A comparative accuracy of several discrete methods for lower confidence limit on system reliability.

Abstract

This thesis is a comparative accuracy study of several discrete methods for lower confidence limits on series system reliability. Computer simulations were used to compare the accuracy of the procedures. Five hundred replications were used in all simulations. Accuracy of each procedure was determined by computing appropriate percentile points of the distributions of the lower confidence limits. A randomization technique was used to improve the performance of one of the procedures. The systems simulated had reliabilities ranging from 0.720 to 0.950. They were composed of five, ten, thirteen, and fifteen components, and had component sample sizes of fifteen, thirty, fifty, and larger in the case of unequal sample sizes. Based on the simulation results the accuracy of the procedures were compared by common comparison with the true system reliabilities which were known in advance prior to the component tests.