A Review of Froth Flotation Control

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

International Journal of Mineral Processing 100 (2011) 5771

Contents lists available at ScienceDirect

International Journal of Mineral Processing


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / i j m i n p r o

A review of froth otation control


B.J. Shean, J.J. Cilliers
Rio Tinto Centre for Advanced Mineral Recovery at Imperial College London, Department of Earth Science and Engineering, Imperial College London, SW7 2AZ, United Kingdom

a r t i c l e

i n f o

Article history:
Received 6 August 2010
Received in revised form 19 April 2011
Accepted 8 May 2011
Available online 14 May 2011
Keywords:
Froth otation
Process control

a b s t r a c t
The last few decades have seen major advances in instrumentation and technology, and simplications and
modications of new otation plant designs. This has allowed for signicant developments in process control.
In particular, the development of base level process control (control of pulp levels, air owrates, reagent
dosing, etc.) has seen signicant progress. Long-term, automated advanced and optimising otation control
strategies have, however, been more difcult to implement. It is hoped that this will change as a result of the
development of new technologies such as machine vision and the measurement of new control variables, such
as air recovery.
This review looks at each of the four essential levels of process control (instrumentation, base level otation
control, advanced otation control and optimising otation control) and examines current and future trends
within each sub-level.
2011 Elsevier B.V. All rights reserved.

Contents
1.
2.

3.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Key variables and considerations in the control of otation . . . . . . . . . . . .
2.1.
Key variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.
Effects of plant layout and the location of the cell in the circuit . . . . . .
2.3.
Types of process input disturbances . . . . . . . . . . . . . . . . . . .
2.4.
System constraints . . . . . . . . . . . . . . . . . . . . . . . . . . .
Instrumentation and base level otation control . . . . . . . . . . . . . . . . .
3.1.
Pulp levels in cells . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.1.
Instrumentation used for pulp level measurement and control. . .
3.1.2.
Base level control systems for pulp level control . . . . . . . . .
3.2.
Air owrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1.
Instrumentation used for air owrate measurement and control . .
3.2.2.
Base level control systems for air owrate control . . . . . . . .
3.3.
Slurry owrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1.
Instrumentation used for slurry owrate measurement and control
3.3.2.
Implementation of slurry owrate in control systems . . . . . . .
3.4.
Elemental assaying . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.1.
Instrumentation used for elemental analysis . . . . . . . . . . .
3.4.2.
Implementation of elemental assaying in control systems . . . . .
3.5.
Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5.1.
Instrumentation used for density measurement. . . . . . . . . .
3.5.2.
Implementation of density measurement in control systems. . . .
3.6.
Reagent addition . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.6.1.
Instrumentation used for reagent addition . . . . . . . . . . . .
3.6.2.
Base level control systems for reagent addition . . . . . . . . . .
3.7.
Eh, pH and conductivity . . . . . . . . . . . . . . . . . . . . . . . . .
3.7.1.
Instrumentation for the measurement of Eh, pH and conductivity .
3.7.2.
Base level control systems for Eh, pH and conductivity control . . .

Corresponding author. Tel.: + 44 20 7594 7360; fax: + 44 20 7594 7403.


E-mail address: j.j.cillifers@imperial.ac.uk (J.J. Cilliers).
0301-7516/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.minpro.2011.05.002

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

58
59
59
59
59
60
60
60
60
60
60
60
61
61
61
61
61
61
62
62
62
62
62
62
62
62
62
62

58

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

3.8.

Gas dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.8.1.
Instrumentation used for the measurement of gas dispersion variables
3.8.2.
Implementation of gas dispersion variables in control systems . . . .
3.9.
Machine vision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.9.1.
Instrumentation and methods used for machine vision . . . . . . .
3.9.2.
Implementation of machine vision in control systems . . . . . . . .
4.
Advanced otation control . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.
Advanced control of mass pull and re-circulating load . . . . . . . . . . . .
4.1.1.
Mass pull control . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.2.
Re-circulating load control . . . . . . . . . . . . . . . . . . . . .
4.2.
Advanced control of grade and/or recovery . . . . . . . . . . . . . . . . .
4.2.1.
Model-based methods in advanced otation control . . . . . . . . .
4.2.2.
Expert systems in advanced otation control . . . . . . . . . . . .
5.
Optimising otation control . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.
Modelling-based methods in optimising otation control . . . . . . . . . . .
5.2.
Expert methods in optimising otation control . . . . . . . . . . . . . . .
6.
Examples of approaches found in advanced/optimising otation control . . . . . . .
7.
Commercial advanced/optimising otation control software . . . . . . . . . . . . .
8.
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Introduction
Froth otation is one of the most broadly used separation methods
in the mineral processing industry. However, despite being introduced in the early 1900s and numerous years of research and
development, otation is still not fully understood and remains
relatively inefcient. As such, large economic gains stand to be made
through optimisation of many present processes (McKee, 1991;
Hodouin et al., 2000; Moilanen and Remes, 2008).
It is important to realise from the outset that process control
consists of several interconnected levels. Several authors, e.g. Roesch
et al. (1976); McKee (1991); Laurila et al. (2002); and Gupta and Yan
(2006), describe the process control of froth otation by a hierarchy of
34 inter-connected layers. The hierarchy described by Laurila et al.
(2002) is presented in Fig. 1.
The lowest level is the instrumentation itself, which is the basis for
all process control. As such, the choice/design and maintenance of the
instrumentation is of central importance to any process control
system. Furthermore, the correct choice of instrumentation can only
be achieved if a detailed understanding of the functioning and
application of the required instrument, within a given process, is
acquired (Laurila et al., 2002).
Base level otation control is focused on maintaining primary
variables at setpoints. These primary variables include: pulp level, air
owrate and reagent addition rate. This is generally achieved through
the usage of conventional SISO PID control; although more advanced

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

63
63
63
63
63
64
64
64
64
64
65
65
66
66
67
67
68
68
68
68
68

methods are now commonly used in modern control strategies.


Similarly, traditional base level otation control was applied to single
cells, although modern control strategies are now regularly applied to
entire banks of cells (e.g. pulp level control).
The two higher tiers of otation control are advanced otation
control (AFC) and optimising otation control (OFC). AFC involves the
rejection of the effects from input disturbances to the process (e.g. a
change in feed grade) and maintaining performance parameters
grade and recovery (although care should be taken when dening
recovery; in a dynamic situation accumulation of material within the
system and lag times should be considered). OFC, on the other hand,
aims to maximise overall nancial protability (commonly by
maximising grade and recovery). Both AFC and OFC attempt to
achieve their objectives through manipulation of lower level control
setpoints. It thus follows that the efciency of AFC and OFC systems
are dependent on satisfactory lower level otation control systems
being in place. Several cells are generally controlled simultaneously;
and advanced control methods (which in control engineering
terminology classically refers to any control strategy more complicated than SISO PID control; and more recently to computer based
technologies) are used as PID control is insufcient.
Numerous years of research into the automation of froth otation
control has been conducted to increase process efciency, with efforts
prior to the 1970s being largely unsuccessful. Several authors, such as
McKee (1991) and Laurila et al. (2002), agree that reasons for this
include: a lack of appropriate instrumentation and technologies, and

Fig. 1. Process control system level hierarchy for otation processes (Laurila et al., 2002).

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

the old design of otation plants; which consisted of numerous


smaller cells making process control more complicated and cumbersome. As such, operator intervention was the only method of process
control (and this is still largely true of many plants today).
However, the early 1970s saw extensive improvements in
available instrumentation and the research into the development of
automatic control of otation began to show increasing promise
(McKee, 1991; Thwaites, 2007). Subsequently, there has been major
development in base level otation control; although the development of robust, long-term, automatic A/OFC systems has proven more
challenging. According to McKee (1991) this is partially due to the
inherent complexity and unpredictability of the response of most otation
circuits to upset conditions, unclear expectations of what a control
system can achieve, unrealistic objectives for control systems and
excessive complexity of the actual control strategies. Osorio et al.
(1999) included coupling among control loops, long varying lag times,
and an imperfect knowledge of the phenomenology of otation and the
lack of appropriate and precise instrumentation as contributors for
complicating control attempts.
Despite this, Laurila et al. (2002) believe that otation is currently
facing a new era in terms of automation and process control, with
there being three main reasons for this:
Henning et al. (1998), and Moilanen and Remes (2008) noted that
otation circuit design is moving away from multiple recycle streams
and towards simpler circuits. This simplies regulation and control
of the over-all process.
Authors such as Kallionen and Heiskanen (1993), and Carr et al.
(2009) have noted that the sizes of otation cells are increasing, the
main benets being: a reduction in capital expenditure and
operational costs, lower energy consumption per cubic metre, less
items to maintain and a lower plant footprint. Fewer cells also mean
less instrumentation is required; allowing for less-intricate process
control systems. However, this does mean more accurate instrumentation is required. Fewer, larger cells also increase the incentive
for better base level control on each individual cell.
Recent developments in instrumentation have seen the development of tools such as eldbus technology and image analysis, and
have allowed for the assembly of smart instruments; devices that
use self-diagnostics to provide information about the equipment
status and measurement quality.
Laurila et al. (2002) also highlight that as each otation process is
unique (e.g. cell conguration, instrumentation, ore, chemistry, etc.) a
large variety of A/OFC strategies have been developed and implemented, and a single, universal, control approach cannot be given.
The aim of this literature review is to broadly explore the various
control strategies that have been/are being, developed and utilised
from a metallurgist's perspective. This work shall begin with a look at
the key variables and considerations in the control of otation. Each of
the four levels of otation control will then be explored in detail.
Lastly, a description of some available commercial control systems
and conclusions are presented. It should also be noted that although
this paper focuses primarily on cell otation devices, as opposed to
column otation devices, much of the literature and theory is
applicable to both.

59

simple other simultaneous sub-processes also occur. Examples


include: entrainment of gangue into the froth phase, coalescence of
bubbles, de-attachment of valuable particles from bubbles as they
impact the froth phase, etc. (Ventura-Medina, 2000). The feed
composition and upstream grinding stages prior to otation also
signicantly affect the process. Thus, in reality, the otation process is
highly complex with Arbiter and Harris (1962) estimating that there
are approximately 100 variables that affect (to varying degrees) the
otation process. Moreover, co-interactions between variables further
complicate control efforts. For example, an increase in air owrate
may well result in a larger bubble size, which will subsequently affect
the bubble rise velocity, rate of attachment, gas holdup, froth depth,
etc. meaning other variables may be affected and need manipulating after a given response time. As such, it is the high number of
variables and the complex, non-linear, inter-relationships between
these variables that make otation control in specic very
challenging.
2.1. Key variables
Laurila et al. (2002) suggest that the following variables
specically from an A/OFC viewpoint are most important:

slurry properties (density, solids content)


slurry ow rate (retention time)
electrochemical parameters/potentials (pH, Eh, conductivity)
chemical reagents and their addition rate (frothers, collectors,
depressants, activators)
pulp levels in cells
air owrates into cells
froth properties (speed, bubble size distribution, froth stability)
particle properties (size distribution, shape, degree of mineral
liberation)
mineralogical composition of the ore
mineral concentrations in the feed, concentrate and tailings
(recovery, grade)
froth wash water rate (especially in otation columns)

Manipulating/measuring each of these variables simultaneously


may well be unnecessary to achieve a good process control result.
However, each of these variables and their effects on the otation
process should be considered.
2.2. Effects of plant layout and the location of the cell in the circuit
The process layout is a key consideration in process control.
Recycle streams in particular can make process control (e.g. pulp level
control) more challenging, especially if accumulation and/or sudden
discharges of material are allowed to occur. Additionally, different
sections of the plant require different process control regimes; the
rougher and scavenger sections being operated at comparatively low
froth depths and high air owrates, to achieve high mineral recovery;
whilst cleaner sections operate with greater froth depths and lower
air owrates in a bid to increase grade (Laurila et al., 2002).

