You are on page 1of 18

12th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference AIAA 2008-5875

10 - 12 September 2008, Victoria, British Columbia Canada

Optimisation of a Scaled Sensorcraft Model with Passive


Gust Alleviation
S. Miller1
University of Manchester, Manchester, M13 9PL, UK.

G.A.Vio2 and J.E. Cooper3


University of Liverpool, Liverpool, L69 7ZF, UK.

and

O. Sensburg4
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Cranfield University, UK.

High altitude long endurance UAVs such as sensorcraft are extremely flexible and
consequently are susceptible to excessive gust loads; therefore it is desirable to incorporate
some form of gust load alleviation system into the air vehicle design. If successful, this could
result in a significant weight reduction. This paper describes part of an EOARD / AFOSR
and ESF supported research project aiming to design, manufacture and test a wind tunnel
model of a sensorcraft structure incorporating a passive gust alleviation device. Various
elements of the project are described, including initial simulated validation of the device,
design and test of a concept prototype and aeroelastic scaling of the sensorcraft wind tunnel
model. Conclusions are made as to the effectiveness and feasibility of incorporating the gust
alleviation device on an aeroelastically scaled wind tunnel model.

I. Introduction
There is much recent interest, particularly at AFRL Air Vehicle Directorate [1-3], in the development of
unmanned High Altitude Long Endurance (HALE) or Sensorcraft air vehicles that are able to provide a 360o sensor
coverage whilst maintaining moderate stealth characteristics. The sensor requirement has naturally led to a re-visit
of the pioneering work into joined-wing aircraft by Wolkovich [4], Gallman et al [5-7] which has been relatively
recently surveyed by Livne [8].
Typical proposed joined-wing sensorcraft structural lay-outs are shown in Figure 1 where it can be seen that the
aeroelastic behaviour is likely to be very different from conventional aircraft configurations.

Figure 1. Examples of Proposed Joined Wing Sensorcraft Aircraft

Much of the work on sensorcraft has been focused upon the optimisation of the aircraft structure. The outer wing
of the sensor craft design leads to the high aspect ratio, which is very favourable for reduction in fuel consumption

1
Research Student, School of Mechanical, Aerospace and Civil Engineering
2
Research Assistant, Department of Engineering. MAIAA.
3
Professor of Aerostructures and Aeroelasticity, AFAIAA
4
Visiting Professor
1
American Institute of Aeronautics and Astronautics

Copyright © 2008 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
and range extension. However, it produces high bending moments stemming from manoeuvres or gusts. Due to its
significant flexibility there have been concerns about:

• the ability to maintain a shape that does not affect the sensor performance
• the loads that may result from gusts
• the need to use non-linear static and dynamic aeroelastic analysis in order to account for the large
geometric deflections.

One of the critical design cases for joined-wing designs is the buckling of the rear wing structure. Previous work
[2] has shown that non-linear buckling analysis is required in order to estimate accurately the deflections and
resulting loads that occur. Linear analysis for buckling under critical gusts loads significantly increased the wing
optimized structural weight, almost doubling it, and when non-linear analysis is used, the wing tip deflection
increases and the optimized wing weight increases significantly again.
The use of some form of gust load alleviation system is therefore very desirable, as this should lead to a
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

significant reduction in structure, and hence weight, that is required. One possible solution is the design of active
load alleviation systems using all of the control surfaces; however, such an approach is complex, requiring the
avionics for such a system to be carried on the sensorcraft, and a certain amount of system redundancy must be
included to allow for system failures.
Previous work under funding from the EOARD [9] has investigated the use of All-Moving Vertical Tails instead
of the conventional fin / rudder arrangement as a means for increasing the aeroelastic effectiveness of vertical tail
structures. It has been shown that this could lead to reduced sizing of vertical tail surfaces with the corresponding
reductions in drag, mass, loads and radar signature.

Figure 2. Graph showing Fin Efficiency vs. Attachment Position for Vertical Tail Model

Figure 2 shows Finite Element and experimental efficiencies for a vertical tail model for a range of different
attachment positions, the left hand side of the graph representing an attachment point that is forward whilst
attachment positions further aft are on the right hand side of the plot. It can be seen that the further aft the
attachment point the better the effectiveness, and at the most rearward positions an efficiency of greater than one can
be achieved.
However, in terms of achieving gust alleviation, the opposite effect is required. Should an all-movable surface be
attached towards the leading edge, then the effectiveness is greatly reduced and indeed lift can actually be lost. This
work investigates the application of forward attached all-moving devices to high aspect ratio sensorcraft wings with

2
American Institute of Aeronautics and Astronautics
the idea of using the structural flexibility to develop a passive gust loads alleviation device which would have none
of the problems associated with active loads alleviation.
In this paper, various elements of the project are described, including a description of the passive device, the
aeroelastic scaling of the sensorcraft wind tunnel model and simulations to show the most effective design.

