Multi-objective seismic design optimization of self-centering bridges with novel mechanical hinges

Abstract

Self-centering rocking bridge piers possess significant potential for improving post-earthquake resilience and reducing the post-earthquake repairs for bridges. However, due to the multi-parameter nature of such piers, challenges remain in designing the self-centering bridge piers and ensuring balanced performance across multiple piers. To addresses these gaps, this study proposes a general multi-objective seismic design optimization framework for self-centering rocking bridges. The framework includes input-output dataset preparation, XGBoost model training, Pareto front calculation, and optimal solution selection. The framework is illustrated on a self-centering bridge equipped with hemisphere-based rocking hinges (HRH), an innovative device previously developed by the authors. Results show that using a machine learning method (XGBoost herein) to develop surrogate model as a substitution for the finite element model improves optimization efficiency by at least 73 %, without compromising accuracy. The optimized bridge with HRHs achieves balanced performance among piers, with a 63.4 % decrease in shear difference and a 11.4 % reduction in displacement difference compared to the benchmark bridge.