A MONTE CARLO APPROACH TO MULTISCALE MODELING OF GRANULAR GAS OF NON-SPHERICAL PARTICLES

Granular flow of non-spherical particles is common in nature and industrial processes. To understand the behavior of granular systems of these non-spherical particles, computational methods are employed to simulate these systems. However, current state-of-the-art simulation methods (TFM and DEM)have two primary drawbacks: (1) high computational cost restricting this method to small-scale systems (DEM) and (2) the use of empirical correlations that cannot be reliably extrapolated to different systems (TFM). Also, due to the statistical limitation and lack of physics-based continuum description, making progress in non-spherical particle flow dynamics with the study of its higher-order moments and transport coefficients is virtually unfeasible. To address these challenges, a DSMC model is developed to simulate a granular gas of spherocylinders with varying aspect ratios. In this work, a 3D classical trajectory calculation (CTC) code is developed to generate pairwise collision data sets. In addition, the Gaussian mixture model, an unsupervised machine learning technique, is used to construct the complex probability distributions required by the DSMC model. Subsequently, the model is implemented and validated against exact solutions derived from equivalent DEM simulations. The model shows high accuracy on both the macroscopic and microscopic scales and is more than 50 times faster than the DEM. The distribution functions of energies and velocities are extracted over time. Following the methodology presented, this approach can be easily adjusted to accommodate different particle shapes.