Aerospace Science and Technology 110 (2021) 106457

Contents lists available at ScienceDirect

Aerospace Science and Technology

www.elsevier.com/locate/aescte

Active flutter suppression of wing with morphing flap

Yan Ouyang, Yingsong Gu, Xiping Kou, Zhichun Yang ∗
Institute of Structural Dynamics and Control, School of Aeronautics, Northwestern Polytechnical University, Xi’an, 710072, PR China

a r t i c l e i n f o a b s t r a c t

Article history: The flutter characteristics of a wing with morphing flap are analyzed based on computational fluid
Received 24 May 2020 dynamics (CFD) method. The unsteady aerodynamic reduced order model (ROM) is constructed to model
Received in revised form 23 September the morphing flap with high accuracy, and the effect of flap camber on the wing flutter boundary is
2020
investigated. Active suppression of aeroelastic response by controlling the flap deflection is validated
Accepted 19 December 2020
Available online 29 December 2020
through numerical simulations. A Linear Quadratic Gaussian (LQG) compensator is designed to explore
Communicated by Z. Weiwei the potential of suppressing wing flutter by deflecting the flap actively. The simulation results show that
the order of morphing flap camber may change the critical flutter speed and flutter frequency of the
Keywords: wing. Compared with the conventional plain flap, the morphing flap shows higher control efficiency with
Morphing flap fewer deflection demands.
Active flutter suppression © 2020 Elsevier Masson SAS. All rights reserved.
Reduced order model
LQG controller

1. Introduction flap with those of the conventional plain flap. They found that
the morphing flap increased the UAV’s performance in terms of
Morphing devices on an aircraft may be expensive and complex larger rolling and pitching moments and smaller control surface
but they may resolve the design conflicts that arise in different deflections. Wind tunnel test on the adaptive compliant wing was
mission requirements [1]. Morphing flap belongs to the airfoil-level performed by a group of researchers, they pointed out that the
morphing device and is an effective means to control aerodynam- morphing flap could get a better flow attachment and maintain
ics [2]. When the conventional rigid plain flap deflects, it will the drag in lower level than that of the plain flap [10]. In the
induce an abrupt change to the camber line of the wing at the Variable Camber Continuous Trailing Edge Flap (VCCTEF) program,
hinge position. This may cause sudden changes in aerodynamic the morphing flap consisted of three chordwise segments of equal
pressure and flow separation, resulting in a large drag increment. length. Kaul and Nguyen [11] found that the VCCTEF with parabolic
In contrast, the wing with morphing flap has a smooth camber line camber had an advantage in lift-to-drag ratio and pressure distri-
and has no gaps on the wing skin panel. To take advantage of the bution. Fincham and Friswell [12] optimized the morphing flap in
morphing flap, even in the early 1920s, aeronautical engineers had the Fishbone Active Camber (FishBAC) system under multiple flight
tried to design some camber morphing wing structures [3]. conditions. They stated that the performance of FishBAC system
From the 1980s, many research interests began to focus on was very close to that of the airfoil which could morph arbitrarily.
the engineering application of morphing control surface. The Mis- In the European Smart Intelligent Aircraft Structure project, mor-
sion Adaptive Wing (MAW) program demonstrated that the F-111 phing flap was used to improve the cruise performance of aircraft
aircraft with leading- and trailing-edge morphing flaps had an ad- by continuously maintaining the optimal shape [13].
vantage in drag reduction especially at conditions away from the The aforementioned studies about the morphing flap have been
design point [4,5]. Sanders et al. [6] studied the aerodynamic and conducted with the main concern of static aeroelastic and aerody-
aeroelastic characteristics of a straight wing and pointed out that namic performances in detail, but studies focusing on the dynamic
the morphing flap might increase the maximum roll rate of the aeroelastic characteristics are very limited. Wing flutter, a typical
wing. Xie et al. [7,8] came to a similar conclusion and trimmed dynamic aeroelastic phenomenon, is a self-excited and potentially
a tailless uninhabited air vehicle (UAV) longitudinally with mini- destructive oscillation and may produce catastrophic failure of the
mal induced drag by multiple morphing flaps. In the Smart Wing aircraft. Questions have been raised about the safety of the us-
program, Kudva et al. [9] compared the functions of morphing age of morphing flaps. If the plain flap is replaced by a morphing
one for better aerodynamic performance, the potential impact on
the wing flutter characteristic should be examined in advance. The
* Corresponding author. main challenge faced by aeroelastic researchers is the calculation
E-mail address: [email protected] (Z. Yang). of unsteady aerodynamics acting on the morphing flap. To the best

