Bayesian procedures for updating deterioration space-time models for optimizing the utility of concrete structures

•Conditional surrogate models for the core sample data are proposed.•Updating accounts for correlations of chloride content in both time and space.•Computationally efficient for portfolio analysis of a large number of structures. The dominant mode of deterioration of concrete bridge structures in North America is corrosion associated with the ingress of chloride ions from salt spreading. Finite element and finite difference models can be utilized to predict the chloride ion content as a function of space and time in concrete and to estimate the time to the corrosion initiation of the reinforcing steel. The input parameters to these models include environmental exposure data, salt spreading protocols and concrete properties. Data on environmental exposure and salt spreading protocols can be obtained from meteorological stations and roadway operators respectively such that the remaining uncertainties are mainly related to concrete properties. Prior distributions on concrete diffusion properties can be specified using compiled databases of experimental data and probabilistic methods are used to propagate uncertainties to derive the prior distribution of chloride ion content as a function of depth and time. When chloride content are available from core samples, Bayesian updating procedures are proposed to update the probability distribution functions of concrete properties considering the type of exposure, the chloride ion profile and time of sampling. The proposed procedure can be applied for any number of core samples sampled at different times and accounts for correlations between chloride content predictions at different times and depths. The proposed procedure is demonstrated for an existing bridge located in Montreal, Canada.