See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/321824098

Experimental evidence of near-wall reverse flow events in a zero pressure

gradient turbulent boundary layer

Article  in  Experimental Thermal and Fluid Science (EXP THERM FLUID SCI) · November 2017
DOI: 10.1016/j.expthermflusci.2017.10.033

CITATIONS READS

15 216

13 authors, including:

Christian Willert Sricharan Srinath

German Aerospace Center (DLR) École Centrale de Lille
404 PUBLICATIONS   9,945 CITATIONS    7 PUBLICATIONS   52 CITATIONS   

SEE PROFILE SEE PROFILE

Jean-Marc Foucaut Joachim Klinner

École Centrale de Lille German Aerospace Center (DLR)
97 PUBLICATIONS   1,394 CITATIONS    98 PUBLICATIONS   472 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

APIAN project View project

Experimental techniques for the investigation of three-dimensional density fields in thermofluids View project

All content following this page was uploaded by Sven Scharnowski on 01 May 2018.

The user has requested enhancement of the downloaded file.

 Experimental Evidence of Near-Wall Reverse Flow
Events in a Zero Pressure Gradient Turbulent
Boundary Layer

C. E. Willerta,∗, C. Cuvierb , J. M. Foucautb , J. Klinnera , M. Stanislasb , J. P.

Lavalb , S. Srinathb , J. Soriac,d , O. Amilic,e , C. Atkinsonc , C. J. Kählerf , S.
Scharnowskif , R. Hainf , A. Schröderg , R. Geislerg , J. Agocsg , A. Röseg
a German Aerospace Center (DLR), Institute of Propulsion Technology, Measurement
Technology, Linder Höhe, D-51147 Köln, Germany
b LML FRE3723, Univ. Lille, Centrale Lille, CNRS, Villeneuve d’Ascq, France
c Laboratory for Turbulence Research in Aerospace & Combustion (LTRAC), Department of

Mechanical and Aerospace Engineering, Monash University (Clayton Campus), VIC 3800,
Australia
d Department of Aeronautical Engineering, King Abdulaziz University, Jeddah, Kingdom of

Saudi Arabia
e Department of Aerospace Engineering and Mechanics, University of Minnesota,

Minneapolis, MN, USA

f Institut fur Strömungsmechanik und Aerodynamik LRT-7, Universität der Bundeswehr

München, 85577 Neubiberg, Germany

g German Aerospace Centre (DLR), Institute of Aerodynamics & Flow Technology,

Göttingen, Germany

Abstract
This study reports on experimentally observed rare near-wall reverse flow events
in a fully developed turbulent flat plate boundary layer at zero pressure gra-
dient with Reynolds numbers between Reθ ≈ 2500 and Reθ ≈ 8000 (Reτ ≈
800 − 2400). The reverse flow events are captured using high magnification
particle image velocimetry sequences with record lengths varying from 50 000
to 126 000 samples. Time resolved particle image sequences allow singular re-
verse flow events to be followed over several time steps whereas long records of
nearly statistically independent samples provide a variety of single snapshots
at a higher spatial resolution. The probability of occurrence lies in the order
of 0.012 – 0.018% which matches predictions from direct numerical simulations
(DNS). The typical size of the reverse flow bubble is about 30 wall units in
length and 5 wall units in height which agrees well with similar observations
made in existing DNS data.
Keywords: particle image velocimetry, turbulence statistics, boundary layer,
flow reversal, wall shear stress
PACS: 47.80.-v 47.80.Jk 47.20.Ib 47.27.nb

∗ Correspondingauthor
Email address: [email protected] (C. E. Willert)

Preprint submitted to Experimental Thermal and Fluid Science October 23, 2017
 1. Introduction

