I’ll start out by saying that if I knew how difficult it would be to achieve what I wanted, I likely wouldn’t have started. That being said, the knowledge and experience I’ve gained has been invaluable, and every failure has been a learning experience that’s made me a better designer and engineer. I owe what small success I’ve had making bicycles to an iterative deisgn process and from making everything possible with my own hands. There is no substitute for the process of designing and building something yourself and using your mistakes and successes to build something better the next time around. Looking through my notes from five years ago is a stark reminder of all the things I’ve learned that aren’t immediately obvious, even to someone with practical and academic experience in composites.
When I first became interested in road bikes (around 2006), the only real aerodynamic bicycles used by the professionals were time trial bikes- with the exception of the gorgeous Cervelo Soloist Carbon. At that time, it seemed like not only were aerodynamics in cycling the next big thing, but that there was a lot of low hanging aerodynamic fruit with regards to frameset and rider integration.
By 2014, not that much had changed on that front. Trek came out with the Speed Concept TT bike at the 2009 Tour de France, introducing the cycling world to the idea of truncated airfoils. What started as a purely aerodynamic innovation moved into the world of road bikes with the Scott Foil (famously debuted by Mark Cavendish) to create high stiffness to weight (STW) shapes still optimized for aerodynamics. I’d met Ben, a PhD candidate studying low speed aerodynamics, on a ride and we decided to build a frame together. At that time, it seemed obvious to us that the next step would be a frame design that integrated the drivetrain+brakes, hydration, food, tools/flat kit, and most importantly the rider into the aerodynamic equation.
Our design process basically follows three steps:
- Maximize the UCI allowed 2D profile for a given frame geometry. This gives us a 2D shape to begin laying airfoils on
- Guess at or measure airflow conditions at various points on the frame. A lot of this is educated guesswork based off of flow visualization and/or some small scale CFD analyses
- Optimize 2D airfoil geometry for the given flow conditions at multiple locations on the frame- usually we have approximately 3-4 airfoils per tube
Initially, we machined the molds direct to female, which created issues because the gelcoat used to seal the mold was too thick and ended up making the frame tubes measurably smaller than intended- which in turn affected the fits of the headset bearing cups, chainstays, and bottom bracket. In addition, we broke several tiny end mills trying to machine the corner radii and had to settle for less than ideal curvatures. Lastly, we had to machine each half of the main frame mold in two pieces, resulting in a slight misalignment near the top of the integrated seatpost. Luckily Ben isn’t that tall!!
At this point in development, our layup “schedule” was basically alternating layers of 0°/90° and 45°/45° woven twill 3k cloth with a central layer of thicker unidirectional along the length of the tube (Frame weight was not a priority). In general, we aimed for a uniform laminate thickness throughout a given tube section and added about 50% thickness in junction/transition areas. With a completed front triangle, it was time to bond on the saddle, chainstays and steatstays.
This was when we first really appreciated locating features as a concept. It is incredibly difficult to position the chainstays and seatstays in plane and at equal heights. Even just cutting these tubes to the correct lengths is very challenging freehand, and the old saying “measure twice, cut once” doesn’t even begin to capture the amount of care required. Another lesson to use in version two of the frame.
Near the end of version 1, we had a ton of lessons we’d learned and were excited to use them on the next version of the frame- combining some structural enhancements (better control over mating surfaces, monolithic chainstay, telescoping seatpost, improved cable routing) with some aerodynamic enhancements (improved shielding of drivetrain and bottles, seatstay drafting, and rider drag attenuation).
Structurally, none of the enhancements we were attempting were groundbreaking, but for us were a step up in complexity. The aerodynamic improvements, however, were ideas not yet seen in the bicycle industry at that time. The most noticeable feature of our new design was the wide, asymmetric downtube with integrated vortex generators- the idea being that we would create a bunch of drag at those features, but would create a boundary layer of energized flow that could pass over both bottles, leaving them effectively invisible to the wind at low yaw. At high yaw, our seattube design was deep enough that we could get a “sailing” effect- propelling the bike forward and negating the drag added from a stalled downtube at high yaw.
In addition, we incorporated a “wing” into the back of the seattube that was designed to create a low pressure zone behind the rider, drawing airflow down from the rider’s back and making the rider/bike system faster. However, this was dependent on the height of the rider- if there was a lot of seatpost, then the feature didn’t work as well and added some extra drag.
In addition, we incorporated a “wing” into the back of the seattube that was designed to create a low pressure zone behind the rider, drawing airflow down from the rider’s back and making the rider/bike system faster. However, this was dependent on the height of the rider- if there was a lot of seatpost, then the feature didn’t work as well and added some extra drag.
Making the CAD model into a working bicycle is another matter entirely. One of our design targets was to reduce corner radii- one of the most important features for a truncated airfoil. Most of the industry generally keeps 3-5mm radii, and we targeted >1 on critical features. In addition, many of the smaller features, such as the vortex generators on the downtube and headtube, required significant experimentation. In FEA, those features caused problems, and manufacturing wasn’t going to make it any easier. In the end, after trying several times- breaking two (!) sets of molds, and 5+ frame layups, I found a solution that worked. Small pieces of expanded polystyrene foam cut precisely to the shape of the problem areas and placed under the first “cosmetic” layer of carbon enabled me to get the shapes I was looking for. It was a shame then, that the best frame I pulled was not structurally sound because of a broken vacuum pump early on in the cure cycle. Pulling this half perfect, half deformed frame out of the mold was even more heartbreaking when I broke the third (and final) set of molds when it was almost out.
Never wallow long in misery. After swearing I’d never make another monocoque frame after that experience (and making a few tube to tube frames in the interim), I’m back and having another crack at it with a few lessons learned:
- Thicker molds- I was using molds about 2-3 times thicker than the frame laminate- talking with some other pattern makers has made me realize I should likely have been using something 8-10x as thick as the laminate instead.
- Fewer sharp corners- while it’s a huge advantage to have aerodynamic features with sharp edges, there were several areas where they weren’t necessary.
- Tube thicness variation should be minimized to reduce the complexity of the layup- if a tube changes along its length, it makes it very difficult to keep the fibers aligned properly
- Use nice materials. Many of my biggest issues stem from being cheap. Having experience doing it the cheap way has taught me where I can skimp and where it’s worth it to spend a little more.
I’ve also come up with some new ideas I can’t wait to try.
- Vaccum assisted resin transfer molding (VARTM). Generally speaking, this process is unsuitable for creating closed shapes. However, I’ve been doing some experimenting with new materials and think I’ve discovered a way to do it. The main advantage to this will be lower void content and a cleaner layup process.
- Cable integration- With the advent of off the shelf systems like FSA’s ACR, it’s possible for custom builders to make something as clean as the best aero bikes on the market. I will be creating some custom bearings to solve issues seen in current designs.
- 3D printed mold making- I will be making the plug/positive shapes for many of my components by additive manufacturing, before making the mold in the traditional manner. Up until now, most printers haven’t been able to achieve a surface finish good enough for mold making, however some recent advancements have made it a viable option. This should give me more flexibility on shapes and reduce my costs.
- Progressive road geometry- most road companies are too afraid to make changes to their traditional “road” geometry- and are possibly missing out on developments seen in mountain bike and gravel markets. Most professional riders use bars that are too low, making for locked out arms which are high drag and bad for handling. This is compounded with long stems taking them too far over the front wheel. I’ve been experimenting with longer reach, slacker headtubes, chainstay length and BB drop on tube to tube frames and think that there are definite improvements that can be made.
I can’t wait to share more developments soon!
OTB Engineering 