Sie sind auf Seite 1von 8

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:
IP Address: 134.117.71.90
This content was downloaded on 09/06/2014 at 17:29
Please note that terms and conditions apply.
Interpretation of a hot wire signal using a universal calibration law
View the table of contents for this issue, or go to the journal homepage for more
1971 J. Phys. E: Sci. Instrum. 4 225
(http://iopscience.iop.org/0022-3735/4/3/016)
Home Search Collections Journals About Contact us My IOPscience
Interpretation of a hot wire signal
using a universal calibration law
H H Bruun
Institute of Sound and Vibration Research, University of
Southampton, Southampton SO9 5NH
MS received 2 October 1970
Abstract This paper describes the interpretation of the hot
wire signal from a single hot wire in terms of a universal
function. The calibration curves for probes with nominally
the same geometry have been studied for deviations in the
shape of calibration curves, yaw dependence and the effect
of angle of incidence. From these results the basic equations
for the interpretation of the signal from normal and yawed
hot wires have been derived.
1 Introduction
The interpretation of hot wire measurements of the turbulent
velocity components depends on a detailed knowledge of the
steady heat transfer from electrically heated cylinders. Investi-
gation of the heat transfer process for flow varying from free
molecular to continuum and speeds from those of natural
convection to supersonic have shown that the Nusselt number
for an electrically heated wire in general is a function of
several parameters :
(1)
where CL is the angle between the velocity vector and the normal
to the wire.
This paper describes the use of hot wires in air flows at
atmospheric pressure (windtunnels, air jet flows, etc.). The
Prandtl number ( Pr) is therefore constant. The effect of the
Grashof number, as shown by Collis and Williams (1959),
will only be significant at extremely low velocities, permitting
this parameter to be omitted in most practical anemometer
applications.
The maximum velocity in this investigation was limited to
150 m s-1, and only a given hot wire type with known nominal
values of Lld and of the overheat ratio TWIT was used. These
restrictions reduce the influence of the Mach number and of
the Knudsen number to second order effects.
The flow temperature during all the experiments was kept
constant and equal to the room temperature. I t is therefore
possible to relate the Nusselt number to the square of the
voltage output E2 and the Reynolds number to pV, giving the
following heat transfer relationship :
(2)
Nu= F (Re, Gr, Pr, Kn, M, Lid, TWIT, E)
E 2 = K (pV, CY, Lld, Tw/T).
To date it has been customary to express the relationship
(2) in the form
(3)
where V, is the effective cooling velocity, often set equal to
the normal component of the velocity, i.e. Ve= Vcos E.
The use of this assumption may lead to a considerable error,
as shown in 03. King (1914) gave the value of 0.5 of the
exponent n, while Collis and Williams (1959) found n=0.45
fitted their data better. The constants A and B are normally
E* =A +B ( p V,).
determined experimentally, with A either as the heat loss at
zero flow speed or as the intersection value of the E2 axis.
Relating equation (3) to measured hot wire calibration
curves has revealed that A and B cannot be assumed constant
over a large velocity range. Kings or Collis and Williams
law are therefore only approximations applicable over a
limited flow range. This aspect is discussed in detail in $2.3.
2 A universal empirical heat transfer law
2.1 Hot wires normal to fl ow direction
The form of equation (2) suggests that for a given hot wire
probe type it may be possible to describe the calibration
curves in terms of a universal heat transfer law, thereby
reducing the necessary calibration work considerably.
The hot wire probe type used in the investigation was the
2 mm ISVR hot wire probe illustrated in figure 1 . The sensing
element of this probe type consists of a 5 pm tungsten wire.
Figure 1 The 2 mm ISVR probe
By a copper plating procedure the active length of the sensing
element has been removed from the prongs (Davies and Davis
1966). The active length obtained by this procedure was found
to vary between 1.8 mm and 2.1 mm. The ratio of the total
length of the wire and the active length is approximately 2.5
(see figure 1). The prongs are 8 mm long. Due to variation of
the active length, the diameter of the wire and the resistivity
of the tungsten material, these probes had a scatter in the cold
resistance of the order of 5-10%. The hot resistance was set
to a constant value of 15 !2giving a nearly constant overheat
ratio Tw/T of 2.
