0
keith82687

Help with airfoil/rib shape making a canopy

Recommended Posts

Heya!

So a little over a year ago I started learning to sew on a single-needle singer 4423. I learned a lot and made a 37.5sq ft canopy using the GOE481A airfoil. Some random notes and a picture of me kiting that canopy can be found here.
http://imgur.com/a/xUEyH

The vents in this picture sucked and the center 3 cells were replaced with a much improved vent design (and in bright pink). It was a proof of concept (and great patience training :S) that I learned a LOT from.

My goal now is to make a ~240-265 sq. ft, 7-cell, vented BASE canopy. I had a couple incredibly insightful and useful meetings with Brian Germain (he lived nearby) and I will have access to a CNC canopy cutting machine to cut all the patterns. I've already done quite a bit of planning, but I'm stuck at choosing the rib/airfoil shape I want. I know most skydiving canopy rib shapes are a copy of one a few different airfoils.

To get a feel for variation in rib shape for BASE canopies, I borrowed several BASE rigs from my friends, pinned the outer-most rib of each canopy out as best I could, and took these pictures:
http://imgur.com/a/nwMk9
http://imgur.com/a/cwu0A

From the pictures I attempted to get a data set so that I could use something like http://airfoiltools.com/userairfoil/index to calculate some airfoil characteristics. It seemed to not be very useful. The crap resolution, not accurate enough coords are attached as an excel spreadsheet.

Here's where I'm stuck. I've debated getting a better trace of the outlaw outer-rib and just going with that, BUT, I really don't want to clone a rib shape. If I wanted to clone a canopy, I'd just buy one. What I'm considering now is to use data representing existing BASE airfoils and simulate/optimize them.

I'm going to mess around with http://www.airfoilevaluator.com/home.php that I just found, but I seriously doubt it'll do anything useful.

Something I'm more interested in using an analysis along these lines:
http://www.sciencedirect.com/science/article/pii/S1018363913000159

https://www.nada.kth.se/utbildning/forsk.utb/avhandlingar/lic/000215.pdf

https://www.mathworks.com/matlabcentral/fileexchange/50142-airfoil-optimization-using-xfoil-and-parsec-geometric-parameterization

But at this point I'm way out of my element and not even sure if there's a point or benefit in using any of these or similar methods. Anyone know someone I could discuss my project with or point me in any direction?

***disclaimer***
I already own a BASE rig, have ~350 BASE jumps, and do not plan to BASE jump this canopy (at least not without a LOT of skydives/testing on it first)

Share this post


Link to post
Share on other sites
You need to meet this crazy guy from england that's doing basically the same thing. He's been going through the process of learning how to build his own canopy for the last couple of years. He's doing well with several jumpable prototypes to show for it. Look up "My Little Project" here in the rigging forum. He's struggled with a lot of the same questions.

Airfoils do make a difference but don't get too hung up on it. For example he is finding that his airfoil has it's center of pressure slightly farther forward then for example, the old PD nine cell airfoil. If he trims his canopy in the same way he winds up with a very flat flying canopy with a very flat glide but a little short on flare athority.

I'm of the build it and jump it school of thought. Their are a lot of things about a canopy that just don't really translate to a program that is built to look at the wing of an airplane. There is a limited value in obsessing about the rib when it's a big floppy bag inflated like a pillow. I mean the "airfoil" is really Poofy between the ribs so you can't really say that the "rib" describes the "airfoil" very well. Some things do translate to the inflated shape of the canopy, at least in the grose since. Maximum thickness, it's location, to some degree the nose shape. But as you go back towards the tail it diverges more. The trailing edge on a normal canopy is very bulbous in comparison to the rib. They are trying to address that more with false ribs now but it's still there. My point is don't obsess too much with your first design.

Having said that, I met a guy from Icarus that was trying to play with iterative analysis of canopy design. He had some good CFD software and he was starting to make progress. He was also trying to model the flight of the canopy to give him some guidance of where to go with his iterative design. Don't think he was very far along with that but I give him credit for trying.

