Optimizing deep drawing parameters for power battery shells through the integration of feature-weighted SVM and genetic algorithm

Abstract. Deep drawing is one of the main forming processes for battery shells, and the rational setting of its process parameters directly affects the quality of the formation. Selecting the optimal deep drawing process parameters requires repeated trials and simulations, which increases the cost, reduces the efficiency, and poses a significant challenge to enterprises. To address this challenge, we focus on the first deep drawing process of the battery shell, proposing a parameter optimization method for battery shell deep drawing based on a Feature-Weighted Support Vector Machine (FWSVM) and Genetic Algorithms (GA). Our aim is twofold: on the one hand, to establish an agent model for finite element analysis of the deep drawing process using the FWSVM technique to enhance prediction accuracy and save simulation time; on the other hand, to derive optimal deep drawing process parameters using GA. The experimental results indicate that the FWSVM can accurately establish the relationship between process input/output parameters, and the optimized process parameters achieved through this method can realize the minimum thinning rate and convex mold contact force.