Measurement Science and

Technology

PAPER • OPEN ACCESS You may also like

- Wired and wireless seamless networks by
Uncertainty assessment of the Ring of Fire photonics
Tetsuya Kawanishi, Atsushi Kanno, Pham
concept for on-site aerodynamic drag evaluation Tien Dat et al.

- Modeling and Analysis of the Receiver

Performance in External OFDM-RoF
To cite this article: A Spoelstra et al 2021 Meas. Sci. Technol. 32 044004 Network Using QAM Modulation
Adnan Hussein Ali and Suhad Hasan
Rhaif

- A hybrid lightwave transport system based

on a BLS with an OSNR enhancement
View the article online for updates and enhancements. scheme
Cheng-Ling Ying, Hai-Han Lu, Chung-Yi Li
et al.

This content was downloaded from IP address 103.109.155.170 on 24/05/2024 at 03:35

 Measurement Science and Technology

Meas. Sci. Technol. 32 (2021) 044004 (14pp) https://doi.org/10.1088/1361-6501/abb50d

Uncertainty assessment of the Ring of

Fire concept for on-site aerodynamic
drag evaluation
A Spoelstra, M Hirsch, A Sciacchitano and F Scarano
Delft University of Technology, Aerospace Engineering Department

E-mail: [email protected]

Received 26 May 2020, revised 10 July 2020

Accepted for publication 3 September 2020
Published 15 February 2021

Abstract
The Ring of Fire (RoF) measurement concept to perform on-site experiments of aerodynamic
drag for transiting objects is investigated with a study that identifies the main sources of
uncertainty.
The main contributors to the uncertainty of the drag measurement are examined for the case
of a cyclist riding through the measurement domain. A sensitivity analysis is conducted that
assesses how the estimated drag is affected by the choice of particle image velocimetry image
processing parameters. The size of the cross-section considered in the control volume
formulation is also investigated. It is found that the accuracy of the estimated drag depends on
the procedure used to detect the edge of the momentum deficit region in the wake. Moreover
imposing mass conservation yields the most accurate drag measurements. The drag estimation
has little dependence upon the spatial resolution of the measurement as long as the interrogation
window size stays within 5% to 25% of the equivalent diameter of the object cross section.
Experiments are conducted in a sport-hall, where the aerodynamic drag estimates from the RoF
are compared to a conventional torque power meter installed on the bicycle, and different rider’s
postures as well as equipment variations are considered. Although the discrepancy in the
absolute value of drag can be as high as 20%, power metering and RoF agree within 4% on
relative drag variations.

Supplementary material for this article is available online

Keywords: large-scale PIV, helium filled soap bubbles, uncertainty PIV, cycling aerodynamics,
on-site aerodynamic measurements

(Some figures may appear in colour only in the online journal)

1. Introduction drag, have been proposed in the past, including wind tun-
nel measurements (Jones 1936, Zdravkovich 1990, Laitone
Aerodynamic loads measurements and optimisation are rel- 1997), towing force measurements (von Carmer et al 2008,
evant in a broad variety of sectors, including aeronautics and Tschepe et al 2019), on-site measurements (Le Good et al
ground transportation, animal flight, and speed sports. Several 1998, Matschke and Heine 2002, Edwards and Byrnes 2007,
approaches for measuring the aerodynamic forces, particularly Usherwood et al 2020) and numerical simulations by Compu-
tational Fluid Dynamics (Griffith et al 2014, Khayrullina et al
2015). Wind tunnel experiments often feature scaled mod-
Original content from this work may be used under the
els where the aerodynamic loads are measured with a force
terms of the Creative Commons Attribution 4.0 licence. Any
further distribution of this work must maintain attribution to the author(s) and balance (Watkins et al 1992). Uncertainties typically arise
the title of the work, journal citation and DOI. from scaling effects, relative-to-ground motion modelling,

1361-6501/21/044004+14$33.00 1 © 2021 The Author(s). Published by IOP Publishing Ltd Printed in the UK
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

