XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 3223
(Vanderbruggen et al. 2022 Ruismäki et al. 2020 Wang,
Gaustad, and Babbitt 2016 Velázquez-Martínez et al.
2019). Size-based composition was mainly obtained from
the work by Vanderbruggens et al., (2022), which was
the most comprehensive source. The additional references
were consulted to verify bulk composition and supple-
ment size-based data. The feed was defined as an artifi-
cial mineral with seven fully liberated materials: casing,
Li(Ni0.33Mn0.33Co0.33)O2 (cathode) here identified as
NMC, graphite (anode), Cu, Al, electrolyte salt and plas-
tics, as illustrated in Figure 1a). It is assumed the mate-
rial has already undergone crushing and that no liquids are
present. We ensured that the particle sizes summarized in
Figure 1b) adhered to a Rosin-Rammler distribution, as
required by HSC Sim.
The process illustrated in Figure 2, comprises eight
distinct processing stages, each characterised the type of
process modeled. We implemented the processing stages
as standard elements using the HSC Sim software, which
is part of the HSC Chemistry package (version 10.2.1,
Metso). In terms of materials flow, the process model
features one input stream of 150 t/h, seven intermediate
streams, and five output streams. For every stream, materi-
als flow details, including the mass of graphite and NMC,
its grade and recovery were calculated. The aim was to
achieve the mechanical separation of materials in the final
output streams (labeled in Figure 2 as “Concentrate 2” and
“Tailings 2”).
Next, the pertinent process parameters impacting
materials flow were identified. Each unit stage in the simu-
lated process contained multiple parameters that control
their functionality. Operational conditions, equipment
specifications, and simulation settings are among the type
of available parameters. In the simulations, most of these
were kept fixed at their encoded default values, with the
exception of Bond’s Work Index (Wuschke et al. 2019) in
milling processes and mass magnetic susceptibilities in the
Magnetic Separator, which were sourced from the CRC
Handbook of Chemistry and Physics (Rumble 2022).
Flotation kinetics were set based on experimental data from
(Vanderbruggen et al. 2022). For the purposes of this case
study, only the 10 parameters considered most influential
on materials flow were modified. Table 1 summarizes the
parameters and their role in the unit stage and process. The
lower and upper boundary values were set to allow a suf-
ficient but significant variability over a large range for each
individual unit.
A high-throughput data generation was performed by
randomly selecting 10,000 combinations of variable process
parameters. Simulations resulted in dataset summarizing
10,000 distinct process outcomes. A randomised approach
ensured a comprehensive and unbiased coverage of many
Figure 1. Input feed materials: a) chemical composition in wt. %and (b) cumulative particle size distribution
(a) (b)
Figure 2. The flowsheet of the initial mechanical separation process modeled with HSC Sim software
(Vanderbruggen et al. 2022 Ruismäki et al. 2020 Wang,
Gaustad, and Babbitt 2016 Velázquez-Martínez et al.
2019). Size-based composition was mainly obtained from
the work by Vanderbruggens et al., (2022), which was
the most comprehensive source. The additional references
were consulted to verify bulk composition and supple-
ment size-based data. The feed was defined as an artifi-
cial mineral with seven fully liberated materials: casing,
Li(Ni0.33Mn0.33Co0.33)O2 (cathode) here identified as
NMC, graphite (anode), Cu, Al, electrolyte salt and plas-
tics, as illustrated in Figure 1a). It is assumed the mate-
rial has already undergone crushing and that no liquids are
present. We ensured that the particle sizes summarized in
Figure 1b) adhered to a Rosin-Rammler distribution, as
required by HSC Sim.
The process illustrated in Figure 2, comprises eight
distinct processing stages, each characterised the type of
process modeled. We implemented the processing stages
as standard elements using the HSC Sim software, which
is part of the HSC Chemistry package (version 10.2.1,
Metso). In terms of materials flow, the process model
features one input stream of 150 t/h, seven intermediate
streams, and five output streams. For every stream, materi-
als flow details, including the mass of graphite and NMC,
its grade and recovery were calculated. The aim was to
achieve the mechanical separation of materials in the final
output streams (labeled in Figure 2 as “Concentrate 2” and
“Tailings 2”).
Next, the pertinent process parameters impacting
materials flow were identified. Each unit stage in the simu-
lated process contained multiple parameters that control
their functionality. Operational conditions, equipment
specifications, and simulation settings are among the type
of available parameters. In the simulations, most of these
were kept fixed at their encoded default values, with the
exception of Bond’s Work Index (Wuschke et al. 2019) in
milling processes and mass magnetic susceptibilities in the
Magnetic Separator, which were sourced from the CRC
Handbook of Chemistry and Physics (Rumble 2022).
Flotation kinetics were set based on experimental data from
(Vanderbruggen et al. 2022). For the purposes of this case
study, only the 10 parameters considered most influential
on materials flow were modified. Table 1 summarizes the
parameters and their role in the unit stage and process. The
lower and upper boundary values were set to allow a suf-
ficient but significant variability over a large range for each
individual unit.
A high-throughput data generation was performed by
randomly selecting 10,000 combinations of variable process
parameters. Simulations resulted in dataset summarizing
10,000 distinct process outcomes. A randomised approach
ensured a comprehensive and unbiased coverage of many
Figure 1. Input feed materials: a) chemical composition in wt. %and (b) cumulative particle size distribution
(a) (b)
Figure 2. The flowsheet of the initial mechanical separation process modeled with HSC Sim software