You are on page 1of 214

MONTENEGRIN ACADEMY

OF SCIENCES AND ARTS








Petar V. Vukoslavevi
Dragan V. Petrovi

M
M
U
U
L
L
T
T
I
I
P
P
L
L
E
E

H
H
O
O
T
T
-
-
W
W
I
I
R
R
E
E
P
P
R
R
O
O
B
B
E
E
S
S

MEASUREMENTS OF TURBULENT
VELOCITY AND VORTICITY
VECTOR FIELDS








Podgorica, 2000











































MONTENEGRIN ACADEMY OF SCIENCES AND ARTS




































And the array of seeming disorder, always follow some definite order,
(Petar Petrovi Njego in the book The Mountain Wreath, 1847)

___________________________________________________________________
MONTENEGRIN ACADEMY OF SCIENCES AND ARTS


Special Editions
Volume 39


Section of Natural Sciences
Edition no. 21




Petar V. Vukoslavevi, Dragan V. Petrovi

M
M
U
U
L
L
T
T
I
I
P
P
L
L
E
E
H
H
O
O
T
T
-
-
W
W
I
I
R
R
E
E
P
P
R
R
O
O
B
B
E
E
S
S

MEASUREMENTS OF TURBULENT VELOCITY
AND VORTICITY VECTOR FIELDS




Editor
Milojica Jaimovi




Podgorica, 2000
_________________________________________________________________________________
_________________________________________________________________________________

AUTHORS



Petar V. Vukoslavevi
Montenegrin Academy of Sciences and Arts
Rista Stijovia 5, PODGORICA, MONTENEGRO
E-mail: petarvuk@ac.me




Dragan V. Petrovi
University of Belgrade - Faculty of Agricutlure
Institute for Agricultural Technique
Chair of Technical Sciences
Nemnajina 6, 11080, BELGRADE ZEMUN, SERBIA
E-mail: epetrodr@agrif.bg.ac.rs;
pdragan59@gmail.com





Reviewers:

Prof. Simeon Oka, Institute Vina, Belgrade


Academician Vladan orevi, Faculty of Mechanical Engineering, Belgrade


Prof. Svetislav antrak, Faculty of Mechanical Engineering, Belgrade



B BO OO OK K R RE EV VI IE EW W, , M Me ea as s. . S Sc ci i. . T Te ec ch hn no ol l. . V Vo ol l. . 1 12 2 N No o. . 3 3 3 35 54 4, , 2 20 00 01 1, ,
d do oi i: : 1 10 0. .1 10 08 88 8/ /0 09 95 57 7- -0 02 23 33 3/ /1 12 2/ /3 3/ /7 70 05 5. .

M Mu ul lt ti ip pl le e H Ho ot t- -w wi ir re e P Pr ro ob be es s
P P V V V Vu uk ko os sl la av vc ce ev vi ic c a an nd d D D V V P Pe et tr ro ov vi ic c
( (P Po od dg go or ri ic ca a: : M Mo on nt te en ne eg gr ri in n A Ac ca ad de em my y o of f S Sc ci ie en nc ce es s a an nd d A Ar rt ts s) )
1 19 94 4 p pp p 2 20 00 00 0 P Pr ri ic ce e D DM M 4 48 8 I IS SB BN N 8 86 6 7 72 21 15 5 1 10 08 8 9 9 ( (h hb bk k) )

H Ho ot t- -w wi ir re e a an ne em mo om me et tr ry y i is s s st ti il ll l a a m ma aj jo or r r re es se ea ar rc ch h t to oo ol l f fo or r t th he e s st tu ud dy y o of f t tu ur rb bu ul le en nt t f fl lo ow w. .
F Fo ol ll lo ow wi in ng g t th he e p pu ub bl li ic ca at ti io on n o of f t th he e g ge en ne er ra al l b bo oo ok k o on n H Ho ot t- -w wi ir re e A An ne em mo om me et tr ry y b by y H H H H B Br ru uu un n i in n
1 19 99 95 5 t th he er re e h ha as s b be ee en n a a n ne ee ed d f fo or r a an n u up p- -t to o- -d da at te e d de es sc cr ri ip pt ti io on n o of f s so om me e o of f t th he e m mo or re e a ad dv va an nc ce ed d h ho ot t- -
w wi ir re e a an ne em mo om me et tr ry y m me et th ho od ds s. . T Th he e b bo oo ok k M Mu ul lt ti ip pl le e H Ho ot t- -w wi ir re e P Pr ro ob be es s b by y e ex xp pe er rt ts s i in n t th hi is s f fi ie el ld d P P V V
V Vu uk ko os sl la av vc ce ev vi ic c a an nd d D D V V P Pe et tr ro ov vi ic c m ma ak ke es s a a s si ig gn ni if fi ic ca an nt t c co on nt tr ri ib bu ut ti io on n t to o t th he e v ve er ry y a ad dv va an nc ce ed d
r re es se ea ar rc ch h t te ec ch hn ni iq qu ue es s o of f m mu ul lt ti ip pl le e h ho ot t- -w wi ir re e p pr ro ob be es s. .
T Th he e b bo oo ok k w wa as s p pr ro od du uc ce ed d ` `i in n h ho ou us se e' ' i in n Y Yu ug go os sl la av vi ia a u un nd de er r r ra at th he er r d di if ff fi ic cu ul lt t c ci ir rc cu um ms st ta an nc ce es s, ,
w wh hi ic ch h u un nf fo or rt tu un na at te el ly y h ha as s a af ff fe ec ct te ed d t th he e q qu ua al li it ty y o of f t th he e E En ng gl li is sh h l la an ng gu ua ag ge e. . H Ho ow we ev ve er r, , i in n m my y
o op pi in ni io on n t th he e b bo oo ok k i is s s sc ci ie en nt ti if fi ic ca al ll ly y o of f a a h hi ig gh h s st ta an nd da ar rd d. .
C Ch ha ap pt te er r 2 2 c co on nt ta ai in ns s a a b br ri ie ef f d de es sc cr ri ip pt ti io on n o of f t tu ur rb bu ul le en nc ce e a an nd d i it ts s m me ea as su ur re em me en nt ts s w wi it th h a a w we el ll l
d di is sc cu us ss se ed d c co om mp pa ar ra at ti iv ve e r re ev vi ie ew w o of f h ho ot t- - w wi ir re e a an ne em mo om me et tr ry y a an nd d l la as se er r D Do op pp pl le er r a an ne em mo om me et tr ry y. .
C Ch ha ap pt te er r 3 3 c co on nt ta ai in ns s a a g ge en ne er ra al l r re ev vi ie ew w o of f c co on ns st ta an nt t- -t te em mp pe er ra at tu ur re e h ho ot t- -w wi ir re e a an ne em mo om me et tr ry y a an nd d t th he e
r re el la at te ed d s si ig gn na al l a an na al ly ys si is s. . T Th he e a au ut th ho or rs s c co ov ve er r a al ll l t th he e r re el le ev va an nt t a as sp pe ec ct ts s w wi it th h s so om me e s si ig gn ni if fi ic ca an nt t
d di is sc cu us ss si io on ns s o on n f fe ee ed db ba ac ck k c ci ir rc cu ui it ts s a an nd d h ho ot t- -w wi ir re e c co oo ol li in ng g v ve el lo oc ci it ti ie es s. . F Fi ig gu ur re es s a ar re e r re el le ev va an nt t, ,
a al lt th ho ou ug gh h s so om me ew wh ha at t c cr ru ud de e i in n p pl la ac ce es s. .
T Th he e m ma ai in n t th he em me e o of f t th he e b bo oo ok k M Mu ul lt ti ip pl le e H Ho ot t- -w wi ir re e P Pr ro ob be es s i is s i in nt tr ro od du uc ce ed d i in n c ch ha ap pt te er r 4 4 w wi it th h
r re ef fe er re en nc ce e t to o t th he e m me ea as su ur re em me en nt t o of f t tu ur rb bu ul le en nt t v ve el lo oc ci it ty y f fi ie el ld ds s. . T Th hi is s s se ec ct ti io on n c co on nt ta ai in ns s a a w we el ll l l la ai id d
o ou ut t p pr re es se en nt ta at ti io on n o of f t th he e m ma an ny y c co on nt tr ri ib bu ut ti io on ns s i in n t th hi is s f fi ie el ld d, , r ra an ng gi in ng g f fr ro om m s si in ng gl le e- -s se en ns so or r t to o e ei ig gh ht t- -
s se en ns so or r p pr ro ob be es s. . T Th he e s si ig gn na al l a an na al ly ys si is s o of f m mu ul lt ti i- -w wi ir re e p pr ro ob be es s i is s o of ft te en n s so om me ew wh ha at t c co om mp pl li ic ca at te ed d a an nd d
t th hi is s c ch ha ap pt te er r c co ou ul ld d h ha av ve e b be en ne ef fi it te ed d f fr ro om m a a l lo on ng ge er r s si ig gn na al l a an na al ly ys si is s c co on nt te en nt t. .
I In n c ch ha ap pt te er r 5 5 t th he e u us se e o of f m mu ul lt ti ip pl le e h ho ot t- -w wi ir re e p pr ro ob be es s f fo or r m me ea as su ur re em me en nt ts s o of f t tu ur rb bu ul le en nt t v vo or rt ti ic ci it ty y
f fi ie el ld ds s i is s d di is sc cu us ss se ed d. . T Th hi is s i is s a a m ma aj jo or r a as sp pe ec ct t o of f t th he e b bo oo ok k w wh hi ic ch h a al ls so o r re ef fl le ec ct ts s t th he e s si ig gn ni if fi ic ca an nt t
r re es se ea ar rc ch h c co on nt tr ri ib bu ut ti io on n b by y t th he e a au ut th ho or rs s i in n t th hi is s a ar re ea a. . T Th hi is s c ch ha ap pt te er r c co or rr re ec ct tl ly y c co on nt ta ai in ns s a a
s si ig gn ni if fi ic ca an nt t d de es sc cr ri ip pt ti io on n o of f t th he e r re el la at te ed d s si ig gn na al l a an na al ly ys si is s f fo or r s so om me e o of f t th he e m mu ul lt ti ip pl le e h ho ot t- -w wi ir re e
p pr ro ob be es s w wi it th h a a n na at tu ur ra al l e em mp ph ha as si is s o on n t th he e a au ut th ho or rs s' ' w wo or rk k. .
C Ch ha ap pt te er r 6 6 c co on nt ta ai in ns s a a d di is sc cu us ss si io on n o of f t th he e o op pe er ra at ti io on na al l p pr ri in nc ci ip pl le es s o of f m mu ul lt ti ip pl le e h ho ot t- -w wi ir re e p pr ro ob be es s. .
T Th he e o op pt ti im ma al l n nu um mb be er r o of f h ho ot t- -w wi ir re es s a an nd d t th he ei ir r c co on nf fi ig gu ur ra at ti io on ns s a ar re e d de ef fi in ne ed d a an nd d t te es st te ed d f fo or r
t tu ur rb bu ul le en nt t f fl lo ow w m me ea as su ur re em me en nt ts s. .
C Ch ha ap pt te er r 7 7 c co on nt ta ai in ns s a a d di is sc cu us ss si io on n o of f g ge en ne er ra al li iz ze ed d n nu um me er ri ic ca al l p pr ro oc ce ed du ur re es s f fo or r v va ar ri io ou us s m mu ul lt ti ip pl le e
h ho ot t- -w wi ir re e p pr ro ob be es s. .
M Mu ul lt ti ip pl le e h ho ot t- -w wi ir re e p pr ro ob be es s a ar re e v ve er ry y u us se ef fu ul l a al lb be ei it t c co om mp pl li ic ca at te ed d r re es se ea ar rc ch h t te ec ch hn ni iq qu ue es s f fo or r t th he e
s st tu ud dy y o of f t tu ur rb bu ul le en nt t f fl lo ow w. . T Th he e b bo oo ok k g gi iv ve es s a a g go oo od d s sc ci ie en nt ti if fi ic c o ov ve er rv vi ie ew w o of f t th hi is s t to op pi ic c a an nd d i is s
r re ec co om mm me en nd de ed d r re ea ad di in ng g f fo or r r re es se ea ar rc ch he er rs s u us si in ng g h ho ot t- -w wi ir re e a an ne em mo om me et tr ry y. .
H H H H B Br ru uu un n
f fa am mo ou us s r re es se ea ar rc ch he er r i in n t th he e a ar re ea a o of f t th he er rm ma al l a an ne em mo om me et tr ry y a an nd d a au ut th ho or r o of f t th he e f fu un nd da am me en nt ta al l b bo oo ok k: :
H Ho ot t- -W Wi ir re e A An ne em mo om me et tr ry y P Pr ri in nc ci ip pl le es s a an nd d S Si ig gn na al l A An na al ly ys si is s , , O Ox xf fo or rd d U Un ni iv ve er rs si it ty y P Pr re es ss s, , 1 19 99 95 5. .










T TA AB BL LE E O OF F C CO ON NT TE EN NT TS S


T TA AB BL LE E O OF F C CO ON NT TE EN NT TS S I
N NO OT TA AT TI IO ON N III

1 1. . I IN NT TR RO OD DU UC CT TI IO ON N 1

2 2. . T TU UR RB BU UL LE EN NC CE E A AN ND D I IT TS S M ME EA AS SU UR RE EM ME EN NT T 4
2.1 Introduction to turbulence 4
2.2 The aims of turbulence research 5
2.3 Basic operational principles of hot-wire anemometers 6
2.4 Basic operational principles of laser-Doppler anemometers 8
2.5 Hot-wire anemometers and laser-Doppler systems: a comparison review
of the operational properties 10
2.6 Hot-wire anemometers and laser-Doppler systems: competitors
and complements 18

3 3. . C CO ON NS ST TA AN NT T T TE EM MP PE ER RA AT TU UR RE E H HO OT T- -W WI IR RE E A AN NE EM MO OM ME ET TE ER RS S: :
G GE EN NE ER RA AL L P PR RI IN NC CI IP PL LE ES S A AN ND D S SI IG GN NA AL L A AN NA AL LY YS SI IS S 20
3.1 General heat-transfer equations 20
3.1.1 Hot-wire cooling mechanism 20
3.1.2 Heat generation the Joules effect 24
3.1.3 Heat balance equation hot-wires cooling law 25
3.2 Basic principles of constant temperature anemometers 27
3.2.1 Development of constant temperature anemometers 27
3.2.2 Basic elements of CTA feedback circuit 28
3.2.3 Basic steps of CTA set-up procedure 30
3.2.4 The operating point of a feedback circuit 32
3.3 Effective cooling velocity 34
3.3.1 The reference coordinate system for fluid velocity vector 34
3.3.2 Effective cooling velocity for a heated wire 36
3.3.3 The generalised law of hot-wire cooling 40
3.3.4 The cosine law and hot-wire effective cooling angle 41
3.4 A typical CTA experimental set-up 44
II P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
4 4. . M ME EA AS SU UR RE EM ME EN NT TS S O OF F T TU UR RB BU UL LE EN NT T V VE EL LO OC CI IT TY Y F FI IE EL LD DS S B BY Y
M MU UL LT TI IP PL LE E H HO OT T- -W WI IR RE E P PR RO OB BE ES S 47
4.1 Introduction 47
4.2 Basic types of the probes for velocity field measurements 48
4.3 Hot-wire probes design 51
4.4 Directional sensitivity of a finite-length-wire probe 53
4.5 Hot-wire probes for 3-D velocity measurements 57
4.6 Influence of hot-wire probe dimensions and geometry on
the measurement accuracy and uniqueness domain 64
4.7 Special hot-wire configurations the eight sensor probe 69

5 5. . M ME EA AS SU UR RE EM ME EN NT TS S O OF F T TU UR RB BU UL LE EN NT T V VO OR RT TI IC CI IT TY Y F FI IE EL LD DS S
B BY Y M MU UL LT TI IP PL LE E H HO OT T- -W WI IR RE E P PR RO OB BE ES S 77
5.1 Vorticity 77
5.2 Spatial resolution of the probes 80
5.3 Measurement of longitudinal vorticity component 90
5.4 Measurement of two cross-stream vorticity components 101
5.5 Simultaneous measurements of two cross-stream vorticity components 104
5.6 Simultaneous measurements of three vorticity components 107

6 6. . O OP PE ER RA AT TI IO ON NA AL L C CH HA AR RA AC CT TE ER RI IS ST TI IC CS S O OF F T TH HE E V VO OR RT TI IC CI IT TY Y- -T TY YP PE E
A AN ND D H HO OT T- -W WI IR RE E P PR RO OB BE ES S S SP PE EC CI IF FI IE ED D F FO OR R V VE EL LO OC CI IT TY Y M ME EA AS SU UR RE EM ME EN NT T 127
6.1 Measurement accuracy of the mean statistical parameters
of turbulence velocity field 127
6.2 Relationships between the higher-order velocity moments 140
6.3 Minimisation of the sampling frequency and time 144

7 7. . G GE EN NE ER RA AL LI IS SE ED D N NU UM ME ER RI IC C P PR RO OC CE ED DU UR RE E 151
7.1 Equations governing the hot-wire cooling 151
7.2 Data reduction for triple and four-wire probes 155
7.3 The algorithm for twelve-sensor probes WP-12+/q(G) 161
7.4 The procedure for two-sensor probes 166
7.5 Possible advances of the generalised numeric procedure 168
7.5.1 General principles 168
7.5.2 Dual directional generalised numeric procedure for quadruple probes 170

8 8. . C CO ON NC CL LU US SI IO ON NS S 177

9 9. . R RE EF FE ER RE EN NC CE ES S 180














N NO OT TA AT TI IO ON N

Numbers in brackets indicate sections in which quantities are defined or first used.

A, A
i
calibration constant (3.1.3-A), constant in a polynomial fit (3.1.3-A
i
)
a, a
T
hot-wire overheat ratio [a=R/R
f
] (3.1.1-a), [a
T
=(T-T
f
)/R
f
)] (3.1.1-a
T
), half-axis
of an offline contraction that describe hot-wire response (4.4)
a
1
-a
5
calibration coefficients of hot-wire cooling law (3.3.3), (6.2)
a
1H
-a
5H,
a
1V
-a
5V
calibration constants for hot-wires placed in a horizontal and vertical
planes, respectively (3.3.3), (7.1)
a
i1
-a
i5
, a
i,j1
-a
i,j5
calibration coefficients of generalised cooling law of hot-wire no. i
(7.2-a
i1
-a
i5
), included in array no. j=1,2,3 of vorticity probe (7.3-a
i,j1
-a
i,j5
)
B calibration constant of hot-wire cooling (3.1.1), (3.1.3), binormal coordinate
axis in a local coordinate system of hot-wire (3.3.1)
B, N, T binormal, normal and tangential axis of wire local coordinate system (3.3.1)
b constant in hot-wire cooling law (3.3.2), half-axis of an offline contraction
describing wire response (4.4), channel half-width (5.2)
b
m
, b
1
-b
3
, n
m
, n
1
-n
3
, t
m
, t
1
-t
3
transformation coefficients between the fluid velocity
components U, V, W in a space-fixed Cartesian coordinate system and
components U
n
, U
b
, U
t
in a local system of hot-wire, where m=1,2,3 (3.3.3)
b
n
, b
in
polynomial constants of 3-D generalised cooling law for i-th hot-wire,
where n=1,2,3,4,5 (7.1)
b
i,j
n
,

b
i,j1
-b
i,j5
polynomial constants of 3-D generalised cooling law for wire no. i=1,2,3,4
included in array no. j=1,2,3 of vorticity probe, where n=1,2,3,4,5 (7.1),(7.3)
0
b ,
0
n ,
0
t unity ort-vectors of B, N, T axes in a local cooridnates of a hot-wire (3.3.1)
C calibration exponent (3.1.1)
C
ij1
-C
ij6
, C
ijl
calibration constants of i-th sensor in j-th array, where l=1,2,3,4,5,6 (5.6)
C
0
, C
0j
geometrical centre of a multiple hot-wire probe and its array no. j=1,2,3 (7.3)
c sound velocity (3.1.1), specific heat of hot-wire material (3.1.3), half-axis of
an offline contraction that describe hot-wire response (4.4)
c
p
, c
v
specific heat of a fluid under constant pressure and volume (3.1.1)
c
1
-c
6
, c
jk
generalised constants of hot-wire cooling law, where j=1,2,3,4,5,6 denotes
a sequential number of a constant, while k=1,2,3 designates a sensor (3.3.3)
D hot-wire diameter (3.1.1), probe sensing volume diameter (4.2)
D
1
, D
2
, D
3
sleeves diameter, prongs diameter and stem diameter (3.3.2)
D
+
,D
q
diameter of a front sensing area of plus and quadrate probe (4.7)
E, E voltage drop along the hot-wire (3.1.2) and its time-averaged value (3.4)
E
i
, E
ij
anemometer output voltage of the i-th sensor of velocity hot-wire probe
(5.6-E
i
), or i-th sensor of j-th array of the vorticity probe (7.3-E
ij
)
IV P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
E
QI
amplifier offset voltage in constant-temperature anemometer (3.2.3)
E
0
output electrical voltage of a constant temperature anemometer at flow
velocity (3.2.2) and at zero-velocity (3.3.4)
E
1
, E
2
characteristic electrical voltage in the branches of Wheatstone bridge (3.2.2)
AE
AC,
, AE
BD
characteristic voltage drops of a Kovasznay probe (5.3)
e calibration constant in a hot-wire cooling law (3.3.2)
e, e
0
fluctuating part of voltage drop E along hot-wire (3.4)
F convection cooling heat flux from a heated wire to the surrounding flow
(3.1.1), flatness factor (4.7)
( ) V F ) E , E , E , V ( F =
3 2 1
characteristic criteria function (5.6)
( ) 0 =
j j
E , W , V , U F characteristic response functions of j-th hot-wire (j = 1,2,3) of a
probe specified for measurements of fluid velocity vector (5.6)
( ) 0
0
1
=
j , i i i
E , V F , ( ) 0
0
3
=
j , i i i
E , V F characteristic criteria functions for the hot-wire array
no. j=1,2,3 of vorticity hot-wire probe, if the vertical wire no. 1 or no. 3
(respectively) are used (7.3)
F
low
(V)F
1
(V), F
hi
(V)F
3
(V) criteria functions for hot-wire no. 1 or 3 used (5.6)
F
U
, F
V
, F
W
skewness factors of probability density distribution functions of longitudinal
(streamwise) U, transversal (normal) V and lateral (spanwise) W velocity
fluctuations, respectively (4.7)
f
S
, f
SLIM
sampling frequency and its limiting value (6.3)
f
t
temporal sampling frequency f
S
criterion q
t
3 U f ~ (6.3)
Af
S
a searching step of sampling frequency in optimisation procedure (6.3)
Gr Grashof number [ ( ) ( )
2 3 2
| / T T D g
f
] (3.12)
G
S
gain of a bucking-up amplifier (3.4)
g gravitational acceleration (3.1.1)
h convection heat-transfer coefficient (3.1.1), calibration constant in hot-wire
cooling law (3.3.2)
h
P
, h
S
prongs separation (4.5- h
P
), wires separation of a parallel-wire probe (5.2-h
S
)
h
i
, h
ij
calibration constant of hot-wire no. i (5.3- h
i
), in the j-th array (5.6- h
ij
)
I, I
0
intensity of electrical current through hot-wire (3.1.2-I), feedback current of a
constant temperature anemometer (3.2.2-I
0
) and electrical current passing
through the prongs of Kovasznay probe (5.3-I
0
)
I
1
, I
2
electric current passing through the branches of Wheatstone bridge of a
constant temperature anemometer (3.2.2)
i , j , k ort-vectors of x, y, z axis in a space-fixed Cartesian coordinate system (3.3.1)
Kn Knudsen number [c/D] (3.1.1)
K
ij1
- K
ij6
calibration constants of the sensor no. j in the i-th array (5.6), (7.3) of a
vorticity probe with nine or twelve hot-wires
k amplifier gain (3.2.4), calibration constant in hot-wire cooling law (3.3.2)
k
i,j1
- k
i,j5
, k
1,j1
,-k
4,j5
calibration coefficients of generalised cooling law hot-wire no.
i=1,2,3,4 included in array no. j=1,2,3 (7.3)
L, L
C
, L
S
active sensing length of hot-wire (3.1.1-L), length over which supports
affects the temperature distribution along hot-wire (3.3.2-L
C
), sampling
length ( )
S S
f U L 2 = (6.3-L
S
)
M calibration constant (3.1.1), grid mesh dimension (5.6)
MULTIPLE HOT-WIRE PROBES V
__________________________________________________________________________________________________________________________
Ma Mach number [U
0
/c] (3.11)
m mass of hot-wire sensor (3.1.3)
N calibration constant (3.1.1), measuring error of turbulence parameter (6.3)
NCD % of non-converged data (5.6)
Nu Nusselt number [ ( )
f
/ D h ] (3.1.1)
P heat-flux induced in hot-wire by Joules effect (3.1.2), pressure (5.1)
Pr Prandtl number [ ( )
f p
/ c ] (3.1.1)
p, q calibration exponents (3.1.1-p, q), pressure (5.5-p)
Q
2
, Q
4
roots introduced in 3-D procedure (7.2), or its 2-D variant (7.4)
R, R
f
hot-wire electrical resistance at operational temperature T (3.1.1-R) and at
fluid temperature (3.1.1-T
f
)
R
H
, R
V
residual term in the generalised hot-wire cooling law of a sensor placed
in a horizontal and vertical plane (7.1)
R
K
,R
P
, R
T
,R
SW
resistive elements in a constant temperature anemometer (3.2.2)
resistance of adjustable potentiometer in a Wheatstone bridge (3.1.3),
residual in the generalised cooling law of vertical hot-wire (7.1)
R
i
residual term in the generalised hot-wire cooling law of a sensor of a
probe specified for fluid velocity measurement (7.2)
R
i,j
, R
i,jP
, R
i,jS
residual terms in the generalised cooling law of vorticity probe sensor,
where i=1,2,3,4 denotes hot-wire and j=1,2,3 is for array (7.3)
R
0
hot-wire electrical resistance at referent temperature T
0
(3.1.1)
R
0
equivalent electrical resistance (3.2.4)
R
1
, R
2
resistances of fixed resistive elements in a Wheatstone bridge of a constant-
temperature anemometer system (3.2.2)
Re Reynolds number [ ( ) / D U
0
)] (3.1.1-Re), defined according to
diameter of a wake cylinder (5.4-Re
d
), grid mesh dimension M (5.6-Re
M
),
boundary layer thickness o (5.6-Re
o
), momentum thickness u (5.2- Re
u
),
friction velocity u
t
(5.2- Re
t
)
S hot-wire surface (3.1.1), prongs separation (3.3.2), skewness factor (4.7)
S
U
, S
V
, S
W
skewness factors of a PDFs of U, V W velocity fluctuations (4.7)
S
Ox
skewness factor of longitudinal/streamwise O
x
vorticity PDF (5.6)
S
1
length of hot-wire prongs (3.3.2)
s virtual length of hot-wire sensor (4.4)
T, T
0
operational and referent temperature of a hot-wire sensor (3.1.1)
T
F
, T
f
film temperature [T
F
=(T + T
f
)/2] (3.1.1), fluid temperature (3.1.1-T
f
)
t, t
S
, t
SLIM
time (5.1-t), sampling time (6.3-t
S
) and its limiting value (6.3-t
SLIM
)
At
S
a searching step of sampling time in optimisation procedure (6.3)
U, V, W longitudinal/streamwise, transversal/normal and lateral/spanwise fluid
velocity vector components in the x,y,z-axis directions of a Cartesian space-
fixed coordinate system (2.5),(3.3.1), hot-wire inner energy (3.1.3-U)
U
N
, U
Ne
total normal component of fluid velocity vector
0
U , perpendicular to hot-
wire axis [
2 2 2
b n N
U U U + = ] (3.1.3-U
N
), and its effective value (3.3.2- U
Ne
)
U
Ni
,U
N1
-U
N4
fluid velocity components orthogonal to hot-wires no. i=1,2,3,4 (5.3)
U
N0
component of instantaneous fluid velocity vector in the measuring direction
of laser-Doppler anemometer (2.4)
VI P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
U
b
, U
n
, U
t
, binormal, normal and tangential component of an instantaneous fluid velocity
vector
0
U
in a local coordinate system of hot-wire (3.3.1)
U
e
, U
eG
, U
ieG
hot-wire effective cooling velocity (3.1.3-U
e
) and its generalised form
VW a UW a UV a W a V a U U
eG
5 4 3 2 1
2 2 2 2
+ + + + + =
(6.2-U
eG
), (7.2-U
ieG
),
where indices i denote sensor number (7.4-U
ieG
)
U
ei
, U
eij
effective cooling velocity of i-th wire (4.5-U
ei
), in j-th array (5.6-U
eij
)
U
e1
-U
e4
effective cooling velocity of wire no. 1,2,3,4 of quadruple probe (4.5)
U
k
, U
i
fluid velocity components in tensor notation, where k = 1,2,3 for U, V and W
component in a Cartesian coordinate system (5.1-U
k
), (5.5-U
i
), respectively
U
i
, V
i
interpretted values (in a 2-D generalised numeric procedure) of U, V
velocity components, which correspond to i-th iteration cycle (7.4)
U
0
, V
0
, W
0
magnitude of a fluid velocity vector
0
U
(3.1.1-U
0
), components U, V, W of
the flow velocity vector at probe centre (5.3-U
0
,V
0
,W
0
)
U
ij
, V
ij
, W
ij
components U, V, W of the flow velocity vector at the centre of hot-wire no.
i=1,2,3,4 of j-th array, where j=1,2,3 (7.3)
U
1
-U
4
, V
1
-V
4
, W
1
-W
4
magnitude of a fluid velocity
0
U
(3.1.1-U
1
-U
4
), components U, V, W
of a flow velocity at the centre of hot-wire no. 1,2,3,4 (5.3)
U
1eG
-U
4eG
, U
i,j
eG
,

U
1,jeG
-U
4,jeG
generalised effective cooling velocity of hot-wire no. 1,2,3
and 4 of a quadruple probe (7.2-U
1eG
-U
4eG
) and of sensors no. i=1,2,3,4
included in array no. j=1,2,3 of vorticity probe (7.3-U
i,jeG
,

U
1,jeG
-U
4,jeG
)
U
0s
orthogonal projection of an instantaneous fluid velocity vector
0
U in a plane
of the hot-wire prongs (3.3.1)
U
0i
,V
0i
,W
0i
longitudinal/streamwise, transversal/normal and lateral/spanwise velocity
component at the centre of i-th array (i=1,2,3) of vorticity probe (7.3)
U
01
, U
02
longitudinal/streamwise component of a fluid velocity, calculated using
different sensor combinations of four hot-wire probe (5.3)
U

free-stream (undisturbed flow) velocity (5.3)


u, v, w instant fluctuations of velocity components in Reynolds decomposition (5.2)
u
0
, v
0
, w
0
instant fluctuations u, v, w at the centre of hot-wire probe (5.3)
u
1
, u
2
, u
3
instant fluctuations u, v and w in tensor notation (5.2)
u
t
friction velocity [
t
t
=
w
u
] (5.2)
AU , Au velocity difference between two normal parallel hot-wires, separated at Ay
(5.2), characteristic velocity difference across a free shear layer (5.4)
AU
i,jP
, AV
i,jP
, AW
i,jP
differences between fluid velocity components U, V, W acting on the
probe measuring volume centre C
0
and that at the centre of i-th (i=1,2,3,4)
sensor included in array no. j=1,2,3 (7.3)
AU
i,jS
, AV
i,jS
, AW
i,jS
differences between components U, V, W acting on the j-th array
centre C
0j
and that at the centre of wire no. i=1,2,3,4 in the same array (7.3)
0
U
instantaneous fluid velocity vector (2.4)
X, Y, Z calibration constants (3.1.3-X,Y), force vector components (5.1-X,Y,Z)
x, y, z coordinate axis (or coordinates) in a space-fixed Cartesian system (3.3.1)
x
i
Cartesian coordinates in tensor notation, where i=1,2,3 for x, y and z (5.1)
x
1
, x
2
, x
3
Cartesian coordinates x, y, z in tensor notation (5.2)
y
i,j
P
, z
i,j
P
y and z coordinate of the centre of i-th (i=1,2,3,4) sensor of a j-th array
(j=1,2,3) with respect to the probe centre C
0
(7.3)
MULTIPLE HOT-WIRE PROBES VII
__________________________________________________________________________________________________________________________
y
i,j
S
,
z
i,j
S
y and z coordinates of i=1,2,3,4-th sensor centre with respect to the centre of
array no j=1,2,3 of multiple vorticity probe (7.3).
y
+
non-dimensional distance from the wall [y
+
= u/u
t
] (3.3.1)
Ay, Ay* sensors separation Ay and its normalised value |Ay*=Ay/q=h
S
/q] (5.2)
o, o
i
temperature coefficient of electrical resistance (3.1.1-o), geometrical angle of
hot-wire (3.3.1-o), where indices i denote hot-wire number (5.3-o
i
)
o
e
,o
ei
,o
eij
,o
e1
,o
e2
effective cooling angle of hot-wire (3.3.2-o
e
), no. i=1,2,3,4 (5.3-o
ei
), in
array no j=1,2,3 (5.6-o
eij
), effective cooling angles of X-wires no. 1,2 (3.3.4)
|, 2|
T
coefficient of volume expansion (3.1.1-|), uniqueness cone angle (4.4-2|
T
)
c molecular free path (3.1.1)
c
ijk
alternating tensor, where i,j,k = 1,2,3 (5.1)
c
REP
(U),c
REP
(V),c
REP
(W) reproduction errors of induced velocity components (7.5.2)
c
Ui
, c
Vi
difference between velocity components in two successive iterations (7.4)
c
Ox
measuring error of longitudinal/streamwise O
x
vorticity component (5.3)
c ,
I
c mean kinetic turbulent dissipation rate and its isotropic value (5.2)
q micro-scale of Kolmogorov [ ( )
4 1
3
c v q / ] (3.1.1), (5.2)
u momentum thickness (5.2), jaw angle (5.6)
u
0
angle between normal fluid velocity component
n n
U n U =
0

(in a local hot-


wire coordinate system) and component
s
U
0
in the prongs plane (3.3.1)
u
0c
effective value of angle u
0
(3.3.2)

f
fluid thermal conductivity (3.1.1)
, v, fluiddynamic (3.1.1), kinematic viscosity (3.1.1) and density (3.1.1)
, A pitch calibration angle of hot-wire probe (2.5) and its step (3.3.4)

0
angle between a fluid velocity vector
0
U and its component
0 0
0
b U n U U
b n
s

+ = in the plane of hot-wire sensor prongs (3.3.1)


angle between a fluid velocity vector
0
U and normal (i.e. its total normal
component
N
U ) to hot-wire axis of symmetry (3.1.1), (3.3.1)

e
effective hot-wire cooling angle (3.3.4)
t
w
wall shear stress (5.2)
O
i
vorticity components in tensor notation, where i=1,2,3 for O
x
, O
y
, O
z
(5.1)
O
x
, O
y
, O
z
vorticity components in a space-fixed Cartesian coordinate system (5.1)
O
x
+
,O
y
+
,O
z
+
normalised non-dimensional vorticity components (5.6)
x
O ,
y
O ,
z
O mean (time-averaged) components of vorticity vector (5.1)
e
x
, e
y
, e
z
fluctuating parts of a vorticity vector components O
x
, O
y
and O
z
(5.1)
E,E(t
S
,f
S
) general designation of a turbulence statistical parameter (6.3)







VIII P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________






















































1 1. . I IN NT TR RO OD DU UC CT TI IO ON N


N No ow w I I t th hi in nk k h hy yd dr ro od dy yn na am mi ic cs s i is s t to o b be e t th he e r ro oo ot t o of f a al ll l p ph hy ys si ic ca al l s sc ci ie en nc ce e, , a an nd d i is s a at t p pr re es se en nt t
s se ec co on nd d t to o n no on ne e i in n t th he e b be ea au ut ty y o of f i it ts s m ma at th he em ma at ti ic cs s. .
( (W Wi il ll li ia am m T Th ho om ms so on n - - L Lo or rd d K Ke el lv vi in n, , 1 18 82 24 4- -1 19 90 07 7) )


The object of present textbook is the measurement methodology of turbulent
velocity and vorticity fields by multiple hot-wire anemometer probes. A detailed
general review of scientific contributions in the area of interest is given. Well-
known complexity of turbulence and complicated measurements of its parameters
limited the manuscript to non-compressible isothermal flows. Within these borders,
the influences of hot-wire probe configuration and characteristic dimensions, as
well as the influences of numeric signal interpretation procedure and sampling
parameters, on the measurement results of turbulence statistical parameters are
analysed. The text contains nine chapters, including the list of references, in order
to achieve higher comprehend and readability.
General description of the textbook represents the object of the 1
-st
(current)
chapter. It provides short basic information on the books content.
Complex nature of turbulence and basic principles of the two most popular
contemporary methods for measurement of its parameters, hot-wire anemometer
and laser-Doppler system, are generally described in the 2
-nd
chapter. In addition, a
short comparison review of operational properties of these instruments is provided,
together with general recommendations for their application.
The 3
-rd
chapter describes hot-wire cooling mechanism and functional principles
of the constant-temperature hot-wire anemometer, which is the most commonly
used for turbulent velocity and vorticity fields measurements. An adequate
mathematical background, necessary for clear understanding of this complex
phenomenon, is also provided.
From the origin of thermal anemometry, functional characteristics of hot-wire
probes have been enormously advanced, especially during the last decade.
Primarily, the probe configurations, designed for measurement of turbulent
velocity field, became more complex and sophisticated. The old models with two
2 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
and three hot-wires are substituted with the new generations of quadruple probes
with four sensors and quintuple configuration containing five hot-wires. A
comprehensive review of existing multiple hot-wire configurations and supporting
numeric procedures, specified for fluid velocity measurements, is provided
together with information on their applicability ranges in the 4
-th
chapter. Special
attention is paid to four-sensor anemometer probes, which seem to be the most
popular configurations for investigation of isothermal turbulent flows of higher
turbulence levels. The own configuration with eight wires, which enables direct
reliable checking of its spatial measurement resolution, is also presented.
Hot-wire probes, specified for turbulent vorticity field measurements, are
presented in the 5
-th
chapter, as well as the adequate supporting numeric algorithms
and fundamental principles of fluid vorticity vectors measurement. Besides the
short historical review of their development, starting from the first configuration
with four sensors (Kovasznay 1950), the newest configurations containing nine and
twelve hot-wires are described in detail. The problem of estimating the probe
spatial resolution is also analysed, as well as the existing methods for its resolving.
Although achieve lower measurement accuracy, hot-wire probes designed for
fluid velocity measurements are more operationally suitable then the vorticity
probes with larger number of sensors and extremely complex signal-interpretation
procedures needed to account the gradients over the sensing volume. A comparison
review of the most important operational properties of the vorticity probes and
configurations specified for fluid velocity measurements is presented in the 6
-th

chapter. The anemometer probes with twelve, four and two hot-wires are
compared, enabling analysis of the influence of neglecting the velocity gradients
and/or one of three fluid velocity components on the mean statistics of turbulent
velocity field. Thus, information on the influence of probe geometry and number of
hot wires on the measurement results of turbulent velocity field is provided. An
algorithm for direct estimating the influence of sampling time and frequency on the
measured values of mean statistical parameters of the turbulence velocity and
vorticity fields is also presented. It reduces the computation time, in comparison to
the total-search method of the time-frequency domain, but achieves the same
evaluation accuracy of sampling parameters.
The 7
-th
chapter explains physical and mathematical background of the own
unique procedures for evaluation of the calibration constants and signals
interpretation of the twelve, four and two hot-wire anemometer probes.
Conclusions, remarks, recommendations and plans for future research work in
the area of interest are, as usual, left for the last chapter.
Such sophisticated and voluminous scientific area is covered by analysing
results originating from the leading world aerodynamics laboratories, published
contributions of other researchers, as well as own experiments performed in the
three laboratories: at University of Maryland (USA), University of Montenegro in
MULTIPLE HOT-WIRE PROBES 3
__________________________________________________________________________________________________________________________
Podgorica (FR Yugoslavia) and Friedrich-Alexander University of Erlangen-
Nrnberg (FR Germany). Our experience, acquired during extensive collaboration
that resulted in Ph.D. thesis of Petrovi 1996a, has also contributed to the content
of the textbook. However, this textbook has primarily benefited from the
experimental results and research work done in the Fluid dynamics laboratory at
the University of Maryland (USA), together with Prof. James M. Wallace.


Laminar separation on a thin ellipse. Source: Bradshaw 1970.










2 2. . T TU UR RB BU UL LE EN NC CE E A AN ND D I IT TS S M ME EA AS SU UR RE EM ME EN NT T


I I a am m a an n o ol ld d m ma an n n no ow w, , a an nd d w wh he en n I I d di ie e a an nd d g go o t to o H He ea av ve en n t th he er re e a ar re e t tw wo o m ma at tt te er rs s o on n
w wh hi ic ch h I I h ho op pe e f fo or r e en nl li ig gh ht te en nm me en nt t. . O On ne e i is s q qu ua an nt tu um m e el le ec ct tr ro od dy yn na am mi ic cs s, , a an nd d t th he e o ot th he er r i is s
t th he e t tu ur rb bu ul le en nt t m mo ot ti io on n o of f f fl lu ui id ds s, , a an nd d a ab bo ou ut t t th he e f fo or rm me er r I I a am m r re ea al ll ly y r ra at th he er r o op pt ti im mi is st ti ic c. .
( (H Ho or ra ac ce e L La am mb b, , a a m me ee et ti in ng g i in n L Lo on nd do on n, , 1 19 93 32 2) )


2.1 INTRODUCTION TO TURBULENCE
Turbulence is an important physical phenomenon. Its name originates from the
Greek word "turbo", meaning vortex. Analysis of turbulent flows is very difficult,
because they contain the most possible complex forms. The application of ordinary
hypothesis (two-dimensionality, stacionarity, non-vorticity), effective in other
flows, is not possible here. This is the main reason why turbulence is still one of
the very rare problems in fluid dynamics that has not been completely resolved yet.
Turbulence is a rather well known term that characterises a special class of
fluid motions. It is not easy to formulate its definition that could cover all details of
flow characteristics. Following Hinze 1959, according to Websters "New
International Dictionary" turbulence denotes agitation, disturbances, boiling, etc.
Unfortunately, this definition is too general and is not quite precise to be
completely acceptable. The most of researches believe that completely adequate
definition of turbulence does not exist. Even more, some of them claim that this
phenomenon can be only described by reviewing its main properties.
Tennekes and Lumley 1978 explained the most important physical properties
of turbulence that enable clear recognition of such flows toward the other existing
classes of fluid motions. They are listed and shortly discussed in the following text
and illustrated in fig. 2.1 by photograph of turbulence initiation, development and
dissipation in a free round jet, produced by a flow visualisation technique.
Irregularity. Behaviour of turbulence can be described as accidental in space
and time. This is one of the main motives for applying the statistical methods for
analysis of turbulent flows.
Diffusivity. Caused by intensive fluid velocity fluctuations, it increases fluid
mixing and transfers of impulse, heat and mass.
MULTIPLE HOT-WIRE PROBES 5
__________________________________________________________________________________________________________________________
High values of Reynolds number.
Turbulence usually initiates at high
Reynolds numbers, as the instability of the
laminar flows described through the
interaction of the viscous and non-linear
terms in the equations of motion.
Furthermore, these type of flows fully and
the most clearly manifest some of their
characteristic properties only at high
enough values of Reynolds numbers.
Three-dimensional vortex fluctuations.
High-levels of vortex fluctuations represent
a typical characteristic of turbulent flows.
They can not exist in two-dimensional
flows, because the crucial mechanism of
vortex generation known as vortex
stretching is three-dimensional by its
nature. The importance of this property
becomes even more clear, having in mind
the origin of term turbulence from the
Greek word turbo meaning vortex.
Dissipation. In opposite to accidental
gravitational and sound waves, turbulence is naturally highly dissipating
phenomenon. Viscous tangential stresses are responsible for deformation work that
increases the inner fluid energy on the count of kinetic energy of turbulent
fluctuations and the mean flow. Consequently, turbulence needs continual supply
of energy to the flow, which compensates viscous looses enabling its maintenance.
Otherwise, the intensity of turbulence decreases and the flow becomes laminar.
Continuum. Turbulence is a phenomenon of continuum mechanics with typical
spatial scales, much greater in comparison to typical dimensions of the molecules
and their characteristic scales.
Primarily, turbulence is not a property of fluid, but a property of its flow. At
high values of Reynolds numbers, dynamics of turbulence is basically the same for
all fluids. General behaviour of turbulence does not depend on the fluid, but on its
environment and flow conditions.

2.2 THE AIMS OF TURBULENCE RESEARCH
Extremely complex structure of turbulent flows demands application of
sophisticated methods for their research. The chances for final success in resolving
any specific problem can not be easily predicted in advance, before detailed
analysis is performed. However, the investigation of turbulent flows is of great

Fig. 2.1: Turbulent free round jet.
Source: Van Duke 1982.
6 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
scientific and practical interest, because turbulence causes strong changes of the
integral flow characteristics. They result in a great increase of the coefficients of
mass, impulse and energy exchange. Nowadays, the importance of turbulence
researching is recognised worldwide. Turbulent flows are frequently present not
only in the nature, but also in the engineering practice.
In some cases turbulence effects are negative, like energy dissipation, heat
losses, generation of pressures and aerodynamic forces by turbulent velocity
fluctuations, etc. In some other situations, turbulence represents the main source of
many useful effects, such as the turbulent transport of heat, mass and impulse,
contaminants dispersion, water circulation in nature, etc. These effects usually have
not the reversal influence on turbulence, as is the case for a turbulent transport of
passive contaminants (temperature, solid particles, etc.). However, interactions
between the turbulence and its consequent effects are also present. The effects
caused by turbulence sometimes directly influence intensifying or depressing of
turbulent flows. The most common examples are heat and mass transfer during the
intensive burning that extremely changes physical properties of the fluid, which
(from the other side) change characteristics of turbulent flow.
The aim of turbulence researching is to provide information that could enable or
improve its control. In some cases, it is useful to initiate turbulence or even amplify
its fluctuations, to suppress it in the processes where its effects are negative and
balances its influences in the cases where turbulence has mixed effects. Typical
illustration of the last case is a thermal apparatus, which needs intensive heat
transfer and low energy dissipation.
The most popular theoretical approaches in the area of turbulence research are
turbulence modelling and direct numerical simulation. However, using these
methods it is not possible to resolve completely the existing problems without
direct support of experimentally verified information. Unfortunately, in spite of
extensive measurements performed in aerodynamic laboratories worldwide for
many years, theoretical methods still suffer from the lack of empirical data,
motivating the researches to perform new experiments.

2.3 BASIC OPERATIONAL PRINCIPLES OF HOT-WIRE ANEMOMETERS
At present, only two instruments for true quantitative measurement of the non-
stationary fluid velocity fields exist: hot-wire and laser-Doppler anemometer. Hot-
wire anemometer is an electronic instrument, primarily designed for measurement
of velocity and temperature fluctuations in the non-stationary flows. Its sensors are
very thin metal wires of tungsten, platinum, iridium, platinum-rhodium, platinum-
iridium alloys, etc., 0.25-10 m in diameter and 0.3-5 mm in length. As it is the
case with all conductors of electric current, hot-wire resistance is a function of its
temperature. Anemometer probes are most frequently designed with one or two
hot-wires, although the last probe versions with larger number of sensors also exist.
MULTIPLE HOT-WIRE PROBES 7
__________________________________________________________________________________________________________________________
Wires are heated by electric current (the Joules effect), and simultaneously
cooled mainly by fluid flow of lower temperature (heat convection). Difference
between generated and convective cooling heat flux is balanced by the inner hot-
wire thermal energy that is directly dependent on its temperature, i.e. electric
resistance. Basically, hot-wire anemometer detects heat transfer from a heated
element placed in the flow toward its environment. It is sensitive, virtually
instantaneously, to any change of the flow conditions that affects the heat exchange
between the heated wire and its environment. Therefore, hot-wire anemometry can
be successfully applied for measurement of the fluid velocity and temperature,
concentration changes in fluid mixtures, phase changes in multiphase flows, etc.
Furthermore, some additional important flow properties, like vorticity, can be also
measured by sophisticated hot-wire techniques. Following the mathematical
definition, vorticity vector components can be evaluated using values of fluid
velocity vector components, measured in the proximate flow locations.
Hot-wire anemometers operate in two modes, depending on the way of control-
ling the electrical heating current through the sensor. If this electrical current is
maintained constant, hot-wire operates in the so-called constant-current mode. In
this case, variations of the sensor temperature caused by the flow generate
corresponding changes of the wire resistance and voltage drop along its length.
Establishing the suitable functional form between the flow properties and sensor
resistance enables evaluation of fluctuating flow properties through monitoring the
fluctuations of voltage drops along the hot-wire. The constant-current (CC) mode
of hot-wire anemometer is used primarily for temperature measurements, when a
low-intensity electric current is supplied to the sensor. Under these conditions, hot-
wire sensitivity toward
fluid velocity is
negligible and hot-wire
sensor functions as a
simple resistance-
temperature device,
usually denoted as
cold wire.
In the constant-
temperature (CT)
mode (fig. 2.2), sensor
is connected in the
electrical feedback
circuit that maintains
its resistance nearly
constant. Having in
mind dependence of
wire resistance on its
temperature, it follows

Fig. 2.2: Constant-temperature anemometers, designed
in the Turbulence Flows Laboratory at the
University of Montenegro in Podgorica:
- CTA-12, containing twelve channels (up) and
- CTA-4ex, possessing four low-noise
electronic measuring circuits (down).
8 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
that sensor temperature is also nearly constant. Fluctuations of hot-wire cooling
flux are registered as variations of the wire heating current, by monitoring the
corresponding voltage drops. The constant-temperature anemometer possesses
higher frequency response in comparison to constant-current mode: nearly constant
wire temperature in the constant-temperature mode decreases its thermal inertia. In
addition, it provides high wire sensitivity to the velocity variations and low
sensitivity to temperature changes if the temperature difference between wire and
flow is maintained high enough. Therefore, constant-temperature anemometer, as
more appropriate, is used almost exclusively for fluid velocity measurements.
The simple measurement principle, together with small measuring volumes of
the probes and low costs of these systems, make the hot-wire anemometer very
suitable and popular instrument, especially for fluctuating air/gas velocity and
temperature measurements. From its origin, estimated to be at the beginning of this
century (Comte-Bellot 1976), hot-wire anemometer characteristics have been
enormously improved. Sophisticated electronic components with high signal-to-
noise ratio, fast dynamic response and high stability toward the changes of
environmental temperature (that are not discussed here) have been developed.
Anemometer probes are also greatly advanced: in opposite to the first
configurations with a single sensor to measure only one (of three) fluid velocity
component, the last generation of the probes with nine and twelve hot-wires is
capable of simultaneous measurement of three-dimensional fluid velocity and
vorticity vectors. Besides these complex configurations, a variety of other hot-wire
probes exists, with different dimensions of measuring volumes, numbers of
sensors, their dispositions, applied materials, fabrication technologies, etc.

2.4 BASIC OPERATIONAL PRINCIPLES OF LASER-DOPPLER ANEMOMETERS
Laser-Doppler anemometer measures indirectly the non-stationary gas and
liquid flow velocities. This instrument registers the phase change of monochrome
laser-light, reflected from the tracer that is currently passing through its measuring
volume. Term tracer assumes various types of optical inhomogenity within the
fluid that trace the flow and provide the reflection of monochromatic laser light of
high enough intensity to be registered. In the liquids, the ordinary present
impurities can be used as tracers. However, researching of gas flows is more
complex, because the introduction of specially prepared liquid or solid tracers is
needed. In these cases serious problems may occur. The necessity of tracer
presence in the fluid flow which velocity field is measured is one of the main
disadvantages of laser-Doppler anemometer. This measuring principle results in the
logical question about the reality of fluid velocity representation by the tracers.
Also, it suffers from the problems of choosing tracers material, technique for their
introducing in the flow, dimensions, quantity, etc.
The velocity of fluid flow and its tracer is very small in comparison to light
velocity of 300 000 km/s. This makes impossible direct registration of Doppler
MULTIPLE HOT-WIRE PROBES 9
__________________________________________________________________________________________________________________________
shift of laser-light, caused by tracers movement. The problem was resolved by
introducing the indirect differential method in the operational practice. Two laser-
beams simultaneously illuminate tracer, while the difference of Doppler shifts is
detected on the photo-detector. The resulting frequency is low, and can be
registered easily. Furthermore, it is proved that connection between this frequency
and fluid velocity is linear in the defined direction.
This connection represents theoretical base of the measurement method of laser-
Doppler anemometer. Generation of the characteristic Doppler signal, which is
registered by the photo-detector, is usually visualised by the model of
interference bars (fig. 2.3b). If the wavelength of both beams is equal, the
stationary dark and light bars arise in the intersection zone. They are orthogonal to
the plane of the laser beams. The tracer travels through the measuring volume,
producing the variations of the intensity of reflected laser light. These variations
are registered on the photo-detector as characteristic Doppler signal.
Typical scheme of laser-Doppler anemometer with two beams is sketched in fig.
2.3a. The basic element of the anemometer is laser (1). It generates polarised
monochrome light (2). Optical prism (4) forms two laser beams of equal intensities,
which are separately modulated in defined frequency by a pair of Brags cells (5).
Prisms (6) set-up beams diameters in front of the output lenses (7), which form the
elliptical measuring volume (8) (fig. 2.3b). The light-stop mask (9) is mounted in
front of the focusing
lens (10). Photo-
electric converter
(12) and its mask
(11) are in the
common housing
(13). Converted
signals are forwarded
in the processing
unit. Described
system measure one
component of
velocity in the
defined direction.
Including the
additional pairs of
differently coloured
(in comparison to the
first pair) laser beams
enable measurement
of other components.


