale otation cell part II: effect on gas
holdup. Minerals Engineering 8 (12), 15571570.
Gorain, B., Franzidis, J., Malanpig, E., 1996. Studies on impeller type, impeller speed
and air ow rate in an industrial scale otation cell part III: effect on
supercial gas velocity. Minerals Engineering 9 (6), 639654.
528
Gorain, B., Franzidis, J., Malanpig, E., 1997. Studies on impeller type, impeller speed
and air ow rate in an industrial scale otation cell part IV: effect bubble
surface area ux on otation performance. Minerals Engineering 10 (4), 364
379.
Gorain, B., Franzidis, J., Malanpig, E., 1999. The empirical prediction of bubble
surface area ux in mechanical otation cells from cell design and operating
data. Minerals Engineering 12 (3), 309322.
Grau, R., Heiskanen, K., 2002. Visual technique for measuring bubble size in
otation machines. Minerals Engineering 15, 507513.
Grau, R., Heiskanen, K., 2003. Gas dispersion measurements in otation cell.
Minerals Engineering 16, 10811089.
Gupta, S., Liu, P.-H., Svoronos, S., 1999. Hybrid rst-principles/neural networks
models for column otation. AIChE Journal 45 (3), 557566.
Harris, M., Runge, K., Morrison, R., 2002. JKSimFloat as a practical tool for otation
process design and optimization. In: Mular, A., Halbe, D., Barratt, D. (Eds.),
Mineral Processing Plant Design & Control. SME Publications, pp. 461478.
Heiskanen, K., 2000. On the relationships between otation rate and bubble surface
area ux. Minerals Engineering 13 (2), 141149.
Henson, M., 1998. Nonlinear model predictive control: current status and future
directions. Computers & Chemical Engineering 23, 187202.
Hernandez-Aguilar, J., Gomez, C., Finch, J., 2002. A technique for direct
measurement of bubble size distributions in industrial otation cells. In:
34th Annual Meeting of the Canadian Mineral Processors, Ottawa, Canada
(Paper 24).
Hernandez-Aguilar, J., Thorpe, R., Martin, C.J., 2006. Experience in using gas
dispersion measurements to understand and modify metallurgical
performance. In: Proceedings 2006 38th Annual Meeting of the Canadian
Mineral Processors. CIM, Ottawa, Canada, pp. 387402.
Hirajima, T., Takamori, T., Tsunekawa, T., Matsubara, T., Oshima, K., Imai, T., Sawaki,
K., Kubo, S., 1991. The application of fuzzy logic to control concentrate grade in
column otation at Toyoha Mines. In: Agar, G., Huls, B., Hyma, D. (Eds.),
Column91 Proceedings of an International Conference on Column Flotation,
vol. 2. Canadian Institute of Mining, Metallurgy and Petroleum, Sudbury,
Hyma, D., Salama, A., 1993. Design and implementation of a process control system
for column otation. CIM Bulletin 86 (973), 5054.
Ityokumbul, M., Salama, A., Al Taweel, A., 1995. Estimation of bubble size in
otation columns. Minerals Engineering 8 (1/2), 7789.
Karr, C., 1996. Adaptive process control using biological paradigms. Journal of
Network and Computer Applications 19, 2144.
Kho, C.-J., Sohn, H.-J., 1989. Column otation of talc. International Journal of Mineral
Processing 27, 157167.
Kosick, G., Dobby, G.S., Young, P., 1991. Columnex: a powerful and affordable
control system for column otation. In: Agar, G., Huls, B., Hyma, D. (Eds.),
Lee, K.Y., Pate, W.T., Oblad, A.E., Herbst, J.A., 1991. Methodology for selecting a
control strategy for a column otation unit. In: Agar, G., Huls, B., Hyma, D. (Eds.),
Lestage, R., Pomerleau, A., Desbiens, A., 1999. Improved constrained cascade control
for parallel processes. Control Engineering Practice 7, 969974.
Lestage, R., Pomerleau, A., Hodouin, D., 2002. Constrained real-time optimization of
a grinding circuit using steady-state linear programming supervisory control.
Powder Technology 124, 254263.
