Development of statistically equivalent representative volume elements for multi-scale modeling of composite materials

The non-uniform spatial arrangement of fibers in composite materials leads to an irregular stress distribution in the microstructure allowing localized microscopic damage mechanisms to occur more easily. In order to accurately predict such damage mechanisms, statistically equivalent representative volume elements (SERVE) are often used. A hard-core random model, where fibers are randomly placed, is generally used to create these SERVE's but such models do not reproduce the microstructure well for high volume fraction composites, such as those used for high strength applications in the aerospace industry. In this paper, a novel method is developed to generate SERVE's for high volume fraction composites. This method uses experimentally measured nearest neighbor distribution functions to define inter-fiber distances. The resulting SERVE is found to show very similar geometric distribution functions (i.e. radial distribution and nearest neighbor distribution) to the actual microstructure. The proposed algorithm is currently being used to generate micromechanical finite element models for multi-scale damage prediction of composite structures.