Reliability Data Assessment for PSA

In accordance with the German Guide for Methods for PSA, in general only plant specific data should be used for reliability analyses. For this purpose GRS has developed a computer based reliability data system which contains databases and software for the collection of maintenance reports, for the assessment of failure data and also mathematical programs for statistical methods. This system also serves for the purpose of a Living PSA. As of today, the system contains plant specific databases for eight NPP (4 PWR and 4 BWR). The generation of reliability data for a PSA is performed in 3 steps: -1. Selection of the components, which are considered to be a homogeneous component group with the same reliability characteristic. 2. Selecting the failure events depending on the required failure mode. 3. Calculating the failure distribution from the selected failure events and related observation times. All mathematical methods for the calculation of the failure distribution are based on the Bayesian approach. In general, only plant-specific data from operational experience of the considered components are applied. In the case that only few plant specific data are available (e.g. stand-by pumps for the failure mode 'Failure to run') generic data of comparable components in other plants is included. In the case of a complete lack of plant specific data, a generic distribution is used. (authors)