The occurrence of near wall flow reversal and with it the presence of negative
values of the local wall shear stress τw of turbulent boundary layers (TBL) have
been subject of debate over the past decades. Eckelmann [1] postulated that
5 near wall reverse flow was not possible and experimentalists have rarely, if at
all, observed this somewhat counter-intuitive flow phenomenon. On the other
hand a variety of direct numerical simulations (DNS) suggest the opposite. For
DNS of zero pressure gradient turbulent boundary-layers (ZPG TBL) events of
negative shear stress have been reported by Spalart and Coleman [2] and also
10 for a turbulent channel flow by Hu et al. [3]. Similar observations have been
made by Lenaers et al. [4] using simulations of turbulent channel flow as well
as ZPG TBL up to shear Reynolds numbers of Reτ = uτ δ/ν = 1000. Negative
wall shear stress events are also documented in turbulent pipe flow [5]. Cardesa
et al. [6] also confirm the existence of areas of vanishing wall shear stress in DNS
15 of turbulent channel flow at Reτ = 934 and Reτ = 1834 and associate these
so-called critical points with large scale structures that extend up to 800 wall
units downstream. More recently reverse flow events have been characterized
through DNS in the adverse pressure gradient (APG) region on the suction side
of an airfoil [7].
20 Common to the observations of the DNS data is that with increasing Reynolds
number both the occurrence and the magnitude of the negative axial/streamwise
velocities increase. Lenaers et al. [4, 7] report reverse flow occurrence of 0.01%
for Reτ = 180 increasing to 0.06% for Reτ = 1000. In their DNS of fully tur-
bulent channel flow Hu et al. [3] report a probability of negative wall shear
25 (τw < 0) of 0.003% at Reτ = 90 increasing to 0.085% at Reτ = 1440.
Due to their predicted low occurrence reverse flow phenomena have only
been observed rather seldom in experiments involving ZPG wall bounded flows.
To properly capture these events long records are necessary which until recently
has only been possible for single point techniques, for instance through the use
30 of laser Doppler velocimetry in a ZPG TBL as reported by Johansson [8]. At
the same time the employed measurement technique needs to provide adequate
spatial resolution as the reverse flow structures observed in DNS data are both
short-lived and restricted to the viscous sublayer (O(5y + )). Using the micro pil-
lar shear stress imaging technique, Brücker [9] has recently been able to visualize
35 the areas of reverse flow on a flat plate turbulent boundary layer at Reτ ≈ 940.
Flow topology can nowadays be obtained through particle image velocimetry
(PIV), yet, in comparison to single point techniques, PIV is generally restricted
in acquisition frequency, number of samples and measurement uncertainty. This
can be partially overcome by restricting the camera field of view which allows
40 both an increase of sample rate and sample count [10]. The following reports
on PIV measurements in the near wall area of a TBL with a negligible pressure
gradient using sample counts exceeding 100 000 which is shown to be sufficient
to capture several instance of reverse flow events.

2
 The PIV measurements were primarily conducted to characterize the up-
45 stream conditions for a different experiment performed further downstream
within the 20 m long test section [11]. Long records, some of which are tempo-
rally resolved, enable the capture of rare events such as those described here.

2. Wind tunnel facility

The measurements were performed at the turbulent boundary layer wind
50 tunnel at the Laboratoire de Mécanique de Lille (LML). The measurement po-
sitions are located at X = 3.2 m and 6.8 m downstream of the boundary layer
trip position which is chosen as the origin of the coordinate system with the
X-axis aligned in the streamwise direction, Y is the wall-normal and Z the
spanwise direction. The tripping device is located at the junction between the
55 contraction nozzle and the 2 × 1 m2 rectangular test section and consists of a
4 mm rod attached to the tunnel wall followed by a 93 mm wide strip of coarse
sandpaper (roughness 40-grit). Full optical access to the 20 m long rectangular
test section is provided by large glass windows on all four sides.
Data was acquired at two free stream velocities of U∞ = 5 m/s and U∞ =
60 9 m/s with the wind tunnel velocities stable to within 0.5%. Temperature stabi-
lization was set at 20.0 ± 0.1 ◦ C. Table 1 provides the relevant parameters of the
turbulent boundary layer at the specific measurement conditions. The friction
velocity can be retrieved directly from the PIV measurements using the method-
ology described in [10]. Other parameters such as the boundary layer thickness
65 are partly estimated from theory (shown in parentheses) since at X = 3.2 m
only the lower portion of boundary layer was captured by the high resolution
PIV measurements. The pressure distribution obtained from discrete positions
along the centerline of the tunnel top wall is presented in Fig. 1 and exhibits
a small, favorable pressure gradient due to the acceleration of the flow by the
70 growing boundary layers on all four sides of the test section.
The recovered data is normalized with inner variables using the traditional
viscous scaling for velocity u+ + +
i (= ui /uτ ), length li (= li uτ /ν), and time t (=
2
tuτ /ν) with uτ being the friction velocity and ν the kinematic viscosity.

3. PIV measurement and data-processing

75 Two PIV measurement configurations were used to characterize the turbu-
lent boundary layer. Stereoscopic PIV captured all three velocity components in
a spanwise wall-normal plane located at X = 6.8 m downstream of the tripping
device. The field of view covered is 300 mm in spanwise direction and 180 mm
in wall normal direction and covers the full height of the boundary layer. A
80 second PIV imaging setup was aimed at capturing the flow field in the imme-
diate vicinity of the wall at a higher magnification. Following the procedures
described by Willert [10] only a narrow wall-normal strip was imaged by the PIV
camera, primarily to obtain the wall-normal velocity profile and related higher
order statistics. For these measurements the plane is aligned in a streamwise
85 wall-normal direction of about 5 mm width and 20 mm height.

