Negative Cdi and Higher CL at wingtip in VSPAERO

[Gordon Yuhim Lai]

Hi,

I found that VSPAERO gives negative Cdi whenever I include the rotors in my model. Yet the results are completely normal when rotors are not included in the model. 

Also I noticed that the CL at wingtips on both side are significantly higher, which is quite abnormal.

What could be possibly wrong?

Attached is the vsp3 file of the model. And I was running on 3.10.0

Thanks.

Settings are as follows:

Sref = 45.765000 

Cref = 1.645333 

Bref = 20.783150 

X_cg = 7.794000 

Y_cg = 0.000000 

Z_cg = 0.914000 

Mach = 0.000000 

AoA = 0.000000 

Beta = 0.000000 

Vinf = 100.000000 

Rho = 0.002377 

ReCref = 10000000.000000 

ClMax = -1.000000 

MaxTurningAngle = -1.000000 

Symmetry = NO 

FarDist = -1.000000 

NumWakeNodes = -1 

WakeIters = 3 

NumberOfRotors = 2 

PropElement_1

1

4.000000 4.000000 1.400000 

1.000000 -0.000000 -0.000000 

1.500000 

0.2500000 

2000.000000 

0.400000 

0.600000 

PropElement_2

2

4.000000 -4.000000 1.400000 

1.000000 -0.000000 -0.000000 

1.500000 

0.250000 

-2000.000000 

0.400000 

0.600000 

NumberOfControlGroups = 0

[Rob McDonald]

Gordon,

You need more spanwise resolution to your models.  The results will be meaningless until they are properly resolved.

Propwash on a wing is an acute case of this.  The propwash causes a localized area of upwash and then downwash.  In order to resolve this effect, you need numerous panels across the diameter of the propeller.

For cases like NASA's X-57 (many small propellers), this means they must use very fine spanwise resolution.

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+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Gordon Yuhim Lai]

Rob,

Thanks for the reply. So does it mean that I have to increase the tessellation of the model to increase the resolution? Or clustering?

Gordon

[Rob McDonald]

Increasing the tessellation will increase the number of points --
clustering can control how they are spaced.

Spanwise, you probably only need to use clustering for the root/tip
segments -- tessellation is sufficient for middle segments.

The best way to start is to experiment with both spanwise and
chordwise tesselation and clustering. Start with a very simple
example (a plain wing) and then build up complexity (wing/tail,
wing/fuselage, wing/propeller, control surfaces, etc.). Then, once
you have a good feel for how all those cases behave, you'll have a
reasonable feel for how to tackle a complex problem.

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.

[Gordon Yuhim Lai]

Rob,

Thanks for the advice. After increasing tessellation/clustering chordwise and spanwise, the VSPAERO finally gives some much more normal results by VLM. However, panel method still gives negative Cdi and zero Cdo - making L/D essentially negative. Should I further increase tessellation/clustering to their maxima?

And there is one side question, I notice there are Re and rho in the VSPAERO command window. What do they do in the solver - especially in a non-viscid model solver?

Thanks again.

Gordon

Rob McDonald於 2017年2月16日星期四 UTC+8上午11時28分46秒寫道：

[Rob McDonald]

I'm not sure, but CD0 may be turned 'off' in the panel method mode --
I wouldn't use it anyway, it is based on a very simple drag buildup
and you would probably rather do your own.

Reynolds number is only used for the Cf equation for the CD0 calculation.

rho and Vinf are only used to convert the CT and CP values for
propellers into the form required for the actuator disk model. If you
aren't using an actuator disk, you don't need them.

I wouldn't simply increase tessellation without bound -- there should
be a sweet spot in terms of results and expense. For clustering,
there should be a sweet spot -- best results for a given tessellation.
Too much clustering isn't good, too little isn't good either.

The best way to build up some confidence on this is to start with
simple cases, develop an understanding and feel for the resolution
requirements -- then apply those rules to more complex cases.

Rob


On Sat, Feb 18, 2017 at 3:52 AM, Gordon Yuhim Lai

[Gordon Yuhim Lai]

Rob,

I have tried the panel method again with the wing (without the rotors). Cd0 still gave 0 but Cdi was positive, so Cd0 might be turned 'off' as what you have mentioned. So how should I find the drag of the model by the solver? Is VLM fairly accurate for that? 

Thanks.

Gordon

Rob McDonald於 2017年2月19日星期日 UTC+8上午12時32分39秒寫道：

[Rob McDonald]

This gets into the realm of analysis that is more art than science.

Whether VLM or Panel methods is more accurate for your situation is a
little tricky to say.

A Panel method is probably easier to make accurate -- it handles
intersections better. It also doesn't make any sort of approximation
when smashing the geometry down to a camber surface.

However, a VLM computation is _much_ faster.

I would probably approach a problem looking for the best way to use
both. If I just wanted a single computation, I would use the Panel
method.

On the other hand, if I knew I was going to be doing a bunch of design
iterations -- or computations across a large flight envelope -- or
computations to fill out a full flight dynamics database (lots of S&C
derivatives, etc). I would take the time to develop confidence in the
VLM -- matching to the Panel method. Then use the VLM for the big
expensive study.

Once I was done with a design study, I would then spot-check the Panel
method again.

The Panel method is still pretty fast. If you are only running a few
cases, then its cost is small. However, if you are running thousands
and thousands of cases, then the speedup of the VLM is well worth
having.

Rob

On Sun, Feb 19, 2017 at 9:29 PM, Gordon Yuhim Lai

[Alexandre Pequeno Antunes]

Dear Lai,
the disc model can alter the local flow angularity in such way that it is possible to get a lifting
force that when decomposed in the wind reference system will yield to a positive CL coefficient
and a negative CD coefficient. Nevertheless, it is a good idea to refine your model.

[Alexandre Pequeno Antunes]

Just a small correction in the posted previous figure....

[corp...@gmail.com]

Here a possible improvement:

- Nacelles removed
- repanneling of all surfaces
- repanelling disks
- increased wake iterations

[corp...@gmail.com]

I add the  figure for a closer comparison



  

[Gordon Yuhim Lai]

Thanks all for the clear explanations and suggestions!

Gordon

Gordon Yuhim Lai於 2017年2月15日星期三 UTC+8下午1時16分42秒寫道：