Predicting copper leaching from slag: an interpretable machine learning approach under oxidative sulfuric acid...

Area of Science:

• Metallurgical Engineering
• Data Science
• Sustainable Resource Management

Background:

• Efficient copper recovery from metallurgical waste is crucial for sustainable resource utilization.
• Copper slag presents a significant, yet underutilized, secondary source of copper.
• Optimizing copper leaching processes requires accurate predictive models.

Purpose of the Study:

• To develop an interpretable machine learning framework for predicting copper leaching efficiency from copper slag.
• To identify the key operational and compositional parameters influencing copper extraction.
• To provide a practical tool for optimizing sustainable copper recovery processes.

Main Methods:

• Compiled a comprehensive dataset of 465 experimental data points from peer-reviewed literature.
• Systematically optimized four machine learning algorithms: Random Forest, Support Vector Regression, XGBoost, and LightGBM, using 10-fold cross-validation.
• Employed SHAP (SHapley Additive exPlanations) for model interpretability analysis.

Main Results:

• XGBoost exhibited superior predictive performance with R² = 0.9794, RMSE = 3.4757, and MAE = 2.3442 on the test set.
• SHAP analysis revealed that leaching time, acid concentration, and temperature are dominant operational parameters influencing copper extraction.
• Compositional variables (Si, S, Al) showed limited direct contribution within the studied range; identified nonlinear trends align with shrinking-core and diffusion-controlled mechanisms.

Conclusions:

• The developed interpretable machine learning framework accurately predicts copper leaching efficiency from copper slag.
• Key operational parameters significantly influence copper extraction, offering targeted optimization strategies.
• The framework provides quantitative guidance for enhancing sustainable metal recovery from metallurgical waste.