(70aa) Product Failure Rates Vs Site-Specific Failure Rates: Implications for Safety

To quantify the safety achieved through the use of automatic safety equipment, one must be able to determine the failure rates of the equipment which are then used to compute average probability of failure on demand (PFDavg).  This is usually done by predicting failure rates via a failure modes, effects and diagnostics analysis (FMEDA) or by estimating the failure rates from failure data obtained from one or more specific end-user sites.  When the predicted failure rates are compared to the estimated failure rates, there are occasions where they differ considerably.  Studies of field failure data have shown the total failure rate for a particular product model number has varied by up to 4 times at different sites.  An understanding of why this number varies from site to site has led to the conclusion that some random failures are caused by operational practices at a particular site.  This means that total failure rates computed from site-specific field failure data consist of two parts:  the product failure rate and the site-generated failure rate.  Product failure rates which are predicted from failure rate and mode databases assume that the site end-user follows a maintenance process per IEC 61511 and that actions to properly store, install, maintain, test and repair are performed accurately.  Site(s)-generated total failure rates account for the effects of random end-user actions that cause failures.    

In this paper we

• Compare product failure rates and  total failure rates for a number of different types of safety equipment and identify the difference as a site-generated failure rate

• Show that estimated total failure rates may be larger than the predicted rates by a factor of 2 or more

• Discuss the safety implications of site-generated failure rates and the difficulties of accurately estimating these rates

• Propose a method for extending prediction techniques to predict site-generated failure rates as well as product specific failure rates

• Apply the new method to an example safety product