Multi-Objective Crashworthiness Optimization of Variable-Thickness Expansion Tubes Using Machine Learning and Decision-Making

While traditional expansion tubes exhibit excellent energy absorption, their uniform wall thickness limits lightweighting and performance optimization. Graded thickness designs can reduce the initial peak crushing force (IPCF) and enhance material efficiency. This paper proposes a variable-thickness expansion tube integrating high energy absorption with tailored mechanical response. Material tensile tests were conducted to determine the constitutive relationship, and axial compression experiments on expansion tubes were performed. Numerical simulations were validated against experimental results, establishing an accurate finite element model. The influence of design parameters on crashworthiness indicators was analyzed via orthogonal experiments. A fully connected neural network with a feature importance layer was then constructed to efficiently replace computationally expensive simulations. Key performance indicators—including IPCF, total energy absorption (EA), and structural mass (m)—were synergistically optimized using a multi-objective genetic algorithm. Finally, the entropy weight–gray relation–TOPSIS method was employed to select the most satisfactory solution from the Pareto front. The relative discrepancies between the selected solution and finite element simulations are 3.65% for EA, 0.23% for mass, and 4.37% for IPCF, confirming the framework’s reliability. This study establishes a systematic design approach combining machine learning, multi-objective optimization, and multi-criteria decision-making for high-performance energy-absorbing structures.