Probabilistic failure prediction for automotive windshields based on strength and flaw distributions

This paper describes a method for predicting structural failure probabilities for automotive windshields. The predictive model is supported by the data from strength tests performed on specimens of automotive glass. Evaluations of stresses can be based on finite element calculations, or measurements of the residual stresses that arise from fabrication. Failure probabilities for each subregion of a windshield are estimated from the local state of stress, the surface area or length (for edge elements) of the subregion, and statistical distributions of glass strengths. Example calculations are presented that show the relative contributions of edge stresses, surface stresses and residual stresses to calculated failure probabilities.