Calibration measurements with hot wire probes having the
probe support placed perpendicular to the mean flow were
carried out in an open circuit wind tunnel and in a 2 in air
jet having a stagnation temperature equal to the room
temperature and expanding into the atmosphere.
225
H H Bruun
In describing the calibration law the heat transfer was
expressed in terms of E*- Eo2 as a function of pV. A small
container having a diameter of 1.5 cm was used to shield the
hot wire for the measurements of the voltage output EO at
zero flow speed. The measured calibration curves indicated
that it was possible to express the heat transfer law for this
hot wire type in terms of a universal function f ( pV) . The
calibration law was therefore expressed as
E2-Eo2= Cf ( p V)
(4)
where C is a constant which must be determined individually
for each wire. A universal function f( V ) has been calculated
for air jet flow from the calibration curves. The function
f ( V ) is given in table 1 in terms of E. The density has been
omitted as a parameter for this flow type as p( r) is identical
for all such calibrations.
Two additional types of measurements were carried out to
check the hypothesis of a universal shape of the calibration
curves. In the first type of experiment two 2mm hot wire
probes A and B were placed simultaneously in the centre of
a 2 in air jet, having the supports perpendicular to the mean
flow. If the hypothesis of a universal shape is valid then the
ratio (E2 - Eo2)a/(E2- Eo~) B should be independent of
"Gx''z5x
4r I f I
80
- 1.10,
I I
0 20 40 60
Velocity (m s-' )
Figure 2 Difference in shape of calibration law for two sets
of two 2 mm long hot wires
velocity. Two typical results of such sets of measurements
have been plotted in figure 2. A small variation in the ratio
(E2- Eo2)a/(E2- Eo~) B with velocity is observed. This is due
to a slight difference in the shape of the calibration curves of
hot wire A and B. The uncertainty in the estimated velocity
caused by this small variation in shape was calculated for
several calibrations. For velocities above 10 m s-l the uncer-
tainty was evaluated to be of the order f +%. Below 10 m s-l
the uncertainty increases slightly with decreasing velocity and
becomes of the order k 2 to 3 % at 1 m s-l. However, if the
maximum velocity of the calibration curve is lowered then the
uncertainty in the lower velocity range is reduced corre-
spondingly.
Due to small imperfections in some wires (dust accumula-
tion, etc.) drift could not always be completely avoided.
Measurement of EO before and after the calibration run gave
the magnitude of the drift. The observed error introduced by
drift was found to be quite consistent with the calculated
additional error in the estimated velocity. The measured
calibration curves showed that a drift in EO of 2 5 mV gave
an additional uncertainty in the velocity of f 1 %. Similarly,
an evaluation based on the calibration curve f ( V ) predicts
that a 1 % change in V corresponded to a change in E of
3-5 mV in the whole velocity range. The additional error
introduced by small amounts of drift can therefore readily
be explained. However, as the source and time of the occur-
rence of the drift is usually unknown, this error can normally
not be compensated for.
The accuracy of the chosen standard functionf( V ) (table 1)
was also investigated. Several calibration curves with different
2 mm hot wires placed one at a time in a 2 in air jet was
measured in terms of E* - EO* as a function of V. The velocity
was determined by accurate manometer readings. The velocity
was then recalculated using the standard function f( V ) and
the assumption of a constant ratio (E2- E02)/f( V) . After
some initial corrections to f ( V ) (incorporated in table 1) the
difference between the velocity calculated by these two
methods could be explained by the total error of the small
difference in shape (error k +%) and the velocity uncertainty
(error f +%) from the manometer reading.
These measurements have justified the use of a universal
shape of the calibration curves for hot wires with the same
nominal geometry. For velocities greater than 10 m s-l the
error in the estimated velocity introduced by this simplifica-
tion will be less than c 1 %. Below 10 m s-1 the uncertainty
increases slightly with decreasing velocity, being of the order
of i: 3 % at 1 m s-1. If the hot wire voltage drifts during the
experiment the uncertainty will increase with 1% for each
5 mV drift.
