www.recumbents.com - Shell Design Software
www.recumbents.com
www.recumbents.com
Home | Profile | Register | Active Topics | Members | Search | FAQ
Username:
Password:
Save Password
 All Forums
 www.recumbents.com
 Technical / Bike Building
 Shell Design Software
 New Topic  Reply to Topic
 Printer Friendly
Previous Page | Next Page
Author Previous Topic Topic Next Topic
Page: of 23

Speedbiker
human power expert

USA
3764 Posts

Posted - 10/28/2015 :  02:16:19  Show Profile  Reply with Quote
CFD software is as good as what you paid for it. Not to mention the system you are running it on. My brother has 5 monster graphics computers crunching numbers under his desk, and he still turned to the pros for CFD. What are you running your CFD on and what do you expect? Even with Predictive Engineering doing pro level CFD on their system, hours of working on the mesh, and starting with a model that had actual performance based numbers, we still only gave the CFD results so much credit. But the low end stuff is fun and it gives you some idea of what's going on.
Go to Top of Page

Balor
recumbent guru

Russia
514 Posts

Posted - 10/28/2015 :  03:18:25  Show Profile  Reply with Quote
In fact, I have a moderately good system: 4core I7, 24gigs of DD3, SSD... too bad that Autodesc does not use CUDA, I have a combatible card too... remnants of the former glory, heh.
But falcon is not just "quick and dirty", it is nearly devoid of options ( like surface rougness).
Hence, it may actually be "worse than useless" in some cases, by giving misleading information... or not. Will try Autodesk CFD.
Go to Top of Page

Speedbiker
human power expert

USA
3764 Posts

Posted - 10/28/2015 :  09:24:35  Show Profile  Reply with Quote
"Misleading" and "worse than useless" are the same terms my brother used for low end CFD. In the end you take your best guess and BUILD IT.
Go to Top of Page

Victor Ragusila
recumbent enthusiast

Canada
402 Posts

Posted - 10/28/2015 :  12:08:32  Show Profile  Reply with Quote
CFD is mostly Colorful Fluid Dynamics...good for pretty pictures, and maybe a good tool to get an intuition for fluids, but not amazing to get good data.

With ETA, we thing we get a lot of laminar flow. We think that because our CFD and custom code tells us so, but also because the shape is pretty much perfectly smooth, sanded to shine and well wax polished, beside using damping elements to prevent the shell from vibrating too much. Still, bug splatter can make the flow turbulent. Our power data also suggests that we get good laminar flow.

If you are not trying to get good laminar flow, the shape of the nose is not very critical. All you need to worry about is separation at the back, and wheel opening drag. Minimize those, and you got a fast bike. Keep the nose small and rounded, and you will probably get good cross wind behaviour, but I think it is easier to experiment with different wings on the back, than to change the shape of the nose to account for that.

Dont trust CFD too much.
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 10/28/2015 :  14:08:40  Show Profile  Reply with Quote
Baylor: Apologies, there were a lot of posts between my visits and I missed yours about exporting both halves to test. There's a lot to do before release and I still have other projects simmering. This code stomped all over things that worked before and there's a lot of safety stuff I got to do so it doesn't crash if you take an errant step. Also it doesn't export the back end yet. That's another step, probably nearly as difficult as this one.

IMO the two halves thing won't work well and here's why I think that. I'm too ignorant to suggest all the effects but here are some I think will be happening.

-- the airflow across two separate halves will be different than one complete shell.

-- viewing the shell as a wing section from above the center of lift on the back side (caused from the forward velocity with a cross wind) will be near the front and will change with... well... everything. That will produce a z-axis torque moment that wouldn't show up with halves.

-- also because, by the time you get to the back, the cross flow air is surely competing with longitudinal flow that is also spilling over the top and bottom (from the side wind pressure) the wind entry into the back half attached to the front surely looks a lot different than if it's standing by itself.

All of that is pretty esoteric and I have no idea if any of it is right but it seems that way to me and seems to muck testing in separate halves to the point I'm not sure I could get anything useful from it. Maybe you can.

Thing is, once you got whatever CFD you want to run, the runs don't cost anything more than time.



Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 10/29/2015 :  06:05:10  Show Profile  Reply with Quote
Oh sheesh, Baylor for some reason I didn't see your pics and posts about "useless". Falcon doesn't assume laminar anything, in fact the math is available at http://www.cfd-online.com/Wiki/Large_eddy_simulation_(LES)

You are running an iso-output. In the settings, you can adjust the pressures (or velocities) to see different visualizations.

[URL=http://s1372.photobucket.com/user/shooky56/media/Recumbent%20Postings/Cross-wind1_zpsei7xa1ej.png.html]
[/URL]

While anybody can slam any tool, mostly the value of the tool is how it is used. I grabbed known models and taught myself how to use the software until I had a good feel that I knew how to set it up to produce the actual wind tunnel results. Now that's not to say that the real visualizations of where the flow needs correcting can be garnered from this tool. It probably can't. I read up on these simulations while I was looking. Many start out to get a rough idea using the same equations, then the next level up process that data further and yet further and further. And at high speeds none of this apply (compressibility, i.e. "supersonic").

But the total drag for what is modeled is pretty darned close for what I've seen and I tried to be as precise as possible. That said, I know I skip a whole lot of drag that needs added in that hence Thom's "double" comment. He's right, these are too low for the real drag but not for what's being modeled. I mean I don't even have a frame in my run... nor a helmet but HEY I GOT ALL THAT FOAM BEHIND ME!!! lol).

Falcon is doing a good job for what I'm trying to get out of it.

Don't throw your skill saw away because it cut the board crooked.
Go to Top of Page

Balor
recumbent guru

Russia
514 Posts

Posted - 10/29/2015 :  06:48:39  Show Profile  Reply with Quote
quote:
Originally posted by shooky56

Oh sheesh, Baylor for some reason I didn't see your pics and posts about "useless". Falcon doesn't assume laminar anything, in fact the math is available at http://www.cfd-online.com/Wiki/Large_eddy_simulation_(LES)
...
Don't throw your skill saw away because it cut the board crooked.



Well, I meant something quite different, actually - namely, transition of boundary layer from laminar to turbulent that behaves quite differently, not that it cannot do turbulence at all.
I can be reasonably sure that it assumes completely smooth and uniform surface, but just please note:

quote:
Originally posted by Victor Ragusila
On a street recumbent it is very difficult to get laminar flow beyond the first couple of cm in the nose anyway, so they would solve a mostly unreal problem.


And it would affect separation point and, hence, optimal Kamm cutoff location.

My major gripe is that it does not allow one to assign surface roughness to the model, unlike 'proper' CFD tools.
And besides, if you intend to make your shell smooth and polished, it may work as intended for you.
In MY case, I'll be using a foam shell with polyuric coating, hence it would not be very smooth.
I'll need to run simulation on smooth and 1mm rougness shells and compare results. If it would result is greatly increased drag, I may actually rethink my idea.
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 10/29/2015 :  08:11:19  Show Profile  Reply with Quote
Surface Roughness: Not sure how they implement that, somebody in here may know. I mean one can either model the surface or estimate the effect. If it's being modeled then the vortex size has to reflect it. Even in Falcon it is turbulent very soon after the beginning on my models. If you tune the setting and see any green that has billowing, even close to the shell, it is turbulent.

Take a look at the F-117A. My models are like that, far too crude to run through any high-end CFD if they want several billion vortexes for a model. Forgot (it's actually pretty easy to check) but I think the shell itself was 60,000 or so. An entire frame, rider and shell was under 200,000 so my dump is far too crude to model for laminar flow. Every joining edge will certainly trip the flow in any CFD even the fake ones like Falcon. I am of the opinion that you expect the turbulence to look much different than it is? You're not going to see billowing swirls two feet behind the nose. That picture you showed is the boundary of turbulence IMO just well behaved turbulence. Between those boundaries shown is swirling air in there somewhere, and if you play around you can see it. If you like, email me that project file and I believe I can show you the turbulence in Falcon, just not the turbulence at 1mm thickness from the shell which is the turbulence we're talking about right? It's there, we just can't see it.

To find your project, click the output button, when the Windows explorer opens (hopefully your explorer has the addy bar), go back up two directories to the "HPV Design" then dive into the Projects folder and email the file for your project (make sure you save it first). It's a small file.

