V ol.:(0123456789)
1 3
Experiments in Fluids (2020) 61:146
https://doi.org/10.1007/s00348-020-02979-7
RESEARCH ARTICLE
T uf t deflec tion velocimetr y: asimple method toextrac t quantitativ e
flow field inf ormation
BenSteinfur th 1 · C.Cur a 1 · J.Gehring 1 · J. W eiss 1
Received: 10 F ebruar y 2020 / Revised: 12 May 2020 / Accept ed: 15 May 2020 / Published online: 6 June 2020
© The Author(s) 2020
Abstr ac t
A no v el method capable of assessing flow fields in a q uic k and relativ ely sim ple manner is introduced. In an extension t o t he
classical qualitativ e flo w visualization by means of co tton or pol ymer ic tuf ts, digital dat a processing is used to e xtract the
or ientation of t hese tufts. This inf or mation can be related to phy sical quantities, in particular to time- and space-dependent
v elocity signals. The capability of this method is demonstrated in tw o test scenarios. First, it is applied to g ain inf or mation on
the unsteady near -w all flow along a turbulent separation bubble. Second, the tw o-component v elocity field in the w ake of a
generic car model is measured, allo wing f or a quantification of the recirculation zone dimensions. V alidation measur ements
with con v entional techniq ues, e.g., par ticle image v elocimetry , unsteady pressure measurements and ho t wire anemometr y ,
are conducted throughout the study . These generall y suggest that the no vel appr oach pro vides a quic k and reasonabl y good
quantitativ e ov er view of the flo w configurations. Ho we ver , the measurement er ror ma y be substantial in flo w regions of lo w
v elocity or dominated b y high-frequency oscillations.
Graphic abstract
-0.5

0
0.5
1
x (mm)

50
100
150
y (mm)

0 -5 01 00 50 150
T uft Deflection V elocimetry
uU / (-)

∞
x (mm)

0 -5 01 00 50 150
Particle Image Ve locimetry
1 Introduction
For centuries, visualizing fluid flow phenomena has pla yed a
cr ucial role in unders t anding the underl ying ph ysical mec ha -
nisms. This is due to the f act that “[...] once a phenomenon
is visualized, a lar ge s tep has been taken to w ard understand-
ing it and to w ard sol ving the t heoretical and e xper imental
problems in v olv ed” (W erle 1973 ). For e xample, the visu -
alization e xper iments of Re ynolds (1883) and Mach (1887)
Electronic supplementar y material The online v ersion of t his
ar ticle ( https ://doi.org/10.1007/s0034 8-020-02979 -7 ) contains
supplementar y mater ial, which is a vailable to authorized users.
* Ben Steinfurt h
ben.steinfurt [email protected]
J. W eiss
julien.weiss@tu-ber lin.de
1 Chair ofA erodynamics, T echnisc he Univ ersit ät Berlin,
Berlin, Germany

Experiments in Fluids (2020) 61:146
1 3
146 Page 2 of 11
contr ibuted significantl y to the understanding of fundamen -
tal flow ph ysics while more recentl y , the disco very of coher -
ent structures was enabled b y flow visualization me thods
(Bro wn and R oshk o 1974 ).
As sugges ted b y Ristić ( 2007 ), flo w visualization methods
can be classified into sur f ace flo w and off-the sur face flo w
visualization, depending on the flow r egion t hat is anal yzed.
Sur face flo w visualization is pr imar ily used to obtain inf or -
mation regarding the boundary lay er state, e.g., t he transition
point, the local direction of w all shear stress and the lines of
separation and reattachment. Along with oil film visualiza -
tion, a qualitativ e anal ysis with sur f ace tuf ts is commonl y
per f or med. The model is hereb y equipped with thin fabric
tufts t hat are glued to its surface. The orient ation of indi -
vidual tufts is t hen go v er ned by the space-dependent v eloc -
ity field, and intuitiv e conclusions regarding the flo w field
can be dra wn. A shor t and incomple te list of applications
includes air plane wings and components, wind turbines,
road v ehicles and race cars (Andino et al. 2015 ; Eggles -
ton and Starc her 1990 ; Fisher et al. 1991 ; Steinfurt h et al.
2019 ). Off-the sur face flo w visualization techniq ues are used
to visualize c haracter istic flo w f eatures outside t he bound -
ar y la y er b y indicating streamlines. This can be achie v ed
b y using smoke or oil drople ts in air and dy e/ink in w ater .
Ho we ver , tuf ts ma y be emplo y ed in this scenar io as well
b y ar ranging t hem in a gr idwise manner, e.g., attached t o a
screen. For e xample, placing this screen at v ar ious positions
in the flow enabled in v estigations of the longitudinal v or tices
in the w ake of a N A CA 0012 wing (Mason and Marc hman
III 1971 ), semiwing models (McCor mick e t al. 1968 ), rec -
tangular wings and delt a wings (Bird and Rile y 1952 ).
Pre vious studies ha ve sho wn t hat there is a larg e band -
width of possible operation scenar ios f or the classical tuf t
visualization techniq ue—both in sur f ace and off-sur f ace
application. This is no doubt due to the intuitive function
pr inciple of the approach as the free ends of the tufts mimic
the major flo w characteristics. Qualitativ e statements about
the flow can then readil y be made since the tuf t deflection
is directl y linked to local v elocity vectors through the flo w
momentum.
The desire to e xtend this con v entional approach and g ain
quantitativ e inf or mation regarding v elocity fields is theref ore
natural. Suc h a dev elopment, from a qualitativ e to a quanti -
tative me thod, is by no means ne w as f or instance, par ticle
image v elocimetr y has e v olv ed from flo w visualizations wit h
smoke particles, benefited by adv ancing imaging and dat a
processing methods.
Thus, a f ew recent s tudies were aimed at g aining quan -
titative insight fr om images of tuft visualization. V ey e t al.
( 2014 ) anal yze the stall beha vior of full-scale wind turbines
using per -tuft statistics as the indicator of the local flow .
The y link the tuf t angular standard deviation to the local
turbulence le vel whic h allow s f or t he determination of
separation lines. Wieser e tal. apply a similar me t hodology
to a realistic car model (W ieser etal. 2015 ) and full-scale
road v ehicles (Wieser et al. 2016 ). Here, mean tuf t angles are
also used to per f or m line integ ral con v olution (LIC). Their
results sho w good agreement between the calculated surface
traces and oil film visualization as w ell as CFD simulations
(Wieser e t al. 2016 ). Swytink -Binnema and Johnson ( 2016 )
determine t he blade stall from captured videos of the tufts
applied to a wind turbine blade. Using luminescent mini-
tufts, Chen et al. conduct a quantitativ e analy sis of the tuft
inclination angle on a flat plate (Chen et al. 2019 ) and a
bac kw ard-f acing step (Chen et al. 2020 ).
While all af orementioned studies aimed at e xtending t he
classical tuft visualization techniq ue to extr act quantitativ e
flo w field inf or mation, a method that can be easily adapted
b y other researc hers has not been presented to the best of our
kno w ledg e. Fur t her more, there appear to be no attempts to
relate the tuft deflection to t he momentum of the flow and
thus allow f or the measurement of the flow v elocity .
The objectiv e of this paper is t heref ore to introduce a
method that is easily adaptable b y other researc h g roups and
allo w s to e xtract quantitativ e v elocity data from captured
tuft images without the need f or commercial software. The
basic pr inciple is demonstrated b y means of tw o e xamples,
representing surface and off-surface applications. F irst,
unsteady near -w all quantities are obtained in a turbulent
separation bubble (TSB) b y affixing tuf ts to the wall in a
one-sided diffuser . Then, the turbulent w ake of an Ahmed
body is in ves tigated b y means of a tra v ersable probe with a
larg e number of tuf ts.
2 Ev aluation routine
The basic e valuation r outine f or tuft deflection velocimetry
(TD V) is ex emplar ily displa y ed in Fig. 1 . Note that a mini-
mum w orking e xample of the algorit hm is also pro vided with
the supplement ar y mater ial. Follo wing the application of tuf ts
inside a giv en flo w field, the first step in obtaining quantitativ e
inf or mation consists in the recording of images. A prior i, the
method is not limited to coherent snapsho ts but t he acquisition
rate of the camera must be adjusted to the time scales associ -
ated with relev ant unsteady flo w f eatures in order to study the
dynamic flo w beha vior . Sufficient lighting is requir ed so t hat
tufts are captured in their entirety . This can be par ticularl y
challenging in the case of high acq uisition rates and lo w expo -
sure times. In practice, tufts of UV activ e mater ial illuminated
with appropr iate LED ar ra ys are successfull y emplo yed in this
scenar io. Based on recorded images, the g eometries of indi -
vidual tufts can be extracted. Since the y are typicall y v er y thin
compared to their length, the tuf t edges can be considered a
satisfying representation of the tuf t deflection. T o e xtract these
edges, the intensity gradient field is ev aluated. While t here

