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
Posted by: Bob
Date posted: Fri Jul 14 6:56:42 US/Eastern 2000
Subject: Component reliability calculation
If a supplier claims that the number of failures observed for his product is 0.009% testing for 1000 hours at 125 deg C and then claims this results in an MTBF of over 11 million hours, how do I determine (i) the confidence limits on this figure, (ii) the calculated MTBF for (my) specified confidence limits and (iii) approximate reliability at specified temperatures?
Or is there insuffucient data here to proceed?
Could someone point me to a useful source of information on how to perform the above type of calculations.
Subject: Component Reliability Calculation
Reply Posted by: Shirish Puranik
Date Posted: Wed Jul 19 16:41:26 US/Eastern 2000
I think supplier has given insufficient information. Normally MTBF values are given with confidence limits. More units yield higher confidence. The length of the test and the number of units have same effect. (More units or more hours yield higher confidence limits). Supplier may have performed Accelerated Life test also. Please inquire the testing methodology and background.
Do the failures fit certain distribution? (Weibull 2 par. or exp, 1 par. etc) Supplier should also be able to give reliability numbers based on the failure rates if mission time is known.
Subject: Component Reliability
Reply Posted by: Bruce Dudley
Organization: Reliability Analysis Center
Date Posted: Thu Jul 27 11:23:41 US/Eastern 2000
Confidence limits can only be calculated if one knows the number of test hours accumulated, test conditions experienced and the number of falures identified. Your information does not state all of these facts so the 0.009% per thousand hours failure rate is not quantifiable by confidence levels. To achieve a confidence level of 90% for this failure rate , one would need more than 25 million hours and zero failures. This confidence level can be calculated using the "CHI" square relationship and statistical tables given that the underlying distribution is exponential. Most statistical books have instructions and tables to calculate exponential confidence levels. With regard to the temperature issue, accelerating values can be developed for given components based on the specific failure mechanisms and failure modes using historical information. Usually, one needs to perform a series of high stress testing to failure to determine accuracy of the acceleration energies. Dr. Nelson's book, "Accelerated Testing, Statistical Models, Testing and Data Analysis", Wiley-Interscience publication, has many acceleration models and methods for calculating results.