Reliability analysis and safety evaluation on a nuclear power plant

In many developed countries, Nuclear Power Plant have been built in the 60's and 70's, especially after the oil crisis. These plants are now often more than 10 or 15 years old and periodic safety re-evaluations are requested to demonstrate their ability to stay in operation with a satisfactory safety level. Among the different aspects of the safety re-evaluation, the analysis of the technical behaviour of the main safety-related components is necessary to demonstrate the conformity with the design hypothesis. In this context, the French nuclear Power plant of St-Laurent-des-Eaux developed tools in order to monitor the reliability parameters of the safety related components (failure rates per demand, hourly failure rates, mean times to repair, unavailability rates....). These parameters are calculated on a yearly basis using the Bayesian statistical method. The study of the evaluation of these parameters, since the start up of the plant in the early 80's, is a useful tool to help in identifying the ageing phenomena and validating the maintenance policies. The paper describes the method used to monitor the reliability of the components and gives some examples of interesting behaviours. The main safety related components are taken into account (more than 2,500 on each unit, grouped in one hundred families). Twelve years after the start up, no ageing phenomena involving any safety problem can be shown. In some cases, the positive effects of design modifications, or operating practices (including preventive maintenance rules) are put in evidence. We have also found in some other cases, generic or typical failures are associated with the same kind of component. These reliability analyses are still under development and must be extended to the periodical test analysis and maintenance optimisation; but the actual results are still encouraging.