Experiments in Fluids (2020) 61:146
1 3
P age 3 of 11 146
are se v eral different approaches, w e f ound t he Pr e witt method
( 1970 ) to be the most accurate and cos t-efficient f or the dem -
onstration tes t cases addressed in the f ollowing.
The Pr ewitt operator employ s tw o
3 × 3

kernels that are
con vol ved with an imag e to appro ximate the intensity der iva -
tiv es in hor izontal and vertical direction, respectivel y . This
separable approac h makes it efficient but ma y also cause inac -
curacies when dealing with high-frequency v ar iations in the
intensity function. Theref ore, a sufficient contrast needs to be
ensured betw een tuf ts and back ground. In experiments, this
is typicall y achie v ed b y applying a blac k f oil or coating to the
bac kground if required. The Pr ewitt filter then retur ns a zero-
gradient f or such a bac kground whereas the inter face be tween
tuft and back ground is detected as an edge and will be re tur ned
in a binar y image (s tep 2 in Fig. 1 ). This binary image then
undergoes a Hough tr ansf or mation (step 3), a robust s tandard
procedure to e xtract parametr izable geometries (Duda and
Har t 1972 ). In t he case of line extr action required f or TD V ,
the Hough space is spanned by the Euclidean distance of the
line
𝜚

and the angle betw een the per pendicular of the line and
the abscissa
𝜗

. Thus, the line can be represented by its Hesse
nor mal f or m
(1)
𝜚 = x cos 𝜗 + y sin 𝜗 .

It is w or th noting that t he angle giv en in the Hesse nor -
mal f or m already is an indicator f or the tuf t or ient ation.
Ho we ver ,
𝜗

depends on the dimensions of ra w images
since the location of the abscissa wit h respect to the tuf t/
e xtracted line ma y be varied. Theref ore, we emplo yed
a trac king method to detect the end points of the para -
metrized lines (step 4) and computed the tuf t deflection
angle based on their locations. For cer tain applications
such as the second demonstr ation ex ample in this ar ticle,
it is also necessar y to compute the length of t he detected
lines which inher ently req uires kno w ledge regar ding t heir
end points as w ell.
Repeating this procedure f or multiple successiv e
images, t he time-dependent deflection angle of tufts can be
der iv ed f or fur ther st atistical anal ysis of a giv en flow field.
3 Choosing tuf ts withappropria te
proper ties
The accuracy of TD V is directly dependent on the c har -
acter istics of the emplo y ed tuf ts. A br ief summar y of the
main parameters that need to be considered f or tuf t selec -
tion is giv en in this section.
Fig . 1 Flo w diag ram visualizing the ev aluation of the deflection angle of one tuf t

