SRC Forum - Message Replies

Forum: Reliability & Maintainability Questions and Answers

Topic: Reliability & Maintainability Questions and Answers

Topic Posted by: Reliability & Maintainability Forum ( )
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: Andrew Comons
Date posted: Fri Mar 29 13:59:26 US/Eastern 2002
Subject: Exponential Conditional Reliability...A Paradox?
I recently came across a derivation of the "Exponential Conditional Reliability" equation which states that regardless of previous accumulated age, the reliability of a system is only dependent on the duration of the present mission, i.e., no "memory" of what occurred prior to the present mission. I fully understand the arithmetic of the derivation, but not the principle. According to the theory, I could momentarily shut a system down in the middle of a mission, turn it back on and expect the same probability of success as at the start of the initial mission. This does not seem practical or realistic, but the math says oterwise. Can anyone explain this paradox?? Thanks in advance


Subject: Exponential Paradox
Reply Posted by: Patrick Hetherington ( )
Organization: Reliability Analysis Center
Date Posted: Tue Apr 2 8:40:07 US/Eastern 2002
The exponential failure distribution is predicated on a random pattern where by the intensity of failures (number or probability) is directly related to the amount of exposure rather than accumulated damage. Items that fail exponentially are generally the result of random overstress. Think of the probability of being hit by lightning (a random overstress situation). The likelihood of getting hit over a ten year period (long period of exposure) is much greater than being hit today (short period). Given that a person was not hit by lightning in the last ten years, however, does not change the probability that they will or will not be hit today.


Subject: Overstress is not a reliability failure
Reply Posted by: tom c
Date Posted: Wed Apr 3 13:33:13 US/Eastern 2002
I have recently become a believer that the exponential distribution is very likely the incorect model for most real complex electronic/mechanical systems. The idea of a constant failure probably isn't valid very often. If the type of overstess that Patrick refers to, one that exceeds the design limit of the device, causes failure - well this is not a reliability failure, but one that the equipment was never intended to withstand. An overstress condition can not be considered a relevant failure and would not be covered by any warranty.


Subject: Exponential Failure Rate
Reply Posted by: Donald L. Meaker ( )Northrop Grumman, Integrated Systems
Date Posted: Fri Jun 14 9:39:00 US/Eastern 2002
There is a difference in using the Exponential Distribution as the really true distribution and as an approximation. If you are performing Weibull Analysis, and have filtered your data so that you are analyzing all the same failure mode, and then your analysis indicates that the shape parameter is 1, (ie, an exponential distribution) I would check the stress level to see if there was a voltage spike or a bump on my course that was the cause of failure. On the other hand, If I am using the exponential distribution as an approximation of the failure rate for a complex component, modeling several different failure modes, (Dreineck's Therom) then no over stress is necessarily present.

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