Simulating progressive failure in laminated glass beams with a layer-wise randomized phase-field solver

TL;DR

This study uses a randomized layer-wise phase-field model to simulate progressive failure in laminated glass beams, capturing multiple crack formations but underestimating ductility and failure sequences compared to experiments.

Contribution

It introduces a computationally efficient, dimensionally-reduced phase-field approach with randomized layer strengths to predict progressive failure in laminated glass.

Findings

Model reproduces progressive failure and multiple cracks.

Response is less ductile than experimental observations.

Cannot fully capture the most common failure sequences.

Abstract

Laminated glass achieves improved post-critical response through the composite effect of stiff glass layers and more compliant polymer films, manifested in progressive layer failure by multiple localized cracks. As a result, laminated glass exhibits greater ductility than non-laminated glass, making structures made with it suitable for safety-critical applications while maintaining their aesthetic qualities. However, such post-critical response is challenging to reproduce using deterministic failure models, which mostly predict failure through a single through-thickness crack localized simultaneously in all layers. This numerical-experimental study explores the extent to which progressive failure can be predicted by a simple randomized model, where layer-wise tensile strength is modeled by independent, identically distributed Weibull variables. On the numerical side, we employ a computationally efficient, dimensionally-reduced phase field formulation -- with each layer considered to be a Timoshenko beam -- to study progressive failure through combinatorial analysis and detailed Monte Carlo simulations. The reference experimental data were obtained from displacement-controlled four-point bending tests performed on multi-layer laminated glass beams. For certain combinations of the glass layer strengths, results show that the randomized model can reproduce progressive structural failure and the formation of multiple localized cracks in the glass layers. However, the predicted response was less ductile than that observed in experiments, and the model could not reproduce the most frequent glass layer failure sequence. These findings highlight the need to consider strength variability along the length of a beam and to include it in phase-field formulations.