Development of a computer code for common cause failure analysis

Abstract

COMCAF, a computer code for the common-cause failure analysis, is developed to treat the common-cause failures in nuclear power plants. The common-cause failures are extremely important because they are the multiplecomponent failures occurring from a single cause that is common to all of the components. In the treatment of common-cause failures, the minimal cut sets of the system are obtained first without changing the fault-tree structure. The occurrence probabilities of the minimal cut sets are then calculated accounting for the common-cause failures between components in the same and/or in the different minimal cut sets. The effects of common-cause failures are confined within the prescribed common-cause component groups. The basic parameter model is used to model the common-cause failures between similar or identical components. For the dissimilar components, the assumption of symmetry used in the basic parameter model is applied to the basic events affecting two or more components. The top event probability is evaluated using the inclusion-exclusion method. In addition to the common-cause failures of components in the same minimal cut sets, failures of components in the different minimal cut sets are also easily accounted for by this method. This study applied this common-cause failure analysis to the PWR auxiliary feedwater system. The top event probabilities are evaluated for the two cases:common-cause component groups having components in the same minimal cut sets and those having components in the different minimal cut sets. The results in the top event probability for the system are compared with those of no common-cause failures. The top event probability increases when we consider common-cause failures between components in the same minimal cut sets, which is of course a well known fact. However, if we consider common-cause failures between components in the different minimal cut sets, the top event probability decreases considerably.