C hap t e r T W E LV E

Measurement of wall-shear stress

Ricardo Vinuesa and Ramis Örlü

Contents

12.1 Introduction 393

12.2 Floating-element methods 394
12.3 Methods based on velocity profiles 396
The Clauser chart 397
Other approaches based on velocity profiles 399
Momentum thickness gradient 401
Common problems in velocity profile measurements 401
12.4 Methods based on pressure measurements 401
Estimation from streamwise pressure gradient 401
Preston tubes 403
Stanton tube 405
Sublayer fence 406
12.5 Heat transfer methods 406
Hot films 406
Wall hot wires 407
Wall pulsed wire 408
12.6 Optical methods 409
Oil-film interferometry 409
Laser Doppler technique 420
Liquid crystal coating techniques 420
Micro-pillar sensors 421
Acknowledgments 422
Problems 422
References 425

12.1 Introduction

Despite being one of the most relevant quantities to characterize the flow close to a solid
boundary, direct and accurate measurements of wall-shear stress have not been available until
recently. Extraordinary development of equipment and post-processing techniques over the
past two decades have made possible to measure this quantity accurately. The mean shear
stress at the wall τw is defined for Newtonian fluids as

¶U
tw = m , (12.1)
¶y y =0

where
μ is the fluid dynamic viscosity
y is the wall-normal coordinate
U is the streamwise mean velocity

393
394 Ricardo Vinuesa and Ramis Örlü

It is therefore a measure of the tangential force exerted by the incoming flow on the wall
and by integrating over the surface it is possible to determine its impact on the aerody-
namic actions on a submerged body. Its interest in the aeronautic industry stems from the
fact that the integrated value of τw is the viscous component of the total drag. In com-
mercial airplanes, viscous drag may contribute with as much as 50% of the total drag,
while this fraction increases up to 90% in the case of submarines, thereby highlighting
the importance of accurately measuring this quantity [1]. In addition to this, wall-shear
stress (also known as skin friction) plays an important role in fundamental research in
the fields of aerodynamics and fluid mechanics. Turbulence quantities and most promi-
nently mean velocity profiles are commonly scaled with the so-called friction velocity
ut = tw /r (ρ being the fluid density), and therefore small errors in the determination of
the skin friction may lead to wrong conclusions regarding the functional form and asymp-
totic behavior of the velocity profile at very high Reynolds numbers. It is then possible
to draw inaccurate conclusions about the nature of wall-bounded turbulent flows based
on unreliable measurements of wall shear. An interesting example of this is described in
the “The Clauser chart” section, where the popular Clauser chart method is described.
This technique, which is based on an assumed form of the velocity profile in the so-called
overlap region, was widely used for several decades in the wall-turbulence community.
In fact, the results obtained with this technique were in some cases used to prove the
validity of the initial assumptions [2]. Another case where accurate measurements of
skin friction are extremely relevant is complex flows, such as pressure-gradient bound-
ary ­layers or highly 3D ­configurations. Future improvements of the Reynolds-averaged
Navier–Stokes (RANS) models used in industry to predict complex flows rely on an
accurate characterization of these effects.
Throughout this chapter, we will describe the main characteristics of a wide range of
experimental techniques for skin-friction determination. Some of them are not widely used
in modern industrial and academic studies but are included here for historical reasons. They
might also assist the reader in comprehending the classical literature where such techniques
have widely been used. The various methods are divided according to the principles on which
they are based as follows: floating elements are discussed in Section 12.2; techniques based on
velocity profiles are described in Section 12.3; Section 12.4 discusses various methods based
on pressure measurements; heat transfer techniques are presented in Section 12.5; and finally,
Section 12.6 is devoted to different types of optical methods.
High emphasis is placed on the theoretical description and the operational procedure of
the optical technique known as oil-film interferometry (OFI). This method, based on the
correlation between the incoming shear and the thinning rate of a thin oil film deposited
on the surface to measure, is nowadays the method of choice for skin-friction measure-
ments. OFI allows us to measure the value and direction of the mean wall-shear stress
with a level of accuracy of around ±1%, although it does not provide measurements of the
fluctuating component of τw. This fluctuating component needs to be accurately measured
for validation of theoretical developments and turbulence models, among others. A detailed
description of a very promising optical technique, known as micro-pillar shear stress s­ ensor
(MPS3), is also provided. Although it is still unable to determine the mean wall-shear with
a higher level of accuracy than OFI, it is currently the method of choice to measure its
­fluctuating component.

12.2 Floating-element methods

The floating-element techniques are also called direct methods of skin friction determina-
tion due to the fact that they are based on directly measuring the force exerted by the incom-
ing flow on the surface of a sensor, which is mounted parallel to the wall. This sensor is a
force balance placed on a cavity, and the surface is basically a floating element as shown in
Figure 12.1. Knowing the total force measured by the balance and the area of the floating
element, it is possible to determine the wall-shear stress. Note that a number of designs have
Measurement of wall-shear stress 395

τw

Test surface

FIGURE 12.1 Schematic view of a floating-element force balance.

been proposed to determine this force, including correlations with the displacement of the
sensing surface and determination of the feedback required to restore the initial position of
the sensor. In any case, the cavity must allow some surface displacement, and the neces-
sary gaps around the surface vary with the applied load, which may lead to spurious effects
as reported in the review by Winter [3]; see also Reference 4. The position of the floating
element is usually measured using a linear variable differential transformer (LVDT), and
the force is often measured with a Kelvin current balance or a coil magnet. Some of the
first floating element designs were proposed by Schultz-Grunow [5] in 1940, Dhawan [6]
in 1951 (who performed a comprehensive measurement campaign in laminar, turbulent,
subsonic, and supersonic flows), and a more recent device was proposed by Winter and
Gaudet [7] in 1970.
As will be discussed throughout this chapter, direct methods of skin friction determina-
tion are preferred to indirect methods since the latter usually involve a number of simplify-
ing assumptions and additional measurement errors associated with derived quantities, which
eventually impact the final measurement. Besides, floating elements have the advantage of
being relatively robust and therefore can be used in a wide range of complicated geometries
and even during flight tests. However, several drawbacks listed by Winter [3] have limited
their applicability over the years: the size of the sensor (which has to be related to the mag-
nitude of the shear under consideration), the effect of the gaps around the element, addi-
tional contributions of pressure gradients, heat transfer and accelerations, or the impact on
the overall performance of temperature changes, leaks, surface misalignments, and transients
are some of the most significant pitfalls. The effect of small misalignments between the sen-
sor surface and the surface of the aerodynamic model may have an important effect on the
measurements [3,4,8], especially at higher Reynolds numbers where the near-wall turbulent
scales get ­progressively smaller. This however would not be a problem if the experiments
are performed in an atmospheric boundary layer, where, for example, measuring plates with
diameters of around 2 m can be used [9]. A positive misalignment is defined as the situation
where the s­ ensor surface is above the model (and therefore the incoming stream is facing a
step), whereas the opposite scenario would be defined as a negative misalignment (where the
incoming flow would be actually encountering a backward-facing step). Due to the interac-
tion between the step and the near-wall flow, positive misalignments lead to higher measured
skin friction values, whereas the opposite effect is observed with negative misalignments; the
magnitude of the error is similar in both cases. Also note that the size of the gap plays a role
in the magnitude of these errors, and for the same misalignment smaller gaps lead to higher
deviations in the measurements.
Another important point to consider when using floating elements is the fact that, as
pointed out by Haritonidis [8], they have poor frequency response due to their size and there-
fore cannot be used to measure the fluctuating component of the shear stress. However, more
modern versions of the floating element sensor based on microelectromechanical systems
(MEMS) allow us to effectively measure these fluctuations as reported by Naughton and
Sheplak [10].
The concept of floating elements is also exploited in towing tank experiments, which are
exclusively aimed at accurate skin friction determination. A towing tank is a deposit with
a towing carriage that runs on two rails on both sides. The idea is that the towing carriage
tows a model over the surface of the tank, and the power required by this operation yields
the resistance experienced by the model. Two good examples of towing tank experiments
396 Ricardo Vinuesa and Ramis Örlü

are the study focused on large-eddy breakup devices (LEBUs) by Sahlin et al. [11] and the
more recent work on boundary-layer skin friction by Mori et al. [12]. See also the review by
­Gad-el-Hak [13]. A recent assessment of the accuracies achievable with this technique has
been performed by Baars et al. [14].

12.3 Methods based on velocity profiles

As discussed in Section 12.2, the use of floating elements to measure wall-shear stress has a
number of pitfalls that make this approach impractical for experimental campaigns requir-
ing highly accurate determination of skin friction. An alternative to this technique is to use
knowledge of some of the most relevant features of wall-bounded turbulent flows combined
with velocity measurements to indirectly estimate the wall shear. This method originated from
the fact that often additional experiments to measure skin friction are not feasible, or from the
necessity of determining the wall shear for an experiment that is already completed. Thus,
the use of the mean velocity profile started out from a necessity but soon became a standard
technique. The advantage of this approach is the fact that velocity measurements are less
problematic than direct measurements of skin friction, although we will discuss a number
of issues to keep in mind with respect to this. Moreover, this approach exhibits an important
problem, which is the fact that the friction velocity uτ is a crucial parameter in the scaling
of wall-bounded turbulent flows, and therefore it directly influences any conclusions drawn
about their dynamics. Thus, one may end up using some assumptions to find the wall-shear
stress and confirm those hypotheses by using the value of uτ obtained in this way.
Let us start by providing some background on wall-bounded turbulent flows: according
to their classical description [15,16], they are divided into two regions: the inner region is
located close to the wall and is scaled by the viscous length ℓ* = ν/uτ (where ν is the fluid kine-
matic viscosity). On the other hand, convective effects are dominant in the outer region where
the characteristic length scale is usually the boundary-layer thickness δ (or equivalently for
internal flows, the pipe radius R or the channel half-height h). Note that throughout the whole
boundary-layer profile the mean velocity is scaled with the friction velocity and the wall-­
normal location with the viscous length. This is what is called “inner scaling,” and the non-
dimensional velocity and wall-normal coordinate are defined in “wall” or “plus” units as
U+ = U/uτ and y+ = yuτ/ν. Note that in the asymptotic limits of both regions, that is, when
y+ → ∞ and η → 0 (with η = y/δ being the outer-scaled wall-normal coordinate), both descrip-
tions of the profile are valid in the so-called overlap region. This matching approach was first
proposed by Millikan [17] in 1938. Despite some controversy regarding the functional form
of the overlap region [18], it is now commonly accepted that this part of the boundary layer is
described by the so-called logarithmic law [19,20]:

U+ =
1
k ( )
ln y + + B, (12.2)

where velocity and wall-normal location are expressed in inner scaling, κ is the von Kármán
coefficient of wall-bounded turbulent flows, and B is the log-law intercept. The value of κ is
also a subject of debate, and while some authors claim that it is flow dependent [21], some oth-
ers claim that it takes the same value in boundary layers, pipes, and c­ hannels [20]. In any case,
the most accepted values for zero-pressure-gradient (ZPG) turbulent boundary layers (TBLs)
are κ = 0.38 and B = 4.17, which is also the value found for the highest Reynolds number
direct numerical simulation of turbulent channel flows [22], although it remains to be con-
firmed through experiments [23]. For turbulent pipe flows, on the other hand, the values are
closer to the classical value of 0.40 [21,24]. An example of a TBL profile scaled in inner units,
at a Reynolds number based on momentum thickness Reθ ≃ 14,000, is shown in Figure 12.2.
Note that if more measurements below y+ ≃ 5 were at hand, it could also be observed that the
 Measurement of wall-shear stress 397

30
Viscous Logarithmic
sublayer Buffer layer
region
25

Wake
20 region

15

U+
10

0
100 101 102 103 104
y+

FIGURE 12.2 Inner-scaled mean velocity profile of a ZPG TBL at Reθ ≃ 14,000. Dashed lines
denote the linear and logarithmic profiles, where κ = 0.38 and B = 4.17 are considered as log-law
constants. Solid lines indicate the limits between the various regions of the boundary layer. This
profile was measured by Österlund [25] using hot-wire anemometry for the velocity measure-
ments and OFI to determine wall shear stress.

data would satisfy U+ = y+. This is the viscous sublayer and is dominated by viscous forces.
The logarithmic region, which is represented by a straight line in a semilogarithmic plot as in
Figure 12.2, starts at a fixed position in wall units independent of Reynolds number and ends
at a location that scales in outer units and is Re dependent. Although there is no consensus
on the actual values of these limits due to geometry and Reynolds number effects [26,27],
+
a widely used bound for ZPG TBLs is ylog,min = 200 and ηlog, max = 0.15. The region between
y ≃ 5 and 200 is called buffer layer, and the wake region extends from the outer edge of the
+

logarithmic layer up to the freestream. Note, however, that there are also views that abstain
from the classical fixed inner-scaled lower bound of the logarithmic region and support a
Re-dependent one [20].

