Differences with AVL for unconventional geometries

[Adithya Ramesh]

Hi Rob, Dave, and the VSPAero community

I have been running analyses to compare AVL, XFLR5, and VSPAero. The geometry is a little unconventional and the left and right wing sections do not meet (there is a disjoint in the profile). I created this disjoint by making the Yloc>0 in the XForm tab of the wing modeller.

Refer to images 01,02,03 for the geometries.

The results are compared and plotted in 04,05,06,07.

VSPAero predicts a much much higher lift coefficient than the other two.

I then ran a comparison analysis by re-setting Yloc=0.

Refer to images 08,09,10 for the geometries.

The results are compared and plotted in 11,12,13,14.   

I only changed the geometries and then ran VSPAero with the same settings (changed nothing).

Could you please shed some light on why there is this huge discrepancy and how (if possible) to correct it?

Version 3.21.2

Thanks and regards

Adithya Ramesh

[C P]

Hi,

have you tried to increase the panel density spanswise  in VSP ? From the picture it seems poorly discretized spannwise, while in AVL panneling is denser.

Corrado

[Rob McDonald]

As was mentioned - the spanwise paneling in the VSPAERO model is not adequate -- particularly for the separated wing case.

It would be good to show plots of lift distribution.  That way you can see how the different codes are handling the load distribution for the separated vs. connected wings.

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 on the web visit https://groups.google.com/d/msgid/openvsp/328515de-c02f-4047-b6eb-c2ca08dc3babn%40googlegroups.com.

[Adithya Ramesh]

So, I re-ran the VSPAero with increased number of tessellations as was suggested. I increased the sections to a maximum of 25. The wing was a lot more dense and the computation time increased. FYI my AVL model only had 22. There is still a considerable difference in VSPAero's predictions. I don't think increasing the tessellation further is the best approach as it just increases the computation time and will probably not significantly affect the results (based on the reduction of C_L between the cases having 6,15,25 sections).  Could this be a limitation of VSPAero?

I haven't developed my batch processing scripts yet for the load distribution. But, I will certainly investigate that as well.

Adi

[Tom]

Hi Ali, curious what airfoil are you using?

Thanks,

Tom

[Adithya Ramesh]

Hi Tom

It's a custom airfoil with a max thickness of around 19% and max camber of around 4%

Adi

[Adithya Ramesh]

Just a follow-up question

Why are there two 'CL', CDi,.. in the exported csv results file? These parameters are mentioned in the initial part of the csv file which shows its variation with each variation. The same parameters are repeated along with load distribution where only a single value is mentioned. Moreover, the latter values are lower than the ones mentioned in the part of the csv having the iterative solution.

[Rob McDonald]

VSPAERO can calculate forces in two ways.  One is by integrating the loads on each edge of the vortex rings of every panel.  The other is by integrating across the wake in a Trefftz plane like manner.  They give slightly different answers.

Integrating across the panels gets you full information -- load distributions spanwise and chordwise, moments, etc.  Integrating the wakes gives higher level information - just forces, but it can be more accurate for a given number of panels.

Rob

To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/256b7fe9-a82e-4a3e-8f6f-92353b75a662n%40googlegroups.com.