II. Gust Alleviation Device

A schematic representation of the device to be investigated in this research work is shown in Figure 3. A wing is
“cut” towards the outer tip and both parts are joined via a torque tube. As the outer part is attached well forward of
the aerodynamic centre, the effect of any vertical upwards gust is to produce a nose down motion in the wing tip,
thereby counteracting the effect of the gust. Due to the gust device being at the wing tip, the load alleviation will
have a significant effect on the resulting loads due to the washout deflection.
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Figure 3. Proposed Load Alleviation Device

With this passive load alleviation device in place it should be possible to avoid the large deflections that result in
the non-linear buckling behaviour described above and therefore to reduce the structural mass. Reference [10]
describes some of the initial studies on the concept. The design task now consisted of investigating the application
of the device on a scaled wind tunnel model.

III. Aeroelastic Scaling of Wind Tunnel Sensorcraft Model


In order to produce an aeroelastically scaled model of the sensorcraft, it is required that the full scale FE and
aerodynamic model is available, however, for this project such a model is not available and instead the University of
Liverpool has been supplied with the natural frequencies, mode shapes, generalised mass and generalised stiffness.

A. Aeroelastic Modelling
Given the standard aeroelastic equation in physical space y

Ayɺɺ + Ey = qQy (1)

where A and E are the mass and stiffness matrices respectively, q is the dynamic pressure and Q is the aerodynamic
force matrix. The system can be transformed into modal space p such that

φ T Aφp
ɺɺ + φ TEφp = qφ TQφp
(2)
A p pɺɺ + Ep p = qφTQφp
where Ap and Ep are the diagonal mass and stiffness matrices respectively. The aerodynamic matrix Q is calculated
from the DLM code in the ZAERO software. This equation can be used to solve the flutter problem. For static
aeroelastic analysis the equation simplifies to

3
American Institute of Aeronautics and Astronautics
Ep p = q φ T Q φ ( p + p 0 ) (3)

where p0 is a vector of wind-off deflections, such as angle of attack.

B. Structural Modelling
The mass and stiffness matrices supplied correspond to a set of mass normalised matrices, giving an identity
matrix for the mass and the stiffness matrix is a diagonal matrix containing the squares of the natural frequencies.

C. Aerodynamic Modelling
The aerodynamic model was generated using a panel method, whose grid covers the planform of the (unknown)
Finite Element (FE) model. The model is symmetric about the X-Z plane. Considering just one half of the model, 5
aerodynamic macro elements were used to describe the entire shape with views of the complete aerodynamic models
presented in figures 4 - 6. Figure 7 shows the location of the 101 structural grid points in comparison to the
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

aerodynamic model. The connection between the front and rear fuselage section has not been modelled due to lack of
data points in this section to estimate the dimensions accurately.
The aerodynamic forces were calculated over a range of reduced frequencies at a series of Mach numbers. Infinite
plate splines were used to transfer the aerodynamic forces to the structure. All non-collinear grid points supplied are
used in the splining process, giving 99 usable points.
Two solution set-ups have been created: 1- Matched Flutter analysis and 2- Static analysis. Flutter analysis
provides the frequency and damping throughout the flight enveloped. A matched solution is used by using the bulk
data card FIXHATM . Trim analysis was set-up for a steady-state level flight condition at a specified angle of attack.
Elastic and rigid deformations are obtained as well as lift and induced drag.
The Flutter analysis was performed at an altitude of 19812 metres (65000 ft) covering the Mach number range
between 0.2 and 0.9. Typical frequency-damping plots are shown in figures 8 and 9 highlight the damping plot near
the flutter crossing.