3
 Table 1: Global parameters of the boundary layer experiments with estimated values given in
parenthesis

measurement location X [m] 3.2 3.2 6.8 6.8

free stream velocity U∞ [m s−1 ] 5.0 (9.0) 5.0 9.0
local free stream velocity Ue [m s−1 ] (5.2) (9.3) 5.4 9.6
boundary layer thickness δ99 [mm] (59) (54) 109 102
displacement thickness δ∗ [mm] (10.4) (9.0) 18.5 16.9
momentum thickness θ [mm] (7.4) (6.5) 13.4 12.4
shape factor H12 = δ ∗ /θ (1.405) (1.376) 1.381 1.359
wall shear rate γ̇ = ∂u/∂y|0 [s]−1 2990 8620 2750 8100
friction velocity uτ [m s−1 ] 0.213 0.362 0.204 0.350
friction coefficient cf (0.00333) (0.00303) 0.00288 0.00262
pressure gradient ∂p/∂x [Pa/m] -0.17 -0.42 -0.17 -0.42
∂p/∂x+ ×104 -2.5 -1.2 -2.5 -1.2
momentum Reynolds number Reθ = Ue θ/ν (2537) (4000) 4767 7952
shear Reynolds number Reτ = uτ δ99 /ν (825) (1293) 1477 2374
wall unit l∗ = ν/uτ µm 71.3 42.0 74.1 43.2

For high magnification PIV, two camera types were used to capture the long
image sequences. A high-speed CMOS camera with 36 GB of RAM (Dimax-
S4, PCO GmbH, Germany) captured more than 126 000 frames at 6.7 kHz to
provide continuous time records. By reducing sample rates to 1-2 kHz statistical
90 independence of the samples was improved while maintaining a similar sample
count. Additional measurements were performed using a scientific CMOS PIV
camera (Edge 5.5, PCO GmbH, Germany) which featured increased sensitivity
and higher spatial resolution. This camera was operated at a double frame rate
of 200 Hz to capture long records of statistically independent samples.
95 The roughly 5 mm wide measurement area was illuminated by a pair of
externally modulated continuous wave lasers (Kvant Laser, Slovakia) with a
combined output power of about 10 W at a wavelength of 520 nm. The non-
collimated laser beam with a size of about 6×2 mm2 was focussed into a uniform
6 mm wide light sheet using a cylindrical lens with focal length of 200 mm. The
100 resulting waist thickness was on the order of 200 µm before entering the wind
tunnel glass panel from below.
Seeding was provided globally in the closed circuit wind tunnel. Consisting
of an evaporated-recondensed water-glycol mixture, it was introduced in the
diffuser downstream of the 20 m long test section just upstream of the fan. The
105 size of the aerosol droplets is estimated at 1µm with a lifetime in the order of
10 minutes.
To provide a sufficiently high magnification at a working distance of ≈ 1.1 m
to the tunnel’s centerline, a telephoto lens (Zeiss Apo-Tessar 300 mm/f2.8) with
a 100 mm extension tube imaged the near wall region with a magnification of
110 m = 0.44. The high-speed camera has a pixel size of 11 µm which corresponds
to a magnification of 25.4 µm per pixel in object space. For the sCMOS camera,
with 6.5 µm pixel pitch, the spatial resolution improves to 14.1 µm per pixel. In
terms of viscous scaling the camera resolution varied from 1.6 to 5.2 pixel per

4
 0.1
U∞ = 5 m/s
U∞ = 9 m/s
0.08

0.06
P −P8.1m
2 ρU∞
1 2

0.04
Cp =

0.02

−0.02
0 1 2 3 4 5 6 7 8 9
X (m)

Figure 1: Streamwise pressure coefficient distribution on the wind tunnel top wall for the two
studied free stream velocities with respect to pressure measurement at X = 8.1 m.

Table 2: PIV parameters of the boundary layer experiments

Camera model PCO Dimax-S4 PCO Edge 5.5

pixel size [µm2 ] 11.0 × 11.0 6.5 × 6.5
magnification m [µm pixel−1 ] 25.4 14.1
image size H ×W [pixel] 200 × 1008 200 × 2560
field of view w×h [mm2 ] 5.08 × 25.6 2.82 × 36.1
pulse separation at 5 m/s [µs] 150 100
at 9 m/s [µs] 100 65

wall unit. To make use of nearly the full aperture of the objective lens the optical
115 axis was inclined about 2◦ with respect to the tunnel wall. At this angle the
200 µm thin light introduced through the windtunnel window has a projected
thickness of about 7 µm (0.25 – 0.5 pixel). Therefore, only a minor perspective
error is introduced by the spanwise distribution of particles illuminated by the
thin light sheet.
120 The acquired data was processed using a conventional 2-C PIV processing
package featuring a coarse-to-fine resolution pyramid with intermediate image
deformation (PIVview2C, PIVTEC GmbH, Germany). To obtain reliable mean
velocity data and statistics within close proximity to the wall a high aspect
ratio image sampling window of 64 pixels in streamwise and 6 pixels in wall-
125 normal direction was chosen. This corresponds to 1.63 × 0.15 mm2 for the high-
speed camera and 0.90 × 0.08 mm2 for the sCMOS camera. For the latter,
the sample has an effective size of 12.6 x+ × 1.2 y + at U∞ = 5 m/s increasing
to 21.1 x+ × 2.0 y + at U∞ = 9 m/s. For the detailed investigation of specific