I've also heard about some of the manufactures trying to address flow separation problems in some slight modes. Base canopies have some really low renalds numbers. It sounded like they were rediscovering... I think it was called laminar bubble separation or some thing like that. It came up back in the 70's when they started to really look at human powered flight. There was a study and a set of airfoils built around avoiding it. Sorry, don't recall the paper.

I noticed in one of your drawing what looked like a simple overlap seam joining your top skins to your loaded rib. I know Germain did that on some of his canopies. Please don't. I had to deal with several of his early canopies that were just falling apart at the seams. I finally refused to repair them any more. He basically sewed the seam too narrow. Think about it. 1/2 inch seam sewed with say a 1/4 inch gauge. How much fabric is left on the edge? So one side only has to pull out 1/8 of an inch and the other side only has to pull out 1/8 of an inch and the whole seam starts coming apart. Canopy was falling apart at the seams. I mean that very literally. Brian is normally a very bright guy but he fucked it bad on that one. Every one is allowed one bad idea and that construction method was his.

Just go do it. Build some shit. Don't ever let any one, particularly any one who has not successfully done it them selves tell you that you can't do some thing.

Lee
Lee
[email protected]
www.velocitysportswear.com

Share this post


Link to post
Share on other sites
Personally, I would start by cloning an existing rib shape just to see how closely your canopy ends up matching the original. I imagine that there are plenty of places for errors, so cloning a known shape will tell you what adjustments are needed to have the plan match the finished product.

After that, I would look at something like OpenFOAM to actually model the airfoil and explore the effects that various changes have on its performance. Though as Lee said, I'm guessing that correlation between a computer model and the real thing will not be that great. There are also some non-free options like MultiElement Airfoils.

Share this post


Link to post
Share on other sites

Thanks for the reply Lee! Was the guy building his own canopy the thread where he called one of his models "the brown thing"? I followed that one and it was AMAZING! So awesome to see him do it over the years.

Thanks for taking a look at the drawings. They were for the little scale model. A lot of rushing/shotty work done it on just since it was for dropping weights, kiting, and learning. For the full scale I'll be using double the margins like you're talking about!

I think we're at a point with fluid dynamics simulations that we can use existing software (COMSOL / matlab / etc) to simulate the airfoil with more realistic parameters to differentiate the simulation from those applied to airplanes, wind turbines, etc. The amount of time it takes me to make a canopy with the equipment I have is very very long :P so I'm taking my time and doing as thorough of planning as possible. Once I actually have the CAD drawings and get the patterns cut, I'll be in no rush and happy with every seam sewn.

Cheers,
-Keith

Share this post


Link to post
Share on other sites
I built a bunch (dozens) of canopies many years ago. I figured for the first one, the best thing to do was copy what I considered to be the best flying canopy of the time (a 7-cell 189 foil). It required much more intricate sewing than others I could have built, but I wanted to get the sewing down before I worried about aerodynamics. I still recommend that approach.

I did a lot of seat-of-the-pants engineering and even my very weirdest canopy flew very nicely. It was a flat-rigged, double-humped, 5-cell, whose first jump was the day Mt. St. Helens blew up. We were in the air, watching the giant cloud of ash and lightning come toward us. I built it out of super cheap, really crappy (tear strength of about 3 pounds) but lightweight fabric. I just hoped it held together for at least one jump, so I could evaluate its performance. It flew and landed quite well I put a small number of jumps on it and sold it really cheap to a teammate, who put hundreds (I think) of jumps on it. It is still in a box somewhere.

One of the problems with you diving into 3-D fluid dynamic simulations is that so many little details such as deformities at line attachments, seam shrinkage, dimples in the top skin, inflation of cells, lines, the jumper, etc. are hard to model. Sure, you'd get some insight to trying a bunch of variations in an extremely complex parameter space, but remember, lots of people have been playing with this stuff for decades, and you might, but probably won't, stumble on anything that hasn't already been stumbled on. BUT, it is about the journey, not the destination.

My advice, don't get bogged down. Just design and build and enjoy.

Also, don't die.

Jeff

-- Jeff
My Skydiving History

Share this post


Link to post
Share on other sites
Short answer: Can't help you with airfoils.

