Large Eddy Simulation
Also known as: LES, subgrid-scale modeling
Large Eddy Simulation (LES) is a turbulence modeling technique that explicitly resolves large-scale turbulent eddies while modeling small-scale subgrid-scale (SGS) motions. Introduced by Joseph Smagorinsky in 1963, LES represents a middle ground between Reynolds-Averaged Navier-Stokes (RANS) and Direct Numerical Simulation (DNS). By capturing the energy-containing scales of turbulence, LES provides superior accuracy for transient flows and complex geometries at computational costs significantly lower than DNS.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
LES is appropriate when unsteady turbulent structures are important, such as flow instabilities, vortex shedding, jet mixing, and cavity acoustics. Use LES for problems where RANS cannot capture the physical phenomena and DNS is computationally prohibitive. LES works best for flows with significant separation, mixing layers, and shear-driven turbulence. Good for transient flows, acoustics, and complex geometries. Avoid LES if only steady-state mean flow is needed or if the domain is too large to afford the computational expense.
Strengths & limitations
- Explicitly resolves large-scale turbulent structures and their dynamics
- Provides unsteady flow information including pressure and velocity fluctuations
- More accurate than RANS for separated flows and complex geometries
- Lower computational cost than DNS while maintaining superior accuracy for most engineering problems
- Applicable to a wider range of flow regimes than RANS-based closures
- Requires finer spatial and temporal resolution than RANS, increasing computational cost tenfold to hundredfold
- Subgrid-scale models introduce additional modeling uncertainty beyond numerical error
- Near-wall modeling remains challenging; wall-resolved LES requires very fine boundary layer meshes
- Strongly dependent on numerical schemes; non-dissipative schemes may accumulate energy in small scales
- Longer simulation times needed to achieve statistically converged averages
Frequently asked
What is the main difference between LES and RANS?
RANS time-averages and models all turbulent scales, yielding only steady-state mean flow statistics. LES spatially filters and resolves large eddies explicitly while modeling small scales, capturing transient eddy dynamics and unsteady phenomena. RANS is cheaper but less accurate; LES costs more but provides superior detail for flows where unsteadiness matters.
How fine must my mesh be for LES?
The mesh size should be on the order of the filter width, typically 1-5% of a characteristic dimension (duct width, jet diameter). In wall-resolved LES, first grid point should satisfy y+ < 1; in wall-modeled LES, y+ can be 40-100 but boundary layer structure is less accurate. For 3D LES, expect 1-100 million cells depending on geometry and desired accuracy.
Why do I need a subgrid-scale model if I am resolving the large eddies?
Resolved eddies interact with unresolved small eddies through the subgrid-scale stress tensor. This interaction removes energy from resolved scales (dissipation) and provides backscatter. A closure model approximates these stresses, usually as eddy viscosity proportional to local strain rate. Without it, energy accumulates at small resolved scales and the simulation diverges.
How long should I run an LES to get converged statistics?
LES must run long enough for turbulent eddies to develop and decorrelate. Typically 10-50 eddy turnover times are needed for first-order statistics and 50-200 for spectra and intermittency. For a jet with exit velocity U and diameter D, one turnover time is roughly D/U. Running a 3D simulation long enough is often the dominant computational cost.
Can LES predict mean flow quantities as well as RANS?
Yes, LES can predict time-averaged quantities accurately provided sufficient grid resolution and simulation time. For well-behaved flows like channel or pipe flow, LES agrees with experimental mean profiles better than RANS. However, for extremely complex geometries or highly separated flows, the added cost may not justify the accuracy gain over optimized RANS models.
Sources
- Smagorinsky, J. (1963). General circulation experiments with the primitive equations: I. The basic experiment. Monthly Weather Review, 91(3), 99-164. DOI: 10.1175/1520-0493(1963)091<0099:GCEWTP>2.3.CO;2 ↗
- Leonard, A. (1974). Energy cascade in large-eddy simulations of turbulent fluid flows. Advances in Geophysics, 18, 237-248. DOI: 10.1016/S0065-2687(08)60464-1 ↗
- Meneveau, C., & Katz, J. (2000). Scale-invariance and turbulence models for large-eddy simulation. Annual Review of Fluid Mechanics, 32, 1-32. DOI: 10.1146/annurev.fluid.32.1.1 ↗
How to cite this page
ScholarGate. (2026, June 3). Large Eddy Simulation. ScholarGate. https://scholargate.app/en/fluid-dynamics/large-eddy-simulation
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Detached Eddy SimulationFluid Dynamics↔ compare
- Direct Numerical SimulationFluid Dynamics↔ compare
- Lattice Boltzmann MethodFluid Dynamics↔ compare
- Reynolds-Averaged Navier-StokesFluid Dynamics↔ compare
- Volume of FluidFluid Dynamics↔ compare