Figure 4. Macro-elements definition Figure 5. X-Y View

4
American Institute of Aeronautics and Astronautics
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Figure 7. Aerodynamic mesh outline and FE grid


Figure 6. X-Z View points location

Figure 8. Frequency-Damping Plot at 19812 m Figure 9. Damping plot at 19812 m - zoomed

D. Aeroelastic Scaling Approach


The approach used in this work is to define a dynamically scaled aeroelastic model that responds identically to a
full-scale design with respect to chosen scale factors (e.g. geometric scaling, velocity scaling etc). This goal can be
achieved on the basis of a set of non-dimensional aeroelastic equations. As the model internal structure is different
compared to the full-scale structure, it is not possible to use the conventional scaling approach of using equivalent
beams.
In practice, the problem is defined as to find an equivalent aeroelastically scaled model structure given the FE and
aerodynamic models of a full scale structure at specified flight conditions. Although the geometric scaling holds for
the planform of the model, the internal structure is likely to be different, possibly not even made of the same material
or construction technique; similarly, the FE and aerodynamic models corresponding to the scaled structure will also
be different. The wind tunnel chosen to perform the tests will constrain the dimensions of the model and the speed it
is tested at.
The differences between the full scale (subscript s) and model (subscript m - physical and computational) mean
that it is not possible to simply scale the FE matrices. Instead comparison must be made of quantities such as modal
quantities (frequencies and mode shapes), influence coefficients and non-dimensionalised deflections.

E. Test Conditions Scaling


There are three scaling values that are defined by the difference in the required design configuration (i.e. wind
tunnel parameters) and the original flight condition. Note that the differences in the Reynolds number will be
ignored and it is assumed that the flow conditions between the full-size and scaled conditions have no effect upon the
aeroelastic parameters.

5
American Institute of Aeronautics and Astronautics
Geometric Scaling
L s = n g Lm (4)
Velocity Scaling
Vs = n v Vm (5)
Air Density Scaling
ρs = n ρ ρm (6)

Typically ng, nv > 1 whereas nρ < 1. Note that the Equivalent Air Speed could be used in order to eliminate the
need for density information.

F. Mass and Stiffness Scaling Parameters


In order to achieve aeroelastic scaling, there is a need to scale in terms of the stiffness and the mass distributions.
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

For an aircraft model that is simply “shrunk”, or has an internal structure that is exactly the same as the full aircraft
(as per Bisplinghoff approach [11]) then this is not a difficult process. However, in the case that we are considering
whereby the internal structure of the scaled model differs from that of the full size model, at least two of the
following parameters must be chosen as part of the scaling process.

Reduced Frequencies
Rewriting the aeroelastic equations in non-dimensional form involves the transformation into non-dimensional
time. The natural frequencies of the full-scale and model structure at the reference test conditions must be the same,
thus

ωs b s ωm b m Vm bs n
= ⇒ ωm = ωs = s ωs (7)
Vs Vm Vs bm nv

i.e. the reduced frequencies must remain the same. The reduced frequencies depend upon a correct mass and
stiffness distribution.

Mode Shapes
The mode shapes corresponding to the various natural frequencies of the full scale and model structure must be
the same. These shapes will be complex to some degree depending upon the characteristics of the unsteady
aerodynamics terms. As with the reduced frequencies, the mode shapes will depend upon obtaining a correct mass
and stiffness distribution. It would be possible to compare the wind-off mode shapes for the full-size and scaled
models.
Flutter Speed
The reduced frequency will remain the same for the full-size and scaled models, even at the flutter speed.
Consequently it is possible to make a comparison of what the flutter speed should be. However, caution should be
made when simply comparing the flutter speed as this does not infer that the entire aeroelastic characteristics have
been captured.
Static Aeroelastic Deflections
The non-dimensionalised static aeroelastic equations show that equivalent non-dimensionalised deflections at the
reference test conditions must be the same. e.g. for the wing tip at the reference conditions

 x tip   x tip 
 b  = b  (8)
 s  m

The static aeroelastic deflections will depend upon getting the stiffness distribution correctly. One advantage of
comparing the static deflections wind-on is that there an implied scaling of the aerodynamic forces which would not
occur for the wind-off case.

Flexibility Influence Coefficients

6
American Institute of Aeronautics and Astronautics
One method to get around the problem of determining the correct scaling of the aerodynamic loads is to determine
the flexibility influence coefficients for the wind-off case at certain reference points on the structure (as per [12]).
This will give the correct stiffness distribution, however, it does not take into account the changes in the
aerodynamic forces.

Mass Scaling
The traditional approach for mass scaling is to non-dimensionalise the relationship between some reference mass
(e.g. total mass) of the aircraft with that of the air. Typically this can be written as



M
b 
3

=
M
b 
3
⇒ M = nMn
m
s
3
(9)
ρ s ρ m ρ g

Although this approach is fine for determining the overall mass, there are concerns when this is applied to a scaled
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

model when the internal structure differs from the full-scale. Although the mass scaling gives a necessary condition
for the scaling, there is no reference to the mass distribution (unless implied by having the same internal structure)
and consequently it is not a sufficient condition for perfect mass scaling.

