Você está na página 1de 17

1

INSTABILITY AND UNSTEADINESS OF AIRCRAFT WAKE VORTICES


L. Jacquin
(1)
, D. Fabre
(1)
, D. Sipp
(1)
, V. Theofilis
(2)
& H. Vollmers
(2)
(1)
Dep
t
of Fundamental and Experimental Aerodynamics
ONERA, 8 rue des Vertugadins, 92190 Meudon - France
(2)
Institute of Aerodynamics and Flow Technology
DLR, Bunsenstrasse 10, D-37073 Goettingen - Germany
SUMMARY
This paper presents a review of theoretical and experimental results on stability and other unsteady properties of
aircraft wakes. Theory mainly concern the dynamics of two-dimensional vortices. The basic mechanisms
responsible for the propagation and the amplification of perturbation along vortices, namely the Kelvin waves
and the co-operative instabilities, are first detailed. These two generic unsteady mechanisms are described by
considering asymptotic linear stability analysis of model flows such as vortex filaments or Lamb-Oseen vortices.
Extension of the linear analysis to more representative flows, using a global stability approach, is also described.
Experimental results obtained using LDV, hot wire and PIV in wind tunnels are presented and they are
commented in the light of theory .
Key words : Wake Vortex - Stability theory Nonparallel global stability theory Unsteady flow
Turbulence Laser velocimetry - Hot wire velocimetry Particle Image Velocimetry .
1. INTRODUCTION
Two strategies may be considered to produce less
harmful wakes behind aircraft, see e.g. Gerz et al.
(2001), Jacquin et al. (2001). The first is to increase
the characteristic radii of the final vortices so as to
decrease the rolling momentum of a following
aircraft during encounters. This may be achieved by
introducing "as much turbulence as possible" into
the vortex system. The second strategy is to promote
the co-operative instabilities which develop in a
system of several (at least two) vortices and which
leads to destructive interactions between the two
halves of the wake.
Theoretical analysis and specific measurements
are both needed to investigate this topic. This paper
aims at presenting the status of our understanding of
unsteadiness in wake vortices which has been gained
both from theoretical analysis of model vortex flows
and from experimental investigations of
representative trailing wakes.
As it will be shown below, useful theoretical
results are provided by linear stability analysis of
generic flows. This enable the introduction of
important physical mechanisms which lead to
unsteadiness. As for the experiments, wind tunnel
tests give access to measurements within
downstream distances limited to typically ten model
spans. In this regime, the wake unsteadiness
amounts to small amplitude displacements of the
vortex cores. Due to the presence of very sharp
velocity gradients within the cores, this
"meandering" of the vortices leads to energetic
velocity temporal fluctuations which can be
measured both by LDV and hot wire techniques.
PIV-measurements also show the fluctuation of tip
and flap tip vortices. Examples will be described in
section 4. These wind tunnel tests can be prolonged
by using a catapult facility, as described in Stuff et
al. (2001).
The paper is organized as follows. Average
properties of a vortex wake are described in section
2. In section 3, basic linear mechanisms which
participate to vortex wake unsteadiness and
instability are reviewed. This comprises the Kelvin
waves and the co-operative instabilities that
develop on long- or short-wavelengths within
multi-polar arrangements of vortices. In section 4,
some experimental observations are presented and
are discussed in the light of theory. Open questions
and perspectives are listed in Section 5.
2. THE MEAN FIELD AND LENGTH-
SCALES
The trailing wake produced by an aircraft amounts
to an antisymmetric distribution of vorticity of
circulation :



0
dz dy (1)
2
where denotes the x-component of the vorticity.
Convention are those of figure 1. Let consider a
wing with lift coefficient
L
C , span b, airfoil surface
S and aspect ratio S b AR
2
which is placed in a
flow with velocity

V . From conservation of
vertical momentum, the lift, S V
L
2
C 2 1

, is equal
to the flux of the wake vertical momentum b
~
V

where :

dzdy y b
~ 1
(2)
is the separation between the vorticity centroids (see
e.g. Saffman, 1992). It follows that b
~
V

Lift .
In the case of an elliptic loading, ( )b b
~
4 , the
circulation is therefore given by AR b V C
L


2 .
For a landing transport aircraft one typically has :
2
L
C , 7 AR .
Figure 1 -Sketch of far-field flow downstream of a wing
The half-wake wake circulation can only
decrease through diffusion and cancellation of
circulation through the plane that separates the two
vortices. In an aircraft vortex wake, this is a very
slow process due to the high value of the Reynolds
number. A decrease of is compensated by an
increase of b
~
in order to maintain the wake vertical
impulse b
~
V

.
In order to characterize the internal structure of the
vortices, several vortex core length-scales may be
defined (see Jacquin et al., 2001). In particular, an
aircraft vortex may be assumed to consist of an
"internal core" (or "viscous core") of radius
1
r ,
rotating as a solid body, and of an "external core" (or
"inviscid core") of radius
2
r , which characterizes
the region containing vorticity surrounding the
internal core (see Spalart, 1998). The internal core
radius
1
r is usually defined as the position of
maximum azimuthal velocity and the external core
radius
2
r is the position where the total circulation
of the vortex is almost attained.
The need of such a two-scale model to describe the
internal structure of a wake vortex has been
illustrated by LDV measurements performed in the
wake of a scale 1:100 scale generic A300 model, see
Jacquin et al. (2001). In this experiment, the model
had a wing span mm b 448 and it was set in two
configurations, a cruise configuration (noted "clean
case") and a high lift configuration (noted "high
lift"), with wings fitted with single block flaps. The
lift coefficients were respectively 7 0. C
L
and
7 1. . Tests have been performed for a free-stream
velocity
1
50

ms V . The Reynolds number


based on the aerodynamic chord mm c 66 was
000 220
c
Re . The Reynolds number based on the
circulation was 000 230

Re in the high
lift case ( s m .
2
46 3 ) and 000 95

Re in the
clean case ( s m .
2
42 1 ). Values of are
obtained using the elliptic expression given above.
Figures 2 and 3 show the tangential velocity

V
after cylindrical averaging of the LDV mean values
obtained in three vertical sections, b x 3,5,9,
behind the model in the high lift and clean cases
respectively. In both cases, the velocity
distributions remain almost unchanged between
3 b x and 9 ; difference in the maximum of

V
at 9 b x results from change in the probing mesh
size.
In the high lift case, see figure 2, the tangential
velocity is characterized by the presence of a
plateau where

V remains around its maximum


value, that is approximately 23% of

V . Such a
velocity plateau was also observed by Devenport et
al. (1996) ; as suggested by these authors, this could
correspond to a remnant of the initial conditions
due to a merging of several cores. In the high lift
case, the merger between the wing tip and the flap
tip vortices occurs around 2 b x . After this
plateau region, the decrease of

V seems to follow
a power law

r with around 0.5 (the exact
value of the slope is indicated in figure 2(b)). This
is in accordance both with the classical model
proposed by Betz (1932), which leads to 5 0.
for a wing with an elliptic load, and with the self-
similar solution of the roll-up given by Kaden (see
Saffman, 1992). Departure from these laws occurs
around 1 0
2
, b r .
In the clean case, see figure 3, no plateau is
observed. A

r law with slightly larger than
0.5, is particularly well adapted here. Vortices are
less intense (the maximum of

V is worth
approximately 15% of

V ) and more concentrated.


