Approach to assess the occurrence probability of a long-term evolution scenario of a long-lived nuclear waste repository

Abstract

The building of a deep geological repository for the management of long-lived radioactive waste requires the implementation of a structure that allows the storage cells and the galleries to be closed after the waste packages have been put into place. These structures are known as ‘seals’ and they play an important role in demonstrating that the repository will not have any effect on people or the environment over the long term. Therefore, assessing how these seals can fail is needed to determine the feasibility of a long-lived, highly radioactive waste repository. Traditionally, safety assessments in the area of nuclear waste are deterministic and therefore overestimate the consequences of failure as they are based on the least favourable hypotheses. The current paper puts forward a complementary approach to determine the probability of occurrence of a ‘sealing fault’ scenario and shows how it has been applied on the basis of expert opinion and Bayesian networks. This work is based on research carried out by ANDRA (the French National Radioactive Waste Management Agency), which is presented in its report Dossier 2005. This report uses a conservative deterministic approach and demonstrates that the sealing fault scenario would not lead to harmful consequences. The present paper adds to the findings of this report by evaluating the probability of occurrence of this scenario based on an expert assessment. Due to the time scales and the related uncertainties, the evaluation of the probability of occurrence of the sealing fault is delicate. Thus, the limits to this exercise and possible improvements are also presented herein.