Evaluation of physics based hard-sphere model with the soft sphere model for dense fluid-particle flow systems

Abstract

In this paper we present the application of a physics-based three dimensional hard-sphere impact model to many-particle systems. The hard sphere model is used with conventional time-advancement which is not event driven. A particle relocation technique based on using both pre- and post-collision velocities is developed and validated for dense systems. Also developed is a parallelization scheme which overcomes the inherent sequential nature of processing collisions with the hard sphere model. The relocation technique is tested in a one-dimensional stack of particles and in the three-dimensional settling of a bed of particles. The technique demonstrates its ability to prevent excessive overlaps and maintaining stable stacks of particles at coefficients of restitutions ranging from 0.0 to 0.98. The model is validated in a bubbling fluidized bed and shows no discernable differences with the soft sphere model but at a computational cost about 23 times less than that of the soft sphere model. The model is used to investigate particle agglomeration of 20 µm size particles in a turbulent ribbed duct flow using LES. It was found that the 20 µm particles were prone to agglomerate in regions of low velocity at the upstream and downstream corners of the rib-wall junction. The hard sphere model enabled time steps an order of magnitude larger than that permissible for the soft sphere model which resulted in DEM speedups of more than 20 times.