A two scale model of the type :
Ar V : r r
1
, Br V : r r r
2 1
2 1

(3)
1
2

Cr V : r r
x
y
z
b
b
~
1
2
3
has been proposed to model the flow, as sketched in
figure 4 . The values of the constants are :
( )
2 1 1
2 r r r A , ( )
2
2 r B , 2 C .
Our results suggests : ( )
2
1
10

O b
~
r ,
( )
1
2
10

O b
~
r . A fourth region, corresponding to
the plateau observed in figure 2, could also be added
for the high lift case.
(a)
0 0.1 0.2 0.3 0.4 0.5
0.00
0.05
0.10
0.15
0.20
0.25
x/b=3
x/b=5
x/b=9
(b)
10
-2
10
-1
0.05
0.10
0.15
0.20
0.25
x/b=3
x/b=5
x/b=9
Figure 2 - Tangential velocity

V V versus the radius
centred on the vortex - high lift configuration : (a) lin-lin
plot, (b) log-log plot (from Jacquin et al., 2001).
In both the clean and high lift cases, the inner core
radius
1
r is very small and the mesh size used
( mm .5 2 ) is not sufficient to discriminate the inner
region
1
r r . As shown by Jacquin et al. (2001), in
this experiment
1
r is close to mm 1 , which means
01 0
1
. b r < .
Another useful lengthscale is the "dispersion
radius", defined as :
( ) { }


t
+

0
2 2
2 1
) z z ( ) y y ( dy dz r
c c
(4)
where
c
y

and
c
z are the coordinates of the vorticity
centroid in half a plane. This provides an evaluation
of the vorticity field dispersion in the plane. If the
evolution of the wake is purely two-dimensional, r
can only increase by viscous diffusion (see
Saffman, 1992). Other definitions for vortex core
length-scales are discussed in Jacquin et al. (2001).
(a)
0 0.1 0.2 0.3 0.4 0.5
0.00
0.05
0.10
0.15
0.20
x/b = 3
x/b = 5
x/b = 9
(b)
10
-2
10
-1
0.05
0.10
0.15
0.20
x/b = 3
x/b = 5
x/b = 9
Figure 3 - Tangential velocity

V V versus the radius
centred on the vortex clean configuration : (a) lin-lin
plot, (b) log-log plot (from Jacquin et al., 2001).
Figure 4 A vortex model.
The available experimental and numerical results
show a large discrepancy on the internal core radius
1
r . On the other hand, the external radius and the
dispersion radius are generally found to be of the
order ( ) b . O 1 0 .
b
r

r
1
r
b
~
. r 1 0
2
b
~
. r 01 0
1

V
V
r
b
r
b
r
b
r
47 0.
57 0.

V
V

V
V

V
V

V
V
4
3. STABILITY OF WAKE VORTICES
3.1. Long-wave instabilities
The wake issuing from a high lift designed wing is
usually compounded of several vortex filaments
whose evolution may be computed using a 3D
vortex method. In general, such a vortex
arrangement is unstable with respect to 3D
perturbations due to mutual straining of the vortices.
Following Crow (1970) and Crouch (1997), a
system of stability equations may be derived by
considering a set of parallel vortex filaments with
slight sinusoidal perturbations of their respective
positions. We suppose that the centerline position
( ) ) t ( Z ), t ( Y
n n
of the vortex filament labeled n is
displaced with an amplitude proportional to
ikx
e
where x is the coordinate in the axial direction :
( ) ( ) ( )
( ) ( )
z
ikx
n n
y
ikx
n n x n
e e t z ) t ( Z
e e t y ) t ( Y e x t , x X
+ +
+ +
(5)
At the leading order, the linearization of the Biot-
Savart law leads to the system :
2
2
2
2
mn
m n
n m
m n
mn
m n
n m
m n
R
Y Y
dt
dZ
R
Z Z
dt
dY

(6)
where ( ) ( )
2 2 2
n m n m mn
Z Z Y Y R + . This system
describes the evolution of an ensemble of N point
vortices. At the following order, the equations which
describes the evolution of the perturbation amplitude
vector ( ) ( )
N N
z , y ,..., z , y t X
1 1
may be put under
the following form :
( )
( ) ( ) t X t L
dt
t X d
(7)
The developed expressions of this linear system are
given in Crow (1970) for the case of a pair of
counter rotating vortex filaments, in Crouch (1997)
for the case of 22 co-rotating filaments and in
Fabre & Jacquin (2000) for 22 contra-rotating
filaments. The right hand side of the system (7)
amounts to a superposition of three effects : (i) the
straining experienced by filament n when displaced
from its mean position in the velocity field of the
unperturbed filaments m, (ii) the self induced
rotation of the filament n and (iii) the velocity field
induced on the filament n by the other vortices when
sinusoidally displaced from their mean positions.
Mechanism (i) may be easily understood when
considering the 2D flow corresponding to a pair of
straight counter-rotating vortex filaments separated
by a distance b
~
h , as in figure 1. In a co-ordinate
system moving downward with a speed
( ) b
~
dt dZ 2 , linearization of (6) around the
center of the vortex labeled 2, i.e. 1 << b
~
r with
( )
2
2
2
Y y r ( )
2
2
Z z + , leads to :
dy dt
dz dt
b
y
z

_
,

_
,

_
,

2
0 1
1 0
2

~
(8)
This velocity field is that of a strain with rate
( )
2
2 b
~
whose axes are oriented t45 . This
leads to amplification of any perturbation of the
vortex centerline, the latter being displaced away
when it leaves its initial position.
The self induced rotation terms, mechanism (ii),
reads :
( )
dy dt
dz dt
a
ka
y
z
n
n
n
n
n
n
n
$
$
$
$

_
,

_
,

_
,

2
0 1
1 0
2

(9)
which introduces a dependence of the solution with
respect to a scale
n
a characteristic of the vortex
thickness. This effect results from the velocity that
a curved vortex filament induces on itself. This self
induced rotation of the perturbed vortex is in the
opposite sense to that of the basic state, see (9).
Widnall et al. (1971) and More & Saffman (1972)
used asymptotic methods to derive the following
expression for the self induced function in (9) :
( )
( )

,
_

+
4
1 2
2
2
n
n
n
ka
ln
ka
ka , (10)
with 58 0. the Eulers constant. This expression
actually corresponds to the frequency of an helical
Kelvin wave, a notion that will be detailed in 3.3
below. Relation (10) is valid in the long-wave limit
1 << ka for a Rankine vortex of radius a
n
or, more
generally, for an arbitrary vortex if a
n
is defined as
the radius of an equivalent Rankine vortex. Let
recall that a Rankine vortex of radius a is defined
by an angular velocity such that
( ) ( )
2
2 a r a r V

, ( ) ( ) r a r V >

2 .
3.2. Example : stability of a four-vortex model
wake
Let consider the example of a four vortex
configuration sketched in figure 5 which comprises
two vortex pairs. Crouch (1997) considered the case
of co-rotating vortices in each half-plane
( 0 0
2 1
> > , ) and Fabre & Jacquin (2000), that
of contra-rotating vortices ( 0 0
2 1
< > , ). The
first case may model wing tip vortices and flap tip
vortices whereas the second one could model
5
merged wing and flap tip vortices and merged inner
flap and horizontal tail vortices. Donaldson &
Bilanin (1975) showed that the evolution of this
system is either divergent (the internal and external
pairs separate) or periodic (the vortices orbit around
the center of vorticity in each half-plane). As shown
by Rennich & Lele (1999), a stationary
configuration, such as Cte dt dZ , dt dY
n n
0 ,
see (6), is possible if :
Figure 5 - Four-vortex arrangement
0 3 3
1
2
1
2
2
1
2
1
2
3
1
2

