Minerals Engineering 79 (2015) 139–142

Contents lists available at ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

Benefits of process control systems in mineral processing grinding

circuits
Sylvie C. Bouffard
BHP Billiton Petroleum and Potash, 200-475 2nd Avenue South, Saskatoon, Saskatchewan S7K 1P4, Canada

a r t i c l e i n f o a b s t r a c t

Article history: In 2009, the survey of Wei and Craig reported qualitatively on the benefits of process control systems in
Received 21 March 2015 grinding and flotation. The present review of about twenty milling operations reports quantitatively on
Revised 2 June 2015 these benefits. Expert systems, model predictive controller, and fuzzy logic control systems, to name only
Accepted 3 June 2015
the most used in the mining industry, have delivered superior operational performance (1–16% gain in
Available online 11 June 2015
ore throughput, at least 40% reduction in mill load variability, up to 1% in metal recovery in flotation)
and reduced operating costs (15% reduction in grinding media consumption 52% reduction of the cyclone
Keywords:
pressure variability).
Process control
Mineral processing
Ó 2015 Elsevier Ltd. All rights reserved.
Quantitative benefits
Business case

1. Introduction reduction by 50%, increase yield of the valuable products by 43%,

This article focuses on quantifying the benefits of process con-
trol systems in the mineral processing industry. The benefits
sought in the mineral processing industry are the same sought in
other commodity industry, namely throughput gain, process sta-
bility, energy consumption reduction, and increased yield. A survey
by Bauer and Craig (2007) of 66 respondents from diverse com-
modity industries, including mineral processing, reported on the
preferred types of control systems and hinted to some gains in
throughput of between three and ten percent. Two years later,
Wei and Craig (2009) narrowed their survey to mineral processors
only, of which 68 responded, the majority of them from the
precious metals and base metals industries in Africa and Europe.
Of all respondents, 35% used single-stage milling in closed circuit,
30% used two-stage milling in open and closed circuits, 22%
used two-stage milling in closed circuits, and 15% used other
circuit configurations. The respondents used one or more control
strategies, PID control being the most common, multivariable
control and expert system-based control being on par, following
distantly by fuzzy logic control, neural networks, and others
(adaptive/self-tuning control, model predictive control, linear pro-
gramming, statistical process control, dead-time compensation
control, constraint control). Eighty-nine per cent of respondents
used on/off trials to demonstrate the benefits of the control system.
Greater process stability was common amongst 71% of respon-
dents, throughput gain by 54% of them, energy consumption
E-mail address: Sylvie.bouffard@bhpbilliton.com
operating labour reduction by 30%, downtime reduction by 21%,
and better plant safety by 16%.
The business case for a control system needs to consider the
costs of control hardware, control software, consultant labour, pro-
duction loss due to installation downtime, and hardware/software
maintenance. Against these costs are the benefits of lower operat-
ing costs and higher metal production. Metal production is given
by the product of head grade, throughput and recovery (metal
produced = grade  throughput  recovery). Assuming grade to
be constant, a key objective of control systems in milling opera-
tions is to establish the ore throughput and recovery relationship,
shown indirectly in Fig. 1 as ore throughput and metal production.
For some ores, increasing throughput outweighs the possible
decrease in recovery caused by a coarser grind and lower grain
liberation. For others, there exists a throughput ‘‘sweet spot’’ above
which recovery drops so sharply that it defeats entirely the
throughput gain. Controlling throughput and grind size with
adjustment of ore feed rate to the mill and water flow rate to
sumps and mills is the foundation of grinding control system.
Whereas Wei and Craig (2009) identified the users of grinding
control systems and cited the qualitative benefits of such systems,
the present article goes one step further, compiling quantitative
data for such benefits. It is the first time that a compilation of this
sort is made. Metallurgists and process engineers interested in
implementing control systems will find much value in this
compilation, having now readily available quantitative evidence
from many different sources to build a business case in support
of continuous improvement initiative for grinding systems.

