Você está na página 1de 14

Minerals Engineering, Vol. 4, Nos 7-11, pp. 6534i66, 1991 0892-6875/91 $3.

00 + 000
Printed in Great Britain © 1991 Pergamon Press plc

AUTOMATIC FLOTATION CONTROL- A REVIEW OF 20 YEARS OF EFFORT

D.J. McKEE

Julius Kruttschnitt MineralResearch Centre, Isles Road, Indooroopilly, Australia.

ABSTRACT

Some twenty years ago, the first on-line devices for measuring the metal
content o f flotation slurries became available. As a result, the first studies in
the automatic or computer control of industrial flotation circuits commenced.
The development of robust and lasting automatic control systems for flotation
circuits has proved difficult. Reasons for this include the inherent complexity
and unpredictability of the response of most flotation circuits to upset
conditions, unclear expectations of what a control system can achieve,
unrealistic objectives for control systems and excessive complexity o f the actual
control strategies. However, the interest in developing control systems has
persisted because the benefits to be gained in terms of improved metallurgical
performance are substantial.

A pattern o f development has emerged for flotation control systems. Most of


the early systems were concerned with some form of stabilizing control,
although a few systems were aimed directly at optimization. It is now generally
accepted that stabilizing control must precede optimization, and the focus has
shifted to a range of increasingly sophisticated approaches to achieve
stabilization by the use of various model based control strategies. A recent
development is the application of expert systems as the crucial role and
knowledge of operators are being appreciated.

Keywords
Flotation control, objectives, stabilizing, optimizing, model based, expert
systems.

INTRODUCTION

Few topics in the metallurgical control area have received the degree of attention which has
been devoted to automatic control systems for flotation over a 20 year period. There has
been sustained effort in many countries, particularly in Australia, Canada and Finland. The
reasons are sensible and easy to understand: flotation is notorious for its susceptibility to
process upsets and resultant poor performance, and control offers the tantalising possibility
of greatly improved metallurgical performance.

With the development of reliable on stream analysis systems almost 20 years ago, the crucial
assay information essential for an on line control system was available. The timing coincided
with the availability of mini computers for process control and the first successful grinding
control systems. All the important requirements for flotation control studies were thus
available, and by 1976 the literature was reporting a number of initial control systems [1].

653
654 D.J. McKEE

It is now history that successful flotation control systems have proved to be an extremely
elusive goal. The early optimism has been replaced by the slow realization that the task of
developing effective systems is very difficult.

One of the difficulties in reviewing the developments in flotation control is that the
literature tends to reflect the development and early application stages of a particular
system. It is unusual to read follow up reports on the same system after a number of years,
and it is therefore extremely difficult to assess the long term effectiveness of a particular
system. However, from personal experience it is known that many systems do not remain
functioning for periods of years after installation. Instead, problems occur which are not
rectified, with the frequent result that the control system is shut down.

The approach adopted in this paper is to present at the outset the factors which have
influenced flotation control system development. The discussion does draw on the
experiences of the Julius Kruttschnitt Mineral Research Centre, much of it unpublished. A
pattern of issues emerges and examples are drawn from the literature to illustrate the
particular points. No attempt is made to review the total literature on flotation control.

G E N E R A L C O N S I D E R A T I O N S FOR F L O T A T I O N C O N T R O L SYSTEMS

Introduction

Rather than launch into an immediate description of different types of flotation control
systems, it is instructive to consider a number of issues which experience has taught to be
crucial in the development of flotation control. It is certainly the experience of the J K M R C
that most of the points to be discussed are important for all circuits, and that time spent in
assessing these factors at the outset is crucial for eventual success.

Control Objectives

In any analysis of control systems, the control objective must be understood and accepted
as an essential first prerequisite. In retrospect, one of the difficulties which has beset
flotation control has been a difficulty in identifying realistic control system objectives. Too
often the objectives have been poorly specified, if at all. The inevitable result has been a
control system without a clear focus.

In the broadest sense, the objectives of a flotation control system can usually be stated as
optimizing the metallurgical performance of the circuit. What this specifically means and
how it should be attempted is open to wide interpretation. However, three objectives can
be stated which cover most possibilities. These are:

To stabilize circuit performance by minimizing the frequency and severity of erratic


operation.

To achieve nominated grade or recovery set points.

To maximize the economic performance of the circuit.

Each of these objectives differs significantly, and it is instructive to examine them in some
detail.

