On The Use of The Extreme Value Distribution in Reliability Analysis and Design

Abstract

Use of the Type 1 Extreme Value Distribution of Maxima to design engineering problems is reviewed. Results of a Monte Carlo study of the statistical behavior of the distribution parameter estimators are summarized. The study compares behavior of the method of moment and maximum likelihood estimators, and the results suggest that the method of moment estimators, a simpler algorithm, is but slightly less efficient than the maximum likelihood estimators. An example is given to illustrate how a design engineer uses these results to translate observations into an estimate of “reliability” or “risk”. Moreover, a scheme for obtaining approximate confidence intervals on reliability is presented.