長時間V-t 特性における最適数学モデル選定と寿命推定の統計的方法

A Statistical Method of Optimum Mathematical Model Selection and Lifetime Estimation in Long-term V - t Characteristics



Hideo Hirose


電気学会論文誌A, Vol.108-A, No.10, pp.451-458 (昭和62.10)

Trans. IEE of Japan, Vol.108-A, No.10, pp.451-458 (1987.10)


This paper describes a new estimating method of elec^"_t~!.cal insulation under given voltage stress. The new technique proposed by the author can not only predict the lifetime under the service voltage stress but also select the most adequate V-t characteristics model objectively. This paper has the four outstanding features as follows. (1) Two V-t characteristic models have been proposed: one is the so-called inverse power law model and the other is the exponential model. The former has a relation with the Weibull distribution, but the latter not. Thus, the author proposed a new exponential model with the Weibull probability distribution. (2) Both two models include the threshold voltage stress parameter in the probability distribution. Thus, we can estimate the threshold value with confidence interval. (3) In order to cut down on time of V-t test, an accelerated stress procedure with type I censoring is considered. (4) We can determine which modcl represents better the V-t test data in the meaning of the most adequate V-t characteristics. In model selection, we use the AIC(Akaike Information Criterion). The idea of AIC will give us deeper knowledge for the voltage deterioration phenomena. Therefore, we can select the probable model, and we can estimate the unknown parameters of the selected model such as the threshold value and the lifetime under the service voltage with confidence interval.

The paper written in English is available upon request.

Key Words



Times Cited in Web of Science:

Cited in Books: