Design and experimental validation of a control system for a morphing wing

T. L Grigorie, R. M. Botez, A.V. Popov

École de Technologie Supérieure, Montréal, Québec H3C 1K3, Canada

The paper presents a smart way to actuate and to control the airfoil shape of a morphing wing.
The actuation system development is based on some smart material actuators like Shape Memory
Alloys, disposed in two parallel actuation lines, and its control is performed by using a fuzzy logic PD
controller of Mamdani type.
Based on the promising benefits related to the drag reduction, fuel consumption economy and
flight envelope increasing, many morphing aircrafts studies were developed in last decade both at the
university and industry levels. The multidisciplinary aspects involved by such studies, bring together
research teams in many fields of the science: aerodynamics and aeroelasticity, automation, electrical
engineering, materials engineering, control and software engineering. Categorized as a part of the
“Smart structures” engineering field, the general concept of morphing aircrafts includes some particular
elements, as a function by the complexity of the developed morphing application. Recent researches in
smart materials and adaptive structures fields have led to a new way to obtain a morphing aircraft by
changing the shape of its wings through the control of the airfoils cambers; the concept was called
“morphing wing”.
In this context, a CRIAQ (Consortium for Research and Innovation in Aerospace in Quebec)
project was developed in Canada, related to “Laminar Flow Improvement on an Aeroelastic Research
Wing”. The research team included university and industrial entities, being initiated and funded by the
Bombardier Aerospace and Thales Avionics aerospace companies, as well as by CRIAQ and the Natural
Sciences and Engineering Research Council of Canada (NSERC), and developed by the Ecole de
Technologie Supérieure in Montréal, Canada, in collaboration with Ecole Polytechnique in Montréal and
the Institute for Aerospace Research at the National Research Council Canada (IAR-NRC).
The project general aim was to reduce the operating costs for the new generation of aircrafts
through a fuel economy in flight, and also to improve aircraft performances, expand its flight envelope,
replace conventional control surfaces, reduce drag to improve range and reduce vibrations and flutter.
To achieve this general aim, two particular objectives for our team (Research Laboratory in Active
Controls, Avionics and Aeroservoelasticity (LARCASE) of the Ecole de Technologie Supérieure in
Montréal, Canada) were established in this project: 1) to detect and visualize the airflow characteristics
using pressure sensors installed on the upper surface of the morphing wing; 2) to develop a system for
active control of the morphing wing during flight to move the transition point from laminar to turbulent
flow closer to the trailing edge, promoting in this way large laminar regions on the wing surface, and
thus reducing drag over an operating range of flow conditions characterized by Mach numbers,
airspeeds, and angles of attack.
The chosen model for the wing was a rectangular one (0.5 m x 0.9 m), with a laminar wing
trailing edge airfoil (WTEA) as reference airfoil; its aerodynamic performance was investigated at IAR-
NRC in the transonic regime. In a preliminary aerodynamic study, thirty five optimized airfoils were
calculated by modifying the reference airfoil for each airflow condition as combinations of seven angles
of attack (-1˚, -0.5˚, 0˚, 0.5˚, 1˚, 1.5˚, 2˚) and five Mach numbers (0.2, 0.225, 0.25, 0.275, 0.3). The
experimentally wing model consisted of two parts: 1) a metal fixed one, designed to sustain the wing
loads at a Mach number of 0.3 and an angle of attack up to 6 deg; and 2) a morphing part, consisting of
a flexible skin installed on the upper surface of the wing. The morphing part was actuated by two shape
memory alloys actuators (SMA) to realize the desired optimized airfoil shapes (Fig. 1).

Actuation Uper surface

Actuation
lines (flexible skin)
Leading point
edge rod
beam Oblique cams roller

Lower part Trailing

(rigid) edge
Actuators support
(rigid)
Cavities for instrumentation

Fig. 1 Morphing wing model – cross section

The flexible skin was manufactured in a 4 ply laminate structure in a polymer matrix, with two
unidirectional carbon fiber inner plies and two hybrid Kevlar®/carbon fiber outer plies. The hybrid
Kevlar®/carbon fiber was used in the chordwise direction, in which flexibility was needed for profile
modification, whereas the low-modulus unidirectional carbon fiber was spanwise installed, in which
case rigidity was preferred. The total thickness of the skin was 1.3 mm, the total Young modulus was 60
GPa, Poisson’s ratios were 0.12 for carbon/Kevlar® hybrid and 0.25 for unidirectional carbon.
The actuation mechanism was designed in order to convert the horizontal movement, along the
wing span, into vertical motion, perpendicular to the chord. In this way, two oblique cams with sliding
rods, spanwise positioned, were included in the actuation system (Fig. 2). The oblique cams with sliding
rods had a horizontal versus vertical ratio 3:1. The actuation distance for each actuator was correlated
with the desired airfoil shape for each of the thirty five airflow conditions, and also, was given by the
mechanical equilibrium between the nickel–titanium alloy SMA wires, that pulled the sliding rod in a
direction, and the gas springs, that pulled the sliding rod in the opposite direction. Each sliding rod was
actuated by means of three parallel SMA wires connected to a current controllable power supply.
Gas springs

