Home > Bike Types > Road Bike

Velocite winds up first aero road bike rims developed by artificial intelligence

Velocite Venn Var carbon filament wound aero road bike rims designed by artificial intelligence
31 Comments
Support us! Bikerumor may earn a small commission from affiliate links in this article. Learn More

The new Velocite Venn Var carbon rim claims to be the first-ever bicycle rim designed completely by A.I. The design was iterated through computational fluid dynamics software, and the project was led by CFD/AI expert and mechanical engineering professor Matthew Smith at the NCKU in Tainan.

Using AI allowed their team to iterate designs much faster than they could have by manually reviewing the results of each and every design change. Smith’s AI optimization routine was first programmed to find the lowest drag cross section possible that would work with a 25mm tire. They constrained the depth to a maximum of 77mm and let it fly.

Velocite Venn Var carbon filament wound aero road bike rims designed by artificial intelligence

The initial result was a rim that measured 42mm at its widest point, so they added a width constraint of 30mm so it would fit in traditional bicycle frames and forks. There were no other constraints to shape or size because Velocite founder Victor Major wanted to give the AI full autonomy to design the best possible shape it could.

Velocite Venn Var carbon filament wound aero road bike rims designed by artificial intelligence

The program went through 675 2D profiles before it found the best ones, then the design was put into 3D simulations to narrow down those candidates. Once the best performing rim shape was found, they machined the mold and got ready for production. But not just any production…

A new type of carbon filament winding

Velocite Venn Var carbon filament wound aero road bike rims designed by artificial intelligence

The prior Venn Rev rims used a very unique construction process called filament winding. We’ve covered it in great detail before, but the nutshell version is this: Carbon strands are pulled through resin then wound around the mold. The process allows extremely precise control of fiber placement and minimal excess overlap.

For the Venn Var rims, they took it one step further. The Var process uses thinly slit carbon fiber prepreg tape rather than individual strands, which they say is less susceptible to environmental changes. Meaning, they can make the rims in a wider range of temperatures and humidity conditions, which should lead to higher production outputs and more reliable supply. They’ll continue to use both Rev and Var methods as the former allows for more shapely designs for rims where aerodynamics are not the top consideration.

Specs for the new rims and wheels are:

  • Width at brake track: 29mm
  • Maximum width: 30mm
  • Depth: 77mm
  • Weight: 625mm per rim
  • Clincher wheelset: Venn Var 77 TCC weight: 1894g
  • Disc brake wheelset: Venn Var 77 TCD weight: 1915g

They’re available now, complete wheels are $/€1,499. Rims are available separately for $/€469 each.

Venn-Cycling.com

Subscribe
Notify of
guest

This site uses Akismet to reduce spam. Learn how your comment data is processed.

31 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
record11
record11
5 years ago

Now, here we have something interesting that is worth a HARD look at…..Look out SRAM/Zipp…..

Aaron Rumple
5 years ago

I’d like to see the 42mm wide version…

BMX
BMX
5 years ago

Should be called the Series 800, missed opportunity there I think!

Larry Falk
Larry Falk
5 years ago

Looks like these have a 21 mm internal width (if memory serves correctly, same as Bontrager carbon rims). Anyone know about their hubs? The 35 and 45 wheels are only $993, so maybe a very good deal.

Tom
Tom
5 years ago
Reply to  Larry Falk

I’m sure the rims are nice, but the hubs are nothing special. Therefore, I don’t really think the price is all that great. Of course, the name brand wheels are astronomically priced.

Victor
Victor
5 years ago
Reply to  Tom

We are using different hubs with the Var wheels. At this price point the hub features that have the most impact on performance are the origin (determining tolerances and consistency) and bearings (determining rolling resistance, lateral play and perceived longevity of the hub). So our hubs are made in Taiwan and the bearings in the Var wheels are Enduro.

