XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 1131
and Biegler 2006). The IDEAS Optimizer is designed to
solve optimization problems where the objective function
and constraints come from a complex model which cannot
be written down as a simple set of analytical equations.
In the use case, the objective function is the summation
of the pump power in each section, the inputs are the vari-
able speed, and the constraints to the algorithm are the most
defining feature for this solution. The constraints in this sit-
uation include the following: The total flow rate of the two
pipelines must be within a small error margin of the flow-
rate in the current operation to work alongside the control
system’s level controller, no optimizer solution can suggest a
different flowrate. The difference in speeds between the two
lines should not be greater than a certain threshold, defined
for each set of pipelines, which helps prevent the optimizer
from driving to unreal solutions or drifting too far from
ideal solutions. Additionally, a constraint is added to the
speed recommendation that constrains how big a change
from the current running status, which is done by updat-
ing speed range constraints to being within a margin of the
current speeds. The Booster Station section has additional
constraints for maximum pressure and flow for each of the
parallel pipelines in order to not trigger a pressure interlock
that shuts down the pumps on that line.
As mentioned in the Online Digital Twin section, part
of the model runs along with the real time condition, but
the other part is trying to solve the optimization problem.
For solving the optimization problem, the model runs as
fast as it can achieve to iterate the optimizer until it finds a
solution. For most optimization Digital Twins, the opera-
tion recommendation is sent out intermittently, for exam-
ple every ten minutes, and is not needed more frequently.
For faster response situations the control system’s logic
and PID loops respond. In this plant, the DCS controls
the total speed demand to control the tank levels, and the
Digital Twin periodically updates the load split between the
lines.
The Digital Twin is continuously running, and the
optimizer has to find a solution repeatedly. The IDEAS
optimizer can be triggered to start calculating an optimal
solution and has code to exit with no solution indicated
at the maximum iterations, so the model can be run con-
tinuously. The trigger to run the optimizer checks that both
trains are running, and the optimizer is ready to run, along
with the main trigger conditions of a time interval, 15 min-
utes in this case, or if the running condition has changed
significantly to need a new optimization run.
The lower power consumption from this optimization
case alone can save an operations team significant energy
consumption.
PROCESS IMPROVEMENT FINDINGS
The Pipeline Digital Twin was implemented at a confiden-
tial American copper mine tailings pipeline. The imple-
mentation had two separate optimization targets as shown
in Figure 1, optimizing the thickener Sump pipeline and
the booster station pipeline separately.
Implementing a Digital Twin is generally a stepwise
process. For the first several months, the Digital Twin pro-
vides the load split to the operator as an advisory setting
for the operators to manually enter. The data is analyzed by
comparing power consumption before and after the opera-
tor implements the advised load split. This step ensures the
solution’s validity without risking the real-time operation.
This also helps build the operation team’s confidence in the
Digital Twin. Once approved by the operation team, in
the next stage the operators can enable the Digital Twin to
directly pass calculated load split setting to the control sys-
tem in a closed loop. Once the loop is closed, it no longer
requires operator intervention.
The initial findings from the full implementation
showed positive feedback from operators, with increasing
utilization of the Digital Twin in steady operating condi-
tions. As Digital Twin utilization increases, the plant comes
closer to autonomous operation. The power consumption
improvements were statistically significant, showing 2.5%
lower kWh consumption when the Digital Twin was online.
The next step of the project is to develop the auto-tun-
ing of the Digital Twin model for its continuous operation
under varying operating conditions, such as changing ore
properties and pump configurations.
CONCLUSIONS
The general approach for an IDEAS Digital Twin involves
incorporating live plant conditions into a simulation model
that is verified to closely match the live process, and pro-
viding to the operations team advisories and recommen-
dations for technical decision-making. Previous IDEAS
Digital Twin presentations focused on providing virtual
instruments that give operations more information about
the process. This presented case takes a high-fidelity model
that contains both the engineering data and historical per-
formance into account and runs an optimizer to provide
direct control in improving energy consumption.
This Digital Twin case is focused on minimizing power
consumption in two sections of parallel pipelines, result-
ing in 2.5% decrease in kWh consumption. A simulation
model can be created for anything, but successful Digital
Twins are created around an operational goal and designed
to answer certain questions. This Digital Twin is config-
ured to solve the optimization problem using the nonlinear
and Biegler 2006). The IDEAS Optimizer is designed to
solve optimization problems where the objective function
and constraints come from a complex model which cannot
be written down as a simple set of analytical equations.