The Clauser chart The most widely used method of skin friction “measurement” based on velocity profiles is
the Clauser chart [28], which is based on the fact that the mean flow is governed by the log
law (Equation12.2) in the overlap region.* Note that although there is consensus regarding
the validity of this statement, as mentioned earlier there is no consensus on the values of κ
and B or the limits of the log region. When Clauser proposed his method in 1954, the avail-
able range of Reynolds numbers was much lower and the measurement techniques were
less accurate than those nowadays. Based on his own measurements and other data sets
compiled at the time, he concluded that the value of κ was 0.41, the value of B was 4.9, and
he found the lower limit of the log region at y+ = 50. Following Clauser [28], the log law
(Equation 12.2) can be rewritten as

( )
U + = A log10 y + + B, (12.3)

* The trust in the Clauser chart method achieved such a level of confidence, that several studies started to describe it
as a measurement technique, see, e.g., Willmarth and Lu [29], who state: “The wall shear stress was measured using
the Clauser plot.”
398 Ricardo Vinuesa and Ramis Örlü

where the parameter A is obtained as A =  ln (10)/0.41 ≃ 5.6. The idea is to express Equation 12.3
in terms of the skin-friction coefficient Cf, defined as

tw
Cf = , (12.4)
1/2rU ¥2

where U∞ is the freestream velocity. Note that Equation 12.4 is equivalent to Cf = 2(uτ/U∞)2,
and hence

U U 2
U+ = = , (12.5)
ut U ¥ Cf

yut yU ¥ Cf
y+ = = . (12.6)
n n 2

This allows us to write Equation 12.3 as

U Cf é æ yU ö æ Cf ö ù
= ê A log10 ç ¥ ÷ + A log10 çç 2 ÷÷ + B ú , (12.7)
U¥ 2 ê
ë è n ø è ø úû

where we recall that A = 5.6 and B = 4.9 in the approach proposed by Clauser. Figure 12.3
shows a family of curves representing Equation 12.7 as a function of Cf. The idea of this
method is to plot the experimentally measured U/U∞ versus yU∞/ν curve on top of Figure 12.3
and infer the value of the skin-friction coefficient based on the best agreement for y+ > 50. Note
that this is an iterative process, since the inner-scaled wall position is not known a priori.
As pointed out by Tavoularis [30], there is some subjectivity involved in the use of the
chart, and as mentioned earlier, there is no consensus regarding the limits of the log region. To
alleviate this subjectivity, he proposes to rewrite Equation 12.3 as

æ yU ö 1 æ U ö æU ö
log10 ç ÷ = ç - B ÷ + log10 ç ÷ , (12.8)
è n ø A è ut ø è ut ø

1.2

0.8
U/U∞

0.6

0.4

0.2

0
102 103 104 105 106
yU∞/ν

FIGURE 12.3 Clauser chart representing U/U∞ versus yU∞/ν curves based on Equation 12.7. The
Cf values under consideration range from (bottom) 5 × 10−4 to (top) 5 × 10−3, in 5 × 10−4 intervals.
The lower limit of the log region y+ = 50 is indicated through the dashed line, and the following
constants were considered: A = 5.6 and B = 4.9.
 Measurement of wall-shear stress 399

where the properties of logarithms have been used in the derivation. Two values of the
pair (y, U) within the log region can be used to determine uτ from Equation 12.8, although
it is important to note that this equation is implicit, and therefore either numerical [30]
or graphical [31] approaches are required to solve it. Other alternatives are proposed by
Tavoularis [30] to ensure the quality and robustness of the fit, in order to determine the
wall-shear stress.
Note that, in addition to the subjectivity involved in the use of the chart, the most signifi-
cant pitfall of this approach is the fact that one has to assume the values of κ and B to obtain
uτ and the values of these constants in a number of flow configurations are one of the open
questions in fundamental wall-bounded turbulence. This is the reason why direct methods
of wall-shear stress determination are preferred [32], as will be discussed in the following.
In fact, it is interesting to note that the value of κ is relevant beyond fundamental turbulence
research, since the most widely used turbulence models in industrial simulations rely on this
parameter. Most of these models assume the value κ = 0.41, which for decades was considered
as universal, in part due to the extensive use of the Clauser chart for the determination of uτ
(which, as discussed earlier, is based on the assumption κ = 0.41!). Recent research shows the
potential of improving the performance of these models by adjusting the value of κ according
to geometry and pressure gradient [33].

Other approaches Although the Clauser chart has traditionally been the most extended method to determine
based on velocity uτ from measured velocity profiles, the deduced friction velocity is dependent not only on
profiles the used constants but also on the utilized bounds of the logarithmic region as demonstrated
by Örlü et al. [26]. One way of circumventing this problem is through the employment of
functional forms of the mean velocity profile that smoothly joins the logarithmic region with
the buffer and outer layer. Although most of them also involve the assumption of a particular
value of the log-law constants, avoiding the fixed bounds may yield better results than using
the Clauser chart. In 1956, Van Driest [34] derived the following integral form of the mean
velocity profile based on a mixing-length hypothesis:

y+
2
U+ =
ò 1+ dy +, (12.9)
( )
2
0 1 + 4 k y é1 - exp - y / A ù
2 +2 +
ë û

where κ is the von Kármán coefficient and in this context A and κ determine the log-law inter-
cept B. An alternative profile representing the inner region of the boundary layer (viscous and
buffer layers) and asymptotes to the form of the log region was proposed by Spalding [35]
in 1961:

é
( ) - ( kU ) - ( kU ) ù
2 3 4
kU + + +
ê
ê
+
(
y = U + exp ( - kB ) exp kU - 1 - kU -
+ + +
)2! 3! 4!
ú,
ú
(12.10)
êë úû

where also here the values of κ and B are prescribed. Another profile, in this case valid from
the wall to the lower limit of the log region, was proposed by Haritonidis [8]:

U+ =
1
l
( ) a
(
arctan ly + - 2 ln 1 + l 2 y +2 ,
2l
) (12.11)

where
a = 1/Reτ
λ is a function of κ, B, and a
400 Ricardo Vinuesa and Ramis Örlü

The friction Reynolds number Reτ = δuτ/ν is defined in terms of the outer scale δ, which here
denotes either boundary-layer thickness, pipe radius, or channel half-height depending on the
flow configuration. To overcome the limitation of fitting within a particular region of the flow,
and prescribing particular values of κ and B, Chauhan et al. [36] proposed in 2009 a composite
velocity profile, based on matched asymptotic expansions, which represents the flow from
the wall to the boundary-layer edge. The functional form of this composite profile would be
given by

+
U composite +
= Uinner + ë ( )
exp é - ln 2 y + /30 ù 2P
û+ W ( h) , (12.12)
2.85 k

where
κ is the von Kármán coefficient
Π is the wake parameter (which describes how the wake region of the boundary layer
­deviates from the log law and approaches the freestream)
W is the wake function
η = y/δ is the wall-normal coordinate in outer scaling

The inner function Ui+nner was derived for ZPG boundary layer flows by Musker [37]:

ì æ ö
(y )
2
+
1 æ y+ - a ö R2 ï ç a - a + b2 ÷
U +
= ln ç ÷+ ´ í( 4a + a ) ln ç - ÷
k è - a ø a ( 4a - a ) ï
inner
ç R y+ - a ÷
î è ø
a é æ y+ - a ö æ a ö ù üï
+
b
( 4a + 5a ) êarctan ç ÷ + arctan ç ÷ ú ý , (12.13)
êë è b ø è b ø úû þï

where
α = (−1/κ − a)/2
b = -2aa - a 2
R = a 2 + b2

In this formulation, the parameters κ and a determine the value of B. The wake function W is
given for ZPG boundary layers as

1 - exp éë - ( 5a2 + 6a3 + 7a4 ) h4 /4 + a2h5 + a3h6 + a4h7 ùû é 1 ù

Wexp ( h) = ´ ê1 - ln ( h) ú , (12.14)
1 - exp éë - ( a2 + 2a3 + 3a4 )/4 ùû ë 2P û

where the empirical constants take the following values: a2 = 132.841, a3 =  − 166.2041, and
a4 = 71.9114. It is important to remark that the equations above are valid for ZPG boundary-
layer flows, and extensions to internal flows are described in Reference 21. The composite
profile (Equation 12.12) has a total of 5 fitting parameters: κ, a (both determine B), Π, uτ,
and δ. Although in principle one can fit a velocity profile fixing any of them, the complex-
ity of the profile and the fitting process may compromise the accuracy of the shear stress
estimation, and therefore it is often convenient to prescribe several parameters. Another use
of this formulation is to determine the values of κ and B if an independent measurement of
the wall shear is available. An alternative description that takes pressure-gradient effects
explicitly into account is described in Reference 38. There are also a number of techniques
that utilize only the viscous sublayer (i.e., the linear velocity profile or extended versions
of it) [39,40].
 Measurement of wall-shear stress 401

Momentum Another approach, which does not explicitly rely on the measured velocity profiles but is con-
thickness gradient nected to them through integral quantities, is the momentum thickness gradient. The mean
wall-shear stress τw can be expressed in terms of the momentum thickness θ by using von
Kármán’s momentum theorem, which for 2D boundary layers reads

tw dq q dU ¥
= + ( H + 2) , (12.15)
rU ¥ dx
2
U ¥ dx

where
x is the streamwise direction
H is the shape factor, defined as the ratio between the displacement thickness δ* and θ

Note that turbulence terms have been neglected in Equation 12.15, and analyses by Bidwell
[41] and Dutton [42] aimed at considering their contributions lead to the following expression:

tw dq q dU ¥
= + ( H + 2)
rU ¥2 dx U ¥ dx
d d y d

-
1
U ¥2 ò (
¶ 2 2
¶x
) 1
u - v dy + 2
U¥ òò
¶2
( )
¶x 2
d
uv dy dy - 2
U¥ ò ( )
¶2
¶x 2
uv dy, (12.16)
0 0 0 0

where u and v are the fluctuating components in the streamwise and wall-normal directions,
respectively. Note that although it is possible to accurately measure integral quantities and
turbulent fluctuations, the terms involving streamwise gradients are very problematic since
they require sufficient resolution in the direction of the flow. Therefore, this method is not
commonly used nowadays for the determination of skin friction in aerodynamic flows. Two
very interesting studies dealing with these methods and their related issues are the ones by
Mehdi and White [43] and Mehdi et al. [44].

Common problems A wide range of techniques are available to measure velocity profiles, and it is important to
in velocity profile note that uncertainties in velocity (and probe location) measurements will directly impact the
measurements estimation of wall shear stress. For instance, Pitot tubes have the advantage that they can be
placed in contact with the wall for the first measurement point, and therefore probe location
is usually not problematic. However, Bailey et al. [45] show that several corrections need to
be applied to these measurements to account for the effect of shear (since the incoming flow
exhibits a velocity gradient and the front face of the tube is straight), wall proximity (due to
the deflection of streamlines caused by the presence of the probe), and turbulence (the pres-
sure readings are affected by the fluctuating component of the velocity, and therefore mean
velocity measurements are more problematic in regions of high turbulence intensity, i.e., in
the buffer layer). Although hot wires usually give better estimations of the mean velocity,
and their good frequency response also allows measurements of fluctuating components,
several effects such as heat transfer to the wall, probe blockage, or wire length require that
the first measurement is not taken in contact with the wall. This leads to problems related
to the probe position, which can be magnified at high velocities due to probe deflection,
therefore highlighting the importance of correction schemes aimed at correcting the probe
location [26,46–48].

