Search

Turbulent two-phase flows occur in a wide range of industrial processes, which strongly encourages the development of numerical methods for such flows. This work focuses on the physical phenomena of dispersion and preferential concentration of solid particles in a gas flow. The main purpose is to extend to Large-Eddy Simulation (LES), the Eulerian mesoscopic formalism introduced by Février et al. (2005) and first implemented by Kaufmann et al. (2006) to perfom Direct Numerical Simulation (DNS). When extending the Eulerian mesoscopic approach to LES, two main issues arise. First, as far as modeling is concerned, two different kinds of unclosed terms appear in the transport equations for the dispersed phase, which is very specific to this approach. Those terms are due either to ensemble averaging introduced by the mesoscopic approach, or by LES spatial filtering. Several closure models are proposed and tested a priori by Moreau (2006) who performs Discrete Particle Simulation (DPS) of particle-laden Homogeneous Isotropic decaying Turbulence (HIT). A posteriori validating these models is then required, which is the aim of the present work. Second, it is delicate to handle numerically the set of transport equations for the dispersed phase. Indeed, there are no physical diffusive terms in the transport equations and strong gradients difficult to represent on the grid must be accounted for. Consequently, first in this work, a new numerical method is proposed. The numerical scheme TTGC (Colin & Rudgyard, 2000) that is known to be few dispersive and few dissipative, is adapted to the dispersed phase and combined with a stabilising numerical method. Then, comparing Direct Numerical Simulation (DNS) of particle-laden decaying HIT flows performed with the Eulerian mesoscopic approach and the Lagrangian one shows the robustness and the accuracy of the numerical method proposed. Finally, the LES Eulerian mesoscopic modeling is validated a posteriori in two different complex geometries. For both, a detailed bank of experimental data are available. The first configuration consists of a particle-laden turbulent jet (Hishida et al., 1987). It requires to develop specific inlet Boundary Conditions (BC) for the dispersed phase. The second configuration is a particle-laden bluff body (Borée et al., 2001) where Eulerian mesoscopic and Lagrangian approaches can be evaluated. Comparing LES using the Eulerian mesoscopic approach with the experiments, and even with the DPS for the second geometry, shows that this new approach is able to accurately capture the dynamics of such particle-laden turbulent flows.