Mesh deformation at boundaries
 

As long as I experience, SU2 does not allow to deform mesh in optimization when the boundary is not defined in FFD box. How can I handle this, since not deforming outlet boundary creates negative cells. I need to SU2 change node locations in outlet boundary.

Hi atelcikti1,



can you please post an example, images, or something else that further describe your problem?


Thanks, Tobi

Hi Tobi,

Geometry is like this. FFD box is defined on divergent part and only yellow part is allowed to be optimized or deformed. When yellow part starts to deform some bad cells are created on the intersection of yellow and red boundary. Since, red boundary node locations are not changing.

I see atelcikti1,


to give a bit of background. Mesh deformation is done in 2 steps:
1. the FFD-box-update morphs only the surfaces/boundary you prescribe (in DV_MARKER)
2. the volume morpher takes the morphed (and unmorphed) boundaries as mostly "clamped" boundary conditions for the linear eleasticity based mesh morpher.


Now I guess you only have "nozzle_wall" as DV_MARKER i.e. "nozzle_wall" is moved by the FFD-box but not "outlet". But both boundaries are taken as "clamped" in the next step of the volume morpher. That is where your bad cells originate from.

Putting "outlet" in DV_MARKER can help your cause but I think if only movement of the FFD-box parallel to the outlet is allowed. Otherwise your outlet boundary will become non-straight most likely.



If you morph your boundary manually using SU2_DEF it probably works if you put "outlet" in MARKER_SYM. Then you don't need to add "outlet" to DV_MARKER. MARKER_SYM boundaries are not treated as "clamped": the nodes can move but only on the straight surface which they are on i.e. no "out-of-plane" movement allowed.


Please test both options using SU2_DEF as starting an optimzation is probably overkill and let me know if it does what you need.

Best, Tobi

Hi Tobi,

Thanks for fast reply, I understand the su2 method to deform mesh.

I tried your first advice and add outlet to dv markers. Dont change the FFD box sizing and place Optimization loop is running now, but I also tried using SU2_DEF. Unfortunately, same issues continue. Outlet boundary nodes dont displace. So skew cells at clamped locations still exist.

I didnt understand the second suggestion, you mean changing the outlet boundary to symmetry will help. But i cannot solve with symmetric boundary in this problem.

Thanks again.

Hey,


you added 'outlet' (which could be a different name for you) to DV_MARKER -> did you also rerun FFD-setting such that this information is written to the FFD part of the mesh?
I think rerunning SU2_DEF with DV_KIND=FFD_SETTING (just as the initial setup) and then SU2_DEF with a sample FFD deformation should do the trick.


The second suggestion with MARKER_SYM: It makes no sense for a primal computation, but the volume mesh morpher treats these boundaries differently. So just running SU2_DEF could then give you a nice deformed mesh. If that works for you we can change a line of code such that the optimization works for you with the correct outlet boundary condition :)




cheers, tobi