Guidelines for reliability analysis of digital systems in PSA context — Phase 4 Status Report
Stefan Authén, Jan-Erik Holmberg, Linda Lanner, Tero Tyrväinen,
DIGREL develops practical guidelines for analysis and modelling of digital systems in probabilistic safety assessment (PSA) for nuclear power plants. The project consists of three interrelated activities. A taxonomy for failure modes of digital I&C systems has been developed by a task group of OECD/NEA Working Group RISK. In the parallel Nordic activity, a fictive digital I&C PSA-model has been developed for the demonstration and testing of modelling approaches. The third activity has been to develop a method for the quantification of software reliability in the context of PSA, which is reported in another publication.
The failure modes taxonomy is based on a failure propagation model and a definition of five levels of abstraction: 1) system, 2) division, 3) I&C unit, 4) I&C unit module, 5) basic component. The failure propagation model constitutes of the following elements: fault location, failure mode, uncovering situation, failure effect and the end effect.
An existing simplified PSA model has been complemented with fault tree models for a four-redundant distributed protection system in order to study and demonstrate the effect of design features and modelling approaches. I&C unit-level and module-level modelling were compared. Modelling on the I&C unit level of abstraction can result in large conservatisms that may produce misleading results e.g. regarding dominating core damage sequences and significance of I&C failure modes. Both undetected and detected failures contribute significantly to the PSA result, indifferently of the assumed fault tolerant design.
Two different modelling approaches were compared. In RiskSpectrum the system failure logic is represented by graphical fault trees while in FinPSA a so called communication network representation was applied. Same minimal cut sets were obtained except some differences in CCF calculations and truncation of minimal cut sets with small probabilities.
Digital I&C system, probabilistic safety assessment, reliability, nuclear power plant safety