model blockage and often interference of the support (Bar- data due to many varying parameters, like rider posture, bikes
low et al 1999, Páscoa et al 2012). Furthermore, some specific geometries and testing conditions.
cases such as the study of accelerating or decelerating condi- Although a number of RoF experiments have returned wake
tions (Coutanceau and Bouard 1977) are difficult to reproduce measurements that agree satisfactorily with wind tunnel data,
in the wind tunnel environment. Finally, the development of the uncertainty of the measured drag and its dependency upon
the wake over a large distance (Scarano et al 2002, Bell et al experimental conditions and the image processing parameters
2015) or the study of the biomechanics of flight (Spedding and have not yet been addressed. In particular, the errors arising
Hedenström 2010, Hedenström and Johansson 2015) are not from the set-up of the PIV measurement, the image recording
easily conducted in wind tunnels. In some of the above cases, and analysis, as well as the application of the control volume
conducting measurements with the object moving through a approach require a detailed scrutiny to support best practice of
quiescent fluid has been pursued. For instance, Henning et al experiments.
(2016) carried out particle image velocimetry (PIV) measure- In this work, linear error propagation is applied to determ-
ments in the lateral underfloor region of a full-scale high-speed ine how the uncertainty of the measured velocity field affects
train transiting at 44 m s−1 ; in their investigation, the high the drag calculation. Furthermore, the effect of the above-
speed of the train and the corresponding Reynolds number mentioned PIV parameters is investigated.
(Re = 9 × 106 ) were crucial for the so-called ballast pro- The drag resolution of the RoF is studied here during large-
jection to occur. When the primary interest of the measure- scale indoor experiments by comparison with simultaneous
ments is not the flow visualisation but rather the evaluation power meter measurements. Absolute and relative drag vari-
of the aerodynamic drag, the steady state torque test and the ations are introduced by different cyclist postures, as well as
coast-down technique are established methods of determining varying garments.
on-road aerodynamic drag of for road vehicles (Passmore and
Jenkins 1988) and of cyclists (Debraux et al 2011).
2. Drag evaluation techniques
In recent years, the steady state torque test became the most
practiced method in cycling. Power meter measurements are This section discusses the principles of drag measurement
carried out to determine the power output of the cyclist, from by use of the integrated power meter and from the Ring of
which the cyclist’s aerodynamic drag is retrieved based on Fire velocity measurements. The example of a cyclist is con-
the balance of power (Grappe et al 1997). When comparing sidered here, although the conclusions can be extended to
these on-site techniques with wind tunnel measurements, the other problems such as ground vehicles and speed sports in
advantages of more realistic flow conditions and lower costs general.
are counterbalanced by an increased uncertainty due to addi-
tional error sources, such as the less-controlled atmospheric
conditions and the physical modelling of non-aerodynamic 2.1. Power meter
resistance forces, e.g. due to rolling resistance and drive train A cyclist travelling on a flat, horizontal road imparts a
losses. Moreover, the on-site techniques adopted so far do force Fcyclist to sustain the motion contrasted by resistive
provide direct access to the aerodynamic drag, but with no forces, namely the aerodynamic drag Daero , external (Drolling )
indication of the flow behaviour. The latter is necessary when and internal frictional forces (Dfriction for the drive-train and
the physical sources of aerodynamic drag need to be identified Dbearing for the wheels). Any unbalance between these forces
and possibly minimised. results in acceleration or deceleration of the cyclist (ma).
In the recent years, a measurement concept has been intro-
duced by the authors’ group (Terra et al 2017, 2018, Spoel- ma = Fcyclist − Drolling − Dfriction − Dbearing − Daero (1)
stra et al 2019), namely the Ring of Fire (RoF), based on
large-scale stereoscopic PIV measurements past an object or a To extract the aerodynamic drag from the total drag value,
vehicle travelling through a quiescent environment. The ana- the other terms are usually modelled by use of semi-empirical
lysis of the momentum difference between the conditions prior expressions. Following the methods described in Martin et al
to and after the transit poses the basis to estimate the aero- (1998) and Lukes et al (2012), the aerodynamic drag can be
dynamic drag. This measurement approach has been used to obtained by:
estimate the aerodynamic drag of cyclists during outdoor and
Pcyclist
indoor sport action returning a quantitative visualization of the Daero = ηdrivetrain · − |{z}
ma
| {z } u
flow field in the wake (Spoelstra et al 2019). Drivetrain efficiency | {zC } Inertia
Knowledge of the drag resolution, or minimum detectable Total resistance
2
drag variation, is relevant when measurements are intended uC
− Crr (m cos α + mg sin α) − (91 + 8.7uC ) · 10−3
to perform aerodynamic optimisation. Terra et al (2018) stud- rm | {z }
| {z } Wheel bearing resistance
ied the drag resolution of a small-scale RoF system analys- Rolling resistance
ing the drag of a towed sphere in a narrow range of velocities, (2)
under the assumption of constant drag coefficient. Spoelstra
et al (2019) attempted drag estimates from a large-scale RoF. Where ηdrivetrain is the drivetrain efficiency, Pcyclist is the
The comparison with literature data could not yield a conclus- mechanical power generated by the cyclist, uC is the cyclist
ive assessment, given the large dispersion (approx. 50%) of the velocity in quiescent air, Crr is the rolling friction coefficient,

2
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

m is the combined mass of rider and bike, g is the gravita- where the cyclist moves along a curvilinear path of the radius
tional acceleration, rm is the radius of curvature for the centre of curvature rm . When the cyclist rides along a straight path,
of mass trajectory and α is the cyclist’s lateral lean angle rel- the expression of the aerodynamic drag simplifies to:
u2
ative to the horizon. The term m rmC cosα accounts for the case

Pcyclist
Daero = ηdrivetrain · ma − Crr mg − (91 + 8.7uC ) · 10−3
− |{z} (3)
| {z } uC | {z } | {z }
Drivetrain efficiency | {z } Inertia Rolling resistance Wheel bearing resistance
Total resistance

The drivetrain efficiency varies between 96% and 98% for upstream and downstream of the transiting object, it can be
power outputs in the range 50–200 W (Kyle 2001, Spicer et al shown that the viscous stresses are negligible (Kurtulus et al
2001). The rolling resistance coefficient is dependent on the 2007) and the drag force can be obtained from the surface
tyre-pressure, -loading, -diameter and -temperature, as well as integral over the inlet and outlet surfaces as illustrated in figure
the surface properties of the ground and the steering conditions 1 (Rival and Oudheusden 2017). The measurements are con-
(Burke 2003). Grappe et al (1997) and Baldissera and Delprete ducted in a fixed frame of reference (that of the laboratory)
(2016) regard the effect of speed on the rolling coefficient in where the cyclist moves at constant speed uC across the meas-
cycling as negligible and therefore use a speed-invariant Crr urement region. As discussed in the initial study of the RoF
value. (Spoelstra et al 2019), air prior to the passage of the cyclist
features inevitable small chaotic motions due to the disturb-
ances in the environment (atmospheric wind, residual motions
2.2. Ring of Fire measurement principle
from previous passages, as depicted in figure 1). The velocity
The drag evaluation through the Ring of Fire has been of such motions is denoted as uenv . After the passage of the cyc-
described in two previous studies from (Terra et al 2017, list, the flow velocity features a coherent wake with a velocity
Spoelstra et al 2019). Conservation of momentum is invoked distribution uwake as a result of the air entrainment produced by
for the stationary problem in a control volume around a mov- the cyclist. The variation of momentum written in the cyclist
ing object. The difference of momentum flux results in the frame of reference, results in the following expression for the
drag force D. When the control volume extends sufficiently instantaneous drag (Terra et al 2017):

¨ ¨ ¨ ¨
2 2
D(t) = ρ (uenv − uC ) dS + p1 dS − ρ (uwake − uC ) dS − p2 dS (4)
S1 S1 S2 S2
| {z } | {z }
1 2

1X
N
where ρ is the air density. This expression is valid at the condi-
tion that the mass flow is conserved across surfaces S1 and S2 . Dmulti = Dsingle,j (6)
N
j=1
Time averaging is performed for every single passage with the
objective of reducing the effect of the unsteady fluctuations.
where N is the number of passages.
1X
M
Dsingle = Di (t) (5)
M
i=1 3. Experimental setup and protocol
Where M is the total number of time instants composing
3.1. Test facility and subject
the measurement. It has been shown in previous works that
the unsteady behaviour in the wake of bluff objects typically Experiments were conducted in a spacious indoor facility
considered for RoF measurements (Terra et al 2017, Spoelstra (figure 2) 39 m wide and 77 m long, with flat concrete surface.
et al 2019) prevents the accurate estimate of the aerodynamic The cyclist rode loops of 190 m length in clockwise direction.
drag from one single passage. Therefore, ensemble averaging The lap can be described as two semi-circles with a radius of
from multiple passages is required to achieve statistical con- 17.8 m, connected by two 39 m long straights. The start point
vergence of the drag estimate. of each lap is located at x= 0 m.

