Design of battery shell stamping parameters for vehicles based on fusion of various artificial neural network models

Abstract

The application of neural network model in engineering prediction is frequent. The BPE shell material was optimized, and the reliability of the new material was verified by modal simulation. The accuracy of finite element modeling was ensured by constrained mode experiments, and all variables were preprocessed by Latin hypercube sampling. The design parameters were determined by Monte Carlo simulation. Four different neural networks, including back propagation (BP), radial basis function (RBF), extreme learning machine (ELM) and wavelet neural network (WNN), are used to train and learn the dataset. The BPE weight reduction ratio was 14.3%, the stress was reduced by 18.6%, deformation displacement was reduced by 14.2%, and the first-order mode was increased by 29.1%.