Getting Good Proof Test Coverage Numbers
Tuesday, April 05, 2011Viewed 9668 times
Proof Testing • (2) Comments • Permalink
Several years ago we recognized that proof test coverage was an important variable that must be considered when doing PFDavg calculations. We ran some models and discovered that the difference between “perfect” proof test coverage (100%) and a very good 90% could mean a whole SIL level in the result. The first question I ask is ‘Why do some engineers still use simplified equations that assume perfect proof test coverage?” Unfortunately the answer I get when I ask them is a question, “What is proof test coverage?” Not a good sign that those doing the work have the competency needed. But that is another topic for another blog.
Given that proof test coverage is so important, how are good numbers obtained? Fortunately exida recognized that the FMEDA technique could be used to calculate a proof test coverage factor (PTC) for any given proof test. So this task was added to our FMEDA procedure several years ago. Now we have a bountiful database of PTC numbers for a variety of proof test techniques for hundreds of various instruments. These have all been incorporated into the proof test generator option for exSILentia, exida’s safety lifecycle engineering tool. In the future look for better tables in our new books.
Tagged as: sil, ptc, proof test coverage factor, proof test coverage, pfdavg, fmeda, exsilentia, dr. william goble,
Hi William,
About proof test, Would you like clarify something for me?
Where I can get the proof test procedure? FMEDA OR somewhere else? most instruemtn manual do not show the detail.
How to make test interval reasonale and effective? for example we use 3051S pressure transmitter in SIL1 function loop and test interval is 12 month. May I extend to 18 month or longer as SIL calculation is not a concern with 36 month test interval.
thanks for your calrify in advance.
How we can estimate the proof test effectiveness? in the SIL Verification Analysis we usually assume perfect proof testing but this is not the case in reality - please explain?