500 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
the quantitative basis for the optimization options that
JMP provides. This experiment had multiple responses
to optimize and as JMP’s support website describes: “The
overall desirability for all responses is defined as the geo-
metric mean of the desirability functions for the individual
responses." For example, in this experiment the response
‘Fe %’ was set to ‘Maximize’ and given the highest rela-
tive importance, because product grade was considered the
most important response. Thus, the desirability of the grade
is described as a linear function with a y-intercept of ~0,
and a slope of ~1.
Figure 4 shows the Prediction Profiler regression model
set to the optimum first pass conditions determined after
the full testing regimen.
The left axis shows the responses, with the average out-
puts in red, and the upper and lower estimates calculated
using the root mean square error (RMSE) for each response.
The bottom axis shows the factors that can be changed and
their respective levels along with the Desirability func-
tions in the far right column. For all the responses, the
Desirability function is a linear function, and Desirability
increases in proportion to the same increase in a response.
The factors can be independently toggled and the responses
will be calculated based on the model. Figure 5 shows the
changes in the responses when the belt speed is changed
from 65 feet per second to 30 feet per second.
The change in belt speed has a significant impact on the
responses: iron enrichment declines while yield improves.
Separator processing conditions can be optimized by
manipulating the various variables using statistical model-
ing as described above. Similar design of experiments and
statistical modeling was conducted on the “air-classified”
iron ore sample.
RESULTS
Ultrafines Iron Ore Separation Results
Different sets of conditions were used for first pass processing
to create multi-stage processing schemes. For Production 1,
a lower belt speed and Feed Port 1 was used to maximize
yield achieving moderate enrichment of Fe. Production 2
used higher belt speed and Feed Port 3 and the polarity was
switched from Top Negative to Top Positive. This set gave
the best first pass results: an increase of 9.61% Fe at 32.8%
yield. These numbers are an average of 8 identical runs per-
formed sequentially. The process conditions and results for
the Production class runs are summarized in Table 5.
Figure 4. Example of effects summary from model
Previous Page Next Page

Extracted Text (may have errors)

500 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
the quantitative basis for the optimization options that
JMP provides. This experiment had multiple responses
to optimize and as JMP’s support website describes: “The
overall desirability for all responses is defined as the geo-
metric mean of the desirability functions for the individual
responses." For example, in this experiment the response
‘Fe %’ was set to ‘Maximize’ and given the highest rela-
tive importance, because product grade was considered the
most important response. Thus, the desirability of the grade
is described as a linear function with a y-intercept of ~0,
and a slope of ~1.
Figure 4 shows the Prediction Profiler regression model
set to the optimum first pass conditions determined after
the full testing regimen.
The left axis shows the responses, with the average out-
puts in red, and the upper and lower estimates calculated
using the root mean square error (RMSE) for each response.
The bottom axis shows the factors that can be changed and
their respective levels along with the Desirability func-
tions in the far right column. For all the responses, the
Desirability function is a linear function, and Desirability
increases in proportion to the same increase in a response.
The factors can be independently toggled and the responses
will be calculated based on the model. Figure 5 shows the
changes in the responses when the belt speed is changed
from 65 feet per second to 30 feet per second.
The change in belt speed has a significant impact on the
responses: iron enrichment declines while yield improves.
Separator processing conditions can be optimized by
manipulating the various variables using statistical model-
ing as described above. Similar design of experiments and
statistical modeling was conducted on the “air-classified”
iron ore sample.
RESULTS
Ultrafines Iron Ore Separation Results
Different sets of conditions were used for first pass processing
to create multi-stage processing schemes. For Production 1,
a lower belt speed and Feed Port 1 was used to maximize
yield achieving moderate enrichment of Fe. Production 2
used higher belt speed and Feed Port 3 and the polarity was
switched from Top Negative to Top Positive. This set gave
the best first pass results: an increase of 9.61% Fe at 32.8%
yield. These numbers are an average of 8 identical runs per-
formed sequentially. The process conditions and results for
the Production class runs are summarized in Table 5.
Figure 4. Example of effects summary from model

Help

loading