In the use case, the objective function is the summation
of the pump power in each section, the inputs are the vari-
able speed, and the constraints to the algorithm are the most
defining feature for this solution. The constraints in this sit-
uation include the following: The total flow rate of the two
pipelines must be within a small error margin of the flow-
rate in the current operation to work alongside the control
system’s level controller, no optimizer solution can suggest a
different flowrate. The difference in speeds between the two
lines should not be greater than a certain threshold, defined
for each set of pipelines, which helps prevent the optimizer
from driving to unreal solutions or drifting too far from
ideal solutions. Additionally, a constraint is added to the
speed recommendation that constrains how big a change
from the current running status, which is done by updat-
ing speed range constraints to being within a margin of the
current speeds. The Booster Station section has additional
constraints for maximum pressure and flow for each of the
parallel pipelines in order to not trigger a pressure interlock
that shuts down the pumps on that line.
As mentioned in the Online Digital Twin section, part
of the model runs along with the real time condition, but
the other part is trying to solve the optimization problem.
For solving the optimization problem, the model runs as
fast as it can achieve to iterate the optimizer until it finds a
solution. For most optimization Digital Twins, the opera-
tion recommendation is sent out intermittently, for exam-
ple every ten minutes, and is not needed more frequently.
For faster response situations the control system’s logic
and PID loops respond. In this plant, the DCS controls
the total speed demand to control the tank levels, and the
Digital Twin periodically updates the load split between the
lines.
The Digital Twin is continuously running, and the
optimizer has to find a solution repeatedly. The IDEAS
optimizer can be triggered to start calculating an optimal
solution and has code to exit with no solution indicated
at the maximum iterations, so the model can be run con-
tinuously. The trigger to run the optimizer checks that both
trains are running, and the optimizer is ready to run, along
with the main trigger conditions of a time interval, 15 min-
utes in this case, or if the running condition has changed
significantly to need a new optimization run.
The lower power consumption from this optimization
case alone can save an operations team significant energy
consumption.
PROCESS IMPROVEMENT FINDINGS
The Pipeline Digital Twin was implemented at a confiden-
tial American copper mine tailings pipeline. The imple-
mentation had two separate optimization targets as shown
in Figure 1, optimizing the thickener Sump pipeline and
the booster station pipeline separately.
Implementing a Digital Twin is generally a stepwise
process. For the first several months, the Digital Twin pro-
vides the load split to the operator as an advisory setting
for the operators to manually enter. The data is analyzed by
comparing power consumption before and after the opera-
tor implements the advised load split. This step ensures the
solution’s validity without risking the real-time operation.
This also helps build the operation team’s confidence in the
Digital Twin. Once approved by the operation team, in
the next stage the operators can enable the Digital Twin to
directly pass calculated load split setting to the control sys-
tem in a closed loop. Once the loop is closed, it no longer
requires operator intervention.
The initial findings from the full implementation
showed positive feedback from operators, with increasing
utilization of the Digital Twin in steady operating condi-
tions. As Digital Twin utilization increases, the plant comes
closer to autonomous operation. The power consumption
improvements were statistically significant, showing 2.5%
lower kWh consumption when the Digital Twin was online.
The next step of the project is to develop the auto-tun-
ing of the Digital Twin model for its continuous operation
under varying operating conditions, such as changing ore
properties and pump configurations.
CONCLUSIONS
The general approach for an IDEAS Digital Twin involves
incorporating live plant conditions into a simulation model
that is verified to closely match the live process, and pro-
viding to the operations team advisories and recommen-
dations for technical decision-making. Previous IDEAS
Digital Twin presentations focused on providing virtual
instruments that give operations more information about
the process. This presented case takes a high-fidelity model
that contains both the engineering data and historical per-
formance into account and runs an optimizer to provide
direct control in improving energy consumption.
This Digital Twin case is focused on minimizing power
consumption in two sections of parallel pipelines, result-
ing in 2.5% decrease in kWh consumption. A simulation
model can be created for anything, but successful Digital
Twins are created around an operational goal and designed
to answer certain questions. This Digital Twin is config-
ured to solve the optimization problem using the nonlinear