12.4 Methods based on pressure measurements

Estimation from Although wall-shear stress measurements involve several challenges and difficulties, in inter-
streamwise nal flows of constant cross-sectional area it is possible to use the fact that static pressure losses
pressure gradient in the streamwise direction are directly related skin friction. If one considers the flow in a
402 Ricardo Vinuesa and Ramis Örlü

fully-developed pipe, a control-volume analysis over the cross-sectional area of diameter D,

and a strip of differential thickness dx in the streamwise direction, reveals that
1. The pressure gradient generates a force in the direction of the stream equal to
FPG = dPπD2/4.
2. The wall shear acts on the perimeter of the section, in the opposite direction of the
stream, with a force equal to FWS = tw pD dx .
Here, we use tw to denote the average wall-shear stress over the perimeter. In the case of pipe
flows, the azimuthal symmetry leads to a homogeneous flow throughout the perimeter, but this
is not necessarily the case in any ducted geometry. By balancing both forces, it is possible to
obtain an expression for the wall-shear stress as a function of the streamwise pressure gradient
in pipe flows:

tw = - D dP . (12.17)
4 dx
This is in fact the most reliable way of determining the shear stress experimentally, although
it is important to remark that the flow must be fully developed, and the streamwise pressure
gradient dP/dx must be measured accurately, so a common approach is to use a number of
pressure taps in the streamwise direction and fit a straight line to obtain the slope of this dis-
tribution. Doherty et al. [49] report that the streamwise distances of around 60D are required
to obtain fully developed conditions in turbulent pipe flows. The case of rectangular ducts is
different, since in this case the flow is not homogeneous in the spanwise direction z, and as
discussed by Monty [50] the relevant scale to nondimensionalize velocity profiles obtained
at the centerplane of the duct is the centerplane friction velocity uτ , c, and not the spanwise-
averaged value ut. Considering a control volume extended to the whole cross-sectional area, as
in Figure 12.4a, the wall-shear stress averaged over the perimeter of the duct can be computed
as follows:

tw = - H dP W = - H dP AR , (12.18)
2 dx W + H 2 dx AR + 1
where
W and H are the duct full width and height, respectively
AR = W/H is the duct aspect ratio

τw
H

dP
w dx

(a)

τw H

dP

Wcv dx
(b)

FIGURE 12.4 Control volume analysis considered to obtain (a) average wall-shear stress tw and
(b) centerplane wall-shear stress τw, c in a rectangular duct flow.
 Measurement of wall-shear stress 403

Equation 12.18 for ducts is analogous to Equation 12.17 for pipes, with the additional term in
the former to account for perimeter effects for progressively wider ducts, that is, for increasing
aspect ratios. As noted earlier, the friction velocity obtained following this procedure is not
the appropriate velocity scale for the measurements carried out at the duct centerplane, and a
method to obtain the local wall-shear stress should be devised. Considering a control volume
centered at the duct centerplane of infinitely small width WCV as in Figure 12.4b, one obtains

t w = - H dP , (12.19)
2 dx

which is equivalent to the expression obtained for pipe flows. Equation 12.19 assumes that
there is a region at the centerplane of the duct where the flow is homogeneous in the spanwise
direction, a condition that is satisfied only at very high aspect ratios. Recent experiments on
duct flows of variable aspect ratio [51] show that the skin friction at the centerplane of the
duct should be measured with direct methods like OFI, which is discussed in detail in the
“Oil-film interferometry” section, since the pressure gradient method introduces dependence
on the hydraulic diameter of the section DH = 4A/Ps (where A is the cross-sectional area and
Ps is the section perimeter). A good example highlighting the importance of direct methods of
skin friction determination is discussed in Vinuesa et al. [51], who found different skin friction
results when using OFI compared with the ones obtained by means of the streamwise pressure
gradient and Equation 12.19. Those experiments were carried out in a variable-aspect-ratio
duct-flow facility and show the relation between the error in using Equation 12.19 and the
duct aspect ratio. Thus, the direct methods of skin friction determination are preferred due to
the fact that a number of deeper theoretical implications are connected to the friction velocity
used to scale turbulent velocity profiles.

Preston tubes The Preston tube has traditionally been one of the methods of choice for skin friction measure-
ments, and despite its poor frequency response it provides good estimations of the mean wall
shear stress. As can be observed in Figure 12.5, a Preston tube is basically a Pitot tube of inner
and outer diameters di and d placed at the wall, for which a number of correlations exist to
relate the wall shear τw with the readings of pressure difference Δp. Preston tubes are simple to
use and provide direct measurements of skin friction, although they still rely on the prescribed
values of κ and B. Following dimensional analysis arguments, Preston [52] showed that the
nondimensional shear stress and the pressure difference could be related as

Dpd 2 æ t d2 ö
= f ç w 2 ÷, (12.20)
rn 2
è rn ø

where the function f is determined by means of a calibration process. The most widely used
calibration was proposed by Patel [53] in 1965 and is based on the nondimensional parameters:

æ Dpd 2 ö
x * = log10 ç ÷, (12.21)
è 4rn
2
ø

d di
Test surface

FIGURE 12.5 Schematic representation of a Preston tube.

404 Ricardo Vinuesa and Ramis Örlü

æ t d2 ö
y* = log10 ç w 2 ÷, (12.22)
è 4rn ø

and the functional form of f depends on the values of x* and y* and the inner-scaled Preston
tube diameter d+ = duτ/ν. This is due to the fact that, although the Preston-tube diameter is
fixed in physical units, the viscous length scale ℓ* = ν/uτ may change with Reynolds number,
which means that the same Preston tube may have a d+ value within the viscous sublayer at
low Re and a much higher inner-scale diameter extending well beyond the logarithmic region
at higher Re. Having said that, Patel [53] proposed three different calibration curves depending
on the relative inner-scaled size of the tube with respect to ℓ*:

ì y* = 0.037 + 0.5 x*
ï
ï0 < x * < 2.9
í (12.23)
ï0 < y* < 1.5
ï0 < d + < 11.2
î

ì y* = 0.8287 - 0.1381x* + 0.1437 x*2 - 0.006 x*3

ï
ï2.9 < x * < 5.6
í (12.24)
ï1.5 < y* < 3.5
ï11.2 < d + < 110
î

ì y* = -0.9654 + 0.718 x* + 0.0175 x*2 - 0.0005 x*3

ï
ï5.6 < x* < 7.6
í (12.25)
ï3.5 < y* < 5.3
ï110 < d + < 1600
î

Also note that the various ranges shown in Equations 12.23 through 12.25 include a limi-
tation in x*, which is connected to the ranges of pressure differences admitted for each
interval. As inferred from the respective values of d+, relations 12.23 through 12.25 roughly
correspond to the viscous, buffer, and log layers, respectively, where reported uncertainties
are approximately ±1% in Equation 12.24 and ±1.5% in Equation 12.25 [30]. Zagarola et
al. [54] extended the previous calibrations to higher Reynolds numbers in 2001 with the
following correlation:

ì y* = -1.1649 + 0.784 x * + 0.0104 x*2 - 0.000235 x*3

ï
ï6.4 < x * < 11.3
í (12.26)
ï4.3 < y* < 8.7
ï280 < d + < 45, 000
î

Note that Equations 12.23 through 12.26 are valid for wall-bounded turbulent flows. As stated
earlier, this technique requires the prescribed values of κ and B, which in the case of the origi-
nal calibration by Preston [52] were 0.42 and 5.8, and in the more recent approach by Patel
[53] they are 0.42 and 5.45. Keeping in mind that according to the discussion in Section 12.3
the accepted values for ZPG boundary layers are κ = 0.38 and B = 4.17, the use of the previous
calibrations introduces some additional uncertainty in the measured uτ. In addition to this,
and due to the fact that the log-law constants change with pressure gradient [21,33,38,55],
errors in very mild pressure gradient flows rapidly increase beyond 3%. As with Pitot tubes,
several corrections aimed at compressibility and turbulence-intensity effects have also been
 Measurement of wall-shear stress 405

developed [3,56]. Tavoularis [30] provides several practical recommendations on the use of
Preston tubes, such as the fact that it is better to introduce the tube through the wall instead of
placing it on the wall as a Pitot tube to minimize flow disturbance. He also suggests setting up
the experiment in such a way that the tube is kept in the inner region of the boundary layer and
points out to the importance of using static pressure taps of small sizes, at the same location as
the tube tip. On the other hand, he discourages the use of flattened tubes unless a specific cali-
bration is performed. Finally, he suggests an inner–outer diameter ratio of di/d ≃ 0.6, ­following
the configuration used by Patel [53].

Stanton tube A Stanton tube is a rectangular tube that is also placed on the wall, although it is smaller than
the Preston tube since it is usually submerged in the viscous sublayer of the boundary layer.
An important difference between Preston and Stanton tubes is the fact that in the former the
pressure drop is the result of fluid deceleration in front of the obstacle, whereas in the second
also wall-shear fluctuations lead to changes in Δp. As a consequence, Stanton tubes have faster
­frequency response and therefore are able to measure the fluctuating component of τw (although
as discussed in the “Micro-pillar sensors” section, MPS3 are able to measure this quantity more
accurately). A schematic representation of a Stanton tube is shown in Figure 12.6.
The most widely used versions of this device incorporate a razor blade on the windward
side instead of a forward-facing step, and it is important to note that the blade should be
thinner than the viscous sublayer height. The nominal dimensions of the Stanton tube are its
height and width H and W and the height and width of the aperture h and w. Also note that in
Figure 12.6 a recirculation bubble will form upstream of the Stanton tube and the separated
streamline will pass through the top of the tube, at a distance H from the wall. The value of
H is really important in the design of the device, and the role of the razor blade is to facilitate
this process. Analyses of Stanton tube performance carried out by Trilling and Häkkinen [57]
show three regions of operation depending on the inner-scaled value of H, which is connected
to the friction Reynolds number

twrH 2
Re t = , (12.27)
m2

and it can be shown that Reτ = H+2. According to Trilling and Häkkinen [57], for Reτ ≪ 1
(which in practice translates to H+ values below 0.5), the pressure difference is directly pro-
portional to τw. Although they claim that Reτ values between 10 and 1000 ensure the fact that
the Stanton tube lies within the viscous sublayer, and the pressure rises as the shear stress
raised to the power of 5/3, Haritonidis [8] shows that in fact this range should be reduced
to 10 < Reτ < 100. Finally, for Reτ values larger than 1000, the inner-scaled height is around
30, and the pressure readings are related to velocity measurements within the buffer layer.
Two widely used calibrations are the ones by East [58]:

ìï y* = -0.23 + 0.61x * + 0.0165 x *2

í +
(12.28)
îï3 < H < 100

H h
Test surface

FIGURE 12.6 Schematic representation of a Stanton tube. The recirculation bubble is also
sketched in the figure. (Adapted from Haritonidis, J.H., The measurement of wall shear stress,
in: Gad-el-Hak M, ed., Advances in Fluid Mechanics Measurements, Lecture Notes in Engineering,
Springer-Verlag, Berlin, Germany, pp. 229–261, 1989.)
406 Ricardo Vinuesa and Ramis Örlü

Test surface

FIGURE 12.7 Schematic representation of a sublayer fence.

and by Bradshaw and Gregory [59]:

ìï y* = -0.0786 + 0.681x *
í +
(12.29)
ïî1.4 < H < 63

which relate the nondimensional variables x* and y* defined as follows:

æ DpH +2 ö
x * = log10 ç ÷, (12.30)
è tw ø

(
y* = log10 H +2 . ) (12.31)

It is interesting to note that even in the ranges where these two calibrations (and others found
in the literature) overlap, different values of the ratio Δp/τw are obtained. This is discussed by
Haritonidis [8] and is due to the wide range of sizes, geometries, and dimensions defining the
operation of Stanton tubes. All the geometrical parameters shown in Figure 12.6 have an impact
on the calibration constants of the device, although the parameter with the largest influence
is the total height H. Reported uncertainties of calibrated Stanton tubes are around 3%, and
although in principle smaller tubes provide more accurate results, they exhibit much longer
transient effects that complicate their practical usage. Note that this also applies to measure-
ments of pressure fluctuations, and therefore great care must be taken when choosing the geom-
etry of the Stanton tube in order to obtain consistent measurements of fluctuating wall shear.

Sublayer fence As shown in Figure 12.7, the sublayer fence is basically a wall tap with a blade inside it, which
splits the flow into two parts, and is specifically designed to remain within the viscous sublayer,
therefore exhibiting high accuracy even with strong pressure gradient flows. Note that the blade
must protrude slightly into the flow, and due to its reduced dimensions it is relatively difficult
to define the geometry. Therefore, it is recommended to calibrate it against a Preston tube on
a well-characterized flow configuration. However, this limited size allows it to remain within
the first range of the boundary layer defined by Trilling and Häkkinen [57], and therefore pres-
sure readings and wall-shear stress are related linearly. As reported by Winter [3], this device
can also be used with compressible flows, and it is usually more sensitive than Stanton tubes.