2. Key variables and considerations in the control of otation


2.3. Types of process input disturbances
Froth otation is a three phase separation process for complex ores
based on the manipulation of the difference in hydrophobicity of the
solids. Suspended, hydrophobic metal-rich particles are contacted
with, and subsequently combine to, air bubbles whilst the more
hydrophilic gangue particles sink and are recovered to the tails
stream. The valuable-mineral loaded bubbles report to a froth phase
and overow into a launder, before being recovered to the
concentrate stream. Although the process may sound relatively

It is important to identify the frequency and severity with which


input disturbances can occur for a given otation process. Wills and
Napier-Munn (2006) suggest that if the grinding circuit control
systems are running efciently, variations in particle feed rate, pulp
density and particle size should be minimal with the otation
circuit being responsible for compensating for variations in mineralogy and oatability of the ore.

60

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

2.4. System constraints


Existing circuit constraints need to be considered when implementing a process control regime. Two types of constraints exist.
Firstly a process may be equipment constrained (e.g. a recovery
cannot be achieved because the required air owrate rate cannot be
supplied). Secondly, a process may be constrained by another part of
the system; an example being the limitation of the maximum
recovery of a desired species, at a high concentrate grade objective,
within a otation plant as a result of insufcient liberation by the
grinding circuit (McKee, 1991). Bergh and Yianatos (2011) add that
in practise the plant operating point that satises the overall economic
goals of the process will lie close to the intersection of constraints.
3. Instrumentation and base level otation control
Information about the input disturbances, process operating
parameters and nal product quality is required before optimisation
and control can be performed; with the quality of measured
information largely determining the efciency of an implemented
control system. However, despite the availability of instrumentation
for the measurement of important parameters such as: ore composition, owrates and less ore specic properties (e.g. pulp levels,
density, pH) essential properties such as liberation degree, surface
chemistry, bubble size distribution, bubble loading, etc. remain
difcult to measure and infer (Bergh and Yianatos, 2011).
Currently, most existing instrumentation on otation plants makes
use of analogue signal technology; with signals requiring conversion
into a digital format before interfacing with automation systems. As
such, it is thought that analogue technology is to be replaced by digital
eldbus technology. This will result in completely digital communication between instrumentation and base level control systems at
source, and has already enabled the decentralisation of pulp level and
air owrate base level control systems. This has also allowed for better
integration of base level control during otation cell design (Laurila
et al., 2002). Continued development of base level control is now
largely focused towards operational aspects that facilitate and speed up
the setting of control loops (Moilanen and Remes, 2008).
3.1. Pulp levels in cells
3.1.1. Instrumentation used for pulp level measurement and control
Laurila et al. (2002) and Carr et al. (2009) suggest that the most
typical methods of pulp level measurement are:
Float with a target plate and ultrasonic transmitter
Float with angle arms and capacitive angle transmitter
Reex radar
Other methods mentioned in literature include:
Hydrostatic pressure measurement. Authors such as Roesch et al.
(1976); Hamilton and Guy (2000); and Maldonado et al. (2008)
describe methods whereby the hydrostatic pressure is measured to
determine the pulp level. Accurate measurement requires that both
the slurry density and air holdup in the pulp be known.
Microwave radar and ultrasonic transmitters. Microwave or ultrasonic beams are emitted towards the froth and are reected at the
slurry surface before being measured, and the pulp level being
inferred (Hamilton and Guy, 2000).
Conductivity and capacitance Hamilton and Guy (2000) describe two
techniques whereby the large difference in dielectric constant
between gasses and liquids is used to determine the pulp level.
Similarly, the difference in electric conductivity, measured using a
conductivity probe, can also be used to locate the frothpulp
interface (Wills and Napier-Munn, 2006; Maldonado et al., 2008).

Laurila et al. (2002) add that accurate level measurement is often


problematic as the slurryfroth transition is not sharp and variations
in slurry density and/or very dense froth layers often exist. This
especially complicates methods using direct ultrasonic or hydrostatic
pressure measurements.
Lastly, Carr et al. (2009) highlight that control valve options are
limited due to the eroding conditions and large variations in owrate.
Dart and pinch valves are generally used (neither being optimal), and
both being occasionally incorrectly sized. The optimal operating range
of these valves is 3060% open, although these valves are commonly
seen operating at below 30% open. Operating a valve near the fully
closed position causes increased wear, while operating near the fully
open position reduces the control range of the valve.
3.1.2. Base level control systems for pulp level control
Stenlund and Medvedev (2000); Kmpjrvi and Jms-Jounela
(2003); and Carr et al. (2009) report that otation cells traditionally use
feedback PI control to ensure pulp levels remain at desired set-points;
although Wills and Napier-Munn (2006) add that feedforward control is
also regularly integrated to account for owrate variations upstream. This
is achieved by manipulating the tailings ow from the cell by adjustment
of the slurry outlet valve. This technique is effective for the control of
isolated cells, but is also commonly used to control a bank of cells in series.
This approach is problematic, as each individual cell control loop attempts
to independently maintain the pulp level at the set-point. As such, a
control action for one cell is a disturbance for the next; the net result
being that each cell drives the following cell off its set-point.
As such, sophisticated multivariable model-based control methods
have been developed; whereby the whole bank of cells is modelled
and compensations between adjacent cells calculated and/or considered. Two examples of multivariable control methods are presented in
Stenlund and Medvedev (2000). The rst makes use of a decoupling
controller model, where compensator parameters are introduced
(representing the dependencies of each cell in a series on prior cells)
to counteract the interactions between cells. The second method
introduced is a multivariable model-based feedback controller, which
manipulates ows out of each cell simultaneously so as to continually
minimise a dened error function. Kmpjrvi and Jms-Jounela
(2003) described an alternative multivariable model whereby a feedforward controller was linked with traditional PI controls on each cell.
Moilanen and Remes (2008) described a similar feed forward control
algorithm. Despite this, Carr et al. (2009) reports that these more
sophisticated methods are rarely used in industrial processing plants.
Commercial control packages such as FloatStar Level Stabiliser by
Mintek, or CellStation by Outotec, also aim to control all cell levels
simultaneously using advanced control techniques. Both have
reportedly been trialled successfully and installed on industrial plants
(www.mintek.co.za; www.outotec.com).
From an A/OFC perspective pulp level control is important as it dictates
the froth depth (dened as the distance from the pulp/froth interface to
the overow lip). Theoretically, a deeper froth allows for increased
drainage of mechanically entrained gangue, and subsequently, a higher
concentrate stream grade (Wills and Napier-Munn, 2006). As such, A/OFC
systems commonly manipulate the setpoints of pulp-level controllers.
3.2. Air owrates
3.2.1. Instrumentation used for air owrate measurement and control
Laurila et al. (2002) report that there are three common methods
of measuring air owrates in otation processes. Two of the methods
make use of differential pressure metres, which are popular in
industry, including otation, due to their low price, simple principle and
fairly low requirement of maintenance. The three methods are:
Thermal gas mass ow sensor. The cooling effect of the air as it ows
past a sensor is measured and correlated to an air owrate. These

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

instruments are unobtrusive to the air ow, although they are


expensive and factory calibrated, making changes difcult.
Differential pressure metre with venturi tube. Green and Perry (2007)
describe that the narrowed venturi tube acts as a restriction to the
ow, resulting in a pressure drop that is measured and a ow rate
determined. This method is reliable, produces tolerable pressure
drops and is accurate, although the venturi tube has large space
requirements and is expensive Laurila et al. (2002) add that an
orice plate is not a suitable solution due to the signicant pressure loss
it causes.
Differential pressure transmitter with Pitot tube or annubar tube. Both
the Pitot tube and annubar element determine the gas owrate, in a
pipe, by comparison of the internal pipe pressure and the static gas
pressure. The difference between the methods is that the Pitot tube
only has one measurement point, while the annubar element has
several measurement points and thus provides an average air
velocity. Both methods are deemed accurate and the observed
pressure drop is small. Fig. 2 illustrates the instrumentation used for
air owrate measurement.
Problems associated with differential pressure metres include
large space requirements, with large sections of straight piping being
needed to ensure a fully developed ow prole. One solution is to
decrease the pipe size, as the required straight pipe length is related to
pipe diameter. Buttery valves are used to control the air ow as they
are cheap and sufcient for the task (Laurila et al., 2002).
3.2.2. Base level control systems for air owrate control
Wills and Napier-Munn (2006) elaborate just how important and
useful aeration control is; adding that otation generally responds
faster to changes in aeration, than to changes in froth depth, and
because of this aeration is often a more effective control variable.
Similarly, they infer (by use of an example) that in comparison to
reagent addition, air is by far the cheaper reagent and leaves no
residual concentration if used in excess. As such, air owrate is
frequently incorporated into A/OFC systems, often in conjunction
with pulp level and/or reagent addition control systems; an example
being the simultaneous manipulation of air-ow rate and pulp level
to control mass pull.
Carr et al. (2009) note that the control of otation aeration is easier
than slurry level control. Luyben and Luyben (1997), Laurila et al.
(2002), and Carr et al. (2009) agree that a simple, well-tuned,
feedback/feedforward PI/PID control loop is adequate to accurately
regulate air ow by manipulation of the control valve setting. Sizing of
the control valve is of central importance for effective control.
Oversized valves are often tted to infer a smaller pressure drop,
but in reality result in a limited control response (Luyben and Luyben,
1997); and can signicantly, and rapidly, affect otation performance
and pulp level control. Smith et al. (2008), suggest that down-the-

61

bank air owrate prole control is also advantageous. If the cell air
owrates are not controlled individually, and air is fed to a group of
cells, buttery valves are often operated manually to adjust the ow
to each cell. Lastly, Laurila et al. (2002) add that otation cells with
self-aspirating aeration mechanisms often do not have automatic airow
control. The available range of airow control is anyhow limited. This
problem is pronounced at high altitude. This limits the potential for the
implementation of advantageous A/OFC strategies.
3.3. Slurry owrates
3.3.1. Instrumentation used for slurry owrate measurement and control
Magnetic ow metres are commonly used to measure slurry
owrates and are based on Faraday's principle of induction, with the
device consisting of an electromagnet coiled around an insulated
length of pipe. Electrodes are installed at opposite sides of the pipe,
which enable an electric current to be generated through the owing
uid and measuring device. From this measured current a owrate
can be determined. This method is non-obtrusive and modern
magnetic owmeters take up to 30 measurements per minute. Slurry
measurement is problematic, however, as solids and air bubbles
decrease performance. Moreover, if magnetic solids (e.g. magnetite)
are present de-magnetisation is required (Laurila et al., 2002). Slurry
owrates can also be controlled by variable or xed speed pumps that
are capable of handling slurries. Suitable valve options have already
been discussed in Section 3.1.1.
3.3.2. Implementation of slurry owrate in control systems
In otation circuits, slurry owrates are generally manipulated to
control pulp levels in cells (see Section 3.1.2), and are not controlled
to a setpoint. Nevertheless, the measurement of slurry owrates are
important, as Laurila et al. (2002) point out, for A/OFC, and allow for
calculating re-circulating loads and performing mass balance calculations. Slurry owrate measurement is also important for reagent
addition base level otation control.
3.4. Elemental assaying
3.4.1. Instrumentation used for elemental analysis
On-line X-ray uorescence (XRF) analysers provide elemental
assays from process ow streams and are now considered standard
hardware on large scale otation plants (Garrido et al., 2008). Several
points of the process can be sampled, with some modern XRF
analysers handling up to 24 streams and most machines being capable
of analysing for several elements and solids content. The time to
analyse a single sample can range from 15 s to a minute, and the
sampling cycle time is between 10 and 20 min depending on the
number of sample points attached to the analyser (Laurila et al., 2002;

Fig. 2. Illustration of (A) thermal gas mass owmeter, (B) venturi tube with differential pressure metre and (C) Pitot tube.