KF
KF
5 years ago

Yes, of course filament winding is the greatest thing since sliced bread according to Venn and FSE for the last few years. Why have I never seen a set in the wild or a pro team on them? I’m sure the Ai claim will finally send them into orbit.. Not holding breath, but they do make for cute stories.

TobinGestatesYou (@TobinHatesYou)
Reply to  KF

I owned a pair of Rev 35 TCDs in 2016. Wasn’t anything wrong with them initially but my front rim failed in an odd way.

Here’s two images of the failure mode:
https://www.instagram.com/p/BJTwDS7A8hg/
https://www.instagram.com/p/BJTwGBDApdp/

Really not sure what happened there, so maybe Victor could give some insight. I initially wanted to warranty them, but the wheel builder I went through didn’t really respond to my requests so I just sent it to Raoul Luescher instead for him to eventually cut up. I’d be willing to try another pair of Venn/Velocite rims, but on the other hand filament winding doesn’t really mean anything to me on its own. Weight, strength, aerodynamic performance, price, customer service, etc. can be measured of course.

Victor
Victor
5 years ago

Sorry to see that. It looks like the spoke hole reinforcement element failed – likely at the point of overlap. It is a ring made of approximately 13 layers of carbon fiber that is bonded onto the rim body to reinforce the rim at the spoke holes. You can warranty this…

Var has a different structure entirely so it will not fail in the same way. I hope they do not fail at all, but that is an unrealistic expectation. In any case they passed the mechanical testing set to a very high standard.

* 0.11mm deformation at brake track when tire is inflated to 150 psi
* 0.01mm maximum deformation at brake track during the braking test when tire is inflated to 120 psi
* UCI impact test pass – no crack, no delamination
* no change in rim geometry after tire inflation, wheel building and side loading (high profile rims’ sidewalls tend to deform under same conditions)
* radial strength of rim – test stopped at 300 kg load
* spoke pull through at 550 kgf (wheel building recommended max is 130 kgf)

TobinGestatesYou (@TobinHatesYou)
Reply to  Victor

Thanks for the quick response. I did end up sending the rim off to Raoul in Australia, and it’s also been over two years since purchase, so I’m guessing a warranty replacement is no longer an option. Let me know if something can be done, and I can contact you / Venn via more appropriate and private media.

Victor
Victor
5 years ago

Send us an email to service@velocite-bikes.com but please be aware that 2 year warranty does mean that. If it has lapsed maybe we can still do something and offer you a lower cost replacement.

MadeInUSA
MadeInUSA
5 years ago

You read it right there. The computer says a 42 mm rim is best–and it’s a matter of time before that’s what we’ll be riding. Not saying this is good or bad, unless you just bought a new set of Zipps, you dentist!

Robin
Robin
5 years ago

The proof will be in test data. Until then, it’s yet to be decided if the “AI” approach is any better than any of the other optimization algorithms/software that other wheel or aero companies use. There’s also a definite lack of data (real world or lab) showing that filament wound wheels are in any superior to CF wheels built with traditional methods.

The wheels do look nice.

Johnny G
Johnny G
5 years ago

should have put a weight constraint as well.

Bob
Bob
5 years ago

So they us AI (what ever BS that is, lots of people use computer modeling. Is that considered AI) and they come up with a rim that looks just like every other deep section wheel.

Since AI is so smart, does that mean that those little saw tooth things on ZIPP’s wheels are dumb and stupid and nothing more than a marketing gimmick?

Robin
Robin
5 years ago
Reply to  Bob

Computer modeling is not AI. Perhaps you should use a search engine of some sort to see what the definition of AI actually is.

Bob
Bob
5 years ago
Reply to  Robin

not sure if their example really qualifies as AI though. they gave a CFD basic parameters and their program just tested a whole bunch more options. the computer was not really thinking for itself. Using the term AI in this context is just click bate. they should just say we test 650 designs and this is the best one.

