A Survey On The Degree of Automation in The Mineral Processing Industry

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

IEEE Africon 2017 Proceedings

A Survey on the Degree of Automation in the


Mineral Processing Industry
Laurentz E. Olivier and Ian K. Craig
Department of Electrical, Electronic, and Computer Engineering
University of Pretoria
Pretoria, 0002
Email: ian.craig@up.ac.za

Abstract—A survey was conducted regarding the degree of pertaining to the processing industry in general, but some
automation in the mineral processing industry. The aim of the of the advanced control results will also be compared to
survey was to gauge the current status of automation in the the results from the survey described in this paper where
industry, and to identify opportunities for improved operation
through process control. Key results indicate that there is still applicable.
a large scope for optimizing control in the industry, especially
regarding proper disturbance rejection, reacting to significant
II. S URVEY RESULTS
changes in operating point, and control in the presence of faults. The objectives of the survey were to determine:
• the degree to which operations in the mineral processing
Index Terms—Automatic control, mineral processing, produc-
tion control, survey. industry are automated;
• how often, and why, operators need to intervene in the

I. I NTRODUCTION process;
• what inhibits further automation; and
Process control is an integral part of the modern processing • what are the challenges on the road to lights-out process
plant. In fact, control is regarded as a mature technology in control?
the process industry, without which the industry cannot operate The survey was distributed (among others) to academics
[1]. There is a natural drive towards automation in processing active in research in the mineral processing industry, control
plants, and it is envisioned that gradually production person- practitioners and consultants, production personnel, as well as
nel will be fewer and more sophisticated, while processing process engineers and metallurgists providing process support.
systems will become more automated and efficient [2]. Respondents completed the survey on-line as a web-based
With increased automation the role of the plant operator questionnaire that was active in March 2015, and again during
will evolve from continuously monitoring and intervening, to a July and August 2016. In total 55 completed questionnaires
manager-by-exception [3]. These automation efforts may then were received. This number compares favourably with the
at some point culminate in “lights-out process control” [4], amount of respondents listed in [5], [6], even though the
which refers to the totally automated operation of a processing number is relatively small for drawing statistically significant
plant (as can only be achieved through a proper automation conclusions. The full survey questionnaire can be seen in [2].
strategy).
This article focuses on the current status and perception of A. Respondent information
automation in the mineral processing industry to gauge how Fig. 1 shows the positions of survey respondents. Note-
far this industry is from achieving lights-out process control, worthy responses in the “Other” category are Engineering
as well as to identify areas for improved operations through Managers and Project Managers. Fig. 2 shows the geographical
automation. A survey on the degree of automation in the locations of respondents and Fig. 3 the commodities that
mineral processing industry was conducted in order to achieve respondents are involved with processing. Most respondents
these aims. are based in Africa, which is largely owing to the questionnaire
A survey on the control and economic concerns of grinding distribution.
mill circuits is presented in [5]. The survey presented here
is not limited to grinding mill circuits, and does not focus B. Results regarding operator actions
specifically on economic performance assessment (see [6] or One of the main indications of the degree of automation of
[7] for an example in the mineral processing context). Some a process is the degree of human intervention in the process,
of the results of this survey are however in line with [5] and shown in Fig. 4. Most respondents report that operators
comparisons will be drawn where applicable. perform 1 action per 10 to 30 minutes. Operator workload has
Another recent survey regarding control in the processing been quantified through the Subjective Workload Assessment
industry is presented in [6]. The survey described in [6] Technique (SWAT) [8]. In [9] the SWAT technique was used to
focuses on economic performance assessment and control find that a good workload level for a distributed control system

404

978-1-5386-2775-4/17/$31.00 ©2017 IEEE


IEEE Africon 2017 Proceedings

How many operator actions are required


during normal plant operation?
What is your position?
5 − Operators are constantly
Researcher busy
Control/Instrumentation
Engineer 4 − At least one operator
Process Engineer action per 10 minutes

Other 3 − One operator action per


30 minutes
Metalurgist
2 − One operator action per
Plant Manager hour
Plant Superintendent
1 − One operator action per
Systems Integrator shift

Sales Engineer 0% 10% 20% 30%

0% 5% 10% 15% 20% Fig. 4.


Fig. 1.
How important are these operator
actions?

