[laurent_soucasse]

Hello,
I am trying to run the tutorial "Bump in a channel" but changing the geometry. I would like to change the smooth bump into a half circle.
Unfortunately, I always get this error at the first iteration when running SU2

CSysSolve::modGramSchmidt: w[i+1] = NaN
terminate called after throwing an instance of 'int'

I use gmsh for creating the geometry and the mesh, I enclose the gmsh file (replace .txt by .geo). The configuration file is the one of the first tutorial "Bump in a channel".

Thanks for your help.
Laurent.

[economon]

Originally Posted by laurent_soucasse

Hello,
I am trying to run the tutorial "Bump in a channel" but changing the geometry. I would like to change the smooth bump into a half circle.
Unfortunately, I always get this error at the first iteration when running SU2

CSysSolve::modGramSchmidt: w[i+1] = NaN
terminate called after throwing an instance of 'int'

I use gmsh for creating the geometry and the mesh, I enclose the gmsh file (replace .txt by .geo). The configuration file is the one of the first tutorial "Bump in a channel".

Thanks for your help.
Laurent.

Dear Laurent,

I believe that the parameters in the bump in a channel config file were optimized for max performance with that test case alone (i.e., the multigrid settings and the CFL number are fairly aggressive). First, I would recommend turning down the CFL a bit and turning multigrid off. Once you have it stable, you might try turning multigrid back on, while adjusting the number of levels for the damping to get performance.

Hope this helps,
Tom

[laurent_soucasse]

Thanks for your answer.
I have already tried to turn the multigrid off and to strongly decrease the CFL number but the I cannot run any iteration.

[economon]

Laurent,

Just to check that the grid is ok, can you please try setting all of the boundaries to far-field BCs and running a simulation? If the grid is ok, you should find that the residuals start around zero and that a uniform free-stream is maintained.

T

[laurent_soucasse]

Originally Posted by economon

Laurent,

Just to check that the grid is ok, can you please try setting all of the boundaries to far-field BCs and running a simulation? If the grid is ok, you should find that the residuals start around zero and that a uniform free-stream is maintained.

T

There is a problem with my grid as your "farfield" test case does not work, but maybe I need a gmsh expert to understand what is going wrong.