0 оценок0% нашли этот документ полезным (0 голосов)

49 просмотров14 страницNew Developments and Applications of Skin-friction Measuring Techniques

© © All Rights Reserved

PDF, TXT или читайте онлайн в Scribd

New Developments and Applications of Skin-friction Measuring Techniques

© All Rights Reserved

0 оценок0% нашли этот документ полезным (0 голосов)

49 просмотров14 страницNew Developments and Applications of Skin-friction Measuring Techniques

© All Rights Reserved

Вы находитесь на странице: 1из 14

Printed in the UK

of skin-friction measuring techniques

H H Fernholz, G Janke, M Schober, P M Wagner and D Warnack

Berlin,

Hermann-Fottinger-Institut

fur

Technische Universitat

Stromungsmechanik,

Strae des 17. Juni 135, 10623 Berlin, Germany

Received 25 March 1996, accepted for publication 12 June 1996

Abstract. This survey covers recent developments and applications of four

skin-friction measurement techniques (oil-film interferometry, wall hot wire, surface

fence and wall pulsed wire). Comparisons of the techniques with each other and

with other methods are presented. Applications in attached and separated fully

turbulent boundary layers and in highly accelerated laminar-like flows will be shown

to demonstrate the application range and the limits of the various techniques.

1. Introduction

One of the most important results of boundary layer theory

has been the determination of the wall shear stress w , a

quantity which largely determines the energy necessary for

moving the flow of liquids and gases over solid walls.

Knowledge of the wall shear stress is very important

for many technical applications and for the understanding

of all wall-bounded shear flows. Therefore one would

like to know the magnitude (mean and fluctuating value)

and the direction of the skin-friction vector w and its

distribution over a surface. The flow may be compressible

or incompressible, laminar or turbulent and can even

reverse its direction in the vicinity of the wall in an adverse

pressure gradient.

The classical survey on skin friction measurements,

though primarily related to the performance of aircraft, was

written by Winter (1977) and there have been two more

recent review papers by Hanratty and Campbell (1983) and

by Haritonidis (1989). This paper differs from the surveys

mentioned above in that it concentrates on measuring

techniques which have been applied in our laboratory for

some years and on their application to standard and complex

flows. A few other methods will be addressed briefly if they

are used for comparison or promise potential for the future.

The discussion will be confined to incompressible flow

of gases along an aerodynamically smooth wall without heat

transfer. Special emphasis is put on a comparison between

the different techniques and on their range of validity.

For a two-dimensional incompressible laminar boundary layer with zero pressure gradient, Blasius (1907) calculated the skin friction, expressed in dimensionless form as

the skin friction coefficient cf ,

cf =

2w

= 0.664(Rex )1/2

u2

(1)

the density and the velocity at the edge of the boundary

c 1996 IOP Publishing Ltd

0957-0233/96/101396+14$19.50

by u , (the kinematic viscosity) and the length x in the

streamwise direction. There is no exact solution for the

equivalent case of a turbulent boundary layer and the semiempirical relationships one finds in the literature are all

based on measurements. For more complex flows even such

relationships do not exist and measurements are a must.

Measuring techniques for the wall shear stress may be

divided into the small group of direct methods (floating

element method and oil film interferometry) and the larger

group of indirect methods which must be calibrated. The

floating element balance is probably the oldest device to

measure the wall shear stress with the implication of a

large area and a very sensitive technique to determine small

forces. Kempf (1929) used such a balance for subsonic

flow and, since the area of the balance could be reduced

due to the high skin friction, the floating element balance

has had its main application in compressible boundary

layers. An excellent survey on the design and application of

these balances was provided by Winter (1977) and a rather

comprehensive list of test cases in compressible boundary

layers was presented by Fernholz and Finley (1977, 1981).

For subsonic flows there are two more recent

developments of floating element balances.

(i) The Bechert balances with relatively large surface

elements with sizes 750 mm 600 mm and 400 mm

500 mm (Bechert et al 1985, 1992). They were designed

mainly to compare the skin friction of aerodynamically

smooth and riblet (small-ribbed) surfaces.

(ii) The balances developed by Dickinson (1965)

are much smaller and can be adapted both to subsonic

and to supersonic boundary layers for local skin-friction

measurements. The latter floating element balances have

the advantage that they were (and possibly still are)

commercially available and their application was described

by Nyguyen et al (1984), for example. These balances can

be used in flows with zero pressure gradient only, since they

occur in an adverse pressure gradient flow, for example.

This problem was solved by Thomann and his co-workers

(Frei and Thomann 1980, Hirt et al 1986), who designed

and built a ring-type floating element balance with a small

extension in the streamwise direction and a liquid seal for

the gaps of the floating element.

All floating element balances mentioned so far have

the disadvantage that they give a value of the skin friction

which is integrated over a larger or smaller area and

necessarily fail to provide exact pointwise measurements.

This problem may be overcome by the use of miniature

surfaces such as a silicon micromachined floating-element

shear-stress sensor. Such devices have been developed

at MIT (Schmidt et al 1988, Padmanabhan et al 1995),

Caltech (Jiang et al 1995) and Case Western University

(Jiang et al 1995), for example, but published results do

not describe more than results obtained in a first laboratory

test phase.

For turbulent boundary layers the Preston tube is

the most commonly used instrument for skin-friction

measurements. The validity of this measuring technique is

dependent on the validity of the logarithmic law of the wall.

The dimensionless diameter d + = du / must lie in the

logarithmic region if the calibration curves of Patel (1965)

or Head and Vasanta Ram (1971) are used. More recently,

user-friendly calibration curves were suggested by Zurfluh

(1984) and Bechert (1995). The accuracy of the Preston

tube method is about 3% and slightly less for adverse

pressure gradients Patel (1965) and Hirt and Thomann

(1986). The state of the art of Preston tube applications

in compressible boundary layers was discussed by Finley

and Gaudet (1995). This update of the floating-element

balance and the Preston tube appeared to be necessary since

both techniques were used for comparisons with the four

techniques discussed in section 2 of this paper.

2. The measuring techniques

2.1. General remarks

It seems appropriate to begin this brief description of the

four techniques to measure skin friction by presenting some

of their properties and their range of application in tabular

form. Table 1 shows that they cover a wide spectrum

of wall-bounded shear flows although no single technique

possesses all desired properties. Figure 1 illustrates the

techniques. None of the four techniques is dependent on

the logarithmic law of the wall. The oil-film interferometry

is the only direct method, needs therefore no calibration and

has the potential to determine the whole field of wall shear

stress.

