[Saideep]

Hi;


I am trying to model a "2D - paddle" as in Fig 1 to use interfoam. Using snappy tool (OF-4x), I come across 2 issues and struggling to find a hack.


1. Using "snappy" with no layer refinement (as it is 2D) and adding layers results in capturing the accurate geometry of the .stl.



I used a 0 contact angle by placing the interface at the mid point of the paddle (along horizontal direction) and surprisingly the meniscus stays flat instead of spreading on the surface (Fig 2(initial), 3(after 1 milli sec)). I guess this is due to the prism cells on the boundary that are not perfectly oriented (they arent exactly a right angles triangle) and the cell faces are slightly inclined (Fig 4) causing issues while computing curvature (as curvature is divergence of interface normal).



2. Now, I try to add layers to have regular cells (rectangular) on the boundary. Now, the prism cells are pushed inwards and causing an issue when one fluid is pushing the other (as you can see there are multiple interfaces and the curvature is potentially completely wrong avoiding the spreading). (Fig 5)



In both the cases, the prism cells look to be an issue. Is there a way to avoid these or atleast merge them with neighbors.



I have tried several combinations of meshQuality controls but none seem to avoid or change the orientation of the prisms.



Any ideas ?!!
Thanks;
Sdp

[GerhardHolzinger]

There is no way in snappy to avoid the prisms with such a geometry, since they are a direct result of the way snappyHexMesh works

You could try cfMesh, this should give you an all-hex mesh

[simrego]

I think cfMesh will do nearly the same.
But for a simple geometry like this you can create a structured mesh with blockMesh pretty easily.


But what the hell is that diagonal stuff in pic 5? That's pretty weird...

[Saideep]

Thanks both!!


I have tried cfMesh earlier but instead of prism cells I have diamond cells and the behavior is similar to Fig 5 as attached in the earlier post. As you see in Fig5, adding layers pushes the prism cells into the domain and causes that sort of an interface (in simple words, the prism cells are trying to divide or layer the flow).

Ofcourse, this looks like a terribly simple geometry to model (and do doubt that sHM is doing its bit to capture the shape of stl) and solve but as you see it is not straight forward. I am not sure if you can mesh with blockmesh as you cannot capture the curves accurately any longer.

There could be several possibilities where it is going wrong here:
1. implementation of the contact angle bc in interfaceProperties.C

2. computation of interface curvature,
3. Prism cells.


I think the first two are not the case as they can be mathematically derived so only these few cells are problematic.



No clue how to deal with these.