62

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

Bergh and Yianatos, 2011). Accuracy ranges from 1 to 6% and


detection limits are as low as 330 ppm (Moilanen and Remes,
2008). Moreover, Haavisto et al. (2008) introduce a new method of
analysis; the visual and near-infrared reectance spectroscopic analysis
of process slurries. This measurement aims to be a supplementary
method, which complements the on-line assay information available
from an XRF analyser. It is further reported that spectral information
can be used to accurately predict element contents in the slurry in
between successive XRF analyses, and as these measurements can be
taken with high frequency as opposed to sparse XRF analysis, a practically
continuous on-line estimate of the slurry contents is reached. This
would allow for any process disruptions to be rapidly identied.
Despite the obvious benets of online XRF analysis Garrido et al.
(2008) report that these analysers are generally under-utilised
because operators do not trust the online information given by estimation
models. They further describe a calibration method for estimation
models that minimise the effects of the uncontrollable disturbances
during the estimation.
3.4.2. Implementation of elemental assaying in control systems
Wills and Napier-Munn (2006) report that the key to effective
(otation) control is online chemical analysis, which produces real-time
analysis of the metal composition of process streams. However, the
relatively long sampling cycle times mean that input disturbances of
high frequency can be missed, making it difcult to capture the
experimental data required to form dynamic models of the process
(Bergh and Yianatos, 2011); and take required control actions. Moreover,
Remes et al. (2007) conducted a study into the effect of speed and
accuracy of on-line elemental analysis on otation control performance.
The study concluded that an increase in sampling cycle time results in a
strong decline in controllability, resulting in negative economic impacts.
3.5. Density
3.5.1. Instrumentation used for density measurement
Roesch et al. (1976) report that nuclear density metres are
commonly used for density measurement on otation plants.
Gamma radiation is emitted by a radioactive isotope and the
attenuation of the radiation by the slurry is measured, from which
the density can be determined from a prior calibration. This method is
non-obtrusive to the process ow. This is in agreement with Laurila
et al. (2002), who add that air bubbles in the slurry often make use of a
nuclear density metre impossible and choice of location in the process
is an important consideration. Additionally, some on-stream XRF
analysers are now capable of density measurement.
3.5.2. Implementation of density measurement in control systems
Density measurements are used in mass balance calculations,
which are associated with A/OFC (Laurila et al., 2002). Furthermore,
control/manipulation is normally preformed in the grinding circuit/s
(Wills and Napier-Munn, 2006).
3.6. Reagent addition
3.6.1. Instrumentation used for reagent addition
A variety of alternative equipment for maintaining/setting reagent
addition rates are used industrially. Laurila et al. (2002) suggest the
main reasons for this are due to the seemingly negligible amounts that
need to be added (often measured in millimetres per minute) and the
large variety of different reagents added, each with their own
chemical properties and attributes. Common methods include:
A simple on-off type dosing system which periodically opens a valve
and allows reagent to enter the process. Regular checks are required
to ensure the correct amount is added, as this method can be very
inaccurate.

Metering pumps are also used, especially if volumes to be added are


very small or cost is of importance. This method is more accurate,
but the pumps are costly and require regular maintenance.

3.6.2. Base level control systems for reagent addition


Base level reagent addition control commonly consists of a
feedforward ratio-type control, as mentioned by Hodouin et al.
(2000), whereby the reagent addition rate is altered according to the
feed rate of ore to achieve a reagent (in slurry) concentration setpoint
(e.g. grammes of reagent per ton of ore). This reagent concentration
setpoint may then be altered retrospectively by operators in a
feedback fashion after considering composition results from an XRF
analyser. Similarly, the setpoint could be manipulated by an A/OFC
system to achieve a required metallurgical result; although Wills and
Napier-Munn (2006) add that this is more common with collector
control, with frother addition rate setpoints usually being set
manually.

3.7. Eh, pH and conductivity


3.7.1. Instrumentation for the measurement of Eh, pH and conductivity
Measurement of electrochemical potential (Eh), pH and conductivity provide information on the surface chemistry of the particles in
the slurry and are the only direct, non-intrusive methods of
determining what is occurring chemically within the otation cell
(Ruonala, 1995). Laurila et al. (2002) report that pH is a commonly
measured electrochemical property, and is related logarithmically to
the hydronium ion activity in solution. Measurement is achieved by
using ion selective electrodes, although this is often problematic as
the electrodes are easily contaminated by active substances in the
slurry. As such, sampling systems are often used for pH measurement, where washing of the electrodes and regular maintenance can
be performed.
Conductivity measurement can often be used instead of, or in
conjunction with, pH measurement as both provide complimentary/
similar information. Conductivity metres are generally cheaper and
are more suitable for highly alkaline solutions, although their use
should be avoided in highly aerated systems. In addition, Wang and
Cilliers (1999), and Bennett et al. (2002) performed studies where
conductivity measurements were used to determine froth density and
ow regimes.
Eh measurement too is problematic, with Woods (2003) adding
that maintaining electrode probes so that they respond appropriately in
these plants is difcult to achieve, and concluding that the choice of
materials in Eh probes requires further research.

3.7.2. Base level control systems for Eh, pH and conductivity control
Base level control of pH involves maintaining the slurry pH at a
desired setpoint, through manipulation of acid or lime addition rates,
with PID control loops being adequate for the task (Laurila et al.,
2002). Long response times of the system mean a lag time should
follow any corrective action/s performed to fully appreciate the effects
it has had (Wills and Napier-Munn, 2006). Commercial controllers,
such as FloatStar pH Controller, are also available and use advanced
control methods.
Control of Eh usually involves the addition of nitrogen or air to the
system to alter the electrochemical potential (Woods, 2003). Much
research into improving metallurgical results through Eh manipulation (A/OFC) has been published over the last four decades. Ruonala
et al. (1997); and Woods (2003) present thorough reviews of this
work on both laboratory and industrial scales, for several ore types. A
summary of a brief literature review, listing the ore-type the studies
were conducted on, is listed below in Table 1.

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771
Table 1
Summary of literature review for Eh measurement/control in froth otation.
Author and date

Mineral type

Berglund (1991)
Bruckard et al. (2007)

Pyrite, sphalerite-galena, chalcopyrite


Arsenopyrite, lollingnite and arsenic from
tin bearing ores
Chander and Fuerstenau (1975) Copper and chalcocite
Clark et al. (2000)
Chalcocite, chalcopyrite and bornite
Guo and Yen (2005)
Enargite and chalcopyrite
Hayes and Ralston (1988)
Galena, chalcopyrite and sphalerite
Hicyilmaz et al. (2004)
Pyrite
Hintikka and Leppinen (1995)
Complex sulphide ores and gold bearing ores
Kirjavainen et al. (1992)
Copper ores with rich copperzinclead pyrite ores
Kocaba and Gler (2007)
Pyrite, galena, chalcopyrite
Leppinen et al. (1997)
Copper and zinc rich complex ores
Qing et al. (2008)
Leadsilverzinc complex ores
Roos et al. (1990)
Chalcocite and covellite
Roos et al. (1990)
Copper and chalcopyrite
Shen et al. (1997)
Sphalerite and pyrite
Uribe-Salas et al. (2000)
Galena, chalcopyrite and pyritic ore
Walker et al. (1984)
Chalcocite
Yuan et al. (1996)
Complex copper/zinc sulphide ore with
pyrite and pyrrhotite

3.8. Gas dispersion


3.8.1. Instrumentation used for the measurement of gas dispersion
variables
Gomez and Finch (2007) report that gas dispersion is the collective
term for supercial gas (air) velocity (volumetric air owrate per unit
cross sectional area of cell, Jg), gas holdup (volumetric fraction of gas in a
gas-slurry mix, g) and bubble size distribution (Db). The publication
also details the equipment used for the measurement of these
variables.
3.8.1.1. Supercial gas velocity measurement sensor. The sensor consists
of a vertically positioned tube; the lower end is partially submerged in
the pulp zone to collect bubbles (see Fig. 3, part A). The continuous
version has an orice valve mounted on the air outlet. When air is
allowed out of the orice valve, time is given for a pressure steadystate to be reached and measured (i.e. the rate of air into the column is
equal to that out the orice valve). The volumetric air owrate is
then inferred from a previous calibration and Jg calculated. Current

63

problems with the design include the requirement of a range of orice


valves to suit all gas velocities and the build up of froth within the
system (Torrealba-Vargas and Finch, 2006; Gomez and Finch, 2007).
3.8.1.2. Gas holdup measurement sensor. The sensor is based on
Maxwell's model that relates the concentration of a non-conducting
dispersed phase to the conductivities of the continuous phase and the
dispersion. This requires two measurements in separate vessels (see
Fig. 3, part B); one vessel measures the conductivity of the aerated pulp;
the other measures the conductivity of the air free pulp (achieved using
a syphon). The ratio of these measurements is used to solve Maxwell's
model and estimate the gas holdup. This method ensures continuous
measurement; although care is required when choosing the opening
sizes of the syphon to ensure no bubbles enter the vessel and result in
inaccuracies (Gomez et al., 2003; Gomez and Finch, 2007).
3.8.1.3. Bubble size measurement sensor. The McGill bubble sizing
device (see Fig. 3, part C) is able to measure the full bubble size
distribution found in the pulp phase. A sample is drawn from the pulp
via a tube and directed into a sloped viewing chamber (the sloped
window allows for a near mono-layer of bubbles to form) exposed to a
pre-set light source. The continuous ow of bubbles is then captured
via image analysis. The accuracy of the measurement is difcult to
establish, although the method is widely used and continues to evolve
by improvements dictated by eld trials (Gomez and Finch, 2007).
3.8.2. Implementation of gas dispersion variables in control systems
A recent publication by Bartolacci et al. (2008) focused on using
gas dispersion sensors in conjunction with a Froth Stability Column
and machine vision (both still to be discussed). Air owrate, pulp
level, reagent dosage and feed rate were varied whilst dispersion
parameters, froth stability, bubble surface area ux and metallurgical
results were measured. The results indicated a high dependence of
grade and recovery on dispersion parameters and froth stability; a
clear indication of the potential of these measurements being
integrated into an A/OFC system.
3.9. Machine vision
3.9.1. Instrumentation and methods used for machine vision
Machine vision makes use of cameras positioned above otation
cells to record digital images of the froth surface. Several froth features
can be extracted from these images and used for control purposes.

Fig. 3. Schematics of (A) gas velocity, (B) gas holdup and (C) bubble size measurement devices (Gomez and Finch, 2007).

64

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

These features are categorised into three types; namely: physical,


statistical and dynamic properties. Within each category several
methods exist for extracting different variables and/or features. For
example bubble size, a physical property of a froth, is one feature that
may be extracted through usage of bubble edge or watershed
algorithms. An extensive amount of literature has been written
about the methods by which features are extracted from froth; with a
full, recent, literature review being carried out by Aldrich et al. (2010)
(see Table 2).
Aldrich et al. (2010) highlight that the use of physical features
for control purposes remains problematic. Issues associated with
physical feature extraction include:
It is commonly observed that the surface bubbles of owing froths
are signicantly larger than those in the layers immediately below;
the lower layers forming the predominant portion of the volume
overowing into the launder. This cannot be corrected for readily.
Neethling et al. (2003) observed that the surface lm size
distribution is not necessarily representative of the bubble size
distribution in the underlying froth layer Wang and Neethling
(2009), however, have determined a method to relate surface lm
size to underlying bubble size distribution.
The watershed method often over-segments larger bubbles and
under-segments smaller bubble sizes.
Mobility feature extraction has been shown to be particularly
useful within the context of A/OFC.
3.9.2. Implementation of machine vision in control systems
It is common in industry for operators to control and regulate a
otation plant by visual inspection of the froth surface. Indeed, an
understanding of processes occurring within the froth phase is central
to understanding the overall behaviour of otation systems (Glembotskii,
1972; Cutting et al., 1986; McKee, 1991; Mathe et al., 1998). The structure
of froths has a signicant effect on grade and recovery; with several
publications (e.g. Supomo et al., 2008; Moilanen and Remes, 2008)
suggesting a link between froth velocity and metallurgical performance.
As such, the development and implementation of machine vision was an
attempt to rene and automate the control based on the froth appearance.
Currently, numerous commercial machine vision systems are available;
making the on-line measurement of froth velocity possible. This is useful
in the implementation of A/OFC, an example being the measurement and
control of mass pull. Moreover, the faster dynamics of machine vision
(1 min) compared to XRF technology (1020 min), open the possibility