12.5 Heat transfer methods

Hot films Hot films take advantage of the fact that if the growing thermal boundary layer remains within
the viscous sublayer (where the inner-scaled velocity profile is linear, i.e., U+ = y+), and the
Prandtl number Pr = ν/α is larger than 1 (note that since α here is the thermal diffusion rate
of the fluid, Pr > 1 implies that the momentum diffusivity dominates over the thermal one),
several correlations hold which relate the heat transfer rate to the fluid and the wall-shear
 Measurement of wall-shear stress 407

δt

Test surface

FIGURE 12.8 Schematic representation of a hot film and the thermal boundary layer growing
over the wall.

stress. Moreover, under certain conditions even the fluctuating component can be measured.
Hot films are basically metallic elements located at the wall, which are heated by means of a
constant temperature anemometer (cf. Chapter 9). A schematic representation of the thermal
boundary layer growing over the wall is shown in Figure 12.8, and it is also important to
note that some of the heat applied to the sensor is dissipated at the substrate and the rest is
transferred to the fluid. Hot films are thus only effective with fluids with larger conductivity
than the wall material, which are usually water and oil, but not air. Under this assumption,
Tavoularis [30] reports that in hot films with thermal boundary layers within the viscous sub-
layer the heat transfer rate from the sensor to the fluid is roughly proportional to the 1/3 power
of the wall-shear stress and formulates an empirical relation reminiscent of King’s law used
in hot-wire anemometry:

E2
= A + Bt1w/ 3 , (12.32)
Tw - T f

where
Tw and Tf are the film and bulk fluid temperatures, respectively
E is the voltage supplied to the sensor
A and B are calibration constants

A common approach to calibrate hot films is to use Preston tubes as a reference, with well-
characterized flows. Note that although they do not work well on the onset of separation and
cannot detect reversed flows, if mounted on a turntable they can identify the orientation of the
mean flow in case of highly 3D flows.

Wall hot wires As mentioned in the “Hot films” section, hot films are effective only when the fluid conduc-
tivity is larger than the one of the wall material and therefore are not recommended for mea-
surements in air. In this situation, it is better to use wall hot wires, which are basically single
hot-wire sensors mounted on the wall, and also lying within the viscous sublayer. A sche-
matic representation of this sensor is shown in Figure 12.9. Although the hot-wire sensitivity
increases with distance from the wall due to higher velocities, their frequency response remains
very good even very close to it, and therefore they are a suitable choice for measurements of
fluctuating components of the velocity. Also note that wall hot wires should be calibrated
against another device, such as Preston tubes, on well-behaved and properly characterized
flow conditions (such as in a canonical ZPG boundary layer). Fernholz et al. [60] show that,
due to heat transfer effects to the wall, the mean velocity measured by a hot-wire anemometer
calibrated in the freestream Um and the true velocity U in the near-wall region are related as
1/ 2
U m y Uy æ Uy ö
= - k1 ç ÷ + k2 , (12.33)
n n è n ø

where
y is the wall-normal coordinate
k1, k2 are calibration constants
408 Ricardo Vinuesa and Ramis Örlü

FIGURE 12.9 Schematic representation of a wall hot wire.

Note that Equation 12.33 holds within the viscous sublayer (y+ ≤ 5), and the constants take
the values k1 = 0.55 and k2 = 3.2 when wall hot wires of 5 μm diameter are used with high
thermal conductivity substrates. Following Fernholz et al. [60], it can be shown that for a
wall hot wire located at a distance h from the wall, the actual skin friction τw is related to the
measured value τm as

1/ 2
æ rn 2 ö rn 2
tm = tw - k1 ç 2 tw ÷ + k2 . (12.34)
è h ø h2

Also note that the measured value τm in Equation 12.34 is obtained in terms of the voltage
supplied to the sensor E by means of calibration laws similar to Equation 12.32, although with
different exponents depending on the experimental setup.

Wall pulsed wire The main difference between wall pulsed wires and wall hot wires discussed in the “Wall hot
wires” section is the fact that here a total of three wires are used as displayed in Figure 12.10:
the ­central wire is heated with an electric current, and the other two are sensing elements.
The measurement procedure is based on applying a very short electric pulse to the wire at the
center, which then heats the fluid around it by means of convection. Then the time that it takes
for the other elements to sense this convection (the so-called “time of flight” T) is related to the
wall shear stress. These devices exhibit excellent frequency response and therefore are capable
of measuring instantaneous wall shear stresses in separated and even reverse flows. It is also

FIGURE 12.10 Schematic representation of a wall pulsed wire.

 Measurement of wall-shear stress 409

possible to detect flow direction with wall pulsed wires. Whereas Castro et al. [61] propose the
following correlation between instantaneous wall shear stress and time of flight:

2 3
æ1ö æ1ö æ1ö
tw = A ç ÷ + B ç ÷ + C ç ÷ , (12.35)
T
è ø T
è ø èT ø

with A, B, and C being calibration constants. Several authors including Fernholz et al. [60]
show that this equation should be time averaged, which also impacts the registered time of
flight by the sensor. Combining calibration and time of flight measurement errors, this tech-
nique shows an overall uncertainty of around 4% in the wall shear stress determination.

12.6 Optical methods

Experimental techniques based on nonintrusive optical devices have become the method of
choice for skin-friction measurements in a wide range of academic applications. This is due
to the fact that they do not rely on assumptions regarding the functional form of the velocity
profile or the extent of its various layers but are based on theoretical relationships of a different
nature instead. For instance, the technique known as oil-film interferometry is based on the
relation between the thinning rate of an oil droplet deposited on the surface of interest and the
shear stress of the incoming flow. Their most significant limitations are the fact that special
coatings or additional equipment are usually required and the fact that additional assumptions
have to be made regarding the physical phenomena used to indirectly estimate wall shear.
Therefore, the price to pay for not avoiding assumptions about the flow to be measured is
the requirement of modeling another phenomenon instead, which may introduce additional
uncertainties.

Oil-film OFI is based on the relation between the motion of a thin film of oil and the wall shear of the
interferometry stream driving it. The initial idea of using thin oil films to measure skin friction was by Tanner
and Blows [62] in 1976, and since the 1990s its use in wall-bounded turbulence research has
increased significantly. Tanner and Blows [62] and Squire [63] show that the motion of the
oil film is determined by wall-shear stress, surface curvature of the interface between the oil
and the driving fluid, surface tension, pressure gradient, and gravity. The shear stress is the
dominant effect in the oil-film motion, and although there are studies aimed at incorporating
other contributions [64], the most widely used approaches nowadays are based on the method
proposed by Tanner and Blows [62] almost 40 years ago.
A schematic representation of the oil film driven by the incoming flow is shown in
Figure 12.11, where the film height is denoted by h and is a function of the streamwise loca-
tion x and time t in this 2D analysis. Wall-shear stress is related to the thinning of the oil,

Oil

Substrate

FIGURE 12.11 Schematic representation of the oil film driven by the incoming flow and Fizeau
interferometry phenomenon.
410 Ricardo Vinuesa and Ramis Örlü

and the method proposed by Tanner and Blows [62] to measure h(x, t) is based on a phe-
nomenon that can also be observed in Figure 12.11: the formation of Fizeau interferomet-
ric fringes. These fringes can be observed if the film is illuminated with a monochromatic
source of light, where a very popular approach is the use of sodium lamps (with wavelength
λsodium = 589.3 nm). These lamps are robust and inexpensive and solve most of the problems
exhibited by the initial approach proposed by Tanner and Blows [62] based on He–Ne lasers
(see the review by Naughton and Sheplak [10]). The thickness of the film determines the total
path length of incoming light within the oil film. As shown in Figure 12.11, if the total path
length is a multiple of λ (2λ and 3λ for the first and third fringes, respectively), the light ray
reflected from the film surface and the one reflected from the bottom wall are in phase, thus
producing a constructive interference. In the case of the second fringe, the total path length of
the light within the oil film is 2.5λ, so film-reflected and wall-reflected rays are out of phase,
leading to a destructive interference and a dark fringe. In fact, a more detailed analysis of
the optical relations between film-reflected and wall-reflected rays as shown in Figure 12.12
reveals that the condition required for a constructive interference to occur is that the difference
in path length satisfies the relation: ABC - AD = nl, where n = 1 , 2 , 3 , …. A series of images
are obtained with a camera during the experiment, resulting in a stack of images showing the
Fizeau interferometric fringes as in Figure 12.13.

D
A C Flow

Oil film

B Substrate

FIGURE 12.12 Optical analysis of the interaction between film-reflected and wall-reflected
light rays in an OFI experiment.

FIGURE 12.13 (See color insert.) Fizeau interferometric pattern exhibited by an oil film driven
by the air flow, where flow is from bottom to top and the time evolution is from left to right.
Measurement of wall-shear stress 411

The motion of an oil film of thickness h(x, t) driven by a fully developed 2D air shear stress
τ(x, h, t) is governed by the following equation:

¶h th ¶h
+ = 0, (12.36)
¶t m ¶x

where μ is the dynamic viscosity of the oil, and this equations neglects the effect of surface
tension, pressure gradient, and gravity. Equation 12.36 can be integrated to yield an expres-
sion of the oil-film thickness h as a function of the constant shear stress from the air stream τ,
the distance from the leading edge of the film x, and the time t as follows:

m x
h= , (12.37)
t t - t0

where t0 is the origin in time. In most applications, t ≫ t0, and thus the time origin can be
ignored. Following the optical setup shown in Figure 12.12 and interference optics rela-
tions (see Hccht [65]), the thickness difference between two consecutive fringes Δh can be
expressed as
l
Dh = , (12.38)
2 n - nair
2 2
oil sin 2a

where
λ is the light wavelength
nair and noil are the respective refractive indices (nair ≃ 1)
α is the angle between the observer and the wall-normal axes

A more detailed discussion about interference optics can be found in Chapter 8. The inter-
ferograms produced using this technique can be analyzed in a number of ways to estimate the
velocity of the fringes, and thus the wall-shear stress. It is important to note that the various
methods available for interferogram analysis require different levels of user interaction (which
introduces some subjectivity in the measurement) and rely on different amounts of informa-
tion from the interferograms.

The XT method The first method was initially proposed by Janke [66] and is based on ana-
lyzing the position (X) of the fringes as a function of time (T) in order to evaluate the fringe
velocity. Let us consider that at the location of a particular fringe k the oil-film thickness is
constant and equal to hk and can be expressed in terms of the film thickness at the location of
the first fringe h0 as

hk = h0 + k Dh. (12.39)

Combining Equations 12.37 and 12.39, it is possible to express the fringe velocity uk = dx/dt as

th t
uk = = ( h0 + k Dh ) . (12.40)
m m

The stack of images similar to the one shown in Figure 12.13 is analyzed, and the user needs
to manually select an interrogation line normal to the orientation of the fringes. Then it is pos-
sible to obtain an X versus T image, which in turn can be used to determine the fringe velocity
uk. With this value, and combining Equations 12.38 and 12.40, the wall-shear stress τ can be
computed as
2
2 noil - nair
2
sin 2 a
t = muk . (12.41)
l ( k + h0 /Dh )
412 Ricardo Vinuesa and Ramis Örlü

Since user interaction is required for determining the interrogation line, the level of error may
vary depending on the user’s experience. Algorithms for automated peak detection are also
available, although in general they do not provide more robust results than the ones obtained
through manual selection. This technique has been successfully used by several research
groups, such as Fernholz et al. [60] and Österlund et al. [67].

