High redundancy contributes to availability of safety systems. It is essential for a realistic assessment of this contribution to take possibilities of common cause failures of symmetrical components into consideration. This failure mechanism breaks the independence assumption for basic events in fault trees, causing the quantification of some events to become conditional on the state of others. This dependency has implications for various applications of a Probabilistic Safety Assessment (PSA) model where different scenarios - each possibly altering the state of events - are evaluated.
In this context, we examine the expected results when using a PSA model within the Significance Determination Process, in online risk monitoring and in different scenarios and expectations when calculating the Risk Increase Factor (also known as Risk Achievement Worth) in a normal PSA calculation. In all these applications, some components are considered unavailable. When these components share common cause failures with other components then there are multiple options how to perform the conditional quantification of common cause failure events. We outline different real-life situations and develop formulas that reflect the common cause logic involved. Furthermore, we argue for an implementation that allows users to specify which of the formulas should be applied, based on their knowledge of the scenario at hand.
Additionally, we discuss the symmetry assumption of standard parametric common cause failure models. The Basic Parametric Model relaxes this assumption because it does not require the total basic event probability in the quantification. This allows us to model situations where a conservative modeling approach would otherwise be necessary if standard parametric models were used. We also describe the new implementation of this common cause failure model in RiskSpectrum PSA.