[Swift]

Dear OpenFoamers,


I have just run a 3D case with overInterDyMFoam in OF1912 and have observed that the mass is not conserved. Over time, the mass has seemed to drain out. There are no inlets and outlets of any form. I observe this in decreasing alpha.water over time. In this case alpha.water dropped from 0.37 to 0.08 in about 20s.


To test if it was a problem with my simulation set up, I tried to replicate the problem in the tutorial.



I have observed the same mass conservation problem with the twoSimpleRotors tutorial which is similar in principle to my 3D case. After 2 s solve time alpha.water has decreased by 6%. I have shown this in the figure named:'twoSimpleRotors.png'



The tutorial mesh is purely hexahedral, so the orthogonality is excellent.



Given that the tutorial settings are not necessarily optimised to give good results, I modified the tutorial fvSchemes, fvSolutions and controlDict for more nOuterIterations, lower tolerance and relTols, lower Co, and vanLeer descretisations. Please see the attached modified tutorial case attached and the results as twoSimpleRotors_modified.png.


I found this discussion regarding a potential bug in OpenFOAM 1706 with overInterDyMFoam, but was unable to find the mentioned bug on the tracker or if/how it was resolved. Does anyone know any more about this?


I have also seen that mass is not necessarily conserved with overset methods (This link with reference to starCCM). Is it possible that my observations are a result of the nature of overset?





So, now for my questions.

1) Is this reduction of mass inherent in overSet methods?
2) If not, how could I improve my simulation results? or,
3) is this a bug?

4) if it is inherent, are there means to reduce this. For example reducing cell size etc, different boundary conditions (totalPressure as an atmosphere (currently testing - will post results later))?
5) if it is inherent, how should the result be interpretted? is there still value to performing this kind of simulation?


If I can provide more details, please let me know.



Thanks,



Thomas

[mAlletto]

you can try this overset interpolation method:


cellVolumeWeight


I think it decomposed the cells in thets and calculates the intersecting volume of the cells involved in the interpolation. The weights of the interpolation are calculated accordingly.



this should be mass conservative. It is however quite slow.



For other method the interpolation error should reduce with difference of cell size of the cells which are involved in the interpolation and of course the size of the cells: So the small the cells and the smaller the difference between the cell size which are involved in the interpolation the better.


Other method are not mass conservative


Best


Michael

[Swift]

Good day Michael,

Thank you very much for your response!


The tutorial case actually already uses "cellVolumeWeight" for the oversetInterpolation in fvSchemes.


I will run a tests with some of the other methods to see if I get better results. I will do this with the base tutorial case. As we speak, I am running with leastSquares and the mass is not conserved.


The tutorial is set up such that the cell sizes are similar. The rotating domains are slightly smaller. For my 3D case where I first observed the non mass conservation, I had identical mesh sizes but they were not aligned. The cells of the rotating domain were rotating 30 degrees to the back ground me. This was done to reduce mesh orthogonality of the rotating parts.



I will try reduce the mesh size of the tutorial case as well as adjust the mesh so that the cell sizes are the same. It is possible the tutorial mesh resolution is too low as it is only to demonstrate overset functionality. It seems to me interFoam is very sensitive to mesh size and quality.

If you have any other recommendations, please let me know. I will post the results as soon as I have them gathered together.


Thanks again for your help,
Thomas

[mAlletto]

you can also try the overset implementation of foam-extend 4.1. I tried it ones (the single phase one) and it was faster and the mass was better conserved compared to the openfoam.com version.

[Swift]

Hi Michael,


Refining the mesh has not shown any improvement in the mass conservation. While refining, I also changed the overset mesh size to match exactly to the background mesh.


I may try using foam-extend.



Beyond this, I don't know what more I can try.


Thanks again for your help.


Thomas

[jfp6]

Thomas,

I can confirm that mass is not conserved for the cases I have run with overInterDyMFoam. I have also looked at different mesh densities and interpolation methods. None seem make much of a difference.

Have you done any more work on this?

Joseph

[Swift]

Hi Joseph,


No I have not looked at it further.



Please let me know if you find anything further. I still would want to use overInterDyMFoam for my application.


Thanks,
Thomas