ABSTRACT
Common Cause Failures (CCFs) are failure events that occur to technological systems that use redundant components to serve as multiple layers of defense and fail multiple components due to the same shared cause. The standard approach to CCF modelling is the use of parametric models (Apostolakis and Moieni 1987; Fleming et al. 1983). Parametric models, such as the Alpha Factor model (Siu and Mosleh 1989), describe mainly the manifestation of CCFs on the system and offer limited insight into the way the system design relates to CCF risk. As such, parametric models are well-suited in analyses within a regulatory context, where the main aim is to demonstrate that existing systems meet specific reliability targets. However, such models have limited use in supporting the design process of a new system. In these cases it is important to be able to model possible design, managerial or operational decisions, and capture the effect of these decisions on the susceptibility of the system to multiple failures. Furthermore, if operational data does not exist, the analysis for the determination of CCF risk is based on observations coming from similar systems. Nevertheless, one needs to take into account the existing differences between the already operating and the new system, such as the different levels of redundancy. Statistical techniques called mapping techniques have been suggested to transform the data so that the information represents a system of the same size as the target system (Vaurio 2006), but these procedures still remain problematic.