9
13
TRACER
U
0
U
N0

U
0
tracer velocity
U
N0
- component in the measuring direction
U
0
U
N0
(a)
(b)
11
3 4 1
2
6 7 8
10 12
5

Fig. 2.3: Laser-Doppler anemometer (LDA):
(a) a functional scheme,
(b) the measuring principle.
Adapted from: Matovi 1988.
10 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
2.5 HOT-WIRE ANEMOMETERS AND LASER-DOPPLER SYSTEMS:
A COMPARISON REVIEW OF THE OPERATIONAL PROPERTIES
Although the laser-Doppler anemometer is a newer and presently very popular
instrument, hot-wire anemometry still represents a very attractive measurement
technique for fluctuating flow properties. The last version of quadruple and
quintuple hot-wire probes (containing four and five hot-wires respectively) can
be successfully applied even in the flows with higher turbulence levels. According
to Dobbeling, Leuckel and Lenze 1990a, quadruple configurations enable
velocity field measurements of turbulence levels up to about 38%. Furthermore,
Holzapfel, Leuckel and Lenze 1994 claimed that their quintuple configuration
enables velocity field measurement in the flows of turbulent levels up to around
fascinating 66%.
Following Bruun 1995, it is likely that hot-wire anemometer still remains the
principal research tool for most of air/gas-flows studies. In addition, for
measurements in the flows of low and moderate turbulence levels below ~25%, it
seems that hot-wire technique is superior in comparison to laser-Doppler
anemometer. Detailed comparison analyses of the typical operational and other
technical properties of hot-wire anemometers and laser-Doppler systems is
presented in the next few paragraphs of the present section.
Cost. Hot-wire systems are cheap in comparison with corresponding laser-
Doppler systems, especially if the latter have to be accompanied with tracers
dozing apparatus (air/gas flow studies).
Frequency response. Most of contemporary constant-temperature hot-wire
anemometers have flat frequency response from 0 Hz to around 20-50 kHz, except
for very low velocities. Furthermore, it is generally accepted worldwide that
reliable measurements of up to several hundred kHz (Bruun 1995), and even up to
1 MHz (in supersonic flows, for example) can be performed. However, this general
belief was little shaded by recent studies of Saddoughi and Veeravalli 1994, 1996,
who pointed out non-adequate hot-wire anemometers response at very high
frequencies. Fortunately, this problem originates from the electronic circuits, not
primarily from hot-wire measurement principle, and therefore can be resolved by
using advanced electronic components and improved anemometers design. In
contrast to hot-wire method, their main competitor laser-Doppler anemometer is
usually restricted to less than 30 kHz.
Size. Typical hot-wires are 5 m in diameter by 1.25 mm in length, although
wire elements as small as 0.25 m by 0.1 mm can be produced using commercially
available materials. Typical measuring volume of laser-Doppler anemometer is
much greater, around 75 m in diameter by 0.25 mm in length. Sensors of multiple
hot-wire probes have to be spatially separated and adequately oriented toward the
mean flow, in order to provide directional information, what increases the space
needed for their placing. Therefore, sensing volume of multiple wire probes has to
MULTIPLE HOT-WIRE PROBES 11
__________________________________________________________________________________________________________________________
be increased. For example, a typical dimensions of measuring volume of
quadruple hot-wire probe, with four 2.5 m tungsten sensors, is 0.5 mm x 0.5
mm x 0.25 mm. Although these dimensions still satisfy the spatial criteria of
Wyngaard 1969 (based on Kolmogorov 1941 micro-scale of the smallest structures
in turbulent flows) in the most cases of practical interest, in some rare occasions
the smaller probes are needed. Recent technological advances enabled fabrication
of the quadruple probe with 0.8 m platinum-alloy hot-wires, which sensing
volume is much smaller. However, smaller probe dimensions simultaneously
increase the sensor and probe fragility, demanding for the probe to be carefully
manipulated by the skilful and experienced persons.
Accuracy. Both laser and hot-wire systems, if properly applied, can provide
very accurate results. In the carefully controlled experiments, measurement errors
can be limited even to 0.1 0.2 %. However, in the practical applications, 1 %
accuracy is usually quite acceptable.
Signal-to-noise-ratio. Hot-wire system generates low-noise output signals. A
resolution of 1:10000 can be easily achieved. With laser-Doppler anemometer,
even 1:1000 is difficult to achieve by present technology (Fingerson and
Freymuth 1983). Laser-Doppler measurements sometimes may also suffer from
the bias problems.
Range of velocity measurements. Both laser-Doppler and hot-wire systems are
effective in a wide fluid velocity range, from very low up to high-speed and
supersonic flows. However, in contrast to laser system, compressibility effects may
influence the measurement results of hot-wire probe (see Durst, Launder and
Kjelstrom 1971 for example).
Simultaneous three-dimensional velocity measurements. From the operational
point of view, simultaneous three-dimensional measurements of fluid velocity
vectors by laser-Doppler systems are very complicated. Practically, a pair of laser
beams is needed for measurement of any fluid velocity component in a specified
direction. Simultaneous measurement of two components is possible by two
differently coloured pairs of laser beams. They can be initiated by the same emitter
unit and easily focused through the same lens to the same measuring location
(volume). However, the third velocity component normal to the plane of the first
two components needs application of an additional pair of laser beams. They have
to be emitted from a different (emitter) position, in the direction orthogonal to
directions of the first two pairs of beams, and focused to the same point in order to
provide a common measuring volume for all three fluid velocity components.
Unfortunately, this can be very complicated to achieve accurately and the most
researches simultaneously measure only the longitudinal and transversal
components, while lateral (which is usually of less importance) is measured
separately, i.e. non-simultaneously with the first two.
Temperature measurements and simultaneous measurements of fluid velocity
and temperature. An important advantage of the laser-Doppler technique originates
12 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
from its non-sensitivity to changes of fluid temperature. Positive consequences of
this laser property are explained in the next paragraph (velocity measurements in
the flows with variable temperature). However, its negative side is that laser-
Doppler systems do not provide a possibility of temperature measurement. In
contrast, it can be successfully measured by hot-wire anemometer operated in the
constant-current mode and equipped with hot-wire sensor usually designated as
cold-wire or resistance-wire.
Hot-wire technique also enables
simultaneous measurement of fluid
temperature and velocity fluctuations.
In such situations special hot-wire
probes are used (see fig. 2.4). Besides
the velocity sensors, usually operated in
the constant-temperature mode, they
contain an additional sensor operating
in the constant-current mode (cold-
wire). In the case of laser-Doppler
system, simultaneous measurement of
fluid temperature and velocity is
impossible. However, it seems that
recently developed technique for in-city
collaboration of cold-wire probe with
laser-Doppler anemometer (see Heist
and Castro 1998) overcomes this
problem. Unfortunately, this approach
may not be sometimes reliable, because
of the cold-wire contamination by
naturally occurring particles in the
water/liquid measurements, or by
seeding particles that have to be
introduced in the air/gas flows.
Velocity measurements in the flows
with variable temperature. In general,
hot-wire readings depend both on the
fluid velocity and temperature. If a classic hot-wire probe is applied for velocity
measurement it is necessary to maintain nearly constant fluid temperature. Even
small temperature changes, of only few degrees centigrade, may produce
significant measurement errors of turbulent velocity field, especially if the higher-
order statistics is analysed. If these changes are very slow, special probes with
elements for automatic temperature compensation of the wire readings are
commercially available (see the catalogues of AUSPEX 1994, DANTEC 1996 and
TSI 1997, etc.). The other possibility is to measure flow temperature by classic
thermometer and using this information digitally correct anemometer output


Fig. 2.4: Triple hot-wire probes,
designed for simultaneous
fluid velocity (two constant-
temperature sensors) and
temperature (a resistance-
wire) measurements.
Source: Fiedler 1978.
MULTIPLE HOT-WIRE PROBES 13
__________________________________________________________________________________________________________________________
voltages in the procedure of their interpretation. However, existing correction
formulas may not be sometimes reliable, especially in the case of the miniature
multiple hot-wire probes. This is the main reason why each of them should be
checked carefully for any specified probe configuration and experimental
conditions before direct application. Many researchers have extensively studied a
variety of different expressions (see Collis and Williams 1959, for example).
Among many of them, an expression of Kaneve and Oka 1973 seems to be very
suitable for practical applications. However, both automatic and digital
compensations of temperature variations are not appropriate for the flows with
simultaneous frequent fluctuations of the fluid velocity and temperature. In such
cases, special hot-wire probes, each of which contains an additional cold-wire for
temperature sensing, have to be used. These hot-wire probes specified for
simultaneous velocity and temperature measurements demand an extremely
complex and time-consuming calibration: besides the velocity (and directional for
the probes with two or more velocity measuring wires) calibration, a temperature
calibration is also needed. In contrast to hot-wire anemometers, laser-Doppler
readings of fluid velocity are independent on the fluid temperature, what represents
an important advantage of this technique in the situations when only flow velocity
measurements (without temperature) are needed.
Near-wall measurements. Both hot-wire and laser-Doppler measurements in the
vicinity of solid surfaces may be followed by problems. In the case of laser
anemometer, tracers may settle down on the wall, especially when they are
artificially introduced in the air/gas flow, as is usually the case. When hot-wire
anemometer is used close to the wall, two additional effects of wire cooling arise.
Besides the heat convection mechanism, which is assumed to be the main process
of the heat exchange with environmental flow and included in the wire response
equation and calibration coefficients, a radiation between the hot-wire and wall
may become significant. This phenomenon changes the wire calibration
characteristics, what can produce significant measurement errors. In addition,
presence of the sensor in the wall vicinity produces local acceleration, because of
the locally decreased free cross-section of the flow. A variety of appropriate
methods for digital corrections of the wall influence on hot-wire readings are
available (Oka and Kosti 1972, Bhatia, Durst and Jovanovi 1982, etc.).
However, all of them include the coefficients that have to be evaluated
experimentally, what represents an additional time-consuming process besides the
calibration of hot-wire probe.
Air/gas flows a seeding problem. It is likely that hot-wire systems are superior
for air/gas flow measurements, in comparison with laser-Doppler anemometers, if
the corresponding turbulence levels are below the limits that enable proper
application of the chosen hot-wire configuration. The main disadvantage of laser
systems is connected with seeding (see Matovi 1988 and Oka and Baki 1996 for
example). It originates basically from the velocity difference between a fluid
control volume and a tracer that represents its velocity. These differences are
14 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
always present to some extent. However, although many extensive experimental
studies have been performed in the best aerodynamic laboratories worldwide, an
accurate and confident general answer has not been found yet. This problem is
important in the air/gas flows, especially if higher-order statistics is analysed,
because of the high accurate velocity measurements needed. An additional seeding
problem expressed through the non-uniform concentration of tracers (especially in
the mixing and recirculation zones) also influences the measurement results.
However, it has been established that accuracy of the tracer representation of the
flow velocity increases at higher fluid viscosity. Therefore, it follows that
measurement accuracy of laser-Doppler increases in the liquid flows. In addition, it
also raises in the high-temperature air/gas flows, because the viscosity of gases
increases with raising their temperature.
Air/gas flows with strong variations of humidity. Hot-wire anemometer is an
indirect instrument, which utilises heat transfer information to deduce desired local
flow properties. Hence, it has to be calibrated before measurement of the real flow
field parameters. However, calibration unit is very often a different device from the
test rig where measurements are performed. Therefore, calibration and measuring
flow conditions may be different and consequently calibration has to be corrected
to account (among other influences) different air/gas humidity. In most of practical
applications influence of humidity variations on the measurement results obtained
by hot-wire anemometer is negligible. However, in some other situations, when
measurements are undertaken in dryers for example, changes of air humidity has to
be accounted in the procedure of hot-wire signals interpretation (Schubauer 1935,
Larsen and Busch 1980, Durst, Noppenberger, Still and Venzke 1996). In
contrast to hot-wire methods, laser-Doppler anemometer do not utilise any heat
transfer mechanism but directly measures tracers velocity, which represent a flow
velocity at specified location. Therefore, for adequately chosen tracing particles
and seeding technique, this instrument is nearly independent on the changes of
air/gas humidity in most of practical applications.
Liquid flows. In opposite to laser systems, applications of hot-film probes in the
single-phase liquid flows suffer from the several possible problems. Accumulation
of deposits on the sensor can be the major disadvantage of hot-wire technique. This
contamination problem can be minimised by introducing a filtration unit in the
flow. In addition, anemometer probes with larger sensors are less sensitive to
contamination mentioned above. A second important problem is temperature
contamination by liquid flow, because of the low overheat ratio used. Even small
changes of liquid temperature will result in substantial changes of the sensor
response characteristics, evaluated by probe calibration. As a direct consequence, a
frequent probe calibration has to be done. Following Bruun 1996, most of these
hot-wire problems can be resolved by taking special precautions. However, at
present, the laser-Doppler anemometer is superior and seems to be more popular
for liquid flow studies, especially in the flows where naturally occurring particles
can produce the required tracers seeding.
MULTIPLE HOT-WIRE PROBES 15
__________________________________________________________________________________________________________________________
Two-phase flows. Hot-film probes can be used in the flows containing one
continuous turbulent phase and distributed bubbles (liquid/gas or liquid/liquid
flows). When a bubble hits the film, an interaction will occur between the sensor
and the interface between the continuous phase and a bubble. If the bubble is larger
than a sensor, this interaction process can be used for signal analysis without
serious restriction on bubble concentration. Laser anemometers demand low bubble
concentration, due to the requirement of clear optical path. However, in the case of
two-phase flows with particles, advantages of laser-Doppler anemometers become
clear. Hot-wire systems are not applicable in such situations, primarily because of
the probe fragility. Furthermore, very small particles can sometimes produce
depositions on the films, which rapidly change their calibration characteristics.
High-temperature flows and combustion. Typical hot-wire probes are specified
for application in the cold or weakly heated flows. However, if iridium sensors are
used, velocity measurements become possible in the gases which temperature does
not exceed 800
0
C. For higher temperatures and flows with combustion, the only
presently available technique is laser-Doppler anemometry (see Matovi, Oka and
Durst 1994, for example).
Measurements in the atmosphere and distant measurements. In opposite to hot-
wire anemometers, contemporary laser-Doppler systems enable reliable fluid
velocity measurements at distances greater than 1000 m. Typical examples are the
atmospheric measurements of air and clouds velocity. However, although hot-wire
anemometry does not enable measurement of fluid properties at large distances, if
used properly, it also enables sophisticated measurements of the atmosphere air
parameters, primarily velocity field fluctuations (Larsen and Busch 1974,1976).
Instruments damage and contamination. Hot-wire probes are very delicate
instruments. However, if handled properly (by skilful persons), they can survive
within many months or even years. For low-speed airflow studies, the most
common case of probe breakage is improper handling by inexperienced operators.
Hot-wire breakage can be also caused by its burn-out. This situation arises if the
overheat ratio is set (accidentally) too high. Furthermore, this will happen in the
ageing process if the sensor works at high overheat ratio setting. In the high-speed
flows, probe can be damaged or destroyed due to impact of relatively fine particles.
Deposition of impurities, usually present in all flows, on the sensor may change its
calibration characteristics. Furthermore, they may reduce the anemometer
frequency response. In these situations probe must be re-calibrated or even cleaned
and then calibrated. Usually ultra-acoustic agitation is used to remove dust from
the sensor. However, all these problems can be avoided by placing the air-filter in
the experimental facility. Due to its non-contact measuring principle none of
mentioned problems influence laser-Doppler anemometer.
Industrial application. Fragility of hot-wire probes, unstable response
characteristics of their sensors and electronic circuits that demand frequent probe
calibration, probe sensitivity to dust presence in the flow and variations of fluid
16 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
temperature etc., make a hot-wire anemometer not very appropriate instrument for
some industrial applications. Of course, some rare exclusions of this rule exist. For
example, Durst, Jovanovi and Unger 1997 designed a special flow meter based
on hot-wire technique. In opposite to hot-wire anemometer, the most of these
problems are not present if the laser-Doppler anemometer is used, making this
technique more suitable for most of industrial purposes.
Signals continuity and their analysis. The output of hot-wire anemometer is a
continuous analogue signal (electric voltage). Consequently, both conventionally
and conditionally sampled signals can be analysed in the time and frequency
domain. In opposite this is sometimes a significant problem with laser-Doppler
anemometer, which signal is discrete.
Spatial information and measurements of special turbulence properties.
Application of two or more spatially separated probes enables measurement of
spatial and temporal correlations. If hot-wire rakes are applied in conjunction with
conditional sampling, the spatial and temporal development of large-scale structures
can be successfully identified within the turbulent flows. Special hot-wire
anemometer probes and adequate numerical algorithms for signal analysis enable
sophisticated evaluation of turbulent quantities such as intermittence, gradients of
turbulent velocity vector components, dissipation rate, vorticity components, etc. In
contrast, discrete character of laser-Doppler signals and the problem of accurate
positioning of two or more pairs of laser beams at close locations makes its
application for measurements of intermittence, vorticity etc., and for providing any
spatial information much more complex and sometimes practically impossible.
Therefore, in these cases, hot-wire anemometer seems to be more suitable.
Highly turbulent and recirculating flows. The applicability of a typical hot-wire
anemometer system with the last version of quintuple stationary probe operated
in the constant-temperature mode is limited to non-recirculation flows which
turbulence levels do not exceed ~66 %. However, typical hot-wire probes that
contain smaller number of sensors are commonly limited more severely: they can
be successfully applied for medium or even for low turbulence level flows only. It
is evident that any hot-wire probe possesses some critical limit of turbulence level,
which define its applicability. Out of this safe area, two possible sources can
produce large measurement errors of turbulence quantities. The first source arises if
the single-wire or X-geometry probes are applied. These configurations demand
neglecting the two or only one (of three) fluid velocity components, what may
cause significant decreasing of measurement accuracy. The second type of error,
usually designated as rectification error, originates from the axial symmetry of hot-
wire sensor. Consequently, the wire element is not sensitive to a reversal of the
flow direction, which may easily occur in the flows with high turbulence levels.
Graphical illustration of this hot-wire property, originating from Willmarth
1985, is presented in fig. 2.5. He analysed the response of idealised hot-wire of
indefinite length. In such case, only the influence of fluid velocity vector component
U
N
, normal to the wire axis of symmetry, is relevant for its cooling. If the tips of
MULTIPLE HOT-WIRE PROBES 17
__________________________________________________________________________________________________________________________
velocity vectors are fixed in
the centre of the (wire)
measuring volume, all their
tails giving the same wire
cooling effect will lie on the
cylinder of infinite length (the
hot-wire coincides with the
cylinder axis of symmetry). It
follows that an infinite number
of velocity vectors with equal
hot-wire cooling effect (giving
the same anemometer output
voltage) exists. This property
restricts the applicability of
hot-wire probes. For any
number of sensors, signal
ambiguity occurs if the attack
angle of fluid velocity vector is
greater than some critical value
of acceptability.
Satisfactory compensation techniques for these errors are not available for
stationary probes at the current level of technology development. A flying hot-
wire method, analogue to the frequency shift in laser-Doppler systems, may be
used for resolving the rectification problem. Various techniques are reported,
providing circular hot-wire probe motion (Cantwell 1976, Wadcock 1978, Coles
and Wadcock 1979, Cantwell and Coles 1983, Walker and Maxey 1985, Sirivat
1989, Hussein and George 1989 and Hussein 1990). Perry 1982, Hussein 1990
and Panchapakesan and Lymley 1993 generated linear probe motion, while
Crouch and Saric 1986 reported flying hot-wire method utilising oscillatory
probe motion. Curvilinear bean-shaped motion mechanisms were used by
Thompson 1984,1987, Thompson and Whitelaw 1984, Jaju 1987, Al-Kayiem
1989, Al-Kayiem and Bruun 1991, Badran 1993. However, all methods demand
complicated traversing mechanisms. Furthermore, corresponding output signal is
not continuous (as it is also the case with the laser-Doppler systems). An additional
solution for preventing the occurrence of the rectification errors is the pulsed-
wire technique. The problem discussed in this paragraph is one of the most serious
disadvantages of hot-wire anemometers, in comparison to laser-Doppler systems
that are capable of measurement of high-turbulence and recirculating flows.
Flow disturbance. Although very small, hot-wire anemometer probe placed in
the fluid modifies the local flow field. However, for aerodynamically well-
designed probe, resulting measurement errors should be very small. Also, correctly
applied probe calibration procedure mainly incorporates the related flow
disturbances in the calibration constants. Still, in some situations even the best hot-
Hot-wire
Tips of fluid
velocity vectors
Tails of fluid
velocity vectors

Fig. 2.5: Fluid velocity vectors, giving the same
output signal of infinitely long wire
(that follows the cosine law), according
to Willmarth 1985.
18 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
wire probes may significantly modify any flow phenomenon sensitive to
disturbances of the flow, such as separation. However, in opposite to any type of
hot-wire probe applied, flow disturbance is negligible only if a laser-Doppler
anemometer with carefully introduced tracers is used.
Nozzle
Mixing zone
Still ambient air
U
0
U
U
0

U
V

V
(b)
Potential core U
0
= const. (a)



Fig. 2.6: A sketch of angular calibration of X-probe: (a) calibration facility; (b)
induced flow velocity vector in the X-probe coordinate system.
Calibration. Hot-wire anemometer is not an absolute instrument, like laser-
Doppler system. It has to be calibrated before measurement, in order to evaluate
the functional dependence of the anemometer output voltage on the fluid velocity.
Inducing fluid velocity fields of known direction and intensity (measured by some
absolute instrument such as Pitot dynamic probe with liquid micro-manometer)
provides this information. The simplest calibration procedure corresponds to the
normal single-wire probe. It assumes only simple velocity calibration - inducing
various intensities of fluid velocity vector without changing their direction.
Calibration of the probes for two-dimensional or three-dimensional flow fields
measurement is much more complex. Besides changing intensity of fluid velocity,
the velocity vectors of various directions have to be induced in order to provide the
information about the probe directional sensitivity. Simply rotation of hot-wire
probe (see fig. 2.6) in the vertical (pitch calibration) and horizontal plane (jaw
calibration) simulate various directions of instant fluid velocity vectors. Hot-wire
anemometer output voltages are registered for each induced velocity vector.
Resulting information pairs (velocity vector and corresponding anemometer
output voltages) are used for calibration curve fitting by the least square method,
splines, special functions of Bessel, Hermite, Chebyshev, etc. In general, hot-wire
probe calibration still introduces some (usually very small) error in the procedure
for anemometer signals interpretation. Furthermore, it is a time consuming process,
which fortunately can be enormously accelerated if the computer-controlled
anemometer circuits are used (see Jorgensen 1996).

2.6 HOT-WIRE ANEMOMETERS AND LASER-DOPPLER SYSTEMS:
COMPETITORS AND COMPLEMENTS
Operational characteristics of hot-wire anemometers and laser-Doppler systems,
explained and discussed in the section 2.4, clearly show that these instruments are
primarily complementary techniques. Although each method possesses own
MULTIPLE HOT-WIRE PROBES 19
__________________________________________________________________________________________________________________________
exclusive advantages and limitations, a significant number of flows where both
instruments can be successfully used exist.
The main hot-wire exclusions, where laser-Doppler systems are easily
applicable, are the high-temperature flows with fluid temperature over 800
0
C,
recirculating and highly turbulent flows and most of two-phase flows with solid
particles. However, it should be noted that recirculating and highly turbulent flows
could be measured by another two sophisticated methods, known as the flying
and pulsed hot-wire techniques. Detailed information of these methods can be
found in comprehensive monographs of Freimuth 1992 and Bruun 1995.
Although both hot-wire and laser-Doppler techniques are applicable for liquid
studies, laser is usually preferred by researchers because the seeding requirement is
satisfied by naturally present fine particles. In addition, problem of the probe
fouling in liquids sometimes makes hot-wire anemometry inappropriate in such
situations. Laser-Doppler anemometer is also superior for velocity measurements at
very large-distances (up to 1000 m and even more, in the atmosphere). However,
although represents a newer instrument with many advantages in comparison with
hot-wire method, disadvantages of laser-Doppler system (explained in the section
2.4), still make hot-wire anemometers to be more popular instrument for a variety
of air/gas flow measurements.
In general, it can be concluded that appearance of the laser-Doppler
anemometer has significantly enlarged the area of possible flow measurements.
However, in many cases, hot-wire anemometer is still more reliable, accurate and
cheaper instrument for real-time measurement of fluctuating characteristics of
turbulence, especially in the studies including vorticity, dissipation terms and
temperature analysis. These reasons represent the main motives for many
researchers in aerodynamic laboratories worldwide to pay special attention for
further development of hot-wire anemometers and their probes. Therefore, these
two sophisticated techniques are not only direct competitors, but the
complementary methods also.


Instability of an axisymmetric jet. Source: Van Duke 1982.










3 3. . C CO ON NS ST TA AN NT T T TE EM MP PE ER RA AT TU UR RE E H HO OT T- -W WI IR RE E A AN NE EM MO OM ME ET TE ER RS S: :
G GE EN NE ER RA AL L P PR RI IN NC CI IP PL LE ES S A AN ND D S SI IG GN NA AL L A AN NA AL LY YS SI IS S


H Ho ow we ev ve er r f fa ar r m mo od de er rn n s sc ci ie en nc ce e a an nd d t te ec ch hn ni ic cs s h ha av ve e f fa al ll le en n s sh ho or rt t o of f t th he ei ir r i in nh he er re en nt t
p po os ss si ib bi il li it ti ie es s, , t th he ey y h ha av ve e t ta au ug gh ht t m ma an nk ki in nd d a at t l le ea as st t o on ne e l le es ss so on n: : N No ot th hi in ng g i is s i im mp po os ss si ib bl le e. .
( (L Le ew wi is s M Mu um mf fo or rd d, , 1 18 89 95 5- -1 19 99 90 0) )


3.1 GENERAL HEAT-TRANSFER EQUATIONS
3.1.1 Hot-wire cooling mechanism
Operation of hot wire anemometer is mainly based on two mechanisms: hot-
wire heating by electrical current (the Joules effect) and convective cooling by
surrounding flow. Their physical connection can be established using the important
and well known hot-wire property: dependence of the sensors electrical resistance
on the temperature. The relation between temperature and electrical resistance can
be defined easily for a known wire material, enabling interpretation of fluid
velocity fluctuations in terms of fluctuations of voltage drop along the wire, which
can be indirectly measured by registration of anemometers output voltages. In the
first approximation, which is valid if the temperature difference T-T
0
is not too
large, hot-wire temperature and electrical resistance can be related by expression
( ) | |
0 0
T - T + R = R o 1 ; o = const., (3.1)
where R is the wire resistance at current temperature T, R
0
represents its resistance
at reference temperature T
0
and o denotes a temperature coefficient of electrical
resistance. For the most common wire materials, tungsten and platinum, o is
approximately 0.004 K
-1
in the temperature range 0 - 300
0
C.
Characteristic ratio of wire resistance R at temperature T and its resistance R
f
at
temperature of surrounding fluid T
f
is an important operational parameter of hot-
wire probe, which determinates the sensor operational temperature T. In the
experimental practice it is known as the hot-wire overheat ratio:
f
R
R
a = . (3.2)
MULTIPLE HOT-WIRE PROBES 21
__________________________________________________________________________________________________________________________
If the reference temperature T
0
is assumed equal to surrounding fluid temperature
T
f
, temperature difference T - T
f
can be directly related to overheat ratio a:
o
1
=
a
T T
f
. (3.3)
The higher wire temperature will intensify its cooling flux and therefore increase
its sensitivity. However, the wire temperature should be kept under some limit
imposed by the oxidation of material used for its manufacturing. For tungsten, this
temperature is around 300
0
C.
Besides overheat ratio in the form (3.2), it can be also expressed as a function of
fluid and wire temperature:
f
f
T
T
T T
a

= . (3.4)
In this case, for T
0
= T
f
,
T f
a T a + = o 1 (3.5)
The heat, generated in the wire by the Joules effect, is transferred to its
environment primarily by convection, but also by conduction through the wire
supports and radiation. For a sufficient wire aspect ratio (length to diameter
quotient L/D), estimated to be around 600, the heat conduction to the wire supports
can be completely neglected (Perry and Morrison 1971). Typical hot wires radiate
only around 10 % of the energy that would be radiated by the corresponding black
body. In the most common operational conditions, radiation flux reaches around
0.1 % of the convective heat cooling flux (Blackwelder 1981). It corresponds to
only 0.01 % of the heat generated in the wire by Joules effect (Comte-Bellot
1976) and, therefore, can be also neglected. The only exception of the conclusions
based on the previous analysis are the velocity measurements in rarefied gases,
where hot-wire convective heat cooling is usually very small. In such cases, as well
as in the vicinity of solid surfaces, heat radiation may become dominant in the
overall heat exchange flux.
Following the well known Newtons cooling law, convective heat flux from a
heated wire toward the surrounding flow can be expressed as
( )
f
T T S h F = , (3.6)
where h is the convective heat transfer coefficient, S wire surface that exchanges
heat with a fluid of temperature T
f
, while T is the wire operational temperature.
Besides the characteristic temperature difference between the heated wire and
surrounding flow, convective heat flux depends on the fluid properties, flow
regime and hot-wire dimensions. Corrsin 1963 proposed the following general
expression
22 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
|
|
.
|

\
|
,
c
c
, a ,
D
L
Kn, Gr, Ma, Pr, Re, Nu = Nu
v
p
T
, (3.7)
based on the non-dimensional terms derived by the theory of similarity:
f
D h
Nu

= (Nusselt number), (3.8)

D U
Re

=
0
(Reynolds number), (3.9)
f
p
c
Pr


= (Prandtl number), (3.10)
c
U
Ma
0
= (Mach number), (3.11)
( )
2
3 2

|
f
T T D g
Gr

= (Grashof number), (3.12)
D
Kn
c
= (Knudsen number). (3.13)
Formulas (3.7) (3.13) represent some kind of general mathematical model of the
hot-wire cooling flux, expressed by convective heat transfer coefficient h, wire
diameter D and fluid thermal conductivity
f
. It is an extremely complex function
of many parameters that characterise fluid properties, flow regime and hot-wire
sensor. Practically, cooling mechanism of a heated sensor can be described as a
multiple dimensional function of fluid density , dynamic viscosity , specific
heats c
p
and c
v
under constant pressure and volume respectively, coefficient of
volume expansion |, molecular free path c and sound velocity c. Fluid parameters
should be evaluated for the film temperature:
2
f
F
T T
T
+
= . (3.14)
In addition, heat flux exchange between the wire and surrounding fluid depends on
many other parameters. As can be seen from (3.7) (3.13), flow velocity U
0
, wire
length L, diameter D and overheat ratio a
T
, as well as the angle between fluid
velocity vector and normal to hot-wire axis (see fig. 3.3, section 3.3.1), also
influence this process. If a wire diameter is much larger than micro-scale q of
Kolmogorov that represents the smallest vortex structures in turbulent flows, hot-
wire surface roughness should be also taken into account.
Fortunately, general heat transfer expression (3.7) can be simplified in the most
of practical applications. For example, hot-wire probes are most commonly used in
MULTIPLE HOT-WIRE PROBES 23
__________________________________________________________________________________________________________________________
flows which velocity is sufficient to create a regime of forced hot-wire convective
cooling. In such cases, in contrast to low speed flows, natural convection becomes
irrelevant factor in the heat exchange mechanism. According to Collis and
Williams 1959 it can be neglected if
N /
r CR
G Re Re
1
> > , (3.15)
where N is about 3 for air. From their measurements, they estimated the value of
Re
CR
to be 0.02. This condition is satisfied for air speed over 20 cm/s and,
therefore, the natural convection (e.g. Grashof number) can be neglected in the
most problems of practical interest. Some researches, like Lekakis 1996, believe
that natural convection is negligible even at lower fluid velocities, of only 5 cm/s,
while the others found the value of Re
CR
up to 0.07.
Furthermore, the molecular free path in the air at standard conditions is around
0.06 m, giving the small Knudsen number that can be also neglected for most
practical situations. The effects of compressibility are negligible if Ma < 0.3
(Blackwelder 1981). The influence of the sensor geometry, i.e. aspect ratio L/D can
be minimised for a sufficiently long wire, and the mechanism of wire heat
exchange with the flow can be considered as two dimensional.
Finally, assuming the constant values of fluid specific heats at constant pressure
c
p
and volume c
v
, as well as the constant overheat ratio a
T
and angle , expression
(3.7) reduces to simple formula
( ) Pr Re, Nu = Nu , (3.16)
which describes the convective heat exchange under limitations listed in the above
text. At present, a variety of different empirical forms of expression (3.16) exist,
depending on the flow conditions.
Kramers 1946 analysed the results of heat-transfer experiments for heated
elements placed in the air, water and oil. Using the formula
50 0 33 0 26 0
57 0 42 0
. . .
Re Pr . Pr . Nu + = , (3.17)
he obtained satisfactory results in the following ranges of non-dimensional
numbers: 0.01 < Re < 10
4
, 0.71 < Pr < 525 and 2 < Nu < 20. Following Hinze
1959, this expression is also valid for many different fluids. However, it is most
appropriate and popular for the liquid flows.
According to Blackwelder 1981, analysis of the results obtained by different
authors within around 40 experiments showed that expression
| ( ) ( )
( )
| ( )
( ) Re , a Pr, q
T
Re , a Pr, p
T T
T T
a + 1 Re a Pr, C + a Pr, B = Nu (3.18)
is more accurate and cover a broad range of experimental data. Coefficients B, C, p
and q are the functions of Prandtl number and the wire overheat ratio. In addition,
coefficient p, as well as q, is dependent on the Reynolds number too. Fortunately
24 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
the last influence is fairly weak and its neglecting is usually acceptable in the
almost all applications.
Having in mind that overheat ratio a
T
is constant for a given measuring
conditions, that Pr number is in the range of 0.7 < P
r
< 1 for the most gases, and
that its value is constant for a given fluid temperature, expression (3.18) can be
rewritten as
p
Re N M Nu + = . (3.19)
Direct comparison of expression (3.17) and (3.19) gives the following formulas for
the constants M and N:
26 0
42 0
.
Pr . M = , (3.20)
33 0
57 0
.
Pr . N = , (3.21)
which can be used for qualitative and preliminary analysis only. However, in the
practical applications, it is strongly recommended to determine the exact values of
the constants M and N experimentally (in the calibration procedure) for each given
hot-wire sensor of the probe before the measurement of the flow properties. This
way, the influences of Prandtl number, overheat ratio and even the influence of
Reynolds number can be accurately taken into account.
The relation (3.19), which can be written in general form of Nu = f(Re), is one
of the basic operational expressions of hot wire anemometry. Many authors
worldwide have examined its properties. Among the first was King 1914, who
found the value for p to be 0.5. It turned out, later on, that a value of 0.45 is much
more accurate. Following the rich experience based on application of modern
multiple hot-wire probes, it seems that value of the exponent p varies in the range
of 0.35 < p < 0.5. It should be determined by calibration, together with constants M
and N, for each sensor and just before each measurement, to prevent temperature
drift of electronic circuits employed. For high accurate measurements, the relation
p = f(Re) should be also determined in the calibration procedure.

3.1.2 Heat generation the Joules effect
The heat is induced in the wires by Joules effect, in accordance to formula:
R I = E I = P
2
, (3.22)
where I is electrical current, E voltage drop along the sensor, and R is the wire
operational resistance.
Although it seems to be very simple, this expression can be very complex
because of the time dependency of generated heat flux P, electrical current I and
voltage drop E. Following (3.22), generated heat P can be expressed implicitly as
MULTIPLE HOT-WIRE PROBES 25
__________________________________________________________________________________________________________________________
( ) R , I P P = (3.23)
or, using (3.1), by the formula:
( ) T , I P P = . (3.24)

3.1.3 Heat balance equation hot-wires cooling law
The internal energy U, accumulated in the wire can be expressed as
T c m U = , (3.25)
where m is the wire mass, c specific heat, while T designates wire temperature.
The time variation of this energy is equal to difference of generated energy P
and energy transferred to the surrounding, mainly by convective heat flux F,
F - P =
dt
dU
. (3.26)
Assuming the constant value for specific heat c and moderate range of
temperature changes, the time variation of inner energy will be
|
.
|

\
|
=
dt
dT
m c
dt
dU
, (3.27)
what gives
F - P =
dt
dT
m c |
.
|

\
|
, (3.28)
and, after substituting (3.1),
F P
dt
dR
R
m c
=

o
0
. (3.29)
Expression (3.6), that governs the convective heat flux F, can be rewritten with
the aid of (3.1), (3.8) - (3.13) and (3.19) as
(
(

|
.
|

\
|
+

=
p
N
f f
D U
N M
R
R R
D
S
F
v o

0
. (3.30)
After substituting terms P (3.22) and F (3.30) in (3.29), one of the basic
equations that describes operation of hot-wire anemometer arises
( ) ( )
dt
dR
R
m c
U Y X R R I R
p
N
f

= +
o
0
2
, (3.31)
26 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
where:
2 0
0
42 0
.
r
f
P .
R D
S
X


=
o

(3.32)
and
33 0
0
57 0
.
r
p
f
P .
D
R D
S
Y |
.
|

\
|



=
v o

. (3.33)
Although some assumptions and approximations has been made in order to get
this expression, it is still assumed to be accurate enough for analysis of hot-wire
anemometer behaviour. In the case of constant wire temperature T, its resistance R
will be also constant. Therefore, dR/dt = 0, and expression (3.31) reduces to
( ) ( )
p
N
f
U Y X R R I R + =
2
. (3.34)
It can be also written in term of voltage drop
( ) ( )
p
N
f
U Y X R R R E + =
2
, (3.35)
or in the general form
p
N
BU A E + =
2
, (3.36)
as King 1914 obtained and widely known as Kings cooling law.
However, presented analysis is based on the assumption that the fluid velocity
vector is perpendicular to the wire longitudinal axis, or that only the component U
N

of fluid velocity vector
0
U contributes to wire cooling. In the general case, when
the orientation of velocity vector is optional, the influence of all three velocity
components should be taken into account by replacing the normal cooling velocity
by so-called effective cooling velocity U
e
giving:
p
e
U B A E + =
2
. (3.37)
The effective cooling velocity U
e
depends on all three components of the fluid
velocity vector (see section 3.8). The constants A and B are functions of constants
X (3.32) and Y (3.33), as well as of the wire resistance R at wire operational
temperature and at fluid temperature R
f
. In the experimental practice, they are
always determined experimentally, by varying fluid velocity U with simultaneous
measurement of voltage drop E.
Besides the expression (3.37), the relation E = f(U) can be expressed in different
forms proposed by many authors. Among the available approaches, the most
convenient methods includes application of polynomial fitting
MULTIPLE HOT-WIRE PROBES 27
__________________________________________________________________________________________________________________________
1
1

=
=

i
n
i
i e
E A U , (3.38)
splines, ortho-normalised functions, etc.

3.2 BASIC PRINCIPLES OF CONSTANT TEMPERATURE ANEMOMETERS
3.2.1 Development of constant temperature anemometers
Constant temperature (CTA) and constant current (CCA) anemometers
represent special instruments for sophisticated real-time measurements of flow
properties, which define two main branches of hot-wire anemometry as a
measurement technique. The basic operational principles and design of the CTA
electrical circuits, which are widely recognised to be superior for measurements of
velocity fluctuations in comparison with CCA, are given in the present chapter.
CTA systems are used in a very broad range of fundamental and applied
investigations of gas and liquid flow phenomena as shortly reviewed by Lekakis
1996 and Bruun 1996. Besides the general hot-wire anemometer characteristics
(small sensing elements, high sensitivity and insignificant interference of hot-wire
probes with the original flow, etc.), reviewed in the 2
-nd
chapter, CTA mode
possesses an additional important property that makes it very convenient for
different type of measurements. Sensors of CTA are maintained at constant
operational temperature, and thereby at constant electrical resistance, what results
in automatically adjusted hot-wire thermal inertia at variable flow conditions.
Therefore, this operational mode is characterised by very fast response toward the
fluctuations of flow properties. Nowadays, CTA is a well-established technique,
whose advantages have been proven by extensive measurements performed by
scientists and engineers in laboratories and industry around the world.
Development of modern CTAs, during nearly a whole century, can be roughly
represented by three main phases, depending on the electronic components used.
The initial period was characterised by the electronic circuits based on the
vacuum electronic tubes. According to Comte-Bellot 1976, the first to utilise the
advantages of such CTA system was Zeigler 1934. Using the same technology,
Weske 1943 and Ossofsky 1948 also designed and tested CTA circuits in order to
overcome the inadequate temporal response of the CCA sensors.
The second phase started after the semiconductors appeared in the early 1950s,
enabling further improvement of CTA electronic circuits up to the level that was
fairly close to the present quality. In the meantime, various CTA systems have
been reported in the literature (see Freimuth 1967, Wyngaard and Lumley 1967,
Perry and Morrison 1971, Weidmann and Browand 1975, Perry 1982, Miller,
Shah and Antonia 1987, etc.). However, the contemporary reliable instruments
became available in the middle 1960s (Bruun 1995), after the analogue low-noise
integrated circuits were introduced in the operational practice.
28 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
The final design phase of CTA is based on integrated digital micro-technology
that became commercially available in the early 1980s. However, digital
microchips were built in the anemometer circuits recently, in the 1990s, giving the
computer-controlled CTA. Besides other digital micro-components, these systems
are equipped with central processing units, integrated memory components and
communication interface specified for direct communication with personal
microcomputer (PC). Introduction of digital technology, as well as the specialised
computer software packages, enables automatic set-up of the operational
parameters of electronic circuits and calibration of arbitrary shaped hot-wire probes
(Steinhilber and Wagner 1994, Jorgensen 1996).
Furthermore, in the middle 1980s appeared first commercial models of
temperature compensated, high-resolution analogue-to-digital converter (ADC)
boards for PCs, making possible fast and reliable simultaneous sampling of
multiple-channel output voltages of CTA. As a result, digital data sampling and
data reduction became the most common approach to manage the output of CTA
systems, either they are computer-controlled or not.
Therefore, both the computer-controlled and classic CTA systems support data
acquisition by ADC board placed in PC. The accuracy of the anemometer analogue
signal (e.g. output voltage), which represents the instantaneous velocity vector in
the flow is mainly dependent on the quality of anemometer circuit used. In
contrast, the resolution of ADC board and sampling time and frequency determine
the accuracy of sampled data time-series. In such a way, the total accuracy is a
hyper-dimensional function of CTA operational characteristics and ADC related to
the actual flow (Jorgensen 1996).

3.2.2 Basic elements of CTA feedback circuit
It is evident that recently developed computer-controlled anemometers are
going to become in the nearest future the only CTA instruments used, rejecting the
classic systems from the experimental practice. However, either digitally supported
or not, all CTA systems utilise the same measurement principle and include
adequate basic analogue measurement circuits. Therefore, analogue electronic
components still represent the most important and unavoidable elements of all
CTA units. They have to be carefully analysed to understand the operation of CTA,
as it is done in this chapter. In contrast, the newest digital electronic components,
specified for computer control and automatic set-up of the anemometer, are not
discussed here. Such decision was made having in mind that they have only a
comfortable role that is useful, but not crucial, in the flow measurements.
An elementary electric scheme of a typical CTA feedback circuit, which still
enables clear explanation of its basic functional principles and operation, is shown
in fig. 3.1. However, all main components of CTA circuit, a Wheatstone bridge, an
integrated amplifier containing one or two stages of voltage gain and a buster-stage
MULTIPLE HOT-WIRE PROBES 29
__________________________________________________________________________________________________________________________
unity-gain amplifier
of electric current
are presented.
Sensor is placed in
one of the bridge
legs. To maintain its
resistance at the
specified value R,
which corresponds
to wire operational
temperature T, an
electrical current of
5-50 mA is usually
required. To supply
that current to the
sensor, resistances
R
2
and R
V
in the left
side of the bridge
have to be much greater then the resistances R
1
and R in the right branch. The ratio
R
2
/R
1
, known as the bridge ratio, is usually around 10, but its value can
sometimes reach even 50.
The bridge can be balanced (E
1
= E
2
) for a given value of sensor resistance R
(depending on its temperature T) by adjustable potentiometer R
V
. When the sensor
is exposed to the surrounding flow, its temperature decreases as well as the
resistance (according to expression (3.1)). As a direct consequence, the bridge
balance is disturbed (E
1
= E
2
). The voltage difference (E
1
- E
2
) is amplified by
differential operational amplifier, which creates feedback current I
0
on the top of
the bridge that heats-up the sensor and increases its resistance till the bridge
balance is achieved again. Taking into account very fast amplifiers response, the
resistance R is restored nearly instantly and the bridge is practically permanently
balanced. In the most common practical applications, this means that the sensor
resistance R and its temperature T are almost constant. For a specified anemometer
circuit, a desired value of sensor resistance R that defines its operational
temperature T may be set by careful choosing the adequate resistance value of
resistor R
P
(as it is explained in section 3.2.3).
In fact, the voltages E
1
and E
2
are never identical. A small voltage difference AE
= E
1
- E
2
is needed to create an adequate input of differential operational amplifier
that will provide a feedback current I
0
. Gains of modern operational amplifiers are
very high (typically around 1000). Therefore, they are capable of generating a
feedback current I
0
of sufficient intensity to heat-up the sensor up to the desired
temperature T even for a very small voltage difference AE. The intensity of I
0

depends primarily on the fluid velocity, which defines cooling flux of hot-wire at
specified temperature T.
R
R
V

E
2

R
2
R
1

E
1

R
P

E
SW
R
K
P
3

P
2

P
1

T
I
K

E
K
E
QI
E
0
I
0
+ +
- -
E
K
E
0
0
Diff.
amplifier
R
T
I
1
I
2

R
SW

Fig. 3.1: An electrical scheme of a typical electronic
feedback circuit of CTA.
30 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
Main elements of CTA, which purposes are explained in the three preceding
paragraphs, provide only the basic hot-wire measurement functions. However, the
electronic feedback circuit has to be accommodated for different measurement
conditions. An additional set-up for the various overheat ratios, probe leading
resistance, as well as for the leading cable resistance, capacitance and inductance,
is also needed. In order to satisfy these requirements, the most common CTA
feedback circuits ordinary contain adequate additional electronic components.
Furthermore, auxiliary systems not directly involved in the measurement of fluid
velocity, which protect sensor against too-high intensity of the heating current, as
well as the square-wave signal generators, low-pass and high-pass electronic filters,
etc., also represent the most commonly built-in circuits in the CTA units.

3.2.3 Basic steps of CTA set-up procedure
To balance the Wheatstone bridge, a current I
K
of low intensity is brought on its
top T through the closed switch P
3
, while the switch P
2
is opened and P
1
is
switched on. If the bridge is unbalanced, (E
1
= E
2
), the output voltage E
0
will
change its value. This change can be annulled and E
0
restored to initial level by
simple adjusting the variable potentiometer R
V
, what will result in E
1
= E
2
, e.g. the
bridge is balanced. During the bridge balancing procedure, electric current through
the sensor is very small because the resistance R
K
is high and intensity of I
K
is low.
Consequently, the wire temperature T is practically identical to the temperature T
f

of the surrounding fluid, giving R = R
f
. Therefore, for a balanced bridge, the
following condition has to be satisfied:
P
P
V f
R R
R R
R R R
+

=
1
1
2
. (3.39)
After the Wheatstone bridge is initially balanced for the wire at fluid
temperature T
f
, the wire has to be heated up to the operational level T. This can be
achieved by turning off the switches P
1
and P
3
, while P
2
has to be switched on.
Turning the P
1
off unbalances the bridge, due to the change of total resistance of
the right-side branch. The voltage difference (E
1
- E
2
) arises again, creating the
feedback current I
0
that is transferred to the bridge top T through the closed
switch P
2
. Consequently, the sensor temperature T (e.g. resistance R) starts
increasing till the bridge balance is not achieved again, as it is described in the
previous paragraph. For a balanced bridge with a resistance R
P
off, it must be:
2 1
R R R R
V
= . (3.40)
Using (3.39) and (3.40), the sensor over-heat ratio a (see (3.2)) can be expressed
as function of the bridge elements resistance
P
P
f
R
R R
R
R
a
+
= =
1
, (3.41)
or as a function of sensor operational temperature T
MULTIPLE HOT-WIRE PROBES 31
__________________________________________________________________________________________________________________________
( )
f
f f
T T
R
R
R
R
a + = =
0
1 o
. (3.42)
Choosing the adequate resistance of resistor R
P
, an optional value of sensor
resistance R (e.g. temperature T) can be established. Contemporary CTA systems
are equipped with adequate sets of built-in resistive elements of carefully chosen
resistance values, which provide a possibility to set-up nearly any desired hot-wire
operational temperature of experimental interest. From the practical point of view,
overheat ratio set-up may be done via the computer code (computer controlled
CTA), or by switching (on/off) the set of switches available on the anemometer
motherboard (classic anemometer). However, some hand-made CTA systems are
designed for a specified overheats ratio, hot-wire probe and its sensors. In such
situation, a single resistor of defined (constant) resistance R
P
is available in the
CTA system, providing a desired value of overheat ratio a that can not be changed
without changing the resistor itself. As it is explained in section 3.1, the wires
sensitivity increases proportionally by increasing its temperature T. However the
sensor may oxidise or even burn out at too high operational temperature. Serious
problems are sometimes caused by the thermal interference between different wires
of multiple hot-wire probes, even at average or low overheat ratios. The wall
influence also becomes more critical for higher sensor temperatures. Therefore, in
accordance to hot-wire probe configuration, flow properties, applied sensor
material and desired measurement accuracy, usual values of overheat ratio should
be in the range between 1.1 and 2.0.
After the Wheatstone bridge is balanced at sensor temperature T, it is
preliminary prepared for flow measurements. However, in order to achieve the best
possible measurement resolution and the highest signal-to-noise ratio, dynamic
response of the whole CTA system has to be checked and optimised in the last step
of the set-up procedure. In general, the response function is dependent on the
functional characteristics of the sensor, Wheatstone bridge, built-in amplifier and
electronic filter, if available.
Simple theoretical analysis of Corsin 1963 showed that hot-wire dynamic
response is not critical in the practical applications. Extensive studies confirmed
that the critical component in the CTA system is differential amplifier. Therefore, it
is necessary to test its operational characteristics and set-up the gain and the
frequency range. This can be achieved by providing accurately controlled
fluctuating input, of known constant frequency and simple shape, for the bridge.
The most commonly the square-wave test signal is used, usually provided by the
generator that is built in the CTA system. The voltage level of this signal is
accommodated by the resistor R
SW
, and then forwarded to the amplifier. In such a
way, by simulating the variable input, it is possible to optimally set-up the
amplifiers gain and frequency range of applicability by simple watching its
response on the electronic oscilloscope. During this phase of CTA set-up
procedure, the sensor should be placed in the uniform flow of the maximal
expected flow velocity.
32 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
3.2.4 The operating point of a feedback circuit
Assuming that the set-up procedure is properly performed, a CTA system with
the sensor heated at desired temperature T is ready for measurement. The fluid
stream will cool the sensor, creating a difference E
1
- E
2
that will be nearly
instantaneously compensated by feedback current I
0
. The voltages E
1
and E
2
are
maintained very close to each other. In contrast, the electric current I
0
and voltage
E
0
vary, depending at first place on the fluid stream velocity. The sensor resistance
R and temperature T are practically constant, except for very high frequencies of
the fluid velocity fluctuations. The frequency limit mainly depends on the sensor
dimensions at first place, overheat ratio and amplifier characteristics.
Using Ohms law, for the amplifier gain equal k, the following relations
between parameters of feedback CTA circuit, shown in fig. 3.1, can be obtained:
( )
QI
E E E k E + =
1 2 0
, (3.43)
T
R
E
I
0
0
= ;
( ) ( )
V
V
T
R R R R
R R R R
R
+ + +
+ +
=
2 1
2 1
, (3.44)
V
V
R R R R
R R
I
R R
E
I
+ + +
+
=
+
=
2 1
2
0
1
0
1
, (3.45)
V V
R R R R
R R
I
R R
E
I
+ + +
+
=
+
=
2 1
1
0
2
0
2
, (3.46)
0
2 1
0
1 2
R
R R R R
I
E E
V

+ + +
= ;
V
R R R R R =
1 2
0
(3.47)
and
|
|
.
|

\
|

+ + +
=
0
2 1
0
0
R
R R R R
I
E k E
V
QI
. (3.48)
Substitution of (3.48) in (3.44) (I
0
= E
0
/R
T
), after some rearranging gives
( )
(
(

+ + +

+

=
v T
T
QI
R R R R R
R k
R
k E
I
2 1
0
0
1
. (3.49)
Using the relation I
1
= f(I
0
) (3.45), the expression (3.49) can be transformed to
( )
( ) ( )
0
2 1
2
1
R k R R R R
R R E k
I
V
V QI
+ + +
+
= . (3.50)
The electric current I
1
passes through the sensor, which resistance R is nearly
constant in the CTA system. In that case, it follows from (3.34) (developed for the
wire of constant resistance) that this current may be also defined as
MULTIPLE HOT-WIRE PROBES 33
__________________________________________________________________________________________________________________________
( ) ( )
p
N f
U Y X R R I R + =
2
1
, (3.51)
or, after simple rearranging
( ) ( ) R / U Y X R R I
p
N
f
=
1
. (3.52)
Formula (3.50),
together with (3.52),
represents a simplified
mathematical model of
CTA system. They
formally describe
operation of a feedback
circuit. Its operating
point is located in the
intersection point of the
curves that graphically
describe these two
functions, as illustrated
in fig. 3.2 for different
values of fluid velocity
U
N
and amplifier offset
voltage E
QI
. The
operating point moves
in the referent
coordinate system (I
1
,
R) when the fluid velocity changes its intensity for a given value of E
QI
. For
example, when a velocity changes from U
N1
to U
N3
, at constant offset voltage E
QI
1

= const., the operation point moves from O
P
1
to O
P
3
. As can be seen in fig. 3.2, the
solid-line curves defining E
QI
=const., are very steep. Therefore, during this change
of the operating regime (from O
P
1
to O
P
3
), hot-wire resistance remains practically
constant (R
1
~ R
3
), while the sensor heating current changes intensity from I
1
1
to I
1
3
.
This result is in the full agreement with the basic operational principle of CTA,
which assumes the constant sensor temperature T (e.g. resistance R).
Previous analysis is focused mainly to the sensor operational resistance R and
voltage drop along its length E. Fortunately, it can be easily shown that intensity of
the sensor heating current I
1
is proportional to the feedback current intensity I
0
, as
well as the voltage drop E along the hot-wire is proportional to the CTA output
voltage E
0
that can be easily recorded. Therefore, the form of expression (3.35),
(3.36), (3.37) and (3.38) will be the same for E and E
0
. Obviously, the coefficients
in these expressions that should be determined by calibration are different.
Therefore, an output voltage E
0
of CTA system represents the information, which is
sufficient for evaluation of the fluid velocity. Following chapter 3.1.3, it is assumed
that velocity vector is orthogonal to the wire axis, or that only normal component U
N

I
1

R
E
QI
=const. exp. (3.50)
U
N
=const.exp.(3.52)
U
N
=0
U
N
3
U
N
2
U
N
1
E
Q
I
=
0

E
QI
1
E
QI
2
E
QI
3
U
N
1
>U
N
2
>U
N
3

E
QI
1
>E
QI
2
>E
QI
3

O
P
1
O
P
3
A
R
13
A
I
1
1
3

E
QI
<0

Fig. 3.2: A sketch defining the operating point of CTA
feedback circuit. Adapted from: Perry 1982.
34 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
of vector
0
U
contributes to the wire cooling. Unfortunately, for high-precision
measurements, the influence of other velocity components on hot-wire cooling flux
should be taken into account. This can be achieved by introducing the effective
cooling velocity U
e
, instead of U
N
, which represents a virtual quantity that relates
hot-wire cooling flux with all three fluid velocity vector components.

3.3 EFFECTIVE COOLING VELOCITY
3.3.1 The reference coordinate system for fluid velocity vector
In the analysis of hot-wire cooling mechanism, that directly governs its
response, it is suitable to represent an arbitrary oriented vector of fluid velocity
0
U
by its components in the local coordinate system (N, B, T) of the sensor:
0 0 0
0
t U b U n U U
t b n
+ + = . (3.53)
As usual, the orientation of normal N, binormal B, and tangential axis T of the
coordinate system is defined by mutually orthogonal unity vectors
0 0 0
t , b , n . As
shown in fig. 3.3, normal component U
n
is orthogonal to the sensor axis and lays
in the plane of its prongs, binormal component U
b
is orthogonal both to the hot-
wire and its supports, and tangential component U
t
is oriented along the sensor.
It is known that hot-wire cooling flux is highly sensitive to the changes of total
normal velocity component U
N
,
0 0
b U n U U
b n
N + = (
2 2 2
b n N
U U U + =
), (3.54)
obviously orthogonal to the sensor axis. Simultaneously, the heated wire is fairly
insensitive to changes of
0
t U U
t
t
= . However, tangential component U
t
still
influences the sensor cooling and has to be taken into account, if accurate
measurement of fluid velocity
vector is needed. This is the
reason why the total normal
component U
N
of fluid velocity
should be replaced by the so-
called effective cooling
velocity U
e
in the basic
equations (as it is stated in the
sections 3.1.3 and 3.2.4). In
such case, an appropriate form
of U
e
dependence of on the
fluid velocity components is
needed.
Most of modern hot-wire
probes contain differently
T(z)

U
0s


U
n

B(y)

U
t
N(x)
o
= 0
U
0


=
Z
(U
0
, U
N
)

u
0

=
Z
(U
0s
, U
n
)

0

=
Z
(U
0
, U
0s
)
u
0

0


U
N


U
b


Fig. 3.3: Components (U
n
, U
b
, U
t
) of an arbitrary
oriented fluid velocity vector
0
U in the
local Cartesian coordinate system (N,
B, T) of the normal single-wire probe.
MULTIPLE HOT-WIRE PROBES 35
__________________________________________________________________________________________________________________________
oriented multiple sensors, what sometimes make the local wire coordinate system
attached to the sensor and expression (3.53) inappropriate. Problem is frequently
resolved by describing a fluid velocity vector in a coordinate system ( k , j , i ),
attached to the probe axis
k W j V i U U + + =
0
. (3.55)
This system of Cartesian coordinates consists of three mutually orthogonal axes:
longitudinal x aligned with the probe axis, transversal y and lateral z axis.
Very often, the y axis is vertical, while the rest two lay in the horizontal plane.
Although systems (N, B, T) and (x, y, z) are differently oriented, in general, a
special case exists when these two Cartesian coordinate systems are identical. Such
situation occurs for the normal single-wire probe (o = 0
0
, as shown in fig. 3.3),
when its hot-wire is oriented in the z-direction, giving:
U U
n
= , V U
b
= , W U
t
= . (3.56)
However, for an arbitrary oriented sensor, adequate relations are more complex,
but can be also simply formulated using basic rules of trigonometry. A typical
probes coordinate system (Oxyz) for a slanted-wire probe in the horizontal plane,
together with a sensor local coordinate system (N, B, T) is sketched in fig. 3.4. In
order to achieve the highest possible probe sensitivity to changes of velocity vector
direction and the largest acceptability cone (explained in the 4
-th
chapter), hot-
wire probe axis is most commonly aligned with the mean flow direction. In such
situations, the coordinate system defined by unity vectors ( k , j , i ) can be also
treated as to be attached to the flow configuration.
T

U
n
B(y)

U
t

N
o
x
z
u
0


U
b
=V

0


U
U
0s


W
z
x

W
o
N
T
o

U
o
AXONOMETRIC VIEW UPPER VIEW

U
N

=
Z
(U
0
, U
N
)
u0
=
Z
(U
0s
, U
n
)
0
=
Z
(U
0
, U
0s
)
U
0


Fig. 3.4: Relations between the sensor Cartesian coordinate system (N, B, T) and
the system (x, y, z) attached to the slanted single-wire probe axis.
36 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
3.3.2 Effective cooling velocity of a heated wire
Idealised analytical analysis
of the cooling mechanism of
uniformly heated indefinitely
long cylinder, placed in the
indefinite flow field of
uniform velocity and
temperature, showed that its
effective cooling velocity U
e
is
equal to the fluid velocity
component U
N
normal to the
wire axis
cos U U U
N e
= =
0
. (3.57)
This expression, very
popular in the past and widely
known as the cosine law of
hot-wire cooling, relates the
effective cooling velocity U
e

and fluid velocity vector
0
U
described by its intensity U
0

and attack angle toward the
normal to the sensor axis (see
fig. 3.3). Up-to-date, it is the
simplest mathematical model
that describes cooling
mechanism of a heated wire.
Although formulated for an
ideal sensor, it is recognised
worldwide that cosine law
holds up for a uniformly
heated wire of sufficient aspect
ratio (L/D) and limited value
of sensor overheat ratio a
(Corrsin 1963, Comte-Bellot
1976, Blackwelder 1981, etc.).
However, having in mind that
turbulent velocity vector temporally changes its direction, a logical question about
the angular applicability range of cosine law arises. Many researches have
extensively tested its validity for a variety of existing hot-wire probe
configurations. Most of them reported that, under normal operating conditions,
formula (3.57) is valid only for the limited variations of velocity angle .
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Cos
Wire 1
Wire 2 U
e
/
U

[
-
]

+o
e
[deg]
-60 -40 -20 0 20 40 60 80 -80

-2.0
-1.0
0.0
1.0
2.0
-30 -20 -10 0 10 20 30
U V
c
(
U
,
V
)

[
%
]

[deg]

Fig. 3.5: Angular applicabillity range of the
cosine law, accompanied with X
probe DANTEC55P61 (pitch angle
defined following fig. 2.6):
(a) deviations of the calibration data
from the cosine law;
(b) signal-interpretation errors of fluid
velocity components, induced by
angular calibration at different
pitchangles u.
Source: Petrovi et al. 1998.
MULTIPLE HOT-WIRE PROBES 37
__________________________________________________________________________________________________________________________
Fortunately, this limited range is not always narrow, especially for well-designed
probes.
A typical illustration of the last statement can be found in recent paper of
Petrovi, Schenck and Durst 1998. In order to verify the technique of Bradshaw
1971, applied for calibration of the DANTEC X probe 55P61, they found that
cosine law holds-up for pitch angles in the range of = 25
0
, (for a velocity
vectors lying in the plane parallel to the wires and their prongs). As can be seen in
fig. 3.5, measurement errors of longitudinal U and transversal V velocity
components are smaller than 1 % in this range of angles , formed by the probe
axis and induced flow direction. Although the high overheat ratio of 1.6 was used,
giving the operational wire temperature of T ~ 250
0
, high accuracy was provided
by large aspect ratio of the wires: L/D = 400 (D = 2.5 m, L = 1 mm). Therefore,
logical conclusion is that carefully calibrated and well-designed probe can follow
tightly the cosine law in a fairly wide angular range of fluid velocity vectors.
Unfortunately, in the critical situations that most frequently occur at low fluid
velocity and especially if sub-miniature probes that contain multiple hot-wires of
small aspect ratio are used, cosine law is not appropriate anymore. It is not
accurate enough and the more sophisticated (e.g. more general) model of hot-wire
cooling mechanism is needed.
For a sensor of finite length,
besides the velocity orientation
parameters
0
and u
0
(figs. 3.3-3.4),
several characteristic scales appear
in the cooling law expression. The
most important are the sensor
length L and diameter D. However,
the other dimensions, like sleeves
diameter D
1
, prongs diameter D
2
,
prongs separation S and length S
1
,
as well as the probe body (stem)
diameter D
3
, all sketched in fig. 3.6,
are also relevant parameters that
influence heat exchange between
the sensor and surrounding flow. According to Corrsin 1963, a characteristic
length L
C
over which the supports affects the temperature distribution along the
sensor should be taken into account too. Formally, these parameters can be related
to effective cooling velocity by general expression proposed by Blackwelder 1981:
( )
1 3 2 1 0 0 0
S , S , D , D , D , D , L , L , , f U / U
C e
u = . (3.58)
However, it is impossible to define the general form of expression (3.58),
because of its complexity: besides the probe configuration, a flow regime and
geometry, as well as many other experimental influences, have to be taken into

D
1

L

D
S

D
3

S
1

D
2


Fig. 3.6: Typical design parameters of a
normal single-wire probe.
38 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
account. The first versions of hot wire probes were equipped with a single sensor,
as sketched in fig. 3.3, and capable of measuring only one velocity component U
N
,
perpendicular to the wire axes. The next types with two sensors were similar to the
letters X or V (see fig. 4.2), looking from a side. These configurations are
capable of simultaneous measurement of two velocity components in the plane
defined by the sensors and their supports. Unfortunately, the third velocity
component, perpendicular to measuring plane, still has to be neglected.
Rather large number of semi-empirical expressions for U
e
was proposed by
different researches for these types of probes. The most known, which formally
describe hot-wire cooling mechanism, are presented in tab. 3.1. The subscript "e"
denotes effective values of the corresponding relevant parameters discussed in the
next chapters, while U
0s
is the velocity component in the plane defined by the
sensor and its supports, as it is sketched in figs. 3.3, 3.4 and 3.7. Unknown
parameters k, e, b and o
e
have to be determined experimentally, by calibration.
However, almost all authors reported the strongest influence of the wire aspect
ratio L/D on their values. As can be seen in the presented formulas, angle is
substituted by u
0
(see fig. 3.7), because all two-wire probes assume the fluid
velocity vector to be in the plane of their sensors, e.g. U
b
= 0 and U
0
= U
0s
(=
0
).
---------------------------------------------------------------------------------------------------
No. EFFECTIVE COOLING VELOCITY U
e
AUTHORS
---------------------------------------------------------------------------------------------------
1.
e s e
cos U U
0 0
u = , ( )
e e e
, , U o u u u
0 0 0 0
= Bradshaw 1971
2. = + =
0
2 2
0 0
1 u u tan k cos U U
s e
Hinze 1959, Webster 1962,

2 2 2
t n
U k U + = Champagne, Sletcher and
Wehrmann 1967
3. ( ) | | { }
0 0 0 0
2 u u u + = cos cos e cos U U
s e
Fujita and Kovasznay 1968
4. ( ) | |
2
0 0
1 1 u cos b U U
s e
= Friehe and Shwartz 1968
---------------------------------------------------------------------------------------------------
Tab. 3.1: Semi-empirical formulas for hot-wire effective cooling velocity U
e
.
The experimental comparison tests of Bruun, Nabhani, Al-Kayiem, Fardad,
Khan and Hogarth 1990 and Adrian, Johnson, Jones, Merati and Tung 1984,
indicate that Hinzes formula (2) gives the best fit in the pure jaw tests, although
the other expressions are more appropriate for certain applications.
However, no one of these expressions describes quite accurately all effects
relevant for wire cooling. For example, the additional aerodynamic blockage
effects of the fluid passing through the opening bounded by the sensor, prongs and
probe stem may also occur (see Comte-Bellot, Strohl and Alcaraz 1971 and
Adrian, Johnson, Jones, Merati and Tung 1984). They are especially important
MULTIPLE HOT-WIRE PROBES 39
__________________________________________________________________________________________________________________________
when the probe axis angle toward the instant flow direction is very high. In order to
account for all of them, a more sophisticated three-dimensional model of hot-wire
cooling mechanism is needed and, obviously, a suitable probe capable of
simultaneous three-dimensional measurement of fluid velocity vector.

