Multiscale modelling of material degradation and failure in plain woven composites: A novel approach for reliable predictions enabled by meta-models

Abstract

In this paper a multiscale method for modelling damage evolution at meso and macro scales with application to plain woven composites is presented for the first time. The method couples the macro and micro-scale simulations using an averaging theorem in which the computations in the two scales are performed simultaneously. Unlike the local–global approach, the developed multiscale analysis is carried out over the entire macro domain rather than a predefined segment. To address the computational challenges in multiscale damage analysis, the proposed method a) employs a non-local approach in constructing the macro–micro interface to avoid the pathological localisation effect introduced by microdamage at the macro continuum; b) reduces the considerable computing effort incurred by the non-linear behaviours in multiscale damage analysis by the introduction of machine learning meta-models. Experimental tensile tests conducted are shown to be in good agreement with the proposed method for both on-axis and off-axis load directions. The biaxial loading cases are also investigated with the results illustrated by the failure envelopes and compared with the existing damage criteria. Finally, an example of open-hole tension test is presented to demonstrate an application of the proposed method to model progressive damage and crack propagation in composite plates.