Numerical investigation on the interaction between particles and eddies in gas-particle flows behind a backward-facing step

2D and 3D numerical investigation of a low speed particle-laden turbulent flow with different Reynolds numbers of 18,400 and 1,290 over a backward-facing step has been carried out. The gas phase is performed by Large Eddy Simulation and the particle phase is solved by a Lagrangian particle tracking model. Both the 2D and 3D simulations predict mean properties for both phases, are in good agreement with experimental results. However, there is large discrepancy for the fluctuating properties between 2D results and experimental results, while 3D results are in good agreement with experimental results. Further comparison indicates that although both 2D and 3D simulations can reveal the evolutions of the turbulent flow of the gas phase, the 3D simulation predicts much frequent activity of vortex evolutions such as rolling up, growing, merging and breaking up.; Simulation also compares the instantaneous concentration of particles with different Stokes numbers, initial velocity slip and the effect of gravitational force. Both 2D and 3D simulations give similar results on particle dispersions. Smallest particles are strongly controlled by the vortex structure of the gas phase and follow closely the gas vortices. Particles with time scales of similar order as the gas time scale are centrifuged out by a vortex and are preferentially concentrated along the edge of the gas vortices. Large particles essentially do not respond to the vortex motion within the gas time scale available and are also not preferentially concentrated.; The success of 3D simulation in predicting a two-phase turbulent flow provides a numerical basis for revisiting the gas-particle covariance model. Several gas-particle covariance models used in second-order closure models are evaluated in the present study. The predicted results by the models are in agreement with the numerical simulation results. However, a proper empirical constant is needed for different cases and there is no formula to determine the constant. A modified model is proposed in this study so that an empirical constant is no longer necessary. The predicted results using our model are as good as those from other models. Therefore, a better closure model is introduced for the gas-particle covariance model.