Local and global wavelength estimation methods Instead of computing the fringe velocity
uk based on the streamwise development of the fringes, it is possible to calculate the wall-shear
stress in terms of the mean peak distance of the interferometric pattern. Note that the wave-
length λf is calculated using geometrical properties as

-1
æ ¶h ö
l f = Dh ç ÷ , (12.42)
è ¶x ø

which can be rewritten as follows taking into account the relation 12.37:

tDht
lf = . (12.43)
m

Rearranging Equation 12.43 and taking the limit for very small t, it can be shown that the local
shear stress τ is computed as

m ¶l f
t= , (12.44)
Dh ¶t

and there are several ways of estimating the wavelength λf. Basically, this quantity can be com-
puted locally (by analyzing the spacing between consecutive fringes) or globally (­considering
several periods). These are the so-called local and global wavelength estimation methods.
A degree of subjectivity through user interaction is introduced in the selection of an image
strip along the centerline of the interferometer. This strip is averaged in the spanwise d­ irection
to reduce noise, resulting in a 1D signal s(x). After this, the signal is high-pass filtered with the
aim of removing any trend in the background illumination produced by l­ ighting of the oil film.
The local wavelength estimation method is based on using the derivative of s(x) as an esti-
mation of the peak positions. This estimation is then refined, and the distance between fringes
is computed at each peak position. These results are then averaged to yield the wavelength of
the signal. Whereas this approach is fast and accurate, it may exhibit false peak detection in
particularly noisy signals. An alternative to this is the global wavelength estimation method,
which relies on Fourier-transform integrals to maximize the correlation between s(x) and a
test function defined as a complex exponential. This approach is also fast and accurate and
presents the advantage of being relatively insensitive to noise.
After comparing the results obtained with the various processing techniques, Vinuesa et al.
[51] concluded that the global wavelength estimation method is most reliable and therefore
the method of choice. Differences between the local and global methods were found only
in moderately noisy signals, and the results were virtually identical when used with signals
exhibiting low noise levels. They also claim that the XT method is slightly less reliable since
its results are more user dependent. Moreover, Vinuesa et al. [51] compared the results also
compared the results from the global wavelength estimation method with the ones obtained by
means of the Hilbert transform approach by Chauhan et al. [68] and found any discrepancies
to lie within the uncertainty of the method.

Procedure for OFI measurements After describing the theoretical background of the
method, in this section we focus on the more practical aspects of the measurements, since to
our knowledge there is a lack of information regarding these practical aspects in the literature.
OFI requires the use of a surface with good reflective properties to allow the interferometric
Measurement of wall-shear stress 413

phenomena described in Figures 12.11 through 12.13 to take place. Common materials are
glass, nickel, and polished stainless steel, all of them with several advantages and drawbacks.
In general, polished stainless steel is a good choice since it exhibits good reflective properties,
is easy to manufacture models with different geometries of this material, and does not exhibit
problems with buildup of static electricity on its surface (unlike glass). A common approach is
to use a transparent upper cover of the test section so that the camera is located directly above
the oil film. However, in setups where the optical access is more difficult it is possible to use
mirrors to project the images on a side of the test section, where the camera is used to obtain
the images. Also, in flat-plate boundary-layer experiments, it is common to use removable
plugs to place oil drops, which are then assembled flush with the test surface for the measure-
ments. Significant errors in skin friction determination may be introduced if the plug surface
is not completely aligned with the test section: misalignments as small as 0.2 mm may lead to
2% error in the measured inner-scaled freestream velocity U ¥+ [46].
The necessary steps to carry out OFI measurements are described next.

Oil calibration A typical oil used in OFI is the Xiameter PMX or Dow Corning silicone
fluid, with viscosities ranging from 10 up to 500 cSt (let us recall that 1 cSt = 10−6 m2/s).
The PMX-200 silicone oil (with ν = 200 cSt) provides reliable and repeatable results, although
the optimum viscosity to use may change depending on the range of friction velocities.
The use of silicone oil is motivated by the fact that its viscosity is much less sensitive to tem-
perature changes than the one of other fluids with similar viscosities. It is extremely important
to calibrate the oil in order to obtain its viscosity as a function of temperature, since the actual
ν may be several percent different from the one reported by the manufacturer. Note that any
error in oil viscosity directly impacts the measured value of wall shear stress, thus highlight-
ing the importance of this measurement. The oil should be carefully calibrated using a con-
trolled bath and a capillary Ubbelohde-type or other viscosimeter (cf. Viswanath et al. [69] for
descriptions of various capillary viscometers) as shown in Figure 12.14. The viscosity curve
is usually represented by an exponential:

n = A exp ( a nT ) , (12.45)

where
the parameter αν is around −0.02°C−1 for silicone oils at ambient temperature
A is a calibration constant, which depends on the nominal viscosity
T is the temperature

Although in principle it should be possible to obtain accuracies on the order of 0.1% in this
kind of viscosimeter, in practice it is difficult to obtain the controlled conditions required for
that level of accuracy. Figure 12.15 shows the calibration curve of a 200 cSt oil in two differ-
ent facilities, which shows a repeatability of around ±0.3%. Along the lines of our previous
discussion regarding the relevance of an accurate calibration, the relative error of viscosity
is ανΔT, which means 2% difference for each degree of temperature. Note that this directly
affects the wall-shear stress measurements, and a 2% error is considerably large, especially as
wall-bounded turbulence conclusions are extrapolated to very high Reynolds numbers.
With respect to oil density, its variation with temperature is on the order of −0.06%/°C [51],
that is, around 30 times smaller than the change of viscosity. Thus, its impact on wall-shear
stress measurements is low, and for practical purposes oil density is assumed to be constant and
equal to the value provided by the manufacturer. Another parameter that does not significantly
vary with temperature is the oil refractive index and only depends weakly on the viscosity. Oil
manufacturers report that the refraction index of a 200 cSt oil may change approximately by
0.14% in a temperature range from 10°C to 50°C, whereas it changes by around 0.2% in the
viscosity range from 20 to 1000 cSt.

Experimental setup The first step after obtaining a proper characterization of the oil to
be used is to calibrate the camera. Before every set of runs, it is necessary to determine
414 Ricardo Vinuesa and Ramis Örlü

(a) (b)

FIGURE 12.14 (a) Ubbelohde viscosimeters and (b) thermal bath used for calibration. Setup
used at the Fluid Physics Laboratory, KTH Mechanics, Sweden.

250
Calibration at IIT
245
Calibration at EPFL
240 Österlund et al. [67] ±1%

235

230
ν (cSt)

225

220

215

210

205

200
14 16 18 20 22 24 26
T (°C)

FIGURE 12.15 Calibration curve of viscosity for silicon oil with nominal ν = 200 cSt at 25°C.
(With kind permission from Springer Science + Business Media: Exp. Fluids, New insight into flow
development and two dimensionality of turbulent channel flows, 55, 2014, 1759, Vinuesa, R.,
Bartrons, E., Chiu, D., Dressler, K.M., Rüedi, J.-D., Suzuki, Y., and Nagib, H.M.)

the calibration factor of the camera by means of a plug with a calibration grid as shown in
Figure 12.16. This plug is placed on the location where oil drops will be tested, and an image
is taken with the camera. This image is then magnified to pixel resolution in order to manu-
ally select marks of the millimeter grid along the line passing through the ­symmetry axis
of the drop. This calibration process exhibits a repeatability better than 0.1%.
Measurement of wall-shear stress 415

FIGURE 12.16 Plug with grid used to obtain the calibration factor of the camera.

Following the experimental procedure described by Bartrons and Muñoz [70], before the
measurements it is necessary to clean the plug (or surface where the oil drop will be deposited)
to prevent the drop from contaminating static particles located at the surface. The surface also
needs to be deionized in order to avoid electrostatic effects, which eventually lead to distorted
fringes in the interferograms. It is also common to use more than one drop per run to increase
the accuracy of the measurements, although the ones closer to the center of the plug are usu-
ally most reliable. A total of three or four drops are placed on the upstream side of the plug
to allow them to develop under the action of the incoming stream, with sufficient separation
between them to avoid mutual interaction. The silicone oil drops are placed on the plug with
a needle. It is important to run the experimental facility for several minutes before the OFI
measurements start for two reasons: first, to allow flow conditions to stabilize, and second to
let the plate heat up to reach the air temperature. Based on our experience in various facilities
of different sizes and velocity ranges, somewhere between 15 and 20 min should be sufficient
to achieve both goals. During the OFI runs, images are obtained at constant time intervals that
depend on oil viscosity and wall shear stress and are usually between 1 and 10 s. The total
duration of the experiment also depends on the characteristics of the oil and flow case but runs
usually last around 15 min approximately. A common configuration includes using a digital
camera to capture images, as well as a low-pressure sodium lamp to illuminate the oil film.
Vinuesa et al. [51] used a Nikon D5000 SLR camera with 18–55 mm zoom lens and a Philips
SOX-35 lamp. It is also important to monitor the temperature during the experiment in order
to ensure stable thermal experimental conditions.

Data processing After acquiring the images, they are cropped around the area of the oil
drops and are stored in a computer file. As discussed in the"Local and global wavelength esti-
mation methods" section, the method of choice is the global wavelength estimation technique
due to its high accuracy and reliability, and thus the procedure described here is based on that
method. The first images of the stack, where the fringes have not been formed yet, should be
removed. Once the final stack is selected, the images are loaded in a post-processing program,
where they are transformed into matrices in which each value represents the light intensity of
a pixel in the original image.
Then an adequate interrogation region must be chosen. A good strategy is to superimpose
the first image (which is brighter) and the last one, as in Figure 12.17. The combination of both
images allows a more objective selection of the interrogation region, valid for the whole run.
Note that in this example a total of four oil drops are used on an aluminum plug, which exhibits
sufficient reflectivity for proper image processing. Then the user selects interrogation regions
on the four drops, which are usually located around the symmetry axis of the drop (avoiding its
leading edge). If the interferometric pattern is noisy around the symmetry line, it is possible to
use another region slightly shifted horizontally from the centerplane of the drop. But in any case,
416 Ricardo Vinuesa and Ramis Örlü

0
Select area to analyze,
press return to continue
200

400

600

x (pixels)
800

1000

1200

1400
0 500 1000 1500 2000
z (pixels)

FIGURE 12.17 Interferogram corresponding to four oil films and selection of an interrogation
region.

it is important to select regions perpendicular to the direction of the shear, not affected by the
sides of the drops and with minimum curvature. This is also done to minimize 3D effects within
the drop, since as discussed earlier the OFI theory assumes 2D oil-film motion. Other aspects
to keep in mind when selecting the interrogation region are the avoidance of wiggly regions
(affected by electrostatic effects), or areas contaminated by particles of dust. Then the selected
interrogation region for each drop is averaged in the spanwise direction, leading to 1D intensity
signals as shown in Figure 12.18. Note that Figure 12.18a shows how the first image, even if it
already shows interferometric fringes, still has not developed a sinusoidal pattern as in the later
stages of the experiment. Figure 12.18b shows a fully developed state of the oil drop, where
the wavelength λf can be clearly determined. It is also useful to allow the flexibility to select
the first and last images to be processed in order to obtain λf. An example of this is presented

120

115
Light intensity

110

105

100
0 50 100 150 200 250 300 350
(a) x (pixels)

100
Light intensity

95

90

85
0 50 100 150 200 250 300 350
(b) x (pixels)

FIGURE 12.18 Spanwise-averaged intensity signal as a function of streamwise distance for

(a) first and (b) last images in the stack.
Measurement of wall-shear stress 417

Selection flag
0.5

0
0 50 100 150 200 250 300 350 400 450
(a) Image #
100

Light intensity
95
90
85
80
0 50 100 150 200 250 300 350
(b) x (pixels)
100
Light intensity

95
90
85
80
0 50 100 150 200 250 300 350
(c) x (pixels)

FIGURE 12.19 Example showing the result of discarding the first 21 images of the stack (a).
In this case, the first image to be processed is number 22 (b), and the last image in the stack is
shown in panel (c). Spanwise-averaged intensity signals shown in both cases.

in Figure 12.19, where the first 21 images in the stack are discarded, and the corresponding light
intensity ­signals are shown. These are then used in the wall-shear stress calculation.
After carefully selecting the set of images to be used, the next stage of the process is to
compute λf using the global wavelength estimation method. Doing so, a representation of λf as
a function of time t can be obtained as in Figure 12.20. Note that as the film develops the dis-
tance between fringes progressively increases, but the relevant parameter here is the thinning
rate dλf/dt, which as shown in Figure 12.20 is approximately constant. The idea is to perform
a linear fit in the λf versus t plot to obtain the slope dλf /dt, which can then be used together
with Equations 12.44 and 12.38 to calculate the wall-shear stress τ. Figure 12.20 also shows
the relative error in dλf/dt from the individual images, and the accumulated value of the fric-
tion velocity uτ in terms of the number of images considered to determine the oil-film thinning
rate. Then the range of images that minimizes the error in dλf /dt is selected, which as a general
criterion should be kept below ±1%. Also, the accumulated uτ trend should be smooth and
converge to a constant value toward the end of the stack under consideration. This process is
repeated for the four drops used in the run, and the average value between them (considering
that outliers may have to be discarded, especially with drops exhibiting problematic interfero-
grams) is the measured friction velocity.

