1564 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
Horizontes samples, with a median of 6.90 kWh/t com-
pared to 4.80 kWh/t for the ABO samples, as showed in
Figure 5.
During desliming, the highest mass partitions for
underflow were observed in the processing of samples
from the ABO mine, indicating a lower content of slimes.
Conversely, samples from the HZT mine exhibited a higher
percentage of slime. In flotation, the highest mass recover-
ies were achieved in the HZT mine, albeit with higher SiO2
concentrations in the final concentrate, surpassing the tar-
get of 2.5%. Overall, processing results from the Abóboras
mine demonstrated superior performance, characterized
by higher metallic Fe recovery and lower SiO2 content in
the concentrate (Figure 6). Mineralogical characterization
of the flotation concentrates with elevated silica levels con-
firmed the low liberation of quartz and hematite observed
in the flotation feed samples.
Statistical Model
Geometallurgical models were established, focusing exclu-
sively on IC and IF. Although the mineralogy was not
mapped in the drill core database of the deposit, the chemi-
cal and granulometric geometallurgical variables, along
with interrelationships reflect the mineralogical composi-
tion, mineral associations, and textural characteristics of
the representative ore samples studied.
The overall circuit recovery was analyzed in terms of
mass recovery. The regression equation characterizing this
parameter, derived from variability tests, incorporates pre-
dictor variables such as Fe grade, the proportion of G1
(%5.3 mm), and the sample origin, represented by the
categorical variable “Deposit” (assigned a value of 1 for
Abóboras and 0 for Horizontes).
..59 *
.8881
31 3
1
%
0.0839
Global mass recovery Deposit
Fe GL
G1
=-33 +
+
-
^h
(1)
The model’s adequacy is reflected in the adjusted coef-
ficient of determination R2(aj), which stands at 86%, with a
predicted R2 of 85%. The predicted R2 indicates the mod-
el’s ability to forecast responses for new data points. The
positive correlation observed between the initial samples’
Fe content and the observed mass recovery underscores
the significant influence of this variable in elucidating the
model. Conversely, for the “G1” variable, a negative cor-
relation is observed, suggesting lower mass recoveries for
samples with a higher proportion of coarse particles.
The metallurgical recovery can be defined by a regres-
sion equation that incorporates various factors, including
the Fe content, loss on ignition, G1/G4 ratio, proportions
of particle size fractions G2 and G3, and the sample ori-
gin represented by the categoric variable “D3_Deposit.”
The adjusted coefficient of determination R2(aj) for ordi-
nary least squares regression was determined to be 69%,
with a predicted R2 of 67%. Notably, the Fe content of
the sample (Fe GL) emerged as the most influential vari-
able in the model, followed by the combined sum of the
intermediate particle size fractions (G2+G3) and the vari-
able D3_Deposit. Furthermore, the findings indicated that
both the mean and median metallurgical recovery of ABO
exceeded those observed for HZT.
Global
metallurgical
recovery (%)
..24
.149
18 41 3
1
– 3_
– 1.913
0.102^G 2 – 0.3131 1/G4
0,588 3_Deposit *2
D Deposit
Fe GL PPC GL
G3h G
D G3h
=
+
++
++^G
(2)
Figure 5. k parameter and specific consumption energy
Horizontes samples, with a median of 6.90 kWh/t com-
pared to 4.80 kWh/t for the ABO samples, as showed in
Figure 5.
During desliming, the highest mass partitions for
underflow were observed in the processing of samples
from the ABO mine, indicating a lower content of slimes.
Conversely, samples from the HZT mine exhibited a higher
percentage of slime. In flotation, the highest mass recover-
ies were achieved in the HZT mine, albeit with higher SiO2
concentrations in the final concentrate, surpassing the tar-
get of 2.5%. Overall, processing results from the Abóboras
mine demonstrated superior performance, characterized
by higher metallic Fe recovery and lower SiO2 content in
the concentrate (Figure 6). Mineralogical characterization
of the flotation concentrates with elevated silica levels con-
firmed the low liberation of quartz and hematite observed
in the flotation feed samples.
Statistical Model
Geometallurgical models were established, focusing exclu-
sively on IC and IF. Although the mineralogy was not
mapped in the drill core database of the deposit, the chemi-
cal and granulometric geometallurgical variables, along
with interrelationships reflect the mineralogical composi-
tion, mineral associations, and textural characteristics of
the representative ore samples studied.
The overall circuit recovery was analyzed in terms of
mass recovery. The regression equation characterizing this
parameter, derived from variability tests, incorporates pre-
dictor variables such as Fe grade, the proportion of G1
(%5.3 mm), and the sample origin, represented by the
categorical variable “Deposit” (assigned a value of 1 for
Abóboras and 0 for Horizontes).
..59 *
.8881
31 3
1
%
0.0839
Global mass recovery Deposit
Fe GL
G1
=-33 +
+
-
^h
(1)
The model’s adequacy is reflected in the adjusted coef-
ficient of determination R2(aj), which stands at 86%, with a
predicted R2 of 85%. The predicted R2 indicates the mod-
el’s ability to forecast responses for new data points. The
positive correlation observed between the initial samples’
Fe content and the observed mass recovery underscores
the significant influence of this variable in elucidating the
model. Conversely, for the “G1” variable, a negative cor-
relation is observed, suggesting lower mass recoveries for
samples with a higher proportion of coarse particles.
The metallurgical recovery can be defined by a regres-
sion equation that incorporates various factors, including
the Fe content, loss on ignition, G1/G4 ratio, proportions
of particle size fractions G2 and G3, and the sample ori-
gin represented by the categoric variable “D3_Deposit.”
The adjusted coefficient of determination R2(aj) for ordi-
nary least squares regression was determined to be 69%,
with a predicted R2 of 67%. Notably, the Fe content of
the sample (Fe GL) emerged as the most influential vari-
able in the model, followed by the combined sum of the
intermediate particle size fractions (G2+G3) and the vari-
able D3_Deposit. Furthermore, the findings indicated that
both the mean and median metallurgical recovery of ABO
exceeded those observed for HZT.
Global
metallurgical
recovery (%)
..24
.149
18 41 3
1
– 3_
– 1.913
0.102^G 2 – 0.3131 1/G4
0,588 3_Deposit *2
D Deposit
Fe GL PPC GL
G3h G
D G3h
=
+
++
++^G
(2)
Figure 5. k parameter and specific consumption energy