Long answer, my thoughts on the topic:

Part of the issue is that the performance of the airfoil is just a part of the performance of the 3-D wing, which is just a part of the performance of the whole system.

For a conventional aircraft, one might know a lot about airplane design in general and then want to tweak the airfoil for a certain purpose. But the interactions between the airfoil and overall system are less clear for parachutes. One can't just pick up a standard aerodynamics text.

The paraglider guys have a bigger industry with traditionally more sophisticated aerodynamics -- using CAD/physics models that don't just draw the constructed parachute but calculate distortions in flight and then presumably overall performance.

I always wondered to what degree one could customize something like the X-Plane simulator to model a parachute. I know it has been done, but don't know if the parameters and aerodynamic modelling are sufficient for testing small variations. One is basically flying a tailless aircraft with a lot of anhedral and a VERY low set center of gravity!

(If one is mathy and wants to approach parachute performance from the basic equations, one could find an old J.S.Lingard paper on parachute design. One or two from the 1990s are around. 'Simulation and control of guided ram air parafoils', a U of Waterloo student paper, is in the same vein.)

Someone with basic airplane aerodynamics experience one might look at an airfoil and say "that's probably aft loaded.. with a lot of trailing edge camber adding negative pitching moment" and have some idea what that implies for an airplane ... but have little clue what effect it will have in practice for the parachute.

If I could magically play with different canopy designs, I'd probably start with a bunch of different trim angles (and thus fore and aft center of gravity) and see what values are acceptable. After that I'd start with airfoil changes, eg chordwise location of maximum thickness, maximum camber, reflex at the back, etc., as well as nose openings.

If you want airfoil contours to work will in the 2-D airfoil simulation programs, I think you'd want to have them really smoothed out. We just don't have airfoils that match the planned dimensions to a millimetre or whatever like a sailplane.

Skydiving airfoils with their distortions etc are kind of dirty, so who knows maybe one just wants to run a simulation with the maximum allowed turbulence or surface roughness. Not sure what they allow adding in.

This is all just off the top of my head. While I know a little aerodynamics I've never built any canopies.

Interesting stuff you're working on.

Share this post


Link to post
Share on other sites
Thanks for the reply! This is a very long term project I don't want to rush. A success is something that flies, but I want to reach as far as I can to learn as much as I can in doing it. Once I have an airfoil choice, everything from there on will be much more fun. I know how to generate the top and bottom skin panel dimensions based on the line set and airfoil top and bottom surfaces. I have vent/cross-port designs. Once I get the canopy sewn, I know that a huge part of the flight characteristics is line trim. For that I plan to use a 4-risers per side, all continuous line system that can be trimmed in-flight (I think that's what Nick Burden does) to come to a decision about trim and then design the actual line set from that.

Another thing he did that I really liked and may do is to have 3 risers per side for the final canopy. The rearmost riser, on which the toggles are stowed, would be a different/bright color and only go to the D lines and control lines. That way riser input is much more efficient.

If/when I have the motivation, I think the direction I'll go next is to take the BASE airfoils I have now, somehow smooth them out (not optimize), get 2D models for those, create another set that are optimized, and then compare existing BASE canopy airfoils as well as their potentially optimized counterparts.

I'd still love any information, suggestions, or points of contact anyone may have.

Share this post


Link to post
Share on other sites
darkwing

I built a bunch (dozens) of canopies many years ago. I figured for the first one, the best thing to do was copy what I considered to be the best flying canopy of the time (a 7-cell 189 foil). It required much more intricate sewing than others I could have built, but I wanted to get the sewing down before I worried about aerodynamics. I still recommend that approach.

I did a lot of seat-of-the-pants engineering and even my very weirdest canopy flew very nicely. It was a flat-rigged, double-humped, 5-cell, whose first jump was the day Mt. St. Helens blew up. We were in the air, watching the giant cloud of ash and lightning come toward us. I built it out of super cheap, really crappy (tear strength of about 3 pounds) but lightweight fabric. I just hoped it held together for at least one jump, so I could evaluate its performance. It flew and landed quite well I put a small number of jumps on it and sold it really cheap to a teammate, who put hundreds (I think) of jumps on it. It is still in a box somewhere.

