Curvature of airfoils

[gerritgroot]

Hi,

Does anyone know whether there exists a standard way, or any other method, to smoothen the curvature of airfoils without changing the geometry too much?

Thanks in advance,

Gerrit

 

[rb1957]

do you mean you've produced a wing skin that has some waviness ?

 

[gerritgroot]

It's a 2D airfoil, I only have the discrete points and the curvature is whimsical or wavy indeed.

I want to slightly change the nodes, especially with respect to eachother, to smooth the curvature, without really changing the geometry too much.

...but not by hand, so I'm looking for a method...

 

[KENAT]

Sounds like whatever software you're using to 'join the dots' maybe needs a better 'line of best fit' type function. Or perhaps you're trying to achieve it using the wrong function or something.

Have you double checked the source data and your transposition of it?

Posting guidelines faq731-376

http://eng-tips.com/market.cfm?

(probably not aimed specifically at you)
What is Engineering anyway: faq1088-1484​

 

[gerritgroot]

I'm not looking for a software package, I'm looking for a method that I can code to process de discretized data directly (reading a text file and writing a corrected text file), without drawing splines or other types of curves by hand.

(The source data is whimsical, that's why)

 

[rb1957]

i get it that you don't want to draw a curve between your data points ... sensible, not very repeatable. i think you need a curve fitting program that will fit splines thru the know points (ie solve the problem mathematically, not graphically).

 

[gerritgroot]

Yes, you're right, but I thought, more people much have dealt with this problem, so there might be an algorithm around.

The point is that continuity in the second derivative doesn't assure smoothness in the curvature at all, it just assures you continuity (and I wonder whether continuity in the 3rd derivative will... so, some kind of filter or smoother has to be applied).

 

[btrueblood]

Yes, continuity in 3rd derivative will work. Google or wikipedia "spline fit".

 

[rb1957]

agree, that wheel has been invented ... i'd look in excelcalcs for a canned answer, wiki might show you how to develop answer yourself, google "spline fitting" should get other hits ...

 

[gerritgroot]

It is not so straightforward as it seems, with local spline fitting you'll end up doing it by hand, because you'll have to decide through which nodes it has to pass and through which nodes it shouldn't. That's why you need a filter judging on local curvature.

 

[rb1957]

yeah, if you've got a bunch of noisy data, maybe the "man in the loop" has to supply the fuzzy logic ?

if you know the airfoil section you wanted, like from NACA data, it'd be alot easier ... but then you wouldn't be fussing with this stuff ...

 

[gerritgroot]

Yeah, I wish it was a NACA, nice analytic...

 

[GregLocock]

The program I'd use is

http://www.octave.org

rather than Excel, as it handles text files gracefully, and has many curve interpolation routines built in. I'd have thought Bezier curves were a natural fit (sorry) for this problem.

Exactly how rubbish are the points you have? Do you have a high density of noisy points, or just a few points that need to be joined up gracefully?

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[gerritgroot]

Thanks,

I'll have a look at octave, anyway I use to use fortran to to this stuff.

And yes, I could do curve fitting, using whatever kind of tool, but I'm looking for a reasonable objective criterion to do local fitting automatically by some fortran code.

Even though the density of the nodes I happen to have now is rather low, I'd prefer a day of coding in stead of starting up some curve tool (having to learn it) and doing the process by hand.

That is why I asked for a method. With a method I meant a mathematical criterion to exclude or deviate or include points in your local fitting scheme to be able to automate the task. To do so, you need to filter the whimsical curvature data and use that as a goal within a maximum allowed deviation from the original geometry.

Obviously, if you do it by hand, and you pull some Beziers, it's straightforward indeed and a natural fit with Beziers is an option, but I'm not looking for that option.

Anyway, I'll look at octave, because I've heard more positive things about it.

Thanks,

Gerrit

 

[GregLocock]

You didn't seem to answer one of my questions, do you have many, noisy, points, or just a few that need to be joined gracefully?

If the latter than you may need to change approach as you work around the perimeter.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[gerritgroot]

It's not that noisy at all, and there're more than a few. I just want to improve the curvature for two reasons.

1. Far better computational convergence at los Reynoldsnumbers in programs like Xfoil.
2. Easier to convert it into a CAD surface automatically.

When you talk about noisy data, you probably refer to measurements from an already manufactured surface, that's not the case.

Gerrit

 

[rb1957]

may i suggest we were mislead by the word "smooth", instead you want to interpolate between sparse data. this clearly (?) is a different problem and depends on the airfoil section you're trying to make, which i'm guessing isn't a standard NACA type. personally, i'd start with a cubic spline through 4 points in the middle of the upper surface. this will define the tangent for the adjacent segments. Use the next 3 points and this tangency to make another cubic spline, etc. plot the mid-points (between your data points) that the splines generate. do these look rational ? i'm willing to bet you could do all this already and want to know what is the best way ? (what'll create the best airfoil) i guess xfoil will tell you that. you can iterate your spline by defining new points. what's the "best" solution ? no-one knows !

 

[sreid]

Ultimately it's NURBS, the mathmatical basis for all "Computer Graphics" today.

 

[mihumihu]

Gerrit,

google for Xfoil or Profili2, IIRC Xfoil can do smoothing of airfoil coordinates.

Michael

 

[gerritgroot]

Hey,

Thanks for all your answers.

@rb1957:
No, I didn't mislead you. I don't want to interpolate between sparse data, I'd like to change the curvature of a discretised airfoil, slightly moving the same nodes. I think this can be done in a discretised way, without any curve fit.

@sreid
Nurbs is about the worst choice for aerodynamic surfaces, the curvature of any automatically generatied nurbs are like the rocky mountains, even if they look nice smooth and shiny at first. A good surface is defined by single bezier surfaces.

@mihumihu
What are these programs about? I found profili, but not IIRC

Gerrit