[Hassan Mandour]

Hello guys,
I'm simulation flow over NACA0012 equipped with vortex generators, constructed many 3D meshes on ICEM with the same cell distribution (and count) and various spanwise lengths, (so the quality metrics aren't varying much for those cases).

Solver: Pressure-Based, Coupled with Pseudo-transient
Model: k-w SST
Re = 1.5 M
B.Cs: Velocity Inlet, Pressure outlet, Periodic sides, walls for Airfoil surface & vortex generators.
Initialization: Hybrid then FMG

for some cases that went completely fine, however, for other cases I encounter the following message between Hybrid initialization iterations:
"Stabilizing ads-0 to enhance linear solver robustness."
and when starting the calculation I encounter another message:
"Stabilizing Pressure coupled to enhance linear solver robustness."
which causes convergence difficulties (may eventually converge tho, for lower angles of attack), what frustrates me is that this setup worked extremely well for some meshes!

I searched alot and can't find any reason for that!
I have attached pictures of the mesh and quality comparison between one of the nice cases (without those messages) and one of the problematic cases.
Thank you for your help ♥

Mesh:
https://ibb.co/98jn9kb
https://ibb.co/8mCq0NG
https://ibb.co/fG2LK1S
Mesh metrics comparison (the beauty and the beast :'D):
https://ibb.co/Y3ysTKM

[LuckyTran]

It's a warning, not necessarily an error. Also if it's only in the initialization, it won't affect the converged solutions afterwards.


The warning is pretty much self-contained. The solver algorithm didn't converge and it had to resort to a more robust algorithm. A kindergarden example would be: when Newton-Raphson fails to find a root, you switch to a binary search algorithm.


If you rely on an automated initialization procedure, eventually you'll run into a case where it isn't smart enough to do its thing.

[Hassan Mandour]

Thank you for your reply, sir.
the message about stabilizing pressure is displayed in the solution itself, indeed it's not affecting the converged solution but it's causing some convergence difficulties (especially in the early iterations), I can live with that, I'm just wondering why it worked fine for some cases and not another, and if there's anything I'm missing that may cause this problem.

Another thing _and excuse my ignorance_, I can't understand what you mean by:

"If you rely on an automated initialization procedure, eventually you'll run into a case where it isn't smart enough to do its thing."

Again thank you for your precious time.

[Far]

I made a detailed study on same issue and here is video: https://youtu.be/HPsBuqWH9FY

[Muhammad Waseem]

Originally Posted by Far

I made a detailed study on same issue and here is video: https://youtu.be/HPsBuqWH9FY

Really informative video, Thank you

[Muhammad Waseem]

Originally Posted by Far

I made a detailed study on same issue and here is video: https://youtu.be/HPsBuqWH9FY

Really informative video, Thank you. I was having the same problem " Stabilizing pressure coupled to enhance linear solver robustness" I changed the the number of processors from 10 to 1, then my problem solved 😁