U
n
y

U
t

o
x
z
u
0

U
b
=V
U
0

0


U
U
0s

W
z
W

U
o
o
o
x

Fig. 3.7: Evaluation scheme for effective cooling velocity components of slanted
hot-wires, placed in a horizontal xOz plane.
The appearance of multiple sensor probes with three or more hot-wires enabled
simultaneous instant measurements of all three components of fluid velocity
vectors. It has been followed by development of corresponding expressions for U
e
,
which also include all three components of fluid velocity. Jorgensen 1971
proposed a popular formula, at present one of the most accurate among other
known expressions that describe hot-wires cooling:
2 2 2 2 2 2
b
t n e
U h U k U U + + = . (3.59)
Expression (3.59) is based on the two calibration constants designated by k and h,
which has to be determined experimentally. It is known that a value of h reaches
minimum for end-plated sensors.
Vukoslavevi, Wallace and Balint 1991 also reported an original expression
that describes relation between the effective hot-wire cooling velocity and three-
dimensional real fluid velocity vector:
2 2
0
2 2
0
2
b
e s e
U h cos U U + = u . (3.60)
Although simpler in comparison to (3.59), expression (3.60) was successfully used
for calibration curve fitting and signal interpretation of the first sub-miniature
probe WP-9t(G) with nine hot-wires, specified for high-accurate simultaneous
measurement of three-dimensional turbulent velocity and vorticity vector fields.
40 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
This formula is based on the effective normal component
e s Ne
cos U U
0 0
u = (3.61)
and neglecting the tangential component U
t
. The effective normal component U
ne

can be also expressed as a function of velocity components U and W and effective
value o
e
of hot-wire geometrical angle o toward the normal of its axis. Effective
angles u
0e
and o
e
has to be determined experimentally. For high-precision
measurements, the influences of velocity vector U
0
magnitude and a real value of
angle u
0
on their value have to be taken into account.

3.3.3 The generalised law" of hot-wire cooling
Effective cooling velocity U
e
is most commonly described by the fluid velocity
components U
n
, U
t
and U
b
, defined in the local Cartesian coordinate system of the
sensor. However, practical reasons impose the necessity of expressing hot-wire
measurement results in the coordinate system attached to the whole probe or, even
more, to the flow configuration generated in the applied laboratory test facility. In
order to achieve this goal, it is necessary to derive a suitable relation between the
velocity vector components described in these two coordinate systems:
0 0 0 0
t U b U n U U
t b n

+ + = , (3.62)
k W j V i U U

+ + =
0
. (3.63)
For a normal single-sensor probe, shown in fig. 3.3, these components are
identical. However, situation is quite different for inclined wires: in the case of X
probe, with horizontal sensors (xOz plane), the following relations are valid:
o o sin W cos U U
n
= ,
V U
b
= (3.64)
o o cos W sin U U
t
= ,
where the upper sign is for wire and fluid velocity vector oriented in accordance to
fig 3.7, and the bottom sign is for the other sensor. Similar expressions can be
derived for X-hot-wires placed in the vertical plane:
o o sin V cos U U
n
= ,
W U
b
= , (3.65)
o o cos V sin U U
t
= .
In general case, for an optional position of coordinate system (Oxyz) and sensor
orientation, it will be:
W n V n U n U
n
+ + =
3 2 1
,
W b V b U b U
b
+ + =
3 2 1
, (3.66)
W t V t U t U
t
+ + =
3 2 1
,
MULTIPLE HOT-WIRE PROBES 41
__________________________________________________________________________________________________________________________
where the coefficients n
i
, b
i
, and t
i
depend on the hot-wire sensor orientation.
In order to measure components U, V and W of fluid velocity vector, it is
necessary to express the wire effective cooling velocity U
e
as an explicit function
of these three velocity components. Simple combining (3.59) and (3.66) can
provide an adequate formula
W V c W U c V U c W c V c U c U
e
+ + + + + =
6 5 4
2
3
2
2
2
1
2
, (3.67)
where constants c
j
are functions of coefficients n
i
, b
i
, t
i
, (i = 1,2,3, for U, V, W,
respectively), as well as of k and h. The most convenient way to determine c
j
(j =
1,2,3,4,5,6) is experimentally, in a calibration procedure. Simultaneous
measurement of three components of fluid velocity vector demands application of a
probe containing at least three hot wires (k = 1,2,3). It means that 3 x 6 = 18
constants c
jk
will appear to be determined in a calibration procedure of such probe.
The situation becomes even more complicated if a high measurement precision
is needed. In such cases, it is necessary to take into account the influence of
intensity of fluid velocity vector on the calibration constants c
jk
, what introduce
additional problems in the calibration procedure. This phenomenon, known as non-
linearity of hot-wire response to intensity and direction of the fluid velocity vector,
is important at fluid velocities below 5 m/s and especially below 2 m/s.
For the probes containing two hot-wires that measure only two components of
fluid velocity, while neglect the third one, for example X probe with sensors lying
in the horizontal (xOz) plane, component V has to be neglected, and therefore:
W n U n U
n
+ =
3 1
, 0 =
b
U , W t U t U
t
+ =
3 1
, (3.68)
what gives a simplified form of (3.67)
2
3 2
2
1
2
W a W U a U a U
e
+ + = . (3.69)
Three constants a
i
(i = 1,2,3) in (3.69) appear for each sensor, what results in six
constants for two sensors of X or V probe. Their values, needed in order to
measure components U and W, are evaluated by calibration, as usual.
The expressions (3.68) and (3.69) can be additionally simplified. Neglecting the
effect of tangential component U
t
on the convective cooling of the wires of X/V
probe, reduce the problem to the simplest possible form: well known cosine law
where only U
n
is taken into account. In such case, for the X probe, the effective
cooling velocities for two sensors are:
o o sin W cos U U U
n e
+ = =
1 1
, o o sin W cos U U U
n e
= =
2 2
. (3.70)

3.3.4 The cosine law and hot-wire effective cooling angle
The cosine law (3.57) is very simple but fairly old and an inaccurate formula.
However, because of its simplicity, it is still in use for the velocity measurements
when the high measurement accuracy is not required. This is the reason why its
42 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
description is included here, as well as because of its connection with the cooling
law of Vukoslavevi, Wallace and Balint 1991 (3.60). It is usually based on the
effective value
e
of the angle between the fluid velocity vector
0
U and the
normal to the wire axis (i.e. total normal velocity component U
N
) (figs. 3.3, 3.4):
e N e
cos U U U = =
0
. (3.71)
Thus, relating the effective velocity in (3.71) to the law of King 1914
(
p
e
U B A E + =
2
, p~0.45 - (3.37)) enables interpretation of hot-wire anemometer
output signals. The effective angle has to be determined experimentally, by
calibration. This can be achieved if a flow stream of known direction and velocity
magnitude is available, or a facility that enables probe directional (pitch)
calibration through the steps A of pitch angle , of about 5
0
from the null-
angle referent position. Following Bradshaw 1971, a sketch of calibration
mechanism that enables manual directional calibration of X hot-wire probe is
presented in fig. 3.8.
Test rig wall seen
from a side
Arm fixed to probe
support rod outside
the test rig
Probe support
inside the test rig
Probe inside
the test rig
Peg to locate arm Holes drilled in the wall
Probe support rod
passing trough the
wall of test rig

Fig. 3.8: A sketch of mechanism specified for manual pitch calibration of hot-
wire probes with two sensors. Adapted from: Bradshaw 1971.
During the probe pitch calibration, induced fluid velocity components U
i
and
V
i
in the calibration plane, are computed for each probe position, defined by
pitch angle ,
cos U U =
0
, (3.72)
sin U V =
0
(3.73)
while simultaneously recording the corresponding output voltage E
i
for each
sensor. In this case, effective attack angle
e
of a fluid velocity vector
0
U
toward a
wire axis (presented in fig. 3.9) is related to the effective slanted hot-wire angle
e e
o =
(the sign depends on the wire orientation). Angle o
e
is the first
approximation of the real geometrical angle o computed by regression methods,
e.g. by fitting the calibration data for each wire of the applied probe toward the
cosine law.
However, the influence of hot-wire supports (prongs) may limit an angular
applicability range of the cosine law, even with the effective cooling angle
MULTIPLE HOT-WIRE PROBES 43
__________________________________________________________________________________________________________________________
used instead of the real geometrical hot-wire angle. According to Bradshaw
1971,1975, a good indication of the cumulative effects of prongs interference and
other sources of hot-wire response departure from the cosine law can be obtained
by plotting the wire (anemometer) voltage E against the angle in the form of:
p
E ) ( E
E ) ( E
cos Q cos
2
0
2
2
0
2
0 =

=

(3.74)
toward sin, as it is presented in fig. (3.10).

(a) (b)
Fig. 3.10: Determination of effetcive cooling angles of X probe DANTEC 55P61:
(a) voltage response; (b) procedure. Adapted from: Petrovi et al., 1998.
In expression (3.74), E
0
is the output anemometer voltage at zero fluid velocity.
Any significant deviation of the calibration point from the lines defined by
(3.74) and drawn in fig. 3.10, indicates that the cosine law is not valid in that
angular range. In the experimental study of Petrovi, Schenck and Durst 1998, the
applicability limit of this cooling law was up to 25
0
.
Analysis of such kind is also important because it enables evaluation of an
effective hot-wire cooling angle. Following Bradshaw 1971,1975 it is the angle
of inclination of the line (3.74), which is drawn toward sin see fig. 3.10b.

x
U
0

U
V
o
e

Flow
direction
x
U
0
U
V
o
e

U
n
=U
N

Flow
direction

e

Normal to the
wire axis

e

o
e

o
e

Normal to
the wire axis
U
n
=U
N

Fig. 3.9: Determination of effective cooling angle of a slanted hot-wire -
directional (pitch) calibration. Adapted from: Petrovi 1991.
44 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
3.4 A TYPICAL CTA EXPERIMENTAL SET-UP

1

HWP
2
CTA BRIDGE
+ AMPLIFIER
6
7
OSCIL-
LOSCOPE
3
SUBTRACTION
UNIT
4
SECONDARY
AMPLIFIER
6
DIGITAL
VOLTMETER
5
RMS
VOLTMETER
II STAGE: ANALOGUE PROCESSING
(BUCKING-UP AMPLIFIER)
AUXILIARY ELECTRONIC EQUIPMENT
SET-UP AND CONTROL
OF THE 1
-ST
& THE 2
-ND
STAGE
I STAGE: MEASUREMENT
(HOT-WIRE ANEMOMETER)

E
0
=E
0
+e
0

U=U+u

E=E+e
E
0

e
0

G
S
*e
0

6
8
ANALOG
TO DIGITAL
CONVERTER

E
0


G
S
*e
0

6
9
PC
+
EXT.MEMORY
III STAGE: DIGITAL PROCESSING
DIGITAL CONVERSION OF ANALOGUE SIGNALS WITH POSSIBLE REDUCTION
BEFORE STORING IN THE EXTERNAL MEMORY UNIT (DISKETTE, COMPACT DISK)

Fig. 3.11: A sketch of typical experimental set-up for constant-temperature
anemometer (CTA) system. Adapted from: Petrovi 1991.
CTA system explained in the section 3.2.2 needs to be supported by additional
electronic equipment. Nor the anemometer set-up procedure, neither the fluid
velocity measurement, can be performed quite independently. A typical
experimental set-up, specified for single-channel manually controlled CTA unit, is
sketched in fig. 3.11. Of course, if a multiple hot-wire probe is used, each sensor
has to be connected to a separate set of supporting electronic equipment.
Single-sensor probe (1) is connected directly to CTA system (2). Under
variations of flow velocity U(t), simply designated in fig. 3.11 as U, the voltage
drop E along hot-wires also fluctuates. Similarly to the concept of Reynolds for
MULTIPLE HOT-WIRE PROBES 45
__________________________________________________________________________________________________________________________
fluid velocity components, it can be also separated to a constant voltage
approximately equal to time average value of E(t) and fluctuating part e(t):
( ) ( ) ( ) t ' e t E t E + = . (3.75)
In order to simplify figure 3.11, time dependent variables are presented as
' e E E + = (3.76)
for a voltage drop along the wire, and
0
0
0
' e E E + = (3.77)
for the output voltage of CTA unit (2).
E
MAX
E
MIN
0
E
0
=E
0
+e
0

e
0

G
S
- e
0

t
I stage II stage
CTA
SUBTRACTION
UNIT
SECONDARY
AMPLIFIER
E
0
-
~E
0
At At At
G
S
AMPLIFIER GAIN

Fig. 3.12: Processing of CTA output voltages in the bucking-up amplifier:
(I) CTA: anemometer output voltage E
0
(t);
(II) BUCKING-UP AMPLIFIER: - signal after subtraction e
0
(t);
- adjusted voltage G
S
- e
0
(t),
forwarded to ADC.
Adapted from: Petrovi 1991.
CTA reacts nearly instantly to variations of the voltage drop E along the heated
wire, what results in the fluctuating anemometer output voltage E
0
that can be
46 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
visually controlled by two-channel oscilloscope. The integrating voltmeter (5)
measures the so-called root-mean-square value of E
0
(RMS(E
0
)), while classic
voltmeter (6) displays the time-averaged value
0
E . This information is very
useful, because it enables preliminary control of the basic functions of hot-wire
probe, CTA system and wind tunnel (i.e. fan rotation rate) stability (not sketched in
fig. 3.12).
The application of so-called bucking-up amplifier is not an obligatory, but an
extremely useful measurement approach. This analogue unit represents the second
stage of analogue signal processing. However, it does not adjust the wire signal E
directly, but the anemometer output voltage E
0
, toward the requirements of
analogue-to-digital converter (ADC). The CTA signal passes two electronic
circuits. The first (3) is used by an experimentalist to subtract some constant
voltage
0
E , which is as close as possible to the time-average of E
0
(see fig. 3.12).
The second circuit (4) contains the low-pass electronic filter that attenuate noise
over some pre-defined frequency (10 kHz for example) and a low-noise amplifier
of gain G
S
. This second amplifier amplifies the filtered fluctuating part ' e of CTA
signal up to measuring limits defined by the range of analogue-to-digital converter
(typically 2.5 V, 5 V, 10 V, 0-10 V, or 0-20 V). As a result, signal ( ) ' e G
S

transformed in the bucking-up amplifier uses nearly the whole measurement range
of ADC (8), increasing its measurement resolution. Furthermore, electronic
filtering before the secondary amplification of anemometer output signal increases
the signal-to-noise ratio.


Hot-wire anemometer with supporting electronic equipment.
University of Montenegro, Department of Mechanical Engineering, Turbulence
Flows Laboratory, Podgorica











4 4. . M ME EA AS SU UR RE EM ME EN NT TS S O OF F T TU UR RB BU UL LE EN NT T V VE EL LO OC CI IT TY Y F FI IE EL LD DS S
B BY Y M MU UL LT TI IP PL LE E H HO OT T- -W WI IR RE E P PR RO OB BE ES S


T Th he e w wo or rl ld d i is s a an n e et te er rn na al l b bu ur rn ni in ng g f fi ir re e, , w wh hi ic ch h b bu ur rn ns s w wi it th h m me ea as su ur re e a an nd d e eq qu ua al ll ly y s so o i in n f fa ad di in ng g. .
( (H He er ra ac cl li it te e, , 5 53 35 5- -4 47 75 5 b be ef fo or re e t th he e N Ne ew w A Ag ge e) )


4.1 INTRODUCTION
Present chapter reviews recently developed multiple hot-wire anemometer
probes specified for subsonic isothermal turbulent velocity field measurements.
Special attention is paid to analysis of the influences of hot-wire probe
configuration and sensing volume dimensions on the probe measurement accuracy
and angular range of applicability, known as the uniqueness domain. However,
ultramodern vorticity probes, which are also capable of very accurate velocity as
well as vorticity fields measurements, are described in the next chapter because of
their multipurpose role and very special design, as well as extremely complex
signal interpretation procedures.
Besides the multiple hot-wire configurations, the older version of single-wire
probes and X and V configurations with two sensors still deserve attention because
of their simplicity and applicability in a wide area of engineering problems.
Triple sensor probes are interesting as the first instruments capable of
simultaneous three-dimensional measurements of fluid velocity vector components,
as well as the logical step toward the configurations with four hot-wires, specified
for the same purpose.
However, a number of excellent textbooks and scientific papers, where
corresponding references about these probes are discussed and summarised,
currently are commercially available. Among many others, the following should be
mentioned: Kovasznay 1954, Hinze 1959, Corrsin 1963, Bradshaw 1971,
Sandborn 1972, Comte-Bellot 1976, Vagt 1979, Blackwelder 1981, Perry 1982,
Freymuth 1983, 1992, Fingerson and Freymuth 1983, Lomas 1986, Muller 1987
and Bruun 1995.
48 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
4.2 BASIC TYPES OF THE PROBES FOR VELOCITY FIELD MEASUREMENTS
A variety of hot-wire probes with different number of sensors and their
configurations exists. Especially after introducing digital computers in the
experimental practice in the early 1970s, the applicability range of hot-wire probes
is crucially enlarged. Digital computers allowed very complex probes to be used
for experimental purposes, as well as in the industry. The most common hot-wire
probe configurations, which typical dimensions based on hot-wire diameters D =
2.5 m, are sketched in figs. 4.1 4.6, and described in the following text.
Single normal-hot-wire
configuration VP-1n (fig.
4.1) enables measurement of
one component of fluid
velocity vector, normal to
the wire. It can be designed
to the smallest dimensions in
comparison to the other
existing probes. However, this configuration demands neglecting of other two
velocity components, what sometimes leads in ambiguity.
Slanted-hot-wire probe VP-1s (fig. 4.1) posses single wire, like a normal-wire
configuration, with the only distinction that the sensor is mounted at angle of 45
0

toward the probe axis. This geometry is capable of measurement of several
statistical moments of turbulent velocity, what can be achieved by placing it
sequentially at various different orientations toward the main flow. In this case,
results suffer from the similar problems, as is the case with the previous probe.
V and X probes with two hot-
wires, VP-2v and VP-2x (fig.
4.2), are used to measure two
components of fluid velocity in
the centres of measuring volumes
formed by sensors. However, the
lack of the third sensor demands
neglecting the third component
during output signals
interpretation, what may produce
significant measurement errors of
the other two measured
components. In the real three-
dimensional flows, such as
turbulence, the fluid velocity vector fluctuates and, obviously, is not exactly in the
plane of sensors. Depending on the misalignment between hot-wires plane and
instant orientation of fluid velocity, neglected component may increase up to the
VP-1n
0
.
5
m
m

VP-1s
0
.
7
m
m


Fig. 4.1: Single hot-wire probes: - normal VP-1n;
- slanted VP-1s.
VP-2v
1
2
1
2
VP-2x
0.5mm
0
.
5
m
m

0.5mm
0.5mm

Fig. 4.2: Typical probes with two hot wires:
- V probeVP-2v and
- X probe VP-2x.
MULTIPLE HOT-WIRE PROBES 49
__________________________________________________________________________________________________________________________
level when velocity measurement by such probes may loose any sense. This is one
of the crucial disadvantages of stationary single and two-wire probes, which
reduces their applicability to flows with low turbulence levels.
T-shaped triple hot-wire configuration VP-3t,
sketched in fig. 4.3, has enough number of sensors
to measure simultaneously all three components of
fluid velocity. Therefore, neglecting of any
component is not necessary, what increases the
measurement accuracy in comparison with the
previously described probes.
"Mercedes" (M) geometry VP-3m (fig. 4.3)
represents the most commonly used triple sensor
probe, containing three mutually orthogonal sensors
that form a cone of half angle o = 54.74
0
. Both T
and M configuration enable simultaneous
instantaneous measurement of three fluctuating
components of fluid velocity. The probe VP-3t can
be manufactured more easily, because its two wires
lay in the same plane. Furthermore, signal
interpretation procedures are also simpler and
converge faster, what was the second reason to use
them in the experimental practice. However, it is
capable of measurement of fluid velocity vectors in
a narrower angular range (toward the probe axis) in
comparison to M configuration. Having in mind fast
development of personal microcomputers and
numeric procedures for interpretation of hot-wire
probes signals, it seems that T probes will be completely substituted by M
configuration in the nearest future.
Four hot-wire probes,
available in the "plus"
VP-4+ and "quadrate"
VP-4q configuration,
illustrated in fig. 4.4, are
capable of simultaneous
measurements of all three
components of fluid
velocity vector. Although
their design is more
complex in comparison to
triple probes, four-wire
probes posses two
1mm
0
.
5
m
m

2
3
1
VP-3t

1
2
3
VP-3m
D = 1 mm

Fig. 4.3: Triple probes:
- probe VP-3t and
- probe VP-3m.
VP-4+
1 mm
1

m
m

2
4
3
1
0.5 mm
1
3
4
2
VP-4q
0.5 mm
0.5 mm
0
.
5

m
m


Fig. 4.4: Four-wire probes: -plus probe VP-4+ and
-quadrate probe VP-4q.
50 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
important advantages: wider angular range of velocity vector directions that can be
measured and higher measurement accuracy. Both the plus and quadrate hot-
wire geometry can be supported by identical calibration procedures and signal
interpretation procedures. They differ only in overall dimensions: quadrate probe
is smaller and, therefore, more sensitive toward thermal interference between the
various sensors. The intensity of this phenomenon increases with decreasing the
fluid velocity and may generate problems in the slow-motion flow regions, such is
the case in the wall vicinity for example. Therefore, quadrate probes has to be
designed with a special care.
Quintuple probe VP-5q (fig. 4.5) containing five wires, represents the newest
design of multiple hot-wire probe specified for fluid velocity measurements in non-
stationary flows. An additional sensor
enabled special radial arrangement of
hot-wires, which removes the prong
interference at high angles of velocity
toward the probe axis that can appear
in the case of four-sensor probes.
Recently developed as an extension
of four-wire probe VP-4+, it also
provides increased measurement
accuracy of fluid velocity vector
components. In such a way, an
applicability range of stationary hot-
wire probes is significantly enlarged.
Besides the commercial hot-wire
configurations, some researches also
use their own technologies
to design and manufacture
various probes. However,
for standard applications,
most frequently used hot-
wire configurations are
those of the world famous
manufacturers DANTEC and
TSI. Their popularity is
based primarily on the
guaranteed quality and
provided technical support.
These probes are presented,
together with plus probe
of Dobelling, Leuckel and
Lenze 1990a,b, in fig. 4.6.

0.5 8.0
72
0
45
0
| 1.1
| 0.02
|
0.1
VP-5q

Fig. 4.5: Quintuple hot-wire probe.
Source: Holzapfel, Leuckel and
Lenze 1994.

Fig. 4.6: Typical hot-wire probes with strait prongs:
a) normal-wire geometry DANTEC 55P11,
b) slanted-wire probe DANTEC 55P12,
c) X-wire configuration DANTEC 55P61,
d) triple hot-film probe TSI-1299 and
e) quadruple plus probe, designed by
Dobbeling Leuckel and Lenze 1990a.
(a)
(b)
(c)
(d)
(e)
MULTIPLE HOT-WIRE PROBES 51
__________________________________________________________________________________________________________________________
4.3 HOT-WIRE PROBES DESIGN
The most commonly used materials for hot-wire sensors are tungsten, platinum,
platinum-rhodium (90% Pt + 10% Ro) and platinum-iridium (80% Pt + 20% Ir).
Their basic physical properties
are listed in table 4.1. Magnified
photographs of the surfaces of
hot wires, containing 80%Pt +
20%Ir (D = 2.5 m) and pure
tungsten (D = 3.8 m), are
presented in fig. 4.7.
Tungsten wires have a high
temperature coefficient of
electric resistance and the
highest tensile strength.
However, according to
Blackwelder 1981, this material
can be used only at fairly low
operating temperatures, under
350
0
C. This restriction
originates from poor oxidation
resistance of tungsten at high
temperatures. An oxidation
process significantly intensifies even over 300
0
C. Platinum possesses a good anti-
oxidation resistance and can be produced in very small diameters, up to 0.5 m.
Unfortunately it is weak, especially at high temperatures. Alloy containing 80 % of
platinum and 20 % of iridium represents a compromising hot-wire material. It has
good anti-oxidation properties, acceptable tensile strength, but low temperature
coefficient of electric resistance. According to its properties, this alloy finds
applications in the situations when wire temperature is too high for tungsten and
platinum. However, tungsten wires are still used in most airflow studies.

PHYSICAL PROPERTY
Tungsten
100%W
Platinum
100%Pt
80%Pt
+20%Ir

Temp. coef. of resistance o(1/K)
0.0045 0.0039 0.0008

Resistivity (O m)
5.5 x 10
-8
10 x 10
-8
31 x 10
-8


Ultimate tensile strength (N/mm
2
) 4120 241 981

Thermal conductivity (W/m K) 197 70 17.6

Tab. 4.1: The most important physical properties of commonly used hot-wire
materials. Source: Fingerson and Freymuth 1983.

Fig. 4.7: Typical hot-wires, under microscope:
(a) 80%Pt + 20%Ir, D =2.5 m (left),
(b) 100% tungsten, D =3.8 m (right).
Adapted from: Blackwelder 1981.
52 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
The choice of hot wire diameter is a question of compromise. Main advantages
of its small value are improved signal-to-noise ratio at high frequencies, increased
frequency response and spatial resolution, reduced flow interference and sensor
end-conduction losses. However, large value of wire diameter increases its
strength, but reduces sensitivity due to the particles presence in fluid etc.
Experimental practice shows that an optimum value of hot-wire diameter should be
in the range between 2 m and 5 m for the most common applications.
The situation is analogue in the case of defining the length of hot-wire. While
short sensors maximise the spatial resolution and minimise the aerodynamic
loading, long wires minimise end-conduction looses and provide more uniform
temperature distribution. The question of the best compromise for the wire length-
to-diameter ratio is differently resolved by various turbulence researches. In
accordance to Ligrani and Bradshaw 1987 and Turan and Azad 1989, it should be
at least around 200. Some other designers of hot-wire probes, like Vukoslavevi,
Wallace and Balint 1991, applied hot-wire probes with the aspect ratio of about
280. Worldwide known manufacturer DANTEC mainly uses the aspect ratio L/D =
250 for the standard end-plated probes with Pt-Ro wires 5 m in diameter, and L/D
= 400 for probes with non-plated 2.5 m tungsten sensors.
Hot-wire is either attached directly to the tips of the prongs or through its plated
ends. The central unplated part of the sensor represents the active sensing length.
Plating of hot-wire ends efficiently define the sensing length of the wire and
reduces the heat flux dissipated by the prongs. It also provides a more uniform
temperature distribution along the unplated part of a heated sensor. The additional
advantage of probes with plated wires ends, toward unplated, is the reduction of
flow disturbances at the measurement point. This property is achieved thanks to the
wider distance between the prongs and active (sensing) part of a wire. Fiedler 1978
verified large discrepancies between the static and dynamic calibration of unplated
hot-wire probes. However, according to Bruun 1976 and Fiedler 1978, both types
of calibration gave nearly the same results for the probes with end-plated sensors.
Elements of hot-wire probe, primarily stem and prongs, aerodynamically disturb
the flow and, therefore, change the velocity field over the sensor in comparison to
the undisturbed flow. Systematic studies of these phenomena are reported by
Comte-Bellot, Strohl and Alcaraz 1971, Strohl and Comte-Bellot 1973, Adrian,
Johnson, Jones, Merati and Tung 1984, Merati and Adrian 1984, Dobbeling,
Lenze and Leuckel 1990a, Holzapfel, Lenze and Leuckel 1994, etc. They
provided general information for hot-wire probes design and formulation of
suitable calibration procedures and output signals interpretation algorithms.
However, one of the most complicated problems in hot-wire anemometry is
connected with thermal interference between various probe elements (sensors
primarily). It is caused by the thermal wake behind a heated wire, which intercepts
the other sensor of the same multiple probe, or the wire in a closely positioned
neighbour probe. Thermal interference has been in the focus of interest of many
MULTIPLE HOT-WIRE PROBES 53
__________________________________________________________________________________________________________________________
researches. Ko and Davis 1971, Jerome, Guitton and Patel 1971 and Strohl and
Comte-Bellot 1973 studied this problem in the case of single and X-wire probes. In
addition, more complex multiple wire configurations with four, nine and twelve
hot-wires are tested (among others) by Vukoslavevi, Wallace and Ballint 1991.

4.4 DIRECTIONAL SENSITIVITY OF FINITE-LENGTH-WIRE PROBE
Conductive heat transfer from hot-wire ends to the prongs can be neglected if
the sensor aspect ratio is sufficiently high. However, following Wyngaard 1969,
spatial resolution criteria imposes sharp restrictions on dimensions of hot-wire
probes and their sensors. These limitations very often demand application of wires
with aspect ratio, which is not large enough to allow neglecting of the conductive
heat transfer from the sensor toward its supports. Following Nitsche and
Haberland 1984 and Pitts and McCaffrey 1986, for example, conductive losses of
a heated wire have to be accounted for in these situations, because of their
significant influence on the wire response to changes of fluid velocity vectors.
A common approach in modelling the mechanism of hot-wires cooling is based
on the well-known term of so-called effective cooling velocity U
e
. Some of the
most popular formulations of relation between U
e
and the components of fluid
velocity vector are reviewed in the section 3.3.2. The expressions, relevant for
single, two and three-dimensional measurements, are included. However, they
were only preliminary analysed, to provide an insight in the fundamental hot-wires
operational principles and signal analysis. The most sophisticated details of hot-
wire response to fluid velocity fluctuations, related to directional response of
multiple sensor probes, are discussed in the present section.
Graphical analysis of hot-wire directional response is very illustrative. Among
the first, Willmarth 1985
applied it. He analysed
angular response of an
indefinitely long heated
cylinder, using the cosine
law (see (2.1) and (3.57))
and showed that indefinite
number of possible fluid
velocity vectors with
various magnitudes and
directions, giving the same
wire response, exist. Their
tips lay on the centre of the
sensor and the tails are on
the cylinder of indefinite
length (fig. 4.8-left). Even

Hot-wire
Tips of fluid
velocity vectors
Tails of fluid
velocity vectors
Hot-wire
SIDE VIEW
Uniqueness
response
cylinder

Fig. 4.8: Fluid velocities giving the same signal of
indefinitely long hot-wire sensor,
according to Willmarth 85. Adapted from:
Dobbeling, Lenze and Leuckel 1990a.
54 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
more, an indefinite number of
possible fluid velocity vectors with
the same intensity but various
directions, giving the same wire
response, exist in the plane normal
to the wire axis (fig. 4.8).
As an extension of Wil-
lamarths 1985 ideas, more
realistic probe behaviour with full
Jorgensens equation (3.59) was
analysed by Dobbeling, Leuckel
and Lenze 1990a. It is graphically
illustrated in fig. 4.9 by an offline
contraction of a rotational
ellipsoid, with characteristic
dimensions giving k = c/b and h =
c/a for Jorgensens constants.
The same authors also analysed
multiple solutions of response
equations for X probe with
orthogonal wires and prongs lying
in the parallel planes (assuming
that one should be very close to
the other). Obtained graphical il-
lustrations of Jorgesen 1971 non-
linear formula (3.59) are presented
in fig. 4.10 for both sensors. In this
case, two intersection curves exist
that satisfy both wire response
equations, giving therefore the
infinite number of solutions for the
instantaneous fluid velocity vector.
Triple hot-wire probes are most
commonly designed with mutually
orthogonal sensors, as
sketched in figs. 4.3 and
4.11. From the condition
of probe summetry
toward the own axis, it
follows that each of
three wires of such
B
N
T
Location of the tips of all
possible fluid velocity
vectors that give the same
wire response.
a
c
b

Fig. 4.9: Fluid velocities giving the same
signal of a finite-length (real) wire.
Adapted from: Dobbeling, Lenze
and Leuckel 1990a.
Intersection curves
of wire 1 and 2.
Orientation
of wire no. 1
Wire no. 1
Orientation
of wire no. 2
Wire no. 2

Fig. 4.10: Response of orthogonal X-probe,
following the law of Jorgensen
1971. Adapted from: Dobbeling,
Lenze and Leuckel 1990a.
A front view!
An axonommetric view!

Fig. 4.11: Orthogonal geometry for triple probes:
Adapted from: Rosemann 1989.
MULTIPLE HOT-WIRE PROBES 55
__________________________________________________________________________________________________________________________
probe form an inclination angle of o = 54.74
0
, measured from the probe axis, as it
is shown at the end of this section and illustrated in fig. 4.13.
For this, so-called "Mercedes geometry, Dobbeling, Lenze and Leuckel 1990a
found it suitable to assume that the wires are oriented in such a way to form a local
Cartesian coordinate system. They showed that symmetrical behaviour of hot-wire
heat-transfer mechanism enables mirror-imaging of the arbitrary oriented fluid
velocity vectors for the wire response equation (3.59) with the planes defined by
two of the local coordinate axes. Therefore, in general, there are eight velocity
vectors of the same magnitude, which give the same hot-wire output signal (fig.
4.12). However, in the special cases when one or two of velocity components are
zero, there will be only four or even only two mirror-imaged vectors, respectively.
Thus, the uniqueness
range of fluid velocity
vectors, measured by a
"triple" orthogonal probe,
is restricted to only one of
the eight existing octants
of orthogonal Cartesian
coordinate system. In
order to avoid uncertainty
about uniqueness, the
measured fluid velocity
vectors must be in the
cone (most commonly
characterised by its half-
angle |
T
) that can be
placed in one of these
octants. This means that
the allowed inclination of
velocity vector toward the
probe axes is limited by
|
T
. If some of the
measured values of fluid
velocity vectors are on the
edge of this cone, the
probe probably does not give unique solutions. Practically this problem is resolved
by introducing the so-called threshold: a special subroutine that controls
orientation of measured fluid velocity vector toward the probe axis and rejects
those data which fall out of the uniqueness cone.
Both o and |
T
angle can be easily evaluated using simple rules of trigonometry.
Assumed ideally orthogonal "triple" hot-wire configuration, what gives three right
angles ZW
1
0W
2
= ZW
2
0W
3
= ZW
2
0W
3
= 90
0
, i.e. three equivalent right-angled
Intersection point 1+2+3
(detected location of
velocity vector tail)
Wire 3
Intersection
curve 1+3
Intersection
curve 2+3
Wire 2
Acceptability octant
Intersection
curve 2+3
Wire 1

Fig. 4.12: Graphical illustration of directional
response of an orthogonal triple probe,
assuming hot-wires cooling law of
Jorgensen 1971. Adapted from: Dob-
beling, Lenze and Leuckel 1990a.
56 P. V. Vukoslavevi and D. V. Petrovi
_________________________________________________________________________________________________________________________
triangles AOW
1
W
2
, AOW
2
W
3
and AOW
1
W
3
(fig. 4.13). Having in mind that these
triangles are defined by virtual length
3 2 1
OW OW OW s = = = of hot wires, it is
evident that each of three triangles posses two sides of equal length s.
The other relevant parameters may be evaluated using simple rules of geomety:
s s
E
= 2 ,
2
E
S
s
h = , s h
E

2
6
, s r
T
=
6
6
, s r
G
=
3
6
(4.1)
and
3 6 = o sin o = 54.74
0
, 3 3 =
T
sin | |
T
= 35.26
0
. (4.2)
Presented theoretical analysis of Dobbeling, Leuckel and Lenze 1990a enables
estimation of the uniqueness cone of an ideally configured orthogonal triple
hot-wire probe. They also claimed that enlarging the geometrical wire-angle o
simultaneously enlarges the half-angle of uniqueness cone |
T
and vice versa.
However, some other researchers verified simultaneous decreasing of probe
directional (angular) sensitivity. At the end, it should be noted that |
T
= 35.26
0
is
just the theoretical value, which can not be reached with the real probes.
Probe axis
O
r
T
r
G
h
h
E
W
2

W
1
s
E
Wire no. 1
s

Wires
no. 2, 3
o =54.74
0
+|
T
=35.26
0
E =90.00
0
B
1
h
S
B
2
B
3
S
W
3
Prongs

Fig. 4.13: A sketch of orthogonal triple hot-wire probe and its
uniqueness cone, prepared following Dobbeling,
Lenze and Leuckel 1990a .
MULTIPLE HOT-WIRE PROBES 57
__________________________________________________________________________________________________________________________

4.5 HOT-WIRE PROBES FOR 3-D VELOCITY MEASUREMENTS
Turbulent fluid motion is three-dimensional by its nature. The special unique
mechanism of turbulence maintaining, known as vortex stretching (Tennekes and
Lumley 1978), can not exist in two-dimensional flows. In addition, contemporary
technical problems in turbulent shear flows very often involve situations in which
three-dimensionality is highly expressed. Typical examples are flows over large
roughness elements, near wakes behind axisymmetric bodies, swirling jets, etc.
In these cases, high turbulence levels make employment of single or two-wire
probes non-adequate, because of their measuring principle. They demand
neglecting of the one or two fluid velocity components, which produce measuring
errors on the measured components. These errors increase with increasing the
magnitude of neglected velocity component(s). Operational applicability of X-wire
probes has been extensively studied in the various flows by many researches:
Chang, Adrian and Jones 1983, Nithianandan, Jones and Adrian 1987, Muller
1992, Ong and Wallace 1996, etc. Using computer-generated Gaussian signals of
specified statistical moments, Kawall, Shokr and Keffer 1983 found that
measurement errors of Reynolds stress-tensor components are small for turbulence
levels below 15%. The error due to the cross-velocity in X-wires becomes
especially important for higher-order velocity moments and, therefore, all three
components of fluid velocity vectors should be measured even in the medium
turbulence level flows.
This problem can be sometimes resolved by applying probes with three hot-
wires, which enable simultaneous instantaneous three-dimensional measurement of
fluid velocity vectors. Fabris 1978, Moffat, Yavuzkurt and Crawford 1978,
Accrivlellis 1979, Huffmann 1980, Accrivlellis 1980, Lakshminarayana 1982,
Andreopulos 1983, Buttler and Wagner 1983, Chang, Adrian and Jones 1983,
Mathioudakis and Breugelmans 1984, Muller 1987, Buddhavarapu and Meinen
1988, Lekakis, Adrian and Jones 1989, etc. reported applications of triple-wire
methods. However, the analysis presented in the previous section, also shows the
serious restriction in applicability of triple-wire configurations. They are limited to
a certain angular range where their output signals enable unique determination of
fluid velocity vector. It is
usually represented by a
conical surface, which axis is
coincident to the probe axis,
and is denoted as the
"uniqueness cone" (fig. 4.14).
In highly turbulent flows,
the possibility of confusion
always exists between
different velocity vectors that
give the same set of triple
y
z
x
|
T

|
T
half-angle of the
uniqueness cone
U
0

Fig. 4.14: A sketch defining the uniqueness cone.
58 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
probe output voltages. Tutu and Chevray 1975 referred this as "rectification error"
in their study of accuracy of turbulent velocity field measurement by X-probe.
Uniqueness domain of response equations of multiple hot-wire probe has been also
investigated by Willmarth 1985, Samet and Einav 1987, Lekakis 1988, Lekakis,
Adrian and Jones 1989, Vukoslavevi and Wallace 1983, Vukoslavevi,
Wallace and Balint 1991, Rosemann 1989, Dobbeling, Lenze and Leuckel
1990a,b, Vukoslavevi and Petrovi 1994, Petrovi and Vukoslavevi 1995, etc.
Difficulties originated from restricted applicability domain of stationary
triple-wire probes can be generally resolved in a few main directions. The first is
to employ the other techniques, such as laser-Doppler anemometer (LDA), which
enable velocity measurement in the highly turbulent and even in some recirculation
flows. However, stationary hot-wire techniques (HWA) are less expensive than
LDA and are more suitable in velocity measurements in gases where high
frequency response and low noise are required (Holzapfel, Lenze and Leuckel
1994). This is particularly the case when flow data are needed to test closure
hypothesis in turbulence models, which demand measurement of higher moments
of velocity fluctuations and power spectra. For the latter, a capability of
measurement technique for coincident and time-equidistant high-frequency data
sampling (not available with LDA) is very useful, in order to perform such data
analysis as the fast Fourier transform (Holzapfel, Lentze, Leuckel 1994).
Besides LDA, a flying hot-wire could be also used in highly turbulent flows.
However, it suffers from some operational restrictions similar to LDA, such as low
frequency of data sampling, for example. Furthermore, flying wires are
operationally complicated, because of the necessity for using special traversing
mechanisms. Their measuring principle, based on continuous oscillatory
movement, usually makes these probes less reliable than stationary probes.
Stationary multiple hot-wire techniques have been extensively used for many
years in a great number of applications. Currently, researches invest great efforts
for further improvement of their design. Triple hot-wire probes are commonly
constructed with orthogonal sensors (fig. 4.11). According to Dobbeling, Lenze
and Leuckel 1990a, this configuration maximises the differences in the wire
voltages because each wire is sensitive only in its normal (N) and binormal (B)
direction and relatively insensitive in its tangential (T) direction (see figs. 3.3 and
3.4). The voltage differences between the wires yield the information, which
determines the direction of the velocity vector. Therefore, triple probes will have
maximum angular sensitivity for orthogonal geometry.
However, Lekakis, Adrian and Jones 1989 and Rosemann, Stager and
Kreplin 1989, claimed different reason for the popularity of orthogonal triple
hot-wire configuration: this design is a compromise between the size of uniqueness
domain and probe angular sensitivity.
At present two general approaches for enlarging the uniqueness range of the
stationary probes for 3-D measurement of turbulent velocity field exist. The first is
MULTIPLE HOT-WIRE PROBES 59
__________________________________________________________________________________________________________________________

to apply special non-orthogonal configurations (Acrivlellis 1980 and Kawall,
Shokr and Keffer 1983). However, Roseman 1989 and Lekakis, Adrian and Jones
1989 showed that such non-orthogonal set-up also reduces the total angular
sensitivity of the probe. The final consequence is expressed in larger angular errors
in evaluation of the velocity vector direction. This error becomes even larger if the
hot-wire output signals are superposed by small measurement errors. In addition,
Roseman 1989 gave a detailed analysis of the uniqueness domain and angular
sensitivity behaviour of a number of triple hot-wire configurations.
Besides orthogonal configuration, triple-wire probes are also designed in T
geometry (fig. 4.3), consisting of V probe and a third sensor inclined at 45
0
with
respect to V-wires plane. Turbulence researchers use such probes in order to
simplify the signal interpretation procedure (Spencer 1970, Chang, Adrian and
Jones 1983 etc.). Vukoslavevi, Balint and Wallace 1991 also applied T probe as
a part of nine-sensor vorticity probe. However, T geometry slightly reduces the
uniqueness domain in comparison to orthogonal. Vukoslavevi and Wallace 1983
found half-angle of uniqueness cone to be |
T
= 26.5
0
for ideal T-probe, which is
not influenced by aerodynamic blockage and has effective hot-wire cooling
angles equal to geometrical (h = 1 and o
e
= o = 45
0
in (3.60)). Unfortunately, real
calibration parameters (effective hot-wire cooling angle o
e
and aerodynamic
blockage constant h) gave |
T
~ 17.5
0
for small probes and low fluid velocities.
However, fast development of digital computers has highly increased the
computation speed, enabling efficient application of very complex procedures for
hot-wire output signals interpretation. In this new situation, simple signal
interpretation algorithms for T configurations do not represent their advantage
anymore and it seems that these probes will be substituted in the practice by
superior orthogonal geometry.
The second and presently the most promising approach in reducing difficulties
of interpreting triple-probe output voltages is to use the special four hot-wire
probes. Kovasznay 1950,1954 designed the first "quadruple" probe (fig. 4.15),
independently from the analysis presented above. He specified this configuration
for measurement of turbulent longitudinal velocity U and vorticity e
X
components,
but it was not capable of
measuring the other two
velocity components V and W.
Vukoslavevi and Wallace
1981 practically finalised its
design. They minimised probe
dimensions (according to
spatial criteria of Wyngaard
1969) and introduced separate
(independent) prongs for each
sensor (fig. 4.16). This way the
B
C
A
D
y
z
x
o
o
o
=45
0

U
0

h
P

h
P


Fig. 4.15: Quadruple probe of Kovasznay 1950.
Adapted from: Vukoslavevi 1981.
60 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
cross-talking between different
wire-signals was eliminated and
probe became capable of
simultaneous measuring of all three
velocity components. The tips of the
prongs were bent according to
recommendations of Strohl and
Comte-Bellot 1973 to reduce
aerodynamic disturbances. However
they showed that this probe was able
to resolve the longitudinal vorticity
only under severe restrictions due to
the neglected velocity gradients. They were among the first who pointed out the
strong influence of velocity gradients not only on vorticity but also on velocity
measurements. This problem can be resolved by either minimising of the probe
dimensions or using configurations with nine or twelve hot-wires.
Lemonis and Dracos 1995 were the first who performed signal interpretation of
a four-wire probe, treating it as a simple combination of four triple-sensor
configurations. The velocity vector solution was obtained by averaging the four
triple-wire solutions that are very close one to another. Ideally, they should be
coincident. The only advantage of this approach is expressed through the
possibility of using the existing software for triple probes, which is generally
simpler than a computer code for more complex quadruple configurations.
However, consequently, most of the advantages of four-wire probe, such as the
higher angular sensitivity to fluid velocity variations, are not optimally accounted.

Fig. 4.17: Sensor arrangement of the probe of Samet and Einav 1987 (left).
Fig. 4.18: The plus probe of Dobbeling, Lenze and Leuckel 1990a,b (right).
0
.
5

m
m

0.5 mm
45
0

45
0

U
0

y
z
x

Fig. 4.16: Modified probe of Kovasznay.
Source: Vukoslavevi 1981.
MULTIPLE HOT-WIRE PROBES 61
__________________________________________________________________________________________________________________________

Phailas and Cousteix 1986, Samet and Einav 1987 (fig. 4.17), Rosemann
1989, Dobbeling, Lenze and Leuckel 1990a,b (fig. 4.18), Nguyen 1993, Marasli,
Nguyen and Wallace 1993, Park and
Wallace 1993, Pompeo and
Thomann 1993 (fig. 4.19), Vukosla-
vevi and Wallace 1996 (as a part of
twelve-wire vorticity probe), Petrovi
and Vukoslavevi 1997, Vukosla-
vevi and Petrovi 1997, Wittmer,
Devenport and Zsoldos 1998 have
also used quadruple probes.
Phailas and Cousteix 1986, as
well as Rosemann 1989, stated that
uniqueness cone of the probes
containing four sensors can be
slightly enlarged by increasing the wire-angles toward the probe axis.
Unfortunately, in analogue to triple geometry, increasing of the wire inclination
angle o toward the probe axis simultaneously reduces the probe angular resolution.
Blanco-Marigota, Ballasteros-Tajadura and Santolaria 1998 recently confirmed
this conclusion on the base of comparison test of three X configurations: the first
was with orthogonal sensors (o = 45
0
) and an additional two with non-orthogonal
(o = 60
0
and 67.5
0
, respectively). Their results are valid for quadruple probes
too, because they contain two pairs of crossed-wires.
A variety of procedures for signal interpretation of the four-wire probes exist.
However, they can be classified in a very few general approaches. Vukoslavevi
1981 developed one of the first and the simplest algorithms, which enabled
Vukoslavevi and Wallace 1981 to perform instantaneous three-dimensional
measurement of fluid velocity field. However, they processed the four available
output signals as to be from two independent X-probes and noticed that this
approach suffered from the restrictions that usually follow two-sensor probes.
Samet and Einav 1987, Rosemann 1989 and Pompeo and Thomann 1993
reduced the four-dimensional problem to two coupled two-dimensional problems.
They are solved by iterative algorithm, such as two coupled X-wire probes.
Dobbeling, Lenze and Leuckel 1990a indicated that this approach sometimes
results in ambiguities. The interpretation procedures of Dobbeling, Lenze and
Leuckel 1990a,b, Marasli, Ngyen and Wallace 1993 and Wittmer, Devenport and
Zsoldos 1998 overcame this problem by involving procedures that solve the
resulting system of four wire-response non-algebraic equations simultaneously.
The first two teams used the least-square minimisation procedure. In contrast, the
latter group applied technique based upon the look-up tables with specially
formulated non-dimensional variables which values tend to vary slowly and,
therefore, allows employment of simple interpolation scheme.

Fig. 4.19: Four-wire probe, analysed by
Pompeo and Thomann 1993.
62 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
All procedures that support quadruple probes utilise simultaneously four
output signals. The only exclusion is the algorithm of Vukoslavevi 1993, based
on a special sub-algorithm that chooses three (of four) output signals, which
correspond to the sensors providing the best angular resolution for the instant
orientation of fluid velocity vector. It was extended for twelve-wire vorticity probe
by Vukoslavevi and Wallace 1996.
The most of designers of four-wire probes based their interpretation procedures
on the average values of calibration coefficients obtained from probe directional
calibrations performed in finite steps for several velocity magnitudes. Directional
calibration involved probe oriented at all possible angles to be encountered in the
measurement, at only one magnitude of calibration velocity. In place of using
average values, the interpretation algorithms of Dobbeling, Lenze and Leuckel
1990a,b, Park and Wallace 1993, Vukoslavevi 1993, Vukoslavevi and
Wallace 1996, Ong and Wallace 1996 and Wittmer, Devenport and Zsoldos 1998
involved calibration coefficients from single-velocity directional (angular)
calibration for each different velocity magnitude. These coefficients can be applied
for hot-wire signals measured at each different location in the flow where the mean
velocity is equal (or at least fairly close) to the magnitude of induced calibration
velocity vector.
The purpose of multiple hot-wire probe calibration is to determine empirical
relationships between the output voltage of each of its sensors and the components
of fluid velocity. This three-dimensional function can be determined by direct
measurement. However, having in mind enormous number of various probe
orientations, which are needed for simulation of variations of fluid velocity
direction, this may be a time-consuming task. Furthermore, the information on hot-
wires response functions dependence on the velocity magnitude (for each fluid
velocity vector orientation) is also necessary, what additionally complicate the
calibration procedure. Unfortunately, this is not the end of problems: multiple hot-
wires probe calibration has to be frequently repeated (so-called recalibration) to
account for CTA circuits drifts and dust deposition and ageing of the sensors.
Therefore, it is suitable to seek a simplification of the wire response function to
a manageable level without significantly compromising the measurement accuracy.
The possible approach to such simplification is to split the calibration in two steps.
The first (so-called "velocity calibration") provides information on the sensor
voltage drop connection with effective cooling velocity U
e
. It assumes varying the
fluid velocity magnitudes at zero pitch and jaw probe angles, what can be
performed easily, but should be frequently repeated. Among others, the main
purpose of this calibration step is to account for the temperature drifts of CTA
electronic components. The second more sophisticated and time consuming step,
"angular calibration", relates the four effective velocities U
ei
(i = 1,2,3,4) and the
three velocity components U, V and W. It accounts for sensor and prongs geometry
effects and requires only occasional repetition. Typical examples for the calibration
procedures explained above were developed and applied by Dobbeling, Lenze and
MULTIPLE HOT-WIRE PROBES 63
__________________________________________________________________________________________________________________________

Leuckel 1990a,b and Wittmer, Devenport and Zsoldos 1998, who assumed a
model for sensor response equation that allow de-coupling of velocity magnitude
from its direction.
In order to improve directional (angular) sensitivity of hot-wire probe output
signals interpretation Petrovi and Vukoslavevi 1997, reported improved version
of the algorithm of Vukoslavevi 1993. The optimisation was performed by taking
into account the calibration coefficients velocity-dependence not only on the mean
but also on the instantaneous velocity magnitude at each measuring location. The
idea originates from the simple algorithm of Durst, Melling and Whitelaw 1970
for the normal single-wire probe and procedure of Petrovi 1991 (that was also
applied by Petrovi, Beniek and Oka 1997) for X-probe. In addition, sensor-
alternation subroutine was extended from two vertical to all four hot-wires of
quadruple probe. Our comparison tests confirmed the superiority of optimised
procedure toward its basic version at low fluid velocities. However, as it was
expected, both algorithms achieved similar results at higher fluid velocity vector
magnitudes.
Uniqueness cone of 40
0
half-angle corresponds to a maximum turbulence level
of 38%, assuming the normal (Gaussian) distribution of isotropic velocity
fluctuations and including 90% of all events (Holzapfel, Lenze and Leuckel 1994).
Still, many practical applications exist, which demand probes with wider
(uniqueness) acceptability range due to increased turbulence levels: turbulence
velocity field measurement in a swirl flow
with vortex breakdown, for example. Using
hot-wire anemometer in any flow of
arbitrary configuration, the researcher has to
carefully check that the acceptability range
is not exceeded, neither by the calibration
procedure nor during measurements.
Almost all measured fluid velocity vectors
have to be within the uniqueness cone,
because even a few erroneous data points
may significantly influence the
measurement accuracy of higher-order
moments statistics.
In order to additionally enlarge the
uniqueness domain, Holzapfel, Lenze
and Leuckel 1994 designed a special
quintuple probe, sketched in fig. 4.20. It
is composed of five hot-wires slanted at o =
45
0
toward the probe axis and evenly
distributed in angles of 72
0
(90
0
is for the
plus probe with four sensors).

Fig. 4.20: A special quintuple
hot-wire probe. Source:
Holzapfel, Lenze and
Leuckel 1994.
64 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Probe designers claimed that this arrangement guarantees that at least four hot-
wire sensors are not in the wake of any prong, for arbitrary oriented fluid velocity
vector within a whole hemisphere. In addition, including the fifth wire into the
probe arrangement provides at least three different output voltages for any flow
direction within a hemisphere and therefore, as they believed, prevents ambiguities.
Under assumptions that turbulent velocity fluctuations are isotropic and follow
Gaussian distribution, 90% of all events can be measured by quintuple probe in a
flow with a maximum turbulence level of 60.6%. However, available experimental
set-up enabled probe calibration only within range of 70
0
for pitch and jaw
angles. They did not give a straight forward proof of the uniqueness cone angle for
the probe, so any assumption of achieving the uniqueness cone half angle higher
then the hot wire angle toward probe axis (e.g. o = 45
0
), should be taken with a
special care.

