XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 449
mass spectrometry (ICP-MS) following lithium borate dis-
solution. Lithium (Li) concentrations were quantified by
ICP-OES following Na2O2 dissolution, while beryllium
(Be) concentrations were determined by ICP-MS follow-
ing LiBO2 dissolution. For sorting products from the REE
carbonatite ore sample, total concentrations of fluorine
(F) were determined by potentiometry, using a combined
fluoride selective electrode, after sample decomposition
through sodium carbonate fusion. For sorting products
from the scandium syenite ore sample, a Satmagan instru-
ment was used to determine the weight percent (wt%) of
magnetic material, assumed here to correspond to the mag-
netite content, while the concentration of scandium (Sc)
was determined by ICP-OES following lithium metaborate
dissolution. Powder X-ray diffraction (XRD) was employed
to ascertain mineral content, which then informed the
determination of mineralogical composition through the
singular-value decomposition quantitative phase analysis
(SVD-QPA) methodology (Couillard &Mercier, 2016).
Quantitative Assessment of Sorting Performance
Comprehensive metallurgical balances were calculated for
each sorting production test conducted. In addition, cumu-
lative recovery versus mass pull curves for specific elements
or mineral components were plotted to further analyze the
test outcomes.
RESULTS AND DISCUSSION
Lithium Pegmatite Ore Sample
Sensor-Based Scanning of Rocks in –4.0/+1.5 and
–1.5/+0.5 Size Fractions
Selected rocks from the –4.0/+1.5 and –1.5/+0.5 size frac-
tions underwent scanning with various sensors to evaluate
the efficacy of different sorting technologies. This included
static tests using DE-XRT, VIS-NIR, color, and SWIR sen-
sors to discern differences between rocks in the ore material.
Image analysis was then conducted using the commercial
software provided by the ore sorter’s manufacturer, with the
key results presented in Figure 1.
The DE-XRT scanning (Figure 1a-b) showed moderate
potential for sorting, identifying some differences between
rocks. However, it was unclear whether these differences
were due to variations in rock density or particle thick-
ness. In contrast, the VIS-NIR (Figure 1c-d) and SWIR
(Figure 1e-f) scans did not demonstrate significant poten-
tial for sorting, as they failed to clearly differentiate between
rocks within the same size fraction.
A more promising result was observed with the color
sensor. By processing the raw image (Figure 1g) from
this sensor and creating a false-color calibration image
(Figure 1h), it was possible to identify rocks containing
spodumene grains. The red areas in the calibration image
corresponded to pixels matching the color of visible spodu-
mene grains, indicating that the color sensor effectively cat-
egorized rocks based on the presence of spodumene.
Given the limited potential of VIS-NIR and SWIR
for sorting, subsequent dynamic sorting tests focused on
combinations of the DE-XRT and color sensors. These tests
aimed to assess the sorting potential more accurately using
these more effective sensors.
DE-XRT and color sensor dynamic tests on –4.0/+1.5
and –1.5/+0.5 size fractions of lithium ore sample
A series of dynamic tests employing either the DE-XRT or
color sensor were completed using the ore sorter conditions
and parameters provided in Table 1. Results from these
dynamic tests are illustrated in Figure 2.
In dynamic sorting tests that utilized only the color
sensor, the results for both –4.0/+1.5 (Figure 2a-b) and
–1.5/+0.5 (Figure 2c-d) size fractions indicated efficient
sorting of rocks. The color sensor successfully segregated
rocks based on the presence of pixels with a greenish-white
color similar to that of spodumene grains. This was evident
when comparing the ejected and unejected rocks a higher
proportion of rocks with white (presumably spodumene-
bearing) surfaces were observed in the ejected fraction.
Conversely, when only the DE-XRT sensor was
used for sorting, the results varied depending on the
size fraction. As visually observable with the unaided
eye, the sorting of rocks from the –4.0/+1.5 size frac-
tion (Figure 2e-f) was more significantly influenced
by variations in rock particle thickness than in the
–1.5/+0.5 size fraction (Figure 2g-h). This was particularly
noticeable in the Sayona ore material.
Given these observations, particularly the influence of
particle thickness on DE-XRT sorting, it was decided to
primarily utilize the color sensor for the ore sorting pro-
duction tests. This approach aimed to minimize the impact
of particle thickness differences, especially when compar-
ing classification performance between the –4.0/+1.5 and
–1.5/+0.5 feed materials.
