Comparison of methods for quantitative analysis of common cause failures - a case study

The beta factor method traditionally used for quantitative analysis of common cause failures (CCFs), does not allow a distinction between different failure multiplicities. Consequently, application of such a simple model to systems characterized by high level of redundancy may lead to excessively conservative estimates of the system failure probability. In the present report several higher-order methods, namely Binomial Failure Rate (BFR) method, Multiple Dependent Failure Fraction (MDFF) method and Multiple Greek Letter (MGL) method are applied to the same set of data. THe MDFF method, originally developed for a system of three identical units, has been modified and extended to a four-channel system. The MGL method gives the best agreement with the results directly assessed from the data set. Generally, as expected, the higher-order methods provide more realistic estimates of system failure probabilities than traditional models using only one CCF-parameter