[abonfi]

Dear developers,
I have been using SU2 to perform a temporal and spatial grid convergence study. I use an exact solution of the Euler eqns consisting in an isentropic vortex convected by a supersonic stream (see e.g. http://dx.doi.org/10.1016/j.jcp.2004.11.023).
The results (in terms of observed order of convergence) are not as expected, but of course I cannot rule out having made some mistake in the configuration file or elsewhere.
Should you be willing to have a look into this, this pdf
https://dl.dropboxusercontent.com/u/60787983/vortex.pdf
explains the testcase and results in detail. Fig. 3.26 and tab. 3.28 apply to the SU2 calculation.

Best Regards,
Aldo

[fpalacios]

Originally Posted by abonfi

Dear developers,
I have been using SU2 to perform a temporal and spatial grid convergence study. I use an exact solution of the Euler eqns consisting in an isentropic vortex convected by a supersonic stream (see e.g. http://dx.doi.org/10.1016/j.jcp.2004.11.023).
The results (in terms of observed order of convergence) are not as expected, but of course I cannot rule out having made some mistake in the configuration file or elsewhere.
Should you be willing to have a look into this, this pdf
https://dl.dropboxusercontent.com/u/60787983/vortex.pdf
explains the testcase and results in detail. Fig. 3.26 and tab. 3.28 apply to the SU2 calculation.

Best Regards,
Aldo

Dear Aldo,

Definitively, there must be something wrong in the input file or the grid in Su2 format.

This seems an interesting problem, if you send us the config file and mesh, it will be a pleasure to take a look at the simulation to be sure that the Roe scheme with Venka's limiter (this is important) is working as expected.

Best Regards,
Francisco

[abonfi]

The grids and initial solutions are accessible through this link
https://docs.google.com/file/d/0BxNU...FqRV93bW8/edit
Once untarred, look at the README file.

Regards,
Aldo

[fpalacios]

Originally Posted by abonfi

The grids and initial solutions are accessible through this link
https://docs.google.com/file/d/0BxNU...FqRV93bW8/edit
Once untarred, look at the README file.

Regards,
Aldo

Thanks a lot, it is a very well done job.

I have run the 40x20 grid, and I have noticed that both grids are composed by triangles. Would it be possible to run a structured grid (without dividing the rectangles)?

Cheers,
Francisco

[abonfi]

I have uploaded here
https://drive.google.com/file/d/0BxNUUuEe1bPmM3A2b0xJU3MxMHM/edit?usp=sharing
a tarfile with the triangular AND quadrilateral grids and initial solutions.
I found out a tiny error in the gas constant that had some influence on the initial solutions for the triangular meshes;
these have been updated.
The calculation of the discretization error was tailored on triangular meshes (using quadrature rules),
so it is not available for quads, though it could be easily done on a pointwise basis.

Aldo

[economon]

Hi Aldo (& Francisco),

I agree! The document is very nicely put together; thanks for sharing it, Aldo.

I would like to add that we should verify the unsteady restart capability as well. It is possible that the initial solutions are not being loaded correctly for the two previous time levels required with the BDF. This is a relatively new feature in the code and could use more testing.

Tom

[abonfi]

After fixing the bug in the initial solution, things have improved: the observed order looks ok up to grid level 3, but on grid level 4 the discretization error has not reduced as it should, see Tabs. 3.28 and 3.29.
The density profile still shows an overshoot upstream of the vortex location, see Fig. 3.26.
The update document is here:
https://dl.dropboxusercontent.com/u/60787983/vortex.pdf
I have also added results obtained using the NSC2KE code from INRIA.

Regards,
Aldo