Color Sensor Production Tests on the –4.0/+1.5 and
–1.5/+0.5 Size Fractions of the Lithium Ore Sample
Sorting production tests with the –4.0/+1.5 and –1.5/+0.5
size fractions were performed using the conditions and
parameters provided in Table 2 according to the process-
ing scheme illustrated in Figure 3. Metallurgical balances
for the production tests are presented in Table 3. Only the
main elements (Li, Si, Al, Fe, Na, K) and minerals (quartz,
mass spectrometry (ICP-MS) following lithium borate dis-
solution. Lithium (Li) concentrations were quantified by
ICP-OES following Na2O2 dissolution, while beryllium
(Be) concentrations were determined by ICP-MS follow-
ing LiBO2 dissolution. For sorting products from the REE
carbonatite ore sample, total concentrations of fluorine
(F) were determined by potentiometry, using a combined
fluoride selective electrode, after sample decomposition
through sodium carbonate fusion. For sorting products
from the scandium syenite ore sample, a Satmagan instru-
ment was used to determine the weight percent (wt%) of
magnetic material, assumed here to correspond to the mag-
netite content, while the concentration of scandium (Sc)
was determined by ICP-OES following lithium metaborate
dissolution. Powder X-ray diffraction (XRD) was employed
to ascertain mineral content, which then informed the
determination of mineralogical composition through the
singular-value decomposition quantitative phase analysis
(SVD-QPA) methodology (Couillard &Mercier, 2016).
Quantitative Assessment of Sorting Performance
Comprehensive metallurgical balances were calculated for
each sorting production test conducted. In addition, cumu-
lative recovery versus mass pull curves for specific elements
or mineral components were plotted to further analyze the
test outcomes.
RESULTS AND DISCUSSION
Lithium Pegmatite Ore Sample
Sensor-Based Scanning of Rocks in –4.0/+1.5 and
–1.5/+0.5 Size Fractions
Selected rocks from the –4.0/+1.5 and –1.5/+0.5 size frac-
tions underwent scanning with various sensors to evaluate
the efficacy of different sorting technologies. This included
static tests using DE-XRT, VIS-NIR, color, and SWIR sen-
sors to discern differences between rocks in the ore material.
Image analysis was then conducted using the commercial
software provided by the ore sorter’s manufacturer, with the
key results presented in Figure 1.
The DE-XRT scanning (Figure 1a-b) showed moderate
potential for sorting, identifying some differences between
rocks. However, it was unclear whether these differences
were due to variations in rock density or particle thick-
ness. In contrast, the VIS-NIR (Figure 1c-d) and SWIR
(Figure 1e-f) scans did not demonstrate significant poten-
tial for sorting, as they failed to clearly differentiate between
rocks within the same size fraction.
A more promising result was observed with the color
sensor. By processing the raw image (Figure 1g) from
this sensor and creating a false-color calibration image
(Figure 1h), it was possible to identify rocks containing
spodumene grains. The red areas in the calibration image
corresponded to pixels matching the color of visible spodu-
mene grains, indicating that the color sensor effectively cat-
egorized rocks based on the presence of spodumene.
Given the limited potential of VIS-NIR and SWIR
for sorting, subsequent dynamic sorting tests focused on
combinations of the DE-XRT and color sensors. These tests
aimed to assess the sorting potential more accurately using
these more effective sensors.
DE-XRT and color sensor dynamic tests on –4.0/+1.5
and –1.5/+0.5 size fractions of lithium ore sample
A series of dynamic tests employing either the DE-XRT or
color sensor were completed using the ore sorter conditions
and parameters provided in Table 1. Results from these
dynamic tests are illustrated in Figure 2.
In dynamic sorting tests that utilized only the color
sensor, the results for both –4.0/+1.5 (Figure 2a-b) and
–1.5/+0.5 (Figure 2c-d) size fractions indicated efficient
sorting of rocks. The color sensor successfully segregated
rocks based on the presence of pixels with a greenish-white
color similar to that of spodumene grains. This was evident
when comparing the ejected and unejected rocks a higher
proportion of rocks with white (presumably spodumene-
bearing) surfaces were observed in the ejected fraction.
Conversely, when only the DE-XRT sensor was
used for sorting, the results varied depending on the
size fraction. As visually observable with the unaided
eye, the sorting of rocks from the –4.0/+1.5 size frac-
tion (Figure 2e-f) was more significantly influenced
by variations in rock particle thickness than in the
–1.5/+0.5 size fraction (Figure 2g-h). This was particularly
noticeable in the Sayona ore material.
Given these observations, particularly the influence of
particle thickness on DE-XRT sorting, it was decided to
primarily utilize the color sensor for the ore sorting pro-
duction tests. This approach aimed to minimize the impact
of particle thickness differences, especially when compar-
ing classification performance between the –4.0/+1.5 and
–1.5/+0.5 feed materials.
Color Sensor Production Tests on the –4.0/+1.5 and
–1.5/+0.5 Size Fractions of the Lithium Ore Sample
Sorting production tests with the –4.0/+1.5 and –1.5/+0.5
size fractions were performed using the conditions and
parameters provided in Table 2 according to the process-
ing scheme illustrated in Figure 3. Metallurgical balances
for the production tests are presented in Table 3. Only the
main elements (Li, Si, Al, Fe, Na, K) and minerals (quartz,