First
actuation line

Leading edge
side

Actuation Trailing edge

points side
Second
actuation line

Three lines
SMA actuators
beam rod

cam
roller

Actuators support
(rigid)
From power From power
supply #1 supply #2

Fig. 2 Actuation system of the morphing wing

 The gas springs role was to counteract the pulling effect of aerodynamical forces acting in wind
tunnel over the flexible skin when the SMAs were inactive. The pulling action of the gas spring retracted
the flexible skin in the un-morphed-reference airfoil position, while the pulling action of the SMA wires
deployed the actuators in the load mode, i.e., morphed airfoil in the optimized airfoil position.
Two architectures were developed for the morphing wing control system: 1) an open loop
architecture, which controlled electrically the SMA actuators system, and 2) a closed loop architecture,
which included the open loop architecture as an internal loop and controlled the transition point
position based on the information from the pressure sensors installed on the flexible skin; 16 Kulite
pressure sensors were initially installed on the flexible skin.
To design and validate the morphing wing control system some numerical simulations, bench
tests and wind tunnel tests were performed.
The here presented work refers to the open loop architecture of the morphing wing, developed
with a fuzzy logic PD controller.
For each of the two actuation lines the open loop control architecture used a controller which
took as a reference value the required displacement of the actuators from a database stored in the
computer memory to obtain the morphing wing optimized airfoil shape (Fig. 3); because the actuation
lines’ structure was identical, both of them used the same controller. As feedback signal the position
signal from a linear variable differential transducer (LVDT) connected to the oblique cam sliding rod of
each actuator was used. This method was called “open-loop control” due to the fact that this control
method does not take direct information from the pressure sensors concerning the wind flow
characteristics.
Plant
Power SMA
dY1 + Controller A
Optimized Supply actuator
α, M (Re)
Flight condition
airfoils
Desired vertical
displacements

database LVDT
+ dY1 measured 3:1 conversion dX1 measured sensor
Reference
Plant
airfoil
Power SMA
dY2 + Controller B
Supply actuator
A B
LVDT
dY2 measured 3:1 conversion dX2 measured sensor

Fig. 3 Open loop control architecture

Due to the strong non-linear character of the smart materials actuators used in our application,
one variant for the controller was developed by using the fuzzy logic techniques. We tried to
counterbalance the existence of a rigorous mathematical model, a prior developed for system, avoiding
in this way the loss of precision from linearization and uncertainties in the system’s parameters, which
negatively influences the quality of the resulting control. In the same time, we used the intuitive
handling, simplicity and flexibility capabilities offered by the fuzzy logic techniques and due to their
closeness to human perception and reasoning; fuzzy logic is an interface between logic an human
reasoning, providing an intuitive method for describing systems in human terms and automates the
conversion of those system specifications into effective models.
The controller chosen structure was a PD fuzzy logic one, having as inputs the error (difference
between the desired and measured vertical displacement) and the change in error (the derivative of the
error), and as output the voltage controlling the Power Supply output current. Widely accepted for
capturing expert knowledge, a Mamdani controller type was used, due to its simple structure of “min-
max” operations.
 Using the Matlab/Simulink simulation scheme in Fig. 4, the characteristics in Fig. 5 were
obtained, when a repeated step signal was used for the desired displacement. In Fig. 4: “Fuzzy
Controller” block models our controller and the conversion voltage - electrical current, obtained from
the power supplies; “SMA model” is an S-function modeling the SMA non-linear behavior, with electrical
current and loading force as inputs, and with displacement and temperature as outputs (the model was
built in the Shape Memory Alloys and Intelligent Systems Laboratory (LAMSI) at ETS, using Lickhatchev’s
theoretical model); “Mechanical system” models the morphing wing system’ mechanical part behavior
in terms of forces (aerodynamic force, skin elastic force, gas spring loading force) and skin vertical
deflection.

