Machine Learning–Driven Hydrometallurgical Optimization for Sustainable Lithium‑Ion Battery Recycling

The rapid expansion of electric vehicle (EV) markets and stationary energy storage systems has created an urgent need for efficient and sustainable recycling technologies to manage growing volumes of end of life lithium ion batteries (LIBs). Here, we propose an AI driven framework for hydrometallurgical LIB recycling that leverages a structured dataset—compiled from peer reviewed literature and laboratory experiments—to train a LightGBM multi target regression model predicting metal recovery efficiencies and impurity levels under varied process conditions. SHapley Additive exPlanations (SHAP) analysis provided mechanistic insights, identifying acid molarity, solid to liquid ratio, leaching time, and pH control as the most influential parameters. Experimental validation of AI recommended conditions yielded lithium, cobalt, and nickel recoveries of 93.2%, 96.1%, and 91.8%, respectively—improvements of 8.5%, 7.8%, and 10.6% over baseline—while maintaining prediction errors within ±1.5% . A comparative life cycle assessment demonstrated a 24.3% reduction in greenhouse gas emissions and 28.6% lower acid consumption, and techno economic analysis revealed a 19.5% decrease in reagent costs. By continuously incorporating new experimental data into a closed loop learning pipeline, the framework self improves—reducing trial and error experimentation and enhancing predictive reliability—and can be generalized to other resource recovery systems, advancing green manufacturing and circular economy infrastructure.