3976 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
reduction ratios are a little lower for the similar product size
distribution) but even this first-generation version of the
CAHM has generated that product size distribution using
significantly less energy. In its best-observed performance,
the CAHM uses ~1/3 of the specific energy consumed in the
HPGR. The SSE calculations at both 75 µm and 150 µm
are also consistent with a finding of some improved energy
efficiency. The SSE calculated for the coarser fragments in
this PSD is dramatically lower for the CAHM. This value
appears to be much more sensitive to the initial feed size
distribution with the best performance of the CAHM to
date requiring ~2/3 the SSE of the HPGR.
DEM MODELING AND EXPERIMENTAL
TESTING RESULTS COMPARISON
Quantitative validation of DEM simulation results can
be achieved through two approaches. The first approach
involves a direct comparison between simulation and
experimental response parameters, followed by the calcu-
lation of relative prediction error. The second approach
involves calculating the relative response values of two or
more experiments and comparing the accuracy of relative
response model predictions. In this section, we present
both approaches for the CAHM and MonoRoll results.
Table 5 presents the DEM and experimental responses
of three CAHM and the HPGR baseline tests. Operating
power and product particle size distributions are the pri-
mary model outputs based on the primary machine condi-
tions (in this case, throughput). The ore characteristics and
operating conditions of the simulations and experiments
were matched as closely as possible.
As evident from Table 5, the DEM simulation of
HPGR predicts throughput and P80 with relative errors of
–9% and –2% respectively. However, the DEM drastically
overpredicts the P50. This over-prediction is a function of
the simulation simplification that truncates the feed PSD
below 6400 µm, which eliminates all fines entering the
machine, and the minimum fragment size simulation set-
ting of 2000 µm, which leads to minimal breakage of par-
ticles finer than 4000 µm in the models. Both limitations
were set for the simulation to reduce computational cost
and execution time. Similar phenomena can be observed
for CAHM simulations overpredicting the P50. Where the
full size distribution is included in the modeling, particu-
larly in tests FG06 and FG07-A, it is clear from the P80
data for the CAHM, that the DEM simulation of the prod-
uct size is excellent with errors below 0.5%. Test FG08 is
discussed in more detail below.
DEM also underpredicts the power draw for the
HPGR and CAHM machines. Given that the DEM
simulation only predicts the power transferred between
geometries and particles, it doesn’t directly account for
Table 4. CAHM and HPGR testing energy efficiency calculations
Test ID
Specific Energy,
kWh/t
Reduction
Ratio, P80
Reduction Ratio,
P50
SSE
75 ,
kWh/t
SSE
150 ,
kWh/t
HPGR Baseline 3.09 2.7 4.4 85 65
FG-06 1.13 2.5 3.8 240 56
FG-07A 1.93 2.5 3.9 214 94
FG-07B 1.4 2.5 3.9 156 68
FG-08 1.08 2.3 3.6 60 40
FG-09A 1.67 2.9 3.8 170 92
Table 5. Summary of HPGR and CAHM simulation and experimental results
Cases Throughput, t/h
Operating
Power, kW
Empty
Power,
kW
Net
Power,
kW
F100,
µm
P50,
µm
P80,
µm
Specific
energy,
kWh/t
HPGR 37.0 63.4 0 63.4 38000 7263 10161 1.714
CAHM FG06 38.8 38.9 0 38.9 38000 6205 8539 1.03
CAHM FG07-A 48.5 58.7 0 58.7 38000 5527 8445 1.21
CAHM FG08 47.6 81.6 0 81.6 38000 6494 8363 1.71
HPGR 40.7 125.7 40.7 85.0 37500 2771 10374 3.09
CAHM FG06 41.5 46.6 6.5 40.1 37500 3029 8509 1.13
CAHM FG07-A 51.4 98.4 10.2 88.2 37500 2826 8418 1.93
CAHM FG08 63.1 68.3 6.3 62.0 37500 3214 8904 1.08
DEM
Experiment
reduction ratios are a little lower for the similar product size
distribution) but even this first-generation version of the
CAHM has generated that product size distribution using
significantly less energy. In its best-observed performance,
the CAHM uses ~1/3 of the specific energy consumed in the
HPGR. The SSE calculations at both 75 µm and 150 µm
are also consistent with a finding of some improved energy
efficiency. The SSE calculated for the coarser fragments in
this PSD is dramatically lower for the CAHM. This value
appears to be much more sensitive to the initial feed size
distribution with the best performance of the CAHM to
date requiring ~2/3 the SSE of the HPGR.