http://dx.doi.org/10.1016/j.mineng.2015.06.006
0892-6875/Ó 2015 Elsevier Ltd. All rights reserved.
140 S.C. Bouffard / Minerals Engineering 79 (2015) 139–142
Metal produced
Ore throughput
Fig. 1. Possible existence of an optimum throughput above which metal production
could decline as a result of coarser grind, lower liberation, and thus lower metal
recovery.
2. Methodology
This study was realized independently of the survey of Wei and
Craig (2009). It pulled information from eighteen published
sources, which included conference proceedings, referred journal
articles, and vendor brochures. No operations survey was
conducted.
The published literature contains many articles about grinding
control systems, many of a theoretical nature with no reference
to any operations, and others presenting simulations that refer to
certain operations. Even though such articles may cite performance
benefits, they are not the subject of this study, concerned only with
actual operational data.
3. Results and discussion
Gold, platinum, copper, nickel, zinc, and iron operations from
across North America, South America, Africa, and Australia, were
examined. There were zero rod mill operations, three ball mill
operations, nine SAG mill operations, three rod and ball mill
operations, zero rod and SAG mill operations, and seven ball and
SAG mill operations.
Table 1 summarizes the information collected as a function of
the type of control system and the commercial control system
used, the objective(s) set out for the control system, the manipu-
lated and controlled variables, and a suite of quantitative metrics,
including mill load, grinding media wear, power consumption,
ore throughput, cyclone performance, recovery benefit, labour,
downtime and utilization, and safety.
3.1. Types of control systems
The types of control system used were: twelve operations using
expert systems, six using model predictive controllers (MPC)
(no mention whether linear or non-linear), five using fuzzy
logic, and occasional applications of multivariate controllers and
other systems (PID control, adaptive/self-tuning control, neural
network-based control, linear programming, statistical process
control, dead-time compensation control, and constraint control).
Emerson Delta V MPC, ABB Linkman Expert Optimizer, Invensys
Connoisseur, Honeywell Profit Suite, Gensym G2, Prediktor, Metso
Adaptive Predictive Model, and Metso OCS process control soft-
ware are common commercial control packages (Thwaites, 2009).
The dominance of the expert system is reinforced by the claim of
White et al. (2004) citing that the former Minnovex (now SGS)
installed fifteen expert crusher control systems from 1995 to
2006. SGS Minnovex competition is Metso OCS. Mintek takes up
a large share of the market for MPCs. Mintek MillStar has been
implemented on many circuit configurations, including platinum
group metals and gold in South Africa, nickel in Zimbabwe,
Botswana and Australia, silver, lead and zinc in Mexico, gold and
copper in Brazil, and copper in Poland.
Whilst this study does not discuss the fundamentals of control
systems, it is worth mentioning some key differences between the
two most commonly used systems: expert system vs MPC. An
expert system uses a model of the operators; MPC uses a mathe-
matical model of the system (Carter, 2010). An expert system is
algebraic and rules-based; MPC is algorithm-based and predictive.
Expert systems have an optimized supervisory control of mineral
processes. MPC track setpoints well, anticipate and reduce
disturbances, and can be used for optimization.
3.2. Benefits
Amongst the references consulted, mill operations imple-
mented control systems to achieve throughput gains (thirteen
instances reporting), increased recovery via grind size control (nine
instances), lesser variability in different areas of the circuit (six
instances), greater control of the mill load (four instances), and
no grind out (three instances). No operations reported testing more
than one control systems in their trials.
The most often cited benefit was an increase in throughput,
followed by a reduction of the throughput variability, a reduction
of the mill load variability, a reduction in power consumption, a
target size achieved, a tightening of the mill feed size distribution,
and a recovery gain. This analysis of the benefits confirms the
previous survey findings of Wei and Craig (2009) reporting most
often process stability improvement, throughput increase, energy
consumption reduction, and increased metal production.
Quantitative data for most of these benefits are given in the
next sub-sections.
3.2.1. Throughput
Throughput gain is the most often cited benefit of process
control systems, followed by lesser throughput variability. Fig. 2
shows a wide range of throughput gains for each type of control
systems. The throughput gain data for expert systems is supported
by Edwards et al. (2002) who quoted four authors having previ-
ously reported ‘‘typical’’ throughput gain of between +4% and
+8% with expert systems. Of those publications consulted, none
reported a loss in throughput; it is construed that operations which
might have experienced a throughput loss did not publish results.
The vertical bars in Fig. 2 represent the average of the range. Five to
seven percent throughput gain is the average of these types of
control system. No system seems to outperform another.
The majority of operations using model predictive controllers
also commented on the significant reduction in the variability of
throughput, by a factor between 22% and 92%. Operations using
other control systems did not make such frequent mention of a
reduction in throughput variability.
3.2.2. Recovery
Table 1 shows significantly fewer references quoting a gain in
recovery. From 0.7% to +1% is the range reported. When using a
fuzzy logic controller, the Xstrata Nickel Raglan operation reported
a 0.7% loss in recovery, but such loss was more than offset by the
3–6% increase in throughput. The operation produced more nickel
as a whole, but that came at the price of losing to tailings 0.7% of
every nickel tonne fed, that would otherwise have been recover-
able under more optimum conditions.
 S.C. Bouffard / Minerals Engineering 79 (2015) 139–142 141