For the application of the other three techniques it is

important to heed that the fence height and the height

of the sensor wires of the probes should lie within the

viscous sublayer and that a probe must be calibrated

against reference measurements at the same position (xo , zo )

because of possible variations of skin friction in the

spanwise direction. For the calibration of the Preston tubes

we used an improved version of Patels calibration curve

(1971) which bridged the gaps between the three branches.

2.2. Oil-film interferometry

The feasibility of measuring the skin friction from the

movement of interference fringes of a thin oil film was

first realized by Tanner and Blows (1976). Oil-film

interferometry is the only direct method for skin-friction

measurement apart from floating-element balances. It has

a high spatial resolution and is capable of measuring reverse

flow. It does not require any assumptions to be made

concerning the flow field. It is easy to apply and needs

but little instrumentation.

A brief description of some features of the method

for two-dimensional flow will be given here. The full

theory and details of the application of this technique have

been described by Janke (1993). For another version of

this method, the dual-laser beam skin-friction measuring

technique, the reader is referred to Monson (1983) and to

Kim (1989), who applied this technique to supersonic flow.

A silicon-oil film is applied to a smooth solid wall.

This film is spread out by the flow into an extremely thin

layer whose thickness is typically of the order of several

wavelengths of visible light. The film thickness can be

visualized by Fizeau fringes, which originate from the

interference of light reflected from the top and from the

bottom of the oil film. Under monochromatic light the

height of the film at the kth black fringe is given by

hk = h0 + k1h

with

1h =

2[n2

k = 0, 1, 2, . . .

.

sin2 ()]1/2

(2)

(3)

film edge (k = 0), 1h the difference in height between

two consecutive fringes, n the refractive index of the oil,

the viewing angle of the observer and the wavelength of

the light. Equations (2) and (3) show that, for a constant

viewing angle, the fringes are contour lines of the oil-film

surface.

Inexpensive equipment (figure 2) is sufficient to obtain

bright fringes. A glass plate blackened on its back side

provides a suitable material for a flat wall, then h0 =

1h/2. For smooth curved walls a piece of video tape

glued on to the surface serves the same purpose as the

glass plate but the ratio h0 /1h will then differ from 0.5

and must be determined for a specific wall material. A

sodium lamp ( = 0.589 m), commercially available at

a marginal price, yields better fringes than does a laser.

The motion of the interference fringes (along a straight

skin-friction line) is recorded in an xt diagram. An

example is shown in figure 3 for the separated flow behind

a backwards-facing step. The diagram was recorded along

the x coordinate by repetitively reading only one line of

a properly adjusted video image into an image processor.

The resulting image consisted of 512 by 512 pixels.

The motion of the oil film will now be analysed. We

assume that the oil layer is so thin that it does not change

1397

H H Fernholz et al

Table 1. Comparison of the four measuring techniques.

Surface fence

Oil-film interferometry

Measured quantity

Calibration necessary

Mean value

Temporal resolution

Cross correlation

Spatial resolution

1x , 1z (mm)

Direction of w

APG

Pressure difference

Yes

Yes

Unclear

No

Heat transfer

Yes

Yes

> 10 kHz

Yes

Time of flight

Yes

Yes

' 20 Hz

No

Movement of fringes

No

Yes

No

No

< 1, 3

Yes

Yes

1.5, 0.5

Yes

Yes

< 1, 1

Possible

Yes

Reverse Flow

FPG

Transitional flow

Laminar flow

3D flow

Accuracy (estimated)

Yes

Yes (but restricted)

Unclear

Yes

Yes

4%

Yes

Yes (but no

instantaneous

reverse flow)

No

Yes

Yes

Yes

Yes

4%

Yes

Probably yes

Probably yes

Probably yes

Yes

4%

Yes

Yes

Yes

Yes

Yes

< 4%

Surface fence

Oil-film interferometry

the flow on top of it but is only driven by the skinfriction distribution of the flow. In this case the temporal

development of the film height in a two-dimensional flow

is described by

1 ( h2 )

h

=

t

2 x

(4)

films higher order terms in h have to be added. These

terms introduce effects of pressure gradient, gravity and

surface tension Janke (1993). The effects can be taken into

account, but for most practical situations h can be made

small enough that they may be neglected.

It can be shown Janke (1993) that, for a sufficiently

viscous oil, the movement of the film surface will be

The

determined by the mean skin friction only.

fluctuating part of , namely the turbulence structure of

the outer flow, will not affect the movement of the film.

Equation (4) can be used in several ways to deduce the

skin friction from an xt diagram not only for spatially

constant but also for skin friction fields with high

gradients. We shall first discuss briefly three of these

methods, then give a short account of an older method

for constant skin friction and finally provide a detailed

description of a fifth method (including an error estimate)

for constant skin friction.

(i) Integration of equation (4) yields

2

Z

2 x h 0

h

dx .

(x) = a

2

ha

h xa t

1398

(5)

interferometry, methods (i)(v).

either from zero height ha or from a known a , by spatial

integration of the temporal change of h at one or several

fixed times. The integral can also be written as

Z x

Z x

h 0

x

dx =

dh.

(6)

xa t

xa t h=constant

Ia denotes the centre of the characteristics, which cannot

be determined in each case.

Figure 11 (later) shows the skin-friction distribution

downstream of a backwards-facing step obtained from the

xt diagram in figure 3 with methods (i)(iii). Good

agreement is observed.

If the relative spatial variation of the wall shear stress

(1 / ) is small compared with the slope of the film surface

(h/x), the wall shear stress can be assumed constant.

Two more methods will now be discussed for this case.

(iv) For constant skin friction near the leading edge of

a wetted surface area the similarity solution has the form

h=

plane.

xt diagram is finite everywhere along a line t = constant.

(ii) It can easily be shown that equation (4) possesses

characteristics in the xt plane which are identical to paths

of the particles that form the oil-film surface (figure 3).

These characteristics can be found by connecting the streaks

crossing the interference fringes in the xt diagram. The

slope of the characteristics in the xt plane is the particle

velocity up ; that is

= up / h.

