< Back to previous page

Project

Compressive failure of unidirectional composites: efficient computational micromechanics and experimental validation

Compressive failure in the fibre direction is a crucial failure mode to consider in the design of fibre-reinforced composite structures. Unfortunately, the lack of reliable models for this failure mode often leads to overdesigned and inefficient components. While the literature has a global idea of the governing parameters, we currently do not know which combination of microstructural features triggers compressive failure. I will therefore develop computational micromechanics tools to predict longitudinal compressive failure of unidirectional composites under quasi-static and fatigue loading. Deep learning will be used to efficiently identify the weakest locations in a virtual coupon (“hot spots”). A convolutional neural network will be developed and trained for this purpose based on coarse finite element (FE) models. The FE models will use microstructures generated by a newly developed 3D microstructure generator, which is trained based on deep learning of computed tomography (CT) data. Once the hot spots have been identified, only a few refined FE models will be required to predict where compressive failure will initiate. I will apply this procedure to quasi-static and fatigue compressive loading, based on input parameters that are objectively measured on the actual constituents. Detailed experimental validation will be performed based on compressive strength, compression-compression fatigue life and the actual micromechanisms observed in CT data.

Date:1 Oct 2021 →  Today
Keywords:compressive failure, unidirectional composites, Computational micromechanics
Disciplines:Polymer composites, Computational materials science, Destructive and non-destructive testing of materials, High performance computing