Table 1
Summary of key operational benefits achieved following implementation of control systems.

Type of control Control system Operation Grinding system Throughput Throughput Recovery Reference
system gain variability gain
Multivariable MantaCube GoldFields St Ives Closed single-stage +6.1% Lesser Wallace (2007)
control SAG mill
MIPAC MML Ltd. Century Zinc Mine Closed SAG mill, ball Lesser Thornton et al. (2009)
mill
Expert system- Gensym G2 Asarco Hayden copper Closed rod mill, ball +3.3% Sotelo et al. (1997)
based control concentrator mill
SGS Barrick Lagunas Norte, Peru +3% Gomez et al. (2010)
SGS Peñoles Gold Mine Ball mill +6.3% Festa et al. (2009)
SGS Peñoles Gold Mine SAG mill, ball mill +5% Festa et al. (2009)
With fuzzy Iron ore plant Ball mill with spiral +8.6% Lesser Chen et al. (2007)
logic classifier
OCS Anglogold Ashanti Noligwa Van Drunick and Penny
(2005)
OCS with fuzzy Anglogold Ashanti SAG mill +3.8% Yes Bouché et al. (2005)
logic Kopanang shaft
OCS with fuzzy Anglogold Ashanti Tau SAG mill +10.2% Yes Bouché et al. (2005)
logic Lekoa shaft
OCS with fuzzy Anglogold Ashanti Mponeng Closed SAG mill +1% Bouché et al. (2005) and
logic mine Van Drunick and Penny
(2005)
Xstrata Nickel Raglan Closed SAG mill, ball +4% to +5% +0.6% Bartsch et al. (2008)
Operation mill
Xstrata Strathcona nickel Closed rod mill, ball +7.7% 0% 0% Bartsch et al. (2008) and
mill mill Nunez et al. (2009)
Fuzzy logic Equigold Mt Rawdon plant SAG mill +2.5% 91% Meech(2006)
control
Leeudoorn gold mine Closed SAG mills +10.8% Van Dyk et al. (2000)
OK Tedi Mines SAG mill Yes McCaffery et al. (2003)
Process vision, Doyon Gold Mine Closed SAG mill, ball +1% to +3% Bourassa et al. (1995)
with PID mill
In-house Xstrata Nickel Raglan Closed SAG mill, ball +3% to +6% 0.7% Bartsch et al. (2008)
Operation mill
Model predictive Closed single-stage Muller et al. (2003)
control ball mill
In-house Palabora Mining Open rod mill, ball mill +2% to +3% +0.9% Du Plessis (2001)
Company
MillStar Platinum plant +10% 46% Mintek brochure
MillStar Gold plant SAG mill +6.4% 66% Mintek brochure
MillStar Not specified Not specified +5% to Smith et al. (2004)
+10%
Other Split Online Morila Gold Mine, Mali SAG mill- ball mill +9.6% 87% Gillot (2005)
cameras
Csense product Not specified Not specified Csense brochure
Matrikon product Not specified Not specified +2.5% Matrikon brochure

