I was recently contacted by one of my validation engineers in Asia, who asked me to help him understand the reliability requirements, submitted by one of the Asian automotive OEM for a passenger compartment electronic module. The requirement specified a B1 life as the warranty period for the module (3 yrs) and also specified the durability target of 500,000 km. The engineer was confused on how to reconcile the requirements of reliability and durability.

In my opinion here we see an example of lack of reliability understanding on the customer’s part, although this was not the first time I came across the product validation requirements listing durability. The problem is that there is no clear technical definition of durability. Sometimes it is associated with the failure-free period, sometimes with the specific section on the bathtub curve, sometimes with the ability of the product to function after repairs, but the truth remains, the universal definition of durability is missing. In layman’s terms this specification means that the module should be capable of surviving 500,000 km. However engineers are not allowed to speak in layman’s terms and are expected to quantify their questions and answers. In this particular case it means that they need to be specific about the population of the parts, which is expected to survive 500,000 km.

In statistical terms, durability is not much different from reliability therefore in order to be able to address the durability we should either have Bx = 500,000 km (x being percent of the population failing at that mark) or the reliability/confidence associated with 500,000 km life.

The bad news is that this information was missing. The good news is that it gave the engineer some flexibility in terms of defining durability the way which would better fit his test plan. For example, this requirement did not specify if the product would need to be actually tested to the equivalent of 500,000 km, which would open the door to the less expensive means of demonstrating the required durability, e.g., extrapolating the shorter test results to the equivalent of 500,000 km using Weibull analysis or utilizing one of the reliability prediction methods. If the test did not produce any failures, then Weibull distribution with assumed beta-slope or exponential distribution could be utilized to extrapolate the results to 500,000 km life. In any scenario, engineer should have a reasonable understanding of reliability and durability requirements and how they can be addressed in terms of test results and test analytics.

Posted by Andre Kleyner

Be Sociable, Share!
  • Twitter
  • Facebook
  • StumbleUpon
  • Delicious
  • LinkedIn
  • Google Bookmarks

Leave a Reply