The grade and recovery performance of a flotation circuit assessed on a short time base
were graphically illustrated when the first on stream analysis data became available. Figure
I illustrates variations in concentrate grade and tail as revealed by OSA for a circuit under
manual control. These results are not abnormal and they provide an obvious incentive to
develop control loops capable of stabilizing the circuit operation. It is clear that the periods
of excessively low recovery must have a detrimental effect on total performance. Correction
of such instability requires an understanding of the causes and the availability of operating
Automatic flotation control 655

variables capable of being manipulated to stabilize the process. The case for stabilizing
control is quite obvious.

. . . . Pb NEW FEED, 2 " 12 %


°°• •'" ••
..... Pb ROUGHER CONC, IO-25 %
°=*Ho Pb SCAVENGER TAIL, 0 " 5 - 2 " 0 % •• ; •,• °•••.
, •o o


\
." / \.
'" /
SO ~
° Q • ° )"~l
o,•)•°° /

i /
/ •• ..•.•'"

\ " ~ _. .- - ~ _
/

zo

i l i i i i l
21oo o ooo o 300 o soo
TIME - HOUR8

Fig.l Typical variation in a lead rougher=scavenger bank performance

The fundamental metallurgical objective of any flotation circuit is usually to achieve a


desired concentrate grade or recovery. In most cases the objective can be reduced to one of
maximizing recovery at a desired concentrate grade. This objective is best attained by
moving circuit performance along a grade-recovery curve to the desired point and
maintaining operation at that point. Having achieved that level of control further
metallurgical gains can be obtained if it is possible to move circuit operation to a higher
level grade-recovery curve.

A control system which is capable of firstly stabilizing circuit performance, and then
driving the circuit to a desired grade or recovery operating point, would undoubtedly be
considered highly successful.

The ultimate control objective must be to achieve the most economic operation of the total
circuit. Such an objective introduces the important concept of how economic performance
is defined, and particularly, the obvious point that an economic criteria is likely to be
highly circuit and operation dependent. In other words, the best economic operating point
will vary from circuit to circuit depending on a host of local factors. Even for a fixed
circuit, the optimum economic target is likely to vary with time, due to changes in price and
cost structures, and other operating and mining strategies, to name just a few of the possible
effects.

All flotation control systems set out to satisfy one or more of the three basic objectives just
discussed. The literature reports many systems fitting these categories. In some early cases,
economic optimization was the sole initial target. It will be argued in this paper that the
sequence of objectives presented: stabilizing, set point grade/recovery control and finally
optimizing control, is the most efficient way to proceed, on the simple basis that it is
difficult to optimize a plant which suffers from severe instability problems.

Another way of approaching the issue of the most appropriate type of control system is to
concentrate initially on defining the correct objectives for the proposed system, While this
appears a self evident point, in practice it is often overlooked with most unfortunate
consequences for the control system, A simple example illustrates the point, There is little
reason to try to develop a rougher-scavenger control system which maximizes recovery if
656 D . J . McKEE

the clear overall objective of the circuit is to achieve a strict grade target. This line of
thinking leads to the necessity to define very precisely the metallurgical objectives of a
circuit as an essential prerequisite to the specification of an appropriate control strategy.

Circuit Constraints

Very few concentrator superintendents would not place maximizing recovery as one of their
prime operating objectives. However, recovery maximization may be a quite unrealistic
control objective for a circuit because of other dominating factors. The operation of
Bougainville Copper Ltd flotation circuit was an excellent example [2]. As a high tonnage,
low grade porphyry copper mine, best overall economics resulted from a maximum tonnage
philosophy. By the time flotation control work was considered, the maximum tonnage
approach meant that the tonnage rate for each of the large ball mills was set at a constant
throughput. As a direct consequence, the grind size to flotation varied with changes in ore
hardness and feed size to the single stage ball mills. Varying grind size had a well defined
effect on rougher-scavenger recovery, the coarser the grind the lower the recovery.

Thus the overall operating philosophy for the plant meant that there would be at times
reductions in recovery quite beyond the ability of any flotation control system to correct.
A system having recovery as a principal objective simply would not be successful. A
knowledge of this important metallurgical constraint was essential in the initial planning of
a control system.

There are other types of constraints, with many related to physical or equipment limitations
in a particular circuit. In such cases, two options are available. Either the constraint is
removed by mechanical modifications, or its presence is accepted, and the control system
objectives are modified accordingly.

