2
capacity) at 67% solids. The grinding conditions were
selected after a series of kinetic grinding tests, with the par-
ticle sizes of the flotation feed corresponding to a P80 of
225 µm. The batch flotation tests were carried out using
a Denver 2.3 L flotation cell at 1200 rpm and 6 L/min air
flow rate. Conditioning time was 3 min, and flotation time
was 11 min pH was adjusted using lime to 10.5.
Flotation Reagents
The flotation experiments involved the use of 13 differ-
ent collectors from different chemical families in Clariant’s
product portfolio, with Flotanol M (MIBC) serving as a
standard frother. In each test, both the collector and the
frother were maintained at a constant dosage of 18 and 13
g/t, respectively. The specific commercial names of these
collectors are omitted from this document to maintain
confidentiality.
Sieving Analysis and Chemical Assay of Flotation
Tailings
Sieving analysis of the flotation tailings was carried out
using standard ASTM sieves for sizes –75 μm, +75 μm,
+150 μm, and +210 μm. The mass fractions obtained were
sent for chemical assay to later determine the Cu distribu-
tion by size.
Dataset
For the statistical analysis conducted in this study, the
thirteen collectors were considered as observations, treat-
ing the copper recovery and copper distribution across size
fractions of –75 μm, +75 μm, +150 μm, and +210 μm as
variables.
Principal Components Analysis (PCA)
The principal components analysis (PCA) was employed
to reduce the dimensionality of the dataset. PCA simpli-
fies and analyses complex datasets by transforming origi-
nal variables into uncorrelated variables known as principal
components (Jolliffe, 2002). It aims to reduce data dimen-
sionality while retaining crucial information for pattern
visualization, relationship identification, and outlier detec-
tion. Natarajan, Nirdosh, Basak, and Mills (2002) applied
PCA to streamline a dataset of flotation collectors’ topolog-
ical indices, successfully pinpointing essential features and
developing accurate regression models for predicting col-
lector efficiencies. Similarly, Moreno, Bournival, and Ata
(2022) used PCA to simplify frother characteristics data,
reducing complexity, and identifying important variables
for frother classification and selection, thereby enhancing
analysis and interpretation efficiency.
Hierarchical Cluster Analysis (HCA)
The hierarchical clustering analysis (HCA) was imple-
mented on the principal components to group collectors
based on similarities and present a collector classification.
HCA is a method that organizes samples into groups, illus-
trating hierarchy (Lee &Yang, 2009). Bournival and Ata
(2021) employed hierarchical clustering with dynamic
time warping (DTW) distance to assess yield–ash curves
across laboratories, distinguishing coal samples and com-
paring factors such as flotation reagents and particle sizes.
Moreno, Bournival, and Ata (2022) combined HCA with
PCA to comprehensively understand frother behavior and
properties, enhancing classification beyond PCA’s dimen-
sionality reduction. This integrated approach visually
represented frother relationships, aiding in selection and
offering insights into behavior and properties, ultimately
providing detailed frother characteristic classification and
understanding.
SME will be working directly from your electronic file
(also include a printed copy of your final paper). Your abil-
ity to follow the general guidelines and specific styles will
result in a more professional-looking finished paper in the
proceedings.
RESULTS AND DISCUSSION
Characterization of Flotation Feed
Chemical Assays
The chemical assay of the Cu-Mo ore sample used in
the batch flotation tests showed 0.80% Cu, 6.87% Fe, and
0.02% Mo.
Particle Size Distribution
The particle size distribution of the flotation feed shows that
50 wt% of particles are smaller than 96 μm (P50), with 80
wt% smaller than 225 μm (P80). Approximately 14 wt%
of the mass consists of particles larger than 210 μm, while
about 43 wt% are smaller than 75 μm. Figure 1 illustrates
this distribution.
Mineralogical Characterization
The mineralogical characterization of the flotation feed
was performed using QEMSCAN. Figure 2(a) shows the
distribution of the sulphides minerals in the sample, with
pyrite (FeS2) being the predominant sulphide at 83.09%,
followed by chalcopyrite (CuFeS2) at 15.95%. Other sul-
phides and minerals are present in much smaller quantities.