for building better predictive otation A/OFC models, based on visual


aspects of the froth as opposed to stream grades (Bergh and Yianatos,
2011).
4. Advanced otation control
AFC, also known as stabilising control, aims to reject the effects of
input disturbances (e.g. a change of ore type) and maintain the
otation process as close to steady state as possible (McKee, 1991;
Laurila et al., 2002). This is generally achieved by controlling mass
pull; recycle load; stream grade and/or recovery to setpoints (usually
set by operators or an optimising controller), with the manipulated
variables being the setpoints of base level controllers (i.e. pulp level,
air owrate, reagent addition rate and pH/conductivity). As such,
effective AFC is only possible if robust, efcient base level otation
control systems are in place.
4.1. Advanced control of mass pull and re-circulating load
An effective method of keeping the plant's mass balance at steady
state is to maintain constant mass pulls and/or re-circulating loads at
setpoints (often in conjunction with maintaining grade and/or recovery).
4.1.1. Mass pull control
Mass pull refers to the amount of concentrate collected. A recent
paper by Supomo et al. (2008) describes the successful installation
and operation of VisioFroth, a commercial machine vision based
control system developed by Metso, at PT Freeport in Indonesia. The
control system measures the velocity of the overowing froth, and
then adjusts the froth-depth to achieve the desired mass pull. An
alternative commercial package, FloatStar Flow Optimiser by Mintek,
uses density and owrate measurements to calculate mass pull. The
required mass pull is then obtained through manipulation of pulp
level and air owrate (www.mintek.co.za).
4.1.2. Re-circulating load control
Re-circulating material through a otation plant results in decreased
residence time of material per otation cell, but does allow for the
valuable material to spend more time in the otation plant as a whole;
and, subsequently, increases the overall recovery (Wills and NapierMunn, 2006). However, this re-circulating load may need to be varied to
account for input disturbances (McKee, 1991). FloatStar Flow Optimiser
is an example of a commercial control package which uses the

Table 2
Overview of methods used for feature extraction by machine vision (Aldrich et al., 2010).
Type

Methods

Froth variables or features

References

Physical

Edge

Bubble size and shape

Watershed

Bubble size and shape

Spectral

RGB

Banford et al. (1998); Forbes and De Jager (2004b); Forbes et al. (2006); Lin et al.
(2007a,b); Wang et al. (2003); Wang and Stephansson (1999)
Forbes and De Jager (2004a); Sadr-Kazemi and Cilliers (1997); Ventura-Medina and
Cilliers (2000); Yang et al. (2008)
Bonifazi et al. (2005a,b); Gebhardt et al. (1993); Hargrave et al. (1996, 1998); Hargrave
and Hall (1997); Morar et al. (2005); Oestreich et al. (1995); Siren (1999);
Vathavooran et al. (2006)
Moolman et al. (1994)
Bartolacci et al. (2006); Liu and MacGregor (2007, 2008)
Hargrave and Hall (1997); Holtham and Nguyen (2002)
Aldrich et al. (1995, 1997); Bezuidenhout et al. (1997); Moolman et al. (1994, 1995a,b)
Bonifazi et al. (2000); Hargrave and Hall (1997); Hargrave et al. (1998)
Bartolacci et al. (2006); Liu et al., (2005); Liu and MacGregor (2007, 2008)
Estrada-Ruiz and Perez-Garibay (2009); Hytyniemi and Ylinen (2000); Jeanmeure
and Zimmerman (1998); Kaartinen and Hytyniemi (2005); Moolman et al. (1995c);
Niemi et al. (1997)
Botha (1999)
Barbian et al. (2007); Forbes and de Jager (2007); Holtham and Nguyen (2002);
Moolman et al. (1994); Supomo et al. (2008)
Nguyen and Holtham (1997)
Barbian et al. (2003, 2005, 2006); Moolman (1995); Morar et al. (2006)

Statistical

Co-occurrence matrices
Fractals
Latent variables

Dynamic

Mobility

FFT
Wavelets
Texture
Spatial and neighbouring grey level
Fractal descriptor
PCA
Neural networks

Bubble tracking
Block matching
Pixel tracing

Stability

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

measurement and manipulation of mass pull rates to regulate the


amount of material recycled (www.mintek.co.za).
4.2. Advanced control of grade and/or recovery
AFC commonly refers to strategies that aim to maintain grade
and/or recovery (Laurila et al., 2002). However, McKee (1991) states
that A [stabilising] control system which is capable of rstly stabilising
circuit performance, and then driving the circuit to a desired grade or
recovery operating point, would undoubtedly be considered highly
successful. highlighting the importance of the stabilising aspect of
AFC.
The review of the successfulness of alternative AFC strategies is
made difcult, as follow-up reports of implemented control strategies
are uncommon, as noted by McKee (1991); who suggested that this
was because most control systems do not remain in operation for
periods of years after installation and are shut down. Furthermore,
Wills and Napier-Munn (2006) add that few (if any) plants can claim
to have fully automated control systems that can run the plant unsupervised for long periods of time. Thus it seems that despite the
large body of literature devoted to the subject, the successful
application of AFC (and indeed OFC) techniques into industrial
application have been largely unsuccessful. Reasons given for this in
literature include:
Design of control systems are insufciently thought out during plant
design stages (Narraway et al., 1991; Thwaites, 2007; Bergh and
Yianatos, 2011); and major control variables are often only
identied once the plant is operational (Wills and Napier-Munn,
2006). This commonly results in poorly designed, incompetent,
control systems being installed on new plants; with later (often
inadequate) modications being required; or additional ad-hoc
control systems being added on as after-thoughts.
A vast knowledge of control systems and jargon is required to
develop, install and maintain A/OFC systems; with the majority of
operators/metallurgists/management not having a background in
control engineering (Hodouin et al., 2001; Thwaites, 2007; Wills
and Napier-Munn, 2006).
As already discussed, the non-linear, complex behaviour of otation
systems complicates modelling attempts; making the design of
robust, effective controllers that can deal with large ranges of
operating conditions difcult.
It is partially for these reasons (specically the latter two) that
Wills and Napier-Munn (2006) argue that the best control systems
are those that interact with the operator (i.e. supervisory control
systems), giving explanations, when alterations to setpoints/variables
of base level otation control systems are required; and as such,
experienced, conscientious operators currently remain a competitive
alternative towards any automated control system.
According to Gupta and Yan (2006), there are broadly two types of
AFC used in mineral processing (a hybrid between the two is also
possible). These are:
Use of model-based methods
Use of expert control systems
The evolution and current trends of each of these branches will
now be discussed.
4.2.1. Model-based methods in advanced otation control
Model-based methods can be further sub-categorised into two
categories, namely: empirical and phenomenological modelling (Polat
and Chander, 2000).
Empirical models make use of statistical methods to relate
measured input and output plant data, such that multivariable models
relating between two or more independent and dependant variables
can be established and used for predictive control (e.g. controlling

65

collector addition rate based on concentrate grade). Furthermore,


continual analysis of plant data and corrective adjustment of the
predictive (model-based) controller make it adaptive to changing
conditions (i.e. adaptive control). Adaptive control is especially
important within the context of otation control, which is prone to
non-linear, complex behaviour. As such, many predictive otation
control systems often (but not always) include adaptive control
aspects. By 1991, Thornton noted that although the amount of
literature devoted to multi-variable model-based control was extensive, the number of applications in industry was still comparatively
small; with McKee (1991) noting that 5 control strategies were
particularly common at the time (and largely remain so). These are:
Feed back control of collector addition to maintain recovery set points.
Feed forward control of collector based on the calculated metal content
of the new feed.
Maintaining concentrate ows within limits, usually by varying
aeration rates or pulp levels.
Maintaining circulating loads within limits, again by variation of
aeration rates or pulp levels.
Controlling aeration rates or pulp level to obtain concentrate grade set
points.
Despite the apparent popularity of adaptive multivariable modelbased controllers, Desbiens et al. (1994), and Gupta and Yan (2006)
note issues with the stability of adaptive control; with the controllers
becoming saturated, and un-adaptive, after a period of time.
Moreover, in spite of the majority of the predictive multivariable
control strategies being based on empirical correlations, research and
development of phenomenological models whereby relationships
between cause and effect are devised through an understanding of the
physics of the otation process for use in predictive controllers was
also conducted. As such, phenomenological modelling methods can
broadly be classied into 3 groups; namely: kinetics, population
balance and probabilistic based modelling (Polat and Chander, 2000).
The success of phenomenological modelling, within an AFC context, is
debatable.
The use of rst-order otation kinetics modelling is thoroughly
covered by Polat and Chander (2000), and has undoubtedly received
the most attention in literature. Kinetics modelling is based on
the assumptions that the rate of the particlebubble collision process is
rst-order with respect to the number of particles and that the
bubble concentration remains constant. Numerous batch experiments
(Imaizumi and Inoue, 1963; Tomlinson and Flemming, 1963; Harris
and Chakravarti, 1970; Jameson et al., 1977; Dowling et al., 1985;
Rastogi and Aplan, 1985) and continuous otation tests (Jowett and
Safvi, 1960) support the use of the rst-order rate equation. This has
allowed a otation cell to be modelled using the chemical reactor
analogy; whereby the removal of solids from the pulp phase is dened
by a rst order rate equation. This means a bank of cells can be
approximated by perfectly mixed CSTR's in series (Gaudin, 1957;
Niemi and Paakkinen, 1969; Atkins et al., 1986; Yianatos and
Hendrquez, 2006). Subsequently, efforts have been made to
accurately determine the so called over-all otation rate constant
(or k). This is not trivial, as k is dependent on particle size, degree of
liberation, air owrate, agitation, etc. To account for this, various
continuous distribution functions of k have been devised (Polat and
Chander, 2000); although previous reviews by Dowling et al. (1985)
conclude that no single distribution model could sufciently represent k; and Roesch et al. (1976), suggest that attempts based on the
rst order kinetics assumption are approximative and hide secondary
details.
Alternatively, a population balance model is presented by Bascar
(1982), Bascar and Herbst (1982), and Bascur (2000). The three phase
model represents each mineralogical species and particle size, with
each particle species being classed according to state in the slurry (i.e.
free in pulp, attached to bubble in pulp, free in the froth, attached to