Li, H., 2003. tude exprimentale des paramtres de mise lchelle de colonnes de
ottation. Ph.D. Thesis, Dpartement de gnie des mines, de la mtallurgie et
des matriaux, Universit Laval, Qubec.
Liu, J.J., MacGregor, J.F., 2008. Froth-based modeling and control of otation
processes. Minerals Engineering 21, 642651.
Liu, J., McGregor, J., Duchesne, C., Bartolacci, G., 2005. Flotation froth monitoring
using multiresolutional multivariate image analysis. Minerals Engineering 18,
6576.
Luttrell, G., Yoon, R., 1991. A otation column simulator based on hydrodynamic
principles. International Journal of Mineral Processing 33, 355368.
Luttrell, G.H., Adel, G.T., Yoon, R.-H., 1987. Modeling of column otation. In: SME
Annual Meeting, Denver, Colorado (preprint number 87130).
Luttrell, G., Mankosa, M., Yoon, R.-H., 1993. Design and scale-up criteria for column
otation. In: XVIII International mineral Processing Congress, Sydney, pp. 785
791.
Maldonado, M., Desbiens, A., del Villar, R., 2007a. Decentralized control of a pilot
otation column: a 3 3 system. In: 6th International Copper Conference,
Toronto, Canada, vol. 2, pp. 241254.
Maldonado, M., Desbiens, A., del Villar, R., Quispe, R., 2007b. Towards the
optimization of otation columns using predictive control. In: IFAC MMM07
on Automation in Mining, Minerals and Metal Processing. Quebec City, Quebec,
Canada, pp. 7580.
Maldonado, M., Desbiens, A., del Villar, R., 2008a. An update on the estimation of the
froth depth using conductivity measurements. Minerals Engineering 21, 935
939.
Maldonado, M., Desbiens, A., del Villar, R., Girgin, H., Gomez, C., 2008b. On-line
estimation of bubble size distributions using Gaussian mixture models. In:
Kuyvenhoven, R., Gomez, C., Casali, A., (Eds.), V International Mineral Processing
Seminar Procemin 2008, Santiago, Chile, pp. 389398.
Maldonado, M., Desbiens, A., del Villar, R., Chirinos, J., 2008c. On-line bias
estimation using conductivity measurements. Minerals Engineering 21, 851
855.
Mauro, F., Grundy, M., 1984. The application of otation columns at Lornex Mining
Corporation Ltd. In: 9th District Sixth Meeting. CIM, B.C., Canada (p. paper no.
23).
McKay, J., Ynchausti, R., 1996. Expert supervisory control of otation columns. In:
Gomez, C.O., Finch, J.A. (Eds.), Column96 Proceedings of the International
Symposium on Column Flotation. Mineral Sciences and Engineering Section of
the Metallurgical Society of CIM and Canadian Mineral Processors Division of
CIM, Montral, Qubec, pp. 353367.
Milot, M., Desbiens, A., del Villar, R., Hodouin, D., 2000. Identication and
multivariable nonlinear predictive control of a pilot otation column. In: XXI
International Mineral Processing Congress. IFAC, pp. 137142.
Moys, M.H., Finch, J.A., 1988a. Developments in the control of otation columns.
Moys, M.H., Finch, J.A., 1988b. The measurement and control of level in otation
column. In: Sastry, K. (Ed.), SME Annual Meeting Column Flotation88. Society
of Mining, Metallurgy and Exploration, Phoenix, Arizona, pp. 103112.
Neethling, S.J., 2008. Simple approximations for estimating froth recovery.
Neethling, S., Cilliers, J., 2001. Simulation of the effect of froth washing on otation
performance. Chemical Engineering Science 56, 63036311.
Neethling, S., Cilliers, J., 2002a. The entrainment of gangue into a otation froth.
Neethling, S., Cilliers, J., 2002b. Solid motion modeling in owing froths. Chemical
Engineering Science 57, 607615.
Nesset, J.E., Gomez, C.O., Finch, J.A., Hernandez-Aguilar, J., Difeo, A., 2005. The use of
gas dispersion measurements to improve otation performance. In:
Proceedings 2005 37th Annual Meeting of the Canadian Mineral Processors.