+ +

,
_

,
_

b
b
b
b
b
b
(11)
A temporal analysis may be then conducted.
Solutions of (7) are seek under the form
( )
t
e X t X

0
. The results of the three dimensional
stability analysis by Fabre & Jacquin (2000) is
described in figure 6 which sketches the
amplification rate of the unstable eigenmodes for
the base flow corresponding to the set of parameters
4 0
1 2
. , 14 0
1 2
. b b , , . b a 1 0
1 1
, . b a 05 0
1 2

a choice suggested by Rennich & Lele (1999). Note
that, according to the above discussion, the method
requires the core radii of the vortices to be small
with respect to their separations, i.e. :
( ) 2
2 1 2 1 2 1
b b , b , b a , a << (12)
The choice 1 0
1 1
. b a is in agreement with the
statements of section 2 concerning core length-
scales ; the choice 05 0
1 2
. b a is a limiting value
with respect to the condition 1
2 2
<< b a .
It is seen in figure 6 that this counter-rotating
vortex arrangement is unstable with respect to two-
dimensional disturbances (k=0). The system may
develop symmetric and antisymmetric perturbations
with a growth rate which decreases as
1 2
b b
increases. The growth rate of the first branch labeled
1
S amounts to 9 for long wavelengths and it
reaches twice this value at wave numbers
corresponding to shorter waves ( ) 7
1
kb . The
second branch, labeled
2
S is limited to long
wavelengths ( ) 2 1
1
. kb < . It is close to the classical
Crow instability that would develop on the outer
vortices on the wavelength 8 0
1
. kb with a growth
rate ( )
2
1 1
2 8 0 b .
Crow
, if theses vortices were
isolated (see Crow, 1970). There is only one branch
for the antisymmetric modes. The shape of the
symmetric and antisymmetric modes corresponding
to 8 0
1
. kb are sketched in figure 7(a)-(b). The
most amplified mode is that sketched in figure 7(c).
This mode appears to be essentially a Crow
instability acting on the inner vortices. After a rapid
linkage of these vortices, this instability let the
outer vortices almost unchanged. A practical
conclusion of this work was that the naturally
emerging mode, obtained without forcing of a long
wavelength mode, will be a short wavelength
instability localized on the inner vortices that will
not affect the outer vortices.
The LDV measurements performed by Jacquin et
al. (2001) in the experiment described in section 2
show that such a four-vortex interaction is likely
participating to the global unsteadiness of the
vortices in the extended near-wake field. This was
shown by means of two-point LDV measurements.

Figure 6 - Growth rate of unstable eigenmodes as
function of the axial wave number for the set of base flow
parameters 4 0
1 2
. , 14 0
1 2
. b b , 1 0
1 1
. b a ,
05 0
1 2
. b a .
1
S ,
2
S : symmetric modes, A :
antisymmetric modes. Mode
2
S is the Crow instability
mode (from Fabre & Jacquin, 2000)
3.3. Kelvin waves
We now focus on unsteady mechanisms with
length-scales of the order of the vortex core or
smaller.
The first basic unsteady mechanism to be
accounted for is inertia waves. Any perturbation in
a rotating flow leads to the propagation of
dispersive waves, called inertia waves. These waves
are equivalent to gravity waves in a stable stratified
medium. The inertia waves propagating along a
vortex are named Kelvin waves. They are described
by solving an eigenvalue problem resulting from
the linearization of the Euler equations, given a
basic field representing the flow within the vortex.
0
5 10
10
0
18

1
2
1
2 b
1
kb
1
b
1

1
2

2

1
2a
2
b
2
2a
6
Figure 7 - Long-wave modes 8 0
1
. kb ,
( )
2
1 1
2 8 9 b . : (a) symmetric
1
S - mode , (b)
antisymmetric A - mode, (c) most amplified
1
S - mode :
( )
2
1 1 1
2 2 18 7 b . , kb . These modes are identified
by a circle in figure 6 (from Fabre & Jacquin, 2000).
The case of a basic flow corresponding to a Rankine
vortex has been extensively described in the
literature (see e.g; Saffman, 1992, Rossi , 2000). We
consider here the case of a Lamb-Oseen vortex, with
angular velocity :
( )
( )

,
_



2 2
1
2
2
a r
e
r
r
r V
r (13)
which has been described by Sipp et al. (1999), see
also Sipp (2000), Fabre et al. (2001). We linearize
the incompressible Euler equations around these
basic flow and introduce the following small
perturbations :
( ) ( )( )
( ) t m kx i
r

x
e r P , H , iG , F p , v , v , v
+

(14)
where

r

x
v , v , v

are the axial, radial and azimuthal
components of the velocity and p the pressure.
i r
i + denotes a complex frequency.
Following Howard & Gupta (1962), Lessen, Singh
& Paillet (1974), we are led to the following second
order equation for the variable ( ) ( ) r r rG Z ,
( ) ( ) r m r :
0 Z B Z A Z

(15)
with A
r
m k r
m k r

+
1 2 2
2
2 2 2

,
2
k B +
2
2
r
m
r
m

2
-
( )
2 2 2
2
4
r k m
mk
+

2
2
2
k

, where ( ) r =
( ) ( ) dr rV d r

1 is the basic flow axial vorticity.
This equation is valid whatever ( ) r . With the
boundary conditions ( ) 0 Z ( ) 0 Z , this equation
constitutes an eigenvalue problem for . This
problem admits a countable infinity of eigenvalues
indexed as ( ) k
n , m
where k and m are the axial
and azimuthal wavenumbers and where the second
index n is related to the number of zeros of the
eigenmode (the higher the label, the more radial
oscillations the mode contains). Resolution is
achieved using a shooting method. Results are
shown in figure 8 for the axisymmetric modes
0 m and the helical modes 1 m . The results are
very similar to those obtained with a Rankine
vortex except for the modes that possess a critical
layer. Critical layers occur for non-axisymmetric
modes ( 0 m ) whenever the angular phase speed
of the perturbation, m
r
, coincides with the
angular velocity of the vortex ( )
c
r at some radius
c
r , i.e. ( ) { } 0
c
r Re . Modes which possess a
critical layer are necessarily damped, i.e. 0 <
i
,
unless if the mean flow vorticity ( )
c
r at the
critical radius is exactly zero, a condition which is
not fulfilled for a Lamb-Oseen vortex. Such
damped critical layer modes occur when
< <
2
2 0 a
r
, and are represented by thin lines
in figure 8(b). Outside this interval, the modes do
not possess a critical layer and are purely
oscillatory, with frequencies represented by the
thick lines in figure 8(b).
From the above described results, the following
properties of the Kelvin waves in a Lamb-Oseen are
found :
- The axisymmetric modes 0 m have a group
velocity dk d
r
which increases with wavelength.
A maximum is obtained for the branch
1 0,
when
0 ka . The slope at origin for this mode in figure
8(a) is found to be :
( )
a
.
dk
k d
,

2
63 0
1 0
(16)
Interestingly, in a Lamb-Oseen vortex, one finds
that the maximum tangential velocity is
( ) a . V
max

2 63 0 . This means that the energy


of axisymmetric perturbations propagates with a
speed smaller than the maximum tangential velocity
of the vortex : ( ) dk k d
n , 0

max
V

. In the case
where 1 <

V V
max
, which holds in trailing
(a)
(b)
(c)
7
vortices, energy of the perturbations is thus
convected downstream. This prevents the possibility
of perturbation energy traveling upstream at a higher
speed than that of the flow, a necessary condition for
occurrence of a "hydraulic jump" leading to vortex
bursting.
(a)
k a

