Three-dimensional physics-based cellular automata model for granular shear flow

Abstract

Granular flows continue to be a complex problem in nature and industry where solid particles exhibit solid, liquid, and gaseous behavior in a manner which is often difficult to predict both locally and globally. In solids processing applications such as pharmaceutical production, food processing, and coal processing, the ability to accurately predict particle flows can improve process efficiency and product effectiveness. The “gold standard” for the modeling and prediction of granular flows is the discrete element method (DEM), which provides a rigorous physical treatment of particle interactions. One possible supplement to DEM-based analysis is lattice-based cellular automata (CA). CA provides a platform for obtaining fast predictions by pairing rule-based mathematics with high-fidelity physics to rapidly model physical processes, such as granular flows. The present work develops a novel three-dimensional (3D) physics-based CA framework for modeling granular shear flows. This 3D CA model extends the framework's modeling capabilities beyond the granular kinetic regime, includes treatments for force chains, and provides reasonably high accuracy in terms of predicting fundamental granular flow experiments. The predictive capabilities and quantitative accuracy of this CA model have been tested and validated against experimental shear flow results from an annular shear cell. Results from this validation work suggest that the model accurately predicts local flow properties. A parametric study varying the particle–particle coefficient of restitution demonstrated the ability of the 3D CA model to capture nonlinear trends in the local solid fraction profiles. A time study was also performed to highlight the computational efficiency of the 3D CA framework developed in this work. Results of this study showed that even at relatively low particle counts (104), the 3D CA model provides for considerable computational savings, with run times of two orders of magnitude less than a 2D DEM model.