Estimating Common Cause Failure Probabilities for a PRA Taking into account Different Detection Methods

The methodology to estimate residual parametric common cause failure (CCF) probabilities consists of the selection of the data source, source plants, source systems and component type, failure mode, assessment of the impact vectors, determination of equivalent observations, calculation of CCF rates of different multiplicities with uncertainties using an empirical Bayes estimation method and finally determining explicit CCF basic events and their probabilities to be used in the probabilistic safety assessment model. The CCF probabilities are obtained as the result of unavailability estimation accounting for different detection methods and corresponding outage times. Typically CCF events of safety system components are detected by tests during plant operation or during annual overhaul. In CCF quantification this is often regarded as the only way of detection. This leads into CCF unavailability quantification in which the CCF rate is based on all kinds of CCF events and the corresponding outage time is always determined by the test interval and testing scheme. This approach might be overly conservative or sometimes optimistic. This paper improves CCF unavailability estimation by taking into account monitoring and different kinds of tests and outage times and considering failure modes in the failure rate estimation.