Modelling of turbulent gas-solid flows from DNS to LES

Abstract

The purpose of this thesis is to explain the underlying physics of turbulent particle­ laden flows in the context of Direct Numerical Simulations (DNS). Moreover , mod­ elling tools, that can be used in Large Eddy Simulations (LES), are developed and validated against experimental and DNS data.

DNS of homogenous and isotropic turbulence are performed in order to explain the turbulence modulation caused by the particles. In order to carry out this analysis, a novel forcing method is developed that retains the statistics of the fluid. Contrary to other forcing schemes, it is shown that this scheme does not affect the statistics of both phases. The two-way coupling spectrum reveals that the particles transfer energy from the large to the small scales, whereby some of this energy is dissipated by the fluid. Additionally, a model two-way coupling spectrum is proposed that considers the observations of the DNS study.

Filtering is performed on the DNS results in order to remove the high frequency velocity components. It is evident from the results that the absence of the subgrid velocity fluctuations have an impact on the particle pair dispersion. Moreover , a novel stochastic model is proposed which reconstructs the subgrid scale turbulence characteristics. The results of the model are very promising as there is good agree­ ment with the DNS data.

LES of a turbulent horizontal particle-laden channel flow is performed in order to compare the soft-sphere and hard-sphere models and coupling strategies. A novel roughness model, used in conjunction with the soft-sphere model, is proposed. The results of the soft sphere and hard sphere models are in very good agreement with the available experimental data. Furthermore, the results, also, show that the wall roughness is an important mechanism in keeping the particles distributed across the channel despite the action of gravity.