Basejumper.com - archive

Inspired by Jaap's post some time ago (http://www.dropzone.com/...ng=kayaking;#1577452) referring to some guy's fluid dynamics solver, I decided to replicate that work and see what it had to say about the kind of stuff I like to look at. Anyway, as a result I've built a virtual wind tunnel. If you own a Mac running OS X, email me and I can fire it off your way. If you own anything else and would like to take a crack at compiling it (requires OpenGL, LibTiff, and Lapack), email me and I'll send you the source.

ANYWAY... I've attached a series of simulations (each is a frame grab from an animation, taken when things had settled out) that came up in a discussion Jaap and I had last night.

The first is a simple circle in the wind tunnel. I've hand-drawn a flow line (and pointed out some vortices) to point out the presence of a Karman street (see http://www.galleryoffluidmechanics.com/vortex/karman.htm for a good explanation and links to a bunch of excellent animations). When something (like a pilot chute) is dragged through the air, it develops a low-pressure region behind it. This region is unstable, and collapses into a Karman street, in which the object sluffs off vortices first from one side, then the other.

Here's the important thing for BASE... It's not the "spilling" of air that causes a PC to oscillate or a round to pendulum. It's the presence of this Karman street, which places alternating sides of the round (or a location on the PC which circles around the perimeter) at low pressure. The spilling of air is a result, and is directly responsible for the higher descent rate on a penduluming round.

The next two attachments give the same simulation for a semicircle and an arc in the tunnel. The Karman street is still there. I half expected these to turn out about the same, since each should generate a dead space in front of itself, but there do seem to be important differences. I'm working on quantifying the force on the object (which I think is the really significant difference here) -- I'll get back to you all with that.

Finally, we have a "vented" arc. Notice that letting a little air through the apex of the arc results in the Karman street being pushed back, so that it no longer affects the object directly. It also has an effect on the dead space inside of the arc, of course, and I assume the pull force will drop a bit accordingly...

I've done some simulations on buildings, wings, cliffs... All with results a lot like one might expect. Let me know if you're interested in seeing any of this stuff, or if you've got any requests.

What fun!

Please could someone convert this to run on Windows?!

This would be such a kick ass tool!

Love it!

Nice one

I have it compiling and linking on Windows except for a Windows version of LibTiff. If somebody can find me a Win32 LibTiff binary or the sources and a Fortran compiler that is easy to set up, I'll make you an executable.

Edit to mention that it's not LibTiff that's the problem but Lapack. I could only find one prebuilt win32 Lapack library but it didn't have the two required functions.

I'll spend some more time looking tonight.

Great work Jason! The images look a lot like what we expected, don't they? I'm getting the semi-circle versus arc as well now.

I'm wondering if you can increase the resolution by an order of magnitude and change the single vent to a buch of smaller vents.

Thanks for writing this, very interesting.

Cheers,

Jaap

I'll work on that. I'm cheating a little on the sparse-matrix stuff right now, though, and as a result increasing the size by an order of magnitude would increase the memory requirements by a factor of 1,000 and slow the thing down by about the same.

As an appeal to all the geeks on here, if you're aware of a sparse linear solver for nice, banded systems which doesn't require the installation of a monster library, I'd love to hear from you...

have PM'd you,

Being a non techie boy, I hope its what you are after?!?!?!

Ive got Mac OS X, email me omarfamada@hotmail.com

Sorry Mac, what I actually meant was Lapack. Win32 binaries for LibTiff are easily obtainable. Thanks though! If any geeks on here want the compilable Visual Studio project (I had to make some minor changes to Jason's code) except for Lapack, let me know. I'll take another look and see if I can find Lapack tonight.

Jason, can you send me the latest version of your code? The one I have is a few weeks old.


You're telling me you're using a dense-matrix? Nice! Nothing like a good ol' float[100][100] eh? Check out the Boost implementation of UBLAS. It's pretty neat.

Not quite that bad. I'm using a band-diagonal solver, so that for a 10x10 grid (for instance), the thing is a float[10][100]. I'll look into your suggestion.

From 'The Gallery of Fluid Mechanics'


Longer delay --> higher airspeed (U)
Longer delay --> smaller PC (D)

Longer delay --> larger f from both factors.

Does this explain why PC oscillations are a major issue at only medium delays? Short delays have a oscillation period that is longer than the opening time and so have no effect and long ones have such a high frequency that nothing actually happens to the large inertial mass of the packjob?

Or am I talking out of my arse again?

Stop getting excited about computers Mac - you can walk outside and find your life again now! Nice work over the weekend though!

Jules

If this is an honest question, then the answer is a big shrug. I've tried to choose the parameters (but may have failed) such that they're appropriate to something maybe a meter in size moving at something on the order of ten meters per second in air. Anyway, it looks good, and isn't that what really matters?

If, however, you're following up on Jaap's success yesterday, I'll only point out that the Reynold's number only applies after nondimensionalizing the Navier-Stokes equations.