Kamm cut-off: It may not apply much at all for very turbulent but I think it does. In the U.S. many semi-trailers (18 wheel trucks) now have a pyramid structure on their back. The pyramid does not close to a point, it only drops to about 1/2 the size of the trailer's rear cross section. That airflow has got to be ugly, yet the cut-off still appears to be about 1/2 the width, just like the Kamm guidelines.
Go to Top of Page

Grant-53
recumbent guru

USA
544 Posts

Posted - 10/29/2015 :  09:29:01  Show Profile  Reply with Quote
Thom, you're right that there is not much reading on how a stall strip works on a liner body. I'll keep looking for useful articles to post. In the mean time I want to mount a 1/2" strip on the silver nose cone I made and do some tuft testing on the road.
Tony, the goal is to reduce lateral drag on the front of the shell and maximize down force on the tire patch. The dimple video shed some light on managing separation. The ribs from Wayne's design are not perfectly round; hopefully the difference in the top and the bottom radius may be used to minimize lift.
Balor, the good news is that a fairly safe vehicle does not have to be very heavy. Take a look at the frame design for the BMW C1 scooter or a modern race car composite tub. It's about converting and directing the kinetic energy of a collision away from the rider and keeping the rider restrained from impacting the shell. There should be studies on accident types and injuries. Drivers communicate with signals and horns. When some one is not paying attention, get their attention! I no longer worry as much about the 5000 lb SUV or trucks as the impaired drivers. The math most cited is for elastic collisions and vehicles are now designed to be semi-elastic. My daughter's 2000 Chevrolet Impala passenger car is much more crash worthy than my 1972 Cadillac which weighed almost twice as much. Safety is about training, energy management, and personal protection equipment.
Go to Top of Page

Grant-53
recumbent guru

USA
544 Posts

Posted - 10/29/2015 :  09:39:56  Show Profile  Reply with Quote
Victor, thank you for your input. Any simulation software is only as good as it matches actual events. I've done a fair amount of calibration work so I wonder if there are some standard shapes that can be used to test the accuracy of various CFD programs?
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 10/29/2015 :  10:24:57  Show Profile  Reply with Quote
I keep think we've got about half dozen topics running here in one thread. Fortunately Warren has posted links to download the msi's where they're easy to find (thanks Warren!). So, I guess it's not so bad except we've hidden other discussions under this one. Oh well, the headaches of administering a website is left to our poor volunteer.

Autodesk comes with some, there are some model sites out there and I pulled one or two in that were free looking for something that had a wind tunnel result. That model->to->wind tunnel link was researched. For instance I found one modeled vehicle model and those numbers are available somewhere (or some are). I pulled in a car for one of the runs. The car modeled well at 150 rez.

The very last thing I did was model a regular upright and I got close to the results for TT position (the only one I modeled). That's probably the worst test because there's a lot a rider can do to change their drag but, still, I was able to hit their numbers. That's where I found out that I had to use hi-rez because I was missing the frame (stays and smaller tubing) at first. When the resolution was upped, the numbers came in line.



Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 11/02/2015 :  11:44:16  Show Profile  Reply with Quote
Still writing code. A year ago when I started this, I plugged in increments along an airfoil chord length to compute y displacement. Since then, there have been a lot of mods to do stuff I did not think of, including the Coro-strips.

The coder's grief I've suffered is that those chord numbers had no real meaning relative to the shell length. Every time I needed a number along the length, it was complicated to derive it. Also, there are no "intermediate points". If you wanted a rib at x-location along the shell, the slices didn't have points corresponding to that x so I had to interpolate to fill in.

But the straws that broke the camel's back were resolution and Kamm'ing a nose-cone (open ended of course). I can't have more than about 60,000 connections comprising the entire shell or it will error out, that leaves about 20,000 points to describe the shell. That was because it was "one graphics element". That makes the average triangle about 6/10ths of a square inch. Doesn't sound bad but the old method had a whole bunch of .02 square inch triangles toward the rear making the front ones about the size of two fingers held close together. That's pretty big.

