20 second Video to CFD (Aero Drag) test is HERE!

[Eric_Semianczuk]

I posted a bit ago to showcase the tangential possibility of creating an aerodynamic test from a video; after spending countless hours refining, building out a team, and a website all for this service, it’s finally here.

Wind Tunnel AI

Yeah… we hopped on the bandwagon and got another .ai domain, but it does use AI! No chatGPT wrappers or anything of that nature though…

So, what is it? This service, which uses technology that barely exists in a commercially viable form (most of it is still cobbled together from various research projects, but it works), enables anyone with a standard camera, to be filmed by a buddy while walking around them in one circular loop, with emphasis on the frontal area, to then be tested by the most powerful computational fluid dynamic software I could get my hands on (Open Foam) at 7.5 degree yaw and 22.3mph windspeed.

I then calculate a Cda value (lower number = more aero), and present that along with various imagery showing the wind turbulence/pressure, and the frontal view image used to determine reference frontal area.

Currently, as this is our first “Alpha/Beta” of any kind, we landed on $30 for each test, making iterative testing to make your setup faster definitely cheaper than it should be… is this some type of sales pitch? Here’s a timelapse with slowdowns at some key points showcasing the hourish of rendering and some of the manual effort that goes into each test:

process is
Video to images → images to where the images are in relation to each other → AI that determines what images would be generated in the scene from viewpoints we have no images, but it “learned” from the current ones → colored cloud of points → custom cloud to watertight polygon program → add back some details, fix random holes manually, scale based off of known wheelsize → test in Open Foam

(Thanks Trainerroad for allowing gifs! Although I did have to decimate the colors and compress the heck out of it to get it to fit the 4mb limit)

How accurate is it? I figured out how to rig the models and test different positions without even needing additional videos:

The difference between these small changes (shoulders in, back straight, head lowered, he was already trying to be as aero as he could, so there wasn’t much to update, but I tried!) at only 22.4mph, the newly adjusted position was 4 watts faster (0.006357 difference in Cda)!

All of the tools I use cost me/us nothing, either Open Source or my custom tools, so the price really comes down to the rendering and manual edits.

The goal here is to get some feedback, I’m no salesman, I’m a developer, if anyone’s curious, or simply want to try it out and any price is a barrier, just let me know, I have time, I’m happy to process some for free. If anyone has business model advice or do bike fittings or anything of that nature and want to connect/partner, let me know, all I want to do is grow this idea.

Also, here’s our insta https://www.instagram.com/windtunnel_ai/

Here’s one of my other sites, a 100% free, no ads or anything, cycling route creation website hosted on a server in my basement: https://sherpa-map.com

[KonaSS]

Is the original video/photo of the rider in a static position? Meaning not pedaling?

My concern would be that alot of people can put themselves in a static position that they can’t hold while pedaling.

[Eric_Semianczuk]

It is a static position, however, this is a solid point, and I may update the “how to” section, as I could add the instruction “for position testing, first start on the trainer while pedaling, when settled into a natural position that is likely aero and can be held while pedaling, cease pedaling and maintain the same feet/leg position for each assumed body position”.

Additionally, I’m experimenting with actually animating legs, but it’s not straightforward in the CFD test portion.

[christophebp]

Since this is static, what would be the best leg position for the rider? Top of the pedal stroke (6/12 o’clock), mid stroke (3/9 o’clock), something else?

[Rizzi]

Just a reference to my question in your previous thread. Maybe you want to play with that idea?

20 sec video to full on CFD AERO test!

Kudos…you need the right idea at the right time and we have a next David Tinker here (intervals.icu). Your cervelo model above has 2 front cables. Could you remove those cables in the model and show us the difference in cda or watts saved at 30 km/h for example. Could you also do that for 4 cables vs 2 cables vs 0 cables? That would be very interesting.

[Eric_Semianczuk]

Yeah, I’ve been very busy, will it show a difference? probably, could I go and try it, yeah, hold on.