5
 reverse flow events the PIV sample window was further reduced to 24 × 8 or
130 32 × 6. The sample overlap varied between 67% and 75%.
Estimates of the mean and unsteady wall shear rate γ̇ = ∂u/∂y were ob-
tained using a single-line cross-correlation approach. In this case the sampling
window has a wall-normal size of only one pixel, which only recover the hori-
zontal displacement.

135 4. Data analysis

To verify that the investigated flow is representative for a ZPG TBL the
following describes some relevant statistics retrieved from the processed data
sets. In this sense the normalized mean streamwise velocity profiles at position
X = 6.8 m for both Reynolds numbers are shown in Fig. 2 along with the corre-
140 sponding variances hu0 u0 i , hv 0 v 0 i and covariances hu0 v 0 i. Both plots also contain
reference data from DNS and LES of a ZPG TBL respectively provided by Sillero
et al. [12] and Eitel-Amor et al. [13]. For the most part, the agreement between
experiment and simulation is very good (the lines practically overlap). Due to
the limited field of view, the outer region of the boundary layer (y + > 400) is
145 not captured by the high magnification PIV setup. Discrepancies can be ob-
served very close to the wall for the high-Re case and can be attributed the loss
of resolution proportional to the reduction in viscous length scales at increas-
ing Reynolds numbers. Due to the coarser resolution of the conventional stereo
PIV imaging setup, reliable measurements are only possible for wall distances
150 greater than y + = 100.
Characteristic quantities such as the skin-friction coefficient cf and the shape
factor H12 listed in Table 1 respectively are within 5% and 1% of the values
suggested by Chauhan et al. [14].
Making use of the high spatial resolution near the wall, both the mean and
155 unsteady wall shear rate γ̇ = ∂u/∂y and with it the corresponding wall shear
stress τw = µ ∂u/∂y can be directly estimated from the velocity gradient at the
wall. Here the reader is referred to [10] for details on the processing scheme.
Fig. 3 shows the probability density functions (PDF) of the wall shear stress τw
extracted from several of available image sequences, which show good agreement
160 with data published in literature [15, 3, 16, 4, 17]. The vertical dashed line marks
the position for τw = 0. When plotted in log-linear form the PDF of the shear
stress exhibits instances of negative shear stress with a probability of less than
0.1%. Since the wall shear stress is directly related to the near-wall velocity,
the PDFs of the streamwise velocity u at wall distances of 1 y + and 5 y + are
165 provided in Fig. 4. Here it can be observed that the reverse flow seems to only
appear very close to the wall while it is practically absent outside of the viscous
sublayer for y > 5 y + .
At this point it should be noted that the data in the tails of the PDFs
have an increasing likelihood of being affected by measurement errors (outliers)
170 rather than representing reliable measurements. Therefore, the underlying data
sets require separate verification to determine a given datum’s validity, which,
taking into account the rather low probability of less than 0.1%, is feasible

