Tero Tyrväinen Risk Importance Measures and Common Cause Failures in Dynamic

Traditionally, fault tree analysis has been the leading method for reliability analysis of complex systems. However, dynamic properties of systems cannot always be described with adequate accuracy using fault trees. Dynamic reliability analysis has been studied widely since the 1990s. Some dynamic calculation tools have been developed but they cannot compete with fault tree analysis tools yet in reliability analysis of nuclear power plants. Dynamic flowgraph modelling (DFM) is an approach for reliability analysis of dynamic systems. DFM models are directed graphs whose nodes can contain a finite number of states. A system’s dynamics is described by discrete state transitions. As in the fault tree analysis, the essential goal of dynamic flowgraph modelling is to identify root causes that lead to a system’s failure. VTT has been developing a DFM-based reliability analysis tool called YADRAT since 2009. DFM models have been analysed previously by transforming them into sets of timed fault trees from which the root causes of the system’s failure have been identified. In YADRAT, the model that describes a system is transformed into a binary decision diagram. Risk importance measures and common cause failures are a significant part of reliability theory and fault tree analysis but they have not been studied much in relation to dynamic flowgraph modelling. Risk importance measures are used to measure how important different components are with regard to the system’s reliability. In this thesis, dynamic risk importance measures based on two traditional risk importance measures are formulated so that they take the multi-valued and dynamic logic of DFM models into account. Dynamic risk importance measures can be calculated separately for different failure states of components so that they provide more detailed information on how different components contribute to the system’s failure. In addition, dynamic generalisations are developed for traditional parametric common cause failure models. In these common cause failure models, the possibility that failure events can occur at different time points is considered. Dynamic risk importance measures and common cause failure models are implemented in YADRAT. Date: 24.11.2011 Language: English Number of pages: 87+12