Droplet Initialization Issue (Regarding Restart Calculation)

[Soobin120596]

Hello everyone,
I have used a random droplet distribution initialized by referring to the file initspraydata_3d.dat located in RegTests\Spray-EB. In the input file, I specify this initialization through the line particles.init_file = "initspraydata_3d.dat".
However, I have encountered an issue during simulation restarts: the code re-reads the initspraydata_3d.dat file instead of restoring the droplet data from the previous time step.
Could someone please advise me on how to resolve this problem so that the restart correctly resumes from the previously saved droplet state, rather than reinitializing?
Thank you in advance for your help.

[dmontgomeryNREL]

Some basic troubleshooting questions:

• Are you restarting from a chk file using amr.restart?
• Do you have particles.write_ascii_files = 1 in your input file?
• Have you tried setting amr.max_step to be one greater than the step you are restarting from to see what the state looks like immediately after restart?

[Soobin120596]

• I am restarting from a checkpoint file (amr.restart) and have enabled ASCII file output (particles.write_ascii_files = 1).

• Yes, I checked, but it seems the simulation is still reading from initspraydata_3d.dat instead of the droplet data from the previous timestep. This is evident because the number of droplets matches the count in initspraydata_3d.dat.

I came up with a somewhat crude but practical solution: I modified the function that outputs droplet ASCII files to simultaneously generate a .dat file in a format that Pele can read. This allows me to use the most up-to-date droplet data from the latest timestep when restarting the simulation.But I believe there must be a more elegant solution to this problem.

[dmontgomeryNREL]

Thanks for the additional information. You should be able to restart from a chk file as it contains particle data. However, if you still have particles.init_file = "initspraydata_3d.dat" in your input file, PelePhysics will restart using the init_file (see here).

You mentioned your case is based on RegTests/Spray-EB, which is from PeleC. Is the case you're running in LMeX or PeleC? I'm able to reproduce the issue in PeleC/Exec/RegTests/Spray-EB. When restarting from a chk file in PeleC, I see the restart_dir variable is empty despite restarting from a chk file. I know others have been able to restart from chk files in LMeX, so I'm wondering if this issue is isolated to PeleC.

[Soobin120596]

I’m running the case in PeleC, and I believe this issue is limited to PeleC.

[Soobin120596]

Thanks for your kind reply. I'm very sorry for not making it clear—this case is run in PeleC. But I believe the spray-related part should be the same between PeleC and PeleLM, right?

I have already tested PeleLMeX/Exec/RegTests/SprayTest and PeleC/Exec/RegTests/Spray-Conv. I think these two cases are quite similar, and both generate an initial droplet field with a uniform distribution. I changed the number of droplets to 2×2 for easier observation. And I continued the simulation from chk0020.

It seems that the droplet part in LMeX can be restarted from chk files. The output information of the droplets is as follows.

But when using PeleC, there seems to be some issues. It can be clearly seen that the droplet temperature has decreased, which indicates that the latest droplet data was not read from the chk file.

I'm very sorry that I'm not familiar with the code of spray, and I hope you can help answer my questions.
Spray-Conv (2).zip
SprayTest (2).zip

[dmontgomeryNREL]

You are correct that LMeX and PeleC use the same spray model from PelePhysics. However, there was a small bug in how PeleC initialized the droplets, specifically when restarting from a check file. I submitted PR #951 yesterday to resolve this issue. Thanks for bringing it to our attention.

[Soobin120596]

If you resolve this problem, please inform me. Thank you in advance!

[dmontgomeryNREL]

The fix has been merged into the development branch of PeleC. Make sure to comment out the 'init_file for the droplets in your input file when you are restarting from check file and everything should work as expected.

[baperry2]

I'm going to the PeleC github page because it's about that code. And it is being fixed by the PeleC PR @dmontgomeryNREL linked.