Gust Response
Gust velocities will scale in the same way as the airspeed. In order to achieve an equivalent scaled response, the
same maximum normalised deflection to the scaled gust (e.g. “1 – cosine”) at certain reference points of the structure
must be obtained.

 x max tip   x max tip 


(V )
gust s = n v ( Vgust )   =  (10)
 b s  b m
m

Non-Linear Static Aeroelastic Deflections


Although the FE model of a non-linear analysis (e.g. the non-linear geometric deflections that occur on a
sensorcraft) is much more complicated than the linear description shown above, a similar scaling approach can be
used. In the same way as for the linear static aeroelastic deflections, the equivalent non-dimensionalised non-linear
deflections at the reference test conditions must be the same. e.g. for the wing tip at the reference conditions

 x tip   x tip 
 b  = b  (11)
 s  m

and this will enable any non-linear deflection behaviour to be scaled.

Buckling
The critical speed at which buckling (linear or non-linear) will occur simply depends upon the velocity scaling
(V )
buckling s = n v ( Vbuckling )
m
(12)

and the critical buckling shape must be the same in a similar way to the mode shapes.

IV. Optimisation Approach


An initial baseline scaled model was set up with the correct geometrically scaled planform and flight condition.
The model is of such a design to enable efficient manufacture. The aeroelastically scaled model was then be found
by determining what sizing of internal structure, skin thickness and possible extra distributed masses (optimisation
parameters) is required in order to achieve some set of defined non-dimensionalised functions. Initially a Genetic
Algorithm based approach was used.

7
American Institute of Aeronautics and Astronautics
There are a large number of possible target functions that can be chosen from the list above and, as the internal
structure will not be a “shrunken” version of the full-scale sensorcraft, it is unlikely that an exact match will be
obtained for all conditions. Initially only the minimum number of parameters (one mass and one stiffness condition)
required at a single test condition and a few modes were considered.

V. Baseline Finite Element Sensorcraft Model


A baseline finite element model of a Sensorcraft type aircraft was developed for use in the aeroelastic scaling
process and eventual manufacture and test of a Sensorcraft wind tunnel model. The Finite Element model has been
developed in a manner that will be consistent with an efficient manufacturing process and consists of two spars along
each wing, whose dimensions can be varied along the span, attached with ribs whose thickness can be altered as part
of the scaling process. It has been constructed in order to size the wind tunnel model dimensions for spar, ribs and
skin thickness.

A. Geometric Scaling
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

The scaled model will be a factor of 15 smaller than the full scale model (semi-span = 18.285m, length = 23.64m)
and so the semi-span of the scaled (half) model becomes 1.5237 metres and the overall length is 1.97 metres.

B. Finite Element Model


A MATLAB code was created in order to define the “deck” used to build the NASTRAN FE model. The model
consists of the following elements:
• Skin modelled using shell elements. The airfoil shape is currently based on the NACA0012 shape,
but this can be changed to any other profile.
• Two plates, modelled as quadrilateral elements, are included to model the front and rear fuselage
sections.
• Beam elements are used to model the spars and are located at 25% and 75% chord along the span
of the model on both front and rear wing sections.
• The junction area between the two wings is modelled by creating an airfoil decreasing in chord
length. It was assumed that the beams would pass through the rib at the joining between the respective rib
and the spars in the front fuselage. This is the most difficult part of the manufacturing process and may
require a further change in the future once the manufacturing part of the project is considered.
• The connection between front and rear fuselage section is modelled using beam elements.
• The model is fully clamped at the front fuselage section, while the rear fuselage section is free to
move in the Z direction.
• The material properties assigned to all the elements is aluminium alloy.

There are a number of variable parameters to approximate the FE model, i.e. the number of nodes describing the
airfoil, the number of beam elements between each rib. Also the number of elements describing the fuselage sections
and the number of quadrilateral elements describing the skin between each rib are all variable. Initial convergence
studies have enabled the correct element density to be determined. The external skin mesh is shown in figure 10 and
figures 11-13 show various views of the internal structure.
Each rib has its own properties (thickness) that can be assigned via the deck. The fuselage plates, skin of the front
wing, the skin of the rear and the beams have different (geometric) properties that can also be assigned. It is assumed
that the shell elements have a uniform thickness throughout each of the elements. The beams that make up the spars
and also the connection between the front and real fuselage are assumed to be rectangular, so that height and width
only need to be defined between each rib. Manufacturing and computational considerations are likely to mean that
the structure is divided into different regions that have the same geometric properties rather than changing the
properties of every element.