The above measurements were carried out for a constant
probe geometry. A few tests were carried out to investigate
whether the concept of a universal shape is applicable to
probes with different geometries. ISVR hot wire probes having
active lengths of 1 mm and 3 mm and run at an overheat
ratio of 2 were used for this purpose. First a 1 mm and a
2 mm hot wire probe was placed in the centre of a 2 in air
jet with their supports perpendicular to the main flow. The
ratio (E2- Eo~) I / ( E* - E092 was recorded as function of the
velocity V, using the 2mm hot wire probe for the velocity
determination. The same measurements were then performed
with a 2 mm and a 3 mm hot wire probe. Both sets of results
showed the same type of curve variation. Only the results for
the 2 mm and 3 mm hot wires have therefore been presented
in figure 3. The figure shows a somewhat greater difference in
shape than in figure 2. The variation is, however, only 2-3
times greater than in figure 2, giving 8 % difference in velocity
estimate at 1 m s-1 if the same universal function f ( V ) is
used. The similar trend of the two sets of experiments, however,
indicate that a different universal function can be used for
other types of hot wires.
Changing the support orientation relative to the mean flow
also has a minor effect on the calibration curve. This is
described in $4.
2.2 Analytic approximation of the calibration law
In evaluating turbulent data the calibration curve f ( p V ) is
often approximated by an analytic expression. To obtain
accurate results, two conditions must be satisfied. First, the
mean voltage E and the mean flow value must satisfy
the analytic expression, and secondly, the slope variation of
Figure 3 Difference in shape of calibration law for two sets
of 2 mm and 3 mm long hot wires
226
Interpretation of a hot wive signal using a unicersal calibration law
the analytic expression must be the same as the calibration
curvef(pV) around the point (E, pV). A convenient way of
expressing f ( p V) is the power law Kl(pV')nl giving
- -
E'- Eo2=K1(p V).l (5)
where KI and nl are functions of the velocity.
I n many hot wire measurements only the velocity fluctua-
tions are of interest. By including the mean density variation
with velocity in the variable K, equation (5) can be rewritten as
E'- Eo2 =KVn. (6)
The change in n with the velocity for air jet flows is shown
in figure 4. The evaluation of n was carried out by the proced-
ures given in the appendix using the function f ( V) (table 1).
I 1 , , # ( I I
IO
0,301 I I I I S I I
I 100
Velocity (m s-l)
Figure 4 Variations of exponent n with velocity
This way of specifying n( V ) ensures that the above mentioned
slope requirement is satisfied. The investigation by Kjellstrom
and Hedberg (1968) with 1 mm DISA hot wire probes gave
a similar variation of nl with pV. Their data evaluation was
based on method (i) in the Appendix and shows some vari-
ation in n1 with prong configuration.
For practical hot wire applications it is necessary to assume
that the variable K and n in equation (6) are constants.
Knowing the mean velocity of interest V, the corresponding
value of n is selected from table 1. The value of K is then
determined from the measurements of EO and one accurate
measurement of corresponding values of E and V. By choosing
these values of n and K the approximation of a constant value
of K and n can be applied over a considerable velocity range.
Two different criteria were used for determining this velocity
range. In the first the limits of the velocity range were deter-
mined by a set maximum deviation (* X%) in the estimated
velocity, in the following denoted as the A ' % velocity range.
Values of 1 % and 5% were used for X. This criterion deter-
mines the maximum error in the mean velocity determination
due to curve approximation. As the second criterion a set
maximum deviation (+ Y%) in the estimated value of the
slope dE/dVof the calibration curve was used, in the following
denoted as the Y% slope range. A value of 5 % was used for
Y. This criterion determines the maximum error in the fluc-
tuating signal interpretation due to curve approximation.
