A problem of numerical method in turbulent channel flow

[jushi-123]

Dear Incompact3d developers,

I have a question to consult you. The hyperviscous 6th scheme is developed for ILES, as mentioned in this paper "Xcompact3D: An open-source framework for solving turbulence problems on a Cartesian mesh[J]. SoftwareX, 2020, 12: 100550". However, the input file for DNS/explicit LES WALE model in the turbulent channel flow example folder all use the hyperviscous 6th scheme. What is the reason behind this choice? Is there any suggestion for the numerical method of DNS/explicit LES in turbulent channel flow?

I'm looking forward to your reply, thanks!

[mathrack]

Hello,

The provided input files are elaborated to illustrate the capabilities of the code. One can

• use only implicit LES with a selected level of numerical dissipation (isecondder = 5) => pure implicit LES
• use only explicit LES with a reduced level of numerical dissipation (isecondder = 4) => pure explicit LES
• use explicit LES with a selected level of numerical dissipation (isecondder = 5)

See https://hal.science/hal-04451824/document for a discussion on the accuracy of mixed Smagorinsky-implicit SVV.

[jushi-123]

Thanks sir. Is there any suggestion for the numerical dissipation of pure explicit LES? In this reference "Straightforward high-order numerical dissipation via the viscous term for direct and large eddy simulation(2011JCP)", the relationship kc''deltax^2=npi*pi is used to control the numerical dissipation in isecondder=4. How to choose n in Explicit LES?

[mathrack]

Hello,

If you use implicit LES (isecondder=5), you can select the level of dissipation using nu0nu and cnu.

If you use explicit LES, look inside les_models.f90 to check the relevant parameter associated with each model.

If you use only explicit LES, you probably use isecondder=4. Then there is no numerical dissipation and nu0nu and cnu will have no impact on the second derivative.

I have no recommendation for the level of dissipation in case of implicit LES. Check the references available. Run some DNS and some implicit LES. Plot the corresponding spectrum. Look at the modified wavenumbers for the second derivative.

Cheers,
Mathrack