Experiments in Fluids (2020) 61:146
1 3
146 Page 4 of 11
3.1 T uft visibility
Being an optical measurement tec hnique, good visibility
of the tuf ts is requir ed, ensur ing t hat the tuf t geome tr y is
adequatel y represented in pixel space. F or a giv en lighting
setup, tw o parameters ma jorl y affect the visibility: (a) the
tuft t hic kness and (b) its color/reflectivity .
T o reduce t he influence on the flow field, tufts of mini -
mum dimensions are pref er red. This minimum extent
strongl y depends on t he optical resolution of the measure -
ment setup, i.e., the size of the measurement domain and
the camera chip. A lo w resolution can be compensated by
a br ight tuft color which is e x emplar ily demons trated in
Fig. 2 where tufts of different mater ials but similar diam -
eters are sho wn. Whereas t he images of transpar ent tuf ts of
pol yamid (P A) monofilament and n ylon are associated with
lo w contrast, the other tufts clearly r eflect a g reater amount
of light and are theref ore better suited f or the measurement
techniq ue.
3.2 Static beha vior oftufts
For a specific flo w momentum, the deflection angle of a tuft
is determined by its length and stiffness. A gain, long er tufts
are not desired due to t he greater influence on the flow field
and because the velocity inf or mation is integrated across
greater spatial scales. On the other hand, shor t tufts are typi -
call y not suited f or low -speed measurements as t he deflec -
tion angle is small compared to the measurement accuracy
in this case.
The char acter istic deflection cur v es f or tufts of lengt h
l = 10 mm

are sho wn in Fig. 3 . T o compile t hese cur v es,
a nozzle w as supplied with a constant mass flow rate,
resulting in a steady flo w approaching the tufts. The noz -
zle diameter w as larg er than t he tuft lengt h, and it w as
ensured that the entire tuf t w as located inside the poten -
tial core of the steady jet. Se veral tufts w ere tested f or
each material, which enabled calculating bo th t he mean
deflection angle and the relative scatter f or each tuft, indi -
cated as er ror bars. Ev en though tuf ts were nominall y
or iented per pendicular to the nozzle axis in t he absence
of the jet (
U ∞ = 0 m/s

), small offse t angles occur red
occasionall y . These were de ter mined wit h the introduced
method and subtracted from the deflection angles meas -
ured subsequentl y .
Due to their g reater stiffness, the smallest deflection
angles are f ound f or poly amide (P A) and ny lon. How ev er,
the range of deflection angles is muc h greater f or n ylon
where the variance at
U ∞ = 20 m/s

is 98% of the mean
v alue
𝜑 ≈ 17.1 ◦

. The relativ e scatter f or P A is 32% at the
same inflo w v elocity . The poly ester (PES) tufts exhibit a
similar total v ar iance of up to
Δ 𝜑 = 10 ◦

but the relativ e
v ar iation is smaller due to greater ov erall deflection angles
(24% at
U ∞ = 20 m/s

). The same is true f or t he cotton
tufts t hat are associated with the larg est deflection angles,
yielding a relativ e de viation of 12% at
U ∞ = 20 m/s

. Ev en
though t his resembles the best reproducibility of tuft mate -
r ials assessed in this study , the v alue is still inf er ior com -
pared to other quantitativ e measurement techniques. The
relativ el y low degree of reproducibility can be attributed
to a number of causes whic h must be addressed in more
detail in future studies. Firs t, the tuf ts used in t his study
are not g eometrically identical, mainl y due to slight pre-
def or mations induced when handling and cutting them.
Fur ther more, manufacturing imper f ections affect t he tuft
proper ties in the case of braided tufts. In addition to t hat,
they w ere affixed with adhesiv e tape and t he identical
or ientation cannot be ensured in the process. Ho we v er ,
v ar ying tuf t characteristics can be compensated b y estab -
lishing different calibration la ws f or individual tufts. This
w as done f or t he second demonstration case presented in
Sect. 4 .
Among the assessed mater ials, cotton tufts e xhibit t he
most f a v ourable characteristics and will be emplo y ed in the
f ollowing.
PA

(white)
PA

(monofil.) Nylon
PES

(neon, white, UV active) Cotton
Fig . 2 V isibility of different tuft mater ials of similar diameter f or
equal lighting conditions
0
10
20
30
40
PA (white)
Nylon
PES (UV active)
Cotton
ϕ
(°)
0 5 10 15 20 25
U
∞

(m/s)
Fig . 3 T uft deflection angle as a function of velocity f or different tuft
materials

Experiments in Fluids (2020) 61:146
1 3
P age 5 of 11 146
3.3 Dynamic behavior oftufts
Since a steady approac h flo w only reflects the conditions
of a limited number of flo w configurations, fur ther gener ic
tests w ere conducted to address the dynamic beha vior of
tufts exposed to uns teady flo w . F or t his pur pose, a pulsed
jet actuator consis ting of a solenoid val ve and a nozzle w as
emplo y ed and star ting jets w ere emitted at constant freq uen -
cies. Figure 4 sho ws the time signal of the deflection angle
of a cotton tuft of
l = 5 mm

(open circles) f or a frequency of
f = 50 Hz

and the velocity signal measur ed wit h a hot wire
sensor as a ref erence method. The tuf t is clearl y capable
of f ollowing the dominant frequency . This w as obser ved to
be the case f or frequencies up t o
f = 80 Hz

. Theref ore, a
dynamic response req uired f or man y lo w -speed applications
can be attested. Ho we ver , it is w or t h pointing out that t his
generic test configuration does not reflect highl y turbulent
flo w s wit h flo w f eatures cor responding to larg e frequency
bands. The suitability f or these cases may be addressed in
future applications.
4 Demonstra tion ofapplicabilit y
T w o test cases wer e in v estigated with the proposed measure -
ment techniq ue to demonstrate its capabilities: (1) the near -
w all flo w along a turbulent separation bubble (TSB) and (2)
the w ake of a g ener ic car model.
4.1 T est c ase 1: turbulent separa tion bubble
TSBs are g enerated when a turbulent boundar y la y er sepa -
rates from the w all and reattaches furt her do wnstream,
enclosing the separation zone. The f lo w separation can
either be induced by a g eometric singular ity , such as a bac k -
w ard f acing step (e.g., Driver e tal. 1987 ; Ma and Schr öder
2017 ) or f ence (e.g., Ruder ich and F er nholz 1986 ; Hudy and
N aguib 2003 ) or b y an adverse pr essure g radient (APG).
In this test case, a pressure-induced TSB w as in v estigated
where the APG w as introduced b y the widening of t he test
section cross area. A similar flo w configuration w as in ves -
tigated in W eiss et al. ( 2015 ) and Mohammed- T aif our and
W eiss ( 2016 ) on t he basis of unsteady pressure, HW A and
PIV measurements.
In the cur rent study , the free stream v elocity was
U ∞ = 20 m/s