desired deflection
Fuzzy controller [mm]
SMA Model SMA elongation skin
1 7 3/1000 Diff error Scope
[m] deflection
Current out Current Displacement
desired [mm]
cam factor Temperature
deflection mm to m Force Temperature
[mm]
0.078 1.8

SMA max set SMA Initial 1.8

[m] length [m] Memory1
273.15
Memory2 SMA Initial Temperature
length Celsius to Kelvin [K]

skin
1 1100 F aero [N] F SMA [N] deflection XY Graph
[mm]
Aerodynamic skin deflection [mm]
x [m] y [mm]
force [N]

Mechanical system

Fig. 4 Matlab/Simulink simulation scheme

Vertical displacement dY [mm]

Vertical displacement dY [mm]

8 desired dY 8
obtained dY
6 6

4 4

2 2

0 0

0 50 100 150 200 250 300 20 40 60 80 100 120

Time [s] SMA temperature [oC]
x 102
120 13
SMA temperature [oC]

SMA loading force [N]

100 12

80 11

60 10

40 9

20 8
0 50 100 150 200 250 300 20 40 60 80 100 120
Time [s] SMA temperature [oC]

Fig. 5 Numerical simulation results

Fig. 5 shows the response of the actuator relative to the desired vertical displacement, the SMA
actuator envelope (obtained vertical displacement vs. temperature), the SMA temperature in time, and
the SMA loading force vs. temperature. Using a preliminary estimation of the forces loading the
mechanical system, the next values were considered in simulations: 1100 N for aerodynamic force; 1250
N for gas spring pretension force; and the linear elastic coefficients of 2.95 N/mm and 100 N/mm, for
the gas spring and for the flexible skin, respectively.
The relative allure of the obtained and desired displacements, proved the good functioning of
the controller; the system’s response is a critically damped one, an easier latency being observed in the
cooling phase of the SMA wires in comparison with theirs heating phase. The SMAs temperature
oscillations in the steady-state of the actuation position are due to theirs thermal inertia, and do not
affect significantly the SMA elongation. The shape of the “displacement vs. temperature” and “loading
force vs. temperature” envelopes highlights the strong nonlinear behavior of the SMA actuators. The
maximum value of the loading force was 1214 N, corresponding to the maximum value of the desired
vertical displacement i.e. 8 mm, while the minimum value was 877 N, being associated with the 0 mm
required displacement. For the SMA temperature, the maximum, respectively the minimum values,
were 104.02 oC and 27.81 oC, and, according to the presented envelopes, were obtained for 8 mm,
respectively for 0 mm desired displacements. We should mention that the interval used to simulate the
desired displacements (0 to 8 mm) was based on the conclusions drawn from the aerodynamic studies
related to the thirty five optimized airfoils corresponding to the thirty five studied flow conditions.
The experimental model developed for the morphing wing open-loop control system is shown in
Fig. 6. The two SMA actuators containing three wires each, were supplied with power by the two AMREL
SPS power supplies, controlled through analog signals by using a NI-DAQ USB 6229 data acquisition card.
The NI-DAQ was connected to a laptop through a USB connection. A control program was implemented
in Simulink, which provided to the power supply unit the needed SMA current intensity through an
analog signal as shown in Fig. 6. The Simulink control program used as feedback three temperature
signals coming from three thermocouples installed on each wire of the SMA actuator, and a position
signal from a linear variable differential transducer (LVDT) connected to the oblique cam sliding rod of
each actuator. The temperature signals were used for the overheat protection system that disconnect
the current supply to the SMA in case of wire temperature pass over the set limit of 120oC. The position
signals served as a feedback for the actuator desired position control.
As can easily be observed from the figure 10, the developed architecture for the open loop
allowed also the wind tunnel experimental validation of the optimized airfoils through the real-time
visualization of the laminar to turbulent transition, starting from the pressure sensors acquired data; in
the bench test phase this feature was disabled because of the airflow absence.
The bench testing was a first experimental validation for the designed controller and its
hardware integration architecture, and, in the same time, constituted a way to see the quality of the
experimental reproducibility of the theoretical reference and optimized airfoils. When the controller
was tested on the bench two steps were performed: 1) Independent testing of the two actuation lines,
without the flexible skin on the model; 2) Simultaneous testing of the two actuation lines with the
flexible skin on the experimental model.
For both of the two steps in bench test was asked to the actuators to perform vertical
displacements corresponding to the pairs (dY 1 , dY 2 ) obtained for each of the thirty five optimized
airfoils. Bench test gave us the opportunity to make some adjustments in the mechanical part of the
model regarding the pretension force of the gas springs, which was finally set to the value of 1000 N; in
lab conditions the mechanical model was not loaded by the aerodynamic forces. Also, the LVDT
transducers were calibrated by using a laser beam which measured in a very precise way the vertical
displacements produced by each of the two actuation lines (Fig. 7). The same laser beam was used to
evaluate the quality of the experimental reproducibility of the theoretical reference and optimized
airfoils by scanning the center line of the wing model.
Once the experimental architecture validated in bench tests, the next phase of the project sent
the morphing wing model in wind tunnel (Fig. 8 – the wing was fixed in vertical position); the wind
tunnel tests were performed at the Institute for Aerospace Research at the National Research Council
Canada (IAR-NRC) in Ottawa. For the open loop architecture validation, all thirty five optimized airfoils
were tested in correspondent flow cases. The here tried configuration included the analysis of the wind
flow characteristics, based on the information provided by the pressure sensors; the transition between
laminar and turbulent flows was real-time visualized.
For all tested cases was noticed an improvement in the laminar regions on the wing surface, the
transition point from laminar to turbulent flow being moved closer to the trailing edge.

