Failures on demand of a reactor core isolation cooling (RCIC) system in BWRs are the most frequent events of limiting conditions for operation during 1982-2006 in Japan, according to data gathered in Nuclear Information Archives (NUCIA). In this work, probabilities of failures of the RCIC system are analyzed by using the hierarchical Bayes method. The failures on demand of the RCIC system are classified into two groups; one is related to the demand at a periodical inspection test, which is performed almost every 13 months at the end of the periodical inspection of the nuclear power plant, and the other is related to the monthly surveillance test during plant operation. The hierarchical Bayes analysis shows the characteristics of probabilities of failures of each Japanese plant and also that probabilities of failures at the periodical inspection test are quite different from those at the surveillance test, comparing Japanese nuclear power plants with American ones. This paper provides a new approach to analyzing sparse failure data taken from nuclear power plants in Japan.
