1774 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
the dynamic feedback response to minimize the error
between the SP and PV.
The Model Predictive Controller (MPC) has been
developed as an advanced tool for process control. MPCs
have been extensively used in the oil and gas industry
(Clark, Mohtadi, &Tuffs 1987 Mayne 2014). In min-
eral processing and metallurgical industry, MPCs are also
used to optimally control critical variables in areas such as
semi-autogenous grinding (SAG) mills, ball mills, cyclones,
leaching circuits, flotation lines, thickeners, and kilns, to
name a few. Using MPCs in leaching circuits and CIL
lines provides certain advantages over PIDs or expert sys-
tems in dealing with process-control complexities. Due to
their model-based approach in control, MPCs estimate the
response of multiple PVs, such as cyanide concentrate and
pH, given process inputs and disturbances, and then use
the model predictions to control sodium cyanide and milk
of lime dosage. These models describe the dynamics and
the gains between the process variables. Once these rela-
tionships are defined, the MPC computes the next control
actions by solving an optimization problem in a receding
horizon to bring the process to the SP. By taking advan-
tage of the identified model for the process dynamics, the
controller can provide corrections and predictions of the
cyanide concentrations, which assures sufficient cyanide
dosage to maximize recovery while saving reagent costs.
ANDRITZ developed a proprietary MPC, called
BrainWave ™, and it has been applied in various process
industries for more than three decades, especially in mining
and mineral processing plants. BrainWave uses Laguerre
basis functions to model process dynamics. The Laguerre
basis functions are formulated in a state-space model
(Zervos &Dumont 1988). BrainWave has inherent opti-
mization features in solving error-based cost functions. It
can optimally calculate the controller’s outputs to move the
process to its SP. MPCs in the CIL leaching circuit can con-
trol pH and cyanide concentration to ensure that reagents
are dosed optimally. This paper first introduces some back-
ground about BrainWave, advanced process control (APC)
and digitalization technologies and then describes an APC
application using BrainWave and other supervisory tech-
niques for controlling and optimizing a CIL circuit at a
gold mine in the United States.
ADVANCED PROCESS CONTROL AND
DIGITALIZATION STRATEGY
The APC technology provides an upper layer for optimi-
zation and autonomy of various mineral-processing areas
to reach operational targets. APCs are often offered as a
solution for tacking difficult real-time control and opti-
mization problems. APC solutions use various techniques
such as fuzzy logic, MPC, machine learning, and artificial
intelligence. MPCs, however, have evolved to become an
effective part of the APC solutions due to their capabil-
ity in managing process dynamics in a practical and effec-
tive way. ANDRITZ has developed various APC solutions,
employing BrainWave as its MPC platform. ANDRITZ’s
APC solutions are called Advanced Control Expert (ACE),
and each ACE solution, such as CIL ACE, includes the
MPC and some supervisory control algorithms. In addi-
tion to advanced control operations, Andritz also provides
real-time, plant-wide optimization and digital twin tech-
nologies through the utilization of IDEAS software, recog-
nized as a proficient process simulator. Furthermore, Metris
as an internet of things (iOT) platform was developed by
Andritz to enable machine learning and advanced analyt-
ics solutions to be tackled using Python programming lan-
guage. Figure 1 summarizes ANDRITZ’s digital solutions
together with the contribution of each software application
to the existing plant resources. In this paper, the CIL ACE
performance is deeply analyzed to demonstrate the effec-
tiveness of ANDRITZ’s APC solution.
The primary objective of all APC solutions is to mini-
mize the variable variability and stabilize the process. Once
the variable variability is reduced, the process can then be
optimized to achieve a specific objective. When a process
is challenging to control, especially when it must respect
a particular high or low limit, operators tend to run the
equipment sub-optimally. This is because running closer to
a high or low limit might result in a safety trigger or inter-
lock trip. Therefore, if variability can be reduced, then the
equipment can be run closer to its limits. In CIL leaching
circuits where a reliable cyanide analyzer is available, opera-
tors tend to overdose sodium cyanide to assure no gold is
lost. An APC solution, such as CIL ACE, leverages MPC
tools, such as BrainWave, to minimize the variability of the
residual cyanide concentration and the pH level in the CIL
tanks. Once variabilities are reduced, it becomes feasible to
lower the target for residual cyanide concentration. Such
a reduction inherently leads to a decrease in reagent con-
sumption. Figure 2 illustrates the contribution that an APC
solution, such as CIL ACE, can provide for a gold leaching
circuit compared to PIDs or expert systems.
BrainWave Model Predictive Controller Application
The BrainWave employs a state-space model derived from
Laguerre basis functions, which precisely models the dynam-
ics and transitory response of the process without requiring
the dynamic feedback response to minimize the error
between the SP and PV.