The velocity ranges for values of the mean velocity P
going from 1 m s-1 to 150 m s-1 has been calculated by using
the function f ( V) . The results are plotted in figure 5 i n terms
of the ratio Vmax/Vmin. Only points up to l ooms-l have
been plotted, as Vmax exceeds 150 m s-l for mean velocities
above 100ms-1. Curves A and B correspond to the 1%
and the 5 % velocity range while curve C corresponds to the
5% slope range. The range corresponding to the 5% uncer-
tainty in the slope (curve C) is seen to be similar to the 1 %
velocity range (curve A) above 30-40 m s-l. Below 30 m s-l,
Table 1 Universal calibration functionf( V ) for a 2 mm hot
wire probe operated at 15 Ll
Air temperature 18C
Velocity V Output E Exponent n Exponent m
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1 *6
1.8
2
2.5
3
3.5
4
5
6
7
8
9
10
12
14
16
18
20
25
30
35
40
45
50
60
70
80
90
100
110
120
130
140
150
1,167
1.269
1.336
1.387
1.425
1.457
1.486
1.511
1.533
1,554
1,570
1.611
1.648
1.681
1,711
1,764
1.810
1.852
1.889
1,923
1.955
2.012
2.062
2.108
2.149
2.187
2.270
2.341
2,405
2.458
2.509
2.553
2.632
2.702
2.761
2.814
2.862
2.906
2,946
2.982
3.017
3,049
(eqn 15)
-
-
0.72
0.65
0.60
0.56
0.54
0.53
0.52
0.52
0.515
0,510
0.505
0,505
0300
0.500
0.495
0.490
0,490
0.485
0.485
0.480
0.480
0.475
0.470
0.465
0.460
0.450
0,445
0.440
0,430
0.425
0.41 5
0.405
0.400
0,390
0.385
0,375
0,370
0.365
0.360
0,360
-
-
-
-
-
0.51
0.50
0.50
0.49
0.49
0.48
0.48
0.465
0.460
0.455
0.445
0.440
0.435
0,430
0,425
0.420
0.415
0.415
0.410
0.405
0.400
0,395
0.390
0,385
0.380
0,380
0.375
0.375
-
-
-
-
-
-
-
-
-
curve C is, on average, midway between the 1 % and 5%
velocity range. The great increase in VmaxlVmin at the lower
velocities is, however, mainly due to a small change in Vmin,
giving large values of Vmax/Vmin at low velocities. The 1 %
velocity range and the 5 % slope range were therefore found
to be nearly equal in terms of the width of the velocity range
which can be defined as +( Vmax - Vmin). For further compari-
son the 1 % velocity range was chosen. This range is described
in more detail in table 2. By assuming a Gaussian probability
function of the turbulent fluctuations it is possible to relate
the velocity range to turbulence intensity. By setting Vmax- V
equal to 3a, where U is the standard deviation, and similarly
setting V- Vmin equal to 3a and averaging the two results,
the turbulence intensities given in table 2 were obtained. At
227
H H Bruun
A I O h velocity range
B 5 '10 velocity range
\
1
1
0 20 40 60 80 100
Velocity (m s-1)
Figure 5 Velocity dependence of the ratio Vmax/Vmin
low flow velocities the turbulent intensity has further been
restricted so that no negative velocities will occur. This table
shows that above 30-40 m s-1 the approximations of constant
K and n can only be applied to flows with less than 15 %
turbulence intensities if less than 5% uncertainty is to be
introduced due to curve approximation. For higher turbulence
intensities more complicated expressions for the calibration
curve must be used to overcome this problem. The use of
Collis and Williams' and King's law in this region will, as
shown in 52.3, give an even worse approximation.
Below 30 m s-1 the approximation of constants K and n
becomes considerably better, permitting flow with up to
30% turbulence intensity to be studied with less than 5%
uncertainty due to curve approximation.
Changing the exponent n in equation (6) to a value different
from the value corresponding to the mean velocity 7 was
Table 2 Velocity ranges corresponding to maximum 1 %
uncertainty in the velocity estimate using the calibration law
equation (6)
1
2
4
6
8
10
20
30
40
50
60
70
80
90
100
105
1.5 0.7
5.0 0.9
10 1 * 5
16 2.5
18 3.5
22 5.0
34 9.6
48 16
65 26
70 30
85 36
95 46
105 54
120 60
130 70
150 85
2.2 15
5.6 30 t
6.7 40f
6.4 351.
5.1 3o.t
4.4 301.