(7)

the contour lines is discernible in the xt diagram, as is

the case in figure 3. The corresponding characteristics are

also shown in figure 3. Distinguishing the characteristics

from the contour lines becomes difficult if the skin friction

approaches a constant value, because characteristics and

contour lines coincide for = constant.

(iii) It can be shown that equation (4) has similarity

solutions h(x, t) = g(x)/t for large values of t. Obviously,

the function g(x) happens to have the same shape as the

contour lines h = constant in the xt diagram. These lines

are also shown in figure 3. It can be seen that the contour

lines are spread out (or drawn together) by characteristics.

From g(x) the skin friction can be obtained via

Rx

Ia + xa g(x 0 ) dx 0

.

(8)

(x) = 2

g 2 (x)

x

.

t

(9)

xt diagram shows a fan of fringes originating at the

leading edge of the film (figure 4). This relation was

already used by Tanner and Blows (1976). Later, Monson

(1983) developed a two-beam laser method for this case

which made the whole principle rather complicated and also

needed a rather complex measuring device. Therefore we

did not follow this route in our investigations but started

again from Tanners basic principles.

(v) For constant wall shear stress the equation (4)

becomes a simple advection equation for the contour lines

hk :

hk

hk

+ uk

=0

(10)

t

x

with uk = hk /. Since uk is equivalent to the slope of the

contour lines in the xt plane, namely

x

uk =

t hk =constant

(11)

yield

2[n2 sin2 ()]1/2

(k + h0 /1h)

(12)

h0

= uk

.

1h

(13)

= uk

and after rearranging

k +

number of fringe velocities uk can be measured from

the xt diagram. Thus equation (13) presents an overdetermined system of linear equations for the unknown

quantities and possibly h0 /1h (if h0 /1h is unknown).

and h0 /1h can then be computed by using a least square

fit. The advantage of this procedure is that the quantity

h0 /1h can be determined from the measurements and does

not need to be known a priori . This was important for

the measurements given below, because wall material other

than glass (video tape) was used.

1399

H H Fernholz et al

Figure 4. The x t diagram showing the development of an oil-film wedge generated in a flow with spatially constant wall

shear stress.

(v). The velocity uk was deduced by measuring its slope

1x/1t in the xt diagram, where 1x is the distance within

which a fringe travels within the time 1t. This slope was

determined by means of a computer program such that lines

could be set parallel to the fringe by hand. The coordinates

of the end points of these lines were then available in pixel

units.

The total duration of the measurement was taken by the

program to determine the conversion factor from time to

pixels. Typical measuring times may lie in the range from

10 min to 3 h, depending on the viscosity of the oil and the

wall shear stress. The speed of the oil film should not be too

high, for the flow needs a minimum of 12 min in order to

attain a steady state. Furthermore, there should be at least

ten fringes in order to obtain a stable mean value. The

conversion factor for the space units and the viewing angle

was determined from an image of an object with known

dimensions placed at the position of measurement. The

viewing angle was obtained from the perspective distortion

with an accuracy of 0.5%.

The physical properties of the oil, oil and noil , were

obtained from tables provided by the manufacturer. Usually

silicon oil is used since its dependence on the temperature

is smaller than that of other oils but still not negligible and

can be interpolated between the tabulated values 0, 25 and

1400

40 C by using

oil = AekT

(14)

dynamic viscosity the density of the tabulated values was

linearly interpolated with an accuracy of better than within

0.5%. Measuring the density of the oil with a Mohr

balance reproduced the interpolated value within 0.2%. The

dynamic viscosity of the oil was measured with a Hoppler

viscosimeter and resulted in differences below 1% of the

computed value. The refractive index is fairly independent

of the temperature and the absolute pressure and can be

taken directly from the tables. Its value is 1.4 0.02 for

most oils.

The assumption of constant wall shear stress in the

area of measurement was checked by comparing method

(i) with method (v). Agreement within 1% between the

two methods was found.

The reproducibility of method (v) is 1%. The oilfilm method and the Preston tube method in a canonical

boundary layer agreed within 2%. Assuming that the

accuracy of the Preston tube method is 3%, the error

of the oil-film method should lie below 4%. Since the

oil-film method does not need to be calibrated, namely is

independent of the Preston tube measurements, its accuracy

can be better than within 4%. This can be shown by

using a method with higher accuracy than the Preston tube

for comparison.

Wall hot-wire and hot-film probes have been used to

measure the instantaneous local wall shear stress. Their

range of application is, however, quite different.

Wall-mounted hot films can be used for time-resolved

measurements in situations in which the conductivity of the

fluid is larger than the conductivity of the wall material.

This is possible for water and oil, but not for air. In

air, the transport of heat to the substrate and back into

the fluid leads to a reduced dynamic sensitivity of the

hot film (Tardu et al 1991). This effect decreases when

the thermal conductivity of the substrate is increased, but

even with walls made from aluminium or copper, the

RMS value of the wall shear stress in turbulent boundary

layers is underestimated by up to 25% (Dengel et al 1987,

Alfredsson et al 1987). Therefore we use only wall hotwire probes which retain their dynamic sensitivity and can

easily be corrected for heat transfer effects to the wall

(Wagner 1991).

The determination of the instantaneous skin-friction

from hot-wire measurements of the velocity gradient in

the viscous sublayer is strongly affected by the proximity

of the wall. However, in this case the distortion of

the temperature field affects only the sensitivity of the

probe, not its frequency response. The need for near-wall

measurements led to several investigations of the walldistance dependence of hot-wire probes. Janke (1987)

discussed these in detail and found a single relationship

to correct the mean velocity measured with constanttemperature anemometers (CTA) both in laminar and in

turbulent boundary layers. Thus one can conclude that the

mean response of the wall hot-wire probe is independent of

the turbulence structure of the flow.

The wall hot wire needs to be calibrated in a fixed

position against another instrument which measures the

local wall shear stress. This was done first by Bradshaw

and Gregory (1959), who determined only the mean skin

friction. Alfredsson et al (1987) described a probe in which

the hot wire is mounted several wire diameters above the

wall and calibrated against a Preston tube in a turbulent

zero-pressure-gradient boundary layer. At relatively high

Reynolds numbers, this technique leads to plausible results

both for the mean and for the RMS value of the wall shear

stress. However, as the Reynolds number is decreased the

sensitivity of the probe is reduced. This can be avoided

when the probe distance from the wall is increased but,

in order to preserve the linear relationship between the

velocity and the wall shear stress, the maximum distance

from the wall is limited. Fortunately, the wall conduction

effect scales with y + . Thus, there will always be a wall

distance at which the sensitivity is high enough and the

probe is still within the linear range of the velocity profile.

