Application of Reliability-Centered Maintenance to Boiling Water Reactor Emergency Core Cooling Systems Fault-Tree Analysis

Reliability-centered maintenance (RCM) methods are applied to boiling water reactor plant-specific emergency core cooling system probabilistic risk assessment (PRA) fault trees. The RCM is a technique that is system functionbased, for improving a preventive maintenance (PM) program, which is applied on a component basis. Many PM programs are based on time-directed maintenance tasks, while RCM methods focus on component condition-directed maintenance tasks. Stroke time test data for motor-operated valves (MOVs) are used to address three aspects concerning RCM: (a) to determine if MOV stroke time testing (as required by Section XI of the American Society of Mechanical Engineers Code) was useful as a condition-directed PM task; (b) to determine and compare the plant-specific MOV failure data from a broad RCM philosophy time period compared with a PM period and, also, compared with generic industry MOV failure data; and (c) to determine the effects and impact of the plant-specific MOV failure data on core damage frequency (CDF) and system unavailabilities for these emergency systems. The MOV stroke time test data from four emergency core cooling systems [i.e., highpressure coolant injection (HPCI), reactor core isolation cooling (RCIC), low-pressure core spray (LPCS), and residual heat removal/low-pressure coolant injection (RHR/ LPCI)] were gathered from Philadelphia Electric Company’s Peach Bottom Atomic Power Station Units 2 and 3 between 1980 and 1992. The analyses showed that MOV stroke time testing was not a predictor for eminent failure and should be considered as a go/no-go test. The failure data from the broad RCM philosophy showed an improvement compared with the PMperiod failure rates in the emergency core cooling system MOVs. Also, the plant-specific MOV failure rates for both maintenance philosophies were shown to be lower than the generic industry estimates. A significant decrease in CDF was found when comparing the plant-specific data with the generic data. Only a small difference in CDF was found between plant-specific RCM data compared with PM data, which can be attributed to the few number of essential MOVs modeled in the HPCI, RCIC, RHR/LPCI, and LPCS PRA fault trees.