Critical issues in the design-by-testing of annealed glass components

•Glass strength depends on so many aspects that make full-scale tests hardly interpretable.•A model is proposed to calculate the strength of a prototype for various failure probabilities.•Tests at constant load or at constant load rate are discussed in three paradigmatic examples.•Failure stresses at different probabilities remarkably vary with the structure type.•Glass strength can be correlated with the target probability for any specimen size and load history.

To assess the reliability of glass components, a common practice is to test full-scale prototypes in the lab, and verify that the failure load is higher than that predicted from the design strength by means of structural calculations. However, any procedure of design-by-testing should be considered with great care because the gross strength of glass, being governed by the opening of pre-existing cracks on the material surface, strongly depends upon the type of defectiveness, the specimen size, the load history and the type of stress field (uniaxial, bi-axial). A model based upon an assumed law of subcritical crack propagation and a distribution à la Weibull of pre-existing flaws is considered for the body strength of annealed glass. This allows to correlate the expected macroscopic strength of glass, measured from testing the prototype, with the target probability of failure, for any type of size and load history. The discussion of paradigmatic examples confirms that appropriate theoretical considerations are needed for the correct interpretation of the experimental results.