Criteria to find an optimum position will be given below.

In the near-wall region of both laminar and turbulent

shear flows, the velocity increases linearly with the

wall distance.

For dimensionless wall distances

y + = y u / < 5 with u = (w /)1/2 this is also valid for

instantaneous velocity profiles. Therefore we can determine

w =

u

y

(15)

distance y above the wall. Due to heat conduction near the

wall the velocity um measured with a hot wire calibrated

in the free stream and the true velocity u in the near-wall

region differ from each other. These two velocities are

related by an empirical relationship (Janke 1987)

um y

uy

=

k1

uy

1/2

+ k2

(16)

the range y + 5. In the following it is assumed that

equation (16) holds also for instantaneous values of u and

um , and so far we know of no experimental data which

contradict this assumption. For substrates with high thermal

conductivity (such as ceramics) and a hot-wire diameter of

5 m, k1 = 0.55 and k2 = 3.2.

For larger wall distances the true velocity is slightly

underestimated. We now consider a probe mounted at a

fixed distance h above the wall. With equations (15) and

(16) a relationship between m and the true skin friction w

can be given:

m = w k1

2

w

h2

1/2

+ k2

2

.

h2

(17)

now be related to the instantaneous wall shear stress w .

This can be done either by using a polynomial or by

applying Kings law. Here we have used Kings law

for the following reason. In all technically interesting

flows, the high skewness of the wall shear stress leads to

instantaneous values well above the highest mean value

that could be realized in the calibration process. Those

values can only be calculated when the calibration function

is extrapolated. Since polynomials tend to deviate strongly

from any physical law outside the calibrated region, we

prefer Kings formula for the linearization.

Kings law reads

e = (A + Bm )1/2 .

(18)

The Preston tube can only measure the mean wall shear

stress. Therefore the calibration function must be timeaveraged. This leads to

m =

e2 A

B

1/

(19)

to determine the coefficients A and B directly. With a

computer-aided measurement system, the squared mean

values of the bridge voltage

h i

h i

e2 = e2 i + e02

(20)

i

1401

H H Fernholz et al

Ruderich and Fernholz (1986) for the calibration of wall

pulsed-wire probes. However, since the values

m i 6= m i

(21)

cannot be determined from the measurements, an error of

m m

r =

m

"

1/

=1

0

1+ m

m

#1/

(22)

estimated to be

1 02

(23)

r '

4 2

by means of a series expansion. If we assume the

turbulence intensity to be of the order of 40%, the resulting

error for the wall shear stress is about 4%. This error can

be further reduced, when the time-records of the bridge

voltage taken at each calibration position are saved. In this

case, the mean values of the linearized bridge voltage can

be calculated and compared with the Preston-tube readings.

The calibration can be further improved by iteratively

correcting the coefficients A and B until good agreement

between the measured and the calculated values of the

mean wall shear stress is reached (Wagner 1991). This is

achieved by (i) calculating the difference between the wall

shear stress w as measured by a Preston tube and the value

computed from the time series e of the bridge voltage using

the first calibration curve, (ii) modifying the Preston-tube

readings in order to double the distance from the calibration

curve and (iii) re-calculating the coefficients A and B from

the modified values. Usually only one or two iterations are

necessary to achieve agreement within 1%.

The constants k1 and k2 in equation (16) hold only

for a specific combination of wall material and hotwire diameter and the calibration curve can be further

improved by determining k1 and k2 . Since deducing

k1 and k2 from velocity hot-wire measurements is very

time-consuming, they are computed directly from the

calibration curve (Warnack 1996). Re-arrangement of

Kings law and substitution of m = (w + k1 w 1/2 + k2 )

by (w + k1 w 1/2 + k2 ) with k1 = k1 1/2 / h and k2 =

k2 2 / h2 yields

w + k1 w 1/2 + k2 =

1 2 A

e

B

B

1/

.

(24)

equation (24) can be used to re-calculate improved values

of k1 and k2 . Usually the correlation coefficient approaches

unity very closely after several iterations.

Finally a few remarks on the design of the wall hot-wire

probe are in order. A sketch is shown in figure 1.

The body of the probe is made of 3 mm diameter steel

tubing with an insert consisting of a ceramic cylinder and

two 0.2 mm diameter Invar wires which serve as prongs.

The sensor wire which is soldered on to the prongs is a

2.5 m diameter platinum-coated tungsten wire with goldplated ends and an active length ` = 0.5 mm. This

1402

constant-temperature anemometer driving a near-wall

hot-wire probe as a function of the dimensionless wall

distance .

ensures an `+ <

20 up to a skin-friction velocity smaller

The optimal wall distance of the hot wire can

be determined from two considerations. One is the

requirement that the hot wire be situated within the viscous

sublayer. The other one is that its sensitivity E = de/d

be near its optimum value, where e is given by equations

(17) and (18) and = h(w /)1/2 /.

Figure 5 presents e() and E() for a typical probe

(ceramic surface and wire diameter d = 2.5 m). The

sensitivity shows a flat maximum in the range 0.5 < <

2.5 and decreases fast for smaller values of . Therefore

the sensor wire must not be too close to the wall. In order

to maintain both sensitivity and linearity, the wall distance

of the probe must be adapted to the mean skin-friction

value. The best results are obtained, if the dimensionless

wall distance is in the range 0.5 h u / 5.

If the probe is built as described and if the above

calibration method is used, this skin-friction measuring

technique has a reproducibility below 1%. The accuracy

of the mean skin friction value w measured by a wall hot

wire is about 4% if the accuracy of the calibration device

(here a Preston tube) is assumed to be 3%.

2.4. The surface fence

The surface fence or sublayer fence, as a means of

measuring the magnitude and direction of the skin friction,

was first applied by Konstantinov and Dragnysh (1960)

and later by Head and Rechenberg (1962) and described

in detail by Vagt and Fernholz (1973). A sketch of a

surface fence is shown in figure 1. The fence height H

should not exceed a value of H + ' 5 and thus remain

within the viscous sublayer. This makes it a device which

is independent of the validity of the logarithmic law of the

wall.