3.8 25
3.0 18
2.5 15
2.4 14
2.4 13
2.1 12
2.0 11
2.0 11
1 *9 11
1 *8 10
+Reverse flow will occur at higher turbulence intensity
found to reduce the velocity range considerably. By adding
0.05 to the value of the exponent n, the 1% velocity range
was reduced from 2.5 to 1.2 for velocities above 30 m s-l.
At the same time the requirement of matching slope of
approximation and calibration curve was no longer satisfied.
2.3 Comparison with King's and Collis and Williams' law
The usual way of expressing the calibration law is
( 7)
E2= A+ BVn
usually known as King's law for n=0.5 and as Collis and
Williams' law for n =0.45.
These laws too have been compared with the calibration
law (6) by calculating their 1 % velocity ranges. By differenti-
ating equation ( 7) one obtains
B=2E(dE/dV)/( Vn-ln). (8)
Inserting the values of n equal to 0.5 and 0.45. and the values
of E and dE/dV from the function f(V) the B variations
shown in figure 6 were obtained. The corresponding values of
A were obtained from equation ( 7) and are also seen plotted
in figure 6. The value of A and B in Collis and Williams'
law below 20 ni s-l is seen to be nearly constant, indicating
a very good approximation uhen A and B are assumed
constant. Evaluation of the 1 % velocity range for King's
law and Collis and Williams' law was carried out using the
values of A and B given in figure 6. The results are given in
table 3. By comparing the velocity ranges with the results in
table 2, King's law is seen to be slightly inferior to equation (6)
at all velocities.
Collis and Williams' law is slightly inferior at higher
velocities and is a better fit below 20 m s-1. In this velocity
region, as pointed out in 52.2, equation (6) can be applied
to Bows with turbulence intensities above 30 % without
introducing more than maximum 5 % error due to curve
approximation. The extension to higher turbulence intensities
by using Collis and Williams' law can only be achieved by
the determination of A( V) and B( V) for each individual
wire. The use of any other value of B than the value given by
equation ( 7) will immediately reduce the applicable velocity
range.
These calculations show that only below 20 m s-l is Collis
and Williams' law a better approximation than the calibration
law (6). This law is, however, such a good fit in this region
that the normal extra calibration procedure necessary for
using Collis and Williams' law seldom can be justified.
228
Znterpretation of a hot wire signal using a uniseusal calibration law
Velocity ( n s-')
A 5
Table 3 Velocity ranges corresponding to maximum 1 %
uncertainty in the velocity estimate using the calibration
law (7)
-
L
cy
4 cy
F
5 0. 90-
U n
3
0.85
King's law Collis and Williams' law
o 24
. x 60 ~
x x x
x X L X 4
* x X
A A X X
r O X
0 0
0 0 A o O 0
0 0
0 I I I I I 01
1 1.5 0.7 2.2 2.0 0.7 2.9
2 4.5 1.0 4.5 10 0.8 12.5
4 9.0 1.5 6.0 30 1.0 30
6 12 2.5 4.8 26 2.0 13
8 16 4.0 4.0 26 2.0 13
10 18 5.0 3.6 26 2.0 13
20 32 12 2-8 36 9.0 4.0
30 46 20 2.3 48 16 2.9
40 65 30 2.2 70 30 2.3
50 70 34 2.1 75 34 2.2
60 85 44 1.9 85 42 2.0
70 95 52 1 *8 100 50 2.0
80 110 60 1 *8 110 60 1.9
90 120 65 1.8 125 65 1 *9
100 135 75 1.8 135 75 1 a8
The universal law (4) requires only one accurate measure-
ment of corresponding E and V value as well as the measure-
ment of EO, but a small error due to small variations in shape
is introduced by this method'(see $2.1). King's law and Collis
and Williams' law has the advantage of permitting an indi-
vidual calibration of each wire. However, unless A( V) and
B( V) are determined for each test run, these laws will not
give any improved accuracy above the universal shape
approach.
3 Hot wires yawed to flow direction
The heat transfer from hot wires yawed to flow direction, in
the following denoted as 'yawed wires', has been studied by
several investigators (Kjellstrom and Hedberg 1968, Sanborn
and Lawrence 1955, Webster 1962, Champagne et al. 1967
Friehe and Schwarz 1968). A summary of these findings has
been given by Bruun (1969).
