“T

he greatest potential for improvement in cycling

speed is aerodynamic” [1]. At racing speeds
(about 54 km/h or 15 m/s in time trails), the
aerodynamic resistance or drag is about 90%
of the total resistance [2-5]. Most previously published
studies on cycling aerodynamics aimed at reducing the
aerodynamic drag of a single cyclist. Fewer publications

Surprises in cycling have addressed the drag reduction due to drafting. In

drafting, two or more cyclists ride close behind each other
to reduce aerodynamic drag. The few published studies

aerodynamics on drafting all confirm the large drag reduction for the
trailing riders (up to 30-40%), whereas there seems to
be a lack of consensus about the effect of drafting on the
leading rider. In this respect, Olds [6] stated:
llBert Blocken1, Thijs Defraeye2, Erwin Koninckx3, “It has been suggested that riding close behind a leading
Jan Carmeliet4 and Peter Hespel5 - DOI: 10.1051/epn/2013102 cyclist will also assist the leading rider in that the low pres-
ll1 Eindhoven Univ. of Technology - Eindhoven - The Netherlands - [email protected] sure area behind the cyclist will be “filled up” by the trailing
ll2 Univ. of Leuven - Leuven - Belgium rider. However, both Kyle (1979) and McCole et al. (1990)
ll3 Flemish Cycling Federation - Brussels - Belgium failed to find any measurable effect either in rolldown ex-
ll4 Swiss Federal Inst. of Technology Zurich (ETHZ) - Zürich - Switzerland and Swiss periments or in field VO2 measurements.”
Federal Lab. for Materials Testing and Research (Empa) - Dübendorf - Switzerland On the other hand, Computational Fluid Dynamics (CFD)
ll5
Bakala Academy - Athletic Performance Center - Univ. of Leuven - Heverlee, Belgium studies on human body models with simplified geometries
(such as elliptical cylinders) found drag reductions for the
leading cylinder up to 5% [9]. However, to the best of our
Drafting is riding close behind each other to knowledge, no studies have yet been published on detailed
3D CFD simulations of drafting cyclists based on realistic
reduce aerodynamic drag. New simulations and
human body geometries. Such 3D CFD simulations were
measurements for drafting cyclists show that also the focus of this new study, which included also wind-tun-
nel measurements. First, the wind-tunnel measurements
the leading cyclist experiences a drag reduction,
were used for CFD validation for single cyclists and two
up to 3.1%. For six or more similarly-sized drafting drafting cyclists. Next, supported by the validation study,
CFD simulations were performed for groups up to eight
cyclists, the position enjoying the largest drag
drafting cyclists, allowing assessment of aerodynamic drag
reduction is the one-but-last position. at every position in the group.