6
 through visual inspection of the data. This can be achieved through velocity-
vs-time plots such as shown in Fig. 5. This image is compiled by extracting a
175 single column of data from each PIV data set of the sequence and placing the
columns side-by-side such that the resulting image has a width of up to 126 000
pixels, depending on the number of samples within a given data set. Therefore,
each horizontal line of pixels in the image represents the velocity record for a
given wall distance.
180 Reverse flow events can be easily detected by highlighting negative velocities
in images such as Fig. 5 and retrieving the corresponding single data sets from
the sequence for closer inspection. One such event is the white spot near the
middle of the bottom edge of Fig. 5. The spot is about 5 wall units high with
a duration of about t ≈ 7 t+ , the former giving an indication on the vertical
185 height of the reverse flow bubble (i.e. about 300µm). In the present sequence of
126 000 images only two such reverse flow events can be detected. Taking into
account the duration of about 10-15 samples per event results in a probability of
about 0.01%. Analysis of the statistically independent sampled image sequences
show between 8 and 12 reverse flow events on record lengths of nearly 64 000
190 samples, corresponding to a probability of 0.012 – 0.018%.
As the image sequences are temporally well resolved, the evolution of a
specific reverse flow event can be observed within the narrow field of view. In
this sense the particle tracks, compiled through the summation of several images,
visualize the shape of the flow structure (Fig. 6). Another way of visualising
195 the reverse flow phenomenon is shown in Fig. 7 which captures the streamwise
motion of particles at four different wall distances for a duration of 0.1 s (667
images). These space-time particle tracks are assembled by extracting a single
fixed row of pixels from each image of the sequence. Particle tracks with a steep
slope indicate slow moving particles; those at rest exhibit a pure vertical slope.
200 Near time t = 0 the particle tracks exhibit an S-shaped motion at wall distances
y = 1.2 y + and y = 3.8 y + , which indicate their brief motion in an upstream
direction. At a wall distance of y = 7.5 y + the particles briefly come to rest and
a flow reversal is not as obvious. At greater wall distances – here y = 14.9 y + –
there is no indication of flow reversal.
205 Magnified views of the flow field surrounding two reverse flow events are
provided in Fig. 8 at two different Reynolds numbers. The thick contour near
the wall at Y = 0 encloses the area with negative streamwise velocity u < 0. To
highlight the flow topology additional plots are provided with the mean local
velocity subtracted. The passage of the reverse flow region through the field of
210 view is shown by a sequence of velocity fields in Fig. 9 for which only every fourth
frame is shown. Finally snapshots of several different “separation bubbles” are
shown in Fig. 10 for the upstream measurement location at X = 3.2 m and free
stream velocities of U∞ = 5 m/s and U∞ = 9 m/s.

5. Results and discussion

215 In total eight sequences at two measurement locations and two free stream
velocities were investigated for the presence of reverse flow. While some se-

7
 quences were affected by low seeding density and limited spatial resolution, all
sequences show multiple incidences of reverse flow in the form of particles mov-
ing upstream for a certain duration. The appearance of the flow features has a
220 probability of 1.2 − 1.8 × 10−4 .
In most cases the reverse flow region has a vertical dimension of 5 y + and
a length of about 30 x+ which is consistent with the DNS results provided by
Lenaers et al [4] (e.g. see Fig. 8 in their publication). Time-resolved sequences
show that the ”separation bubble” traverses downstream through the field of
225 view at a convection speed Uc of about 0.1 Ue or Uc /uτ ≈ 2.5 (estimated from
Fig. 9) which corresponds to the mean velocity of the viscous sublayer. While not
captured through the present measurements, the corresponding DNS indicate
that the structures have a spanwise dimension on the order of 30 z + , that is,
their xz shape is roughly circular. Although the number of samples is limited
230 the flow field surrounding the ”separation bubble” can be considered to be self
similar, in particular when plotted in fluctuating velocity (mean local velocity
subtracted). The reverse flow is associated with a local deceleration of the flow
of up to 8 U + extending well into the buffer layer. Contrary to the mean velocity
profile, the local velocity profile exhibits an inflection point within the buffer
235 layer (i.e. y > 10 y + ). When plotted in fluctuating velocity a vortical structure
is present above the reverse flow patch within the buffer layer. This was also
reported by Lenaers et al [4] who state that the backflow is induced by strong
oblique vortices located above. These oblique vortex structures are believed to
be a result of streak instabilities in the turbulent boundary layer [18].
240 Aside from the rare occurrence of the reverse flow events the DNS results
by Lenaers et al [4] indicate that they appear quite sudden. While the number
of events captured with PIV are too few to allow for a statistical analysis, the
passage of the reversed flow structure through the field of view as in Fig. 9
suggests a persistence on the order of 3 ms or t ≈ 10 t+ . Here the question
245 arises whether the particles used to visualize the events can faithfully follow
the flow. The water-glycol droplets used in this investigation have a diameter
of about 1µm and a relaxation time τ ≈ 10 µs or t ≈ 0.03 t+ . In the viscous
sublayer the Stokes number St = τ U d−1 based on friction velocity uτ and
viscous sublayer thickness d = 5 y + reduces to 0.005, which indicates that the
250 particles can faithfully follow the flow reversal events.

6. Summary and outlook

Through analysis of long PIV data sequences rare events such as small-scale
near wall flow reversal could be documented at two measurement locations and
two free stream velocities of a ZPG TBL. Both the probability of occurrence
255 as well as the shape of the observed reverse flow structures agree with previous
DNS by Lenaers et al [4]. Similar reverse flow events could also be observed
in time-resolved, high resolution 3-D PTV measurement data obtained from a
ZPG TBL in a different wind tunnel facility at Reθ = 2770 (Reτ = 930, see
Fig. 8 in [19]).