Process Disturbances Affecting Flotation Circuits

If it is accepted that the starting point for most flotation control systems is stabilizing
control, then one of the crucial needs it to understand the types of disturbances affecting
the circuit and the way in which the disturbances cause instability.

Although there are numerous individual upsets which affect flotation circuits, most fall
within one of the following categories.

Changes in throughput.

Variations of ore types and flotation response.

Deliberate alteration of the metallurgical objectives of the circuit (grade or recovery).

Despite the proven ability to use automatic control systems in grinding circuits to achieve
reasonably constant product size, in practice few flotation circuits receive a steady grind
size. Variations in grind size affect not only the rougher and scavenger recovery, but also
the cleaning sections of circuits, where liberation may be too coarse to achieve final
concentrate targets without high cleaner tail assays and subsequent large circulating loads.
In reality, while laboratory testing and historical plant records may have quantified grind
effects quite accurately, it is more difficult to make quantified judgements in the operating
circuits, generally because the grind size is not known.

The effects of ore type variations are generally more serious. The mysteries of changing ore
types are particular to each deposit. One of the great problems which has confronted
flotation systems is that experienced operators can often detect that the ore has changed, but
recognition of the change by the control system has proved enormously difficult. Early
flotation control work at Mount Isa [3] recognized the need to detect ore type changes and
attempted to achieve on line recognition by monitoring the behaviour of the first four
rougher cells. The problem is quite simply that a major shift in flotation response of an ore
Automatic flotation control 657

may require both changes in metallurgical and therefore control targets as well as a changed
reagent regime.

There is no single, effective measure of ore type. There are only two possibilities: either
establish a measurement (usually chemical) or the feed from which changes may be
inferred, or deduce the magnitude or type of change from subsequent circuit measurements,
usually from particular assay patterns. Neither approach has been very successful. Ore type
recognition remains an area in which the good operator is still well in advance of an on
line system. Recognition of this situation is one of the principal reasons for the move
towards expert systems, as will be discussed later.

The final disturbance category, that of changes in metallurgical targets, has the advantage
of being operator induced and therefore quantified. The important point here is to recognize
that such changes do constitute disturbances to the circuit, with the resultant need for some
corrective control action.

Types o f Flotation Circuits

The first 15 years of flotation control was concerned with the development of systems for
circuits consisting of banks of conventional flotation cells. The picture has changed
dramatically in the last five years with the widespread advent of flotation columns. Most
new circuits, and many older ones, now contain one or more columns, either for
conventional cleaning duties or for the treatment of specialized middling streams. The
presence of columns has altered the task of flotation control considerably.

It was quickly established that simple stabilizing control of columns was essential for
effective operation of single columns. The control concepts most commonly developed have
been presented [4]. In one sense, control of columns has proved to be deceptively simple.
Classic stabilizing loops, operating to local set points, are used to successfully control froth
depth, aeration rate and wash water additions. Concentrate grade control using the column
aeration has been quite effective.

However, there are no known reports of the wider goal of controlling a total circuit which
contains conventional cells and columns. There seems no reason that the circulating loads
from columns will not cause the same control problems for total circuits that are
encountered in circuits of conventional cleaners.

S u m m a r y - An Approach to Flotation Control

The foregoing discussion suggests a structured approach to the development of a flotation


control system. In order, the factors to be considered are listed below:

Define the precise metallurgical objectives of the circuit.

Isolate and understand the constraints present in the circuit.

Develop at least an initial understanding of the types of disturbances to be


encountered.

Only after a thorough examination of the above factors, formulate an initial set of
objectives for the desired control system.

Decide on the general control approach - stabilizing, set point control or optimizing.

Finally, select a control approach or mechanism, and commence the development of


strategies.
658 D . J . McKEE

S T A B I L I Z I N G C O N T R O L SYSTEMS

Introductory Comments

The previous discussion presented an approach to the development of a flotation control


system. The comment "decide on the general control approach - stabilizing, set point control
or optimizing" is deceptively simple, for it makes no reference to the actual methods to be
used to implement the chosen approach. The comment also suggests that there is a choice
in the approach to be selected initially. Originally, this was true, in that some early control
studies aimed directly at optimizing circuit performance, as in the case of the K i d d Creek
Concentrator [5]. Other operations began with various forms of stabilizing control, based on
the reasoning that it was first necessary to stabilize a circuit before attempting optimization.

