A min cut-set-wise truncation procedure for importance measures computation in probabilistic safety assessment

A truncation process aims to determine among the set of minimal cut-sets (MCS) produced by a probabilistic safety assessment (PSA) model which of them are significant. Several truncation processes have been proposed for the evaluation of the probability of core damage ensuring a fixed accuracy level. However, the evaluation of new risk indicators as importance measures requires to re-examine the truncation process in order to ensure that the produced estimates will be accurate enough. In this paper a new truncation process is developed permitting to estimate from a single set of MCS the importance measure of any basic event with the desired accuracy level. The main contribution of this new method is to propose an MCS-wise truncation criterion involving two thresholds: an absolute threshold in addition to a new relative threshold concerning the potential probability of the MCS of interest. The method has been tested on a complete level 1 PSA model of a 900 MWe NPP developed by “Electricité de France” (EDF) and the results presented in this paper indicate that to reach the same accuracy level the proposed method produces a set of MCS whose size is significantly reduced.