XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 937
Figure 5 shows the independent variable selections for
Figure 4. It is important to note that different bubble tex-
tures, different air flowrates, and different mill feed Cu val-
ues reveal distinct contours in the multidimensional model
form. Air impacts flotation hydrodynamics deep in cell.
This highly dynamic behavior is indicative of the remark-
able complexity of flotation.
Copper in Lead Concentrate—Introducing Neural
Network and Decision Tree Modeling
Cu/Pb separation is sensitive to the ratio of the elements
in the mill feed, as shown in Figures 6 and 7 by fitting
mill feed assays (percent Pb/Zn/Cu/Fe) to the Cu% in the
Pb Concentrate data using neural network (NN) modeling
method. Figures 6 and 7 show NN modeling of Cu in Pb
concentrate versus Feed Cu% and Feed Fe% and versus
Feed Cu% and Feed Pb% respectively, with the NN model
architecture shown in Figure 8.
MULTIVARIATE DATA ANALYSIS AND
MODELING
Bubble Texture -Introducing Response Surface
Modeling (RSM) without factor interactions
Figures 4 and 5 first introduce the linear RSM, with Cu
in Pb concentrate as response and Cu in mill feed, bub-
ble textures and air as factors but without factor interac-
tions. String factors from PI database are used to best
trace and indicate meaning of selected factors. The bubble
texture is a method of determining repeating patterns of
froth types to correlate the effectiveness of mineral load-
ing with the reagents and concentrate quality (Mang et al.,
2024a). Bubble texture directly measures mineral loading
and Figure 4 shows how Cu deports to the Pb concentrate
based on the bubble texture and mill feed Cu grade. Higher
bubble textures correlate directly with less Cu going to the
Pb concentrate, which in turn implies better Cu recovery.
Figure 4. Cu in the Pb concentrate in the Cu/Pb separation
versus bubble texture in the first rougher and mill feed Cu
grade
Figure 5. Contour plot variables for Figure 4
Figure 6. Cu% in the Pb Concentrate vs Feed Cu% and Feed
Fe%. In this contour, the Feed Pb and Feed Zn are fixed at
2.59% and 0.56% respectively
Figure 5 shows the independent variable selections for
Figure 4. It is important to note that different bubble tex-
tures, different air flowrates, and different mill feed Cu val-
ues reveal distinct contours in the multidimensional model
form. Air impacts flotation hydrodynamics deep in cell.
This highly dynamic behavior is indicative of the remark-
able complexity of flotation.
Copper in Lead Concentrate—Introducing Neural
Network and Decision Tree Modeling
Cu/Pb separation is sensitive to the ratio of the elements
in the mill feed, as shown in Figures 6 and 7 by fitting
mill feed assays (percent Pb/Zn/Cu/Fe) to the Cu% in the
Pb Concentrate data using neural network (NN) modeling
method. Figures 6 and 7 show NN modeling of Cu in Pb
concentrate versus Feed Cu% and Feed Fe% and versus
Feed Cu% and Feed Pb% respectively, with the NN model
architecture shown in Figure 8.
MULTIVARIATE DATA ANALYSIS AND
MODELING
Bubble Texture -Introducing Response Surface
Modeling (RSM) without factor interactions
Figures 4 and 5 first introduce the linear RSM, with Cu
in Pb concentrate as response and Cu in mill feed, bub-
ble textures and air as factors but without factor interac-
tions. String factors from PI database are used to best
trace and indicate meaning of selected factors. The bubble
texture is a method of determining repeating patterns of
froth types to correlate the effectiveness of mineral load-
ing with the reagents and concentrate quality (Mang et al.,
2024a). Bubble texture directly measures mineral loading
and Figure 4 shows how Cu deports to the Pb concentrate
based on the bubble texture and mill feed Cu grade. Higher
bubble textures correlate directly with less Cu going to the
Pb concentrate, which in turn implies better Cu recovery.
Figure 4. Cu in the Pb concentrate in the Cu/Pb separation
versus bubble texture in the first rougher and mill feed Cu
grade
Figure 5. Contour plot variables for Figure 4
Figure 6. Cu% in the Pb Concentrate vs Feed Cu% and Feed
Fe%. In this contour, the Feed Pb and Feed Zn are fixed at
2.59% and 0.56% respectively