There is now little argument that stabilizing control is the area to start. What has emerged
is a considerable range of methods of actually implementing a stabilizing control strategy.
Some of the significant developments in this area are now discussed.

Approaches to Stabilizing Control

A summary of simple stabilizing systems [6] outlined the following common control
objectives:

Feed back control of collector additions to maintain recovery set points.

Feed forward control of collector based on the calculated metal content of the new
feed.

Maintaining concentrate flows 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.

To illustrate the development cycle of a reasonably comprehensive stabilizing control


system, some of the experiences at Renison Ltd [7] will be discussed. The sulphide flotation
circuit at Renison, as illustrated in Figure 2, removes sulphides in the form of pyrrhotite
from a cassiterite ore prior to gravity and flotation recovery of cassiterite. The objective of
the sulphide flotation circuit is to maximize the recovery of sulphide to the concentrate,
while keeping the cassiterite recovery to the concentrate below maximum limits. The
sulphide concentrate is a discard product, so any cassiterite reporting to the concentrate is
lost. It is also necessary to maintain the sulphide content of the scavenger tail below
maximum limits for effective performance in the gravity circuit.

Because of the need to minimise over-grinding prior to sulphide flotation, much of the
rougher and scavenger concentrate contains sulphide-cassiterite composites, as well as
entrained fine liberated cassiterite. The regrind was located to liberate cassiterite from the
composites, while the duty of the cleaners was to reject entrained and newly liberated
cassiterite back to the head of roughing via the cleaner tail and the thickener.

A concise metallurgical objective for a stabilizing control system was formulated as follows:
to reduce the sulphur content of the rougher tail to a set point value provided the tin
content of the sulphide concentrate did not exceed a maximum value. Prior testwork
indicated that the sulphur in the rougher tail was controllable using the rougher aeration
rate. The sulphur content was inferred from an iron assay. Testwork also indicated that in
this case the rougher tail sulphur was not responsive to collector addition.
Automatic flotation control 659

NEWFEED
v
C Y ~

GRAVITY
CIRCUIT

--0 r

I--ol
Legend

Fe, Sn assays

G Mass Flow

FINAL SULPHIDE CONCENTRATE •

Fig.2 Initial sulphide flotation circuit at Renison Ltd.

A simple feedback control system was implemented using a PI controller to manipulate the
aeration rate to achieve a sulphur set point. However, testing was relatively brief because
of an expansion of the circuit (throughput increase) and the fact that a new process control
computer was to be installed.

The revised sulphide circuit is shown in Figure 3. A scavenging stage and an extra cleaning
stage were added to provide more flotation capacity, and the position of the regrind was
altered within the cleaner circuit.

Flotation control studies resumed after some months. The metallurgical objective of the
circuit remained unaltered, so it was logical to reintroduce the initial rougher aeration
control loop. However, simple response testing showed that while the rougher concentrate
volume flow responded to changes in rougher air, there was little change in the scavenger
tail assay. Testwork indicated that the scavenger assay tail did respond to changes in
scavenger aeration over short periods, but over longer time periods, circulating load effects
from the cleaners and the thickener tended to become dominant. An increase in scavenger
air increased the scavenger concentrate flow, and with the cleaner section air at constant
values, the cleaner tail flow increased. This effect passed quickly through the low residence
time thickener back to the roughers.
660 D.J. McKEE

TOGRAVITY CIRCUIT

Legend T ~
d
v
2ND
CLEANER i
i
O Fe, Sn assays

O Mass Flow 3RD

FINALSULPHIDECONCE

Fig.3 Revised sulphide flotation circuit at Renison Ltd.

A second problem, which related to the selection of a suitable scavenger tail sulphur set
point, also became evident. Effective operation of the scavenger loops was dependent on the
selection of a "correct" set point and while it was possible to state a usual range of set
points, selecting the appropriate value on a day by day basis was anything but simple.

As a result of these problems, control investigations were shifted to the cleaner section, as
the direct source of the circulating load. In summary, response testing indicated that the
circulating load was responsive to aeration rates in the three cleaner stages. Furthermore,
operating experience indicated that there were acceptable limits to the thickener underflow
mass flow rate and that these limits would be useful as set points. A control strategy was
developed from these concepts, and is shown in block diagram form in Figure 4. The
cleaner tail density modification to the strategy was based on the observation that operators
were using the density measurement as an early indication of changes in the thickener
underflow mass flow. The other point indicated in Figure 4 is the use of simple dynamic
compensation techniques to improve the behaviour of the standard PI controllers which
always have difficulties in controlling processes with significant time lags.