Outline of post-processing code In this section, we highlight the most relevant ­features
of a post-processing code to analyze interferograms using the global wavelength e­ stimation
method:
1. Calibration factor of the camera.
a. Transform image to matrix data in the format of the post-processing program.
b. Select reference distance on the digitized image.
c. Introduce calibration factor connecting the number of pixels with the actual ­physical
distance Δx.
418 Ricardo Vinuesa and Ramis Örlü

×10–3
2

1.5

λf (m)
1

0.5

0 50 100 150 200 250 300 350 400 450

(a) t (s)

0.5

0
e (%)
–0.5

–1
0 50 100 150 200 250
(b) Image #
uτ (m/s)

1.75

1.7
0 50 100 150 200 250
(c) Accum. # of images

FIGURE 12.20 Computed fringe wavelength λf as a function of time t; the range used to obtain
the slope dλf /dt highlighted in black (a). Relative error between the dλf /dt obtained from a
­particular image and the one computed from the whole stack (b). Accumulated value of friction
velocity uτ as a function of the number of images considered for its calculation (c). Note that in
this particular example the time interval between images was 1 s.

2. Generation of a stack of images to process.

a. Select a data set to process and transform all the images to matrix data in the format
of the post-processing program.
b. Use the first and last images of the stack to select a useful region containing the
drops to process, and crop all the images.
c. Sort files and save the cropped stack.
3. Introduce parameters characterizing the OFI run.
a. Optical parameters: camera calibration factor Δx (m/pixel), time-step between
images Δt (s), light wavelength λ (nm), angle between camera axis and wall-normal
direction α (°).
b. Oil properties: oil refraction index noil, oil density ρ (kg/m3), oil calibration ­parameters
A (cSt) and B (°C−1), oil temperature Toil (°C), and oil kinematic viscosity ν (cSt).
c. Air properties: atmospheric pressure Patm (Pa), air temperature Tair (°C), and
freestream velocity U∞ (m/s).
4. Calculate wall-shear stress using the global wavelength estimation method.
a. Load configuration file and image stack for the run to post-process.
b. Initialize a graphic user interface (GUI) where the user can select the interroga-
tion region for each of the drops under consideration. A superposition of the first
(brighter) and last images of the stack is usually considered in this step.
Measurement of wall-shear stress 419

c. Average light intensity in the spanwise direction to obtain a 1D ­intensity signal s(x).
d. Obtain, for each image in the stack, the fringe wavelength λf by maximizing the
correlation between s(x) and a complex integral. A fast Fourier transform (FFT) is
combined with signal smoothing to yield the best estimation of λf(t).
e. Show the resulting λf as a function of time t and allow to select a different range of
images within the stack to minimize the error in the slope dλf/dt.
f. Display the resulting friction velocity uτ as a function of the images selected from
the stack for this calculation.

Comparison of OFI with other shear-stress measurement techniques After having

described several widely used methods for skin-friction measurement, here we compare
the most relevant features of OFI with other techniques. Fernholz et al. [60] performed a
comparative study of sublayer fence, wall hot wire, wall pulsed wire, and OFI methods,
which is summarized in Table 12.1. An interesting conclusion of this study is the fact that
Preston tubes should not be used in reversed flows or downstream of reattachment. In
these flows, wall hot wires, wall pulsed wires, and OFI perform well, although sublayer
fences show small deviations with respect to the results from the other techniques due to
small asymmetries of the probe. The uncertainty of sublayer fences, wall hot wires, and
wall pulsed wires was quantified by Fernholz et al. [60] to be around ±4%, whereas they
show that OFI exhibits uncertainty values below this level. Interestingly, the uncertainty
of OFI has reduced significantly over the last decades, from the value of around 5%
reported by Ferhnolz and Finley [71] in 1996 to 1%, documented in the recent studies by
Rüedi et al. [72] in 2003 and Nagib et al. [32] in 2004. These studies also show that the
1% uncertainty can be potentially reduced to 0.5% with careful experimental procedures
and calibration. Vinuesa et al. [51] recently estimated the uncertainty of various OFI
quantities in a rectangular duct flow facility: 0.85% in the wall shear stress τw, 0.58%
in the friction velocity uτ, and 0.87% in the inner-scaled centerline velocity U c+ . This
highlights the fact that OFI is the most accurate way of measuring local mean wall-shear
stress in turbulent flows.
Some extensions of the standard procedure of the OFI technique described in this sec-
tion include its application using white light (proposed by Desse [73]), its implementation
on aerodynamic models in larger wind tunnels closer to industrial applications (discussed
by Driver [74]), and the possibility of using dual- and multi-image analysis as suggested by
Naughton and Hind [75].

Table 12.1 Comparison of features from skin friction measuring methods

Feature Sublayer fence Wall hot wire Wall pulsed wire OFI

Classification Pressure Heat transfer Heat transfer Optical

Measurement Pressure Heat transfer Time of flight Fringe
difference rate advection
Calibration Yes Yes Yes No
Mean τw Yes Yes Yes Yes
Temporal resolution Unclear 10 kHz 20 Hz No
Spatial resolution Δx 1 mm 0.5 mm 1.5 mm 1 mm
Spatial resolution Δz 3 mm 0.5 mm 0.5 mm 1 mm
Direction of τw Yes Yes Yes Yes
APG Yes Yes Yes Yes
Reverse flow Yes Yes Yes Yes
FPG Yes Yes Yes Yes
3D flows Yes Yes Yes Yes
Source: Adapted from Fernholz, H.H. et al., Meas. Sci. Technol., 7, 1396, 1996.
420 Ricardo Vinuesa and Ramis Örlü

Laser Doppler As sketched in Figure 12.21, in this approach a laser beam is passed through a diffractive
technique lens below the test section, where two narrow gaps are placed. The beam is diffracted by
the two gaps, leading to interferometric patterns where the fringes exhibit a spacing d that is
proportional to the laser wavelength λ and inversely proportional to the separation between
the gaps s. In fact, this technique is an adaptation of the popular laser Doppler velocimetry
(LDV) method used for velocity measurements, which is discussed in detail in Chapter 10.
The idea is to seed particles on the flow, which will scatter the light from the laser beam at the
Doppler frequency fD (see the work by Naqwi and Reynolds [76] and Fourguette et al. [77]
for additional reference information). Then the light is received with a lens and projected on
a photodetector that allows the measurement of fD. In the viscous sublayer the velocity profile
is linear, and therefore all the seeded particles scatter light with the same Doppler frequency,
a fact that can be used to determine the wall-shear stress as
ml
tw = fD , (12.46)
s

where μ is the fluid dynamic viscosity. Although this method exhibits problems associated
with seeding and shows limitations due to the behavior of the particles close to the wall, it is
able to detect flow reversal, does not require calibration, and shows good frequency response.
Thus, it allows to obtain accurate measurements of the fluctuating component of the wall-
shear stress. Another related optical approach is the use of micro particle image velocimetry
(PIV) for velocity and wall-shear stress measurements, as reported by Kähler et al. [78].

Liquid crystal The use of liquid crystal coatings was traditionally aimed at temperature mapping and flow
coating techniques visualization purposes. However, more recent studies starting with the work by Reda and
Muratore [79] in 1994 have extended their applicability to measurements of skin friction. In
this ­technique, the surface under consideration is covered with a liquid crystal coating and is
illuminated with white light in a direction normal to the surface forming an angle of around
15° [10]. If no shear stress is present the coating exhibits a red-orange color, whereas an
observer positioned upstream of the flow will perceive a tendency toward blue as wall shear
is increased. Interestingly, no color changes will be perceived by an observer located down-
stream. The shear stress acting on the crystal modifies its physical properties, thus affecting
the spectral properties of the light reflected from it. A number of images are acquired with
digital cameras from different angles and orientations in order to properly resolve the effect of
wall shear on the crystal. Besides, color change depends not only on the magnitude but also
on the orientation of the shear.

s
Test surface

Photodetector

Laser beam

FIGURE 12.21 Schematic representation of the experimental setup of a laser Doppler device for
wall-shear stress measurements. (Adapted from Tavoularis, S., Measurement in Fluid Mechanics,
Cambridge University Press, Cambridge, UK, 2005.)
 Measurement of wall-shear stress 421

Although this method still shows a number of challenges in its practical implementation,
there is potential for future developments in terms of accuracy and reliability [80,81].

Micro-pillar In the previous sections, we discussed several methods for skin friction measurement, exhibit-
sensors ing different levels of accuracy when determining both the mean and the fluctuating compo-
nents of the wall shear stress. We showed that the most accurate way of determining mean
wall-shear stress is through the use of OFI, which combined with the global wavelength esti-
mation method may reach uncertainty levels on the order of ±1%. However, this method is
unable to predict the fluctuating wall shear, and the ones that are able to do it (wall hot wire
and wall pulsed wire) show much larger error levels on the order of ±4%. In this section, we
present another optical method that is able to accurately measure the fluctuating wall shear:
the micro-pillar shear stress sensor. Introduced by Brücker et al. [82] and Große and Schröder
[83], this sensor consists of an array of flexible micro-pillars on the wall of the flow under
consideration. The idea behind this technique is to exploit the relation between the deflection
of these micro-pillars and the shear stress, a concept inspired by the motion of corn heads
under the action of the wind.
A schematic description of this sensor, which is able to measure unsteadiness and the
direction of wall-shear stress in wall-bounded turbulent flows, is shown in Figure 12.22. The
deformation of the pillar, characterized by the displacement of the pillar tip Δt, is connected
with the magnitude of the shear and also with its direction. This is due to the fact that the
cylindrical cross section leads to the same stiffness in all wall-parallel directions, thus leading
to high directional sensitivity. Gnanamanickam et al. [84] showed that assuming the linear
bending theory of a circular beam the tip displacement can be connected with the wall-shear
stress τw as follows:
4
112 1 æ L p ö
D t  tw ç ÷ Lp, (12.47)
9 E è Dp ø

where
Lp and Dp are the length and diameter of the pillar, respectively
E is its Young’s modulus

Note that the micro-pillar has to be immersed in the viscous sublayer so that it is subjected
to a linear velocity profile. This limits the value of Lp to around 60–1000 μm [84]. Although
Equation 12.47 gives an indication of the relation between Δt/Lp and τw, micro-pillars have to
be calibrated, usually on Couette flows where the velocity profile is linear. Gnanamanickam
et al. [84] showed that Δt/Lp is linearly proportional to τw when Δt/Lp ≃ 0.2, but higher tip dis-
placement values lead to nonlinear relations. This is why a specific micro-pillar sensor has to
be designed and calibrated for the particular flow configuration being tested. Optical systems
and high-resolution cameras are used to measure the deflection of micro-pillars. The tip is
usually coated with some particular color to easily identify its deflection Δt. Whereas shear
stresses as low as 10−2 N/m2 can usually be measured with this technique, its spatial resolution
is on the order of 5 viscous units.

Δt

Lp

Dp
Test surface

FIGURE 12.22 Schematic representation of an MPS3.

422 Ricardo Vinuesa and Ramis Örlü

It is important to note that due to the multi-scale nature of turbulent flows, the motion of the
micro-pillar will be determined by a range of temporal frequencies that may be on the order
of the kHz in the case of the smallest scales. This means that when used with fluids with high
viscosity like water the system is overdamped, but with air the system exhibits a low-pass filter
behavior, with a prominent resonance peak. Although Große et al. [85] ­considered a frequency-
dependent added mass to encapsulate the dynamic behavior of the sensor, Gnanamanickam
et al. [84] proposed the use of dynamic calibrations. They showed that the dynamic behavior
of the micro-pillar can be described by

¶ 4 w ( y, t ) ¶ 2 w ( y, t )  ¶w ( y, t )
EI
¶y 4
+ 
m ( St ) ¶t 2
+ D ( St )
¶t
= F ( y, t ) , (12.48)

where
EI is the pillar stiffness
w is the lateral displacement at a particular wall-normal location y at an instant t
St is the Strouhal number
F(y, t) is the excitation, whereas m and D are the reduced mass and damping coefficients,
respectively

Note that the Strouhal number is defined as St = fDp/U∞, where f is the frequency. A dynamic
calibration analysis [84] shows that micro-pillars exhibit a roughly constant transfer func-
tion in all the frequencies below 0.3–0.4f0. Note that this frequency needs to be determined
by ­considering an aeroelastic problem (due to the added mass consequence of the displaced
fluid), as confirmed by comparison with experimental boundary-layer data [86].
The MPS3 is a reliable and robust experimental method that provides accurate measure-
ments of the fluctuating component of τw and is the only technique able to measure the spatial
correlations of wall shear stress fluctuations. Although OFI is still the method of choice to
accurately determine the mean wall-shear stress, this technique is very promising and will
progressively become more widely used with further development of the method and post-
processing techniques in the coming years. For instance, micro-pillars allow to measure
backflow events, which are difficult to measure using other techniques [87], and are very
relevant to the mechanisms of flow separations [88]. Some of the issues to address would be
the problems with resonance frequency or the need to have the pillar fully submerged in the
viscous sublayer, which still limits its applicability at low Reynolds numbers. As discussed
by Brücker [89], in high-Re applications, the micro-pillar is too long, and therefore it can
influence the flow near the wall.

