Interpretable machine learning for battery capacities prediction and coating parameters analysis

Battery manufacturing plays a direct and pivotal role in determining battery performance, which, turn, significantly affects the applications of battery-related energy storage systems. As complicated process that involves chemical, mechanical electrical operations, effective property predictions reliable analysis strongly-coupled parameters or variables become key but challenging issues for wider applications. In this paper, an interpretable machine learning framework could effectively predict product properties explain dynamic effects, as well interactions is proposed. Due to data-driven nature, can be easily adopted by engineers no specific mechanism knowledge required. Reliable dataset particularly coating (one stage) collected from real chain evaluate proposed framework. Illustrative results demonstrate three types capacities including cell capacity, gravimetric volumetric capacity accurately predicted with R2 over 0.98 at early-manufacturing stage. Besides, information regarding how variations mass, thickness, porosity affect these identified, while also quantified. The developed makes model more opens promising way quantify final properties. This assist obtain critical insights understand underlying material behavior, further benefiting smart control manufacturing.