A further important feature of the control system is the way in which operators have been
involved as an integral part of the strategy. It was recognized that conditions would be
encountered which were beyond the capability of the circulating load loops. A
comprehensive alarm system was initiated to alert operators to potential problems and to
provide some guidance for possible reaction by operators. A further significant input from
operators was the selection of a compatible thickener underflow mass flow set point.
Automatic flotation control 661

Thickener underflow
I massflow setpoint

o or
with dynamic compensation
for time between first
cleaner tail density
I +
CONTRO[.!
Thickener underflow
massflow measurement
change and thickener
underflow massflow
response /
IFirst cleaner tail density setpoint

T
Density c°ntr°ller with I .t
dynamic compensation First cleaner tail density
for time between first CONTROIJ -~ measurement
cleaner air change and
first cleaner tail
density r e s p o n s e /
[First cleaner air setpoint

air controller measurement

Signal to air control valve

Fig.4 Block diagram of circulating load control strategy


The circulating load control strategy has proved remarkably successful in stabilizing the
operation of the sulphide circuit to such an extent that the original scavenger air control has
not been reintroduced.
The Renison study illustrates a number of very important factors as listed below:
The appropriate loops for satisfactory stabilizing control are not always immediately
obvious.
Selection of set points which are appropriate for the circuit and the strategy can be
very difficult.

Simple PI control loops often do not provide acceptable control response.


Operators have an important role, both in understanding the characteristics of the
circuit and most importantly, interacting with the control system to recognize the
need to alter limits and set points.
The development of successful strategies requires consistent effort over a period of
many months.
662 D.J. MCKEE

Refinements in Stabilizing Control

The Renison control example serves to highlight two major challenges in developing
successful stabilizing control systems. The first of these refers to the problems in tuning
effective loops in the face of long time delays, variable process responses and interactions
between variables. The variable process response means for example that the magnitude of
a change of grade in response to a given change in aeration rate will probably vary from one
day to the next, for totally unknown reasons. The interaction problem can occur if variables
such as aeration rate and pulp level are used together. In practice, these problems have
caused many difficulties in attaining effective stabilizing control and it has been argued
[8] that the classical control approach must be replaced by some form of model based
control.

Use of Models to Enhance Stabilizing Control

A thorough treatment of model based control in crushing, grinding and flotation has been
presented [8]. The following overview is taken from that publication.

The overall framework of model based control is presented in Figure 5. The process model
must be capable of reproducing the dynamic behaviour of the flotation circuit. The
estimator takes information from the process and the model to provide estimates of the state
of the process. The optimizer selects the best control path for the process, based on overall
control objectives and the current state of the process. The optimizer varies the set points
of the stabilizing loops and gains of the final control elements.

) J PLANT I
"1 I

T STANDARD
LOOPS

[,dl

T
OPTIMIZER

T MODFI.

h,~l ESTIMATOR
vI
Fig.5 General form of a model based controller (after [8])
Automatic flotation control 663

There are a limited number of reports of actual model based control systems for flotation
circuits, [9, 10, 11, 12]. A number of these studies involved the application of m i n i m u m
variance adaptive control. In this case a predictive model of the process is constructed with
the following form:

yl (t+k+l) --- A 1 y(t)+A 2 y (t-1)+...+Aj y (t-j+l)+B 0 u(t)+B 1 u (t-l)+...+B i u (t-i)+D (1)

where yl (t + k + 1) is the predicted output k +1 steps into the future (variable to be


controlled).
u(t) is the process input at time t (variable to be manipulated).
k is the process dead time.

The model parameters in Eq. 1 (A i, B i and D) are estimated on line by using a recursive
least squares algorithm. A control law is developed by minimizing the following function
with respect to u (t).

j=[(yl(t+k+ l)-y*(t+k+ l))2/2+L(u(t))Z/2] (2)

where y* (t + k + 1) is the value from which the process set point (value of u) is determined
at t i m e r + k + 1.
L is a weighting parameter used to adjust the magnitude of the control action.

An example of such a control action is a loop to control the collector addition to obtain a
set point value of a tail assay. The problem in assessing the effectiveness of minimum
variance control is that the literature does not contain follow up publications which report
on the long term success of the approach. A recent publication [12] claims 75% utilization
for an adaptive loop, with the revealing comment that such operating time is indicative of
Success.