66

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

bubble in froth). Kinetic equations relate transfer of particles between


slurry states and hydrodynamic considerations (e.g. power dissipation into pulp, gas holdup in pulp, etc.) are incorporated into the
population balance. This makes it possible to simulate the effects of
manipulated variables such as air owrate, pulp level, agitation, etc.
on a otation process.
Despite much research into both empirical and phenomenological
based modelling, authors such as Bergh and Yianatos (2011) and
McKee (1991) highlight several issues that still exist; the latter
concluding that Multivariate predictive control is ideally the solution
for high quality control. However, to be applicable without losing its
benets, good measurements, acceptable regulatory control of local
objectives (i.e. base level controls), reliable dynamic models, explicitly
stated process constraints and new methods to promote robustness are
needed. Flotation processes have weaknesses in most of those aspects. It
is for reasons such as these that expert control systems (where
decision making by operators is automated by use of articial
intelligence) are used.
4.2.2. Expert systems in advanced otation control
The potential of expert systems in the mineral industry was
recognised as early as 1983 (Bearman and Milne, 1992); and were
introduced into A/OFC as modelling of otation systems is difcult,
they are suitable for the handling of non-linear systems, and they
automate (and standardise) decision making by operators. Although
several methods of Articial Intelligence (AI) based control systems
exist, three important techniques include:
Articial neural networks (ANN) data-driven computing devices,
comprising of a large number of neurons, inter-connected by a predetermined network of synapses on a large scale. These neurons are
arranged in several layers and adjustable numerical weights are
associated with the connecting synapse network (see Fig. 3). Neural
networks are trained by iteratively updating the associated weight
matrix, such that a set of outputs can be predicted for a given set of
inputs. As such, an ANN based controller can be taught how to
manipulate base level control system's setpoints to maintain a given
metallurgical objective; an example being the changing of collector
addition rate to maintain recovery (Aldrich et al., 1997; Gupta and
Yan, 2006) (Fig. 4).
Inductive machine learning makes use of mathematical models to
generate rules and form induced decision trees, and is based on the
concept of information entropy. A set of samples, each with a
corresponding vector of classifying attributes (e.g. bubble size, froth
velocity, etc.) is assessed. The vector is then split according to the
most informative attribute (each split forming a branch of the
decision tree); with each newly formed subset subsequently being
re-split (according to a different attribute) until each subset consists

of examples of a specic class. As such, a given system (e.g. a froth


surface) can be categorised and appropriate control action taken if
required (Aldrich et al., 1997; Filipic and Junkar, 2000).
Fuzzy logic reasoning that serves to be approximate rather than
precise. Compared with binary logic, where a result is false or true
(or quantitatively 0 or 1), fuzzy logic caters for a degree of truth
(anywhere between 0 and 1). Flotation system parameters (e.g.
pulp level) are divided into fuzzy sets (according to dened
membership functions), which can subsequently be combined to
form fuzzy subsets. A response to a fuzzy set/subset is then applied,
using IFTHEN rule based strategies (Gupta and Yan, 2006). For
example, if froth velocity and air rate are the variables being
arranged into fuzzy subsets, the control logic might be something
like: IF froth velocity is low AND air rate is high, THEN decrease froth
depth; ELSE IF froth velocity is low AND air rate is low/medium,
THEN increase air rate; ELSE do nothing.
One area in which AI has been used extensively is the identication
and categorisation of froth images from machine vision. Aldrich et al.
(1997) trialled and compared inductive learning techniques with a
back-propagation neural net method in industry; with all methods
being found to be equally capable of classication of various froth
features. Similarly, Cipriano et al. (1998) used rule-based expert
control, combined with machine vision, to control rougher cells. The
supervisory controller was able to identify froth characteristics and
subsequently suggest actions to be taken to the operators. More
recently, Supomo et al. (2008) reported on the successful use of the
commercial control software VisioFroth on PT Freeport, in Indonesia.
The system combines expert control with machine vision to control
mass pull, and has reportedly resulted in increased recovery.
Similarly, PlantVision by KnowledgeScape also makes use of expert
control.
In a comparison between expert control and multivariable modelbased control methods, Zavala et al. (1995) compared a supervisory
system of multiple SISO PID regulators, a multivariable model-based
predictive controller and an expert rule-based controller on a
simulated otation system. The ndings included that: the expert
system readily became saturated, the PID controllers were difcult to
tune, while the model-based controller required a linear model of the
system; the suitability of such a model to handle all disturbance types/
magnitudes being unlikely. Tighter control was obtained with the
model-based controllers. Subsequent studies by Prez-correa et al.
(1998) and Osorio et al. (1999) altered the model-based controllers to
become more exible to varying conditions (although a high degree of
on-line mathematical manipulation was required); and altered the
expert controllers to avoid control saturation while achieving high
recoveries despite severe simulated input disturbances.
The combination of AI and model-based methods is also possible.
Cubillos and Lima (1997, 1998) noted that using ANN to model
otation systems is problematic (due to the many associated degrees
of freedom and the heavy computational requirements). However,
such issues are averted, and the full advantages of ANN realised, by
implementing neural network systems into predictive model-based
control systems, such that the AI system is responsible for updating/
modifying constants within the model with time; thus allowing the
hybrid controller to remain adaptive. Testing of this hybrid control
strategy offered promising results, with the controller reportedly
being robust and exible.
5. Optimising otation control

Fig. 4. Simplied diagram of a neural network (Gupta and Yan, 2006).

The upper-most tier of process control in otation is OFC, which by


denition aims to maximise the nancial feasibility of the process.
This is achieved by determining where on the theoretical grade
recovery curve is most protable to operate and, subsequently, shifting
the operating point orthogonally to further maximise prot (see Fig. 5).

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

67

varied to four cells in a rougher bank, and the air rate that yielded the
PAR identied in each of the cells. Moreover it was shown that
operating the cells at their PAR air rates resulted in a higher mineral
recovery being obtained.
The link between operating cells at PAR air rates and improved
otation performance was also shown in studies presented by Smith
et al. (2010) and Hadler et al. (2010), where it was shown that
optimising air recovery in a bank of cells resulted in either a higher
concentrate grade, a higher mineral recovery, or in many cases both.
This has important implications for control as air recovery is a single
quantitative variable that can be measured and maximised; and the
process optimised in terms of grade and recovery.
Other methods of OFC exist; some using modelling and others
resorting to expert control methods.
Fig. 5. Graderecovery curve illustrating optimising control objective (Wills and
Napier-Munn, 2006).

This in turn creates recovery and/or concentrate grade setpoints for the
lower AFC/base level control structures (Laurila et al., 2002). However,
OFC should only affect lower control levels if the process is at steady
state (McKee, 1991).
Graderecovery curves vary according to feed grade and can be
altered through manipulation of plant operating variables such as air
rates. In this way, graderecovery curves can be optimised (Neethling
and Cilliers, 2008). An example of this is presented by Hadler and
Cilliers (2009), whereby the graderecovery curve for a bank of four
rougher cells was optimised by maximising the stability of the froth
within each cell.
The use of froth stability as a parameter was rst presented by
Moys (1984), who published a study whereby the horizontal velocity
of the froth could be related to the froth stability, (also dened as
the air recovered into the launder in the form of unbroken bubbles).
Using this quantitative measure of froth stability, Woodburn et al.
(1994) developed a semi-empirical froth-based otation model that
combines a conceptual froth structure with the kinetics of otation;
the latter being based on the ux of bubble surface area overowing
from the cell (see Eq. (1)).
B = Q a Sb vf hw Sb

In this model the ux of bubble surface area (B) is calculated


from the volumetric air owrate into the cell (Qa), the specic bubble
area (Sb) and the air recovery, . This is approximately the same as
relating B to the specic bubble area, froth velocity (vf), froth height
(h), weir length (w) and the volume fraction of the froth that is air (),
usually taken as unity. This relationship was simplied by Barbian
et al. (2003), who suggested the value of can be calculated by
Eq. (2).
vf h w=Q a

Barbian et al. (2003, 2005, 2006) measured froth stability at


laboratory and industrial scales using two different methods; air
recovery and the Froth Stability Column. The air recovery was
determined using image analysis to measure the froth velocity, with
the overowing froth height being measured visually. The Froth
Stability Column is an alternative measure of froth stability comprising a non-overowing column in which the froth rises unhindered.
The rate of froth growth and maximum froth height achieved are used
to give a quantitative measure of froth stability. A good correlation
was shown between the two froth stability measures, suggesting the
potential for both as future measures of froth stability at an industrial
scale. Of greater interest, however, were the results shown by Barbian
et al. (2006), where a peak in froth stability was shown as the air
owrate increased. A peak in air recovery (or PAR) was also shown in
the study by Hadler and Cilliers (2009), in which air owrate was

5.1. Modelling-based methods in optimising otation control


Many model-based optimising otation controllers are algorithms
that locate the optimal operating point on the graderecovery curve;
and then present recovery and grade setpoints to lower control
systems and/or plant operators/management. Flintoff (1992) presents
the principal of iso-economic contours; which are presented as
negatively sloped straight lines on a graderecovery curve that are
calculated according to the price of nal product, smelting costs,
transportation costs, etc. The optimal operating position is then found
by locating the point at which the calculated iso-economic contour is
tangent to the graderecovery curve (generated from plant data).
Other modelling-based optimising controllers are more complex,
and also consider operating parameter limits and technical details (i.e.
not solely economical factors). Muoz and Cipriano (1999) present a
model-based control strategy that aims to both regulate and optimise
a combined primary grinding and otation circuit. The optimising
branch of the controller aims to maximise nancial prot using nonlinear dynamic modelling; encompassing both technical (e.g. mill
power limits, sump level limits, etc.) and economic criteria (e.g. metal
prices, grinding costs, etc.). More recently, Maldonado et al. (2007),
proposed a method considering phenomenological models for each
otation bank of the circuit, validated using process data obtained from
several sampling surveys.... The control objective is the minimisation of
the Cu tailing grade in each bank given a nal Cu concentrate grade. This
is achieved through dynamic programming methods, such that nonlinear behaviour can be accounted for; with promising simulation
results being obtained.
5.2. Expert methods in optimising otation control
Laurila et al. (2002) suggest new expert systems are concentrating
to solve the issue of feed type classication, which is a challenging and
important task. It has already been mentioned that the grade
recovery curve is a function of the ore type, and hence it is logical that
operating parameters and dosing rates should be altered accordingly
if possible. Laine (1995) adds that the performance of such a control
strategy is decided by the ability of the system to effectively classify
the feed ore, which in turn is dependent on the information received
by the classication algorithm. The on-line measurement technique
must extract adequate useful information; whilst the algorithm must
be able to effectively (and readily) classify ore into one of the dened
ore type classes. This would allow for a feedforward type controller to
alter upstream operating parameters so as to optimally process each
dened ore type. An example of such an expert system is presented by
Jms-Jounela et al. (2000); where Kohonen self organising maps
(a type of ANN) are used for classication of the ore feed information
gathered from instrumentation. Once the ore is classied, the expert
control system alters the setpoints of lower control systems and sends
information to operators; with the performance being indicated by an
economical success index display.

68

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

Table 3
Summary of literature review focusing on various types of control.
Branch

Type

References

Modelling based

Feedback predictive MIMO


Feedback and feedforward predictive MIMO

Hodouin et al. (1993)


Del Villar et al.(1999); Desbiens et al. (1998b); Ding and Gustafsson
(1999); Hodouin et al. (2000); Hulbert (1995); Zavala et al. (1995)
Benaskeur and Desbiens (1999); Delport (2005); Desbiens et al. (1998a);
Maldonado et al. (2007)
Desbiens et al. (1994); Jms-Jounela (1992); Sbarbaro (1999); Thornton
(1991)
Bergh et al. (1995); Cubillos and Lima (1997, 1998); Gaulocher et al.
(2008); Nez et al. (2010)
Benford and Meech (1992); Bergh et al. (1996, 1998, 1999); Bergh and
Yianatos (1999); Cipriano et al. (1991); McKay and Ynchausti (1996)
Carvalho and Duro (1999, 2000, 2002); Cipriano et al. (1998); Hirajima
et al. (1991); Osorio et al. (1999); Suichies et al. (1998, 2000)
Aldrich et al. (1997); Cortez and Duro (1995); Duro and Cortez (1995);
Moolman et al. (1995d)
Aldrich et al. (1997)
Bascur (1991); Muoz and Cipriano (1999); Pulkkinen et al. (1993); SosaBlanco et al. (2000)

Non-linear mathematical modelling


Adaptive control
Modelling/AI based hybrids
AI based

Hierarchical combination of expert


and modelling
Supervisory control
Fuzzy logic
Neural networks

Integrated otation and grinding control

Inductive learning
Modelling and/or AI based

However, despite the various OFC strategies outlined in literature,


many mineral processing plants currently rely on operators and plant
management to manually select setpoints of lower control systems,
based on past experience, in order to optimise the process (Laurila
et al., 2002).
6. Examples of approaches found in advanced/optimising
otation control
Various methods of achieving AFC and OFC have been discussed. A
summary of examples using some of these different advanced
methods are presented in Table 3 (adapted from (Hodouin et al.,
2001)). It should be noted that although the methods are divided into
groups, some overlap between methods is inevitable (e.g. adaptive
model-based controllers are inevitably also classed as predictive
controllers).
7. Commercial advanced/optimising otation control software
Various commercial A/OFC systems are available on the market
and have been trialled and implemented in industry; many of which
have already been mentioned in this communication.
One such control package is the FloatStar suite; which consists of
FloatStar Level Stabiliser, FloatStar pH controller, FloatStar Flow

