So after getting a prototype made up we finally made it to the wind tunnel yesterday up in Charlotte. The A2 wind tunnel is a bit of a commercial wind tunnel, renting out wind tunnel on a time basis. If you read any cycling publications you’ve seen its fans as backdrop for any number of “Aero” tests. It was a really nice facility. It was an open circuit wind tunnel that was sized almost perfectly for bicycles (although maybe a bit small for cars). It was a really professional facility.
We ran our prototype vs a few of the top aero wheels currently in the US with VERY good results, but we can still use a little refinement.
On the way back from the wind tunnel I had some good car time to think about why my CFD was missing some important things that we were seeing in the model. That’s when I stopped at a bike shop and happened to start talking to a bike builder and we got on the topic of CFD and wind tunnels. I mentioned that we did CFD analysis on a bunch of rim designs before heading to the wind tunnel with a prototype, he looked at me incredulously that we would even consider wasting our time at a wind tunnel when we could just model everything in CFD.
I just about threw a chair at him.
Yes computers now are very powerful, and a lot of CFD out there is VERY good at approximating scenarios. But, CFD will never replace a good wind tunnel test, ESPECIALLY with something like bicycle components. There are a few reasons for this.
First of all is Reynolds number similarity in the wind tunnel, which basically means matching accurately simulating flow phenomenon in a wind tunnel. Essentially you want your Reynolds Number and Mach number to be similar between your model in the wind tunnel and real life. For aircraft and rockets, this is very difficult. You cannot put a 1/50 scale aircraft in a wind tunnel and run it at flight speed, that would reduce your Reynolds number by a factor of 50 relative to real world, changing all the flow phenomenon. Your alternative is to increase airspeed by 50 times, this obviously is very difficult to physically do for any flight type air speed, but it will also entirely change the mach regime your model is in, rendering your model useless. Bicycles on the other hand are VERY easy to put in a wind tunnel. You can put full sized models in a relatively small tunnel, then run them at real world air speeds, giving you essentially real world drag results.
Second is that fluid flow, specifically turbulence is INCREDIBLY complex:
“I am an old man now, and when I die and go to heaven there are two matters on which I hope for enlightenment. One is quantum electrodynamics, and the other is the turbulent motion of fluids. And about the former I am rather optimistic” - Horice Lamb (1932) British applied Mathematician
“Perhaps the biggest fallacy about turbulence is that it can be reliably described (statistically) by a system of equations which is far easier to solve than the full time-dependent three-dimensional Navier-Stokes equation” - Peter Bradshaw Professor of Engineering @ Standford University
It might seem that bicycles, traveling at a relatively low speed of 20-30 mph are simple things, but the fact of the matter is that there are a ton of very complex things going on in the bike system. Take spokes for instance. They are not simply wires traveling through air, they are rotating, and they are small relative to the rest of the system. This means that they create entirely different types of flows than the rim. There’s also the interface of the rotating wheel and the fork, the stationary ground. Don’t even start on the body of the cyclist (which accounts for 80-90% of all drag anyways).
The solution to this is approximation. In every field of complex engineering we try to break up the model in to manageable parts, then figure out the interactions between these parts. Currently I’m modeling the rim shapes in a 2 dimensional domain. This has some very obvious shortfalls, but to run an accurate model of a 3D rotating wheel (just the front), is beyond the computational power that I, and probably most bike companies, have on hand. It is a light model that allows us to run through a relatively large design space with essentially a souped up gaming computer. That’s one of the reasons why we go to the wind tunnel in the first place, to refine the CFD. Yesterday I found out that my model was predicting the performance of some of the wheels incorrectly, and now that I have other data to reference the CFD against, I can refine the model.
There’s nothing wrong with approximation, it is how design is done. However if you forget that the model within your computer is just a model, you will make missteps or worse. CFD a.k.a Cleverly Forged Data, Colorful Figure Delivery, etc is great for marketing, it shows that you have made the steps to improve your product methodically. However it is very easily altered to suit your needs, I can make a CFD model show that our product is better than brand X easily with the right assumptions and models. You will never have a fully realized fluid model of a bike or any other complex design, in sports or aerospace. In order to close the gap between the simplified CFD world we work in and the real time gains on the road, the wind tunnel will never be replaced.