The angle a between the flow direction and the normal to
the wire is in the following used to describe the positioning
of the wire.
Introducing the effective cooling velocity
Ve =Vf ( a) (9)
in the calibration law equation (7), gives
E,'=A+BVen. (10)
In many hot wire measurements normal component cooling
or cosine law cooling, i.e. Ve= Vcos a, has been assumed.
The above mentioned investigations, however, have shown
that a considerable error can be introduced with this assump-
tion.
To correct for this deviation, several expressions for the
function f ( a ) have been suggested, normally applicable for
a e 60". Some of these are, however, too cumbersome for
practical applications.
Two applicable expressions for f ( a) are
f1(a) =cos mq a ) (1 1)
f2(a) =(cos +k2 sin 2a)l;'. (12)
The yaw parameters k and m1will, in general, be functions
of a: and V and to a minor extent of Lid, TWIT, wire material
and prong configuration.
3.1 Experimental investigation of yaw parameters
The heat transfer measurements for the yawed wires were
performed in the potential core of a 2 in air jet. The velocity
range investigated was 1 m s-1 to 60 m s-1. The velocity was
determined with a 2mm hot wire probe using the universal
calibration law (equation (4)). The yawed hot wire probe
was placed horizontally, having the support perpendicular to
the flow velocity. The measurements of 01 were performed
with a telescope fitted with a graticule and mounted on the
opposite side of the jet from the probe. In this way a could
be determined to within f +'.
The yaw parameters k and ml were determined from the
universal calibration law in the form
Ea2 - Eo2 =Cf( Ve).
(1 3)
E, was measured as a function of the angle a and of the
velocity V. By measuring EO and Ea=o as a function of V,
the constant C in equation (13) could be calculated, Knowing
these quantities, equation (13) was then used to calculate Ve
as a function of V and a. By using equations (lo), (11) and
(12) the dependence of the yaw parameters k and ml on a
and V was evaluated.
The result for the yaw parameter k is plotted in figure 7
for three different velocities, showing a large variation in k
with a for this probe type and probe support orientation.
The dependence of the yaw parameter ml on a: is plotted in
figure 8 showing ml to be nearly independent of a for a <70".
The velocity dependence of this yaw parameter is seen from
figure 9 to be very small.
0.3
E
n
0.
X
X
A
(m 5-11
* *
4
I I I I I I
20 40 60 80
Angl e of y a w
Figure 7 Dependence of k on angle of yaw and velocity
Angl e of ya w
Figure 8 Dependence of ml on the angle of yaw
229
H H Bruun
I
I I I
0 I O 20 30 40 50 60
Velocity (m s-1)
Figure 9 Variation in m1 with velocity
For practical yaw measurements it is necessary to express
equation (13) in an analytic form. Using the power law the
equation takes the form
where K and n are constant values corresponding to
Ve= V(x=O0). This law is, as mentioned in $2.2, only an
approximation to equation (13). For large values of V/ Ve
(i.e. a large), deviations between Vecalculated from equation
(13) and equation (14) will occur and therefore also deviations
in the yaw parameters. The difference in the yaw parameters
calculated from (13) and (14) was negligible at small values of
a. At 45" the difference amounted to approximately 5% and
at 60" to 10%. Remembering that the yaw parameters only
are corrections to the cosine law, this uncertainty in the yaw
parameter will only amount to 1-2% relative uncertairdy in
the magnitude of the turbulent quantities.
The yaw parameter ml is seen to be nearly independent of
a and V ( a <70") for this probe type and probe orientation.
The following analytic representation of the yaw calibration
law U as therefore chosen for calculations
E,'-EO'=K{VCOS ~~( ~L ) }~=K V ~C OS ma , (15)
The velocity dependence of m is given in table 1 for values of
V < 60 m s-1.