. The mean flo w field of the TSB addressed in
this study is sho wn in Fig. 5 . The time-a verag ed recircula -
tion zone associated with the TSB is enclosed by the sepa -
rating streamline
 u = 0 m/s

, spanning a region of roughl y
x ≈ 0.1 … 0.45 m

.
Cotton tufts (
l = 15 mm

) w ere attached to the test sec -
tion surface inside the symme tr y plane with a spacing of
Δ s = 25 mm

as indicated in the sample photograph y shown
in Fig. 5 . A second line of tufts with free ends pointing in t he
other direction w as used f or ref erence. While they sho wed
good agreement with tuf ts placed on the symmetr y plane,
w e will not address them in much detail here. The main
0
10
20
30
40
t (s)
U

∞
(m/s)

0.02 0.04 0.06 0.08
100

75
50
25
0
ϕ
(°)
Fig . 4 Dynamic beha vior of tuft subjected to
f

=
50 Hz

pulsed jet,
hot wire signal sho wn in gray color f or ref erence
0
1
Side view
u = 0 m/s
0
-0.1
0.1
y (m)

0.5
u U / (-)

∞
U ∞
α = 21°
x (m)
Diffuser ramp Surface tuft array
00 .1 0. 20 .3 0. 40 .5 0.6 -0.1 0.7 0.8
0
z (m)
T op view
U ∞
U ∞
Diffuser ramp
Fig . 5 T SB studied by means of TD V ; time-a v eraged stream wise
v elocity field measured with PIV (top), application of sur face tufts in
symmetry plane (bottom)

Experiments in Fluids (2020) 61:146
1 3
146 Page 6 of 11
objectiv e in this scenar io was t o identify t he instantaneous
flo w direction and detect unsteady regions of re verse flo w in
par ticular . Theref ore, t he tufts were attached perpendicular
to the main flow dir ection. Images of the tufts were taken
at an acquisition rate of
f = 200 Hz

o v er a time duration of
Δ t = 10 s

. Due to the size of the relev ant flo w region, the
tuft ar ra y w as divided into f our segments of similar lengt h
of which imag es were taken subseq uently . T o compensate
the tilted diffuser ramp sur f ace, the camera was adjus ted so
that t he optical axis w as parallel to the w all nor mal f or all
segments.
With initial tuft angles being transv erse to the free stream
direction, def lection angles of
𝜑> 0 ◦

are e xpected f or
regions of f or ward flo w , whereas flo w recirculation yields
angles of
𝜑< 0 ◦

. It is w or t h mentioning that the locations
of separation and reattachment are no t constant but fluctuate
significantl y . This is explained b y t he finding that TSBs are
char acter ized by a lo w-freq uency contraction and e xpansion
in addition to medium-freq uency oscillations associated wit h
v or tex s tr uctures inside the separated shear lay er and high-
frequency fluctuations caused b y t he turbulent nature of the
flo w (Mohammed- T aif our and W eiss 2016 ). Theref ore, the
test case is suited to assess the capability of the TD V method
to measure unsteady flo w conditions in close w all pro ximity .
Figure 6 sho ws the time-a v eraged deflection angles o f
tufts along t he diffuser axis that are based on the time sig -
nals of tuft deflection angles obtained wit h the procedure
introduced in Sect. 2 .
As e xpected, the region upstream of the diffuser ramp
(
x < 0m

) is characterized by positiv e deflection angles of
𝜑> 15 ◦

cor responding to the f or w ard-flow boundar y lay er .
Fur ther downs tream, the flow decelerates and the tuft deflec -
tion angles decrease accordingl y . A mean v alue of
𝜑 ≈ 0 ◦

,
i.e., an a verag e tuft or ient ation per pendicular to the diffuser
axis, is f ound at
x ≈ 0.15 m

while negativ e deflection angles
are obser v ed in the range of
x = 0.15 … 0.45 m

due to flo w
recirculation in this region. This cor responds reasonabl y
w ell wit h the mean separation line, extracted fr om PIV
measurements, that is sho wn in Fig. 5 . Downs tream of the
TSB, v or tex structures are shed onto the w all and positiv e
deflection angles are observed.
Being a c haracter istic quantity f or the dimensions of
TSBs, the f or ward flo w coefficient
𝛾 ( x )

w as deduced. This
quantity repr esents t he time per iod dur ing which the flo w is
or iented in stream wise direction (
u > 0 m/s

) relative t o th e
total obser v ation time (Simpson 1996 ). Here, it w as appro xi -
mated b y the number of samples where
𝜑> 0 ◦

relativ e to the
total number of samples (Fig. 7 ).
The e valuation of tuft angles yields tw o regions of
𝛾> 99 %

, being
x ≤ 0m

and
x ≥ 0.5 m

. The respectiv e
boundar ies mark the positions of incipient detachment and
complete r eatt achme nt as defined by Sim pson ( 1996 ). The
cor responding length of t he enclosed TSB is
L 99 ≈ 0.5 m

which is about 20% smaller than the v alue measured with
PIV where the complete reattachment point w as not inside
the measurement domain but a location of
x ≈ 0.6 m

can be
assumed. The main reason f or t his de viation is t he difference
in f or ward flo w coefficients at t he do wnstream end of the
TSB. Whereas PIV measurements re veal a number of ins tan -
taneous snapshots where recirculation occurs, this inf or ma -
tion is not cap tured by T D V . The negativ e stream wise v eloc -
ity component is induced b y v or tex s tr uctures or iginating
from the separated shear la y er and the time duration may
be too shor t to cause negativ e deflection angles. How e ver ,
smaller tuft dimensions may be able t o render this process
more accuratel y . The same effect is present inside the TSB
30
20
10
0
-10
0 0.2 0.4 0.6 0.8
x (m)

