Development of GPU Parallel Multiphase flow solver for turbulent slurry flows in cyclone

Abstract

The development of GPU parallelized unstructured multiphase solver and its application in predicting turbulent swirling flow of slurries inside cyclones is presented. Algebraic slip mixture model (ASM) is modified with additional shear lift forces and slurry rheology is corrected with fines fraction. The modified ASM model coupled with LES is used to predict particle classification inside hydrocyclone. During hydrocyclone operation the residence time of the fluid is very small and hence there is insufficient time for the larger eddies to cascade into smaller eddies. LES can accurately resolve flow structures that are few times the Kolmogorov scale at an increased computational cost due to finer mesh requirement. Therefore, the solver has been parallelized using general purpose graphics processing units (GPGPUs). In the current solver, the Pressure Poisson equation has been parallelized with an algebraic multigrid solver on GPU architecture using CUDA programming language for unstructured grids. The single phase flow field predicted by LES shows good agreement with experimental results obtained from open literature. The turbulent flow fields, the size segregation and the particle efficiency curve obtained from multiphase simulations are presented. Additionally, computational speedup due to GPU parallelization is reported over its serial version of the solver.