20 Article available at http://www.europhysicsnews.org or http://dx.doi.org/10.1051/epn/2013102

 cycling aerodynamics features

Wind-tunnel measurements CFD simulations versus wind-tunnel

Three sets of wind-tunnel measurements were done. The measurements
first set included overall drag force measurements as well The CFD simulations only consider the body of the cyclist.
as point measurements with 30 pressure plates on the body Therefore, the corresponding experimental drag area of the
of a real cyclist (Cyclist A: height 1.83 m, weight 72 kg) in cyclist body is obtained by subtracting the experimental drag
different positions, including the upright position (UP), the area of the bicycle configuration plus force platform, which
dropped position (DP) with straight arms and the time- was measured separately, from the total experimental drag
trial position (TTP) (Figs. 1, 2a). The second set comprised area of the cyclist body, bicycle and platform. The deviations
overall drag force measurements and point measurements between CFD simulations and measurements are 10.5%,3.5%
with 115 pressure taps on two reduced-scale (1/2) models and 0.7% for the UP, DP and TTP, respectively. Given the very
of cyclist A, obtained by rapid prototyping (Fig. 2c,d). The low percentage deviation for TTP, it is likely that some errors
third set consisted of overall drag force measurements on have cancelled each other. Similar simulations were made for
two drafting cyclists (Cyclists B and C) behind each other the two drafting cyclists in DP at d = 0.15 m, yielding a drag
at a wheel-to-wheel separation distance d = 0.15 m. In all reduction for the leading cyclist of 1.3% versus 1.6% from the
measurements, the wheels and legs were static (i.e., no pedal- wind-tunnel measurements. Both the CFD simulations and
ling). For brevity, only the results of the overall drag force the wind-tunnel measurements confirm the drag reduction
measurements on the real cyclists are presented in this ar- of the leading rider due to the presence of a trailing rider in
ticle. A conservative estimate of the measurement error of his wake. The agreement between the CFD simulations and
the drag force is 0.3% at approach-flow air speed U∞ = 15 m/s. the wind-tunnel measurements is considered to be very good,
The measurement results are reported together with the which justifies using these simulations for further analysis of
CFD results in the next sections. More detailed information the flow field and also using the same computational mod-
about the wind-tunnel measurements can be found in [4,5]. els (grid, turbulence model, etc.) for the CFD simulations of
groups of up to eight drafting cyclists.
Computational models
Digital models of the cyclist (Cyclist A) were obtained with Flow-field analysis for two drafting cyclists
high-resolution 3D laser scanning, capturing the specific Compared to the single cyclist in TTP, the drag reductions
body characteristics in UP, DP and TTP (Fig. 2b). To gener- for two drafting cyclists (TTP, d = 0.01 m) are 2.6% for the
ate groups of up to eight riders, the cyclist geometry (only leading and 13.9% for the trailing cyclist. The validated
cyclist body, not bicycle) was copied and the cyclists were CFD simulations are used to explain these drafting effects.
placed behind each other with a wheel-to-wheel separation Figure 4 displays the pressure coefficient Cp in the vertical
distance d = 0.01 m. The cyclists were placed in a computa- centre plane and in a horizontal plane at waist height of the
tional domain with dimensions and spatial discretisation cyclist(s). Cp is defined as (P-P0)/(0.5ρU∞²) with P the static
according to best practice guidelines in CFD and based on pressure, P0 the reference static pressure, ρ the air density and
grid-sensitivity analysis [10-13] (Fig. 3).Very small control U∞ the approach-flow air speed. The figure legend is limited
volumes of 30 μm were applied at the cyclist body surface
b Fig. 1:
to resolve the boundary layer down to the thin viscous sub- Three cycling
layer (Fig. 3b) This is important because boundary layer positions with
separation determines to a large extent the aerodynamic indication of bicycle
drag. Further away from the surface, tetrahedral cells were separation distance d
(wheel-to-wheel):
used with an average size of about 0.03 m. The grids for the (a) Upright position
single cyclist contained about 7.7×106 cells versus 35.6 ×106 (UP); (b) Dropped
cells for the eight drafting cyclists. The simulations were position (DP) with
made with a uniform inlet velocity of 15 m/s and a turbu- straight arms;
(c) Time-trial
lence intensity of 0.02% as in the wind tunnel, representing position (TTP).
the relative air movement when cycling at this velocity in
still air (zero wind speed). The 3D steady Reynolds-aver-
aged Navier-Stokes (RANS) equations were solved with the
standard k-ε turbulence model [14], near-wall modelling
with the one-equation Wolfshtein model [15], pressure-
velocity coupling with the SIMPLE algorithm, second-order
pressure interpolation and second-order discretisation
schemes using the commercial CFD code ANSYS/Fluent
12. Convergence was monitored carefully and the itera-
tions were terminated when all residuals showed no further
reduction with increasing number of iterations.