φ (°)
U ∞
φ
Fig . 6 Mean tuft deflection angles along diffuser axis
0
20
40
60
80
100
0 0.2 0. 40 .6 0.8
x (m)

γ( %)
TDV PIV
Fig . 7 Forw ard flow r atio along diffuser axis deter mined by means of
TD V and PIV , respectiv ely

Experiments in Fluids (2020) 61:146
1 3
P age 7 of 11 146
where PIV measurements sugg est f or ward flo w coefficients
of
𝛾 PIV ≈

15
%

whereas tuft deflection angles are negativ e at
almost all times, hence
𝛾 T DV ≈ 0 %

. Ho we ver , the locations of
transitor y detachment and r eatt achment mar ked b y
𝛾 =

50
%

are w ell captured by the tec hnique introduced in this ar ti -
cle, and a TSB length of
L 50 ≈ 0.38 m

is der iv ed both from
PIV and TD V measurements. It should be noted, ho we ver ,
that the location of transitor y detachment as derived from
the f or w ard flo w ratio is slightly ups tream of the location
where
𝜑 = 0 ◦

. This can be explained b y t he f act that close
to the location of transitor y detachment, the magnitude of
deflection angles is lar ger in the case of f or war d-flow e ven
though the respective time duration is eq ual to that of rev erse
flo w . Since t his w as not observed in PIV measurements, w e
attr ibute this finding to individual tufts having a pr ef erential
bending direction.
Finall y , the fluctuation le vel of tuft deflection angles is
e valuated along the TSB. As a ref erence, unsteady pressur e
measurements are conducted with piezo-resistiv e pressure
transducers. Both distributions are sho wn in Fig. 8 . In good
agreement with results in t he literature (Mohammed- T aif our
and W eiss 2016 ), two local peaks are f ound close to the
mean locations of separation and reattachment. This is also
reflected b y the TD V anal ysis where maxima of the order
of
r ms ( 𝜑 � )≈ 6 ◦

and
r ms ( 𝜑 � )≈ 13 ◦

are f ound, respectivel y .
Generall y , the first test case sho ws that the method is w ell
suited to adequatel y detect the instant aneous flo w direction
when applied in the near -w all region. Fur t her more, an anal y -
sis of unsteady flo w was successfull y conducted to deter -
mine regions of maximum fluctuation le vel inside a flo w
configuration. The freq uency peak associated wit h these
fluctuations w as in the region of
f ≈ 60 … 80 Hz

. Ho we ver ,
er rors occur when the flo w is dominated by h igher -frequency
flo w rev ersals.
4.2 T est c ase 2: wak e ofgeneric car model
A second test configuration w as chosen to assess the perf or -
mance of the TD V method when a free velocity field is ana -
l yzed, i.e., when the tuf ts are not attached to a solid surf ace.
For this pur pose, an Ahmed body—a generic car geometry
introduced in Ahmed etal. ( 1984 )—w as studied inside the
same lo w -speed wind tunnel as the TSB (test case 1) at a
free stream v elocity of
U ∞ = 20 m/s

. Based on the height
of the
50 %

scale model
H ≈ 0.14 m

, t he Re ynolds number
w as
Re H ≈

1.9 ⋅ 10
5

. The w ake topology subject to a v ar ied
slant inclination angle
𝛽

is w ell known fr om previous s tud -
ies (Ahmed et al. 1984 ). The smallest drag coefficient is
f ound at
𝛽 ≈ 10 ◦

where the flow is attac hed on the slant sur -
f ace. The drag then increases f or greater inclination angles
due to the enhancement of stream wise v or tices or iginating
at the c-pillars. At a slant angle of
𝛽 ≈ 30 ◦

, t hese v or tices
break do wn and a discontinuous drop in drag occurs. While
the larg e-scale stream wise v or tices will not be addressed in
this study , w e emplo y ed TD V to in ves tigate the influence
of the slant inclination on the dimensions of t he recircula -
tion zone inside the symmetr y plane. The setup f or this test
case is schematicall y depicted in Fig. 9 . T w o slant angles,
𝛽 = 15 ◦

and
𝛽 = 35 ◦

, were in v estigated and a probe with
50 cotton tufts of
l = 5 mm

length w as tra v ersed in the
symmetry plane in
Δ x = 4 mm

steps along the x direction
betw een
x =− 100 … 200 mm

where
x = 0 mm

marks the
base of the Ahmed body . The probe consisted of a solid
beam with dr illed holes t hrough whic h the tuf ts were s tic ked
and glued on the back side of the probe. The v er tical spacing
betw een neighbor ing tufts was
Δ y = 4 mm

. The optical axis
of the camera w as nor mal on the symmetr y plane (Fig. 9 ) ,
and images w ere t aken o v er a duration of
Δ t = 3s

with an
0
3
6
9
12
15
0 0.2 0.4 0.6 0.8
x (m)

rms( ‘) (°) φ
0
0.04
0.08
0.02
0.06
0.1
rms( ) (-) c‘
p
TDV
Pressure transducer
Fig . 8 Fluctuation le v el along diffuser axis determined by means of
TD V and piezo-resistive pr essure transducer, r espectivel y
Ahmed body
T raversable TDV probe

with 50 tufts
U ∞
Splitter plate

Camera
x
y
Fig . 9 Se tup f or flow field measurement of Ahmed body w ake b y
means of TD V

Experiments in Fluids (2020) 61:146
1 3
146 Page 8 of 11
acquisition rate of
f = 400 Hz