4 Effect of probe support orientation on calibration law
Recently a number of investigators (Hoole and Calvert 1967,
Gilmore 1967, Dahm and Rasmussen 1969, Eyre 1967,
Guitton 1968) have shown that the hot wire response depends
on the angle of incidence P. P is defined for a normal wire as
the angle between the mean velocity vector and the direction
of the hot wire support. Dahm and Rasmussen (1969)
observed for probes with the active length extending right to
the prongs, a difference of approximately 20% in velocity
estimates for P=O" and P=90c. By using hot wires with
plated ends, removing the active length from the prongs, the
I .76
0 20 40 60 80
Angle of incidence
Figure 10 Variation in hot wire voltage with angle of
incidence
P dependence is reduced considerably. A typical variation in
E with is for a 2 mm ISVR probe is shown in figure 10 for a
velocity of 30 m s-l. There is a 2.5 % difference in the esti-
mated velocity between P=O" , 45" and 8 % difference between
/3=0" and 90".
A series of tests was carried out to investigate whether this
difference in estimated velocity between P=O" and 90" is
independent of velocity. If the difference is constant the same
universal functionf( V ) can be applied both when the support
is parallel and when it is perpendicular to the flow. Two
different 2 mm hot wires were placed normal to the flow in a
2 in air jet. One hot wire A had the support perpendicular to
the flow and the other B the support parallel to the flow. The
ratio (E2 - Eo2)~/(E2 - Eo~) B was calculated as a function of
the velocity. Some consistent difference between the estimated
velocity from hot wire A and B using the function f(V)
emerged. The result in terms of difference in estimated
velocity for several such tests are plotted in figure 11. The
figure shows that the two support orientations give slightly
different types of calibration curves. The similar trend of all
the tests prove that the concept of a (different) universal
functionf*( V ) can also be applied for this support orientation.
For the sake of brevity the value of this function has not been
given.
x Test I
o Test 2
I I I I I 1 I
0 20 40 60 80
Velocity (m 5- 9
Figure 11 Error in the estimated velocity for parallel
support V, using the universal function f( V )
5 Measurements of turbulent quantities in flows with low
turbulence intensity by single hot wire techniques
In measurements of the turbulent structure of a flow it is
assumed that the calibration law for steady flow (equation
(15)) is valid when E, Vand a are replaced by the instantaneous
values. For flows with small turbulence intensity (< 10%)
only first order fluctuation terms need to be taken into account.
Representing the instantaneous values by their mean values
and fluctuations, equation (15) becomes
(1 6)
(I?,+e)z- Eo2= K( o+ cosm 6
where 0 is the angle between the instantaneous flow direction
and the normal to the wire. By using the series expansion
derived by Bruun (1969) for Vn and COS^6 and only retaining
first order terms this equation can be written as
U
Ea2 -t 2Eae - Eo2=KOn COS^a
5.1 Normal wire measuremenfs
Applying equation (17) to a normal wire (a=O) gives the
following relationship :
By taking the time average, the equation for the mean value
is seen to satisfy the steady flow equation (6). Subtracting this
equation from equation (18), the equation for the fluctuating
velocity component is obtained:
(19)
U
2Ee =K ofin =
U'
230
Interpretation of a hot wire signal using a universal calibration law
By squaring and taking time averages, the equation for the
time mean square value becomes
5.2 Yawed wire measurements
Subtracting the steady state equation (15) from equation (17),
the equation for the fluctuating quantities is obtained:
2E,e= KO" cos ma n -- m tan ( a) " ) -= (21)
By squaring equation (21), taking time averages and intro-
ducing E* =(Ea2 - Eo2)/i?, 0, the equation for the time mean
square value can be written as
4p=E*2{n22+m2 tan2(a)F2-2mntan(a)E}. (22)
If one measures the quantity 3 corresponding to +01 and
the similar quantity z2 corresponding to - a at the same
flow condition, their difference will be
( e T - g ) = E * 2 mn tan(a)m (23)
( r C' .
while their sum mill be
2(e7++7)=E*2(n2 s +m2 tan * ( a ) 2 ) . (24)
The three equations (20), (5) and (24) thus enable us to
calculate the Reynolds stress ZiT and the mean square values
i? and 2.
