F
Member
【Editorial Discussion】
A proof-test interval in a safety calculation is only as meaningful as the test that is really performed. If the procedure does not exercise a stuck contact, a bypassed channel, or a portion of the final element, the coverage you entered in the workbook is not the coverage you have. The editorial question is not “what the standard allows,” but what is physically toggled in your plant test.
1) List the dangerous undetected failures you are claiming to find, and walk the procedure against that list. A test that trips the channel but never moves the final element, or that uses a simulated input only, leaves a different residue than a test that strokes the valve or de-energizes the actual actuator.
2) Record what is bypassed to make the test possible. Temporary overrides, forced values, and bridged contacts are sometimes necessary; they also define what the test cannot see. The restoration of those bypasses is part of the test, not an afterthought.
3) Diagnostics that run online and the periodic proof test should be written as complementary, not as two copies of the same claim. If a diagnostic already covers a failure mode, the proof test should spend its time on what the diagnostic cannot see — and vice versa.
4) Evidence is the procedure, the as-found / as-left, and the identity of the equipment tested. A tick in a CMMS row named “proof test” is not coverage.
Open question: which part of the safety function do your proof tests most often fail to move — input, logic, or final element? Have you found a stuck condition that the periodic test had been silently missing? Practical experiences, without plant names, are welcome.
A proof-test interval in a safety calculation is only as meaningful as the test that is really performed. If the procedure does not exercise a stuck contact, a bypassed channel, or a portion of the final element, the coverage you entered in the workbook is not the coverage you have. The editorial question is not “what the standard allows,” but what is physically toggled in your plant test.
1) List the dangerous undetected failures you are claiming to find, and walk the procedure against that list. A test that trips the channel but never moves the final element, or that uses a simulated input only, leaves a different residue than a test that strokes the valve or de-energizes the actual actuator.
2) Record what is bypassed to make the test possible. Temporary overrides, forced values, and bridged contacts are sometimes necessary; they also define what the test cannot see. The restoration of those bypasses is part of the test, not an afterthought.
3) Diagnostics that run online and the periodic proof test should be written as complementary, not as two copies of the same claim. If a diagnostic already covers a failure mode, the proof test should spend its time on what the diagnostic cannot see — and vice versa.
4) Evidence is the procedure, the as-found / as-left, and the identity of the equipment tested. A tick in a CMMS row named “proof test” is not coverage.
Open question: which part of the safety function do your proof tests most often fail to move — input, logic, or final element? Have you found a stuck condition that the periodic test had been silently missing? Practical experiences, without plant names, are welcome.