Surface fences are easier to build than are wall hotwire probes and need only a precise manometer to read

the pressure difference 1p upstream and downstream of

the fence. Due to the small size of the fence and the

manufacturing tolerances, it is necessary in general to

calibrate each fence, although the flow around the fence

can also be calculated in principle.

surface fence it is appropriate to determine the relevant

parameters. We only consider the influence of the mean

velocity profile and exclude turbulence effects from our

analysis. We assume that the pressure difference 1p

between the front and the back of the fence located in the

viscous sublayer depends on

1p = 1p(uH , H, , )

(25)

friction w is given implicitly through the linear relationship

u+ = y +

uH = H + u =

H u

H w

u =

.

(26)

determine whether all terms of equation (25) are of equal

importance. The Reynolds number is defined as

Re =

H uH

2

= H+ .

(27)

0.5 < H + < 5. For H + < 0.5 the pressure difference

becomes so small that it cannot be measured accurately.

At H + 5 the buffer layer begins and equation (26) no

longer applies. Therefore, the Reynolds number lies in the

range 0.25 < ReH < 25.

At the lower end of the Reynolds number range we

may assume creeping flow and neglect inertia forces in

comparison with viscous forces. Equation (25) then reduces

to

1p = 1p(uH , H, )

(28)

which gives

1p H

= F(1)

uH

(29)

1p w .

(30)

the height of the fence, a fact first pointed out by Taylor

(1935) discussing the half-Pitot tube.

For high Reynolds numbers, we may neglect the

viscous forces in comparison with the inertia forces, which

is a crude assumption for ReH ' 25. Equation (25) then

reduces to

1p = 1p(uH , H, )

(31)

1p

= F(1).

u2H

(32)

1p

w2 H 2

.

2

(33)

at the lower and upper ends of its range of applicability,

equations (30) and (33) and obtain

1p = C1 w + C2 w2

H2

.

2

(34)

1pH 2 /( 2 ) and w = w H 2 /( 2 ), we write

equation (34) in non-dimensionalized form as

1p = C1 w + C2 (w )2

(35)

calibration process.

The desired relationship of the skin friction as a

function of the pressure difference 1p may now be

calculated:

"

#1/2

1 C1

1 C1 2

1

+

+

1p

.

(36)

w =

2 C2

4 C2

C2

The advantage of equation (36) over Patels (1965)

power law 1p = wn ; n ' 1.5 is that it exhibits also the

correct asymptotic behaviour at the upper and lower ends

of the Reynolds number range, as suggested by dimensional

analysis.

Figure 6 shows a typical calibration curve obtained

using a least square fit based on equation (35). The

agreement between the fitted curve and the calibration

points is better than 1%. Together with the accuracy

of the Preston tube used as a calibration device, the

accuracy of a surface fence is about 4%. As will be

shown in section 3, the accuracy of the mean skin friction

measurement may decrease if the turbulence structure

of the flow under investigation differs strongly from

the turbulence structure of the flow of calibration. An

explanation could be that the the flow field around the fence,

the tubing and the pressure transducer presents a complex

nonlinear system.

It should be noted that Gur (1993) investigated the

applicability of the surface fence to measure fluctuating

values of the wall shear stress by using microphones built

into the probe.

1403

H H Fernholz et al

Table 2. Applications of the measuring techniques.

Type of flow

(authors)

Turbulent boundary layer

(Dengel et al 1987)

Turbulent boundary layer

(Warnack 1996)

Separation and recovery

regions downstream of a

normal plate with a

splitter plate

(Ruderich and Fernholz 1986,

Dengel et al 1987)

Separation and recovery

regions downstream of a

backwards-facing step

(Janke 1993)

Turbulent boundary layer with

and without reverse flow

(Dengel and Fernholz

1989, 1990)

Turbulent boundary layer with

and without reverse flow

(Gasser 1992,

Gasser et al 1993)

Wall

hot

wire

Wall

pulsed

wire

Pressure

gradient

Surface

fence

Oil-film

interferometry

Adverse

Favourable

Adverse

Adverse

Adverse

Adverse

X

X

Preston

tube

Method (v)

Hot

film

Floating

element

balance

Methods (i)(iii)

Figure 7. A comparison of three different measuring techniques for skin friction in an adverse pressure gradient (APG)

axisymmetric boundary layer layer. Data from Dengel et al (1987).

The wall pulsed-wire probe, first investigated by Ginder

and Bradbury (1973), is one of the few probes capable

of measuring the instantaneous wall shear stress in highly

turbulent flows with flow reversal. Methods like oil-film

interferometry or surface fences can be applied to flows

with reverse flow regions but have poor temporal resolution

at their present stage of development.

The principle of the velocity pulsed-wire anemometry

was described in detail by Bradbury and Castro (1971) and

Castro (1992), for example and applications of the wall

pulsed-wire probe were given by Castro and Dianat (1983),

Castro et al (1987) and Dengel et al (1987). Therefore

1404

probe, as shown in figure 1, consists of three parallel wires

mounted in a plane parallel to the wall and at a height

within the viscous sublayer. The central wire is heated by

a very short (5 s) electrical pulse, which generates a heat

tracer. The two sensor wires are operated as temperature

sensors and note the arrival of the heat tracer. The time of

flight T of the heat tracer is a measure of the instantaneous

wall shear stress.

The probe is conveniently calibrated against a Preston

tube. Thereafter, the probe can be used to determine the

mean and fluctuating part of the skin friction in regions

where the use of a Preston tube would lead to erroneous

is limited by the requirement that the probe not extend

the viscous sublayer. Furthermore, the instantaneous shear

stress must not create a time of flight shorter than the

shortest time which the pulsed wire anemometer is able

to measure (typically 100 s). Otherwise the truncation of

these high shear stresses will lead to severe errors.

It is thus necessary to adjust the probe geometry,

namely the wire spacing and wall distance, to the special

requirements of the flow under investigation. Typically,

however, the wires are located at a wall distance of 50 m

and separated by 0.7 mm from each other. The range of

skin friction of such a probe lies within 2 N m2 < w <

2 N m2 . The sensor wires are made of 2.5 m platinumcoated tungsten wires with gold-plated ends and an active

length of 0.5 mm. The pulsed wire is made of 5 or 9 m

platinum-coated tungsten wire. Thick pulsed wires give

strong signals but also produce larger wakes. For details

of the probe design the reader is referred to Dengel et al

(1987) or Castro et al (1987).

