Multiscale modeling of segregation in granular flows

Modeling and simulation of segregation phenomena in granular flows are investigated. Computational models at different scales ranging from particle level (microscale) to continuum level (macroscale) are employed in order to determine the important microscale physics relevant to macroscale modeling. The capability of a multi-fluid model to capture segregation caused by density difference is demonstrated by simulating grain-chaff biomass flows in a laboratory-scale air column and in a combine harvester. The multi-fluid model treats gas and solid phases as interpenetrating continua in an Eulerian frame. This model is further improved by incorporating particle rotation using kinetic theory for rapid granular flow of slightly frictional spheres. A simplified model is implemented without changing the current kinetic theory framework by introducing an effective coefficient of restitution to account for additional energy dissipation due to frictional collisions. The accuracy of predicting segregation rate in a gas-fluidized bed is improved by the implementation. This result indicates that particle rotation is important microscopic physics to be incorporated into the hydrodynamic model. Segregation of a large particle in a dense granular bed of small particles under vertical vibration is studied using molecular dynamics simulations. Wall friction is identified as a necessary condition for the segregation. Large-scale force networks bearing larger-than-average forces are found with the presence of wall friction. The role of force networks in assisting rising of the large particle is analyzed. Single-point force distribution and two-point spatial force correlation are computed. The results show the heterogeneity of forces and a short-range correlation. The short correlation length implies that even dense granular flows may admit local constitutive relations. A modified minimum spanning tree (MST) algorithm is developed to asymptotically recover the force statistics in the force networks. This algorithm provides a possible route to constructing