You are on page 1of 13

48th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference<br>15th AIAA 2007-1889

23 - 26 April 2007, Honolulu, Hawaii

AIAA-2007-1889

Aeroelastic Scaling and Optimization


of a Joined-Wing Aircraft Concept
P. Pereira, L. Almeida and A. Suleman
IDMEC-Instituto Superior Tecnico
Lisbon, Portugal
V. Bond, R. Canfield and M. Blair
AFRL/Air Force Institute of Technology
Wright-Patterson AFB, OH

3rd AIAA MultiDisciplinary


Design Optimization
Specialists Conference
23-26 April 2007
Honolulu, Hawaii

Copyright © 2007 by A. Suleman, P. Pereira, M. Blair, R. Canfield and V. Bond. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
AEROELASTIC SCALING AND OPTIMIZATION OF A JOINED-WING
AIRCRAFT

P. Pereira, L. Almeida and A. Suleman


IDMEC-Instituto Superior Tecnico
Lisboa, Portugal,
Email: suleman@ist.utl.pt

R. Canfield, V. Bond, M. Blair


AFIT-WPAFB/AFRL
Dayton, OH, USA
E-mail : robert.canfield@afit.edu

Abstract
An aeroelastic scaling method is described to develop a scaled wind-tunnel test model for a joined-wing
aircraft. The scaling technique aims to reproduce the aeroelastic equations of motion of a full-sized
baseline aircraft configuration using a low-cost wind-tunnel model. The scaled model was developed by
matching flight Mach numbers while maintaining a constant dynamic pressure scale factor and matching
the reduced natural frequencies. Ground vibration tests were conducted and the computational and
experimental results between the baseline configuration and the scaled wind tunnel model were in good
agreement.

Introduction L length
T,t time
Flutter clearance is critical in developing new
g gravitational acceleration
aircraft configurations and typically involves the
E Young´s modulus
development and testing of dynamically and
Fr Froude number
aeroelastically scaled wind tunnel models prior to
q dynamic pressure
full-scale prototype development. However,
p total pressure
experimental aeroelasticity is expensive and typically
γ specific heat ratio
involves flight testing of the full-scale aircraft.
ρ density
This paper describes how the classical aeroelastic
Subscripts
scaling laws are applied to a scaled model of an
W pertaining to the aircraft (wing)
unconventional joined-wing aircraft configuration.
M pertaining to the wind tunnel model
The model was scaled such that its characteristics
under steady loads match those of the full-scale
Background
aircraft. The model was also scaled to match
normalized natural frequencies. Relevant Design of lifting devices has not changed
assumptions, restrictions, limitations and implications dramatically in the last fifty years. However, a new
of this methodology are also discussed. concept, based on an aerodynamic effect presented by
Prandtl, has been object of study by researchers, since
Wolkovich proposed the joined-wing concept in 1986
Nomenclature
[1]. This concept features diamond-shapes in plan and
b semi span of wing front views and provides advantages such as thin
Z function that depends of time airfoils, low induced drag, high altitude operation,
ω frequencies and a stiffer and potentially lighter wing structure.
k reduced frequency Various studies have documented interesting
α angle of attack design issues related to the joined-wing behavior.
i imaginary number Weisshaar and Lee provided considerable insight into
c chord of the wing the important role of flutter in constrained joined-
U velocity wing design [2]. Blair, Canfield and Roberts [3]
M mass examined the buckling response of a linear and

