Large Eddy Simulation (LES) Turbulence Models

Apart from the popular RANS turbulence models, SimScale also offers Large Eddy Simulation (LES) turbulence models like LES Smagorinsky and LES Spalart-Allmaras. The following simulation types can use at least one of the listed LES turbulence models:

• Incompressible
• Incompressible (LBM)
• Compressible
• Convective Heat Transfer
• Multiphase

Large Eddy Simulation (LES)

Large Eddy Simulation (LES) is a turbulence modeling approach used in CFD to simulate turbulent flows by resolving large eddies and modeling smaller ones. The large eddies, which carry most of the turbulent energy, are directly computed by computing the Navier-Stokes equations for each mesh cell, while the smaller, more universal eddies are approximated using Subgrid-Scale (SGS) models.

LES excels in capturing unsteady, large-scale flow phenomena such as vortex shedding and flow separation, making it ideal for complex aerodynamic studies. It offers greater accuracy than traditional methods by allowing the large eddies to evolve naturally based on flow conditions, while only approximating the smaller scales of turbulence.

Large Eddy Simulation (LES) Turbulence Models in SimScale

There are currently three LES turbulence models in SimScale as follows:

LES Smagorinsky Model

This is one of the most commonly used subgrid-scale models for LES. It assumes that turbulence at scales smaller than the grid resolution (subgrid scales) can be modeled with an eddy viscosity proportional to the square of the local strain rate. It introduces a constant viscosity term to account for unresolved turbulent structures

LES Spalart-Allmaras Model

This is a hybrid approach combining Spalart-Allmaras, a well-known Reynolds-averaged Navier-Stokes (RANS) turbulence model, with Large Eddy Simulation. The Spalart-Allmaras component is usually used near walls, while LES is used in the bulk flow regions. It is a one-equation LES turbulence model simplifying computations by solving for a single variable, the turbulent viscosity, to model various turbulence effects

LES Smagorinsky (Direct) Model

This version of the LES Smagorinsky model is only available for the Incompressible (LBM) simulations. While the Smagorinsky model strictly follows the original formulation and LES idea, the Smagorinsky (Direct) is a bit cheaper, but a bit modified. For Smagorinsky (Direct), only the Lattice Boltzmann Method (LBM) mesh has to be computed during a time step, while for Smagorinsky, the LBM and the Finite-Difference meshes have to be computed during a time step making it comparatively costlier.

How LES is Different from RANS?

RANS (Reynolds-Averaged Navier-Stokes) averages all scales of turbulence, focusing on time-averaged equations that model the entire turbulent flow. In contrast, LES resolves large eddies and models only the small ones. This gives LES the ability to capture complex, transient flow features that RANS often oversimplifies.

The main differences are as follows:

RANS: Efficient, lower computational cost, suited for steady-state flows but struggles with complex or unsteady turbulent flows.

LES: Higher accuracy in resolving unsteady flows and complex turbulence but requires significantly more computational power and finer grids.

You can also read more about the difference between LES, RANS, and DNS in the following SimWiki article:

What is Large Eddy Simulation (LES) in CFD?

How to Decide the Mesh Fineness?

The accuracy of LES depends heavily on the mesh resolution. Key factors in mesh fineness include:

1. Grid Resolution: The grid must be fine enough to resolve large eddies while relying on SGS models for smaller eddies. A finer mesh is required in regions with high turbulence, like wakes and separation zones.
2. Near-Wall Regions: Near walls, turbulence is more complex. A finer grid is necessary to resolve the boundary layer and small-scale eddies. In practice, wall-adapted LES may be used to balance accuracy and computational cost.
3. Critical Flow Areas: Apply finer mesh in regions of interest (e.g., around the diffuser or wake regions). Use SimScale’s automatic mesh refinement features for a broader control, but manually refine critical zones for more detailed resolution.

Balancing mesh fineness and computational cost is key—start with a coarse mesh for general flow insight, then iteratively refine important areas for a more accurate LES.

Last updated: February 11th, 2025