8
American Institute of Aeronautics and Astronautics
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Figure 10. XYZ View Full FE model external surface


Figure 11 XY View internal structure FE model

Figure 12. XZ View internal structure FE model Figure 13 XYZ View internal structure of FE model

VI. Matching of Wind Tunnel Scale Model to Full size Model of Sensorcraft

The Sensorcraft Model was divided into 4 sections: Front Fuselage (FF); Rear Fuselage (RF), Front Wing (FW);
Rear Wing (RW) with the following construction:
• The front fuselage has a solid plate at mid thickness, 8 ribs and skin.
• The rear fuselage has a solid plate at mid thickness, 8 ribs and skin.
• The front wing has 2 beams located at 25% and 75% chord, 11 ribs and skin.
• The rear wing has 2 beams located at 25% and 75% chord, 9 ribs and skin.
• The beams are of rectangular cross-section and have a constant cross-section throughout the span. The
maximum height is dictated by the height of the smallest rib.

The variables for the optimization were:


• FF, RF, FW and RW ribs thickness
• FF and RF block thickness
• FF, RF, FW and RW skin thickness
• FW, RW beams and Fuselage beam section

9
American Institute of Aeronautics and Astronautics
The initial objective was to match the first three reduced frequency (only symmetric modes considered) and static
displacement at the leading edge front wing tip. Two models are considered: One with skin and ribs on fuselage
section as well and a modified version with no ribs or skin over the fuselage section. The cost function utilized was

f = W [1 − M FM − M SM κ (1 − k FM − k SM ) ]
T
1 − δ FM − δ SM
 V f − VMS (13)
1 + V f < VMS
κ = VMS
 1 V f > VMS

where W is a vector of weighting factors, and κ is a penalty function. M refers to the mass, k to the reduced
frequency and δ to the static displacement ratio. The penalty function is introduced to force the flutter speed to be
above the Scale Model flight test speed. Vf is the model flutter speed and VMS is the test flight speed.
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

The subscript FM refers to the Full Scale Model and SM to the Scaled Model. For the following cases the vector
W=[0 0.5 0.5]. In the following table the assigned values to optimisation are listed. Where a number between
brackets is present, corresponds to a different value assigned to the modified model.
Variable Minimum (m) Maximum (m) Divisions (2n) Material
FF skin 0.0001 0.001 3 Nylon
RF skin 0.0001 0.001 3 Nylon
FF block 0.0011 0.015 3 Aluminium
RF block 0.0011 0.015 3 Aluminium
WF skin 0.0001 0.001 3 Nylon
WR skin 0.0001 0.001 3 Nylon
WF beam 25% 0.0011 0.0011 0.015 0.015 3 Aluminium
[width height]
WF beam 75% 0.0011 0.0011 0.015 0.015 3 Aluminium
[width height]
WR beam 25% 0.0011 0.0011 0.015 0.015 3 (4) Aluminium
[width height]
WR beam 75% 0.0011 0.0011 0.015 0.015 3 (4) Aluminium
[width height]
Beam Fuselage 0.01 0.01 0.03 0.03 3 Aluminium
[width height]
Ribs 0.0011 0.015 3 (4) Aluminium

The following parameters were utilized for the models.


Parameter Full Scale Model Model Scale
Mass 1.55365E5 lb
Span 74.9852 ft 1.5237 m
Airspeed (m/s) 130.41 30
Mach Number 0.39 0.0882
Altitude 5000 ft 0m
Static Displacement 29.4887 in

Material properties of Nylon and Aluminium were fixed as follows


Material Young’s Modulus (N/m2) Poisson’s Ratio Density (kg/m3)
Nylon 3x109 0.42 1100
Aluminium 72.9x109 0.33 2700

10
American Institute of Aeronautics and Astronautics
The frequencies of the full scale model at the target airspeed are
Mode Frequency (Hz) Mode Frequency (Hz)
1 0.7881 11 (Sym) 5.9117
2 (Sym) 1.1307 12 6.0483
3 (Sym) 1.2371 13 6.3137
4 1.3665 14 (Sym) 6.5428
5 2.1782 15 7.7942
6 (Sym) 2.7110 16 (Sym) 7.7455
7 2.5988 17 8.2135
8 (Sym) 3.9146 18 (Sym) 9.4046
9 4.1680 19 9.7755
10 (Sym) 4.6459 20 (Sym) 10.1409

