Machine Learning-Based Crashworthiness Optimization of Crash Box Geometry by Adding Corner Shape Reinforcement Variations

This research presents a framework for estimating the crashworthiness of a crash box shape using a machine learning approach. The first shape mode used in this study was straight-line reinforcement with two different variant models that are triangle corner and square corner. Curved angle reinforcement was also used as the second shape mode with the convex and concave models as the variations. Both shape modes were compared to the conventional square tube to find out how this optimized structure gives an impact on reducing the peak force during the crash. In order to increase the crashworthiness of crash box shape, structural characteristics were employed as input data. The part machine learning technique’s Gate Recurrent Unit (GRU) model was used to predict the energy-absorbing characteristics and the prediction accuracy of different models, which were also utilized as optimization input. The results showed that the convex form generated a lower peak force of 282.1kN when compared to the concave shape, a difference of roughly 7%. As a result, it can be stated that optimizing crash box geometry employing angle reinforcement with a certain design will minimize peak forces for improved crashworthiness.