Comparative analysis of porosity coarse-graining techniques for discrete element simulations of dense particulate systems

Abstract

The Discrete Element Method (DEM) is a well-established approach to study granular materials in numerous fields of application; modelling each granular particle individually to predict the overall behavior. This behavior can be then extracted by averaging, or coarse graining, the sample using a suitable method. The choice of appropriate coarse-graining method entails a compromise between accuracy and computational cost, especially in the large-scale simulation typically required by industry. A number of coarse-graining methods have been proposed in the literature, these are reviewed and categorized in this work. Within this contribution two novel porosity coarse-graining strategies are proposed including a Voxel method where a secondary dense grid of "pixel-cells" is implemented adopting a binary logic for the coarse graining and a Hybrid method where both analytical formulas and pixels are utilized. The proposed methods are compared with four coarsegraining schemes that have been documented in the literature, including the Particle Centroid Method (PCM), an Analytical method, a method which solves the diffusion equation and an approach which employs averaging using kernels. The novel methods are validated for problems in both two and three dimensions through comparison with the "accurate" Analytical method. It is shown that, once validated, both the proposed schemes can approximate the exact solutions quite accurately, however there is a high computational cost associated with the Voxel method. The accuracy of both methods can be adjusted allowing the user to decide between accuracy and computational time. A detailed comparison is then presented for all six schemes considering "accuracy", "smoothness" and "computational cost". Optimal parameters are obtained for all six methods and recommendations for coarse graining DEM samples are discussed.

Pyrolysis of Spherical Wood Particles in a Packed Bed – Comparison between Resolved and Unresolved Discrete Element Method/Computational Fluid Dynamics

“…In AVM, convective heat transfer _ Q p between particle p and surrounding fluid is calculated as follows: Here, htc is the heat transfer coefficient, T fl is an averaged fluid temperature, T pf is the triangulated surface temperature, and A pf denotes the surface area. Fluid properties (e.g., temperature, thermal conductivity) are averaged for the whole particle, using a kernel approach as proposed by Kalderon et al [22].…”

“…Fluid properties (e.g., temperature, thermal conductivity) are averaged for the whole particle, using a kernel approach as proposed by Kalderon et al. 22. …”

Locally Resolved Simulation of Gas Mixing and Combustion Inside Static and Moving Particle Assemblies

“…Thus, differences between unresolved methods exist in the algorithm computing the porosity distribution as well as in the way source terms are computed and distributed. A comparative study between methods to compute the local porosity in dense particulate systems can be found in [22]. A more detailed overview of the status of unresolved DEM-CFD simulations targeting reactive particle systems are given in [21,[23][24][25].…”