5 − Essential to safe and


profitable operation
What is you geographical location? 4 − Without these it would be
difficult to keep the unit
Africa running
3 − Necessary but without
South America them overrides would still
keep the unit running
2 − Overrides and safety
Asia systems could also have kept
the unit running profitably
1 − These actions have little
Europe impact on the profitability
of the unit
North America 0% 10% 20% 30%

Fig. 5.
Oceania

0% 20% 40% 60%


(DCS) operator performing under steady-state conditions is
Fig. 2. about 4.4 control moves per hour. During process transients
a moderate workload of 13.6 moves per hour is reported. An
excessive workload was marked at about 18.6 moves per hour,
above which operator errors increase to an unacceptable level
What commodity do you extract? [9]. This result shows that operators in the mineral processing
Copper industry have acceptable workload levels.
Gold Fig. 5 shows the perceived level of importance of operator
Platinum actions, with most respondents rating on the higher end of the
Iron scale. If operators are taking a fair amount of actions per hour
Zinc
and these actions are very important to the unit operation,
an important question is what these actions are. The result
Coal
is shown in Fig. 6 with most respondents stating that these
Lead
actions are for changes in operating conditions or for normal
Other operational requirements.
Silver Changes made for normal operational requirements include
Phosphor actions to keep the plant operating at setpoint; the vigilant
Diamonds operator may also make some optimizing moves when the
0% 20% 40% plant is at steady state. Moves made for changes in operating
conditions show that plants experience frequent transients.
Fig. 3.
Respondents indicated a relatively large variance in the
amount of times control functions are disabled on their plants

405
IEEE Africon 2017 Proceedings

What is the main reason for operator What types of control technologies does
actions? your plant/unit use?
Changes in operating PID control
conditions
Model predictive control
Normal operational Expert system−based control
requirements
Fuzzy logic
Failures of
instruments/equipment Multivariable control
Constraint control
Poor control setup
Linear programming
Other Non−linear models/algorithms
Statistical process control
Safety considerations
Neural networks
0% 10% 20% 30% 40% 50% 0% 25% 50% 75%

Fig. 6. Fig. 8.

How often are control functions How many control loops are on your
disabled and operated manually by the plant?
operator?

5 − At least once per hour Fewer than 20

4 − About once every 4 hours


20 − 100

3 − About once per shift


100 − 1000
2 − About once per day

More than 1000


1 − Fewer than once per day

0% 10% 20% 30% 0% 10% 20% 30% 40%

Fig. 7. Fig. 9.

(see Fig. 7). Respondents were asked to disregard control based control and fuzzy logic, together outnumber the use of
functions that are not set up properly and are disabled perma- MPC, and these are largely preferred over MPC when only
nently. Other than incorrect set-up, there can be a wide array of poor process models are available.
reasons why control functions are disabled corresponding to The scope for lights-out process control on a per-plant basis
the difficult nature of controlling mineral processing plants. is related to the relative size of control installations typically
This difficulty is attributed to strong external disturbances, encountered. Fig. 9 shows the number of basic controllers and
poor process modelling, and difficulty in measuring process Fig. 10 shows the number of advanced controllers typically
variables [10]. The implementation of peripheral control tools encountered.
(see [10] for the definition and [11] for an implementation ex- Regarding on-line process measurements (see Fig. 11); the
ample) to overcome these difficulties may help with alleviating fact that respondents feel more measurements are made than
the need for disabling control functions. Faults on instruments needed for basic control functions indicates that there is room
and actuating elements (as well as the maintenance thereof) are for implementation of optimizing controllers. Some important
also large contributors to why control functions are disabled. process variables are measured manually (see Fig. 12) which
corresponds with the statement by [10] that measurements are
C. Results regarding automatic control installations difficult to make. Poor on-line measurement quality will also
The control technologies used by mineral processing plants limit implementation of optimizing controllers.
is shown in Fig. 8. This figure indicates that PID control really It is shown in Fig. 13 that most respondents are happy with
is ubiquitous. In [5] PID was also found to be by far the most the performance of their control systems. The main inhibitor
common for grinding mills. Model predictive control (MPC) of improved control performance is illustrated by Fig. 14 to
is reported as the most common advanced control method, be a lack of understanding of the process dynamics. Reasons
which is in line with the results of [6] for the process industry pertaining to automation, namely setting up and maintaining
as a whole. The heuristic approaches, namely expert system- control systems and a lack of on-line measurements, are listed

406
IEEE Africon 2017 Proceedings

How efficient would you rate the


control setup on your plant/unit to be?
How many advanced process controllers
are on your plant? 5 − We owe a big portion of
turnover to our efficient
controls
Fewer than 5
4 − Our control systems
perform well
5 − 10 3 − The control setup helps
us to keep the unit
operational
10 − 100
2 − We struggle with
efficient control
More than 100
1 − Control functions are
barely active
Not sure
0% 10% 20% 30% 40% 50%

0% 10% 20% 30% Fig. 13.