3
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Figure 1. Schematic view of the velocity distribution before and after the passage of the cyclist.

The rider was a professional athlete; his body mass and an external torque dynamometer, the 95% confidence level
height were 79 kg and 187 cm, respectively, and his shoulder uncertainty of the SRM power meter was estimated as 2%
width was equal to 50 cm. He wore a short leg and short arm of the measured value over a range of 0–4096 W (Bertucci
time trial skin suit from Team Sunweb. Two helmet types were et al 2005). After each complete crank revolution, power and
tested as shown in figure 3. The rider wore over-shoes extend- cadence measurements were obtained. Data was recorded by
ing to half of their calves, as well as laser protection goggles. the head unit at a rate of 1 Hz after being linearly interpolated
A Team Sunweb time trial bike, model Trinity Advanced in time (Underwood 2012).
SL 2018 from Giant was used during experiments of 1.7 m
length and 8.8 kg weight. The rear wheel was a PRO Tubular
disc. The wheels mounted a Tubular Vittoria Corsa G 23 mm 3.3. Ring of Fire measurement system
tyre set at 5 bar pressure. The estimated rolling resistance coef- Velocity measurements upstream and downstream of the cyc-
ficient for this tire and conditions is Crr = 0.0045 (Bierman list were performed with a large-scale stereoscopic-PIV sys-
2016). The cyclist’s velocity was monitored with a magnetic tem based on neutrally buoyant helium-filled soap bubbles
sensor. A magnet was placed on the rear wheel of the bike and (HFSB) of 0.3 mm diameter (Bosbach et al 2009). The tracers
scanned by a magnetic sensor to retrieve information about were produced by a 200 nozzles rake installed inside the tun-
displacement and velocity. The bike GPS device stored these nel. A LaVision HFSB fluid supply unit (FSU) controls soap,
data with a frequency of 1 Hz. air and helium flow rates supplied to the seeding rake. A
The cyclist maintained a constant speed of 8.3 m s−1 , 10 m long tunnel structure of 4 × 3 m2 cross-section was
with a normalised pedalling frequency (Crouch et al 2014) of built that confines the tracers around the measurement plane.
k = 2πrf/uc = 0.15, where r is the bike crank length, f the The tunnel was built out of wooden panels integrated in an
cadence and uc the cyclist velocity. Three different configur- aluminium frame. Experiments were performed at a tracer
ations were examined (see figure 3): (1) the cyclist in upright concentration of approximately 13 bubbles/cm3 . To quantify
position with an aerodynamic helmet; (2) the cyclist in time the tracing fidelity, the tracer’s Stokes number, St , is con-
trial position with the same helmet and (3) the cyclist in time sidered, which is defined as the ratio of the tracer response
trial position with a road helmet. For each configuration, meas- time, τp , over the flow characteristic time, τf . Samimy and
urements were collected during 40 loops to build an ensemble Lele (1991) showed that a particle is a faithful flow tracer
average estimate of the aerodynamic drag from the RoF and when the condition St < 0.1 is satisfied. Based on previous
gather data from the power meter installed on the bicycle. studies from our group (Scarano et al 2015, Faleiros et al
2019), the helium-filled soap bubbles feature a tracer response
time in the order of 10 to 100 µs, yielding a tracer’s Stokes
3.2. Power meter measurement system
number in the order of 10–3 based on cyclist torso length and
The bike was equipped with an SRM Road Pro crank-spider- velocity. The light source was a Quantel Nd:YAG Evergreen
based power, widely regarded as the benchmark for power 200 laser (2 × 200 mJ at 15 Hz). A laser sheet thickness of
meter devices (Duc et al 2007, Passfield et al 2017). The 5 cm was selected to guarantee a sufficient number of tracer
device recorded concurrent measurements of the athlete’s particles in each interrogation window, as well as to comply
mechanical power output, ground velocity and cadence in with the one-quarter rule (Raffel et al 2018) when images
time. Before commencing trials, all units used during testing were recorded with ∆t = 2 ms pulse separation. Based on the
were calibrated against a zero torque reference, while ped- study of Terra et al (2019), a maximum out-of-plane velocity
als were stationary and unloaded as indicated by the man- of 5 m s−1 was expected in the cyclist’s wake, thus requir-
ufacturer. For the calculation of ground velocity, the meas- ing a laser sheet thickness of at least 4 cm to comply with
ured wheel circumference value of 2096 mm was used. Using the one-quarter rule. It should be noted that the selected laser

4
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Figure 2. Isometric view of the testing facility and measurement apparatus layout.

Figure 3. Riding configurations and helmets used. Upright position with aerodynamic helmet (left). Time trial position with aerodynamic
helmet (middle). Time trial position with conventional road helmet (right).

sheet thickness is about 1/40th of the in-plane dimensions of ∆t = 2 ms. A detailed sketch of the RoF setup is shown in
the measurement domain, which is consistent with many PIV figure 4.
experiments conducted at smaller scale (Raffel et al 2018).
Two LaVision Imager sCMOS cameras (2560 × 2160 pixels 3.4. Measurement protocol
at 50 fps, 16 bit, pixel pitch 6.5 µm) were equipped with
AF Nikkor 35 mm objectives and daylight optical filters. A The PIV system and SRM power meter were calibrated at the
lens-tilt mechanism allowed complying with the Scheimpflug beginning of each measurement day. Bubbles production was
condition for in-focus imaging in stereoscopic conditions. The initiated about two minutes before the start of each run to
lens aperture was set to f# = 8 ensuring that particles in the achieve a uniform tracer distribution with sufficient concen-
illuminated region were imaged in focus. The cameras were tration in the measurement domain. The cyclist started riding
placed 5.2 m upstream of the measurement plane at a relat- from the opposite side of the hall with respect to the measure-
ive angle of 35 degrees. The field of view captured by both ment region, accelerating to the desired speed of 8 m s−1 . At
cameras was 2.4 × 1.9 m2 , yielding a magnification factor each lap the image acquisition was triggered by the PHD, after
M = 0.0065 and a digital imaging resolution of 1.01 px mm−1 . which 40 image pairs are recorded and saved to a mass storage
The measurements were synchronised with the transit of the device before the next lap (typically 20 s). For all tests, all the
athlete using a photodetector (PHD) placed 20.5 m upstream of doors of the hall were closed to minimise externally generated
the measurement plane, which triggers the PIV system through airflows. However, as the cyclist circled the hall, the induced
a LaVision programmable time unit (PTU 9). Image pairs were air entrainment resulted in some systematic tailwind. The lat-
acquired at a rate of 15 Hz with a pulse separation time of ter effect is unwanted, first as it tends to transport the seeding

