Reliability prediction of bridge structures based on bayesian dynamic nonlinear models and MCMC simulation

Nowadays, the health monitoring systems of bridge structures in many countries have collected a large number of structural response data in the long-term monitoring period. However, how to use health monitoring data to assess the safety and serviceability of bridge structures has become the bottleneck in the field of structural health monitoring. What's more, the monitoring information-based time-dependent reliability prediction and assessment of existing bridges has been also one of the world-wide concerned problems in the field of infrastructure engineering. In this paper, to incorporate both historical monitoring data and realtime monitoring data in the prediction of time-dependent structural reliability, the Bayesian dynamic nonlinear model (BDNM) is introduced. For the Bayesian dynamic nonlinear model, the traditional way is the linearization of nonlinear model by Taylor series expansion technique, but its application range is small. In this paper, in consideration of the limitations in linearization of Bayesian dynamic nonlinear model, a more reasonable Bayesian nonlinear dynamic model is established based on the monitored stress data of bridges and a probabilistic recursive process of the nonlinear Bayesian dynamic model are completed through Markov Chain Monte Carlo (MCMC) simulation. Based on the built Bayesian dynamic nonlinear model of the stress and real-time monitored stress data, the reliability indices of bridge structures is solved and predicted real-timely with the first order and second moment method (FOSM). An actual example is provided to illustrate the application and feasibility of the Bayesian dynamic nonlinear model built in this paper.