4.6 INFLUENCE OF HOT-WIRE PROBE DIMENSIONS AND GEOMETRY
ON THE MEASUREMENT ACCURACY AND UNIQUENESS DOMAIN
Finite dimensions of hot-wire probes may sometimes lead to errors induced by
strong gradients of the velocity fluctuations and especially of the mean-velocity
gradients in the probe sensing volume. This phenomenon has been in the focus of
turbulence researchers interest for many years, but has not been completely
resolved yet. Wyngaard 1969 reported very important results, which are frequently
cited even nowadays.
In order to resolve problems connected with vorticity measurements, reported
by Vukoslavevi and Wallace 1981, Wassmann and Wallace 1979,1980
proposed a new configuration of vorticity probe, which should (besides all three
velocity components) be capable of measurements of velocity gradients. The
proposed design, containing three arrays with three hot-wires each, was developed
by the new research team over a period of ten years. Balint, Vukoslavevi and
Wallace 1987 and Vukoslavevi, Balint and Wallace 1989 performed the first
test-measurements of instant values of all three components of fluid vorticity
vectors using preliminary versions of the nine-wire vorticity probe. Its final design
was reported by Vukoslavevi, Wallace and Balint 1991. This was the first probe
capable of simultaneous measurements of all three components of velocity without
neglecting either any of velocity components or the velocity gradients inside of
measuring volume.
Besides this team (Department of mechanical engineering at the University of
Maryland, USA), a few other research groups have also used probes with nine or
twelve hot-wires in various flow configurations (see Tsinober, Kit and Dracos
1992, Honkan 1993 and Park and Wallace 1993). Finally, Vukoslavevi and
Wallace 1996 and Vukoslavevi and Petrovi 1998, reported clear advantages of
twelve-sensor probes in comparison to the configurations with nine hot-wires.
MULTIPLE HOT-WIRE PROBES 65
__________________________________________________________________________________________________________________________

They were the first who used a special algorithm, which takes into account the non-
uniformity of velocity field not only between arrays but also within each individual
array. The main purpose of probes with nine and twelve wires was to measure
vorticity. Therefore, detailed information about these probes is provided in the next
(5
-th
) chapter.
Due to chosen probe geometry and its finite size, the particular wires sense
different fluid velocities. Furthermore, each sensor averages the velocity over its
finite length. Both the finite probe dimensions and finite length of its wires may in
the worst case lead to complete misinterpretation and definitely alter the results of
turbulent velocity field measurements, especially for higher-moments of velocity
fluctuations (Rosemann 1989 and Pompeo and Thomann 1993). To reduce this
influence, designers try to minimise the probes, what is not always simply. Small
dimensions sometimes cause thermal and aerodynamic interference of probe
elements, short sensors suffer from non-uniform temperature distribution over their
length, because of end-conduction losses, etc.
Following Wyngaard 1968,1969 and knowing Kolmogorov micro-scale that
describes the smallest turbulence coherent structures, most of existing four-wire
probes seem to be fairly large for application in turbulent flows. Diameter of
measuring volume of the probe designed by Dobbeling, Lenze and Leuckel
1990a,b (fig. 4.18) is equal 2 mm. Samet and Einav 1987 (fig. 4.17) designed even
larger configuration of 2.5 mm in diameter.
To suppress the influence of mean velocity gradient, normal to the wall,
Pompeo and Thomann 1993 analysed a special quadruple arrangement sketched
in fig. 4.19. They decreased height of the probe measuring volume to 0.5 mm (that
may be acceptable in well-designed wind tunnels and adequate flow regimes),
increasing the width to fairly large value of about 2 mm.
Rosemann, Stager and Kreplin 1996 designed the probe of quadrate
geometry that generally minimises the sensing volume. However, the measuring
volume of their probe was still fairly large: a cube of (1 mm)
3
. Unfortunately, none
of these designers provided reliable information on Kolmogorov micro-scale at
relevant flow locations, preventing any estimation of probe spatial resolution.
Some other designers reported smaller probes. Typical examples, available at
the University of Montenegro in Podgorica, are presented in fig. 4.4. Measuring
volume of the probe VP-4+ is 1 mm in diameter and 0.5 mm in length, while
VP-4q is even smaller: a cube of only (0.5 mm)
3
, giving a front-side diameter of
only 0.7 mm. Both probes posses 2.5 m tungsten wires. If thinner commercially
available sensors would be used, current technology will enable further
minimisation of their dimensions (probes with even twice smaller diameters of the
sensing volume can be produced).
Besides VP-4q, AUSPEX configuration AVOP-4-100, used by Wittmer,
Devenport and Zsoldos 1998, is also a miniature four-wire probe. Although
66 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
equipped with tungsten wires of 5.0 m (instead of 2.5 m in the case of VP-4q) its
outer dimensions are equal to VP-4q, what decreases the probe sensitivity to fluid
velocity gradients in the measuring volume. Therefore, suitably configured sensors
in a small measuring volume of quadrate probes appear to overcome some of the
gradient error problems associated with triple-wire probes (as it was reported by
Devenport, Glegg and Sharma 1992).
However, it seems that the newest quintuple probe of Holzapfel, Lenze and
Leuckel 1994, specified for velocity field measurement, currently represents the
best available choice for highly turbulent flows. This configuration provides the
highest accuracy and the largest uniqueness domain, having the same dimension as
four-sensor probe. Diameter of its measuring volume is only 1.1 mm and length is
around 0.5 mm, with 2.5 m tungsten sensors. Unfortunately, Holzapfel, Lenze
and Leuckel 1994 did not provide information of possible further probe
minimising if thinner hot-wires would be used. Upper limit for the half-angle of
uniqueness cone of this probe configuration, declared to be 90
0
, is also
questionable and demands an additional verification.
Besides overall dimensions of the probe measuring volume, its configuration
(geometry) also influences the operational characteristics. Among them, very
important is the probe response to the instantaneous fluid velocity vector, usually
expressed by the size of uniqueness domain (most commonly described by
so-called uniqueness cone) and probe angular resolution. Although a variety of
adequate papers exists, it seems that the most illustrative are works of Rosemann
1989, Lekakis, Adrian and Jones 1989 and Rosemann, Stager and Kreplin 1996.
They found orthogonal geometry to be optimal for triple probes. It compromises
the size of uniqueness domain and probe overall angular resolution.
In the case of quadruple probes, Rosemann, Stager and Kreplin 1996
reported wire angle o = 45
0
(see fig. 4.20) toward the probe axis as optimal. The
uniqueness domains of four-wire probes are generally larger than domains of
triple hot-wire probes. They are less affected by finite wire length and provide
higher measurement accuracy of turbulence quantities than triples. Whether with
three or four sensors, sharper probes (with smaller angle o) exhibit narrower
uniqueness range and increased angular sensitivity toward changes of the
instantaneous flow direction.
Therefore, for a given value of the maximum turbulence intensity in a flow of
interest, an optimal probe geometry should have the smallest hot-wire angle o,
which still guarantee that specified percentage of all velocity vectors (events) are
within the uniqueness domain. This guideline compromises probes directional
sensitivity and percentage of fluid velocity vectors, which directions are within the
uniqueness cone.
Figs. 4.21 and 4.22, undertaken from Rosemann, Stager and Kreplin 1996, are
very illustrative. They represents the uniqueness domains for various triple and
MULTIPLE HOT-WIRE PROBES 67
__________________________________________________________________________________________________________________________

[a] [b]
[c] [d]
[e] [f]
Fig. 4.21: Lines U
E,2
/U
E1
=tano
2
=const. ------ and U
E,3
/U
E1
=tano
3
=const. - - - for
orthogonal "triple" probes. Borders of uniqueness domain are marked
by tick solid lines. Source: Rosemann, Stager and Kreplin 1996.
[a] o =-35.26
0
; k=0.00; h=1.0; [b] o =-35.26
0
; k=0.04; h=1.0;
[c] o =-40.00
0
; k=0.00; h=1.0; [d] o =-30.00
0
; k=0.00; h=1.0;
[e] o =-35.26
0
; k=0.00; h=1.4; [f] o =-35.26
0
; k=0.00; h=0.7.
68 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
quadruple (respectively) hot-wire probes, at different wire angles in the direction
space of pitch and jaw u angles. In addition, these figures confirm previous
conclusions about the influence of hot-wire inclination angle o toward the probe
axis and number of sensors on the measurement accuracy and especially on the
uniqueness range of the solutions of hot-wire response equations.



[a] [b]
[c] [d]
Fig. 4.22: Lines U
E,3
/U
E1
= tano
1
= const. ------ and U
E,4
/U
E2
= tano
2
= const. - - -
for "quadruple" (four-wire) probes. Borders of the uniqueness
domains are marked with solid tick lines.
[a] o =+/-45
0
; k=0; h=1.0; [b] o =+/-45
0
; k=0; h=1.4;
[c] o =+/-60
0
; k=0; h=1.0; [d] o =+/-30
0
; k=0; h=1.0.
Source: Rosemann, Stager and Kreplin 1996.
MULTIPLE HOT-WIRE PROBES 69
__________________________________________________________________________________________________________________________

4.7 SPECIAL HOT-WIRE CONFIGURATIONS THE EIGHT-SENSOR PROBE
Results presented in the previous section show that the influences of hot-wire
probe configuration and dimensions on the measurement results of the turbulent
velocity field have not been
completely resolved yet. The
main important question is:
"How to ensure high accuracy
of turbulent velocity field
measurement in the flows with
high velocity gradients and high
turbulence levels?" It is evident
that the probe with at least four
sensors should be used, because
of its possibility to measure all
three components of fluid
velocity vector through the
fairly wide uniqueness cone.
In order to minimise the
influence of velocity field non-
homogeneity over the
measuring volume, the
quadrate configuration VP-4q
should be used, instead of
plus probe VP-4+ (for
illustration see fig. 4.23).
Unfortunately, the four-wire
probes do not provide
information on the gradients of
velocity field within the probe
sensing volume and its
influence can not be taken into
account. This question reaches
one of the topics of
contemporary fluid mechanics. Currently, a few general approaches to estimate the
probe spatial resolution are available.
The most common is based on the estimation of Kolmogorov micro-scale and
criteria of Wyngaard 1969 (explained in the next chapter). However, this scale is
unknown before the experiment is performed (finished), because it has to be
estimated on the base of experimental results. Thus, it seems that researcher runs in
some kind of circle, which represents the relationships between dimensions of
hot-wire probe measuring volume and corresponding experimental results, based
(again) on the geometry and dimensions of the probe sensing volume.
(a)

D
+
~ 1mm
D
q
~ 0.7mm
VP-4+
VP-4q
(b)
(c)
VP-4+
D
+
= 1mm

VP-4q
D
q
~ 0.7mm
(d)
Fig. 4.23: A sketch defining front sensing areas
of the plus and quadrate
configuration, assuming hot-wires of
length L = 0.7 mm and geometrical
angle o = 45
0
toward the probe axis:
(a), (c) - plus probe VP-4+ and
(b), (d) - quadrate probe VP-4q.
70 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
The other approach of resolving the problem of velocity gradients in the sensing
volume assumes the application of the newest nine-wire or twelve-sensor probe.
These hot-wire configurations, if designed with properly distanced arrays, can be
used to measure not only the all three fluid velocity components, but also the
corresponding velocity gradients. In this case, the results of velocity measurement
by four-wire configuration can be treated as reliable if the differences between
statistic parameters, measured by vorticity probe WP-12+(G) and its arrays,
numerically treated as independent probes VP-4+, are negligible.
Nevertheless, the basic purpose of the probe WP-12+(G) is not to measure
velocity. It is designed mainly for vorticity measurement. In addition, WP-12+(G)
possesses greater number of sensors, demanding additional measurement channels
of supporting electronic devices, in comparison to four sensor probes. Furthermore,
sensing area (and volume, obviously) of the vorticity anemometer probe WP-
12+(G) is larger in comparison to VP-4+. Although it does not need neglecting of
velocity gradients, the new hypothesis has to be applied that assumes constant
gradients in the probe sensing volume, what introduces new very complex
algorithm for output signals interpretation. In opposite, four-sensor probes "use"
less acceptable assumption about uniform velocity in its own measuring volume,
but the diameter of the measuring volume of the probe VP-4+ is about 2.5 times
smaller in comparison to vorticity configuration WP-12+(G).
The third method of testing the influence of velocity gradients over the probe
sensing volume on the measurement results of turbulent velocity field is reported
by Vukoslavevi and Petrovi 1996. It is based on the idea to use simultaneously
two probes of different dimensions but the same geometry, what enables
application of the same output signals interpretation procedure. Therefore, the
influence of probe-calibration and signal-interpretation procedure, as well as the
probe configuration, would disappear.
Following this concept,
the unique probe VP-8qd
was designed and tested
(see figs. 4.24 and 4.25). It
consists of two coaxial
quadrate hot-wire
arrays configured as probe
VP-4q, placed one inside
other. As can be seen from
fig. 4.25, a typical
dimension of the smaller
array is half the size of the
outer one. This gives
significantly smaller area
of the sensing cross-
VP-8qd
1.00
0.70
0.36
0.50

Fig. 4.24: A sketch of the probe VP-8qd (left).
Fig. 4.25: End-view of the probe VP-8qd (right).
Source: Vukoslavevi and Petrovi 1996.
MULTIPLE HOT-WIRE PROBES 71
__________________________________________________________________________________________________________________________

section and smaller the measuring volume. The criterion for estimating the
influence of velocity gradients is as follows. If the differences between results from
two arrays are within desired limits, spatial resolution of the inner (smaller) array is
correct and measurement results are accepted.
Both arrays, configured as the VP-4q probe, are very similar to modified type of
Kovasznay 1950 probe, designed by Vukoslavevi 1981. The sensors are
positioned at angles 45
0
toward probe axis. Eight tungsten 2.5 m wires are placed
on stainless steal prongs, 0.25 mm in diameter, tapered to about 40 m on their
tips. The fabrication of the probe is similar to
vorticity probe WP-12+(G).

Fig. 4.27: A special coordinate microscope.
Turbulence flows laboratory (TFL)
at the University of Montenegro.
As it can be seen in fig. 4.26, the special
multiple hot-wire probe VP-8qd enables
simultaneous examination of the quadrate
configuration VP-4q response, as well as X
probes in both vertical (VP-2x/V) and
horizontal plane (VP-2x/H). Typical
dimensions of the front sensing areas of these
configurations are shown in fig. 4.25.

1
3
4
2
VP-4q

1
3
VP-2x/V

4
2
VP-2x/H

Fig. 4.26: Simulation sketches of
hot-wire probes with
four (VP-4q) and two
sensors, placed in the
vertical (VP-2x/V) and
horizontal plane (VP-
2x/H). Adapted from:
Vukoslavevi and
Petrovi 1996.
72 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Hot-wire probe VP-8qd
was manufactured and tested
in the Turbulence flow
laboratory (TFL) at the
University of Montenegro in
Podgorica. It possesses
sophisticated optical,
mechanical and electronic
devices, specified for
manufacturing, testing and
repairing hot-wire probes
(figs. 4.27 4.30). Using
special coordinate traversing
microscope (fig. 4.27) and
electronic welder with stereo
microscope (fig. 4.28), 16
prongs and 8 wires were
accurately positioned in the
miniature probe measuring
volume of only 0.5 mm
3
(1 mm
x 1 mm x 0.5 mm).
The probe VP-8qd was
calibrated in the potential core
of the free round jet (fig. 4.29),
of the open wind tunnel (fig.
4.30) that provided flat (hat-
type) exit velocity profile, with
turbulence level on the centre-
line below 0.5%. Desired
magnitude of the nozzle exit
velocity profile was maintained
constant within the limits of
0.1% by electronically
controlled fan rotation rate.
All measurements were
performed by simultaneous
sampling of eight output
voltages of hot-wire
anemometer with built-in
bucking-up amplifier (fig.
4.31), designed and
manufactured at the University
of Montenegro, TFL Podgorica.

Fig. 4.28: Electronic hot-wire welder with stereo
microscope. TFL Podgorica.


Fig. 4.29: Special mechanism for angular
calibration of hot-wire probes,
attached to the round nozzle producing
uniform and nominally zero-level
turbulence flow. TFL Podgorica.
MULTIPLE HOT-WIRE PROBES 73
__________________________________________________________________________________________________________________________

Angular calibration was performed by pitching and jawing the probe, placed in
the suitable calibration mechanism (see fig. 4.29). This procedure has been
repeated for various combinations of jaw and pitch angles, until the optimal
number of n different three-dimensional flow realisations was provided. This
known flow realisations and measured anemometer output voltages yield a system
of linear algebraic equations, which can be used to determine unknown calibration
coefficients.


Fig. 4.30: Open wind tunnel. TFL Podgorica.

Fig. 4.31: Hot-wire anemometer and two-channel oscilloscope. TFL Podgorica.
Influence of probe dimensions, i.e. velocity gradients over the sensing volume,
on the measurement results of the real turbulent velocity field was only preliminary
tested. Vukoslavevi and Petrovi 1996 placed the probe in the axis of three
turbulent wakes, 130mm behind the cylinders of 2.0mm (CYL1), 2.2mm (CYL2)
and 3.5mm (CYL3) in diameter. Some differences between corresponding results of
inner/smaller (I) and outer/larger (O) array were evidenced for all three flows,
generated by these three cylinders (even for mean velocity).
74 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
However, as usual,
they were largest for
higher-order velocity
field statistics. For
illustration, the
skewness S
U
and
flatness F
U
factors of
the longitudinal
component U are
presented in fig. 4.32.
It is obvious that their
absolute values are
larger, if measured
with the inner/smaller
array, in comparison to
the outer. These
differences represent
the influences of fluid
velocity gradients over
the probe sensing
volume on the
measurement results of
turbulence velocity
field statistics. They
originate from the
artificial spatial
averaging of
fluctuations of velocity
components by the
outer/larger array, in
contrast to the inner
array which front-view
sensing area is about four times smaller (see fig. 4.25).
The influences of neglected fluid velocity components on the measured values
of turbulence statistical parameters were examined by analysing existing
differences between results originating from different hot-wire configurations
(VP-4q, VP-2x/H and VP-2x/V) of the same size (i.e. array). These differences are
also illustrated in fig. 4.32 for skewness S
U
and flatness F
U
factors of the
longitudinal velocity. In such a way, a superiority of quadrate geometry that
does not demands neglecting of any velocity component, toward two-sensor VP-
2x/V and VP-2x/H probes, was simply verified.
Furthermore, differences between measurement results of the inner and outer
array of the same configurations (four-sensor toward four-wire geometry or two-
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
CYL1 CYL2 CYL3
S
U

(
-
)


0.0
1.0
2.0
3.0
4.0
5.0
4q-I 4q-O 2x/V-I 2x/V-O 2x/H-I 2x/H-O
CYL1 CYL2 CYL3
F
U

(
-
)


Fig. 4.32: Skewness S
U
and flatness F
U
factors of the
longitudinal U velocity in the wakes behind
three cylinders (CYL1, CYL2, CYL3),
measured by VP-4q probes and two-wire
probes VP-2x/V and VP-2x/H placed in the
vertical and horizontal plane, respectively.
Source: Vukoslavevi and Petrovi 1996.
MULTIPLE HOT-WIRE PROBES 75
__________________________________________________________________________________________________________________________

wire toward two-sensor configuration) are nearly equal to differences between vari-
ous configurations on the same array (two-sensor probe VP-2x toward four-wire
quadrate configuration VP-4q), where their dimensions are equal. It follows that,
in this case, the influences of velocity gradients in the measuring volume of hot-
wire probes are of the same order as the influences of neglected fluid velocity
vector component.
An indirect proof of the reliability of these results can be found after careful
analysis of the values of skewness S
U
and flatness F
U
factors that correspond to
three applied cylinders. It is evident that higher deviations of their values from the
values of Gaussian distribution correspond to the first (the largest) cylinder. This
was expected and is quite logical because this cylinder generates coherent
structures of the largest scales (in the flow region of interest), which represent the
main source of turbulence non-Gaussian properties. In contrast, in the case of the
smallest cylinder (no. 3) these deviations were the smallest.
Although these results represent only a preceding information, unique multiple
sensor probe VP-8qd with eight wires enabled simultaneous testing of two various
types of hot-wire geometry in two different dimensions. It is shown that especially
higher-order statistical moments are sensitive not only to neglecting the one
velocity component (X-probe) but also to velocity gradients in the measuring
volumes of two and four-wire probes.
However, it is well known that not only the configuration and size of hot-wire
probe influences the measurement results, but also the flow regime and cross-
section dimensions of
the wind tunnel.
Influence of flow
conditions (i.e. strong
favourable or adverse
pressure gradient) on
the length scales of
vortex structures and
development of a
turbulent boundary
layer is illustrated in
fig. 4.33.
In order to enhance
spatial resolution of
measurement, design
of wind tunnel
characterised by large
cross-section area and
application of low-
velocity flows are


Fig. 4.33: Influence of pressure gradient on turbulent
vortexes development in a boundary layer:
(a) strong favourable pressure gradient (up);
(b) strong adverse pressure gradient (down).
Source: Head 1982.
76 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
suitable, because they result in low gradients of fluid velocity components and
large coherent vortex structures (i.e. large micro-scale q of Kolmogorov).
Unfortunately, slow turbulent flows are very complicated for measurement by hot-
wire probes, because of some heat-transfer phenomena typical for the low-
magnitude fluctuations of velocity vectors, which are explained in the section 7.5).
More details on the estimation of hot-wire probe spatial resolution in a specified
flow are provided in the 5
-th
chapter. Besides definition of the term vorticity, in
the section 5.2 is provided mathematical definition of the scale q, according to
Kolmogorov 1941. The most popular, less or more complex and accurate,
measurement techniques for its evaluation are also presented and discussed.
Influence of the flow regime is also illustrated in fig. 5.2, which clearly shows that
increasing the flow velocity does not increases the large-scale vortexes, while
significantly changing the smallest structures that control dissipation of the
turbulent kinetic energy.

Subsonic jet becoming turbulent. Source: Bradshaw, Ferris and Johnson 1964.










5 5. . M ME EA AS SU UR RE EM ME EN NT TS S O OF F T TU UR RB BU UL LE EN NT T V VO OR RT TI IC CI IT TY Y F FI IE EL LD DS S
B BY Y M MU UL LT TI IP PL LE E H HO OT T- -W WI IR RE E P PR RO OB BE ES S


M Ma an n i is s t th he e m me ea as su ur re e o of f a al ll l t th hi in ng gs s: : t th he e o on ne es s t th ha at t a ar re e, , t to o b be e s so o, , t th he e o on ne es s t th ha at t a ar re e
n no ot t, , n no ot t t to o b be e s so o. .
( (P Pr ro ot ta ag go or ra as s, , ~ ~ 4 48 81 1- -4 41 11 1 b be ef fo or re e t th he e N Ne ew w A Ag ge e) )


5.1 VORTICITY
The fundamental role that rotation, e.g. vorticity, of fluid particles plays in
turbulent flows has been emphasised by many authors during the past fifty years.
Vorticity is a defining property of turbulence. As stated in a variety of famous
textbooks and review papers, such as Hinze 1959, Tennekes and Lumley 1978,
Wallace and Foss 1995 and many others, its presence in a flow should be
considered essential to identifying it as a true turbulent motion. Mathematical
definition of vorticity assumes the curl of velocity vector
i
k
ijk i
x
U
c
c
= c O , (i,j,k = 1,2,3), (5.1)
which is equal to the twice of the mean angular velocity of the fluid particle. In
accordance to the tensor notation, used in (5.1), the term c
ijk
represents the
alternating tensor and
U
1
= U,
U
2
= V, (5.2)
U
3
= W,
are the scalar components of velocity vector
k W j V i U U + + =
0
(5.3)
in an orthogonal Cartesian (Oxyz) coordinate system. Therefore, the longitudinal
(streamwise) O
x
, transversal (normal) O
y
and lateral (spanwise) O
z
components of
vorticity vector in the same (Cartesian) coordinate system will be:
78 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
z
U
y
W
x
c
c

c
c
= O ,
x
W
z
U
y
c
c

c
c
= O , (5.4)
y
U
x
V
z
c
c

c
c
= O .
According to Panton 1984, the physical picture of vorticity can be understood if
one imagines that a spherical fluid particle is instantaneously frozen and, therefore,
starts behaving as a solid body. This frozen ball will then continue to move as a
result of previous motion. If this motion is separated into the translation and
rotation, the latter will occur at an angular speed of O
i
/2. This means that vorticity
can be interpreted more precisely: as just twice the rotation rate of fluid particle.
According to Wallace 1986, it can be alternatively interpreted as the circulation per
unit area of a surface perpendicular to the vorticity field.
Starting from the basic equation of fluid motion that describes the momentum
transport, well known Navier-Stokes equation
j j
i
i
i
j
i
j
i
x x
U
x
P
X
x
U
U
t
U
c c
c
v
c
c
c
c
c
c
2
1
+ = + , (5.5)
and differentiating its terms alternatively with c/cx
i
, after some simple
transformations the dynamics equation for vorticity transport
j j
i
j
i
i
j
i
j
i
x x x
U
x
U
t c c
O c
v
c
c
O
c
O c
c
O c
2
+ = + (5.6)
can be obtained. It is also known as the equation of fluid motion in vorticity form
and shows that the rate of change of a fluid particle vorticity is equal to the rate of
deformation of the vortex line passing through its centre, plus the rate of viscous
diffusion of its vorticity.
This equation is as important in Fluid mechanics as it is the basic equation of
fluid motion (5.5). However, unlikely to basic equation, it does not contain
pressure directly, showing that viscous forces generate rotation (vorticity) of fluid
particles. This makes sense because the pressure and gravitational forces, acting in
the centres of fluid particles, can not create their rotation. The absence of pressure
term in vorticity dynamics equation (5.6) provides some simplification in
computations and measurements.
Diffusion of vorticity, described by the last term in (5.6), is a relatively slow
process governed by the same diffusivity constant as that for momentum in (5.5):
kinematic viscosity v. In some cases, this property allows neglecting the diffusion
MULTIPLE HOT-WIRE PROBES 79
__________________________________________________________________________________________________________________________

terms in (5.6) and inviscid treatment of vorticity dynamics in viscous flows.
Consequently, theorems of Kelvin and Helmholtz, governing the inviscid motion of
vortices, can be applied. When viscous diffusion is validly neglected, vortex lines
can be considered as coincident to material lines that move with the fluid.
However, curvilinear motion of fluid particle does not necessarily indicate the
presence of vorticity and vice versa. Typical examples are ideal vortices with zero
vorticity, and rectilinear channel flow that has considerable level of vorticity
(Wallace 1986). Vorticity dynamics equation (5.6) is more suitable from the
experimental point of view too, because the measurements of pressure fluctuations
in turbulent flows seem to be even more difficult then vorticity measurements
(Wallace and Foss 1995).
In analogue to measurements of the terms in the equation that describes the
turbulent kinetic energy for steady mean flow, analysis of the balance of the terms
in the averaged transport equation
|
|
.
|

\
|
c
c

c
c

c c
|
|
.
|

\
|
c
+
c
c
=
|
|
.
|

\
|

c
c

j
i
j
i
j j
i
j
i
j i
i
j
j
x x x x x
U
x
U
O O
v
O
v O O
O
2
2
2
2
2
, (5.7)
[ [1 1] ] [ [2 2] ] [ [3 3] ] [ [4 4] ]
for a magnitude of vorticity, known as enstrophy ( ) 2
2
i
O , has been performed.
The four averaged terms indicated by overbars are:
[ [1 1] ] advection,
[ [2 2] ] rotation and stretching/compression by the velocity gradient field,
[ [3 3] ] viscous diffusion and
[ [4 4] ] viscous dissipation of total enstrophy.
They contain different combinations of terms in equation (5.6) for time-averaged
2
i i
O O and fluctuating 2
2
i
e enstrophy (see Balint, Vukoslavevi and
Wallace 1990).
Recording the turbulent flow field by a moving camera, Niquradse 1929
noticed more than 60 year ago that the vortex structures of different sizes and
intensities make the base of turbulent flow field. In fact, it seems that Leonardo da
Vinci (1452-1519) was the first systematic observer of turbulent flows. His
sketches of the vortex formation in separation zones, velocity distribution in a
vortex and water exiting from a hole into a pool (fig. 5.1) are masterfully accurate.
Later on, fast development of various sophisticated techniques and scientific
instruments intensified the experimental research. Consequently, it became
recognised worldwide that knowledge of vorticity dynamics is essential to a more
complete understanding of the process that governs turbulent shear flows. Lighthill
1963 makes this convincingly clear. Wllmarth 1975, in his survey of turbulent
boundary layer research, says: "Turbulent structure can not be easily understood
80 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
without the concept of vorticity".
Laufer 1975 stated that "vorticity
measurements suggest themselves as
the most promising method for the
quantitative study of turbulent
motion".
The occurrence of coherent vortex
structures in different types of
turbulent flows has been detected by
many visual studies worldwide. Van
Duke 1982 presented very detailed
information on the existence of vortex
structures in turbulent flows. His
album of fluid motion is one of the
most beautiful books in the area of
Fluid mechanics, which provides
clear and reliable proofs for vortex
structure of turbulence. Some of these
pictures are included in the present book.
Experiments clearly showed, even more than twenty years ago, that turbulent
motion, far of being chaotic, contains coherent structures as the basic flow
elements in the near-wall region of bounded flows. Far from the wall, large-scale
vortexes have been detected and studied as possible source of noise. Coherent
structures appear randomly in space and time. Therefore, it is very difficult to
detect them by simple measurements at a given point of flow field because the
probe is exposed alternatively to coherent and random fluid motion. So it was
necessary to develop a measurement method and instrument, capable of direct
simultaneous measurements of all three components of fluid vorticity vectors.
This has been an elusive goal of experimental turbulence research for a long
time. Among different techniques and methods that have been used, hot-wire
anemometry has played the most important role in vorticity measurements.
Detailed reviews of these techniques, as well as the other methods of vorticity
measurements, are given by Wallace 1986, Foss and Wallace 1989 and more
recently by Wallace and Foss 1995. In addition, Bruun 1995 provided up-dated
information on hot-wire techniques specified for vorticity measurements.

5.2 SPATIAL RESOLUTION OF THE PROBES
One of the main problems that appear in vorticity measurements (as well as in
any measurement of fluid velocity components and their gradients), whether hot-
wire anemometry or other available techniques are used, is related to spatial
resolution of measuring instrument applied. In order to measure velocity at a given

Fig. 5.1: A sketch of water exiting from a
square hole into a pool, as seen
by Leonardo da Vinci ~1500.
Source: Gad-el-Hak 1998.
MULTIPLE HOT-WIRE PROBES 81
__________________________________________________________________________________________________________________________

point of turbulent flow field, it is necessary to have the probe of as small dimension
as possible, and (in order to measure velocity gradients) at least two probes placed
in a small distance from each other. The maximum allowed distance, over which
the smallest turbulence structure can be measured, has been in focus of interest of
many experimentalists.
A basic and unavoidable parameter of nearly all experimental as well as
theoretical studies on turbulent flows is micro-scale of Kolmogorov 1941. Among
other important roles, it represents a crucial parameter for estimation of hot-wire
probe spatial resolution and is defined as
4 1
3
|
|
.
|

\
|

c
v
q , (5.8)
where v is the kinematic fluid viscosity, while c designates the mean (time
averaged) kinetic turbulent dissipation rate. This scale indicates the smallest
coherent vortex structures encountered in the turbulent flow characterised by the
values of v and c . Therefore, determining the Kolmogorov scale (5.8) requests
values of these two flow characteristics. Reliable and fairly simple measurement
techniques for kinematic viscosity still exists, while the turbulent dissipation is
very difficult to measure directly, even in the best world laboratories.
According to Hinze 1975 and Browne, Antonia and Shah 1987, turbulent
kinetic energy dissipation per unit time per unit mass can be expressed as
i
j
i
j
j
i
x
u
x
u
x
u
c
c

|
|
.
|

\
|
c
c
+
c
c
=v c , (5.9)
using standard Cartesian tensor notation and summation convention. Applying the
more usual notation for velocity fluctuations (u, v, w instead of u
1
, u
2
, u
3
) and
coordinate axes (x, y, z instead of x
1
, x
2
, x
3
), the following expression arises:

+ |
.
|

\
|
c
c
+ |
.
|

\
|
c
c
+ |
.
|

\
|
c
c
=
2 2 2
2
x
w
x
v
x
u
v c +
|
|
.
|

\
|
c
c
+
|
|
.
|

\
|
c
c
+
|
|
.
|

\
|
c
c
2 2 2
2
y
w
y
v
y
u

[ [1 1] ] [ [2 2] ] [ [3 3] ] [ [4 4] ] [ [5 5] ] [ [6 6] ]
+ |
.
|

\
|
c
c
+ |
.
|

\
|
c
c
+ |
.
|

\
|
c
c
+
2 2 2
2
z
w
z
v
z
u

[ [7 7] ] [ [8 8] ] [ [9 9] ]

|
|
.
|

\
|
c
c
|
.
|

\
|
c
c
+ |
.
|

\
|
c
c
|
.
|

\
|
c
c
+ |
.
|

\
|
c
c

|
|
.
|

\
|
c
c
+
y
w
z
v
x
w
z
u
x
v
y
u
2 2 2 (5.10)
[ [1 10 0] ] [ [1 11 1] ] [ [1 12 2] ]
82 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Having in mind the complexity of expression (5.10), the value of c obtained using
all twelve terms is most commonly designated as total dissipation. Evaluation of
terms [ [1 1] ]- -[ [1 12 2] ] requires instantaneous values of nine gradients of fluctuating fluid
velocity vector components: cu/cx, cu/cy, cu/cz, cv/cx, cv/cy, cv/cz, cw/cx, cw/cy
and cw/cz. They can be provided experimentally. However, measurement of all
terms [ [1 1] ]- -[ [1 12 2] ] is, even nowadays, very complex and time-consuming task that
demands sophisticated instruments, signal interpretation procedures and very
skilful experimentalists. In order to resolve a problem related primarily to
experimental technique, most of previous studies have treated c by making less or
more severe (and adequate) assumptions that enable neglecting some of the terms
in (5.10) or its suitable rearrangement.
The simplest one, but the least accurate, formula can be derived assuming
isotropic turbulence, as explained by Browne, Antonia and Shah 1987. However,
Taylor 1935 was the first who obtained isotropic value
I
c for c :
2
15 |
.
|

\
|
c
c
=
x
u
I
v c . (5.11)
Having in mind that assumption of isotropy is strictly valid only in extremely small
number of real situations, such is a grid decaying turbulence for example, this
approach is not a reliable one, and can be applied only for preliminary analysis of
the probe spatial resolution.
The first three quantities in (5.10), [ [1 1] ], [ [2 2] ] and [ [3 3] ], can be derived from a time
derivatives of velocity components u, v and w, e.g. cu/ct, cv/ct, cw/ct, using well
known and the simplest form of Taylors (frozen field) hypothesis
t
U
x c
c

c
c 1
, (5.12)
where U is the local mean velocity. However, the validity of this hypothesis is
very questionable in most of practical applications. Many experimental studies
have shown that its applicability is severely restricted to special situations, which
are very rare in the experimental practice. Unfortunately, a more accurate and more
general solution has not been found yet, demanding application of (5.12) even in
the occasions when it does not guarantee the highest possible measurement
accuracy. Term [ [1 1] ] can be obtained from a single normal-wire probe, while the
other two require one X or V probe with two sensors. However, it should be kept in
mind that all restrictions of the single-wire and X/V probes applicability,
originating from their measurement principles, are present here too. These probes
demand neglecting of two (single normal-wire probe, see VP-1n in fig. 4.1) and
one fluid velocity components (X and V probes VP-2x and VP-2v, see fig. 4.2).
Practically, this measurement technique provides high accuracy only in the flows
with low turbulence-level, up to around 5 - 7 % (or 10 % maximally).
MULTIPLE HOT-WIRE PROBES 83
__________________________________________________________________________________________________________________________

In contrast to the first three quantities, the term [ [4 4] ] and [ [7 7] ] can be measured
independently from Taylors hypothesis. In this case, two parallel single normal-
wire probes are needed. If a quantity ( )
2
y / u c c is to be measured, the probes are
separated in the direction of y axis, while for ( )
2
z / u c c this separation is along
the z axis. In this case, the same restrictions of single-wire probes are also
present. In addition, a new question arises about the optimal distance between the
single normal-wire probes and other possible errors, as it is discussed by
Antonia and Browne 1986.
Measurement of term no. [ [5 5] ], [ [6 6] ], [ [8 8] ], [ [9 9] ], [ [1 10 0] ] and [ [1 11 1] ] requires the use of two
X or V probes placed close to each other in the parallel planes. They has to be
separated along the y axis if the quantities ( )
2
y / v c c , ( )
2
y / w c c and
( ) ( ) x / v y / u c c c c are measured. In contrast, measurement of ( )
2
z / v c c ,
( )
2
z / w c c and ( ) ( ) x / w z / u c c c c can be performed by the probes separated along
the z axis. All restrictions, obviously present in experiments employing
configurations with two wires, exist here too.
The last term in expression (5.10) can be measured using four X/V probes, as
well as with two "triple" probes. However, this is practically very complicated to
perform, especially to meet the criterion of spatial resolution of measurement.
Presented measurement techniques for various quantities included in (5.12)
clearly show that neglecting of some terms of c may decrease the complexity of
the experiment needed for their measurement. Consequently, the accuracy of
evaluating c simultaneously decreases. Therefore, some compromise between the
measurement accuracy and complexity of required experimental technique is
needed. Having in mind that this textbook is focused to the probes capable of
accurate simultaneous measurements of all terms included in c in higher
turbulence-level flows, these problems are not discussed here. However, a
voluminous scientific literature, concerned to problems related to measurements of
velocity gradients and turbulent dissipation by one or more simply configured
probes containing single-normal wire or two sensors, is available. Typical
examples, published in the past fifteen years are: Antonia, Brown and Chambers
1984, Antonia, Anselmet and Chambers 1986, Browne, Antonia and Shah 1987,
Antonia, Shah and Browne 1987, Kida and Murakami 1989, Antonia and Kim
1994, Antonia, Zhu and Kim 1994, Jovanovi, Ye and Durst 1995, Ewing,
Hussein and George 1995, Ye 1996, Antonia, Zhou and Zhu 1998, Schenck,
Petrovi, Jovanovi and Durst 1998a,b, etc.
Besides the measurement technique of Browne, Antonia and Shah 1987, more
sophisticated methods and special multiple hot-wire probes were recently
developed. They enable simultaneous measurements of all twelve terms in (5.12),
84 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
even in the flows of higher turbulence levels. Such probes enable instantaneous
three-dimensional measurements of fluid velocity components at three locations
closely separated along the y and z axis, which form a plane perpendicular to
probe axis. They are described in further sections of this chapter.
Turbulent fluid
motion is very
complex, containing
contributions from a
wide range of length
scales (see fig. 5.2).
The largest coherent
vortex structures,
which length scale
is obviously of the
order of the flow
configuration size,
defined by the size
of wind tunnel (test
facility), contain
most of the kinetic
energy and perform
most of the
turbulent transport.
In contrast, the smallest structures that are represented by the scale q of
Kolmogorov 1941 are responsible for dissipation of the turbulent kinetic energy of
fluid motion into internal fluid energy.
A hot-wire can not effectively resolve motions of turbulent coherent structures,
which scales are smaller from its length (Wyngaard 1968). In order to resolve a
non-uniform velocity field of a vortex structure, hot-wire obviously must be much
shorter in comparison to spatial scale of that structure. In most of practical
applications, single-wire probe is capable of (under usual restrictions connected
with neglecting two components of fluid velocity vector) accurate resolving the
large-scale coherent structures that occur in turbulent flows. In contrast, this is not
always easy to achieve for the smallest structures. Whether the probe is capable of
resolving the smallest vortex structures in the flow or not depends on the ratio of
hot-wire length toward the micro-scale q of Kolmogorov 1941.
This inherent limitation becomes more severely expressed if the multiple hot-
wire probes are used to measure statistics of turbulence at nominally one point in
the space. The problem additionally increases in the laboratory conditions, where
large Reynolds numbers can be usually obtained only by increasing the flow speed.
Consequently, typical scale q of the smallest structures become extremely small
and velocity fields of such structures can be hardly measured with high accuracy.
Typical illustration of this phenomenon is illustrated in fig. 5.2: doubling the


Fig. 5.2: Reynolds number influence on the structure of
turbulent mixing layer of nitrogen flowing above
and helium-argon mixture flowing below at the
same density. Photograph below corresponds to
double value of Reynolds number in comparison to
the photograph above. Source: Konrad 1976 and
Brown and Roshko 1974.
MULTIPLE HOT-WIRE PROBES 85
__________________________________________________________________________________________________________________________

Reynolds number by increasing the pressure (and velocity) does not significantly
influence the large-scale structures, while strongly influence the smallest (turbulent
energy dissipating) structures in the flow.
In order to provide more clear insight in this phenomenon, Frenkiel 1948,
Uberoi and Kovasznay 1953, Silverman 1968, Wyngaard 1968,1969,1971,
Roberts 1973, Browne, Antonia and Chambers 1983 and others, have extensively
studied the spectral transfer characteristics of various hot-wire probes. The
contributions of Wyngaard 1968,1969,1971 represent a framework for estimation
of the spectral signal attenuation by finite-size hot-wire probes, caused by spatial
resolution effects.
However, in the area of turbulent velocity and vorticity fields measurements,
the most popular and frequently cited is the paper of Wyngaard 1969. To obtain
numerical results, he assumed isotropy of turbulence and three-dimensional
spectrum of velocity according to
Pao 1965.
He studied a vorticity meter, well-
known probe of Kovasznay 1954
sketched in fig. 5.3(a). In order to
simplify the problem, an ideal
geometrical summetry of the probe
was assumed, e.g.:
1. all hot-wires have the same length
L and the same (geometrical)
angle o toward the probe axis,
coincident to the mean flow
direction;
2. the wire prongs are in the corners
of a square of side h
P
;
3. diagonally opposite wire supports
are of the equal length, one set
being shorter than the other two
(the latter hypothesis is valid for
each quadrate hot-wire
configuration);
4. probe is placed in a low
turbulence-level flow (well-
known restriction of applicability
of Kovasznay 1954 probe), where
each wire responds to the
component of fluid velocity vector
normal to its axis (the cosine
law (3.57)).
y
z
x
U
0
h
S

L
y
z
x
U
0
h
P

L
o
(a)
(b)
h
P


Fig. 5.3: Sketches of hot-wire probes
analysed by Wyngaard 1969:
(a) quadrate hot-wire probe of
Kovasznay 1954;
(b) probe with two parallel
normal hot-wires, specified
for measurement of fluid
velocity gradients.
Adapted from: Wyngaard 1969.
86 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Under explained simplifications, the probe spectral vorticity response is
analysed for different L, h
P
, o and q. It was shown that half the contributions to
2
x
O comes from the vortex structures with scales greater than 10q, what means
that a Kovasznay-type probe with L = h
P
and sensors of length L > 10q will
measure only half of the real value of
2
x
O . Furthermore, the ratio of measured and
true
2
x
O decreases sharply as q/L decreases. It is essentially 1.0 for q/L = 1, drops
to 0.86 for q/L = 0.32 and after this limit serious losses occur at smaller values of
q/L. He concluded that a value of q/L should not be smaller than 0.3, as later
confirmed Browne et al 1988.
Wyngaard 1969 also found that ratio h
P
/L has strong influence on the
measurement results. The analysis was performed for a fixed value of q/L = 0.32.
The minimum value for assumed geometry, h
P
/L = 0.707, provides the best probe
response, while the wider sensor spacing results in sharp loss of spatial resolution.
The effect of changing wire angle toward the mean flow on the value of
2
x
O is
rather weak. For a fixed ratios h
P
/L = 1.0 and q/L = 0.32, changing the o between
30
0
, 45
0
and 60
0
practically has no effect for the longitudinal (streamwise)
normalised wave numbers below 0.1. However, the differences occur and slowly
increase, reaching about 10% for normalised wave number of 1.0.
Wyngaard 1969 also analysed the spatial resolution of gradient probe (fig.
5.3b), considering two parallel wires of length L, with separation h
S
. He showed
that wires lengths should not exceed L = 4q, for h
S
/L = 0.25, in order to provide a
90% response to the quantities ( )
2
x u c c and ( )
2
y u c c . The restrictions are much
more severe if X probes are used for measurements of ( )
2
x v c c : for a 90%
response, wires must not be longer than around 2.5q for h
S
/L = 0.5.
Many other researches have also examined the parameters relevant for hot-wire
probe spatial resolution and possible consequent hot-wire signal attenuation.
Klewicki and Falco 1990 used hot-wire probe that enables systematic variation of
a separation between the two parallel sensors. The examination showed, for a
boundary layer at a Reynolds number
1010
0
=

=
v
u
u
U
Re (5.13)
based on the momentum thickness u, that just above the buffer layer the variance of
the velocity gradient ( )
2
y / u c c is attenuated by 5 % for h
S
/q ~ 3.3. However, they
also found rapid increasing of this attenuation, to about 35%, for h
S
/q ~ 5.9.
Antonia, Zhu and Kim 1993 performed a test of Wyngaards 1969 parallel-
sensor probe analysis. They used data of Kim, Moin and Moser 1987 and Kim
MULTIPLE HOT-WIRE PROBES 87
__________________________________________________________________________________________________________________________

1989, originating from direct numerical simulations (DNS) of turbulent channel
flow at centreline, with Re
t
=180 and 400 respectively:
v
t
t
b u
Re

= , (5.14)
where t
t
=
w
u (5.15)
is the friction velocity, t
w
represents wall shear stress, b is the channel half-width
and is the fluid density.
(
A
u
/
A
y
)
2


(
c
u
/
c
y
)

2

Ay
*
Ay/q
0 2 4 6 8 10 12
0.6
0.8
1.0
0.4

Fig. 5.4: Attenuation of measured fluid velocity gradients, as a function of the
hot wires separation Ay
*
= Ay/q = h
S
/q (Antonia, Zhy and Kim 1993):
A - DNS at Re
t
= 180;
- DNS at Re
t
= 400;
. . . . .
- analysis of Wyngaard 1969 with Paos 1965 spectrum;
____
- analysis of Wyngaard 1969 with DNS spectrum;
- - - - - analysis of Husain & George 1990 with Paos 1965 spectrum.
Adapted from: Wallace and Foss 1995.

They reported (fig. 5.4) decreasing of the ratio between the measured variance
( )
2
y / u A A
(assuming finite probe separation Ay = h
S
) and measured variance of
fluid velocity gradient
( )
2
y / u c c
with increasing the wire separation Ay = h
S
, as it
was qualitatively predicted by Wyngaard 1969 and experimentally verified by
Klewicki and Falco 1990. Furthermore, it was noticed that analysis of Wyngaard
1969, based on the assumptions of local isotropy and spectrum of Pao 1965, agree
well with DNS data only for h
S
/q < 2.0. However, Wyngaard 1969 stated that his
results strongly depend on the assumed form of three-dimensional spectrum.
Following his statement, Antonia, Zhu and Kim 1993 repeated the calculation of
88 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Wyngaard 1969 using the spectrum from DNS itself this time, and achieved quite
good agreement in the whole range of h
S
/q.
Most of presented results are related to measurement of fluid vorticity and
velocity gradients by the specified probe configurations. Therefore, for arbitrary
hot-wire configurations different from those of Wyngaard 1969, they can be used
only as a preliminary direction line. Final decision about characteristic hot-wire
probe dimensions still depends on many parameters that can not be evaluated quite
accurately and, therefore, has to be provided using intuition and experience of the
probe designer. Of course, any analysis has to be based on the reliable
experimental information on the scale of Kolmogorov 1941 and specified hot-wire
probe configuration.
Since the vorticity vector is computed from the gradients of velocity
components, adequate multiple hot-wire probes have to be suitably designed to be
capable of providing information on the velocity gradients. Therefore, they consist
of at least two spatially separated arrays, which form the common configuration
for vorticity measurements.

100
10
1
0.1
Ay*
=
Ay/q
0 2 4 6 8 10 12
(
A
u
/
A
y
)
2


(
c
u
/
c
y
)

2


Fig. 5.5: Ratio of experimental () and DNS (

) measured estimations of the


variance of velocity gradients toward DNS real values, as a function of
hot wires separation Ay
*
= Ay/q = h
S
/q over which the velocity difference
is determined. Data that correspond to analysis of Wyngaard 1969, with
DNS spectrum used, are presented by solid line. Adapted from: Antonia,
Zhy and Kim 1993.
MULTIPLE HOT-WIRE PROBES 89
__________________________________________________________________________________________________________________________

This separation has to be finite, in order to detect difference between velocity
vectors in two different spatial points. If it is too large, the smallest structure will
not be detected correctly: their influence will be spatially averaged within the probe
sensing volume. Therefore, measurement results may not provide a good
approximation to a measurement point. In contrast, decreasing the distance
between measuring points results in increasing the absolute error in gradients
(vorticity) measurements, although the absolute measurement error of velocity is
the same, because of increased relative uncertainties.
A typical illustration of this effect is presented in fig. 5.5 of Antonia, Zhu and
Kim 1993. The ratio of the measured velocity gradient variance (based on the
finite-difference approximation) to true variances of velocity gradient, for the
turbulent channel-flow DNS on the centreline (Re
t
= 180) are compared to the
analogue ratio but for the gradient physically measured by hot-wires with variable
separation distances. The ratio obtained by applying the Wyngaards 1969 analysis,
but which utilises three-dimensional DNS spectrum instead of spectrum of Pao
1965, is also presented. The experimental values agree well with DNS finite-dif-
ference approximations for Ay
*
= Ay/q = h
S
/q > 2.0. However, for smaller values
of Ay
*
, hot-wire measurement technique provide values of this ratio much over 1.0.
The accuracy problem, related to the separation of hot-wires and their arrays,
was discussed in detail and explained by Vukoslavevi, Wallace and Balint 1991.
Furthermore, Antonia, Zhy and Kim 1993, Andreopulos and Honkan 1996 and
Wallace and Foss 1995 reported that effects of noise might even overwhelm the
wire signal in such cases.
Therefore, the analysis of spatial resolution of hot-wire probes, but also the
other instruments specified for measurement of turbulence properties, has been and
still is one of the main problems in their design. Presently, there exist a common
opinion that probe dimension should be of the same order as it is q. More
precisely, according to Wallace and Foss 1995, it must not exceed 2-4q when both
the resolution and accuracy constraints are accounted.
Unfortunately, it is not easy to satisfy this criterion in practical applications and,
therefore, researchers try to provide an acceptable spatial resolution by choosing
the suitable flow regime. Desired Reynolds number can be achieved by using large
wind tunnels combined with slow fluid motion, instead of increasing fluid velocity
in the test rigs of small cross-section area. Consequently, coherent vortex structures
of larger scales will occur, resulting in greater q and improved probe spatial
resolution. It means that measurements should be performed at the lowest possible
mean flow velocities, which still guarantee required flow conditions. In addition,
wind tunnels should be designed with the largest possible cross-section area.


90 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
5.3 MEASUREMENT OF LONGITUDINAL VORTICITY COMPONENT
The first hot-wire probe to
measure vorticity, sketched in fig.
5.6, was designed by Kovasznay
1950,1954. This configuration had
four sensors supported by the same
number of prongs, what means that
each prong was common for two
sensors. Hot-wires were operated
as four legs of a Wheatstone bridge,
in a constant-current anemometer
circuit. It was believed at that time
that, if all four hot-wires are
geometrically and electrically
identical, the voltage drop AE
BD

across points B and D uniquely depends on longitudinal velocity component U, and
the voltage drop AE
AC
between points A and C is a function of streamwise vorticity
component O
x
only. The influence of other two velocity components (V and W) on
the measured quantities was considered to be negligible. Dependence of
longitudinal velocity U and vorticity O
x
on measured voltages was approximated
by linear function, which was determined experimentally. However, it was very
complicated, practically impossible with available technology, to manufacture this
configuration to be small enough to measure the smallest turbulence structures with
all wires geometrically and electrically identical.
Uberoi and Corrsin 1951 were the first who used this type of probe. It was
applied in the study of the transition from laminar to turbulent flow. The probe was
calibrated by spinning it at different angular speed in a stream of uniform fluid
velocity. That way, the voltage drop AE
AC
between points A and C can be directly
related to longitudinal (streamwise) vorticity component O
x
, by a suitable
mathematical expression. The applied calibration procedure was simple only at
first site, because it required a sophisticated mechanism to spin the probe and
transmit the output wire signals through a mercury-bath commutator.
Kistler 1952 analysed the sensitivity of Kovasznay 1950 probe to different
parameters. He found out that, even for a perfect geometrical and electrical probe
symmetry, voltage drop E
AC
is not only sensitive to longitudinal (streamwise)
vorticity component O
x
, but also to the second-order derivatives of longitudinal U
velocity component:
|
|
.
|

\
|
c
c
+
c
c
+ =
2
2
2
2
2
2
z
U
y
U
cos h A sin h A E
P x P BD
o O o A
. (5.16)
Coefficient A in (5.16) is a function of operating electric current I and magnitude of
longitudinal velocity component U, while h
P
and o are separation and angle
B
C
A
D
y
z
x
U
0
I
0

I
0
h
P

h
P

o
o
o
o

Fig. 5.6: A sketch of four hot-wire probe,
designed by Kovasznay 1950, 1954.
MULTIPLE HOT-WIRE PROBES 91
__________________________________________________________________________________________________________________________

defined in fig. 5.6. In addition, the same author noticed several effects caused by
ordinarily present geometrical and electrical asymmetries of the probe. He also
reported that voltage drop E
CA
, and therefore O
x
, is not only sensitive to
longitudinal U, but also depends on transversal V and lateral W velocity
components. This parasitic influence increases with increasing the wire angle o. He
believed that this influence is negligible in turbulent flows with reasonably low
levels of V and W. Kistler 1952 measured the one-dimensional vorticity spectrum
in a turbulent grid flow and compared it with spectrum computed utilising the
known one-dimensional velocity spectrum. The agreement was good only for high
wave numbers.
Corrsin and Kistler 1954,1955 used the same type of hot-wire probe to study
vorticity dynamics near the free stream boundaries of turbulent flows. They
measured the RMS values of longitudinal vorticity O
x
distribution at different
distances from the wall. As a result, it was claimed that the probe had adequate
electrical and geometrical symmetry and that influence of cross-stream velocity
components was negligible in the flow of low turbulence level, where the
measurements were performed.
Kastrinakis, Eckelmann and Willmarth 1979 showed that the conclusion of
Corrsin and Kistler 1955 is valid only under some severe limitations, e.g. for
specific combinations of magnitudes of cross-stream velocity components.
Therefore, if a probe of Kovasznay type is used, the influences of V and W
components on the measured values of U and O
x
are negligible only at some
specific orientations of the fluid velocity vector toward the probe axis. However,
the turbulent velocity vector changes its direction and intensity randomly in time,
so the most of the signal will be contaminated by this influence.
Willmarth and Lu 1972 measured the correlation between longitudinal velocity
U and vorticity O
x
components in a high-speed turbulent boundary layer, using
their own probe of Kovasznay type. However, its spatial resolution was not
adequate for the smallest vortices that occur in the turbulent flow.
Wynagard 1969 successfully analysed spatial resolutions of four-wire vorticity
probe, configured according to Kovasznay 1950,1954, and other hot-wire arrays.
His important results, presented in more details in section 5.2, are very often cited
even nowadays.
Kastrinakis 1976 used a pair of probes of Kovasznay type, containing hot wires
2.8mm long, to measure the distribution of RMS(O
x
) across the fully developed
channel flow and the two-point longitudinal-vorticity correlation function RO
x1
O
x2

for laterally separated probes. The parasitic sensitivity of instantaneous O
x
signals
to the signals of streamwise U velocity component (predicted by Kistler 1952
analysis) was noticed and corrected, since the probe was capable of measuring that
velocity component (U). However, the influence of cross-stream components (V,
W) was not corrected, because the applied probe could not determine it. It was
expected that, due to the probe symmetry, this influence should cancel out in
92 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
measuring the mean values of O
x
and its add moments. According to
Vukoslavevi and Wallace 1981 the measurement of even moments of O
x

distributions by this probe would not be reliable.
Utilising the same two probes, which were previously used by Kastrinakis
1976, and pattern recognition algorithm of Wallace, Brodkey and Eckelmann
1977, Kastrinakis, Wallace, Willmarth, Ghorashi and Brodkey 1977 detected the
coherent structures in the near wall region. Due to the probability of being equally
positive and negative, they expected that the influence of V and W components to
measured hot-wire signals would cancel out in conditional averaging of large
number of data.
Kastrinakis, Eckelman and Willmarth 1979 analysed analytically and
experimentally the influence of streamwise V and W components of fluid velocity
on the measured signal of O
x
. In order to correct this parasitic sensitivity of O
x
,
they proposed modification of the probe design originating from Kovasznay
1950,1954. The new configuration should have the same geometrical arrangement,
but each sensor should be supported by separate prongs and operated in an
independent electronic circuit. To achieve this goal, it was necessary to place the
eight prongs, in place of previous four, and avoid the possible significant
increasing the aerodynamic disturbance of the flow. In fact, the modified probe
should consist of a pair of X probes placed in a vertical and horizontal plane. They
are, in principle, capable of measuring the transversal V and lateral W components
of fluid velocity and thus, their influence on hot-wire signals can be accounted.

h
P
h
P
o=45
0

h
P
=0.5mm

(a) (b)
Fig. 5.7: Quadruple hot-wire probe of Vukoslavevi 1981 for measurement of
longitudinal vorticity O
x
and three velocity components: a side-view
sketch and a front-view. Source: Vukoslavevi and Wallace 1981.