0
0 1 2 3 4 5
-1.5
-1
-0.5
0
0.5
1
1.5

0,2

0,1

0,-1

0,-2

0,-3

0,-4

0,-5

0,5

0,4

0,3
C
o
n
t
i
n
u
o
u
s
s
p
e
c
t
r
u
m
k a

0
0 1 2 3 4 5
-1.5
-1
-0.5
0
0.5
1
1.5

0,2

0,1

0,-1

0,-2

0,-3

0,-4

0,-5

0,5

0,4

0,3
(b)
k a
0 1 2 3 4 5
-1
-0.5
0
0.5
1
1.5
2

1,2

1,1

1,5

1,4

1,3

1,-2

1,-1

1,0

0
0 1 2 3 4 5
-1
-0.5
0
0.5
1
1.5
2
Figure 8 - Frequencies
r
of the Kelvin waves in a
Lamb-Oseen vortex : (a) axisymmetric mode 0 m , (b)
helical mode 1 m .
2
0
2 a (from Fabre et al.
2001).
- Let consider now the case of helical modes
1 t m . According to the symmetries of the base
flow, one has
n , n , 1 1
. This means that the
helical waves must be considered by pairs, the left-
handed modes ( 1 m ) propagating along the vortex
axis in opposite directions than their right-handed
counterpart ( 1 m ). As shown by Widnall et al.
(1971), the asymptotic behavior for 0 ka of the
branch
0 1, t
is :
( )

,
_

+
,
_


t
Cte
ka
ln
k
k
,
1
4
0
2
0 1
m (17)
This branch is called the "slow branch" since both
the frequency and the phase velocity k c
r

(and also the group velocity dk d
r
) tend towards
zero when k goes to zero. The corresponding mode,
which has the least complex spatial structure
(eigenfunctions with no zero), is responsible for the
main part of the vortex core displacement. This
mode corresponds to the self-induced oscillation
mode of a filament vortex introduced in 3.1, see
relations (9)-(10).
- The oscillations which correspond to the
branches
n , 1 t
, n=1, 2, in figure 8(b), verify :
( )
( )
( )( )

,
_

+ +
+

t
t
n n
ka
a
k
n ,
2 1
2
1
2
0
2
2
1
, (18)
see Leibovich et al. (1986). The angular velocity
( ) m m , k 0 of these modes exactly matches
that of the vortex axis ( ) 0 r . They are called
"fast branches" because their phase velocity tends
to as 0 k (note however that, according to
(18), the group velocity of these "fast waves" tends
to zero with k). Calculation of their spatial structure
shows that in the limit 0 k , these modes are
concentrated in a very small region near the vortex
axis and that, consequently, they are likely
eliminated by viscosity, see Fabre et al. (2001).
- It exists wavenumbers where 0
r
, for which
the helical perturbations 1 t m are steady. For a
Lamb-Oseen vortex, see figure 8(b), these
wavenumbers are found to be 26 2. ka , 96 3. ,
61 5. , etc. The superposition of these steady waves
with their counterparts 1 m leads to steady
untwisted perturbations. As an example, the vortex
deformation resulting from the superposition of the
two branches
1 1 t

,
is visualized in figure 9 for the
case of a Rankine vortex (in a Rankine vortex, the
wave number for which 0
r
is 5 2. ka , a
value slightly higher than that of the Lamb-Oseen
which is 26 2. ka ). The vorticity field associated
to this wave is shown in figure 10 for the case of a
Lamb-Oseen vortex.
An interesting mechanism is now a possible
exponential amplification of these steady Kelvin
waves by straining imposed by other vortices,
which means instability.
3.4. Core instabilities
As seen in 3.1, the presence of a second vortex
leads to imposition of a strain field on the current
vortex. Such a field has been described in (8) for a
counter-rotating vortex pair ; it corresponds to a
strain of rate ( )
2
2 b
~
with axes oriented t 45
and its is responsible for the displacement of the
two vortex filaments in the Crow instability
mechanism.
0 <
i
8
Figure 9 - Wave motion produced by the combination of
two steady Kelvin waves
1 1

,
and
1 1

,
for 5 2. ka
on a Rankine vortex. The vortex core boundary is
displayed with an arbitrary amplitude (from Eloy, 2000).
Now, considering the flow within the core, the strain
field amplifies, by stretching, any vorticity
perturbations aligned with the strain axis, such as
that in figure 10(a). This mechanism, called Widnall
instability, may be quantified by means of a multiple
time scale analysis, see Moore & Saffman (1975),
Tsai & Widnall (1976). Considering the flow
evolution on the characteristic length and time scales
a L and
2
2 a T , the strain amounts to a
perturbation at order ( )
2
b
~
a . The assumption :
1
2
<<
,
_


b
~
a
(19)
allows an asymptotic approach. Using polar co-
ordinates, the strain field depends on the second
harmonic ( ) i exp 2 . The total divergence-free
velocity field in the plane normal to the vortex axis
may then be expressed into the form :
( ) + r U U
0
( ) ... , r U +
1
(20)
with ( ) r U
0
( ) ( )
t
, r r , 0 0 , ( ) r U
1
= ( , R sin f 2
( ) )
t
, r cos f 0 2 2 . The stream-function at order
is ( ) ( ) 2 2
1
cos r f , r . The function ( ) r f
must be determined so that (20) fulfils the Euler
equations at order , see Moore & Saffman (1975).
Considering 3D perturbations, i.e. two Kelvin waves
with azimuthal wavenumber
1
m and
2
m , with an
amplitude of the order 1 << , interaction leading to
instability occurs at the order . The mechanism
described with this method is a global resonant
interaction between three steady perturbations, i.e.
two Kelvin waves and the strain, which takes place
as soon as the triadic resonance relation 2
2 1
m m
is satisfied. For the Lamb-Oseen vortex, this only
occurs for 1
1
m , 1
2
m . In the Rankine vortex
other Kelvin modes may be involved, such as
2
1
m , 0
2
m (see Moore & Saffman, 1975,
Eloy, 2000). Note that the resonant condition is also
fulfilled in the limit 0 k for the slow branch
Kelvin modes
0 1, t
, see figure 8(b). This is related
to the Crow instability and requires a different
asymptotic treatment as that presented in 3.1.
(a)
(b)
Figure 10 - 3D-vorticity perturbation field of the wave
motion produced by the combination of two opposite
kelvin waves 1 m and 1 m on a Lamb-Oseen
vortex. (a) vorticity in the plane orthogonal to the vortex
axis, (b) axial vorticity. (from Sipp, 2000).
The amplification rate
i
of these co-operative
instabilities, as obtained using the Moore &
Saffman's method described above, reads :
( ) ( ) [ ]
2 2 2
2 Q a k ka R b
~
c
(21)
This corresponds to a narrow band of instability of
width Q R a k ka
c
< where R and Q are two
constants of order 1 which depend on the actual
vortex model ;
c
k is the wave number such that
( ) 0
1

t c n ,
k . The peak of instability is reached
4
4
9
for a k ka
c
in which case
2
2 b
~
R . As an
example, figure 11 shows the amplification rate of
the instabilities due to resonance of the straining
field with the helical waves ( 0
r
,
1 t m , 61 5 96 3 26 2 . , . , . ka ) for a lamb-Oseen
dipole of aspect ratio 2 0. b
~
a . The results are
plotted here versus kb
~
. The first lobe, close to the
origin 0 b
~
k , corresponds to the Crow instability.
The short-wave instabilities concerns wave numbers
12 b
~
k . The status about the co-operative
instabilities is the following :
- Amplification rate of these short wave co-
operative instabilities is ( )
2
2 4 1 b
~
. , see
figure 11. It is slightly larger than the amplification
rate of the Crow instability, ( )
2
2 8 0 b
~
.
Crow
.
The time scale
1
of the two instabilities are
thus equivalent and of the order ( )
2
2 b
~
O .
0 10 20 30 40 50
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Figure 11 - Amplification rate of the co-operative
instabilities due to resonance of the straining field with
the steady helical waves ( ) k
, 1 1 t
such that 0
r
in a
Lamb-Oseen vortex pair ( 2 0. b
~
a ).
- Compared with the viscous time scale

