MODELING AND ESTIMATION OF COMMON CAUSE FAILURE LIKE PRE-INITIATOR HUMAN FAILURE EVENTS USING EXPERIENCE DATA

摘要

Motivated by an idea that CCF-like pre-initiator HFEs might already have been counted in CCF probabilities, CCF experience data for pressure transmitters, level transmitters, and channel bistables from the CCF Database and Analysis System developed by the US NRC were investigated. The investigation confirmed that CCF-like pre-initiator HFEs were included in the CCF data base for the selected components. Various HFEs caused CCFs and they can be grouped into two general categories, HFEs during calibration and HFEs in restoring components after previous activities. Also, for all three components, CCFs caused by HFEs, if ranked by the number of retrieved events, were the second major events. However, if ranked by the contributions to the CCF basic events probabilities which were measured by the contributions to the expected number of CCF basic events, N_k (k = the number of failed components), CCFs caused by the HFEs were the most dominant contributors in level transmitters and channel bistable cases. For CCF group size of 4, the contributions of CCFs caused by HFEs to N_4 was 86.1 % and 100% for level transmitters and channel bistables, respectively. For pressure transmitters, CCFs caused by HFEs contributed to about 2.5% of N_4 and about 20% for N_2 and N_3. Based on the results of this study, it was concluded that, for those three components, it would be better not to model CCF-like pre-initiator HFEs as separate HFE basic events but let them be included in CCF basic events because: 1) Contributions from CCF-like pre-initiator HFEs were already included in their CCF probabilities which can be estimated based on actual CCF experience data and 2) a simple HFE basic event alone may not be able to represent well the complexity and variety of HFEs which can cause CCFs. It is believed that the same conclusion can be applicable to other types of instrumentation components.