Downloaded from https://www.everyspec.com A AMCP 700r198 Is more complicated in this case than in par. sonic measure of its uncertainty, 10-2.1. and probably isn't worth the extra trouble, since the uncertainties are going to One big difficulty with choosing a model be large in either case. The computer pro- is to be aware of all the answers that are gram in Ref.
4 would have to be modified to part of the model, rather than part of the account for the new variables, data, As an absurd extreme, if the model Is chosen so that there is a lower limit on fail- 10.3 OTHER MODELS ure rate, one ought not to be surprised at the statistical predictions from the model and A large number of mathematical models data that there is a lower limit on failure are possible for reliability growth. Many of rate.
them can fit any particular set of historical data reasonably well, The real question, how- ever, is how well they allow you to estimate A reliability statistician can help you current reliability and to predict its future create a growth raodel that you are willing value. to live with. The cost of analyzing the model will generally be negligible compared to the Refs. 5 and 6 each compare several models cobt of the physical tests. In the absence of with simulated data, Ref. 5 is limited to such help, one can plot - log)h against "some binomial data (success-failure). Ref. 6 in. measure of program resources consumed so cludes the Duane model in its comparisons.
far" where R is the estimate of reliability Ref. 7 analyzes the Duane model. Unfortu- from the most recent test. If there are a nately, very few references show how to great many data, the form of the curve will estimate the uncertainty in reliability-growth be obvious. If there are very few data, the predictions. No statigtical point estimate is scatter on the graph will emphasize your worth very much unless it tm accompanied by irreducible uncertainty, REFERENCES 1. 3. T. Duane, Technical Information Series 5. R. C. Dahiya, A. J. Gross, "Adaptive Ex- Report 62MD390, General Electric Co., ponential Smoothing Models for Reliability DCM&G Dept., Erie, PA; 1962. Estimation", IEEE Trans.
Reliability, Vol. R-23, December 1974, pp. 332-334, 2. J. T. Duane, "Learning Curve Approach to Reliability Monitoring", IEEE Trans. Aero- 6. T. Jayachandran, L. R. Moore "A Coin- space, Vol. 2, 1964. parison of Reliability Growth Models", IEEE Trans. Reliability, Vol. R-25, April 3, E. Parzen, Stochastic Processes, Holden- 1976. Day, Inc., San Francisco, 1965. 7. L. H. Crow, "On Tracking Reliability 4. R. A. Evans, "The Analysis of Acceler- Growth", Proc 1975 Annual Reliabhairy ated Temperature-tests", Proc. 1969 and Maintainability Symposium, pp, 438- Annual Symposium on Reliability, 443 Available from the IEEE. pp. 294-302. Available from the IEEE. 4 (I! !