Prediction of Deflection for Prestressed Concrete Girders Using a Bayesian Approach

To control the alignment of prestressed concrete bridge during construction, it is reasonable to treat the prediction of the deflection as an uncertainty. This paper presents a Bayesian updating approach to predict the deflection of prestressed concrete girders. A prior distribution is developed by using Monte Carlo stimulation with a proposed deterministic model under a variety of prestressing levels, material properties and environmental conditions. Then the posterior distribution is obtained by updating the prior distribution based on a limited number of initial measurements, thus greatly reducing the uncertainty of the deflection prediction. The method is applied to predict the camber of two actual prestressed T girders, and the predictions are satisfied.