Numerical Study of Buffet and Transonic Flutter On The NLR 7301 Airfoil

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Aerospace Science and Technology 7 (2003) 540550

www.elsevier.com/locate/aescte

Numerical study of buffet and transonic flutter on the NLR 7301 airfoil
W. Geissler
DLR-Institute of Aerodynamics and Flow Technology, Bunsenstr. 10, Gttingen 37073, Germany
Received 25 July 2002; received in revised form 3 July 2003; accepted 7 July 2003

Abstract
Intensive numerical calculations have been carried out on the NLR 7301 supercritical airfoil section under transonic flow conditions. The
airfoil is assumed to be suspended in a two degree of freedom aero elastic system allowing the airfoil to move in both pitching and plunging
modes respectively. The numerical code used for the present investigation is a two-dimensional time accurate NavierStokes solver based on
the full equations. The SpalartAllmaras turbulence model has been used throughout the calculations.
The elastic constraints, mass ratio, etc. have been taken from frequent tests in the DLR-DNW transonic wind tunnel carried out by the
DLR Institute of Aeroelasticity in Gttingen, Germany.
The numerical calculations have two goals:
(1) investigation of the buffet boundary on the fixed airfoil under transonic flow conditions,
(2) flutter investigations on the airfoil under aero elastic conditions.
From the flutter calculations it has been shown that in addition to the Mach number, the neutral position of the torsion spring is of considerable
importance for the investigation. Therefore this parameter has been studied in some detail.
Numerical results obtained are presented in suitable form and compared with available experimental data.
2003 ditions scientifiques et mdicales Elsevier SAS. All rights reserved.
Keywords: Aeroelasticity; Unsteady Aerodynamics; Fluide Structure Interaction; Buffet; Limit Cycle Oscillations; Stock Induced Separation

1. Introduction
The present numerical investigations are motivated by
extensive transonic flutter experiments on the NLR 7301
supercritical airfoil section as done by the DLR Institute
of Aeroelasticity, [14,15]. These tests have been carried out
in the transonic wind tunnel facility TWG at the DLR in
Gttingen, Germany. The airfoil model has been suspended
in an oscillatory test stand allowing the model to move as a
two degree of freedom aero elastic system.
Several test periods have already been completed, starting
with first measurements in the perforated (1 m 1 m)
test section of the tunnel. Comparisons of these earlier
tests with present numerical results make a correction for
wind tunnel wall interference effects a necessity. Wall
correction procedures for the TWG have been described i.e.
in [11]. However the correct representation of perforated
wall interference effects on unsteady loads is a formidable
E-mail address: wolfgang.geissler@dlr.de (W. Geissler).

task. More recent experiments with the same test stand and
model have been carried out in the new adaptive wall test
section of the tunnel.
The present numerical code assumes free flight conditions. Due to the fact that the different non-linear transonic
effects measured in the perforated wall test section have also
been found in the adaptive wall test section, wind tunnel wall
corrections have not been tried in the present investigations.
The main results of the flutter experiments are the
observations of limit cycle oscillations (LCOs), which the
model has been experienced in different flow regimes. The
measured phenomena could be identified as of different
types:
Combined pitching and plunging oscillations;
Pitching oscillations alone;
Indifferent changes between pitching and plunging
modes.
Further it was observed, that different levels of amplitudes
might occur at the same parameter set (coexisting LCOs). In

1270-9638/$ see front matter 2003 ditions scientifiques et mdicales Elsevier SAS. All rights reserved.
doi:10.1016/S1270-9638(03)00065-8

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

541

Nomenclature
Dimensional

0
s
a
b
c
f
gh
g
h
I
Kh
K
L
M
m
p0

S
h
a

airfoil incidence . . . . . . . . . . . . . . . . . . . . . . . . . . .
zero force torsion spring incidence . . . . . . . . . .
incidence for steady start condition . . . . . . . . .
free stream speed of sound . . . . . . . . . . . . . m s1
half chord (b = c/2) . . . . . . . . . . . . . . . . . . . . . . m
chord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m
frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hz
damping coefficient, heave . . . . . . . . . . . . kg s1
damping coefficient, torsion . . . . . . . . kg m2 s1
plunge displacement . . . . . . . . . . . . . . . . . . . . . . m
mass moment of inertia about elastic axis kg m2
bending stiffness . . . . . . . . . . . . . . . . . . . . . kg s2
torsional stiffness . . . . . . . . . . . . . . . . . kg m2 s2
aerodynamic lift per unit span . . . . . . . . . . . . . N
aerodynamic moment per unit span (with
respect to the elastic axis) . . . . . . . . . . . . . . . Nm
mass of airfoil structure . . . . . . . . . . . . . . . . . . kg
total pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . bar
inter modal phase shift . . . . . . . . . . . . . . . . . . . . .
static mass moment about elastic axis . . . . kg m
natural bending frequency . . . . . . . . . . . . . . . s1
natural torsion frequency . . . . . . . . . . . . . . . . s1