An appropriate relation between the wall shear stress

and the time of flight was given by Castro et al (1987):

w = A

2

3

1

1

1

+B

+C

T

T

T

(37)

calibration process. The advantages and disadvantages

of second- or third-order polynomials are discussed by

Handford and Bradshaw (1989). The calibration inside

a turbulent boundary layer makes it important to timeaverage equation (37), as shown by Ruderich and Fernholz

(1986). Due to the turbulent fluctuations, the instantaneous

time of flight recorded by the pulsed-wire probe may vary

widely for one calibration point, that is one mean shearstress value known from the Preston tube measurement.

The corresponding equation used during the least square fit

of the calibration process is

2

3

1

1

1

+B

w = A

+C

.

T

T

T

measuring techniques in a favourable pressure gradient

(FPG) turbulent boundary layer without re-laminarization

FPG. Data from Warnack (1996).

measuring techniques in a FPG turbulent boundary layer

with re-laminarization FPG. Data from Warnack (1996).

(38)

such that the probe may also be used in flows with

turbulence structures that differ strongly from the flow used

for the calibration.

Note, that due to its low sampling rate ( 40 Hz)

the wall pulsed-wire probe is not well suited to obtain

spectra. Besides the error caused by the calibration device

the resolution of the time of flight counter becomes a source

of error for high values of the skin friction, typically 1% for

w > 1.5 N m2 . The combined errors of calibration and

time of flight measurement can accumulate to an overall

error of about 4%.

3. Application and comparison of the measuring

techniques

All measuring techniques for the wall shear stress presented

in section 2 have been applied to turbulent boundary layers

with strong reverse-flow regions. In table 2 the different

types of flow and the measuring techniques are compared.

We shall discuss first measuring techniques applied to

attached boundary layers with adverse and favourable

pressure gradients and then flows with strong and weak

reverse flow.

The first comparative tests were performed in an

axisymmetric adverse pressure gradient turbulent boundary

layer on a cylinder with its axis in the streamwise direction

Dengel et al (1987). Three techniques were compared

(Preston tube, surface fence and wall pulsed wire) upstream

of separation. The results are presented in figure 7 and show

on the whole good agreement.

Skin-friction measurements in favourable pressure

gradient boundary layers are still scarce and therefore

it seemed to be appropriate to compare four measuring

techniquessurface fence, wall hot-wire, Preston tube

and oil-film interferometry in accelerated flows (Warnack

1405

H H Fernholz et al

Figure 10. A comparison of three measuring techniques for skin friction in a wall-bounded turbulent shear layer with strong

reverse flow (xR = 0.38 m). Data from Dengel et al (1987).

already the breakdown of the standard logarithmic law

of the wall in a highly accelerated boundary layer. In

such a flow the Preston-tube method fails. Therefore

Patel and Head (1968) used a surface fence for their

measurements. In our experiment the favourable pressure

gradient boundary layers were generated in a 0.44 m

Plexiglas pipe by means of two centre bodies and the skin

friction was measured on the pipe wall. The mean skin

friction w was made dimensionless with the reference

quantity u20 at the first measuring station where the

pressure gradient is zero and plotted against the streamwise

distance x in figures 8 and 9. Figure 8 shows case 1

which has the lowest acceleration (k 2 106 , k =

[/(u3 )]dp/dx). It reveals the breakdown of the

Preston-tube method in the favourable pressure gradient

region. The wall hot-wire and the surface-fence data

collapse on each other. Figure 9 presents case 2 which

has a highly accelerated flow region k 4 106 and

shows the deviation of the surface-fence data from those of

the wall hot-wire and the oil-film technique in the region

around x 1.6 m. In this region, the turbulence structure

differs strongly from the zero-pressure-gradient boundary

layer, where the fence was calibrated. This is indicated

by the skewness of the wall shear stress. The surface-fence

data are supposed to be incorrect due to the nonlinear effects

caused by the change in the turbulence structure. Neither

the oil-film nor the wall hot-wire technique is affected by

these effects (see section 2). Obviously, the change in

turbulence structure in case 1 is not large enough to cause

discrepancies between the surface-fence and the wall hotwire methods.

The next two flows are wall-bounded turbulent shear

flows with strong reverse-flow regions (Fernholz 1994).

The first flow is generated by a normal plate with a long

splitter plate Ruderich and Fernholz (1986), whereby the

1406

for skin friction in a wall-bounded turbulent shear layer with

strong reverse flow. Data from Janke (1993).

plate and re-attached on the splitter plate. Three measuring

techniques (wall pulsed-wire, Preston tube oriented in the

respective flow direction and surface fence) are compared

in the reverse-flow region and in the recovery region

downstream of re-attachment (figure 10). Only the wall

pulsed-wire and the surface-fence methods show the correct

skin-friction distribution. The Preston tube fails because the

logarithmic law does not exist in the entire region presented

here (Ruderich and Fernholz 1986).

The second strong reverse-flow (figure 11) is

downstream of a straight backward-facing step (Janke

1993). Here we have compared the wall pulsed-wire and

the surface-fence with the oil-film method using evaluation

methods (i)(iii) (see section 2). The three oil-film methods

agree well with each other and with the wall pulsed-wire

Figure 12. The distribution of the pressure coefficient cp , skin-friction coefficient cf and reverse-flow parameter w for three

APG boundary layers. All tagged data (O0 , 0 , 0 ) were measured by a wall-pulsed wire probe. From Dengel and Fernholz

(1990).

Figure 13. A comparison of four measuring techniques for skin friction in an APG-bounded turbulent boundary layer. From

Gasser et al (1993).

measurements close to separation and re-attachment. The

interpolated data from the wall pulsed-wire probe show

re-attachment slightly upstream of those of the oil film.

Both discrepancies could be caused by the probe sizes in

the x direction which were used in a rather small reverseflow region (about 40 mm total length) so that the oil film

technique probably gives the true re-attachment point.

The ability of the wall pulsed-wire probe to distinguish

between small positive and negative values in a highly

turbulent flow, even to determine w = 0, is shown in

figure 12. The three boundary layers (Dengel and Fernholz

1990) which were measured in the same axisymmetric test

section as the one described in figure 7 differ slightly in

their pressure distribution upstream and thus in their skin-

slightly positive, negative or zero over a longer distance. It

is important to note that instantaneous reverse flow occurred

in all three cases as shown by the distribution of the reverseflow parameter w (measured with the wall pulsed-wire

method).

