A voxel-based clump generation method used for DEM simulations

摘要

Abstract Using a cluster of pebbles to generate clumps is a common procedure to explore granular particles' micro- or macro-scale mechanical responses in DEM simulations. The struggle is to balance the number of pebbles and the shape accuracy of generated clumps compared to the original particles. A low-cost multi-image-based method is adopted in this paper to extract 3D triangular networks of original particles. A voxel-based clump generation method, namely V-CLUMP, is developed to generate clumps based on triangular networks. The operations, such as gradual discretization, surface cell optimization, and candidate pebble hollowing out, are proposed to improve the accuracy and efficiency of clump generation. The ideal of corner preserving is also adopted, and a novel image-based method to detect corners and ridges of particles or clumps is proposed in this paper. Clumps for four idealized geometrical shapes and five quintessential particles are generated using the V-CLUMP. The parameters, including the number of pebbles, volume, projection area, sphericity, roundness, and roughness, are used to evaluate the performance of the clump generation method proposed in this study. The result shows that the absolute value of the errors for most generated clumps with 30 to 120 pebbles in macro-scale descriptors, i.e., volume, projection area, and sphericity, are less than 6%. A better result of meso- and micro- descriptors, i.e., roundness and roughness, can be obtained by selecting optimum clump generation parameters. The code takes less than one second to a few minutes, within the acceptable limits. However, pebble hollowing out and corner detecting are still time-consuming, and more efficient algorithms are needed to improve the performance of the operations.