Cleveland 1979 reported the first version of modified Kovasznay probe. He
formulated an original expression to measure longitudinal vorticity component O
x

in a flow with zero cross-stream velocity components and tested its applicability by
spinning the probe in uniform flow with V and W components equal to zero.
Although it was possible to determine the influence of cross-stream velocity
components by this probe, under usual assumptions, such analysis was not
MULTIPLE HOT-WIRE PROBES 93
__________________________________________________________________________________________________________________________

performed. The probe spatial resolution was not estimated too.
Vukoslavevi 1981 also constructed a miniature modified probe of Kovasznay
1950 (fig. 5.7). Using his probe, Vukoslavevi and Wallace 1981 analysed the
parasitic influence of transversal V and lateral W velocity components on the
measurement results of longitudinal velocity U and vorticity O
x
components. The
probe, applied in a low-velocity (U

= 2.6m/s) boundary layer of a wind tunnel at


the University of Maryland (USA), had spatial resolution that satisfied the criteria
defined by Wyngaard 1969. Four CTA circuits heated the sensors, configured as
two X arrays. Under usual assumptions, the probe should provide the instantaneous
values of transversal V and lateral W velocity components, enabling estimation of
their influence on the longitudinal vorticity O
x
signal and its correction. The prong
tips were bent in order to reduce the aerodynamic blockage (disturbance) of the
flow, according to recommendation of Strohl and Comte-Bellot 1973. At that time,
this probe represented the most advanced hot-wire configuration, specified for
measurement of O
x
.
Assuming constant gradients of fluid velocity components within the sensing
volume (cU/cy=const., cU/cz=const., cV/cy=const., cV/cz=const., cW/cy=const.
and cW/cz=const.), the U, V and W components may be expanded in the first-order
Taylor series around the centre of the probe measuring volume O (fig. 5.8):

y
z
U
0

h
P
L
o
h
P
x
W
V
O
U

y
z O
h
P
h
P
1
3
2
4

(a) (b)
Fig. 5.8: A sketch of reference orthogonal Cartesian coordinate system for
vorticity measurement by multiple hot-wire probe containing four-wires:
(a) relative orientation of the coordinate system to the quadruple probe;
(b) front view of the probe and adopted Cartesian coordinate system.
Adapted from: Vukoslavevi and Wallace 1981:
2
0 1
P
h
y
U
U U
c
c
+ = ,
2
0 1
P
h
y
V
V V
c
c
+ = ,
2
0 1
P
h
y
W
W W
c
c
+ = , (5.17)
2
0 2
P
h
z
U
U U
c
c
= ,
2
0 2
P
h
z
V
V V
c
c
= ,
2
0 2
P
h
z
W
W W
c
c
= , (5.18)
94 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
2
0 3
P
h
y
U
U U
c
c
= ,
2
0 3
P
h
y
V
V V
c
c
= ,
2
0 3
P
h
y
W
W W
c
c
= , (5.19)
2
0 4
P
h
z
U
U U
c
c
+ = ,
2
0 4
P
h
z
V
V V
c
c
+ = ,
2
0 4
P
h
z
W
W W
c
c
+ = . (5.20)
For a probe perfectly aligned with the mean flow direction, what means that only
the mean value of longitudinal (streamwise) component U is different from zero
( 0 = =W V ), Reynolds decomposition may be simplified to:
v v V V = + = , w w W W = + = , (5.21)
Therefore, the velocity components U
Ni
(i=1,2,3,4), orthogonal to the wire, will be:
2
0
2
0 0
2
1
2 2 2
|
|
.
|

\
|

c
c
+ +
(

|
|
.
|

\
|

c
c
+ +
|
|
.
|

\
|

c
c
+ =
P P P
N
h
y
v
v sin
h
y
w
w cos
h
y
U
U U o o ,
2
0
2
0 0
2
2
2 2 2
|
.
|

\
|

c
c
+
(

|
.
|

\
|

c
c
+ |
.
|

\
|

c
c
=
P P P
N
h
z
w
w sin
h
z
v
v cos
h
z
U
U U o o
,
2
0
2
0 0
2
3
2 2 2
|
|
.
|

\
|

c
c
+
(

|
|
.
|

\
|

c
c

|
|
.
|

\
|

c
c
=
P P P
N
h
y
v
v sin
h
y
w
w cos
h
y
U
U U o o
,
2
0
2
0 0
2
4
2 2 2
|
.
|

\
|

c
c
+ +
(

|
.
|

\
|

c
c
+ + |
.
|

\
|

c
c
+ =
P P P
N
h
z
w
w sin
h
z
v
v cos
h
z
U
U U o o
.
(5.22-5.25)
The terms in the first (square) brackets represent the normal U
n
and the terms in the
second brackets are the binormal U
b
components of velocity in the local coordinate
system of the wire (figs. 3.3 and 3.4). In place of the real angle o, an effective
value o
e
may be also used. Other parameters are designated according to fig. 5.8.
Combining the equation (5.22) to (5.25), the following expression for the
streamwise (longitudinal) component of vorticity vector arises:
o
O
sin h
v U v U w U w U
P
N N N N
x

+
=
2
0
2
1
2
0
2
3
2
0
2
2
2
0
2
4
. (5.26)
It was derived neglecting the terms containing h
P
2
(assuming that h
P
is
reasonably small). In addition, terms containing velocity gradient cv/cy and cw/cy
were assumed to approximately cancel each others when expanded using the
binomial theorem, by putting the radicals in the form (1+x)
n
. Besides the effective
cooling velocity components U
Ni
, the cross-stream fluctuations at the centre O of
the measuring volume v
0
and w
0
also appears in (5.26). It became quite clear that
their influence is not of the second order, as thought previously, and that it is
MULTIPLE HOT-WIRE PROBES 95
__________________________________________________________________________________________________________________________

minimised only if they are of the same intensity. The following expression for
cross-stream velocity components can be derived from (5.22-5.25)
o
o
sin
cos
z
U
h w U w U
v
P N N
2
2
0
2
4
2
0
2
2
0

c
c
+
= (5.27)
o
o
sin
cos
y
U
h v U v U
w
P N N
2
2
0
2
1
2
0
2
3
0

c
c
+
= (5.28)
if the gradients cV/cy, cV/cz, cW/cy and cW/cz are neglected, what is usual
assumption for any type of these probes.
For a probe perfectly aligned with a mean flow, the transversal v and lateral w
fluctuations of turbulent fluid velocity are much smaller than cooling velocity U
N

(v
0
<< U
N1
, U
N3
and w
0
<< U
N2
, U
N4
) and, therefore
o
o
sin
cos
z
U
h U U
~ v
P N N
2
4 2
0

c
c
+
,
o
o
sin
cos
y
U
h U U
~ w
P N N
2
1 3
0

c
c
+
. (5.29-5.30)
Expressions (5.29) and (5.30) shows that streamwise (transversal and lateral)
velocity components can be accurately measured only if the streamwise gradients
of longitudinal velocity component cU/cy and cU/cz are negligible. Although this
is also a basic assumption for any type of X probe, Vukoslavevi and Wallace
1981 showed that it may lead to a large error in O
x
measurements, but also in
measurements of cross-stream V and W velocity components. To determine this
error, they designed and applied a gradient probe with two sensors spaced at the
same distance (4.6 viscous lengths in their boundary layer) as the four-wire
vorticity probe. Measured maximum value of instantaneous gradients cU/cy and
cU/cz showed that the instantaneous measurement error in O
x
, due to neglecting of
these gradients, could be of the same order as the measured values are.
Park and Wallace 1993 emphasised an important contribution of Vukoslavevi
and Wallace 1981, who recognised the problem of vorticity measurement by a
four-wire probe. They pointed out the logical inconsistency to assume that
Kovasznay-type probe and its modified version are sensitive only to the velocity
gradients of the longitudinal vorticity O
x
(cW/cy, cV/cz) and simultaneously
insensitive to the other instantaneous gradients that are also present in the turbulent
flow. Very important is also the evidence that wire separation affects not only the
vorticity measurement by multiple hot-wire probes, but also the measurement of
velocity components with X or V probes possessing only two sensors.
In a subsequent paper, Wallace and Vukoslavevi 1982 derived an explicit
formula for estimation of the measurement error of longitudinal vorticity O
x
, due to
neglecting the transversal V and lateral W components of fluid velocity:
96 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
|
|
.
|

\
|
+

=
4
2
2
2
3
2
1
2
2
1
N N N N P
x
U
w
U
w
U
v
U
v
sin h o
c
O
. (5.31)
They estimated the probability density distribution of c
Ox
, using triple hot-wire
probe to measure cross-stream V and W velocity components in a turbulent
boundary layer at y
+
= 20. Although the probability of large measurement errors
was small, extremely large errors, even larger than the value of mean vorticity
close to the wall, occurred occasionally. Therefore, they should not influence the
mean and RMS values of longitudinal vorticity, but the higher-order statistics of
the same vorticity component (O
x
) can be strongly affected.
Park and Wallace 1993 performed more refined analysis of the modified
Kovasznay-type probe designed by Vukoslavevi and Wallace 1981. Following
the approach of Vukoslavevi, Wallace and Balint 1991, they introduced
effective cooling angles o
ei
in place of the wire (geometrical) angles o
i
and
different calibration constants h
i
for each sensor. Assuming the probe perfectly
aligned with a mean flow direction and expanding the fluid velocity components
around the probe centre in the first-order Taylor series, they obtained the following
expressions for hot-wire effective cooling velocities:
2
0 1
2
1 0 1 0 1
2
1 0 1 0
2
1
2 2 2
2 2
|
|
.
|

\
|
c
c
+ +
)
`

|
|
.
|

\
|
c
c
+ +
|
|
.
|

\
|
c
c
+ +
+
)
`

|
|
.
|

\
|

c
c
+
|
|
.
|

\
|

c
c
+ =
P P P
P P
e
h
y
v
v h cos
h
y
w
w sin
h
y
U
U k
sin
h
y
w
w cos
h
y
U
U U
o o
o o
,

2
0 2
2
2 0 2 0 2
2
2 0 2 0
2
2
2 2 2
2 2
|
.
|

\
|
c
c
+
)
`

|
.
|

\
|
c
c
|
.
|

\
|
c
c
+
+
)
`

|
.
|

\
|
c
c
+ |
.
|

\
|
c
c
=
P P P
P P
e
h
z
w
w h cos
h
z
v
v sin
h
z
U
U k
sin
h
z
v
v cos
h
z
U
U U
o o
o o
,

2
0 3
2
3 0 3 0 3
2
3 0 3 0
2
3
2 2 2
2 2
|
|
.
|

\
|
c
c
+
)
`

|
|
.
|

\
|
c
c

|
|
.
|

\
|
c
c
+
+
)
`

|
|
.
|

\
|

c
c
+
|
|
.
|

\
|
c
c
=
P P P
P P
e
h
y
v
v h cos
h
y
w
w sin
h
y
U
U k
sin
h
y
w
w cos
h
y
U
U U
o o
o o
,

2
0 4
2
4 0 4 0 4
2
4 0 4 0
2
4
2 2 2
2 2
|
.
|

\
|
c
c
+ +
)
`

|
.
|

\
|
c
c
+ + |
.
|

\
|
c
c
+ +
+
)
`

|
.
|

\
|
c
c
+ |
.
|

\
|
c
c
+ =
P P P
P P
e
h
z
w
w h cos
h
z
v
v sin
h
z
U
U k
sin
h
z
v
v cos
h
z
U
U U
o o
o o
.

(5.32-5.35)
MULTIPLE HOT-WIRE PROBES 97
__________________________________________________________________________________________________________________________

and derived the velocity and vorticity components in the following forms:
( ) ( )
)
`

+
c
c
+ +
+


+

=
3 1 3 1 0
3
2
0 3
2
3
1
2
0 1
2
1
01
2
2
1
e e
P
e e
e
e
e
e
tan tan
h
y
w
tan tan w
cos
v h U
cos
v h U
U
o o o o
o o
,
(5.36)

( ) ( )
)
`

+
c
c
+
+


+

=
4 2 4 2 0
4
2
0 4
2
4
2
2
0 2
2
2
02
2
2
1
e e
P
e e
e
e
e
e
tan tan
h
z
v
tan tan v
cos
w h U
cos
w h U
U
o o o o
o o
,
(5.37)

|
|
.
|

\
|
+
c
c
+
|
|
.
|

\
|

+


=
4 2 4 2
02
4
2
0 4
2
4
2
2
0 2
2
2
0
1 1
2
1 1
2
1
e e
P
e e
e
e
e
e
tan tan
h
z
U
tan tan
U
sin
w h U
sin
w h U
v
o o o o
o o
,
(5.38)

|
|
.
|

\
|
+
c
c
+
|
|
.
|

\
|
+


=
3 1 3 1
01
1
2
0 1
2
1
3
2
0 3
2
3
0
1 1
2
1 1
2
1
e e
P
e e
e
e
e
e
tan tan
h
y
U
tan tan
U
sin
v h U
sin
v h U
w
o o o o
o o
,
(5.39)

|
|
.
|

\
|

c
c
+
|
|
.
|

\
|

c
c
+
+
|
|
.
|

\
|
+ +


+

=
4 2 3 1
4 2 3 1
0
3
2
0 3
2
3
1
2
0 1
2
1
4
2
0 4
2
4
2
2
0 2
2
2
1 1
2
1 1
2
1 1 1 1
1
e e
P
e e
P
e e e e e
e
e
e
e
e
e
e
P
x
tan tan
h
z
U
tan tan
h
y
U
tan tan tan tan
U
sin
v h U
sin
v h U
sin
w h U
sin
w h U
h
o o o o
o o o o o
o o o
O
.
(5.40)

Ideally, for a perfectly calibrated probe and geometrically and electrically identical
sensors placed in the flow of uniform velocity, the values U
01
and U
02
should be
98 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
identical. However, the presence of the velocity gradients in the probe measuring
volume, as well as some deviation of its design from ideal, produce difference
between the values of U
01
and U
02
. Therefore, it is appropriate to compute the
corresponding mean value:
2
02 01
0
U U
U
+
= . (5.41)
Besides the effective wire cooling velocities, equation (5.40) also include all
three components of velocity U
0
, v
0
and w
0
and cross-stream gradients y / U c c and
z / U c c of longitudinal velocity. In addition, it is also dependent on the gradients
y / w c c and z / v c c , which influence is introduced indirectly, by U
0
that depends
on these streamwise fluid velocity gradients (see expressions (5.36) and (5.37)). A
coupled set (5.36)-(5.40) of equations can provide values of U
0
, v
0
, w
0
and O
x
.
Obviously, as is the case with all X-arrays (the assumption of uniformity of
instantaneous fluid velocity field within the probe sensing volume), the gradients
included in these equations have to be neglected. Their exclusion from (5.36)-
(5.40) enables iterative solving of the equation system. If the cross-stream fluid
velocity components v
0
, w
0
are also neglected, and the effective cooling angles
assumed to be identical, the expression (5.40) reduces to the (much simpler)
original expression proposed by Kovasznay 1950,1954.
Although the explicit influence of y / U c c and z / U c c on O
x
is rather weak,
their effects are introduced through their presence in equation (5.38) and (5.39) for
streamwise v
0
and w
0
components. Thus, if the characteristic dimension h
P
(fig.
5.8) is too large, the required assumption of the uniform flow within the probe
sensing volume is not correct and the cross-stream velocity components v
0
and w
0

will be poorly measured. Consequently, an attempt to account them in equation
(5.40) will result in the large measurement errors of longitudinal vorticity O
x
, and
even more in the case of its higher-order statistics.
Using experimental data of Balint, Wallace and Vukoslavevi 1991, measured
in a zero-pressure-gradient turbulent boundary layer of R
u
=2700 by nine-sensor
probe at y
+
=30, Park and Wallace 1993 analysed influence of neglected gradients
of the modified Kovasznay-type probe on U
0
, w
0
and v
0
. The time series of three-
dimensional velocity vectors and their gradients, x / U c c , y / U c c , z / U c c ,
x / V c c , y / V c c , z / V c c , x / W c c , y / W c c and z / W c c , were available from
this data-base permitting the computation of cooling velocities of four hot-wires of
Kovasznay-type probe for each time step from the response equations (5.32)-(5.35).
In the further step, the values of U
0
, w
0
, v
0
and O
x
were computed for the
modified Kovasznay-type probe from the equations (5.36)-(5.40), by neglecting the
velocity gradients.
MULTIPLE HOT-WIRE PROBES 99
__________________________________________________________________________________________________________________________

Later on, they were
compared with adequate
values of Balint, Wallace
and Vukoslavevi 1991,
measured by the nine-wire
probe (fig. 5.9). The agre-
ement between longitudinal
U velocity measured by nine-
wire probe and those from a
simulated response of four-
wire modified Kovasznay
configuration, was very good.
In contrast, cross-stream
velocity components, as well
as the Reynolds shear stress
and longitudinal vorticity, did
not compare well during
several time intervals of the
analysed time series. In order
to check whether these
differences are caused by the
neglected fluid velocity
gradients in equations (5.36)-
(5.40), Park and Wallace
1993 also calculated the
simulated wire response
including these gradients in
(5.36)-(5.40). The results for
each variable computed with
accounted velocity gradients
closely matched that of the
nine-wire probe. The authors
concluded that, to the first
order, expressions (5.36)-
(5.40) would accurately yield the dependent variables U
0
, v
0
, w
0
and O
x
if the
instantaneous values of all independent variables could be measured in each time
step. Unfortunately, this is not possible with the four-wire probe. Thus, the
necessary omission of the velocity gradients, which instantaneous magnitudes are
often large in turbulent flows, can quite adversely affect the inferred values of v
0
,
w
0
and O
x
. Effects of these neglected fluid velocity gradients on various statistical
flow properties were also discussed in the paper.
Willmarth and Bogar 1977 also analysed the influence of velocity gradients on
measured signals of X probe. They found out the minimum distance between the

t (ms)

Fig. 5.9: Instantaneous time-series of the velocity
components U(t), v(t) and w(t), correlation
) t ( v ) t ( u
and longitudinal vorticity O
X
(t):
__

__
a simulated response of a
modified Kovasznay probe with
four wires for h
P
/q = 5.2, Park
and Wallace 1993;
______
a nine-sensor vorticity probe
measurements of Balint, Wal-
lace and Vukoslavevi 1991.
Source: Wallace and Foss 1995.
100 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
X-sensors, necessary to avoid the influence of velocity gradients. Unfortunately,
this is impossible to achieve for a vorticity probe, because the distance between one
of X arrays is determined by the sensors length L of the other array. However, this
was the initial motivation to design a nine-sensor probe that should be capable of
measuring vorticity and velocity field without neglecting any gradient of velocity
vector components.
Kastrinakis and Eckelmann 1983 measured statistical parameters of turbulent
channel flow of low Reynolds number by the modified Kovasznay probe. Its spatial
resolution was about 5 times of Kolmogorov micro-scale, what does not satisfy
criteria of Wyngaard 1969. The probe was tested in a uniform jet flow. Pitching
and yawing the probe, cross-stream velocity components were induced and cor-
responding expression was derived in order to correct their influence on output
signals of hot-wire probe.
Kastrinakis, Nychas and Eckelmann 1983 used the data-base of previous
experiment to determine length scales from auto-correlation functions, and Nychas,
Kastrinakis and Eckelmann 1985 used it to investigate components of the
production and dissipation terms of the transport equation for fluctuating
enstrophy. Menevay, Screenivasan, Kailasnath and Fan 1990 used a modified
probe of Kovasznay-type for multi-fractal analysis of O
x
(t) in the atmospheric
surface layer. A variety of other papers exist, which presents experimentally
studied effects of wire separation on X-array measurements of turbulent velocity
components: Bremhorst 1972, Tagawa, Tsuji, Nagano 1992, Nakayama and
Westphal 1986, Browne, Antonia and Shah 1988, etc.
Recently, a number of numerical-simulation data-basis of turbulent flows
became available. Thus, it become possible to simulate the effects of neglected
fluid velocity components, as well as the gradients of fluid velocity components
within the probe sensing volume, on the values of U
0
, V
0
, W
0
and O
x
, measured by
the finite-dimension modified Kovasznay probe. Moin and Spalart 1987, Suzuki
and Kasagi 1992 and Pompeo and Thomann 1993 also have been carried out
studies of such kind. All of them found the cross-stream velocity components to be
most affected by wire separation, with obvious further adverse consequences to
vorticity measurement techniques that rely on finite-difference estimations of
gradients of these components of fluid velocity vectors. According to Wallace and
Foss 1995, these effects of sensor separation represent the mayor aspect of probe
resolution problems, which may result in decreased measurement accuracy.
Kock am Brink and Foss 1993 have used an alternative design to measure a
longitudinal vorticity O
x
in the flow downstream of a rippled trailing edge of a
splitter plate creating two-stream mixing layer. The values of O
x
for a plane and
rippled trailed edge conditions have been compared. Their probe was very
complex: its eight sensors were configured in four X-arrays placed on a circuit of
5.5mm diameter in the cross-stream plane. Each sensor had active length of 1mm
and was supported by prongs separated by 3mm at their tips. Separation between
MULTIPLE HOT-WIRE PROBES 101
__________________________________________________________________________________________________________________________

wires of each X-array was 1mm too. Depending on their orientation, X-array
signals were processed giving the values of U and V, or U and W. Therefore, a
correction of binormal cooling velocity was possible for each array, depending on
its orientation. However, the probe sensing volume was too large.

5.4 MEASUREMENT OF TWO CROSS-STREAM VORTICITY COMPONENTS
Hot-wire probes,
specified for
measurement of, one
at a time, transversal
(normal) O
y
and lateral
(spanwise) O
z

components of fluid
vorticity vector have
been designed and
refined for a number
of years by Foss
1976,1981, 1994,
Foss, Ali and Haw
1987 and Foss and
Haw 1990a,b. Initially
designed probe (Foss
1976, 1981),
containing two closely
placed two-wire
arrays, is presented in
fig. 5.10a. The first
array of two parallel
hot-wire sensors,
employed to measure
gradient cU/cy, is
placed besides an X-
array specified for
determination of U
and V components of
fluid velocity vectors.
Hot-wires are end-
plated and widely
separated, according to
recommendations of
Comte-Bellot 1973, in
order to ensure the
(a)
1mm
20mm
1mm
4
3
2
1
3mm
U
0

y
x
U
V
|
1

|
2

3
m
m



(b)
3
m
m

1mm
Parallel wires
4
3
1
2
z
y
x
y
O
O
x
z
O
The streamwise
plane
Ox - longitudinal (streamwise) axis
Oy - transversal (spanwise) axis
Oz - lateral (normal) axis
X-wires
3
m
m

1
m
m


(b)

Fig. 5.10: Sketches of the front and side views of four
hot-wire probes, specified for measurement of cross-
stream component of vorticity, O
y
or O
z
:
(a) an initial version, designed by Foss 1981;
(b) a compact design of Foss and Haw 1990b.
Adapted from: Wallace and Foss 1995.
102 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
probe applicability in the free-shear flows characterised by relatively large attack
angles of fluid velocity vectors toward the mean flow direction (usually aligned
with the probe axis). Active sensors lengths and spatial distances between hot-
wires of each array are 1mm, while the distance between array centres and the total
lengths of hot-wires (including plated ends) are 3mm. The estimations of vorticity
components are based on Taylors hypothesis. Furthermore, during signal
interpretation, some of velocity gradients over the probe sensing volume have to be
neglected, as it was the case for Kovasznay-type probes.
Foss, Klewicki and Disimile 1986 applied this probe to measure the
fluctuating velocity components U and V and the lateral vorticity component O
z

at the entraining of a large plane shear layer. Falco 1983 also used the same type
of hot-wire configuration to evaluate ensemble averages of time traces of O
z
in a
boundary layer that was visually conditioned on the passage of pocket
structures. Klewicki 1989a,b, Klewiski and Falco 1990, Klewicki, Gendrich,
Foss and Falco 1990, Klewicki, Falco and Foss 1992 and Klewicki, Murray and
Falco 1994, used the probe designed by Foss to measure statistical parameters of
lateral vorticity and its correlation with velocity components in a thick boundary
layer characterised by Re
u
= 1010-4850. Based on the estimation of turbulent
dissipation of Balint, Wallace and Vukoslavevi 1991, the distance between the
centres of the X-arrays and the parallel-wire array was around 6q near the wall,
decreasing to about 2.4q at the outer part of the flow. However, for Re
u
= 2870,
these values were 13.3q and 3.6q, respectively, and 18.7q and 4.6q, for the
highest Re
u
.
Adverse effects of sensor separation, found by Foss, Ali and Haw 1987 for the
array spatial distance of 3mm, motivated Haw, Foss and Foss 1989 to change the
probe of Foss 1981 (see fig. 5.10b). Two sensors in the X-array of the new probe
are threaded within the parallel array. This way, the probe volume was made much
more compact, what reduced the effects of unknown velocity gradients within the
sensing volume. The compact probe, sketched in fig. 5.10b, was applied to measure
distributions of transversal vorticity O
y
and corresponding spatial derivations of
fluid velocity components at the high-speed edge of the large planar shear layer.
Foss and Haw 1990a gave a detailed description of probe design, data processing
algorithms and the uncertainty considerations related to its use in turbulent flows.
Foss and Haw 1990b used the same (compact) probe to measure the RMS
values of O
z
probability density distribution in a 2:1 velocity-ratio mixing layer.
The spatial resolution of the probe was around 7q at the splitter plate plane.
Antonia and Rajagopalan 1990 measured statistical properties of O
y
and O
z
,
one at a time, by rotating a probe for 90
0
, in the wake behind a cylinder. They used
a compact version of hot-wire configuration designed by Foss. The experiment was
performed at Reynolds number Re
d
= 1070, where d was the diameter of a cylinder,
at 420 diameters downstream of the cylinder. The spatial distance between the
MULTIPLE HOT-WIRE PROBES 103
__________________________________________________________________________________________________________________________

parallel wires was about 3.5q, while the separation between the wires of the X-
array was around 5.1q. Evaluating both sides of the approximate equation
( )
y z
w v
y
v u
e e ~
c
c
, (5.42)
which neglects the longitudinal gradients of the variances of velocity components
on the right-hand side, they tested their hot-wire probe. Full explanation of the
approximation (5.42) can be found in Tennekes and Lumley 1972, for example.
Good agreement was found in the fully turbulent part of the wake, what was not the
case in the intermittent region. The authors believed that the lack of agreement in
that flow region originates from the resolution problems, related to domination of
small-scale structures produced by vortex-stretching mechanism, but they did not
prove it.
Rajagopalan and Antonia 1993a used a version of the four-wire probe
presented in fig. 5.10b to measure RMS values of O
z
distributions in a boundary
layer at Re
u
=1450. This flow regime was chosen to match approximately the
highest Re number of the direct numerical simulation (DNS) of Spalart 1988. At
the non-dimensional distance from the wall y
+
= 10, separations between the pairs
of parallel and X wires were 6q and 7.4q, respectively. The spatial resolution
enchanced in the intermittent region of the flow to 1.5q and 1.8q, respectively.
Among other contributions, these authors have verified experimentally that a good
approximation to the RMS values of lateral vorticity component can be obtained
using expression
2
2
|
.
|

\
|
c
c
+
|
|
.
|

\
|
c
c
~
x
u
y
u
RMS
z
o e , (5.43)
where the coefficient o = 1.4 fits over much of the turbulent boundary layer and
goes to zero at the wall.
Rajagopalan and Antonia 1993b applied the same four-wire vorticity probe in
the identical boundary layer to measure single-point correlations between the O
z

and fluid velocity components. Furthermore, the two-point correlations between
the O
z
and longitudinal velocity U, measured by a single-wire probe, placed at
the edge of the viscous sub-layer, were researched experimentally.
Foss 1994 summarised measurements of cross-stream vorticity components in a
free shear layers for different values of Reynolds number
v
u A
u

=
U
Re =4800-78000, (5.44)
where AU is the characteristic velocity difference across the layer. The spatial
resolutions of probes, applied in these measurements, varied between 7q and 13q.
104 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Antonia, Browne and Shah 1988 used a pair of X-probes to measure, one at a
time, cross-stream vorticity components O
z
and O
y
in a turbulent wake, using
Taylors hypothesis. The sensor separation was 1mm, while the distance between
centres of X probes was 1.6mm. It means that, for the Kolmogorov micro-scale
estimated to be q ~ 0.45mm, spatial resolution was between 2.2 and 3.6 q.

5.5 SIMULTANEOUS MEASUREMENT OF TWO
CROSS-STREAM VORTICITY COMPONENTS
Eckelmann, Nuchas, Brodkey and Wallace 1977 constructed a probe
containing five hot films (fig. 5.11) in order to measure simultaneously transversal
O
y
and lateral O
z
components of vorticity vector. The measurements were
performed in an oil channel flow, with a micro-scale of Kolmogorov of about 1mm
in the wall region, which provided a good spatial resolution of the applied probe.
The single-film and two sensors of V-array were used to obtain cU/cz gradient,
while cV/cx was determined by applying Taylors hypothesis to the time derivation
of V, e.g. cV/ct. Similar approach was applied in the case of cW/cx. The lateral
gradient cU/cz was calculated from the longitudinal velocities U
1
and U
2
, measured
by the pair of V-configured films and two sensors included in the X-array,
respectively.

y x
z
90
0

90
0

1
2
3
4
5
45
0
4
1
2
3
5
x
y
w
a
l
l

4mm
1mm
z

Fig. 5.11: A sketch of the five hot-films probe, specified for simultaneous
measurement of cross-stream vorticity components O
y
and O
z
, designed
by Eckelmann, Nuchas, Brodkey and Wallace 1977. Adapted from:
Wallace 1986.
These authors utilised the pattern recognition procedure, proposed by Wallace,
Brodkey and Eckelmann 1977, for determination of ensemble averaged cross-
stream components (O
y
and O
z
) of vorticity patterns, which occurred
MULTIPLE HOT-WIRE PROBES 105
__________________________________________________________________________________________________________________________

simultaneously with
ejection/sweep pattern
of longitudinal U
velocity component.
However, the authors
found out that, for y
+
>
10, their probe was
unable to resolve
correctly the mean
velocity cU/cy gradient,
probably because of too
small sensor separation
in the y direction, of
only 1q. The problems
related to small
separation of the probe
sensors are discussed in
the section 5.2.
J.-H Kim 1989
designed and used a
special six-wire probe to
measure simultaneously
O
x
and O
z
or, by
rotating the probe
through 90
0
around its
longitudinal axis, O
x

and O
y
. Kim and Fiedler 1989 also applied the same configuration. The probe
sensors were specially arranged to form four pairs of X arrays, as can be seen in
fig. 5.12. Four adjacent hot-wires, separated for 1mm (wires no. 1,2,3 and 4) and
lying in the planes parallel to xOz plane (or xOy plane, after rotation), form three
X-arrays. The third pair of hot-wires, separated at 2mm, is threaded through them,
in the xOy plane (or xOz plane, after rotation for 90
0
).
J.-H. Kim 1989 assumed the separation of 1mm for the sensors of the three
parallel X arrays to be sufficiently small to allow consideration of the flow field
over the sensing volumes of each of these arrays as uniform. Thus, U and V (or U
and W, for rotated probe) velocity components were determined from the simple X-
array sum and difference response equations. In contrast, for the additional X array
in the orthogonal planes, he concluded that the spacing of 2mm was too large to
allow the application of the previous hypothesis on the uniform velocity field.
Therefore, he derived expressions for the effective cooling velocity for each of
these sensors, which were similar to those of Vukoslavevi and Wallace 1981. He
used Taylor series of the first order to expand velocity components about the centre
of the probe in the cross-stream (zOy) plane. Thus, when obtaining cV/cz (or cW/cy
x
z
O
h
1
5 6
2h
1,3 2,4
5
6
h
h
h
h
2
3
4
x
y
O
z
y
O
h
h
h
h
h
2h
1
2
3
4
5 6
SCALE: h = 1 mm

Fig. 5.12: The 6-sensor probe of J.-H Kim 1989 designed
to simultaneously measure cross-stream
vorticity components: O
x
and O
y
, or O
x
and O
z
.
Adapted from: Wallace and Foss 1995.
106 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
with rotated probe), the non-uniformity of the fluid velocity field was taken into
account in the (zOy) plane. Vorticity vector components were computed by
integration of the fluid velocity components around the 2mm x 2mm square, shown
in the fig. 5.12 by dotted lines, to find the micro-circulation using algorithm similar
to those of Foss and his research team. They followed Foss 1981, who assumed an
accumulation of large sample intervals at the high sampling rate, of about 20kHz in
this case, for each computation step. That way, a convection length that
approximately matched the cross-stream integration lengths of 2mm was obtained.
In this interpretation procedure, the convection velocity in Taylors hypothesis was
assumed to be instantaneous velocity measured by the probe, causing the obtained
time series to be of the non-equidistant time intervals.
J.-H. Kim 1989 applied his probe in a two stream mixing layer with velocity
ratio 2:1 and Re
u
= 4300, possessing turbulent boundary layers on both sides of a
splitter plate at the trailing edge. The vorticity components were measured at a
downstream distance of 1.25m, where a length of 1mm corresponded to
approximately 3.8q. The RMS values of O
z
, measured with this probe were
compared well with data of several different experiments, like Lang 1985, as well
as to DNS data-basis of Rogers and Moser 1994.
a
b
t-----------t

Fig. 5.13: Turbulent isothermal mixing layer of Helium and Nitrogen, flowing at free
(undisturbed) stream mean velocities U
1
= 1m/s and U
2
= 0.38m/s at p =
4bar: (a) a side view and (b) an upper view. Source: Bernal et al 1979.
However, in order to use any of these presented probes and corresponding
measurement methods, it was necessary to neglect one or more components of
fluid velocity vector and their gradients. In the real turbulent flows, containing
three-dimensional vortices illustrated in fig. 5.13, these approximations affect to
some extent the measurement accuracy of vorticity components.
MULTIPLE HOT-WIRE PROBES 107
__________________________________________________________________________________________________________________________

5.6 SIMULTANEOUS MEASUREMENTS OF THREE VORTICITY COMPONENTS
Having in mind that the cooling flux of any heated wire depends on all three
components of flow velocity it is necessary to use at least three sensors in order to
measure the velocity vector at a given point in a flow. However, with only three
wires, the velocity gradients over the space occupied by hot-wires have to be
neglected, what may cause large measurement errors. In order to determine the
velocity gradients, it is necessary to measure the velocity components
simultaneously at one additional point closely separated to the measuring point in
the defined direction. This was the main idea for construction of the probe with
nine hot-wires, specified for simultaneous three-dimensional measurement of fluid
vorticity vectors. Three arrays, each containing three wires, are closely separated in
the directions of y and z axis, in order to measure the velocity gradients cU
i
/cy and
cU
i
/cz (i = 1,2,3 for the axes x, y and z fig. 5.14). In this case, the assumption of a
uniform velocity field over the probe sensing area, which is unavoidable for the
probes with three wires, is replaced by the assumption of the constant gradients of
fluid velocity components over the probe sensing volume.
(a) (b) (c)
(d) (e)
Fig. 5.14: A nine-wire probe WP-9t of Vukoslavevi, Balint and Wallace 1991:
a side-view sketch (a,d), probe assembly (b) and a front view (c,e).
Adapted from: Vukoslavevi, Balint and Wallace 1991.
108 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
The first attempt to measure all three components of the vorticity vector
simultaneously was made by Wassman and Wallace 1979,1980, who used a nine-
sensor probe. It was configured of three identical arrays, spaced over small
distances from each other to measure the velocity vectors at three spatially close
points and calculate velocity gradients over the sensing volume, in order to
determine vorticity vector components. To make the probe as compact as possible,
the supporting prong at centre of each array was common for the three sensors of
that array. They solved the problem of cross talk between electrical circuits, due to
the common prongs, but they were not able to develop the probe to the operation
level to measure accurately three-dimensional velocity and vorticity fields.
To simplify the response equations of the sensors, Vukoslavevi and Wallace
1984 designed a nine-wire probe with two sensors in each array at 45
0
to the probe
axis, forming a V-array in the xOz plane, and the third wire at the same angle in
xOy plane. The probe was further developed up to the end of 1980s, by the same
team. The advances in its design were reported by Vukoslavevi, Wallace and
Balint 1989,1990, while the last version, shown in fig. 5.14 and designated as WP-
9t(G), was presented by Vukoslavevi, Wallace and Balint 1991. The final
version of nine-wire probe WP-9t(G), also consisted of three T-configured arrays,
each containing three wires inclined at 45
0
to the probe axis. Total dimensions of
the sensing area were 1.7mm vertically and 2.2mm horizontally, with an average
distance between the hot-wire centres of approximately 1.2mm and a distance h
between the supporting prongs of each array equal to 0.5mm. Diameter of the
tungsten sensors was 2.5m and their length was about 0.7mm, giving an aspect
ratio of about 280.
Fabrication of the probe was difficult because of its small size, complex design
and fragility. The main problem was to fit twelve supporting prongs in the smallest
feasible space, while still satisfying accuracy as well as the spatial resolution
criteria. Arranging the prongs with very small distances between them may result
in two difficult problems: preventing electrical shorts between the prongs or
between the prongs and the probe body and maintaining the prongs in place so that
they keep the desired array geometry. Several technical solutions and various prong
materials have been tried. For this version of the probe, a nickel-plated tungsten
wire of about 0.25mm diameter was used for supporting prongs. Short pieces of
this wire were chemically tapered from their original diameter to a tip diameter of
about 75m. The four prongs of each array are then drawn through round four-bore
ceramic insulating tubes and finally placed in a stainless steel tube. The prong tips
were manipulated under a measuring microscope into their proper geometric
position within about a 0.01mm tolerance. After nickel plating the prong, the
2.5m diameter unsheathed tungsten-wires were welded to the tips.
The choice of the probe dimensions has been guided by two space-resolution
criteria, both of which need to be satisfied. In order to estimate with good
resolution the velocity that effectively cools each sensor, its length should be
MULTIPLE HOT-WIRE PROBES 109
__________________________________________________________________________________________________________________________

smaller than the size of the smallest turbulent vortex structures encountered in the
flow i.e. the micro-scale q of Kolmogorov. Balint, Wallace and Vukoslavevi
1991 applied the probe in a zero-pressure-gradient turbulent boundary layer of a
low speed wind tunnel at Re
u
= 2685, and found that the sensor lengths were about
3.6q at y
+
= 11.2 at the location closest to the wall. For positions further from the
wall, q became larger as the kinetic energy dissipation rate c (directly related to the
gradients of fluid velocity vector components) became smaller, thus giving better
spatial resolution of the sensors. Clearly, hot-wires did not resolve the smallest
turbulence scales (i. e. coherent vortex structures), as is the case in almost all
turbulence measurements.
Besides the first resolution criterion, the probe should also be able to resolve
adequately the velocity gradient field in the flow of interest. A measure of this
resolution criterion is the average spacing between sensor centres over which
gradients are estimated. This average spacing for this probe was about 6.3q, or
10.9 viscous lengths, at y
+
= 11.2 in the boundary-layer flow described above.
Following the analysis given by Wyngaard 1969, Klewiski and Falco 1990
estimated from the data obtained by Balint, Vukoslavevi and Wallace 1987 that
a previous version of nine-sensor probe attenuated the RMS(O
z
) measurements by
only about 10-15% at Re
u
= 2100. The dimensions of the last version, shown in fig.
5.14, were slightly smaller, so it is reasonable to conclude that, for the presented
measurements this probe was able to resolve all but the smallest scale of the
studied boundary layer.
As discussed by designers of this probe, it was possible to make it much
smaller. However, the spatial resolution requirements of the probe must be
balanced by other criteria that affect measurement accuracy. The velocity and
vorticity measurement precision increases with decreased array spacing to the
extent when the neglected second and higher-order terms in the Taylor series
expansion of the velocity around the probe centre become negligible. However, the
measurement accuracy will decrease with reducing the spacing between arrays, for
two reasons. At first, the relative accuracy of measuring the velocity differences
decreases as the values of the two velocities to be subtracted become close to each
other, what may increase the relative measurement error of their difference. This is
because the absolute accuracy of each velocity measurement should be independent
of spatial separation to the extent that the first-order gradient approximation is
nearly valid. It means that, for the same absolute error in velocity measurements,
the absolute error in gradient measurements will increase proportionally to the
space decreasing. The second reason is related to relative accuracy of estimation of
the spacing between arrays (denominator in the first-order velocity gradient
estimate), which also decreases with decreased spacing by the same reasoning. In
addition, the effects of aerodynamic and thermal interference of these closely
spaced prongs and wires must be considered too. Therefore, all these competing
considerations must be balanced in the choice of sensor lengths and spacing.
110 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
A special numeric algorithm was developed to support the output signal
interpretation of the nine-wire probe. The following expressions for effective
cooling velocity U
eij
of the sensor j in the i-th array were used for the wires in
the xOy and xOz plane, respectively:
( )
2 2 2 2
W h sin V cos U U
ij eij eij
eij
+ = o o , (5.45)
( )
2 2 2 2
V h sin W cos U U
ij eij
eij eij
+ = o o . (5.46)
They are equivalent to formula (
2 2 2 2 2
0
2 2
0
2
b
ne
b
e
s
e
U h U U h cos U U + = + = u (3.60)),
given in the 3
-rd
chapter for the effective hot-wire cooling velocity. The expressions
(5.45) and (5.46) can be derived, under some simple assumptions related to the
mechanism of hot-wire cooling, from the well known cooling law (3.59) of
Jorgensen 1971 (
2 2 2 2 2 2
b t n e
U h U k U U + + = ), also discussed in the 3
-rd

chapter. The coefficient k varies from about 0.2, for hot-wire aspect ratio L/D =
200, to about zero for L/D = 600 (Champagne, Sletcher and Wehrmann 1967).
For the nine-sensor probe, the ratio L/D = 280 required accounting for the
tangential cooling of the wires. A good way to achieve this and to account the other
effects is the concept of an effective sensor cooling angle o
e
explained in the
section 3.3.4 (see also Bradshaw 1971,1975 and Bruun and Tropea 1980).
Introducing the effective angle accounts for tangential wire cooling and, therefore,
the tangential component U
t
can be formally neglected, while the normal
effective cooling component will be
e e ne
sin V cos U U o o = , for a sensor in
xOy plane. A similar expression will follow for the sensors in xOz plane what, after
replacing in the formula (3.59) of Jorgensen 1971, gives the expression (5.45) and
(5.46) for the effective cooling velocities.
The effective cooling velocity U
eij
of the sensor j of i-th array (i,j = 1,2,3) of
the nine-wire probe can be defined in terms of the three velocity components U
0
, V
0

and W
0
at the probe centre C
0
(see fig. 5.14c), six streamwise velocity gradients (in
the plane normal to the probe axis passing through the sensor centres), the distance
projected on this plane between the prong tips of any array, the individual sensor
effective angle o
eij
and the blockage coefficients h
ij
. The velocity components U
0
,
V
0
, and W
0
are expanded to the first order in a Taylor series about C
0
, giving the
following set of nine non-linear algebraic equations with nine velocity and cross-
stream velocity gradient component unknowns:
( ) 47 5
2
6 5 0 3
2
4 3 0 2 2 1 0 1
2
.
y
W
C
z
W
C W K
y
V
C
z
V
C V K
y
U
C
z
U
C U K U
ij ij ij
ij ij ij ij ij ij eij
(

|
|
.
|

\
|
c
c
+
c
c
+ +
+
(

(
|
|
.
|

\
|
c
c
+
c
c
+

+
|
|
.
|

\
|
c
c
+
c
c
+ =

MULTIPLE HOT-WIRE PROBES 111
__________________________________________________________________________________________________________________________

It is therefore clear that nine hot-wires are required in order to account to first order
for the non-uniformity of the velocity field across the plane perpendicular to the
probe axis. In the equation (5.47), transversal component V is interchanged with W
for hot-wires j = 2,3 (see fig. 5.14b).
The twenty-seven coefficients K
ijk
have to be determined experimentally, from
calibration. The constants K
ij1
, by their nature, are cosines of the effective angles
o
eij
, the K
ij2
coefficients are sines of the effective angles, while the K
ij3
coefficients
account for aerodynamic blockage as described above. The C
ijl
(l = 1,2,3,4,5,6)
constants are positive or negative fractions of the projected prong spacing h for a
given geometry of arrays.
A unique numeric algorithm is applied to solve these equations. Expression
(5.47) represents a non-linear algebraic function for which the imaginary roots can
be ignored. However, there still exists the problem of choosing between the
remaining possible real roots, known as the uniqueness solution problem. The
critical value (V/U)
cr
, which determines the half-angle of the uniqueness cone for
this probe was found in the form:
1 2
2
2
1
ei i
ei ei
cr
cos h
cos sin
U
V
o
o o
=
|
.
|

\
|
. (5.48)
In the ideal case, where all the effective angles are the same and equal to 45
0
and
there is no aerodynamic blockage effect, i.e. h
i2
= 1, this value of (V/U)
cr
is 0.5,
corresponding to an angle of attack of velocity vector to the probe axis of 26.6
0
.
For the real values of o
eij
and h
ij
, determined from calibration, the average value of
(V/U)
cr
for all three arrays is about 0.36, corresponding to an angle of attack of only
19.8
0
, what is very close to the maximum values of V/U found by previous
measurements in the near-wall region. Within this range of attack angles that
defines the uniqueness cone, fluid velocity vectors and gradients of velocity
components physically occurring in the flow can be uniquely determined from the
solution of equations system (5.47). At the edge of uniqueness cone, besides the
problem of uniqueness, the problems of measurement accuracy and convergence of
solution procedure also appear. At the measurement location closest to the wall,
defined by y
+
= 11, Balint, Wallace and Vukoslavevi 1991 found that 7.1% of
measured data do not converge. However, the percentage of non-converged data
drop to only 0.9% at y
+
= 27.
These authors used two approaches in order to check the measurement accuracy
of vorticity by this probe. The first is the comparison of the values of y U c c ,
obtained by time-averaging the instantaneous values of ( ) y t U c c across the
boundary layer, with the derivative of the average velocity distribution y / U c c .
This comparison is illustrated in fig. 5.15, showing quite good agreement between
the experimental data and theoretical curves.
112 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
An indirect second
criterion of the probe
measurement accuracy was
to determine what spurious
value of vorticity
components the probe
measures in nominally
irrotational (potential) flow.
Using the data obtained in
the (irrotational) core-flow
above boundary layer, the
probability density
functions of spurious
vorticity vector components
were computed and
compared to the values
obtained deep in a
boundary layer, at y
+
= 41.
As can be seen in fig. 5.16, the
signal-to-noise ratio was in the
range from 12 to 25.
In addition to turbulent
boundary layer measurements of
Balint, Wallace and Vukoslave-
vi 1991, this type of nine-sensor
probe was used by Ong 1992 and
Piomelli, Ong, Wallace and
Ladhari 1993 to study momentum
and mass transport by vortex
motions in a turbulent boundary
layer. Wallace, Balint and Ong
1992 used grid flow data obtained
with this nine sensor probe,
together with mixing layer and
boundary layer data bases of
Balint et al., 1989, to examine
scalar properties of helicity
density.
Honkan 1993 has also
constructed a nine-sensor probe
capable of measuring all three
vorticity components. The sensors
in each array of his probe were of
P
D
F
(
O

x
)

0.02
0.00
0.04
0.02
0.00
0.06
0.04
0.02
0.00
0.06
-1000 1000 500 0 500
O
x
, O
y
, O
z
[1/s]
P
D
F
(
O

y
)

P
D
F
(
O

z
)


Fig. 5.16: Comparison of the measured PDFs
of vorticity components at y
+
= 41
in the boundary layer and in a free
non-irrotational stream. Adapted
from: Balint et al., 1991.
Vukoslav evi , Wallace
and Balint 1991
1.0
0.8
0.6
0.4
0.2
0.0
1 3 10 30 100 300 1000
_ +
c
U

------
c
y

y
+
[-]
Coles 1962
Spalding 1961

Fig. 5.15: Mean velocity gradient y / U c c across the
boundary layer. Adapted from: Wallace
and Foss 1995.
MULTIPLE HOT-WIRE PROBES 113
__________________________________________________________________________________________________________________________

orthogonal (Mercedes) configuration VP-3m, sketched in fig. 4.3. In general, he
built the same type of the probe as Wasmann and Wallace 1979 attempted
previously. This configuration increases the cone of uniqueness in comparison to T
geometry VP-3t, as it is discussed in chapter 4. In this case, a separate pair of
prongs supported each of nine sensors, eliminating any possibility of electrical
cross talk between different hot-wires.
He attempted to calibrate the probe using a look-up table mapping of relation
between the measured voltages and induced velocity vectors, but found out that
this approach was too complicated. Therefore, the probe was calibrated by
simultaneous pitch and yaw variation using Jorgensens 1971 expression for
effective cooling velocity. The flow was assumed to be uniform over the sensing
area of each array, but the flow field variation between arrays was accounted.
Having in mind that the sensing diameter of each array is greater then the distance
between array centres, these two assumptions are contradictory. Honkan 1993
measured simultaneously the velocity and vorticity vectors in turbulent boundary
layer, at Re
u
= 2790, and over a delta-
wing at 15
0
angle of attack and compared
it with own boundary-layer
measurements, as well as with
measurements and direct numerical
simulations of other researches. However,
it seems that the probe spatial resolution
was not quite adequate: at y
+
= 12.5 in the
boundary layer, the distance between
array centres over which the gradients
were estimated were 6.4q. At the same
measuring location, the diameter of the
sensing volume of each array, over which
gradients of fluid velocity components
were neglected, was about 10.7q.
Tsinober, Kit and Dracos 1992
constructed a twelve-sensor probe
sketched in fig. 5.17. They added an
additional tungsten sensor of 2.5 m to
each array and used separated pair of
prongs for each sensor, making a total of
twelve hot-wires supported by 24 prongs.
The probe directional calibration was
performed by its pitching and yawing
over 81 angle combinations at each of 7
velocity magnitudes. Resulting
calibration data were used to construct a

1
2
3
4
ARRAY 1
ARRAY 2
1
.
4
m
m

1.4mm
y
z
ARRAY 3
(a)


y
z
x
o
=45
0

L=0.4mm
(b)

Fig. 5.17: Hot-wire probe of Tsinober,
Kit and Dracos 1992:
(a) an end view and
(b) a probe array with four
hot-wires.
Adapted from: Tsinober,
Kit and Dracos 1992.
114 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
look-up table by fitting the calibration data set with Chebychev polynomials,
what was an improvement in comparison to calibration procedure of nine-sensor
probe. However, in place of using the fourth sensor of each array to increase the
uniqueness cone angle, they averaged the velocity components obtained using
four combinations of three sensors of each array. They even did not mention any
investigation of uniqueness range of their hot-wire probe.
Fortunately, the probe was used in a grid flow characterised by low turbulence
levels (and, therefore, small attack angles of velocity vectors to the mean flow
direction), where their approach was probably good. However, using their
approach, the full advantage of the fourth sensor in the near wall region can not be
used. Like Honkan 1993, they assumed the flow to be uniform over the sensing
area of each array. The diameter of the projection of sensing area of each array was
0.7 times the distance over which the gradients were estimated. It was better
resolution then it was for Honkans probe, but still the dimensions of arrays
sensing area and its separations were of the same order. Neglecting the velocity
variations over one of them and taking it into account over the other is
contradictory and deserves further analysis. Tsinober, Kit and Dracos 1992
measured gradients of velocity components and vorticity of a turbulent grid flow
(Re
M
= 26580) at several different stations downstream of the grid mesh dimension
M. At distance of 90 M downstream, the separation of the probe arrays was 1.9
Kolmogorov scale q and the diameter of the array cross-stream projected area was
1.4q. Although these values are close, it seems that the probe resolution was
excellent. However, the RMS vorticity value measured at that location was only 17
s
-1
, resulting in signal which was of the same order as noise, considering typical
hot-wire sensitivity and noise level.
Shortly after finishing the nine-sensor probe Vukoslavevi and Wallace started
developing a twelve-sensor probe in order to increase the range of uniqueness and
the measurement accuracy. It was a logical next step, following the previously
designed probe with nine hot-wires. The probe was configured and constructed
differently from the probes of Tsinober, Kit and Dracos 1992 and Honkan 1993
and the data reduction method was completely different too. This probe was first
used, with permission of Vukoslavevi, by Nguyen 1993 for measurements in the
near-wake of circular cylinder. Measured output anemometer signals in his study
were reduced and interpreted with an error minimisation algorithm described by
Marasli, Nguyen and Wallace 1993.
A detailed description of the probe and corresponding data reduction algorithm
which gives the full advantage of the fourth sensor is reported by Vukoslavevi
and Wallace 1996. The advantage of adding the fourth sensor to each array became
clear from the analysis of the cooling law of any j-th sensor of an arbitrary array,
what can be expressed in general form as
( ) 0 =
j j
E , W , V , U F , (5.49)
MULTIPLE HOT-WIRE PROBES 115
__________________________________________________________________________________________________________________________

where U, V and W are the velocity components at a given location in the flow and
E
ij
are the voltage responses of each sensor at the same location. The shape of this
function is not universal as it was discussed in chapter 3. In the algorithm that
supports nine-sensor probe, Vukoslavevi, Wallace and Balint 1991 used a
special form of this function (5.45) - (5.46), based on effective angles. For three
sensors (j = 1,2,3), two components of the fluid velocity vector in equation (5.49)
can be eliminated and a single expression for the third V, in the present case,
remains in the following form:
0
3 2 1
= ) E , E , E , V ( F (5.50)
The shape of this function, for different values of V, is shown in fig. 5.18. For
each of the curves shown in this figure, the voltages E
1
, E
2
and E
3
are constant.
Thus, the (5.50) reduces to single-variable function
) V ( F ) E , E , E , V ( F =
3 2 1
. (5.51)
It is qualitatively
similar for different
types of cooling
expression (5.49), but
differs somewhat for
different sensor
orientations. The
uniqueness problem
appears in the range
for V/U > 0 where the
curve turns downward,
giving two possible
real roots for F(V).
One of these solutions
is always larger than a
value corresponding to
the maximum of the
curve (5.51), which is
designated as (V/U)
cr
.
This critical value
determines the range of uniqueness of any specific hot-wire probe, depending
primarily on its configuration, but also, on the applied fabrication technology and
procedure for signal interpretation. This means that the probe can be used in a flow
in which the largest value of (V/U)
max
is smaller than the critical value (V/U)
cr
.
Otherwise it is impossible to recognise which of the two possible solutions
physically occurs in the flow. The critical value can be determined analytically. In
the boundary layer investigation of Vukoslavevi et al., 1991 it was around 0.36
and even smaller close to the wall.
-1.0 1.0 0.5 -0.5 0.0
2.5
0.0
1.5
0.5
-0.5
-1.5
-2.5
F
(

V

)

[
-
]

V [m/s]
2
3
1
y(V)
VP-3t
z(W)
V/U=-0.36
V/U=-0.176
V/U=0
V/U=0.176
(V/U)
cr
=0.36

Fig. 5.18: Function curves F = F(V,E
1
,E
2
,E
3
), for various
V/U ratios. Adapted from: Vukoslavevi and
Wallace 1996.
116 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
The fact that (V/U)
max
can be close to the critical value may also lead to an
additional problem. Any error in measuring the values of E
j
voltage or the
dependence of the calibration constants on the variations of fluid velocity vectors that
is not taken into account, moves the F(V) curve up or down. As a direct consequence
of these movements, the intersection point of F(V) with the abscissa changes, what
results in an error in determining V. It is obvious that this error will be higher for
(V/U) close to (V/U)
cr
, where the curve is flat and nearly horizontal. In the worst
situations, the curve F(V) can even fall under the abscissa, giving no solution. The
number of no-solution occurrences increases as the wall is approached, due to the
decreasing U velocity component, which increases the (V/U) ratio. If the number of
non-converged data is too large, then essential information about the flow field will
be lost, and the resulting measurement statistics will be biased.
An interesting property of an array of three sensors is that the critical value of
V/U appears only in the range V > 0, for the geometrical arrangement shown in fig.
5.18 (right-down corner). This probe is characterised by a vertical sensor no. 1
mounted below horizontal V-sensors no. 2 and 3, in the negative y direction of V
component of fluid velocity vector. It is easy to show that the critical range will
shift to V < 0, if the wire no. 1 is above sensors no. 2 and 3 (e.g. if mounted in the
positive y direction of V component. Both situations are illustrated in fig. 5.19,
where F
low
(V)=F
1
(V) denotes sensor 1 below hot-wires 2 and 3 and F
hi
(V)=F
3
(V)
denotes hot-wire no. 1
above sensors no. 2 and 3.
The critical values of
V component of fluid
velocity vector are
negative in the second
case: the maximum
measurement error and
the no-solution occur-
rences will be for V < 0.
These facts provided the
impetus to the authors to
add an additional hot-wire
to each of three arrays of
the nine-sensor probe. In
such a way, having at the
same time sensors in both
the positive and negative
y (V) directions, both
F
low
(V) and F
hi
(V) curves
are available.

-1.0 1.0 0.5 -0.5 0.0
2.5
0.0
1.5
0.5
-0.5
-1.5
-2.5
F
(
V
)

[
-
]

V [m/s]
F
1
(V)=F
hi
(V) F
3
(V)=F
low
(V)
V/U=const.
-0.36 +0.36

Fig. 5.19: Criteria function for different positions of
the vertical wires of plus probe:
________
( ) V F
low
wire no. 1 is below the
sensors no. 2 and 3 and
_ _ _ _
( ) V F
hi
sensor no. 1 is above hot-
wires no. 2 and 3.
Source: Vukoslavevi and Wallace 1996.
MULTIPLE HOT-WIRE PROBES 117
__________________________________________________________________________________________________________________________

Following this analysis, Vukoslavevi
and Wallace 1996 have developed a special
algorithm for the new vorticity probe with
twelve hot-wires (the probe was later
designated as WP-12+(G)). Although much
more complex in comparison to nine-wire
probe, this configuration enabled choosing
the appropriate curve F(V) (F
1
(V) or F
3
(V))
depending on the algebraic sign of the
instant V component of fluid velocity
vector. This new probe consists of three
arrays of four hot-wires, oriented at 45
0

according to the probe axis, as can be seen
in the sketches in fig. 5.20.
Besides the fourth sensor, added to each
of three arrays of the previous version of
the probe WP-9t(G) containing 3 x 3 = 9
sensors, there is an additional important
advantage of the new twelve-wire probe
WP-12+(G). The common central prong for
each array of the nine-sensor probe have
been replaced by four separated thin central
prongs, as Tsinober et al., 1992 and
Honkan 1993 done.
Thus, the new probe contains twelve-
sensors and twenty-four prongs placed in a
sensing area of approximately a 2.5mm diameter. In fact probe dimensions are
practically the same as for the old configuration WP-9t(G), although the distance
between the prongs is a little smaller, being 0.44mm compared to 0.5mm for the
nine-sensor probe. However, this is large enough to provide a reasonable aspect
(L/D) ratio of about 250 for tungsten hot-wires: of 2.5 m diameter.
The prongs are produced of stainless steel, 0.25mm in diameter, and copper and
silver-plated all the way up to their tips to reduce their resistance. The prong tips
are not plated in order to make welding of the wire to its prongs easier and more
robust. To reduce the blockage of the sensing area, the tips of the prongs are
tapered to about 50 m. Finally, the prongs are drawn through ceramic insulating
tubes, which were placed and sealed in a common stainless steel tube and plastic
probe holder.
The choice of the WP-12+(G) probe dimensions has been guided by the same
two resolution criteria as for the nine-sensor probe. The average spacing between
hot-wires centres is about 6.3q at y
+
= 11.2 and Re
u
= 2685, just as it was for the
nine-sensor probe. Following the analysis of Wyngaard 1969, assuming Pao's

1
2
3
4
1
2
3
4
1
2
3
4
ARRAY 1
ARRAY 3 ARRAY 2
C
0
1
.
1
4

0.09 0.44
1.32
y
0
z
0

(a)

y
z
x
o
Dimensions
are in mm!
(b)

Fig. 5.20: Schematics of the twelve-
wire probe WP-12+(G):
(a) end-view projection;
(b) a probe array.
Adapted from: Vukosla-
vevi and Wallace 1996.
118 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
1965 spectrum, and the
experiment of Klewicki
and Falco 1990, it seems
that measurement of RMS
values of O
z
will be
attenuated by ~10-15% at
Re
u
= 2100, just as was
estimated for the nine-
sensor probe WP-9t(G)
(Balint, Wallace and
Vukoslavevi 1991).
The probe WP-12+(G)
was tested in order to
evaluate the measurement
errors of the fluid velocity
components, as well as the
spurious gradients of
velocity vector components
and spurious vorticity
components measured by
the probe in irrotational
flow where this should, in
principle, be zero. In fact,
this test is a good measure
of the accuracy of the
measuring system. The
probe was tested for flow
conditions that cover
combinations of U, V and
W, expected in the
turbulent boundary layer.
Fig. 5.21 shows the results
obtained for a zero yaw
angle (W = 0) and different
pitch angles in the range
= 20
0
with

U = 1.43
m/s. In fig. 5.22, the results
for yaw angles equal to
pitch angles in the range
20
0
are presented for the
same

U . Similar results are obtained for zero pitch and variable yaw angle,
as well as for the pitch angles equal to negative values of yaw angles.