https://doi.org/10.1016/j.ast.2020.106457
1270-9638/© 2020 Elsevier Masson SAS. All rights reserved.
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

of the authors’ knowledge, the effect of flap camber on the wing

flutter characteristics is seldom investigated in the literatures.
With the development of modern control theory, the applica-
tion of control surface in active flutter suppression is of particular
research interest because it can improve the aeroelastic perfor-
mance without weight penalty. In the published literatures, there
are many examples of active flutter suppression by the plain flap
[14–18]. In recent years, some researchers begun to use the mor- Fig. 1. Equivalent deflection angle of morphing flap.
phing flap in active flutter suppression. Marques et al. [19] first
used the morphing flap to suppress the limit cycle oscillation of compared. The aim of this study is to develop a better understand-
a two-dimensional wing model with bilinear and cubic nonlinear- ing of dynamic aeroelastic characteristics of wing with morphing
ity. The variation of flap camber was described by a third-order flap.
polynomial and the aerodynamics was calculated by distributed The rest of this paper is organized as follows. In Sec. 2, the mor-
lumped vortex method. Stanford [20] examined the use of VCCTEF phing flap is described mathematically, and the aeroelastic wing
in aeroservoelastic optimization. The morphing flap could be used model with a flap is established. Sec. 3 describes the procedure of
to minimize the structure weight while increasing flutter speed. ROM-based aeroelastic analysis and the process of LQG controller
The classical flutter analysis is usually based on the linear po- design. Sec. 4 shows the simulation results and explores the con-
tential flow theories, such as Theodorsen’s theory and doublet lat- trol effect. The conclusions are drawn in Sec. 5.
tice method. These aerodynamic models have been proven to be
adequate for compressible flows in the subsonic and supersonic re- 2. Aeroelastic model
gions, but they are developed for a rigid plain flap, and cannot be
used to model the morphing flap exactly. Therefore, a high-fidelity 2.1. The definition of morphing flap
aerodynamic model should be used in the flutter analysis of the
wing with a morphing flap. As shown in Fig. 1, when the plain flap (solid line) deflects,
The CFD method makes the fewest assumptions about the it rotates around the hinge axis as a rigid body. The deflection is
flow field and is capable of accurately calculating aerodynamics measured by the rotation angle β . xc is the location of the hinge
of structure with complex configuration. This method has already axis position, the flap length is c f . The morphing flap (dashed line)
been used to study the aerodynamic and static aeroelastic charac- does not have a distinct hinge axis, so in order to quantify the de-
teristics of morphing flap. Huntley et al. [21] studied the effect of flection achieved by the morphing flap, an equivalent deflection
morphing flap camber on the lift to drag performance and pointed angle is defined based on the vertical displacement of the trailing-
out that for much of the lift range the 3rd order polynomial shape edge point. This angle is equal to the deflection obtained by the
generated the best performance. Burdette and Martins [22] opti-
plain flap with the same chord length. When the plain flap de-
mized the wing structure of commercial transport, and found that
flects downward, the vertical displacement of the TE point equals
the morphing flap could reduce the cruise fuel burn effectively.
yte . The equivalent deflection angle of the morphing flap can be
Tian et al. [23] introduced the morphing flap into a high-lift sys-
calculated as:
tem and obtained higher lift coefficient and lift-to-drag ratio than
the original system. Within this context, CFD is probably the most yte
β = arctan (1)
reliable tool for calculating unsteady aerodynamic loads on the cf
morphing flap in the current study.
A generic fourth order equation, Eq. (2), is used to describe the
The CFD/CSD coupling method exhibits capability to predict
camber line of the deflected flap [33]:
flutter characteristics in an accurate manner. Many results obtained
by this method showed good agreement with experiment data tan β
y c (x) = − (x − xc )n , x ∈ [xc , 1], n = 1, 2, 3, 4 (2)
[24,25]. However, even with the impressive improvements of mod- (1 − xc )n−1
ern computer hardware, the computational cost of CFD/CSD cou-
where n is the order of the equation, and determines the kind of
pling method is still too prohibitive in industrial analysis. Further
flap. The case of n = 1 is the plain flap, while n = 2, 3 and 4 corre-
more, the full order CFD model cannot be used in the controller
spond to three kinds of morphing flaps. In the following sections,
design process.
the order of the Eq. (2) is used to represent the kind of flap.
To address these limitations, several kinds of aerodynamic
The NACA Four-Digit Series thickness distribution, yt (x), is de-
ROMs were developed and the amount of time spent on analyzing
fined by following equation [34]:
the flutter characteristics was significantly reduced [26–30]. The
ROM-based methodology has the advantage of producing a simpli- t  √
yt (x) = 0.2969 x − 0.1260x − 0.3516x2
fied mathematical model for the aerodynamics while still retaining 0.2

