A corrugated steel fender for bridge pier protection against truck collision: Multi-objective optimization design

This paper proposes a multi-objective optimization design method for bridge anti-collision fenders that aims to ensure the post-impact serviceability of the bridge pier while simultaneously enhancing vehicle safety. To this end, a finite element model of the vehicle-fender-bridge was established. Based on simulated results from 40 crash scenarios, three design objectives were identified: the post-impact damage level of the bridge pier, the peak impact force sustained by the vehicle, and the energy dissipation of the fender. Subsequently, a comprehensive dataset was constructed, comprising 112 samples detailing the vehicle mass and velocity, the bridge pier and fender design parameters, and the corresponding values of the three design objectives. Machine learning algorithms, including Decision Trees, Random Forest, and Extreme Gradient Boosting Trees (XGBoost), were then employed to develop surrogate models for predicting three design objectives. The results demonstrate that the Random Forest classification model achieves the highest prediction accuracy for post-impact performance levels, while the XGBoost and Random Forest regression models excel at predicting the peak impact force and energy dissipation, respectively. The SHapley Additive exPlanations (SHAP) algorithm was subsequently used to determine recommended ranges for fender design parameters, tailored to different post-impact serviceability levels of the pier. Finally, integrating the developed models with SHAP-derived recommendations, this paper establishes a multi-objective design framework to guide the selection of anti-collision fender parameters. Under the constraint of a specified post-impact serviceability level for the pier, the framework seeks to minimize the peak impact force on the vehicle and maximize the energy dissipated by the fender.