The increasing global demand for rare earth elements (REEs) presents significant mining and extraction challenges. Extraction of REEs from secondary resources such as mine tailings, electronic waste, and industrial by-products becomes an alternative solution. This study introduces an explainable artificial intelligence (AI) system designed to (i) predict the leaching efficiency rates and (ii) provide real-time explanations of the key extraction factors during the leaching process, while also recommending condition adjustments to optimize the leaching efficiency of REEs from secondary resources. Trained on 572 experimental datasets sourced from the Web of Science database, the system enhances leaching performance by offering explainable recovery rates and identifying influential process parameters. The results showed that silica concentration is the most critical factor, followed by REEs classification (light vs. heavy). In contrast, acid strength (pH), aluminum content, and temperature exhibited moderate but comparatively lower contributions to the overall leaching performance. By identifying suboptimal parameters and suggesting adjustments that lead to improvements in the predicted recovery (R2=0.81), the approach exemplifies how explainable AI can bridge the gap between empirical data and process innovation. This methodology offers a broadly applicable framework for enhancing decision-making and process efficiency in the complex extraction systems, with the potential to extend beyond REEs into other resource- and energy-intensive industries.
Vo et al. (Sat,) studied this question.