8
260 In the present measurement configuration the spanwise extension of the
structures could not be measured. Multiple-camera (photogrammetric), time-
resolved techniques such as tomographic PIV [20] or 3-D PTV [21, 22, 23], or
digital holography [24] are ideal candidates to capture the fully resolved velocity
field of the small separation bubble.
265 The presented PIV measurement technique and associated post-processing
methods are believed to be valuable tools in the investigation of rare flow phe-
nomena that could not be reliably captured before. While not subject of the
present study, the available measurement data also exhibits rare strong wall-
normal velocity events very close to the wall that have been investigated through
270 DNS by Lenaers et al [4].

Acknowledgements

The authors would like to acknowledge the financial support provided through
the European High performance Infrastructures in Turbulence (EuHIT) Transna-
tional Access Program (European Grant Agreement no. 312278, www.euhit.org).
275 The LML members also acknowledge the regional councils of Nord-Pas-de-
Calais, the CNRS and the European Community for providing funding for the
refurbishment of the wind tunnel facility through the CISIT project.

References

[1] H. Eckelmann, The structure of the viscous sublayer and the adjacent wall
280 region in a turbulent channel flow, Journal of Fluid Mechanics 65 (1974)
439–459. doi:10.1017/S0022112074001479.
URL http://journals.cambridge.org/article_S0022112074001479
[2] P. Spalart, G. Coleman, Numerical study of a separation bubble with heat
transfer, European Journal of Mechanics B - Fluids 16 (2) (1997) 169–189.
285 URL http://eprints.soton.ac.uk/71967/
[3] Z. Hu, C. L. Morfey, N. D. Sandham, Wall pressure and shear stress spectra
from direct simulations of channel flow, AIAA Journal 44 (7) (2006) 1541–
1549.
URL https://doi.org/10.2514/1.17638
290 [4] P. Lenaers, Q. Li, G. Brethouwer, P. Schlatter, R. Örlü, Rare backflow and
extreme wall-normal velocity fluctuations in near-wall turbulence, Physics
of Fluids 24. doi:10.1063/1.3696304.
URL http://dx.doi.org/10.1063/1.3696304
[5] G. K. E. Khoury, P. Schlatter, G. Brethouwer, A. V. Johansson, Turbu-
295 lent pipe flow: Statistics, Re-dependence, structures and similarities with
channel and boundary layer flows, Journal of Physics: Conference Series
506 (1) (2014) 012010. doi:10.1088/1742-6596/506/1/012010.
URL http://stacks.iop.org/1742-6596/506/i=1/a=012010

9
 [6] J. I. Cardesa, J. P. Monty, J. Soria, M. S. Chong, Skin-friction critical
300 points in wall-bounded flows, Journal of Physics: Conference Series 506 (1)
(2014) 012009.
URL http://stacks.iop.org/1742-6596/506/i=1/a=012009
[7] R. Vinuesa, R. Örlü, P. Schlatter, Characterisation of backflow events over
a wing section, Journal of Turbulence 18 (2) (2017) 170–185. doi:10.1080/
305 14685248.2016.1259626.
URL http://dx.doi.org/10.1080/14685248.2016.1259626
[8] G. Johansson, An experimental study of the structure of a flat plate
turbulent boundary layer, using laser-doppler velocimetry, Ph.D. thesis,
Chalmers University of Technology, Göteborg, Sweden (1988).
310 [9] C. Brücker, Evidence of rare backflow and skin-friction critical points in
near-wall turbulence using micropillar imaging, Physics of Fluids 27 (3)
(2015) –. doi:10.1063/1.4916768.
URL http://dx.doi.org/10.1063/1.4916768

[10] C. E. Willert, High-speed particle image velocimetry for the efficient mea-
315 surement of turbulence statistics, Experiments in Fluids 56 (1). doi:
10.1007/s00348-014-1892-4.
URL http://dx.doi.org/10.1007/s00348-014-1892-4
[11] C. Cuvier, S. Srinath, M. Stanislas, J. M. Foucaut, J. P. Laval, C. J.
Kähler, R. Hain, S. Scharnowski, A. Schrder, R. Geisler, J. Agocs, A. Röse,
320 C. Willert, J. Klinner, O. Amili, C. Atkinson, J. Soria, Extensive charac-
terisation of a high reynolds number decelerating boundary layer using
advanced optical metrology, Journal of Turbulence 18 (10) (2017) 929–972.
doi:10.1080/14685248.2017.1342827.
URL http://dx.doi.org/10.1080/14685248.2017.1342827
325 [12] J. A. Sillero, J. Jimnez, R. D. Moser, One-point statistics for turbulent
wall-bounded flows at reynolds numbers up to δ + ≈ 2000, Physics of Flu-
ids 25 (10) (2013) –. doi:10.1063/1.4823831.
URL http://scitation.aip.org/content/aip/journal/pof2/25/10/
10.1063/1.4823831
330 [13] G. Eitel-Amor, R. Örlü, P. Schlatter, Simulation and validation of a
spatially evolving turbulent boundary layer up to Re θ = 8300, Interna-
tional Journal of Heat and Fluid Flow 47 (2014) 57–69. doi:{10.1016/j.
ijheatfluidflow.2014.02.006}.
URL https://doi.org/10.1016/j.ijheatfluidflow.2014.02.006