the accuracy of the complete unsteady CFD calculation. +0.2843x3 − 0.1015x4 (3)
The ROM can be formed by system identification theory. The
autoregressive moving average (ARMA) model is typical of this the- where t is the non-dimensional maximum thickness (e.g. t = 0.12
ory and has been widely used in ROM-based aeroelastic analysis for the NACA 0012 airfoil). x ∈ [0, 1] is the non-dimensional dis-
[26,31]. The ARMA model establishes the relationship between the tance along the chord. If the flap maintains its thickness distri-
structural displacement and the aerodynamic load that acting on bution after deflection, the final coordinates for the wing section
the wing structure. Moreover, it has an explicit mathematical for- upper surface (xu , y u ) and lower surface (xl , yl ) can be determined
mula and can also be used in aeroservoelastic analysis [32]. by the following relationships:
In this paper, the original model presented in the work of ⎧
Ref. [31] is adopted while the rigid plain flap is replaced by a mor- ⎪
⎪ xu = x − yt (x) sin(θ(x))
⎪
⎪
phing one. Aerodynamic ROM is used to guarantee the accurate ⎨ y u = y c (x) + yt (x) cos(θ(x))
calculation of unsteady aerodynamics of the morphing flap. Three (4)
⎪
⎪ xl = x + yt (x) sin(θ(x))
configurations of morphing flaps are used to suppress the flutter ⎪
⎪
⎩
by the LQG controller, and the performances of different flaps are yl = y c (x) − yt (x) cos(θ(x))

2
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

Fig. 2. NACA 0012 wing section with different kinds of flaps, β = 15◦ .

S α and S β are the static moments of wing about e.a. and of flap
about hinge axis, respectively. The right-hand side of Eq. (6) is the
aerodynamic loads which are computed by integrating the pres-
sure distribution around the main wing and flap.
Defining a vector ξ = [ h/b α β ] T , Eq. (6) can be written as
the following non-dimensional form:

Mξ̈ + Kξ = F (7)
⎡ ⎤
1 xα xβ
where M = ⎣ xα (a f − a)xβ + rβ2 ⎦ is the mass
2
rα
Fig. 3. Sketch of the wing with flap.
xβ (a f − a)xβ + rβ2 rβ2
matrix,
⎡ ⎤
where θ(x) is the local slope of the camber line, and can be calcu-
(ωh /ωα )2 0 0
lated as:
K = ⎣ 0 2
rα 0 ⎦ is the stiffness matrix, F =
θ(x) = arctan( y c (x)) (5) 0 0 rβ2 (ωβ /ωα )2
⎡ ⎤
2
−C l
When the order of flap and deflection angle are determined, v∗ ⎣ 2C m ⎦ is the aerodynamic coefficients vector. v ∗ = v√ is
the coordinates of the whole wing can be calculated. As shown in π ωα b μ
2C β
Fig. 2, the four flaps are deflecting with the same angle. The plain m
flap deflects with a distinct convex point on the airfoil surface,
the non-dimensional velocity, μ= πρ b2
is called the mass ratio.
while all of the three morphing flaps keep smooth wing surface, By defining a structural state vector
and the curvature of camber line becomes greater with the in-
creasing order of the flap. xs = [ h/b α β ḣ/b α̇ β̇ ]T ,
Eq. (7) can be further rewritten in state-space form as:
2.2. Governing equation of motion
ẋs = As xs + qBs fa
As shown in Fig. 3, a two-dimensional aeroelastic wing has (8)
ξ = Cs xs + qDs fa
three degrees of freedom (DOF) and is supported by translational  
and rotational springs. The wing chord length is 2b. K h and K α are 0 I 0  
where As = , Bs = , Cs = I 0 , Ds = [0], 0 and
the linear plunging and pitching stiffness, respectively. The elastic −M−1 K 0 M−1
axis (e.a.) is located at a distance ab from the mid-chord point, ∗2
I are zero matrix and identity matrix, respectively. q = vπ is the
and the center of gravity (c.g.) of entire wing is located at a dis- non-dimensional dynamic pressure.
tance xα from the e.a. The distance between e.a. and flap hinge Structural parameters in the Ref. [31] were adopted in the cur-
axis is a f b. K β is rotational stiffness of flap around hinge axis. rent study, as listed in Table 1. ωh , ωα and ωβ are the uncoupled
xβ b is the distance of c.g. of flap from the hinge axis. The pitch-
natural frequency of plunging, pitching about the elastic axis, and
ing angle α is positive nose-up and the plunging displacement h
of flap rotating around the hinge axis, respectively. Though the ge-
is positive downward. β is the flap defection angle (downward de-
ometry of the morphing flap is different from that of the plain flap,
flection is positive). Usually, there is a gap between the main wing
it has been validated that this difference has a very limited effect
and plain flap in practice. Because the camber effect is the main
on the c.g. position (xβ b) and the second moment of inertia of flap
concern in this study, the actual physical gap will be ignored for
about the hinge axis (I β ) when the deflection angle is small. So it
the plain flap.
is reasonable to neglect the difference in structural dynamics be-
Neglecting structural damping and using the Lagrange’s equa- tween the plain and morphing flaps in a linear aeroelastic analysis,
tions yields the equations of motion for this three-DOF aeroelastic and Eq. (8) is also used in the calculation of wing with morphing
system [31]: flap in the present paper.
⎧
⎨ mḧ + S α α̈ + S β β̈ + K h h = − L
3. Flutter prediction and controller design
S α ḧ + I α α̈ + [ I β + b(a f − a) S β ]β̈ + K α α = M α (6)
⎩
S β ḧ + [ I β + b(a f − a) S β ]α̈ + I β β̈ + K β β = M β 3.1. Flutter analysis based on ROM technique
where m is the total mass of wing. I α = mrα b and I β = mrβ2 b2
2 2

