I/O NO_WRITE, eliminating some output for very large lagrangian simulation
 

Hello, new here. Apologize if I’m not in the correct thread.

Little background:

Running very large simulation using primarily icoUncoupledKinematicParcelFoam.
I am post processing the data with my own program, and really only need the Lagrangian data output. I have eliminated the U, Rho, and Mu outputs by altering the createFields.H for myIcoUncoupledKinematicParcelFoam.

I need to also eliminate Phi output at the very least, and would also like to get rid of epsilon, k, and nut.

The problem is that in createFields.H phi is included with #include createPhi.H (which is in the shallowWater solver). And I haven't found yet where the output for epsilon, k, and nut is being handled.

I have tried editing createPhi.H in the source code (not in my user directory) but I am unable to compile (wmake) for some reason, and unsure what I would need to do to copy createPhi.H over to my user directory, and #include that createPhi.H in the createFields.H of myIcoUncoupledKinematicParcelFoam solver (instead of the one it is referencing currently, which I assume is the one in the source).

So help with createPhi.H would be much appreciated, as well as any help locating and altering the I/O for epsilon, k, and nut.

Thanks so much,

Steven

Well, no one seems to be able to help me, thank you those who have read this post though.

UPDATE:
I figured out from my solvers dependency file where exactly createPhi.H was for this case and edited that to get rid of Phi.

However, epsilon, k and nut files are still being output to every time step. With my number of timesteps, this takes up way too much HDD space and slows down the simulation by writing unnecessary info. It further slows down the simulation because I am required to write to a compressed format. If I could get rid of the nut, epsilon and k outputs at each timestep, this would bring my uncompressed simulation size (basically only outputting the Lagrangian, particle dynamics data) to like 6 GB.

The nut, epsilon and k are part of the incompressible turbulence model, but I cannot figure out how to set these fields to NO_WRITE. I tried editing kEpsilon.C for RAS which I thought my solver was using, but nowhere in the dependency file (for my solver, myIcoUncoupledKinematicParcelFoam) does it mention RAS or kEpsilon.

Thanks in advance, and again, if anyone has any input it would be much appreciated.

So you don't need the turbulence data at all, just initializing the simulation? What RASModel are you using in the first place. If you turn turbulence off this would prevent files from being written, but I don't think that's what you want.

Correct, I need turbulence included for the particle dynamics to be correct, so I can't turn it off. However, I do not need the turbulence data, I only need the particle data as I am post processing it with my own program to approximate triboelectric charging of the particles.

So ideally I want to edit OpenFOAM so that it simply doesn't output the turbulence data.

I'm using incompressible RAS kEpsilon.

From what I can tell, editing the kEpsilon.C file in the source (root) and recompiling seems to do nothing. Even after I set the fields to NO_WRITE for k,epsilon, and nut, and recompile both the turbulence model and my solver, without fail, it is still writing k,epsilon and nut (which are very large files, >1MB compressed, each) in every timestep.

The next approach I was considering is creating my own RAS model and using that instead, but I don't really see how that would do anything different.

Wow, I'm sorry, you are right. I don't need to turbulence anymore. icoUncoupledKinematicParcelFoam doesn't solve any fluid equations anyway. I can just set to laminar and not have to deal with the turbulence model at all. I'm an idiot haha. I needed the turbulence model for the first fraction of a second while it reached a steady state, before I even introduced the particles. All it was doing with icoUncoupledKinematicParcelFoam was copying the same k, epsilon, and nut fields from the initial timestep to all future timesteps.

Okay, problem solved, unless I find I need to use the coupled solver, because that will use the turbulence model. But, now that I think of it, the alterations I have already made to kEpsilon.C will likely take care of that. Sorry for the nonsense, and thanks for reading.

Hope this helps someone else in the future.

Steven