335 [14] K. A. Chauhan, P. A. Monkewitz, H. M. Nagib, Criteria for assessing exper-

iments in zero pressure gradient boundary layers, Fluid Dynamics Research
41 (2) (2009) 021404. doi:10.1088/0169-5983/41/2/021404.
URL http://stacks.iop.org/1873-7005/41/i=2/a=021404

10
 [15] S. Große, W. Schröder, High reynolds number turbulent wind tunnel
340 boundary layer wall-shear stress sensor, Journal of Turbulence 10 (14)
(2009) 1–12. doi:10.1080/14685240902953798.
URL http://dx.doi.org/10.1080/14685240902953798
[16] L. Keirsbulck, L. Labraga, M. G. el Hak, Statistical properties
of wall shear stress fluctuations in turbulent channel flows, In-
345 ternational Journal of Heat and Fluid Flow 37 (0) (2012) 1 – 8.
doi:{10.1016/j.ijheatfluidflow.2012.04.004}.
URL http://www.sciencedirect.com/science/article/pii/
S0142727X12000537

[17] N. Miyagi, M. Kimura, H. Shoji, A. Saima, C.-M. Ho, S. Tung, Y.-C.

350 Tai, Statistical analysis on wall shear stress of turbulent boundary layer
in a channel flow using micro-shear stress imager, International Journal of
Heat and Fluid Flow 21 (2000) 576–581. doi:10.1016/S0142-727X(00)
00047-3.
URL https://doi.org/10.1016/S0142-727X(00)00047-3

355 [18] W. Schoppa, F. Hussain, Coherent structure generation in near-wall tur-

bulence, Journal of Fluid Mechanics 453 (2002) 57–108. doi:10.1017/
S002211200100667X.
URL http://journals.cambridge.org/article_S002211200100667X

[19] A. Schröder, D. Schanz, R. Geisler, S. Gesemann, C. Willert, Near-wall

360 turbulence characterization using 4D-PTV Shake-The-Box, in: 11th Inter-
national Symposium on Particle Image Velocimetry PIV15, Santa Barbara
(CA), USA, 2015.
URL http://www.piv2015.org/

[20] F. Scarano, Tomographic PIV: principles and practice, Measurement Sci-

365 ence and Technology 24 (1). doi:10.1088/0957-0233/24/1/012001.
URL http://stacks.iop.org/0957-0233/24/i=1/a=012001
[21] M. Bross, C. J. Kähler, Time-resolved 3D-PTV analysis of near wall reverse
flow events in APG turbulent boundary layers, in: 18th International
Symposium on Applications of Lasers and Imaging Techniques to Fluid
370 Mechanics, Lisbon, Portugal, 2016.
URL http://ltces.dem.ist.utl.pt/lxlaser/lxlaser2016/
finalworks2016/papers/03.7_3_71paper.pdf
[22] D. Schanz, A. Schröder, S. Gesemann, ’Shake The Box’ - a 4D PTV
algorithm: Accurate and ghostless reconstruction of lagrangian tracks in
375 densely seeded flows, in: 17th International Symposium on Applications
of Laser Techniques to Fluid Mechanics, Lisbon, Portugal, 2014.
URL http://ltces.dem.ist.utl.pt/lxlaser/lxlaser2014/
finalworks2014/papers/01.5_5_328paper.pdf

11
 [23] D. Schanz, S. Gesemann, A. Schröder, Shake-The-Box: Lagrangian particle
380 tracking at high particle image densities, Experiments in Fluids 57 (5)
(2016) 1–27. doi:10.1007/s00348-016-2157-1.
URL http://dx.doi.org/10.1007/s00348-016-2157-1
[24] J. Sheng, E. Malkiel, J. Katz, Using digital holographic microscopy for
simultaneous measurements of 3D near wall velocity and wall shear stress
385 in a turbulent boundary layer, Experiments in Fluids 45 (6) (2008) 1023–
1035. doi:10.1007/s00348-008-0524-2.
URL http://dx.doi.org/10.1007/s00348-008-0524-2

12
a)

U+ = y+
U + = 1/ log(y + ) + B
 '16Reθ = 4500
/(6Reθ = 8180
63,9U∞ = 5PV = 
 63,9U∞ = 9PV B = 
µPIV, U∞ = 5PV
µPIV, U∞ = 9PV
U+








       
y+
b)

'16Reθ = 4500
/(6Reθ = 8180
h 0 0i +
uu
 63,9U∞ = 5PV
63,9U∞ = 9PV
µPIV, U∞ = 5PV
 µPIV, U∞ = 9PV
+
E
ui 0 uj 0


D

 h 0 0i +
vv
 h 0 0i +
uv
       
y+

Figure 2: Profiles scaled with inner variables of the mean streamwise velocity (a) and Reynolds
stresses (b) for two different free stream velocities obtained X = 6.8 m downstream of the
tripping device.

