Comparison to actual flight testing results - DG-100 glider - Questions

[Luka]

Hi to all,

I started using OpenVSP about a month ago, for a project. Extremely satisfied with my

experience so far, very easy to use and quick for getting results.

After finishing the project model and performing some analysis, I needed to know what

are the tendencies of OpenVSP (optimistic, pesimistic, etc), in order to know how much I

can rely on OpenVSP results.

I watched most of the tutorials and have been checking out this group as well, for

additional information.

Anyhow, I decided to compare OpenVSP results to real-life data. I wanted to test a clean

configuration with available measured data for reference, so I decided to try it out on the

DG-101G glider, since I have a pilots operating manual where the measured flight polar

of the aircraft is present (measured by Akaflieg).

I managed to find quite detailed drawings (attached) and modeled the glider. Only two

pieces of information I needed were missing - the main wing's angle of incidence and

twist angle. Correct airfoils were modeled in.

I assumed twist angle to be zero (even though I know it has a fairly large twist angle) and

set the angle of incidence in such a way that the wings provide the required CL for the

best glide ratio  ~0.82 (@ wing 4°AOA). At a later point I modeled in a twist of -1.5° with

a negligible difference as well.

I then refined the mesh and performed the VLM analysis.

VLM gave a best glide ratio of ~26 @ 0°AOA, and the measured L/D for that glider at

those conditions is 38.3. That is more than a 30% difference and I'm wondering what

might be the cause.

I expected the difference to go in the other direction, for the VLM to show a higher

efficiency.

I assumed that the OpenVSP VSPAERO VLM module would be fairly accurate in

predicting CDi and CL and went on to check what is happening with CD0.

VLM calculated CD0 to be around 0.0194, and the parasite drag tool calculated it to be

around 0.0162.

I then calculated the allowed total drag, based on the known CL and L/D ratio of the

measured glider polar and subtracted the CDi acquired from VLM (which I assume to be

fairly accurate).

This gave me a CD0 of 0.009. This value could be achieved in the Parasite drag tool with

setting laminar flow percentage to be 65 for the wing, HT and VT and 30% for the

fuselage, with a Q of 1.05 for the wing and Q = 1.0 for all other components.

I've attached the 3-view document and my vsp3 model can be found here ->

http://hangar.openvsp.org/vspfiles/538

Since I did not introduce any twist to the main wings, I assumed L/D values would be

larger than real-life.

So, here are my questions:

1. Do you agree with my lines of thought described above? Is there something that I'm

missing or a way to improve my results and my understanding of them?

2. Why are there such large differences between VLM and real-life

3.  What part of VLM 'can be trusted', by your experience? Is there a general rule like 'CL

and CDi results are ok, disregard CD0'? Does it make sense to input the CD0 from the

parasite drag tool, as I have done?

4. How can this knowledge be transferred to another project, which is a very different

type of aircraft (tandem-wing with V-tail and a not-that-slender-body, lower aspect ratio

wings)? Can the same percentage of laminar flow be transferred to that project

(assuming the wing is in a clean air-stream and not at high angles of attack)?

I'm mainly interested in optimizing for the cruise condition, which means that high AOA

considerations are not an issue for me at this point.

5. Does anyone have a good tip for the utilization of the 'Q' parameter in the parasite

drag tool? Examples or something similar?

6. Is the fact that VLM under-estimated the drag for this case transferable to any other

case? Or, more exactly, does that mean that the VLM method generally gives pessimistic

results?

7. Do my assumptions in the parasite drag tool seem optimistic to you? Mainly the

Q = 1.05 and 65% laminar flow cases.

The flight condition was 105 km/h @ 1500m ALT ISA, or 1.5e6 Re.

My goal is to have a better understanding of where and how can OpenVSP be utilized

and with what kind of margin of error. To know where and by how much VLM differs

from real-word and in what area, in order to address those issues or just simply keep in

mind.

I fully understand that this is a tool for quick assessment of an idea and that it can not

substitute CFD, wind-tunnel tests and flight tests, so I guess what I'd like to have is a feel

for the limitations of OpenVSP.