Acknowledgments

The authors thank H.M. Nagib, J.-D. Rüedi, E. Bartrons, and M. Muñoz for sharing some of
the material related to oil-film interferometry measurements.

Problems

12.1 Derive the thin oil-film equation 12.36 discussed in this chapter. To do so, consider the
following steps:
(i) Use a control volume enclosing a thin oil film developing in the streamwise direc-
tion. The differential oil film will have lengths dx and dz in the streamwise and
spanwise directions and will have an initial thickness h. A convective velocity
h

geometry.
0 ò
U c = 1/h u dy defines the oil-film motion in x. Perform a mass balance in this
Measurement of wall-shear stress 423

(ii) An oil film with h  = 1 μm and ν  = 100 cSt has an approximate Reynolds number
of 10−8 [10]. Keeping this in mind, use the x-momentum equation to find an expres-
sion for the streamwise velocity within the oil film u.
(iii) Integrate the result obtained in (ii) to compute the convective velocity and find an
expression for the thin oil-film equation.
(iv) Note that the equation found in (iii) involves the streamwise pressure gradient, but
not surface tension effects. According to Brown and Naughton [90], the oil-film
surface curvature ∂2h/∂x2 can be used to calculate the oil-film pressure as

¶ 2h
P0 = P - s , (12.49)
¶x 2

where
P is the aerodynamic pressure
σ is the surface tension

Using this definition, extend the previous thin oil-film equation to incorporate
­surface-tension effects.
(v) Assess the relative importance of the various terms in the previous equation and the
validity of Equation 12.36 discussed in the “Oil-film interferometry” section.
12.2 An MPS3 system is under consideration to accurately determine fluctuating stresses in
wind-tunnel experiments of TBLs. The first step is to characterize the static response of
the pillar in a well-known flow configuration such as the Couette flow, which exhibits
a linear velocity profile. A 200 μm long pillar of aspect ratio Lp/Dp = 10 is calibrated in
such configuration, with the following results:

Δt (μm) 8.2 16.4 26.4 35 40.2 46.2 52.4 58.4

τw (N/m2) 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Δt (μm) 63.4 69.1 75.6 78.4 83.6 87.4 88.4 92.6
τw (N/m2) 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6

(i) Determine Young’s modulus of the pillar.

(ii) Obtain the static calibration curve of the sensor.
12.3 A wind-tunnel experiment is being carried out to measure the inner-scaled velocity pro-
file of a TBL subjected to a zero streamwise pressure gradient. Velocity measurements
are performed by means of hot-wire anemometry at a distance x = 9 m from the leading
edge of the flat plate where the boundary layer is developing. The freestream velocity
is U∞ = 40 m/s and the operating fluid is air. Previous experience with this configura-
tion shows that an approximate Reynolds number based on momentum thickness of
Reθ = 30,000 will be obtained. In order to determine the friction velocity, the possibility
of using two Preston tubes of two different diameters (0.5 and 5 mm, respectively) is
considered. When the smaller Preston tube is used, the measurement is 209.13 Pa above
the static pressure, whereas in the larger one the reading is 416.98 Pa.
(i) Determine the friction velocity measured by each of the two Preston tubes.
(ii) OFI measurements on the same flow case yield a friction velocity value of 1.3 m/s.
Discuss the accuracy of the previous results and the impact of the various factors
influencing these particular Preston tube measurements.
12.4 An experimental setup to perform OFI measurements in a wind tunnel has been devel-
oped in order to accurately determine the wall-shear stress. A Philips SOX-35 low-­
pressure sodium lamp is used for illumination, and the angle between the camera axis
and the wall-normal direction is 24°. A Nikon D5000 SLR camera with an 18–55 mm
424 Ricardo Vinuesa and Ramis Örlü

zoom lens is used to capture images. Its calibration factor is 2.7 × 10−5 m/pixel, and
images are taken every 2 s. A 200 cSt Xiameter PMX silicone fluid is used for the
experiments, and a careful characterization of its properties is required to ensure accu-
rate measurements. The measured oil density and refraction index agree very well with
the values provided by the manufacturer: 970 kg/m3 and 1.403. The oil viscosity is
carefully measured in a wide range of temperatures, leading to the calibration curve:
ν = A exp (ανT), where A = 0.00032 m2/s and αν =  − 0.02°C−1.
During the experiment, a single drop was deposited on the plug, and the air tempera-
ture was 25°C. After processing the whole stack of images with an appropriate inter-
rogation region, the global wavelength estimation method is used to obtain the fringe
wavelength distribution with time λf(t). The following values were extracted from the
complete data set:

t (s) 0 12 30 40 50 60 70 80
λf × 104 (m) 1.1 1.8 3.1901 3.6 4.007 4.4094 4.7999 5.1982
t (s) 90 100 110 120 130 140 150 160
λf × 104 (m) 5.6107 6.04 6.3871 6.7922 7.2162 7.6118 8.0158 8.4337
t (s) 170 180 190 200 210 220 230 240
λf × 104 (m) 8.8201 9.2602 9.5782 9.9755 10.446 10.81 11.196 11.617
t (s) 250 260 270 274 282 292 298 300
λf × 104 (m) 12.010 12.412 12.859 12.920 14.3 15.8 14.8 14.0

Given the information described earlier, determine the friction velocity uτ.
12.5 The ZPG TBL is one of the most widely studied canonical cases due to its relatively
simple geometry and the interesting features associated with the streamwise develop-
ment. The following velocity profile was measured in a wind tunnel test using hot-wire
anemometry [25]:

y (mm) 0.0815 0.1145 0.1542 0.2037 0.2653 0.3402 0.4339 0.5495

U (m/s) 4.526 5.947 7.508 8.873 10.028 10.959 11.702 12.489
y (mm) 0.6895 0.8644 1.0790 1.3457 1.6712 2.0763 2.5729 3.1863
U (m/s) 13.044 13.602 14.049 14.574 15.011 15.495 16.024 16.464
y (mm) 3.9413 4.8690 6.0149 7.4275 9.1689 11.3165 13.9648 17.2277
U (m/s) 16.958 17.500 17.986 18.505 18.969 19.482 20.096 20.622
y (mm) 21.2473 26.2047 32.3163 39.8496 49.1368 60.5865 74.7006 92.0996
U (m/s) 21.278 22.037 22.824 23.740 24.722 25.701 26.407 26.565
y (mm) 113.5452 139.9897
U (m/s) 26.545 26.526

The wind tunnel was operated with a freestream velocity of U∞ = 26.55 m/s, and the
resulting Reynolds number based on momentum thickness was Reθ = 14,300.
(i) Use the Clauser chart to estimate the friction velocity.
(ii) Compare the result obtained in (i) with the one obtained during the experiment
using OFI: uτ = 0.9175 m/s. What is the relative error? Plot the velocity profile in
inner scaling using both friction velocities, and compare both distributions. Discuss
the most relevant mean flow features, such as the extent of the logarithmic layer,
values of the log-law parameters, etc.
(iii) Is it possible to use any points within the viscous sublayer to determine the friction
velocity? Comment on the accuracy of this approach, and its potential application
in high Reynolds number flows.
Measurement of wall-shear stress 425

References

1. M a r u s ic I (2009). Unravelling turbulence near walls, J. Fluid Mech., 630, 1–4.

2. George WK (2006). Recent advancements toward the understanding of turbulent boundary
­layers, AIAA J., 44, 2435–2449.
3. Winter KG (1977). An outline of the techniques available for the measurement of skin friction in
turbulent boundary layers, Prog. Aerosp. Sci., 18, 1–57.
4. Allen JM (1977). Experimental study of error sources in skin-friction balance measurements,
J. Fluids Eng., 99, 197–204.
5. Schultz-Grunow F (1940). New frictional resistance law for smooth plates, NACA Technical
Memorandum 986.
6. Dhawan S (1951). Direct measurements of skin friction, PhD thesis, California Institute of
Technology, Pasadena, CA.
7. Winter KG, Gaudet L (1970). Turbulent boundary-layer studies at high Reynolds numbers
at Mach numbers between 0.2 and 2.8, RAE Technical Report No. 70251. Ministry of Aviation
Supply, Royal Aircraft Establishment, RAE.
8. Haritonidis JH (1989). The measurement of wall shear stress, in: Gad-el-Hak M, ed., Advances
in Fluid Mechanics Measurements, Lecture Notes in Engineering, Springer-Verlag, Berlin,
Germany, pp. 229–261.
9. S a d r R, K l e wicki JC (2000). Surface shear stress measurement system for boundary layer
flow over a salt playa, Meas. Sci. Technol., 11, 1403–1413.
10. N au g h t o n JW, Shep lak M. (2002). Modern developments in shear-stress measurement,
Prog. Aerosp. Sci., 38, 515–570.
11. S a h l in A, J o h ans s on AV, A lf reds s on PH (1988). The possibility of drag reduction by
outer layer manipulators in turbulent boundary layers, Phys. Fluids, 31, 2814–2820.
12. M o r i K, Ima n is hi H, Ts uji Y, H attori T, M ats ubara M, M ochizuki S, Inada M,
K a s iwag i T (2007). Direct total skin-friction measurement of a flat plate in zero-pressure-­
gradient boundary layers, Fluid Dynamics Res., 41, 021406.
13. G a d -e l -H a k M (1987). The water towing tank as an experimental facility, Exp. Fluids, 5,
289–297.
14. Baars WJ, Squire DT, Talluru KM, Abbassi MR, Hutchins N, Marusic I (2016). Wall-drag
­measurements of smooth-and rough-wall turbulent bo undary layers using a floating element.
Exp. Fluids, 57, 90.
15. Te n n e k e s H, L um ley JL (1972). A First Course in Turbulence, MIT Press, Cambridge, MA.
16. P o pe S (2000). Turbulent Flows, Cambridge University Press, New York.
17. Millikan CB (1938). A critical discussion of turbulent flows in channels and circular tubes,
Proceedings of the Fifth International Congress on Applied Mechanics, Cambridge, MA,
pp. 386–392.
18. George WK, Castillo L (1997). Zero-pressure-gradient turbulent boundary layer, Appl. Mech.
Rev., 50, 689.
19. Monkewitz PA, Chauhan KA, Nagib HM (2008). Comparison of mean flow similarity laws in
zero pressure gradient turbulent boundary layers, Phys. Fluids, 20, 105102.
20. M a r u s ic I, M o nty JP, H ultm ark M, Smits A (2013). On the logarithmic region in wall
turbulence, J. Fluid Mech., 716, R3.
21. N ag ib HM, C h auhan KA (2008). Variations of von Kármán coefficient in canonical flows,
Phys. Fluids, 20, 101518.
22. Lee M, Moser RD (2015). Direct numerical simulation of turbulent channel flow up to
Reτ = 5200, J. Fluid Mech., 774, 395–415.
23. Schulz MP, Flack KA (2013). Reynolds-number scaling of turbulent channel flow, Phys. Fluids,
25, 025104.
24. Bailey SCC, Vallikivi M, Hultmark M, Smits AJ (2014). Estimating the value of von Kármán’s
constant in turbulent pipe flow, J. Fluid Mech., 749, 79–98.
25. Österlund JM (1999). Experimental studies of zero pressure-gradient turbulent boundary-layer
flow, PhD thesis, Royal Institute of Technology, Stockholm, Sweden.
26. Örlü R, Fransson JHM, Alfredsson PH (2010). On near wall measurements of wall bounded
flows—The necessity of an accurate determination of the wall position, Prog. Aerosp. Sci., 46,
353–387.
27. Vinuesa R, Schlatter P, Nagib HM (2014). Role of data uncertainties in identifying the
logarithmic region of turbulent boundary layers, Exp. Fluids, 55, 1751.
28. Clauser FH (1954). Turbulent boundary layers in adverse pressure gradients, J. Aero. Sci., 21,
91–108.
426 Ricardo Vinuesa and Ramis Örlü