A similar adverse pressure gradient turbulent boundary

layer with separation and re-attachment and downstream

recovery of the flow in a mild favourable pressure

gradient was investigated by Gasser (1992). A comparison

among the results of four measuring techniques used

in this boundary layer (surface-fence (the probes were

manufactured at the HFI), wall pulsed-wire (the probes

were manufactured at the HFI), Preston tube and floatingelement balance) was given by Gasser et al (1993).

1407

H H Fernholz et al

an annular pipe of 200 mm diameter where the pressure

distribution was generated by suction of air through a

perforated concentric inner tube. Thus a family of boundary

layers could be generated and at the same time the locations

of the balance and the other skin-friction probes were fixed

while the pressure distribution was moved. The readings

of the different probes were averaged over 72 points,

distributed along the circumference of the pipe, compared

with the reading of the ring-type floating-element balance

of 10 mm axial length and accuracy 0.01 Pa. Figure 13

shows the comparison of the results obtained with the

four devices in four different pressure distributions (Gasser

et al 1993). The deviations among the four techniques

were not completely explained but the wall pulsed-wire

and the floating-element-balance methods show fairly good

agreement at all measuring positions whereas the Preston

tube reads consistently lower than the balance in the region

xA > 200 mm where the boundary layer is in a mild

favourable pressure gradient. The probes read higher

than the balance in regions with strongly decreasing w

(xA < 100 mm).

A more limited comparison (Preston tube and surface

fence) in the flow downstream of a backwards-facing step

was performed by Kiske et al (1981), who already found

discrepancies between the two methods in disturbed wall

bounded turbulent flows. We have not shown detailed

measurements of the fluctuating shear stress (w02 )1/2 in this

review. Such data exist and may be found, for example, in

the investigation of Ruderich and Fernholz (1986), Dengel

et al (1987), Dengel and Fernholz (1990), Janke (1993),

Wagner (1995) and Warnack (1996). There are hardly any

comparisons between measurements of wall pulsed-wire

and wall hot-wire probes, since the application range of the

two probes lies in separated or attached flows, respectively.

4. Conclusions

Four skin-friction measuring techniques in turbulent wall

bounded shear flows were presented. The techniques have

been developed further, probes were designed and built

and they were used in normal and complex flows. The

range of applicability of these measuring techniques was

investigated and they were compared with each other and

with data from a floating-element balance and from Preston

tubes. The Preston tube was also used to calibrate three of

the four techniques investigated herethe oil-film method

is absolute.

All four techniques can measure the mean skin friction,

wall hot-wire and wall pulsed-wire fluctuating quantities,

and oil-film and wall pulsed-wire skin friction in flows in

which the flow direction reverses. They are all independent

of the validity of the standard logarithmic law of the

wall. The four techniques complement each other and have

ranges in which they overlap, but there are other ranges

within which only one or two may be used.

In complex flows no general rule for an application can

be given but some specific points should be addressed.

1408

safely be applied in attached boundary layers and wallbounded shear layers with variable pressure gradients.

For wall hot-wires there must be no flow reversal.

In highly accelerated boundary layers the surface-fence

measurements may no longer be correct if the fence was

calibrated in a canonical boundary layer. The breakdown

of the Preston-tube method in highly accelerated boundary

layers, as already found by Patel (1965), was confirmed.

(ii) In flows in which the flow direction reverses

or downstream of re-attachment Preston tubes must not

be used. The wall-pulsed wire, the oil-film and the

surface-fence techniques may be used in such flows but

small deviations occur for the surface-fence measurements

close to separation and re-attachment because of small

asymmetries of the probe.

Acknowledgment

The authors acknowledge the financial support of the DFG.

References

Alfredsson P H, Johansson A V, Haritonidis J H and Eckelmann

H 1987 The fluctuation wallshear stress and the velocity

field in the viscous sublayer Phys. Fluids A 31 102633

Bechert D W 1995 On the calibration of Preston tubes AIAA J.

34 2056

Bechert D W, Hoppe G and Reif W-E 1985 On the drag

reduction of the shark skin AIAA Paper 85-0546

Bechert D W, Hoppe G, van der Hoeven J G H and Makris R

1992 The Berlin oil channel for drag reduction research

Exp. Fluids 12 25160

Blasius H 1907 Grenzschichten in Flussigkeiten mit kleiner

Reibung Dissertation Gottingen.

Bradbury L J S and Castro I P 1971 A pulsed-wire technique for

velocity measurements in highly turbulent flow J. Fluid

Mech. 22 67987

Bradshaw P and Gregory N 1959 The determination of local

turbulent skin friction from observations in the viscous

sub-layer, Reports and Memoranda 3202, ARC, London

Castro I P 1992 Pulsed-wire anemometry Exp. Therm. Fluid Sci.

5 77080

Castro I P and Dianat M 1983 Surface flow patterns on

rectangular bodies in thick boundary layers J. Wind. Eng.

Ind. Aero. 11 10719

Castro I P, Dianat M and Bradbury L J S 1987 The pulsed-wire

skin-friction measurement technique Proc. 5th Symp. on

Turbulence and Shear Flows

Dengel P and Fernholz H H 1989 Generation of and

measurements in a turbulent boundary layer with zero skin

friction Advances in Turbulence vol 2, ed H H Fernholz and

H E Fiedler (Berlin: Springer) pp 4327

1990 An experimental investigation of an incompressible

turbulent boundary layer in the vincinity of separation

J. Fluid Mech. 212 61536

Dengel P, Fernholz H H and Hess M 1987 Skin-friction

measurements in two- and three-dimensional highly

turbulent flows with separation Advances in Turbulence

vol 40, ed G Compte-Bellot and J Mathieu (Berlin:

Springer) pp 4709

Dickinson J 1965 The determination of turbulent skin friction

PhD Thesis Laval University Quebec (now at the

Department de Genie Mecanique)

Fernholz H H 1994 Near-wall phenomena in turbulent separated

flows Acta Mechanica 4 5767

Fernholz H H and Finley P J 1977 An initial compilation of

compressible turbulent boundary layer data AGARDograph

223 Advisory Group Aerospace Research and Development

1981 A further compilation of compressible turbulent

boundary layer data with a survey of turbulence data

AGARDograph 263 Advisory Group Aerospace Research

and Development

Fernholz H H and Warnack D 1996 The effect of a favourable

pressure gradient and of the Reynolds number on an

incompressible axisymmetric turbulent boundary layer

Part. 1 The turbulent boundary layer, submitted for

publication

Finley P J and Gaudet L 1995 The Preston tube in adiabatic

compressible flow Exp. Fluids 19 13341

Frei D and Thomann H 1980 Direct measurements of skin

friction in a turbulent boundary layer with a strong adverse

pressure gradient J. Fluid Mech. 101 7995

Gasser D 1992 Experimentelle Untersuchung stark verzogerter

turbulenter Grenzschichten Dissertation ETH Zurich.

