Rate-dependent continuum damage modeling of composite materials

A three-dimensional, phenomenological, tensorial, isotropic, damage model is developed in the framework of continuum damage mechanics for materials whose behavior is governed by elastic deformation coupled with damage. The model was then extended to include isotropic damage for anisotropic materials, as well as rate-dependency behavior caused by damage evolution. The shortcomings of the commonly used scalar variable, as representative of isotropic damage are discussed. It is shown that isotropic damage is best represented by an isotropic tensor of rank four. The damage evolution equations are postulated using strain tensor invariants, based on decomposition of strain energy. The model simulates well the results of static and dynamic uniaxial tension tests on quasiisotropic laminated graphite-epoxy obtained in this study and results, from literature, of uniaxial compression tests on quartzite rock.; To determine the material parameters used in the model and to validate the model, a set of material and structural tests, testing a laminate containing a hole, were performed under static and dynamic loading conditions. A tensile version of the Hopkinson bar, suitable for testing of laminated composite materials, is developed to perform dynamic tests. A pulse duration of 200--250 microseconds and peak strain rates of up to 350 s--1 are obtained. Tests performed on a quasi-isotropic lay-up of graphite-epoxy show good repeatability. Comparison of Hopkinson bar tests results with results of tests performed at a quasi-static rate on a hydraulic test machine shows the rate-dependency of this lay-up of graphite-epoxy. Tensile strength and fracture strain are found to be higher for dynamic testing.; The model was evaluated for structural analysis, by implementing the model into a finite element code and analysing a laminate containing a hole. Two techniques are investigated in evaluating the model for structural analysis: stress limiter and mesh limiter. The model is found to be objective with respect to the mesh size. The predicted failure loads using both techniques conform well to the experiments and to the results obtained using one of the existing models.