I have the new method that uses a collection of graphics elements to build up the shell. Problem is the old method was re-used dozens of times and there's a lot of patch work to get it all working. The new shells look very smooth though and are memory limited in resolution (probably billions of points). But the part I like the best from my standpoint is the points are spaced exactly 1/4 inch along the shell length. That means every thing like ribs or a Kamm will be "snapped" to 1/4" precision along the length. The shell itself can be any length as the last tail point is whatever it is.

Been a big PITA but it's getting there. Probably no one is eagerly awaiting the next release but there's a lot of stuff coming, including frame part dimensions (individual tubes so far).
Go to Top of Page

Balor
recumbent guru

Russia
514 Posts

Posted - 11/03/2015 :  04:00:18  Show Profile  Reply with Quote
Hey, I do :)
Go to Top of Page

warren
human power expert

USA
6120 Posts

Posted - 11/03/2015 :  07:52:56  Show Profile  Visit warren's Homepage  Reply with Quote
Sure we are all eager to see it. I think Thom bought a PC just so he could run this app. :)
Go to Top of Page

Speedbiker
human power expert

USA
3764 Posts

Posted - 11/03/2015 :  10:30:23  Show Profile  Reply with Quote
Ha ha! I'm not remotely computer savy enough to run that app! This Peaks Training project I'm on has my little brain boiling enough. But I hope some people do use it to produce either composite or coro liners. If they knew how much fun the will have, they would get to it. Those years I raced a variety of coros were some of my more enjoyable. Faster than a fast lowracer and no fear of crashing. Crashes in a coroliner are fun! Remember that one I had with the clear sides? That was a great bike. All this coroplast talk and that thread on Stix's amazing street liner about have me ready to wrap my cromoly midracer in coro. But I will use a composite nose and light aluminum stringers under the coro to build it. Not Wayne's amazing software.
I think a rich dude like Wayne should sponsor a coroplast build contest for NB next year. $10,000 to the nicest coroliner. Count me in!
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 11/03/2015 :  11:09:28  Show Profile  Reply with Quote
LOL, thanks guys. I want to add tubing type/size picker to report frame weight but that needs another popup window to let you pick mats. Would be pretty easy to add your own DB of "bike parts I like to use" to auto-tally the weight for you.

Right now though I'll focus on getting what's started out the door.

Even been thinking about making a frame simulator. Not suspension behavior as that would be a whole 'nother app, but a steering simulator.

I know how to do the momentum, centrifugal force, center of gravity stuff because that's all just ordinary physics. Have absolutely no idea how to do "no-hands" steering input from lean based off head tube angle and trail.

What I had in mind was a no-hands radio control simulator. If you've ever driven and RC motorcycle, you use weight shift and throttle to enter and tighten turns.

Since it's there's not much weight shift to a recumbent, I'm just guessing we ("we" is mostly "you guys" because I have about 20 miles in my life on a two-wheeler recumbent) we counter steer by instinct starting a left turn with a little right jig to get the weight leaning? I have solicited Andrew Dressel (who authored a simulator for wikipedia) but it's only been 20 minutes ago :)

About the only data I think I could gather is "speed to no-hands". Anybody think that would be a decent indicator of stability?

Probably could create a "computer pilot" and report just how skitty it was based on how much the pilot had to do to keep it upright.
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 11/03/2015 :  11:11:34  Show Profile  Reply with Quote
LOL, "rich dude": You meant yen not dollars right?
Go to Top of Page

Jerry
recumbent guru

USA
988 Posts

Posted - 11/03/2015 :  16:07:28  Show Profile  Reply with Quote
Wayne, you really do amaze me. When you are done, I am sure even dumb computer me can design and build using your app! If I run into to too much trouble, I might be a surprise at your front door! lol
You are only a little over 100 miles away. You don't seem to be too busy. lol.

Really though, next year I am seriously thinking of using the app and building a coro velomobile. Never owned a trike before, but sure loved trying out a couple of Catrike 700's and an ICE trike. I could race and tour in a velo! How much is 4 yens? I think I have a ruble also. Or is it a Paso?
Go to Top of Page

warren
human power expert

USA
6120 Posts

Posted - 11/03/2015 :  16:37:05  Show Profile  Visit warren's Homepage  Reply with Quote
Wayne, that's a cool image of your virtual velo hitting the blob in a crosswind!
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 11/03/2015 :  19:37:25  Show Profile  Reply with Quote
...the blob won...
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 11/04/2015 :  13:13:40  Show Profile  Reply with Quote
Mr. Dressel was kind enough to point me in the direction, app is free and reusable with appropriate credits. Not sure if I can use it but they have several validations from different places that the formulas work. Basically it says the speed the bike goes unstable, they were running bare frames, of course I'd like the rider on board. Haven't looked at the software yet, they claim it's not real easy to set up. But, to repeat, my thoughts are "the slower this thing handles with no-hands, the easier it is to ride". I also guess that some designs (reverse trail for example) can be shown to ... using Ralph Nader's title, "Unsafe at Any Speed".

