Chelating extraction of critical metals from cathode of end-of-life lithium titanium oxide batteries: Experiments, machine learning, and validation

Disposal of lithium-ion batteries (LIBs) at their end-of-life (EoL) calls for their recycling to recirculate valuable metals and protect the environment. Therefore, we extracted metals from cathodes of EoL lithium-titanate batteries using ethylene-di-amine-tetra acetic acid disodium (EDTA-2Na). An orthogonal array was used to design experiments and signal-to-noise calculations were used to define optimal conditions, which were 0.50 mol/L EDTA-2Na, pH 6, 75°C, 180 min, 2% pulp density, and 300 rpm, resulting in 97.96%, 94.79%, 96.45%, and 98.89% leaching efficiencies for Li, Ni, Co, and Mn, respectively. Then, statistically significant variable interactions were investigated using Pearson correlation at 95% confidence. Accordingly, pH and temperature were significant. Extraction efficiency decreased as pH increased, whereas it increased as temperature increased. Machine learning fitting using linear regression for multioutput prediction was unsatisfactory, whereas random forest regression (RFR) was satisfactory. The fitted RFR was used for permutation variable importance, confirming pH and temperature as influential variables, but time and pulp density were also noted. Subsequently, the fitted RFR failed to satisfactorily predict additional validation experiments. Therefore, we recommended increasing the number of experiments and using other fitting models. Additional analysis that included the initial oxidation-reduction potential (optimal of 33.3 mV) revealed this to be the most important variable, largely overshadowing the effects of all other variables. Finally, environmental assessment highlighted benefits of chelating extraction, but economic assessment indicated room for improvement.