DEM MODELING AND EXPERIMENTAL
TESTING RESULTS COMPARISON
Quantitative validation of DEM simulation results can
be achieved through two approaches. The first approach
involves a direct comparison between simulation and
experimental response parameters, followed by the calcu-
lation of relative prediction error. The second approach
involves calculating the relative response values of two or
more experiments and comparing the accuracy of relative
response model predictions. In this section, we present
both approaches for the CAHM and MonoRoll results.
Table 5 presents the DEM and experimental responses
of three CAHM and the HPGR baseline tests. Operating
power and product particle size distributions are the pri-
mary model outputs based on the primary machine condi-
tions (in this case, throughput). The ore characteristics and
operating conditions of the simulations and experiments
were matched as closely as possible.
As evident from Table 5, the DEM simulation of
HPGR predicts throughput and P80 with relative errors of
–9% and –2% respectively. However, the DEM drastically
overpredicts the P50. This over-prediction is a function of
the simulation simplification that truncates the feed PSD
below 6400 µm, which eliminates all fines entering the
machine, and the minimum fragment size simulation set-
ting of 2000 µm, which leads to minimal breakage of par-
ticles finer than 4000 µm in the models. Both limitations
were set for the simulation to reduce computational cost
and execution time. Similar phenomena can be observed
for CAHM simulations overpredicting the P50. Where the
full size distribution is included in the modeling, particu-
larly in tests FG06 and FG07-A, it is clear from the P80
data for the CAHM, that the DEM simulation of the prod-
uct size is excellent with errors below 0.5%. Test FG08 is
discussed in more detail below.
DEM also underpredicts the power draw for the
HPGR and CAHM machines. Given that the DEM
simulation only predicts the power transferred between
geometries and particles, it doesn’t directly account for
Table 4. CAHM and HPGR testing energy efficiency calculations
Test ID
Specific Energy,
kWh/t
Reduction
Ratio, P80
Reduction Ratio,
P50
SSE
75 ,
kWh/t
SSE
150 ,
kWh/t
HPGR Baseline 3.09 2.7 4.4 85 65
FG-06 1.13 2.5 3.8 240 56
FG-07A 1.93 2.5 3.9 214 94
FG-07B 1.4 2.5 3.9 156 68
FG-08 1.08 2.3 3.6 60 40
FG-09A 1.67 2.9 3.8 170 92
Table 5. Summary of HPGR and CAHM simulation and experimental results
Cases Throughput, t/h
Operating
Power, kW
Empty
Power,
kW
Net
Power,
kW
F100,
µm
P50,
µm
P80,
µm
Specific
energy,
kWh/t
HPGR 37.0 63.4 0 63.4 38000 7263 10161 1.714
CAHM FG06 38.8 38.9 0 38.9 38000 6205 8539 1.03
CAHM FG07-A 48.5 58.7 0 58.7 38000 5527 8445 1.21
CAHM FG08 47.6 81.6 0 81.6 38000 6494 8363 1.71
HPGR 40.7 125.7 40.7 85.0 37500 2771 10374 3.09
CAHM FG06 41.5 46.6 6.5 40.1 37500 3029 8509 1.13
CAHM FG07-A 51.4 98.4 10.2 88.2 37500 2826 8418 1.93
CAHM FG08 63.1 68.3 6.3 62.0 37500 3214 8904 1.08
DEM
Experiment