HIGH PROFILE – Comex
The benefits of AI powered, multi-sensor sorting solutions
Pre-concentration through sensorbased sorting( SBS) has an increasing potential in mineral processing, allowing the elimination of waste particles in the early stage and, thus, reduced processing costs, lower energy consumption, reduced water and flotation chemical requirements, and a reduction in tailings.
To provide these benefits, however, the sorting process needs to be accurate enough to also eliminate losses of valuable ore or other minerals within the waste fraction.
This makes SBS a complex process in need of optimisation and, potentially, the application of several advanced sensors.
Sorting technology
The newly developed Comex sorting system can be configured with different sensors, depending on material types and analysed parameters, to provide the best identification and separation efficiency at various processing capacities.
The sorting system includes the X-ray attenuation analysis carried out by the X-ray source and the X-ray Transmission( XRT) sensors. In addition, X-ray Fluorescence( XRF) sensors are applied to the same area of analysis. Further image analysis is realised by a camera system installed over the belt conveyor, employing visible light( RGB) or SWIR( Short Wave Infrared) hyperspectral cameras, which can be used simultaneously
40
Comex CXR sorting system in multi-sensory configuration
Jacek Kolacz, CEO Comex Group, talks through the multi-sensor and AI-infused options for sensor-based sorting or separately depending on the application.“ This brings an important advantage of higher definition of the analysed material, allowing high accuracy results related to chemical composition of the sorted particles,” Kolacz says.
Sensor types
The output signals from different sensors provide various information about the processed material. The XRT sensor( XRT- DE Dual Energy) brings the most important information about the processed material. In this case, the transmitted X-ray photons are analysed with 0.4-0.8 mm spatial resolution and their energy is measured by X-ray sensors. X-ray attenuation depends on the atomic number of the material, with this effect used to recognise different material groups. The resultant image is presented in the form of separate colours where, in this case, the blue colour shows low density materials, and the yellow and green colours present high-density particles. Each particle is described by a fraction of the defined colour content within its contours, providing information about the content of the specified atomic number group in such particles. This information can further be related to the ore content inside the particle.
“ By applying a threshold, and comparing it with the calculated ore content, such particles are classified by the software to the rejected or non-rejected fractions,” Kolacz explained.
Combining the XRT image with other sensor information can result in higher accuracy analysis and better sorting efficiency, according to Comex. For instance, the optical RGB camera can provide surface analysis for different light frequencies, intensities and colour saturation that is simply not possible with the human eye. Depending on the content of each type of colour class, when compared with the particle contours, it is possible to describe such particles by areas related to these characteristic values.
In recent years, new types of cameras were introduced to the market, using hyperspectral imaging in the infrared light spectrum. While standard colour cameras produce three colour planes( RGB), the hyperspectral cameras generate image cubes where over 250 wavelength channels are recognised, typically in the range of 900-2500 nm( SWIR cameras). The light reflectance in this wavelength range is different for various materials and, therefore, such cameras can be used for recognising specific materials.
“ This type of recognition is already used for drill core scanning, where mineralogical composition of the processed material can be identified,” Kolacz says.“ Copper ore preconcentration is an example here, whereby the system was first calibrated to recognise copper ore( sulphides) within the regular sandstone, based on reflectance differences. In this example, the system recognised the areas being chemically similar to the copper ore( shown in green), where the waste particles that contained significantly less copper were shown in brown.“ This brings another possibility to identify particles based on their mineral composition and, indirectly, by their chemical composition,” he said.
XRF sensors can provide another type of data for some chemical elements where specific metals can be identified on the surface of the sorted particles. Due to the very short time of exposure during sorting( usually a few milliseconds), this analysis can only be carried out for heavier metals( like Cu, Fe, Zn, etc), however, this
Comex multi-sensory sorter during factory testing
International Mining | MAY 2025