(47b) Validation of Spring Operated Pressure Relief Valve Time to Failure and the Importance of Statistically Supported Maintenance Intervals

The Savannah River Site operates a Relief Valve Repair Shop qualified by the National Board of Pressure Vessel Inspectors to NB-23, The National Board Inspection Code. Local maintenance forces perform inspection, testing, and repair of approximately 1200 spring-operated relief valves (SORV) each year as the valves cycle in from the field. The Site now has over 6000 certified test records in the Computerized Maintenance Management System (CMMS); a statistical summary of that data is presented in this paper. In previous papers, several statistical models were used to investigate failure rates including a quantal response method that was used to predict the failure probability as a function of time in service. Actual time to failure is in most cases not known; we only know that if the valve failed to pass the proof test then the ?valve stuck? or failed shut some time since the last proof test (censored data). ?Stuck shut? is industry defined as the valve not opening at less than or equal to 1.5 times the set pressure. This paper attempts to validate the assumptions underlying the statistical lifetime prediction results using Monte Carlo simulation. It explores the use of different failure probability distributions over discrete time intervals to evaluate the robustness of the applied statistical methodologies. This paper answers two questions: ?what are the survival functions over the chosen maintenance (inspection) intervals?, and ?what confidence should we have in basing proof test intervals on proof test results.?