A problem with the form of adaptive control outlined is the continuing need to specify
process set points. For example, for the collector - assay loops, the assay set point must be
specified. In practice good plant operators are often the best judge of the need for set point
changes. Recognition of the role of operators, both in terms of their knowledge of best
control strategies, and their skill in knowing when to adjust control limits and set points,
has been the catalyst for considering the application of expert systems in flotation control.

Expert Systems in Stabilizing Control

The operator has a crucial role in interacting with a control system. The experience of
Outokumpu at the Pyhasalmi Concentrator illustrated this point very clearly, [13] when the
following comment was made: "the importance of the role of the operator cannot be
overstressed even today; operating on the highest level of the control hierarchy all the
responsibility rests in his hands when changing set points and alarm limits of the control
variables". The same point was recognized in the development of the Renison control system
described earlier.

Good operators possess two skills of relevance to flotation control systems. The first is their
knowledge of the circuit, the disturbances encountered and the most useful corrective
actions. In planning an automatic control strategy, one of the golden rules is to first check
with the best operators. The second skill is the ability to recognize the type of disturbance
and to make the connection between the disturbance and necessary changes to set points and
limits.

Such reasoning leads directly into the realms of expert system technology. Outokumpu has
developed an operator advisory system for assisting the control of the pyrite flotation circuit
at Pyhasalmi, [13]. This expert system is connected to a speech synthesizer which outputs
messages for alarm conditions and for disturbances in feed conditions. Operator actions,
particularly set point changes, are also recorded and analysed to provide the basis for better
understanding of the reasons for set point alterations.
664 D.J. McKEE

A second expert system application has been reported at the Polaris lead-zinc concentrator,
[14]. Stated reasons for the system were:

To standardize operating practice.


To stabilize circuit operator.
To train inexperienced operators.
To provide on-line operating support for process operators.
To provide a means of analysing, evaluating and modifying operating strategies.

The above objectives fit very well with the requirements for effective flotation control
outlined earlier.

OPTIMIZING CONTROL OF FLOTATION CIRCUITS

The goal of optimizing control is to determine the most efficient operating point on the
circuit grade recovery-curve. The actual determination of the best point is usually based on
economic criteria. For example, at the Pyhasalmi Concentrator, the best economic operating
point is determined on the basis of an economic recovery equation [15].

P1 = f (Cu in Cu conc, Zn in Cu Conc, Cu Recovery) (3)


The functional relationships in Eq. 3 take into account the following factors:

Price of copper.
Price of zinc.
Transportation, smelting and refining costs as a function of copper concentrate grade.

Costs and prices are updated periodically.

In practice, the optimizing control determines the copper concentrate grade set point and
stabilizing control manipulates process variables to attain that set point. However, the
stabilizing system only seeks the final grade set point if important assays and flows are
within limits. The optimizing control updates the desired grade set point on a two hourly
interval.

An optimizing control scheme also has been developed for the roughers and scavengers
within the zinc circuit. Cleaner control is aimed at stabilizing circulating loads. The
optimizing control determines the best copper sulphate additions to maximize the value of
zinc in the zinc concentrate. Once again, the optimizing section of the strategy only operates
when the important circuit variables are within limits.

The important aspects of the Pyhasalmi control system is that optimization is considered
secondary to stabilizing the circuit. The Outokumpu approach is one which recognizes that
the critical starting point must be a stable circuit.

A SUMMARY OF THE PRESENT STATUS OF FLOTATION CONTROL

What has been achieved after 20 years of effort to develop flotation control systems? Firstly,
a great deal has been learned through hard won experience, particularly that the
development of lasting control systems is difficult. If recent literature is a true reflection
of developments, then there are two themes to current work. The first is to use model based
control to achieve better performance from stabilizing loops. This work is an unstated
recognition that stabilization is the most important problem to be considered and that the
simple, classical control approach is often found wanting in flotation applications. The
second trend is to recognize the role of operators, with expert systems being the convenient
vehicle for operator interaction.
Automatic flotation control 665

There is no doubt that the Outokumpu operation at Pyhasalmi has the record of the longest
consistent effort towards control system development. The Outokumpu experience, gleaned
from their published literature, and observations of the plant and discussions with technical
personnel, can be summarised as follows:

An initial vision of stabilizing control followed by economic optimization has


remained as the goal over approximately 20 years.

Stabilization of the circuit using the most appropriate control variables is the essential
requirement.

