Anybody who has bicycled with a group on a windy day has enjoyed the benefit of sheltering behind the person in front of them. Similarly, it is well-known in cycling that drafting riders (cyclists who ride behind another cyclist) benefit from the slipstream of the front rider. That is why in races, like the Tour de France, cyclists try to ride in a line to save strength for the final part of the race.
A collaborative research effort was conducted to study cycling aerodynamics with our colleagues at KU Leuven (Thijs Defraeye and Peter Hespel), the Flemish Cycling Union (Erwin Koninckx), ETH Zurich (Jan Carmeliet) and the Dutch-German Wind Tunnels (Eddy Willemsen). We are all now continuing these studies with the ANSYS CFD code to look into the aerodynamic effects in time trial races and in full peloton sprints.
CFD simulations: Large Eddy Simulation based on grid-sensitivity analysis, with 15 micrometer near-body grid cells and time steps of 0.3 milliseconds based on CFL number. Figure: time-averaged velocity vectors, for dropped position, interbicycle spacing is 1 cm.
We decided to conduct our own research by modeling a group of real cyclists, since there was a lack of consensus in previously published scientific research about the effect of the trailing rider on the leading rider. While some studies suggest there is no effect, others argue that the effect might be substantial, maybe up to 5 percent*. However, previous studies were conducted based either on field tests, which are very difficult to measure, or on CFD studies of simplified models of human bodies, such as cylinders.
Using ANSYS CFD, we performed — to the best of our knowledge — the first CFD study on aerodynamics of multiple cyclists, based on geometry of real cyclists’ bodies. Our results not only confirmed a 30 to 35 percent reduction in air resistance for the trailing (second) rider, but they revealed that the air resistance of the first (leading) rider decreases, by about 2 to 2.5 percent. We successfully validated these simulations with a few carefully selected wind tunnel measurements on two cyclists in the Dutch-German Wind Tunnels in Marknesse, The Netherlands.
So how new is this result? The effect itself is just physics — it has always existed. But to the best of our knowledge, this is the first time that the type and size of this effect has been assessed based on studies on real cyclists’ bodies (not cylinders), and that its size has also been publicly announced.
And, how significant is this 2.5 percent reduction in drag? Because time trial races are often won or lost in a matter of seconds, the 2.5 percent difference on the leading rider can be extremely valuable when competing in a cut-throat environment. Here is what that percentage means in terms of distances and time at 54 km/h (typical sprinting or time-trial speed)*:
- 1 second gain on a distance of 1 km
- 10 seconds gain on a distance of 10 km
- 50 seconds gain on a distance of 50 km (e.g. time trial distance in Tour de France)
- 15 m gain on a distance of 1 km
- 150 m gain on a distance of 10 km
- 750 m gain on a distance of 50 km (e.g. time trial distance in Tour de France)
Some might argue that 2.5 percent for the leading cyclist is only a minor difference — particularly when compared to the 30 to 35 percent benefit of the trailing cyclist. However, remember that professional cyclists, cycling teams, bicycle manufacturers and race clothing manufacturers are all spending large amounts of funding and effort these days to achieve gains on the order of 0.1 to 1 percent. Efforts like these remain very valuable and should certainly be continued and pursued, and indeed can be decisive. But this also provides a reference to put the 2.5 percent benefit you could get “for free”, as shown by this study, in the right perspective. Note that the late Belgian top-tier cyclist Frank Vandenbroucke, just before each race, cut off the tiny side edges of the small number plate, fixed to his bicycle, since he was focused on reducing weight and aerodynamic resistance (in this case, estimated to provide a gain of the order of 0.1 to 0.01 percent).
Note that the 2.5 percent applies for cyclists that we tested with identical body shape and size. When the second cyclist is larger/wider/taller than the first one, the effect (reduction in air resistance for the first one) will be larger, and vice versa: A second cyclist who is smaller/shorter than the first will have an effect that is less than 2.5 percent. This means that for time-trial races, where the cyclists are all different shapes and sizes, cyclists order plays a very important role. As far as we know, this knowledge is not currently taken into account; it could potentially change the strategy and the way teams prepare for a race.
One way for cyclists to take advantage of these findings in sprints, theoretically of course, is to position the largest team member just behind the front runner during the final sprint, as the induced gain of 2 to 3 percent could make the difference between winning and losing. But we realize that, in actual practice, many other aspects play a role. Therefore, we expect that the benefits of this study will be limited to time trials or team pursuit.
