Lessons drawn from probabilistic studies carried out by Electricite de France

Since the beginnings of the French nuclear power programme, Electricite de France has made special efforts to develop probabilistic methods and has undertaken a comprehensive programme of reliability studies on the safety systems at the Fessenheim nuclear power station. Appropriate analysis techniques have been developed for this purpose, and a reliability data bank and a system for collecting such data at the power stations have been set up. The results of this programme confirm that deterministic design rules are justified and lead to a high level of safety in nuclear power plants. The probabilistic studies also showed that too much attention is given to highly improbable events. This might detract from the ''homogeneous safety'' of facilities if insufficient efforts are made to analyse sequences initiated by more frequent transients. For the designer, the greatest interest of probabilistic methods lies in the qualitative analysis preceding any quantification procedure. It constitutes a synthesis of the design of the system, its installation, its operation, and its interactions with other systems. In particular, it makes it possible to identify the possible causes of common failure of any origin. Recognizing the helpfulness of reliability studies, Electricite de France nevertheless considers that deterministic rules and criteria must remain the basis of design methods. Efforts to make these rules and criteria more precise and in some instances more complete have increased as a result of the accident which occurred at the Three Mile Island nuclear power station in 1979. Examples are the studies concerning assistance to diagnostics and the use and improvement of operating procedures. (author)