EPN 44/1 21
Features cycling aerodynamics

assumed – but by the one-but-last cyclist. Indeed, while the

last cyclist benefits from the leading riders in front of him,
the one-but-last cyclist benefits from both the riders in front
of him and from the rider behind him. This is confirmed by
the CFD simulations: Figure 5 shows the drag reduction for
every cyclist in a group of 2, 4, 6 and 8 riders. For groups of 6
or more similarly-sized riders, the one-but-last rider experi-
ences the largest drag reduction. For smaller groups, it is the
last rider that has the largest drag reduction. The reason is
that the wake behind the riders widens with downstream
position. Therefore, as an example, the last rider in a group of
3 benefits more than the last one in a group of 2. The widen-
ing of the wake becomes less pronounced from about the 5th
position, and the beneficial effect of having a trailing rider
in your wake then becomes comparatively more important.
In addition, note that the leading cyclist in a team of three
or more experiences a larger benefit (3.1%) than in a team
of two (2.6%), due to the upstream disturbance of the flow
(overpressure area) by the 3rd rider that extends up to the
position of the 1st rider.

m Fig. 2: to the interval [-0.05; 0.1] in order to highlight the changes Strategy in team time trials
(a) Cyclist A in the in the pressure field due to drafting; the actual maximum Many factors determine the outcome of a race. In regular
DNW LST wind tunnel
in Marknesse. The and minimum (absolute) values of Cp are much larger. The races, the above-mentioned drag reductions for the leading
cyclist and bicycle are figures clearly show the area of overpressure in front of the cyclist are probably too low for this knowledge to lead to
positioned on the cyclists and the area of negative pressure behind them. In different racing behaviour. This is certainly the case for a
force balance and
equipped with 30 case of two drafting cyclists, the wake behind the leading chaotic peloton sprint. However, team time trials are much
pressure plates. (b) 3D cyclist interacts with the overpressure area in front of the more organised than peloton sprints. In these races, a group
laser scanning of body trailing cyclist, which not only results in a drag reduction for of cyclists of the same team try to apply drafting in the best
geometry of cyclist A.
(c) Half-scale model of the trailing cyclist, but also for the leading cyclist. Note that possible way to achieve the best possible team performance.
cyclist A in time-trial also the size of the negative pressure area behind the lead- During the race, the members of the team alternate to take
position with 115 ing rider decreases due to a trailing rider in his wake. This the lead role.Apart from the alternating order, of course also
pressure taps. (d) Half-
scale model of cyclist effect shows a striking correspondence with the statement body size and shape and position on the bike are important.
A in upright position by Olds [6] about the low-pressure area behind the leading A larger trailing rider will provide a larger drag reduction
with 115 pressure taps. cyclist being “filled up” by the trailing cyclist. for a smaller rider in front of him, while the trailing rider
Photo shows pressure
tubes exiting from himself will benefit less because of the smaller wake from
the hollow model. Drag reductions for larger groups the smaller leading rider. In addition, different cyclists also
Because the aerodynamic drag of a leading cyclist is sig- have different power curves. The best strategy and alternat-
nificantly reduced by a trailing cyclist in his wake, it can be ing sequence in team time trials can be determined based
expected that in larger groups of cyclists, the largest drag re- on the combination of aerodynamic drag simulations and
duction is not experienced by the last cyclist – as is generally power performance curves.

c Fig. 3:
(a) Computational grid
for two drafting cyclists
(body of cyclist A) in
time-trial position and
with bicycle separation
distance d = 0.01m.
The grid on the cyclist
body and in the
vertical centre plane is
illustrated. (b) Detailed
view of computational
grid near upper body,
head and helmet.
Note the very high
grid resolution at
the body surface.

22 EPN 44/1
 cycling aerodynamics features

Limitations
The two main limitations of the study are, first, that all
cyclists had identical body geometry and position on
the bicycle, and second, that only static positions (i.e.,
no pedalling) were evaluated. The simulations were also
performed for cyclists directly behind each other at a
separation distance of only 0.01 m. While this distance
is unrealistically low when riding precisely behind each
other, it should be noted that cyclists often ride much
closer to each other, be it in a slightly staggered arrange-
ment, with the front wheel of the trailing rider next to the
back wheel of the leading rider. In such cases, the drag
reductions might be even larger. Future work will include
analysis of drag effects in such arrangements.