Data acquisition system for

Transition position
pressure sensors
visualization
Signal processing
Matlab/Simulink

Real
Pressure
dY 1 dY 2 airfoill sensors

Flight Optimised
SMA1 SMA2 Reference
conditions airfoils Desired airfoil

α, M
database
Human dY 1, dY 2

Operator (Re) (dY 1opt,

dY 2opt)
dY 1real ,
e dY 2real From position
sensors (LVDT)
Data
acquisition From thermocouples
Temperatures
card Power
Controller
supplies Electrical
Computer current

Fig. 6 Experimental model developed for the morphing wing open-loop control system

along the ch
ord

Laser
Electric motor
scanner
and encoder Actuation points
Laser along the span axis
Morphed beam
Upper surface for
airfoil
Reference morphed airfoil
airfoil
Flexible skin
Center line
Leading of the wing
edge model
∆Y1
∆Y2

Upper surface for Trailing

Actuation edge
reference airfoil Lower part Actuators support lines
(rigid) (rigid)

Fig. 7 Bench test morphing wing laser scan

 Trailing edge
Leading edge

Flexible skin Lower part

(rigid)

Power supplies

Data acquisition Computer for control

card and for the analysis
of the wind flow
characteristics

SMA wires
Flexible skin

Actuators
Oblique cams support

Kulite pressure
sensors

Fig. 8 Morphing wing model during wind tunnel tests

As an example, we expose here the wind tunnel test results for M=0.3, α=-1 deg airflow case
(dY 1 =6.26 mm, dY 2 =7.88 mm). Shown are: the FFT of the Kulite pressure sensors data (Fig. 9), and the
pressure data RMSs and the C p distribution (Fig. 10), for un-morphed and morphed airfoils.
 Unmorphed airfoil Morphed airfoil

Eleventh sensor

Turbulence

Fig. 9 FFT of the Kulite pressure sensors data

Unmorphed airfoil Morphed airfoil

Morphed airfoil
Optimized airfoil

Reference airfoil Unmorphed airfoil

Turbulence

Eleventh
sensor
RMS curve

RMS curve
...
Available pressure sensors

Fig. 10 The pressure data RMSs and the C p and N factor distributions

In Fig. 10 can be observed the C p and N factor (for transition positioning) distributions for
reference airfoil and optimized airfoil, which was calculated for this airflow case. The distributions were
estimated by using the XFoil computational fluid dynamics; XFoil code is free licensed software in which
the en transition criterion is used. In these graphs, the C p and N values calculated by XFoil for various
sensors are defined by circles. The critical value N cr = 7.34 was used in the simulation to match the
turbulence level T = 0.14%, measured in the wind tunnel using Mack’s correlation.
 The predicted transition position calculated by XFoil for the reference and optimized airfoils was
found to be the seventh position of the available sensor positions, corresponding to 0.44·c, respectively
the tenth position of the available sensor positions, corresponding to 0.59·c. In the morphed-to-
optimized airfoil case, the RMS plot displayed in Fig. 10 with star symbols, showed that the sensor with
the maximum RMS became the eleventh sensor plotted, which was located at 0.63·c. The spectral
decomposition of the pressure signals in Fig. 9 confirmed the Tollmien–Schlichting wave’s occurrence in
the eleventh sensor, visible in the highest power spectra (eleventh channel in the right hand side plots)
in the frequency band of 3–5 kHz.