A Comparison of the Accuracy of Univariate and Bivariate Techniques for Finding the Lower Confidence Limits of System Reliability.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH SCHOOL OF ENGINEERING
Pagination or Media Count:
The purpose of this thesis is to compare the accuracy of two Monte Carlo simulation techniques of finding lower system reliability confidence limits the bivariate technique and the univariate technique. The actual results compared are the confidence interval coverages of the true system reliability associated with the confidence limits. The bivariate technique is based upon the assumption that the maximum likelihood estimators of the component shape and scale parameters have an asymptotic normal distribution. The univariate technique uses the assumption that the component reliability estimates have a normal distribution. Two variations of the univariate technique are also examined. The first variations assumes that component reliability estimates follow a beta distribution instead of a normal distribution. The second variation replaces all perfect system reliability estimates with new, adjusted reliability values.
- Statistics and Probability