[kbziya]

Hi everyone,

I am setting up a 2D solver, flow around 3 bodies (2 prisms and one cylinder, stl's are made with FreeCAD). I want to run firstly a stationary case with 1rst order discretizations, SIMPLE algorithm, and then move to second order discretizations before starting the PISO algorithm to calculate the evolution in time.

When meshing with snappyHexMesh in parallel I get the following error just before the layer addition subroutine:

"face ... area does not match neighbour by ..."

Before moving on, I would like to mention that I have double checked the stl's, the CAD generation process, snappy parameters and so on. (I will attach snappyHexMeshDict, controlDict, decomposeParDict, surfaceFeatureExtractDict, blockMeshDict)

The process that I follow is:
1. surfaceFeatureExtract
2. blockMesh
3. decomposePar
4. mpirun -n 4 snappyHexMesh -overwrite -parallel > log.sHM
If snappy is succesfull
5. checkMesh
6. extrudeMesh
7. checkMesh (to see that I get only one cell in z-direction, a proper 2D mesh for OpenFOAM)
8. mpirun -n 4 simpleFoam -parallel > log.simpleFoam

Being only an issue related to decomposition in several processors, I am able to run in single core, snappy and CFD solvers. But, as you could understand, this is far from ideal. I have also tried to decompose after meshing in single core, but the obtained residuals clearly shows that something is going wrong.

Residuals - meshing single core, running simpleFoam single core.


Residuals - meshing single core, running simpleFoam in 4 cores.


Changing the stl's for snappyHexMesh "searchable boxes/cylinder" I have been able to finish snappyHexMesh in some occasions, but without been able to add the extendedFeature feature. That gives bad snapping in the prisms (boxes) that I want.

I have read all the info posted in this forum related to the aforementioned error (in this and this other) and I'm still a bit confused. I tried to relax the matchTol of all the processors (in their respectives boundary files) and snappy finishes correctly, but the residuals shows again that is not a good mesh.

So, if I cannot play with that parameter, where should I set a higher tolerance? I haven't got any problem before with the standard values for mergeTol in snappyHexMeshDict and writePrecission in controlDict. I increased these tolerances without any good result. Can anyone explain me how this tolerances are related? Am I missing some instructions for the utilities?

I'd like to attach some figures with the faces that don't match between processors.


Trying only with one of the bodies (cylinder, although I've tried with each one of them individually).


You can notice that this conflictive faces can be in regions where no prisms nor cylinder are near.

If anyone can shed light on this, I would be very gratefull.

Victor.