Velocite plan to “revolutionize aero bicycle design”

Taiwanese brand uses supercomputer for latest project

by Mat Brett   May 4, 2012  

ZX Spectrum

Bike brand Velocite plan to develop “the most aerodynamic, non-recumbent bicycle possible” with the help of the Taiwanese National Centre for High Performance Computing (NCHC) and a supercomputer. The bike will not be legal for UCI (cycle sport’s world governing body) competition.

According to a statement from the Taiwan-based manufacturer, “Velocite has partnered with the NCHC to develop the next generation of aerodynamic bicycles using specially developed algorithms employing both computational fluid dynamics (CFD) simulation and genetic algorithm (GA) optimization engines conceived and written by NCHC researcher Dr Matthew Smith and a team of researchers from the Applied Computational Engineering (ACE) and Applied Scientific Computing (ASC) divisions.”

That’s a lot of abbreviations and not a lot of punctuation. The main point is that Velocite are hooking up with some computer boffins (that’s what you have to call experts of a scientific nature when you’re in the media) to develop a new aero bike.

The statement continues, “Among the significant resources that are made available to this joint project is the GPU (graphics processing unit) supercomputer housed within the NCHC location in the Tainan Science Park. This GPU based supercomputer was constructed and configured by NCHC staff and possesses more than 240 Tesla computing devices and approximately 8.4 TB of total system memory, making it one of the most powerful computer systems in the world.”

A TB is a terabyte. A terabyte is a trillion bytes. A trillion (in this case) is a million million. If, like us, you’re getting a bit lost in the detail, they’re doing the work on a massive computer, the aim being to simulate a complete bike in motion, including a moving and pedalling rider. Essentially, they’re planning to create a virtual wind tunnel in far more depth than anyone has done before.

“Optimization of the design will be conducted by a genetic algorithm where the final solution evolves over successive breeding generations,” the statement continues. “It is expected that approximately 500 generations will be required, resulting in at least 7500 individual CFD simulations. This means that during the course of optimization at least 7500 individual aerodynamic bicycle designs will be evaluated and tested automatically by the artificial intelligence GA for aerodynamic performance before the optimal design is found.”

The final simulation run will take over 25 days to compute. On a standard desktop PC it would take over 150 years.

Velocite won’t consider UCI regulations during the design process so the resulting bike will be aimed at triathletes and cyclists who don’t take part in UCI-sanctioned events. They expect technological and design breakthroughs from the project and plan to publish some of the findings in relevant scientific and engineering journals.

So, there you go. We’ll show you the design the computer comes up with as soon as Velocite send it our way. Watch this space.

12 user comments

Oldest firstNewest firstBest rated

Interesting stuff. Be surprised if someone like Trek hadn't already used something similar though?

(8.4TB = 8602GB, for anyone wondering. You've probably got between 2 and 4GB in your PC.)

Last night I would have considered trading a very loud baby for a really nice bike.

posted by notfastenough [3322 posts]
4th May 2012 - 12:09

2 Likes

The computer is a cluster at a national research centre, and I'm sure it's used for a lot more than just modelling airflow around recumbents Wink.

Basically, they've gotten a bit of run time on a super-computer.

posted by Paul J [646 posts]
4th May 2012 - 12:21

3 Likes

It's a bit more than some run time on a super computer! As far as I know no one has ever done this before. IMHO it's about time someone tried something new, hats off to a small brand like Velocite to try it. Takes balls to do that.

posted by steviebike [30 posts]
4th May 2012 - 12:38

5 Likes

Although a lot of work has been done on aerodynamics, it's so tied down by regulations. Even non-UCI sports like Triathlon have regs. Looking at motorbikes, it seems pretty obvious that some kind of fairing is what's needed to divert more airflow around the bike/rider, rather than trying to make the bike itself create less turbulence.

posted by outOfPhase [12 posts]
4th May 2012 - 13:23

3 Likes

That zx spectrum looks almost modern in a wierd kinda way. Even got world champ colours. Nice

Fringe's picture

posted by Fringe [1082 posts]
4th May 2012 - 14:15

2 Likes

Fairings are the answer. Just ask Graeme Obree. They should get him involved with this project, too.

posted by Yennings [224 posts]
4th May 2012 - 15:48

4 Likes

Yawwwwn.....
As long as it has more torsional rigidity whilst being more vertically compliant who cares.
I suspect the marketing men might have a hand in the unnecessary use of a super computer. Are we loosing sight of the fact, it's a push bike not a jet fighter?

posted by Glossies [29 posts]
4th May 2012 - 15:57

2 Likes

Yennings wrote:
Fairings are the answer. Just ask Graeme Obree. They should get him involved with this project, too.

