Statistical Analysis of Thermal Shock Data

Abstract

Hazard plotting and maximum likelihood estimation (MLE) are used to determine Weibull parameters for single cycle thermal shock tests performed on WC-Co specimens. Data for each sample fall into three classes: tests below a threshold prequench temperature for which failure never occurs; tests above some higher prequench temperature for which failure is universal; and tests at intermediate prequench temperatures for which both failures and nonfailures are interspersed. Hazard plots in the region of the third class show a good Weibull fit. For some samples, the shape of the hazard plots indicates competing risks. MLE of the Weibull parameters for each sample is performed and the estimates are compared. Failure models which might account for the shape of the hazard plots are proposed.