CIM, Ottawa, Canada, pp. 401421.
Nguyen, A.V., Harvey, P.A., Jameson, G.J., 2003. Inuence of gas ow rate and
frothers on water recovery in a froth column. Minerals Engineering 16, 1143
1147.
Nicol, S., Roberts, T., Bensley, C., Kidd, G., Lamb, R., 1988. Column otation of
ultrane coal: experience at BHP-Utah Coal Limiteds Riverside mine. In: Sastry,
K. (Ed.), SME Annual Meeting Column Flotation88. Society of Mining,
Metallurgy and Exploration, Phoenix, Arizona, pp. 711.
Nez, E., Chirinos, J., Desbiens, A., del Villar, R., Duchesne, C., 2006a. Multivariable
predictive control of a pilot otation column. part 1: on-line sensors for froth
depth and air hold-up. In: International Conference on Mineral Process
Modelling, Simulation and Control, Sudbury, Ontario, Canada, pp. 277290.
Nez, E., Desbiens, A., del Villar, R., Duchesne, C., 2006b. Multivariable predictive
control of a pilot otation column. Part 2: identication and control. In:
International Conference on Mineral Process Modelling, Simulation and Control,
Sudbury, Ontario, Canada, pp. 291301.
OConnor, C., Mills, P., Cilliers, J., 1995. Prediction of large scale column otation
cells using pilot plant data. The Chemical Engineering Journal 59, 16.
OKeefe, C., Viega, J., Fernald, M., 2007. Application of passive sonar technology to
mineral processing and oil sands applications if you can measure it, you can
manage it. In: 39th Annual Meeting of the Canadian Mineral Processors,
Ottawa, Ontario, Canada, pp. 429457.
OKeefe, C., Poplawski, J., Maron, R., 2008. Application of non-intrusive sonar arraybased technology to solve unique and difcult measurement situations. In: 40th
Annual Meeting of the Canadian Mineral Processors, Ottawa, Ontario, Canada,
pp. 401419.
teyaka, B., Soto, H., 1995. Modelling of negative bias column for coarse particles
otation. Minerals Engineering 8 (1/2), 91100.
Pal, R., Masliyah, J., 1990. Process dynamics and control of a pilot otation column.
Canadian Metallurgical Quarterly 30 (2), 8794.
Pate, W.T., Herbst, J.A., 1989. Dynamic simulation and control of column otation
units. In: Chandler, S., Klimpel, K. (Eds.), Advances in coal and mineral
processing using otation. SME, Littelton, CO, USA, pp. 315321.
Prez, R., del Villar, R., 1996. Measurement of bias and water entrainment in
otation columns using conductivity. In: Gomez, C.O., Finch, J.A. (Eds.),
Column96 Proceedings of the International Symposium on Column
Flotation. Mineral Sciences and Engineering Section of the Metallurgical
Society of CIM and Canadian Mineral Processors Division of CIM, Montral,
Qubec, pp. 327338.
Prez-Garibay, R., del Villar, R., 1999. On-line gas hold-up measurement in otation
columns. Canadian Metallurgical Quarterly 38 (2), 141148.
Persechini, M., Jota, F., Peres, A., 2000. Dynamic model of a otation column.
Persechini, M., Peres, A., Jota, F., 2004. Control strategy for a column otation
process. Control Engineering Practice 12, 963976.
Pu, M., Gupta, Y., Taweel, A.A., 1991. Model predictive control of otation columns.
In: Agar, G., Huls, B., Hyma, D. (Eds.), Column91 Proceedings of an
International Conference on Column Flotation, vol. 2. Canadian Institute of
Mining, Metallurgy and Petroleum, Sudbury, Ontario, pp. 467478.
Randall, E.W., Goodall, C.M., Fairlamb, P.M., Dold, P.L., OConnor, C.T., 1989. A
method for measuring the sizes of bubbles in two- and three-phase systems.
Journal of Physics E: Scientic Instruments 22, 827833.
Rodrigues, R., Rubio, J., 2003. New basis for measuring the size distribution of
bubbles. Minerals Engineering 16, 757765.
Rubinstein, J.B., 1995. Column Flotation Process. Designs and Practices. Gordon
and Breach Science Publishers, Amsterdam B.V.