Fig. 10.
What do you think is the main factor
that inhibits your control system from
performing better?
Lack of understanding of the
process dynamics
How many online measurements are made Lack of expertise in
on your plant/unit? operating the system
Lack of expertise in setting
5 − There is an abundance of up and maintaining the
measurements for control, control system
optimization and planning Lack of instrumentation for
functions online measurements
4 − More measurements are Focus is placed on other
made than are required for areas
basic control functions Difficulty of the control
3 − Just enough measurements problem
are made for basic control Other
functions
Lack of actuators to
2 − Fewer measurements than implement control signals
are needed for basic control
functions 0% 10% 20% 30%
1 − Just some measurements
are made, other than that the Fig. 14.
unit is operated on operator
experience
0% 10% 20% 30% 40%
as third and fourth. The main inclusion in the “Other” category
Fig. 11. is instrument failures.
Much work has gone into modelling processes in the min-
eral processing industry. An overview of modelling techniques
for a variety of mineral processing operations is shown in [12].
How many manual measurements are made A validated model for a grinding mill circuit is presented
on your plant/unit?
by [13]. Black-box modelling has also been used to good
5 − Most of our important
process variables are
effect for MPC (see for example [14]). It does therefore seem
measured manually that these techniques have not yet found widespread use in
4 − Many important process improving the lack of understanding of process dynamics the
variables are measured
manually mineral processing industry.
The fact that a lack of expertise in operating the system
3 − Some process variables
are measured manually is listed as the second largest inhibitor of improved control
performance also indicates that there is a large scope for
2 − Very few manual improved operation through increased automation.
measurements are made
Fig. 15 shows the functions that advanced controllers per-
1 − We do not make use of form on mineral processing plants. Most respondents indicated
manual measurements
that advanced controllers perform regulation and optimization
0% 10% 20% 30% 40% 50% tasks on their plants. Basic regulation tasks should however not
solely be performed by advanced control. Advanced control
Fig. 12.
should be implemented on top of properly functioning regu-
latory controls, the correct implementation of which usually

407
IEEE Africon 2017 Proceedings

What functions do advanced controllers Does your control system give you the
perform on your plant/unit? ability to specify objectives on the
following control layers?
Regulatory control

Optimization Regulatory control

Switching of control
philosophies
Optimization
Planning functions

None
Planning functions
Not sure

0% 20% 40% 0% 25% 50% 75% 100%

Fig. 15. Fig. 17.

What is the main benefit you gain from How are faults on instruments and
advanced process control? actuators detected?
Faults are only detected by
Reduction in variability of operators through their
process variables effects
Transmitters and actuators
are used that provide some
Not sure feedback regarding faults
(e.g. open circuit)
Through specialized fault
Safety detection and isolation
software
Faults often go undetected
Other for extended periods of time
Only during plant walks or
Environmental protection routine inspections are
faults detected
0% 25% 50% 75% 0% 10% 20% 30% 40%

Fig. 16. Fig. 18.

leads to significant control benefits. indicate that the automated achievement of planning objectives
It is less common to see advanced controllers facilitating is widespread in the industry. A plant may have one planning
the switching of control philosophies, or to perform planning objective achieved through control and many others achieved
functions. Even though advanced planning functions can be through manual action, and the respondent would have still
automated [15], advanced control may also perform basic answered “yes” to this question.
planning functions, such as managing plant throughput. The
fact that few respondents marked “None” on this question D. Results regarding fault detection
indicates the penetration of advanced control into the mineral Faults are reportedly only detected through their effects by
processing industry. operators as shown in Fig. 18. This implies that in general fault
Many control implementations and advanced control detection and isolation is a manual action. To a certain extent
projects are justified on the promise of reduction in variability plants also commonly make use of transmitters and actuators
of process variables [16] which is perceived by respondents that provide some form of fault feedback. This helps to isolate
(see Fig. 16). Even though environmental compliance is listed some common control system failures.
as one of the main challenges for control in the future of
mineral processing [1], no respondents marked environmental E. Results regarding lights-out process control
protection as their main benefit from advanced control. This The statement was made that “It will in future be possible to
situation may however change in the future. run our mineral processing plant(s) completely autonomously,
Fig. 17 indicates whether implemented control systems give i.e. with no human intervention.” Fig. 19 shows the agreement
users the ability to achieve regulatory control functions, opti- of respondents with this statement. Almost half of the respon-
mization functions, and planning functions (such as specifying dents either agreed or strongly agreed with this statement.
the throughput of a unit). All respondents indicated that they About one third of respondents either disagreed or strongly
achieve regulation through control. Fig. 17 does not however disagreed with the statement.

