NACA Sections

[Eric Sponberg]

Eric: surely keel thickness would be heavily influenced by the aspect ratio, from a structural point of view?

I can see how the thicker section would be desirable in disturbed water.

In aircraft thick sectons are more common in slow craft. Would this not also apply to a boat's keel?

Doug: are the Rave's foil sections asymmetrical?

Right, keel thickness also depends on aspect ratio and structure. And most boats are very slow, compared to aircraft, so there are any number of factors that come into play.

Paul B, you made the point in yesterday's post that many cruising designs have longer chords and shallower draft and that, therefore, a narrower keel would suffice. I agree. You design for the case at hand. For example, Project Amazon had a very short chord keel and had a high aspect ratio. The keel, as I said had a 30% chord trim tab, but it was also going to be made out of aluminum. The interior of the keel blade was to be a diesel fuel tank. Yet, the boat was a round-the-world racer, always sailing in waves of all sizes. All factors related to the keel spoke for a fat section, 21% aerofoil, and it worked well on all counts.

On the other hand, Eagle, my Globetrotter 45 cruising sailboat, does have a longer keel, shorter draft, and the volume necessary for the ballast speaks for a narrower keel chord, 10% thickness aerofoil. The rudder is a 15% aerofoil.

I am not afraid of fat sections. I like them fat or thin, depending on what is required. Bagatelle, for example, a 44' Long Island Sound daysailer and weekend cruiser, originally had a 10% aerofoil carbon fiber/wood-epoxy blade on a lifting keel. This worked very well. Later, the owner wanted a fixed keel at shallower draft, with a bigger bulb, and made out of stainless steel. I went to a 20% aerofoil section in order to get enough section modulus for strength. This keel worked equally well. They were vastly different in section, however.

Saint Barbara, a 37.5' sailboat design for a client sailing in Lake Michigan, similar in purpose to Bagatelle, also has a lifting keel with a carbon fiber/wood-epoxy keel blade. The owner wanted to engineer the blade himself using 18" wide carbon fiber bands set into each side of the otherwise wood keel blade, whereas Bagatelle's original keel blade had a complete carbon fiber skin over an internal wood core. I had the benefit of a total carbon fiber skin for the structural strength and stiffness, whereas Saint Barbara's owner limited himself to 18" wide bands. He needed more width for his keel structure to work, so we settled on a 17% aerofoil for that keel. That boat sails extremely well, according to the owner.

The point is, the selection of which foil section to use comes part and parcel with what the design is trying to achieve, and I have no qualms about using fat sections when the design demands them.

Eric

 

[peterraymond]

Reynolds number and foil points

Rick,

Thanks for the reply and the link. I think the link covers JavaFoil vs Xfoil pretty well and leaves me with a quandary. Since stall is pretty important, JavaFoil doesn't look too good for wing design. For instance, it's hard to keep thick foils from stalling, so the typical thick front element is hard to analyze.

I've expected critical transitions to be someplace between 10^5 and 10^6. For instance, figure 2 in chapter 6 in Hoerner, Fluid-Dynamic Drag. I thought past this performance would be reasonably stable and I could get well past by going to 10^7, but the polar was radically different. I then tried 140 points (JavaFoil changed that to 141) and looked at the polar again. Significantly different yet again, but perhaps slightly closer to what I would expect. The whole process is not as straightforward as one might expect.

Of course, Xfoil won't do double element foils, so it's starting to look like you have to spend money to get a good wing design, or at least to show that you've got one.

 

[peterraymond]

I tried 51, 101, 201, 401 and 801 points. The polar seemed reasonably stable between 401 and 801, but it does run slower at 801.

At Re = 1e6 it did change some with different numbers of points, but much less across the full range.

 

[daiquiri]

Peter,
This is what you should compare your graphs with (turbulent Re range), in order to find the correct parameters for the simulation:

 

[peterraymond]

Peter,
This is what you should compare your graphs with (turbulent Re range), in order to find the correct parameters for the simulation:

You're right of course and I have that book. I think I got a little ahead of myself, because I dove into JavaFoil and started playing with two element wings, before I compared the output to experimental results for a simple shape.

I now more or less know how to use JavaFoil, but I'm not really happy with the results. Even for a symmetric single element, unflapped dinghy DB, I'd want to use Xfoil or something similar for performance at very low speeds.

I realize that wind shear, plan form, boat motion and a bunch of other things still mean that the model results will miss reality, but it's nice to get as close as you can.