One of the problems with you diving into 3-D fluid dynamic simulations is that so many little details such as deformities at line attachments, seam shrinkage, dimples in the top skin, inflation of cells, lines, the jumper, etc. are hard to model. Sure, you'd get some insight to trying a bunch of variations in an extremely complex parameter space, but remember, lots of people have been playing with this stuff for decades, and you might, but probably won't, stumble on anything that hasn't already been stumbled on. BUT, it is about the journey, not the destination.

My advice, don't get bogged down. Just design and build and enjoy.

Also, don't die.

Jeff



Aside from fluid simulating software being EXTREMELY expensive (if you want good ones) its also incredibly difficult and time consuming to learn and has a very steep learning curve. Expect to play with it for a year to just start to figure out how to get a simulation going. Alot of those guys who have 5-10 years experience with that software are really only scratching the surface on what it can do...and they are experts.

Share this post


Link to post
Share on other sites
To clarify the above comment, I don't think you find it difficult to get some of these packages "going" but getting from them any thing resimbaling the truth out of them is another story. Models are famous for lying to you. They can produce tremendous amounts of fabulous looking data that has nothing to do with reality. The "Expert" part is being able to get them to give you real answers. Alot of that has to do with learning to understand where they break down. And models do break down. You've got this idea in your head that CFD is some magical thing which can give you answers from on high like holly writ. It's not like that. Their are points where it "breaks", shit it just can't handle. And it will feed you garbage on a silver platter. For example, the nose. That big open hole with shit washing in and out of it . Total fucking disaster. any normal CFD program will shit it self trying to deal with that. Their was a study where they tried to simulate it. Really hard problem. Obviously no one from the industry. They chose a very pore airfoil and had the nose cut all wrong and only go some thing vaguely decent when they almost recreated the angle on a canopy. interesting study. Really showed the importance of the nose and nose cut in terms of the in and out flow in relation to the stagnation point. But it was a bear. Absolute monster of a problem. And that was in 2D.

So how do you actually think you're going to do this? One way, a classic way, would be to do a two dimensional section and then expand that into the finite wing. That's how a lot of planes are done. Thing is a lot of the techniques for doing this break down at low aspect ratios. Any thing below about an AR of 5 is low. You're talking 2. And keep in mind that you can't even do the wing section, see above study and how far it diverged from the original wing section. So what do you think that you're going to do? A big huge panel problem in a big 3D space? Very complicated and remember the nose. It shit it self trying to do it in 2D. Doing some problems in CFD are not hard but trying to get truth out of it on some thing like this is hard, as in fucking impossible. And how do you know if any thing it's saying is real? Well, you have to validate it, ie. build a canopy. Smartest thing would be to take an existing canopy. Di-sect it, take it apart, it's the only way you'll really be able to measure it, and then try to model it. Can you actually get your model to recreate the "real" data? And if it can't do that then your model is bull shit. And where does that real data come from. People have rented time in BIG wind tunnels where they could kite full size canopies. That's "real" data that you might be able to try to match your model against. And I think you'll find that you're mashing things around so much that you might as well abandon the CFD and just do curve fitting to the real data that you have any way.

My take, learn to build a canopy. Get a little data recorder and get some real world data from it. Tweak it and jump again. Repeat. Actual real data. Let's say you spend a year trying to make your CFD work. Well you've had to do the other to validate it any ways. In the mean time you could have spent a year building canopies. That's a dozen canopies maybe two dozen if you were really set up and on the ball. And you would have real world data from all of them over a spread of trims for each of your models. And it's real. Real data not made up bull shit and you've learned to build a canopy which is the other half of it. But that's just my take.

My recommendation. Get a data recorder. Start recording data on a base canopy. Start tweaking the trim on that canopy to see how it changes and works. Do this for a couple of canopies. Then try drawing your own rib. Build the canopy and run it over the same spread of trims as the others you have played with. See how they compare. Start building up a library of real flight data for all of these designs. It will give you a good base of real data as you incrementally create your own. And you learn to sew on things. And you get to jump and play with them. Doesn't that sound better then jerking off in front of a computer screen. If you really want to do that then you should at least watch porn and not a CFD program trying to crunch numbers.

