Modeling of subgrid-scale effects on particles in large-eddy simulation of turbulent two-phase flows

Most of the previous works in the LES of particle-laden turbulent flows, which use the trajectory approach, have assumed that the effect of subgrid scales on particles is negligible. This assumption is not accurate, particularly, when the particle time constant is small and/or the filtered energy from the carrier phase is significant. Two models are developed and validated via a priori and a posteriori tests to relax this assumption.; In the stochastic model,{09}the particle position and velocity and the seen fluid particle velocity combine into a random vector which evolves based on a Lagenvin type of stochastic differential equation. The model is validated in the decaying isotropic turbulence and its performance is very promising far the particles with small particle time constants regardless how much energy is filtered. However, discrepancy is observed between the model and DNS results for particles with large time constants.; In the deterministic model, the approximate deconvoluted velocity is obtained by applying the approximate deconvolution on the resolved velocity and then the approximate velocity is used to advance particles. The model is successfully validated on the homogeneous isotropic and shear turbulence for all considered particle time constants.; Non-isothermal particle-laden homogeneous shear turbulent flows are also studied by DNS. Both one- and two-way couplings are considered. The results indicate that the increase of the mass loading ratio or the particle time constant generally reduce the temperature variance and the magnitude of the turbulent heat flux of both carrier and dispersed phases. The increase of the ratio of specific heats increases the particle temperature variance but exhibits an opposite effect on the fluid. The magnitude of the turbulent heat flux of the carrier phase is not influenced by the specific heat ratio.; All DNS and LES simulations are conducted using a pseudo-spectral code on the homogeneous turbulence. The subgrid-scale model utilized in LES is dynamic Smagorinsky.