[andre.weiner]

Hello FOAMers!

I try to connect inlet and outlet of my flow domain with a cyclic/cyclicAMI bc. (picture attached)

When I define the bc as follows, the patches point in the opposite direction.

    inlet
    {
        type cyclicAMI;
        neighbourPatch outlet;
        transform        rotational;
        rotationAxis     (0 0 1);
        rotationCentre   (11.25e-3 11.25e-3 0.5e-3); // center of domain
        //rotationAngle    180;
        faces
        (
            (0 10 12 2)
            (2 12 14 4)
        );
    }

A possible solution would be to transform two times (one as above and one to flip the direction), but when I define another transform type, the first one gets overwritten.
Another solution would be to extend the domain, what I want to avoid.

Maybe there is a very simple way to fix this and I just don't want to see it

Thanks in advance for any ideas!

Best regards!

[RodriguezFatz]

Hi, I don't really get the picture... Are inlet and outlet "around the corner"? Thus, everything that leaves on the lower horizontal line comes back through the top and everything that leaves on the lower vertical (right) line comes back trough the upper horizontal (left) line?

[andre.weiner]

I try to simulate flow through staggered tube bundle form the ERCOFTAC database. (Case C.78)

The main flow direction is x (as in the picture) with an average velocity of 1.06m/s.
The sides y-max and y-min are symmetry planes.
I could extend the flow domain in x direction (imagine the complete mesh mirrored on the outlet), then inlet and oulet could be connected with a cyclic bc.
The smallest possible flow domain, when the flow is assumed to be fully periodic and steady, is what you see in the picture. The oulet-velocity corresponds to the inlet-velocity mirrored on the x-z-plane through the center of the domain.

I hope this helps to understand the case better.
Thanks!

[RodriguezFatz]

Ok, looks like you would need to tell openfoam the "direction" for a regular cyclic patch (not cyclicAMI). I don't know how this can be done. Would it be sufficient to let the outlet "line" have a direction from bottom to top in you meshing tool?

But anyway: Why this crazy blocking? Don't you have distorted cells in the upper left and lower right corner? This doesn't look right...

Edit: It looks like there is a way to set the transformation tensor, so it must be something like
1 0 0
0 -1 0
0 0 1

[andre.weiner]

When using cyclic bc I always get the following error: (I tried many different transformations and changed the order of the vertices)

face 0 area does not match neighbour by 120% -- possible face ordering problem.
patch:inlet_1 my area:6.87521e-08 neighbour area:2.75008e-07 matching tolerance:0.0001
Mesh face:11028 fc:(0 0.0108844 0.0005)
Neighbour fc:(0.0225 0.0087875 0.0005)

cyclicAMI is not necessary as the blocking is conform. I was just playing around to see if there were any improvements.

For the mesh generation I used blockMesh. The cells you pointed out are not optimal but ok. Any idea how to devide this geometry in different blocks?
There is still the opportunity to switch to an unstructured mesh but it wont solve the bc problem.

Best regards

Andre

[RodriguezFatz]

You get this error when something doesn't match 100%. I make the grids with ICEM (ansys) and if I have one point at 1.210000 and the other one at 1.210004 I get these kinds of errors in OpenFoam.

Here you go:
mesh.jpg

[andre.weiner]

Many thanks for the hint to the blocking! It looks indeed much better

Regrading to the cyclic bc I ran blockMesh in debug mode. In the attached picture you can see how cell-faces are connected. The picture shows a mesh without grading. When I use grading, of course an error is thrown as the wrong cell-faces are coupled.

I couldn't figure out how to change this. Extending the flow domain is probably the easiest solution.

Thanks again for your help!