Planning accelerated life tests with type II censored data

Abstract

This paper presents methods for planning accelerated life tests for models in which the logarithm (or some other appropriate transformation) of time-to-failure follows a location-scale distribution and the location parameter is a function of stress. We give exact expressions for the BLUE of the lOOPth percentile of the life distribution at design conditions and for the variance of this estimator. Then we describe methods for evaluating the important properties of test plans and for finding approximate continuous and exact discrete optimum test plans. A numerical example illustrates the methods.

Keywords:

• Weibull distribution
• lognormal distribution
• optimum design
• reliability
• accelerated test
• experimental design
• BLUE