Corrosion of reinforced concrete in nuclear plants: application of Bayesian networks to the risk management of cooling towers

Degradation modelling of concrete structures uses uncertain variables and leads, using reliability assessment, to time dependant evolution of failure probabilities. However, only few data are generally available to feed models leading to two types of uncertainties: an intrinsic one depending on the modelled phenomena and one related to the precision of the measurements. Each new data available is a piece of information which allows updating the initial prediction. In this article, an example of updating process, based on a Bayesian network, is presented and applied on the corrosion risk of a cooling tower in a nuclear plant. (authors)