The static calculations were performed with a 5 degree Angle of Attack. The reference length used to calculate
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

the reduced frequency is the span. The optimisation loop was fixed with 33 genes plus 7 new genes for each iteration
and 50 generations. Figure 14 shows the convergence for the best 7 genes throughout the generations whereas in
figure 15 the best cost is plotted against the history of the elements of the cost function.
The following agreement with the target values was reached
Parameter Target value Achieved value Achieved value
(Modified Model)
Mass (kg) 20.88 13.3376 23.755
Reduced Frequency 1 0.19817 0.20686 0.20516
Reduced Frequency 2 0.21681 0.3494 0.38112
Reduced Frequency 3 0.47513 0.47543 0.47679
Static Displacement Ratio 0.0328 0.042208 0.03142

The following table gives the resulting dimensions of all the elements.
Element Dimension (m) Dimensions (m) (Modified Model)
FF skin 0.0001 N/A
RF skin 0.0001 N/A
FF block 0.0011 0.0186
RF block 0.0090 0.0300
WF skin 0.0001 0.0005
WR skin 0.0001 0.0005
WF beam 25% 0.0051 0.0031 0.0048 0.0020
[width height]
WF beam 75% 0.0011 0.0090 0.0011 0.0030
[width height]
WR beam 25% 0.0051 0.0130 0.0030 0.0020
[width height]
WR beam 75% 0.0011 0.0031 0.0011 0.0048
[width height]
Beam Fuselage 0.01 0.01 0.01 0.01
[width height]
Ribs FF 0.0090 0.0011 0.0130 0.0150 0.0051 N/A
0.0011 0.0110 0.0031
Ribs RF 0.0071 0.0090 0.0130 0.0051 0.0031 N/A
0.0130 0.0071 0.0051
Ribs FW 0.0130 0.0150 0.0150 0.0130 0.0150 0.0137 0.0112 0.0049 0.0124
0.0090 0.0110 0.0011 0.0031 0.0011 0.0074 0.0036 0.0137 0.0011 0.0150
0.0130 0.0175 0.0187 0.0187
Ribs RW 0.0130 0.0130 0.0150 0.0130 0.0130 0.0124 0.0162 0.0162 0.0124
0.0150 0.0110 0.0071 0.0175 0.0200 0.0187 0.0061 0.0036

11
American Institute of Aeronautics and Astronautics
These results are not as good as expected and work is ongoing to establish why the discrepancies, particularly in
mode 2, occur.
0.98

0.96

0.94

0.92
Cost Function

0.9

0.88

0.86

0.84
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

0.82

0.8
0 10 20 30 40 50 60 70 80 90 100
Generation

Figure 14. Optimisation evolution


1

0.95

0.9

0.85
Cost Function

0.8

0.75

0.7

0.65

0 10 20 30 40 50 60 70 80 90 100
Generation

Figure 15. Optimisation cost function. Black line – Best cost function. Blue line – 1st reduced frequency.
Green line – 2nd reduced frequency. Red line – 3rd reduced frequency. Cyan line – Static displacement

VII. Application of Gust Device to Scaled Sensorcraft Model

A. Original Wing
Firstly, the divergence, flutter and gust behaviour of the original wing was investigated. It is intended to use the
model to design a wind-tunnel model and so all aerodynamic analyses are done using sea-level atmospheric
conditions and at a velocity of 30m/s.
The divergence speed was found to be 80.846m/s, with bending of the forward wing being the critical mode.
The structure was found to flutter at 66.373m/s, with the coalescing frequencies associated with mode 1 (rear
wing plunge and pitch) and mode 2 (first wing bending of both forward and rear wings).
The gust analysis was performed with a freestream velocity of 30m/s. A ‘1-cosine’ gust was used for the gust
analysis. Using FAA regulations, the maximum gust that the scaled aircraft should be designed for is 2.957m/s. The
specified gust gradient distance (the distance required for the gust to build to a peak) is 12.5 mean chord lengths of
the aircraft, with this chord measurement being 0.27356m.

