Optimal linear regression estimator in the fitting of Weibull strength distribution

Abstract

The strength of ceramics is characterized by a wide scatter because of pre-existing cracks that occur during the manufacturing and machining processes. The Weibull distribution is one of the most widely used functions for the characterization of strength data. A linear regression method is generally applied in the estimation of Weibull parameters for its simplicity and low overestimation, in which probability estimators play an important role. In this paper, an optimal probability estimator for different sample sizes is obtained by using alumina strength data. In comparison with other commonly used estimators, the optimal probability estimator shows less bias and higher safety. The performance of the optimal probability estimator is also verified by other experimental strength data. In conclusion, an optimal probability estimator constant of 0.25 is suggested in practical applications.