Usage of .chkp and field0.f0000* files.

[Bkv-1915]

Hi creators,
I am new to Neko and I am working on turb_channel example. I want to know certain things:

1. Say I run simulation for 40 simulation time and get the .chkp file. How can i use this file as initial condition to run simulation from 40 - 80 simulation time and so on.
2. How to use these filed0.000* files or some other file (say .nek5000) to postprocess the results. Because, I am working on Ensemble Kalman filter and I need the values of the velocities. Where can I find these so that I can use this data on Kalman filter.
   Thank you for your help

[timofeymukha]

Hi!

1. Yeah, that is what checkpoints are for. You can use the restart_file keyword in the case file and point it to your checkpoint. Take a look at the case file documentation, that should be there.
2. You can open the .nek5000 file in Paraview or Visit. There is also a Python package called pymech, which can open them.

[MartinKarp]

Information on restarts should be available here: https://neko.cfd/docs/develop/dd/d33/case-file.html

By setting the "restart_file" keyword you can restart your simulation.

As for the field files, you can load them into paraview or visit, or postprocess directly in fortran or us for example pymech to process the files in python https://pymech.readthedocs.io/en/stable/usage.html

Please observe that by default neko only writes the mesh information in the first field file, so the most reliable way to read the fiels is with the readnek function.

[Bkv-1915]

@MartinKarp , thank you for your kind reply. Regarding the fortran part, is there a usage manual for the same?

[Bkv-1915]

I want to read and modify the variables - u,v,w, p and T in the checkpoint files, so that I can restart my simulation with a slightly modified initial conditions. I was able to read the filed files using fortran and I was able to understand it a bit. However, I am not able to understand the structure of the checkpoint files. I am not able to extract the velocity components, pressure and temperature. Is there a manual for the checkpoint files?

[scarybenjz]

@timofeymukha But the GS is being executed at each of those 1000 sims. If its once, then my prob would be solved. Currently I am having a do loop to execute the mpirun command 1000 times. Now I feel that its the wrong way of doing it.

[pschlatt1]

Then my suggestion is to modify the code such that you run all your simuations in one "run", i.e. by replacing the initial conditions whenever a previous simulation is done. How exactly to do that in Neko I would not know (in Nek5000 it would be quite straightforward by just changing the temporal order to 1 and replacing the velocity field). Then you do not need to restart anything, mesh etc. is all the same.

[timofeymukha]

I still don't get it, to be honest. Either the init is a significant part of the total time, and then the total time has to be short, so the optimization is not worth it. Or each simulation is not short, and then the init is negligible... I guess I am missing something. Anyway, I guess you can try to hack your way in ;-).

[scarybenjz]

@timofeymukha and @pschlatt1 , thank you for the quick replies. I appreciate it. I will now dive into Neko, hope I get this by this weekend. Fingers crossed.

[Bkv-1915]

Hi again. I was able to modify the code using your suggestions. Now the code is much faster as needed. I only made slight modification in the simulation.f90 file. i.e, I included another loop as shown in the screenshot to read field files and checkpoint files outside the time loop.

What I want: When I do mpirun, it should read the case file once, and perform the simulation from t_0 to t_end several times ( modifying the initial condition according to the field files and reseting the time and time-step to zero)

I wanted to ask if additional modifications are required (screenshot included).