12
American Institute of Aeronautics and Astronautics
The maximum (absolute) stress encountered during the response (see figure 16) to the gust (which occurs in the
rear spar of the forward wing at the root) was found to be 146.361MPa. A peak (absolute) twist of the wing at the tip
was found to be 1.6805°. The response was also measured at the point on the span where the wing would be
modified; at this point the maximum (absolute) twist encounter was 1.4825° and the maximum (absolute) plunge of
the wing (at a point where the wing will be modified to attach the device i.e. 86% span) was 0.050316m.
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Figure 16. Gust Response for Baseline Wing

B. Wing with Gust Alleviation Device

The wing was altered to include a gust-alleviation device. This device replaced part of the outboard section of the
wing (86% - 100% span) and is attached to the rest of the wing via a torsional spring close to the leading edge,
allowing it to pitch relative to the rest of the wing with an associated stiffness (see figure 17). As the main wing
pitches up in response to a gust (and by doing so increasing aerodynamic loads and resultant stresses), the device
rotates about the spring (as the aerodynamic centre lies aft of this axis), decreasing its incidence and therefore
decreasing aerodynamic loads.
Of interest in this study is the aerodynamic behaviour of the wing as both the stiffness of the attachment, as well
as the chordwise position of the attachment point are varied. Three attachment points that coincide with existing
nodes in the model are considered; x/c = 0.0, x/c = 0.08307 and x/c = 0.1666, with c the local chord length and x the
chordwise distance from the leading edge.
Figure 18 illustrates the variation of the divergence speed with attachment stiffness for several different chordwise
positions of attachment. The critical divergence mode was found to be bending of the forward wing for stiffnesses
approximately greater than 50Nm/rad and pitching of the device as well as rear wing bending for stiffnesses below
this value.
Although the divergence speed barely alters over the range of stiffnesses considered, the device actually improves
the divergence behaviour of the wing slightly; this is due the fact that the attachment points lies ahead of the
aerodynamic centre of the device and so a nose-down moment results, decreasing the angle of incidence of the
device as the stiffness decreases. The result is reduced aerodynamic loads on the wing. Furthermore, the plot shows
that in terms of divergence behaviour, the closer to the leading edge of the wing the attachment point is, the better;
this reflects the greater length of the moment arm between the aerodynamic centre and the axis of rotation. The
difference in the divergence speed at high stiffnesses from the original model (80.846m/s) is most likely due to
structural modifications to include the device. However, the critical modes are identical.
Figure 19 illustrates the modified wing’s flutter speed variation with attachment stiffness, for various chordwise
attachment points. It is clear that the gust alleviation device has a detrimental effect on the flutter behaviour of the

13
American Institute of Aeronautics and Astronautics
aircraft; as the stiffness of the device decreases, so too does the flutter speed. The significant range of stiffnesses that
have a major impact on the flutter speed is between around 5 - 40 Nm/rad. At the upper end of this range a sharp
levelling-off of the flutter speed is observed. This is caused by a change of modes causing the flutter. Also evident is
that the closer the attachment point is to the leading edge, the better, with regards to flutter behaviour.
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Figure 17. Sensorcraft Model With Gust Alleviation Device

Figure 18. Divergence Speeds When Gust Device is Employed

Figure 20 shows the maximum stress in the wing during the response to the gust. For attachment points at x/c =
0.08307 and x/c = 0.1666 the plot ends abruptly because below the corresponding value of torsional stiffness the
wing flutters. At 30m/s with the attachment point at the leading edge (x/c = 0.0), the wing is free from flutter over the
entire range of stiffnesses considered. The device is able to reduce the maximum gust response of the baseline
structure by 19%.
As expected, above a certain stiffness value, a decrease in the stiffness alleviates the stress in the structure.
However, below this point the stress begins to increase again. This is due to the decrease of damping in the system. It
can also be seen that the further forward the attachment point is, the more effective the device is at alleviating the
stress. Figures 21 and 23 also show that a forward-placed attachment point is best at minimising displacements of the
main wing.

14
American Institute of Aeronautics and Astronautics
Figure 24 shows the behaviour of the wing due to the gust for stiffness value k = 30Nm/rad at 30m/s and with an
attachment point of x/c = 0.0. It was found that the device pitched to much higher incidences when the torsional
stiffness was lowered.
The high stiffness values of maximum stress (one again found in the rear spar of the forward wing at the root) are
slightly below the original wing’s value of 146.361MPa; this can be attributed to structural modifications in the
model. However, the corresponding values of wing twist, tip twist and plunge (analysed at attachment point to
eliminate angular contributions from the device) are in good agreement with the original wing (1.4825°, 1.6805°,
0.050316m respectively).
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Figure 19. Flutter Speeds When Gust Device is Employed

Figure 20. Variation of Max Stress due to Gust vs. Stiffness

15
American Institute of Aeronautics and Astronautics
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Figure 21. Variation of Wing Twist due to Gust vs. Stiffness

