A Bayesian strategy for forecasting the leakage rate of concrete containment buildings – Application to nuclear containment buildings

Abstract

The forecast of large concrete structures behavior such as Nuclear Containment Buildings (NCB) is a critical safety issue, whether for the structure itself or for its surrounding environment. Over the past few decades, complex finite element models have been developed for assessing the mechanical behavior and the leak tightness of large concrete structures. These models involve numerous parameters known with uncertainties, which importance could jeopardize the reliability of forecasts. Therefore, this contribution aims to present a weakly coupled Thermo-Hydro-Mechanical-Leakage methodology to forecast the evolution of dry air tightness of large concrete structures. The proposed approach is based on a Bayesian updating strategy aiming to reduce uncertainties on forecasts by combining information provided by both a physical model and in situ observations acquired up to date. The Bayesian strategy is coupled with a scale reduction method for ensuring applicability at the scale of large structures. The methodology is applied to a 1:3 scale NCB, in order to improve confidence level of leakage rate forecasts by taking advantage of monitoring measurements. The forecasted leakage rates are in good agreement with the measurements, and the proposed methodology enables to significantly reduce the uncertainty throughout the operational phase of the structure, even with a limited amount of monitoring data.