Model Updating and Seismic Analysis of Concrete Filled Steel Tubular Arch Bridge Based on Data Driven

To enable finite element model (FEM) updating and seismic vulnerability analysis of long-span CFST arch bridges, this study proposes a data-driven method combining an optimized BP neural network surrogate model and an improved grey wolf algorithm (IGWO). The IGWO integrates dynamic mutation, fuzzy convergence factors, and an elite archive strategy to enhance parameter optimization. A multi-objective function is constructed using static and dynamic response data under varied loads. Applied to a CFST arch bridge on the Nanning South Ring Expressway, the updated FEM improves accuracy, with surrogate model prediction errors below 4%. The IGWO demonstrates superior optimization efficiency and robustness. Seismic vulnerability analysis reveals that post-update vulnerability curves align with pre-update trends but show slightly higher damage probabilities, reflecting calibrated energy dissipation and deformation capacity. The results validate the method’s effectiveness in refining structural seismic performance assessments.