Annealed Glass Failure Modelling : A Weakest-Link Approach with Statistical Analysis

Abstract: Novel and increased use of glass in building construction has been made possible, as a result of methods of mass production introduced during the 20th century and advanced computational structural analyses. With modern use of glass structures, a range of demands emerge which need to be addressed in the strength design process. The aim of the thesis work has been to develop experimentally verified prediction models that can be used as design tools for advanced glass structures. The first part of the thesis consists of an introduction and overview with a background and motivation for the work carried out, the aim and objectives, along with a summary of the appended papers, and a conclusion and outlook with suggestions for further work. The second part of the thesis consists of five appended publications, Paper A to E. Paper A presents results from a wide-ranging survey of laboratory tests which was conducted pertaining to the strength of new annealed soda-lime float glass tested in an ambient environment. With a basis in the survey, four standard statistical distributions were compared with respect to their performance as strength models, namely the normal, lognormal, Weibull, and Gumbel. It was concluded that the extreme value distributions provide basic models for edge failures but perform poorly for surface fracture origins. In Paper B, C, and E, various numerical implementations of finite-size and flaw-size based weakest-link systems were developed and applied to model the failure stress and fracture origins on the surface of plates subject to lateral loading, in addition to a consideration of strength-scaling size effects. Applications were made to a range of different load cases including small plates subject to ring-on-ring and ball-on-ring loading, large linearly supported plates subject to uniform pressure, and large panels with complex geometry subject to impact loading. In Paper E, results were presented from laboratory tests which were carried out on two series of annealed glass plates subject to ring-on-ring and ball-on-ring loading in an investigation of the distribution of failure stress and fracture origin, and their dependence on the surface area exposed to greatest tension. The Weibull effective areas were expressed in closed-form and employed to calculate a strength-scaling size effect. According to the observed surface strength data, the weakest-link premise of the ordinary Weibull model is rendered intractable and more sophisticated approaches are warranted. Finally, it was concluded that there is a need for additional research on the surface condition of glass that can lead to more reliable information about the suitable choice of model parameters. In Paper D it was considered that the Weibull distribution parameters that are fitted to laboratory measurements of as-cut, arrised, ground, and polished edge strength exhibit considerable variability. Estimates for the characteristic 5%-fractile edge strength were obtained in a hierarchical modelling approach by considering the Weibull parameters as nested random variables. It was shown that glass supplier random effects are important to consider in addition to effects on the observed strength due to environmentally assisted crack growth, applied stress rate, and edge length exposed to maximum stress.