Problems linked to sawtooth waves

PFDavg and PFH are both average values of time-dependent parameters U(t) for PFDavg and w(t) for the PFH. When the SIS components are tested periodically, U(t) and w(t) take the shape of typical sawtooth curves that increase gradually, peaking just before the tests and falhng to low values just after the tests. 

The first idea is to hmit the acceptable maxima values of U(t) or w(t) but this is not enough because we can accept them too high for short periods (during testing for example) if particular precautions are implemented. The second idea is to limit (e.g. 10%) the cumulative time spent above the upper hmit of the announced SIL. This 

If dangerous failures of lEC standards match the unsafe failures above, this is not the case for safe failures. Indeed, in their current editions, any non-dangerous failure is considered safe. This allows the suppher to consider safe any kind of failure (e.g. SU failure) without any relationship to the safety function in question and to announce the SFF safe failure fraction, i.e. proportion) to be much higher than in reality. 

Beyond the safe/unsafe partition, two major classifications of failures occur 

Even if failures in operating are most often revealed, and failure on stand-by failures unrevealed, they are different concepts that should not be confused. A (too) fast reading of standards often leads to an assimilation of revealed failure s safe failure, which of course is true only after an action has been successfully undertaken 

---