all cases investigated in the wind tunnel tests the amplitudes


of both pitching and plunging modes were rather limited
in size: The pitch amplitude did not exceed about 0.5 ; the
plunge amplitude was limited to about 1 mm.
In addition to the experiments several numerical treatments of the problem have been tried.
In [15] first calculations have been carried out with
two different codes based on the NavierStokes equations
but the measured LCOs could not be found as measured.
In [17] first results have been presented which clearly
show limit cycle oscillations. However the amplitudes of
the pitching mode increased 4o and therefore exceeded
the experimentally observed values by an order of magnitude.
Later comprehensive calculations have been performed
[3] to systematically investigate the influence of perforated
wind tunnel walls. Modelling the porosity of the wind
tunnel walls showed a significant influence of the viscosity
parameter on the prediction of the limit cycle amplitude. The
computed phase angle between pitch and plunge motions as
well as the flutter frequency were not significantly affected
(see [3] for details).
The present study has focused on a different parameter.
It has been pointed out from the experimental side that the
Mach number is of considerable importance. Very small
variations of Mach may change the output of the system
completely. Assuming that the total pressure is constant, the

t
U

time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . s
free stream velocity . . . . . . . . . . . . . . . . . . . m s1

Non-dimensional
CL
CM
cp
h

x

V
g h
g
M
Re
T

LCO

2 c)
lift coefficient, CL = L/( /2U
2 c2 )
moment coefficient, CM = M/( /2U
pressure coefficient
h/c
airfoil-air/mass density ratio, m/( b2 )
radius
of gyration about elastic axis

(I /(mb2))
distance of mass centre from elastic axis
S /(mb)
dimensionless airspeed, U /(b )
structural damping coefficient (heave)
gh c/(ma )
structural damping coefficient (pitch),
g /(ma c)
free stream Mach number
Reynolds number, Re = U c/
time, T = tU /c (steady: T = ta /c)
reduced frequency, 2f c/U
Limit Cycle Oscillation

second similar important parameter is the incidence of the


airfoil. But it should be emphasised that the final incidence
is an output parameter, i.e. a result of either the experiment
or the calculation.
The parameter to be selected in advance is the neutral
angle of the torsion spring (0 ).
0 determines where in the complex non-linear flow
environment the solution is finally found.
The final incidence either steady or oscillatory is adjusted
due to the M-0 parameter set selected.
The present calculations are therefore concentrated on
a systematic variation of 0 keeping the Mach number
constant.
Before the set of aero elastic equations will be solved,
the airfoil will be investigated rigidly within the high Mach
number regime where the so-called transonic dip does exist
limiting stable from unstable flutter conditions.
To find the buffet boundaries of the airfoil is a first
important test to investigate the suitability of the numerical
code in combination with the turbulence model. In all
present calculations the SpalartAllmaras (SA) one equation
model [16] has been used.
In [8] similar calculations have been carried out for the
18% circular arc airfoil which has been investigated in detail
in [12]. The results have shown that the Mach number for
buffet-onset, buffet frequency and amplitude, etc. matched

542

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

the experimental data closely, if the SA-turbulence model is


used.