 

[daiquiri]

I think Javafoil might become a nice tool for airfoil analysis, but it would be necessary to allow users to have more control over the turbulence and separation models.
Please read my short note (posts #3 and #6) in this thread: http://www.boatdesign.net/forums/design-software/javafoil-31386.html#post343257
and the excellent explanation by T. Speer (post #15) about the mathematical models and computational methods on which Javafoil and Xfoil are based:
http://www.boatdesign.net/forums/design-software/reynolds-number-javafoil-31832.html#post351198
You really cannot expect too much from a 2D software near the stall angles of attack (a stall is a 3D phenomenon, involving mass and vorticity transfer along the wing span, in the direction perpendicular to the airflow/waterflow).
but you should be able to adjust the simulation parameters in order to obtain good agreement up to 7-8° AoA.

 

[Guest625101138]

Rick,

Thanks for the reply and the link. I think the link covers JavaFoil vs Xfoil pretty well and leaves me with a quandary. Since stall is pretty important, JavaFoil doesn't look too good for wing design. For instance, it's hard to keep thick foils from stalling, so the typical thick front element is hard to analyze.

I've expected critical transitions to be someplace between 10^5 and 10^6. For instance, figure 2 in chapter 6 in Hoerner, Fluid-Dynamic Drag. I thought past this performance would be reasonably stable and I could get well past by going to 10^7, but the polar was radically different. I then tried 140 points (JavaFoil changed that to 141) and looked at the polar again. Significantly different yet again, but perhaps slightly closer to what I would expect. The whole process is not as straightforward as one might expect.

Of course, Xfoil won't do double element foils, so it's starting to look like you have to spend money to get a good wing design, or at least to show that you've got one.

Peter
I have attached the results from 3 different resolutions - 61, 141 and 401 for a NACA0012. These are at 6e6 Re#. I cannot reproduce the plots you have for 1e7. Have you checked the options settings you have.

I have done considerable comparison against Selig data and JavaFoil and found it to be very reliable. In fact it is more usable because it does not jump about as much as the test data for very small changes in Re#. Admittedly most of my interest is at the point of maximum L/D rather than at stall and beyond.

The comparison I linked to earlier showed JavaFoil was conservative on the onset of stall and was spot on past stall with the test data. I feel it would give a slightly more conservative and realistic result than you would get from Xfoil in the case shown there.

Rick

 

[daiquiri]

Rick,
can you please post the relevant Selig data, so that a comparison can be made?
Those curves you have just posted are way off from both Abbott & Doenhoff and from Gregory & O'Reilly test data made at UK National Physical Laboratory. See the attachment below. It also contains a very interesting introductory note about some laminar-bubble and 3D-crossflow related problematics observed during NACA 0012 section testing, for Re below 2.88e6.

Peter,
I have read some claims that Rfoil software, made by danish Risoe research laboratory uses an improved mathematical model in the near-stall region. I have never tried it and don't know too much about it, but it seems that it can be requested via e-mail from Risoe. I understand it might be free of charge if you are a research worker or a student:
http://www.risoe.dk/?sc_lang=en

 

[tspeer]

Good paper!

...
I have read some claims that Rfoil software, made by danish Risoe research laboratory uses an improved mathematical model in the near-stall region. I have never tried it and don't know too much about it, but it seems that it can be requested via e-mail from Risoe. I understand it might be free of charge if you are a research worker or a student:
http://www.risoe.dk/?sc_lang=en

I've never received any response from them when I've requested rfoil (more than once). Anyone else have any luck?

 

[Guest625101138]

Rick,
can you please post the relevant Selig data, so that a comparison can be made?
Those curves you have just posted are way off the test data from both Abbott&Doenhoff and from Gregory and O'Reilly test data made at UK National Physical Laboratory. See the attachment below. It also contains a very interesting introductory note about some laminar-bubble and 3D-crossflow related problematics observed during NACA 0012 section testing, for Re below 2.88e6.

Peter,
I have read some claims that Rfoil software, made by danish Risoe research laboratory uses an improved mathematical model in the near-stall region. I have never tried it and don't know too much about it, but it seems that it can be requested via e-mail from Risoe. I understand it might be free of charge if you are a research worker or a student:
http://www.risoe.dk/?sc_lang=en

Give some examples of the variation with JavaFoil. The test data seems reliable against what JavaFoil is predicting although the data does show wide variation based on surface condition. Do your own analysis at the various Re# against the test data you have and see if there is significant variance. Once again do not assume the test data is accurate. There is great difficulty in measuring these things as the instrumentation can influence the results.

My interest has been high L/D foils at low Re#. The one I have used most often in the early stages before finding JavaFoil was the MA409. I have scanned the Selig data for that foil. I will leave you to run the comparison with JavaFoil.

If you are going to use test data in a meaningful way as the basis for speed prediction in different conditions it needs to be used interactively throughout the range of operation in the boat model. You need sufficient data to do a regression on all the variables that have significant influence on the results. You will go nuts trying to find enough foil test data to do this - Selig is the only source of suitable foil data in my range of interest that I have found.