Sastry, K.V., Lofftus, D.D., 1988. Mathematical modeling and computer simulation of
column otation. In: Sastry, K.V.S. (Ed.), Column Flotation88. SME, Littleton, CO,
USA, pp. 5768.
Schwarz, S., Alexander, D., 2006a. Gas dispersion measurements in industrial
otation cells. Minerals Engineering 19, 554560.
Schwarz, S., Alexander, D., 2006b. JKSimFloat V6:1plus : improving otation circuit
performance by simulation. In: International Conference on Mineral Process
Modelling, Simulation and Control, Sudbury, Ontario, Canada, pp. 3547.
Schwarz, S., Crnkovic, I., Alexander, D., 2008. Flotation circuit analysis at Zinifex
Century Mine. In: 40th Annual Meeting of the Canadian Mineral Processors,
Ottawa, Ontario, Canada, pp. 297310.
Skogestad, S., Postlethwaite, I., 2005. Multivariable Feedback Control Analysis and
Design. second ed.. Wiley.
Stevenson, P., 2007. Hydrodynamic theory of rising foam. Minerals Engineering 20,
282289.
Stevenson, P., Stevanov, C., Jameson, G.J., 2003. Liquid overow from a column of
rising aqueous froth. Minerals Engineering 16, 10451053.
Tavera, F., Escudero, R., 2002. Gas hold-up and solids hold-up in otation columns:
on-line measurement based on electrical conductivity. Transactions of the
Institutions in Mining and Metallurgy, Section C: Mineral Processing and
Extractive Metallurgy 111, C94C99.
529
Tavera, F., Gomez, C., Finch, J., 1998. Estimation of gas holdup in froths by electrical
conductivity: application of the standard addition method. Minerals
Engineering 11 (10), 941947.
Tuteja, R., Spottiswood, D., Misra, V., 1994. Mathematical models of the column
otation process a review. Minerals Engineering 7 (12), 1459
1472.
Uribe-Salas, A., Gomez, C.O., Finch, J.A., 1991. Bias detection in otation columns. In:
Agar, G., Huls, B., Hyma, D. (Eds.), Column91 Proceedings of an International
Conference on Column Flotation, vol. 2. Canadian Institute of Mining,
Metallurgy and Petroleum, Sudbury, Ontario, pp. 391407.
Vieira, S., Sousa, J., Duro, F., 2005. Fuzzy modelling strategies applied to a column
otation process. Minerals Engineering 18, 725729.
Xu, M., Finch, J., 1990. Simplication of bubble size estimation in a bubble swarm.
Journal of Colloid and Interface Science 140 (1), 298299.
Yianatos, J.B., Finch, J.A., Dobby, G.S., Xu, M., 1988. Bubble size estimation in a
bubble swarm. Journal of Colloid and Interface Science 126 (1), 3744.
Yianatos, J., Bergh, L., CortTs, G., 1998. Froth zone modelling of an industrial
Zhou, Z., Egiebor, N., Plitt, L., 1993. Frother effects on bubble size estimation in a

Column Flotation Simulation and Control An Overview

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Column Flotation Simulation and Control An Overview

Încărcat de

Drepturi de autor:

Formate disponibile

Minerals Engineering 22 (2009) 519529

Contents lists available at ScienceDirect

Column otation simulation and control: An overview

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

depressants and pH modiers). These chemicals are added in a

Fig. 1. Flotation column.

where J g is the gas supercial velocity. Fig. 2 illustrates the bubble

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

ux) to the metallurgical performance of otation columns (grade

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

two-phase system. The latter process variables were inferred using

ing. An optimisation refers to the objective search for the

simplistic models were used to make up for the absence of pulp

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

4.1. Intermediate level control

Concentrate grade control in a zinc cleaning column using fuzzy

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

the shape of the bubble size distribution, which must be estimated

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

(1) they allow the development and study of control strategies

monitor concentrate ow-rates if the velocity requirements are

may exceed original expectations and the payback period tends to

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

J. Bouchard et al. / Minerals Engineering 22 (2009) 519529

S-ar putea să vă placă și