interTwoPhaseCentralFoam v2412 running in parallel

[rebshannon]

Hello,

I am trying to run a shock-droplet interaction using interTwoPhaseCentralFoam on OpenFOAM-v2412. I was able to run the blastToCyl tutorial in serial, but it fails with a sigFpe error related to the speedOfSound() function (sqrt specifically) when I run in parallel. This occurs just after time = 5.5e-06s, which is just as the pressure wave reaches the leading edge of the droplet. I've done some testing and found that if I split my domain in a way that the processor boundary doesn't split the water droplet it runs, but this significantly limits the parallel capabilities of the code. I also tried decreasing the CFL to 0.1, but I got the same error at approximately the same time.

I've included the case files, log file for the parallel run (that fails), a log file from a parallel run with additional print statements in the speedOfSound() function, and the truncated log file for the serial run (cut off after the time where the parallel run fails). The newly printed values in the verbose log file are the minimum values. On the last timestep, it shows the volume fraction going out of bounds, leading to a negative gammaM.

The only thing I've changed in the case file is whether it runs on 1 or 4 processors; everything else matches the tutorial case.

I've run tests on the damBreak and movingContact tutorials, and I was able to run them both in parallel. Do you have any advice on how to get blastToCyl-like cases to run in parallel?

Thank you!

case_and_log_files.zip

[mkraposhin]

Hello, @rebshannon . Thank you for the the report. This is certainly a bug and I hope we'll fix it together, especially because you've provided enough details.

However, I'd like to point your attention to the next important points:

1. I correct bugs only for the last version of the library, which is 2512 today, therefore, I hope you'll dedicate some efforts to reproduce the problem for this version.
2. interTwoPhaseCentralFoam is still experimental, i.e., whereas it has some attractive properties (if you compare L1 norms for Sod's problem among all solvers, then interTwoPhaseCentralFoam shows the least value), it still has inherent problems, and one is associated with the overdetermied system for pressure equation (oops, sorry). The working horses are vofTwoPhaseCentralFoam and Lubing Xu's twoPhaseMixingHLLCFoam.
3. My notebook is not very powerful, thus I expect that you prepare an archive (the case + the time directory) which is 1 - 2 time steps before the crash so that it doesn't take a lot of time to receive results.
4. Finally, there is a Telegram group https://t.me/hybridCentralSolvers where you can discuss the solver with me and other users directly.

[rebshannon]

Thank you for your time and feedback. It was very useful and has helped me to get some things moving on my project! I've responded to each of your points below:

1. I moved to the 2512 version of the library, ran on OpenFOAM-v2512, and didn't run into the error I previously described.
2. I ran a quick comparison between interTwoPhaseCentralFoam and vofTwoPhaseCentralFoam. Looking at the blastToCyl tutorial, I found that the vofTwoPhaseCentralFoam seemed to track droplet deformation better, so I have been moving forward with that solver. I haven't looked at twoPhaseMixingHLLCFoam yet, but plan to explore that after I get a handle on this one.
3. I have not included any case files because I am focusing more on vofTwoPhaseCentralFoam, and I found that switching to v2512 helped with my initial problem. However, if this is still something of interest to you, I am happy to provide case files from the test I mentioned above.
4. I am very interested in joining the Telegram group, but I was not able to join via the link. Is there another way to join or get added to it? Or may I message you directly on Telegram?

[mkraposhin]

For the 4, you can DM me (linkedin or gmail).