2. Numerical code
The present numerical code is based on the 2D-time
accurate NavierStokes equations: The complete system is
taken into account [5] different from other codes, which
are only based on the thin layer approximation of the
equations.
The code is based on the Approximate Factorisation
Implicit methodology originally developed by Beam and
Warming, [2]. The implicit feature of the solution procedure
allows using a rather large time-step without exceeding
the stability boundary. However in the present flow cases
strong separation effects are involved. In these cases the
time step is not only restricted by stability boundaries but by
physical constraints: the time resolution of the flow has to
be high enough in order to resolve the smallest frequencies
sufficiently. The number of time steps is therefore dictated
by the physical problem involved rather than by numerical
stability bounds.
One special feature of the code is the possibility to
deform the grids with respect to time. This feature allows for
the representation of oscillating flaps [9] or time-dependent
deformation of the airfoil leading edge [10] in order to
improve separation characteristics of the airfoil in case of
a helicopter rotor blade.
A C-type grid topology has been used throughout the
present study. From the 361 71 grid 271 points represent
the airfoil surface. The wall distance of the first grid line has
been set to 5 105 chord keeping y + to about 1.
A variety of turbulence models has been included in the
code: from the algebraic BaldwinLomax model, [1], the
SpalartAllmaras one equation model [16] to the k two
equation model [13] respectively.
First steps have also been introduced in order to take
into account transition effects in the unsteady code, [6]. In
the present calculations however fixed transition has been
assumed on both lower surface (14%) and upper surface
(7%) as it has been verified in the experiments.
The numerical code used for the present calculations is
represented by three different components:

3. Buffet investigations
For the successful calculation of the buffet phenomenon
the numerical code has to be time-accurate and the number
of time-steps has to be sufficiently high in order to resolve all
the time scales involved in the unsteady and separated flow
environment.
In [12] the buffet boundary for the 18% thick circular arc
airfoil has been experimentally investigated by McDevitt.
The test results could be calculated accurately [8]. The
procedure applied was a stepwise increase of the Mach
number keeping the incidence at zero.
In the following the same approach will also be applied
for the NLR 7301 airfoil section.
Fig. 1 shows the lift time histories with respect to the
number of time-steps for a variation of increasing Mach
numbers, starting at M = 0.77 to M = 0.80 respectively. Up
to M = 0.78 the lift curves are damped after less than 30 000
time steps. For M = 0.79 the lift curve shows oscillations
which have the tendency to be damped out after a large
number of time steps. The calculations have been stopped
after 60 000 time steps.
At M = 0.80 the lift curve shows stable buffet oscillations
with a frequency of 81.1 Hz if the wind tunnel model chord
of c = 0.3 m is taken into account. In the experiment a buffet
frequency of 71 Hz ( = 0.55) has been measured in the
adaptive wall test section of the TWG, [15].
It is of considerable interest to investigate the buffet
behaviour of the NLR 7301 airfoil if the Mach number is
further increased.
Fig. 2 shows the corresponding lift time histories up
to M = 0.83 without changing either incidence or total
pressure.
At M = 0.81 well established buffet oscillations are
found with a slightly higher frequency of f = 88.5 Hz. If
M is increased further, to M = 0.82 and 0.83 a remarkable
change is observed:

The grid calculation procedure to develop structured


C-grid topologies;
The time-accurate NavierStokes code for airfoils in
static fixed position;
The time-accurate NavierStokes code coupled with the
two-degree of freedom elastic equations for unsteady
aero elastic investigations.
In the following section buffet and buffet boundaries will
first be investigated by means of the first and second code
component.

Fig. 1. Lift time histories at various Mach numbers, the incidence is kept
constant at = 0.5 , the total pressure is p0 = 0.5 bar.

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

543

Fig. 3. Coordinate system.

Fig. 2. Lift time histories at various Mach numbers, high Mach regime.

The amplitudes of the lift oscillations are reduced considerably and seem even to be damped at increasing time.
This behaviour may be closely related to the occurrence of
a transonic dip as reported frequently in [15]. Buffet oscillations occur within a small band of M 0.02 only. At
lower Mach numbers as well as at higher Mach numbers the
buffet phenomenon does not exist.

4. Transonic flutter investigations


For flutter investigations on a two-degree of freedom
(pitching and plunging) system the aero elastic equations
and the fluid dynamic equations have to be solved simultaneously.
For the pitch and plunge system the following set of aero
elastic equations has to be solved:
Aero elastic equations:
m h + S + gh h + Kh h = L(t),
S h + I + g + K ( 0 ) = M(t),

(1)
(2)

where in Eq. (1) each term has the dimension [N] and in
Eq. (2) each term has the dimension [Nm].
Using the dimensionless terms as defined in the Nomenclature the following non-dimensional set is derived:


1
h M 2

h
h + x + g h h + 4
2
V
= 2

2
M
CL (T ),

(3)



1 2
M 2
1

x h + + g +
( 0 )
2
4
V
M2
(4)
= 2 CM (T ).