Lee
Lee
[email protected]
www.velocitysportswear.com

Share this post


Link to post
Share on other sites
I designed and built two parachutes and put about a thousand jumps on each. I certainly agree with all that Lee, Jeff and Peter have said, cloth will form its own 3D shape and the finer details of the airfoil shape of the ribs will be lost. However if you are thinking of a significant change to the airfoil (as I did) it is important to remember that it needs to have strong lift very near the top front of the wing in all modes of flight to "pull" itself though the air, there is no hard structure like on a rigid wing to push the nose forward. You can search the forum for Nova problems. Stay safe.

Dave

Share this post


Link to post
Share on other sites
You might want to talk with other people who started building canopies after throwing a lot of CFD at them. Fluid Wings people are skydivers with a paragliding background they're applying to skydiving canopies, including working around the traditionally loose tolerances allowed in skydiving, as opposed to very tight ones in paragliding. They might be able to give you useful pointers. And if you do learn anything interesting that isn't their trade secret, please share :)
"Skydivers are highly emotional people. They get all excited about their magical black box full of mysterious life saving forces."

Share this post


Link to post
Share on other sites
Hey Lee,

Thanks so much for your detailed replies (I always follow your posts on BASEjumper as well).

I don't wanna make any changes too drastic to the airfoil of an existing canopy. This is mostly for me to push myself to learn more and build a rudimentary set of skills and understanding that maybe one day I could refine, learn from others, and actually pass the point of just being a tinkerer.

I'm going to start in 2D (likely with and without the nose cut) with the existing 7 airfoils I have and go from there. Maybe I'll get nothing useful...maybe I can see some differences in existing airfoils (especially considering how different some of them are from each other like the outlaw and the dagger).

I definitely know what you're talking about with models and shit data! A computer will spit results at you all day, ignore all the bajillion variables you didn't tell it about, and make you think you have something when, in reality, it only works for a spherical cow in a vacuum. I'm not expecting to get anything useful out of the modeling, but, it will be a fun, rewarding endeavor and maybe I might be able to get some small modicum of interesting information out of it. If nothing else, I may just change an existing airfoil very slightly (within the variation between existing airfoils I have). Who knows! I would "just go build it", but it'll take me a LONG time (my sewing machine and sewing space leaves MUCH to be desired). I'm not saying I want to improve or make something better...I just want to make something that's mine. In saying that I "don't wanna copy" a canopy, what I'm realistically expecting is that it'll probably be such a small change to an existing airfoil that it's negligible (especially considering effects of trim).

I would LOVE to take some real data! I won't delude myself and say it's not an option, but it's kinda unrealistic for me in my current situation. I was just daydreaming today about how awesome it'd be to have an array of pressure sensors I can put on the top and bottom skin of canopies and take real flight data over the whole control range. Especially considering that the "optimizing" often takes as an input the minimum pressure areas on the top and bottom surface :(

Speaking of...this is where I'm at now. I have photoshop files of each of the 7 BASE canopies in which the outer-most rib is cropped out. Here is an amazon drive link to them

https://www.amazon.com/clouddrive/share/EhvSHOvhzhCf6YZ2qp82T82IeNvzYw9yBcyollDGi8f?ref_=cd_ph_share_link_copy

I did a stroke layer to get the outline

http://i.imgur.com/8ekGV4x.jpg
(this is the blackjack)

Obviously this isn't exactly smooth...

After making two separate image files (one for the top surface and bottom surface), I used matlab to read in the image pixel by pixel and created two arrays of x,y coordinates which are scatter plots of the top and bottom surfaces. Averaged all non-unique x-values using accumarray. Then did a fit function (polyfit 8 coefficients)

http://i.imgur.com/QxMziDC.jpg

something about the scaling is off...so I have to figure out why, but you get the general idea.

I need to play with different fit options and which best smooths out the rib outline. My current goal is to get a smooth curve that best represents the rib of each of the 7 canopies.