408
IEEE Africon 2017 Proceedings

Please record your level of agreement III. C ONCLUSION


with the following statement: It will
in future be possible to run our A survey was conducted on the degree of automation in
minerals processing plant(s) completely
autonomously, i.e. with no human the mineral processing industry. The survey results show that
intervention. operators still intervene frequently in the process, and that their
5 − Strongly agree actions are considered important. Most operator actions relate
to changes in the operating region, maintaining important
4 − Agree process variables around their targets, and reacting to process
faults. These results indicate that there exists a large scope for
3 − Neutral further implementation of optimizing control.
Many important process variables are only measured man-
2 − Disagree
ually. On-line measurements (where possible) can help with
1 − Strongly disagree optimization. When on-line measurements are not possible,
soft sensors can serve as a relatively inexpensive alternative.
0% 10% 20% 30% 40%
In order to approach lights-out process control, automatic
Fig. 19. compensation for changes in operating region, optimizing
control, and automatic fault tolerant systems will need to see
much more application in the mineral processing industry.
If you agree with the statement in the
previous question, by when do you think R EFERENCES
it will be possible to run your
minerals processing plant(s) completely [1] I. Craig, C. Aldrich, R. Braatz, F. Cuzzola, E. Domlan, S. Engell, and
autonomously? J. Hahn et al, “Control in the process industries,” The impact of control
technology. IEEE control systems society, 2011.
Already possible [2] L. E. Olivier, “On lights-out process control in the minerals processing
industry,” Ph.D. dissertation, University of Pretoria, 2017.
[3] D. M. Brann, D. A. Thurman, and C. M. Mitchell, “Human interaction
Within the next 5 years
with lights-out automation: A field study,” in IEEE Third Anuual
Symposium on Human Interaction with Complex Systems, 1996, pp. 276
Within the next 10 years – 283.
[4] L. E. Olivier and I. K. Craig, “Lights-out process control – analysis and
framework,” in IEEE Africon, 2017.
Within the next 20 years
[5] D. Wei and I. K. Craig, “Grinding mill circuits - a survey of control and
economic concerns,” Int. J. Miner. Process., vol. 90, pp. 56 – 66, 2009.
It will take more than 20
[6] M. Bauer and I. K. Craig, “Economic assessment of advanced process
years
control - a survey and framework,” Journal of Process Control, vol. 18,
0% 10% 20% 30% pp. 2 – 18, 2008.
[7] D. Wei and I. K. Craig, “Economic performance assessment of two ROM
Fig. 20. ore milling circuit controllers,” Minerals Engineering, vol. 22, pp. 826
– 839, 2009.
[8] G. B. Reid and T. E. Nygren, “The subjective workload assessment tech-
nique: A scaling procedure for measuring mental workload,” Advances
in psychology, vol. 52, pp. 185 – 218, 1988.
Respondents were also asked to briefly explain their answer. [9] C. S. Connelly, “Toward an understanding of DCS control operator
Respondents in agreement with the statement cited that the workload,” ISA Transactions, vol. 34, pp. 175 – 184, 1995.
[10] D. Hodouin, “Methods for automatic control, observation, and optimiza-
increased sophistication of modern control systems will lead tion in mineral processing plants,” Journal of Process Control, vol. 21,
to increased levels of automation. Efficiency of advanced pp. 211 – 225, 2011.
controllers over manual optimization actions was also men- [11] L. E. Olivier, B. Huang, and I. K. Craig, “Dual particle filters for state
and parameter estimation with application to a run-of-mine ore mill,”
tioned as a driver towards total automation. Respondents Journal of Process Control, vol. 22, pp. 710 – 717, 2012.
disagreeing with the statement mostly listed large changes in [12] R. P. King, Modeling and simulation of mineral processing systems.
operating conditions, and equipment failures as reasons for Elsevier, 2012.
[13] J. D. Le Roux, I. K. Craig, D. G. Hulbert, and A. L. Hinde, “Analysis and
their disagreement. These opinions indicate that fault tolerant validation of a run-of-mine ore grinding mill circuit model for process
control [17] has not yet achieved widespread use throughout control,” Minerals Engineering, vol. 43-44, pp. 121–134, 2013.
the mineral processing industry. [14] X. Chen, J. Zhai, S. Li, and Q. Li, “Application of model predictive
control in ball mill grinding circuit,” Minerals Engineering, vol. 20, pp.
Lastly respondents were asked, if they agreed with the 1099 – 1108, 2007.
[15] J. Z. Lu, “Closing the gap between planning and control: A multiscale
statement regarding total automation, by when they thought it MPC cascade approach,” Annual Reviews in Control, vol. 40, pp. 3 –
would be possible to run mineral processing plants completely 13, 2015.
autonomously; Fig. 20 shows the result. Most respondents [16] F. Xu, B. Huang, and S. Akande, “Performance assessment of model
pedictive control for variability and constraint tuning,” Industrial and
indicated that this would be possible within the next 20 years. Engineering Chemistry Research, vol. 46, pp. 1208 – 1219, 2007.
Considering the amount of time it takes for sound control [17] Y. Zhang and J. Jiang, “Bibliographical review on reconfigurable fault-
philosophies and technologies to be widely implemented in tolerant control systems,” Annual reviews in control, vol. 32, pp. 229 –
252, 2008.
industry from the time it was first presented in academic
literature, this estimate may not be far fetched.

409

You might also like