Trend analysis in input data for PSA

Abstract

The focus of this paper is to form a conclusion on the validity of the postulate on absence of trends in reliability data for Nordic nuclear power plants. Trends in this case imply an increasing reliability parameter with time, corresponding to degradation in component malfunction frequency. A non parametric test method, the Wilcoxon rank sum test, was applied to data sets representing observations of component malfunctions presented in the T-book. Linear trends of magnitude corresponding to that can barely be detected by the test were postulated and added to tabulated reliability parameters. The trend containing parameters were extrapolated in time till today and used in PSA models, comparing results to those obtained while running the same sequences with tabulated parameters. The conclusions of this thesis are that time dependant trends can be observed in reliability data today. These trends are, in one case, insignificant with respect to PSA results. In the other case the trend is most likely decreasing with respect to frequency, yielding a lower malfunction probability, instead of a heightened as was postulated.