[Eric_Semianczuk]

I’d imagine mid pedal stroke, feet aligned parallel to the ground, but as long as it’s static and the same it shouldn’t distract from the overall form/other changes to a noticeable degree.

We do have an aerodynamic specialist on the team and he’s advising on those particulars. This is why we’re in early beta/alpha stage.

[Eric_Semianczuk]

Okay, would have been faster but I was hungry…

So, I don’t have the bike model from the other thread, I deleted it at one point.

So I just used this one:

I got rid of the wires outlined.

Now, to be fair, they are a tiny bit bigger when I from points to mesh (still attenuating that part, but it works well enough)

A little choppy choppy:

Difference in CDA?

Cd : 0.527263
CdA : 0.226792 m²

to

Cd : 0.5264281
CdA : 0.226301 m²

at 22.3mph, that’s a difference of
0.298W

[Rizzi]

Thanks a lot! That is so amazing yet disappointing at the same time.

4 cables less and not even a single watt gained, that’s hard. Suddenly my desire for a new bike with a slick cockpit has vaporized…I was hoping for new records

Not sure what the difference of an integrated cockpit vs stem+bar is and / or a 2mm smaller top part of an aero bar / cockpit … but given the cable example above I’d assume your model will give less gains than all the aero claims and promises…

[jkarrasch]

0.226 is pretty fast.

Can you scale this gravel bike to a pro tour spec TT bike? Would prob be pretty low

[Eric_Semianczuk]

For reference, the rider was my twin, we’re both 5’6", and, as you can likely guess, obsessed with aero gains, and that’s the most aggressive position I could get on that bike, aerobars under handlebars, short, -30 degree stem, seat froward, tilted down etc.

I actually have a full TT setup I use for Ironman, and it only gets marginally better results, it’s actually a tad less aggressive because there’s really only one position, in the aero bars, unlike the gravel bike, without accounting for drivetrain efficiency difference (mtb clutch on gravel bike), rolling resistance, etc. yeah, gravel setups can get staggeringly aero, especially when aerobars are allowed.

I digress.

To your point, absolutely! The precision of this software can show the .29 watt difference of getting rid of some cables, the 4 watt difference of a 10 mm larger tire size, a 4 watt difference in a minute body position change (that looks almost the same to the naked eye).

Additonally, with ParaView, I can showcase actual models with the high pressure/low pressure, streatracers, colored by windspeed (magnitude) so you really get an in-depth analysis as to what may require adjusting.

[Eric_Semianczuk]

we’re disappointed too, we even have an aero specialist on board (we’re aiming to go to a wind tunnel the moment we can afford it) and this was already a matter of fact to him.

The thing is, I love Aero! I have a disc wheel, TT bike, spent so much time and effort working on form, built out analysis tools, have a nearby outdoor velodrome, and, over the years, I’ve discovered it’s only a minute upgrade in many capacities over 60mm deep wheels on my endurance road bike with a slammed stem and aero bars, replicated a similar position.

The thing that really kills it for me, is I’ve always done my best in Gravel racing with my hard tail with aggressive aerobars over my much more refined, position wise, aero focused gravel bike.

It really was a small difference in watts between the setups, especially at gravel speeds, and the handling capability of a mountain bike is just ridiculous in comparison, pulling off minutes against myself in sections.

I still couldn’t believe it, and, time after time, have regretted taking my gravel bike to races, now, this would be a different story if they disallowed aerobars in gravel races for non-elite, BUT, I’d probably just dropbar my hardtail then.

[Eric_Semianczuk]

I spent some time making a higher end version of the model to test:

Less than 2% difference in CdA between the two, if anyone was curious.

[splash]

Definitively an interesting approach. If you did get enough datasets, could you work out common factors in especially good positions, and then maybe even be able to give recommendations?

And yes, would be good to check the modelling results against actual data. How about using the Chung test on the road?

[Eric_Semianczuk]

Oh certainly, especially with things like water bottle placement/hydration pack placement, if I had enough data I could make generalizations like “road cyclists between x and y height and z weight are typically more aerodynamic with this type of helmet”.

A list like this would do a great job driving people to the site.