GPU-based unresolved LBM-DEM for fast simulation of gas-solid flows

A new discrete particle simulation algorithm, namely unresolved LBM-DEM method, is proposed. In this method, fluid phase and particle phase are calculated by lattice Boltzmann method (LBM) and discrete element method (DEM) respectively, and immersed moving boundary (IMB) is used for treating gas-solid coupling. The addi-tional collision term in IMB is regarded as force source term, and grid equivalent drag force is introduced to modify the weighting function. Due to the intrinsic mesoscopic properties of LBM, unresolved LBM-DEM method is suitable for simulating the hydrodynamics of Geldart A particles. In addition, GPU parallel computation of unresolved LBM-DEM is implemented to improve computational efficiency. To validate GPU-based unresolved LBM-DEM, it is used to simulate three classical gas-solid fluidizations. It was found that the results predicted by GPU-based LBM-DEM are all in good agreement with experimental data. Meanwhile, GPU-based LBM-DEM method brings one to two orders of magnitude speed up ratio compared with CPU or CPU-GPU heterogeneous CFD-DEM algorithm. The proposed unresolved LBM-DEM method has significant advantages over the traditional CFD-DEM in resolution scale and computational efficiency, suggesting it to be a promising computational strategy for fluidization and multiphase flows.