XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 3977
any mechanical, electrical, and other losses in the system.
Corem’s HPGR draws 40.7 kW of power when operating at
set speed without any ore in the machine (“Empty Power”
in the tables). This is an indication of all mechanical and
electrical losses in the machine and its power train. If we
subtract the empty power draw from the operating power,
the net power draw would be 85 kW. Although this power
more accurately represents the power required to process
the ore in HPGR, some additional frictional losses, when
operating under a load, beyond 40.7 kW of running empty
are not accounted for in the modelled power calculation.
Comparing the DEM power draw prediction with the net
power draw from the experiment, the relative error reduces
to below 25% for the HPGR and still relates to a simula-
tion that doesn’t account for the fine particles in the feed.
Similar to HPGR, the CAHM machine has mechani-
cal, electrical, and other power losses that are not predicted
by the DEM simulation, although these losses are much
smaller than the HPGR machine. Given the power draw
under prediction, the CAHM simulations predict the net
power draw with a relative error of –2.75%, –33.6%, and
31.6% for FG06, FG07-A, and FG08 respectively.
The FG-08 experimental test started with stable opera-
tion but after a few seconds, there was a buildup forming
in the machine due to a high feed rate of 63.1 t/h, leading
to a halt of the test’s execution. The DEM simulation pre-
dicted the same buildup in the mill for equivalent operat-
ing conditions, hence the low modelled throughput of 47.6
t/h. This buildup makes the calculated specific energy for
this experimental test unreliable. This highlights the impor-
tance of considering feed rate and machine capacity in the
design and operation of comminution machinery and the
fact that DEM simulation is a powerful tool for predicting
operating performance and limitations.
At this point in the test program, a comprehensive,
quantitative analysis of the relative performance predic-
tions has not been completed but a qualitative review of
the data suggests that “successful” tests (those completed
without operational upset or other interruptions) follow
the simulated trends for both energy and product particle
size distributions.
DEM simulations of tumbling mills have demon-
strated a high correlation to actual mills (Powell et al. 2011
Tavares &de Carvalho, 2009) but they still require consid-
erable time and computing resources. As presented earlier
here, most MonoRoll DEM simulations were conducted
on consecutive mill slices with a minimum fragmentation
size set to 1000 µm to reduce the number of particles in
consideration and the computation demand. Here we com-
pare the results of one full mill simulation with experimen-
tal results. Even considering the truncated feed and product
size distributions along with the very low mill holdup
(compared to a ball mill), the simulation required almost 3
months to simulate 20s of MonoRoll operation. A full ball
mill simulation was not completed but a single slice simu-
lation was used at 35% total fill with 26% balls to predict
a power draw of 14.25 kW. The test data for the ball mill
benchmarking tests ranged between 13.6 and 14.4 kW.
Table 6 presents the MonoRoll DEM simulation
and experimental test results. It should be noted that the
MonoRoll simulation was conducted with a feed rate of 1.8
t/h and a mill speed of 38 rpm, while the presented experi-
ment had a feed rate of 1.5 t/h and a mill speed of 35 rpm.
The feed size distribution is also slightly different with the
simulation having sharper size distribution curve (more
coarse material but a smaller F50). As discussed above for
CAHM and HPGR power predictions, not including the
mechanical and electrical losses, MonoRoll power is under-
predicted and the P50 is over-predicted by the simulation
as it doesn’t account for the actual energy consumed to
break material below the threshold size (1000 µm). Having
said that, the results are similar and consistent and clearly
a step change from the ball mill baseline. Again, a compre-
hensive review of relative performance predictions has not
been completed here but using the slice simulations was
very effective in guiding the testing program.
CONCLUSION AND FUTURE WORK
In this paper, we have presented the Conjugate Anvil
Hammer Mill (CAHM) technology platform and its two
prototype machines, CAHM and MonoRoll. We have also
discussed how the Discrete Element Method (DEM) mod-
eling has been utilized as a powerful tool in their develop-
ment and design. High-Pressure Grinding Rolls (HPGRs)
were chosen as the reference equipment for both calibrating
DEM models and benchmarking CAHM’s performance.
Table 6. Summary of MonoRoll simulation and experimental results
MonoRoll
Feed Rate,
t/h
Rotational
Speed, rpm
Average
Power, kW
Average
F80, µm
Average
F50, µm
Average
P80, µm
Average
P50, µm
Specific
Energy,
kWh/t
DEM 1.8 38 3.7 6300 4700 3170 2560 2.1
Experiment 1.5 35 5.48 7052 3419 5700 1140 3.7
any mechanical, electrical, and other losses in the system.
