[Neka]

Hello everybody!

I've observed a very strange behavior at the processor boundaries when using reactingFoam in parallel calculation.

I work with foam-extend- 4.0. My test case is a wedge and what we see in the attached image is the upper side of this wedge. The flow comes from the left. I use reactingFoam here and use the settings (fvSchemes, fvSolution etc.) from the reactingFoam tutorial. The gas is molecular nitrogen (N2) and the velocity is of 200 m/s. On the wedge, where the processor boundary is located, I have set a small quadrangular fraction of atomic nitrogen (N) of a mass fraction of 1.0e-04. So, I work here with small numbers. Then, I’ve run the simulation just for a couple of time steps (deltaT = 1.0e-07). I don’t use adjustTimeStep but a fixed time step. The N mass fraction started moving along the processor boundary – see in the attached image a strip on the left hand side (for making it visible I reduced the maximum of the color scale to 1.0e-16). I do not observe this behavior when running this test case in serial.

I discovered this problem because for my actual work I use a kind of a mix between reactingFoam and sonicFoam for calculating chemical kinetics at supersonic velocites.

Does anybody know what happens with my mass fractions?
Is it a bug?
Has anyone observed the same error?

Any suggestion or advice are welcome!

Regards,
Alex

[chicky]

Hi,

I have the same problem than you. Using mpirun -np 4 reactingFoam with laminar combustion model with the tutorial case SandiaDFlame.

I don't know why it is doing that. I will relaunch the calculation with only one processor to see if it's working.

[Neka]

Hello Nicky,
so, I'm not the only one who faces this problem.
Do you use foam-extend or OpenFOAM? I'm asking because it would be interesting to know whether this is a foam-extend specific problem.

[chicky]

I'm using OpenFoam 5.0 from openfoam.ORG

When i launched my parallel simulation the flame goes correctly ignited and then when steady state reached i see the same problem as you. My mesh is axis-symmetric so I wonder if my mesh goes wrong but checkMesh say that it is OK.

Tonight i just launched the calculation with one processor and the problem doesn't occur yet (the calculation is not converged totally).

[chicky]

[deleted by me, wrong information of myself. SORRY]

[chicky]

I have some news. If you check the pressure fields with paraview you will see that the fields is uniform and is set to 1e+05 Pa.

I think that's the reason of the bug. To change this you should allow pressure to change in specifying Pmin and Pmax into the PIMPLE option of fvSolution.

Refer to this topic I found on openFoam bugs website

https://bugs.openfoam.org/view.php?id=2566

That's change the result for me and seem's to be more convinient

[Neka]

Hello Nicky,

thank you for your reply.

I just have some questions:

I don’t understand how you solved the problem. You said you have the same problem as me. I have problem with mass fractions. My mass fractions move along the processor boundaries. As far as I understood from the link you posted, you solved this problem by bounding the pressure? Am I correct? And It really worked for you? I just don’t understand how the pressure field can influence my mass fraction fields. If I take a look at my pressure field it is not uniform. My pressure field in the boundary layer and on the surface is higher than in the free stream. As far as I know it should be like this. Can you give me your fvSolution settings for PIMPLE so I can try if it works for me please?

Regards,

Alex

[chicky]

my fvSolution is from the SandiaD_LTS tutoriel case. I just add Pmin and Pmax.

Perhaps our two cases are not similar. I have one cell where all the fields had a very high value but it is not at the junction between two processor. When I saw your image I was thinking that you had the same problem as me but that's apparently not the case because We aren't studying the same problem.

Sorry for the misunderstanding

[Neka]

No Problem.

Yes, what you have is something different.