29. Wil l m a r t h WW, Lu SS (1972). Structure of the Reynolds stress near the wall, J. Fluid Mech.,
55, 65–92.
30. Tavo u l a r is S (2005). Measurement in Fluid Mechanics, Cambridge University Press,
Cambridge, UK.
31. Ozarapoglu V (1973). Measurements in incompressible turbulent flows, PhD thesis, Laval
University, Quebec City, Quebec, Canada.
32. N ag ib HM, Chris top horou C, R üedi J-D, M onkewitz PA, Ö s terlund JM,
G r ava n t e S (2004). Can we ever rely on results from wall-bounded turbulent flows without
direct measurements of wall shear stress? 24th AIAA Aerodynamic Measurement Technology and
Ground Testing Conference, June 28–July 1, 2004, Portland, OR.
33. Vinuesa R, Rozier P, Schlatter P, Nagib HM (2014). Experiments and computations of
localized pressure gradients with different history effects, AIAA J., 52, 368–384.
34. Va n D r ie s t ER (1956). On turbulent flow near a wall, J. Aero. Sci., 23, 1007–1011.
35. S pa l d in g DB (1961). A single formula for the “law of the wall”, J. Appl. Mech., 28, 455–458.
36. Ch au h a n KA, M onkewitz PA, Nagib HM (2009). Criteria for assessing experiments in
zero pressure gradient boundary layers, Fluid Dyn. Res., 41, 021404.
37. M u s k e r AJ (1979). Explicit expression for the smooth wall velocity distribution in a turbulent
boundary layer, AIAA J., 17, 655–657.
38. N ic k e l s TB (2004). Inner scaling for wall-bounded flows subject to large pressure gradients,
J. Fluid Mech., 521, 217–239.
39. D u r s t F, K ik u ra H, Lekakis I, J ovanovic, Ye Q (1996). Wall shear stress determination
from near-wall mean velocity data in turbulent pipe and channel flows, Exp. Fluids, 20, 417–428.
40. Alfredsson PH, Örlü R, Schlatter P (2011). The viscous sublayer revisited–exploiting self-
similarity to determine the wall position and friction velocity, Exp. Fluids, 51, 271-280.
41. B id we l l JM (1951). Application of the von Kármán momentum theorem to turbulent boundary
layers, NACA Technical Note 2571. Langley Aeronautical Laboratory, Langley Field, VA.
42. Dutton RA (1956). The accuracy of the measurement of turbulent skin friction by means of
surface Pitot-tubes and the distribution of skin friction on a flat plate, Aeronautical Research
Council Reports and Memoranda, 3058. Ministry of Supply, London, UK.
43. Mehdi F, White CM (2011). Integral form of the skin friction coefficient suitable for experimental
data, Exp. Fluids, 50, 43–51.
44. M e h d i F, J o h a ns s on TG, W hite CM, N aughton JW (2013). On determining wall shear
stress in spatially developing two-dimensional wall-bounded flows, Exp. Fluids, 55, 1656.
45. Ba il e y SCC, H ultm ark M, M onty JP, A lf reds s on PH, C hong MS, D uncan RD
et al. (2013). Obtaining accurate mean velocity measurements in high Reynolds number turbulent
boundary layers using Pitot tubes, J. Fluid Mech., 715, 642–670.
46. Vin u e s a R (2013). Synergetic computational and experimental studies of wall-bounded turbu-
lent flows and their two-dimensionality, PhD thesis, Illinois Institute of Technology, Chicago, IL.
47. Vinuesa R, Nagib HM (2016). Enhancing the accuracy of measurement techniques in high
Reynolds number turbulent boundary layers for more representative comparison to their canonical
representations, Eur. J. Fluid Mech. B/Fluids, 55, 300–312.
48. Vinuesa R, Duncan RD, Nagib HM (2016). Alternative interpretation of the Superpipe data
and motivation for CICLoPE: The effect of a decreasing viscous length scale, Eur. J. Fluid Mech.
B/Fluids, 58, 109-116.
49. D o h e r t y J, N g a n P, M onty JP, C hong M (2007). The development of turbulent pipe flow,
Sixteenth Australasian Fluid Mechanics Conference, Crown Plaza, Gold Coast, Queensland,
Australia, December 2–7, 2007.
50. M o n t y JP (2005). Developments in smooth wall turbulent duct flows, PhD thesis, University of
Melbourne, Melbourne, Victoria, Australia.
51. Vin u e s a R, Ba r trons E, C hiu D, D res s ler KM, R üedi J-D, Suzuki Y, N agi b HM
(2014). New insight into flow development and two dimensionality of turbulent channel flows,
Exp. Fluids, 55, 1759.
52. Preston JH (1954). The determination of turbulent skin friction by means of Pitot tubes, Journal
of the Royal Aeronautical Society, 58, 109–121.
53. Pat e l VC (1965). Calibration of the Preston tube and limitations on its use in pressure gradients,
J. Fluid Mech., 23, 185–208.
54. Z ag a r o l a MV, William s DR, Sm its AJ (2001). Calibration of the Preston probe for high
Reynolds number flows, Meas. Sci. Technol., 12, 495–501.
55. Hosseini SM, Vinuesa R, Schlatter P, Hanifi A, Henningson DS (2016). Direct numerical
simulation of the flow around a wing section at moderate Reynolds number, Int. J. Heat Fluid
Flow, doi:10.1016/j.ijheatfluidflow.2016.02.001.
Measurement of wall-shear stress 427

56. H a n r at t y TJ, Cam p bell JA (1996). Measurement of wall shear stress. in: Goldstein RJ, ed.,
Fluid Mechanics Measurements, 2nd edn., Taylor & Francis, Washington, DC, pp. 575–648.
57. Tr il l in g L, H ä kkinen RJ (1955). The calibration of the Stanton tube as a skin-friction meter,
in: G o rtler H, Tollmie n W, eds., 50 Jahre Grenzschichtforschung, Friedr. Vieweg and Sohn,
Braunschweig, Germany, pp. 201–209.
58. E a s t LF (1967). Measurement of skin friction at low subsonic speeds by the razor blade tech-
nique, Technical Report 3525, Aeronautic Research Council, London, U.K.
59. Br a d s h aw P, Gregory N (1959). The determination of local turbulent skin friction from
observations in the viscous sub-layer, Technical Report 3202, Aeronautic Research Council,
London, U.K.
60. F e r n h o l z HH, J anke G, Schober M, Wagner PM, Warnack D (1996). New develop-
ments and applications of skin-friction measuring techniques, Meas. Sci. Technol., 7, 1396–1409.
61. Ca s t r o IP, D ia nat M, B radbury LJS (1987). The pulsed-wire skin-friction measurement
technique, in Proceedings of the Fifth Symposium on Turbulent Shear Flows, Durst F et al. ed.,
Vol. 5, Springer-Verlag, Berlin, pp. 278–290.
62. Ta n n e r LH, B l ows LG (1976). A study of the motion of oil films on surfaces in air flow, with
application to the measurement of skin friction, J. Phys. E, 9, 194–202.
63. S q u ir e LC (1962). The motion of a thin oil sheet under the boundary layer on a body, in:
M a lt b y RL, ed., Flow Visualization In Wind Tunnels Using Indicators, AGARDo-Graph,
Vol. 70, North Atlantic Treaty Organization Advisory Group for Aeronautical Research and
Development, Bedford, England, pp. 7–23.
64. Segalini A, Rüedi J-D, Monkewitz PA (2015). Systematic errors of skin-friction measurements
by oil-film interferometry, J. Fluid Mech., 773, 298–326.
65. H c c h t E (1987). Optics, 2nd edn., Addison-Wesley, New York, pp. 270–314, 346–361.
66. Janke G (1993). Über die Grundlagen und einige Anwendungen der Ölfilminterferometrie
zur Messung von Wandreibungsfeldern in Luftströmungen, PhD thesis, TU-Berlin, Berlin,
Germany.
67. Ö s t e r l u n d JM, J ohans s on AV, N agib HM, H ites MH (2000). A note on the overlap
region in turbulent boundary layers, Phys. Fluids, 12, 1–4.
68. Chauhan K, Ng HCH, Marusic I (2010). Empirical mode decomposition and Hilbert trans-
forms for analysis of oil-film interferograms, Meas. Sci. Technol., 21, 105405.
69. Vis wa nat h DS, G hos h T, Pras ad DHL, D utt NVK, R ani KY (2007). Viscosity of
Liquids: Theory, Estimation, Experiment, and Data, Springer, Netherlands.
70. Ba r t r o n s E, M uñoz M (2012). Aspect ratio and perimeter effects on turbulent channel flows,
Technical Report, Illinois Institute of Technology, Chicago, IL.
71. F e r n h o l z HH, Finley PJ (1996). The incompressible zero-pressure-gradient turbulent
boundary layer: An assessment of the data, Prog. Aerosp. Sci., 32, 245–311.
72. R ü e d i J-D, N ag ib HM, Ö s terlund J, M onkewitz PA (2003). Evaluation of three tech-
niques for wall-shear measurements in three-dimensional flows, Exp. Fluids, 35, 389–396.
73. D e s s e J-M (2003). Oil-film interferometry skin-friction measurement under white light, AIAA J.,
41, 2468–2477.
74. D r iv e r DM (2003). Application of oil-film interferometry skin-friction measurement to large
wind tunnels, Exp. Fluids, 34, 717–725.
75. N au g h t o n JW, H ind MD (2013). Multi-image oil-film interferometry skin friction measure-
ments, Meas. Sci. Technol., 24, 124003.
76. N aq wi AA, Re y nolds WC (1991). Measurement of turbulent wall velocity gradients using
cylindrical waves of laser light, Exp. Fluids, 10, 257–266.
77. Fo u r g u e t t e D, M odarres s D, Taugwalder F, Wils on D, Kooches fahani M,
G h a r ib M (2001). Miniature and MOEMS flow sensors, Proceedings of the 31st AIAA Fluid
Dynamics Conference and Exhibit, AIAA Paper 2001-2982, American Institute of Aeronautics
and Astronautics, Reston, VA.
78. K ä h l e r CJ, S cholz U, O rtm anns J (2006). Wall-shear-stress and near-wall turbulence
measurements up to single pixel resolution by means of long-distance micro-PIV, Exp. Fluids, 41,
327–341.
79. Re d a DC, M u r atore JJ (1994). Measurements of surface shear stress vectors using liquid
crystal coatings, AIAA J., 32, 1576–1582.
80. R e d a DC, Wil d er MC, M ehta R, Zilliac G (1998). Measurement of continuous pressure
and shear distributions using coatings and imaging techniques, AIAA Journal, 36, 895–899.
81. S m it s AJ, L im TT, eds. (2000). Flow Visualization Techniques and Examples, Imperial College
Press, London, U.K.
82. B r ü c k e r Ch, Bauer D, C haves H (2007). Dynamic response of micro-pillar sensors
­measuring fluctuating wall-shear-stress, Exp. Fluids, 42, 737–749.
428 Ricardo Vinuesa and Ramis Örlü

83. Große S, S c h r ö der W (2008). Mean wall-shear stress measurements using the micro-pillar
shear-stress sensor MPS3, Meas. Sci. Technol., 19, 015403.
84. G na na m a n ic k am EP, N ottebrock B, Große S, Sullivan JP, Schröder W (2013).
Measurement of turbulent wall shear-stress using micro-pillars, Meas. Sci. Technol., 24, 124002.
85. Große S, S o o d t T, Schröder W (2008). Dynamic calibration technique for the micro-pillar
shear-stress tensor MPS3, Meas. Sci. Technol., 19, 105201.
86. N o t t e b r o c k B, Schröder W (2012). Improvement of the measurement range of the micro-
pillar shear-stress sensor MPS3, in Proceedings of the 28th AIAA Aerodynamic Measurement
Technology, Ground Testing, and Flight Testing Conference. New Orleans, LO, June 25–28, 2012,
Technical Report AIAA-2012-3011.
87. BrÜcker C. Evidence of rare backflow and skin-friction critical points in near-wall turbulence
using micropillar imaging, Phys. Fluids, 27, 031705 (2015).
88. Vinuesa R, OrlÜ R, Schlatter P. Characterisation of backflow events over a wing section,
Journal of Turbulence, Accepted (2016). Available at: http://dx.doi.org/10.1080/14685248.2016.1
259626.
89. Br ü c k e r C (2011). Interaction of flexible surface hairs with near-wall turbulence, J. Phys.:
Condensed Matter, 23, 184120.
90. B r own JL, N aughton JW (1999). The thin-oil-film equation, Technical Report NASA/TM
1999-208767, NASA-Ames Research Center, Moffett Field, CA.