Thanks for your feedback and kind regards,

Luka

[C P]

Hi,

in order to get some useful result, I have attached the file with some quick modification as follows:

some cleaning and refinement at the intersection of Vtail and fuselage, a better distribution of a reduced number of wing panels (spannwise and chordwise), used Panel method instead of VLM method , ” X-Z Symmetry”= on, increased wake iteration to 12,added CDo from Parasite Drag tool to CDi from Panel method results in order to have CDtot, divided CL by CDtot that gives CL/CD=33.2 about

I think it can be a first good improvement waiting for further refinement/coarsening, wake iteration changes, etc..etc.

…and don’t expect to replicate the reality J.

I hope it helps

Corrado

[Rob McDonald]

The CD0 model in the VLM is very simple.  It is based on empirical data for the NACA 0012.

So, it does not surprise me that the VLM CD0 estimate is high compared to what you would expect for a very clean sailplane with decent amounts of laminar flow.

There is a very simple stall model -- there has been some change here and I am not 100% sure where it stands right now (or maybe the changes are still in development).  In any case, the stall models are very primitive and might produce reasonable results -- but I would want to test and calibrate it before trusting it.

In the good news department, you're looking at an entirely subsonic aircraft with no sweep and no exotic high lift devices -- so it should be a best-case scenario for a simple stall model.

For fully attached flow, the VLM should do a great job of predicting lift, lift distribution, induced drag, moments, load distributions, control surface effectiveness, and S&C derivatives.

It won't predict details of wing-body interaction (or anything else dominated by the fuselage thickness).  For example, I would upgrade to the panel method to look at where to mount an antenna on the surface of the fuselage, or to get an idea of the velocity magnitudes around a wing-body join.

I don't think we output hinge moments -- but I wouldn't trust them if we did.  Hinge moments are one of the toughest things to predict for any tool.

Your model has an open nose -- you should make the fuselage water tight.  There are some strange artifacts of the skinning / lofting around the nose of the fuselage.  In general, less is more -- I would try to eliminate one or two of the XSecs in the fuselage to make it easier to model.

Overall, I think your model has too much resolution -- you should be able to de-res the wireframe and still get essentially the same answer out of VSPAERO.

I don't have great guidance for the Q parameter in the drag buildup.  My best advice would be to read the chapter on interference drag in Hoerner.  The parasite drag tool is no more sophisticated than the kinds of drag buildups engineers have been doing manually or in spreadsheets for decades.  Our goal with that tool was not to revolutionize understanding of physics -- but instead to simply make it easy to do what engineers have been doing all along.  Most companies have a set of best practices or adjustments that they have developed to match the data they've obtained in flight and tunnel testing of their aircraft.  If you are involved with an Akaflieg, I would expect they should have a similar experience base to draw from.

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/d200e752-fada-4f5c-875a-7b08d76aee75n%40googlegroups.com.

[Brandon Litherland]

The Q factors can also be found in Raymer under "Component Interference Factors" 

Copied from Raymer, "Aircraft Design", 2nd Ed. pp. 284-285: (excuse the OCR copy issues)

Parasite drag is increased due to the mutual interference between components.

For a nacelle or external store mounted directly on the fuselage or

wing, the interference factor Q is about 1.5. If the nacelle or store is

mounted less than about one diameter away, the Q factor is about 1. 3. If it

is mounted much beyond one diameter, the Q factor approaches 1.0. Wing

tip-mounted missiles have a Q factor of about 1.25. . .

For a high-wing, a mid-wing, or a well-filletted low wmg, the mterfere1:1ce

will be negligible so the Q factor will be about 1.0. An unfilletted low wmg

can have a Q factor from about 1.1-1.4. .

The fuselage has a negligible interference factor (Q = 1.0) m _most cases.

Also, Q = 1.0 for a boundary-layer diverter. For tail surfa~es, mterfer~nce

ranges from about three percent ( Q = 1.03) t:or a clean Y_-tatl to about eight

percent for an H-tail. For a conventional tail, four to five percent may be

assumed (Ref. 8).

[Luka]

Thanks to all for your helpful answers!

Regards,

Luka