Predicting copper recovery from flotation using machine learning and laboratory-generated data

The efficient extraction of copper is becoming increasingly complex due to the declining availability of high-grade ore deposits and the implementation of more rigorous environmental standards. These constraints have heightened the demand for advanced technologies that optimize copper recovery through processes, such as flotation. Flotation, a widely employed physicochemical separation technique, is highly sensitive to operational parameters, making its optimization essential for maximizing metal recovery, reducing operational costs, and promoting environmentally responsible resource utilization and conservation. This study investigates applying Machine Learning (ML) techniques to improve flotation process performance. Specifically, this study assesses the predictive performance of four ML algorithms: random forest, support vector machine, K-means clustering, and Artificial Neural Networks (ANNs) for estimating copper recovery in flotation processes. The models were trained and validated using experimental data from a laboratory-scale flotation system. Among the evaluated algorithms, the ANN achieved the highest prediction accuracy of 98.69%, demonstrating a strong capacity to model complex nonlinear interactions among critical process variables. Complementary, disequilibrium, and entropic measures validate the results using the probability selection between three classes. These results highlight the potential of ML-based approaches to support process optimization, enhance recovery efficiency, and contribute to the sustainable development of copper extraction technologies.