Figure 22. Variation of Device Twist due to Gust vs. Stiffness

In terms of divergence, flutter and gust response behaviour, it is the recommendation of the authors that an
attachment point as close to the wing’s leading edge as possible is used; this raises both the divergence and flutter
dynamic pressures, and decreases stresses and deflections caused by gust inputs.
Further to this, in terms of divergence behaviour, it is apparent that the gust-alleviation device is beneficial to the
wing; as the stiffness of the attachment decreases, the divergence dynamic pressure increases. However, this is not
the case with flutter; in general, a decrease in attachment stiffness lowers the flutter dynamic pressure. However,
above a certain stiffness (approximately 40Nm/rad), a change in the stiffness has little effect on the flutter behaviour,
as neither of the two modes involved in flutter at these stiffness values is associated with the device. With regards to
gust response behaviour, in general the lower the stiffness of the device attachment, the more effective the device is
at reducing gust-induced stress peaks; however, below a certain stiffness value the stress and deflections of the
response increase again with decreasing stiffness, which reflects the fact that damping in the system is increasing.

16
American Institute of Aeronautics and Astronautics
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

Figure 23. Variation of Max Plunge at Attachment Point due to Gust vs. Stiffness

Figure 24. Gust History and Response of Wing for k = 30Nm/rad, x/c = 0.0.

For this model only, modified specifically as above, and at the considered atmospheric conditions (sea-level,
30m/s) with the specific gust as described previously, a torsional stiffness of 10-40Nm/rad will be used for the
attachment. Future work will investigate the use of the an adaptive device to enable the stiffness to be varied in flight
Experimentally, the attachment will be realised by the use of an aluminium beam, clamped at either end by the
device and wing respectively. Using an appropriate cross-section (square with side-length 5mm), the above torsional
stiffness range can be obtained by separating the clamps by 57-229mm.

VIII. Conclusions

The aeroelastic scaling of a sensorcraft structure has been performed and the design for a wind-tunnel model
achieved. It has been shown that the addition of a passive device at the wing-tip is beneficial for reducing the
maximum stress due to the gust response, 19% in this example, while having little effect upon the flutter and
divergence characteristics.

17
American Institute of Aeronautics and Astronautics
Acknowledgment
The authors would like to thank the EOARD / AFOSR for its support via technical monitor Dr Surya Surampudi,
and also the technical input from Dr Max Blair, Dr Robert Canfield and Lt/Col.Vanessa Bond. The project was also
supported by the EPSRC through the ESF Eurocores SMorph programme.

References
1
M. Blair & R. A. Canfield, “A Joined Wing Structural Weight Modeling Study” AIAA-2002-1337. SDM Conf 2002 .
2
M. Blair, R A Canfield & R.W Roberts, “Joined wing aeroelastic design with geometric non-linearitity”. IFASD Conf. June
2003.
3
R.W. Roberts, R A Canfield & M Blair, “Sensorcraft Structural Optimisation and Analytical Certification” SDM Conf April
2004.
4
J. Wolkovich, “The Joined Wing – An Overview” J. Aircraft v23 n3 1986 pp 161 – 178.
5
J W Gallman & I M Kroo, “Structural Optimisation for Joined Wing Structures” J. Aircraft v33 n1 1996 pp 214 – 223.
Downloaded by CRANFIELD UNIVERSITY on January 16, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2008-5875

6
I M Kroo, J W Gallman & S C Smith, “Aerodynamic and Structural Studies of Joined Wing Aircraft” J. Aircraft v28 n1
1991 pp 74 – 81.
7
J W Gallman, I M Kroo & S C Smith, “Optimisation of Joined Wing Aircraft” J.Aircraft v30 n6 1993 pp 897 – 905.
8
E Livne, “ Aeroelasticity of Joined-Wing Airplane Configurations: Past Work and Future Challenges – A Survey” SDM
Conf 2001. AIAA2001-1370.
9
J.E.Cooper, S.Bureerat, O Sensburg & G.Dimitriadis, “Aeroelastic and Structural Design of an All-Movable Aircraft Fin”
CAES Conf on Multidisciplinary Aircraft Design and Optimisation, 2001 pp 177 - 188.
10
A.Ainul, J.E.Cooper & Otto Sensburg, “Development of a Gust Alleviation Device” Int Forum on Aeroelasticity and
Structural Dynamics 2007.
11
RL Bisplinghoff, H Ashley & R L Halfman, “Aeroelasticity” Dover Edition 1996.
12
M French & F Eastep, “Aeroelastic Scaling of Wind Tunnel Models” AGARD SMP meeting on Advanced
Aeroservoelastic Testing. Rotterdam. 1995.

18
American Institute of Aeronautics and Astronautics

You might also like