Data-driven modelling and evaluation of a battery-pack system's mechanical safety against bottom cone impact

The exploration of the mechanical responses of the battery pack system (BPS) when subjected to dynamic impact loads is crucial for the safety of power batteries during collisions. To elucidate the mechanical response mechanism of the BPS under bottom impact, in this study, a series of finite element (FE) analysis tests were conducted in simulation environments with various combinations of the top radius of the cone model simulating objects, shell thickness, impact velocity, and impact angle. In the FE simulation, the BPS's nonlinear model was developed and then validated through modal analysis. Subsequently, five machine learning (ML) models enhanced by optimized algorithms were used to evaluate the mechanical response to the bottom impact of BPS. The training data of the ML model is obtained through the optimal Latin hypercube experimental design based on the FE dynamic simulation process. By comparison between cases, the predictive capabilities of various ML models are examined and the robustness of these models are assessed through adding Gaussian noise to ML models. In conclusion, the results demonstrate that ML methods can effectively evaluate the mechanical response of BPS bottom impact and furnish critical design inputs for manufacturing next-generation batteries with durability and safety limitations.