Mesh type plays a pivotal role in the accuracy, efficiency, and stability of CFD-DEM (Computational Fluid Dynamics–Discrete Element Method) simulations. While tetrahedral meshes are frequently adopted for their automation and flexibility, other mesh types—including hexahedral, polyhedral, and hybrid (hexahedral–polyhedral)—offer distinct trade-offs in terms of mesh quality metrics, computational cost, and numerical robustness.
This study compares the effects of four mesh types—tetrahedral, hexahedral, polyhedral, and hybrid—on key CFD-DEM simulation outcomes in a mechanically stirred mixing system. Evaluation criteria include mesh quality indicators such as skewness, non-orthogonality, and aspect ratio, as well as simulation runtime and convergence behavior. By using a standardized impeller-baffled tank configuration, the study ensures fair comparison across all mesh scenarios.
The results reveal that hybrid (hexahedral–polyhedral) meshes strike a favorable balance, offering superior orthogonality and lower skewness, which lead to enhanced numerical stability and reduced simulation divergence. Hexahedral meshes, while yielding the lowest numerical diffusion and highest accuracy in velocity fields, are constrained by limited geometric flexibility and higher meshing effort. Tetrahedral meshes, although easy to generate, suffer from higher skewness and reduced stability in capturing near-wall interactions and vortical structures. Hybrid meshes combine the geometric adaptability of tetrahedrals with the accuracy and stability of polyhedrals or hexahedrals, showing promise for complex geometries without significantly compromising performance. In terms of simulation runtime, polyhedral and hybrid meshes show reduced computational load due to better convergence and fewer iterations per timestep.
This work provides critical guidance on mesh selection strategies in CFD-DEM modeling of multiphase mixing systems. Future investigations will focus on mesh adaptation techniques and their impact on simulations involving non-spherical particles and multiphase turbulence modeling.
