RELIABILITY CONSIDERATIONS IN ACCELERATED LIFE TESTING OF ELECTRICAL INSULATION WITH GENERALIZED LIFE DISTRIBUTION FUNCTION

This paper introduces a new approach to estimate life distributions at nominal conditions from the results of accelerated life testing of electrical insulating materials. A very general family of probability distributions is introduced and a best fit member of this family is used to represent life data at each stress level. Nonlinear optimization techniques are applied in conjunction with linear regression analysis. In any accelerated life testing study important questions pertain to the minimal and maximal stress levels to be applied. A method of determination of the minimal stress level as well as the suitable number of tests based on reliability considerations is presented in the paper. A numerical example based on actual test data and a user-friendly computer program are presented.