Assessment of common cause failures and defensive measures for the representation of I&C in probabilistic models

This article presents an approach to improve the representation of digital I&C, while keeping the models simple and usable in probabilistic models of an installation, the so called SPINOSA approach. It relies on the combined use of a particular representation of I&C effects, the "Compact Model", and a sensitivity analysis based on "Beta Factors" representing potential dependencies due to hardware, software, human actions or interactions. It considers random mechanisms and systematic mechanisms, assessed by a combination of probabilistic and deterministic approaches. The framework used to assess the systematic failures due to hardware, software and human actions is partly presented here, i.e., the taxonomy of software related failure mechanisms and associated defence measures necessary to assess associated factors. The taxonomy is based on a general system failure model and an identification of faults, effects, activating events, common cause contexts. We expect from this framework an analytical approach representing a significant improvement compared to holistic assessment approaches such as IEC 61508 and its declinations. It will also be the starting point of effective FMEA dedicated to digital systems.