5
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Figure 4. Schematic representation of the experimental layout of the RoF.

particles out of the RoF-tunnel, and secondly because it intro- an image taken in absence of seeding. Particle intensity is
duces larger fluctuations in the air motions prior to the passage homogenised by a min/max-filter (Westerweel 1993). The
of the cyclist. This effect was mitigated by carrying a blanket two-frame recordings are interrogated with iterative cross-
through the tunnel in the opposite direction after every pas- correlation algorithm with window deformation (Scarano
sage. A movie recorded during the experimental campaign is 2001). The initial interrogation window (IW) size is at least
available online. equal to or larger than ¼ of the particles image displacement
(Adrian & Westerweel, 2011), whereas the final interrogation
window size is varied to study the effects of spatial resolu-
4. Data processing tion (section 5.3). Spurious vectors identification is based on
the universal outlier detection method proposed by Wester-
4.1. Power meter weel and Scarano (2005). To assess the out-of-plane velocity
The power output and measured bike velocity of all runs are scales that the PIV system is able to resolve, the dynamic velo-
post-processed to obtain instantaneous drag area values. The city range (DVR, (Adrian 1997)) is determined as the ratio
power meter data are synchronised with the data recorded between the maximum velocity in the near wake of the cyc-
from the PIV measurements, including the time stamp of the list (≈8 m s−1 ) and the standard deviation of the velocity dis-
laser illumination at the cyclist passage. Traces of velocity and tribution in the quiescent flow prior to the cyclist’s passage
power around the track are shown in figure 5 for the individual (≈0.03 m s−1 ). This leads to a DVR of 266.
upright measurements. Besides the individual traces of each The results are presented in the coordinate system as shown
loop, the mean value of all loops is included. in figures 1 and 2. Non-dimensional relative velocity and non-
The mean velocity and mean power during the lap vary by dimensional time are defined respectively as:
approximately 5% and 45%, respectively. The loop-to-loop
variations are up to 10% for the velocity and 70% for the uwake − uenv − |uC | t × |uC |
u∗x = t* = (7)
power. Olds (2001) and Lukes et al (2012) have discussed |uC | D
the relation between these variations and the movement of the
centre of gravity (CG) towards the centre of the track during where c = 0.5 m is the shoulder width of the cyclist and t = 0 is
corners. Due to this movement, the CG travels at lower speed the time instant when the rearmost point of the bicycle saddle
than the tyre contact point. Therefore, the comparison to the crosses the laser sheet
RoF is based on the power and velocity data recorded within
the straight segment that includes the RoF measurement sta-
4.3. Ring of Fire: control volume approach
tion (dotted box in figure 5). Within this portion, the relative
variation of mean velocity and power stays within 1% and 8% 4.3.1. Wake identification. The evaluation of the cyclist drag
respectively. via the control volume approach requires the flow velocity
An additional correction needed for the power meter needs measurements before and after the passage of the cyclist. In
to account for the velocity of the air (uenv ). The model as the ideal case when the cyclist is moving through quiescent air
described by equation (3) assumes surrounding air at rest to and the velocity measurements are noise-free, the drag estim-
calculate drag from power and velocity. The relative velocity ate is not affected by the cross-sectional size of the control
between cyclist and air is estimated by PIV measurements volume. However, in practice, environmental flow fluctuations
prior to the cyclist passage. and noise in the velocity measurements affect the estimated
drag value and, based on equation (4) and as discussed in more
detail in section 5.1, their effect increases with increasing size
4.2. Ring of Fire: particle image velocimetry
of the cross-sectional areas S1 and S2 . Hence, a dedicated wake
The recorded images are analysed with the LaVision DaVis contouring approach is applied, which isolates the cyclist’s
8 software. Background light is removed by subtracting wake from the outer flow region. Several steps are performed

6
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Figure 5. Instantaneous (thin lines) and phase-averaged (thick lines) velocity (top) and power (bottom) traces over the circuit. Data from
the upright individual test. The RoF measurement plane location is indicated with a green vertical line.

to define the wake region behind the cyclist, which are presen- through-plane particles motion (Sciacchitano 2019). In this
ted in figure 6. The wake is preliminary identified with the flow section, the uncertainty of the estimated drag is evaluated
region where the velocity is below a certain fixed percentage, based on linear error propagation for the case where the velo-
arbitrarily set to 30%, of the minimum velocity (maximum city measurements are affected by random errors ε, whereas
deficit) in the flow field. Such region is then spatially dilated by the systematic errors are negligible. The linear error propaga-
a flat disk-shaped structuring element with a specified radius. tion is performed in the wind tunnel frame of reference
The dilation length is chosen such that the entire wake and the (frame of reference moving with the model). Furthermore, two
shear layers are included in the region to be selected for the simplifying assumptions are made: (a) the upstream and down-
momentum analysis. The result is the control surface at the stream planes are sufficiently far from the object, so that the
outlet S2 , recalling equation (4). The procedure for wake con- static pressure in both planes is undisturbed and equal to p∞ ;
touring is summarised in figure 6. (b) there is a uniform inflow. Based on these assumptions, the
aerodynamic drag of the cyclist simplifies to:
¨
4.3.2. Mass conservation. The control volume analysis is D=ρ (u∞ − u) · udS (8)
based on the hypothesis that the net mass flow is zero across
A
the side and top boundaries of the domain (Anderson 2011).
Therefore, the shape and size of the inlet plane S1 (figure 1) Where u∞ is the freestream velocity seen by the cyclist,
must be adapted to ensure that the mass flow rate across S1 and u is the streamwise velocity component behind the cyclist,
is equal to that across S2 . The wake contour at each measure- in a cross-section of area A. Assuming that the latter velocity
ment plane downstream (viz. after the passage) of the cyclist component is affected by a (spatially varying) random error ε:
is adapted following the contouring approach discussed above.
As initial contour upstream, the projected wake contour at the
u = utrue + ε (9)
plane behind the cyclist is taken. The contour of the inlet plane
is then narrowed or broadened one row of vectors at the time being utrue the actual velocity in the wake of the cyclist, the
to reduce the mass flow difference from about 20% to below expression of the drag becomes:
0.1%. A graphical representation of this approach is presented ¨
in figure 7. D=ρ (u∞ − utrue − ε) · (utrue + ε) dA =
¨A ¨
5. Results
=ρ (u∞ − utrue ) · utrue dA − ρ ε (utrue + ε) dA = (10)
A A
5.1. A-priori uncertainty estimation ¨
Several error sources can affect the PIV measurements, = Dtrue − ρ ε (utrue + ε) dA
from noise in the image recordings, to peak locking and A

7
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Figure 6. Flow schematic of wake contouring procedure. Explanatory movie available online.

