VSPAERO Issues with wing-body in 3.45.X

[Brian Davis]

I'm current running into some weird behavior with vspaero in 3.45. I have experienced issues in all the newer versions but testing now the most recent 3.45.2. 

Based on my testing, this seems to stem from intersections but not totally sure. Behavior is described below.

I have a relatively simple wing-body geometry, attached for reference. I have only tested in thin surface mode, as I need to ultimately run supersonic. All my testing has been done at Mach 0.3.

1. Running wing only, everything runs fine with both symmetry and full span.

2. Running with wing-body with default settings and full span, the solver NaNs and produces no output.

3. Running wing-body with default settings and XZ symmetry, it appears to run fine (maybe, I can't be sure). But that precludes me from using any of the other run types, like pitch stability for example to get neutral point.

4. Running wing-body with full span and cull orphans on (using its default settings), it runs but produces quite asymmetric results in the span loading with beta = 0.

5. Increasing the cull orphans parameters up from the default, say to 0.5, starts to improve the asymmetry but does not eliminate it.

My model is built as part of an automated upstream process (that writes a vspscript file) that runs fine in previous versions (e.g. 3.43). I've attached resulting vsp3 file.

Any ideas?

Brian

[Rob McDonald]

It looks like the new mesh stuff is struggling with the thin fuselage representation.  You might be able to play around with either mesh resolution and/or model scale to get better behavior.

Cull Orphans should not be needed on this case.  If everything was working right, this geometry would not have any orphans to cull.

When you do use Cull Orphans, it actually puts a hole in the fuselage -- which is clearly bad and is likely a source of the asymmetries, etc.

These are most easily seen by turning the NGonMeshGeom view mode to Hidden or Shaded after the geometry is processed for VSPAERO.  I often use 'Prepare Solution' to generate the mesh and see how it looks before I commit to running VSPAERO.

I can work to make this better, but in the meantime, I suggest you use a mixed thick-thin representation.

Put the Fuselage in Set_0

Put the Wing in Set_1

You can go into the Set Editor and name these sets 'Thick' and 'Thin' if you want.

Then, in the VSPAERO setup GUI, choose the appropriate thick and thin sets.

Rob

[Brian Davis]

Thanks for the quick response.

Has anything changed in the recent  vspaero updates such that thick bodies will work in supersonic flow? I’ve not had very good luck with it in previous releases.  

[Rob McDonald]

Not yet.

Dave has done some proof of concept work about 18 months ago that showed promise for thick body supersonics.  However, that was before the adjoint rewrite and all the current effort.

Hopefully things will settle down a bit and I suspect he will eventually be able to prioritize things like that, but on a years timescale (not months) and only a maybe basis....

Rob

--
You received this message because you are subscribed to the Google Groups "OpenVSP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openvsp+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/openvsp/bea6b3f8-af73-44a1-99d8-d7ad08d9c05dn%40googlegroups.com.

[ReynoldsNo]

Hi Rob, Brian, 

    I am currently using OpenVSP 3.46.0 and have experienced similar issues as Brian with modeling a Wing/Body/Belly Fairing, ALL in thin surface representation. 

    Perhaps this is associated with the centerline overlaps of various symmetric components?   

    I noticed that when I try to generate a simple Fuselage/BellyFairing model in thin surface mode, the overlapping portions of Fuselage and Belly fairing lying along the centerline

    are actually overlapping instead of having the clearly delineated portions associated with the fuselage and belly fairing.   Image is attached. 

     

[Rob McDonald]

My recommendation is to use the thick representation for the body and fairing, and thin for the wing.

The thin-thin intersection for perfectly overlapping surfaces is miraculous that it works at all.  You can post an example file and maybe I can try to do some deeper debugging with the example.  However, this is one of the reasons the thick / thin meshing and formulation is a thing.

Rob