In the LES approach, large eddies are resolved directly, while small eddies are modeled. Therefore, modeling approximations are required only for part of the turbulent spectrum that are less likely to be influenced by the flow and boundary conditions. If the mesh and time step allow a large proportion of eddies to be resolved, the LES model produce better resolution of transient, recirculation and vortex structures than RANS models. However, the challenge of a multitude of length and time scales present in turbulent flows is intensified due to the additional length and time scales caused by the secondary phases.
Within the Euler-Euler approach, the effects of the phase interfaces are averaged, and semi-empirical formulations are used to account for phase interactions. These models are generally developed for a certain type of multiphase flow regimes and, therefore, suffer from inaccurate prediction of flow-regime transitions, particularly when the flow geometry changes. The use of LES combined with data obtained from DNS and experiments has the potential of developing multiphase flow strategies that rely on underlying microscale physics and may accurately predict the flow-regime transition [1].
Numerous applications of the Euler-Euler based LES to dispersed multiphase flows are reported in literature [2], [3]. [4], and [5]. In existing interaction models, drag force depends on a power of slip velocity between phases, which, if averaged, produces extra turbulent correlations that are not directly modeled, but indirectly accounted for through other terms. Since the resolved velocities are used within LES, the ignored correlations are less significant and, therefore, LES has an impact on model constants used in the interaction models. Better interaction models are very much so a research topic at present [1], and the publications sited above more or less use the same interaction models as are used with RANS models.
The transport equations are filtered in LES. For CFD analyses, the most appropriate filter is the cell control volume. The filtered equations are the same as ensemble-averaged equations with the following differences:
the variables represent resolved values rather than ensemble-averaged RANS values
the turbulent correlations are the subgrid correlations rather than RANS correlations
In Ansys Fluent, the following methods are available for modeling the subgrid correlations:
Mixture LES multiphase model
This model uses mixture properties and velocities. It is applicable to multiphase flows when phases separate as in stratified or nearly stratified multiphase flows, and when the density ratio between phases is close to 1.
Dispersed LES multiphase model
This model Is applicable when there is clearly one carrier phase and the concentrations of the secondary phases are dilute. The turbulence correlations within the carrier phase are modeled, and the dispersed phase kinematic viscosity is assumed to be equal to the carrier phase kinematic viscosity or ignored altogether.
The LES subgrid models available for the Eulerian multiphase model are:
Smagorinsky-Lilly model
Dynamic Smagorinsky-Lilly model
Wall adapting local eddy-viscosity (WALE) model
Algebraic wall-modeled LES model (WMLES)
Dynamic kinetic energy subgrid-scale model
The above models are described in section Subgrid-Scale Models in the Ansys Fluent Theory Guide.
The first four models use algebraic relations to obtain the subgrid correlations from resolved dependent gradients (such as velocity). Gradients are obtained:
from mixture variables for the Mixture LES multiphase model
from the carrier phase variables (such as velocity gradients) for the Dispersed LES multiphase model
For the dynamic kinetic energy subgrid-scale model, the transport equation is the same as its single-phase version described in section Dynamic Kinetic Energy Subgrid-Scale Model in the Ansys Fluent Theory Guide. However, like the RANS transport equation for the turbulent kinetic energy, it contains an extra source term for modeling the interaction between the dispersed phases and the continuous phase. These models are the same as their counterparts in RANS models described in section k- ε Dispersed Turbulence Model in the Ansys Fluent Theory Guide.
The turbulent dispersion forces and turbulent interaction models used with LES are the same as those available for RANS models. However, unlike the RANS models that uses averages for all turbulence spectrum, here all turbulent parameters are based on values obtained for subgrid scales. Therefore, the impact of any errors resulting from these models is expected to be less significant for LES than for RANS. For more information about these models, see section Turbulent Dispersion Force and section Turbulence Interaction Models in the Ansys Fluent Theory Guide.
After enabling beta feature access (Introduction), the Large Eddy Simulation (LES) option becomes available in the Viscous Model dialog box (Model group box) for Eulerian multiphase cases.
Note: The Large Eddy Simulation (LES) model is not compatible with the Population Balance model.
Once you select Large Eddy Simulation (LES), you can choose from the following LES subgrid-scale models:
Smagorinsky-Lilly
WALE
WMLES
WMLES S-Omega
Kinetic-Energy Transport
For modeling the subgrid correlations, you can select from the following models in the LES Multiphase Model group box:
Mixture
Dispersed
See Theory for details about these models.
[2] "Large Eddy Simulation of the Gas-Liquid Flow in a Square Cross-Sectioned Bubble Column". UChemical Engineering Science. 6341–6349. 2001.
[3] "Numerical Simulation of Dynamic Flow Behaviour in a Bubbly Column: A Study of Closure for Turbulence and Interface Forces". Chemical Engineering Science. 7593–7608. 2006.