With the following abbreviations:


1
c
h = h
, h = h 2 , T = t a /c, etc.
a
a

0 in Eq. (4) corresponds to the zero force torsion spring


incidence, and are the first and second time derivatives
respectively.
The reason for taking the speed of sound a as a
reference velocity in Eqs. (3)(4) is to coincide with the fluid
dynamic equations.
The system of equations (3), (4) is discretized and rearranged such that the displacement h and the incidence
can directly be calculated for the next time level. These values are then used to determine the fluid dynamic quantities,
i.e. lift and moment for this time level and resolve Eqs. (3),
(4) again.
Fig. 3 shows a sketch of the airfoil suspension and the
coordinate definitions used in the calculations.
The overall calculation procedure is as follows:
First step: Calculation of two grids at assumed maximum
and minimum incidences. All grids used for the instantaneous calculations are simple linear interpolations (or extrapolations) of these two grids.
Second step: Calculation of steady start conditions. For
steady calculations a value for the incidence (s ) has to
be estimated. It is known from experiments that a negative
pitching moment is developed changing the incidence from
the neutral incidence of the torsion spring (0 ) in counter
clockwise direction, i.e. nose down. The amount of this
incidence reduction is about 0.7 . With a selected 0 the
incidence s for the steady start conditions is such defined.
It should be pointed out here that this procedure is done
only to be not too far away with steady start conditions from
the final solution. Taking an arbitrary s may only cause
a larger amount of time steps necessary to derive a final
converged solution.
Third step: The final step includes the unsteady calculation procedure solving the aero elastic system (3), (4). The
output of this calculation is beside all fluid dynamic quantities also the time histories of h and .
In the present paper the Mach number will be kept fixed
(M = 0.77) but the neutral incidence of the torsion spring 0
will be varied systematically.
Tables 1 and 2 include the necessary input data for the
elastic system, [4]. The damping coefficients have been set

544

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

Table 1
Input data
x []

[]

f [Hz]

[kg m3 ]

0.086

0.3686

43.61

411.71

Table 2
Input data
T0 [K]

p0 [bar]

g h a [m s1 ]

g a [m s1 ]

310

0.45

0.893

0.0229

For M = 0.77 the shock locations for both the lower and
upper shocks are matched quite well however in the rear
loading area still some deviations occur. Increasing to M =
0.78 is already too much. The best fit is obviously found in
between at probably M = 0.775 (not indicated). This can
also be verified from the lift and moment coefficients: At
M = 0.77 the lift is still too high, for M = 0.78 it is too
low compared to the experimental value. Similar results are
obtained with the pitching moment (see table in Fig. 4).
Of considerable interest is the cf -behavior for the different numerical results:
For M = 0.77 shock-induced separation occurs on the
upper surface. No separation exists on the lower surface.
At M = 0.78 no much changes can be detected on the
upper surface, however on the lower surface shock induced
separation occurs now in addition.
This behaviour seems to be typical for the supercritical
airfoil section NLR 7301. The flow behaves very sensitive in
transonic flow environments.
The reason why a best fit of pressures between experiment and calculation is achieved at a slightly higher Mach
number (the incidence was exactly the same) is not known
at the present time. It may be explained by a sort of jet effect occurring along the sidewalls of the long adaptive wall
test section. Further studies of this effect are of considerable
interest.
4.2. Unsteady calculations

Fig. 4. Steady results.

to zero in all present calculations, the total pressure p0 has


also been unchanged.
The chord c of the wind tunnel model is 0.3 m.
4.1. Steady calculations
First calculations have been carried out in a steady
flow regime and comparisons of the numerical data with
experimental results from the TWG with adaptive wall test
section [4] are presented in Fig. 4.
The experimental data have been measured at M = 0.76
at an incidence of = 0.5 .
The aim of this study was to find the set of input
parameters for the numerical calculations, which deliver a
best fit of the measured pressure distributions.
A rather fine grid with 361 71 grid points has been used
for the present calculations. The number of time-steps was
30 000 (T = 0.002, T = ta /c), with a well converged
steady solution obtained after 60 chord length.
Trying the parameters ( = 0.5 , M = 0.76) of the
experiment first, the calculated pressures show some slight
deviations, which indicate that the Mach number is too low
to get a good fit.
Therefore the Mach number has been increased step by
step to first M = 0.77 and then to M = 0.78 respectively.
The results of these calculations are also included in Fig. 4.

