Non-crimp 3D woven composites unit cell: from geometric modelling to damage simulation

Abstract

In the last twenty years, the research on composite materials has increased and many progresses have been made. However, there are still unresolved issues concerning the geometric modelling of a material at the meso-level (i.e. on a unit cell) and its damage simulation. In particular, the complexity of the internal geometry of some composite materials, such as 3D textiles, yields to new challenges for the research community. A correct definition of the internal structure in all the important parameters (fibre volume fraction, yarn section dimensions, unit cell thickness, crimp ratio) is the crucial starting point for the evaluation of the homogenized mechanical properties and damage evolution of the unit cell. The present PhD Thesis is composed by two parts. The aim of the first one is to generate a unit cell (UC) of a non-crimp 3d woven composite with a realistic over-all and intra yarn volume fraction and to develop a FE-model for the evaluation of the elastic constants. The aim of the second part is to implement a new damage model, suitable for the 3D textile composites. In the first part, a WiseTex model is generated and, in order to better represent the real cross section, the elliptical original shapes are modified according to optical microscopy measurements on the textile cross-section. A semiautomatic solution for the yarn interpenetration is developed and a FE model is created. Periodic boundary conditions are applied to the UC and the homogenized elastic constants of the unit cell are then evaluated. A geometry with a realistic over-all (50%) and intra-yarn (75%) volume fraction is obtained and the comparison between the FE-model results and experimental data shows good agreement. In the second part, a damage model appropriate for 3D textile composites is chosen from literature. The damage model is implemented into the commercial code ABAQUS using FORTRAN subroutines (UMAT). A validation of the subroutine is conducted on a single element loaded with simple tension and shear stresses, the results are compared with the numerical solution obtained with a Matlab code. Some tests are then made on a more complex geometry (UD plate with a hole) applying various loading conditions. The damage model implemented presents a very good agreement with the Matlab numerical solution in the case of the single element while it shows some convergence issues in the case of the more complex model, for which further investigations are under way.