Although artificial neural networks are widely used in various fields, including mining and mineral processing, they can be problematic for appropriately choosing the model architecture and parameters. In this article, we describe a procedure for the optimization of the structure of a convolutional neural network model developed for the purposes of metallic ore pre-concentration. The developed model takes as an input two-band X-ray scans of ore grains, and for each scan two values corresponding to concentrations of zinc and lead are returned by the model. The whole process of sample preparation and data augmentation, optimization of the model hyperparameters and training of selected models is described. The ten best models were trained ten times each in order to select the best possible one. We were able to achieve a Pearson coefficient of R = 0.944 for the best model. The detailed results of this model are shown, and finally, its applicability and limitations in real-world scenarios are discussed.

This study evaluates sensor-based pre-concentration using XRT technology as an alternative to selective mining for low-grade European copper ores (0.48% Cu), addressing the need for sustainable beneficiation amid declining ore grades and environmental pressures in green mining initiatives. Copper ore samples from Złote Hory mine (Czech Republic) were selectively extracted, mixed (1:1:1 ore 8–16 mm/ore 16–32 mm/waste rock 8–32 mm), and analyzed on Comex’s LSX-MAX laboratory sorter with dual-energy XRT sensors, calibrated for maximum product recovery via density-based classification into High-Density (product) and Low-Density (waste) fractions. Sorting achieved a 1:1 product-to-waste mass split from feed (Cu = 0.5%, 100% mass), yielding pre-concentrate at 0.91% Cu (52.08% mass yield, 95.67% recovery) and waste at 0.04% Cu (47.92% mass, 4.33% loss)—a 1.82x grade upgrade superior to mixed feed and 1.42x superior to selective mining (0.64% Cu at 66.21% yield). Combined approaches promise further optimization; future work will assess downstream grinding/flotation impacts, industrial scaling, and economic/environmental benefits.

Flotation is one of the most widely used separation methods in mineral processing, and its performance is strongly influenced by complex physicochemical and hydrodynamic conditions. Flotation behaviour is commonly described using kinetic models. This study investigates the application of laboratory-derived flotation kinetic models for the simulation of a pilot-scale flotation circuit treating a polymetallic sulfide ore containing critical raw materials. Laboratory flotation experiments were conducted using Denver flotation cells, and kinetic parameters were determined by fitting first-order, two-component, and three-component kinetic models. The obtained parameters were implemented into an HSC SIM process model to evaluate the influence of flotation circuit configuration and residence time on process performance. The simulation indicated that the optimum circuit configuration consisted of two rougher and two scavenger flotation cells with an overall residence time of 28 min. The multi-component kinetic models provided a substantially better description of flotation behaviour than the classical first-order model and enabled a more realistic prediction of flotation performance. The simulation results were subsequently validated using a continuous pilot-scale flotation experiment. The simulated cleaner concentrate Cu grade (9.02 wt.%) was in good agreement with the experimental value (9.27 wt.%), confirming the capability of the proposed methodology to predict pilot-scale flotation performance. Although minor differences in product composition were observed, the overall flotation trends were reproduced successfully. The remaining discrepancies were mainly attributed to scale-up effects and differences in hydrodynamic conditions. The proposed methodology demonstrates that laboratory-derived kinetic parameters can be successfully applied to pilot-scale flotation process simulation and provides a practical tool for flotation circuit design and process optimization.