Guidelines for reliability analysis of digital systems in PSA context — Final report
Stefan Authén, Jan-Erik Holmberg, Tero Tyrväinen, Lisa Zamani,
The objective of the DIGREL project has been to provide guidelines to analyse and model digital systems in the context of probabilistic safety assessment (PSA). A failure modes taxonomy for digital I&C systems has been developed jointly with OECD/NEA Working Group on Risk Assessment. Reliability modelling has been studied by developing a fictive, simplified PSA model representing a four-redundant distributed protection system. The evaluation of the example PSA has demonstrated the developed taxonomy and verified that it is suitable for PSA purpose. The evaluation shows that the choice of the level of abstraction for the modelling of digital I&C is of high importance for the results. Module level is recommended. Both undetected and detected hardware as well as software failures contribute significantly to the PSA results, indifferently of the assumed fault tolerant design. Similar conclusion can be drawn from the test of using different CCF parameters for undetected and detected failures. Software faults have a non-negligible effect on the results due to their functional impact on all divisions. In order to develop a realistic fault tree model for a digital I&C protection system it is vital that the chosen fault tolerant design is fully understood and correctly described in the model. The treatment of faulty inputs and degraded voting logic sets the foundation of the fault tree analysis.
Digital I&C system, probabilistic safety assessment, reliability, nuclear power plant safety