XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 429
(sulfide type—mainly chalcopyrite). In this case, the XRT
sensor had a high resolution of 50 microns (Kolacz, 2019),
allowing identification of a fine and disseminated material
with higher density. In addition, the SWIR hyperspectral
camera was applied to identify copper bearing minerals on
the surface of the sorted material, which were not visible for
the XRT sensor. Both signals were correlated in the sort-
ing model to allow final identification of the product and
waste fractions. Table 6, shows the results achieved during
sorting using XRT sensor in combination with the SWIR
hyperspectral camera (assays done by XRF spectroscopy).
In this case, the sorting process was done twice to sepa-
rate three fractions and allow achieving more information
about the process flexibility and potential limitations. Two
different system settings of thresholds for SWIR and XRT
analysis were used to allow separation of the different cop-
per grades. The feed material had a copper concentration of
0.71%. At the first threshold stetting, when the waste frac-
tion was separated, it was possible to eliminate over 15% of
the material mass, which contained only 0.02% of copper,
which was far below the economic level. At the same time,
only 0.4% of copper was lost in this waste fraction. Further
adjustment of the threshold in the sorting model and
reprocessing of the remaining material, provided higher
waste rock removal, however, with more lost copper in the
middlings fraction. In this case, the copper content in the
middlings was 0.31%, which corresponded to 19.5% metal
recovery. If acceptable for a user, this can bring a significant
reduction of the low grade ore (about 60% of the fraction
having 0.02–0.31% of Cu) from further processing stages
and the final product can reach 1.43% of Cu instead of
0.71% as in the original feed material. However, the meal
losses can reach about 19.5% (Cu recovery). In any case,
sorting brings new possibilities for optimization of the
plant operations including sorting and further processing
steps as the overall process.
This type of separation is very new and still requires a
lot of testing, especially regarding the results obtained in
the laboratory and industrial scale. Each case of the particle
analysis by hyperspectral SWIR camera, generates a lot of
data, which requires very precise analysis to tune the sorting
model and to get the best result from such technique.
CONCLUSIONS
Sensor based sorting can be the new effective method for
pre-concentration of minerals. The XRT sorting detec-
tors with AI support can provide a very efficient pre-con-
centration of copper and gold ores and it is expected to
provide similar advantages for sorting of other metal ores.
This brings an enormous advantage over the current sort-
ing equipment (standard XRT sensors), which provide sig-
nificant meal recovery losses when processing particles with
disseminated ores.
Detection of ore particles smaller than 50 µm is not
possible in the new XRT detection system, however, this
may provide another positive effect. Copper ore with high
dissemination is very problematic for further processing
due to necessary fine grinding and fine particle flotation,
and it is often present in the flotation waste streams. By
applying the mentioned sorting process (XRT and SWIR),
even when the sorting waste fraction may contain a small
amount of copper, this ore would be lost in the traditional
processing circuits operating today, due to the fine size.
This effect is the subject of the separate study carried out
at Comex, including ore nature, sorting efficiency, flota-
tion efficiency and the overall process calculations includ-
ing environmental aspects, for large scale operations. This
brings new possibilities to reduce the processing cost and
gain the important environmental benefits.
Combination of the XRT sensors with the visible light
cameras and hyperspectral infra-red cameras in SWIR
range, can become a very powerful system for identifying
mineralogical composition of the sorted material instead of
only density or other single features of the analyzed mate-
rial. Application of the AI models for data analysis becomes
a powerful tool for handling complicated and demanding
sorting cases. This approach will bring even more possibili-
ties for further material identification and more efficient
sorting of complex minerals or materials. It can make many
current mining operations more economic or bring new
possibilities for new potential mines, where the current eco-
nomic or environmental restrictions hinder their operation.
