Direct Numerical Simulation
Also known as: DNS, resolved turbulence simulation
Direct Numerical Simulation (DNS) is a computational approach that solves the Navier-Stokes equations without turbulence models, resolving all scales of motion from the largest energy-containing eddies down to the smallest dissipative scales (Kolmogorov microscales). Pioneered by Steven Orszag in 1971, DNS provides complete information about turbulent flow fields and serves as a reference solution for validating turbulence models. However, extreme computational demands limit DNS to relatively simple geometries and low to moderate Reynolds numbers.
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
DNS should be used only when complete turbulence information is essential and computational resources permit, typically in academic research and fundamental studies. Use DNS for validation studies, to understand turbulence mechanisms, or when turbulence models are inadequate for new flow regimes. DNS is impractical for engineering design, product optimization, or complex geometries. Restrict DNS to low-to-moderate Reynolds numbers, simple domains (channels, pipes, periodic boxes), and problems where the added insight justifies weeks of supercomputer time.
Strengths & limitations
- Requires no turbulence model; solves the exact Navier-Stokes equations
- Provides complete statistical information including spectra, correlations, and intermittency
- Ideal for validating and improving turbulence models for other methods
- Reveals fundamental mechanisms and eddy structures not accessible otherwise
- Eliminates modeling uncertainty from results; errors are purely numerical
- Computational cost scales poorly with Reynolds number, limiting applicability to Re ≤ 10,000 in practical scenarios
- Requires very fine spatial grids: Kolmogorov scale must be resolved, demanding millions of cells even for simple geometries
- Very restrictive boundary conditions: periodic domains or simple walls; complex industrial geometries are infeasible
- Time integration must be very fine-grained; total simulation time can be days to weeks on supercomputers
- Storage of instantaneous 3D velocity fields requires substantial data infrastructure
Frequently asked
Why is DNS so expensive compared to RANS or LES?
DNS must resolve the Kolmogorov microscale, the smallest energy-dissipating scale. For a channel at Re=10,000, this scale is thousands of times smaller than the channel height. You need millions of grid points in 3D, and the time step must be proportionally small to resolve viscous dynamics. A RANS case uses 100,000 cells in minutes; DNS uses 100 million cells and runs for days.
What is the relationship between Reynolds number and DNS feasibility?
Grid requirements scale roughly as Re^(9/4) for 3D isotropic turbulence, making DNS quadratically more expensive with Reynolds number. A DNS at Re=5,000 costs 10-100 times more than at Re=1,000. Current supercomputers can handle Re~40,000 for channel flow, but only the simplest geometries. Industrial problems at Re>100,000 remain far beyond reach.
What boundary conditions are typically used in DNS?
Periodic boundaries (for decaying turbulence or channel flow) are most common because they avoid complex boundary layer handling. Channel and pipe DNS use periodic streamwise and spanwise directions with no-slip walls in the wall-normal direction. Open domains and inflow-outflow boundaries are extremely challenging and rarely used in DNS due to the difficulty of maintaining turbulence without artificial forcing.
How do I extract turbulence statistics from DNS?
Compute ensemble averages (over time or space, depending on stationary assumptions) of velocities and their products. Second-order statistics (Reynolds stress tensor, kinetic energy) require 10-50 integral timescales of data. Spectra and higher-order moments need proportionally more. Most DNS codes include online averaging routines to avoid storing entire 3D fields.
Can DNS be used for practical engineering applications?
Not directly. DNS provides reference data for validation and understanding, not design solutions. Engineers use DNS results to develop and validate RANS and LES models, which then provide fast solutions for actual design problems. A DNS study might take months; a RANS analysis of the same geometry takes hours. The investment in DNS pays off indirectly through better predictive tools.
Sources
- Orszag, S. A. (1971). Numerical simulation of incompressible flows within simple boundaries: accuracy. Journal of Fluid Mechanics, 49(1), 75-112. DOI: 10.1017/S0022112071001940 ↗
- Moin, P., & Mahesh, K. (1998). Direct numerical simulation: a tool in turbulence research. Annual Review of Fluid Mechanics, 30, 539-578. DOI: 10.1146/annurev.fluid.30.1.539 ↗
- Kim, J., Moin, P., & Moser, R. (1987). Turbulence statistics in fully developed channel flow at low Reynolds number. Journal of Fluid Mechanics, 177, 133-166. DOI: 10.1017/S0022112087000892 ↗
How to cite this page
ScholarGate. (2026, June 3). Direct Numerical Simulation. ScholarGate. https://scholargate.app/en/fluid-dynamics/direct-numerical-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.
- Boundary Layer TheoryFluid Dynamics↔ compare
- Large Eddy SimulationFluid Dynamics↔ compare
- Lattice Boltzmann MethodFluid Dynamics↔ compare
- Reynolds-Averaged Navier-StokesFluid Dynamics↔ compare
- Smoothed Particle HydrodynamicsFluid Dynamics↔ compare