A machine learning-based crashworthiness optimization for a novel pine cone-inspired multi-cell tubes under bending

Bionic tubes are of interest in vehicle engineering due to their superior crashworthiness potential. This study proposes a crashworthiness response investigation and machine learning-based multi-objective optimization of pine cone-inspired muti-celled tubes (PCMTs). The base computer PCMT model was correlated using existing experiments, followed by a dynamic response evaluation of different PCMT geometrical and thickness configurations to assess their structural performance. Surrogate models of these PCMTs were then constructed using machine learning algorithms, and their main and interaction effects were analyzed. A non-dominated sorting genetic algorithm II (NSGA-II) approach was employed to perform a multi-objective optimization. The results demonstrate that thickness change had more effect on the initial peak force (IPF) and the mean crushing force (MCF) than the specific energy absorption (SEA). Besides, due to the coupling effect, IPF, MCF and SEA of the optimal design solution of the PCMTs could reach a 36.82 %, 61.66 % and 72.95 % increase than the sum case, suggesting that embedding inner tubes could significantly increase energy absorption with a relative minor IPF increase. Moreover, the MCF and SEA of optimal design gave an average difference of 18.01 % and 5.91 % from the original tubes. PCMTs, therefore, could be used as an ideal energy absorption structure in the vehicle body structures.