LESSONS LEARNED IN REVIEWING CCF EVENTS

The U.S. Nuclear Regulatory Commission (NRC) has an ongoing Common Cause Failure (CCF) data analysis program that periodically collects and evaluates information on component failures at U.S. commercial Nuclear Power Plants (NPPs). The primary information sources include the Licensee Event Reports (LER) and records from the Equipment Performance Information Exchange (EPIX) program. Once the information is collected, the failure records are evaluated to identify potential CCF events. CCF events are then coded, reviewed, and loaded into the NRC's database. Verification of the CCF events is performed with the intended purpose of ensuring that events entered into the CCF database are indeed CCF events and that the event coding is consistent and correct. To ensure technical accuracy and correctness of the events loaded into the CCF database, the NRC requested the Pressurized Water Reactors Owners Group (PWROG) support in reviewing these events. Reviews of multiple data sets of CCF events were conducted on behalf of the PWROG. The data sets included CCF events that have occurred at U.S. commercial nuclear power plants. CCF events that occurred during 2006 through 2007 were included in the most recent data set that was reviewed. The level of information provided for reported CCF events varies from utility-to-utility. Without utility participation or input, the lack of consistency and varying level of detail can lead to incorrect interpretation and classification of a CCF event regarding its Probabilistic Risk Assessment (PRA) impact. This paper offers lessons learned from the reviews that were conducted. Insights for improving the consistency and level of detail related to the PRA information are summarized in this paper. The leading causes of initial misclassification of CCF events and patterns observed in conducting the reviews are discussed. The resolutions of misclassified CCF events are also discussed as part of the evaluation process to enhance the pedigree of the CCF database.