The absolute errors between reality and the JavaFoil become insignificant in the scheme of things once the boat model is expanded to take all important variables into account.

Like all test data I expect the absolute error between the above reference and reality would not be insignificant either - particularly when you consider the instrumentation available in 1973.

A good model will provide sufficient accuracy to make relative comparison of alternatives. It might not give results to 4 places but it will give relative merits that are borne out by reality. A poor model will lead you in the wrong direction. I have found JavaFoil to be a good model. It makes it easy to quickly manipulate a huge amount of foil data and make useful comparisons. The end result with props is error less than 1% between reality and model.

Rick W

 

[Guest625101138]

When you do the comparison with the test data on the NACA0012 you will need to set the correct Mach number as some of the data is at relatively high Mach. The data I gave previously was for water.

Rick W

 

[daiquiri]

My interest has been high L/D foils at low Re#. The one I have used most often in the early stages before finding JavaFoil was the MA409. I have scanned the Selig data for that foil.

Well, I am reading "NACA0012" in the graphs you have posted above.
I am also reading Re=6,000,000 and Re=10,000,000 in your graphs.

So what are you actually talking about here? The talk is about NACA sections, and in particular about Peter's difficulty with simulating the NACA 0012 section test data with Javafoil. And you are comparing it to a completely different foil type in a totally different Reynolds range...? What's the sense of it?

Give some examples of the variation with JavaFoil. The test data seems reliable against what JavaFoil is predicting

I'm affraid your sight must be resenting too much time before the PC monitor because of this forum.
Well, you could ask someone with good eyes to describe you, for example, the immediately noticeable huge difference between the drag curves in your graphs and the experimental drag curves from the 2 references attached in my previous posts.

Like all test data I expect the absolute error between the above reference and reality would not be insignificant either - particularly when you consider the instrumentation available in 1973.

1973, not the 1773. The Eppler code, by the way, which is the base of the Javafoil, was made in 1980 - based on test data gathered prior to 1979, even back to 1949.

 

[daiquiri]

When you do the comparison with the test data on the NACA0012 you will need to set the correct Mach number as some of the data is at relatively high Mach. The data I gave previously was for water.

But I don't need to do any comparison. If you want, you can make the comparison. The comparison should be made between your graphs in the post #22, and the experimental data in the posts #19 and #23. Please let us know the results of the comparison.
Oh - and don't forget to make a correction for the Mach number, it is very, very important.

 

[Guest625101138]

Well, I am reading "NACA0012" in the graphs you have posted above.
I am also reading Re=6,000,000 and Re=10,000,000 in your graphs.

So what are you actually talking about here? The talk is about NACA sections, and in particular about Peter's difficulty with simulating the NACA 0012 section test data with Javafoil. And you are comparing it to a completely different foil type in a totally different Reynolds range...? What's the sense of it?


I'm affraid your sight must be resenting too much time before the PC monitor because of this forum.
Well, you could ask someone with good eyes to describe you, for example, the immediately noticeable huge difference between the drag curves in your graphs and the experimental drag curves from the 2 references attached in my previous posts.


1973, not the 1773. The Eppler code, by the way, which is the base of the Javafoil, was made in 1980 - based on test data gathered prior to 1979, even back to 1949.

The talk had progressed to validating JavaFoil. Peter's interest is with dual foils working at very high lift I believe.

The data you have provided is in air at high Mach number. I do not think that is relevant to what he is doing. The data I provided was at 0 Mach for water. If you want to see how JavaFoil compares with the data you have you can set it up and make the comparison. You need to compare eggs with eggs. You can see how much variation there is at different Re# and also different Mach if you care to look. He needs to be using test data around what he wants to do in reality and model with Javafoil accordingly.

JavaFoil allows different stall models if you want to update it over the Eppler.

I do not know why Peter's results are so different from what I get. He may have an old version of JavaFoil or he has has included boundary conditions that I have not.

Rick W

 

[Guest625101138]

Attached is the comparison chart I referenced earlier. You can take your pick of what is going to be closest to reality. For my mind JavaFoil looks impressive. It hits the middle of the other two. If you averaged the results from all three of them JavaFoil would be the closest to the average. The biggest difference would be conservative. You could also look at the variation between stall models.

If you go through any paper on testing foils that gives a description of the test set up and the instrumentation you become very skeptical of the results. It is as bad as tank testing tiny model boats and scaling up. How can the testers hope to get precise 2D data from a relatively confined tunnel.

There are a huge number of people using JavaFoil for designing and operating foils. It is being constantly validated.

Rick W