Because exact Direct Numerical Simulation (DNS) resolving all scales down to the Kolmogorov microscale scales as (mathrm{Re}^{9/4}), simulating high-Reynolds-number industrial flows directly remains mathematically impossible for decades to come. Computational aerodynamics resolves this bottleneck through a hierarchical triad of turbulence modeling strategies: RANS, LES, and Hybrid RANS-LES.
1. The Spectrum of Turbulence Approaches
| Paradigm | Governing Philosophy | Resolved Scales | Computational Cost | Primary Use Case |
|---|---|---|---|---|
| RANS | Time/Ensemble Averaged | Mean flow only; all turbulent eddies modeled | Low ((sim 10^0)) | Standard industrial design, steady aerodynamics |
| LES | Spatial Low-Pass Filtering | Large energy-containing eddies resolved; subgrid eddies modeled | Medium-High ((sim 10^3 ext{–}10^4)) | Unsteady separation, acoustics, combustion mixing |
| DNS | Exact Navier-Stokes | All scales resolved from integral scale (L) to Kolmogorov (ta) | Astronomical ((sim 10^7 ext{–}10^9)) | Fundamental physics research, benchmark validation |
2. RANS Closures: The Boussinesq Eddy Viscosity Hypothesis
In 1877, Joseph Boussinesq proposed that turbulent Reynolds stresses are linearly related to the mean rate-of-strain tensor through an apparent scalar property called the turbulent eddy viscosity (mu_t) (or (
u_t = mu_t /
ho)):
[
–
ho overline{u’_i u’_j} = 2mu_t bar{S}_{ij} – rac{2}{3}
ho k_t delta_{ij}
]
Unlike molecular dynamic viscosity (mu) (an intrinsic fluid property), eddy viscosity (mu_t) is a flow property that varies from point to point depending on local turbulent intensity. To determine (
u_t), two-equation transport models solve transport PDEs for two turbulent scalars:
The (k ext{-}psilon) Model (Launder & Sharma)
Solves transport equations for turbulent kinetic energy (k_t) and its rate of dissipation (psilon):
[
u_t = C_mu rac{k_t^2}{psilon}, quad (C_mu approx 0.09)
]
The (k ext{-}psilon) model excels in free-shear flows (jets, wakes, mixing layers) but performs notoriously poorly near walls and under adverse pressure gradients.
The (k ext{-}omega) SST Model (Menter)
Florian Menter’s Shear Stress Transport (SST) model represents the gold standard in modern aerospace CFD. It blends the Wilcox (k ext{-}omega) model (which solves for specific dissipation rate (omega = psilon / k_t)) in the near-wall boundary layer with the standard (k ext{-}psilon) model in the outer freestream using spatial blending functions (F_1). Furthermore, it limits eddy viscosity to prevent unphysical over-prediction of shear stresses under adverse pressure gradients, accurately predicting flow separation and stall.
3. Large Eddy Simulation (LES): Resolving the Anisotropic Core
Large Eddy Simulation recognizes that large eddies are anisotropic, geometry-dependent, and carry (sim 80%) of turbulent kinetic energy, while small subgrid eddies are statistically isotropic and universal. LES applies a spatial low-pass filter (G(mathbf{x}, Delta)) of width (Delta = (Delta x Delta y Delta z)^{1/3}) to the Navier-Stokes equations:
[
rac{partial bar{u}_i}{partial t} + rac{partial(bar{u}_i bar{u}_j)}{partial x_j} = -rac{1}{
ho}rac{partial bar{p}}{partial x_i} +
u
abla^2 bar{u}_i – rac{partial au_{ij}^{ ext{SGS}}}{partial x_j}
]
The unclosed Subgrid-Scale (SGS) stress tensor ( au_{ij}^{ ext{SGS}} = overline{u_i u_j} – bar{u}_i bar{u}_j) represents momentum transfer between resolved large scales and unresolved subgrid scales.
The Smagorinsky Subgrid Model
The classical Smagorinsky model expresses SGS stress via an eddy viscosity scaled by the filter width (Delta) and the magnitude of the resolved strain rate tensor (|bar{S}| = sqrt{2 bar{S}_{ij}bar{S}_{ij}}):
[
au_{ij}^{ ext{SGS}} – rac{1}{3} au_{kk}^{ ext{SGS}}delta_{ij} = -2
u_{ ext{SGS}} bar{S}_{ij}, quad
u_{ ext{SGS}} = (C_s Delta)^2 |bar{S}|
]
where (C_s approx 0.1 ext{–}0.17) is the Smagorinsky coefficient. In modern dynamic Smagorinsky models (Germano identity), (C_s) is computed dynamically at every time step from resolved scales, allowing natural representation of laminar-to-turbulent transition and backscatter (energy transfer from small scales back to large scales).
4. Hybrid RANS-LES: Detached Eddy Simulation (DES)
While LES drastically reduces computational cost compared to DNS in free shear flows, near solid walls resolving the tiny near-wall turbulent streaks requires grid spacing scaling as (Delta x^+ sim 30, Delta y^+ sim 1), making wall-resolved LES prohibitively expensive at flight Reynolds numbers. Detached Eddy Simulation (DES) uses RANS inside the thin attached boundary layer and seamlessly switches to LES in separated wake regions, providing an optimal compromise for industrial aerodynamics.