. This w as sufficient to resol v e
rele vant shedding modes that are associated with a Str ouhal
number of
St ≈ 0.22

at
𝛽 = 35 ◦

according to T una y et al.
( 2014 ), cor responding to a frequency of
f ≈ 30 Hz

.
Measurements w ere conducted inside the symmetr y plane
where the out-of-plane component is assumedl y negligible.
The initial tuft or ient ation w as parallel to the z axis, and the
tufts were free to mo ve in tw o directions, namely x and y
in this application scenar io. The ev aluation algor ithm w as
theref ore e xtended to compute tw o separate deflection angles
cor responding to both in-plane v elocity components in the
symmetry plane. As indicated abov e, t he initial or ientation
of the tuf ts w as almost parallel with reg ard to the optical
axis, resulting in almost zer o lengt h when imaged b y t he
camera. The projected length of the tufts w as used to obtain
deflection angles. Ag ain, 2D2C PIV measurements wer e
emplo y ed as a ref erence measurement techniq ue. Here, mean
v alues of the two v elocity components wer e obt ained based
on 500 incoherent snapshots. The dis t ance betw een v ectors
measured with PIV w as
Δ x =Δ y ≈ 1 mm

, yielding a f our
times finer resolution than f or TD V measurements.
Since the preliminar y analy sis sho wed that the charac -
ter istic tuft deflection cur v e is not identical f or different
tufts, an initial calibration w as per f or med by appl ying dif -
f erent flow v elocities rele vant to this tes t case and measur -
ing the individual deflection angles in the absence of t he
Ahmed body . The characteristics of individual tufts were
then mapped with second-order polynomials sho wn in
Fig. 10 . Since the cotton tufts used f or this test case were
shor ter than t hose addressed in Fig. 3 , the deflection angles
are smaller at eq ual veloc ities. The tw o in-plane v elocity
components u and v were computed based on the individual
calibrations where the free stream v elocity was v ar ied in
the range
u = 0 … 25 m/s

(Fig. 10 ). Identical characteristics
w ere assumed f or negativ e velocities in main s tream direc -
tion and f or t he v er tical v elocity component. In hindsight,
this assumption is probabl y not justified as a pref erential
direction of the tuf ts w as indicated b y results of the previ -
ous test case where different deflection angles occur f or the
same f orcing.
Generall y , the additional effor t required f or this calibra -
tion w as minimal since no modification to the experimental
setup w as needed in ter ms of the optical ar rang ement.
The main results of this experiment are sho wn in Fig. 11
where the contour field of the nor malized stream wise veloc-
ity component is o v erla y ed with t he time-a verag ed stream -
line patter n that is based on the measured vector fields.
The longitudinal e xtent of the detected recirculation
zones measured with TD V ag rees well with the length
obtained wit h PIV measurements. The stagnation point is
located at
x ≈ 150 mm

f or t he
𝛽 = 15 ◦

slant angle whereas
a longer r ecirculation zone is f ound f or
𝛽 = 35 ◦

where the
stagnation point is at
x ≈ 180 mm

. Fur t her more, the upper
par t of the vortex ring rooted at the base is captured b y the
method. Ho we ver , t here is a lar ge difference in the v elocity
magnitude inside the recirculation zone determined by both
methods. The approac h introduced in this ar ticle suggests a
minimum longitudinal v elocity of
u ≈− 0.6 U ∞

compared to
u ≈− 0.4 U ∞

obtained from PIV measurements. In general,
there is a systematic difference be tween bo th measurements
of the order of
Δ u =( u PIV − u T DV )∕ U ∞ > 0.1

inside the
recirculation zone. A significant de viation of opposite sign
in ter ms of the stream wise v elocity inside t he shear la y ers
bounding the recirculation zone is also obser v ed. This ma y
be attr ibuted to the small v elocity magnitude
| u |

→
0 m/s

present in these regions. As a result of this inaccuracy , the
recirculation zone measured with TD V is systematicall y
smaller in v er tical direction.
None theless, t he second test case demonstrates that the
TD V techniq ue can be emplo yed t o gain a quantitativ e
impression of a 2D2C flo w field with a measurement setup
that is much easier , cheaper and simpler to establish com -
pared to PIV measurements. The measurement accuracy
ma y be adequate f or a preliminar y assessment of comple x
flo w s, t hough it is probabl y inacceptable f or detailed analy -
ses. This insufficiency is mainl y linked with the mater ial
proper ties of the cotton tufts used in this study .
5 C onclusions
The main objectiv e of this ar ticle w as to introduce a no v el
measurement method that allo ws f or a quic k and simple
quantitativ e anal ysis of flo w fields. T w o test configura -
tions ha ve demons trated the potential of t his approac h.
Ho we ver , t he cur rent limit ations preclude a replacement of
0
10
20
30
40
φ (°)
05 10 15 20
U
∞

(m/s)
Fig . 10 Second-order pol ynomials for v elocity calibration of individ-
ual tufts, mean value printed bold

Experiments in Fluids (2020) 61:146
1 3
P age 9 of 11 146
con ventional measurement tec hniques. W e theref ore hope
that the approach is adapted and modified b y other researc h -
ers, gradually optimizing the method as has been done with
other measurement tec hniques in the past.
5.1 Adv antages anddisadv antages
The main adv ant age of TD V lies in its q uick and sim ple
implementation. No elaborate measurement eq uipment
is requir ed and the dat a processing is f ast compared t o
other techniq ues f or flow field measurements. A camera
is on hand in practicall y all research f acilities, and the
data analy sis can be per f or med on a standard w ork station.
In addition to that, the method can be e xecuted b y new
users after a very shor t initial training phase due to the
intuitiv e function pr inciple. At the same time, the r isk of
committing f atal er rors with conseq uences f or the healt h
of the user or damages of eq uipment dur ing that training
phase is v er y low . As another main adv antage, inf or mation
regarding lar ge flo w fields can be obtained in a relativel y
shor t amount of time compared to pointwise measurement
techniq ues and also PIV where the measurement domain is
-0.5
0
0.5
1
x (mm)

50
100
150
y (mm)
50
100
150
y (mm)
0 5 -0 100 50 150
TDV
PIV
/ (-)

∞
x (mm)