2
nonlinear fully stressed design, concluding that the the structure. The vibration equation requires mass
buckling mode shape tends to unload the outboard and stiffness distributions be maintained.
wing tip. The mass ratio is a ratio of vehicle mass to mass
of the surrounding airspace. The static aeroelastic
Theory equilibrium equations require that mass ratio and
Froude number be maintained. Dynamic
Aeroelastic similitude between flight vehicles
aeroelasticity equations require that the reduced
and wind tunnel models can be attained by matching
frequency, which represents the ratio of speed
non-dimensional parameters that govern the
associated with the oscillations to the forward speed
aerodynamics, the structure and their coupling [4–6].
of the vehicle, be maintained.
The parameters to be matched are those that are free
For similitude of the aerodynamics, the Mach number
after constraints are applied to the model up to the
and Reynolds number must match and the bodies
number of primary dimensions involved in the
must be geometrically similar. This implies that not
governing equations. For aeroelastic wind tunnel
only must the shapes of the different bodies be the
models no more than three constraints, corresponding
same, but also there must be similarity in their
to the primary dimensions of length, time, and mass,
incidence to the flow and in their static elastic
are applied. The wind tunnel size limitation dictates
deformations. Nevertheless, design, construction and
the length scale as the ratio of the allowable model
testing of aeroelastic wind tunnel model allow
span in the wind tunnel to the actual wing span The
neglecting Reynolds number under certain
wind tunnel’s operating limits determine the speed.
circumstances. The aerodynamic equations state that
The combination of speed and length scales together
the airfoil shape must be preserved, along with the
determines the time scale. The mass scale for
Reynolds and Mach number. The Reynolds number
aeroelasticity is governed by the mass ratio, that is,
establishes the flow behavior in the boundary layer.
the ratio of the model mass to the mass of a
For the scaled model it will usually be much lower
characteristic volume of the air flowing over it. In the
than that of the wing at flight altitude and this is
tunnel used here, there is no control of the air density
expected. The three limitations already imposed in
through the tunnel. Therefore, the ratio of wind tunnel
the model test, those of length, speed and air density,
air density to the air density at altitude for the full-
together with the known viscosity of the air in the test
scale aircraft determines the mass scale. Once these
section, completely determines the Reynolds number
three scales are fixed, the remaining parameters
of the model. According to Molyneux, in order for the
appearing in the relevant physical equations must be
test model to represent full-scaled conditions at
matched in non-dimensional form, when creating the
reasonable air density (altitude) conditions, a
model, in order to duplicate the aeroelastic response.
compressed air tunnel is required. The difficulty
In summary, the three constrained parameters in this
arising from an excessive density scale derives from
study are span, speed, and air density.
the necessity to satisfy the Reynolds number
bM UM "M parameter, which affects the boundary layer.
, , (1)
bW UW "W According to the literature, experience has shown that
The subscripts M and W indicate model and full-scale it is generally of minor importance as far as
wing, respectively. The constrained ratios (1) fix the aeroelastic effects for main lifting surfaces are
scales of concerned. If the Reynolds number requirement is
! the primary dimensions, length, L, time, T, ignored, the density scale becomes a free parameter.
and mass, M.
Therefore, the Reynolds number effects are neglected
LM bM in the current design and testing procedure. Also,
= (2)
LW bW since air density and hence speed of sound cannot be
T M UW bM controlled in the tunnel, Mach number cannot be
= (3) matched. Therefore, compressibility effects are also
TW U M bW
neglected.
! # b &3
M M "M
= % M( (4) Once the equations for the wing motion are in a
MW "W $ bW '
! form that is independent of primary unit size, all other
Once the primary dimension ratios (2–4) are fixed, wings that have the same dimensionless properties
the scaled parameters appearing in the governing under the same dimensionless flight conditions will
equations
! may be determined give the same dimensionless answers. The mass ratio
Aeroelastic similitude produces a long list of must be the same for the model to be aeroelastically
required similarity parameters. For the structural similar to the original wing, as dictated by (4). Any
force deflection equations, similitude requires that the model that has the same mass distribution and the
ratio of stiffness to aerodynamic forces be maintained same flexibility distribution over the same geometry
as well as the distribution of the stiffness throughout will produce the same scaled natural frequencies and
mode shapes as the full-scale aircraft. In addition, the

