A Study on Predicting Ship Hull Structural Responses in Collisions Based on Machine Learning

Ship collisions represent a prevalent and critical category of maritime accidents, often resulting in substantial equipment damage and significant human casualties. This study aims to evaluate the safety risks associated with ship collisions by employing finite element simulations to investigate the dynamic structural responses under various typical collision scenarios. Furthermore, a BP neural network model was developed based on the simulation data to predict structural damage outcomes across diverse collision conditions. The findings indicate that the model achieves a prediction error within 25% for post-collision structural damages, demonstrating robust reliability. However, the model’s accuracy in predicting scenarios involving minimal rupture remains a challenge and warrants further refinement.