AI-Assisted Prediction of Modal and Impact Performance in EV Battery Packs Under Multi-Scenario Conditions

Abstract

This study focuses on the dynamic performance prediction of electric vehicle battery packs in complex service environments, and
constructs a multi scenario modal and impact response prediction system based on artificial intelligence. By establishing a refined finite element model of the battery pack, modal frequencies, vibration characteristics, and impact response data under different operating conditions are
obtained. A fusion architecture of deep convolutional networks and recurrent neural networks is used to achieve rapid prediction of modal parameters. A multi parameter coupled impact performance prediction model was established by combining support vector machine and random
forest ensemble algorithm. The research results show that the proposed AI prediction method achieves a modal frequency prediction accuracy
of 95.2%, and the prediction error of impact peak stress is controlled within 8%, significantly improving the efficiency of battery pack dynamic performance evaluation. The structural optimization based on genetic algorithm increases the first-order modal frequency of the battery
pack by 12.3% and enhances the impact energy absorption capacity by 18.7%, providing effective technical support for the engineering design
of electric vehicle battery packs.