Machine Learning Methods for the Design of Battery Manufacturing Processes

Batteries are key enablers for the electrification of transportation systems and a contributor when moving towards a net-zero-carbon future. A clean future is only achievable via the mass production of high quality and efficient battery cells which are low in cost and have reduced environmental footprints. However, battery manufacturing is a complex process with a large number of control variables and interconnected steps, which affect the characteristics of the final cell such as capacity, lifetime, energy density, and thermal behaviour. Due to the strongly coupled interdependencies that involve material, chemical, mechanical and electrical operations, the in-depth understanding of various key production variables, parameters, their correlations and effects towards resulting manufactured electrode properties or final battery performance is urgently needed but still remains a significant challenge. Recent advancements in “Big Data” analytics have raised interest in designing machine learning approaches to conduct sensitivity analysis and predictions for battery manufacturing. This chapter presents a critical introduction of using state-of-the-art machine learning for predicting links between battery manufacturing and electrochemical performance. First, we showcase the key steps for battery production. We then review the key objectives, general framework and some widely-adopted machine learning methods for battery manufacturing applications. Finally, two case studies of designing suitable machine learning solutions to benefit the analysis of manufacturing dependency and the prediction of battery capacities are presented and discussed. This chapter provides insights into the use of advanced machine learning with interpretability for the effective prediction of manufactured battery performance, and sensitivity analyses for different material formulations and manufacturing parameters.