3
reduced frequency, k, the non-dimensional ratio ! M U M bW
associated with the frequency, ω, of aeroelastic = (10)
!W UW bM
oscillation appearing in the aeroelastic equation must
also be matched.
"b Scaling Methodology
k= (5)
U
The methodology used to develop the scaled models
So, the requirement for reduced frequencies is as follows:
similitude is given by • Calculate the natural frequencies and modes of
! #" b & = #" b & vibration, and static deformations of the baseline
% ( % ( (6) model by Ansys and Nastran.
$ U 'M $ U 'W
• Make the aeroelastic analysis of the baseline
Two approaches for achieving (6) will be model by ZAERO.
examined in this paper. One is to design the scaled • By modal analysis of the baseline model and the
model to! match scaled natural frequencies and natural scale factors, calculate the geometric and
mode shapes, so that in light of incompressible dynamic parameters of the wind tunnel model.
aerodynamics the reduced frequencies should match. • Design the wind tunnel model as well as make
The second approach will be to match both the scaled the static and dynamic analysis by Ansys.
natural frequencies and reduced flutter frequency in • Compare all results of the wind tunnel model
an optimization of the scaled model. Preliminary with the objective parameters obtained by scale
results in the next section were obtained by designed factors and modal analysis of the baseline model.
the scaled model to match the scaled natural • Optimization process to reduce the discrepancies
frequencies only. between objective parameters and the results.
• If the discrepancies are relatively small, make the
Similitude of Froude Number aeroelastic analysis of the wind tunnel model and
There are several types of static aeroelastic compare the results obtained with the results
model programs where the velocity ratio, or Froude obtained by baseline model.
number, can be important. The Froude number • If the discrepancies between the two aeroelastic
determines the ratio of the deflection under steady analyses are small, build the wind tunnel model.
gravitational load to deflections due to aerodynamic If not, back the step associated the design of the
and inertial loads. For a Froude-scaled model, the wind tunnel test.
aerodynamic and elastic forces are in proportion to • Perform the ground vibration tests and wind
the gravitational forces. The gravitational tunnel test.
contribution to the loading is important near zero
velocity. For low speed aeroelastic models Froude
scaling may be important, because the deflection due Modal Analysis Results for the Baseline Wing
to gravitational loading for a model will be significant Model
compared to the aerodynamic loading. MSC.NASTRAN was used to calculate the natural
U frequencies and modes of vibration, and static
Fr = (7) deformations using a linear static analysis. This
gb
software produces the analysis using four basic steps:
Now, one can neglect the variation of the 1. Modeling, in which the structure is subdivided
gravitational acceleration. So, the matching of Froude into an assemblage of discrete volumes called finite
number can be given by elements, and properties are assigned to each
element.
&U # &U #
$$ !! = $$ !! (8) 2. Evaluation of element characteristics, such as
% b " M % b "W stiffness and mass matrices, followed by
With length and speed already fixed as two of the assembling the element characteristic matrices to
three basic model constraints, the Froude number can obtain the assembled or so-called ”global” matrices
be matched only within the limit of the wind tunnel characteristic of the entire structure.
maximum speed. 3. Solution of the system equations for
displacements, natural frequencies and mode
Scale Factors in terms of Parameters of the Wing shapes.
• The Velocity Ratio 4. Calculating other quantities of interest, such as
strains, stresses and strain energy.
UM b
= M (9) The results presented in Fig. 1 illustrate the modal
UW bW
analysis results for composite baselines aircraft based
• The Frequency Ratio on input data provided by Ramussen, Canfield and

