Running NS LES and NSSA with a NACA 0012 GMSH file.

[StepTang17]

To whom it may concern,

I am in the midst of trying to apply a simple 2D NACA 0012 simulation to match with NASA NACA0012 simulation as a validation test case before proceeding with my test case. Referring to Discussion #151, although the test NACA0012 provided in Solver/test/NavierStokes serves as a sanity check, I've referred to the that mesh to create a new one using GMSH. Pre-analysis shows for polynomial order of 2, this mesh should have the first node well within y+<1 which meant the mesh should be fine enough for wall resolved RANS applications.

I ran the case with NSSA (modified the Problem File for the UserDefinedInitialCondition for SpalartAllmaras) with Gauss instead of Gauss Lobatto. I generally encounter two issues:

1.) Iteration speed too slow: I understand that with finer mesh leads to higher computational cost, but I've tried various changes to the control file and it is too slow (Example: 4.28x10(-8) [simulation time/iteration] ). I believe I am missing something in my control file but can't seem to find the main cause.

2.) Numerical divergence: I still get case diverging despite trying various Re, CFL, and DCFL, settings. I haven't tried FAS settings but I don't think it will resolve the diverging issue. Just in case, I've also tried exporting the .geo file in .msh file format of 2.2 ASCII and 4.1 ASCII where both lead to the same results.

I apologize for trauma-dumping but I was stuck on this issue for awhile, please let me know if I am missing something either in my mesh or control file (I am pretty sure it is my stupidity is the cause of this issue)

Many thanks for any help and guidance
P.S: (I am a noob when it comes to Github and realized I can't upload my mesh and control files here) I have the mesh files (.geo & .msh) and control file said case located in the following link)

https://github.com/StepTang17/H3D-NACA-0012-Proto-C

[loganoz]

Hi @StepTang17,

Converging NSSA equations is always hard. The issues you are facing with explicit time integrators are expected. I think your best bet would be to try multigrid or/and implicit time integrators included in horses.
For details, you can check @wj-laskowski phd thesis https://oa.upm.es/72084/, as he implemented those schemes in horses.

Please take a look, try those schemes and let us know if that helps.

Best,
Gonzalo