A problem of ILES in Incompact3D

[jushi-123]

Dear Incompact3d developers,

I have a question to consult you. Based on my understanding of Incompact3D, coefficients of the second derivative can introduce numerical dissipation which is suitable for ILES. So what is the difference between the ILES and the coarse DNS(no model LES)? If the DNS equation with coarse grid is implemented using the high numerical dissipation, whether such method is ILES?

I'm looking forward to your reply, thanks!

[mathrack]

Hello,

In the context of implicit LES, there is no explicit LES model : dissipation at small scale is provided by the modified viscous operator. Thus, there is no fundamental difference between

• DNS (very fine mesh, negligible extra dissipation at small scale)
• quasi-DNS (fine mesh, limited extra dissipation at small scale)
• LES (medium to coarse mesh, extra dissipation at small scale)

However, one should carefully tune the viscous operator to provide an amount of dissipation at small scale adapted to the mesh used. See https://hal.science/hal-04452192v1/ for instance.

[jushi-123]

Thanks for your reply. So in the practical application, DNS/quasi-DNS/ILES are implemented with parameter "ilesmod=0" in input file and adjust the "nu0nu, cnu" in each case, while explicit LES is implemented with parameter "ilesmod=1"?

[mathrack]

To add an explicit LES model, you should set ilesmod to 1 yes. See https://github.com/xcompact3d/Incompact3d/blob/master/src/transeq.f90#L498 for instance.

For implicit LES, you should set the parameters ilesmod to 0 and isecondder to 5 and adjust the values of nu0nu and cnu. See https://github.com/xcompact3d/Incompact3d/blob/master/src/schemes.f90#L674