2
a , we have ( ) Re b
~
a
2

. Therefore,
if we assume that ( ) 1 0. O b
~
a , for 100 >> Re , the
co-operative instabilities are free from viscous
damping.
- Dependence of co-operative instabilities on the
vortex model is a question of importance. Only
small differences are found when comparing two
models based on a single scale, such as the Rankine
vortex and the Lamb-Oseen vortex for instance : the
most amplified wave numbers
c
k and the
amplification rates are comparable. The main
difference lies in the suppression of resonances
(such as 2 0
2 1
m , m ) due to critical layers. Note
that, given any of the one scale vortex models
considered so far, one may replace the core scale a
by the dispersion radius r , see (4), these two scales
being of the same order.
- Things become different when considering more
representative models such as that suggested by our
experimental results, see figure 2 to 4. Figure 12
shows the first amplification lobe of the short wave
instabilities obtained when considering three
simplified vortex models. The first one is the
Lamb-Oseen vortex pair, such that 2 0. b
~
a , the
second one is a "plateau" vortex pair with a
constant tangential velocity region from
02 0
1
. b
~
r to 2 0. b
~
r and the third is a vortex
0 25 50 75 100
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Figure 12 - Amplification rate of the first Widnall
instability in pairs of vortices corresponding to different
models : Lamb-Oseen ( 2 0. b
~
a ), "plateau" and Betz-
like, with 02 0
1
. b
~
r , 2 0
2
. b
~
r . For the Betz-like
model which is defined by equation (3), different values
of the exponent have been considered.
pair corresponding to the model (3), with internal
radius 02 0
1
. b
~
r and external radius 2 0
2
. b
~
r .
In accordance with figure 11, the short-wave
instabilities in the Lamb-Oseen dipole starts around
12 b
~
k . Compared to this reference case, it is seen
that the two scale models leads to a shift towards
smaller wavelengths of the instability regions. The
"plateau" model leads to occurrence of the first
Widnall instability at a slightly higher wavenumber
( 20 b
~
k ). In the case of the Betz-vortex dipole,
the higher , the smaller are the unstable
b
~
. a 2 0
Oseen - Lamb

V
1
"plateau"

r
1
r

V
b
~
. r 02 0
1
b
~
. r 2 0
2

4 0.
45 0.
5 0.
55 0.

2
2 b
~
b
~
k
b
~
k

2
2 b
~
b
~
. r 02 0
1
b
~
. r 2 0
2

Equation (3)
10
wavelengths. For instance, using 5 0. , it is found
that the first shortwave co-operative instability is
85 b
~
k
c
. This corresponds to 7 1
1
. r k
c
which
means that the instability scales with the inner core
radius. Note that, interestingly, the wave length and
the width of the instability band are extremely
sensitive to the value of within the range of
values observed in the experiments, see section 2.
This topic is the object of ongoing work.
- A last interesting theoretical result is that
obtained by Sipp (2000) using a weakly non-linear
analysis of the Widnall instabilities of a strained
Lamb-Oseen vortex. The result is that short-wave
instabilities saturate very quickly. The mechanism
responsible for this saturation is a sudden self
induced rotation of the plane waves that occurs
when their amplitude become large. Detailed results
shows that the maximum vortex core distortions
induced by the Widnall instability is then limited to
values smaller that b
~
.01 0 before non-linear
interactions occurs, see Sipp (2000).
In conclusion, the above results suggest that, in
representative vortex wakes, the shortwave co-
operative instabilities are likely confined in the very
center region of the vortex cores and that they
cannot lead to significant core distortions before
they saturate.
3.5. Global instability analysis
The analysis of long- and short-wave perturbations
conducted above can be merged using a more
general approach, by relaxing the assumptions (12)
and (19). Any flow quantity ( )
t
p , w , v , u q is again
considered as :
( ) ( ) ( ) t , z , y , x q
~
z , y q t , z , y , x q + , (22)
with a small-amplitude perturbation of order and
form :
( ) ( )
( )
. c . c e y , x q t , z , y , x q
~ t kx i
+

(23)
where { } Re
r
is related with the frequency of a
global eigenmode q
~
while the imaginary part,
{ } Im
i
is its growth/damping rate. The
objective of the analysis is still identification of
unstable eigenvalues and associated eigenvector
amplitude functions q
for a given basic state q
describing the wake-vortex system. The system for
the determination of and the associated
eigenfunctions q
in its most general form can be
written as the complex nonsymmetric generalised
EVP :
( ) ( )
( ) [ ] ( )
( ) ( ) [ ]

'

+


+ +
w i p D w w D L v w D
v i p D w v D v v D L
u i p ik w u D v u D u L
w D v D u ik
z z y
y z y
z y
z y
0
(24)
where the linear operator is
( )( )
z y z y
D w D v u ik D D k Re L + +
2 2 2
1 (25)
and y D
y
, y D
y
2 2 2
, Significant
comments regarding the global instability analysis
are, first, that the two-dimensional eigenvalue
problem (24) permits considering wake-vortex
systems having a velocity component u in the
direction of the aircraft motion, x, in addition to
those defined on the Oyz plane, v and w ; the
only assumptions of the analysis are
0 x q t q (26)
the first of which may be relaxed in case of a time-
periodic basic state by employing Floquet theory
(Herbert, 1983, Barkley & Henderson, 1996). It
should also be noted here that solution of one of the
alternative simplified forms of the partial derivative
eigenvalue problem (24) valid in case of a single
velocity component (Tatsumi & Yoshimura, 1990)
or, additionally, in the inviscid limit (Hall &
Horseman, 1991) is not permissible in the wake-
vortex stability problem.
Successful applications of global instability
analysis based on (24) are the studies of the swept
attachment-line boundary layer, Lin & Malik
(1996), who exploited the symmetries of that
problem to reduce the computing effort, that of
Theofilis (1998) who solved (24) for the same
configuration without resort to symmetries as well
as the analyses of Theofilis (2000) in open and lid-
driven cavity flows and that of Theofilis et al.
(2000) in boundary layer-flow which encompasses
a closed recirculation bubble. Compared with the
latter applications, in which a matrix eigenvalue
problem of leading dimension in excess of
4
10 was
solved using state-of-the-art numerical algorithms
for the discretisation of (24) and efficient
algorithms for the eigenspectrum, the wake-vortex
system presents the additional challenge of yet
higher resolution being necessary for the adequate
description of the wake-vortex system basic flow
itself and, consequently, the sought global
instabilities.
One simplification which halves the storage
requirements for the solution of (24) that are
typically of the order of several gigabytes, is the
case in which the basic flow velocity component u
is absent in the wake-vortex system. In conjunction
of the redefinitions i
~
, w i w
~
, this results in
11
a real partial-derivative eigenvalue problem
enabling storage of real arrays alone, as opposed to
the complex arrays appearing in (24). Freeing half of
the necessary storage results in the ability to address
flow instability at substantially higher resolutions
and/or higher Reynolds numbers compared with an
analysis based on solution of (24). However, from a
physical point of view, neglecting the axial velocity
component u in the basic flow restricts the classes
of flows that can be addressed by a global instability
analysis. With this consideration in mind, the
potentiality of the method will be illustrated by
considering the case of a system composed of
Batchelor vortices, each of which is characterized
by :
( )
2
r
e z , y u

