Printer-friendly versionPDF version

As the dust settles on the Tour de France, and before it is stirred up again in Rio for the 2016 Olympic games, cycling sure is in the news of late. Unlike most sports in the summer Olympics, cycling is probably only one of two sports (the other being sailing) where aerodynamic performance directly impacts the athlete. So much so that at 50km/h 90% of the energy loss is related to aerodynamic drag. Cycling aerodynamics has been a widely studied topic for many years now, and its measurement takes on three forms: on-track testing (using power meters in the wheels), wind tunnel testing on cyclist mannequins (for direct force measurement) and CFD which, in comparison to other sports, is a relatively recent development in the last 15 years or so.

There are many factors in all three methods that give rise to measurement repeatability issues: rider positioning discrepancies, physiology differences between riders, bike equipment differences and the fact that no two wind tunnels are the same. Having designed and built a wind tunnel myself, I can attest that they are somewhat like children - you always overprotective of your baby! Unlike children however there are conscious design decisions that mount up in a making a wind tunnel that you have control over, hence the over protection. However, CFD offers the one thing traditional “physical” testing in the literature I reviewed lacks – increased repeatability and the ability to analyse flow and force results together without adding stings or fixing wheels to measure drag. One study impressively had a dynamic cycling mannequin to analyse the flow but due to the motor required to drive the model through the rear wheel force results were not gathered. CFD also offers insight into the flow field all the way up to the cyclist’s body, as laser diagnostic tools such as PIV and LDA used to unlock this in the wind tunnel would tend to scare the Lycra shorts off any living cyclist they were aimed at and none would agree to having any pressure tapings installed. A combination of all three test methods are required to fully understand, quantify and validate any simulation.

The one area, to my knowledge and brief literature review, that CFD has yet to be able to fully realise is a complete transient model with prescribed rider motion. This was until recently when a novel method* of using computer animation software to extract data from and use within the mesh morphing and overset meshing capabilities inside STAR-CCM+®; a feature that has been lying dormant waiting to be discovered. Armed with this approach, a curiosity to step into the unknown and a reasonable amount of patience to learn a new piece of animation software I set up a transient kω-SST model with 70 million cells (4mm sizing in the wake around and downstream of the rider) of a rider morphing the legs at 1.5-degree pedal crank angle intervals with full rotation of the pedals and wheels. The results in the video show the flow structure at 50km/h behind the rider for the 5th and 6th complete crank angles, it is interesting to actually see:

  • How the approach is able to morph the hips and stomach around the legs to increase the realism as opposed to using the overset mesh technique with a series of simplistic ellipsoids
  • How highly interactive all the vortex structures are off the shoulders, upper arms and elbows, hips, feet and helmet are – indeed the shoulder and hip vortices (for this position) tend to wrap around each other, a truly transient problem
  • The vortex “tendrils” forming behind the moving wheels as a result of the wheel rotation and the presence of the wheel spokes
  • How the large flow separation behind the leg in its full extension is destroyed when the leg is at the top of the pedalling cycle
  • Results for the drag coefficient area “CDA” showed some slight variation during the cycle, but were consistently around 0.292. A drag peak that was approximately 20% higher occurred whenever the down-stroke leg was at its full extension over a short duration
  • Drag was mostly stable as when one leg was doing one thing the opposing leg’s drag was doing the opposite. This compared to a CDA obtained from similar speed and position track testing of 0.276 and wind tunnel testing 0.296 which long term cycle averaged values

Needless to say these results are fairly dependent on this particular rider and position for this speed with this bike. Whilst there is more data to analyse from this simulation, at least the first step towards a “revolution” in analysing cycling aerodynamics within STAR-CCM+ is afoot, it seems that CFD just got that little bit more human after all.

* thanks to Rafael Ritterbusch and Gabriel Amine-Eddine for the “walker” demo that inspired this



  1. "A Phase-Averaged Analysis of the Pedalling Cyclist Wake", Crouch et. al, 19th Australasian Fluid Mechanics Conference, 2014
  2. Aerodynamic drag in field cycling with special reference to the Obree’s position”, Grappe et al., Ergonomics, 1997
  3. "Reference values and improvement of aerodynamic drag in professional cyclists", Garcia-Lopez, Journal of Sports Sciences, 2008
Products New: 
Matthew Godo's picture
Matthew Godo
STAR-CCM+ Product Manager
Stephen Ferguson's picture
Stephen Ferguson
Marketing Director
James Clement's picture
James Clement
STAR-CCM+ Product Manager
Dr Mesh's picture
Dr Mesh
Meshing Guru
Joel Davison's picture
Joel Davison
Lead Product Manager, STAR-CCM+
Ravindra Aglave's picture
Ravindra Aglave
Director - Chemical Processing
Sabine Goodwin's picture
Sabine Goodwin
Director, Product Marketing
Karin Frojd's picture
Karin Frojd