On the Reliability of Large-Eddy Simulation for Dispersed Two-Phase Flows

This work deals with the numerical description of dilute polydispersed and turbulent two-phase flows. The method used to describe the two-phase flow is based on the Eulerian-Lagrangian approach. The carrier phase turbulent flow behavior is modeled utilizing Large Eddy Simulation (LES). Most of the recent DNS investigations attempting to provide insight into particle-fluid interaction mechanisms are performed on simple, mostly homogeneous turbulent flows with very low Reynolds numbers. In contrast to this, LES is an opportunity to extend this studies to flows with significantly higher Reynolds numbers and to more complex geometries. Moreover, it is less restrictive to flows containing particles, larger than the Kolmogorov length scale. The main objective of this work is to assess the state of the art capabilities of the two-phase LES and to identify weak points. Recommendations for additional model development for increasing the predictive ability are derived. The reliability of transfer of findings from one configuration to others is assessed. The conclusions are based on a systematic variation of relevant flow parameters, such as the Reynolds number and the Stokes number, so that a wide range of applications is covered. Therefore, several particle-laden flow configurations (two plane channel flows, a free jet and an evaporating spray at low temperature) are investigated. Special attention is paid to the prediction of the particle preferential accumulation, because of its importance for simulations of mixing and combustion in turbulent flows. By performing a systematic variation of the particle Stokes number (here defined as the ratio of the particle time scale to the Kolmogorov time scale of the fluid) and the flow Reynolds number, it is observed that the maximum preferential accumulation occurs at a constant Stokes number and that this number does not depend on the Reynolds number. The magnitude of the accumulation, however, depends on the Reynolds number of the flow. The effect of an additional sub-grid scale dispersion model on the particle accumulation is found to be less pronounced for particles with characteristic time scales in the order of the Kolmogorov scale. An additional issue arises due to unphysical particle number densities, when preferential accumulation occurs. In this context, the usage of a model, accounting for natural dispersion due to particle collisions, is inevitable. The influence of the dispersed phase on the fluid turbulence is considered in a wall bounded and a free shear flow. The improper modeling of the particle-wall interaction mechanisms can lead to different solutions for the particle velocity field and therefore to different levels of turbulence modification in the near wall region. It should be noted that these solutions are not necessarily unphysical, as some experimental investigations confirm. In contrast to this, the turbulence modification in a free shear flow is predicted very well, moreover, the prediction is found to be reliable for different mass loadings of the dispersed phase. Within the scope of this work, a time efficient non-iterative Fractional Step procedure is applied for the pressure-velocity coupling of the carrier fluid governing equations. The coupling of the method with two-way coupled particle tracking simulations is validated successfully on configurations of isothermal flows and such with low density variation using experimental and DNS data