Figure 2(b) shows the distribution of chalcopyrite (CuFeS2)
particles based on their degree of liberation. Most of the
capacity) at 67% solids. The grinding conditions were
selected after a series of kinetic grinding tests, with the par-
ticle sizes of the flotation feed corresponding to a P80 of
225 µm. The batch flotation tests were carried out using
a Denver 2.3 L flotation cell at 1200 rpm and 6 L/min air
flow rate. Conditioning time was 3 min, and flotation time
was 11 min pH was adjusted using lime to 10.5.
Flotation Reagents
The flotation experiments involved the use of 13 differ-
ent collectors from different chemical families in Clariant’s
product portfolio, with Flotanol M (MIBC) serving as a
standard frother. In each test, both the collector and the
frother were maintained at a constant dosage of 18 and 13
g/t, respectively. The specific commercial names of these
collectors are omitted from this document to maintain
confidentiality.
Sieving Analysis and Chemical Assay of Flotation
Tailings
Sieving analysis of the flotation tailings was carried out
using standard ASTM sieves for sizes –75 μm, +75 μm,
+150 μm, and +210 μm. The mass fractions obtained were
sent for chemical assay to later determine the Cu distribu-
tion by size.
Dataset
For the statistical analysis conducted in this study, the
thirteen collectors were considered as observations, treat-
ing the copper recovery and copper distribution across size
fractions of –75 μm, +75 μm, +150 μm, and +210 μm as
variables.
Principal Components Analysis (PCA)
The principal components analysis (PCA) was employed
to reduce the dimensionality of the dataset. PCA simpli-
fies and analyses complex datasets by transforming origi-
nal variables into uncorrelated variables known as principal
components (Jolliffe, 2002). It aims to reduce data dimen-
sionality while retaining crucial information for pattern
visualization, relationship identification, and outlier detec-
tion. Natarajan, Nirdosh, Basak, and Mills (2002) applied
PCA to streamline a dataset of flotation collectors’ topolog-
ical indices, successfully pinpointing essential features and
developing accurate regression models for predicting col-
lector efficiencies. Similarly, Moreno, Bournival, and Ata
(2022) used PCA to simplify frother characteristics data,
reducing complexity, and identifying important variables
for frother classification and selection, thereby enhancing
analysis and interpretation efficiency.
Hierarchical Cluster Analysis (HCA)
The hierarchical clustering analysis (HCA) was imple-
mented on the principal components to group collectors
based on similarities and present a collector classification.
HCA is a method that organizes samples into groups, illus-
trating hierarchy (Lee &Yang, 2009). Bournival and Ata
(2021) employed hierarchical clustering with dynamic
time warping (DTW) distance to assess yield–ash curves
across laboratories, distinguishing coal samples and com-
paring factors such as flotation reagents and particle sizes.
Moreno, Bournival, and Ata (2022) combined HCA with
PCA to comprehensively understand frother behavior and
properties, enhancing classification beyond PCA’s dimen-
sionality reduction. This integrated approach visually
represented frother relationships, aiding in selection and
offering insights into behavior and properties, ultimately
providing detailed frother characteristic classification and
understanding.
SME will be working directly from your electronic file
(also include a printed copy of your final paper). Your abil-
ity to follow the general guidelines and specific styles will
result in a more professional-looking finished paper in the
proceedings.
RESULTS AND DISCUSSION
Characterization of Flotation Feed
Chemical Assays
The chemical assay of the Cu-Mo ore sample used in
the batch flotation tests showed 0.80% Cu, 6.87% Fe, and
0.02% Mo.
Particle Size Distribution
The particle size distribution of the flotation feed shows that
50 wt% of particles are smaller than 96 μm (P50), with 80
wt% smaller than 225 μm (P80). Approximately 14 wt%
of the mass consists of particles larger than 210 μm, while
about 43 wt% are smaller than 75 μm. Figure 1 illustrates
this distribution.
Mineralogical Characterization
The mineralogical characterization of the flotation feed
was performed using QEMSCAN. Figure 2(a) shows the
distribution of the sulphides minerals in the sample, with
pyrite (FeS2) being the predominant sulphide at 83.09%,
followed by chalcopyrite (CuFeS2) at 15.95%. Other sul-
phides and minerals are present in much smaller quantities.
Figure 2(b) shows the distribution of chalcopyrite (CuFeS2)
particles based on their degree of liberation. Most of the