I think that's a great thought. One could also treat the pilot chute as a pendulum (with the force of gravity replaced by the total force on the PC) and find out what its resonant frequency is at each speed. Pilot chute oscillations will only be important near the speed(s?) at which that frequency is near that of the vortex street.

Hey Jaap, I did a little digging and found this FAQ page for Lapack.

A little ways down there is a question regarding Lapack for windows and a link to a Sourceforge LISP library, which isn't really helpful. However, I got a ton of hits when I did a search for Lapack off of Sourceforge's main page. You may want to give that a try.

Thanks! I managed to find one prebuild Lapack library but when I tried it, I still got linker errors, meaning the two functions that Jason used were either not in there, or their names were decorated differently.

I think I'll try and get a copy of Visual Fortran and compile my own Lapack to link in. Stay tuned, I'll hack something together tonight.

please keep in mind...
when I toss my pc out, I really hope it extracts my canopy before "things had settled out." the pc functions in tubulent conditions, not "steady-state."

also, remember the pc is not fixed it space.
the calculations are relative to the pc.
what does the airflow look like behind an oscillating pc?
-----

still, the screenshots demonstrate how a vented pc can be more efficient than a non-vented pc.

now screenshots of a steady wind around a building would be easier to interpret... that could be VERY illuminating.

And that brings up an idea that might just be crazy enough to work. What if minor differences (up to an inch) can have a profound enough effect on the size of the street in relation to its speed? Sort of like wave theory where tiny differences can cancel each other out resulting in a flat signal. Maybe we should have been jumping 43 and 47 inch PCs all along.

I suppose there are still too many variables (speed, weight and shape of body, weight of parachute, density of air) to do anything useful with this information, but it's fun to ramble about anyway.

Interesting point, and it made me think of something else. A PC most likely will not have perfectly clean air going into it. Does that utility support two objects?

I'd be interested in seeing the vented arc being simulated 9-10 feet after some representation of the human body (like an oblong oval or something).

We mean different things by "settled out". The simulations do, in fact, include turbulence. They are not steady-state; if you watch the full movie from which the above frames were grabbed (which I haven't posted), you can see vortices developing, being shed, etc.. The simulation begins with a set of initial conditions which are entirely unrealistic, however (uniform velocity), and it takes time for the system to develop to a more realistic (but still dynamic) state.

A pilot chute oscillates so slowly relative to the speed of the airstream that I'd be surprised if its lateral motion made much difference.

The problem is that the more precise you try to make the simulation, the less trustworthy it gets. The increase in complexity, resolution and dimensions when you add just the simplest extra variable is exponential in computational fluid dynamics (CFD). I'd say that Jason's 2D simulation of a single object is about as far as we can go while still being able to let skeptical laymen like us trust the results. And even then the results are still nothing more than an interesting thought experiment.


Interesting for sure. But not super useful for BASE. If anything, wind simulations like this and the Gallery of Fluid Motion book just scare the crap out of me and make me realize I should only jump buildings and cliffs in zero wind conditions. Then again, I'm a pussy that way. Nothing like a good ol' crosswinder to ruin your day.

Well -- let experience, caution and your WDI guide you for sure -- but it could be useful to see if maybe a cross-wind corner might be preferable to the leeward corner (in light winds, of course).

I can't guarantee the 9-10 feet, and clearly the circle I've got here is smaller than a human body, but as Jaap says, one shouldn't take these things too seriously anyway. But here's something...

The vortices, of course, aren't right where they were in the unperturbed case, but they're there and they're a lot like they were before...

Edit: Oops. Vented arc. I'll get on that.

What for? Mac OS X is where it's at.

Here's a block of four buildings with a wind coming in ten degrees off of one street. Synopsis, I think, goes like this... (1) Winds are funneled down the streets, the effect being larger the better the wind is lined up with the street. (2) Crosswind streets don't see much wind at all, but what is there might well be turbulent. (3) Downwind, even by ten degrees, is bad news -- a world of swirling vortices and smaller turbulence.

I'll let you draw your own conclusions about crosswind options.

I've been getting a few PM's and emails about a Windows executable of Jason's virtual windtunnel.

I mucked around with it for another two hours last night and managed to get Lapack and Blas out of the way. I'm now left with a dependency on libf2c which seems to be very Unix oriented (it requires headers the Win32 SDK doesn't supply).

And then it was bedtime...

I'll try some more tonight. After that I'm off to Twin Falls so it might become next week before I have something up and running. If and when that happens, I'll post a link to a download over here. Stay tuned.

Interesting stuff. I spent a few years solving 2nd-order partial differential equations (studying superconducting vortices). I used fast C++ code to crunch numbers and MATLAB to visualize the data.

The difference between superconductivity and BASE aerodynamics, I guess, is that there's almost nothing a piece of toilet paper won't tell you about the latter.