( ) r e cos q z , y v
r

,
_



2
1 (27)
( ) r e sin q z , y w
r

,
_



2
1
where ( ) ( )
n n n
a z z y y r
2 2
+ , ( )
n n
z , y
denotes the centre and
n
a the radius of vortex n.
Here the centreline axial velocity of the vortices has
been taken equal to unity. In constructing a basic
flow composed of several Batchelor vortices
satisfying (27), the additional freedom exists in the
choice of the relative circulation, radius and location
of the vortices. The basic flow analysed here was
constructed along the lines of those discussed in
3.2. It consists in two pairs of co-/counter-rotating
vortices. A first case was constructed by taking
1
1
q and 5 0
2
. q and assuming a stationary
configuration fixed by the Rennich & Lele condition
(11). The first vortex was placed at the outmost
starboard location ( ) ( ) 0 7
1 1
, y , x , lengths being
made nondimensional using the radius parameter of
the outer vortices. The radii were chosen
consistently with the definition (27), 1
1
a ,
5 0
2
. a . The Rennich-Lele condition delivers the
location of the second vortex ( ) ( ) 27 1 0
2 2
. , y , x in
this case. The locations and swirl of these two
vortices were mirrored with respect to the centreline
0 x to construct the fields which are illustrated in
figure 13 (a) through the distribution of the vertical
velocity ( ) z , y w . A second case was constructed
using the same vortex locations and 1
1
q ,
5 0
2
. q . Such a configuration is not stationary.
A Reynolds number can be defined as

1 1
q Re . The global instability analysis was
performed at
3
1
10 Re and several axial periodicity
wavenumbers k, of which results at a short wave
number, 3 2 k i.e. 4 9 2 . k L
x
, are
(a)
(b)
Figure 13 Vertical velocity ( ) z , y w of the basic flow
considered : (a) 0
2
> q (co-rotating), (b) 0
2
< q (contra-
rotating).
indicatively presented. Homogeneous Dirichlet
boundary conditions have been imposed on all
disturbance velocity components at the far-field and
a compatibility condition was used on the
disturbance pressure at the boundaries. A resolution
analysis study was performed to ensure integrity of
the results presented, using 64 16 32 64 16 32 ) ( ) (
Legendre collocation points to resolve the two-
dimensional domain considered, [ ] 10 10, z
[ ] 5 5, y . Accordingly, the Krylov subspace
dimension was increased from 200 m at the
lowest- to 400 m at the highest-resolution runs.
The resulting memory requirements for the
recovery of the most interesting window of leading
eigenvalues and eigenvectors at a single pair of the
parameters ( ) k , Re ranged from 300 Mbytes to 4.5
Gbytes and the corresponding runtime from 1.5 to
65 mins at 3.5 Gflops on a supercomputer. In all
runs a shift parameter 0 was used, ensuring
resolution of the eigenspectrum in the
neighbourhood of ( ) 0
i r
, . We present results
of the two cases 0
2 1
> q q and 0
2 1
< q q in order to
facilitate qualitative and quantitative comparisons.
The symmetry of the basic flow suggests that either
stationary or complex conjugate pairs of
eigenmodes are to be found in the spectrum.
Indeed, this result can be seen in figure 14 in which
the parts of the eigenspectra recovered at the two
highest resolutions utilised,
2
56 and
2
64 are
presented. Interesting observations are the
following. First, at these parameters the global
instability of the flow 0
2 1
> q q can be easier
12
(a)
(b)
Figure 14 Global eigenspectra at
3
10 Re ,
3 2 k : (a) 0
2
> q (co-rotating), (b) 0
2
< q (contra-
rotating). The symbols correspond to different resolutions
(see text).
resolved in comparison with that in which 0
2 1
< q q .
The lower resolution suffices to deliver several
converged eigenvalues in the first case, while the
higher resolution appears sufficient only for
qualitative statements to be made in the second case.
In both cases the most unstable mode is a stationary
disturbance, i.e 0
r
. The spatial structure of
which may be found in figure 15 through
eigenfunctions ( ) z , y p .
The key statement here is that the destabilisation
of the model wake-vortex system can be adequately
described by numerical means, using global linear
theory. Depending on the relative sign of the swirl
parameters of the outer and inner pairs of vortices in
the example presented either vortex pair system may
be destabilized and eventually lose its coherence on
account of the linear mechanism discussed. In both
occasions the periodicity length of instability,
x
L is
comparable with the spacing of the outer vortices,
and is an order of magnitude smaller than that of
Crow instability. Results of comparisons with eralier
works will be presented in due course. Here it
suffices to stress that global instability theory based
on numerical solution of (24) emerges as one viable
alternative to assist the current efforts to minimise
the coherence of the wake-vortex system, which
delivers results of the quality of direct numerical
simulations at a negligible fraction of the cost of
the latter approach, while retaining the generality
that is necessary to address realistic aircraft wake
configurations.
(a)
(b)
Figure 15 Spatial structure of the most unstable global
eigenmodes : ( ) z , y p : (a) 0
2
> q (co-rotating), (b)
0
2
< q (contra-rotating) .
4. EXPERIMENTAL OBSERVATIONS
Lets have a look now on the unsteady properties
of wake vortices found in experiments.
LDV measurements
It is usually found that the energy of the velocity
perturbations within a vortex reaches its maximum
in the vortex center. Figure 16 shows the variation
with the downstream distance of the peak
turbulence rate

V k measured with a 3D-LDV


system in the center of the wing tip vortex of the
A300 model presented in section 2. It is observed
that the turbulent kinetic energy k first decays and
then remains almost constant beyond 3 spans. In the
high lift case, the merger takes place at 2 b x .
The damping of the energy decrease beyond this
distance shows existence of a mechanism that
produces perturbations. If not,

V k would
decrease monotonously under diffusion and
dissipation. This production could come from the
development of long-wave and short wave
cooperative instabilities.
Hot wire measurements
The perturbations characterized above do not
correspond to an equilibrium turbulence but are
characteristic of a global "meandering" of the
r

13
vortices. This is put in evidence by considering the
spectral contain of the perturbations. Figure 17
shows the energy density of the axial component of
the velocity measured using a standard single wire
DISA P11 probe placed in the center of the high lift
and clean configuration vortices at 5 b x . Some
0 2 4 6 8 10
0
0.025
0.05
0.075
0.1
0.125
0.15
0.175
0.2
k
max
(HL)/V

k
max
(C)/V

Figure 16 - Downstream evolution of the peak value of


the turbulence rate

V k measured with a 3D-LDV


system in the centre of the wing tip vortex of a A300
model in the high lift and clean configurations (in the
high lift case, merging with the flap vortex occurs around
2 b x ; from Jacquin et al., 2001).
characteristic slopes are indicated. It is seen that the
vortex meandering corresponds to a broadband
spectrum which exhibits a sharp energy excess for
frequencies smaller than, say Hz f 1000 (red
curves). Searching for possible relationships with
instabilities lead to identification of energy
overshoots in the spectra. In the high lift wake
vortex, three energy accumulations may be
identified (see arrows in figure 17). They are located
at, approximately, Hz f 15 , Hz 55 and Hz 400 .
These energy peaks are not found in the clean case
shown in figure 17(b), except an energy bump
around the first of these frequencies. It is also very
clear from these figures that the small scales of the
clean case vortex are much less energetic than in the
high lift case.
Thus, one may wonder if these features are related
to linear mechanisms. Some answers were proposed
by Jacquin et al. (2001). They are summed-up
bellow with some additional remarks that can be
made in the light of the new theoretical results
presented in figure 12 concerning short-wave
instabilities.
- As seen in 3.2, the theoretical wavelength for
the Crow instability which may develop in a dipole
or in a four-vortex arrangement is 8 0. b
~
k which
means b
~
Crow
8 . At 5 b
~
x , one has
mm b
~
340 which gives
Crow
f
Crow
V

