Statistical approach of elastic properties of continuous fiber composite

This paper is an attempt to predict the elastic properties of plain-weave continuous fiber-reinforced composites by coupling data of microstructure analysis and mechanical tests with a stochastic and multi-scale finite element analysis. The difference of this study over the previous one relies in the fact that these microstructural data are used directly in the proposed procedure without any pre-processing. Thus, a random procedure is used to build the mesh of the representative volume element (RVE), the description of which is based on the mosaic model. The geometry of this RVE is described by a random choice of data provided by image analysis. Variables as the fiber modulus and the local fiber volume fraction are randomly assigned to each element of the mesh. Finally finite elements calculations are performed with the homogenization technique to obtain the macroscopic elastic properties. The results show that geometry fluctuations of the RVE have a significant impact on the material macroscopic moduli. Furthermore, the method is more efficient to predict the mean values of the elastic moduli than the experimental dispersions (NB: the aim of this approach is to predict the scatter of elastic properties. An elastic homogenization procedure is sufficient to obtain numerically “mean” values). The results also show that the variations of the local fiber volume fraction have a small influence on the elastic properties because the mean value of the fiber volume fraction over the whole RVE varies very little from a draw to another. Finally, such a proposed method if it is fully automated could be very useful in preliminary design because cheaper than to perform experimental tests.