The present work focuses on improving the understanding and modelling of the mechanical response
of woven composites. It comprises of original experimental observations together with novel numerical
and analytical developments.
The experimental work focused on the investigation of damage initiation and propagation under
compression. Detailed microscopy shows that damage is controlled by the individual failure of the load-aligned
tows through kink-band formation. Moreover, both weave architecture/internal geometry and
support provided by the adjacent layers are seen to affect damage location and morphology, suggesting
that, to capture the failure mechanisms in compression, they should be explicitly modelled.
Subsequently a theoretical framework to derive periodic boundary conditions for the mechanical
analysis of periodic structures, using domains smaller then the unit cells, is presented. These domains,
named reduced Unit Cells, lead to significant gains in efficiency, which enable the use of refined
numerical/analytical models.
The framework mentioned above was used to develop a detailed nonlinear numerical model of a
2D woven composite. Weave architecture was modelled explicitly and the effect of the support provided
by the adjacent layers was taken into account. The constitutive response and ultimate strength
values predicted numerically agree well with experimental results for both tension and compression.
Additionally, the model was used to investigate the effect of the biaxial loading ratio on the failure
strength.
Finally, an analytical model to predict the compressive and tensile response of woven composites
is presented. The load-aligned tow was modelled as an Euler-Bernoulli beam supported by an elastic
foundation. The properties of the latter are fully derived from kinematic models for the deformation
of the weave, providing great insight into the deformation mechanisms of woven composites. The
analytical predictions agree well with both experimental and numerical results. Overall, its flexibility
offers an alternative and valuable addition to detailed numerical models, particularly when performing
sensitivities studies.