The Model Predictive Controller (MPC) has been
developed as an advanced tool for process control. MPCs
have been extensively used in the oil and gas industry
(Clark, Mohtadi, &Tuffs 1987 Mayne 2014). In min-
eral processing and metallurgical industry, MPCs are also
used to optimally control critical variables in areas such as
semi-autogenous grinding (SAG) mills, ball mills, cyclones,
leaching circuits, flotation lines, thickeners, and kilns, to
name a few. Using MPCs in leaching circuits and CIL
lines provides certain advantages over PIDs or expert sys-
tems in dealing with process-control complexities. Due to
their model-based approach in control, MPCs estimate the
response of multiple PVs, such as cyanide concentrate and
pH, given process inputs and disturbances, and then use
the model predictions to control sodium cyanide and milk
of lime dosage. These models describe the dynamics and
the gains between the process variables. Once these rela-
tionships are defined, the MPC computes the next control
actions by solving an optimization problem in a receding
horizon to bring the process to the SP. By taking advan-
tage of the identified model for the process dynamics, the
controller can provide corrections and predictions of the
cyanide concentrations, which assures sufficient cyanide
dosage to maximize recovery while saving reagent costs.
ANDRITZ developed a proprietary MPC, called
BrainWave ™, and it has been applied in various process
industries for more than three decades, especially in mining
and mineral processing plants. BrainWave uses Laguerre
basis functions to model process dynamics. The Laguerre
basis functions are formulated in a state-space model
(Zervos &Dumont 1988). BrainWave has inherent opti-
mization features in solving error-based cost functions. It
can optimally calculate the controller’s outputs to move the
process to its SP. MPCs in the CIL leaching circuit can con-
trol pH and cyanide concentration to ensure that reagents
are dosed optimally. This paper first introduces some back-
ground about BrainWave, advanced process control (APC)
and digitalization technologies and then describes an APC
application using BrainWave and other supervisory tech-
niques for controlling and optimizing a CIL circuit at a
gold mine in the United States.
ADVANCED PROCESS CONTROL AND
DIGITALIZATION STRATEGY
The APC technology provides an upper layer for optimi-
zation and autonomy of various mineral-processing areas
to reach operational targets. APCs are often offered as a
solution for tacking difficult real-time control and opti-
mization problems. APC solutions use various techniques
such as fuzzy logic, MPC, machine learning, and artificial
intelligence. MPCs, however, have evolved to become an
effective part of the APC solutions due to their capabil-
ity in managing process dynamics in a practical and effec-
tive way. ANDRITZ has developed various APC solutions,
employing BrainWave as its MPC platform. ANDRITZ’s
APC solutions are called Advanced Control Expert (ACE),
and each ACE solution, such as CIL ACE, includes the
MPC and some supervisory control algorithms. In addi-
tion to advanced control operations, Andritz also provides
real-time, plant-wide optimization and digital twin tech-
nologies through the utilization of IDEAS software, recog-
nized as a proficient process simulator. Furthermore, Metris
as an internet of things (iOT) platform was developed by
Andritz to enable machine learning and advanced analyt-
ics solutions to be tackled using Python programming lan-
guage. Figure 1 summarizes ANDRITZ’s digital solutions
together with the contribution of each software application
to the existing plant resources. In this paper, the CIL ACE
performance is deeply analyzed to demonstrate the effec-
tiveness of ANDRITZ’s APC solution.
The primary objective of all APC solutions is to mini-
mize the variable variability and stabilize the process. Once
the variable variability is reduced, the process can then be
optimized to achieve a specific objective. When a process
is challenging to control, especially when it must respect
a particular high or low limit, operators tend to run the
equipment sub-optimally. This is because running closer to
a high or low limit might result in a safety trigger or inter-
lock trip. Therefore, if variability can be reduced, then the
equipment can be run closer to its limits. In CIL leaching
circuits where a reliable cyanide analyzer is available, opera-
tors tend to overdose sodium cyanide to assure no gold is
lost. An APC solution, such as CIL ACE, leverages MPC
tools, such as BrainWave, to minimize the variability of the
residual cyanide concentration and the pH level in the CIL
tanks. Once variabilities are reduced, it becomes feasible to
lower the target for residual cyanide concentration. Such
a reduction inherently leads to a decrease in reagent con-
sumption. Figure 2 illustrates the contribution that an APC
solution, such as CIL ACE, can provide for a gold leaching
circuit compared to PIDs or expert systems.
BrainWave Model Predictive Controller Application
The BrainWave employs a state-space model derived from
Laguerre basis functions, which precisely models the dynam-
ics and transitory response of the process without requiring