XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 1565
The energy requirement of representative samples from
the ABO and HZT ores can be explained through a linear
regression equation, which considers G1 and G2+G3 as
variables. The model exhibits an adjusted R2
(aj) fit of 68%,
with a predicted R2 of 66%.
Energy
requirement (kwh/t)
..02268
.03246^G
g1
G3h
0 800 0
0 2 2 =
+
++
=G (3)
The predictive model for SiO2 content in the concentrate
was individually tailored for the Abóboras and Horizontes
mineral deposits. For the Abóboras deposit, a regression
tree (refer to Figure 7) was constructed utilizing the vari-
ables X[0]= (G2+G3) and X[1]= PPC. Regression Trees,
also known as CART (Classification and Regression Trees),
represent a non-parametric supervised machine learning
approach utilized for decision tree generation (Breiman et
al., 1984).
For the Horizontes mineral deposit, a multivariate
regression equation was formulated using the variables Fe,
G1 and PPC, yielding a R2
(aj) fit of 58% and a predicted
R2 of 52%.
Figure 6. Mass recovery for desliming, flotation, global, Fe recovery and SiO2 grade in concentrate
The energy requirement of representative samples from
the ABO and HZT ores can be explained through a linear
regression equation, which considers G1 and G2+G3 as
variables. The model exhibits an adjusted R2
(aj) fit of 68%,
with a predicted R2 of 66%.
Energy
requirement (kwh/t)
..02268
.03246^G
g1
G3h
0 800 0
0 2 2 =
+
++
=G (3)
The predictive model for SiO2 content in the concentrate
was individually tailored for the Abóboras and Horizontes
mineral deposits. For the Abóboras deposit, a regression
tree (refer to Figure 7) was constructed utilizing the vari-
ables X[0]= (G2+G3) and X[1]= PPC. Regression Trees,
also known as CART (Classification and Regression Trees),
represent a non-parametric supervised machine learning
approach utilized for decision tree generation (Breiman et
al., 1984).
For the Horizontes mineral deposit, a multivariate
regression equation was formulated using the variables Fe,
G1 and PPC, yielding a R2
(aj) fit of 58% and a predicted
R2 of 52%.
Figure 6. Mass recovery for desliming, flotation, global, Fe recovery and SiO2 grade in concentrate