Share this post


Link to post
Share on other sites
keith82687

After making two separate image files (one for the top surface and bottom surface), I used matlab to read in the image pixel by pixel and created two arrays of x,y coordinates which are scatter plots of the top and bottom surfaces. Averaged all non-unique x-values using accumarray. Then did a fit function (polyfit 8 coefficients)

http://i.imgur.com/QxMziDC.jpg

something about the scaling is off...so I have to figure out why, but you get the general idea.



See this: http://www.edmundoptics.com/resources/application-notes/imaging/advantages-of-telecentricity/

A photograph without any special lenses or post-processing will not give you an accurate 2D representation of the rib shape. Why not do this with a protractor and a ruler?

Share this post


Link to post
Share on other sites
mxk

***After making two separate image files (one for the top surface and bottom surface), I used matlab to read in the image pixel by pixel and created two arrays of x,y coordinates which are scatter plots of the top and bottom surfaces. Averaged all non-unique x-values using accumarray. Then did a fit function (polyfit 8 coefficients)

http://i.imgur.com/QxMziDC.jpg

something about the scaling is off...so I have to figure out why, but you get the general idea.



See this: http://www.edmundoptics.com/resources/application-notes/imaging/advantages-of-telecentricity/

A photograph without any special lenses or post-processing will not give you an accurate 2D representation of the rib shape. Why not do this with a protractor and a ruler?

By "the scaling is off", I meant that there is something wrong with my matlab code (I'll figure that out another day...i just needed to stop working on it so I could get my actual work done :P). I knew before I even took the pictures that images do not portray 2D shapes accurately. I'd say that me using a protractor and a ruler on a pinned-out canopy would not necessarily be better than the image method (especially considering the size and the difficulty of managing the fabric/tension/etc). Someone also suggested that I count the rip-stop squares and use them as reference to some points...but they aren't exactly "square". I think it seems easier to get the rib shape than it actually is. In a perfect world, I'd be able to stitch rip it out and use a vacuum table + (a most gentle) iron + a giant scanner...or just have the manufacturer send the rib pattern :P

I think the image way will work well enough though.

Share this post


Link to post
Share on other sites
In case anyone is interested, I'm still making progress (albeit slowly with work, play, and all the other things I'm doing atm).

I started using smoothing splines with variable knot positioning which led to a much better fit for the top and bottom surfaces than the polyfit function.

I'm keeping the same amazon drive folder updated (link at the bottom of this post). The main folder has untiitled.m which is my matlab script file I've been using to run the code.

The folders labeled "5 knots", "10 knots", etc have the m-files for each of the 7 canopies (with each folder corresponding to how many knots are used in the smoothing splines...more knots doesn't mean a better fit so I'm going to go back through them for each canopy's top and bottom surface to figure out how many knots give the best fit). If you want to check them out, you'll need the shape modeling language files from the matlab file exchange here (or just download the entire amazon drive folder, which has the shape modeling language files):

https://www.mathworks.com/matlabcentral/fileexchange/24443-slm-shape-language-modeling/content/SLMtools/slmengine.m

I really wanted to use the 'regularization','cross' options to get a really superb data fit, but trying to do that with 15 knots kept either taking too long (over 2 days and my computer would be restarted when I came back) or, most likely, crashing.

The link for my amazon drive folder for the project is available to anyone here:

https://www.amazon.com/clouddrive/share/12tPElq7QHwIdnZWQuQbDLflYpFlwuFWZQgIc9Hksjq?ref_=cd_ph_share_link_copy

I won't post another update until I have something more. It's been nice to get back used to using matlab for this project!

Share this post


Link to post
Share on other sites
My best translation:

He's working on his data set of existing airfoil shapes. As I recall he was photographing the end ribs of canopies or tracing them. This gave him kind of a wiggle out line.