4
Blair [7]. Table 1 presents the natural and normalized long and a section 1.2m x 0.8m and can be used in an
frequencies of the baseline model. These results are open or closed configuration. A uniform flow
very important in order to apply the scaling factors velocity with less than 0.8% can be obtained in a
and obtain the corresponding parameters for the wind cubic zone 1.1m x 0.6m x 1.4m. The wing semi-span
tunnel model. must be no more than 1.0m long, in order to maintain
uniform flow over the entire wing.
Table 1: Natural and reduced frequencies of the The resulting geometry the scaled wing model is
baseline model given in Figure 2. Table 3 shows the geometric
dimensions of the scaled model; these values depend
Mode ωW [Hz] ωW [rad/s] only on the ratio bM/bW=0.0310:
1 0.79 0.04
2 2.01 0.10 Table 3: Scaled wing dimensions
3 2.86 0.14
4 4.33 0.20 Baseline Wing Dimensions
5 7.01 0.33 Sib 26.0 m xfa 22.0 m
6 7.31 0.34 Sob 6.25 m zfa 7.0 m
7 9.09 0.43 crf 2.50 m Λ_ib 30 deg
8 11.30 0.53 cra 2.50 m Λ _ob 30 deg
9 12.72 0.60 cm 2.50 m Airfoil FX-60-126-1
10 13.87 0.65 ct 2.50 m
Scaled Wing Dimensions
Sib 0.810m xfa 0.682 m
Aeroelastic Analysis of the Baseline Wing Model Sob 0.190 m zfa 0.217 m
crf 0.077 m Λ_ib 30 deg
The aeroelastic calculations were performed to
determine if flutter or divergence of the scaled model cra 0.077 m Λ _ob 30 deg
reproduce the aeroelastic behavior of baseline wing cm 0.077 m Airfoil FX-60-126-1
model. The aeroelastic simulations were done using ct 0.077 m
ZAERO. Note that the scaled model represents the
behavior of baseline model when the wing is The mass and velocity of the wind tunnel test are
operating at an altitude of 15,240 m. For the baseline calculated using Eqs. (25) and (27), respectively; and
model, we fixed the density in 0.19 kg/m3 and then using " M = 1.22 kg/m3 and "W = 0.19 kg/m3
increased the velocity from 1 m/s to 330 m/s, and for giving MM=0.4326 kg and UM = 31.17 m/s. The
the scaled model we fixed the density in 1.225 Kg/m3 natural frequencies of the wind tunnel test are
and the velocity was changed from 1m/s to 50 m/s. calculated using Eq. (26) and are tabulated in Table 4.
The results obtained for the baseline model are! !
presented the Table 2: Table 4: Natural frequencies of the scaled wing

Table 2: Aeroelastic results of the baseline model Natural Frequencies of the Baseline Wing
Mode ωW [Hz] Mode ωW [Hz]
Baseline Wing Model 1 0.79 6 7.31
Mode f(Hz) V(m/s) Mach Alt(m) 2 2.01 7 9.10
2 2.14 298.69 0.60 15240 3 2.86 8 11.30
1 Flutter
5 5.48 331.79 0.60 15240
4 4.34 9 12.72
5 7.01 10 13.87
Now, The objective is to manufacture an
aeroelastic model to be tested in the wind tunnel. To ↓Natural Frequencies of the Scaled Wing
construct an aeroelastic model, the proper scaling of Mode ωM [Hz] Mode ωM [Hz]
model characteristics is important. Since the model 1 4.49 6 41.52
geometric ratio affects other parameters, it is usually 2 11.54 7 51.67
fixed by consideration of wind tunnel limitations. 3 16.24 8 64.18
The maximum model span that the tunnel can 4 24.26 9 72.26
accommodate fixed the ratio bM/bW, where the 5 39.84 10 78.77
subscripts refer to the model and full scale wing. The
quantity b can be any linear dimension. Here, the Aeroelastic Models
wing span was used.
A closed section horizontal Gottingen-type In order to properly represent the baseline
tunnel is to be used for the wind tunnel tests. It is configuration model by the scaled model, it was
capable of operating between 5 and 70 m/s with decided to match the first five reduced frequencies for
temperature control air stream. The test section is 2m the model. The development of the models was