Beside Mach number and total pressure p0 which both


are kept constant for the present investigation the incidence
of the airfoil is of major concern. Different to fixed wings
where the incidence is input parameter, the elastic suspension in torsion of the wing allows an adjustment of the incidence only without wind. The corresponding parameter is
0 as mentioned before. 0 is also input parameter into the
aero elastic equations (see Eqs. (3) and (4)).
At wind-on conditions the airfoil will be rotated from
its initial incidence position nose-downwards due to a
negative (counter clockwise or nose-down) pitching moment
developing on this airfoil.
Fig. 5 shows calculated pitching moments versus incidence. The slope of this curve is known as the static
stability of the configuration, with unstable conditions at
dCM /d > 0 and stable conditions at dCM /d < 0. In the
stable case the increase of the incidence develops a nosedown pitching moment, which tries to reduce the incidence.
The moment variation with respect to incidence is strongly
non-linear and has a minimum at about = 0.25 . The unsteady calculations have been oriented with respect to the
CM -variation in Fig. 5 in the following way: the numerical calculations are started with steady start conditions. The
incidences s for these calculations are estimated from the
selected 0 by subtracting an angle of about 0.50.7 from
the neutral torsion spring incidence. The amount subtracted
is arbitrary and has been oriented on the experimental data.

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

545

The aim of the unsteady calculation procedure is, to find


solutions in the following three regions of Fig. 5:

(3) At approximately the minimum of the CM versus


curve.

(1) Along the negative branch of dCM /d;


(2) Along the positive branch of dCM /d;

The number of time steps used for a single run on the DLRSX5 supercomputer was 60 000. With the airfoil chord of
0.3 m this number of time steps corresponds to an elapsed
time of 0.142 s. A total of 7 runs have been carried out as
a minimum for the various cases investigated corresponding
to a physical time of about 1 s.
Compared to the experiments this time seems to be
rather small, however the tendencies and flow details to be
investigated are already included in the results representing
about 30 cycles of oscillations.
It is assumed that increasing the calculation time further
may result in accumulating numerical errors, which may
cause either a breakdown of the calculation or lead to
erroneous results.
In the following the calculation will first be shifted into
the stable region of Fig. 5.
The neutral torsion spring incidence is chosen to be 0 =
0.5 .
The steady start condition is s = 0.0 .
Fig. 6 shows the time-histories of lift-, drag and pitchingmoment as well as the incidence and vertical displacement
variations respectively.

Fig. 5. Static stability curve.

Fig. 6. Time-histories of lift-, drag- and pitching-moment (left) and incidence and vertical displacement variations (right) for 0 = 0.5 (s = 0.0 ).

546

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

It can be clearly detected from Fig. 6 that after about


2 s of elapsed time, limit cycle oscillations occur. The final
displacements shown in Fig. 6 are:
max = 4.53,
min = 4.71,
mean = 0.091,
hmax = 16 mm,
hmin = 17.7 mm,
hmean = 0.6 mm.
The dimensionless time period of oscillation is T = 26.1
corresponding to a frequency of 32.7 Hz.
The incidence variation in Fig. 6 shows also the start of
the calculation: The incidence is set to the initial condition
s = 0.0 . Then at time T = 0 the system is set free to
oscillate in both pitching and plunging modes respectively.
The final mean incidence of the calculation is located
at the stable branch of the static stability curve of Fig. 5
and marked as attenuated. A further calculation (not
displayed) has been carried out with a resulting mean closer
to the minimum. However the results still show attenuation
to high values similar to the results in Fig. 6.
High amplitudes of oscillations have already been reported in [17] where also amplitudes of incidence variation
of about 4 have been calculated. In the experiments LCOs
with such high levels in pitch and plunge response have not
been documented. One reason may be that in cases where
in the experiments the incidence is increased beyond about
2 amplitude the mechanical flutter brakes of the test stand
has been activated in order not to destroy the model and its
suspension.
A closer look at the physics of the flow during oscillation
is displayed through Fig. 7, where both pressure- and skinfriction distributions are plotted for a representative period.
The period has been subdivided into 5 constant time
steps; the solid curves display the situations at the minimum and maximum time instants respectively. From both
pressure- and skin-friction distributions it is obvious that
both shock waves on lower and upper surface of the airfoil are involved in a strong forward and backward movement. This movement takes place in anti-phase: if the upper
surface shock has its maximum strength, the lower surface
shock almost disappears and vice versa.
If the shocks strengthen, shock induced separation occurs
(see cf -plot in Fig. 7). The shock induced separation opens
all the way to the trailing edge over part of the cycle. This
phenomenon also appears phase shifted on both upper and
lower surface of the airfoil.
The details discussed so far do coincide closely with
the flow situation, which has been found during buffet (see
Section 3). Buffet oscillations are also characterized by a
phase shifted strong oscillation of both upper and lower
surface shock waves. Due to a complete separation behind
shocks it comes to an interaction between both shocks via

