XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 469
associated with the composition of the material. Thirdly, a
data processor receives and analyzes the information trans-
mitted by the sensor. Finally, based on the comparison
between the measured property and the established thresh-
old, the material is directed either towards further process-
ing or into a waste stream.
The effectiveness of implementing SBS relies on the
heterogeneity of the material, the capability of the sensor to
identify the differences between the specimens, the perfor-
mance of the sensor and the economic feasibility (Li, Klein,
Sun, &Kou, 2020). To effectively develop a sorting system,
it is crucial to select a proper sensor. A wide range of sensing
technologies are available, however, not all the sensors are
appropriate to identify all the materials (Modise, Zungeru,
Mtengi, &Ude, 2020). In this way, a study needs to be
conducted to determine whether the sensor is suitable for
sorting the material of interest. Several studies are avail-
able focusing on the sortability of precious and base met-
als. However, limited information is available regarding the
applicability of SBS to industrial minerals, such as calcite
or quartz. SBS has the potential to enhance the economics
of deposits where the main limitation is the grade material
rather than its recovery, due to its low operating costs.
Initially, the variability of the material was evaluated by
determining the constituent heterogeneity (CH). This met-
ric reflects the relative variance of heterogeneities between
particles within each sample to be compared, thereby indi-
cating the potential for sorting the material based on its
composition (Smith, 2001). Subsequently, a sensor that
aligns with the specific requirements of the material is cho-
sen for optimal performance in the sorting process.
This study is focused on assessing the sortability of
Queguay limestone with the aim of enhancing its calcite
grade. The evaluation encompasses the determination of
the CH value of the sample and the identification of a
suitable sensor. Four sensors were assessed in this process:
x-ray transmission (XRT), x-ray fluorescence (XRF), laser
and colorimeter. Additionally, two distinct approaches were
studied: receiver operating characteristic (ROC) analysis
and a conventional algorithm, specifically, simple linear
regression (SLR). Through ROC analyses, graphs which
serve as a visual aid for evaluating the effectiveness of sen-
sors could be created to select the classifier whose perfor-
mance excel (Fawcett, 2006). The main difference between
these two methods is that ROC uses the sensor response
while through SRL the sensor response is transformed into
a calcite grade to make a decision.
MATERIALS AND METHODS
Sample Preparation and Selection of Specimens
A 70 kg representative sample of crushed limestone was
obtained from a deposit within the Queguay Formation,
which is currently mined for the cement production. To
reduce the mass of the sample, the coning and quartering
method was employed, resulting in a subsample weighing
approximately 35 kg. The subsample was then screened into
three size fractions ranging from 0 to 5.1 cm. The material
with a size greater than 2.5 cm and smaller than 5.1 cm was
designated as fraction C, while the material ranging from
1.9 cm to 2.5 cm was labeled as fraction F (Figure 1). The
fraction with particles smaller than 1.9 cm was considered
unsuitable for sorting evaluation. Subsequently, 100 rocks
were randomly chosen from fractions F and C for further
analysis.
Testing Procedure
Firstly, each specimen was washed with tap water to clean
its surface, preventing that external fine particles interfere
with the measurement procedures.
Sensory testing was conducted using four distinct sen-
sors: XRT, XRF, laser and colorimeter. The measurements
for XRT, XRF, laser and colorimeter were obtained utiliz-
ing the following equipment: a COMEX Lab-Sorter MSX-
400-VL-XR-3D, a Thermo Scientific Niton XL2 analyzer,
a laser of type 650 nm class II (1mW), and a Photovolt
Model 577, respectively.
In the case of XRF, laser, and colorimeter measure-
ments, readings were taken from four faces of each rock of
C fraction, while for those that were part of the F fraction,
measurements were collected from two faces. Regarding
the XRF scans, the equipment was configured to operate in
“mining mode Cu/Zn” with a counting time of 40 seconds.
The color measurements were conducted with the use of a
blue filter. The laser responses were obtained from process-
ing images of each specimen when they were exposed to the
incident laser.
Following the sensor measurements, each specimen
was pulverized to a particle size of 150 μm and subjected
to elemental analysis. The calcite grade was estimated based
on the calcium content of the sample. It was assumed that
the calcium present in the sample is only hosted as cal-
cite. The average calcite grade of two measurements was
reported as the assay rock calcite grade. Additionally, to ver-
ify these results, approximately 1 g of the pulverized sample
was utilized for confirmation through the loss on ignition
(LOI) method. The LOI quantification was performed at
1000 °C.
associated with the composition of the material. Thirdly, a
data processor receives and analyzes the information trans-
mitted by the sensor. Finally, based on the comparison
between the measured property and the established thresh-
old, the material is directed either towards further process-
ing or into a waste stream.
