XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 1121
flocculated correctly to maximize the copper production
rate as close or better than the plant schedule daily targets.
All violations are reported and the time series data is
aggregated at the desired degree of details using a modern
process historian and advanced analytics tool. It is impos-
sible for people to do this task, so automating many of the
traditional exercises used in a typical spreadsheet should be
made. This strategy is sometimes difficult to understand.
The computer is trained to do many of these process engi-
neering calculations (Bascur, et al., 2020, Concha and
Bascur, 2024).
The creation of the Digital Twin Plant Model starts
with a good up to date process flow diagram of the plant,
the best way of communicating the process knowledge
(Turton et al., 2018) for the creation of valid predictive
operational models. The process variables and key process
controllers should be shown to build on these variables the
necessary process mass balances and on-line state simula-
tion models of the plant. In this way a Digital Twin is cre-
ated by an online observer predicting the outcomes from
the operational state and the process variables.
DYNAMIC GRINDING CLASSIFICATION
CIRCUIT MODEL
To maximize the metal production, it is a top priority to
maintain the desired throughput with the right cut size
and the proper particle size distribution to maximize the
rougher flotation recovery with a control strategy able to
absorb the disturbances caused by the variability of the ore.
The size distribution soft sensors, in conjunction with dual
measurement sensors and the water addition to the sump,
in agreement to its availability, have been used to main-
tain the particle size in control. A selected algorithm can be
used in practice to generate the right grind cut index P80
and particle size distribution shape (PSDS). The PSDS can
be represented by the Gaudin-Schumann model via the M
parameter (F3(d) =(D/Dmax)^M as shown in Figure 6.
The target is to maintain a stable product to flotation with
the right particle size distribution for the best attachment of
the hydrophobic particle with the air bubbles in the flota-
tion bank. In the past, grinding and flotation was treated
independently, however today they are tied together.
The hydrocyclone feed sump hydrodynamics plays an
important stabilizing action in a ball-mill circuit. This pro-
cess element is carefully included in the off-line and real
time simulator to achieve the best hydrocyclone efficiency.
The large circulating load and the sump water addition can
create instabilities in the circuits. Dynamic grinding cir-
cuits simulators assist the design of the proper best com-
binations of hydrocyclone, pump size and sump level for a
stable operation of a grinding circuit (Bascur et al., 1986
Bascur and Junge 2011). Figures 6 shows comminution
system with the key measurements for particle size and the
estimation of particle size distribution shape using in, this
Figure 6. Mineral processing plant grinding and flotation predictive model (updated Dec 2023)
flocculated correctly to maximize the copper production
rate as close or better than the plant schedule daily targets.
All violations are reported and the time series data is
aggregated at the desired degree of details using a modern
process historian and advanced analytics tool. It is impos-
sible for people to do this task, so automating many of the
traditional exercises used in a typical spreadsheet should be
made. This strategy is sometimes difficult to understand.
The computer is trained to do many of these process engi-
neering calculations (Bascur, et al., 2020, Concha and
Bascur, 2024).
The creation of the Digital Twin Plant Model starts
with a good up to date process flow diagram of the plant,
the best way of communicating the process knowledge
(Turton et al., 2018) for the creation of valid predictive
operational models. The process variables and key process
controllers should be shown to build on these variables the
necessary process mass balances and on-line state simula-
tion models of the plant. In this way a Digital Twin is cre-
ated by an online observer predicting the outcomes from
the operational state and the process variables.
DYNAMIC GRINDING CLASSIFICATION
CIRCUIT MODEL
To maximize the metal production, it is a top priority to
maintain the desired throughput with the right cut size
and the proper particle size distribution to maximize the
rougher flotation recovery with a control strategy able to
absorb the disturbances caused by the variability of the ore.
The size distribution soft sensors, in conjunction with dual
measurement sensors and the water addition to the sump,
in agreement to its availability, have been used to main-
tain the particle size in control. A selected algorithm can be
used in practice to generate the right grind cut index P80
and particle size distribution shape (PSDS). The PSDS can
be represented by the Gaudin-Schumann model via the M
parameter (F3(d) =(D/Dmax)^M as shown in Figure 6.
The target is to maintain a stable product to flotation with
the right particle size distribution for the best attachment of
the hydrophobic particle with the air bubbles in the flota-
tion bank. In the past, grinding and flotation was treated
independently, however today they are tied together.
The hydrocyclone feed sump hydrodynamics plays an
important stabilizing action in a ball-mill circuit. This pro-
cess element is carefully included in the off-line and real
time simulator to achieve the best hydrocyclone efficiency.
The large circulating load and the sump water addition can
create instabilities in the circuits. Dynamic grinding cir-
cuits simulators assist the design of the proper best com-
binations of hydrocyclone, pump size and sump level for a
stable operation of a grinding circuit (Bascur et al., 1986
Bascur and Junge 2011). Figures 6 shows comminution
system with the key measurements for particle size and the
estimation of particle size distribution shape using in, this
Figure 6. Mineral processing plant grinding and flotation predictive model (updated Dec 2023)