Voronoi-based periodic representative volume element generation method for efficient multiscale analysis of random particulate composite

This study proposes a Voronoi-based modelling method to efficiently generate periodic representative volume elements (RVEs) for particulate composites containing particles with random distributions, highly varied shapes, large size differences and high volume fractions, e.g., asphalt concrete. This method develops a replicating and partitioning procedure to initially generate periodic Voronoi diagrams and a repartitioning procedure to further acquire Voronoi cells with the required size distributions. A surface smoothing procedure is presented to reshape the Voronoi cells into particles with multiple levels of smoothness. Further, a picking procedure is proposed to select the particles with the required volume fractions. In these procedures, the Voronoi cells directly impacting the model periodicity are grouped, and those belonging to the same group should undergo consistent manipulations to ensure the periodicity. The proposed modelling method enables efficient and reliable multiscale simulations of particulate composites by using the assembly of periodic RVEs within the region of interest.