[arieljeds]

I am trying to set up a case using waveFoam to run in parallel on a cluster. First though I am trying to test it on my own machine as I've had some issues reconstructing the case for post-processing. I'm attaching the images from paraview following reconstruction:

reconstructedCase_bottom.jpg

reconstructedCase_inlet.jpg

reconstructedCase_side.jpg

reconstructedCase_top.jpg

As you can see, the case *almost* looks ok from the top (except for some issues at the walls) but then is completely wrong from all other views. My decomposeParDict file looks like:

numberOfSubdomains 2;

method          scotch;

simpleCoeffs
{
    n               ( 2 1 1 );
    delta           0.001;
}

hierarchicalCoeffs
{
    n               ( 1 1 1 );
    delta           0.001;
    order           xyz;
}

manualCoeffs
{
    dataFile        "cellDecomposition";
}

I will eventually decompose for 6 processors but I'm not understanding how to decompose properly. Does anyone have some advice on how to fix this?

Thank you!!


EDIT: Ok I just tried to see where the issue is and it is definitely in how I am decomposing the relaxation zones...... is there an example of how to properly decompose the relaxation zones somewhere?

[arieljeds]

Hi there, I'm going to answer my own questions since I seem to have figured out the problem...

I changed to scotch decomposition as I was not confident on how to properly decompose the geometry otherwise. This fixed the relaxation zone issue but then I realised that the problem had to do with setWaveField so I ran:

blockMesh
waveGaugesNProbes
setWaveParameters 
mpirun -np 2 setWaveField - parallel 
mpirun -np 2 waveFoam -parallel > log

And (so far) this seems to have fixed the issue

[ngj]

Hi Ariel,

Your observations makes very little sense in the context of how the relaxation zones are supposed to work (and based on my long experience). Therefore, can I ask you to prepare a case including an Allrun-script, where the problem can be reproduced. Then I will have a look.

Furthermore, which OF version are you using?

Kind regards

Niels

[arieljeds]

Hi Niels,

As far as I can tell, what I was doing before was decomposing it strangely along the y (and I also tried z) axis, which, when I then used relaxationZoneLayout, seemed to decompose the relaxation zone itself weirdly. When I changed this and left it as scotch decomposition (meaning that I didn't put how to decompose the geometry in myself), the relaxation zones appeared correctly (as I think you said in an earlier post, the relaxation zones are local and shouldn't be affected by decomposition, so I think that I was confused in which applications/utilities to run using mpirun as well as the actual way the geometry was decomposed).

I believe I have fixed the problem now by running setWaveField using mpirun -np 2. Before I did it this way, the wave field was not set in the separate processors.. I hope I haven't confused you more now!

Would you still like me to provide a case with the allrun script? I can check shortly if the current simulation did actually work as expected.

I am using OF version 2.4.0

Thank you for getting back to me,

Ariel

[ngj]

Yes, if the error is reproduceable, I would like to have the case.

Thanks,

Niels