Corem’s HPGR draws 40.7 kW of power when operating at
set speed without any ore in the machine (“Empty Power”
in the tables). This is an indication of all mechanical and
electrical losses in the machine and its power train. If we
subtract the empty power draw from the operating power,
the net power draw would be 85 kW. Although this power
more accurately represents the power required to process
the ore in HPGR, some additional frictional losses, when
operating under a load, beyond 40.7 kW of running empty
are not accounted for in the modelled power calculation.
Comparing the DEM power draw prediction with the net
power draw from the experiment, the relative error reduces
to below 25% for the HPGR and still relates to a simula-
tion that doesn’t account for the fine particles in the feed.
Similar to HPGR, the CAHM machine has mechani-
cal, electrical, and other power losses that are not predicted
by the DEM simulation, although these losses are much
smaller than the HPGR machine. Given the power draw
under prediction, the CAHM simulations predict the net
power draw with a relative error of –2.75%, –33.6%, and
31.6% for FG06, FG07-A, and FG08 respectively.
The FG-08 experimental test started with stable opera-
tion but after a few seconds, there was a buildup forming
in the machine due to a high feed rate of 63.1 t/h, leading
to a halt of the test’s execution. The DEM simulation pre-
dicted the same buildup in the mill for equivalent operat-
ing conditions, hence the low modelled throughput of 47.6
t/h. This buildup makes the calculated specific energy for
this experimental test unreliable. This highlights the impor-
tance of considering feed rate and machine capacity in the
design and operation of comminution machinery and the
fact that DEM simulation is a powerful tool for predicting
operating performance and limitations.
At this point in the test program, a comprehensive,
quantitative analysis of the relative performance predic-
tions has not been completed but a qualitative review of
the data suggests that “successful” tests (those completed
without operational upset or other interruptions) follow
the simulated trends for both energy and product particle
size distributions.
DEM simulations of tumbling mills have demon-
strated a high correlation to actual mills (Powell et al. 2011
Tavares &de Carvalho, 2009) but they still require consid-
erable time and computing resources. As presented earlier
here, most MonoRoll DEM simulations were conducted
on consecutive mill slices with a minimum fragmentation
size set to 1000 µm to reduce the number of particles in
consideration and the computation demand. Here we com-
pare the results of one full mill simulation with experimen-
tal results. Even considering the truncated feed and product
size distributions along with the very low mill holdup
(compared to a ball mill), the simulation required almost 3
months to simulate 20s of MonoRoll operation. A full ball
mill simulation was not completed but a single slice simu-
lation was used at 35% total fill with 26% balls to predict
a power draw of 14.25 kW. The test data for the ball mill
benchmarking tests ranged between 13.6 and 14.4 kW.
Table 6 presents the MonoRoll DEM simulation
and experimental test results. It should be noted that the
MonoRoll simulation was conducted with a feed rate of 1.8
t/h and a mill speed of 38 rpm, while the presented experi-
ment had a feed rate of 1.5 t/h and a mill speed of 35 rpm.
The feed size distribution is also slightly different with the
simulation having sharper size distribution curve (more
coarse material but a smaller F50). As discussed above for
CAHM and HPGR power predictions, not including the
mechanical and electrical losses, MonoRoll power is under-
predicted and the P50 is over-predicted by the simulation
as it doesn’t account for the actual energy consumed to
break material below the threshold size (1000 µm). Having
said that, the results are similar and consistent and clearly
a step change from the ball mill baseline. Again, a compre-
hensive review of relative performance predictions has not
been completed here but using the slice simulations was
very effective in guiding the testing program.
CONCLUSION AND FUTURE WORK
In this paper, we have presented the Conjugate Anvil
Hammer Mill (CAHM) technology platform and its two
prototype machines, CAHM and MonoRoll. We have also
discussed how the Discrete Element Method (DEM) mod-
eling has been utilized as a powerful tool in their develop-
ment and design. High-Pressure Grinding Rolls (HPGRs)
were chosen as the reference equipment for both calibrating
DEM models and benchmarking CAHM’s performance.
Table 6. Summary of MonoRoll simulation and experimental results
MonoRoll
Feed Rate,
t/h
Rotational
Speed, rpm
Average
Power, kW
Average
F80, µm
Average
F50, µm
Average
P80, µm
Average
P50, µm
Specific
Energy,
kWh/t
DEM 1.8 38 3.7 6300 4700 3170 2560 2.1
Experiment 1.5 35 5.48 7052 3419 5700 1140 3.7