Tolerance of the mesh size factor
 

Hi SU2 users,
Hope you all are doing good. For few days I've been asking a lot of doubts here so please bare me a little.;)
I've another doubt to clarify today. I've been trying to run a steady state simulation in the transonic regime with gmsh as the mesh generation tool. I've a mesh size of 0.0008m near the wall and the simulation runs successfully until convergence is reached but as I reduce the mesh size to 0.0005m it gives an error of "Zero CV area" which I'm supposing because the mesh area < tolerance control volume. If I set any value 0.0008> mesh size > 0.0005 I get an error that the solution has diverged after 20-25 iterations. And I guess it is again due to excessive decrease in mesh size.
And the same happens when I run the same geometry in openfoam.
I'm wondering is there any way to decrease the tolerance more in SU2? Also, if someone can give me an insight of the utility of this tolerance it will of great help.
Thank you

Hi there,

I've seen your post on MAIN CFD FORUM.

I am afraid your problem has nothing to do with SU2 or in general your solver. Please note SU2 is a node based solver and it is double precision.

I am acquainted with Gmesh but I assume there should be something letting you set your node tolerance there. This is most probably what is happening for you. The default node tolerance in your mesh generator is greater than 0.0005 and less than or equal to 0.0008.

Then when you export your mesh and in your mesh there are nodes whose distance are less than the node tolerance, they are treated as same nodes in your exported mesh. That's why when you run it with a solver like SU2, solver doesn't see any difference between those nodes and gives you negative volume/area warning.

I have attached an SU2 example (i.e. naca6412) that has initial delta_s around 1e-06 in the mesh and has neither convergence nor running problems.


example ---> here


Good luck,
Pay