Ensure Matching mesh for MARKER_PERIODIC in mesh deformation

[TobiKattmann]

Hello you,

I want to deform a pin shape in a streamwise periodic setup:

Here I have 3 pin sections with periodicity in streamwise direction and symmetry on the upper and lower wall. So when mirrored and translated appropriately you will get a pin array and I want to ensure that there is only 1 updated Pin shape in the whole array, i.e. the front quarter pin has to have the geometry of the back half of the half pin in the middle (and the other way for the back quarter pin).

For now I put the same FFD box around all 3 pins like so:

which works fine in this specific case (I'll explain why later). For a working setup I need:

1. The partly-pins should have matching geometry deformation (as explained above)
2. Point 1 ensures that the periodic interface geometry is matching
3. For a periodic boundary also all mesh nodes have to match at the interface -> which is the pain point

The setup above seems to guarantee me a Point 1 and 2 in the case where I put the only the pin-surfaces in DV_MARKER and prescribe the 2 periodic markers as symmetry such that they can freely move in-plane (which they need to in order to not break the mesh). A simple workaround for this would be to fix this pin midplane such that that point is not allowed to move -> then the periodic surfaces can retain fixed and the mesh will work but the pin shapes are limited.

In the above case BSPLINE_UNFORM is used with 3rd order in x-direction where I need to fix the middle + the neighbors to ensure a fixed midplane. In case of order 2 that reduces to just the midplane.

With BEZIER I cannot really get this to work as further away FFD-deformations will also influence the position of the midpoint.

The problem I face is, that the Volume morpher's boundary conditions are allowing the periodic interface nodes to move freely in its plane (which is exactly what it should do) but the inlet and outlet nodes will not match anymore because the FFD-box is not symmetric (which I dont want to constrain). But then the Periodic Marker will not match and therefore fail:

So what can I do in order to enusure a matching mesh at the interface while allowing the pin midline to change. I will provide the case files to recreate the shown cases later .

Simplification: I am just interested in translational periodicty and just assume SU2_DEF is running in serial to avoid any of those complications.

[TobiKattmann]

Alright, as suggested in the Developers meeting I did the following:
Extend the FFD-Box to fully submerge the periodic markers and add those markers to the DV_MARKER list such that they are deformed by the FFD-algorithm and set as a clamped boundary for the volume mesher. This guarantees, that both markers undergo the same deformation.

A test deformation in the 5th row of the FFD-Box (only in y-direction!) showed nice results for FFD_BLENDING= BEZIER and BSPLINE_UNIFORM. The periodic interface stays intact in two aspects:

1. inlet and outlet points are perfectly matching (despite undergoing a shrinkage of the boundary length)
2. The in and outlet stay a straight line, i.e. the x-coordinates of the points are not altered

The periodic marker reports in a preprocessing step to the screen output whether there are matching interface points. So it is easy for any user to check validity of a periodic marker pair.

Test deformation applied the same way to all 3 pins:

Deformation of the interface (with a downstream copied domain) with BSPLINES (order 3,4):

The white mesh being the undeformed mesh

The same deformation but now using BEZIER (which also respects the interface's x-coordinate in this setup):

Two more notes

1. in order to retain good results I will not allow deformation of FFD-points in x-direction. With BSPLINE_UNIFORM that would theoretically be possible because one can dictate there sphere of influence by their degree. i.e. at order 2 I would need to fix the x-value of the FFD points in the center that run directly over the interface. With order 3 the center +1 column left and right (and so on)
2. I will keep the top two (or 3) rows of the FFD-box that do not cut the pin-surface fix. That ensures a smoother mesh. The absolute top row needs to stay fix in order to keep the symmetry boundary a single straight. Here I guess one needs to be careful again with the BSPLINE order in y-direction: So increasing that increases the points that are possibly moved and then 2 fixed rows provide a nice security against tearing the mesh apart. (Edit: it seems only to be necessary to fix the most outer FFD rows with BSPLINE in order to keeo the symmety boundary, even at higher Bspline orders in y-direction!)

[TobiKattmann]

Here the 9x9 lattice version that works nicely:
2b__Multiple_FFD_fluid_2.zip

And here the version from the initial post that did not produce a matching interface:
2__Multiple_FFD_fluid.zip

Thanks @pcarruscag and the others that provided this working version in the Dev-meeting discussion ☺️ 👍

In case there are any questions concerning a similar issue feel free to open another discussion and link me, or ask here