Crashworthiness Prediction and Optimization of Rib-Reinforcement Multi-Cell Tubes Based on Machine Learning Under Axial and Lateral Loadings

In recent decades, the search for lightweight and high-performance energy-absorbing structures and materials has made porous thin-walled structures a prominent research area due to their light weight and high energy-absorbing properties. However, due to the more intricate nonlinear interaction between the energy absorption properties and the multicellular structure topologies, it is still challenging to identify the optimal configuration effectively. Therefore, a deep-learning-based technique for predicting the energy absorption properties of thin-walled structures is presented using rib-reinforcement multi-cell tubes (RMTs) as an example. The energy absorption parameters of the thin-walled structure with varied reinforced rib configurations were derived through finite element analysis. Based on the acquired energy absorption properties, the neural network was trained and verified. The rib configuration with the optimum energy absorption performance was identified using a trained neural network. Finally, using the non-dominated sorting genetic algorithm II (NSGA-II), a thickness gradient based on multi-objective optimization of the discovered optimal configuration was carried out. The results indicate that the artificial neural network (ANN) is able to predict the energy absorption performance of RMTs with a high degree of precision, and the optimized structure effectively suppresses the peak crushing force (PCF) with a significant improvement in both the specific energy absorption (SEA) and the crushing force efficiency (CFE). In summary, the prediction capabilities of MLP in addressing difficult engineering issues are not only proven in this study, but an effective prediction tool is also offered for finding the optimal design of thin-walled structures in the future.