Probability of Initial Failure for Spring Operated Pressure Relief Valves

In several previous papers, we performed extensive analyses on CCPS PERD quality data from thousands of proof tests of spring operated pressure relief (SOPR) valves. In our studies, “stuck shut” failures have been of particular interest as they affect the safety integrity levels (SIL) achievable by safety functions which include SOPR valves. An important finding of past studies has been the severe negative effects on safety of so called “initial failures” – instances where a SOPR valve is installed in the “stuck shut” condition and, therefore, never provides any protection against overpressure events. This is a controversial issue because many believe that the probability of initial failure (PIF) is either zero or negligibly small; others believe PIF to be significant enough to require modeling in SIL calculations. In this paper we examine the latest proof test data set available from Savannah River Site (SRS) for evidence of the significance of PIF. This data set is especially important because SRS has a strictly enforced requirement that all valves (both new and used) be subjected to on-site proof testing prior to installation. These tests have revealed a number of stuck shut failures that would have been installed if on-site testing had not occurred. This paper will include specific examples of the underlying causes of stuck shut valve failures which offer incontrovertible evidence that initial failures do exist and, had they been installed, would have represented a significant number of the total stuck shut valves in the plant. Further we examine these initial failures for their association with valves of specific material compositions. We conclude that PIF is sufficiently significant that it should be included in models that calculate SIL for SOPR valves that are not tested immediately prior to installation.