Figure 7. Contour adaptation based on the identification of the wake region (downstream plane) and compliance with the conservation of
mass (upstream plane; contour colour corresponds to colour framing of the wake plane).

Where Dtrue is the true aerodynamic drag, in absence of using the SIMPLE algorithm with the standard k-Omega
measurement errors on the velocity. The expression of the SST turbulence model (Wilcox 2008). The simulation is
time-averaged drag thus becomes: performed in a volume of 20 × 20 × 25 sphere diameters
¨ (W × H × L). The inlet velocity is set to 2 m s−1 , resulting
D̄ = D̄true − ρ σε2 dA (11) in a Reynolds number of 1.4 × 104 . Errors with Gaussian
A
distribution are imposed to the streamwise velocity com-
ponent in the wake plane 7.5 diameters downstream of the
Being σε2 = ε2 the variance of the velocity error, and having sphere. The relative standard deviation of the random error
assumed that error and velocity are uncorrelated: ε · utrue = 0. in the streamwise velocity (σε /u∞ ) is varied in the range
From equation (11), it follows that a random error in the velo- from 0 to 3.5%. Consistently with equation (11), figure 8 con-
city field leads to an underestimation of the drag. The latter firms that the measured drag is underestimated in presence
scales with the variance of the random error and with the area of measurement errors in the velocity, and that the measured
of integration. This result clarifies the importance of redu- drag decreases quadratically with increasing measurement
cing the region of momentum analysis to the minimum, i.e. errors in the velocity. Furthermore, the effect of the size of
only encompassing the region of deficit. It is, however, of the cross-sectional area is investigated by cropping the ori-
great importance that the domain captures the full wake for ginal measurement region from all sides. Terra et al (2018)
the entire duration of the measurement, otherwise an even lar- already identified the issue of errors arising from the size
ger underestimation of the drag may occur. For this reason, it of the domain used for the momentum analysis; the authors
is concluded that for the use of the Ring of Fire in an in-vivo showed that a reduction in cross-sectional area of the meas-
environment, the best results in terms of accuracy of the drag urement domain could potentially lead to a reduction in the
evaluation are obtained after applying a dedicated wake con- uncertainty of the measured drag by 10%. From the current
tour as described in section 4.3. analysis it is observed that, as expected, the systematic errors
In order to confirm the results from the a-priori uncertainty scale with the extent of the measurement domain considered
estimation, a Monte Carlo simulation is conducted on the for the drag estimation. If however, the domain is cropped
flow field around a sphere with diameter d = 10 cm, obtained so that part of the wake velocity deficit is cut off (area 4 in
from a steady-state solver for incompressible, turbulent flow, figure 8), consequently, the drag value is underestimated. In

8
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

the current case of the sphere, this led to an underestimation the control volume approach. Although the simplest way to
of 30% even in absence of any measurement errors enhance the spatial resolution is to reduce the interrogation
in the velocity. window size, this is accomplished at the cost of increasing
Assuming a typical uncertainty of the in-plane velocity uncertainty (Sciacchitano et al 2013). Hence, given the camera
components (σuy , σuz ) measured by the Ring of Fire sys- resolution, a compromise needs to be found between an image
tem equal to 0.1 pixel (Westerweel 1993), the uncertainty size large enough to capture the full wake and an interrogation
of the out-of-plane velocity component can be estimated window small enough to capture the small scale structures,
using equation (12) (Prasad 2000), and is equal to 1% and while still providing an appropriate signal to noise ratio to min-
0.4% of the cyclist’s velocity for θ = 20 and 45 degrees, imise the number of spurious velocity vectors as well as the
respectively. uncertainty on the estimated drag. The effects of the IW size
σu on the velocity fields and on the estimated drag are investigated
σε = √ y,z (12) for 30 runs of the baseline case (time trial posture + aerody-
2 tan θ namic helmet). The size of the IW is varied from 8 × 8 pixels2
Considering a ratio of measurement area over cyclist (8 × 8 mm2 ) to 512 × 512 pixels2 (512 × 512 mm2 ). The
frontal area of 47, then, based on the Monte Carlo simula- interrogation windows are weighted with a Gaussian function,
tion results, the measured drag is underestimated by 3.5% and the overlap factor between adjacent windows is kept con-
and 1% for stereoscopic angles of 20 and 45 degrees, stant at 75% for all cases. The details of the spatial resolution
respectively. analysis are summarized in table 1.
The velocity fields reported in figure 9 show that the use
of a large interrogation window (512 × 512 mm2 ) yields an
5.2. Effect of mass conservation
underestimation of the peak entrainment velocity in the cyc-
In previous Ring of Fire experiments (Terra et al 2017, 2018, list’s wake. The latter is caused by spatial modulation whereby
Spoelstra et al 2019), mass conservation in the measurement the cross-correlation estimation of the convex velocity distri-
plane between before and after the passage was only assumed bution produces a less-than-average value; conversely, the use
but never imposed. However, such assumption is not gener- of a small interrogation window (16 × 16 mm2 ) is not visibly
ally valid. In the current experiment, due to the rider’s motion, affected by spatial modulation, but random errors occasion-
an out-of-plane velocity of 2 m s−1 and a 0.4 m s−1 in-plane ally appear due to the spurious occurrence of region with a
motion of the surrounding air were induced in the measure- low seeding concentration. The spatial modulation in the velo-
ment plane. This led to a difference in the mass flow rate before city field has clear consequences on the drag area: over the
and after passage of the cyclist of the order of 20%. The sensit- first 5 m of the wake, the multi-passage average drag area (red
ivity of the drag estimate to the mass conservation is presented curve in figure 9-right) for the 512 × 512 mm2 IW is lower
hereafter for the cyclist in time trial position with aerodynamic than that computed with the 64 × 64 and 16 × 16 mm2 win-
helmet. In literature, the drag area of a cyclist in time trial dows, especially in the near wake where the peak velocities are
position is reported to be between 0.2–0.3 m2 (Crouch et al higher. The uncertainty of the measured drag area is approx-
2017); the value measured via power meter measurements imately constant (~0.016 m2 , or 7% of the measured value)
falls in that range, being 0.247 ± 0.008 m2 . Without impos- for interrogation window sizes between 16 × 16 mm2 and
ing mass conservation, the drag obtained by the Ring of Fire 128 × 128 mm2 , which indicates the low sensitivity to the
is largely overestimated (0.447 ± 0.015 m2 ). Instead, when PIV spatial resolution in this range of interrogation window
conservation of mass is imposed by applying the approach dis- sizes. In contrast, higher uncertainty is retrieved for smaller
cussed in section 4.3, the estimated drag area becomes equal interrogation windows (8 × 8 mm2 , uncertainty of 0.018 m2
to 0.211 ± 0.008 m2 , showing much better agreement with or ~10% of the measured value) due to the dramatic loss of
the power meter measurement. This same trend is observed the cross-correlation signal-to-noise ratio which causes large
for the other two test cases, namely cyclist in time trial posi- measurement errors in the velocity fields, as well as for larger
tion with road helmet the cyclist in upright position: without interrogation windows (exceeding 256 × 256 mm2 ) due to spa-
mass conservation the value of the drag area is overestimated tial modulation effects that cause a larger spread in drag area
by approximately 100%, whereas when mass conservation is between the different runs. Hence, the size of IW should be
imposed, the estimated drag area agrees with the power meter within 0.05 c and 0.25 c, where c is the characteristic length
measurements within 20%. scale representative for the wake topology (shoulder width
in this case). Choosing a larger IW leads to errors due to
modulation; smaller IW size, on the other hand, leads to an
5.3. Sensitivity to the measurement’s spatial resolution
increase in uncertainty due to random errors. This is, how-
The spatial resolution of the PIV technique is an important ever, very dependent on experimental settings such as seed-
parameter characterizing the overall measurement perform- ing density and pixel size of the camera. Based on the con-
ance. The PIV cross-correlation analysis with a finite interrog- siderations above, the final interrogation window size value of
ation window (IW) size is known to return a spatially filtered 64 × 64 mm2 has been selected for the results presented in the
velocity field (Raffel et al 2018); the amount of spatial filter- remainder of this work as a compromise between high spatial
ing is expected to affect the accuracy of the drag estimate via resolution and low measurement errors in the velocity fields.

