Surrogate optimization of lithium-ion battery coating process

In the Li-ion battery manufacturing process, uniform coating thickness is essential for ensuring high-quality electrode production. Elevated or scalloped coating edges are often formed because of inadequate coater design. Traditional coater design approaches entail resource-intensive coating experiments or time-consuming simulations. In this study, we present a five-step optimization framework to achieve uniform coating thickness in the cross-web direction. First, we conducted computational fluid dynamics (CFD) simulations by using a preselected set of 13 variables related to coater design and rheological properties of the slurry. Non-uniform coating characteristics were captured as dimensionless features derived from the CFD data. Then, we constructed a surrogate model to accurately replicate the CFD simulation and evaluate the dimensionless features. The surrogate model exhibited a high level of consistency with the original CFD data. The importance of the design variables was assessed in terms of accumulated local effects and Shapley values. On the basis of this assessment, six design variables related to coater geometry were selected to determine the optimal coater design given the coater width and slurry properties. Finally, genetic algorithms were employed to minimize the dimensionless features associated with defective coating edges. Statistically, the solutions reduced the number of dimensionless edge features by more than 90%. A comparison between the velocity profile data obtained by CFD and the surrogate model for the optimized solutions demonstrated the successful elimination of super-elevated edges in the coating. The proposed framework offers an effective optimization strategy that can be applied to practical coater design to minimize the occurrence of edge defects in the battery manufacturing industry.