U

[
m
/
s
]

1.16
1.32
1.48
1.64

(a)
V
,
W

[
m
/
s
]

-0.80
-0.48
-0.16
0.16
0.48
0.80

(b)

O
x

[
1
/
s
]

0 -10 10 20 -20
Pitch angle - [
0
]
-30
-10
10
30

(c)
Fig. 5.21: Induced and measured velocity and
vorticity components during pitch angle
variation: (^) induced values, (+) probe
WP-12+, ( ) probe WP-9t. Adapted
from: Vukoslavevi and Wallace 1996.
MULTIPLE HOT-WIRE PROBES 119
__________________________________________________________________________________________________________________________

It is evident that values
of the velocity components,
measured by the twelve-
sensor probe, match the
induced values within about
0.7 %. To compare the
relative accuracy of the
twelve and 9-wire probes,
the signals from hot-wire
no. 3 of each array were
neglected, transforming the
twelve-sensor probe to a
nine-wire one. The
corresponding data are
shown together with twelve-
sensor probe data in the
same figures 5.21 and 5.22.
For pitch and yaw
angles smaller than 15
0
, the
results agree well, but above
15
0
, an error of up to 5%
appears in fig. 5.21a,b for
the nine-sensor probe.
Moreover, there are some
pitch and yaw angles
for which the nine-sensor
probe algorithm does not
converge at all, e.g. = u =
-15
0
and = u = -20
0
in fig.
5.22a,b,c. This is not
surprising, because these
angles correspond to the
most critical angular range,
where the induced cross-
stream components of fluid
velocity vector are of the
order of U/2, corresponding to a cone angle of attack of about 27.5
0
. This is well
outside the uniqueness range of the nine-sensor probe WP-9t(G) (estimated to be
about 17.5
0
at this flow velocity magnitude). Similar results are obtained neglecting
the signals from sensor no. 3 on the other two arrays.
Superior performances of the twelve-sensor vorticity probe WP-12+(G),
compared to the probe WP-9t(G) that contains nine wires (simulated by neglecting
the signal from hot-wire no 3 see fig. 5.20) is also clear from the fig. 5.23. A
U

[
m
/
s
]

1.32
1.48
1.64
1.80
(a)

V
,
W

[
m
/
s
]

-0.80
-0.48
-0.16
0.16
0.48
0.80
(b)

O
x

[
1
/
s
]

0 -10 10 20 -20
-30
-10
10
30
(c)
Pitch angle equal
to jaw angle.

Fig. 5.22: Induced and measured velocity and
vorticity components during pitch
angle variation: (^) induced values, (+)
probe WP-12+(G), ( ) probe WP-9t(G).
Adapted from: Vukoslavevi and
Wallace 1996.
120 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
slope of the curve
F
1
(V), obtained from
the response equations
for effective cooling of
the sensors no. 1, 2 and
4, provides higher
accuracy of root
evaluation than F
3
(V),
obtained from response
expressions for ef-
fective cooling of hot-
wires no. 2, 3 and 4.
The F
1
(V) curve, for
the case of an induced
value of V=-0.24m/s,
has two intersection
points with abscissa:
one of them at V=-
0.23m/s, giving the
physical solution, and
the other at V= -0.5m/s.
The solution is,
obviously, not unique.
However, in contrast to
F
1
(V), the F
3
(V) curve
has only one
intersection point, giving a unique and accurate solution for fluid velocity
component V. Increasing the induced value of V up to -0.46m/s, the F
1
(V) curve
becomes more tangential to the abscissa giving no intersections, while F
3
(V) gives
a unique and accurate solution for V.
The spurious streamwise component of vorticity O
x
measured by the twelve-
sensor probe in the irrotational flow during calibration (fig. 5.21c and 5.22c) were
under 10s
-1
, except for one case ( = u = 15
0
) where it was 18s
1
. This spurious
value corresponds to an accuracy of 0.7% in velocity measurements: simultaneous
velocity measurement errors of +0.7% at one array and -0.7% at the second array,
with a spacing of 1.14mm, results in a gradient error of 18s
-1
. The maximum nine-
sensor probe error was of the same order, but occurred much more often, with
some cases not converging at all, e.g. = u = -20
0
, -15
0
, 20
0
. Thus it is obvious that
a small velocity error corresponds to a significant spurious vorticity error. Having
this in mind, the signal-to-noise ratio of measurements of Tsinober, Kit and
Dracos 1992 behind the grid, with the RMS value of vorticity of 17s
-1
, is of the
same order.
-1.0 1.0 0.5 -0.5 0.0
2.5
0.0
1.5
0.5
-0.5
-1.5
-2.5
F
(
V
)

[
-
]

F
3
(V)=F
low
(V)
V [m/s]
2
4
3
y(V)
VP-3t
1
z(W)
V/U=const.
F
1
(V)=F
hi
(V)

Fig. 5.23: Functions F(V) for different combinations of
hot-wires of the probe WP-12+(G), arranged
according to fig. 5.20:
_ _ _ _
F
1
(V)=F
hi
(V), for the wires no. 1, 2 and
4 (upper sensor, no. 1, is used);
_______
F
3
(V)=F
low
(V), for the wires no. 3, 2 and
4 (lower sensor, no. 3, is used).
Adapted from: Vukoslavevi & Wallace 1996.
MULTIPLE HOT-WIRE PROBES 121
__________________________________________________________________________________________________________________________

The probability density distribution
function (PDF) of the spurious vorticity
component O
x
, measured in the
potential core of the calibration-facility
jet (fig. 5.25-up) with the probe WP-
12+(G), is shown in fig. 5.24. The
width of this PDF is another measure of
the noise that characterises the entire
measuring system. The longitudinal
vorticity has rare occurrences of
spurious values of about 25s
-1
, with a
probability of 98.74% of being in the
range of 15s
-1
. Similar results are
obtained simulating the nine-sensor
probe WP-9t(G) by neglecting the
signal from the vertical sensor no. 1 or
hot-wire no. 3 of each array. The
maximum spurious value was ~25s
-1
,
with a probability of 95.65% of being in
the range of 15s
-1
. The spurious value
RMS(O
x
) is 6.3s
-1
for the probe
WP-12+(G) and 7.5s
-1
for the nine-
sensor one. These values indicate a high signal-to-noise ratio when compared to
peak values of O
x
of about 1000s
-1
and RMS(O
x
) values of about 200s
-1
near the
wall in the boundary layer. The maximum spurious value of 90s
-1
that Balint,
Wallace and Vukoslavevi 1991 obtained is reduced by a factor of 3 in this
experiment, due to improvements in hot-wire probe design and calibration
procedure.
In order to compare performances of the new twelve-sensor probe to the
analogue performances of the previously designed configuration with nine wires,
data were taken at several positions near the wall in a low-speed open return wind
tunnel, designed to create a thick turbulent boundary layer (fig. 5.25-down). The
free stream speed can be controlled over a 1-7m/s range with a free-stream
turbulence level of ~0.5%.
As it was described above, the twelve-sensor probe was treated as a nine-sensor
probe here also, by rejecting signals from vertical hot-wire no. 1 or 3 of each array.
In this way a simultaneous analysis of the same data in the vorticity twelve-wire
probe (WP-12+(G)) and probe (WP-9t(G)) configurations was possible and a
straight-forward comparison of both probes under the identical flow and
experimental conditions could be achieved. In addition, the results were compared
to the previous measurements with the nine-sensor probe of Balint et al., 1991 that
possessed common central prongs at each array. The free stream velocity was set to
3.47m/s, what resulted in thick turbulent boundary layer characterised by
P
D
F
(
O

x
)


[
%
]

20
15
10
5
0
-30 -20 20 30 0
O
x
[1/s]

Fig. 5.24: PDF of the spurious
longitudinal vorticity:
+ probe WP-12+(G) and
probe WP-9t(G).
Adapted from: Vukoslavevi
and Wallace 1996.
122 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Re
o
= 27650. The room
temperature was controlled
in the limits of 20
0
C
0.5
0
C, giving air viscosity
v of 1.5110
-5
m
2
/s.
Applying Clausers
method, Vukoslavevi
and Wallace 1996 found a
friction velocity to be u
t
=
0.135m/s. The sensors were
heated by constant
temperature hot-wire
anemometer circuits built
by A.A. Lab Systems with
a flat response up to about
4000Hz, well enough over
the highest expected
frequency in the specified
flow.
One of the
characteristics that
demonstrate the probe
performances is the number
of non-converged data in
the measured time-series.
As it is shown in figs. 5.18
and 5.19, when the flow
conditions approach the
critical value two problems
appear: the measurement
accuracy is decreased and no convergence for high V/U or W/U values can occur.
The percentage of these non-converged time steps is a good indication of the flow
being close to the critical condition and, therefore, one indication of measurement
accuracy. These non-converged data are lost, influencing the measured statistical
properties of the turbulence. Additionally, as the critical point is approached the
accuracy goes down and the measurement error is increasing.
Fig. 5.26 shows the percentage of non-converged data from the total data
sample at different locations from the wall, for the twelve and nine-sensor probes.
The percentage of non-converged data for the twelve-sensor probe goes from 5% at
y
+
= 14.2 to 2.4% at y
+
= 18.7 and approaches zero further from the wall. In the
case of the nine-sensor probe, the percentage of non-converged data is about three
times larger.


Fig. 5.25: Hydrodynamic laboratory, University of
Maryland - USA:
- calibration facility (up)
- open return wind tunnel for boundary
layer research (down).
Photographed in 1991, under permission
of Prof. J. M. Wallace.
MULTIPLE HOT-WIRE PROBES 123
__________________________________________________________________________________________________________________________

Fig. 5.27 presents a short
real-time series of longitudinal
U velocity component,
measured in the near wall
region using twelve (WP-
12+(G)) and nine-sensor
(WP-9t(G)) hot-wire probes.
Although there are significant
gaps in the data because of
non-convergence of data
interpretation originating from
the probe WP-9t(G), its
converged values compare
well to the values measured
by WP-12+(G). The same
conclusion applies to the V
and W velocity components
except for rare occasions. This
demonstrates a good design of
older probe with nine hot-wires: if the numeric interpretation procedure converge,
the results generally converge to the correct solutions.
90 130 170 210 250 290
t [ms]
U

[
m
/
s
]

0.90
1.40
1.90
0.65
1.15
1.65

Fig. 5.27: Time series of longitudinal U velocity vector component, measured in a
turbulent zero-pressure-gradient boundary layer by:
(
_____
) twelve-sensors probe WP-12+(G) and
( ) nine-wire probe WP-9t(G).
Adapted from: Vukoslavevi and Wallace 1996.
0 10 20 30 40 50
y
+
[-]
0
4
8
12
16
N
C
D

[
%
]


Fig. 5.26: Non-converged data at different
locations from the wall:
+ twelve-wire probe WP-12+(G),
nine-sensor probe WP-9t(G).
Adapted from: Vukoslavevi and
Wallace 1996.
124 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Some of the results obtained
with the nine-sensor probe that
were not consistent with the
constraints of symmetry, like
skewness factor of probability
density distribution of the
vorticity component O
x
, are
consistent in measurements
with the twelve-sensor probe.
For measuring mean and RMS
values, both the twelve and
nine-sensor probes have
acceptable accuracy. But, for
some higher-order moments,
the probe WP-12+(G) is much
more accurate, especially in the
flows with a high ratio of V/U
or W/U, or both of them
simultaneously. It appears that
the upper value of these ratios is
for the nine-sensor probe
around 0.3, and for the twelve-
wire probe it is about 0.45.
Therefore, the twelve-sensor
probe is more suitable for
measurement of statistical
properties of the turbulent
boundary layer velocity and
vorticity fields than the nine-
wire one, especially in the near
wall region where the nine-
sensors probe does not behave
properly.
Complementary analysis of
the experimental data of Vuko-
slavevi and Wallace 1996 is
available in Vukoslavevi and
Petrovi 1998. They computed
the probability density
distribution functions (PDFs) of
the vorticity vector components
at various locations in the near-
wall region of a turbulent
0.0
1.0
2.0
3.0
4.0
5.0
y
+
=14.22
0.0
1.0
2.0
3.0
4.0
5.0
y
+
=18.72
y
+
=27.72
0.0
1.0
2.0
3.0
4.0
5.0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
-0.5
0.0
0.5
y
+
=41.22
O
x
+
O
y
+
O
z
+
P
D
F

[
%
]

O
+
[-]
0.75
-0.75
P
D
F

[
%
]

P
D
F

[
%
]

P
D
F

[
%
]


Fig. 5.28: PDFs of vorticity components, nor-
malised by u
t
and v. Adapted from:
Vukoslavevi and Petrovi 1998.
MULTIPLE HOT-WIRE PROBES 125
__________________________________________________________________________________________________________________________

boundary layer (fig. 5.28).
Superimposed on the PDF of
spurious vorticity in a potential
core flow, as given in fig. 5.29,
they clearly show the high
accuracy of measurements.
Namely, the maximum values
of the vorticity components are
very much larger then largest
value of spurious vorticity
components arising from
facilities and instrumentation
system errors. The narrowing
of the range of vorticity
component values with
increasing the distance from
the wall is evident, showing
decreasing the vorticity
magnitude and, therefore, its
RMS values. The PDF of
longitudinal vorticity
component O
x
, as well as
transversal component O
y

seems to be symmetric. It means that the probabilities of appearing the negative
and positive values are equal and, therefore, the skewness factor of these two
components of vorticity vectors should be zero. The skewness factor of PDF(O
x
)
measured with 12-sensor probe is compared to the one obtained by 9-wire probe of
Balint et al. 1991 (fig.5. 30).
The skewness factor
obtained with the probe WP-
12+(G) is close to zero in the
near wall region, what is in
accordance with the symmetry
of PDF. It makes sense
because the appearance of
negative and positive values of
O
x
should be of the same
probability. It was not the case
with the nine-wire probe WP-
9t(G), where an erroneous
negative skewness is present in
the near wall region.
0
5
10
15
20
25
30
35
40
-0.75 -0.5
0.0
0.5 0.75
O
x
+

[-]
P
D
F

[
%
]

y
+
=14.22
pot. core

Fig. 5.29: PDFs of O
x
, at y
+
=14.22 in the
turbulent boundary layer and in the
potential core, measured by the probe
WP-12+(G). Adapted from:
Vukoslavevi and Petrovi 1998.
-1.0
-0.5
0.0
0.5
0 10 20 30 40 50
y
+
S
O

x
+


Fig. 5.30: Skewness factors of longitudinal
vorticity O
x
, measured by:
+ twelve-wire probe and
nine-sensor probe.
Adapted from: Vukoslavevi and
Petrovi 1998.
126 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
This clearly demonstrates the
superiority of twelve-wire probe
compared to the nine-sensor probe. An
asymmetry of PDFs of O
z
is evident,
especially in the near wall region but
not as much as it was in measurements
of Balint, Wallace and Vukoslavevi
1991. Following the same conclusion
based on O
x
measurements, the results
obtained by twelve-sensor probe should
be much more reliable.
Therefore, these results confirm
clear superiority of the probe WP-
12+(G) in vorticity measurements,
comparing to the nine-sensors probe
WP-9t(G), the only instrument that is,
besides twelve-wire probe, capable of simultaneous measurements of all three
velocity and vorticity components. Some erroneous asymmetries, that are apparent
in results obtained with nine-wire probe, are not present in the case of WP-12+(G)
measurements. From the PDFs of vorticity components, presented in fig. 5.28, it is
evident the rate of vorticity decreasing with increasing the distance from the wall,
as well as the probability of the rotation directions.
Besides hot-wire probes described in fig. 5.20, Vukoslavevi and Wallace have
developed a new configuration WP-12q(G), with three quadrate arrays. This probe
is, at present, in the testing procedure.

Turbulent vortex structures in a free turbulent round jet. Source: Van Duke 1982.

1
2
3
4
ARRAY 1
ARRAY 3 ARRAY 2
C
0

1
.
0
2

1.16
y
z
WP-12q
1
2 4
1
2 4
0
.
3
5

0
.
4
8

0
.
3
4

0.08

Fig. 5.31: A sketch of the hot-wire
probe WP-12q(G).













6 6. . O OP PE ER RA AT TI IO ON NA AL L C CH HA AR RA AC CT TE ER RI IS ST TI IC CS S O OF F T TH HE E
V VO OR RT TI IC CI IT TY Y- -T TY YP PE E A AN ND D H HO OT T- -W WI IR RE E P PR RO OB BE ES S
S SP PE EC CI IF FI IE ED D F FO OR R V VE EL LO OC CI IT TY Y M ME EA AS SU UR RE EM ME EN NT T


T Th he er re e i is s n no o g gr re ea at te er r i im mp pe ed di im me en nt t t to o p pr ro og gr re es ss s i in n t th he e s sc ci ie en nc ce es s t th ha an n t th he e d de es si ir re e t to o s se ee e i it t
t ta ak ke e p pl la ac ce e t to oo o q qu ui ic ck kl ly y. .
( (G Ge eo or rg ge e C Ch hr ri is st to op ph h L Li ic ch ht te en nb be er rg g, , 1 17 74 42 2- -1 17 79 99 9) )


6.1 MEASUREMENT ACCURACY OF THE MEAN STATISTICAL
PARAMETERS OF TURBULENCE VELOCITY FIELD
Defining the optimal number of hot-wires and their configuration represents the
non-avoidable crossroad of all experimental studies involving the thermal
anemometry. The goal is to use the simplest possible probe that still provides
acceptable measurement accuracy of the flow properties. However, most of
reported hot-wire probe comparison studies were incomplete. Some of them did not
provide detailed results of testing, like Nithianandan, Jones and Adrian 1987 for
example. The others have tested different probes non-simultaneously (like Chang,
Adrian and Jones 1983) or compared them only in the nominally irrotational flow
(Petrovi and Vukoslavevi 1995). Generally, it can be concluded that, in almost
all reported comparison-studies, different hot-wire probe configurations were not
tested simultaneously at the same measuring point. Consequently, hot-wire
anemometer output signals of compared probes were not spatially and temporally
identical. In addition, mainly the older types of probes, with up to three hot-wires,
have been tested.
Simultaneous testing of the various configurations, positioned on exactly the
same measuring location in the turbulent flow, represents the most reliable
approach for evaluating the optimal number of hot-wires and their optimal
configuration. It provides the most realistic working conditions, with examined
probes exposed to spatially and temporally identical flow. However, these
experiments are very complex, demanding application of special multiple hot-wire
128 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
probes, which became operationally available at the beginning of current decade
(Wallace and Foss 1995). Therefore, very small number of such experiments has
been reported. Comparison of the nine-wire and twelve-sensor vorticity probes,
performed by Vukoslavevi and Wallace 1996 and Vukoslavevi and Petrovi
1998, currently represent the very rare exceptions of this practice, especially in the
area of multiple sensors configurations.
The analysis of measurement accuracy of vorticity-type and velocity-type hot-
wire probes is presented in this chapter. The influence of neglecting the velocity
gradients over the probe sensing volume, on the measured values of turbulent
velocity field statistical parameters, is analysed and compared to corresponding
influence of neglecting the one of fluid velocity vector components. The final goal
was to experimentally check the validity of applying the vorticity probes for
measurement of turbulent velocity field statistics, which can be verified only
through increased measurement accuracy that should balance problems connected
with their complex structure.
Presented results are based mainly on the studies of Petrovi 1996a and Vuko-
slavevi and Petrovi 1997. They processed the twelve-wire WP-12+(G) probe
(fig. 5.20) signals, as Vukoslavevi and Wallace 1996 done also. However, the
approaches were different, because the latter two researchers compared only
vorticity probes, with nine and twelve wires. The various anemometer probes were
emulated by selective taking
into account different
combinations of the output
anemometer signals of the
probe WP-12+(G). Thus, the
following probes were tested
simultaneously (fig. 6.1):
two-wire probes placed in
vertical VP-2v/V and
horizontal plane VP-2v/H,
three-sensor configurations
VP-3t/T and VP-3t/, probe
VP-4+ with four hot-wires
and original vorticity
configuration WP-12+(G).
Algorithm for
interpretation of the
anemometer output voltages
of the vorticity probe
WP-12+(G) was slightly
modified to provide results
that correspond to the centres
of its three arrays, which are

4
2
1
3
4
2
3
4
2
1
VP-3t/

)
VP-2v/H(VH)
4
2
1
3
VP-2v/V(VV)
VP-3t/T(T)
VP-4+(+)

Fig. 6.1: Software-emulation sketches of the tested
probes. Adapted from: Petrovi 1996a.
MULTIPLE HOT-WIRE PROBES 129
__________________________________________________________________________________________________________________________

spatially identical to the centres of emulated configurations with two and four
wires. This approach is different from Vukoslavevi and Wallace 1996, who
calculated fluid velocity and vorticity vectors at the centre of the sensing area that
corresponds to all twelve sensors. All applied algorithms are described in the
chapter no. 7.
The output anemometer signals of the probe WP-12+(G) originate from the
buffer region of a low-speed zero-pressure-gradient turbulent boundary layer,
generated in the wind tunnel (fig. 5.25) in the Hydrodynamics laboratory of the
Department of mechanical engineering at the University of Maryland, USA. The
flow regime was identical as in the case of Ballint, Wallace and Vukoslavevi
1991, giving the Kolmogorov micro-scale of q = 0.192 mm at y
+
= 11.2, i.e.
sensors length was 3.2q and the average spacing between the sensor centres over
which gradients were estimated was 6.3q. Going upward, the spatial resolution was
better. Following Wyngaard 1969 and Falco 1983, the probe resolved all but the
smallest turbulent scales in the boundary layer of this flow.
Hot-wire probe WP-12+(G) was calibrated in the potential core of a round jet
(fig 5.25-up). Magnitude of induced calibration velocity vectors was varied in the
range between 0.4m/s and 2.8m/s. Angular sensitivity of the WP-12+(G) vorticity
probe was evaluated by four angular calibrations at 0.6m/s, 1.0m/s, 1.4m/s and
2.0m/s. Using appropriate numerical methods, calibration signals were selectively
used to reproduce the components of induced calibration velocities. Under regular
conditions all emulated probes achieved very high measurement accuracy, with
relative errors below 1%.
Typical parameters of the turbulent boundary layer were:
- free-stream velocity: U

= 3.47 m/s;
- shape factor: H = 1.29;
- friction velocity: u
t
= 0.135 m/s;
- momentum thickness: u = 12.1 mm;
- displacement thickness: o
-
= 15.6 mm;
- boundary layer thickness: o = 125 mm;
- Re numbers: Re
u
= 2685 and Re
o
= 27650.
Real turbulence data were sampled at 1kHz in the turbulent boundary layer, at
various distances from the wall. A special DATA TRANSLATION analogue-to-
digital converter, with twelve simultaneous sample-and-hold channels of sixteen-
bit resolution was used, together with twelve-channel hot-wire anemometer.
Figs. 6.2-6.5 illustrate real-time measurement of turbulent velocity field by
different hot-wire probes emulated by selective application of the WP-12+(G)
probe output signals. Fig. 6.2 presents a At = 200 ms sequence of instantaneous
fluid velocity components at y
+
=7.15, viewed by two-sensor probes placed in the
vertical (VV) and horizontal plane (VH) and "triple" probes (T and ).
130 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
U
(
t
)


[
m
/
s
]

V
(
t
)


[
m
/
s
]

t (ms)
W
(
t
)


[
m
/
s
]

VV VH
T
1.5
1.0
0.5
-0.5
1.0
0.0
-0.5
1.5
1.0
0.5
-0.5
1.0
0.0
-0.5
0.0 0.0
200 100
-1.0 -1.0
0.0 0.0
DIVERGENCE PROBE -
DIVERGENCE PROBE -

Fig. 6.2: Instantaneous values of fluid velocity vector components, simultaneously
measured in a boundary layer at y
+
= 7.15 by V and T probes. Hot-wire
probes are designated according to fig. 6.1.
Orientation of fluid velocity vectors is defined by ratios of V/U and W/U and
presented in fig. 6.3. The areas out from angular acceptance ranges for triple
wire probes (estimated to be in the range of about 17.5
0
) are shaded. The
differences between values of velocity components, measured by different probes,
MULTIPLE HOT-WIRE PROBES 131
__________________________________________________________________________________________________________________________

are evident for U and V component, while values of lateral component W are much
closer to each other in all cases. The same figure also illustrates the low efficiency
of triple probes near the wall, caused by narrow angular acceptance range.
-20
-10
0
10
-40
-30
-50
30
VV T
-20
-10
0
10
-40
-30
-50
30
a
t
a
n
(
V
/
U
)


[
0
]

20 20
DIVERGENCE PROBE

200 50 100 150
t (ms)
20
30
-20
-10
0
10
VH

T
DIVERGENCE PROBE
50
20
30
-20
-10
0
10
50
20 20
-30 -30
0
a
t
a
n
(
W
/
U
)


[
0
]


Fig. 6.3: Instantaneous orientations of the fluid velocity vectors shown in fig. 6.2.
Turbulent velocity components measured at y
+
= 10.14 are presented in the figs.
6.4 and 6.5. The instantaneous values of U, measured by (VV and VH) probes with
two wires, four-sensor (+) configuration and original twelve-wire probe (G), are
similar. Except for the rare occasions, the difference is under 4% and is coming
from either gradients or one of the components neglected. The instantaneous W can
132 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
be in error up to 50% of its maximum values due to velocity gradients neglecting,
while it is practically the same no mater whether V component is neglected or not,
showing similar trend in time in all cases.
V
(
t
)

m
/
s

0.0
W
(
t
)

m
/
s

-0.5
0.0
-0.5
0.0
250 300 350 400 450 500 550
t [ms]
0.5
0.5
-0.5
-0.5
0.5 0.5
1.0
1.5
2.0
0.5
0.0
U
(
t
)

[
m
/
s
]

VV VH
G
+
1.0
1.5
2.0
0.5

Fig. 6.4: Instantaneous values of turbulent velocity components, simultaneously
measured by V and + hot-wire configurations and by original twelve-wire
probe in a boundary layer at y
+
= 10.14. Probe designations follow fig.
6.1. Source: Vukoslavevi and Petrovi 1997.
MULTIPLE HOT-WIRE PROBES 133
__________________________________________________________________________________________________________________________


VV + G
a
t
a
n
(
V
/
U
)

[
0
]

-40
-30
0
10
30
-40
-30
0
10
30
20
a
t
a
n
(
W
/
U
)

[
0
]

-40
-30
10
30
20
-40
-30
0
10
30
250 300 350 400 450 500 550
t [ms]
VH + G
0
-10
-20
-20
-10
20
-10
-20
-10
-20
20

Fig. 6.5: Instantaneous orientations of velocity vectors presented in fig. 6.4, as
viewed by various hot-wire probes designated according to fig. 6.1.
The instantaneous values of transversal V component also show similar trend in
all three cases, although not as much as longitudinal U and lateral W components.
However, these instantaneous values of V are mutually shifted, what results in
different mean values shown in the fig. 6.7. The discrepancies from values
measured by twelve-sensor WP-12+(G) probe are bigger if only two sensors are
used, in comparison to "triple" (T configured) probe. The reason is not in
neglecting the lateral W component, what has to be done in the case of two-sensor
probe (V configuration). It originates from the worse spatial resolution of two-
sensor probe in the y direction. If neglecting of W component had any significant
influence, similar effect would appear in W measurement when V component is
134 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
neglected. Therefore, the errors in measuring instantaneous values of turbulent
velocity field components are more influenced by neglecting of fluid velocity
gradients then any of its spanwise (transversal or lateral) components.
Measuring of streamwise (longitudinal) U component is mainly influenced by
gradients cV/cy or cV/cz, while the transversal V component is influenced
primarily by cU/cy and W by cU/cz. It follows that the error in V measuring in the
turbulent boundary layer should be the highest, because the instantaneous gradient
cU/cy is much larger then other gradients in the near-wall region. Fig. 6.5 shows
that measured velocity vectors were mainly in the acceptance range of tested
probes, what means that operational (working) conditions were regular. Therefore,
it became clear that differences are caused only by influence of different probes
geometry (i.e. neglected one of three fluid velocity vector components and/or fluid
velocity vector gradients).
Fig. 6.6 shows the mean longitudinal U velocity profile toward the normalised
distance y
+
= y/(u
t.
*
v) from the wall, measured by the probes containing two, four
and twelve hot-wires. The agreement of U component profile is very well for all
probes. Differences are of the order of 1%, what means that the error due to
neglecting of velocity gradients cancels out in the time-averaging procedure. The
error due to neglecting one of velocity components is small in this case, because
the ratio of V/U and
W/U is almost
always under 0.4,
while the RMS
values of V or W
components are less
then 0.2.
Petrovi 1996a
directly tested V
hot-wire
configuration (VP-
2v) in the
irrotational flow and
verified that
neglecting one of
the velocity
components results
in the higher
measured value of U
then the real one,
while the other
transversal V or
lateral W component
0
5
10
15
20
1 10 100
LINEAR ***
COLES 61 **
SPALDING 61*
9T/BVW 87
9T/BWV 91
12+/VW 96
VV VH
4+
U
+

[
-
]

y
+
[-]
U
+
=2.44 ln(y
+
) + 5** y
+
=U
+
***;
y
+
=U
+
+ 0.1108(e
0.4U+
-1 - 0.4U
+
)*
12+G

Fig. 6.6: Mean longitudinal U velocity, measured by:
VV/VH - two-wire probe VP-2v, placed in
the vertical/horizontal plane,
+ - four-wire configuration VP-4+,
12+G - vorticity probe WP-12+(G).
Source: Vukoslavevi and Petrovi 1997.
MULTIPLE HOT-WIRE PROBES 135
__________________________________________________________________________________________________________________________

is not so strongly affected. As can be seen in fig. 6.6, vertical VP-2v/V and
horizontal VP-2v/H probe gave slightly higher mean value U, than three-
dimensional probes with four (VP-4+) and twelve wires (WP-12+(G)). Their
difference is of the order of only 1%. In the case of three-dimensional probes,
neglecting the gradients (probe VP-4+) or taking them into account (probe WP-
12+(G)) practically did not influence the measured mean value U.
In the isothermal turbulent boundary layer flow, the mean value of W
component should be zero and mean V close to zero. Really, the measured values
of W are close to zero for all three probes, but the mean value of V is different (see
fig. 6.7). The best results are obtained when neither the gradients nor the W
component are neglected. However, a mean transversal V component different
from zero (V = 0) appears near the wall, showing that probe was slightly pitched
down at that location, or some of the wall effects cause the error.
Neglecting the velocity gradients generates error giving a spurious value of
mean V, up to 20% of mean U for two-wire probe. Absence of this error in the case
of mean W measurement and its decreasing further from the wall in the case of
mean V component is very indicative. It shows that mean non-zero V values are
caused by mean gradient cU/cy, while the instantaneous velocity gradients do not
significantly affect the measurement accuracy. The error is smaller for the triple
hot-wire probes VP-3T and VP-3, because of their better spatial resolution
(smaller overall dimension, i. e. height) in the y direction.
Although the
instantaneous
values were
sometimes
badly in error, it
was expected
that turbulence
intensities,
skewness S and
flatness F
factors would
not be affected
significantly by
neglecting the
velocity
gradients or one
of velocity
components.
The reason for
this expectation
originates from
-0.3
-0.2
-0.1
0.0
0.1
0
10 20 30 40 50
VH
4+
12+/G
V

[
m
/
s
]

y
+
[-]

Fig 6.7: Mean transversal V velocity component, as viewed
by different hot-wire probes. For captions see fig. 6.6.
136 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
the fact that all three velocity components in each of three analysed cases had
similar trend. However, it comes out to be truth for RMS values of all three
velocity components, shown in the fig. 6.8 and for skewness and flatness factors of
longitudinal U and lateral W component, presented in the figs. 6.9 and 6.10. For the
S and F factors of the transversal component V of the fluid velocity vector, that was
not the case.

1
2
3
u
'
/
u
t


0
1
2
WWd89 WWg89 WWm89
KMM87 SPAL.88 KE83
9T/BVW91 9T/BWV87 VV
VH 4+ 12+G
v
'
/
u
t


0
1
2
0 10 20 30 40 50 60 70 80
w
'
/
u
t


Fig. 6.8: Longitudinal u'/u
t
, transversal v'/u
t
and lateral w'/u
t
turbulence
intensities, normalised by the friction velocity u
t
.
VV/VH - probe VP-2v, placed in a vertical/horizontal plane,
+ - four-wire configuration VP-4+ and
12+G - vorticity probe WP-12+(G).
Adapted from: Vukoslavevi and Petrovi 1997.
MULTIPLE HOT-WIRE PROBES 137
__________________________________________________________________________________________________________________________

The RMS values of all three-velocity components, at different locations near the
wall, are nearly the same for all emulated probes. In addition, present results also
agree well with the results of Klebanoff 1954, Wei 1987, Balint, Wallace and
Vukoslavevi 1991, Spalart 1988, Kastrinakis and Eckelmann 1983 and Kim,
Moin and Moser 1987. The skewness S velocity factors are presented in fig. 6.9.
The skewness factors of U and W are the same for all three tested types of probes.
It follows that neglecting of velocity gradients or one of velocity components does
not influence significantly the third moments of probability density distributions of
U and W fluctuations.
-1
0
1
WW KMM87 KE83
9T/BVW91 9T/BWV87 9T/VW96
9

/VW96
12+/VW96
VV
VH 4+ 12+G
S
U


-0.5
0
1
S
V


-0.5
0
0.5
0 10 20 30 40 50 60 70 80
S
W

y
+

Fig. 6.9: Profiles of skewness factors of PDF of longitudinal (S
U
), transversal (S
V
),
and lateral (S
W
) velocity vector components, measured in a turbulent
zero-pressure-gradient boundary layer by:
VV - probe VP-2v, placed in a vertical plane,
VH - probe VP-2v, placed in a horizontal plane,
+ - four-wire configuration VP-4+ and
12+G - vorticity probe WP-12+(G).
Adapted from: Vukoslavevi and Petrovi 1997.
138 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Profiles of the third moment S
U
agree well with those of Balint, Vukoslavevi
and Wallace 1987, Balint, Wallace and Vukoslavevi 1991, Wei 1987, Wei and
Willmarth 1989 and Vukoslavevi and Wallace 1996. However, direct numerical
simulation of Kim, Moin and Moser 1987 and Kastrinakis and Eckelmann 1983
is considerably more negative (see fig. 6.9). The S
W
is close to zero in all three
cases, as it should be if we have in mind that fluctuations of this velocity
component follow the Gaussian distribution.
It is well known that S
V
measurement suffers much more from errors, then U or
W statistics (see Vukoslavevi and Wallace 1996 for example). Klewicky 1989
tabulated this statistics for numerous experiments but did not find any clear trend to
explain rather large differences between them. Data obtained with LDV
measurement by Karlsson and Johansson 1988, being about -0.2 for y
+

= 3-20 and
crossing abscissa at y
+
= 40, agree well to the present results of S
V
measurement
with V-probe. When the velocity gradients are taken into account, the results have
the same trend as two-wire data, being slightly more negative. If only the gradients
are neglected, what corresponds to the VP-4+ probe, the S
V
becomes high positive
as Kastrinakis and Eckelmann 1983 and Klewicky 1989 found it.
Flatness factors of probability density distributions of turbulent velocity
components are shown in the fig. 6.10. Longitudinal F
U
and lateral F
W
factors are
practically the same in all three analysed cases: without neglecting one of the fluid
velocity component or the velocity gradients (probe WP-12+(G)), with neglecting
only velocity gradients (probe VP-4+) and with neglecting simultaneously one
velocity component and all velocity gradients (two-wire probes VP-2v). Whatever
of three tested configurations was used, measured values of flatness factors were
close to 3, what corresponds to the Gaussian probability density distribution.
The analysis of flatness F factors, presented above, is in accordance with the
analogue measurement results of skewness S factors. It follows that neglecting of
fluid velocity gradients or one of velocity components does not significantly affect
measurement accuracy in the case of longitudinal U and lateral W velocity
component. This conclusion agrees well with results of Balint, Wallace and Vuko-
slavevi 1991 and Wei and Willmarth 1989. However, the factor F
V
exhibits
considerable variations near the wall. When neither gradients, nor W component is
neglected the value of F
V
is almost unchanged at different distances from the wall,
having a value around 4. If gradients are neglected, it strongly increases in the near
wall region, as it was found by Wei 1987 and Kim, Moin and Moser 1987, while in
the case of two-sensor probe has quite opposite trend.
Presented results show that mean velocity gradient over the probe sensing area
has strong influence on some of the measurement results. In that case, the sensor
centres are exposed to different values of mean velocity. It means that in a
boundary layer shear flow, where only cU/cy = 0, a V probe should be used for
measurement of U and W. However, an X probe should be applied for researching
the U and V statistics, if the probe has reasonable spatial resolution. In both cases,
the sensors centres are exposed to the same mean U.
MULTIPLE HOT-WIRE PROBES 139
__________________________________________________________________________________________________________________________

2
3
4
5
WW
KMM87
KE83
9T/BVW91 9T/BWV87 9T/VW96
9/VW96
12+/VW96 VV
VH 4+ 12+G
F
U


2
3
4
5
6
F
V


2
3
4
0 10 20 30 40 50
60
70 80
F
W

y
+

Fig. 6.10: Profiles of flatness factors of probability density distributions of
longitudinal (F
U
), transversal (F
V
), and lateral (F
W
) components of
velocity, measured in a turbulent boundary layer by.
VV - probe VP-2v, placed in a vertical plane,
VH - probe VP-2v, placed in a horizontal plane,
+ - four-wire configuration VP-4+ and
12+G - vorticity probe WP-12+(G).
Adapted from: Vukoslavevi and Petrovi 1997.
Remains unclear why the results of S
V
measurement differs significantly for
numerous investigations. It is also surprising that fluctuating velocity gradients,
with rather high RMS values (half the size of the mean U gradient), does not affect
the rest of turbulence statistics parameters up to the fourth-order moments,
although the instantaneous measured values of velocity vector components can be
much in error.
140 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
6.2 RELATIONSHIPS BETWEEN THE HIGHER-ORDER VELOCITY MOMENTS
Analysis of the relations between higher-order moments of turbulent velocity
field represents a sophisticated goal, based on complex experiments. Their
importance is recognised worldwide and explained in Zari 1974, Durst, Jovano-
vi and Kaneve 1987, Durst, Jovanovi and Johanson 1992, Durst and Jovano-
vi 1995, etc. In general, values of higher-order velocity moments, measured by
various researchers, are significantly different. These discrepancies are caused
mostly by the influence of different flow configurations and conditions (see Hus-
sain and Zedan 1978a,b and Petrovi 1991) and by influence of hot-wire probe
configuration, as shown in section 6.1.
However, although different measurements of higher-order moments in the
various flow configurations show significant discrepancies, especially for cross-
stream velocity close to the wall, the existence of their interrelations has been
confirmed. When data originating from different experiments are cross-plotted all
together, they are highly concentrated around the simple curves, as reviewed by
Durst and Jovanovi 1995 for the bounded flows.
Fig. 6.11 shows relations between flatness F and skewness S factors of the
longitudinal U and transversal V velocity. Data measured in the round jet (J) at
Z/D
0
= 3, 4.5, 6, 10 and 20 and in the plain wake (W) at distances Z/D = 200, 300
and 400 downstream a cylinder, follow the parabolic trend lines:
2
62 1 65 2
U U
S . . F + = , (6.1)
2
48 2 13 3
V V
S . . F + = (6.2)
for U and V component respectively. However, wake-data are more dispersed
around lines defined by (6.1) and (6.2), and their interval ranges are wider.

0
5
10
15
20
25
30
35
0 2 4 8 10
W-z/D=200 W-z/D=300
W-z/D=400
F=F(S)
J-z/D=3
J-z/D=4.5
J-z/D=6
J-z/D=10 J-z/D=20
F
U

(
-
)

S
U
2
(-)
F
U
=2.65+1.62*S
U
2
0
5
10
15
20
25
30
35
0 2 4 6 10
W-z/D=200 W-z/D=300
W-z/D=400
F=F(S)
J-z/D=3
J-z/D=4.5
J-z/D=6
J-z/D=10
J-z/D=20
F
V

(
-
)

S
V
2
(-)
F
V
=3.13+2.48*S
V
2

8

Fig. 6.11: Parabolic relationships between the flatness and skewness factors
(F=F(S)): (J) free round jet and (W) plain wake behind a cylinder.
Source: Petrovi, Schenck and Vukoslavevi 1997.
MULTIPLE HOT-WIRE PROBES 141
__________________________________________________________________________________________________________________________

The situation becomes even more clear after analysis of fig. 6.12, which
illustrates the relationships between the skewness S and flatness F factors of
turbulent velocity components in the boundary layer. Data of Petrovi, Schenck
and Vukoslavevi 1997 (VV, VH, 4+ and 12+G) are presented together with
vorticity probes results of Balint, Vukoslavevi and Wallace 1987 (9T/BWV87),
Balint, Wallace and Vukoslavevi 1991 (9T/BVW91) and Vukoslavevi and
Wallace 1987 (9T/9NT/VW96, 12+/VW96). The experimental data of Jovanovi
1984 (J84), and Hedley and Keffer 1974 (HK74), Kreplin 1973/1986 (K73/86) and
Gupta and Kaplan 1972 (GK72) (adopted from Durst and Jovanovi 1995) are
also included in this figure. Comparison of figs. 6.11 and 6.12 shows that
deviations of skewness S and flatness F factors, from corresponding Gaussian
values (0 and 3 respectively), are larger in the free flows than in the bounded.

2
3
4
5
6
7
8
9
0 1 2
3
4
9T/BVW91
9T/BWV87
9T/VW96
9NT/VW96
12+/VW96
VV VH
4+ 12+G
J84
HK74
K86/73
GK72
F=F(S)
F
U

(
-
)

S
U
2
(-)

2
3
4
5
6
7
8
9
0 1 2 3 4
9T/BVW91
9T/BWV87
9T/VW96
9NT/VW96
12+/VW96
VV
VH
4+
12+G
J84
GK72
F=F(S)
F
V

(
-
)

S
V
2
(-)

Fig. 6.12: Parabolic relationships between the flatness and skewness factors
(F=F(S)) in the turbulent boundary layer.
VV/VH - probe VP-2v, placed in a vertical and horizontal plane,
+ - four-wire configuration VP-4+ and
12+G - vorticity probe WP-12+(G).
Source: Petrovi, Schenck and Vukoslavevi 1997.
Relations between skewness S and superskewness SS, and flatness F and
superflatness SF factors of velocity components in the free jet and wake are
presented in figs. 6.13 and 6.14. It is evident that theoretically formulated lines:
S SS =10 , (6.3)
30 15 = F SF (6.4)
formulated on the base of the truncated Gram-Charlier series expansions, quite
correctly represent the interrelations in the inner parts (near the axis of symmetry)
of the free jet and wake. However, in the outer parts of these free flows,
expressions (6.3) and (6.4) represent only general trend-lines of measured data.
142 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________

-100
-75
-50
-25
0
25
100
-3 -2 -1 0 1 3
W-z/D=200
W-z/D=300
W-z/D=400
SS=10S
J-z/D=3
J-z/D=10 J-z/D=20
S
S
U

(
-
)

S
U
(-)
SSU=10S
U

J-z/D=6
J-z/D=4.5

SS
V
=10S
V

-100
-75
-50
-25
0
25
100
-3 -2 -1 0 1 3 S
V
(-)
75
50
S
S
V

(
-
)


Fig. 6.13: Linear relationship SS = 10S between the skewness and superskewness
factors: (J) free round jet and (W) plain wake behind a cylinder. Source:
Petrovi, Schenck and Vukoslavevi 1997.

0
1000
2000
4000
0 5 10 15 20 25
W-z/D=200 W-z/D=300
W-z/D=400 SF=SF(F)
J-z/D=4.5 J-z/D=6
J-z/D=10 J-z/D=20
S
F
U

(
-
)

F
U
(-)
SF=15F-30

0
1000
2000
4000
0 5 10 15 20 25
W-z/D=200 W-z/D=300
W-z/D=400 SF=15F-30
J-z/D=4.5 J-z/D=6
J-z/D=10 J-z/D=20
S
F
V

(
-
)

F
V
(-)
SF=15F-30

Fig. 6.14: Linear relationship SF = 15F30 between the flatness and superflatness
factors: (J) free round jet and (W) plain wake behind a cylinder. Source:
Petrovi, Schenck and Vukoslavevi 1997.
As it can be seen in figs. 6.15 and 6.16, high-precision four-wire probe VP-4+
measurements confirm relations (6.3) and (6.4). In addition, anemometer output
signals were interpreted by three different algorithms: generalised procedure (4+V)
of Vukoslavevi and Wallace 1996 and its modified variants (4+W) and (4+D)
(Petrovi 1996a). Although they gave different values of S, SS, F and SF factors,
corresponding cross-plots in the form of SS=SS(S) and SF=SF(F) follow the same
trend lines defined by (6.3) and (6.4). Therefore, it follows that these interrelations
are nearly independent on the signal interpretation procedure.
MULTIPLE HOT-WIRE PROBES 143
__________________________________________________________________________________________________________________________


-5
0
5
10
10S
4+V
4+W
4+D
S
S
U

(
-
)

S
U
(-)

10
20
30
40
50
15F-30
4+V
4+W
4+D
S
F
U

(
-
)

F
U
(-)


-5
0
5
10
10S
4+V
4+W
4+D
S
S
V

(
-
)

S
V
(-)

10
20
30
40
50
15F-30
4+V
4+W
4+D
S
F
V

(
-
)

F
V
(-)


-5
0
5
10
-0.5 0.0 0.5 1.0
10S
4+V
4+W
4+D
S
S
W

(
-
)

S
W
(-)

10
20
30
40
50
2 3 4 5
15F-30
4+V
4+W
4+D
S
F
W

(
-
)

F
W
(-)

Fig. 6.15: Relationship SS = 10S in a turbulent boundary layer (left).
Fig. 6.16: Relationship SF = 15F-30 in a turbulent boundary layer (right).
Source: Petrovi 1996a.
Presented results show that interrelations of higher-order velocity moments are
practically independent on hot-wires probe configurations and the signal
interpretation procedure. This is in a way surprising, because Vukoslavevi and
Petrovi 1997 proved significant influence of probe configuration on the measured
values of some higher-order velocity moments. Although these shapes slightly
differ in the various flow configurations, they still follow the same trend lines.
144 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
6.3 MINIMISATION OF THE SAMPLING FREQUENCY AND TIME
Turbulence is a typical non-stationary phenomenon, demanding computer-aided
measurements of its parameters. This is the main reason why preliminary
estimation of acquisition time and frequency ordinarily precede experimental
research of turbulent flows. The accurate evaluation of optimal values of these two
sampling parameters has the crucial role. Low sampling frequency excludes high-
frequency sub-harmonics of turbulent velocity fluctuations from the analysis. In
opposite, short acquisition period does not comprehend completely the low-
frequency (long period) sub-harmonics (Hinze 1959). On the other side, increasing
the frequency and acquisition period results in proportionally greater number of
sampled data, which have to be stored and numerically interpreted.
The situation may become critical if the multiple hot-wire probes with nine and
twelve sensors, specified for three-dimensional fluid velocity and vorticity field
measurements, are used. Besides a few times larger number of measuring channels,
than velocity-type probes possessing up to five sensors, the signals of these
vorticity-type probes has to be interpreted by special numeric procedures. Such
algorithms take into account the fluid velocity gradients within the probe sensing
volume and, therefore, involve additional iteration cycles in comparison to the
procedures specified to support only measurements of fluid velocity.
Although the vorticity hot-wire probes achieve higher measurement accuracy in
comparison to configurations specified for velocity measurements, as discussed in
the section 6.1, the highly-increased number of their output signals and computation
efforts needed for their interpretation and management still represent a serious
disadvantage. In these situations, overestimation of acquisition parameters results in
exponentially increased data-processing time and expensive measuring equipment.
Therefore, in practical application of vorticity probes, it is convenient to
decrease the number of samples in measured time-series. Thus, it is desirable to
define the lowest acceptable measurement precision of turbulence parameters and
derive reliable method for estimation of minimum sampling time and frequency of
hot-wire probe anemometer output signals, which enable such level of
measurement accuracy.
Ideally, digital acquisition of the hot-wire anemometer output signals should
temporally resolve the frequency (Wallace and Foss 1995)
q
t
3
U
f ~
, (6.5)
where U is the local mean fluid velocity and q represents the scale of
Kolmogorov. The sampling frequency f
S
has to be at least the twice f
t
in order to
satisfy the well-known criterion of Nyquist.
However, sometimes may be very difficult to reach this requirement, especially
in the flows of higher velocity magnitude. It is easier to involve the hypothesis of
MULTIPLE HOT-WIRE PROBES 145
__________________________________________________________________________________________________________________________

Taylor 1938 and make the sampling length (associated with twice the sample
period, i.e. 2/f
S
)
S
S
f
U
L 2 = , (6.6)
at least approximately equal to the sensing length of the probe in the cross-stream
plane. In this case, two sample periods have to be used because the central-
difference temporal differentiation is needed to avoid phase shifts between the
terms of velocity gradients included in the components of vorticity vectors (Wal-
lace and Foss 1995).
Besides these two approaches, a sampling frequency is sometimes estimated on
the base of the temporal auto-correlation function. However, most researchers did
not check directly the measured values of statistical parameters of turbulence
velocity field. Probably, that is a reason why experimentalists mainly did not
explain details of the applied procedure for the estimation of optimal sampling
parameters (time and frequency).
A possible reliable way for accurate evaluation of sampling parameters is to
apply the method of total searching" of the time-frequency domain. In that case,
the relevant turbulence statistical parameter E is treated as a function of two
independent variables E(t
S
,f
S
): sampling time t
S
and frequency f
S
. Both
independent variables (t
S
,f
S
) should be varied (i.e. increased) simultaneously with
small enough but constant steps, At
S
and Af
S
respectively. Of course,
corresponding values E(t
S
,f
S
) has to be calculated for any existing combination
of (t
S
,f
S
) in the whole domain and compared. When all relevant turbulence
parameters stabilise their values within a desired limit (1% for example), the
current values of sampling time and frequency are accepted as optimal. If the
searching step is chosen to be small enough, presented algorithm guarantees
maximal possible precision of estimating the optimal acquisition parameters.
However, total-search method possesses a serious disadvantage: usually, the nu-
mber of necessary calculations is enormous, especially if the researcher is not
experienced enough. The problem becomes even more critical with increasing the
number of hot-wires.
Petrovi 1996b analysed this problem using the signals of hot-wire probe
WP-12+(G) (fig. 5.20). Output anemometer voltages were sampled at 20 kHz per
measurement channel, at the measurement location distanced 0.9 m from the wall
of a large wind tunnel. Mean flow velocity was set to about 9 m/s. The special
analogue to digital converter with twelve simultaneous sample-and-hold channels
of sixteen-bit resolution (DATA TRANSLATION) was used for this purpose. The
algorithm of Wallace and Vukoslavevi 1996 (described in the chapter no. 7) was
applied for evaluation of the calibration constants and interpretation of the probe
anemometer output voltages.
146 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
The variation of acquisition time and
frequency was performed by simple software
simulation. Frequency is varied by choosing
every N-th data from the original sample
series (N = 1, 2, 3, 4, 5, 6, 8, 16, 32, 64, 128,
256, 512), according to tab. 6.1. In addition,
the number of processed data was used to
control the length of acquired time-series.
Directly speaking, number of data was divided
by current simulated frequency, giving the
sampling time interval. Varying the sampling
time intervals with increment of At = 1 s, gave
the illustrations presented in the figs 6.17,
6.18 and 6.19. Real-time series of 10 s was
long enough for graphic illustration of the new optimised algorithm.

Fig. 6.17: A topography map of % deviations of the longitudinal turbulence
intensities o
U
(f
S
,t
S
) from their values at 20kHz and 10s (o
U
(20kHz,10s)).
Source: Petrovi 1996b.
log
2
(f
S
/f
S0
) [-] f
S
[Hz]
0 39.0625
1 78.1250
. .
. .
. .
8 10 000.0000
9 20 000.0000
Tab. 6.1: Frequency conversion
to logarithm values in
figs. 6.17-6.19, where
f
S0
= 39.0625 Hz.
MULTIPLE HOT-WIRE PROBES 147
__________________________________________________________________________________________________________________________



Fig. 6.18: Dependence of the measured values of the mean longitudinal velocity
( )
S S
f , t U and longitudinal turbulence intensity ( )
S
,
S
f t
U
o on the
sampling parameters. Source: Petrovi 1996b.
148 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________


Fig. 6.19: Dependence of the mean longitudinal vorticity ( )
S S X
f , t e and its RMS
value ( )
S S
f , t
x e
o on the sampling parameters. Source: Petrovi 1996b.
MULTIPLE HOT-WIRE PROBES 149
__________________________________________________________________________________________________________________________

Relevant parameters of turbulence were calculated for every single pair of
independent variables: sampling time t
S
and frequency f
S
. Some parameters, such as
the mean longitudinal velocity and vorticity, together with their RMS values, are
illustrated in figs. 6.18, and 6.19. They are presented in the form of three-
dimensional surfaces that represent functions of two independent variables:
sampling time and frequency. As an additional illustration, a topography map for
the longitudinal turbulence intensity is presented in fig. 6.17. The applied algorithm
is known as total-search optimisation method.
Figs. 6.17, 6.18 and 6.19 show that in this case very short sampling time
interval t
SLIM
~ 2-3 s, designated as limiting, should be enough for estimating the
optimal acquisition frequency. After this time limit, measured statistical parameters
of turbulence stabilise their values over some critical frequency f
SLIM
(see fig.
6.17). Further increasing of sampling frequency (f
S
> f
SLIM
) does not change
significantly the measurement results. This conclusion is very important, because it
shows the way for estimating the sampling parameters.
This approach
should be performed
in two phases. In the
first, the initial
sampling time t
S
is set
to be constant and
equal only a few
seconds (t
SLIM
), while
increasing the
acquisition frequency
f
S
until the all
relevant turbulence
parameters stabilise
their values within
the N% (the vertical
bar in fig. 6.20).
The lowest
allowed frequency
that guarantees the
convergence should
be adopted as
limiting frequency
f
SLIM
. If this process diverge, i.e. if the increasing of frequency f
S
does not result
with convergence of the measured values of analysed parameters, limiting
frequency" f
SLIM
can not be evaluated. In that case, it follows that initially adopted
value of the limiting sampling time t
SLIM
is too low. It has to be increased, until
convergence of all measured turbulence parameters is not achieved (within the
specified relative error of N%).

t
S
(s)
f
S
(Hz)
t
Smax
f
Smax

f
Slim
=f
Sopt
~ const.
t
Slim

A
f
S

A
t
S

Fig. 6.20: A general scheme of reduced searching
method of the time-frequency domain. Source:
Petrovi 1996b.
150 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
In the second step, sampling frequency f
S
is set to be constant and equal to
evaluated value of f
SLIM
. However, sampling time t
S
should be varied (increased)
over t
SLIM
(this process is symbolised by horizontal quadrate in the fig. 6.20) until
the convergence of all relevant parameters is not achieved. If this process becomes
unstable, then a few neighbour frequencies must be controlled simultaneously
(dotted squares in fig. 6.20) to prevent possible errors.
This algorithm is designated as method of reduced searching of the time-
frequency domain. In order to check its applicability, it is comparatively tested
with total-search method. Optimal values of sampling parameters, estimated by
both methods, were the same. However, reduced-search method demanded lower
number of calculations.
At the end of this chapter, it can be concluded that increased measurement
accuracy of turbulence velocity field statistics can be achieved by using the probes
designed primarily for fluid vorticity field measurements. However, they are much
more complicated for operational application and require large number of sampled
data, in comparison to the probes specified for measurement of turbulent velocity
field. This disadvantage of vorticity probes can be minimised by optimising the
sampling parameters.



Karman vortex street. Source: Van Duke 1982.













