Crashworthiness performance prediction of square tube with elliptical cutout as crush initiator via FEA-ANN technique

This study investigates the crashworthiness performance of thin-walled square tubes used as crash boxes, focusing on optimising performance indices through the introduction of elliptical cut-outs. A computationally efficient method is proposed to predict crashworthiness under axial impact, based on the vertical and horizontal dimensions and the position of the elliptical cut-outs. This method employs an artificial neural network (ANN) trained on historical finite element analysis (FEA) data using the Levenberg-Marquardt algorithm. The trained ANN effectively modelled the relationship between the input parameters (cut-out dimensions and position) and the output parameters (crashworthiness performance indices). Subsequently, a multi-objective genetic algorithm (MOGA) was applied to the ANN model to maximise crush force efficiency and specific energy absorption. The resulting Pareto-optimal solutions were then verified using FEA, demonstrating a high level of accuracy. Optimised elliptical cut-outs significantly enhanced the crush force efficiency and specific energy absorption of the square tube, yielding increases of 17.72% and 6.03%, respectively, compared to the intact tube.