Optimiser (these rst three have already been discussed), FloatStar


GradeRecovery Optimiser and FloatStar Reagent Optimiser. FloatStar
GradeRecovery Optimiser, an OFC, uses online grade analysis to
ensure that recovery is maximised for a specied grade; through
manipulation of level, air owrate, re-circulating load and reagent
addition setpoints across the plant. FloatStar Reagent Optimiser uses a
combination of control approaches (such as fuzzy logic and non-linear
multivariable predictive control) to automate the manipulation of
reagent addition rates (see Fig. 6; www.mintek.co.za).
Several commercial systems make use of machine vision; these
include: VisioFroth by Metso (www.metso.com), FrothMaster by
Outotec (www.outotec.com) and PlantVision by KnowledgeScape
(www.kscape.com). Measured variables include froth velocity, bubble
size distribution, stability and colour. Expert systems are used to
manipulate variables such as pulp level, air addition rate, reagent
addition and/or water addition in a bid to increase recovery at a set (or
improved) concentrate grade. Each of these systems has been
successfully trialled and incorporated on plants. For example, VisioFroth
was successfully implemented on the PT Freeport plant in Indonesia
(Supomo et al., 2008) (Table 4).
8. Conclusions
Despite several advances in base level controls since the 1970s,
reports of fully automated advanced and optimising otation control
systems operating successfully (and unassisted) for long periods
remain scarce. It is hoped, however, that through continued
development of new, robust technologies (e.g. machine vision and
air recovery measurement) and the continued simplication/modications of plant designs (requiring less-intricate control systems),
long term, automated advanced and optimising otation control
will be achievable. Such an outcome would indeed be nancially
rewarding.
Acknowledgements
The authors would like to thank Prof. Raymond Shaw, Dr. Stephen
Neethling and Dr. Kathryn Hadler of Imperial College London for their
valuable input and advice.
References

Fig. 6. Graphical representation of a fuzzy logic generated reagent addition calculation, based
on concentrate and tailings grade by FloatStar Reagent Optimiser (www.mintek.co.za).

Atkins, J.E., Barton, G.W., Johnston, R.D., 1986. Synthesis and evaluation of alternative
control schemes for a chalcopyrite otation circuit. International Journal of Mineral
Processing 16, 2942.

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

69

Table 4
Summary of commercial packages (Aldrich et al., 2010).
System
Company/institution
Year of launch
Major installations

AceFlot

Frothmaster

JKFrothCam

SmartFroth

VisioFroth

1994
Chile

Outokumpu, Finland
1996
Finland, Cadia Valley (Australia),
Macraes (New Zealand)

JKMRC, Australia
1999
Australia

UCT, South Africa


2000
Anglo Platinum (South Africa)

Metso
2001
Escondida (Chile), Freeport
(Indonesia), Australia

Aldrich, C., Moolman, D.W., Eksteen, J.J., Van Deventer, J.S.J., 1995. Characterization of
otation processes with self-organizing neural nets. Chemical Engineering
Communications 139, 2539.
Aldrich, C., Moolman, D.W., Douws, F.S., Schmitz, G.P.J., 1997. Machine learning strategies
for control of otation plants. Control Engineering Practise 5 (2), 263269.
Aldrich, C., Marais, C., Shean, B.J., Cilliers, J.J., 2010. On-line monitoring and control of
froth otation systems with machine vision: a review. International Journal of
Mineral Processing 96 (1-4), 113.
Arbiter, N., Harris, C.C., 1962. Flotation kinetics. In: Fuerstenau, D.W. (Ed.), Froth
Flotation. Edward Brothers Inc., New York, pp. 215246.
Banford, A.W., Aktas, Z., Woodburn, E.T., 1998. Interpretation of the effect of froth
structure on the performance of froth otation using image analysis. Powder
Technology 98 (1), 6173.
Barbian, N., Ventura-Medina, E., Cilliers, J.J., 2003. Dynamic froth stability in froth
otation. Minerals Engineering 16, 11111116.
Barbian, N., Hadler, K., Ventura-Medina, E., Cilliers, J.J., 2005. The froth stability column:
linking froth stability and otation performance. Minerals Engineering 18 (3),
317324.
Barbian, N., Hadler, K., Cilliers, J.J., 2006. The froth stability column: measuring froth
stability at an industrial scale. Minerals Engineering 19 (68), 713718.
Barbian, N., Cilliers, J.J., Morar, S.H., Bradshaw, D.J., 2007. Froth imaging, air recovery and
bubble loading to describe otation bank performance. International Journal of
Mineral Processing 84 (14), 8188.
Bartolacci, G., Ourriban, M., Bouajila, A., Gomez, C., Finch, J., Goyette, G., 2008. On-site
use of hydrodynamic sensors to improve metallurgical performance of otation
machines. In: Kuyvenhoven, R., Gomez, C., Casali, A. (Eds.), Procemin2008, V
International Mineral Processing Seminar, pp. 357366. Chapter 6.
Bartolacci, G., Pelletier Jr., 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 (68), 734747.
Bascar, O.A., 1982. Modelling and computer control of a otation cell. PhD thesis,
University of Utah, Salt Lake City, Utah.
Bascar, O.A., Herbst, J.A., 1982. Dynamic modelling of a otation cell with a view toward
automatic control. XIV IMPC, CIM, Toronto, Canada.
Bascur, O.A., 1991. Integrated grinding/otation controls and management. In: Dobby,
G.S., Argyropoulos, S.A., Rao, S.R. (Eds.), Proceedings of copper 91, 2. Pergamon
Press, Oxford, pp. 411428.
Bascur, O.A., 2000. An interactive dynamic otation model framework. Proceedings of
XIV International Mineral Processing Congress, Rome, Italy, pp. 2131.
Bearman, R.A., Milne, R.W., 1992. Expert systems: opportunities in the minerals
industry. Minerals Engineering 5 (1012), 13071323.
Benaskeur, A.R., Desbiens, A., 1999. Lyapunov-based decentralised control of a rougher
otation phenomenological simulator. In: Hodouin, D., Bazin, C., Desbiens, A. (Eds.),
Control and Optimisation in Minerals, Metals and Materials Processing, pp. 323332.
Benford, P.M., Meech, J.A., 1992. An advisory package for otation operators using a
real-time expert system. Mineral Engineering 5 (1012), 13251331.
Bennett, M.A., West, R.M., Luke, S.P., Williams, R.A., 2002. The investigation of bubble
column and foam processes using electrical capacitance tomography. Minerals
Engineering 15, 225234.
Bergh, L.G., Yianatos, J.B., Acua, C.P., 1995. Hierarchical control strategy for otation
columns. Minerals Engineering 8 (12), 15831591.
Bergh, L.G., Yianatos, J.B., Cartes, F., 1996. Hierarchical control strategy in columns at El
Teniente. In: Gomez, C.O., Finch, J.A. (Eds.), Proceedings of the International
Symposium on Column Flotation, Column '96. CIM, Montreal, pp. 369380.
Bergh, L.G., Yianatos, J.B., Leiva, C.A., 1998. Fuzzy supervisory control of otation
columns. Minerals Engineering 11 (8), 739748.
Bergh, L.G., Yianatos, J.B., Acua, C.A., Prez, H., Lpez, F., 1999. Supervisory control at
Salvador otation columns. Minerals Engineering 12 (7), 733744.
Bergh, L.G., Yianatos, J.B., 1999. Supervisory control experience on large industrial
otation columns. In: Hodouin, D., Bazin, C., Desbiens, A. (Eds.), Control and
Optimisation in Minerals, Metals and Materials Processing, pp. 299310.
Bergh, L.G., Yianatos, J.B., 2011. The long way toward multivariate predictive control of
otation processes. Journal of Process Control 21, 226234.
Berglund, G., 1991. Pulp chemistry in sulphide mineral otation. International Journal
of Mineral Processing 33, 2131.
Bezuidenhout, M., Van Deventer, J.S.J., Moolman, D.W., 1997. The identication of
perturbations in a base metal otation plant using computer vision of the froth
surface. Minerals Engineering 10 (10), 10571073.
Bonifazi, G., Massicci, P., Meloni, A., 2000. Prediction of complex sulde otation
performances by a combined 3D fractal and colour analysis of the froths. Minerals
Engineering 13 (7), 737746.
Bonifazi, G., Giacontieri, V., Serranti, S., Volpe, F., 2005a. A full color digital imaging
based approach to characterize otation froth: an experience in Pyhaslami and

Garpenberg plants. Beijing International Conference on Imaging: Technology and


Applications for the 21st Century, pp. 172173.
Bonifazi, G., Serranti, S., Volpe, F. And, Zuco, R., 2005b. Characterization of otation froth
colour and structure by machine vision. Computers and Geosciences 27 (9), 11111117.
Botha, C.P. 1999. An online machine vision froth otation analysis platform. M.Sc.
thesis, University of Stellenbosch, Stellenbosch, South Africa.
Bruckard, W.J., Kyriakidis, I., Woodcock, J.T., 2007. The otation of metallic arsenic as a
function of pH and pulp potential a single mineral study. International Journal of
Mineral Processing 84, 2532.
Carr, D., Dixon, A., Tiili, O., 2009. Optimising large otation cell performance through
advanced instrumentation and control. Proceedings of 10th Mill Operators
Conference, Adelaide, 2009, pp. 299304.
Carvalho, M.T., Duro, F., 1999. Performance of a otation column fuzzy controller.
In: Mastorakis, M.E. (Ed.), Computer and Computational Engineering in Control,
pp. 220225. Athens.
Carvalho, M.T., Duro, F., 2000. Strategies for fuzzy control of a water/air column.
Proceedings of the XXI International Mineral Processing Congress, pp. 17-23.
Carvalho, M.T., Duro, F., 2002. Control of a otation column using fuzzy logic inference.
Fuzzy Sets and Systems 125, 121133.
Chander, S., Fuerstenau, D.W., 1975. Electrochemical reaction control of contact angles
on copper and synthetic chalcocite in aqueous potassium diethyldithiophosphate
solutions. International Journal of Mineral Processing 2, 333352.
Cipriano, A., Medina, A., Martinez, C., Gonzalez, G., 1991. Supervisor and control
strategy for a otation plant. Proceedings of the IFAC Workshop on Expert Systems
in Minerals and Metals Process, Helsinki.
Cipriano, A., Guarini, M., Vidal, R., Soto, A., Seplveda, C., Mery, D., Briseo, H., 1998. A real time
visual sensor for supervision of otation cells. Minerals Engineering 11 (6), 489499.
Clark, D.W., Newell, A.J.H., Chilman, G.F., Capps, P.G., 2000. Improving otation recovery
of copper sulphides by nitrogen gas and sulphidisation conditioning. Minerals
Engineering 13 (12), 11971206.
Cortez, L., Duro, F., 1995. An expert neural network to control a mineral otation
process. Proceedings of the XXV APCOM, Brisbane, pp. 431436.
Cubillos, F.A., Lima, E.L., 1997. Identication and optimising control of a rougher
otation circuit using an adaptable hybrid-neural model. Mineral Engineering 10
(7), 707721.
Cubillos, F.A., Lima, E.L., 1998. Adaptive hybrid neural models for process control.
Computers and Chemical Engineering 22, S989S992 (suppl.).
Cutting, G.W., Barber, S.P., Newton, S., 1986. Effects of froth structure and mobility on
the performance and simulation of continuously operated otation cells.
International Journal of Mineral Processing 16, 4361.
Delport, R., 2005, Process identication using second order volterra models for
nonlinear model predictive control design of otation circuits M.Sc. thesis,
Department of Chemical Engineering, University of Pretoria.
Del Villar, R., Grgoire, M., Pomerleau, A., 1999. Control of bias and level in a laboratory
otation column. Proceedings of the 31st Annual Meeting of the Canadian Mineral
Processors. CIM, Ottawa, pp. 425442.
Desbiens, A., Hodouin, D., Najim, K., Flament, F., 1994. Long-range predictive control of a
rougher otation unit. Mineral Engineering 7 (1), 2137.
Desbiens, A., del Villar, R., Milot, M., 1998a. Identication and gain-scheduled control of
a pilot otation column. In automation in mining, mineral and metal processing.
Proceedings of an IFAC Symposium. Pergamon, Oxford, pp. 337342.
Desbiens, A., Hodouin, D., Mailloux, M., 1998b. Non-linear predictive control of a rougher
otation unit using local models. Automation in Mining, Mineral and Metal Processing.
Proceedings of an IFAC Symposium. Pergamon, Oxford, pp. 297302.
Ding, L., Gustafsson, T., 1999. Modelling and control of a otation process. In: Hodouin,
D., Bazin, Desbiens, A. (Eds.), Control and Optimization in Minerals, Metals, and
Materials Processing. Metallurgical Society of the CIM, Montral, pp. 285298.
Dowling, E.C., Klimpel, R.R., Aplan, F.F., 1985. Model discrimination in the otation of a
porphyry copper ore. Mineral and Metallurgical 87101 (May).
Duro, F., Cortez, L., 1995. Neural network controller of a otation process. Proceedings
of the XIXth IMPC, Vol. 1, pp. 245250.
Estrada-Ruiz, R.H., Perez-Garibay, R., 2009. Effect of particle size on the otation
performance of a complex sulphide ore. Journal of the Southern African Institute of
Mining and Metallurgy 109 (7), 441446.
Filipic, B., Junkar, M., 2000. Using inductive machine learning to support decision
making in machining processes. Computers in Industry 42, 3141.
Flintoff, B.C., 1992. Measurement issues in quality control. Presented at the 1992
Toronto CMP Branch Meeting.
Forbes, G., De Jager, G., 2004a. Bubble size distribution for froth classication.
Sixteenth Annual Symposium of the Pattern Recognition Association of South
Africa, pp. 99104. Langebaan, South Africa.
Forbes, G., De Jager, G., 2004b. Texture measures for improved watershed segmentation
of froth images. Fifteenth Annual Symposium of the Pattern Recognition
Association of South Africa, pp. 16. Grabouw, South Africa.

