| description abstract | Over the past four decades, surface flaw parameters for the two-parameter glass failure prediction model were estimated based on statistical measures and the researcher’s interpretation of the failure data rather than an objective and repeatable method. A robust statistical method, the maximum likelihood estimate, is advanced in this work, which is tailored to the two-parameter glass failure prediction model for rectangular glass plates to objectively and repeatably estimate the surface flaw parameters. The surface flaw parameters represent the number, type, depth, and orientation of microscopic flaws present on the surface of the glass, and hence, it is impractical to physically measure them. Instead, they are inferred from a destructive testing of identical dimensions of glass plates monotonically loaded to failure. Historical failure test data consisting of weathered and new annealed monolithic glass specimens are collected and used to illustrate the working principle of the maximum likelihood estimator method. However, the published experimental data of these tests are not consistently complete. Therefore, several critical requirements are noted along with strategies for using incomplete data sets. Current design standards are based on a 3-s basic wind load, and hence, surface flaw parameters based on 3-s time duration are reported. Several example designs are used to illustrate a comparison of the historical sample glass failure prediction model (GFPM) surface parameters with the current GFPM parameters used in, ASTM E1300, the standard in the United States used to determine the load resistance of a glass component. | |