I've got a bug in what I already have, the normals (normals are lines perpendicular to the surface for reflection calculations) are getting screwed up when I do the back half. It's been pesky and I've been on it for two days. It works just fine but it'd be like building a bike and hearing a squeak and not being able to track it down.
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 11/09/2015 :  15:18:12  Show Profile  Reply with Quote
Just an update: Found the problem with the shell creation not long after my last post. Since then I have been working to integrate the new shell into the rest of the code. Going pretty smoothly but much to do. In the rounds I did run an output on an upright nose and am positive that Falcon failed miserably to analyze that one correctly. Despite the rather turbulent air flow, it said Cd = 0.11. This is the same rider and bike that ran a 0.51 earlier this year, only change is the nose cone. Can't be positive but since Falcon has reliably and repeatedly agreed with other, more notable CFD runs for like models my guess is that the mesh import created closed solids where they did not occur. In particular, rather than an "umbrella in front", it thought it had a solid barrier across the back of the umbrella but even that's a stretch to get down that low.

[URL=http://s1372.photobucket.com/user/shooky56/media/Recumbent%20Postings/Upright%20Nose%20Cone%203-30mph_zpspbzksqjc.png.html]
[/URL]

[URL=http://s1372.photobucket.com/user/shooky56/media/Recumbent%20Postings/Upright%20Nose%20Cone%204-30mph_zpswvh4myvz.png.html]
[/URL]

Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 11/09/2015 :  15:23:25  Show Profile  Reply with Quote
You know, thinking about it, I had to screw that CFD run up because I had let it run 5 hours before this 15 hour run and had a .31 going. Accidentally shut that one down (clicked the X on the wrong app) and restarted it again. I bet I had 4 inch model or something. I'll rerun but got other stuff going on (it bogs my PC down pretty bad, shows 100% CPU use most of the time).

Go to Top of Page

Jerry
recumbent guru

USA
988 Posts

Posted - 11/09/2015 :  15:43:34  Show Profile  Reply with Quote
Wayne, hurry up! lol I bought a new Apple MacBook Air just to run your app. Not really, but my old antique computer keeps crashing. I don't know half of what you tech guys are talking about. But I think it is pretty darn awesome the amount of time and energy you are putting into it. And then sharing it with the world. Good on you!

Thom, I might let you use my new laptop. I will even show you how to upload pictures! lol
Go to Top of Page

shooky56
recumbent guru

USA
545 Posts

Posted - 11/14/2015 :  07:18:57  Show Profile  Reply with Quote
Jerry: The software will work to do the things you need as it is. Kamming can easily be done at the physical, rather than digital level (if that is desired).

Believe it's all working now but it takes me a couple days to test things. Since IMO, nothing in this release adds much of value (unless you need the digital Kamm for CFD runs), I'm going to work to unify my parallel versions. What had happened was I was working to remove the GDI frame drawing routines and replace them all with DirectX. GDI is used in the actor and frame designer.

Most users probably don't need to do the actor designer much. Once they have correctly entered their own data, they probably won't re-design themselves. But the frame designer could use some revamping. I've used the frame designer extensively and had it nearly converted before starting a divergent branch for the shell strip code.

Plans are to finish the frame conversion then merge the Coroplast and Kamm code with it. Fortunately both versions don't have a lot of overlapping code changes. Those who have actually used the software will probably like the changes.
Go to Top of Page
Page: of 23 Previous Topic Topic Next Topic  
Previous Page | Next Page
 New Topic  Reply to Topic
 Printer Friendly
Jump To:
www.recumbents.com © 2017 www.recumbents.com Go To Top Of Page
Snitz Forums 2000