1122 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
case, the Gaudin-Schumann equation to track the number
of fines and coarse particles to include their effect in the
rheological model (Bascur, 1991b).
DYNAMIC HYDRODYNAMIC FLOTATION
MODEL
Flotation dynamic models were developed by Bascur
(1982), Bascur and Herbst (1986), Bascur (2005, 2013).
More recently by Bascur and Junge (2011) implemented
Dynaflote as dynamic interactive model that integrates
Bascur’ the hydrodynamic model with the particle popula-
tion balance described by Mika and Fuerstenau (1968) and
Fuerstenau (1999) using the turbulent aggregate velocity
proposed by Liepe in Bischofberger and Schubert (1978)
and Schuber and Bischofberger (1978, 1979). Figure 7
shows the Dynaflote simulator having the pulp and froth
phases describing the mayor flotation subprocesses in
terms of the hydrodynamics (power, volumes, air flow rate,
frother and physicochemical regimes).
The flotation bank dynamics are shown in Figure 8,
which, for a given mineral type, maximizes the metal pro-
duction rate. In this particular case, the first cells operate to
recover all fast coarse particles, the second set of flotation
cells are set to recover the fines particles and the last cells
to recover the slow floating particles. Having the rougher
metal assays, particle size P80, PSDS (M), estimated air
hold up cell profile, flotation cell power, froth height, feed
flow, and reagent flows a predictive model is derived using
machine learning tools.
Figure 8 shows Copper Production Rate predictive
model excellent results showing the deeper understanding
on how to the set grinding/flotation and thickening operat-
ing variables to maximize the Copper Production Rate. The
machine learning tools use the real-time data cleansed data
set using the running Ok operating model to predict the
production rate using a multiple linear regression.
The disturbances manipulated and observed variables
are used to find the best set of operating condition for a
given ore type. The most important result is the online Profit
Gain or Lost comparing the actual production results with
the planning production targets. The Net Metal Production
Rate is plotted with the Target as you can the plant is able
to follow the plant very closely and actually achieve greater
results (Current 12 TPH is the Target with a 15 Tph result.
This is due to highly variable ore which can produce more
than the planned target. Operations increase up to 6 TPH
more copper from the 15 TPH targets. The water was a
critical issue until they started to produce the right particle
size distribution shape above M 0.37 as suggested by the
Digital Twin analysis. As you can see in Figure 8 the GAIN
and LOSSES are estimated in real time to define the opti-
mal mill throughput and water additions which provides
the best particle size distribution shape (M) and not the
P80 SIZE. The size is not as important as the creation of
fines which are difficult be floated or flocculated using the
traditional installed equipment in the plant. The fines with
the valuable metal content easily lost to the tailings and the
Figure 7. Operational dynamic flotation models summary and rougher flotation bank performance monitoring
case, the Gaudin-Schumann equation to track the number
of fines and coarse particles to include their effect in the
rheological model (Bascur, 1991b).
DYNAMIC HYDRODYNAMIC FLOTATION
MODEL
Flotation dynamic models were developed by Bascur
(1982), Bascur and Herbst (1986), Bascur (2005, 2013).
More recently by Bascur and Junge (2011) implemented
Dynaflote as dynamic interactive model that integrates
Bascur’ the hydrodynamic model with the particle popula-
tion balance described by Mika and Fuerstenau (1968) and
Fuerstenau (1999) using the turbulent aggregate velocity
proposed by Liepe in Bischofberger and Schubert (1978)
and Schuber and Bischofberger (1978, 1979). Figure 7
shows the Dynaflote simulator having the pulp and froth
phases describing the mayor flotation subprocesses in
terms of the hydrodynamics (power, volumes, air flow rate,
frother and physicochemical regimes).
The flotation bank dynamics are shown in Figure 8,
which, for a given mineral type, maximizes the metal pro-
duction rate. In this particular case, the first cells operate to
recover all fast coarse particles, the second set of flotation
cells are set to recover the fines particles and the last cells
to recover the slow floating particles. Having the rougher
metal assays, particle size P80, PSDS (M), estimated air
hold up cell profile, flotation cell power, froth height, feed
flow, and reagent flows a predictive model is derived using
machine learning tools.
Figure 8 shows Copper Production Rate predictive
model excellent results showing the deeper understanding
on how to the set grinding/flotation and thickening operat-
ing variables to maximize the Copper Production Rate. The
machine learning tools use the real-time data cleansed data
set using the running Ok operating model to predict the
production rate using a multiple linear regression.
The disturbances manipulated and observed variables
are used to find the best set of operating condition for a
given ore type. The most important result is the online Profit
Gain or Lost comparing the actual production results with
the planning production targets. The Net Metal Production
Rate is plotted with the Target as you can the plant is able
to follow the plant very closely and actually achieve greater
results (Current 12 TPH is the Target with a 15 Tph result.
This is due to highly variable ore which can produce more
than the planned target. Operations increase up to 6 TPH
more copper from the 15 TPH targets. The water was a
critical issue until they started to produce the right particle
size distribution shape above M 0.37 as suggested by the
Digital Twin analysis. As you can see in Figure 8 the GAIN
and LOSSES are estimated in real time to define the opti-
mal mill throughput and water additions which provides
the best particle size distribution shape (M) and not the
P80 SIZE. The size is not as important as the creation of
fines which are difficult be floated or flocculated using the
traditional installed equipment in the plant. The fines with
the valuable metal content easily lost to the tailings and the
Figure 7. Operational dynamic flotation models summary and rougher flotation bank performance monitoring