CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Meshing & Mesh Conversion (http://www.cfd-online.com/Forums/openfoam-meshing/)
-   -   [fluent3DMeshToFoam] High aspect ratio cells found (mesh from ICEM) (http://www.cfd-online.com/Forums/openfoam-meshing/102189-fluent3dmeshtofoam-high-aspect-ratio-cells-found-mesh-icem.html)

alquimista May 24, 2012 10:34

[fluent3DMeshToFoam] High aspect ratio cells found (mesh from ICEM)
 
Hello,

I'm converting a mesh with the fluent3DMeshToFoam tool. Everything seems to be ok but checkMesh gives me:

***High aspect ratio cells found, Max aspect ratio: 4053.52, number of cells 1 Threshold = 1000

I tried to repair it in ICEM CFD but I don't obtain errors checking the mesh with that software.

How can I repair it in ICEM or just see if there's some bad elements after conversion?

or
How can I repair it in OpenFOAM?

lth May 24, 2012 14:51

Perhaps increase the number of cells in ICEM and see if that reduces your aspect ratio issue. Also, in ICEM, there is an option to check aspect ratios in the Blocking / premesh-Q tab. Scroll down to aspect ratio under criterion and click on bars that are highest in the results window. This will highlight the elements in question.

Best, Lori

gonpe May 25, 2012 09:07

Check out the refineMesh and refineHexMesh commands.

Also, the mesh checking algorithms used in Icem and Openfoam have different thresholds. Aspect ratio is a quality issue, not so much an error in my mind.

A worse problem would be a negative volume or wrong oriented face error.

alquimista May 25, 2012 11:10

Quote:

Originally Posted by lth (Post 362971)
Perhaps increase the number of cells in ICEM and see if that reduces your aspect ratio issue. Also, in ICEM, there is an option to check aspect ratios in the Blocking / premesh-Q tab. Scroll down to aspect ratio under criterion and click on bars that are highest in the results window. This will highlight the elements in question.

Best, Lori

Thanks Lori. My mesh is unstructured and I already checked the quality mesh in Edit Mesh -> Check Quality, I can see the distribution of aspect ratios but ICEM don't show bad values. I'll try to refine the mesh and check it, but I'm doing a model with a complex geometry and near 20 millions of elements. When I mesh something with blocks I don't obtain errors with checkMesh but with other methods I do. Although normally I get convergence.


Quote:

Originally Posted by gonpe (Post 363109)
Check out the refineMesh and refineHexMesh commands.

Also, the mesh checking algorithms used in Icem and Openfoam have different thresholds. Aspect ratio is a quality issue, not so much an error in my mind.

A worse problem would be a negative volume or wrong oriented face error.

Thans gonpe. I'm trying to delete that errors related with the quality because I get divergence problems running simpleFoam, pisoFoam or even potentialFoam. My BC are ok (I checked it in a simplified model), but when I add some componentes the mesh quality descreases a bit, but I run without problems and good convergence in ANSYS CFX even using LES turbulence model.

Checking in OpenFOAM the solution beforte divergence starts (time step continuity errors increase and maximum vallues increasing fastly) the field data have reasonable values but just fail in some cells. I work-around playing with fvSchemes adding some limiters in divergence, gradient, laplacians.... and fvSolution.

ANSYS CFX has some secrets tricks to get convergence in "bad meshes", but I have problems to control it in OpenFOAM or check when mesh is not good enough before simulate it.

gonpe May 25, 2012 11:19

Hi alquimista

Is the high aspect ratio cell in a critical location (high gradients)? Then I could see some issues develop.

You could cellSet those cells and refine them to test that theory.

I don't know what your solver settings are but adding nOrthogonalCorrector (pressure) steps sometimes helps.

On CFX's secret convergence tricks ... don't let them trick you ;)

Goncalo

lth May 25, 2012 11:26

I agree that nOrthogonalCorrector can sometimes help in high gradient areas.

Best, Lori

alquimista May 25, 2012 11:35

Quote:

Originally Posted by gonpe (Post 363130)
Hi alquimista
Hi alquimista

Is the high aspect ratio cell in a critical location (high gradients)? Then I could see some issues develop.

You could cellSet those cells and refine them to test that theory.

I don't know what your solver settings are but adding nOrthogonalCorrector (pressure) steps sometimes helps.

On CFX's secret convergence tricks ... don't let them trick you ;)

Goncalo


Yes Goncalo, the high aspect ratios are just in a gradient area (some kind of wings) and cilindrical shapes (after some simplifications)

So... I'm going to follow the way to cellSet bad values and refineMesh that set in order to increase the quality, is this possible?

I tried with nOrthogonalCorrector but nothing happens.

ANSYS CFX has a lot of disadvantages, but the control of the convergence is a great thing, maybe state scret, without it they would be forced to go out of business :D

alquimista May 25, 2012 11:38

Quote:

Originally Posted by lth (Post 363132)
I agree that nOrthogonalCorrector can sometimes help in high gradient areas.

Best, Lori

I tried more than 5 correctors in potentialFoam! Do you think that it's enough?

lth May 25, 2012 11:47

yes, in the past for me 4 correctors has converged for high aspect ratios with highly stressed materials.

kalyangoparaju May 27, 2012 01:48

Guys,

I've been facing the same problem for the past couple of days. I have a very decent mesh built in ICEM which runs flawlessly in Fluent but doesn't even come close to converging in OpenFOAM.

I've tried the following.

1. Decreasing the under-relaxation factors.
2. Playing with the pressure laplacian settings
3. Increasing nNonOrthogonalCorrectors.

But to no avail. The time step continuity just shoots up after a few iterations and the solution fails. One thing which I have to say is that, my mesh has a max non-orthogonality of around 88 and a max aspect ratio of 27000 according to checkMesh.

I am simulating a case where I require a mesh of y+=1 and hence I don't think I can afford to have lower aspect ratios without drastically increasing the number of cells. Is there any other solution to this problem other than refining the mesh ?

Thanks,
Kalyan

Industrial_CFD January 29, 2013 15:28

Hi Kalyan,

I am having similar issues. I made a nice hex mesh in ICEM, used fluent3dtofoam to convert it to openfoam, and it blows up. If I use snappyHexMesh, the solution is stable. I wonder if something is going on. Did you solve your problem?


All times are GMT -4. The time now is 10:21.