I’ve been receiving several inquiries lately regarding the impact of the Partial Valve Stroke Test (PVST) interval on the failure rates of a final element. Specifically asking how does exSILentia® deal with this? exSILentia® allows users to specify if partial stroke testing is performed on their final element configuration. In addition to the “yes/no” checkbox, the tool provides some additional features. For example, users can specify the Partial Valve Stroke Test interval, i.e. the frequency at which the partial stroke test is performed.
The frequency of the PVST interval can have an interesting impact on the way the PVST test is implemented in the SIL verification calculations, specifically relating to the question if the test can be considered “automatic” or if it should be considered an additional manual proof test. The rule of thumb is that if a diagnostic test is performed at least an order of magnitude more frequent than the proof test, the performance evaluation does not need to consider the interval at which the diagnostic test is performed. In other words, if the diagnostic test is performed 10 times within one proof test interval, the diagnostic test can be considered automatic from a reliability calculation perspective.
If a test is considered automatic, the effectiveness of the test is represented in the failure rates. For a final element, one will see that with PVST the final element will have both detected and undetected failures, whereas without PVST all failures are undetected (assuming a typical pneumatic actuator-valve combination). Assuming a 60% PVST coverage, one will see that of the original dangerous undetected failures, 60% is classified as dangerous detected and 40% remains classified as dangerous undetected.
If a test is not considered automatic, the effectiveness of the test is not represented in the failure rates. In this case the SIL verification calculation will consider an extra proof test interval, for example performed after 4 months within a 12-month proof test interval, with a coverage of the extra proof test equal to 60% (per the example above). In this case, despite that PVST is implemented, the failure rates of the final element remain classified as undetected.
When plotting the Probability of Failure on Demand (PFD) as a function of time for a final element we can clearly recognize the differences in calculation implementation. Figure 2 shows a final element with a 12-month proof test interval and PVST performed monthly. Figure 3 shows a final element with a 12- month proof test interval and PVST performed every 4 months. The latter figure clearly shows the additional “saw tooth” due to the extra proof test.
Check back tomorrow to learn why Architectural Constraints may change…