Hz 18 , a value close to Hz 15 . Consequently, the


(a)
10
0
10
1
10
2
10
3
10
4
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
(b)
10
0
10
1
10
2
10
3
10
4
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
Figure 17 Spectral densities of the axial component of
velocity measured with a hot-wire probe in the center of
the vortex at 5 b x . (a) high lift case, (b) clean case.
The red and black curves correspond to 20kHz and 2kHz
sampling, respectively (from Jacquin et al., 2001)
peak located around Hz f 15 may correspond to
the emergence of a long-wave co-operative
instability of this type. As mentioned in 3.2, the
"signature" of a four vortex instability was detected
in two-point correlations measured in the high lift
case. This may contribute to differences in the low
frequency energy contain between figures 17(a) and
17(b).
- The amplification rate of the Crow instability is
( )
2
2 8 0 b
~
.
Crow
. As seen in 3.2, it may
reach a value ten times higher when co-operation
with an inner counter-rotating vortex pair is
accounted for. Development of such long-wave
instabilities takes place on a characteristic distance
b x
high lift
clean
( ) f S
uu
( ) Hz f
-4/3
-3
-5/3
-3
( ) Hz f
10 1 0.1 100
b
~
k
10 1 0.1 100
b
~
k
14


V x . Using an elliptic model for evaluating
and b
~
(see section 2), the characteristic distances
corresponding to an
1
e amplification of the Crow
instability is found to be 30 b x in the high lift
case ( 7 1. C
L
) and 70 b x in the clean case
( 7 0. C
L
). These distances become ten times
smaller if a four vortex system is considered. In
conclusion, only the very first stages
of the development of the Crow instability can be
felt at 5 b x . But the vortices are very
concentrated and slight vortex displacements may
generate strong fluctuations on a fixed hot wire
probe. So, there is no impossibility to establish a
relationship between the first energy peak observed
in the spectrum in figure 17 and a long-wave co-
operative instability, even if the latter only emerges.
- Let consider now the short-wave instabilities. In
the clean case, we saw that a Betz-like vortex with
55 0. fits correctly the velocity profile of the
clean case vortex at 5 b x , see figure 3(b). In this
case, from figure 12 it is found that the first
shortwave co-operative instability occurs around
95 b
~
k
c
that is Hz f
c
2200 using mm b
~
340 .
In figure 17(b), such high frequencies correspond to
very low levels of energy compared with those of
the low frequency part. The conclusion is that, as
suggested by theory, short wave instability
contributions are almost undetectable in the clean
case.
- Things are different in the high lift case. First, the
presence of a plateau region shown in figure 2 lets
room to occurrence of short-wave instabilities at
smaller frequencies than in the clean case. This is
suggested by figure 12. The third energy peak found
at Hz f 400 in figure 17(a) correspond to a
wavelength m .125 0 . This correspond to
17 b
~
k which is not incompatible with the
tendencies depicted by the plateau model in figure
12. The much higher energy contained in these
intermediate frequencies in figure 17(a) compared
with figure 17(b) does suggest that short wave
instabilities could be more active there. This is also
illustrated in figure 18 which shows the product
( ) f S f
uu
obtained when the hot wire is moved
along a vertical line crossing the vortex core up to its
center. Each curve correspond to a displacement of
mm z 1 . Integral under these curves corresponds
to the signal energy. It is seen that energy of the
perturbations undergo a dramatic increase within a
distance of mm 2 1 when approaching the vortex
axis. The energy increase essentially comes from
short-wave contributions.
- Note finally that neutral Kelvin waves, initiated by
perturbations emanating from the model (turbulence,
separations..) and propagating downstream along the
vortex', see 3.3, may also contribute to the
broadband spectra shown in both figures 17(a) and
(b).
10
0
10
1
10
2
10
3
10
4
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Figure 18 - Lin-log plot of ( ) f S f
uu
obtained by
translation of the hot wire in the vortex of the high lift
case at 5 b x , along a vertical line crossing the vortex
core. The highest energy is obtained on the vortex axis.
Each curve corresponds to a displacement of mm z 1
with respect to the vortex axis.
PIV measurements
PIV measurements enable to characterize vortex
unsteadiness in a different way. One considers here
the wake generated by a 1:13.6 A320 half-model
with a semi-span equal to 1.25m. The experiment
was conducted in DNW-LLF. For three different
configurations of the wing, samples of one to two
hundred particle images were taken and the vector
fields evaluated. Besides some exceptions which
showed three vortices, the topology of the field in
the frame of reference of the camera was formed by
two vortices in each half plane, as shown in figure
19. In order to estimate the position and strength of
the vortices, the velocity field was approximated by
two Lamb-Oseen vortices. The distributions of
circulation and radius a, cf. equation (13), were
investigated. The mean value and the standard
deviation of both quantities are given in Table 1.
Though this overlay of velocity fields is not
consistent with Navier-Stokes equations the
approximation is acceptable as shown in figure
19(b), see also Vollmers (2001). In figure 20,
variation of the vortex center points for three
different model configurations have been plotted.
Both the vortex displacements depicted in figure 20
and the variations in the model parameters in
table 1define measures of the instationarity of the
vortices.
Figure 20 gives an indication about meandering
amplitude whose typical value is found to be
m .05 0 , that is 02 0. b . This is much higher
than the value which were inferred from LDV
measurements in the experiment presented above. It
( ) f S f
uu

( ) Hz f
10 1 0.1 100
b
~
kb
~
k
15
(a)
(b)
Figure 19 (a) the velocity field with two vortices
generated by flap and tip of a model wing is
approximated by superposing the velocity fields of two
Lamb-Oseen vortices, (b) comparison of v-component of
velocity in cuts through the approximated vortices
(dashed lines) given in figure (a). PIV data in solid lines.
Circles represent location and radius of Lamb-Oseen
vortices. The respective lines of zero values indicate the
center of vortices as determined by the approximation
with two Lamb-Oseen vortices.
is seen that the flap vortex is subjected to larger
excursions than the wing tip and that its movements
admit a preferential orientation which suggest a co-
operative instability. This could confirm that multi-
polar wakes are much more unstable than dipolar
ones. From figure 20, it can be also inferred than the
"blue" and "green" model planforms produce more
unsteady wakes than the "red" one. At last, the
fluctuations of circulation and vortex core radius
which are reported in table 1 show that the
configuration corresponding to case m05 is more
unsteady than the others.
5. CONCLUSIONS
The characterization of the unsteady properties of
the wake of a given aircraft configuration may be
tested in wind tunnels using conventional unsteady
measurement techniques, i.e. LDV, hot wire and
PIV. Sample results have been presented.
Figure 20 - Distribution of center points from the
approximation by two Lamb-Oseen vortices. The
different colors refer to different configurations of the
same model (95 samples in blue, 298 in green, 100 in
red). The different rectangles refer to the same region of
the flow.
( )

1
2 1
m s

( )