7 7. . G GE EN NE ER RA AL LI IS SE ED D N NU UM ME ER RI IC C P PR RO OC CE ED DU UR RE E


M Me ec ch ha an ni ic cs s i is s t th he e p pa ar ra ad di is se e o of f t th he e m ma at th he em ma at ti ic ca al l s sc ci ie en nc ce es s b be ec ca au us se e b by y m me ea an ns s o of f i it t o on ne e
c co om me es s t to o t th he e f fr ru ui it ts s o of f m ma at th he em ma at ti ic cs s. .
( (L Le eo on na ar rd do o d da a V Vi in nc ci i, , 1 14 45 52 2- -1 15 51 19 9) )


7.1 EQUATIONS GOVERNING THE HOT-WIRE COOLING
Together with developing the four-sensor and twelve-sensor hot-wire probes, a
unique numeric algorithm to evaluate calibration constants and signal interpretation
has been developed by Vukoslavevi and Wallace since 1992. It was finally
reported by Vukoslavevi and Wallace 1996. The procedure is designated as
generalised, because of the general form of calibration constants used, in three
variants adjusted for different measuring conditions and accuracy. Although
primarily specified for evaluation of calibration constants and signal interpretation
of anemometer probes with four and twelve wires, generalised algorithm can be
also successfully used for triple T and configurations, as well as for the two
sensor probes of V and X configuration.
Generalised numeric procedure originates from the older algorithms. One of
them was specified for numeric support of the vorticity probe with four wires
(Vukoslavevi and Wallace 1981), and the other to support the nine-sensor
configuration WP-9t(G) of Vukoslavevi, Wallace and Balint 1991. Probe with
twelve hot-wires WP-12+(G) and corresponding procedure were developed to
improve the measurement precision of turbulent velocity and vorticity field in
comparison to the older configuration WP-9t(G). Furthermore, the acceptability
range of velocity angles toward the probe axis is enlarged. As usual, the experience
acquired during testing and using the previous models of vorticity hot-wire probes
was used wherever it was possible. Hot-wire response equations
( )
2 2 2 2
W h sin V cos U U
ij eij eij eij
+ = o o , (5.45)
( )
2 2 2 2
V h sin W cos U U
ij eij eij eij
+ = o o , (5.46)
152 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
of Vukoslavevi, Wallace and Balint 1991 had the crucial role in developing the
generalised procedure. Assuming the wires mounted exactly in the vertical xOy
and horizontal xOz planes, they are based on the wire effective cooling angle o
e
.
However, further experimental tests and analysis clearly showed that
expressions (5.45) and (5.46) should be replaced with more suitable descriptions of
hot-wire effective cooling velocity U
e
. The most important reason for such
conclusion was the deviation of probe geometry from the ideal shape. Although
miniature multiple-sensor probes are fabricated carefully, with high precision of
wires positioning, it is practically impossible to mount the sensors quite exactly in
the desired orientation, i.e. in the vertical or horizontal plane. It is verified that
these errors can produce measurement errors of velocity and vorticity components,
if these expressions are applied. In order to improve the accuracy of signal
interpretation, the generalised hot-wire cooling law was applied
W V c W U c V U c W c V c U c U
e
+ + + + + =
6 5 4
2
3
2
2
2
1
2
, (3.67)
which was derived starting from the empirical law of Jorgensen 1971
2 2 2 2
b t n e
U h U k U U + + = , (3.59)
as described in the section 3.3.2. Therefore, generalised hot-wire cooling law
also, but indirectly, relates effective cooling velocity U
e
and normal U
n
, tangential
U
t
and binormal U
b
fluid velocity components in the wire local coordinate system
(see figs. 3.3 and 3.4). Obviously, factors k and h, the tangential and binormal
cooling coefficients respectively, are included in the new calibration coefficients c
i
.
They take into account, among many other unknown effects, the aerodynamic
blockage resulting from the sizes and arrangements of the sensors and prongs.
For relatively small angles of fluid velocity toward the probe axis, smaller than
20
0
approximately, k and h can be assumed to be constants (see, for example, Vagt
1979, Adrian, Johnson, Jones, Merati, Tung 1984, Lekakis, Adrian and Jones
1989 and Marasli, Nguen and Wallace 1993). For higher attack angles of velocity
vector toward the probe axis, corrected values of k and h should be used to achieve
high measurement accuracy. The correction is also necessary if the magnitude of
U
e
changes over a broad range.
As it is discussed in the section 3.3.3, in a general case for optionally oriented
Cartesian coordinate system and arbitrary oriented sensor, relations between
velocity components U, V, W and U
n
, U
t
and U
b
will be in the form (3.66):
W n V n U n U
n 3 2 1
+ + = ,
W b V b U b U
b 3 2 1
+ + = , (3.66)
W t V t U t U
t 3 2 1
+ + = .
MULTIPLE HOT-WIRE PROBES 153
__________________________________________________________________________________________________________________________

Constants n
m
, t
m
and b
m
(m = 1,2,3) are the
transformation coefficients from the probe
coordinates to the local coordinates of
each hot-wire. They depend on the sensor
relative orientation toward the probe axis,
which can be determined by careful
measurement. For illustration of relation
(3.66), the horizontal wire inclined at
angle o toward the normal to the probe
axis is analysed according to fig. 7.1.
Depending on hot-wire orientation, the
normal U
n
, tangential U
t
and binormal U
b

fluid velocity components are:
o o sin W cos U W U U
N N n
+ = + = ,
V U
b
= , (7.1)
o o cos W sin U W U U
T T t
= + = .
These expressions give the following coef-
ficients of coordinates transformation:
n
1
= coso, n
2
= 0, n
3
= sino,
b
1
= 0, b
2
= 1, b
3
=0, (7.2)
t
1
= sino, t
2
= 0, t
3
= -coso..
It means that it is necessary to know exact spatial orientation of each sensor in
order to determine the coefficient n
m
, t
m
and b
m
.
For a miniature multiple hot-wire probe, accurate determination of individual
sensor orientation is possible, but very difficult. Some additional effects, connected
with aerodynamic blockage of the flow by probe elements, cooling of the sensor
not only by the flow but also by cold prongs etc., change the real heat flux from the
expected. These effects should be also taken into account during evaluation of the
calibration coefficients and signals interpretation. Therefore, constants n
m
, t
m
and
b
m
(m = 1,2,3) should be treated as unknowns and determined through the
appropriate calibration process. Having in mind that generalised cooling law
(3.67) is derived by introducing (3.66) in expression (3.59), it is clear that factors
c
m
(m = 0,1,2,...,5) are functions of n
m
, t
m
, b
m
, k and h. Hence, the geometrical and
thermal coefficients are lumped together to be determined by direct calibration.
After dividing (3.67) by c
1
, introducing factors a
m
= c
m
/c
1
and the generalised
form of the effective cooling velocity U
eG
2
= U
e
2
/c
1
, response equation becomes
W V a W U a V U a W a V a U U
eG
+ + + + + =
5 4 3 2 1
2 2 2 2
, (7.3)
z
x
o
U
1
W
U
T
U
N

W
N

W
T

y
V=U
b

Fig. 7.1: A definition sketch of fluid
velocity vector in the local
coordinate system of a
slanted hot-wire.
154 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
for an arbitrary oriented sensor. Effective cooling velocity has traditionally been
related to the anemometer output voltage by the law (3.37) proposed by King 1914.
However, sometimes the polynomial fit (3.38) is used, as it was the case in the
present scheme. It seems that the fourth-degree polynomial provides the highest
accuracy of hot-wire signals interpretation. In that case, expression (7.3) becomes:
W V a W U a V U a W a V a U E b
n
n
n
+ + + + + =

=

5 4 3 2 1
2 2 2
5
1
1
. (7.4)
Formula (7.4) simplifies for hot-wire sensors lying exactly in the vertical xOy
and horizontal xOz plane, to expressions:
V U a W a V a U E b U
V V V
n
n
Vn eGV
+ + + = =

=

3 2 1
2 2 2
5
1
1 2
, (7.5)
W U a W a V a U E b U
H H H
n
n
Hn eGH
+ + + = =

=

4 2 1
2 2 2
5
1
1 2
, (7.6)
because the coefficients a
4V
, a
5V
, a
3H
and a
5H
are identically zero in that case.
In order to formulate an iterative algorithm (7.5) and (7.6) should be rearranged:
2
2
2
2
2
4 2
3
1 2
3
V
a
a W a V
a
U U
V
V V
V
eGV

|
|
.
|

\
|
+ + |
.
|

\
|
+ = , (7.7)
2
2
2
2
2
4 2
4
2 1
4
W
a
a V a W
a
U U
H
H H
H
eGH

|
|
.
|

\
|
+ + |
.
|

\
|
+ = . (7.8)
Comparison of equations (7.7) and (7.8) to the analogue expressions (5.45) and
(5.46) shows that last terms in (7.7) and (7.8) should be zero for hot-wires
positioned exactly in the vertical and horizontal plane, respectively. If they are
mounted nearly in the horizontal or vertical plane, these terms, as well as the
calibration constants a
4V
, a
5V
, a
3H
and a
5H
should be very small. In that case, the
expression for effective cooling velocity, for any wire close to the vertical plane,
could be written in the general form
V eGV
R W a V
a
U U
V
V
+ + |
.
|

\
|
+ =
2
2
2
2
3
2
, (7.9)
W U a V U a V
a
a R
V V V
V
V
+ +
|
|
.
|

\
|
=
5 4
2
2
4
3
1
. (7.10)
Residual (7.10) represents the errors in wire positioning and is very small for
ordinary probe sensor. The analogue formula follows for the hot-wire mounted
fairly close to the horizontal plain:
MULTIPLE HOT-WIRE PROBES 155
__________________________________________________________________________________________________________________________

H H
H
eGH
R V a W
a
U U + + |
.
|

\
|
+ =
2
2
2
1
4
2
, (7.11)
W U a V U a W
a
a R
H H H
H
H
+ +
|
|
.
|

\
|
=
5 3
2
2
4
4
2
. (7.12)
In fact, these expressions will be valid for arbitrary sensor orientation, but R
V
and
R
H
will be small only if the wires are close to vertical or horizontal plane.
Equations (7.9) and (7.11), for the sensors positioned in the vertical and
horizontal plane, respectively, are similar to the nine-sensor probe equations (5.45)
and (5.46). However, they incorporate the residuals (7.10) and (7.12) that has not
been previously used, providing their applicability for any sensor oriented nearly in
the vertical and horizontal plane. This is a great advantage in comparison to the
approach used by Vukoslavevi, Wallace and Balint 1991 and Balint, Wallace
and Vukoslavevi 1991 for the nine-sensor probe WP-9t(G), which is very
sensitive to any misalignment of the sensor from the horizontal or vertical plane.

7.2 DATA REDUCTION FOR TRIPLE AND FOUR-WIRE PROBES
4
2
1
3
VP-4+ VP-4q
4
2
1
VP-3t/

4
2
3
VP-3t/T

Fig. 7.2: Sketches of triple probes VP-3t/T, VP-3t/ and quadruple hot-wire
configurations VP-4+, VP-4q. Adapted from: Petrovi 1996a.
Numeric procedures that support "triple"-wire probes VP-3t/T and VP-3t/, as
well as the configurations VP-4+ and VP-4q with four wires (sketched in fig. 7.2),
assume the uniform fluid velocity vector field over the whole probe sensing
volume. In these cases, the expression (7.4) is applied for all available hot sensors,
giving the following system of equations describing the hot-wire response to fluid
velocity vector:
( )
W V a W U a V U a W a V a U
, , , i E b U
i i i i i
n
n
i i i
n
eG
+ + + + + =
= = =

=

5 4 3
2 1
2 2 2
5
1
1
4 3 2 1
. (7.13)
156 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Generalised procedure assumes probe calibration to be performed in two
phases: velocity and angular (directional). In the first step, probe is subjected
to a uniform flow of constant direction and variable but known magnitude. The
second calibration phase provides information on the wire directional response.
The magnitude of velocity
is maintained constant, while
its direction is varied over the
angular range expected in
turbulent flow. This can be
achieved by pitching and
yawing the probe for angles
and u in the nominally
irrotational flow with known
uniform speed, which may be
generated by a suitably
profiled nozzle. Also, a free-
stream part of a tunnel flow
can be used for this purpose.
Depending on the type of applied calibration mechanism for directional calibration,
components U, V and W are related to the fluid velocity magnitude U
0
by sine and
cosine of pitch and jaw u angles. A sketch of a probe, positioned in front of a
round nozzle, is shown in fig. 7.3, while the directional calibration data reduction
map is presented in fig. 7.4.
As usual, calibration coefficients are evaluated by multiple linear regression
method. It results in the system of ten linear algebraic equations for every sensor,
which enable evaluation of ten unknown calibration constants in (7.13). This
system can be easily solved by standard Gauss elimination method and, in
general, pivoting procedure does not
increase the computation accuracy.
Practically, the evaluation of the
calibration coefficients is quite the same
for the probes with three and four wires,
with the only difference in the number of
sensors. However, interpretation of the
corresponding signals measured in the real
turbulent flow is qualitatively different.
The equivalent equations are used, but the
algorithms differ because the four-sensor
probe has additional fourth hot-wire. This
enables the choice of three the most
suitable signals that provide the best
possible directional sensitivity of the probe
toward the instant direction of the fluid
U
V
W
z
x

y
u
U
0

Fig. 7.3: A sketch defining hot-wire probe
calibration in the jet potential core.

(

,

V
>
0
)

(u, W>0)
Pitch angle

0
Jaw angle
u

0
20 10 0 -10 -20
-10
-20
10
20

Fig. 7.4: Angular calibration map for
quadruple probes.
MULTIPLE HOT-WIRE PROBES 157
__________________________________________________________________________________________________________________________

velocity vector fluctuations, what results in the increased measurement accuracy of
turbulent velocity field.
Signal interpretation starts from the formulas (7.9 7.12), rewritten with
included indices (i = 1,2,3,4) in order to make distinction between various sensors:
i i
i
n
n
i n
eG
i
R W a V
a
U E b U + +
|
|
.
|

\
|
+ = =

=
2
2
5
1
1 2
2
3
2
, (7.14)

W U a V U a V
a
a R
i i
i
i i
+ +
|
|
.
|

\
|
=
5 4
2
2
4
3
1
(i = 1,3)
, (7.15)

i i
i
n
n
i n
eG
i
R V a W
a
U E b U + +
|
|
.
|

\
|
+ = =

=
2
2
5
1
1 2
1
4
2
, (7.16)

W U a V U a W
a
a R
i i
i
i i
+ +
|
|
.
|

\
|
=
5 3
2
2
4
4
2
(i = 2,4)
. (7.17)

Expression (7.16) and (7.17) are used for numeric support of both triple and
four-wire probes, because all of them posses two horizontal sensors (see fig. 7.2).
However, T and configurations posses only one hot-wire in the vertical plane:
no. 3 or no. 1, respectively. In that sense, for the last two probes, only one set of
expression (7.14) and (7.15) can be used. In opposite, plus (+) and quadrate
(q) probes are equipped with four hot wires that provides possibility of choosing
the three (of four available hot-wire anemometer output voltages) that provide the
highest possible measurement accuracy of fluid velocity vector components, and
their further processing through expressions (7.14) - (7.17) (for i = 1,2,3,4).
The first operational step in the procedure of interpretation of hot-wire signals
sampled in the real turbulent flow is the calculation of generalised cooling
velocities U
ieG
2
. It is performed for all hot-wires, available on the specific probe
configuration, using the fourth-order polynomial regression

=

=
5
1
1 2
n
n
i n
eG
i
E b U (i = 1,2,3,4). (7.18)
In order to express U and W variables (values of longitudinal and lateral
velocity components), expression (7.16) are rearranged for i = 2 and i = 4, giving
the system of two simultaneous linear equations:
2
4
2
2
2
1
2
2
2
W a
U R V a U
eG

+ =
, (7.19)
2
4
4
4
2
1
4
2
4
W a
U R V a U
eG

+ =
. (7.20)

158 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Introduction of the new combined terms, so-called roots",
2
2
1
2
2
2 2
R V a U Q
eG
=
, (7.21)

4
2
1
4
2
4 4
R V a U Q
eG
=
, (7.22)

in (7.19) and (7.20), gives the system of two linear algebraic equations:
2
4
2
2
Q W
a
U = +
, (7.23)

4
4
4
2
Q W
a
U = +
. (7.24)

Its solutions, U and W, can be expressed as functions of transversal velocity V and
unknown residuals R
2
= R
2
(U,V,W) and R
4
= R
4
(U,V,W):
4
2
4
4
4
4
2 2
4
4
a a
Q a Q a
U


= ,
(7.25)

4
4
4
2
4 2
2
a a
Q Q
W

= .
(7.26)

Small absolute values of residuals guarantee the algorithm convergence. In the
first iteration cycle, they are set to zero, while in the further iterations, residuals
values are up-dated according to expression (7.15) and (7.17).
Equations (7.14) and (7.15), which describe responses of hot-wire sensors no. 1
and 3, can be used to find transversal V velocity component. Substitution of (7.25)
and (7.26) in expression (7.14) gives the following characteristic equation for the
first hot-wire (i = 1):
0
4
4
2
2
1 1
2
4
2 4
4 2
2
1
2
3
1
4
2
4
4
4
4
2 2
4
4
1
= +
(
(

|
|
.
|

\
|

+
+
(
(

+
|
|
.
|

\
|


=
eG
U R
a a
Q Q
a
V
a
a a
Q a Q a
) V ( F
(i = 1) (7.27)
and substituting in (7.16), characteristic equation for the third (i = 3) wire:
0 4
2
2
3 3
2
4
2
4
4
4 2
2
3
2
3
3
4
2
4
4
4
4
2 2
4
4
3
= +
(
(

|
|
.
|

\
|

+
+
(
(

+
|
|
.
|

\
|


=
eG
U R
a a
Q Q
a
V
a
a a
Q a Q a
) V ( F
. (i = 3) (7.28)
MULTIPLE HOT-WIRE PROBES 159
__________________________________________________________________________________________________________________________


Pitch angle

Jaw u angle
20 10 0 -10 -20
-10
-20
10
20
(a)
=15
0
u= 0
0

,
V

>

0

(u, W>0)


U
V
W=0
U
0
u
=0
0

1
`3
2
4
=15
0

(b)
F
1
(V)
F
3
(V)
4 2
3
(c)

=15
0

15
0



-2
F
1,3
(V)
V [m/s]
F
1
(V)
F
3
(V)
F
1
(0)>0
F
3
(0)<0
-50
-100
50
100
-2 -4 2 4
F
1,3
(V)
V [m/s]
F
1
(0) < 0
F
3
(0) > 0
F
1
(V)
F
3
(V)
-4 0 2 4
50
-100
-150
(e)
(d)
V/U = 0.268
W/U = 0.000
V/U = -0.268
W/U = 0.000
U
0
=10 m/s
BETTER IS SIGNAL
OF THE WIRE NO. 3
BETTER IS SIGNAL
OF THE WIRE NO. 1
0

Fig. 7.5: Testing the procedure of choosing the optimal hot-wire signals of the
probe with four hot-wires. Source: Petrovi 1996a.
(a) probe position in the calibration map,
(b) + probe pitched at = 15
0
,
(c) a sketch of alternative sensors no. 1 and no. 3,
(d) characteristic (criteria) functions F
1,3
(V) at u = 0
0
, = 15
0
and
(e) characteristic (criteria) functions F
1,3
(V) at u = 0
0
, = -15
0
.
160 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Flow-charts of the characteristic functions F
1
(V) and F
3
(V), which correspond to
vertical sensors no. 1 and 3 (fig. 7.5c), are sketched in figs. 7.5d,e. For positive
values of V (fig. 7.5d), corresponding to clear "pitch" angle = 15
0
(this probe
orientation is illustrated in figs. 7.5a,b), it is better to use function F
1
(V). The
reason is simple. The slope of this curve toward abscissa is greater at the point of
their intersection, providing the higher precision of evaluating V. Any error in
measuring the anemometer output-voltage value E
i
or the dependence of calibration
constants on velocity variation that is not taken into account moves the F
1,3
(V)
curves up or down. Consequently, the intersection points of these curves with
abscissa are changed, resulting in the error in determining V and, therefore, U and
W. The errors of velocity components (U, V, W) increase with decreasing the slope
of the curve. For negative V, equation F
3
(V) = 0 should be used instead of F
1
(V) =
0 for the same reason (fig. 7.5e). The highest error can occur in the range close to
the (V/U)
CR
, where one of the curves is almost flat. It can even fall under abscissa
giving no solution, so the choice of better curve is crucial in that range.
Solutions of the equations (7.27) and (7.28) should be identical only in the
irrotational flow, what results in zero-velocity-gradients over the probe sensing
volume and calibration coefficients independent on the velocity magnitude and
direction. Such situation is typical only for calibration, usually performed in the
uniform flow with relatively small attack angle toward hot-wire probe axis.
However, in the turbulent flow, solutions of F
1,3
(V) = 0 are ordinarily different,
depending on the instant velocity gradients and directions of velocity fluctuations.
The advantage of four-wire plus (+) and quadrate (q) probes, comparing to
"triple"-wire T and configurations, becomes clear now. The four-sensor probe
enables choice between signals of the wire no. 1 or 3, according to direction of the
instant fluid velocity vector. This approach results in the higher precision of
anemometer output signals interpretation, e.g. higher measurement accuracy of
fluid velocity vector components. Furthermore, an additional advantage of four-
wire probe is evident from the figs. 7.5d,e. Equations (7.27) and (7.28) have not the
unique solution in the whole domain of argument V. In the region of V > 0, the
equation F
3
(V) = 0 has two possible solutions (see fig. 7.5d). In opposite, in the
region of V < 0 this is the case with F
1
(V) = 0 (see fig. 7.5e).
From analysis presented above, the problem of recognising the better output
signal (of sensor no. 1 or wire no. 3 see figs. 7.2 and 7.5c) arises. The question is
how to select this signal before the instant value of V component is known, what is
a case in the real turbulent velocity field measurements. This can be achieved by
using the values of characteristic functions F
1 1
(V) and F
3 3
(V), for V = 0, i.e. F
1 1
(0)
and F
3 3
(0). Characteristic function, which value is positive for V = 0, (F
1 1
(0) > 0 or
F
3 3
(0) > 0) corresponds to a wire-signal that can provide more accurate solution for
measured fluid velocity vector (see figs. 7.5d,e). Following this important property
of F
1 1
(0) and F
3 3
(0) that enables optimal hot-wire signals selection, the functions
F
1
(V) and F
3
(V) were denoted as criteria functions.
MULTIPLE HOT-WIRE PROBES 161
__________________________________________________________________________________________________________________________

Generalised procedure is iterative by its nature. Iteration cycles should be
repeated until the difference of U, V and W values from two successive iterations is
smaller then an accepted relative measurement error c (of 1%, for example). In
some occasions, usually caused by the high attack angles of the fluid velocity
vector toward the probe axis, or by high fluid velocity gradients over the probe
sensing volume, the algorithm does not converge. This may also happen if the
measured velocity vectors are out of the calibration range. In order to prevent the
indefinitely long calculation, the control counter is introduced and limited to 10.
However, our experience shows that usual number of iterations is smaller than 4.

7.3 THE ALGORITHM FOR TWELVE-SENSOR PROBES WP-12+/q(G)
Data reduction for
the vorticity probe
WP-12+(G) follows a
similar algorithm to
that of the plus
probe VP-4+,
presented in the
chapter 7.2, except for
the gradients across
the measuring volume
which are taken into
account. However,
calibration procedures
for both probes are
equivalent, with the
only distinction in the
number of sensors
(see fig. 7.6). Both
algorithms use quite
the same formulation
of hot-wire cooling law, but with slightly different organisation of indices. For
twelve-wire probe, two indices are used to define the sensor position and velocity
components that are basically different for each sensor in a turbulent flow. In that
sense, applying expression 7.4 for the wire no. "i" (i = 1,2,3,4) of the j
-th
array (j
= 1,2,3), gives the following system of hot-wire response equations:
j , i j , i j , i j , i j , i j , i j , i j , i j , i
j , i j , i j , i j , i j , i
n
n
j , i
n
j , i
eG
j , i
W V a W U a V U a
W a V a U E b U
+ + +
+ + + = =

=

5 4 3
2
2 2
1
2
5
1
1 2

(i = 1,2,3,4; j = 1,2,3). (7.29)
1 2
3
4
2
3
4
1
4
ARRAY 1
ARRAY 3
ARRAY 2
1
2 C
0

1
.
1
4

0.09 0.44
1.32
y
z
3
W
V
U
(a)
(b)
3
4 2
1

Fig. 7.6: A sketch of the vorticity-type hot-wire probe
WP-12+(G) (all dimensions are in mm):
(a) the end-view orthogonal projection of the
whole configuration and
(b) a quadruple array, configured identically
as VP-4+ hot-wire probe.
Adapted from: Vukoslavevi and Wallace 1996.
162 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Similarly, like for the four-wire probes, these expressions can be written in the
following form for the sensors mounted nearly in the vertical planes:
( )
j , i j , i j , i j , i j , i j , i
j , i j , i j , i j , i
j , i
j , i
eG
j , i
R W k V k U
R W a V
a
U U
+ + + =
= + +
|
|
.
|

\
|
+ =
2 2
1
2
2
2
2
2
3
2
(i=1,2,3,4; j=1,2,3), (7.30)
where
j , i j , i j , i j , i j , i j , i j , i j , i
j , i j , i j , i j , i j , i j , i j , i
j , i
j , i j , i
W U k V U k V k
W U a V U a V
a
a R
+ + =
= + +
|
|
|
.
|

\
|
=
5 4 3
5 4 1
2
2
2
3
4
. (7.31)
For horizontal or nearly horizontal sensors, the following expressions arise:
( )
j , i j , i j , i ij j , i j , i
j , i j , i j , i j , i
j , i
j , i
eG
j , i
R V k W k U
R V a W
a
U U
+ + + =
= + +
|
|
.
|

\
|
+ =
2 2
2
2
2
2 1
1
4
2
(i=1,2,3,4; j=1,2,3), (7.32)
where
j , i j , i j , i j , i j , i j , i j , i j , i
j , i j , i j , i j , i j , i j , i j , i
j , i
j , i j , i
V U k W U k W k
W U a V U a W
a
a R
+ + =
= + +
|
|
|
.
|

\
|
=
5 4 3
5 3
4
2
2
2
2
4
. (7.33)
These expressions are also valid for any arbitrary sensor orientation, but the
residuals R
i,j
will be small enough only if the sensors are mounted nearly in
horizontal or vertical planes.
The velocity components U
i,j
, V
i,j
and W
i,j
, acting on each sensor of the twelve-
wire probe, can be defined in terms of the velocity components U
0
, V
0
, W
0
at the
probe centre C
0
(see fig. 7.6), their gradients in the plane normal to the probe axis
passing through the sensor centres and the distances of each wire centre from the
centre of the measuring volume of the whole probe (e.g. twelve hot-wires).
Following Vukoslavevi and Wallace 1996, the velocity components U
0
, V
0
, W
0

are expanded to the first order in a Taylor series around C
0
, giving the following set
of twelve non-linear equations, with 3 velocity components and 6 cross stream
velocity-gradient component unknowns:
MULTIPLE HOT-WIRE PROBES 163
__________________________________________________________________________________________________________________________


P
j , i j , i j , i j , i
j , i j , i j , i j , i j , i
eG
j , i
R
y
W
y
z
W
z W k
y
V
y
z
V
z V k
y
U
y
z
U
z U U
P P
P P P P
+
|
|
.
|

\
|
+ + +
+
(

|
|
.
|

\
|
+ + + + + =
2
0
2
2
0
1
0
2
c
c
c
c
c
c
c
c
c
c
c
c

(i = 1,2,3,4; j = 1,2,3), (7.34)
where
|
|
.
|

\
|
+ +
|
|
.
|

\
|
+ + +
+
|
|
.
|

\
|
+ +
|
|
.
|

\
|
+ + +
+
|
|
.
|

\
|
+ + =
y
W
y
z
W
z W
y
V
y
z
V
z V k
y
W
y
z
W
z W
y
U
y
z
U
z U k
y
V
y
z
V
z V k R
P P P P
P P P P
P P
j , i j , i j , i j , i j , i
j , i j , i j , i j , i j , i
j , i j , i j , i
P
j , i
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
0 0
5
0 0
4
2
0
3

(i = 1,2,3,4; j = 1,2,3). (7.35)
are residuals, defined as function of velocity components at the probe centre and
velocity increments from the probe centre to the centre of each sensor.
The constants z
i,j
P
and y
i,j
P
are the z or y coordinates of the i,j-th sensor centre
with respect to the probe centre C
0
. Analogue to equations (7.30-7.31) and (7.32-
7.33), V should be interchanged by W for horizontal hot-wires (j = 2 and j = 4).
Introducing the increments of fluid velocity vectors, in the form:
y
U
y
z
U
z U
P
j , i
P
j , i
P
j , i
c
c
+
c
c
= A ,
y
V
y
z
V
z V
P
j , i
P
j , i
P
j , i
c
c
+
c
c
= A , (i = 1,2,3,4; j = 1,2,3) (7.36)
y
W
y
z
W
z W
P
j , i
P
j , i
P
j , i
c
c
+
c
c
= A ,
the expressions (7.34) and (7.35) become
( ) | | ( )
P
j , i
P
j , i j , i
P
j , i j , i
P
j , i
eG
j , i
R W W k V V k U U U + + + + + + =
2 2
0
2
0
1
0
2
A A A
(i = 1,2,3,4; j = 1,2,3), (7.37)
( ) ( ) ( )
( ) ( )
P
j , i
P
j , i j , i
P
j , i
P
j , i j , i
P
j , i j , i
P
j , i
W W V V k
W W U U k V V k R
A A
A A A
+ + +
+ + + + + =
0 0
5
0 0
4
0
3
. (7.38)
164 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Effective cooling velocities can be also written as functions of the velocity
components U
0j
, V
0j
, W
0j
at the centre C
0j
of j-th array and the corresponding
velocity increments from the centre of that array to the centre of the i-th wire:
( ) | | ( )
S
j , i
S
j , i j j , i
S
j , i j j , i
S
j , i j
eG
j , i
R W W k V V k U U U + + + + + + =
2 2
0
2
0
1
0
2
A A A
(i = 1,2,3,4; j = 1,2,3), (7.39)
where
( ) ( ) ( )
( ) ( )
S
j , i j
S
j , i j j , i
S
j , i j
S
j , i j j , i
S
j , i j j , i
S
j , i
W W V V k
W W U U k V V k R
A A
A A A
+ + +
+ + + + + =
0 0
5
0 0
4
2
0
3

(7.40)
and
y
U
y
z
U
z U
S
j , i
S
j , i
S
j , i
c
c
+
c
c
= A ,
y
V
y
z
V
z V
S
j , i
S
j , i
S
j , i
c
c
+
c
c
= A , (i = 1,2,3,4; j = 1,2,3), (7.41)
y
W
y
z
W
z W
S
j , i
S
j , i
S
j , i
c
c
+
c
c
= A .
The constants z
i,jS
and y
i,jS
are the z and y coordinates of the i,j-th sensor centre
with respect to the centre of j-th array. Therefore, z
i,jS
< z
i,jP
, y
i,jS
< y
i,jP
and
P S
j , i j , i
U U A A << ,
P S
j , i j , i
V V A A << , (i = 1,2,3,4; j = 1,2,3) (7.42)
P S
j , i j , i
W W A A << .
Using this formulation, a successive-iteration method, which clearly illustrates
the character of the equations and limitations of the probe, can be applied to solve
equation (7.39). In the first iteration the velocity increments from the array centres
to the sensor centres, AU
i,jS
, AV
i,jS
, AW
i,jS
, as well as R
i,jS
can be neglected (taken
equal to zero), giving three sets of four equations with three unknown velocity
components at each array centre:
( )
j j , j j , j
eG
j ,
W k V k U U
0
2
1 0
1
1 0
2
1
+ + = ,
( )
j j , j j , j
eG
j ,
V k W k U U
0
2
2 0
1
2 0
2
2
+ + = ,
( )
j j , j j , j
eG
j ,
W k V k U U
0
2
3 0
1
3 0
2
3
+ + = ,
( )
j j , j j , j
eG
j ,
V k W k U U
0
2
4 0
1
4 0
2
4
+ + = . (j = 1,2,3) (7.43)
MULTIPLE HOT-WIRE PROBES 165
__________________________________________________________________________________________________________________________

These equations can be solved for the three velocity components at the centre of
each array, in the first iteration. This, initial, phase of procedure is practically
equivalent to the algorithm for the four-sensor probe. It is obvious that four
equations are available to find three unknowns at each array. Combining the
second and fourth equation with the first or third respectively, and having in mind
that U
i,jeG
is related to E
i,j
through polynomial fit, two functions of the V
0j
velocity
component can be obtained:
( ) 0
0 1
=
j , i j j
E , V F , ( ) 0
0 3
=
j , i j j
E , V F . (7.44/45)
In these formulas, subscripts 1 or 3 indicate the corresponding sensor in the vertical
plane (see figs. 7.2 and 7.5c). The choice of the best three out of four signals is
analogue to the procedure for four-sensor probes, explained in the section 7.2.
The first iterative approximation of the velocity gradients can be found as the
velocity differences between arrays divided by the corresponding distances
between the array centres. So, the small quantities AU
i,jS
, AV
i,jS
, AW
i,jS
(expressions
(7.41)), as well as R
i,jS
(from (7.40)), can be determined. The velocity components
in the second and next iterations can be obtained by solving the set of equations
(7.39) for i = 1,2,3,4 and j = 1,2,3 with known velocity increments and residuals
from the previous iteration. The velocity gradients in the second and other
iterations can be determined the same way as for the first iteration, dividing the
velocity difference between arrays by corresponding arrays separations. The
calculation proceeds until the differences of two successive gradients or velocity
iterations become negligible. The method has been found to converge very quickly,
requiring only a few iterations.

7.4 THE PROCEDURE FOR TWO-SENSOR PROBES
This variant of generalised numeric procedure was developed starting from
the procedure for quadruple hot-wire probes, VP-4+ and VP-4q. It is known that
two-sensor X and V probes are incompatible with three-dimensional formulations
of hot-wire cooling law, because of the third sensor missing. Therefore, two
available output voltages demands neglecting of one velocity component, normal
to the wires, within the calibration algorithm and signal interpretation procedure.
The consequence is well known simplification of the general three-dimensional
hot-wire cooling law, resulting with the corresponding two-dimensional variant. In
the present case, the expression (7.3) is transformed, giving the new approximate
expression, denoted here as two-dimensional generalised hot-wire cooling law.
For the wire placed exactly in the vertical plane, the following formula is valid:
V U a V a U U
eGV
+ + =
3 1
2 2 2
, (7.46)
while for horizontal sensor, the analogue expression can be used:
W U a W a U U
eGH
+ + =
4 2
2 2 2
. (7.47)
166 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
The numeric procedure is based on the expres-
sions (7.46-7.47) that govern the cooling
mechanism of a sensor exposed to (idealised) two-
dimensional fluid velocity field. Having in mind
that signal interpretation of X and V probes (fig.
7.7) does not depend on their vertical or horizontal
orientation, further analysis assumes wires lying in
the vertical plane(s). In that case, generalised
effective cooling velocity

=

= + + =
5
1
1 2 2 2
3 1
n
n
i n
i i
eG
E b V U a V a U U
(i = 1,2) (7.48)
does not take into account the velocity component normal to the sensors plane(s).
Gradients of fluid velocity vector components, over hot-wire probe sensing
volume, are also neglected.
The effective cooling velocity U
eG
can be represented as sum of two terms:
dominant and residual R, which can be neglected in the first iteration cycle.
For two hot-wires (i = 1,2) the following two expressions arise:
i
i i
i
i
EG
i
R V
a
U V
a
a V
a
U U +
|
|
.
|

\
|
+ =
|
|
.
|

\
|
+
|
|
.
|

\
|
+ =
2
2
2
2
2
2 4 2
3 3
1
3
. (7.49)
After simple transformations, for i = 1,2, expression (7.49) gives:
1
2
1
1
2
3
R U V
a
U
EG
= + , (7.50)
2
2
2
2
2
3
R U V
a
U
EG
= + , (7.51)
where residuals R
1
and R
2
are defined as
2
2
1
1 1
4
3
1
V
a
a R
|
|
.
|

\
|
= , (7.52)
2
2
2
2 2
4
3
1
V
a
a R
|
|
.
|

\
|
= . (7.53)
Following the four-wire algorithm, after introducing the new formal variables,
designated as "roots"
1
2
1 1
R U Q
EG
= , (7.54)
VP-2v
1
2
1
2
VP-2x

Fig. 7.7: Two-sensor probes.
MULTIPLE HOT-WIRE PROBES 167
__________________________________________________________________________________________________________________________

2
2
2 2
R U Q
EG
= (7.55)
the system of linear equations (7.50-7.51) gives the following solutions:
3
2
3
1
1
3
2 2
3
1
a a
Q a Q a
U


= , (7.56)
( )
3
2
3
1
2 1
2
a a
Q Q
V


= (7.57)
Expressions (7.56) and (7.57) represent the formal base of the two-dimensional
generalised algorithm for signal interpretation of X and V hot-wires probes.
In accordance to probe geometry, calibration assumes two-dimensional fluid
velocity field. Thus, all the calibration velocity vectors has to be induced in the
same plane, in order to provide the information about the two-dimensional
response of the applied hot-wire probe. The third velocity component has to be
maintained equal to zero, because of missing the third sensor for its measurement.
The calibration coefficients can be evaluated easily by non-linear regression
method, following the expression (7.48). The algorithm is very simple and fast
converging, demanding solving of the system of seven linear algebraic equations
for both sensors, which can be done easily by Gauss elimination method.
Numeric interpretation of hot-wire anemometer output voltages, measured in
the turbulent flow, is slightly more complex, but still very simple. It starts from the
generalised effective cooling velocities U
eG1
and U
eG2
, evaluated by substituting the
X or V probe output signals E
1
and E
2
in the right-hand side of expression (7.48).
In the first iteration cycle, the residuals R
1
and R
2
are set to zero. Longitudinal
U and transversal V velocity component are calculated from the expression (7.56)
and (7.57). In further iterations, the residuals are updated by using the expression
(7.52) and (7.53) with "roots" up-dated through the equations (7.54) and (7.55).
The iterative procedure is finished under standard condition: when relative
differences of the values U and V, resulting from two successive iterations, are
simultaneously smaller from the desired limit (1% for example):
% %
U
U U
i
i i
Ui
1 100
1
<

=

c , (7.58)
% %
U
V V
i
i i
Vi
1 100
1
<

=

c . (7.59)
The two-dimensional generalised procedure is characterised with the
limitation ordinary present in all situations involving anemometer probes with two
hot-wires: it achieves high measurement accuracy of the turbulent velocity vector
only if this vector lies in the plane of probe sensors or is very close to it. In
opposite, as shown by many authors, serious errors may arise. Therefore, two-
168 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
dimensional measurement of turbulent velocity field is reliable only in the flow of
low turbulence levels (up to approximately 10%).

7.5 POSSIBLE ADVANCES OF THE GENERALISED NUMERIC PROCEDURE
7.5.1 General principles
Besides hot-wire probe configuration, the algorithm for evaluation of the
calibration constants and interpretation of hot-wire output voltages are directly
related to the measurement accuracy of turbulence parameters. Although these
procedures strongly depend on the probe design, still a wide area for their
improvement exists. An appropriate procedure for numeric support of hot-wire
anemometer probe should be balanced between the following goals:
- the highest measurement accuracy that can be achieved by applied probe,
- smallest possible percentage of rejected hot-wire signals, i.e. non-resolved
anemometer output voltages, caused by divergence of the numeric procedure,
- fast converging (procedures are mostly iterative by their nature),
- short, simple and easy understandable structure of the algorithm,
- low consumption of computer resources: processing time, memory (RAM),
external storage space (hard disk, compact disk, magnetic tape), etc.
It is well known that measurement of turbulent velocity field becomes more
difficult with decreasing the magnitude of measured fluid velocity vectors.
Furthermore, in the flow regions with low velocities (the wall vicinity, for
example) high velocity gradients are ordinary present, which additionally increase
the existing problems. Current efforts for resolving this situation are concentrated
in further improvement not only the probes but also the existing procedures for
their numeric support. Measurements of low-intensity fluctuating velocities by
multiple hot-wire probes are very complex. At fluid velocities below 5m/s
approximately, and especially below 2m/s, simultaneous thermodynamic and aero-
dynamic effects can produce serious problems. They initiate deviation of the real
hot-wires cooling law from its theoretical form. Among many other phenomena, it
seems that the strongest influence has:
- decreasing of forced convection domination in the overall sensor cooling flux,
- increasing the free convection cooling flux of hot-wire,
- intensified downstream growth of wakes, behind hot-wires and prongs and
- complex probes calibration at low fluid velocities (slow fluid motion results
in extremely small dynamic pressures producing additional difficulties in
their measurements and controlling the air speed in the wind tunnel).
From the numerical point of view, the additional problems are caused by strong
dependence of hot-wire cooling law on the magnitude of flow velocity at slow fluid
MULTIPLE HOT-WIRE PROBES 169
__________________________________________________________________________________________________________________________

motion. In order to evaluate dependence of calibration constants of the adopted
hot-wires cooling law on the flow velocity, it is necessary to provide information
about hot-wire probe directional sensitivity at various intensities of fluid velocity.
To achieve this goal, hot-wire probe directional calibration has to be repeated at
different magnitudes of calibration velocity.
Unfortunately, the problem has not been completely resolved yet. Turbulent
velocity field is non-stationary. The instantaneous magnitude of fluid velocity
vector cannot be estimated accurately before finishing the interpretation of
corresponding hot-wires signals. Therefore, instantaneous values of fluctuating
velocity components must be preliminary evaluated by interpreting available hot-
wire signals using the appropriate calibration constants that correspond to current
fluid velocity vector. However, the choice of these calibration coefficients depends
on the instantaneous magnitude of velocity vector for which evaluation the
calibration constants are used. Thus, it is necessary to apply special numeric
procedures for interpretation of hot-wire output voltages in the slow turbulent
flows. However, although very complex, experiments in the slow turbulent flows
are very important from a few reasons.
The first is strongly connected with spatial resolution of measuring instruments.
It is well known that turbulent flows contain so-called quasi-deterministic vortex
structures, imposed to the main flow. Slow turbulent flows generate larger
coherent vortex structures in comparison to faster flows, providing better spatial
resolution of the applied measuring instruments.
The second reason lays in their importance for future development of turbulence
theory. Fluid velocity fluctuates up to zero value at the solid borders and in the
outer zones of some boundary-free flows (free jets and wakes, etc.) where
turbulence is generating. In addition, recently developed refined experimental and
numeric methods, which enabled reliable measurements and reliable flow
predictions close to walls (see Antonia and Kim 1994, Smith 1994, Fernholz,
Krause, Nockemann and Schober 1995, Zagarola and Smits 1998, Fischer,
Durst and Jovanovi 1998, etc.) revealed interesting results. They showed a low-
Reynolds-number dependence of the statistical parameters of the near-wall
turbulence (time-mean velocity, turbulence intensity, RMS values of longitudinal
streamwise vorticity and pressure fluctuations, as well as the Reynolds shear stress
and average turbulence dissipation rate, etc.).
This phenomenon is illustrated in fig. 7.8 (Schenck 1997), which presents
turbulence intensity measured close to the walls of the plane-channel flows for
various (low) values of Reynolds numbers (see, for example, Kuroda, Kasagi. and
Hirata 1989, Kuroda, Kasagi and Hirata 1993, Horiuchi 1992, Gilbert and
Kleiser 1991 and Antonia, Teitel, Kim and Browne 1992). Therefore, appropriate
hot-wire methods for measurements of turbulent velocity and vorticity fields in the
slow turbulent flows are still of great scientific interest.
170 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
7.5.2 Dual directional generalised numeric procedure for quadruple probes
Petrovi 1996a and Petrovi and Vukoslavevi 1997 formulated a unique self-
optimising procedure for numeric interpretation of anemometer signals of
miniature four-sensor probes VP-4+ and VP-4q. Although primarily specified for
the application in the slow turbulent flows (below 5m/s), this algorithm also
achieves high precision and efficiency of hot-wire output signals interpretation at
higher velocities. New procedure possesses two crucial advantages in comparison
to the previous (basic) variant of Vukoslavevi 1993.
1. Calibration coefficients, used for hot-wire anemometer signals
interpretation, are additionally optimised according to instant magnitude
of the fluctuating fluid velocity vector.
2. The choice of output hot-wire anemometer signals, according to instant
direction of fluid velocity vector, is fully optimised. Instead of selecting
only the output voltages of vertical sensors (no. 1 or no. 3 see fig. 7.2),
new algorithm uses any combination of the best three of four signals (no.
1, 2, 3 or no. 4), whether they belong to vertical or horizontal hot-wires.
In order to expose these two properties of the new algorithm, it is designated
here as dual directional generalised procedure (DD) while the basic variant is
designated as single basic generalised procedure (SB).
Having in mind that the basic procedure is explained in the section 7.2, here are
presented only the principles, necessary for explanation of the advanced features of
DD algorithm. The signal selection within the basic procedure is illustrated in the
figs. 7.5d,e. The higher probe directional sensitivity and wider uniqueness range
correspond to signal (sensor), which gave positive value of characteristic criteria
functions F
1
(V=0) or F
3
(V=0) on the ordinate axis.
It means that
this procedure was
not fully optimised.
The horizontal hot-
wires no. 2 and 4
(see fig. 7.2) should
be also included in
the optimisation
sub-algorithm
specified for
selection of the
signals that give the
best probe
directional
sensitivity to the
instant direction of
fluid velocity. An
0
0.1
0.2
1 10 100
y
+
Re=3300 Re=4940
Re=6790 Re=8260
Re=10040 Re=12180
Re=15000 Re=17500
Kuroda et al. 1989
Kuroda et al. 1993
Horiuchi et al. 1992
Gilbert and Kleiser 1991
Antonia et al. 1992
u
c
'
/
U

0.3
0.4

Fig. 7.8: Turbulence intensity close to the wall, as a function
of Reynolds number, obtained by numerical and
experimental studies. Source: Schenck 1997.
MULTIPLE HOT-WIRE PROBES 171
__________________________________________________________________________________________________________________________

additional sub-algorithm, which chooses the best signal, not only between F
1
(V)
and F
3
(V), but also between F
2
(W) and F
4
(W) was developed.

u
=0
0
; =20
0

Vertical sensors optimizing: F
1,3
(V)=0

1
3
2 4

u
=0
0
; =20
0

Horisontal sens. optimizing: F
2,4
(W)=0

1
3
2 4

u
=0
0
; =0
0

Vertical sensors optimizing: F
1,3
(V)=0

1
3
2 4

u
=0
0
; =0
0

Horisontal sens. optimizing: F
2,4
(W)=0

1
3
2 4


u
=0
0
; =-20
0

Vertical sensors optimizing: F
1,3
(V)=0

1
3
2 4

u
=0
0
; =-20
0

Horisontal sens. optimizing: F
2,4
(W)=0

1
3
2 4

Fig. 7.9: Criteria functions F
1,3
(V), which correspond to probe VP-4+
calibration in the vertical plane of hot-wires 1 and 3 (left).
Fig. 7.10: Criteria functions F
2,4
(W), which correspond to probe VP-4+ pitch
calibration in the (horizontal) plane of wires 2 and 4 (right).
Source (both figures): Petrovi 1996a.
172 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Figs. 7.9 and 7.10 present flow charts of the functions F
1,3
(V) and F
2,4
(W),
respectively, in the case of "pitch" calibration for the different values of V velocity
component and W = 0. This functions are obviously different, except for V = W.
For high values of V (positive or negative) and W = 0, one of the functions F
1
(V) or
F
3
(V) are near the critical range what is not the case with F
2,4
(W) and vice versa.
This can affect the convergence (the number of rejected points) and the accuracy of
the data reduction procedure.
However, one other effect can also have strong influence on the choice of the
best combination of available signals. As it is explained in section 7.2, the basic
SB procedure assumes zero velocity gradients over the probe sensing volume. In
turbulent flows, these gradients are random quantities changing its values
temporally in all directions. The best choice will be the sensor combination that are
less affected by velocity gradients. In the near-wall region, this will be mainly the
combination of sensors 2 and 4, with one of wires 1 or 3. So, a choice between
characteristic functions F
1
and F
3
will be the most appropriate although a cases
with functions F
2
and F
4
are also possible. In a grid-turbulence, any choice of the
best three signals is of the same probability.
Unfortunately, the orientation of velocity vector and intensity of velocity
gradients are unknown in advance. The problem can be resolved in different ways.
One of the possibilities is to optimise by both, F
1,3
(V) and F
2,4
(W). If only one
procedure converges the available result should be accepted as final. If both of
them converge, the obtained results should be very close, and an average value can
be accepted or additional criteria may be introduced to choose the better one.
In order to account the influence of velocity magnitudes, the average calibration
coefficients of the adopted hot-wire cooling law, which correspond to wide
velocity range are used. They are evaluated using all calibration output voltages,
measured through minimum three (low, medium and high-velocity) directional
calibrations, which should "cover" the whole range of fluid velocity intensities that
are of interest on the current measuring location. This (first) step achieves low
precision of signal interpretation, but still high enough to estimate the velocity
range and choose the local coefficients for the second high-precision stage. Local
calibration coefficients, used in the second stage of interpretation, are sorted in
three groups that correspond to closer sub-intervals of the velocity magnitudes and
minimum one directional probe calibration per each group.
Direct and simultaneous testing of the SB and DD procedure gave quite
encouraging results. It was performed by comparing reproduced (REP) and real
(IND) values of fluid velocity components, induced during calibration, e.g. by
computing the resulting reproduction errors:
( ) ( ) 100 * U / U U % U
IND IND REP REP
= c ,
( ) ( ) 100 * U / V V % V
IND IND REP REP
= c , (7.60)
( ) ( ) 100 * U / W W % W
IND IND REP REP
= c .
MULTIPLE HOT-WIRE PROBES 173
__________________________________________________________________________________________________________________________

Two multiple-
velocity calibration data
files were used to meet
the expected conditions
in the real turbulent
flow. The first, HIG
file, contains output
voltages sampled during
angular calibrations at:
1.0, 1.4 and 1.9m/s. In
the LOW calibration
file, these velocities are:
0.6, 1.0 and 1.4m/s.
Typical map (fig. 7.4) of
angular calibration is
equal to the map of
SB algorithm.
Mean reproduction
errors in a calibration
procedure (fig. 7.11)
were roughly 1-2.5% in
the case of SB
algorithm and 0.5-0.8%
for DD procedure.
However, the analysis of
average errors is not
always the best way to
estimate accuracy of
different procedures,
because their differences
decrease through the
averaging process. In
that sense, fig. 7.12 that
presents maximum
reproduction errors of
induced velocity vector
components seems to be
much more useful: their
values are generally
between 5 and 20% if
SB procedure is used,
while they reach 2.5%
extremely rare with DD algorithm applied.
0
1
2
HIG-1 HIG-2 HIG-3 LOW-1 LOW-2 LOW-3
4+SB
4+DD
R
M
S
(
c
R
E
P
)
S
U
M

(
%
)


Fig. 7.11: RMS values of reproduction errors of all three
components of the velocity vectors induced
during probe calibration.
0
5
10
15
4+SB
4+DD
c
R
E
P
(
U
)
M
A
X

(
%
)


c
R
E
P
(
V
)
M
A
X
(
%
)

0
5
10
4+SB
4+DD
20%

0
5
10
HIG-1 HIG-2 HIG-3 LOW-1 LOW-2 LOW-3 c
R
E
P
(
W
)
M
A
X

(
%
)

4+SB
4+DD

Fig. 7.12: Maximum reproducing errors of calibration
velocity components in the HIG-1,2,3 and
LOW-1,2,3 range, for arrays no. 1, 2, 3 of the
probe WP-12+(G).
Source: Petrovi and Vukoslavevi 1997.
174 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Preliminary results show that DD procedure generates lower average
reproduction errors, but still fairly close to corresponding errors of SB algorithm.
However, significant differences arise when maximal reproduction errors are
compared. Therefore, following these results, it seems that measurement results of
the mean (time-averaged) flow velocity and lower-order velocity correlations
should be reliable, even at low velocities, whether the older SB or new DD
procedure is applied. However, in the case of higher-order statistics of turbulent
velocity field, the new algorithm should be applied to achieve high accuracy of hot-
wire signals interpretation.
The basic SB and improved DD algorithms were simultaneously applied for
interpretation of anemometer output signals of the VP-4+ probe sampled in the
turbulent boundary layer. Real-time series of velocity components that correspond
to the location at y
+
=10.73 and the second array of the probe WP-12+(G) are
presented in figs. 7.13, 7.14 and 7.15. They illustrate wide variations of velocity
components in the turbulent boundary layer. It can be seen in fig. 7.13 that
longitudinal U velocity component vary from 0.5m/s to about 1.7m/s in the interval
between 50ms and 70ms. During the same interval, magnitude of V component
varies in the range between 0.6m/s and +0.1m/s and W component fluctuates
between 0.7m/s and +0.2m/s approximately.
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
0 25 50 75 100 125 150 175 200 225
SB
DD
t (ms)
U (m/s)
Divergence of
"SB" procedure

Fig. 7.13: Simultaneous real-time series of longitudinal U velocity component,
computed by SB and DD procedure.
U

(
m
/
s
)

t (s)
MULTIPLE HOT-WIRE PROBES 175
__________________________________________________________________________________________________________________________

-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 25 50 75 100 125 150 175 200 225
SB
DD
t (ms)
V(t) (m/s)
Divergence of
"SB" procedure

Fig. 7.14: Simultaneous real-time series of transversal V component, computed by
SB and DD procedure.
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 25 50 75 100 125 150 175 200 225
SB
DD
t (ms)
Divergence of "SB" procedure
W(t) (m/s)

Fig. 7.15: Simultaneous real-time series of lateral W component, computed by
SB and DD procedure.
V

(
m
/
s
)

W

(
m
/
s
)

t (s)
t (s)
176 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
These figures also show a superiority of the DD procedure in resolving the
convergence problem of signal interpretation. At 1, 2, 56 and 57ms, the SB
procedure diverged, while the improved signal-selection sub-algorithm in DD
procedure enabled its convergence.
The analysis presented in this section clearly shows that numeric procedure for
anemometer signals interpretation may significantly influence the results of
turbulent velocity field measurement. The advantages of new fully optimised DD
algorithm, compared to the basic variant are evident. It provides higher accuracy of
hot-wire output signals interpretation, as well as the higher efficiency in resolving
the fluid velocity field, expressed through smaller percentage of rejected data sets
measured in the real turbulent flow. Therefore, similar advances should be
expected for the variant of generalised numeric procedure specified for the twelve-
sensor probes: WP-12+(G) and WP-12q(G).




Laminar boundary layer separation ahead of the sphere equator at Re = 15000.
Source: Werle 1980.













8 8. . C CO ON NC CL LU US SI IO ON NS S


P Pa an nt ta a r re ei i. .
( (H He er ra ac cl li it te e f fr ro om m E Ef ff fe es s, , a ab bo ou ut t 5 53 35 5- -4 47 75 5 b be ef fo or re e t th he e N Ne ew w A Ag ge e) )


Methodology of three-dimensional simultaneous measurements of isothermal
turbulent velocity and vorticity fields by multiple hot-wire probes is presented in
this manuscript. The analysis is focused on the influences of hot-wire probe
geometry, sensing volume dimensions, calibration procedure, output signals
interpretation algorithm and sampling parameters, on the measured values of the
corresponding statistical parameters.
The first three chapters represent the first part of the textbook, providing
foundations in the scientific area of interest. Besides an attempt to define the
extremely complex phenomenon and basic properties of turbulence, fundamental
operational principles of thermal and laser-Doppler anemometers are generally
analysed and compared from the practical point of view, showing the importance
of both techniques and their complementary status. Theoretical foundations of
constant-temperature hot-wire anemometer are also provided, as well as the
information on its practical operation: electronic circuit design, set-up procedure,
calibration, measurement principles, output voltages interpretation, connecting with
auxiliary electronic equipment, etc. Furthermore, fabrication technology of hot-
wire probes and applied materials are described, taking into account temperature
ranges of their applicability.
The second part, consisting of the rest five chapters, reviews two main groups
of contemporary multiple hot-wire probes and adequate measurement techniques.
The first comprises hot-wire configurations and supporting numeric procedures
specified for turbulent velocity field measurements only, while the probes and
algorithms from the second group are capable of simultaneous three-dimensional
measurements of both the fluid velocity and vorticity fields. Besides reviewing the
recently developed multiple hot-wire probes, a detailed information on their
practical applicability is provided. A special attention is paid to evaluation of the
uniqueness angular range of each specific hot-wire probe configuration.
Furthermore, existing methods for estimation of hot-wire probe spatial resolution
178 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
are also discussed. In both cases, clear recommendations related to multiple sensor
probes selection depending on the number of hot-wires, geometry and
characteristic dimensions, are given.
In that sense, a unique experimental results of comparison testing the vorticity-
type and hot-wire configurations specified for velocity measurements are also
presented and analysed. They show that vorticity probes provide higher accuracy of
turbulent velocity field measurement. They are very important for future design
and employment of adequate hot-wire anemometer probes in the area of turbulence
research. The influence of mean streamwise velocity gradient over the probe
sensing area on turbulence velocity field statistics is analysed. In addition,
performed analysis shows that influence of velocity gradients within the probe
sensing area (four-wire probe) reaches the level of influence of neglected velocity
component (V-probe).
However, although vorticity configurations provide higher measurement
accuracy than velocity probes, they suffer from the problem of interpretation of
enormous number of sampled signals because of the great number of their sensors
(up to twelve) and very complex algorithms needed. Fortunately, it can be
minimised by algorithm of reduced searching of time-frequency domain that is
also presented.
Besides hot-wire probe configuration and procedure for output voltages
interpretation, probe dimensions also represent crucial parameters that strongly
influence the accuracy of turbulent velocity field measurement. A special unique
probe with eight hot-wires is also described together with preliminary results of its
testing. Reviewed results are encouraging, showing the influence of velocity
gradients on the accuracy of velocity field measurements.
Depending on the sophistication level of numeric support, measurement
accuracy of a probe can be significantly increased. As a result of own investigation
through nearly two decades, an original generalised numeric procedure,
specified for evaluation of calibration constants and anemometer signals
interpretation of the probes with two, three, four, nine and twelve hot-wires was
developed. Using existing own reports, it is also presented, together with results of
experimental testing and recommendations for application and future
improvements. Thus, the importance of choosing the optimal calibration constants
and hot-wire probe signals, toward the magnitude of instant velocity vector and its
direction respectively, is discussed in detail. Both optimisations results in improved
angular resolution of four-wire probes and increased interpreting precision of their
output signals. They are included in the new, fully-optimised variant that supports
"quadruple" probes: dual directional generalised procedure. Its advantages are
visible especially at low fluid velocities, as its authors experimentally verified. At
present, this is unique algorithm experimentally verified for efficient application of
a four-sensor probe at velocities below 2m/s (the procedure was tested in the range
of 0.5m/s - 3.0m/s approximately).
MULTIPLE HOT-WIRE PROBES 179
__________________________________________________________________________________________________________________________

Future researching plans in this area should include simultaneous comparison
tests of the newest five-sensor probe VP-5p of pentagram geometry, with
quadrate VP-4q and plus VP-4+ four-wire configurations. In addition, the
probe WP-12q(G) with three quadrate arrays VP-4q, which advantages have not
been fully used yet, should be also tested together with WP-12+(G).
Besides direct analysis of the influence of hot-wire probe configuration on the
turbulence velocity field statistics, it is also interesting to analyse its influence on
the functional shape of higher-order moments interrelations, as it was done in the
work of Petrovi, Schenck and Vukoslavevi 1997. Their results are quite
unexpected: although the values of statistical parameters, measured by different
hot-wire configurations were different, their interrelations were in nearly perfect
agreement.
Also, positive experience of employing dual directional generalised
procedure with the probe VP-4+, especially in the slow turbulent flow, provides
possibility for additional improving the existing algorithm that supports twelve-
wire vorticity probe WP-12+(G).


