[Adam_K]

I'm using OpenFOAM 5 with blueCFD to simulate the temperature response of a material initially at 300K subjected to left and right sides at 310 K when t>0s. I would like the geometry to translate infinitely in the vertical direction, so I am using cyclicAMI boundary conditions. This is an initial case that I would like to eventually extend towards more complicated ones.

I think that there is a problem with the way I've used cyclicAMI boundary conditions. In using probes to follow the temperature at the top and bottom of the simulation domain, I'm seeing differences that I didn't expect as I thought the surfaces were linked.

This is my geometry, if need be my full casefile and mesh are on DropBox.



In the changeDictionaryDict file, I've described the relationship between the top and bottom patches.

  
    matrixTop
    {
        type            cyclicAMI;
        neighbourPatch  matrixBottom;
        transform       translational;
        separationVector (0 -0.0002 0);
        matchTolerance  0.0001;
    }
    matrixBottom
    {
        type            cyclicAMI;
        neighbourPatch  matrixTop;
        transform       translational;
        separationVector (0 0.0002 0);
        matchTolerance  0.0001;
    }

The simulation works, and if I look at the central probe, it follows the analytical solution.

Here is a plot of the temperature difference between the probe located at a given y-position and the probe at the bottom (y = 0). I'm not surprised that there is some difference between the probes in the interior, but I expected the top and bottom probes to match much more closely. At the centre (x = 100 µm) they are very close, but at X = 50 µm, there is a considerable difference between the top and bottom of the domain, up to 0.3 K.


Have I messed up my treatment of the cyclicAMI? Or am I missing something here?

Thanks in advance!
Adam

[shach934]

Hi
Have you found a solution to your problem? Is the cyclicAMI interface ok with heat transfer or it needs special setting?
Thanks!

[Adam_K]

Originally Posted by shach934

Hi
Have you found a solution to your problem? Is the cyclicAMI interface ok with heat transfer or it needs special setting?
Thanks!

No, I never really found an explanation for this.

[shach934]

Ok....Then at last are you still using cyclicAMI and how do you treate it?

[ykanani]

Originally Posted by Adam_K

I'm using OpenFOAM 5 with blueCFD to simulate the temperature response of a material initially at 300K subjected to left and right sides at 310 K when t>0s. I would like the geometry to translate infinitely in the vertical direction, so I am using cyclicAMI boundary conditions. This is an initial case that I would like to eventually extend towards more complicated ones.

I think that there is a problem with the way I've used cyclicAMI boundary conditions. In using probes to follow the temperature at the top and bottom of the simulation domain, I'm seeing differences that I didn't expect as I thought the surfaces were linked.

This is my geometry, if need be my full casefile and mesh are on DropBox.



In the changeDictionaryDict file, I've described the relationship between the top and bottom patches.

  
    matrixTop
    {
        type            cyclicAMI;
        neighbourPatch  matrixBottom;
        transform       translational;
        separationVector (0 -0.0002 0);
        matchTolerance  0.0001;
    }
    matrixBottom
    {
        type            cyclicAMI;
        neighbourPatch  matrixTop;
        transform       translational;
        separationVector (0 0.0002 0);
        matchTolerance  0.0001;
    }

The simulation works, and if I look at the central probe, it follows the analytical solution.

Here is a plot of the temperature difference between the probe located at a given y-position and the probe at the bottom (y = 0). I'm not surprised that there is some difference between the probes in the interior, but I expected the top and bottom probes to match much more closely. At the centre (x = 100 µm) they are very close, but at X = 50 µm, there is a considerable difference between the top and bottom of the domain, up to 0.3 K.


Have I messed up my treatment of the cyclicAMI? Or am I missing something here?

Thanks in advance!
Adam

I am not sure how the BlueCFD works in regards to the probing the data, However, for the cyclic patches you should expect the patch values exactly math and not the cell center values near the patch. Please note that, the patch values for cyclic cases are not written in saved data files. Paraview, for example, assumes zerogradient if the patch values does not exist (e.g. cyclic). I suspect that the way that data is probed is causing this mismatch rather than a problem in CyclicAMI itself.
You may need to directly probe the boundary values using boundaryCloud (see OpenFOAM-5.x/etc/caseDicts/postProcessing/probes/). Otherwise, I believe it uses the cell center data in someway to return the value.

Hope this helps,
Regards

[Adam_K]

Originally Posted by ykanani

I am not sure how the BlueCFD works in regards to the probing the data, However, for the cyclic patches you should expect the patch values exactly math and not the cell center values near the patch. Please note that, the patch values for cyclic cases are not written in saved data files. Paraview, for example, assumes zerogradient if the patch values does not exist (e.g. cyclic). I suspect that the way that data is probed is causing this mismatch rather than a problem in CyclicAMI itself.
You may need to directly probe the boundary values using boundaryCloud (see OpenFOAM-5.x/etc/caseDicts/postProcessing/probes/). Otherwise, I believe it uses the cell center data in someway to return the value.

Hope this helps,
Regards

This makes sense and is along the lines with what I was thinking (but not super confident about).

In the end, the actual probe values at those points isn't very important to me and I've realized that using cyclicAMI vs zeroGradient BCs doesn't impact my final simulation results (temperature profile at the center of the slab).