Abstract
We propose the successive inverse polynomial interpolation method to optimize model parameters in subgrid parameterization for large-eddy simulation. This approach is illustrated for the Smagorinsky eddy-viscosity model used in homogeneous decaying turbulence. The optimal Smagorinsky parameter is resolution dependent and provides minimal total error in the resolved kinetic energy. It is approximated by starting with a “bracketing interval��? that is obtained from separate “no-model��? and “dynamic eddy-viscosity��? large-eddy simulations. The total error level is reduced 3–6 times compared to the maximal initial errors. The computational overhead of the full optimization at resolution N3 is comparable to a single simulation at $(3N/2)^3$ grid cells. The increased accuracy is higher than obtained with dynamic modeling at a resolution of $(4N)^3$.
Original language | Undefined |
---|---|
Article number | 10.1063/1.2391840 |
Pages (from-to) | 118102 |
Number of pages | 4 |
Journal | Physics of fluids |
Volume | 18 |
Issue number | 11 |
DOIs | |
Publication status | Published - Nov 2006 |
Keywords
- Interpolation
- Turbulence
- Viscosity
- Flow simulation
- METIS-238743
- IR-63903
- EWI-9015