Accelerated carbonation tests for the probabilistic prediction of the durability of concrete structures

This paper proposes an efficient approach based on Bayesian networks for designing the cover of CEMI-based concrete structures for a prescribed durability under carbonation exposure. The approach combines the use of a carbonation model, whose correlated parameters are the nodes of a Bayesian network, and an experimental investigation on accelerated carbonation tests under high pressure (50% atmospheric pressure), the results of which allow the distributions to be updated. Several updates are performed at 5 days, 7 days and 10 days of exposure. The reliability-based design of cover is subsequently carried out using posterior distributions selected on the basis of the Bayes factor and the maximum likelihood criteria. Percentile values of the carbonation depth resulting from measurement and output of the model are compared, underlining the interest of the approach, which is valuable for practical considerations in engineering design.