The Weibull Distribution: Some Dangers With Its Use in Insulation Studies

Abstract

It is evidenced that analysis of dielectric behavior with uncritical use of the Weibull Distribution tends to draw attention away from physical phenomena in general; that it tends specifically to obscure aging processes; and that it obscures the equivalence that can exist between ramp, step, and constant voltage test results. To offset these dangers, it is recommended that failure probabilities be expressed in the form P(V,T,t) with aging time T and test time t; that the smallest statistically-independent time of voltage application be one focal point of investigation; and that test equivalences (eg., ramp to constant voltage) be applied. Analytical procedures are given to determine for ramp and step tests, the equivalent time for a test at constant voltage. Examples are shown, including a method for accounting for physical aging processes in such comparisons.