Fig. 7. Pressure- and skin-friction distributions at representative period of


Fig. 6.

the airfoil trailing edge. Animation of the flow shows that


separation reaches the trailing edge and influences the other
side of the airfoil respectively.
The second unsteady calculation has been carried out
along the unstable branch of the static stability curve (Fig. 5)
with 0 = 1.9 and s = 1.5 .
The corresponding results are displayed in Fig. 8. Note
that the scaling is changed compared to Fig. 6. The oscillations first of all start with high amplitudes but then are
damped out rather fast. The number of time steps has been
limited in this case because the trend of the calculation is
already obvious after about 0.5 s running time.
The mean values for incidence and vertical displacement
of this case are:
mean = 1.23,
hmean = 1.14 mm.
The mean incidence has been indicated in Fig. 5 and marked
as damped.
In this case the zero neutral incidence of the torsion spring
shifts the solution into a range, where the corresponding
fluid dynamic response leads to stable conditions.
It should be pointed out that the initial incidence s is
obviously chosen too high. The final mean value is 1.23 .
Therefore considerable oscillation amplitudes occur at the
beginning of the calculations.
So far the responses show cases of attenuation and damping of oscillation amplitudes. In the case of attenuation the
system finally arrives at limit cycles however the amplitudes
of modes as well as aerodynamic forces and moment exceed
by far the values, which have been observed in the measurements.
The question now arises: do low amplitude level LCOs
occur somewhere between the two extreme cases discussed

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

547

Fig. 8. Time-histories of lift-, drag- and pitching-moment (left) and incidence and vertical displacement variations (right) for 0 = 1.9 (s = 1.5 ).
Table 3
Main results of Fig. 9 compared to experimental data

Calculation
Experiment [15]

0 [ ]

mean [ ]

hmean [mm]

/2 [ ]

h/2 [mm]

f [Hz]

 [ ]

Time-steps

1.35
1.92

0.642
1.30

1
0

0.29
0.30

1.0
1.0

32.7
34.0

161.8
174

420 000

so far and is there probably a better match with the


experimental observations?
Taking 0 = 1.35 and s = 0.65 the time histories
displayed in Fig. 9 are obtained.
Note again that the scaling is completely different compared to Fig. 6.
The final mean incidence for this case is mean = 0.641
and hmean = 1 mm, the final mean incidence is therefore
very close to the estimated start value. As can be seen in
Fig. 9 the amplitudes of oscillations are not very much
different from the later converged values.
The results of Fig. 9 clearly show limit cycles in a range,
which has also been found in the experiments, [15].
The main calculated results are summarized in Table 3,
with /2 and h/2 as the amplitudes of incidence
and vertical displacement respectively.  is the inter
modal phase shift between pitch and plunge response.

The corresponding experimental data [15] are included for


comparison.
It is obvious in the different plots of Fig. 9 that the results
stay stable for quite a long time or number of time steps
(420 000 in the present case). However it is not completely
clear whether the oscillations remain stable for all time
without any changes.
It seems that the present stage describes an indifferent
flow situation where either damping or attenuation may
occur at a later time.
Therefore the result of Fig. 9 is indicated in Fig. 5 with
the term indifferent as the limit between attenuation and
damping.
The comparison of the results with experiments (Table 3)
shows very good correspondence for most of the different
data. Exceptions are the neutral incidence of the torsion
spring and the final mean angle of incidence which are 0 =

548

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

Fig. 9. Time-histories of lift-, drag- and pitching-moment (left) and incidence and vertical displacement variations (right) for 0 = 1.35 (s = 0.65 ).

1.92 and mean = 1.3 respectively. These higher values


must obviously be attributed to the influence of wind tunnel
wall interference effects of the perforated wall test section
used in the experiments.
If the main data between calculation and measurement
coincide well which is the case here, a closer look into the
flow details is worthwhile since the numerical results include
all instantaneous details.
As in the previous cases the instantaneous pressure- and
skin-friction distributions will be given a closer look.
Fig. 10 shows first of all the overall pressure and skin
friction data as derived from a representative period of
oscillation. Completely different from Fig. 7 where strong
oscillations on both sides of the airfoil occur, the shock
on the upper surface is moving alone with small amplitude
whereas the shock on the airfoil lower surface is almost
fixed. Although some differences occur in detail, similar
shock behaviour can be detected from the experimental data,
Fig. 11.
Fig. 11 shows a series of instantaneous pressure distributions for one representative period of oscillation of a flow
case closely corresponding to the numerical case of Fig. 10.