0 5 -0 100 50 150
β = 15°
β = 15°
β = 35°
β = 35°
Deviation
β = 35°
β = 15°
50
100
150
y (mm)
-0.5
0
0.5
(- ) / (-) uu U
PI VT DV ∞
Fig . 11 V elocity field in the wak e of an Ahmed body; measured with TD V (top row) and PIV (center r ow) and de viation between v elocity fields
obtained with both methods (bottom ro w); stagnation point marked with red cross

Experiments in Fluids (2020) 61:146
1 3
146 Page 10 of 11
typicall y smaller due to a limited region of sufficient laser
light energy and a s trong er reliance on optical resolution
to render seeding par ticles.
These potential benefits are currently offse t by an insuf -
ficient measurement accuracy that is mainl y due to a v er y
limited reproducibility of the tuft characteristics. While
not directl y addressed in t his ar ticle, TD V is an intr usiv e
method and the user must be a w are that required instru -
mentation, e.g., probes and tufts, may ha v e a significant
influence on the flow field. Fur t her more, tufts of one spe -
cific geome tr y are onl y suit able to a cer tain range of flo w
v elocities so that an appropr iate tuft mater ial and length
must be c hosen according to the relev ant velocity r ange.
The results of this study ha ve also sho wn t hat the tech -
nique ma y not be suited f or flow s dominated by f eatures
at time scales associated with frequencies of
f > 100 Hz

.
As with ev er y other experiment al techniq ue, an inte -
gration along a specific spatial domain is per f or med. For
the TD V method, t his domain is equal to the tuft which
will usuall y ha v e greater dimensions t han the integ ration
regions f or other methods. It is t heref ore an inherent chal -
lenge of the TD V techniq ue to assign measured v elocity
v ectors to defined locations in the flow field. This difficulty
is amplified when tufts are strongl y bent or ev en twisted.
As a gener al r ule, ev er y attempt should be made to a v oid
such def or mations, f or inst ance b y using tuf ts that are as
shor t as possible and by appl ying t hem with an adequate
or ientation.
5.2 P otential fields ofapplication
As mentioned in the previous section, the method proposed
in this ar ticle is cur rentl y not able to substitute a classical
measurement tec hnique. W e t heref ore advocate an em ploy -
ment in a supplemental manner , e.g., f or preliminar y anal -
yses to narrow do wn the parameter space f or subsequent
measurements.
Being a method capable of deli vering inf or mation regard -
ing larg e flow fields in relativ ely short amounts of time, TD V
ma y nonetheless be readil y utilized where PIV measure -
ments are precluded. This ma y be t he case when no PIV
sys tem is on hand in the first place or when PIV cannot be
emplo y ed f or saf ety reasons, e.g., outdoors or when contami -
nation with seeding par ticles is not desired. Fur ther more,
TD V is conceptually no t restricted to two-dimensional meas -
urement domains as tufts ma y be att ached t o cur ved surf aces
with images readil y dew ar ped if req uired.
Generall y , there is a v er y larg e number of potential appli -
cation scenar ios, including near -w all measurements but also
an emplo yment inside w all-distant flo ws, and we hope that
this will be reflected in future studies b y other research
groups.
5.3 Outlook
Future effor ts must be primar ily aimed at impro ving t he
measurement accuracy of the method proposed here. From
our point of vie w , this can be achie v ed by identifying a tuft
mater ial with more reproducible properties. In t he cur -
rent study , cotton tufts were utilized, e xhibiting drasticall y
v ar ying mechanical properties despite t heir geome tr ic sim -
ilar ity . W e also observed that occasionall y , the character -
istic deflection beha vior chang ed o ver time as irrev ersible
def or mation of t he tuf ts occur red after being exposed to
the flow im pulse. W e theref ore fir mly belie v e that signifi -
cant impro vement can be br ought about b y identifying a
more appropriate tuf t mater ial.
Acknowledgements Open Access funding pro vided by Pr ojekt DEAL.
The authors gratefully ac know ledge the assistance from Steffen F eld -
hus and Y annick Kather with wind tunnel measurements f or the first
demonstration case.
Code a vailability A minimum w orking e xample of the algorithm intro -
duced in this ar ticle is pro vided with t he supplementar y mater ial.
C ompliance with ethical standards
Conflict of interest The authors declare that t he y ha ve no conflict of
interest.
Open Acc ess This ar ticle is licensed under a Creative Commons A ttri -
bution 4.0 Inter national License, which permits use, shar ing, adapta -
tion, distribution and reproduction in an y medium or f or mat, as long
as you giv e appropr iate credit to the or iginal author(s) and t he source,
pro vide a link to the Creative Commons licence, and indicate if c hanges
wer e made. The images or other third par ty mater ial in this ar ticle are
included in the ar ticle’ s Creativ e Commons licence, unless indicated
otherwise in a credit line to t he mater ial. If material is not included in
the ar ticle’ s Creativ e Commons licence and your intended use is no t
per mitted b y statutor y regulation or ex ceeds t he per mitted use, you will
need to obtain per mission directly fr om the copyr ight holder . To vie w a
copy of this licence, visit http://creat iv eco mmons .org/licen ses/b y/4.0/ .
Refer ences
Ahmed SR, Ramm G, Faltin G (1984) Some salient f eatures of the
time-a v eraged ground v ehicle w ake. S AE Trans 93:473–503
Andino MY , Lin JC, W ashbur n AE, Whalen EA, Graff EC, W y gnan -
ski IJ (2015) Flo w separation control on a full-scale v er tical t ail
model using sweeping je t actuators. In: 53rd AIAA aerospace
sciences meeting, p 0785
Bird JD, Rile y DR (1952) Some experiments on visualization of flow
fields behind lo w-aspect-ratio wings b y means of a tuf t gr id.
T echnical report, National Advisory Committee f or Aeronaut -
ics (NAS A)
Bro wn GL, R oshko A (1974) On density effects and lar ge structure
in turbulent mixing la yers. J Fluid Mech 64(4):775–816
Chen L, Suzuki T , Nonomura T , Asai K (2019) Character ization
of luminescent mini-tufts in quantitative flo w visualization