These results have opened the door to many more investigations. A system of several cyclists under varying wind conditions may give some insightful recommendations about the order of the cyclists as well as the rider’s position on the bike in order to minimize pressure/drag on the entire team and to maximize the pressure applied to competitors’ cyclists.
Editor’s Note: Guest blogger Bert Blocken is a professor in urban physics at Eindhoven University of Technology, The Netherlands, and worked in concert with Thierry Marchal, industry director at ANSYS in Belgium, to produce this post.
*View the list of references for this article
Dear Bert,
Very interesting post! As many I’m sure, I am really surprised of the 2.5% reduction in drag experienced by the first cyclist. What I sometimes saw watching the Tour de France is teams cycling together in a group formation and alternating the leading rider. I think (but I am no expert) that they do so to maximize drag reduction for the group without exhausting a single leading rider. Only the key rider of each team never takes the lead position as he is supposed to keep the maximum amount of energy for the finish line.
Now looking at the huge reduction in drag for both the group and leading rider, do you think your study could change some tactics team use during races? For example, would they start using a team/pack strategy even more? Will we see the next Tour de France winner using that strategy thanks to Bert Blocken and his critical strategy input?
Thanks again for this great read!
Gilles E.
Hi Gilles
Many thanks. Indeed you are fully right about the lead rider and his team members. In a regular race with a sprinting peloton at the end, the lead rider will be sheltered by team members until the final 500 m or so. Therefore I do not think that the knowledge of the 2.5% reduction will make a different in such regular races. Because the cyclists will ride close together anyway, whether they are your team members or not.
In team time trials however, the situation is different, because the whole team races together and the time of the fourth of fifth cyclist crossing the finish line counts and you have to perform as a team. Here, CFD can make a big difference because every team member has a different body size and height. With a larger cyclist behind you, your drag reduction will be larger than 2.5%. With a smaller one behind one, it will be less. Therefore the order in which the team members ride and alternate will influence the dag of every member of team and also of the team as a whole. With CFD simulation, the theoretically best racing order of a given team can be determined. My colleague Thijs Defraeye is now performing such simulations for the Belgian national cycling team. We see that every different order gives a different total drag of the team. Of course also the power performance curve of each cyclist will be different, and this also has to be taken into account.
Hi Bert,
Nice results!
This reminded me of an episode of Mythbusters where they regarded airplanes flying in V-formation, equal to a flock of birds. They found, in a less scientific and more experimental method, results that are similar to the results you found. Not only the second and third plane, but also the lead plane experiences a drop in fuel consumption when flying in formation.
These results will probably be valid for a lot of different formations in a flow (different type of sports, transportation, migrating animals and probably natural formations) because of the influence of the second and third object on the vortices behind the first object, reducing the size of the slipstream and the amount of drag.
Good luck with the simulations for the Belgium team and I think the Dutch cycling team needs some help too…
With kind regards,
Reinier Maas
Hi Reinier
Thank you. The V formation indeed also has a very interesting aerodynamic background, although it is based on a slightly different principle. As in cycling, also in the V formation of birds, the lead bird is the one doing most of the hard work. What is different however, is that all birds behind the lead bird are not positioned exactly in the wake, but staggered, to take advantage of the wingtip vortex. This wingtip vortex is very effective, and every bird except the first makes sure to fly in the upwash flow that is caused by this wingtip vortex of the bird in front of it. It is (much) more effective than the effect of just flying straight behind each other. The V-shape can provides much more benefit than the 30% drag reduction effect that cyclists have on each other. For a more streamlined creature such as a bird, just flying behind each other will lead to even less than 30% reduction in drag. That’s why birds of the same species will almost never fly straight behind each other. The V-shape and the very large drag reduction is crucial for birds to be able to perform their very long migration routes. They also alternate in cycles. Interestingly, they are even known to help weaker members of their group by not forcing them to take the lead.
However, the lead bird would indeed have more advantage if the second bird would fly straight behind him/her – because of the same overpressure-underpressure effect as with cyclists. But overall, the group would not benefit from this. Mathematical models have been developed to assess the optimum flight configurations for birds, which are surprisingly similar to their actual flight behavior. A similar and very nice exercise for cycling races was done by Tim Olds, in 1982, who has actually provided mathematical models for cyclists to be successful (or not) in a break-away. The reference is:
Olds, T., 1998. The mathematics of breaking away and chasing in cycling. Eur. J. Appl. Physiology 77: 492-497.
