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 Show all replies Topics List About this forum
Original Message:

Posted by: John K. (kelsmit@aol.com )
Date posted: Thu Jan 16 22:00:40 US/Eastern 2003
Subject: Reliability Growth
Message:
Ok- I have a test project with five units to test. The units will accumulate hours in test at different rates (I have to share with design engineering - I break it, they fix it). If my first cumulative MTBF calculation is 200 hours (1000 hrs total/ 5 failures), What amount of time do I need before I should run another calculation? Is there some statistics or rational to support a minimum amount of time? Is there a better way to report or calculate reliability? All I have is MS Excel. Thank you - John


Reply:

Subject: Test Time
Reply Posted by: Jorge Romeu (jromeu@alionscience.com )
Organization: RAC
Date Posted: Mon Jan 27 14:19:08 US/Eastern 2003
Message:
1) It is always better to test each unit until it fails, and then record the failure time. 2) If the experiment is terminated on time, then it is better to test at least until after one failure occurs. Also, the treatment of the data analysis will vary according to whether the unit lives are distributed Exponential or as some other distribution. 3) The two above points constitute a basis to determine the minimum time required for testing, before another calculation is obtained. In addition, 4) MTBF is only a point estimator and by itself is a poor indicator. It is better to use Confidence Interval. 5) Finally, and for all the above-mentioned reasons, it is very important to establish a statistical distribution for the lives of the device and to estimate its parameters. Start by obtaining this, if you can.


Reply to this message