Victor
Victor
5 years ago
Reply to  Bob

Hi Bob, in the context of the short article you are correct. The actual work that the AI did was to find a self guided solution in just the 625 or so runs. Brute force optimisation would have required between 10000 to 15000 simulations.

Bizzybob
Bizzybob
5 years ago
Reply to  Victor

10,000 to 15,000 iterations for a rim limited to 29x77mm? That seems highly unlikely.

Victor
Victor
5 years ago
Reply to  Bizzybob

I am not sure why it seems unlikely since the optimisation that we performed is parametric (ie. contunuous) not discreet, and I am telling you facts not opinions. I wanted to find a simple link that talks about AI versus brute force, but it is a complex field. So I encourage you to read more about it. It is in fact a cutting edge branch of science, however most examples that you will come across have to do with processing of population data (consumer preferences) and machine learning.

Cheese
Cheese
5 years ago
Reply to  Victor

How does this differ from conventional ANSYS parametric optimization sweeps, Victor?

Victor
Victor
5 years ago
Reply to  Cheese

I cannot say for sure as we did not use Ansys for optimisations. The optimisation methodology is described in this paper so have a look and compare: https://www.researchgate.net/publication/223614463_Rapid_optimization_of_blast_wave_mitigation_strategies_using_Quiet_Direct_Simulation_and_Genetic_Algorithm

We used Comsol as the 2D CFD solver, and then our own CFD package that uses the LBM method to simulate the 3D models.

Cheese
Cheese
5 years ago
Reply to  Victor

Thanks for the link, Victor, but I’m not clear how blast barrier simulations relate to rim wind resistance.

Victor
Victor
5 years ago
Reply to  Cheese

Well the specific result sought from the methodology is different, but the method is the same. Think of it like using a fork to eat meat or potatoes. It’s still a fork and it still does the same thing even though you use it for something different.

Eli
Eli
5 years ago

Looks like it was also restricted to needing a brake track. Be more interesting if it didn’t have to keep the parallel or at least mostly parallel sides

Victor
Victor
5 years ago
Reply to  Eli

We did not tell it to keep the brake track, it is just there and is not entirely parallel either as that was the best result. The only “weirdness” that we noticed in some of the abandoned optimisation branches that the AI tried was to introduce a concave shape. That branch was quite successful but ultimately the more normal looking profile that you see here won. I was also hoping that we would end up with something a lot weirder, alas we had no say in it.

Ian Grant
Ian Grant
5 years ago

I have Venn 35’s and now that i’ve finally gone tubeless would say they are better than my old style Enve 45’s and nowhere near the cost.

blahblahblah
blahblahblah
5 years ago

could have saved themselves lots of money and copied everyone else designs, hmm now that i look at them…..

Garth Magee
5 years ago

If you started with the assumption that the rim cross-section must be optimized for minimal drag located at the elevation of the axle, then all your subsequent effort was pointless. Bad initial assumptions will lead to the wrong result.

Wheel drag as it actually affects overall vehicle drag is highly concentrated near the top of the wheel, as taught by my patent US 9,878,745 – not at the axle level. One needs to minimize drag on the upper wheel surfaces primarily, which are actually magnified against propulsive counterforces at the axle. Any drag on lower wheel surfaces enjoys a mechanical advantage over propulsive counterforces, and thereby adds very little overall drag to the vehicle. (Just try resisting a toy car’s motion first on the upper wheel versus then on the lower wheel, and you can prove this key point to yourself ! You don’t need to be an engineer.)

Victor
Victor
5 years ago
Reply to  Garth Magee

Thanks for that. We use z-plane projections of the 3D profile. That is also how I designed the Velocite Syn bike frame and the Venn 507 rim, but without the AI.

Gary
Gary
4 years ago

How closely does the rim profile match the new ETRTO standard that is soon to be released?

Subscribe Now

Sign up to receive BikeRumor content direct to your inbox.

Subscribe Now

Sign up to receive BikeRumor content direct to your inbox.