A versatile and highly efficient algorithm to generate representative microstructures for heterogeneous materials

In this study, a novel approach based on the maximum penetration-biased (MPB) algorithm is proposed to rapidly generate representative volume elements (RVEs) for advanced composites. This method is capable of handling various microstructural features such as nonuniform distribution of high-volume fractions, concave or convex inclusions and the control of inter-inclusion distances. Statistical functions from short-range to long-range demonstrate that the resulting microstructures are complete spatial random, and the microstructures observed in realistic composites can be reproduced by the proposed algorithm. Accurate prediction for the elastic properties and transverse isotropy of the generated microstructures with varying inclusion shapes using homogenization method further verify the validity of the developed method. A nonlinear damage study compared to the circular inclusion reveals that the capsule shape leads to a reduction of strength of unidirectional fiber reinforced composites under transverse tension or compression while the lobular inclusion presents a deterioration of strength under tension but an enhancement of strength in compression. The MPB algorithm provides an effective tool for micromechanical assessment, quantitative research and data-driven studies such as machine learning of composites.