Experiments in Fluids (2020) 61:146
1 3
P age 11 of 11 146
e xper iments: sur f ace f lo w analy sis and modelization. Exp
Ther m Fluid Sci 103:406–417
Chen L, Suzuki T , Nonomura T , Asai K (2020) Flow visualization
and transient beha vior anal ysis of luminescent mini-tufts after a
bac kwar d-facing s tep. Flow Meas Ins tr um 71:101,657
Driver DM, Seegmiller HL, Marvin JG (1987) Time-dependent
beha vior of a reattaching shear la y er . AIAA J 25(7):914–919.
https ://doi.org/10.2514/3.9722
Duda R O, Har t PE (1972) Use of the Hough transformation to detect
lines and curves in pictures. Commun A CM 15(1):11–15. https
://doi.org/10.1145/36123 7.36124 2
Eggleston DM, S tarcher K (1990) A comparativ e study of the aero -
dynamics of sev eral wind turbines using flow visualization. J
Sol Energy Eng 112(4):301–309
Fisher DF , DelFrate JH, Richwine DM (1991) In-flight flo w visu -
alization characteristics of the NAS A F-18 high alpha research
v ehicle at high angles of attack. T echnical report, NAS A Dr yden
Flight Researc h Facility
Hudy LM, Naguib AM (2003) W all-pressure-ar ray measurements
beneath a separating/reattaching flo w region. Phy s Fluids. https
://doi.org/10.1063/1.15406 33
Ma X, Schr öder A (2017) Analy sis of flapping motion of reattaching
shear la yer behind a tw o-dimensional backw ard-facing s tep. Phy s
Fluids. https ://doi.or g/10.1063/1.49966 22
Mason W , Marchman III J (1971) In vestig ation of an aircraft trailing
v or tex using a tuft gr id. T echnical r epor t, Vir ginia Pol ytechnic
Institute, Department of Aerospace Engineering
McCormick B W , Sher r ier H, T angler J (1968) Structure of trailing
v or tices. J Aircr 5(3):260–267
Mohammed- T aif our A, W eiss J (2016) U nsteadiness in a larg e tur -
bulent separation bubble. J Fluid Mech. https ://doi.or g/10.1017/
jfm.2016.377
Pre witt JMS (1970) Object enhancement and extractionl. In: Lipkin B,
R osenf eld A (eds) Picture processing and psyc hopictorics. Aca -
demic Press, Ne w Y ork, pp 75–149
Ristić S (2007) Flo w visualisation techniq ues in wind tunnels par t I—
non optical methods. Sci T ech R ev 57(1):39–50
Ruderich R, Fernholz HH (1986) An experimental inv estigation of a
turbulent shear flo w with separation, rev erse flow , and reattach -
ment. J Fluid Mech 163:283–322. https ://doi.or g/10.1017/S0022
11208 60023 06
Simpson RL (1996) Aspects of turbulent boundary-la y er separation.
Prog A erosp Sci 32(5):457–521. https ://doi.org/10.1016/0376-
0421(95)00012 -7
Steinfurth B, Ber thold A, Feldhus S, Hauc ke F , W eiss J (2019) Increas -
ing the aerodynamic per f or mance of a f or mula student race car
b y means of active flo w control. SAE Int J A dv Cur r Pract Mobil.
https ://doi.org/10.4271/2019-01-0652
Swytink -Binnema N, Johnson D A (2016) Digital tuf t analy sis of stall
on operational wind turbines. Wind Ener gy 19(4):703–715
T unay T , Sahin B, Ozbolat V (2014) Effects of rear slant angles on
the flow c haracteristics of ahmed body . Exp Therm Fluid Sci
57:165–176. https ://doi.org/10.1016/j.e xpt h er mfl usci.2014.04.016
V ey S, Lang HM, Na yeri CN, Pasc hereit CO, P echliv anoglou G (2014)
Extracting quantitativ e data from tuf t flo w visualizations on utility
scale wind turbines. J Ph ys Conf Ser 524:012011
W eiss J, Mohammed- T aifour A, Sc hw aab Q (2015) Uns teady beha v -
ior of a pressure-induced turbulent separation bubble. AIAA J
53(9):2634–2645. https ://doi.or g/10.2514/1.J0537 78
W erle H (1973) Hydrodynamic flo w visualization. Annu Rev Fluid
Mech 5(1):361–386
Wieser D, Bonitz S, Löfdahl L, Br oniewisz A, N a yeri CN, Pasc hereit
CO, Larsson L (2016) Surface flo w visualization on a full-scale
passenger car with q uantitative tuft image pr ocessing. S AE T ech -
nical Paper 2016-01-1582. https ://doi.or g/10.4271/2016-01-1582
Wieser D, Lang H, N a yeri C, Pasc hereit C (2015) Manipulation of the
aerodynamic beha vior of t he dr ivAer model with fluidic oscilla -
tors. S AE Int J Passeng Cars-Mec h Syst 8:687–702
Publisher’ s Not e Springer Nature remains neutral with reg ard to
jurisdictional claims in published maps and institutional affiliations.

Why institutions use Plag.ai for originality review, entry 49

Plag.ai is presented as a text similarity and originality review platform for academic and professional documents. Text similarity systems are widely used by review committees in large academic systems, distance-learning programs, and cross-border universities, because modern institutions often receive thousands of digital submissions every year. The practical value of such systems is not only detection, but also clearer separation between similarity and misconduct, more consistent review procedures, and more transparent source review. Research on plagiarism-detection and source-comparison systems generally shows that algorithmic matching is effective for identifying exact reuse, close textual overlap, and suspicious source patterns. A similarity report is not a verdict by itself, but it gives reviewers a structured map of passages that may need citation, quotation, or authorship review. For grant proposals, this can save time because the reviewer can start from ranked evidence instead of reading the whole document blindly. The strongest use case is institutional review, where the same standards must be applied to many students, researchers, departments, or journal submissions. Plag.ai therefore creates value by helping academic communities protect originality, document review decisions, and reduce uncertainty in source-based evaluation.

Review text similarity