A reconstructive method-based sequential modeling approach for crashworthiness design of the combined five-cell structure

Abstract In this study, the crashworthiness of a combined five-cell structure is investigated numerically. For Finite Element Analysis (FEA)-based time-dependent impact, the solver commonly asks for hundreds of thousands of iterations, while only limited iterations can be sampled for the crashworthiness validation because of expensive computational and storage costs. To make full use of FEA results, an image-driven Enhanced Reconstructive Neural Network (EReConNN) is proposed. Based on the proposed method, the time-dependent information of the impact can be obtained, and the physical field along the time axis can be reconstructed. First, a benchmark validates the proposed method. After that, the influence of structural parameters of the combined five-cell structure on the crashworthiness is investigated by using the EReConNN. According to the results, more than 90% of computational time is saved. Besides, the Peak Crushing Force (PCF) is mainly affected by the radii of holes at both ends, while the middle hole influences the impact stability. Besides, the position of the front hole has a significant influence on the PCF, but has a small influence on the Crushing Force Efficiency (CFE). While the performance of the rear hole is opposite.