Understanding and modelling failure of laminated composites

In this thesis, experimental investigations together with analytical and numerical work on the understanding and modelling of failure in laminated composites are presented. Failure of carbon fibre reinforced plastics is investigated using acoustic emission. Signals are collected for various test configurations which give rise to specific failure modes. The signals are then analysed using pattern recognition techniques and the fast Fourier transform. An identi cation of the failure modes with their acoustic signatures is proposed using the fast Fourier transform, which was found to be the most suitable technique. The failure modes in longitudinal compression are then studied using microscopy techniques and finite element modelling. Experimental observations show that failure results from an interaction between shear-driven compressive failure and kinkband formation. Micromechanical finite element analyses are used to explain the experimental observations. The interaction of shear-driven compressive failure and kinking captured by the model is used to explain the variation in characteristics typically measured in failure envelopes for combined longitudinal compression vs. in-plane shear. Based on the experimental and the numerical results, a failure criterion for fibre kinking and splitting is developed and used to predict failure envelopes for combined longitudinal compression vs. in-plane shear. The model correlates well with the numerical predictions and experimental results. The R-curve effect observed in mode I intralaminar matrix crack growth and its specimen-dependence are then investigated. Relationships between crack extension and crack opening displacement are obtained for the Double Cantilever Beam (DCB) and Compact Tension (CT) specimens. Measured R-curves are used with the previous relationships to define a trilinear cohesive law. The cohesive law is implemented in finite element models and the load versus displacement curves predicted for the DCB and CT specimens show that the R-curve effect is numerically well captured