[amin.z]

Hi everybody,

I'm doing a mesh study for a simple flow using 4 different meshes - Course, Medium, Fine and Finest-
The results for three cases are rational, while for one of them (the Fine one) although velocity distribution is correct, the pressure distribution is wrong, as you can see in the attached plot (The gray one is the aforementioned case)
the geometry is the same, the solvers for all cases are thoroughly the same, transport and turbulence specifications as well, the boundary conditions for velocity and pressure are driven from mapping the fields of other meshes, so these two couldn't be the reason. I've also already checked the type of boundary conditions more than 10 times -4 walls and 2 cyclic periodic BCs-
Since everything is quite the same -of course except the grid- I'm confused what could result in such a strange results.
Any idea is highly appreciated,

[piu58]

It may be that the mesh gets too fine near the boundaries, so the wall functions works not in the way it should. You may check the yPlus values for that.

[amin.z]

Originally Posted by piu58

It may be that the mesh gets too fine near the boundaries, so the wall functions works not in the way it should. You may check the yPlus values for that.

Thanks for your response,
Actually, the flow is laminar and incompressible, so I think y+ doesn't play a role in this case, and the interesting part is that the Coarser and Finer grids both yield correct results (while for the finer one the boundary elements are closer to the wall comparing with this mesh)

[piu58]

Thank you. I missed your point: The FINE and not the FINEST mesh has a different behavior. I mixed that.

I recommend to copy one of the three nearly identical cases, say the medium mesh, change only the mesh to the fine one and let it run again. May be, there is something different in the fine case, which can not be seen at a glance.

[alexeym]

Hi all,

@amin.z

Though I was able to find, that the flow is incompressible and laminar, I was not able to find name of a solver you are using, could you disclose it?

What convergence criteria do you utilise? Basically, are you sure you have converged?

What linear solvers your are using? I.e. depending on your problem GAMG and PCG can give different results with all other parameters kept the same. Due to all these interpolations GAMG (at least OpenFOAM's implementation) introduces more errors in solution.

[amin.z]

Originally Posted by alexeym

Hi all,

@amin.z

Though I was able to find, that the flow is incompressible and laminar, I was not able to find name of a solver you are using, could you disclose it?

What convergence criteria do you utilise? Basically, are you sure you have converged?

What linear solvers your are using? I.e. depending on your problem GAMG and PCG can give different results with all other parameters kept the same. Due to all these interpolations GAMG (at least OpenFOAM's implementation) introduces more errors in solution.

Thanks for ur reply,
It's a simple flow and based on its specifications (as I already mentioned) I'm using steady-state solver simpleFoam. The convergence criteria are not ideal enough it has converged till 1e-05 so far, but I'm not sure if it could result in such a considerable difference in pressure.
I'm using GAMG and smoothSolver for pressure and velocity, respectively. Yeah, you're right, BUT the weird part is that the velocity distribution for all cases are almost the same and this so-called wrong result only occurs in pressure distribution. I don't know if it's possible that for the same geometry, solver and of course boundary conditions, the solver yield same velocity distribution with a difference pressure distribution.

It's now running according what Uwe proposed.

[alexeym]

Well, in case of pressure only gradient matters. In terms of pressure gradient all curves looks the same.

What if you change GAMG/smoothSolver pair to PCG/PBiCG? GAMG maps data between grids and MAYBE fine grid is a case, which produces something weird with you GAMG settings. What happens if you make grid 0.5*(MEDIUM + FINE) or 0.5*(FINE + FINEST)?

[amin.z]

Originally Posted by piu58

Thank you. I missed your point: The FINE and not the FINEST mesh has a different behavior. I mixed that.

I recommend to copy one of the three nearly identical cases, say the medium mesh, change only the mesh to the fine one and let it run again. May be, there is something different in the fine case, which can not be seen at a glance.

I tried what you suggested and the results were the same,
I'm not sure but it sounds there is something wrong with the mesh

[amin.z]

Originally Posted by alexeym

Well, in case of pressure only gradient matters. In terms of pressure gradient all curves looks the same.

What if you change GAMG/smoothSolver pair to PCG/PBiCG? GAMG maps data between grids and MAYBE fine grid is a case, which produces something weird with you GAMG settings. What happens if you make grid 0.5*(MEDIUM + FINE) or 0.5*(FINE + FINEST)?

Tnx for ur recommended modification but replacing GAMG with PCG didn't work too.

[AR91]

Hello,
I am a student at University of Dundee. I am also facing a similar issue. The pressure values from my coarse and fine mesh make sense but the intermediate mesh results in wrong pressure values. Have you found a solution to this problem?
Thanks,
Arun.