Predictive insights for copper recovery: A synergistic approach integrating variability data and machine learning in the geometallurgical study of the Tizert deposit, Morocco

Abstract

This work involved the application of machine learning (ML) to predict copper recovery through variability data in the context of the geometallurgical study of the Tizert deposit in Morocco. For this purpose, geological and metallurgical variability investigations were conducted on drilling samples from Tizert ore, leading to the consolidation of a robust database for ML algorithms. The dataset comprised 345 observations across seven variables, namely GPS coordinates, geological units, lithological facies, the main copper deportment, and metallurgical responses from flotation tests, as well as the copper grade and the oxidation rate. The geological and mineralogical variability revealed two geological units: The Basal Series and Tamjout Dolomite unit, hosting copper mineralization dominated by chalcocite and malachite. Besides, the metallurgical evaluation demonstrated recovery variability, primarily influenced by oxidized copper. Copper oxides exhibit less efficient flotation than sulfides, with recovery decreasing as the oxidation rate increases. The prediction of copper recovery was performed using the variability data through the XGBoost algorithm. The resulting model showed good performance on the test dataset, with a coefficient of determination R2 of 0.83 and a RMSE of 2.47, improving the accuracy of copper recovery predictions. In addition, the model elucidated the most significant variables impacting Tizert ore behavior, notably the casing, copper-bearing mineral, and oxidation rate. These variables were considered key factors for advancing Tizert ore valuation and geometallurgical model development. Overall, metallurgical modeling based on the ML approach provided an innovative way to integrate variables into mine planning and processing. The use of such a predictive model gave valuable insights for improving efficiency, understanding the metallurgical behavior, optimizing production, and reducing costs along the value chain of the Tizert deposit. Further data processing efforts are underway, to improve the assessment of copper recovery quality for the Tizert ore and the development of a complete geometallurgical model on a plant scale.