70

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771

Forbes, G., De Jager, G., Bradshaw, D.J., 2006. Effective use of bubble size distribution
measurements. XXIII International Mineral Processing Congress, pp. 554559.
Istanbul, Turkey.
Forbes, G., de Jager, G., 2007. Unsupervised classication of dynamic froths. SAIEE Africa
Research Journal 98 (2), 3844.
Garrido, C., Toledo, H., San Martin, P., Villanueva, A., 2008. Calibration methodology and
calibration maintenance in courier analyzers for copper molybdenum otation
process. In: Kuyvenhoven, R., Gomez, C., Casali, A. (Eds.), Procemin2008, V
International Mineral Processing Seminar, pp. 339345. Chapter 6.
Gaudin, A.M., 1957. Flotation. McGraw-Hill, New York.
Gaulocher, S., Gallestey, E., Lindvall, H., 2008. Advanced process control of a froth
otation circuit. In: Kuyvenhoven, R., Gomez, C., Casali, A. (Eds.), Procemin2008, V
International Mineral Processing Seminar, pp. 327337. Chapter 6.
Gebhardt, J.E., Tolley, W.K., Ahn, J.H., 1993. Color measurements of minerals and
mineralized froths. Minerals and Metallurgical Processing 10 (2), 9699.
Glembotskii, V.A., 1972. Flotation. Primary Sources, New York.
Gomez, C.O., Finch, J.A., 2007. Gas dispersion measurements in otation cells.
International Journal of Mineral Processing 84, 5158.
Gomez, C.O., Cortes-Lopez, F., Finch, J.A., 2003. Industrial testing of a gas holdup sensor
for otation systems. Minerals Engineering 16, 493501.
Green, D.W., Perry, R.H., 2007. Perry's chemical engineers' handbook, 8th edition. McGraw Hill. Section 8 Process Control.
Guo, H., Yen, W.-T., 2005. Selective otation of enargite from chalcopyrite by
electrochemical control. Minerals Engineering 18, 605612.
Gupta, A., Yan, D.S., 2006. Mineral processing design and operation, First edition.
pp. 622671. Chapter 18.
Haavisto, O., Kaartinen, J., Hytyniemi, H., 2008. Optical spectrum based measurement
of otation slurry contents. International Journal of Mineral Processing 88, 8088.
Hadler, K., Cilliers, J.J., 2009. The relationship between the peak in air recovery and
otation bank performance. Minerals Engineering 22, 451455.
Hadler, K., Smith, C.D., Cilliers, J.J., 2010. Recovery vs. mass pull: the link to air recovery.
Minerals Engineering 22, 9941002.
Hamilton, J.A., Guy, P.J., 2000. Pulp level control for otation options and a CSIRO
laboratory perspective. Minerals Engineering 14 (1), 7786.
Hargrave, J.M., Miles, N.J., Hall, S.T., 1996. The use of grey level measurement in
predicting coal otation performance. Minerals Engineering 9 (6), 667674.
Hargrave, J.M., Hall, S.T., 1997. Diagnosis of concentrate grade and mass owrate in tin
otation from colour and surface texture analysis. Minerals Engineering 10 (6),
613621.
Hargrave, J.M., Brown, G.J., Hall, S.T., 1998. A fractal characterisation of the structure of
coal froths. Coal Preparation 19 (12), 6982.
Harris, C.C., Chakravarti, A., 1970. Semi-batch otation kinetics: species distribution
analysis. Transactions AIME 247, 162176.
Hayes, R.A., Ralston, J., 1988. The collectorless otation and separation of sulphide
minerals by Eh control. International Journal of Mineral Processing 23, 5584.
Henning, R.G.D., Schubert, J.H., Atasoy, Y., 1998. Improved otation performance at
Fimiston plant through better level control. Presented at International Symposium
on Gold Recovery, Montreal.
Hicyilmaz, C., Emre Altun, N., Ekmekci, Z., Gokagac, G., 2004. Quantifying hydrophobicity of pyrite after copper activation and DTPI addition under electrochemically
controlled conditions. Minerals Engineering 17, 879890.
Hintikka, V.V., Leppinen, J.O., 1995. Potential control in the otation of sulphide
minerals and precious metals. Minerals Engineering 8 (10), 11511158.
Hirajima, T., Takamori, T., Tsunekawa, M., 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. Proceedings of international conference on column
otation '91, Sudbury, Canada, Vol. 2, pp. 375389.
Hodouin, D., Flament, F., Bazin, C., 1993. Predictive control of otation plants.
Emerging Computer Techniques for the Minerals Industry. SME, Littleton,
Colarado, pp. 312.
Hodouin, D., Bazin, C., Gagnon, E., Flament, F., 2000. Feedforwardfeedback predictive
control of a simulated otation bank. Powder Technology 108, 173179.
Hodouin, D., Jms-Jounela, S.-L., Carvalho, M.T., Bergh, L., 2001. State of the art and
challenges in mineral processing control. Control Engineering Practise 9, 9951005.
Holtham, P.N., Nguyen, K.K., 2002. On-line analysis of froth surface in coal and mineral
otation using JKFrothCam. International Journal of Mineral Processing 64 (23),
163180.
Hulbert, D.G., 1995. Multivariable control of pulp levels in otation circuits. Automation
in Mining, Mineral and Metal Processing, Proceedings of an IFAC Symposium.
Pergamon Press, Oxford, pp. 309314.
Hytyniemi, H., Ylinen, R., 2000. Modelling of visual otation froth data. Control
Engineering Practice 8 (3), 313318.
Imaizumi, T., Inoue, T., 1963. Kinetic consideration of froth otation as a rate process.
Proc. VI Int. Min. Proc. Congr., Cannes, p. 581.
Jameson, G.J., Nam, S., Moo Young, M., 1977. Physical factors affecting recovery rates in
otation. Mineral Science and Engineering 9, 103108.
Jms-Jounela, S.-L., 1992. Simulation study of self-tuning adaptive control for rougher
otation. Powder Technology 69, 3346.
Jms-Jounela, S.-L., Karesvuori, J., Laurila, H., 2000. Flotation process neural data analysis
and on-line monitoring. In: Tagami, M. (Ed.), Proceedings of the 32nd Annual
Operator's Conference of the Canadian Mineral Processors, pp. 441457. Ottawa.
Jeanmeure, L.F.C., Zimmerman, W.B.J., 1998. CNN video based control system for a coal
froth otation. Proceedings of the IEEE International Workshop on Cellular Neural
Networks and Their Applications, pp. 192197.
Jowett, A., Safvi, S.M., 1960. Renement in methods of determining otation rates.
Transactions AIME 217, 351357.

Kaartinen, J., Hytyniemi, H., 2005. Combining multi-camera-data of otation circuit


with PCA and PLS. Australasian Institute of Mining and Metallurgy Publication
Series, pp. 121125.
Kallionen, J., Heiskanen, K., 1993. Effective otation of a difcult nickel-ore based on
intelligent mineral technology. Minerals Engineering 6 (810), 917928.
Kmpjrvi, P., Jms-Jounela, S.-L., 2003. Level control strategies for otation cells.
Minerals Engineering 16, 10611068.
Kirjavainen, V., Lyyra, M., Laaplas, H., Heiskanen, K., 1992. Study on using potentiostatic
control in selective sulphide otation. Minerals Engineering 5 (1012), 12791286.
Kocaba, D., Gler, T., 2007. Two-liquid otation of sulphides: an electrochemical
approach. Minerals Engineering 20, 12461254.
Laine, S., 1995. Ore type based expert control system for Hitura concentrator. In: Barker,
I.J. (Ed.), Preprints of the 8th IFAC International Symposium on Automation in
Mining, Mineral and Metal Processing, pp. 321327. Sun City.
Laurila, H., Karesvuori, J., Tiili, O., 2002. Strategies for instrumentation and control of
otation circuits. Mineral Processing Plant Design, Practise and Control, Volume 1,
pp. 21742195. ISBN-10: 0873352238.
Leppinen, J.O., Hintikka, V.V., Kalapudas, R.P., 1997. Effect of electrochemical control
on selective otation of copper and zinc from complex ores. Minerals Engineering
11 (1), 3951.
Lin, X.-Z., Zhao, G.A., Gu, Y.B., 2007a. A classication of otation froth based on
geometry. Proceedings of the 2007 IEEE International Conference on Mechatronics
and Automation, pp. 27162720. ICMA 2007, Art. No. 4303987.
Lin, X.-Z., Gu, Y.-Y., Zhao, G.-Q., 2007b. Calculation of the specic area of the froth in
otation. Meitan Xuebao/Journal of the Chinese Coal Society 32 (3), 304308.
Liu, J.J., MacGregor, J.F., Duchesne, C., Bartolacci, G., 2005. Flotation froth monitoring
using multiresolutional multivariate image analysis. Minerals Engineering 18 (1),
6576.
Liu, J.J., MacGregor, J.F., 2007. On the extraction of spectral and spatial information from
images. Chemometrics and Intelligent Laboratory Systems 85 (1), 119130.
Liu, J.J., MacGregor, J.F., 2008. Froth-based modeling and control of otation processes.
Minerals Engineering 21, 642651.
Luyben, M., Luyben, W., 1997. Essentials of process control. McGraw-Hill. ISBN: 0-07039172-6.
Maldonado, M., Sbarbaro, D., Lizama, E., 2007. Optimal control of a rougher otation
process based on dynamic programming. Mineral Engineering 20, 221232.
Maldonado, M., Desbiens, A., del Villar, R., 2008. An update on the estimation of the
froth depth using conductivity measurements. Minerals Engineering 21, 935939.
Mathe, Z.T., Harris, M.C., O'Connor, C.T., Franzidis, J.-P., 1998. Review of froth modelling
in steady state otation systems. Minerals Engineering 11 (5), 397421.
McKay, J.D., Ynchausti, R.A., 1996. Expert supervisory control of otation columns. In:
Gomeze, C.O., Finch, J.A. (Eds.), Proceedings of the International Symposium on
Column Flotation, Column '96. Montral, CIM, pp. 353367.
McKee, D.J., 1991. Automatic otation control a review of 20 years of effort. Minerals
Engineering 4 (711), 653666.
Moilanen, J., Remes, A., 2008. Control of the otation process. In: Kuyvenhoven, R.,
Gomez, C., Casali, A. (Eds.), Procemin2008, V International Mineral Processing
Seminar, pp. 305313. Chapter 6.
Moolman, D.W., Aldrich, C., Van Deventer, J.S.J., Stange, W.W., 1994. Digital image
processing as a tool for on-line monitoring of froth in otation plants. Minerals
Engineering 7 (9), 11491164.
Moolman, D.W. 1995. The on-line control of an industrial otation plant using
videographic images and signal processing, PhD thesis, University of Stellenbosch,
South Africa.
Moolman, D.W., Aldrich, C., Van Deventer, J.S.J., Bradshaw, D.J., 1995a. Interpretation of
otation froth surfaces by using digital image analysis and neural networks.
Chemical Engineering Science 50 (22) 35012513.
Moolman, D.W., Aldrich, C., Van Deventer, J.S.J., Stange, W.W., 1995b. The classication
of froth structures in a copper otation plant by means of a neural net. International
Journal of Mineral Processing 43 (34), 193208.
Moolman, D.W., Aldrich, C., Van Deventer, J.S.J., 1995c. The analysis of videographic data
with neural nets. Chimica Acta Slovenica 42 (1), 137142.
Moolman, D.W., Aldrich, C., Van Deventer, J.S.J., 1995d. Monitoring of froth surfaces on
industrial otation plants using connectionist image processing techniques.
Minerals Engineering 8 (12), 2330.
Morar, S.H., Forbes, G., Heinrich, G.S., Bradshaw, D.J., King, D., Adair, B.J.I., Esdaile, L.,
2005. The use of a colour parameter in a machine vision system, smartFroth, to
evaluate copper otation performance at Rio Tinto's Kennecott Utah copper
concentrator. Australasian Institute of Mining and Metallurgy Publication Series,
pp. 147151.
Morar, S.H., Hateld, D.P., Barbian, N., Bradshaw, D.J., Cilliers, J.J., Triffet, B., 2006. A
comparison of otation froth stability measurements and their use in the
prediction of concentrate grade. Proceedings of the XXIII International Minerals
Processing Congress, pp. 739744. Istanbul, Turkey.
Moys, M.H., 1984. Residence time distributions and mass transport in the froth phase of
the otation process. International Journal of Mineral Processing 13, 117142.
Muoz, C., Cipriano, A., 1999. An integrated system for supervision and economic
optimal control of mineral processing plants. Mineral Engineering 12 (6), 627643.
Narraway, L.T., Perkins, J.D., Barton, G.W., 1991. Interaction between process design and
process control: economic analysis of process dynamics. Journal of Process Control
1, 243250.
Neethling, S.J., Lee, H.T., Cilliers, J.J., 2003. Simple relationships for predicting the
recovery of liquid from owing foams and froths. Minerals Engineering 16 (110),
11231130.
Neethling, S.J., Cilliers, J.J., 2008. Predicting and correcting graderecovery curves:
theoretical aspects. International Journal of Mineral Processing 80, 1722.