The actual control strategies are likely to be specific to each individual circuit.

It has been recognized that a control system is also an information system, and the
credibility of the system can be enhanced by the appropriate presentation of both
raw and calculated data to operators and management.

It has also been acknowledged that operators play a crucial role in recognition of
particular circuit conditions still well beyond the ability of a control system to
comprehend. Likewise, operators are best placed to make intelligent adjustment to
set points and limits.

There is a role for on line models to improve control performance.

Expert systems have a role, particularly in association with the operator.

There is no substitute for persistence in the development of floatation control


systems. 20 years of sustained commitment at the one site is ample evidence of the
point.

FUTURE DEVELOPMENTS

The economic incentive for flotation control systems will continue, and as a result, the
effort to develop systems will be maintained. Given the incentive, there are three necessary
factors, namely the availability of:

Sufficient process information.


The appropriate control techniques.
Trained technical personnel.

As a general rule, the raw process measurements are available for a control system. The
literature does not reflect any real impediments in this area, although there is no doubt that
local problems of instrument maintenance do occur. There is also scope for the inclusion of
instruments which measure chemical parameters of the pulp.

An overall approach to the development of control systems exists which recognizes the
importance of realistic objectives and the need to achieve stable circuit operation. There is
no doubt that robust methods to achieve a satisfactory level of control performance are still
wanting, and work will continue in the use of model based control. However, model based
control techniques are not simple and their successful application requires skilled control
engineers. The shortage of suitable personnel has always been a serious impediment to
control system development, and unfortunately, there are no signs of a real improvement
in this position.

The fourth necessity for success is persistence. Experience indicates that successful control
systems take time to develop and furthermore, that continued maintenance and updating is
essential. The most successful installations have all recognized this factor, and have
maintained consistent effort. There is no reason why this situation will change.

HE 4 : 7 / 1 1 - C
666 D . J . MCKEE

REFERENCES

. Smith H.W., Computer control in flotation plants. Flotation, Volume 2, 963. AIME.
New York (1976).

. Plavina P., Automatic control of rougher-scavenger flotation at Bougainville Copper


Ltd. JKMRC Report No MS-12-1982. (unpublished). (1982).

. McKee D.J., Fewings J,H., Manlapig E.V. & Lynch A.J., Computer control of
chalcopyrite flotation at Mount Isa Mines Ltd. Flotation, Volume 2,994. AIME. New
York (1976).

.
Finch J.A. & Dobby G.S., Column Flotation. 137. Pergamon Press, Oxford (1989).

5. Amsden M.P., Chapman C. & Reading M.G., Computer control of flotation at the
Ecstall Concentrator. CIM Bulletin, 66, 84 (1973).

. Cooper H.R., Trends in computer automation of flotation circuits. AIME, Annual


Meeting. New Orleans (1979).

. Casey M.G., Automatic control of circulating load in a flotation circuit. M.Eng.Sc.


Thesis, University of Queensland (unpublished), (1984).
. Herbst J.A. & Bascur O.A., Mineral processing control in the 1980s - realities and
dreams. Control 84, 197. AIME. New York (1984).

. Henriksen H., Kaggerud J. & Olsen T.O., Application of on line estimation of a


rougher flotation process. Proceedings 6th IFAC Symposium on Identificative and
System Parameter Estimation, Washington (1982).
10. Hammoude A. & Smith H.W., Experiments in self-tuning control of flotation.
Proceedings 3rd IFAC Symposium in Mining, Mineral and Metal Processing, Montreal
(1980).

11. Franklin M.W. & Hendriks D.W., Minimum variance control of flotation circuits at
Brunswick Mining. CIM 24th Annual Conference of Metallurgists, Vancouver (1985).

12. Thorton A.J., Cautious adaptive control of an industrial flotation circuit. Minerals
Engineering, (in press).

13. Kallioinen J., Kesti E., Miettunen J. & Mullen I., A sophisticated information system
for process control. Copper 87, CIM-Chilean Institute of Mining Engineers, Vina der
Mar (1987).

14. Harris C.A. & Kosick G.A., Expert system technology at the Polaris Mine. Canadian
Mineral Processors Annual Operators' Conference, Ottawa (1988).

15. Miettunen J., The Pyhasaimi Concentrate - 13 years of computer control. Proceedings
4th IFAC Symposium on Automation in Mining, Mineral and Metal Processing.
Hensinki (1983).

Você também pode gostar