[Carno]

Hi Experts,

I am solving a problem involving a simpeFoam solver with pressure jump in between the inlet and outlet. Inlet and outlet are atm. pressure BCs.
I am seeing very strange residual behavior of the solver. Right now it has not diverged, but the it is not going the good way. Earlier simple cases on the problem have diverged in a similar way?

Few more information:

The mesh is all tetra: 110 million cells
Run diverges immediately in decompose split (8 2 4) = 64 cores
Run marches good on 16 cores (1 1 16) - 16 cores- The flow is in the direction of Z.

Is there in any relation of convergence how we split the domain?

Please find attached the residual plot of the (1 1 16) split. The results at 300 iterations makes perfect sense. But the residuals are not correct, so I am not able to declare the convergence.

What is the way around? The current run also may diverge..

Best regards,
Carno

[kera]

Hallo,

Have you tried the simulation with hexa mesh instead of tetra?

I've had problems with the tetra meshes in the past.

Regards,
Ricky

Originally Posted by Carno

Hi Experts,

I am solving a problem involving a simpeFoam solver with pressure jump in between the inlet and outlet. Inlet and outlet are atm. pressure BCs.
I am seeing very strange residual behavior of the solver. Right now it has not diverged, but the it is not going the good way. Earlier simple cases on the problem have diverged in a similar way?

Few more information:

The mesh is all tetra: 110 million cells
Run diverges immediately in decompose split (8 2 4) = 64 cores
Run marches good on 16 cores (1 1 16) - 16 cores- The flow is in the direction of Z.

Is there in any relation of convergence how we split the domain?

Please find attached the residual plot of the (1 1 16) split. The results at 300 iterations makes perfect sense. But the residuals are not correct, so I am not able to declare the convergence.

What is the way around? The current run also may diverge..

Best regards,
Carno

[Carno]

Thanks for the reply.
I also suspect the same. Mesh issue. The geometry is very complicated, so it is very difficult or not feasible to mesh 100%hexa. I am trying atleast hexa dominant mesh. Let us see.

[kera]

I agree, that it is difficult to come up with a 100% hexa mesh sometimes. If the geometry is too complicated, you can use "snappyHexMesh" to get an hexa dominant mesh.

Regards,
Ricky

Originally Posted by Carno

Thanks for the reply.
I also suspect the same. Mesh issue. The geometry is very complicated, so it is very difficult or not feasible to mesh 100%hexa. I am trying atleast hexa dominant mesh. Let us see.

[Carno]

The problem is solved, when I used polyhedra mesh.
Snappy is a great tool, but it was also not helping as the castellation was not happening. We tried and spent lot of time to resolve the leak.
Then polyDualMesh script helped.

Thanks

[kera]

Hallo Carno,

good to know, that you were able to solve the problem.

Haha! the thing with snappyHexMesh is, it is a great tool if it works out for us, else you will end up banging your head against the wall :P.

For snappyHexMesh you need waterproof stl. I have struggled with it a lot as well. However there is a work around which works some times (before I involve myself in playing around with salome), try to create a very coarse mesh for your geometry using snappyHexMesh and extract the surface using surfaceMeshTriangulate and keep working iteratively, its time consuming but it works.

A more detailed explanation is given here (also a video tutorial is available, It was very helpful when I started learning things) from Tobi, I think you might have crossed paths with this link as well.

Regards,
Ricky

Originally Posted by Carno

Snappy is a great tool, but it was also not helping as the castellation was not happening.
Thanks