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:Marc Hughes
Date posted: Wed Jul 14 14:52:22 US/Eastern 1999
Subject: Confidence Limits on Reliability Predictions
I have been asked about confidence limits on failure rates calculated using MIL-HDBK-217 & Bellcore, using the NPRD and vendor data. My experience in determining confidence limits is based on statistical type testing as described by MIL-STD-781 and not based on the reliability prediction.
The requestor is continuing to state that you can determine confidence intervals on reliability calculations without test data. Do you know of a method, or a publication that would help?
I thought at one time there was a paper written concerning confidence limits on the formulas presented in MIL-HDBK-217 since these formulae are based on test data of components?
Subject: Prediction confidence
Reply Posted by: (email@example.com
Organization: Reliability Analysis Center
Date Posted: Fri Jul 16 9:03:21 US/Eastern 1999
MIL-HDBK-217, as well as other reliability prediction models, are based on data from various sources and complete models are not always developed using a single source of data. For example, all MIL-HDBK-217 environmental factors, which are a direct multiplier to part failure rates, were developed under separate study efforts. Because of the fragmented nature of the data and the fact that it is often necessary to interpolate or extrapolate from available data when developing new models, no statistical confidence intervals should be associated with the overall model results for any given prediction. In addition to the variability associated with developing the models, there is human variability in making prediction assumptions, counting of field failures, and failure definitions.
In the late 1980's a study was conducted that compared predicted-to-field MTBFs in an attempt to quantify the uncertainty associated with reliability predictions. This study was a "snap shot" that analyzed data on systems for which both predicted and observed MTBF data was available. The data in the two columns below show the multipliers of a failure rate point estimate as a function of confidence level that was derived from analysis of this data. For example, using this data set one could be 90% certain that the field failure rate was less than 7.575 times the predicted value, for this sample.
Uncertainty Level Multiplier
The data used to develop the above multipliers was based on approximately 200 reliability predictions performed during the 1970's and 1980's and documented in a study sponsored by Rome Air Development Center entitled "Reliability/Maintainability Operational Parameter Translation II," RADC-TR-89-299. It should be remembered that the predictions on these 200 systems were developed by a wide range of individuals making many assumptions. Field MTBFs used in the study introduce more variability with a wide range of operating hours, failure counts and maintenance policies for each system. Therefore, the study results could well be different if reconstructed today using a new data set. It serves only to provide a notion of the variability possible across a wide range of systems, companies, individuals and field maintenance policies. The results could be much better, say, if a single experienced reliability engineer was applying a standard prediction tool over a long period of time and there was like consistency in field failure counting practices.