Subgrid modeling in particle-laden channel flow

Abstract

Direct numerical simulation (DNS) and large-eddy simulation (LES) of particle-laden turbulent channel flow, in which the particles experience a drag force, are investigated for two subgrid models and several Reynolds and Stokes numbers. In this flow, turbophoresis leads to an accumulation of particles near the walls. The objectives of the work are to investigate the accuracy of the subgrid models studied with respect to particle behavior and to explain the observed particle behavior predicted by the different models. The focus is on particle dispersion and mean particle motion in the direction normal to the walls of the channel. For a low Reynolds number, it is shown that the turbophoresis and particle velocity fluctuations are reduced compared to DNS, if the filtered fluid velocity calculated in the LES is used in the particle equation of motion. This is a combined effect of the disregard of the subgrid scales in the fluid velocity and the inadequacy of the subgrid model. Better agreement with DNS is obtained if an inverse filtering model, which was recently proposed, is incorporated into the particle equation. This model is shown to enhance turbophoresis and particle velocity fluctuations in actual LES. The results of the approximate deconvolution model (ADM) agree better with DNS results than results of the dynamic eddy-viscosity model. This can be explained from the better prediction of the fluid velocity statistics by ADM and the better correspondence of the subgrid models adopted in the fluid and particle equations. Although the differences between the two subgrid models become smaller, similar conclusions are obtained at a higher Reynolds number. Compared to fourth-order interpolation of the fluid velocity to the particle position, second-order interpolation approximately cancels the effect of the subgrid model in the particle equation of motion.