The effectiveness of implementing SBS relies on the
heterogeneity of the material, the capability of the sensor to
identify the differences between the specimens, the perfor-
mance of the sensor and the economic feasibility (Li, Klein,
Sun, &Kou, 2020). To effectively develop a sorting system,
it is crucial to select a proper sensor. A wide range of sensing
technologies are available, however, not all the sensors are
appropriate to identify all the materials (Modise, Zungeru,
Mtengi, &Ude, 2020). In this way, a study needs to be
conducted to determine whether the sensor is suitable for
sorting the material of interest. Several studies are avail-
able focusing on the sortability of precious and base met-
als. However, limited information is available regarding the
applicability of SBS to industrial minerals, such as calcite
or quartz. SBS has the potential to enhance the economics
of deposits where the main limitation is the grade material
rather than its recovery, due to its low operating costs.
Initially, the variability of the material was evaluated by
determining the constituent heterogeneity (CH). This met-
ric reflects the relative variance of heterogeneities between
particles within each sample to be compared, thereby indi-
cating the potential for sorting the material based on its
composition (Smith, 2001). Subsequently, a sensor that
aligns with the specific requirements of the material is cho-
sen for optimal performance in the sorting process.
This study is focused on assessing the sortability of
Queguay limestone with the aim of enhancing its calcite
grade. The evaluation encompasses the determination of
the CH value of the sample and the identification of a
suitable sensor. Four sensors were assessed in this process:
x-ray transmission (XRT), x-ray fluorescence (XRF), laser
and colorimeter. Additionally, two distinct approaches were
studied: receiver operating characteristic (ROC) analysis
and a conventional algorithm, specifically, simple linear
regression (SLR). Through ROC analyses, graphs which
serve as a visual aid for evaluating the effectiveness of sen-
sors could be created to select the classifier whose perfor-
mance excel (Fawcett, 2006). The main difference between
these two methods is that ROC uses the sensor response
while through SRL the sensor response is transformed into
a calcite grade to make a decision.
MATERIALS AND METHODS
Sample Preparation and Selection of Specimens
A 70 kg representative sample of crushed limestone was
obtained from a deposit within the Queguay Formation,
which is currently mined for the cement production. To
reduce the mass of the sample, the coning and quartering
method was employed, resulting in a subsample weighing
approximately 35 kg. The subsample was then screened into
three size fractions ranging from 0 to 5.1 cm. The material
with a size greater than 2.5 cm and smaller than 5.1 cm was
designated as fraction C, while the material ranging from
1.9 cm to 2.5 cm was labeled as fraction F (Figure 1). The
fraction with particles smaller than 1.9 cm was considered
unsuitable for sorting evaluation. Subsequently, 100 rocks
were randomly chosen from fractions F and C for further
analysis.
Testing Procedure
Firstly, each specimen was washed with tap water to clean
its surface, preventing that external fine particles interfere
with the measurement procedures.
Sensory testing was conducted using four distinct sen-
sors: XRT, XRF, laser and colorimeter. The measurements
for XRT, XRF, laser and colorimeter were obtained utiliz-
ing the following equipment: a COMEX Lab-Sorter MSX-
400-VL-XR-3D, a Thermo Scientific Niton XL2 analyzer,
a laser of type 650 nm class II (1mW), and a Photovolt
Model 577, respectively.
In the case of XRF, laser, and colorimeter measure-
ments, readings were taken from four faces of each rock of
C fraction, while for those that were part of the F fraction,
measurements were collected from two faces. Regarding
the XRF scans, the equipment was configured to operate in
“mining mode Cu/Zn” with a counting time of 40 seconds.
The color measurements were conducted with the use of a
blue filter. The laser responses were obtained from process-
ing images of each specimen when they were exposed to the
incident laser.
Following the sensor measurements, each specimen
was pulverized to a particle size of 150 μm and subjected
to elemental analysis. The calcite grade was estimated based
on the calcium content of the sample. It was assumed that
the calcium present in the sample is only hosted as cal-
cite. The average calcite grade of two measurements was
reported as the assay rock calcite grade. Additionally, to ver-
ify these results, approximately 1 g of the pulverized sample
was utilized for confirmation through the loss on ignition
(LOI) method. The LOI quantification was performed at
1000 °C.