m Fig. 4:
Conclusions References Pressure coefficients
CFD simulations validated by wind-tunnel measure- CP in the vertical
ments show that drafting also benefits the leading rider, [1] D.G. Wilson, Bicycling science (2004). Third Edition, Cambridge, centre plane (a,c)
MA: MIT Press. and in a horizontal
due to the presence of trailing riders in his wake. The
[2] C.R. Kyle, E.R. Burke, Improving the racing bicycle. Mech Eng plane (b,d) at waist
drag reduction for the leading rider is up to 2.6% with height for (a,b)
106(9), 34 (1984)
one trailing rider and up to 3.1% with two or more trail- isolated (single)
[3] R.A. Lukes, S.B. Chin, S.J. Haake, Sports Eng 8, 59 (2005) cyclist and (c,d) two
ing riders. The same effects imply that in a group of 6 or
[4] T. Defraeye, B. Blocken, E. Koninckx, P. Hespel, J. Carmeliet, drafting cyclists in
more similarly-sized riders, the position with the largest time-trial position
J Biomech 43(7), 1262 (2010a).
drag reduction is not the last, but the one-but-last. The (TTP) with bicycle
[5] T. Defraeye, B. Blocken, E. Koninckx, P. Hespel, J. Carmeliet,
difference in drag reduction between both positions is J Biomech 43(12), 2281 (2010b).
separation distance
d=0.01 m. The cycling
about 1%. In top competitions, where winning or losing
[6] T. Olds, Eur J Appl Phys 77, 492 (1998). speed is 54 km/h.
is often a matter of seconds, these drag reductions due
[7] C.R. Kyle, Ergonomics 22(4), 387 (1979).
to trailing riders are significant and can be decisive. n
[8] S.D. McCole, K. Claney, J.-C Conte, R. Anderson, J.M. Hagberg,
J Appl Phys 68(2), 748(1990).
About the Authors
[9] A. Iniguez-de-la-Torre, J. Iniguez, Eur J Phys 30, 1365 (2009).
[10] M. Casey, T. Wintergerste, Best Practice Guidelines (2000).
ERCOFTAC Special Interest Group on “Quality and Trust in
Industrial CFD”, ERCOFTAC.

1 2 3 4 5 [11] J. Franke, A. Hellsten, H. Schlünzen, B. Carissimo, Best practice

guideline for the CFD simulation of flows in the urban environ-
(1) Bert Blocken is full professor and chair of building ment, COST Action 732: Quality assurance and improvement of
microscale meteorological models, Hamburg, Germany (2007).
physics at Eindhoven University of Technology, with spe-
cific expertise in (urban) aerodynamics. (2) Thijs Defraeye [12] Y. Tominaga, A. Mochida, R. Yoshie, H. Kataoka, T. Nozu, M. Yoshi-
kawa, T. Shirasawa, J Wind Eng Ind Aerodyn 96(10-11), 1749 (2008).
is a postdoctoral fellow of the FWO-Flanders in biosystems
engineering at Leuven University in Belgium. (3) Erwin [13] B. Blocken, C. Gualtieri, Environ Modell Softw 33, 1 (2012).

Koninckx is a postdoctoral researcher with a background [14] W.P. Jones, B.E. Launder, Int J Heat Mass Transfer 15, 301 (1972).
in both engineering and biomedical kinesiology working [15] M. Wolfshtein, Int J Heat Mass Transfer 12(3), 301 (1969).
at the Flemish Cycling Federation. (4) Jan Carmeliet is full
b Fig. 5: Numerically
professor and chair of building physics at ETH Zurich. (5) simulated drag
Peter Hespel is full professor in biomedical kinesiology at reduction in groups
the Bakala Academy at Leuven University. of 2, 4, 6 and 8
cyclists, in time-trial
position with bicycle
Acknowledgements separation distance
The hardware support of the Laboratory of the Unit d = 0.01 m.
Building Physics and Services for the CFD simulations
and the high-quality assistance of the Dutch-German
Wind Tunnel team (DNW) headed by Eddy Willemsen
are gratefully acknowledged. Thijs Defraeye is a post-
doctoral fellow of the Research Foundation – Flanders
(FWO) and gratefully acknowledges its support.

EPN 44/1 23