13
   


 


3τw 0 

3τw 0 
 


 


  

         
τw /στ
0
τw 0 /στ

Figure 3: Probability density functions of wall shear stress τw for three separate image se-
quences obtained for Reθ = 4000 (X = 3.2 m) in linear (left) and logarithmic scaling (right).

 
u + DW5y +
 
3u 0 

 
u + DW0. 5y +

 

 
     
u 0 /σu

Figure 4: Probability density functions of the fluctuating streamwise velocity u+ at wall

distances of 0.5 y + and 5 y + for Reθ = 2537 (X = 3.2 m).

14
     
7LPHt + u 0 +
 
 

 
 
 
y+

 

 
 
 
    
7LPHt +

Figure 5: Time trace of streamwise velocity u0 at U∞ = 5 m/s (Reθ = 4767, X = 6.8 m)

covering 0.75 s (5000 samples, top) and 0.075 s (500 samples, bottom). The white region at
time t+ = 0 indicates a single reverse flow event. In both sub-plots the vertical axis represents
wall distance 0 < y < 80+ (≈ 6 mm).

 

 
\>PP@

\>PP@

 

 

 
       
[>PP@ [>PP@

Figure 6: Multi-exposed particle images showing a near wall flow reversal event in a ZPG
TBL at Reθ = 4767 (X = 6.8 m, U∞ = 5 m/s). Full image view at left with detail on right.
The tunnel glass wall is located at Y = 0 with reflections of particle images visible at Y < 0.

15
 1. 2 y + 3. 8 y + 7. 5 y + 14. 9 y +







7LPH>PV@










                   
[>PP@ [>PP@ [>PP@ [>PP@
Figure 7: Particle streaks extracted at various wall-normal distances at U∞ = 5 m/s (Reθ =
4767) covering 0.1 s (667 samples; 274 t+ ) on vertical axis. A single reverse flow event is
located near t = 0.

16
 90 90
20 U + 5 U+
80 80

70 70 8
18
60 60 6

15 4
50 50

Velocity in [U + ]

Velocity in [U + ]
y+

y+
12 2
40 40
0
9
30 30
−2
20 6 20
−4

10 3 10 −6

0 0 0 −8
−20 −10 0 10 20 −20 −10 0 10 20
x+ x+
90 90
20 U + 5 U+
80 80

70 70 8
18
60 60 6

15 4
50 50
Velocity in [U + ]

Velocity in [U + ]
y+

+
y

12 2
40 40
0
9
30 30
−2
20 6 20
−4

10 3 10 −6

0 0 0 −8
−20 −10 0 10 20 30 40 −20 −10 0 10 20 30 40
x+ x+

Figure 8: Single reverse flow events at Reθ = 4767 (top) and Reθ = 7952 (bottom) shown
in viscous scaled units (left) and with mean local streamwise velocity subtracted (right).
Contours near y = 0 represent horizontal velocity U ≤ 0 at increments of 0.2U + .

17
 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20
x+ x+ x+ x+ x+ x+ x+

−20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20
x+ x+ x+ x+ x+ x+ x+

Figure 9: Sequence showing the temporal evolution of a flow reversal event at Reθ = 4767.
Contours, vector and vertical axis scaling as in Fig. 8. Red contour near the wall indicates
area of negative streamwise velocity. Temporal separation between frames is 600 µs; PIV
sampling window of 32 × 8 pixels with 75% overlap, vectors down-sampled 4× horizontally,
3× vertically. Bottom row is plotted in fluctuating velocity (mean local streamwise velocity
subtracted).

18
 a) b) c) d)
90
20 U +
80

70
18
60
15
50

Velocity in [U + ]
y+
12
40

9
30

20 6

10 3

0 0
−20 −10 0 10 20 −15−10−5 0 5 10 15 −20 −10 0 10 20 −20 −10 0 10 20
x
+
x+
x+
x+

90
5 U+
80

70 8

60 6

50 4

Velocity in [U + ]
y+

2
40
0
30
−2
20
−4

10 −6

0 −8
−20 −10 0 10 20 −15−10−5 0 5 10 15 −20 −10 0 10 20 −20 −10 0 10 20
x+ x+ x+ x+

Figure 10: Various instances of near wall reverse flow obtained at measurement position
x = 3.2 m at Reθ = 2537 (a,b) and Reθ = 4000 (c,d). Bottom row is plotted in fluctuating
velocity (mean local streamwise velocity subtracted).

19

View publication stats