Reliability prediction of bridge structures based on BDNM

To predict the bridge structural reliability based on the monitoring information and the priori model of stress parameters (mean), the dynamic measure of structural stress over time is treated as a time series, and considering the limitation of the BDLM, a Bayesian dynamic nonlinear model (BDNM) is then introduced. State equation and monitoring equation of monitoring stress are established with BDNM. Then the BDNM is approximately transferred into Bayesian dynamic linear model (BDLM) by Taylor series expansion technique, and the monitoring information is monitored by bayes factor. Combining parameters' prior information with the early stress data containing noise, the stress state parameters are deduced with Bayesian Posterior Probability. A dynamic model is built to forecast the changing trend of structural stress. To allow for the epistemic uncertainty in variance estimation based on monitoring information, a discount factor approach is made for specification of unknown variance. Finally based on the built BDNM and the FOSM method, the structural reliability is predicted, and the feasibility and application of the built model is illustrated by an actual example.