Physics-informed Gaussian process regression for identifying longitudinal collaborative working performance parameters in continuous precast concrete girder bridges

Abstract

This study constructs a simplified mechanical model that characterizes the longitudinal collaborative working performance (LCWP) between spans through transverse wet joints at pier-top sections of the continuous precast concrete girder bridge. The model represents the load-transfer mechanism of wet joints using bending spring stiffness, thereby establishing a quantitative parameter for evaluating LCWP. Based on this, a mechanical system parameter identification model for the stiffness was established by combining whole-bridge traffic flow load and synchronous monitoring data of dynamic strain at the bottom of the middle span. Finally, the identification was achieved via a novel physics-informed Gaussian process regression parameter estimation algorithm proposed by the authors. Numerical simulations demonstrate the method’s capability for accurate direct identification of bending spring stiffness, confirming its robustness and reliability. Field application validates the method’s effectiveness in supporting practical maintenance strategies, highlighting its significant research value and broad engineering applicability.