[Jonathan]

hi,
I was wondering whether anyone could help me troubleshoot a turbomachinery test case I am trying to get working in SU2.

I get the following error when i run my case:

---------------------- Turbomachinery Preprocessing ---------------------

Initialize Turbo Vertex Structure.


Error in "virtual void CPhysicalGeometry::ComputeNSpan(CConfig*, short unsigned int, short unsigned int, bool)":
-------------------------------------------------------------------------
At the moment only turbomachinery with the same amount of span-wise section can be simulated
nSpan inflow 37, nSpan outflow 0
------------------------------ Error Exit -------------------------------

I've seen one or two previous posts and Github issues where this appeared to be as a result of orientation of the rotation axis as 'x' or 'y' and not 'z-axis'.

So i reorientated my grid (to the z-axis) but still get the same error. The case is a test case, so I can post the entire case, but the grid is 50MB.

The grid generator is a custom, but very simple H-grid generator which i wrote a custom converter to, to get it into .su2 format.

Perhaps some can try to run it, and check the setup? As i said its an open case, so happy to share the mesh / config, assuming i can upload the mesh here.

[joshkellyjak]

Hi Jonathan,


Firstly, you are correct in reorienting your grid around the z-axis, this is essential for turbomachinery. Secondly, ensure the flow path follows a positive z-direction. You also need to ensure the 'inflow' and 'outflow' boundaries have the same number of spanwise divisions. The error you show here indicates the code cannot find the start and end points between the two periodic boundaries at the outflow. This could be a mesh issue, make sure your grid is periodic and the boundaries are defined as you would expect. Can you upload an image of your domain with boundary labels, it is difficult to work out what is going on. I suspect the issue may be you are using multiple periodic boundaries in a single zone, this is probably why the turbovertex structure can't be built.

[Jonathan]

Hi Joshua,

Originally Posted by joshkellyjak

Hi Jonathan,


Firstly, you are correct in reorienting your grid around the z-axis, this is essential for turbomachinery. Secondly, ensure the flow path follows a positive z-direction. You also need to ensure the 'inflow' and 'outflow' boundaries have the same number of spanwise divisions. The error you show here indicates the code cannot find the start and end points between the two periodic boundaries at the outflow. This could be a mesh issue, make sure your grid is periodic and the boundaries are defined as you would expect. Can you upload an image of your domain with boundary labels, it is difficult to work out what is going on. I suspect the issue may be you are using multiple periodic boundaries in a single zone, this is probably why the turbovertex structure can't be built.

Thanks a lot for the response. Nice to make your acquaintance since I was just watching your talk for a conference on SU2 Coding (massflow BC). Very good and clearly presented. Much more tractable than OF tbh.

Attached is a screenshot of the mesh. Luckily flow is in the positive z-direction, but yeah, the periodics are split due to the way the mesh is exported from the grid generator. Hopefully that isnt a problem.

The mesh is generated as a circumferential sweep from a 2D surface in the so it should be 1-to-1 periodic. The grid in the axial direction is a pure sweep in that direction too, so the input / output cell counts in the radial direction should be the same, unless i messed something up in the mesh conversation tool. But from the paraview images, it looks fine ...

[joshkellyjak]

Ah nice to get a positive review my youtube tutorial career is beginning to take-off The mass flow BC is now in the master branch in a bit of a better form than that tutorial and I would reccomend using that version for your calculations.



The periodics may be the cause of the issue, it should be fairly simple to concatenate periodics directly in the .su2 mesh file. You can modify the MARKER_TAG directly to include the corresponding elements. If that doesn't work you should try modifying your mesh script to include them as one boundary. The turbovertex structure implementation currently needs a bit of a reformating but it is not a trivial task.

[Jonathan]

Ok, thanks a lot. I can definitely concatenate the periodics into a single marker ... I hadnt thought about putting them together, but I suppose numerically there potentially might not be an issue ... I dont know enough of how its coded up but that would be much easier than hacking at the mesh ...

thanks again a lot, i'll give it a go and report back as to how it went. hopefully thats the issue, and its not something else i've done wrong.

Cheers
Jon

Originally Posted by joshkellyjak

Ah nice to get a positive review my youtube tutorial career is beginning to take-off The mass flow BC is now in the master branch in a bit of a better form than that tutorial and I would reccomend using that version for your calculations.



The periodics may be the cause of the issue, it should be fairly simple to concatenate periodics directly in the .su2 mesh file. You can modify the MARKER_TAG directly to include the corresponding elements. If that doesn't work you should try modifying your mesh script to include them as one boundary. The turbovertex structure implementation currently needs a bit of a reformating but it is not a trivial task.

[joshkellyjak]

No worries, if that doesn't work update here or contact me on the Slack channel. I can show a simple way to hack the mesh that should work pretty easily

[Jonathan]

Thanks mate

Originally Posted by joshkellyjak

No worries, if that doesn't work update here or contact me on the Slack channel. I can show a simple way to hack the mesh that should work pretty easily

[Jonathan]

Just an update, so merging all the periodic patches into the same periodic + periodic_shadow, and reorienting things into the positive z-direction managed to get things working.

So in summary for anyone else who runs into this:
1) Rotation axis must be positive z-axis using right hand rule
2) Flow must be in positive z-direction
3) periodic faces cant be split i.e. you must have only x1 pair of periodic faces per blade row

Cheers
Jon

[bigfootedrockmidget]

Great to hear that things are working now.


@Josh: maybe we can add some warning and error messages for each of these points to guide users.

[Jonathan]

Originally Posted by bigfootedrockmidget

Great to hear that things are working now.


@Josh: maybe we can add some warning and error messages for each of these points to guide users.

An error message at runtime would be great. the error i got was quite cryptic and took quite a bit of hunting to start tracking things down.

[bigfootedrockmidget]

https://github.com/su2code/SU2/issues/2378

[joshkellyjak]

Agree on the error message, but also think a refactor of the whole turbovertext structure is probably more beneficial in the long run.

[bigfootedrockmidget]

I agree, Josh. My guess is that you already have a pretty good idea what needs to be done. If you can start a PR, that would be great!