Accurate prediction of NCM batteries recovery process under machine learning: Mechanism analysis and industrial application.

Effective recycling of spent LiNixCoyMn1-x-yO2 (NCM) battery is crucial to ensure sustainability of the lithium-ion battery industry. However, recycling is inherent with multiple operational steps and many effective factors. It is difficult to optimize the whole recycling process and identify the controlling steps, especially when the compositions and features of the raw materials are turbulent. This research demonstrates a machine learning (ML) strategy by mechanism analyzation to more accurately predict a spent NCM battery recycling process. Considering 28 input features under three categories (i.e., raw material properties, leaching reagent properties, operating conditions), Li, Ni, Co, and Mn leaching efficiency were analyzed with 4 typical ML models where extreme gradient boosting performed best. The leaching efficiency can be significantly improved when optimizing the leaching process by ML precisely forecasting. In addition to conventional operating conditions, the average key length of acid also significantly impacts metal leaching efficiency. Efficient leaching of Li can be achieved under malic acid (2.27 mol/L), S/L (48.25 g/L), stirring speed (528 rpm), temperature (55 ℃) and pH (2.15). This research could accurate predict NCM battery recovery process and pave the way for mechanism analyzation and industrial application under big data analyzation.