9
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Figure 8. Uncertainty on drag area propagated from random error in the out-of-plane velocity for different wake plane sizes.

Table 1. Effect of the interrogation window size on the cross-correlation signal-to-noise ratio (SNR) and the estimated drag area (Cd A).

IW size [mm2 ] Image density [particles/IW] Vector pitch [mm] CorrelationSNR Cd A [m2 ] ±95% CI [m2 ]

8×8 0.5–1 2 1.6 0.191 ±0.027

16 × 16 2–5 4 1.9 0.204 ±0.018
32 × 32 8–20 8 2.1 0.203 ±0.016
64 × 64 30–70 16 3.0 0.204 ±0.007
128 × 128 100–300 32 4 0.202 ±0.007
256 × 256 500–1200 64 6 0.198 ±0.008
512 × 512 2000–5000 128 8 0.183 ±0.016

5.4. Comparison to drag estimation from power meter data these values are 0.008 m2 and 0.010 m2 , respectively, at 95%

confidence level.
The multi-passage average drag Cd Amulti of the three differ- The average drag areas per test cases computed from the
ent test cases obtained from the Ring of Fire is compared to the power meter data follow the methodology and processing steps
average drag estimated from the power meter data. The meas- explained in sections 2.1 and 4.1. The final values are presen-
urements were acquired simultaneously, so the average results ted in figure 10 together with the above mentioned values from
are obtained from the same set of samples for both the Ring of the RoF.
Fire and the power meter. The results in figure 10 can be analysed in two differ-
Firstly, the drag areas obtained from the RoF are con- ent ways, namely by assessing the relative difference of the
sidered. As was presented in section 2.2,
in order to obtain measurement techniques between each test condition, or by
the multi-passage average drag Cd Amulti per test case, first evaluating the absolute values of the predicted drag area.
the drag area in the wake of the single passages needed to be Regarding the absolute values, it is observed that the power
time-averaged to reduce the effect of the unsteady fluctuations. meter approach on average overestimates the drag by 20%
The wake is divided into two regions, namely the near and the compared to the RoF. Additionally, the drag values obtained
far wake. In the near wake region, within five characteristic with the two techniques do not agree within the respective
length scales from the cyclist, pressure effects cannot be dis- uncertainty bands. These disagreements can be ascribed to
regarded according to Terra et al (2017). Considering as char- systematic errors in both the RoF approach, as described in
acteristic length the shoulder width of the cyclist c = 0.5 m, section 5.1, and in the power meter measurements due to the
it follows that the static pressure in the flow affects the cyc- simplified power meter model (e.g. flat road) and of the uncer-
list’s drag estimate for the first 2.5 m downstream of the rider. tainty in the model constants (e.g. rolling resistance coef-
Furthermore, the rider transited the laser sheet with no pre- ficient), as discussed in section 2.1. While the latter error
defined crank-angle, meaning that the crank-angle at the laser sources affect the absolute drag estimates obtained with the
sheet location varied from run to run. Spoelstra et al (2019) power meter, they cancel out when considering relative drag
report the information of the pedal position is maintained in variations. Therefore, considering the relative performance,
the near wake, but not in the far wake due to turbulent mix- the trends of the power meter and the Ring of Fire measure-
ing of the flow. For the upright case, this led to a computed ments show good agreement, as a large-scale drag increase
drag area of 0.257 m2 , with a 95% confidence level uncertainty from time-trial to upright position is obtained. While the Ring
of 0.012 m2 . For the time-trial position, the drag are reduces of Fire predicts an increase in drag area of 0.049 m2 (23%), the
to 0.211 m2 when the rider wears a time trial helmet, and to power meter results increase by 0.066 m2 (27%). Between the
0.226 m2 when the road helmet is used. The uncertainties of two helmet types, a small-scale increase of 0.015 m2 (7%) can

10
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Figure 9. Dimensionless instantaneous streamwise velocity (u∗x ) at t∗ = 1 (left) and streamwise Cd A evolution in the wake (right). Data
from the individual test in time trial position with aerodynamic helmet. The black lines represent the wake contours used for the drag
analysis. (a) Window size: 16 × 16 [mm2 ]. (b) Window size: 64 × 64 [mm2 ]. (c) Window size: 512 × 512 [mm2 ].

