Optimizing seismic fragility assessments for high-speed railway track bridges: a novel multi-parameter methodology and intensity measure selection

This research innovatively presents a multi-parameter seismic fragility analysis methodology anchored on Blending ensemble learning tailored for the high-speed railway track-bridge (HSRTB) system. Leveraging ensemble machine learning models, a seismic demand model encapsulating multiple earthquake intensity measures (IMs) is formulated. The research progresses to deduce the mean seismic fragility representations for single, dual, and composite earthquake parameters. Key findings of this investigation underscore that the advanced methodology aligns well with the Monte Carlo (MC) fragility, signifying its robustness. Crucially, earthquake IM parameters, Sa(0.5) and Sa(0.3)_ASI, emerged as the optimal choices for singular and dual-parameter seismic fragility delineations. Further enriching this research, a composite parameter optimization technique rooted in multi-CPU genetic algorithms is proposed. This avant-garde method remarkably minimizes the fragility estimation error, emphasizing its unparalleled efficacy in enhancing the precision of fragility evaluations.