It is obvious from this plot that the upper shock wave is


moving in time over a small distance (smeared black area)
whereas the shock on the lower surface is represented by a
narrow line indicating that it is not moving. The white curves
represent the average pressure distributions for the period.
Comparing Fig. 10 (calculation) and Fig. 11 (experiment)
the details of the pressure curves show some differences,
however the small amplitude motion of the upper shock
wave and the almost fixed lower shock wave are present in
both investigations.
Very important is the behaviour of the skin friction
distributions in Fig. 10 (lower): shock induced separation
is found only on the upper surface of the airfoil (note that
negative skin friction on the upper surface and positive
skin friction on the lower surface indicate separation). This
behaviour shows a different type of limit cycle oscillations:
in case of high amplitude LCOs, Fig. 7, both shocks have
been involved. In the present case the upper shock is moving
almost alone.
Fig. 12 gives a closer look at the calculated shock regions
of both lower and upper surface: The numbering refers
to distributions obtained at constant time steps during one

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

Fig. 10. Pressure- and skin-friction distributions at representative period of


Fig. 9.

549

Fig. 12. Detail of Fig. 10 at shock position.

5. Conclusions
Numerical investigations have been carried out for the
NLR 7301 supercritical airfoil section in the transonic flow
regime by means of a 2D time accurate NavierStokes
solver. The aim of this study was twofold:

Fig. 11. Measured instantaneous pressure distributions from one period of


oscillation [15].

representative period of oscillations. 1 and 6 are related


to max , 4 is related to min . The lower surface skin
friction distributions do not cross the zero line although
they are almost touching it. The upper surface skin friction
distributions cut the zero line over a part of the oscillation
cycle and it comes to shock induced separation, which is
partly extended beyond the trailing edge. It is further seen
in Fig. 12 that incidence and shock position are not in phase.
The phase relationships between the movement of the
airfoil and the fluid dynamic response can better be studied
in flow animations. From these animations it is found that
the upper shock is moving with shock induced separation
matching the trailing edge separation over part of the cycle,
whereas the shock on the lower surface shows only a small
movement without a sign of separation.

(1) Investigation of the buffet boundary by systematic


variation of Mach number.
(2) Investigation of the flutter behaviour of the airfoil by
coupling the fluid flow solver with the two degree of
freedom pitching and plunging elastic equations. Here
the Mach number has been kept constant whereas the
zero force torsion incidence 0 was used as the main
parameter to be changed.
The following results have been found during the present
study:
Increasing the Mach number in the transonic flow
regime step by step, the buffet boundary was found.
Well-established buffet oscillations occur beyond the
buffet boundary. A further stepwise increase of the
Mach number led to a second boundary. Beyond this
second boundary the buffet oscillation amplitudes are
almost vanishing or damped out, a behaviour, which has
also been found in experiments.
For flutter investigations the parameter 0 , i.e. the
neutral incidence of the torsion spring has been varied
systematically. Mach number and total pressure were
assumed to be constant. Three flow regimes have been
distinguished:

550

W. Geissler / Aerospace Science and Technology 7 (2003) 540550

Attenuated oscillations up to high amplitudes for


small 0 ;
Damped oscillations for high 0 ;
Indifferent case with low amplitude oscillations in
between.
The three areas of 0 have been indicated in the static
stability curve: CM versus .
In the case of the high amplitude response (small 0 )
LCOs have been calculated with amplitudes of more than
4 . Similar results are reported in other studies however they
have not been documented in experiments. One explanation
may be that the experimental test stand was not able to
handle such high amplitudes.
For the high 0 regime a damped response of the aero
elastic system is detected.
With these two results in mind the search for a case
of small amplitude response similar to experimental data
was successful in the vicinity of the minimum of the static
stability curve.
For this case the amplitudes of both pitching and plunging
modes are limited to about 0.5 and 1 mm respectively.
The oscillation frequency also coincides closely with the
experimental result.
Although the present results are very promising there still
remain problems, which have to be looked at in the future.
Of major concern is the application of turbulence modelling
in unsteady and separated flows. Although considerable
progress has been achieved in recent years it must be kept
in mind that turbulence modelling for the present kind of
flow problems remains a formidable task.
The present application of the SpalartAllmaras turbulence model leads to quite successful results. However it
was found for other supercritical airfoil sections with only
one shock wave on the upper surface, that a buffet boundary could not be found with the SA-model, [7]. The shock
remained in a fixed position if the incidence was increased
step by step. Experiments for this case however have clearly
shown a buffet boundary and well-defined buffet oscillations
beyond this boundary.
The application of a more sophisticated turbulence model,
i.e. the kSST two-equation model was able to predict a
buffet boundary in this case.
It may be worthwhile in future studies of the present
flutter problem to use more complex turbulence models as
well.

