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: Gupta
Date posted: Wed May 28 1:05:53 US/Eastern 2003
Subject: Usage of Reliability
While Reliability predication, how do i know that the component is following
1. Exponential Reliability or
2. Weibull reliability or
3. Log Normal Reliability.
Is it avl. in PRISM package.
If Suppose I am taking the field failure data, than how come I decide it.
Subject: Prediction Vs. Distribution
Reply Posted by: BWD
Date Posted: Tue Jun 3 10:15:49 US/Eastern 2003
The best way to determine the failure distribution is to test the items to failure in the expected environment. One can then analyze the failure distribution using Weibull characteristics to find the wearout points. The PRISM program is a prediction model based on past historical chance failure information. The PRISM results are exponentially distributed in that each failure is random in nature. Wearout characteristics are not included in PRISM. If you have test data that details the item life times to failure, then you can categorize the information using a Weibull plot to determine the distribution.
Subject: Best way to analyze field data
Reply Posted by: Larry George
Organization: Problem Solving Tools
Date Posted: Fri Jun 6 12:50:37 US/Eastern 2003
Sorry to disagree, but the best way to determine field reliability is to analyze field data, not test data. Customer and environment factors determine reliability, in addition to design and process. The variation is customer and environment factors is difficult to capture in test conditions. Why pay for tests if you already have field data?
Furthermore, let the field data tell you what the field reliability function is rather than make some assumption, which may hide actionable information and requires that you justify the assumption.
Field data for field reliability analysis can take at least two forms:
1. ages at failures and survivors' ages. If you're lucky enough to have this data, there are standard statistics for its analysis. Start with nonparametric statistics; perhaps you'll see that some commonly used parametric reliability function will fit. Usually real field reliability is far more interesting.
2. ships (production, installed base, installations and so on) and returns (complaints, repairs, warranty actions, failures, spares sales and so on). Such data are required by generally accepted accounting principles, so you should find it; you may have to dig. If you have this data, enter it in http://www.fieldreliability.com/Table.htm or send it to me for nonparametric reliability estimates and analysis.