From equation (23) and (24) it is obvious that the assump-
tion of normal component cooling, i.e. m=n, can lead to a
considerable error even at very low turbulence intensity. For
a 2 mm ISVR probe, placed with the support perpendicular
to a flow having a velocity of 30 m s-l, the value of n-m will
be approximately 0.055. This difference gives 11 % deviation
for and 21% deviation for 7 from the corresponding
normal component cooling quantities. Similar deviation was
estimated by Champagne and Sleicher (1967) using the yaw
parameter k.
6 Conclusion
This investigation has shown that the calibration law for
probes with nominally the same geometry can be expressed
in terms of a universal calibration law. The necessary cali-
bration work can therefore be reduced considerably by using
the values given in table 2 for the 2 mm ISVR probe.
Comparison with King's and with Collis and Williams'
law shows that as light improvement in the accuracy for
velocities less than 20 m s-l may be obtained uith the latter
but only at the expense of considerably more calibration work.
A method for calculating the Reynolds stresses 2, 2 and
based on the universal calibration law concept has been
and derived. This method gives deviations of 11 % for
21 % for ? from the simple cosine law results.
Acknowledgments
The author wishes to thank Dr P 0 A L Davies and other
members of the Institute for their advice and help, and the
British Council for providing a research grant.
Appendix Three methods for calculating the exponent n in
the calibration law (6)
(i) In this procedure the exponent n is calculated from two
points 1 and 2 on the calibration curve E=f ( V) by assuming
n and K to be constant between 1 and 2.
By subtracting equation (26) from (25) and taking the
logarithm of both sides, n can be evaluated from
(27)
lg (EL' - Eo2) - lg (E2 - Eo2) .
n =
lg v1-lg vz
This procedure has been used by Kjellstrom and Hedberg
(1968) for a 1 mm DISA hot wire, giving a similar variation
of n with velocity as in figure 4.
(ii) The universal calibration law
with K and n as functions of V can be used for optimizing n.
Differentiating equation (28) with respect to V,
E2-Eo2=KV'=Cf(V) (28)
inserting Vn =(E' - Eo2)/K, and rearranging, the exponent n
becomes
Using the universal function f (V) for determining
(E2-Eo2)- 1 d(E2- Eo2)/dV this equation and an iterative
procedure were used for obtaining the optimum value of n
for a given calibration curve.
(iii) When n varies slowly with the velocity, K and n in the
calibration law
E' - Eo2=KV'
(30)
can be assumed to be constant over a limited velocity range.
In this case the following relationship for the derivatives is
obtained:
dE
dV
2E - =KVn-ln.
(31)
By introducing equation (30) into equation (31) one obtains
2EV dE
____
E2-Eo2 dV'
Using the function E= f ( V ) and approximating dE/d V by
(hE/hV) n(V) was obtained.
References
Bruun H H 1969 I SVR Technical Report No. 21
Champagne F H Sleicher C A and Whermnnn 0 H 1967
J. Fluid Mech. 28 153-75
Champagne F H and Sleicher C A 1967 J. Fluid Mech. 28
Collis D E and Williams M J 1959 J. Fluid Mech. 6 357-84
Dahm M and Rasmussen C G 1969 DI SA Information No. 7
Davies P 0 A L and Davis M R 1966 Institute of Sound and
Vibration Research (Southampton University) Report No. 155
Eyre D 1967 United Kingdom Atomic Energy Authority,
TRG Report 1521 (W)
Friehe C A and Schwarz W H 1968 Trans. Am. Soc. Mech.
Engrs 35 655-62
Gilmore D C 1967 Mechanical Engineering Research
Laboratories (McGill University) Report TN 67-3
Guitton D E 1968 Mechanical Engineering Research
Laboratories (McGill Unioersity) Report TN 68-6
Hoole B J and Calvert J R 1967 J. R. Aero. Soc. 71 511-3
King L V 1914 Phil. Trans. R. Soc. A214 373-433
Kjellstrom B and Hedberg S 1968 Aktiebolaget Atomenergi
(Stockholm) Report AE-3 38
Sanborn V A and Laurence J C 1955 NACA Report TN3563
Webster C A G 1962 J. Fluid Mech. 13 307-12
J ournal of Physics E: Scientific Instruments 1971 Volume 4
Printed in Great Britain
177-82
19-24
23 1

Das könnte Ihnen auch gefallen