5
carried out in an incremental order of complexity in Table 6: Final characteristics of the scaled wing after
order to enable us to attain an appreciation of the optimization
design. Therefore, it was decided to start the study
with a simple model, and incremental modifications
Target Scaled Model Error (%)
were introduced with the objective of reaching an
optimal scaled model that is a good approximation to Mass (Kg) 0.43 0.44 1.62
the real structure, and most importantly it has to be f_1 4.49 4.57 1.81
manfacturable. f_2 11.45 11.00 3.95
The model is constituted by 6 spars, 29 ribs and f_3 16.24 17.50 7.74
the shell. The design variables of the optimization f_4 24.26 25.06 3.23
process are the all the rib thicknesses. Figure 3 shows f_5 39.8 39.10 1.84
the internal structure of the scaled model. f_6 41.52 40.58 2.28
The spars are made of balsa wood and they are f_7 51.67 51.35 0.62
located in the leading and trailing edges. Ribs 1 to 6 f_8 64.18 62.89 2.01
are made of stainless steel, and the ribs 7 to 29 are f_9 72.26 74.11 2.56
made of aluminum. Note that selection of the f_10 78.77 79.53 0.96
materials is based on the need to match the reduced
frequencies. The shell is made of polycarbonate and These results are the optimal values obtained by
oracover. The polycarbonate is used in the lower the iterative process. One can see the progressive
surface and the oracover is used on upper surface of evolution of the optimization process. One can also
the scaled model. In the tip of the scaled model, a see that the major error (7.74%) is associated at the
connection is located between the front and the rear third natural frequency. The remainder errors are
spars. This connection is made of stainless steel. lower. The average error between the theoretical
The initial values of the parameters were value and computational value of the natural
arbitrarily assigned and then a sensitivity analysis was frequencies is 2.7%. With these observations, one can
carried out to interpret the results followed by a conclude that the model is a good representation in
manual optimization process. With this information a terms of dynamic representation of the baseline wing.
selection of the optimal scaled model was possible. In conclusion, the optimization of the wing has
With these parameters, we obtained the following enabled the design and manufacture of an optimal and
characteristics of the scaled wing: manfacturable aeroeastically scaled wing with low
errors associated with mass and natural frequencies
Table 5: Initial characteristics of the scaled wing calculations. The aeroelastic results associated to the
before optimization baseline wing are listed in Table 2. With these values
and the scale factor presented by Eq. (26), one can
Scaled Error obtain the objective values that the scaled model
Target
Model (%) should have. These values are listed the following
Mass (Kg) 0.43 0.38 12.14 table: The results obtained with the scaled model by
f_1 4.49 4.79 6.64 ZAERO are presented in Table 7:
f_2 11.45 11.48 0.27
f_3 16.24 22.50 38.56 Table 7: Target and computational values of the
f_4 24.26 25.46 4.96 scaled model
f_5 39.8 41.91 5.20
f_6 41.52 43.69 5.21 Target Values
f_7 51.67 58.40 13.03 Mode f(Hz) V(m/s) M h(m)
f_8 64.18 74.06 15.39 2 12.15 52.60 0.10 0.0
V_f
f_9 72.26 80.81 11.84 5 31.12 58.43 0.10 0.0
f_10 78.77 82.53 4.77 Computational Values
Mode f(Hz) V(m/s) M h(m)
Table 5 shows that the major error is associated 2 22.90 50.33 0.10 0.0
to the third frequency. Therefore, the new objective V_f
5 37.13 58.72 0.10 0.0
function is the third frequency, while the other
frequencies and the weight become the constraints.
Table 8 shows the discrepancy between the target
Figures 4 to 7 illustrate the iteration history for the
values and the computational values of the scaled
design variables, weight and frequencies during the
wing between the design target and the test article
iterative process. The final characteristics of the
models.
scaled model after this optimization process are listed
in Table 6.

6
Table 8: Discrepancies between the target and test Table 9: Comparison between computational and
article values experimental natural frequencies
Scaled Wing Model Experimental Computational Discrepancy
(Hz) (Hz) [%]
Mode ∆f (%) ∆V (%) f_1 4.40 4.57 3.77
1 Flutter 2 88.4 0.54 f_2 11.00 11.00 0.03
5 19.33 0.88 f_3 16.20 17.50 7.42
f_4 25.06 25.06 0.02
One can see that these values are in good f_5 34.50 39.10 11.8
agreement in terms of the difference in velocity.
f_6 40.50 40.58 0.18
Relative to frequencies, it is noted that the scaling
f_7 49.88 51.35 2.87
process used the reduced natural frequencies and not
f_8 65.13 62.89 3.56
the reduced aeroelastic flutter frequencies, as is
f_9 69.00 74.11 6.90
usually documented in the literature. The next stage
of the research will be to compare both techniques f_10 81.00 79.53 1.85
and validate the approach proposed in this paper.