Table 6. Sorting results using X-ray sensor and SWIR camera
Fraction Yield [%]Cu Content [%]Cu Recovery [%]
Feed 100 0.71 100
Product 39.75 1.43 80.1
Middlings 44.65 0.31 19.5
Waste 15.55 0.02 0.4
(sulfide type—mainly chalcopyrite). In this case, the XRT
sensor had a high resolution of 50 microns (Kolacz, 2019),
allowing identification of a fine and disseminated material
with higher density. In addition, the SWIR hyperspectral
camera was applied to identify copper bearing minerals on
the surface of the sorted material, which were not visible for
the XRT sensor. Both signals were correlated in the sort-
ing model to allow final identification of the product and
waste fractions. Table 6, shows the results achieved during
sorting using XRT sensor in combination with the SWIR
hyperspectral camera (assays done by XRF spectroscopy).
In this case, the sorting process was done twice to sepa-
rate three fractions and allow achieving more information
about the process flexibility and potential limitations. Two
different system settings of thresholds for SWIR and XRT
analysis were used to allow separation of the different cop-
per grades. The feed material had a copper concentration of
0.71%. At the first threshold stetting, when the waste frac-
tion was separated, it was possible to eliminate over 15% of
the material mass, which contained only 0.02% of copper,
which was far below the economic level. At the same time,
only 0.4% of copper was lost in this waste fraction. Further
adjustment of the threshold in the sorting model and
reprocessing of the remaining material, provided higher
waste rock removal, however, with more lost copper in the
middlings fraction. In this case, the copper content in the
middlings was 0.31%, which corresponded to 19.5% metal
recovery. If acceptable for a user, this can bring a significant
reduction of the low grade ore (about 60% of the fraction
having 0.02–0.31% of Cu) from further processing stages
and the final product can reach 1.43% of Cu instead of
0.71% as in the original feed material. However, the meal
losses can reach about 19.5% (Cu recovery). In any case,
sorting brings new possibilities for optimization of the
plant operations including sorting and further processing
steps as the overall process.
This type of separation is very new and still requires a
lot of testing, especially regarding the results obtained in
the laboratory and industrial scale. Each case of the particle
analysis by hyperspectral SWIR camera, generates a lot of
data, which requires very precise analysis to tune the sorting
model and to get the best result from such technique.
CONCLUSIONS
Sensor based sorting can be the new effective method for
pre-concentration of minerals. The XRT sorting detec-
tors with AI support can provide a very efficient pre-con-
centration of copper and gold ores and it is expected to
provide similar advantages for sorting of other metal ores.
This brings an enormous advantage over the current sort-
ing equipment (standard XRT sensors), which provide sig-
nificant meal recovery losses when processing particles with
disseminated ores.
Detection of ore particles smaller than 50 µm is not
possible in the new XRT detection system, however, this
may provide another positive effect. Copper ore with high
dissemination is very problematic for further processing
due to necessary fine grinding and fine particle flotation,
and it is often present in the flotation waste streams. By
applying the mentioned sorting process (XRT and SWIR),
even when the sorting waste fraction may contain a small
amount of copper, this ore would be lost in the traditional
processing circuits operating today, due to the fine size.
This effect is the subject of the separate study carried out
at Comex, including ore nature, sorting efficiency, flota-
tion efficiency and the overall process calculations includ-
ing environmental aspects, for large scale operations. This
brings new possibilities to reduce the processing cost and
gain the important environmental benefits.
Combination of the XRT sensors with the visible light
cameras and hyperspectral infra-red cameras in SWIR
range, can become a very powerful system for identifying
mineralogical composition of the sorted material instead of
only density or other single features of the analyzed mate-
rial. Application of the AI models for data analysis becomes
a powerful tool for handling complicated and demanding
sorting cases. This approach will bring even more possibili-
ties for further material identification and more efficient
sorting of complex minerals or materials. It can make many
current mining operations more economic or bring new
possibilities for new potential mines, where the current eco-
nomic or environmental restrictions hinder their operation.
Table 6. Sorting results using X-ray sensor and SWIR camera
Fraction Yield [%]Cu Content [%]Cu Recovery [%]
Feed 100 0.71 100
Product 39.75 1.43 80.1
Middlings 44.65 0.31 19.5
Waste 15.55 0.02 0.4