References
[1] B.S. Baldwin, H. Lomax, Thin layer approximation and algebraic
model for separated turbulent flow, AIAA Paper 78-257, 1978.
[2] R. Beam, R.F. Warming, An implicit factored scheme for the compressible NavierStokes equations, AIAA J. 16 (4) (1978).
[3] B.M. Castro, K.D. Jones, J.A. Ekaterinaris, M.F. Platzer, Analysis of
the effect of porous wall interference on transonic airfoil flutter, in:
31st AIAA Fluid Dynamics Conference & Exhibit, Anaheim, CA,
2001.
[4] G. Dietz, Privat communication, 2001.
[5] W. Geissler, Instationres NavierStokes Verfahren fr beschleunigt
bewegte Profile mit Ablsung (Unsteady NavierStokes code for
accelerated moving airfoils including separation), DLR-FB 92-03
(1992).
[6] W. Geissler, M.S. Chandrasekhara, M.F. Platzer, L.W. Carr, The
effect of transition modelling on the prediction of compressible deep
dynamic stall, in: The Seventh Asian Congress of Fluid Mechanics,
Chennai (Madras), India, 1997.
[7] W. Geissler, S. Koch, Adaptive airfoil, in: IUTAM Symposium
Transonicum IV, DLR-Gttingen, Germany, 2002.
[8] W. Geissler, L.P. Ruiz-Calavera, Transition and turbulence modelling
of dynamic stall and buffet, 4th International Symposium on Engineering Turbulence Modelling and Measurements, Frantour&CCAS,
Porticcio-Ajaccio, Corsica, France.
[9] W. Geissler, H. Sobieczky, H. Vollmers, Numerical study of the
unsteady flow on a pitching airfoil with oscillating flap, in: 24th
European Rotorcraft Forum, Marsseilles, France, 1998, Paper AE09.
[10] W. Geissler, M. Trenker, H. Sobieczky, Active dynamic stall control
studies on rotor blades, in: Research and Technology Agency, Spring
2000 Symposium on Active Control Technologies, Braunschweig,
Germany, 2000.
[11] P. Mackrodt, Windkanalkorrekturen bei Messungen an zweidimensionalen Profilen im Transsonischen Windkanal, Gttingen, ZFW 19
(1971) 449454.
[12] J.B. McDevitt, Supercritical flow about a thick circular arc airfoil,
NASA TM-78549, 1979.
[13] F.R. Menter, Improved two-equation k-omega turbulence models for
aerodynamic flows, NASA TM-103975, 1992.
[14] G. Schewe, H. Deyhle, Experiments on transonic flutter of a twodimensional supercritical wing with emphasis on the non-linear effect,
in: Proceedings of the Royal Aeronautical Society Conference on
Unsteady Aerodynamics, London, UK, 1996.
[15] G. Schewe, A. Knipfer, H. Mai, G. Dietz, Experimental and
numerical investigation of non-linear effects in transonic flutter, DLR IB 232-2002 J 01 (2002), DLR-Institute of Aeroelasticety, http://www.ae.go.dlr.de/exp/pub/DLR_IB_232_2002J01; Nonlinear effects in transonic flutter with emphasis on manifestation of
limit cycle oscillations, Appears in Journal of Fluid & Structures.
[16] P.R. Spalart, S.R. Allmaras, A one-equation turbulence model for
aerodynamic flows, AIAA-Paper 92-0439, 1992.
[17] S. Weber, K.D. Jones, J.A. Ekaterinaris, M.F. Platzer, Transonic flutter
computations for a 2-D supercritical wing, AIAA Paper 99-0798,
1999.

You might also like