Version 3.48.2 : Static margin jumps around AOA = 3 deg in stability analysis

[Erik Fougner Arnesen]

Hi,

I am running a steady stability sweep in VSPAERO for a missile-like configuration at Mach 0.75, beta = 0 deg, with alpha swept from -4 to +6 deg.

The overall base coefficients look good:
*  CL vs alpha is smooth and nearly linear
*  Cm vs alpha decreases smoothly with alpha
*  CD vs alpha also behaves smoothly

However, around alpha = 2 to 3 deg, the stability derivative output becomes very noisy, even though the base coefficients stay smooth.

Example:
*  at 2 deg: CL_alpha = 5.87, Cm_alpha = +2.17, Cm_q = -17.5, SM = -0.37  
*  at 2.5 deg: CL_alpha = 6.46, Cm_alpha = -2.81, Cm_q = -114.6, SM = 0.434

* at 3 deg: CL_alpha = 9.06, Cm_alpha = -13.78, Cm_q = -329.2, SM = 1.521

*  at 3.5 deg: CL_alpha = 6.20, Cm_alpha = -1.31, Cm_q = -83.4, SM = 0.211  
*  at 4 deg: CL_alpha = 6.21, Cm_alpha = -1.37, Cm_q = -83.3, SM = 0.220  
Meanwhile the base coefficients in the same region remain smooth:     
* CL rises smoothly

*  CD rises smoothly

*  Cm transitions smoothly through zero

So this looks more like noisy derivative estimation than a real aerodynamic discontinuity.  

Has anyone else seen this in VSPAERO steady stability analysis? If yes, are there known causes or recommended fixes?  

I have added the .stab files from both the wider AOA sweep from -4 to 6 deg and more focused 2 to 4 deg test. Both showing instances at around 3 deg specifically. Also added the model if that might also be the cause

With kind regards

[Rob McDonald]

I would recommend more spanwise resolution -- particularly on the wing.  I would also recommend using less (or no) LE/TE clustering when you're using thin-surface wings.  The process of constructing the thin wing from a curved leading edge increases the effective clustering.

Have you tried the adjoint based static stability?  I would recommend you use that.

When you sweep alpha, the wing wake will cross over the horizontal tail at some condition.  When that happens, you will usually see some irregularity in the response.  If the mesh resolution is appropriate and the wake is converged, then this can represent real behavior.  If you run a sweep, you can click through the solutions in viewer to see what is happening with the solution.

Rob

[Erik Fougner Arnesen]

Thank you for your recommendation!

I switched to the following on the main wing:

* Spanwise resolution from Num U = 6 -> 9
* LE Clustering from 0.25 -> 0.00100

* TE Clustering from 0.25 -> 0.00100

* Switched over to Steady adjoint 

Trying this with AOA from 0 to 4 and looking at the results there are some irregularities at 4 degrees with extreme hot spots especially at the elevators and vertical rudders. 
Looking into the .adjoint.stab file available to see the results from AOA 4, but all derivatives have become -nan(ind) along with SM.

I then reduced the LE/TE Clustering on both vertical and horizontal fins to 0.00100 and performed the test again and observed the same issue, only here the large parts of the base coefficients jumped by an extreme factor example:
CMy from -2.4887778 to -5777.8609998

--
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/a6e74360-3e1e-4526-9e41-b41140269489n%40googlegroups.com.

[Brandon Litherland]

Apologies for any confusion.  When Rob said to reduce the LE/TE clustering, the intent was to lessen the grouping, not to lower the number. Here are some (somewhat outdated but still useful) slides from 2021 about modeling for VSPAERO. https://openvsp.org/wiki/lib/exe/fetch.php?media=workshop21:2021_vspws_-_modeling_best-practices_for_vspaero.pdf

No clustering means to have the value set to 1.0.  Try using Num W of around 33 to start (and may need to go higher) with Num U roughly 11.  Avoid very high aspect ratio (very thin/narrow) panels.

[Erik Fougner Arnesen]

Ohhh my apologies my bad, trying it out now!

To view this discussion visit https://groups.google.com/d/msgid/openvsp/c5893664-455b-4fd6-8432-a93e188d80bcn%40googlegroups.com.

[Rob McDonald]

Thanks for clarifying that Brandon.

The clustering parameters are normalized by what would be uniform spacing.

Setting clustering to 0.5 results in a first panel that is 0.5 that of uniform spacing.

So, in my confusing message, 'less' clustering meant to increase the clustering parameter towards 1.0.

Sorry about that.

Rob