ANSYS Fluent Turbulence Modeling
#+CATEGORY: ansys
ANSYS Fluent Turbulence Modeling :: RANS, LES, DES, and Hybrid Approaches
ANSYS Fluent provides many turbulence models. They range from RANS (Reynolds-averaged) to DNS (direct numerical simulation). Hybrid RANS/LES (DES, SAS) bridges the gap between scale-resolving and scale-modeling approaches. This note catalogs all available turbulence models. It includes their governing equations, assumptions, recommended applications, and comparison with OpenFOAM equivalents.
RANS Turbulence Models
| Model | Equations | Closed Variables | Default Diffusion | Best For |
| ------- | ----------- | ----------------- | ------------------- | ---------- |
| Spalart-Allmaras | 1 | ν̃ | Prandt | External aerodynamics, boundary layers |
| Standard k-epsilon | 2 (k, epsilon) | σₖ, σε | Standard | Industrial, high-Re, attached flows |
| RNG k-epsilon | 2 (k, epsilon) | Modified RNG theory | RNG enhanced | Moderate swirl, high strain |
| Realizable k-epsilon | 2 (k, epsilon) | Modified Cμ | Modified turbulence diffusivity | Jet flows, boundary layers with pressure gradient |
| Reynolds Stress Model (RSM) | 7 (6 RNG + k) | Full stress tensor transport | Full differential pressure | Strong swirl, rotating flows, secondary flow |
| v2-f | 5 (k, epsilon, v, f) | Near-wall correction | Near-wall v²-f model | Near-wall attached flows with strong curvature |
| k-omega SST | 2 (k, omega) | Blended k-epsilon/k-omega | Shear stress transport | General engineering, adverse pressure gradients, separation |
Spalart-Allmaras
The one-equation model targets external aerodynamics (aerofoil, aircraft). It solves a transport equation for a modified eddy viscosity ν̃ (related to actual turbulent viscosity νₜ):
where $P_{\tilde{\nu}}$ is production and $D_{\tilde{\nu}}$ is destruction/dissipation.
Turbulent viscosity comes from $\tilde{\nu}$:
with $f_{v1}$ a damping function. It vanishes near walls.
Fluent implementation: Spalart-Allmaras model. Skip low-Re correction. The model works on y₊ ≈ 1 walls. Users typically apply wall functions (y₊ ≈ 30–300).
RNG k-epsilon
The RNG $k$-$\epsilon$ model derives from Renormalization Group (RNG) theory. It modifies the standard $k$-$\epsilon$ equations. It incorporates small-scale turbulence effects analytically. Key differences from Standard $k$-$\epsilon$:
1. New term in the $\epsilon$-equation improves accuracy for high strain rates and moderate swirl 2. $C_\mu$ is no longer constant. It depends on the dimensionless strain rate:
Where $P_{ij}$ is production, $D_{ij}$ is diffusion, $\Pi_{ij}$ is pressure-strain correlation, and $\epsilon_{ij}$ is dissipation. A model handles the pressure-strain term (usually Rotta + Daly-Harlow closure).
Fluent RSM is the most complete RANS model (second-moment closure). It is also the most expensive (9 equations total: 7 transport + 2 for k and $\epsilon$). It converges significantly worse than two-equation models.
k-omega SST (Shear Stress Transport)
Use k-omega SST as the default RANS model for general engineering in Fluent. It blends the $k$-$\omega$ model near the wall with the $k$-$\epsilon$ model in the free stream. The $k$-$\omega$ model has correct near-wall asymptotic behavior. The $k$-$\epsilon$ model works better in the free stream (where $k$-$\omega$ is over-sensitive).
The blending function approaches 1 (pure $k$-$\omega$) near walls and 0 (pure $k$-$\epsilon$) in the free stream. The SST modification adds a limiter on eddy viscosity. This prevents over-prediction of turbulent shear stress:
where $\alpha$ is a blending function. This limiter makes k-omega SST superior to standard $k$-$\omega$ for adverse pressure gradient flows.
Fluent implementation: kOmegaSST model. Omega at walls defaults to $S\frac{\nu}{\beta_1 y^2}$, where $S$ is the specific rotation rate and $y$ is near-wall distance.
LES (Large Eddy Simulation)
Fluent's LES solves filtered Navier-Stokes equations with an SFS (SubFilter-Scale) model. Available SFS models:
| Model | Description |
| ------- | ------------- |
| Smagorinsky | Classic eddy-viscosity model (constant Cs) |
| Dynamic Smagorinsky | Dynamically computed Cs via test-filter |
| WALE (Wall-Adapting Local Eddy-viscosity) | Improved near-wall behavior; no damping needed |
| Vreman | Similar to WALE, slightly different formulation |
LES in Fluent needs:
- Fine mesh (typically $10^7$–$10^9$ cells for engineering applications)
- Small time step (CFL ≤ 1)
- No wall functions unless using hybrid RANS/LES
- Prescribed inlet turbulence (synthetic turbulence, precursor simulation, or recycled data)
Use the WALE model for LES with Fluent. It provides correct near-wall scaling. It needs no damping functions:
where $S_{ij}^d$ is the symmetric part of the squared velocity gradient.
DES / DDES / SAS (Hybrid RANS/LES)
Fluent supports these hybrid RANS/LES approaches:
| Model | Description |
| ------- | ------------- |
| SST-DES | Delayed Eddy Simulation using SST-RANS away from wall, LES in separated regions |
| SST-DDES | Delayed DES (avoids LES buffer layer issues near walls) |
| SST-SAS | Scale-Adaptive Simulation (RANS model with SAS term for enhanced LES behavior in unsteady regions) |
| IDDES | Improved DDES (includes near-wall modeling capability) |
Summary (ANSYS vs OpenFOAM)
| Model/Model | ANSYS Fluent | OpenFOAM Equivalent |
| ------------- | ------------- | ------------------- |
| k-omega SST | kOmegaSST (RANS) | kOmegaSST, kOmegaSSTIDDES |
| k-epsilon (Standard) | StandardKEpsilon | kEpsilon |
| k-epsilon (RNG) | RNGKEpsilon | RNGkEpsilon (available via kEpsilon model selection) |
| k-epsilon (Realizable) | RealizableKEpsilon | RealizableKEpsilon |
| Reynolds Stress Model (RSM) | ReynoldsStressModel | RASModel with RSM |
| Spalart-Allmaras | SpalartAllmaras | SpalartAllmaras |
| Smagorinsky (LES) | Smagorinsky | Smagorinsky |
| Dynamic Smagorinsky (LES) | DynamicSmagorinsky | DynamicSmagorinsky |
| WALE (LES) | WALES | WALES |
| SAS | SSTSAS | kOmegaSSTSAS |
| SST-DES | SSTDES | DES |
| SST-DDES | SSTDDES | DDES |
| SST-IDDES | SSTIDDES | IDDES |
See Also