Gasser D, Thomann H and Dengel P 1993 Comparison of four

methods to measure the wall shear stress in a turbulent

boundary layer with separation Exp. Fluids 15 2732

Ginder R B and Bradbury L J 1973 Preliminary investigation for

skin friction measurements in highly turbulent flows, ARC

Report

Gur Y 1993 Mean and fluctuating wall shear stress

measurements PhD Thesis Massachusetts Institute of

Technology, Mechanical Engineering Department

Handford P M and Bradshaw P 1989 The pulsed-wire

anemometer Exp. Fluids 7 12532

Hanratty T J and Campbell J A 1983 Measurement of wall shear

stress Fluid Mechanics Measurements ed R J Goldstein

(New York: Hemisphere) pp 559615

Haritonidis J H 1989 The measurement of wall shear stress

Advances in Fluid Mechanics Measurements ed

M Gad-el-Hak (Berlin: Springer) pp 22961

Head M R and Rechenberg I 1962 The Preston tube as a means

of measuring skin friction J. Fluid Mech. 14 117

Head M R and Vasanta Ram V 1971 Simplified presentation of

the Preston tube calibration Aeronautical Quarterly XXII

295300

Hirt F and Thomann H 1986 Measurements of wall shear stress

in turbulent boundary layers subject to strong pressure

gradients J. Fluid Mech. 171 54762

Hirt F, Zurfluh U E and Thomann H 1986 Skin-friction balances

for strong pressure gradients Exp. Fluids 4 296300

Janke G 1987 Hot wire in wall proximity Advances in

Turbulence ed G Compte-Bellot and J Mathieu (Berlin:

Springer) pp 48898

die Grundlagen und einige Anwendungen

der Olfilm-interferometrie

zur Messung von

Wandreibungsfeldern in Luftstromungen Dissertation

Technische Universitat Berlin

Jiang F, Gupta B, Tai Y C and Goodman R 1995 Measurement

of instantaneous turbulent shear stress distribution by

MEMS based sensors Bull. APS 40 12, FC 2

Kempf G 1929 Neue Ergebnisse der Widerstandsforschung

Werft, Reederei, Hafen II 2349

Kim K-S 1989 Skin friction friction measurements by laser

interferometry in supersonic flows PhD Thesis Pennsylvania

State University

disturbed turbulence structure on the Preston-tube method

of measuring wall shear stress Aeronautical Quarterly

XXXII 35467

Konstantinov N I and Dragnysh G L 1960 The measurement of

friction stress on a surface, English translation, DSIR RTS

1499

Monson D J 1983 A nonintrusive laser interferometer method

for the measurement of skin friction Exp. Fluids 1 1522

Nguyen V D, Dickinson J, Jean Y, Chalifour Y, Anderson J,

Lemay J, Haeberle D and Larose G 1984 Some

experimental observations of the law of the wall behind

large-eddy breakup devices using servo-controlled skin

friction balances AIAA Paper 840346

Padmanabhan A, Goldberg H D, Breuer K S and Schmidt M A

1995 A silicon micromachined floating-element shear-stress

sensor with optical position sensoring by photodiodes

(private communication) Bull. APS 40 12, FI 1

Pan T, Hyman D, Mehregang M, Reshotko E and Garverik S

1995 Micro shear stress sensors with direct electrical

readout Bull. APS 40 12, GE 5

Patel V C 1965 Calibration of the Preston tube and limitations

on its use in pressure gradients J. Fluid Mech. 23 185208

Patel V C and Head M R 1968 Reversion of turbulent to laminar

flow J. Fluid Mech. 34 37192

Ruderich R and Fernholz H H 1986 An experimental

investigation of the structure of a turbulent shear flow with

separation, reverse flow, and re-attachment J. Fluid Mech.

163 283322

Schmidt M A, Howe R T, Senturia S D and Haritonidis J H 1988

Design and calibration of a microfabricated floating-element

shear-stress sensor IEEE Trans. Electron Devices 35 7507

Tanner L and Blows L 1976 A study on the motion of oil films

on surfaces in air flow, with application to the measurement

of skin friction J. Phys. E: Sci. Instrum. 9 194202

Tardu S, Pham C T and Binder G 1991 Effects of longitudinal

diffusion in the fluid and of heat conduction to the substrate

on the response of wall hot-film gauges Advances in

Turbulence vol 3, ed A V Johansson and P H Alfredsson

(Berlin: Springer) pp 50613

Taylor G I 1935 Measurements with a half-Pitot tube Proc. R.

Soc. A 166 476

Vagt J D and Fernholz H H 1973 Use of surface fences to

measure wall shear stress in three-dimensional boundary

layers Aeronautical Quarterly XXIV 8791

Wagner P M 1991 The use of near-wall hot-wire probes for

time-resolved skin-friction measurements Advances in

Turbulence vol 3, ed A V Johansson and P H Alfredsson

(Berlin: Springer) pp 5249

Wagner P M 1995 Koharente Strukturen der Turbulenz im

wandnahen Bereich von Ablosegebieten Dissertation

Technische Universitat Berlin

Warnack D 1996 Eine experimentelle Untersuchung

beschleunigter turbulenter Wandgrenzschichten Dissertation

Technische Universitat Berlin

Winter K G 1977 An outline of the techniques available for the

measurement of skin friction in turbulent boundary layers

Prog. Aerospace Sci. 18 157

Zurfluh U E 1984 Experimentelle Bestimmung der

Wandschubspannung in turbulenten Grenzschichten

Dissertation ETH Zurich

1409