Definition:
Probability of Failure on Demand (PFD) measures the likelihood that a safety function will fail when required to operate. In reality, when discussing PFD, it is really PFDavg. It is the most core mathematical concept of all of functional safety. It is essentially the probability of something failing when it is needed.
Different levels of PFDavg relate to permitted safety integrity levels (SIL) and risk reduction factor (RRF). PFDavg serves as one of the three main pillars of every SIF in a SIS, see three pillars article.
The calculation of PFD average is its own very complex subject. First, there are various approaches to do the math. Such as simple equation, complex, and there’s also something called a Markov model which is the advanced method. The terms that go into PFDavg include things such as: Proof-test coverage (CPT), Proof-test interval (TI), failure rates, lifetime (LT), and mean time to restore (MTTR).
Key Points:
- Primary risk metric for low-demand SIFs.
- Calculated based on failure rates, proof testing, diagnostics, and architecture.
- The calculation of PFD average is its own very complex subject.
Example:
A SIF is fully calculated to have a PFDavg is 1E-2, the safety function has a 1% chance of failing on demand.
See Also: SIL, probability of failure per hour (PFH), failure rate (λ), RRF, markov model
Cited Source:
- IEC 61511-1:2016, Table 4
- IEC 61508-6:2010, Annex B
- Internal article – PFDavg Explained
- NTNU training deck