SU2 Nonequilibrium solver residual stalling
susmitsj opened this issue · 7 comments
Hello,
I have been using SU2 NEMO to run cases for steady, inviscid, hypersonic flow around blunt objects. Previously I have been able to successfully run cases using SU2 version 7.1.1. However, I have been running into residual stalling with the newer versions (7.4.1 and 7.5.1). Recently, I tried two different meshes which worked for me before (TestCases/euler/bluntbody/blunt_91.su2 and a RAM C-II mesh I previously used). I also tried updating the confguration file to match selections in TestCases/nonequilibrium/invwedge_ausm.cfg but I still face this issue. I have attached the relevant files here. Please let me know if this is a solver issue or if I need to change any selections in the configuration file.
residual_stalling.zip
Hello, can you send the .cfg and .su2 files of a case that previously worked well and now it doesn't? I can check it. Thanks
Hi Catarina. Thanks for your help. Since there's a attachment file type limitation, I am attaching the files in a zip folder here. In the folder, the configuration files marked as '_old' are the ones that worked well before. 'invbb' refers to the blunt body case and 'invram' refers to the RAM C-II case. The mesh files are the same as before (blunt body is in .su2 format and ram c-ii in .cgns).
Hello,
I have been debugging the differences between the two versions. I found 5 differences that lead to slightly different results. Two of them have been verified to be an improvement of physical modelling. The other 3 I am still trying to figure out. Could you please run the same tests with MUSCL_FLOW=NO and see if you're still facing this issue?
I see that the MUSCL reconstruction is one of the 3 things that have been modified and not sure if could be the cause of the problem or not. I will keep working on understanding the impact of the other 2 things.
When you do this - could you please save the simulation output log of both versions and send those back to me too (in this way I'll have a better feeling of what is happening to the residuals throughout the simulation)? We can just focus on the invbb case to make it simpler.
Sorry to piggyback on this, I hope it is okay. I think I have the same issue...
I have been trying to get a hyper-sonic flow around the cylinder to converge but unfortunately to no avail. I assumed my setup was incorrect so I tried to look for an already tested case to run. I tried the case with the files uploaded in #1625 but that seems to not converge for me either. I have attached the residual plot
I am willing to run a few cases on my workstation to help debug this. Kindly let me know which files to use and with which settings toggled/changed.
Please let me know if you need any more information from my side.
Hi @aditya12398
That would be helpful as it has been a bit difficult to identify the problem between versions of the code that have a lot of differences.
As mentioned in my prev message, one of these differences is the MUSCL implementation and I am leaving this out for now. If you could ran the case attached in the following manner, would be useful:
- sending cfg files invbb_old_711.cfg and invbb_old_751.cfg (for versions 7.1.1 and 7.5.1 respectively) + mesh file blunt_91.su2
- could you run 2 simulations, one with 7.1.1 another with 7.5.1 with those cfg files pls?
MUSCL is disabled, so if we obtained very different results between those 2 simulations we have a good hint that the problem is not MUSCL. and we can then decide what to do next.
Please confirm if you can do this and send me updates once you have them - thanks! I keep working on this in different fronts, more hands-on debug.