ABSTRACT
This study aims to improve the evaluation of structural performance in prestressed concrete (PC) girder bridges by employing a calibrated FE model. This is achieved through Bayesian model updating, utilizing measured responses in the elastic stress state. Bayes’ theorem is employed to quantify uncertainties in FE model parameters by inferring the joint posterior probability density function (PDF). The Transitional Markov Chain Monte Carlo (TMCMC) sampler is utilized to generate samples for estimating the posterior PDFs of model parameters. The selection of a suitable model class is guided by sensitivity analysis of model parameters and a comprehensive discussion of the updating outcomes from linear models. The nonlinear static simulation results obtained through the calibrated model are rigorously compared with the measured crack pattern and load-deflection curve to validate the effectiveness of Bayesian model updating. Observations showed that structural damage resulting from concrete cracking minimally affects the 1st bending frequency. Consequently, vibration characteristics exhibit specific limitations in their application for assessing damages caused by concrete cracking. The proposed high-fidelity model updating approach, driven by multi-source data, establishes a reliable digital twin model suitable for SHM of PC bridges.