Not if Velocite are hoping to appeal to triathletes... and they say they are. Triathlon has plenty of rules relating to bikes. One ITU triathlon rule is: "Windbreaks, other bodywork or other substantial means of reducing wind resistance on machines are prohibited."

Course, they might want to define "substantial" at some stage, unless they want to end up in a UCI-style pickle.

posted by Mat Brett [1906 posts]
4th May 2012 - 16:08

3 Likes

It'll be interesting to see what they come up with, but, utimately, it's restricted if they only consider the bike. It says they're going to run the simulations with a dynamic rider, but not if the rider's position will be considered/changed through the 7500 iterations.
Beyond fairings, which were mentioned above and available back in the '80's (I think Moulton even supplied them as an option on the AM7), the most influential/obvious factor has to be rider position (Superman position may be legal for triathlon?). Everything else is just p**sing in the wind, so to speak!

posted by pwake [309 posts]
4th May 2012 - 18:22

4 Likes

I'd guess this a probably an interesting computational problem and there's a bike-mad professor / PhD student in Taipei who;s persuaded Velocite to chip in some money to support the project. And they're getting some nice marketing in return. What would be interesting to know is the constraints they're putting around the design - does it have to be a standard double-doamond frame? Can it be a receumbent? Otherwise they'll just end up with a teardrop-shaped, faired recumbent.

Whether it produces something radically different will be interesting too. I suspect a lot of bike companies have thought about this then concluded it's actually cheaper and easier to hire some time in a windtunnel where you can simulate a bunch of different wind conditions fairly quickly.

posted by thebongolian [41 posts]
4th May 2012 - 18:41

6 Likes

Victor from Velocite here.

Thank you all for the interest in this story, and your comments.

I'll add some more information and direct you to the full press release that due to it being a joint press release,is indeed heavy on detail and technical jargon:

http://blog.velocite-bikes.com/2012/05/new-era-for-bicycle-aerodynamics-...

There are actually several supercomputers and high powered workstations that are used in this project. The final run, once everything is set up will be executed on the "real" supercomputer and that will occupy a large part of it for about 25 days solid. We now also have access to a wind tunnel, and other validation tools to establish perfect correlation between simulation and wind tunnel data (I cannot say reality since wind tunnels themselves are not perfect and introduce error).

The project is very complex with all software being custom written for a particular purpose. This specificity makes the project possible. Off-the-shelf software is not optimised for massively parallel supercomputers.

To the bike. We do not know what the end result will be, we have no preconceptions beyond the basic mechanical requirements. This alone makes the project unique. There is no "starting point" as such, where we look to optimise some aspects of it. The entire system comprising of the bike, fork and wheels is being developed from a blank slate. Nobody has done this yet, in any field, let alone another bike company. The resources required are well beyond commercial sensibilities.

Of course, there are constraints. It would be pointless if we ended up with a perfectly aerodynamic blob. Thus, the project will deliver a non-recumbent bike that can be ridden comfortably, in a familiar way, legally and safely in non UCI races. I specifically say non UCI since that allows us to push the envelope a little bit further. Of course we'll have sufficient data to make a UCI legal bike too, but that is not the initial focus.

The bike will be stiff, lightweight and will handle well. That's a promise, not a marketing goal Smile

I'll be happy to answer any further questions.

We will also be sharing some initial information, and hopefully have something to show in time for the Cycle Show held at the NEC Birmingham, 27-30th September.

Velocite Bikes CEO - my opinions may be biased Smile

vmajor's picture

posted by vmajor [4 posts]
4th May 2012 - 23:39

5 Likes

I would just like to point out that a certain G.Obree is doing something very similar to this, using all of the computer firepower that his kitchen can provide...
Check out the Vimeo clip/interview from a few weeks ago.

Currently going slower than I'd like...

posted by stealth [193 posts]
5th May 2012 - 20:27

4 Likes