11
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Figure 10. Comparison between the average drag area from the RoF and Power meter for the different configurations; uncertainty bars for
95% confidence interval.

be extracted from the Ring of Fire measurements, compared shows great potential for a range of other applications (drones,
to a delta of 0.020 m2 (8%) for the power meter approach. cars, trains, birds, …) due to the possibility to determine the
aerodynamic drag in-field rather than in the lab environment
and simultaneously obtain flow visualization.
6. Conclusions

In this work, large-scale stereo-PIV measurements are con- Acknowledgments

ducted to determine the aerodynamic drag of a moving cyclist
using the control volume approach. The flow is measured in This research is supported by the Netherlands Organisation for
the wake of a cyclist moving at 8.3 m s−1 . The single-passage Scientific Research (NWO) Domain Applied and Engineering
and multi-passage average aerodynamic drag is evaluated via Sciences (TTW), project 15583 ‘Enabling on-site sport aero-
a control volume approach along the wake behind the cyc- dynamics with the Ring of Fire’. The Team Sunweb provided
list, accounting for the non-uniform flow conditions prior to part of the cycling equipment for both experiments. We thank
the cyclist’s passage. A sensitivity analysis takes into account Albert Timmer and Floortje Mackaij riding a combined 550
key parameters of the PIV technique, namely the interrogation laps or 105 km during the experiments. Edoaro Saredi is kindly
window size and the control volume formulation, specifically acknowledged for supporting the CFD simulations.
determining the boundaries of the control volume. It is found
that applying a dedicated wake contour and imposing the con- Data and supplementary movies
servation of mass results in the most accurate drag measure-
ments. Furthermore, the IW size should be within 0.05 c and The data presented in this publication is avail-
0.25 c, where c is the characteristic length scale representative able at: doi:10.4121/uuid:c8a6c2a7-7278-43eb-b8aa-
for the wake topology. Finally, the drag values obtained with 7a0fb4b2160d. Supplementary movies are (available online at
the RoF are compared against the drag estimates from simul- stacks.iop.org/MST/32/044004/mmedia)
taneously acquired power meter data. To assess the agreement
between the two approaches in different regimes, three indi-
ORCID iDs
vidual tests are performed where small drag variations due to
different helmets as well as large drag variations due to differ- A Spoelstra  https://orcid.org/0000-0002-1876-1612
ent cyclist postures are produced. Regardless of the underlying A Sciacchitano  https://orcid.org/0000-0003-4627-3787
input parameters in the power meter model, both small- and F Scarano  https://orcid.org/0000-0003-2755-6669
large scale deltas are well captured by both the Ring of Fire
technique and the power meter approach and agree with avail-
able literature values (Blair et al 2009, Barry et al 2014, Spoel- References
stra et al 2019). The uncertainty on the average drag meas-
Adrian L, Adrian R J and Westerweel J 2011 Particle Image
urements from the RoF is within 5%. Although such value Velocimetry (Cambridge: Cambridge University Press)
is considered rather coarse when compared with state-of-the- Adrian 1997 Dynamic ranges of velocity and spatial resolution of
art force balance measurements conducted in a wind tunnel, it particle image velocimetry Meas. Sci. Technol. 8 1393–8

