CFD Online Discussion Forums - Sliding interface in parallel

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)

- OpenFOAM Running, Solving & CFD (https://www.cfd-online.com/Forums/openfoam-solving/)

- - Sliding interface in parallel (https://www.cfd-online.com/Forums/openfoam-solving/105569-sliding-interface-parallel.html)

Sliding interface in parallel

Hello,
I am trying to run the mixer2D tutorial in OF1.6-ext (with sliding interface, not GGI) in parallel.
After the first time step I get this error:

Quote:

[1] Cannot find point in pts1 matching point 15 coord: 0.287492 0.00143987 0.025) in pts0 when using tolerance 2.79913e-06
[1] Searching started from:5 in pts1
[1] Compared coord: 0.262486 0.00274884 0.025) with difference to point 0.0250411
[1] Compared coord: -0.287492 -0.00143987 0.025) with difference to point 0.574992
[1] Compared coord: 0.287484 0.00301064 0.025) with difference to point 0.00157079
[1] Cannot find point in pts1 matching point 14 coord:-0.287484 -0.00301064 0.025) in pts0 when using tolerance 2.79508e-06
[1] Searching started from:5 in pts1
[1] Compared coord:0.262486 0.00274884 0.025) with difference to point 0.55
[1] Compared coord:-0.287492 -0.00143987 0.025) with difference to point 0.00157079
[1] Compared coord:0.287484 0.00301064 0.025) with difference to point 0.575
[1] processorPolyPatch:Order : Writing faces to OBJ file procBoundary1to0_faces.obj
[1] processorPolyPatch::Order : Dumping newly found match as lines between corresponding face centres to OBJ file "/home/ivan/OpenFOAM/ivan-1.6-ext/run/tutorials/incompressible/icoDyMFoam/mixer2D/processor1/procBoundary1to0_faceCentresConnections.obj"
[1]
[1]
[1] --> FOAM FATAL ERROR:
[1] in patchProcBoundary1to0 : Cannot match vectors to faces on both sides of patch
masterCtrs[0]:-0.262486 -0.00274884 0.025)
ctrs[0]:-0.262486 -0.00274884 0.025)
Please check your topology changes or maybe you have multiple separated (from cyclics) processor patches
Continuing with incorrect face ordering from now on!

If I use only 1 processor, it runs perfectly. I get the error only in parallel.
Does anyone know where is the problem?
As far as I know the Sliding Interface modifier is already fully parallelized. Or am I wrong?

Thanks a lot
Ivan

Hello Ivan,

I am also facing the similar problem, did you find any solution for it ?

Regards
Sayali

Dear Sayali,
no, no solution so far :-(
Ivan

Same issue in foam-extend 4.0

Hi All,
I am facing this issue with newer version of foam-extend. Please could anybody kindly advice the best practice for running sliding interface in parallel

Many thanks

Michael

Quote:

Originally Posted by michvip (Post 694176)

Hi All,
I am facing this issue with newer version of foam-extend. Please could anybody kindly advice the best practice for running sliding interface in parallel

Many thanks

Michael

With ggi at least you need to create global cell zones with the sliding patches. Also, these patches have to live in the same address space, i.e. in the same processor.

Quote:

Originally Posted by Santiago (Post 694383)

With ggi at least you need to create global cell zones with the sliding patches. Also, these patches have to live in the same address space, i.e. in the same processor.

Hi Santiago

Thanks for the tip. I have managed to do the above successfully through the decomposeParDict file. However on running using mpirun I get a segmentation fault error after
"Selecting dynamicFvMesh mixerFvMesh" line in the log file

Not sure if there is something obvious I am missing, and also how to analyse the error log/trace

Thanks

Michael

Quote:

Originally Posted by michvip (Post 694176)

Hi All,
I am facing this issue with newer version of foam-extend. Please could anybody kindly advice the best practice for running sliding interface in parallel

Many thanks

Michael

Hi Michael,

I am working on sliding interface case.
I was able to run it using two processors only (simple decomposition).

Have you solved the problem?

Hi Saleh
I don't use foam-extend much anymore, but from looking at this old case I seem to have it run ok using 4 processors

Hi Michael,

Is it possible to share the dynamicMeshDict ?

In my case, only the simple algorithm works for 2 partitions only.
Each partition contains one pair of the sliding without crossing the sliding interface and without using any constrains.

I've tried almost every possibility without success.

Regards,

Quote:

Originally Posted by Saleh Abuhanieh (Post 728483)

Hi Michael,

Is it possible to share the dynamicMeshDict ?

Regards,

File is attached.

Quote:

Originally Posted by michvip (Post 728711)

File is attached.

Thank you Michael,

Can I ask the decomposeParDict as well :rolleyes: ?

Regards,