Best regards
Bert
This is very interesting and fun!
Out of curiosity, what is the margin of error on the CFD and Wind Tunnel results? Clearly, the flow field around the cyclists must be very complicated with all the non linearities and challenging for the CFD to model.
Oh, if you don’t mind, two more questions. What was your Reynolds number and how well were you able to predict the absolute value of drag? I assume your Re is in the range of transition.
Sorry for all the questions! I tried finding free copies of the referenced papers, but couldn’t find any. And, in general, I am hesitant to buy CFD papers since there is so much written about it and the cost is hard to justify considering how much fluff is out there. But, was this the LLF or LST? Was a moving ground plane used? Was the Re used a value typical of what a cyclist would experience? Unfortunately, at these possibly low Re values, the results can be so very sensitive! Thanks, and again, sorry for all the questions.
Hi Martin
I will send you the three papers by email.
The Reynolds numbers are very large, due to the high air speed: typically 10e6. So fully turbulent flow, apart from the viscous sublayer and buffer layer at the solid surfaces.
The absolute accuracy CFD vs. wind tunnel experiments is about 10%. The comparative (relative) accuracy however is larger, but very difficult to determine. Detailed answers to the other questions can be found in the papers.
Hi Bert, thanks and I did receive the papers. The three that I got were for a single rider, do you have a paper which describes the tandem (drafting) rider results?
I hope you don’t mind that I make the following comments. Some of my bicycle riding friends are a little nuts (they won’t mind me saying this) so I thought I would post my thoughts on the matter. The comments are directed to those who know a little about aero, but are not hard core.
First, my thoughts are my opinion on this matter and are based on my experience with other geometries at high Reynolds numbers. I have never done an analysis for tandem bicyclists (even though I would love too). And I can’t say for certain that my experience carries over to this.
Anyway, any aero solution has, what sometimes I like to call, layers. Sort of like the transparent anatomy drawings. But, really, they are different effects. In this case there are the inviscid and viscous effects. And for viscous effects, there are the boundary layer, base flow, and wake effects. For this post I’ll only touch on the inviscid, base, and wake affects.
For inviscid effects, as two bodies get closer together they will try to push themselves apart for reasons Bert mentioned. Also, as the rear body gets larger compared to the front, the front body is pushed forward more. And I believe that is the effect being discussed here.
For wake effects, the rear body is in an area of reduced dynamic pressure, therefore, the rear body experiences less drag. Thus, drafting.
On the other hand, base affects are more tricky. As the rear body gets closer to the forward body it gets in the region of the base and the rear body will modify the shape and aerodynamics of the base flow. This can cause the two bodies to be sucked together, in other words, the drag on the front body begins to increase from some value, i.e. a minimum occurs. And, in this case, a larger body can increase the lower pressure region between the two bodies, in other words increase the drag on the front body even more. (Think of it in terms of an extreme example. If an object is sliced vertically [i.e. y-z plane] at any given x station, the pressure in the gap created by that slice will be some pressure reflecting the pressure at the outer edges of that slice. If the Cp at the outer edges is negative, the bodies probably will be pulled together, it the Cp is positive, then the bodies are pushed apart.) In the LES CFD picture above, it seems that the rear rider is starting to get very close to that recirculating, turbulent, base flow. It could be that if the rear rider gets closer to the front one that the axial forces on the two riders change such that they get pulled together. But, I want to emphasize “could”. It is not given. The effect is dependent on the geometry and Reynolds number.
And, to the general reader, I would like to emphasize that this is very hard to predict. RANS is unreliable for base flows and can be deceptive since frequently it gives OK integrated loads but poor pressure distributions. It does not completely capture the physics. On the other hand, LES/DES (as was stated by Bert) gives better pressure distribution results, i.e. it captures the physics better. However, it takes a lot of resources and it can be a pain to achieve numerically good unsteady results. Again, this is something Bert mentioned.
Bert, sorry if I stole your podium. If you would like to correct me on anything, or add to it, please do so.
To all, enjoy your cycling, and don’t hurt yourself!
A little more clarification, two similar sized objects should probably not travel so close that a significant about of separated flow on the lead body reattaches itself to the rear body.