There are a lot of ways to draw a curve. One is to set some points and generate the curves between them. The points are his "knots" and the "splines" are his curves. There are lots of preramiters you can set. Like the line goes straight through the points so the slope is the same on both sides. And how the curve changes between the points. Add infinitum. He's basically trying to draw smooth curves through the lines he traced off other canopies or photographed. There are even ways to automate this to let a computer optimize the shape for you to get as close to all the points as possible. You can if you wish make this as complicated as you like. His computer is having a hernia trying to crunch the numbers for him. The guy clearly has a computer fetish and worships at the alter of Intel. Not judging. It's not my place to criticize his religious views or his sexual deveancies.

As some one who was actually really into math, differential geomitry, and curves back when I was in school, I got two words for him. Ship Curve. Look it up. Hell, make your own. Find some thing springy and use it to smooth out the edge of your pattern. Done. Now take it and go start cutting fabric.

Lee
Lee
[email protected]
www.velocitysportswear.com

Share this post


Link to post
Share on other sites

Pretty good summary :p


regarding "knots": let's say you plot just the top surface of the ram-air airfoil...the theoretical plot goes from x = 0 to 1 and the y values give you the top surface. If I use 5 knots it means that the interval x = 0 to 1 gets broken up into 4 sections. For each section (between each knot), matlab finds the best 3-degree polynomial fit for your upper surface with supplied constraints. One constraint I used is that the top surface can only be concave down (forcing the second derivative to always be negative).

The "smoothing spline" consists of the 3 degree polynomials for each section between knots.

So if I add more knots it subdivides the top (or bottom) of the airfoil into more and more, smaller, and smaller, pieces. For each piece it does the best it can to give me a 3-degree polynomial fit. If you use too few knots, you get a bad fit (you can't model an airfoil upper surface well with just a 3 degree polynomial)...BUT, also, if you use too many knots, you get into issues with the resolution of your data and you start to get a fit that may seem pretty good when you zoom out, but as you zoom in you notice there are shit ton of squiggles as matlab tries to fit a 3rd degree polynomial between just a handful of data points. So the idea is to use constraints and an appropriate number of knots to best model the airfoil shape as smoothly as possible.

I'm working on figuring out that part now. I did a little more work on it tonight:

In the amazon drive folder there are now a lot of ".png" image files. If you open, for an ideal example, "20 Mojo 240top.png" you'll see a scatter plot. The x-values are the same as they were for the top surface of the airfoil. The y-values are = [the 20 knot smoothing spline with constraints function] MINUS the original scanned image values. It's an easy way to see visually where, and by how much, my model differs from my originally analyzed image.

I'm still working on the code so the pngs aren't complete (I need to make sure my fitting options and constraints are good before I can say those pngs are useful)

From these, for each canopy AND for each of the upper and lower curves of the airfoil, I'll choose whichever # of knots gives the best fit. Theeeeeen I'll have my representation of each of the 7 canopies.

Please know, I realize the accuracy of this is limited and it's likely overkill, but it's how my brain works and the best method I came up with at the time. But I could modify the code and use it on a more accurate image of an existing airfoil and immediately turn that into a 2D model, which I'm happy about.

The flow "simulation" will be an absolute joke since it's a shitty 2D representation of an infinitely more complex system, but I'll still do it. As long as I keep within my original stipulation (don't change an existing airfoil more than the variance between airfoils), it'll be fine.

Oh ya, and to Quagmirian..."adequate resources" are relative :P Regarding sewing, I have the most basic supplies (materials and a heavy duty / cheap single needle machine i made a 37.5sq ft model BASE canopy on), BUT I'm going to get the patterns CNC cut (an actual ZP/F-111 cutting CNC machine from a canopy manufacturer) when I get to that point

Share this post


Link to post
Share on other sites
Another logbook entry on the project.

After going through all the different fits for different canopies I've settled on an airfoil/rib shape:

http://i.imgur.com/C2qw5ql.jpg

Now I'm going to use that to get top and bottom skin panel dimensions. Then I add in cross-ports, vents, and sewing margins to everything. I updated the amazon drive folder to have the most current files.

Share this post


Link to post
Share on other sites
I think an interesting learning part of this project will be that the inflated and loaded shape can be vastly different from the theoretical shapes you draw. Although I am not a fan of copying an existing design, I do think it's a good starting point.

-Michael

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

0