The machine safety community has defined a number called the “B10 number.” It is a measure of time where 10% of a population of devices should have failed. Generally it is a measure of expected end of life or “useful life” as defined by the reliability engineering community.
The IEC/ISO 13849-1 functional safety standard for simple machines has defined equations directly relating the B10 number to the average random failure rate in the time period before end of life. For any given application, the cycle rate is estimated and the “random failure rate” is calculated. I heard one engineer state in an email “This is great. We finally are getting failure rate data from these safety relay companies.” Whoa!
There is an assumption in this whole concept of relating B10 numbers to average random failure rates. The assumption is that the all random failures are due to premature wear-out mechanisms. That assumption can be valid when high cycle rates occur in an application but NEVER when low cycle rates are present. There are limits in the valid range of cycles per hour, but I cannot find them stated in the standard. However one thing is certain: these numbers cannot be used for low demand safety functions where cycles per decade might be a good metric.
NEVER use failure rate data based on a B10 calculation to determine PFDavg.