MatGBM: A Computer Vision-Aided Triangular Mesh Generator for High-Fidelity Grain-Based Model

Abstract

The grain-based model (GBM) stands as a renowned model for polycrystalline simulations in computational mechanics. Despite its popularity, there remains a critical need for a more advanced and user-friendly tool to generate high-fidelity microstructures with specified grain size distributions. Addressing this need, this paper introduces ’MatGBM’, an innovative modeling tool that aspires to enhance numerical simulations of polycrystalline materials. MatGBM seamlessly integrates three modules: a computer vision-aided mineral grain distribution detection, Voronoi tessellation processing, and triangular mesh generation. To accurately capture the two-dimensional structural characteristics of polycrystalline materials, the mineral grain distribution detection module employs computer vision functions to pinpoint particle coordinates and areas. The weighted Voronoi tessellation is generated and processed based on the original grain distribution features, resembling the original image of the polycrystalline material more closely than basic Voronoi tessellation. Finally, MatGBM directly outputs triangular mesh using two optional meshing tools based on the Voronoi polygons. Our rigorous testing via uniaxial compressive tests, Brazilian splitting tests, and three-point bending tests in crystalline rocks and metals, using the combined finite-discrete element method, validates that MatGBM can reliably reproduce the key deformation, damage, and failure characteristics of polycrystalline materials. Overall, MatGBM emerges not only as a promising tool for numerical simulations of rock, metallurgic, and ceramic materials, but also as a potent pre-processing tool for multiple numerical methods.