2
2 1
m s

( )
a
m
1
( )
a
m
2
Case m06
Mean 12.3 5.1 0.0501 0.0251
0.32 0.258 0.00197 0.00245
% 2.6% 5.1% 3.9% 9.7%
Case m04
Mean 10.4 5.6 0.0504 0.0345
1.08 0.711 0.00789 0.00869
% 10.4% 12.7% 15.6% 25.2%
Case m05
Mean 10.4 5.6 0.0506 0.0352
1.00 1.02 0.00737 0.0136
% 9.6% 18.1% 14.5% 38.6%
Table 1 - Mean values and standard deviation of
approximations of velocity fields by two Lamb-Oseen
vortices for different configurations of a half model.
Asymptotic linear stability theory leads to the
characterization of important physical mechanisms
that promote unsteadiness in wake vortices. This
theory provides useful reference guidelines to
understand some of the experimental observations.
Application of such quasi-analytical methods to
characterize representative wake flows is underway.
A more complete and predictive approach for
realistic flows, based on a global stability analysis,
is also in progress. Results of comparisons with
asymptotic results and experiments will be
presented in due course.
Acknowledgments
The authors gratefully acknowledge the European
Community and the French-Service des Programmes
Aronautiques (SPA) for their support.
16
6. BIBLIOGRAPHY
[1] Betz A., Verhalten von Wirbelsystemen, Z. Ang.,
Math. Mech., 12, pp164,174, 1931.
[2] Barkley & Henderson R., Three-dimensional
floquet stability analysis of the wake of a circular
cylinder. J. Fluid Mech. 322, 215-241, 1996.
[3] Crow S.C., Stability theory for a pair of trailing
vortices, AIAA Journal, 8 (12). 2172-2179, 1970
[4] Crouch J.D., Instability and transient growth for
two trailing vortex pairs, J.Fluid Mech., vol. 350,
pp. 311-330, 1997.
[5] Devenport W.J., Rife M.C., Liaps S.I. and Follin
G.J., The structure and development of a wing
tip vortex, J. Fluid Mech. Vol. 312, pp. 67-106,
1996.
[6] Donaldson C.P. & Bilanin A.J., Vortex wakes of
conventional aircraft, AGARDograph 204, 1975.
[7] Duck P.W. & Foster M.R., The inviscid stability
of a trailing vortex, ZAMP, 31, pp 524-532,
1980.
[8] Eloy C. & Le Dizes S., Three dimensional
instability of Burgers and Lamb-Oseen vortices
in a strain field, J. Fluid Mech. Vol. 378, pp.
145-166, 1999
[9] Eloy C., Instabilit multipolaire de tourbillons,
PhD Universit d'Aix-Marseille, 2000.
[10] Fabre D. & Jacquin L., Stability of a four-
vortex aircraft wake model, Phys. Fluids, vol 12,
N10, pp. 4238-4243, 2000.
[11] Fabre D., Jacquin L. & Sipp D., The response
of columnar vortices to three dimensional
perturbations, in preparation
[12] Gerz T., Holzpfel F. & Darracq D., Aircraft
wake vortices a position paper, WakeNet
position paper to appear.
[13] Hall P. and Horseman N.J., The linear inviscid
secondary instability of longitudinal vortex
structures in boundary layers, JFM, 232, p. 357-
375, 1991.
[14] Herbert Th., Secondary instability of plane
channel flow to subharmonic 3D-disturbances,
Phys.Fluids,26,1983p.871-874, 1983.
[15] Howard L.N. and Gupta A.S., On the
hydrodynamic and hydromagnetic stability of
swirling flows, J. Fluid Mech., 14, 463-476,
1962.
[16] Jacquin L., Fabre D., Geffroy P. & Coustols E.,
The properties of a transport aircraft wake in the
extended near field; AIAA paper 2001-1038
[17] Leibovich S.N., Brown S.N. & Patel Y., Short-
waves on inviscid columnar vortices, J.Fluid
Mech., 173, 595-624, 1986.
[18] Lessen M., Singh P.J., and Paillet F., The
stability of a trailing line vortex. Part 1.
Inviscid theory, J. Fluid Mech., 63(4):753-763,
1974.
[19] Lin, R.-S. and Malik, M. R., On the stability of
attachment-line boundary layers, Part 1. The
incompressible swept Hiemenz flow, J. Fluid
Mech., 311, p. 239-255,1996.
[20] Mayer E.W. & Powell K.G., Viscous and
inviscid instabilities of a trailing vortex, J. Fluid
Mech., vol 245, pp 91-114, 1992.
[21] Moore D.W., Saffman P.G., Motion of a
vortex filament with axial flow, Proc. R. Soc.
Lond., Ser. A 272, pp. 403-429, 1972.
[22] Moore D.W., Saffman P.G., The instability of
a straight vortex filament in a strain field, Proc.
R. Soc. Lond., A346, pp.413-425, 1975.
[23] Orlandi P., Carnevale G.F., Lele S., Shariff
K., Reynolds W.C., DNS study of stability of
trailing-vortices, Proceedings of the Summer
Program 1998, Center of Turbulence Research,
1998.
[24] Rennich S.C. & Lele S.K., A method for
accelerating the destruction of aircraft wake
vortices, AIAA A98-16497, 1998.
[25] Rossi M., Of vorticies and vortical layers : an
overview, in Vortex Structure and dynamics,
Lecture Notes in Physics 555, Maurel &
Petijeans (Eds.), Springer, 2000.
[26] Saffman P.G., Vortex Dynamics , Cambridge
Monographs on Mechanics and Applied
Mathematics, Cambridge University Press,
1992.
[27] Sipp D., Weakly nonlinear saturation of
shortwave instabilities in a strained Lamb-
Oseen vortex. Physics of Fluids, Vol. 12, N 7,
pp. 1715 - 1729, 2000
[28] Sipp D., Coppens, F. & Jacquin, L.,
Theoretical and numerical analysis of the
dynamics of wake vortices, in Third
International Workshop on Vortex Flows and
Related Methods, Toulouse, France, 1998.
European Series in Applied and Industrial
Mathematics (ESAIM),
www.emath.fr/Maths/Proc
[29] Spalart P., Airplane trailing vortices, Annual
Review of Fluid Mechanics, vol. 30, pp. 107-
138, 1998.
[30] Stuff R., Dieterle L., Horstmann K-H., Pailhas
G., Coton P. & , Lozier J-F., Present issue
[31] Tatsumi T. and Yoshimura T., Stability of the
laminar flow in a rectangular duct, J.Fluid
Mech.,212, p. 437-449, 1990.
[32] Theofilis V., Globally unstable basic flows in
open cavities, AIAA, 2000.
[33] Theofilis V., Linear instability in two spatial
dimensions, Proc. of the European
Computational Fluid Dynamics, Conference
ECCOMAS '98, p. 547-552, editor K. Papailiou
et. al, Athens, Greece, 1998.
[34] Theofilis V. and Hein S. and Dallmann U.Ch.,
On the origins of unsteadiness and three-
dimensionality in a laminar separation bubble,
Phil. Trans. Roy. Soc. London (A), 358, p.
3229--3246, 2000.
[35] Tsai C.Y., Widnall S.E., The instability of
short waves on a straight vortex filament in a
weak externally imposed strain field, J. Fluid
Mech. 73 , pp. 721,733,1976.
17
[36] Vollmers H., Detection of vortices and
quantitative evaluation of their main parameters
from experimental velocity data. Meas. Sci.
Technol. 12, pp. 1-9, 2001.
[37] Widnall S.E., Bliss D. & Zalay A., Theoretical
and experimental study of the stability of a
vortex pair, in "Aircraft wake turbulence and
its detection ", Plenum press, 1971

Você também pode gostar