Validation with experiments & other methods

[kaoanej]

Hi all ! so to explain we were trying to validate the VSM on experimental datas from Belloc's arched wing. Our approach is to run VSM with XFoil database, to extract the effective AoAs and then apply the XFoil distributions of velocities and Cps (interpolated at the right effective angle) on the geometry. Note that we use the uncorrected angles distribution. By doing so, we match the accuracy of our 3DPM method in terms of cd-cl polar (see green vs light blue curves).

We remark that the alpha-cd and apha-cl curves are not the same but still the cd-cl curves are matching perfectly, which is to us what matters at the end.

The point is that the polar we get inside our FSI code GENEPI is far from the raw one from the VSM (orange curve). As you can see, the raw drag from VSM is underestimated by a lot. We suspect an issue in the drag calculation of the VSM and so we were wondering if you guys had already noticed such behavior.

[jellepoland]

Thanks for bringing this up and sharing your initial results!

I have a few questions:

• What do you mean by VSM in GENEPI vs VSM outside GENEPI?
  Is "in" an FSI run with VSM as the aero model & "out" a static-aero run without a structural model (no FSI)?
• Would it be possible to share the input file you use to generate H. Belloc's case?
• Which version of VSM are you running?
• Why have you decided to use the uncorrected angle of attack? Because it gave better results?

Some intermediate answers: yes, I have experienced a lower drag prediction on the V3 kite when comparing to OpenFOAM RANS CFD, and when comparing to wind tunnel testing.

[kaoanej]

Hi @jellepoland , thanks for your response ! About your questions :

• VSM outside GENEPI is the direct polar given by the generate_3D_polar_data() method of the VSM. VSM inside GENEPI is also a static aero-run, but where I extract the induced angles (through uncorrected_alpha) from the VSM, to apply the corresponding XFoil distributions of Cps and Cfs to my wing aero mesh and recalculate the loadings by integration.
• Of course, I will provide you with the geometry and wind tunnel data asap, so that you can try it yourself. For GENEPI we use a very special homemade format and I think I should first try to convert it to a more classic one. I could also provide you with the VSM-ready files that include the wing sections LE & TE coordinates and the XFoil polar data for each section.
• I was using the old VSM version from @ocayon 's git folder but I updated to the new version this morning and the curves are sensibly the same.
• I use the uncorrected angle of attack for multiple reasons, but I must admit that I'm not sure about the correctness of my approach. What I understood from the VSM theory is that the efforts were computed at the 3/4 chord and then the local AoAs were recomputed to the 1/4 chord using the same circulation distribution to have the right orientation of lift and drag. However, what I try to extract here are the angles that gives the correct efforts, so for me I have to take the uncorrected alphas. When testing my approach, I noticed that the corrected angles had some artefacts on the tips which were problematic. When comparing the cl vs alpha distributions, I noticed that taking the uncorrected angles for my interpolation gave the same results as the raw VSM, while the corrected angles did not.

Thanks for your validation data ! As you said it seems that there also is a drag issue. What I see is that your case is close to my "raw VSM" case : it looks like the part of the drag that is not captured correctly is a part of the induced drag, as the drag gap with WT data increases with lift. I would be curious to see a cl-cd comparative plot based on your data.
In my case, the lift is overestimated. This leads to an overestimation of the induced drag with my GENEPI approach, but still yields a correct cl-cd polar. This is why I'm suspecting a problem in the drag calculation in the VSM despite computing correct angles. But it can also be that I get the correct cl-cd curve by chance, and that there is a deeper problem with the VSM theory, as it fails to estimate the right oswald factor.

[kaoanej]

Hi all, I could regroup the geometry and the wind tunnel data. I uploaded them to a git repository that you can find here. You will find the meshed geometry, a list of airfoils, a VSM-ready format with XFoil polars and the experimental wind tunnel data from H. Belloc.

[ocayon]

Hi all,

Thanks for looking into this, Tom.
Your analysis prompted me to revisit the assumption of using the 3/4-chord for the force magnitude and the 1/4-chord for the force direction in the aerodynamic model.

I was surprised to find that this assumption might not actually be the most accurate. The original reference for this approach is Li et al., 2022
. In their conclusions, they advocate for this correction, but the results shown in the paper don’t demonstrate a clear improvement in accuracy — in some cases, using only the 3/4-chord values gives slightly better agreement.

Motivated by that, I computed the V3 kite polars both with and without the AoA correction. Interestingly, when the correction is omitted, the drag prediction aligns much better with the wind-tunnel data, while the lift remains consistent.

As a bonus, the simulation runs slightly faster without the correction.

(See plot below — the uncorrected version matches the experimental curve much more closely.)

[jellepoland]

Thanks for uploading the data!

I am still unsure about the underlying physics, but based solely on comparing the V3 and Belloc paradliger case with their respective wind tunnel data measurements, one concludes that not using the correction is preferable.

I also added code to the latest VSM release v2.0.2 to generate this respective plot.