[klaus]

Dear OpenFoam friends,

can somebody tell me if i can use the boundary type "cyclic" for 3D
axi-symmetic cases. I know that there is the "wedge" type, but this is only
for 2D. Please see the scetch for clarification. Or is there some other way?



Greetings,
Klaus

[henry]

Yes you can and should use "cyclic" for this case.

[mattos]

Dear WonderFoam friends!

Please, I need some help.
1) Case:
I'm testing a very simple axi-symmetric teste case. It's a sector of a cylinder with 50 mm of radius and 300 mm long. The angle is 15 degree because the converter crashes announcing very small angle between the faces when I define a 5 deg sector. And the mesh have only one division in the circumferencial direction, as explained in the UG man. The front and back planes are imposed as imposed as wedge type BC and all fluid properties (U T p etc) connected with these patches have been their BC type assigned as "wedge".

The velocity at inlet has their value assined as "(1 0 0.5)" bacause U would like to test if Foam is able to capture the swirl effect. In the Paraview visualisation, the calculations do not present the velocity vectors with circumferential components (They are zero).

Q1: Why can't I see the circumferential component velocity?

Q2: Should I do something more to have the circumferential velocity?

However, I just checked my lastTime U file and the Z component velocity is prsent and amost equal to the inlet values. Is it related with paraFoam limitations?

Q3: What the minimum angle allowed to the edge in order to have success in the checkmesh and mesh converter? (I read the previous discussion in the board.)

Q4: This is a stupid question and out of scope of this topic, but I need! How can us to attach a file or picture in this message board?

Many Tanks in advance for your help!

Wladimyr

[mattos]

Dear WonderFoam friends!

Please, I need some help.
1) Case:
I'm testing a very simple axi-symmetric teste case. It's a sector of a cylinder with 50 mm of radius and 300 mm long. The angle is 15 degree because the converter crashes announcing very small angle between the faces when I define a 5 deg sector. And the mesh have only one division in the circumferencial direction, as explained in the UG man. The front and back planes are imposed as imposed as wedge type BC and all fluid properties (U T p etc) connected with these patches have been their BC type assigned as "wedge".

The velocity at inlet has their value assined as "(1 0 0.5)" bacause U would like to test if Foam is able to capture the swirl effect. In the Paraview visualisation, the calculations do not present the velocity vectors with circumferential components (They are zero).

Q1: Why can't I see the circumferential component velocity?

Q2: Should I do something more to have the circumferential velocity?

However, I just checked my lastTime U file and the Z component velocity is prsent and amost equal to the inlet values. Is it related with paraFoam limitations?

Q3: What the minimum angle allowed to the edge in order to have success in the checkmesh and mesh converter? (I read the previous discussion in the board.)

Q4: This is a stupid question and out of scope of this topic, but I need! How can us to attach a file or picture in this message board?

Many Tanks in advance for your help!

Wladimyr

[steja]

I have an axisymmetric geometry, i.e. a blade passage where I want to use cyclics for the BC of the rotational periodicity. My problem is, that the faceList of both cyclic patches is not numbered in the way OF needs it.

So, I wonder if there is a simple possibility to perform the mapping of both non-planar patches which are defined by one rotation about a coordinate axis and thus define the renumbering with the 1-to-1 mapping. I know that couplePatches is not able to handle this problem because the patches are not planar.

1)
Could you point me to some effective coding in order to get the correct faceList by mapping the one patch to the other with some tolerance and a predefined rotation.
The transformation is not the problem, but what mesh manipulation methods are available to handle the numbering of the faceList?

2) Is it enough to reorder the faceList or are there further connections I have to think about?

Thanks for your answers/help

[steja]

Looking into the code of couplePatches & cyclicPolyPatch I found that maybe for none-planar rotational cyclics instead of using n0 and n1 to define the rotation tensor it might be a solution to simply define the rotation tensor manually because it is known for such configurations.

Thus it is not necessary to have planar cyclics. Can someone comment on this.

[rmagnan]

I have been having problems with some Foam meshes imported through CGNS. These meshes have (rotational) cyclic patches and their faceList are defined in the proper order (face i matches face n/2+i). However, I found that some Foam classes (see globalPoints and cyclicPolyPatch) also requires that the first vertex of each face matches the first vertex of its corresponding face. When this requirement is not met, checkMesh does not complain, almost everything in Foam works fine but I get bad results from paraFoam and volPointInterpolation.

I would like to know if this vertex ordering is actually a requirement on Foam meshes or should that be taken care of "on the fly" within Foam when loading a mesh.

Thanks for your help.

[mattijs]

It is a requirement for anything 'to do with points' (e.g. volPointInterpolation, mesh motion).

Take 1.4.1 'couplePatches', replace line 107 with meshChanged=true. (since operator= of face does not detect rotation of the vertices)

See if running that couplePatches helps.

[rmagnan]

Great! Yes, this solves my problem.
However, isn't this sort of thing something which should be done on the fly when reading a mesh with cyclics? Especially, if the mechanics to do it is already in there.

Many thanks

[mattijs]

> which should be done ..

Pro: it does not change the mesh. No addressing changes apart from the start of the face.
Against: it can fail. It uses a geometric tolerance on point locations. Also it is reasonably time consuming and would need to be done every time the mesh gets loaded.