B.J. Shean, J.J. Cilliers / International Journal of Mineral Processing 100 (2011) 5771
Niemi, A., Paakkinen, U., 1969. Simulation and control of otation circuits. Automatica
5, 551561.
Niemi, A.J., Ylinen, R., Hytyniemi, H., 1997. On characterization of pulp and froth in
cells of otation plant. International Journal of Mineral Processing 51 (14), 5165.
Nez, F., Tapia, L., Cipriano, A., 2010. Hierarchical hybrid fuzzy strategy for column
otation control. Minerals Engineering 23, 117124.
Nguyen, K.K., Holtham, P., 1997. The application of pixel tracing techniques in the otation
process. Proceedings of the First Joint Australian and New Zealand Biennial Conference
on Digital Imaging and Vision Computing and Applications, pp. 207212.
Oestreich, J.M., Tolley, W.K., Rice, D.A., 1995. Development of a color sensor system to
measure mineral compositions. Minerals Engineering 8 (12), 3139.
Osorio, D., Prez-correa, J.R., Cipriano, A., 1999. Assessment of expert fuzzy controllers
for conventional otation plants. Minerals Engineering 12 (11), 13271338.
Prez-correa, R., Gonzlez, G., Casali, A., Cipriano, A., Barrera, R., Zavala, E., 1998.
Dynamic modelling and advanced multivariable control of conventional otation
circuits. Minerals Engineering 11 (4), 333346.
Polat, M., Chander, S., 2000. First-order otation kinetics models and methods for
estimation of the true distribution of otation rate constants. International Journal
of Mineral Processing 58, 145166.
Pulkkinen, K., Ylinen, R., J.ams.a-Jounela, S.L., J.arvensivu, M., 1993. Integrated expert
control system for grinding and otation. Proceedings of the XVIIIth IMPC
Congress; The Australasian Institute of Minerals and Metals, Sydney, Australia,
pp. 325334.
Qing, W.-Q., He, M.-F., Chen, Y.-P., 2008. Improvement of otation behaviour of Mengzi
leadsilverzinc ore by pulp potential control otation. Transactions of Nonferrous
Metals Society of China 18, 949954.
Rastogi, R.C., Aplan, F.F., 1985. Coal otation as a rate process. Minerals and
Metallurgical Processing 137145 (Aug.).
Remes, A., Saloheimo, K., Jms-Jounela, S.-L., 2007. Effect of speed and accuracy of online elemental analysis on otation control performance. Minerals Engineering 20,
10551066.
Roesch, M., Ragot, J., Degoul, P., 1976. Modeling and control in the mineral processing
industries. International Journal of Mineral Processing 3, 219246.
Roos, J.R., Celis, J.P., Sudarsono, A.S., 1990a. Electrochemical control of chalcocite and
covellite-xanthate otation. International Journal of Mineral Processing 29, 1730.
Roos, J.R., Celis, J.P., Sudarsono, A.S., 1990b. Electrochemical control of metallic copper
and chalcopyrite-xanthate otation. International Journal of Mineral Processing 28,
231245.
Ruonala, M., 1995. The use of electrochemical mixed potential measurements for the
process control and expert system development at the Hitura mine. Proceedings of
Automation in Mining, Mineral and Metal Processing.
Ruonala, M., Heimala, S., Jounela, S., 1997. Different aspects of using electrochemical
potential measurements in mineral processing. International Journal of Mineral
Processing 51, 97110.
Sadr-Kazemi, N., Cilliers, J.J., 1997. An image processing algorithm for measurement of
otation froth bubble size and shape distributions. Minerals Engineering 10 (10),
10751083.
Sbarbaro, D., 1999. Adaptive non-linear predictive control of a simulated otation
cell. In: Hodouin, D., Bazin, C., Desbiens, A. (Eds.), Control and Optimization in
Minerals, Metals, and Materials Processing. Metallurgical Society of the CIM,
Montral, pp. 311321.
Shen, W.Z., Fornasiero, D., Ralston, J., 1997. Effect of collectors, conditioning pH and
gases in the separation of sphalerite from pyrite. Minerals Engineering 11 (2),
145158.
Siren, A.K., 1999. The characterization of otation by colour information and selecting
the proper equipment. Proceedings of the 2nd International Conference on
Intelligent Processing and Manufacturing of Materials, vol. 1, pp. 471476.
Smith, C., Neethling, S., Cilliers, J.J., 2008. Air-rate prole optimisation: from simulation
to bank improvement. Minerals Engineering 21, 973981.
Smith, C.D., Hadler, K., Cilliers, J.J., 2010. Flotation bank air addition and distribution for
optimal performance. Minerals Engineering 23, 10231029.

71

Sosa-Blanco, C., Hodouin, D., Bazin, C., Lara-Valenzuela, C., Salazar, J., 2000. Economic
optimisation of a grinding and otation plant. Minerals Engineering 13 (1011),
9991018.
Stenlund, B., Medvedev, A., 2000. Level control of cascade coupled otation tanks.
Control Engineering Practice 10, 443448.
Suichies, M., Leroux, D., Dechert, C., 1998. An implementation of generalised predictive
control in a otation plant. Automation in Mining, Mineral and Metal Processing,
Proceedings of an IFAC Symposium, pp. 425442.
Suichies, M., Leroux, D., Dechert, C., Trusiak, A., 2000. An implementation of generalised
predictive control in a otation plant. Control Engineering Practise 8, 319325.
Supomo, A., Yap, E., Zheng, X., Banini, G., Mosher, J., Partanen, A., 2008. PT Freeport
Indonesia's mass-pull control strategy for rougher otation. Mineral Engineering
21, 808816.
Thornton, A.J., 1991. Cautious adaptive control of an industrial otation circuit.
Minerals Engineering 412, 12271242.
Thwaites, P., 2007. Process control in metallurgical plants. Annual Reviews in Control
31, 221239.
Tomlinson, H.S., Flemming, M.G., 1963. Flotation rate studies. Proc. VI Int. Min. Proc.
Congr., Cannes, p. 563.
Torrealba-Vargas, J.A., Finch, J.A., 2006. Continuous air rate measurement in otation
cells: some fundamental considerations. International Journal of Mineral Processing 81, 8592.
Uribe-Salas, A., Martnez-Cavazos, T.E., Nava-Alonso, F.C., Mndez-Nonell, J., LaraValenzuela, C., 2000. Metallurgical improvement of a lead/copper otation stage by
pulp potential control. International Journal of Mineral Processing 59, 6983.
Vathavooran, A., Batchelor, A., Miles, N.J., Kingman, S.W., 2006. Applying froth imaging
techniques to assess ne coal dewatering behaviour. Coal Preparation 26 (2), 103121.
Ventura-Medina, E., 2000. Flotation froth structure and performance. PhD Thesis,
Chemical Engineering Department, University of Manchester.
Ventura-Medina, E., Cilliers, J.J., 2000. Calculation of the specic surface area in
otation. Minerals Engineering 13 (3), 265275.
Walker, G.W., Stout, J.V., Richardson, P.E., 1984. Electrochemical otation of sulphides:
reactions of chalcocite in aqueous solution. International Journal of Mineral
Processing 12, 5572.
Wang, W., Bergholm, F., Yang, B., 2003. Froth delineation based on image classication.
Minerals Engineering 16 (11), 11831192.
Wang, M., Cilliers, J.J., 1999. Detecting non-uniform foam density using electrical
resistance tomography. Chemical Engineering Science 54, 707712.
Wang, Y.A., Neethling, S.J., 2009. The relationship between the surface and internal
structure of dry foam. Colloids and Surfaces A: Physicochemical and Engineering
Aspects 339, 7381.
Wang, W.X., Stephansson, O., 1999. A robust bubble delineation algorithm for froth
images. Proceedings of the 2nd International Conference on Intelligent Processing
and Manufacturing of Materials, vol. 1, pp. 471476.
Wills, B.A., Napier-Munn, T.J., 2006. Wills' Mineral Processing Technology, 7th edition.
Elsevier, pp. 267344. Chapter 12.
Woodburn, E.T., Austin, L.G., Stockton, J.B., 1994. Froth based otation kinetic model.
Chemical Engineering Research and Design 72 (A2), 211226.
Woods, R., 2003. Electrochemical potential controlling otation. International Journal
of Mineral Processing 72, 151162.
Yianatos, J.B., Hendrquez, F.D., 2006. Short-cut method for otation rates modelling of
industrial otation banks. Minerals Engineering 19, 13361340.
Yuan, X.-M., Plsson, B.I., Forssberg, K.S.E., 1996. Flotation of a complex sulphide ore I.
Cu/Zn selectivity control by adjusting pulp potential with different gases.
International Journal of Mineral Processing 46, 155179.
Zanin, M., Wightman, E., Grano, S.R., Franzidis, J.-P., 2008. Quantifying contributions to
froth stability in porphyry copper plants. International Journal of Mineral
Processing 91, 1927.
Zeng, X., Franzidis, J.-P., Manlapig, E., 2004. Modelling of froth transportation in
industrial otation cells Part I. Development of froth transportation models for
attached particles. Minerals Engineering 17, 981988.

You might also like