XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 931
will highlight the required setpoint changes that operators
should consider.
CASE-STUDY
A 10+ Mtpa flotation circuit of a South American iron ore
mine is shown as a case study in this paper. Note that our
Flotation Optimizer has been commissioned for other com-
modities as well. The circuit consists of four reverse-flotation
stages: rougher, scavenger, cleaner and re-cleaner. The whole
flotation circuit consists of two parallel lines of 11 flotation
columns (5-2-2-2) where three reagents are used to separate
quartz and other silicates from iron oxide-bearing particles.
Following our standard configuration and commissioning
procedures, the mining technology engineering and data
science teams performed a data quality and availability anal-
ysis on process variables, such as feed material properties,
air flows, wash water, froth/pulp level, and pulp feed flow
to the column cells, to assess the viability of implementing
the Flotation Optimization App. The analysis informs the
team and the client about faulty sensors and the instrumen-
tation maturity level of the plant. The main conclusion was
that, with certain assumptions, the data quality and avail-
ability were enough to carry on with the configuration of
the Hydrodynamic and Kinetic Virtual Sensors.
Following a path of incremental value capture for
operators and metallurgists, the Flotation Optimization
App is configured and implemented in sequential stages
of increasing value to operations. The initial stage consists
in deploying the App’s real-time Hydrodynamic Virtual
Sensors to provide metallurgists with vital parameters like
bubble diameter, bubble surface area flux, gas holdup for
every flotation cell in the circuit. This allows, for example,
to visualise in real-time profiles of bubble surface area flux
across banks of cells as shown in Figure 4.
Utilising historical process data plus historical labo-
ratory data for concentrate and tails grade the Flotation
Optimization App’s Kinetic Virtual Sensors can be cali-
brated. Implementation of these metrics allows the Flotation
Optimization App to act as a Digital Process Model of the
circuit, providing real-time information to operators and
metallurgists about mass pull, recovery and grade for every
cell in the circuit. Additionally, this information is aggre-
gated to provide recovery and grade for rougher, scaven-
ger, and any other area. The information provided by these
Virtual Sensors, plus already existing operating parameters
and reagent addition are used to inform the ML models
developed to predict laboratory analyses (elemental com-
position). These models are used for both: 1) to provide the
operation with laboratory results ahead of time, and 2) to
enable reagent optimization.
Once the above model outputs were validated with the
client, using historical and live data, the next step was to
Figure 3. Flotation Optimizer screen for operators. Performance gains are summarised to highlight the benefits of following
the recommended setpoints
will highlight the required setpoint changes that operators
should consider.
CASE-STUDY
A 10+ Mtpa flotation circuit of a South American iron ore
mine is shown as a case study in this paper. Note that our
Flotation Optimizer has been commissioned for other com-
modities as well. The circuit consists of four reverse-flotation
stages: rougher, scavenger, cleaner and re-cleaner. The whole
flotation circuit consists of two parallel lines of 11 flotation
columns (5-2-2-2) where three reagents are used to separate
quartz and other silicates from iron oxide-bearing particles.
Following our standard configuration and commissioning
procedures, the mining technology engineering and data
science teams performed a data quality and availability anal-
ysis on process variables, such as feed material properties,
air flows, wash water, froth/pulp level, and pulp feed flow
to the column cells, to assess the viability of implementing
the Flotation Optimization App. The analysis informs the
team and the client about faulty sensors and the instrumen-
tation maturity level of the plant. The main conclusion was
that, with certain assumptions, the data quality and avail-
ability were enough to carry on with the configuration of
the Hydrodynamic and Kinetic Virtual Sensors.
Following a path of incremental value capture for
operators and metallurgists, the Flotation Optimization
App is configured and implemented in sequential stages
of increasing value to operations. The initial stage consists
in deploying the App’s real-time Hydrodynamic Virtual
Sensors to provide metallurgists with vital parameters like
bubble diameter, bubble surface area flux, gas holdup for
every flotation cell in the circuit. This allows, for example,
to visualise in real-time profiles of bubble surface area flux
across banks of cells as shown in Figure 4.
Utilising historical process data plus historical labo-
ratory data for concentrate and tails grade the Flotation
Optimization App’s Kinetic Virtual Sensors can be cali-
brated. Implementation of these metrics allows the Flotation
Optimization App to act as a Digital Process Model of the
circuit, providing real-time information to operators and
metallurgists about mass pull, recovery and grade for every
cell in the circuit. Additionally, this information is aggre-
gated to provide recovery and grade for rougher, scaven-
ger, and any other area. The information provided by these
Virtual Sensors, plus already existing operating parameters
and reagent addition are used to inform the ML models
developed to predict laboratory analyses (elemental com-
position). These models are used for both: 1) to provide the
operation with laboratory results ahead of time, and 2) to
enable reagent optimization.
Once the above model outputs were validated with the
client, using historical and live data, the next step was to
Figure 3. Flotation Optimizer screen for operators. Performance gains are summarised to highlight the benefits of following
the recommended setpoints