A new statistical methodology for assessing mechanical survivability of brittle materials

Abstract

Brittle materials such as ceramics are subject to fracture without warning. Because non-destructive techniques are unreliable for determining potential fracture sources in ceramic materials one must rely on statistical analysis of laboratory strength data. This data is used to determine the minimum strength, and its uncertainty, of a set of specimens, and then must translate this laboratory data into a projection of the reliability of components manufactured from the material. This paper sets forth new guidelines for the choice of a statistical methodology to fit the laboratory data and puts forth a procedure – known as tolerance limits and coverage – to extrapolate this data to predict component reliability. Data on a borosilicate glass is used to demonstrate the usefulness of this procedure.