Five operations using expert systems reported recovery gains of

Other up to 1%. One operation cited no gain but saw benefits in higher
throughput, lower power consumption, and fewer mill shutdowns.
Two operations cited some gain but did not quote the value, and
Model predicve
the remaining two cited a gain of 0.6% and 1%, in addition to
reporting a throughput increase. The increase in recovery must
Fuzzy logic have been the direct outcome of a reduction in grind size; some
operations increased by up to 1.4% the percentage of ore passing
75 lm, while another lowered the P80 by 10 lm. This combination
Expert system of higher throughput and higher recovery is the best outcome that
could have been expected, guaranteed to generate higher metal
0 2 4 6 8 10 12 production.
Gain in throughput (%) One of Palabora Mining Company’s operations had six parallel
milling circuits consisting each of a rod mill followed by a ball mill
Fig. 2. Range of throughput gain achieved with various control systems. The short
in open circuit (du Plessis, 2001). It reported 0.9% gain in recovery,
vertical bar is the average gain.
as well as a throughput gain, using an in-house MPC designed to
At the Peñoles Gold Mine in Chile, the SGS expert system was either reduce particle size for a given throughput, or increase
implemented, resulting in higher throughput without sacrificing throughput for a given particle size. Unfortunately, metallurgists
grind size or recovery. Another benefit of the system was in fact could not determine whether the controller or a change in miner-
a reduction in the variability of the grind size. alogy was responsible for this gain.
142 S.C. Bouffard / Minerals Engineering 79 (2015) 139–142
3.2.3. Other benefits
Six operations reported greater than 90% utilization of the con-
trol system. While encouraging, this finding is bias; it is likely that
plants using control systems more intermittently would not have
published their performance. Mintek identified that the main
causes for not utilizing the controller are: (1) a change in ore type
or particle size, (2) operators reacting too quickly to a small upset
that may not be caused by the controller, and (3) the controller
being untuned.
Five operations reporting on the mill load commented that
when the control system was in use, the variability of the load
was reduced by 40–90%. Only the Leeudoorn gold mine indicated
that the control system increased the mill load by 2.9%. The expert
system of Anglogold Ashanti Kopanang and Xstrata Strathcona
prevented mill overload.
Two operations reported up to 15% reduction in the grinding
media consumption. Xstrata Strathcona and a platinum operation
in South Africa experienced fewer mill grindouts.
Operations cited fewer alarms, fewer mill stop/start, fewer mill
shutdowns, and reduced maintenance on the mill feed hoppers
and seals. Two plants specifically reported setpoints being reached
80% faster with the control system in use.
With respect to the cyclones performance, the benefits were
achieved mostly by reducing the variability of the cyclone
pressure by 32–52%, the cyclone density by 50–85%, and the
cyclone overflow rate by 20–92%.
No plant reported a reduction of labour or an increase in safety
with the use of the control system, though one operation cited
better crew morale.
4. Conclusions
In grinding and flotation processes, the new era of operational
efficiency seeks greater productivity and recovery from ore
reserves. Implementing a control system for typically less than
one million dollars achieves this goal. The mining industry has
used primarily expert system, model predictive controller, and
fuzzy logic control systems. Of the twenty mining operations
worldwide commenting on the performance of their control sys-
tem, benefits are undisputable: 1–16% gain in throughput, up to
1% gain in recovery, many other simultaneous benefits in grinding
and flotation, fewer operator interventions, and a payback of less
than six months (Bouché et al., 2005; Meech, 2006).
Acknowledgement
The author thanks Mrs. Elizabeth Amann, Strategic Knowledge
Leader at BHP Billiton, for collecting publications, as well as two
(2) anonymous reviewers for their suggestions.
References
Bartsch, E., Comeau, G., Hardie, C., 2008. Evolution of SAG mill process control at the
