SRC Forum - Message Replies


Forum: Reliability & Maintainability Questions and Answers

Topic: Reliability & Maintainability Questions and Answers

Topic Posted by: Reliability & Maintainability Forum (src_forum@alionscience.com )
Organization: System Reliability Center
Date Posted: Mon Aug 31 12:47:36 US/Eastern 1998

Back to message list Topics List About this forum
Original Message:

Posted by: Ed
Date posted: Mon Jan 6 13:17:18 US/Eastern 2003
Subject: Wearout or random failure
Message:
The failure criteria of a device is 10% change in one parameter. Is it a wearout or random failure? Thanks.


Reply:

Subject: 10% Parameter Change
Reply Posted by: Joe Dzekevich (jdzekevich@ieee.org )
Organization: Raytheon RAL
Date Posted: Mon Jan 6 13:25:21 US/Eastern 2003
Message:
My take on it is: it depends. If it degrades over time, it's probably wearout. If it happens quickly, then it may be do to a failure. Rule of thumb - do not guess. Do a failure analysis. Joe


Reply:

Subject: Again: Wearout or random failure
Reply Posted by: Ed
Date Posted: Tue Jan 7 15:26:08 US/Eastern 2003
Message:
The failure is degradation over time. So if it is the case, how to predict MTBF of a system which includes some this kind of device. 217 or Telcordia's prediction is based on constant failure rates, however the failure mechanism of a lot of electronic devices is parametric degradation. How to perform the MTBF? Thanks.


Reply:

Subject: Wear-Out MTTF
Reply Posted by: Joe Dzekevich (jdzekevich@ieee.org )
Organization: Raytheon RAL
Date Posted: Tue Jan 7 16:41:56 US/Eastern 2003
Message:
Ed, since it is a wear-out device, then you should treat it like you would with a fan or e-cap. For the unit, quote two numbers: a MTBF (includes all non wear-out components and the random f/r of wear-out components) and a MTTF for the wear-out. Now, if you want to combine them into one number, you then must quote it over some time period, like x days, y warranty, z mission time. Many preduction tools allow you to mix MTBF (exponential) with MTTF (Weibull), but the Weibull part needs a mission time. In effect, you are adding the Weibull f/r at time t to the exponential random failure rate. Joe Joe


Back to message list Topics List About this forum
Reply to this message