Hybrid Uncertainty Quantification for Probabilistic Corrosion Damage Prediction for Aging RC Bridges

Abstract

A systematic framework is proposed to quantify hybrid uncertainties (i.e., aleatory and epistemic uncertainty) for the probabilistic prediction of corrosion damage in aging reinforced concrete (RC) bridges when the initial statistical parameters of variable are not available, sparse or cannot be accurately obtained. The key idea is to use a likelihood-based approach to calculate the probability distribution function (PDF) of the variable described by sparse data and an entropy-based transformation method to obtain the PDF of variable described by expert-based information. Following this, a hybrid description of uncertainties is proposed using the marginal integration. The uncertainty quantification of important factors on corrosion initiation and propagation are discussed, and a time-variant corrosion cracking model is developed. The proposed methodology is illustrated and demonstrated by a numerical example of corrosion damage prediction of an existing RC bridge. The prediction results of corrosion damage agree well with the experimental observations.