12
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Anderson J D 2011 Fundamentals of Aerodynamics (New York: Kurtulus D F, Scarano F and David L 2007 Unsteady aerodynamic
McGraw-Hill) forces estimation on a square cylinder by TR-PIV Exp. Fluids
Baldissera P and Delprete C 2016 Rolling resistance, vertical load 42 185–96
and optimal number of wheels in human-powered vehicle Kyle C R 2001 The mechanical efficiency of bicycle derailleur and
design Proc. Inst. Mech. Eng. P 231 33–42 hub gear transmissions Human Power 52 3
Barlow J B, Rae W H, Pope A and Pope A 1999 Low-speed Wind Laitone E V 1997 Wind tunnel tests of wings at Reynolds numbers
Tunnel Testing (New York: Wiley) below 70 000 Exp. Fluids 23 405–9
Barry N, Burton D, Sheridan J, Thompson M and Brown N 2014 Le Good G M, Howell J P, Passmore M A and Cogotti A 1998 A
Aerodynamic performance and riding posture in road cycling comparison of on-road aerodynamic drag measurements with
and triathlon Proc. Inst. Mech. Eng. p 229 28–38 wind tunnel data from pininfarina and MIRA SAE Trans.
Bell J R, Burton D, Thompson M C, Herbst A H and Sheridan J 107 967–76
2015 Moving model analysis of the slipstream and wake of a Lukes R, Hart J and Haake S 2012 An analytical model for track
high-speed train J. Wind Eng. Ind. Aerodyn. 136 127–37 cycling Proc. Inst. Mech. Eng. P 226 143–51
Bertucci W, Duc S, Villerius V, Pernin J N and Grappe F 2005 Martin J, Milliken D, Cobb J, Mcfadden K and Coggan A 1998
Validity and reliability of the powertap mobile cycling Validation of a mathematical model for road cycling power J.
powermeter when compared with the SRM device Int J Sports Appl. Biomech. 14 276–91
Med 26 868–73 Matschke G and Heine C 2002 Full Scale Tests on Side Wind Effects
Bierman J 2016 Tire Test - Vittoria Corsa G+ (Tubular) (available on Trains. Evaluation of Aerodynamic Coefficients and
at: www.bicyclerollingresistance.com) Efficiency of Wind Breaking Devices (Berlin, Heidelberg:
Blair K and Sidelko S 2009 Aerodynamic performance of cycling Springer) (https://doi.org/10.1007/978-3-540-45854-8_3)
time trial helmets (P76) The Engineering of Sport 7 (Berlin: Olds T 2001 Modelling Human Locomotion Sports Med. 31
Springer) pp 371–8 497–509
Bosbach J, Kühn M and Wagner C J E I F 2009 Large scale Páscoa J, Brojo F, Charrua Santos F and Fael P 2012 An innovative
particle image velocimetry with helium filled soap bubbles experimental on-road testing method and its demonstration on
Exp. Fluids 46 539–47 a prototype vehicle J. Mech. Sci. Technol. 26 1663–70
Burke E R 2003 High Tech Cycling the Science of Riding Faster Passfield L, Hopker J G, Jobson S, Friel D and Zabala M 2017
(Champaign, IL: Human Kinetics) Knowledge is power: issues of measuring training and
Coutanceau M and Bouard R 1977 Experimental determination of performance in cycling J. Sports Sci. 35 1426–34
the main features of the viscous flow in the wake of a circular Passmore M A and Jenkins E G 1988 A Comparison of the
cylinder in uniform translation. Part 2. Unsteady flow J. Fluid Coastdown and Steady State Torque Methods of Estimating
Mech. 79 257–72 Vehicle Drag Forces (https://doi.org/10.4271/880475)
Crouch T N, Burton D, Brown N A T, Thompson M C and Prasad A K 2000 Stereoscopic particle image velocimetry Exp.
Sheridan J 2014 Flow topology in the wake of a cyclist and its Fluids 29 103–16
effect on aerodynamic drag J. Fluid Mech. 748 5–35 Raffel M, Willert C E, Scarano F, Kähler C J, Wereley S T and
Crouch T N, Burton D, Labry Z A and Blair K B 2017 Riding Kompenhans J 2018 Particle Image Velocimetry: A Practical
against the wind: a review of competition cycling Guide (Berlin: Springer)
aerodynamics Sports Eng. 20 81–110 Rival D E and Oudheusden B 2017 Load-estimation techniques for
Debraux P, Grappe F, Manolova A V and Bertucci W 2011 unsteady incompressible flows Exp. Fluids 58 20
Aerodynamic drag in cycling: methods of assessment Sports Samimy M and Lele S K 1991 Motion of particles with inertia in a
Biomech. 10 197–218 compressible free shear layer Phys. Fluids 3 1915–23
Duc S, Villerius V, William B and Grappe F 2007 Validity and Scarano F 2001 Iterative image deformation methods in PIV Meas.
reproducibility of the ergomo® pro power meter compared with Sci. Technol. 13 R1–R19
the SRM and powertap power meters Int. J. Sports Physiol. Scarano F, Ghaemi S, Caridi G, Bosbach J, Dierksheide U and
Perform 2 270–81 Sciacchitano A 2015 On the use of helium-filled soap bubbles
Edwards A G and Byrnes W C 2007 Aerodynamic characteristics as for large-scale tomographic PIV in wind tunnel experiments
determinants of the drafting effect in cycling Med. Sci. Sports Exp. Fluids 56 42
Exerc 39 170–6 Scarano F, van Wijk C and Veldhuis L J E I F 2002 Traversing field
Faleiros D E, Tuinstra M, Sciacchitano A and Scarano F 2019 of view and AR-PIV for mid-field wake vortex investigation in
Generation and control of helium-filled soap bubbles for PIV a towing tank Exp. Fluids 33 950–61
Exp. Fluids 60 40 Sciacchitano A 2019 Uncertainty quantification in particle image
Grappe F, Candau R, Belli A and Rouillon J D 1997 Aerodynamic velocimetry Meas. Sci. Technol. 30 092001
drag in field cycling with special reference to the Obree’s Sciacchitano A, Wieneke B and Scarano F 2013 PIV uncertainty
position Ergonomics 40 1299–311 quantification by image matching Meas. Sci. Technol. 24
Griffith M, Crouch T, Thompson M, Burton D, Sheridan J and 045302
Brown N 2014 Computational fluid dynamics study of the Spedding G R and Hedenström A 2010 PIV-based investigations of
effect of leg position on cyclist aerodynamic drag J. Fluids animal flight Animal Locomotion, ed G K Taylor, M S
Eng. 136 101105 Triantafyllou and C Tropea (Berlin, Heidelberg: Springer
Hedenström A and Johansson L C 2015 Bat flight: aerodynamics, Berlin Heidelberg) pp 187–201
kinematics and flight morphology J. Exp. Biol. 218 653–63 Spicer J, Richardson C, Ehrlich M, Bernstein J, Fukuda M and
Henning A, Richard H, Kowalski T, Gries T, Huntgeburth S and Terada M 2001 Effects of frictional loss on bicycle chain drive
Loose S 2016 Full scaled high speed train PIV measurement efficiency J. Mech. Design 123 598–605
Jones B 1936 Measurement of Profile Drag by the Pitot-traverse Spoelstra A, de Martino Norante L, Terra W, Sciacchitano A and
Method 1688 (London: H.M. Stationery Office) Scarano F 2019 On-site cycling drag analysis with the Ring of
Khayrullina A, Blocken B, Janssen W and Straathof J 2015 CFD Fire Exp. Fluids 60 90
simulation of train aerodynamics: train-induced wind Terra W, Sciacchitano A and Scarano F 2017 Aerodynamic drag of
conditions at an underground railroad passenger platform J. a transiting sphere by large-scale tomographic-PIV Exp. Fluids
Wind Eng. Ind. Aerodyn. 139 100–10 58 83

13
Meas. Sci. Technol. 32 (2021) 044004 A Spoelstra et al

Terra W, Sciacchitano A, Scarano F and van Oudheusden B W 2018 vortex wakesfrom multi-camera PIV measur-ements in
Drag resolution of a PIV wake rake for transiting models Exp. free-flight laboratory Particle Image Velocimetry:
Fluids 59 120 New Developments and Recent Applications
Terra W, Sciacchitano A and Shah Y H 2019 Aerodynamic drag (Berlin, Heidelberg: Springer Berlin Heidelberg)
determination of a full-scale cyclist mannequin from pp 377–94
large-scale PTV measurements Exp. Fluids 60 29 Watkins S, Saunders J W and Kumar H 1992 Aerodynamic
Tschepe J, Nayeri C N and Paschereit C O 2019 Analysis of moving drag reduction of goods trains J. Wind Eng. Ind. Aerodyn.
model experiments in a towing tank for aerodynamic drag 40 147–78
measurement of high-speed trains Exp. Fluids 60 98 Westerweel J 1993 Digital particle image velocimetry: Theory and
Underwood L 2012 Aerodynamics of Track Cycling Doctoral, application (Netherlands: Delft University)
University of Canterbury, Canterbury Westerweel J and Scarano F 2005 Universal outlier detection for
Usherwood J R, Cheney J A, Song J, Windsor S P, Stevenson J P J, PIV data Exp. Fluids 39 1096–100
Dierksheide U and Bomphrey R J 2020 High aerodynamic lift Wilcox D C 2008 Formulation of the k-w turbulence model revisited
from the tail reduces drag in gliding raptors J. Exp. Biol. 223 AIAA 46 2823–38
jeb214809 Zdravkovich M M 1990 Conceptual overview of laminar and
von Carmer C F, Heider A, Schröder A, Konrath R, Agocs J, turbulent flows past smooth and rough circular cylinders J.
Gilliot A and Monnier J-C 2008 Evaluation of large-scale wing Wind Eng. Ind. Aerodyn. 33 53–62

14