Ground Vibration Tests Concluding Remarks

Reproducing the aeroelastic response of the full-


In order to validate the computational results for
scale vehicle in a wind tunnel model required several
the scaled wing model, and to make some detailed
non-dimensional parameters to be identical for both
system identification, impulsive response vibration
the vehicle and the scaled model. Similitude
tests were conducted with the wing mounted in the
requirements for testing implied matching Froude
vertical position and inside the wind tunnel. These
numbers and the ratio of stiffness to aerodynamic
impulsive tests were made using instrumentation
forces. Using the proposed methodology, it was
described next.
possible to reproduce the scaled natural frequencies
The impulse response was measured to obtain the
and scaled flutter speed of the full scale vehicle, but
frequency response (amplitude vs. frequency or
not the reduced frequencies, using a low-cost scaled
power vs. frequency) by spectral analysis. The
wind tunnel model. The ground vibration tests
complete system frequency spectrum is then excited,
exhibited excellent correlation between the baseline
and all of its natural frequencies are highlighted,
and scaled models. Current efforts are being focused
showing peak amplitudes in the frequency response.
in the wind tunnel testing and flutter speed and
To make sure all the vibration modes were adequately
frequency determination.
detected several run tests were performed, each for a
distinct position of the piezoelectric accelerometer.
Acknowledgements
Figure 8 shows the different positions used to find the
natural modes of the scaled wing model. Table 9
presents the natural modes obtained by OROS system This research was supported by the U.S. Air Force
and the respective discrepancies relative to the (AFRL-WPAFB) under the EAORD Grant/
computational values. Cooperative Agreement Award No. FA8655-05-1-
The discrepancies between the natural 3076.
frequencies for both models are relatively small,
which means that the scaled wing model is a good References
representation of the aeroelastic behavior of the
baseline wing. In terms of mass, the scaled wing 1. J. Wolkovitch, "The Joined Wing: An
model has 0.490 Kg. This value presents a Overview," J. of Aircraft, March 1986, 161-178
discrepancy of 11.44% relative to computational mass 2. T.A. Weisshaar and D.H. Lee, “Aeroelastic
model, possibly due to the glue required to make the Tailoring of Joined-Wing Configurations”,
scaled wing test article. AIAA2002-1207, 43rd AIAA/ASME/ASCE/
AHS/ASC Structures, Structural Dynamics and
Materials Conference, Denver, CO, April 2002.
3. Maxwell Blair; Robert A. Canfield; Ronald W.
Roberts, “Joined-Wing Aeroelastic Design with
Geometric Nonlinearity,” Journal of Aircraft,
Vol. 42, no.4, 2005, pp. 832-848
4. Bisplinghoff, Ashley and Halfman,
“Aeroelasticity,” Addison-Wesley Publishing,
Reading, Massachusetts, 1955.

7
5. Molyneux, W. G., “Aeroelastic modeling” RAE
Technical Note 353, March 1964.
6. J. Heeg, C. V. Spain and J.A. Rivera, “Wind
Tunnel to Atmospheric Mapping for Static
Aeroelastic Scaling”, 45th AIAA/ASME/
ASCE/AHS/ASC Structures, Structural
Dynamics and Materials Conference, Palm
Springs, California, Apr. 19-22, 2004.
7. C. Rasmussen and R. Canfield; M. Blair,
“Joined-Wing Sensor-Craft Configuration
Design,” AIAA Paper 2004-1760, 45th
AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics and Materials Conference,
Palm Springs, California, Apr. 19-22, 2004

8
VIBRATION MODELS OF THE COMPOSITE STUCTURE
Mode 1, ω=0.79092Hz Mode 2, ω=2.0163Hz

Mode 3, ω=2.8602Hz Mode 4, ω=4.3354Hz

Figure 1: Modes of vibration 1 –4 for the Baseline Model

Figure 2: Geometric characteristics of the joined-wing

9
Figure 3: Geometry of the Scaled Wing

10
Figure 4 - Weight and Frequencies vs. Iteration Number

11
Figure 5 – Error for the Frequencies vs. Iteration Number

Figure 6 - Design Variables Associate of Main/Aft Wing vs. Iteration Number

Figure 7 - Design Variables Associate of Body 3/Body 4 vs. Iteration Number

12
Figure 8: The scaled wing model showing the position of the accelerometers

13

You might also like