Instability of a round jet. Source: Van Duke 1982.










9 9. . R RE EF FE ER RE EN NC CE ES S


E Ea ac ch h f fl la am me e i is s f fa ad di in ng g a aw wa ay y. . S Sh ho or rt t i is s i it ts s l li if fe e, , a an nd d w wh he en n i it t g go oe es s o ou ut t a an n a as sh h r re em ma ai in ns s o on nl ly y - - i it ts s
t tr ra ac ce e. . H Ho ow we ev ve er r, , t th he e a as sh h d de et te er rm mi in ne es s t th he e f fl la am me es s s se en ns se e a an nd d i im mp po or rt ta an nc ce e. .
( (V Vu ul le et ta a M M. . i iv vo oj ji in n, , i in n t th he e b bo oo ok k A A F Fl la am me e a an nd d i it ts s A As sh h , , 1 19 98 88 8) )


Accrivlellis M. (1979): A New Evaluation Method of Hot-Wire Anemometer Measure-
ments in Low and Highly Turbulent Flows". Report, Karlsruhe University, Institute of
Flow Mechanics and Flow Machinery. Karlsruhe.
Acrivlellis M. (1980): Measurement of Turbulent Velocity Fields by Means of Triple
Sensor Probes. J. Phys. E: Sci. Instrum., no. 13, p. 986-992.
Adrian R. J., Johnson R. E., Jones B. G., Merati P. and Tung A. T.-C. (1984):
Aerodynamic Disturbances of Hot-Wire Probes and Directional Sensitivity. J. Phys.
E: Sci. Instrum. no. 17., p. 62-71.
Al-Kayiem H. H. (1989): Separated Flow on a High Lift Wing.Ph. D. Theiss,
Department of Mechanical Engineering, University of Bradford.
Al-Kayiem H. H. and Bruun H. H. (1991): Evaluation of a Flying Hot-Wire Probe
System, Meas. Sci. Technol. 2, p. 374-380.
Andreopulos Y. (1983): Improvements of the Performance of Three Hot-Wire Probes.
Rev. of Sci. Instrum. 54, p. 733-740.
Andreopulos Y. and Honkan A. (1996): Experimental Techniques for Highly Resolved
Measurements of Rotation, Strain and Dissipation-Rate Tensors in Turbulent Flows.
Meas. Sci. Technol. 7, p. 1462-1476.
Antonia R. A., Anselmet F. and Chambers A. J. (1986): Assesment of Local Isotropy
Using Measurements in a Turbulent Plain Jet. J. Fluid Mech., vol. 163, p. 365-391.
Antonia R. A., Browne L. W. B. and Chambers A. J. (1984): On the Spectrum of the
Transverse Derivatives of the Streamwise Velocity in a Turbulent Flow". Phys. Fluids
no. 27, p. 2628-2631.
Antonia R. A., Browne L. W. B. and Shah D. A. (1988): Characteristics of Vorticty
Fluctuations in a Turbulent Wake. J. Fluid Mech. vol. 189, p. 349-365.
Antonia R. A. and Kim J. (1994): A Numerical Study of Local Isotropy of Turbulence.
Phys. Fluids A 6, p. 834-841.
Antonia R. A. and Rajagopalan S. (1990): Performance of lateral Vorticity Probe in a
Turbulent Wake. Exp. in Fluids no. 9, p. 118-120.
Antonia R. A., Shah D. A. and Browne L. W. B. (1987): Spectra of Velocity Derivatives
in a Turbulent Wake. Phys. Fluids 30, p. 3455-3462.
MULTIPLE HOT-WIRE PROBES 181
__________________________________________________________________________________________________________________________

Antonia R. A., Teitel M., Kim M. and Browne L.W.B. (1992): J. Fluid Mech., vol. 236,
p. 579-605.
Antonia R. A., Zhu Y. and Kim J. (1994): Corrections for Spatial Velocity Derivatives
in a Turbulent Shear Flow. Exp. in Fluids 16, p. 411-413.
Antonia R. A., Zhu Y. and Kim J. (1993): On the Measurement of Lateral Velocity
derivatives in Turbulent Flows. Exp. in Fluids 15, p. 65-69.
Antonia R. A., Zhou T. and Zhu Y. (1998): Three-Component Vorticity Measurem. in a
Turbulent Grid Flow. J. Fluid Mech., vol. 374, p. 29-57.
AUSPEX (1994): AUSPEX Standar Hot-Wire Probes (DANTEC Compatible).
AUSPEX Corporation, Bethlehem, USA.
Badran O. O. (1993): A Flying Hot-Wire Study of Separated Flows.Ph. D. Thesis,
Department of Mechanical and Manufacturing Engineering, University of Bradford.
Balint J. -L., Vukoslavevi P. and Wallace J. M. (1987): A Study of the Vortical
Structure of the Turbulent Boundary Layer. In: Advances in Turbulence (ed. Comte-
Bellot and J. Mathieu), p. 456-464, Springer.
Balint J. -L., Wallace J. M. and Vukoslavevi P. (1989): The Statistical Properties of
the Vorticity Field of a Two-Stream Mixing Layer. Advances in Turbulence 2. Second
European Turbulence Conference 1988 (ed. H. H. Fernholtz, H. E. Fiedler), p. 74-78.
Springer-Verlag, Berlin/Heidelberg/New York.
Balint J. -L., Vukoslavevi P. and Wallace J. M. (1990): The Transport of Enstrophy
in a Turbulent Boundary Layer. In: Near Wall Turbulence. (1988 Zoran Zari
Memorial Conference, ed. S. J. Kline and N. H. Agfan), p. 932-950. New
York/Washington, Philadelphia/London: Hemisphere.
Balint J. L., Wallace J. M. and Vukoslavevi V. P. (1991): The Velocity and Vorticity
Vector Fields of a Turbulent Boundary Layer. Part 2: Statistical Properties. J. Fluid
Mech. no. 228, p. 53-86.
Bernal P. L., Breidenthal E. R., Brown. L. G., Conrad H. J. and Roshko A. (1979).
Selected Papers from the Second International Symposium on Turbulent Shear Flows,
Imperial College, London.
Bhatia J. C., Durst F. and Jovanovi J. (1982): Corrections of Hot-Wire Anemometer
Measurements Near Wall. J. Fluid Mech. no. 122, p. 411-431.
Blackwelder R. F. (1981): Hot-Wire and Hot-Film Anemometers. In: Methods of
Experimental Fluids: Fluid Dynamics, (ed. R. J. Emrich), vol. 18, Part A, p. 259-314.
Academic Press.
Blanco-Marigota E., Ballasteros-Tajadura R. and Santolaria C. (1998): Angular
Range and Uncertainty Analysis of Nonorthogonal Crossed Hot Wire Probes. Trans. of
ASME: J. Fl. Engineering, vol. 120, p. 90-94.
Bradshaw P. (1970): Experimental Fluid mechanics. Oxford: Pergamon Press.
Bradshaw P. (1971): An Introduction to Turbnulence and its Measurement. Pergamon
Press, Oxford.
Bradshaw P. (1975): An Introduction to Turbnulence and Its Measurements, Oxford
/New Yourk/Toronto/Sydney/Braunschweig: Pergamon.
Bradshaw P., Ferris D. H. and Johnson R. F. (1964): J. Fluid Mechanics 19, p. 591-624
(according to Van Duke 1982).
Bremhorst K. (1972): The Effect of Wire Length and Separation on X-Array Hot-Wire
Anemometer Measurements. IEEE Trans. Instr. Measurements, IM-21(3): p. 244-248.
Brown G. L. and Roshko A. (1974): On Density Effects in Turbulent Mixing Layers. J.
Fluid Mech. vol. 64, p. 775-816.
182 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Browne L. W. B. (1988): Selection of Wires and Wire Spacing for X-wires. Exp. in
Fluids 6, p. 286-288.
Browne L. W. B., Antonia R. A. and Chambers A. J. (1983): Effects of the Separation
Between Cold Wires on the Spatial Derivatives of Temperature in a Turbulent Flow.
Boundary Layer Meteorology 27, p. 129-139.
Browne L. W. B., Antonia R. A. and Shah D. A. (1987): Turbulent Energy Dissipation
in a Wake. J. Fluid Mech. 179, p. 307-326.
Browne L. W. B., Antonia R. A. and Shah D. A. (1988): Selection of Wires and Wire
Spacing for X-Wires. Exp. in Fluids no. 6, p. 286-288.
Bruun H. H. (1976): A Note on Static and Dynamic Calibration of Constant-Temperature
Hot-Wire Probes. J. Fluid Mech. 76, p. 145-155.
Bruun H. H. (1995): Hot-Wire Anemometry, Principles and Signal Analysis. Oxford
University Press, Oxford/New York/Tokio.
Bruun H. H. (1996): Hot-Wire Anemometry in Liquid Flows. Meas. Sci. Technol. 7, p.
1301-1312.
Bruun H. H., Nabhani N., Al-Kayiem H. H., Fardad A. A., Khan M. A. and Hogarth
E. (1990): Calibration and Analysis of X Hot-Wire Probe Signals. Meas. Sci. Tech. 1,
p. 782-785.
Bruun H. H. and Tropea C. (1980): Calibration of Normal, Inclined and X-Array Hot
Wire Probe. Sonderforschungbereich 80. University of Karlsruhe, FR Germany.
Buddhavarapu J. and Meinen D. (1988): A New Data Analysis Scheme and a Three
Sensor Probe for Measurements in a 3-D flow. TSI Flow lines.
Buttler T. L. and Wagner J. W. (1983): Application of a Three Sensor Hot-Wire Probe
for Incompressible Flow. AIAA J. 21, p. 726-732.
Cantwell B. J. (1976): A Flying Hot-Wire Study of the Turbulent Near-Wake of a
Circular Cylinder at a Reynolds Number of 140000. Ph. D. Thesis, California Institute
of Technology, USA.
Cantwell B. J. and Coles D. E. (1983): An Experimental Study of Entrainment and
Transport in the Turbulent Near Wake of a Circular Cylinder. J. Fluid Mech., vol. 136,
p. 321-374.
Champagne H. F., Sletcher C. A. and Wehrmann O. H. (1967): Turbulence
Measurement with Inclined Hot-Wires. Part I: Heat Transfer Experiments with Inclined
Hot-Wires. J. Fluid Mech., vol. 28, p. 153-176.
Chang P. H., Adrian R. J. and Jones B. G. (1983): Comparison Between Triple-Wire
and X-wire Meas. Techniques in High Intensity Turbulent Shear Flow. Proceedings of
The 8
th
Biennal Symposium on Turbulence, University of Missouri-Rolla, USA.
Cleveland W. G. (1979): M. Sc. Thesis, The University of Maryland, Maryland, USA
(according to Vukoslavevi 1981).
Coles D. (1962): The Turbulent Boundary Layer in Compressible Fluid. Appendix A: A
Manual of Experimental Practice for Low Speed Flow. Random Report R403R-PR,
ARC 24473.
Coles D. and Wadcock A. J. (1979): Flying Hot-Wire Study of Flow past a NACA 4412
Airfoilat at Maximum Lift. J. AIAA no. 17, p. 321-329.
Collis D. C. and Williams M. J. (1959): Two-Dimensional Convection from Heated
Wires at Low Reynolds Numbers. J. Fluid Mech. 6, p. 357-384.
Comte-Bellot G. (1976): Hot-Wire Anemometry. Annual Rev. Fl. Mech. 8, p. 209-231.
Comte-Bellot G., Strohl A. and Alcaraz E. (1971): On Aerodynamic Disturbances
Caused by Single Hot-Wire Probes. J. Appl. Mech. 38, p. 767-774.
MULTIPLE HOT-WIRE PROBES 183
__________________________________________________________________________________________________________________________

Corsin S. (1963): Turbulence: Experimental Methods. Handbuch der Physik. vol. 8.2, p.
523-590. Springer-Berlin.
Corrsin S. and Kistler A. L. (1954): The Free-Stream Boundaries of Turbulent Flows.
NACA TN 3133.
Corrsin S. and Kistler A. L. (1955): Free-Stream Boundaries of Turbulence Flows.
NACA Report 1244.
Crouch J. D. and Saric W. S. (1986): Oscillating Hot-Wire Measurements above an
FX63 airfoil. AIAA 24
-th
Aerospace Sciences Meeting, January 6.-9., Nevada, AIAA-
86-0012.
DANTEC (1996): Probe Catalog - CTA, Dantec Measurement Technology A/S,
Skovlunde, Denmark.
Devenport W. J., Glegg S. A. L. and Sharma G. (1992): Turbulence Measurements in
Trailing Vortices for BWI Noise Prediction. Report to NASA Langley, under grant
NAG-1-1119.
Dobbeling K., Lenze B. and Leuckel W. (1990a): Basic Considerations Concerning the
Construction and Usage of Multiple Hot-Wire Probes for Highly Turbulent 3-
Dimensional Flows. Meas. Sci. Technol. 1, p. 924-933.
Dobbeling K., Lenze B. and Leuckel W. (1990b): Computer-Aided Calibration and
Measurements with a Quadruple Hotwire Probe. Exp. in Fluids 8, p. 257-262.
Durst F. and Jovanovi J. (1995): Investigations of Reynolds-Averaged Turbulence
Quantities, Proc. Royal Soc. London A 451, p. 105-120.
Durst F., Jovanovi J. and Johanson G. T. (1992): On the Statistical Properties of
Truncated Gram-Chalier Series Expansion in Turbulent Wall-Bounded Flows. Phys.
Fluids, A4(1), p. :118-126.
Durst F., Jovanovi J. and Kaneve Lj. (1987): Probability Density Distribution in
Turbulent Wall-Bounded Shear Layer Flows. Turbulent Shear Flows, Springer-Verlag
no. 5, p.197-220.
Durst F., Jovanovi J. and Unger T. (1997): Method for Very Low Volume Rate
Measurements and its Application. Proc. of the 8
-th
Internationale Messe mit Kongress
fur Sensoren, Messaufnehmer and Systems, Nurnberg 13.-15. Mai, p. A6.3.1-A6.3.5.
Durst F., Launder B. E. and Kjelstrom B. (1971): The Effect of Compressibility on the
Shear-Stress Distribution in Turbulent Pipe Flow. Aeronaut. J. R. Aeronaut. Soc. 75, p.
55-57.
Durst F., Melling A. and Whitelaw H. J. (1971): The Interpretation of Hot-wire Signals
in Low Turb. Flows. Report of Department of Mechanical Engineering, Imperial
College, London, S.W.7.
Durst F., Noppenberger S., Still M. and Venzke H. (1996): Influence of Hymidity on
Hot-Wire Measurements. Meas. Sci. Technol. 7, p. 1517-1528.
Eckelmann H. , Nuchas S. G., Brodkey R. S. and Wallace J. M. (1977): Vorticity and
Turbulence Production in Pattern Recognised Turbulent Flow Structures. Phys. Fluids
no. 20, p. 225-231.
Ewing D., Hussein H. J. and George W. K. (1995): Spatial Resolution of Parallel Hot-
Wire Probes for Derivative Measurements. Exp. Thermal Fluid Sci., 11, p. 155-173.
Fabris G. (1978): Probe and Method for Simultaneous Measurements of True Instantane-
ous Temperature and Three Velocity Components in Turbulent Flow. Rev. of Sci.
Instr. 49, p. 654-664.
Falco R. E. (1983): New Results - A Review and Synthesis of the Mechanism of
Turbulence Production in Boundary Layer and its Modification. Rep.: AIAA-83-0377.
184 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Fernholz H. H., Krause E., Nockemann M. and Schober M. (1995): Phys. Fluids, vol.
7(6), pp. 1275-1281.
Fiedler H. (1978): Data Acquisition and Signal Processing in Heated Turbulent Flows.
Proceedings of Dynamic Flows Conference (Marseille) p. 81-100.
Fingerson L. M. and Freimuth P. (1983): Thermal Anemometers. Fluid Mechanics
Measurements, (ed. R. J. Goldstein et al), p. 99-154. Hemisphere, Washington.
Schenck T. (1997): Personal communication, LSTM Erlangen, Technical Faculty,
University of Erlangen-Nurnberg, FR Germany.
Fisher M., Durst F. and Jovanovi J. (1998): Reynolds Number-Influence on Near-Wall
Behaviour of Statistical Properties in Channel Turbulent Flows, Proc. of ASME
FEDSM98, Washington DC, p. 4956.1-6.
Foss J. F. (1976): Accuracy and Uncertainty of Transverse Vorticity Measurements.
Bull. Am. Phys. Soc. No. 21, p. 1237 (abstract).
Foss J. F. (1981): Advanced Techniques for Transverse Vorticity Measurements.
Proceedings of the 7-th Biennal Symposium on Turbulence, University of Misouri Rolla
1980, p. 208-218.
Foss J. F. (1994): Vorticity Considerations and Planar Shear Layers. Exp. Therm. Fluid
Sci., no. 8, p. 260-270.
Foss J. F., Ali S. K. and Haw R. C. (1987): A critical Analysis of Transverse Vorticity
Measurements in a Large Plane Shear Layer. In: Advances in Turbulence. First
European Turbulence Conference 1986, ed. Comte-Bellot, J. Mathieu, p. 446-455,
Berlin/Heidelberg/New York: Springer-Verlag.
Foss J. F. and Haw R. C. (1990a): Transverse Vorticity Measurements Using a Compact
Array of Four Wires. In: The Heuristics of Thermal Anemometry (ed. D. E. Stock, S.
A. Sherif, A. J. Smits), ASME-FED 97, p. 71-76.
Foss J. F. and Haw R. C. (1990b): Vorticity and Velocity Measurements in a 2:1 Mixing
Layer. In; Forum on Turbulent Flows, ed. W. W. Bowers, M. J. Morris, M. Samminy,
ASME-FED 94, p. 115-120.
Foss J. F., Klewicki C. J. and Disimile P. J. (1986): Transverse Vorticity measurements
Using an Array of Four Hot-Wires. NASA CR 178098.
Foss J. F. and Wallace J. M. (1989): The Measurement of Vorticity in Transitional and
Fully Developed Turbulent Flows. In: Advances in Fluid Mechanics Measurements
(ed. Gad-el-Hak M.). Lect. Notes in Engineering, vol. 45, p. 263-321, Springer-Verlag.
Frenkiel F. N (1948): Etude statistique de la turbulence. 1. Mesure de la turbulence avec
un fil chaud non-compense. 2. Influence de la longueur dun fil chaud compense sur la
mesure de la turbulence Rapport 37, ONERA.
Frenkiel F. N and Klebanoff P. S. (1973): Probability Distributions and Correlations in a
Turbulent Boundary Layer. Phys. Fluids, vol. 16, no. 6, p. 725-737.
Freimuth P. (1967): Feedback Controll Theory for Constant-Temperature Hot-Wire
Anemometers. Rev. Sci. Instr. 38, p. 677-681.
Freimuth P. (1983): History of Thermal Anemometry. In: Handbook of Fluids in
Motion (ed. N. S. Cherenisimoff and R. A. Gupta), p. 79-91. Ann Arbor Science
Publishers, Ann Arbor.
Freimuth P. (1992): Bibliography of Thermal Anemometry. TSI Inc.
Friehe C. A. and Schwartz W. H. (1968): Deviations from the Cosine Law for Yawed
Cylindrical Anemometer Sensors. J. Appl. Mech. 35, p. 655-662.
Fujita H. and Kovasznay L. S. G. (1968): Measurement of Reynolds Stress by a Single
Rotated Hot-Wire Anemometer. Rev. Sci. Instr. 39, p. 1351-1355.
MULTIPLE HOT-WIRE PROBES 185
__________________________________________________________________________________________________________________________

Gad-el-Hak M. (1998): Fluid Mechanics from the Beginning to the Third Millenium,
Proc. of ASME FEDSM98, Washington DC), p. 5134/1-8.
Gilbert N. and Kleiser L. (1991): Proceedings of the 8-th Symposium on Turbulent Shear
Flows, TU Munich, Germany, pp. 26.1.1-6.
Haw R. C., Foss J. K. and Foss J. F. (1989): The Vortical Properties of the High-Speed
Region in a Plane Shear Layer and its Parent Boundary Layer. Advances in Turbulence
2. Second European Turbulence Conference 1988 (ed. H. H. Fernholtz, H. E. Fiedler),
p. 90-95. Springer-Verlag, Berlin/Heidelberg/New York.
Head M. R. and Bandyopadhyay P. (1981): J. Fluid Mech., vol. 107, p. 297-338
(according to Van Duke 1982).
Heist D. K. and Castro I. P. (1998): Combined Laser-Doppler and Cold Wire
Anemometry for Turbulent Heat Flux measurement. Exp. in Fluids 24, p. 375-381.
Hinze O. J. (1959/1975): Turbulence. An Introduction to its Mechanism and Theory.
McGraw-Hill Book Company, New York.
Honkan, A. (1993): An Experimental Study of the Vortical Structure of Turbulent
Flows. Ph.D. Thesis, The City University of New York, NewYork.
Holzapfel F., Lenze B. and Leuckel W. (1994): Assessment of a Quintuple Hot-Wire
Measurement Technique for Highly Turbulent Flows. Exp. in Fluids 18, p. 100-106.
Horiuchi K. (1992): Report Gants-in-Aid for Scientific Research, No. 02302043.
Huffmann G. D. (1980): Calibration of Triaxial Hot-Wire Probes Using Numerical
Search Algorithm. J. of Physics E: Sci. Instrum. 13. p. 1177-1182.
Hussein H. J. (1990): Measurements of Turbulent Flows with Flying Hot-Wire
Anemometry, ASME Proc. of Symposium on the Heyristics of Thermal Anemometry,
FED, vol. 97, University of Toronto, p. 77-80..
Hussein H. J. and George W. K. (1989): Measurement of Small-Scale Turbulence in an
Axisymmetric Jet Using Moving Hot-Wires. Proceedings of the Seventh Symposium
on Turbulent Shear Flow, Stanford University, p. 30.2.1-30.2.6.
Hussein H. J. and George W. K. (1990): Influence of Wire Spacing on Derivative
Measurements with Parallel Hot-Wire Probes. In: Forum on Turbulent Flows-1990
(ed. W. M. Bower, M. Samminy), no. 94, p. 121-124.
Hussain A. K. M. F. and Zedan F. (1978a): Effect of the Initial Condition on the
Axisymmetric Free Shear Layer: Effect of the Initial Fluctuation Level, Phys. Fluids,
Vol. 21 no. 7, p. 1100-1112.
Hussain A. K. M. F. and Zedan F. (1978b): Effect of the Initial Condition on the
Axisymmetric Free Shear Layer: Effect of the Initial Momentum Thickness, Phys.
Fluids, Vol. 21 no. 9 1475-1481.
Jaju A. A. R. (1987): Development of a Flying Hot-Wire probe System, M. Phil. Thesis,
Department of Mechanical and Manufacturing Engineering, University of Bradford.
Jerome F. E., Guitton D. E. and Patel R. P. (1971): Experimental Study of the Thermal
Wake Interference Between Closely Spaced Wires of an X-type Hot-Wire Probe.
Aeronautical Quarterly 22, p. 119-126.
Jorgensen F. (1971): Directional Sensitivity of Wire and Fibre Film Probes. DISA Inf.
no. 11, p. 31-37.
Jorgensen F. (1996): The Computer-Controlled Constant-Temperature Anemometer.
Aspects of Set-Up, Probe Calibration, Data Acquisition and Data Conversion. Meas
Sci. Technol. 7, p. 1378-1387.
Jovanovi J. (1984): Statistical Analysis and Structure of Wall Turbulence. Ph.D.
Thesis, Faculty of Mechanical Engineering, University of Belgrade, Yugoslavia.
186 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Jovanovi J., Ye Q.-Y and Durst F. (1995): Statistical Interpretation on the Turbulent
Dissipation Rate in the Wall Bounded Flows. J. Fluid Mech. vol. 293, p. 321-347.
Kampe de Feriet J. (1966): David Taylor model basin report no. 2013, Naval Ship
Research and Development Centre, Washington D. C.
Kaneve G. and Oka S. (1973): Correcting of Hot-Wire Readings for Temperature
Variations. DISA Inf. no. 15, p. 21-24.
Karlsson R. I. and Johansson T. G. (1988): LDV Measurements of Higher Order
Moments of Velocity Fluctuations in Turbulent Boundary Layer. In: Laser Anemo-
metry in Fluid Mechanics". Ladoan-Instituto Superior Technico, Lisbon, Portugal.
Kastrinakis E. G. (1976): An Experimental Investigation of the Fluctuations of the
Streamwise Components of the Velocity and Vorticity Vectors in a Fully Developed
Turbulent Channel Flow. Ph.D. Dissertation, Georg-August University in Gottingen.
Kastrinakis E. G. and Eckelmann H. (1983): Measurement of Streamwise Vorticity
Fluctuations in a Turbulent Channel Flow. J. Fluid Mech. 137, p. 165-186.
Kastrinakis E. G., Eckelmann H. and Willmarth W. W. (1979): Influence of the Flow
Velocity on a Kovasznay Type Vorticity Probe. Rev. Sci. Instr., vol. 50, p. 759-767.
Kastrinakis E. G., Nychas S. G. and Eckelmann H. (1983): Some Streamwise Vorticity
Characteristics of Coherent Structures. In: Structure of the Compelx Turbulent Shear
Flow, (ed. R. Dumas and L. Fulachier), p. 31-40.
Kastrinakis E. G., Wallace J. M., Willmarth W. W., Ghorashi B. and Brodkey R. S.
(1977): On the Mechanism of Bounded Turbulent Shear Flows. Lecture notes in
Physics 75, p. 175-189.
Kawall J. G., Shokr M. and Keffer J. F. (1983): A Digital Technique for the
Simultaneous Measurement of Streamwise and Lateral Velocities in Turbulent Flows.
J. of Fluid Mech., 133, p. 83-112.
Kida S. and Murakami Y. (1989): Statistics of Velocity Gradients in Turbulence at
Moderate Reynolds Numbers. Fluid Dyn. Res. 4, p. 347-370.
Kim J. (1989): On the Structure of Pressure Fluctuations in Simulated Turbulent Channel
Flow. J. Fluid Mech. 205, p. 421-451.
Kim J. H. (1989): Wirbelstarkemessungen in Einer Turbulenten Scherschicht. Doktor-
Ing. Thesis, Tech. Univ. Berlin, Germany.
Kim J. H. and Fiedler H. E. (1989): Vorticity Measurements in a Turbulent Mixing
Layer. In: Advances in Turbulence 2. Second European Turbulence Conference 1988
(ed. H. H. Fernholtz, H. E. Fiedler), Berlin/Heidelberg/New Yourk: Springer-Verlag, p.
267-271.
Kim J., Moin P. and Moser R. (1987): Turbulence Statistics in Fully Developed Channel
Flow at Low Reynolds Numbers. J. Fluid Mech. 177, p. 133-166.
King L. V. (1914): On the Convection of Heat from Small Cylinders in a Stream of
Fluid. Physical Transactions of Royal Society London, A214, p. 373-432.
King L. V. (1915): Phil. Mag. 29(6), p. 556-577 (according to Comte-Bellot 1976).
King L. V. (1916): J. Franklin Inst. 181, p. 1-25 (according to Comte-Bellot 1976).
Kistler A. L. (1952): The Vorticity meter. M. Sc. Thesis, The John Hopkins University,
USA.
Klebanoff P. S. (1954): Characteristics of Turbulence in a Boundary Layer with Zero
Pressure Gradient. NACA TN3178.
Klewicky J. C. (1989a): On the Interactions Between the Inner and Outer Region Motions
in Turbulent Boundary Layer. Ph.D. Thesis, Michigan State University, East Lansing,
USA.
MULTIPLE HOT-WIRE PROBES 187
__________________________________________________________________________________________________________________________

Klewicky J. C. (1989b): Velocity-Vorticity Correlations Related to the Gradients of the
Reynolds Stresses in Parallel Turbulent Flows. Phys. of Fluids A no. 1, p. 1285-1288.
Klewiski J. C. and Falco R. E. (1990): On Accurately Measuring Statistics Associated
with Small Scale Structure in Turbulent Boundary Layers Using Hot-Wire Probes. J.
Fluid Mech. 219, p. 119-143.
Klewiski J. C., Falco R. E. and Foss J. F. (1992): Some Characteristics of the Vortical
Motions in the Outer Region of Turbulent Boundary Layers. J. Fluids Eng. no. 114, p.
530-536.
Klewiski J. C., Gendrich C. P., Foss J. F. and Falco R. E. (1990): On the Sign of the
Instantaneoys Spanwise Vorticity Component in the Near-Wall Region of Turbulent
Boundary Layers. Phys. Fluids A, 2, p. 1497-1500.
Klewiski J. C., Murray J. A. and Falco R. E. (1994): Vortical Motion Contributions to
Stress Transport in Turbulent Boundary Layers. Phys. Fluids A no. 6, p. 277-286.
Ko N. W. M. and Davies P. O. A. L. (1971): Interference Effect of Hot-Wires. IEEE
Trans. Instr. and Meas. 20, p. 76-78.
Kock am Brink B. and Foss J. F. (1993): Enchanced Mixing via Geometric
Manipulation of a Splitter Plate. AIAA Pap. 93-3244 (according to Wallace and Foss
1995).
Kolmogorov A. N. (1941): The local Structure of Turbulence in Incompressible Viscous
Fluid for Very Large Reynolds Numbers. C. R. Academy of Sciences URSS no. 30,
301.
Konrad J. H. (1976): Ph. D. Thesis, California Institute of Technology, California, USA
(according to Van Duke 1982).
Kovasznay L. S. G. (1950): Q Program Report Aerospace Department Contract NORD-
8036-JHB-39. The John Hopkins University, USA.
Kovasznay L. S. G. (1954): Hot-Wire Method. In: Physical Measurements in Gasdyna-
mics and Combustion" (ed. R. W. Landenburg) vol. 9, Section F2, p. 219-285. Princeton
University Press.
Kramers H. (1946): Heat Transfer from Spheres to Flowing Media. Physica no. 12, p.
61-80.
Kuroda A., Kasagi N. and Hirata M. (1989): Proc. of International Symposium on
Computational Fluid Dynamics, Nagoya, Japan, p. 1174-1179.
Kuroda A., Kasagi N. and Hirata M. (1993): Proc. of the 9-th Symposium on Turbulent
Shear flows, Kyoto, Japan, p. 8.4.1-6.
Lakshminarayana B. (1982): Three Sensor Hot wire/Film Technique for Three-
Dimensional Mean and Turbulence Flow Field. TSI Q. 8, p. 3-13.
Lang D. B. (1985): Laser-Doppler Velocity and Vorticity Measurements in Turbulent
Shear Layers. Ph. D. Dissertation, California Institute of Technology, California.
Larsen S. E. and Busch N. E. (1974): Hot-Wire Measurements in the Atmosphere. Part I:
Calibration and Response Characteristics. DISA Info 16, p. 1534.
Larsen S. E. and Busch N. E. (1976): Hot-Wire Measurements in the Atmosphere. Part
II: A Field Experiments in the Surface Boundary Layer. DISA Info 20, p. 521.
Larsen S. E. and Busch N. E. (1980): On the Humidity Sensitivity of Hot-Wire
Measurements. DISA Info 25, p. 4-5.
Laufer J. (1975): New Trends in Experimental Turbulence Research. Annual Rev. of
Fluid Mech., no. 7, p. 307-326.
Lekakis I. C. (1988): Coherent Structures in Fully Developed Turbulent Pipe Flow.
Ph.D. Thesis, University of Illinois at Urbana-Champaign.
188 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Lekakis I. (1996): Calibration and Signal Interpretation for Single and Multiple Hot-Wire
and Hot-Film Probes. Meas. Sci. Tech. 7, p. 1313-1333.
Lekakis I. C., Adrian R. J. and Jones B. G. (1989): Measurements of Velocity Vectors
with Orthog. and Non-Orthogonal Triple-Sensor Probes. Exp. in Fluids 7, p. 228-232.
Lemonis G. and Dracos T. (1995): A New Calibration and Data Reduction Method for
Turbulence Measurements by Multihotwire Probes. Exp. in Fluids 18, p. 319-328.
Lighthill M. J. (1963): In: Laminar Boundary Layers, (ed. Rosenhead L.), chapter II,
Oxford University Press, Oxford.
Ligrani P. M. and Bradshaw P. (1987): Spatial Resolution and Measurement of
Turbulence in the Viscous Sublayer, Using Subminiature Hot-Wire Probes. Exp. in
Fluids 5, p. 407-417.
Lomas C. G. (1986): Fundamentals of Hot-wire Anemometry. Cambridge University
Press, Cambridge.
Matovi M. (1988): Experimental Investigation of Free Premixed Flame Flow Field, by
Laser Anemometer (in Serbian). M.Sc. Thesis, Faculty of Mechanical Engineering,
University of Belgrade.
Matovi M., Oka S. and Durst F. (1994): Structure of the Velocity Temperature Field in
Premixed Axisymetric Acetilen Flame. J. of Fluids Eng., vol. 116, no. 3, p. 631-642.
Marasli B., Nguyen P. and Wallace J. M. (1993): A Calibration Technique for Multiple-
Sensor Hot-Wire Probes and its Application to Vorticity Measurements in the Wake of
a Circular Cylinder. Exp. in Fluids 15, p. 209-218.
Mathioudakis H. and Breugelmans F. (1984): Use of Triple Hot-Wires to Measure
Unsteady Flows with Large Direction Changes. J. of Physics E: Sci. Instrum. 18, p.
410-414.
Menevay C., Screenivasan K. R., Kailasnath P. and Fan M. S. (1990): Joint
Multifractal Measures: Theory and Applications to Turbulence. Phys. Rev. A41, p.
894-912
Merati P. and Adrian R. J. (1984): Directional Sensitivity of Single and Multiple Sensor
Probes. TSI Quarterly X, p. 3-12.
Miller I. S., Shah D. A. and Antonia R. A. (1987): A Constant Temperature Hot-Wire
Anemometer. J. of Physics E: Sci. Instr. vol. 20, p. 311-340.
Moin P. and Spalart P. R. (1987): Contributions of Numerical Simulation Data-Base to
Physics, Modelling, and Measurement of turbulence. NASA Tech. Memo. 1000022.
Moffat R. J., Yavuzkurt S. and Crawford M. E. (1978): Real Time Measurement of
Turbulent Quantities with Triple Hot-Wire Systems. Proc. of Dynamic Flow
Conference on Dynamic Measurements in Unsteady Flows, p.1013-1018, Marseilles
and Baltimore/Skovlunde.
Muller U. R. (1987): Developments in Measuring Reynolds Stresses. In: Perspectives in
Turbulence Studies (eds. Meier H. U. and Bradshaw P.), Springer-Verlag, Berlin.
Muller U. R. (1992): Comparison of Turbulence Measurements with Single X and Triple
Hot-Wire Probes. Exp. in Fluids 13, p. 208-216.
Nakayama A. and Westphal R. V. (1986): The Effects of Sensor Length and Spacing on
X-Wire Measurements in a Boundary Layer. NASA Tech.. Memo. 88352.
Nguyen P. N. (1993): Simultaneous Measurements of the Velocity and Vorticity Vector
Fields in the Turbulent Near Wake of a Circular Cylinder. Ph.D. Thesis, University of
Maryland, Maryland.
Nikuradse, J. (1929): Kinematographische Aufnahme einer turbulenten Stromung.
ZAMM, p. 495-496.
MULTIPLE HOT-WIRE PROBES 189
__________________________________________________________________________________________________________________________

Nithianandan K. C., Jones G. B. and Adrian J. R. (1987): Turbulent Velocity-Pressure
Field Structure in an Axisymmetric Mixing Layer. In: Nuclear Thermal-Hydraulics
Report 1, Univ. of Illinois at Urbana Chapaign.
Nitsche W. und Haberland C. (1984): Ein vereinfachtes Eichverfahren fur
Hitzdrahtanemometer. Z. Flugwiss. Welltraumforsch. 8, p. 264-270.
Nychas S. G., Kastrinakis E. G. and Eckelmann H. (1985): On Certain Aspects of
Vorticity Dynamics and Turbulent Energy Production. Lecture Notes Phys. (ed. GEA
Meier, F. Obermeier) 235, p. 269-278. Berlin/Heidelberg/New Yourk: Springer-Verlag.
Oka N. S. and Baki V. V. (1996): Investigation of Turbulent Field in Premixed
Acetylene Flame Using Conditional and Unconditional LDA Statistics. Proc. of the 3-
rd Int. Symp. on Engineering Turb. Modelling and Measurements, Crete, p. 807-817.
Oka S. and Kosti . (1972): Influence of wall proximity on hot-wire velocity
measurements. DISA Inf. 13, 1972., p. 29 - 33.
Ong L. (1992): Visualisation of Turbulent Flows with Simultaneous Velocity and
Vorticity Measurements. Ph. D. Thesis, University of Maryland, Maryland.
Ong L. and Wallace J. M. (1996): The velocity Field of the Turbulent Very Near Wake
of a Circular Cylinder. Exp. in Fluids 20, p. 441-453.
Ong L. and Wallace J. M. (1998): Joint Probability Density Analysis of the Structure and
Dynamics of the Vorticity Field of the Turbulent Boundary Layer. J. of Fluid Mech.,
vol. 367, p. 291-328.
Ossofsky E. (1948): Constant Temperature Operation of the Hot-Wire Anemometer at
High Frequency. Rew. Sci. Instr. 19, p. 881-889.
Panton R. L. (1984): Vorticity Dynamics. In: Incompressible Flow, p. 322-348. John
Wiley and Sons, New York.
Pao Y. H. (1965): Structure of Turbulent Velocity and Scalar Fields at Large Wave
Numbers. Physics of Fluids 8, p. 1063-1075.
Panchapaesan N. R. and Lumley J. L. (1993): Turbulence Measurements in
Axisymmetric Jets of Air and Helium. Part 1, Air Jet. J. Fluid Mech., 246, p. 197-223.
Park S.-R. and Wallace J. M. (1993): The Influence of Velocity Gradients on
Turbulence Properties Measured with Multi-Sensor Hot-Wire Probes. Exp. in Fluids
16, p. 17-26.
Perry A. E. (1982): Hot-Wire Anemometry. Claredon Press, Oxford.
Perry A. E. and Morrison G. L. (1971): A Study of the Constant-Temperature Hot-Wire
Anemometer. J. Fluid Mech. vol. 47, p. 577-599.
Petrovi V. D. (1991): Investigation of Turbulent Fluid Flow in the Free Round
Isothermal Jet by Hot-Wire Anemometer. M. Sc. Thesis, Faculty of Mechanical
Engineering, University of Belgrade.
Petrovi V. D. (1996a): Influence of the Hot Wire Anemometer Probe Design and
Acquisition Parameters on Measurement Results of Turbulent Velocity Field. Ph.D.
Thesis, Univ. of Montenegro in Podgorica, Department of Mechanical Engineering.
Petrovi V. D. (1996b): Optimal Acquisition Parameters of Hot Wire Probes Signals in
the Isothermal Turbulent Velocity Field. Proc. of The Second International
Symposium "Contemporary Problems of Fluid Mechanics", p. 109-112, Belgrade.
Petrovi D. and Beniek M. (1992): Turbulence Structure in the Isothermal
Axisymmetric Free Jet Near-Field. In monograph: "Spatio-Temporal Structure and
Chaos in Heat and Mass Transfer Processes" (eds. L. Pismen and M. Todorovic), p. 41-
53, Mrlje{ and Sons (selected lectures from the ICHMT International Symposium
held in Athena).
190 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Petrovi D., Jovanovi J. and Oka S. (1992): Higher-Order Correlations in the Near
Field of the Turbulent Isothermal Axisymmetric Free Jet". In: "Spatio-Temporal
Structure and Chaos in Heat and Mass Transfer Processes" (eds. L. Pismen and M.
Todorovic), p. 55-61, Mrlje{ and Sons (selected lectures from the ICHMT
International Symposium held in Athena).
Petrovi V. D., Beniek M. and Oka N. S. (1997): Statistical Relationships in a
Turbulent Free Round Jet. In: Turbulence, Heat and mass Transfer 2: Proc. of the 2
nd

Int. Symp. on Turbul., Heat and Mass Transfer, Delft, (Monograph of selected papers,
eds. K. Hanjalic and T. W. J. Peeters), p. 173 178, Delft University Press, Delft.
Petrovi V. D, Schenck C. T. Durst F. (1998): Velocity Correlations in the Wake Behind
a Cylinder. The Scientific Journal Facta Universitatis, series Mechanics, Robotics
and Automatic Control vol. 2, no. 8, pp. 655-668.
Petrovi V. D., Schenck C. T. and Vukoslavevi V. P (1997): An additional Look in
Turbulence Statistics. Thermal Science, vol. 1 no.2, p. 27-42.
Petrovi V. D. and Vukoslavevi V. P. (1995): Experimental Test of 4 Hot-Wire
Anemometer Probes Applicability for Incompressible Fluid Turbulent Velocity Field
Measurement. KGH (HVAC) no. 2, p. 43-48, Belgrade.
Petrovi V. D. and Vukoslavevi V. P. (1997): Interpretation of Hot Wire Probes
Signals in Slow Turbulent Flows. Proc. (Section B Fluid Mechanics) of the 22-th YU
Congress of Theoretical and Applied Mechanics, p. 118-123, V. Banja.
Phailas G. and Cousteix J. (1986): Methode d'exploitation des donnes d'une sonde
anemometrique a quartre fils chauds. La Recherche Aerospatiale 2, p. 161-168.
Pitts W. M.. and McCaffrey B. J. (1986): Response Behaviour of Hot-Wires and Films
to Flows of Different Gases. J. Fluid Mech. 169, p. 465-512.
Planchon H. P. (1974): The Fluctuating Static Pressure Field in a Round Jet Turbulent
Mixing Layer. Ph. D. thesis, University at Urbana Champaign.
Piomelli U., Balint J.-L. and Wallace J. M. (1989): On the Validity of Taylors
Hypothesis for Wall-Bounded Turb. Flows. Phys. Fluids A1, p. 609-611.
Piomelli U., Ong L., Wallace J. M. and Ladhari F. (1993): Reynolds Stress and
Vorticity in Turbulent Wall Flows. Appl. Sci. Res. no. 51, p. 365-370.
Pompeo R. and Thomann H. (1993): Quadruple Hot-Wire probes in a Simulated Wall
Flow. Exp. in Fluids 14, p. 145-153.
Rajagopalan S. and Antonia R. A. (1993a): RMS Spanwise Vorticity Measurements in a
Turbulent Boundary Layer. Exp. in Fluids no. 14, p. 142-144.
Rajagopalan S. and Antonia R. A. (1993b): Structure of the Velocity Field Associated
with the Spanwise Vorticity in Wall region of a Turbulent Boundary Layer Phys.
Fluids A, no. 15, p. 2502-2510.
Roberts J. B. (1973): On the Correction of Hot-Wire Turbulence Measurements for
Spatial Resolution Errors. Aeronautical. J. 77, p. 406-412.
Rogers M. M. and Moser R. D. (1994): Direct Simulation of a Self-Simmilar Turbulent
Mixing Layer. Phys. Fluids A, no. 6, p. 902-923.
Rosemann H. (1989): Einflu| der Geometrie von Mehrfach-Hitzdrahtsonden auf die
Me|ergebnisse in Turbulenten Strmungen. Deutsche Forschungs-anstalt fur luft und
raumfahrt, Forschungsbericht DLR-FB 89-26, Gottingen.
Rosemann H., Stager R. and Kreplin H-P. (1996): Development and Application of a
Quadruple Hot-Wire Technique for Turb. Flows. Meas. Sci. Technol. 7, p. 1477-1491.
Saddoughi S. G. and Veeravalli S. V. (1994): Local Isotropy in Turbulent Boundary
Layers at High Reynolds Number. J. Fluid Mech. vol. 268, p. 333-372.
MULTIPLE HOT-WIRE PROBES 191
__________________________________________________________________________________________________________________________

Saddoughi S. G. and Veeravalli S. V. (1996): Hot-Wire Anemometry Behaviour at Very
High Frequencies. Meas Sci. Technol. 7, p. 1297-1300.
Samet M. and Einav S. (1987): A Hot-Wire Technique for Simultaneous Measuring of
Instantaneous Velocities in 3-D Flows. J. Physics E: Sci. Instrum. 20, p. 683-690.
Sandborn V. A. (1972): Resistance Temperature Transducers. Metrology Press, Ft. Collins
CO.
Schenck C. T., Petrovi V. D., Jovanovi J and Durst F. (1998a): Turbulent Dissipation
Measurements in a Plane Wake Flow. Proc. of ASME FEDSM98, p. 4976/1-6,
Washington DC.
Schenck C. T., Petrovi V. D., Jovanovi J. and Durst F. (1998b): Measurement of
Dissipation Correlations in a Turbulent Plane Wake. Proc. of ISFMTM98, National
Cheng Kung University of Taiwan, Tainan, Taiwan R.O.C., p. 507-514.
Schubauer G. B. (1935): Effect of Humidity in Hot-Wire Anemometry. J. Res. Nat. Bur.
Stand. 15, p. 575-578.
Silverman B. A. (1968): The Effect of Spatial Averaging on Spectrum Estimation J.
Appl. Metheorology 7, p. 168-172.
Sirivat A. (1989): Measurement and Interpretation of Space-Time Correlation Functions
and derivative Statistics from a Rotating Hot-Wire in a Grid Turbulence. Exp. in Fluids
7, p. 361-370.
Smith R. W. (1994): Ph.D. Thesis, Princeton University, Princeton, New Yourk, USA.
Spalart P. R. (1988): Direct Simulation of a Turbulent Boundary Layer up to R
u
= 1410.
J. Fluid Mech. vol. 187, p. 61-98.
Spalding D.B. (1961): A Single Formula for the Law of the Wall. Transactions of ASME
E: J. of Appl. Mech. 28, p. 455-457.
Spencer B. W. (1970): Statistical Investigation of Turbulent Velocity and Pressure Fields
in a Two-Stream Layer. Ph. D. thesis, Nuclear Engineering Program, University of
Illinois at Urbana-Champaign, USA.
Steinhilber R. and Wagner P. M. (1994): A Digital Processor for Anemometer
Controll. Exp. in Fluids 17, p. 302-306.
Strohl A. and Comte-Bellot G. (1973): Aerodynamic Effects due to Configuration of X-
Wire Anemometers. J. of Appl. Mech. 40, p. 661-666.
Suzuki Y. and Kasagi N. (1992): Evaluation of Hot-Wire Measurements in Wall Shear
Turbulence Using a Direct Numerical Simulation Database. Exp. Therm. Fluid. Sci. 5,
p. 66-77.
Tagawa M., Tsuji T., Nagano Y. (1992): Evaluation of X-Probe Response to Wire
Separation for Wall Turbulence measurements. Exp. in Fluids no. 12, p. 413-421.
Taylor G. I. (1935): Statistical Theory of Turbulence. Parts 1-4. Proc. R. Soc. London A
151, p. 421-478.
Taylor G. I. (1938): Production and Dissipation of Vorticity in a Turbulent Fluid. Proc.
R. Soc. London Ser. A 164, p. 15-23.
TSI (1997): Innovation in Thermal Technology (A Complete Family of Thermal
Anemometry Systems, Probes, and Accessories). TSI Incorporated Fluid Mechanics
Instrument Division, St. Paul, USA.
Tennekes H. and Lumley L. J. (1978): A first course in turbulence. The MIT Press,
Cambridge, Massachusetts and London, England.
Thomson B. E. (1984): The Turbulent Separating Boundary Layer and Downstream
Wake. Ph. D. Thesis, Department of Mechanical Engineering, Imperial College,
London.
192 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Thomson B. E. (1987): Appraisal of a Flying Hot-Wire Anemometer, DANTEC Info.
No. 4, p. 14-18.
Thomson B. E. and Whitelaw J. H. (1984): Flying Hot-Wire Anemometry, Exp. in
Fluids, p. 47-55.
Tsinober A., Kit E. and Dracos T. (1992): Experimental Investigation of Field of
Velocity Gradients in Turbulent Flows. J. Fluid Mech. 242, p. 169-192.
Turan O. and Azad R. S. (1989): Effect of Hot-Wire Probe Defects on a New Method of
Evaluating Turbulence Dissipation. J. of Phys. E: Sci. Instr. 24, p. 1278-1280.
Tutu N. K. and Chevray R. (1975): Cross-Wire Anemometer in High Intensity
Turbulence. J. of Fluid Mech. 71, p. 785-800.
Uberoi M. S. and Corrsin S. (1951): Prog. Report for Contract NAW 5504 for NACA.
The John Hopkins University, USA.
Uberoi M. S. and Kovasznay L. S. G. (1953): On Mapping and Measurement of Random
Fields. Quart. Appl. Math. 10, p. 375-393.
Vagt J. D. (1979): Hot-Wire Probes in Low Speed Flow. Programm Aerospace Science
no. 18, p. 271-323.
Van Duke M. (1982): An Album of Fluid motion. Department of Mechanical
Engineering, Stanford University: The Parabolic Press, Stanford, California, USA.
Vukoslavevi V. P. (1981): Development of the Hot-Wire Probes with X-Sensors and
Corresponding Methods of Measuring the Longitudinal Vorticity and Velocity
Components in a Turbulent Fluid Flow. Ph.D. Thesis, Department of Mechanical
Engineering, University "Veljko Vlahovi", Titograd.
Vukoslavevi V. P. (1993): Influence of the 4.-th Sensor on Precision of Flow Field
Parameters Measurements. Proc. of the 17
-th
Yugoslav Congress of Theoretical and
Applied Mechanics, p. 71-78, Kragujevac.
Vukoslavevi P. and Petrovi V. D. (1994): Generalized Cooling Law of Small
Multisensor Anemometer Hot-Wire Probes. Proc. of The 25
-th
YU Congress of HVAC,
p. 299-307, Belgrade.
Vukoslavevi V. P., Petrovi V. D. (1996): Eight Hot Wire Probe. Proc. of the Second
Int. Symposium "Contemporary Problems of Fluid Mechanics", p. 105-108, Belgrade.
Vukoslavevi V. P. and Petrovi V. D (1997): Influence of Hot-wire Probe
Configuration on the Results of 3-d Fluid Velocity Field Measurements. In:
Turbulence, Heat and Mass Transfer 2: Proc. of the 2
nd
International Symposium on
Turbulence, Heat and Mass Transfer Delft, The Netherlands 9-12, 1997 (Monograph of
selected papers, eds. K. Hanjalic and T. W. J. Peeters), p. 189 199, The University
Press, Delft.
Vukoslavevi V. P. and Petrovi V. D (1998): The Probability Density Distribution of
the Vorticity Vector Field in the Near Wall Region of Turbulent Boundary Layer.
Proc. of the 5
-th
National Congress on Mechanics, Ioannina, p. 617-624.
Vukoslavevi V. P. and Wallace J. M. (1981): Influence of Velocity Gradients on
Measurements of Velocity and Streamwise Vorticity with Hot-Wire X-Array Probes.
Rev. Sci. Instr. 52(6), p. 869-879.
Vukoslavevi V. P. and Wallace J. M. (1983): The Range of Uniqueness of the
Solution Obtained Using 3 or 9 Sensor Probe. Gessellshaft fur Angewandte
Mathematik und Mechanik (GAMM Tagung, Kurzfassungen der vortage Abstacts).
Vukoslavevi V. P. and Wallace J. M. (1984): A multi sensor probe for measuring the
velocity and vorticity components. Proc. of the YU Congress of Theoretical and
Applied Mechanics, Beii.
MULTIPLE HOT-WIRE PROBES 193
__________________________________________________________________________________________________________________________

Vukoslavevi P., Balint J.-L. and Wallace M. J. (1989): A Multisensor Hot Wires
Probe to Measure Vorticity and Velocity in Turbulent Flows. Trans. of ASME: J. of
Fluids Eng., 111, p. 220-224.
Vukoslavevi P. and Wallace J. M. (1996): A 12-sensor Hot-Wire Probe to Measure the
Velocity and Vorticity Vectors in Turbulent Flow. Meas. Sci. Tech. 7, p. 1451-1461.
Vukoslavevi V. P., Wallace J. M. and Balint J.-L. (1990): The Simultaneous
Measurement of the Velocity and Vorticity Vectors in a Turbulent Flow Using Hot-
Wire Anemometry. Rep. TRL-90-1, University of Maryland, Maryland, USA.
Vukoslavevi V. P., Wallace J. M. and Balint J.-L. (1991): The Velocity and Vorticity
Vector Fields of a Turbulent Boundary Layer. Part 1. Simultaneous Measur. by Hot
Wire Anemometry". J. Fluid Mech., 228, p. 25-51.
Wadcock A. (1978): Flying Hot-Wire Study of Two-Dimensional Turbulent Separation
on an NACA 4412 Airfoil at Maximum Lift. Ph. D. Thesis, California Institute of
Technology, USA.
Walker M. D. and Maxey M. R. (1985): A Whirling Hot-Wire Anemometer with Optical
Data Transmission. J. Physics E.: Sci. Instr., 18, p 516-521.
Wallace J. M. (1986): Methods of Measuring Vortivity in Turbulent Flows. Exp. in
Fluids no. 4, p 61-71.
Wallace J. M., Ballint J.-L. and Ong L. (1992): An Experimental Study of Helicity
Density in Turbulent Flows. Phys. Fluids A 4, p.2013-2026.
Wallace J. M., Brodkey R. S. and Eckelmann H. (1977): Pattern Recognition in
Bounded Turbulent Shear flows.J. Fluid Mech., vol. 83, p. 673-693.
Wallace J. M., Foss J. F. (1995): The Measurement of Vorticity in Turbulent Flows.
Annual Rev. of Fluid Mech. 27, p. 469-514.
Wallace J. M. and Vukoslavevi P. (1982): Measurement of the Structure of the
Streamwise Vorticity Field in a Turbulent Boundary Layer. In: Structure of
Turbulence in Heat and Mass Transfer (ed. Zari} Z. P.), p. 29-41. Washington, D.C.:
Hemisphere.
Wasmann W. W. and Wallace J. M. (1979): Measurement of Vorticity in Turbulent
Flows. Exp. in Fluids 4, p. 61-70.
Wasmann W. W. and Wallace J. M. (1980): A Multisensor Hot Wires Probe for
Measuring the Velocity and Vorticity Components. EUROMECH 132, Lyon.
Weber D. P (1974): Turbulent Velocity Field Structure in Round Jet and its Relation to
Fluctuating Pressure. Ph.D. thesis, University of Illinois at Urbana Champaign.
Webster C. A. G. (1962): Note on Sensitivity to Jaw of Hot-Wire Anemometer. J. Fluid
Mech., vol. 13, p. 307-312.
Wei T. (1987): Reynolds Number Effects on the Small Scale Structure of a Turbulent
Channel Flow. Ph.D. Thesis. The University of Michigan.
Wei T. and Willmarth W. W (1989): Reynolds Number Effects on the Structure of
Turbulent Channel Flow. J. of Fluid Mech. 204, p. 57-95.
Weidmann P. D. and Browand F. K. (1975): Analysis of a Simple Circuit for Constant
Temperature Anemometry. J. of Physics E: Sci. Instr., vol. 8, p. 553-560.
Werle H. (1980): Rech. Aerosp. 1980-5, p. 35-49 (acc. to Van Duke 1982).
Weske J. R. (1943): A Hot-Wire Circuit with Very Small Time Lag. NACA Technical
Note No. 881.
Winant D. C. and Browand K. F. (1974): Vortex Pairing: The Mechanism of Turbulent
Mixing Layer Growth at Moderate Reynolds Number. J. Fluid Mech., vol. 63, part 2,
p. 237-255.
194 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________
Willmarth W. W. (1975): In: Advances in Applied Fluid Mechanics (ed. Yih C. S.), vol.
15, p. 159-198, Academic Press, New Yourk.
Willmarth W. W. (1985): Geometric Interpretation of the Possible Velocity Vectors
Obtained with Multiple Sensor Probe. J. of Physics of Fluids 28, p. 462-465.
Willmarth W. W. and Bogar T. J. (1977): Survey and New Measurements of Turbulent
Structure Near the Wall. Phys. Fluids 20, p. 9-21.
Willmarth W. W. and Lu S. S. (1972): Structure of the Reynolds Stress near the Wall.
J. Fluid Mech., vol. 55, p. 65-92.
Wittmer K. S., Devenport W. J. and Zsoldos J. S. (1998): A Four Hot-Wire Probe
System for Three-Component Velocity Measurement. Exp. in Fluids 24, p. 416-423.
Wygnaski I. and Fiedler H. (1969): Some Measurements in Self-Preserving Jet. J. of
Fluid Mech. 38, p. 577-612.
Wyngaard J. C. (1968): Measurement of Small-Scale Turbulence Structure with Hot-
Wires. J. of Physics E: J. of Sci. Instr., series 2, vol.1, p. 1105-1108.
Wyngaard J. C. (1969): Spatial Resolution of the Vorticity Meter and Other Hot-Wire
Arrays. J. of Physics E: J. of Sci. Instr., series 2, vol.2, p. 983-987.
Wyngaard J. C. (1971): Spatial Resolution of a Resistance Wire Temperature sensor.
Phys. Fluids 14, p. 2052-2054.
Wyngaard J. C. and Lumley (1967: A Constant Temperature Hot-Wire Anemometer. J.
Sci. Instrum. vol. 44, p. 363-365.
Ye Q.-Y. (1996): Die turbulente Dissipation mechanischer Energie in Scherschicthen,
Ph.D. Thesis, LSTM, University of Erlangen-Nurnberg.
Zagarola M. and Smits A. (1998): Proc. ASME FEDSM98, Washington DC, p. 4950.1-6.
Zari Z. (1974): Etude Statistique Dela Turbulence Parietale. Ph. D. Thesis, University
of Paris (in French).
Ziegler M. (1934): Verh. K. Neth. Akad. Wet. Amsterdam 15 (3), p. 1-22 (according to
Corrsin 1963 and Blackwelder 1981).


Computer-controlled anemometer of DANTEC with a calibration facility.
Source: Jorgensen 1996.
MULTIPLE HOT-WIRE PROBES 195
__________________________________________________________________________________________________________________________



All objects are characterised by their own frequencies and vibrations.

(Nikola Tesla, 10 July 1856 7 January 1943)
196 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________

MULTIPLE HOT-WIRE PROBES 197
__________________________________________________________________________________________________________________________


198 P. V. Vukoslavevi and D. V. Petrovi
__________________________________________________________________________________________________________________________

You might also like