Xstrata Nickel Raglan Operation. In: Paper #28 presented at the 40th Annual
Meeting of the Canadian Mineral Processors, Jan 22–24, 2008, Ottawa, ON,
Canada, pp. 445–464.
Bauer, M., Craig, I.K., 2007. Economic assessment of advanced process control – a
survey and framework. J. Process Control 18, 2–18.
Bouché, C., Brandt, C., Broussaud, A., Drunick, V.W., 2005. Advanced control of gold
ore grinding plants in South Africa. Miner. Eng. 18, 866–876.
Bourassa, M., Lanthier, R., Beaudoin, P., Chéfai, S., 1995. Grinding circuit control
optimization at Mine Doyon. In: Proc. Of the 27th Annual Meeting of the Canadian
Mineral Processors, Paper #22, Presented January 18, 1995, pp. 295–310.
Carter, R.A., 2010. Process control: what’s new, what’s next. Eng. Min. J., 42, January/
February
Chen, X., Zhai, J., Li, Q., Fei, S., 2007. Fuzzy logic based on-line efficiency
optimization control of a ball mill grinding circuit. 4th Int. Conf. on Fuzzy
Systems and Knowledge Discovery, vol. 2. The Computer Society, pp. 575–580.
du Plessis, B.J., 2001. Grinding control strategy on the conventional milling circuit of
Palabora Mining Company. J. S. Afr. Inst. Min. Metall., 165–168
Edwards, R., Vien, A., Perry, R., 2002. Strategies for the instrumentation and control
of grinding circuits. In: Mular, A.L., Halbe, D.N., Barratt, D.J. (Eds.), Mineral
Processing Plant Design, Practice, and Control. Society for Mining, Metallurgy
and Exploration, pp. 2130–2151.
Festa, A., Cornejo, F., Orrante, F., Alanis, R., Gutierrez, B., 2009. Expert system
implementation at Penoles Group Concentrators. In: Presented at the 42nd
Annual Meeting of Canadian Mineral Processors, Ottawa, Canada, January 2009,
13 pp.
Gillot, P.L., 2005. Pit-to-plant optimization at Morila Gold Mine. <http://www.
spliteng.com/downloads/Pit%202%20Plant%20-%20P.L.Gillot.pdf>, 14 pp.
Gomez, A., Aracena, C., Cornejo, F., Festa, A., Vasquez, A., 2010. Rule and fuzzy-logic
based expert control of the Barrick Lagunas Norte mine, Automining 2010,
Chapter II, pp. 28–29.
McCaffery, K.M., Katom, M., Craven, J.W., 2003. Ongoing evolution of advanced SAG
mill control at OK Tedi. In: SME Transactions, vol. 312, 9 pp.
Meech, J.A., 2006. The evolution of intelligent systems in the mining industry. In:
Presented at the Mineral Process Modelling, Simulation and Control Conference,
Laurentian University, Sudbury, ON, Canada, June 5–7, 2006, 30 pp.
Muller, B., Singh, A., de van der Spuy, D., Smith, V.C., 2003. Model predictive control
in the minerals processing industry. In: Lorenzen, L., Bradshaw, D.J. (Eds.),
Proceedings of the XXII International Mineral Processing Congress, 29 Sept–3
Oct., 2003, pp. 1692–1702.
Nunez, E., MacPherson, G., Graffi, D., Tuzun, A., 2009. Self-optimizing grinding
control for maximizing throughput while maintaining cyclone overflow
specifications. In: Paper #32 Presented at the 41st Annual CMP Conference,
Jan 19–22, 2009.
Smith, G.C., Jordaan, L., Singh, A., Vandayar, V., Smith, V.C., Muller, B., Hulbert, D.G.,
2004. Innovative process control technology for milling and flotation circuit
operations. J. S. Afr. Inst. Min. Metall., 353–365
Sotelo, R.R., Gordon, L.M., Jewart, C., 1997. Implementing advanced control on a
copper ore grinding circuit, ISA TECH/EXPO Technology Update Conference, vol.
1, pp. 157–165.
Thornton, A.J., Prinsloo, A., Tewu, R., 2009. Evolution of semi-autogenous grinding
mill control at Century Zinc, AusIMM Mill Operators Conference, 14 p.
Thwaites, P., 2009. Developments in process control with particular reference to
comminution and flotation, McGill Professional Development Seminar –
Mineral Processing Systems, 72 slides.
Van Drunick, W.I., Penny, B., 2005. Expert mill control at Anglogold Ashanti. J. S. Afr.
Inst. Min. Metall. 105, 497–506.
Van Dyk, W., Stange, W., Derbyshire, J., Williams, H., 2000. Optimum control of the
Leeudoorn semi-autogenous milling circuit. CIM Bull. 93 (1043), 106–110.
Wallace, P., 2007. Stabilising the SAG at St Ives. Automation Today, Asia Pacific,
pp. 12–13.
Wei, D., Craig, I.K., 2009. Grinding mill circuits – a survey of control and economic
concerns. Int. J. Miner. Process 90, 56–66.
White, L., Barone, L., Hingston, P., Huband, S., Tuppurainen, D., Bearman, R., 2004. A
multi-objective evolutionary algorithm approach for crusher optimization and
flowsheet design. Min. Eng. 17 (11–12), 1063–1074.