are the second moment of inertia of wing about e.a. and of flap When forming the ROM of aerodynamics, the ARMA model can
about hinge axis, respectively. rα and rβ are the radius of gyra- be used to describe the response of aerodynamic system as a sum
tion of wing about e.a. and of flap about hinge axis, respectively. of scaled previous inputs and outputs of the system [35]. For a

3
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

Table 1
Structure parameters of three-DOF wing.

b = 0.127 m span = 0.52 m a = − 0.5 a f = 0.5

xα = 0.4316 xβ = 0.01985 2
rα = 0.5331 rβ2 = 0.01292
μ = 25.24 ωh = 4.45 Hz ωα = 9.21 Hz ωβ = 19.44 Hz

Fig. 5. Comparison of unsteady Euler and ROM model solutions for Ma = 0.4.

Fig. 4. Prescribed 3211 input signal for unsteady CFD solver. where ξ is the generalized structural displacement vector, fa is the
generalized aerodynamic coefficient vector, fa0 is the static offsets
multi-input/multi-output (MIMO) system in the discrete time do- of fa . Because fa0 has no effect on the stability characteristics of
main, the ARMA model is formulated as: system, it will be omitted in the following analysis. The four coef-
ficient matrices are

na 
nb −1 ⎡ ⎤
y(k) = Ai y(k − i ) + Bi u(k − i ) (9) A1 A2 · · · Ana−1 Ana B1 B2 ... Bnb−1 Bnb
⎢ I 0 ··· 0 0 0 0 ··· 0 0 ⎥
i =1 i =0 ⎢ ⎥
⎢ .. ⎥
where y(k) is the system output vector (non-dimensional aerody- ⎢ . I ··· 0 0 0 0 ··· 0 0 ⎥
⎢ ⎥
namic force coefficients), u is the system input vector (generalized ⎢ .. .. . . .. ..
.. . . ..
.. .. ⎥
⎢ . . . . . ⎥
structural displacements). Ai and Bi are constant coefficient matri- ⎢ . . . .
. ⎥
⎢ 0 ··· 0 0 0 ··· 0 ⎥
ces need to be calculated for an prescribed model order containing Ãa = ⎢ 0 I 0 ⎥,
⎢ 0 0 ··· 0 0 0 0 ··· 0 0 ⎥
na previous outputs and nb inputs. A major advantage of ARMA ⎢ ⎥
⎢ 0 0 ··· 0 0 I 0 ··· 0 0 ⎥
model is that the system response at any time step k is a linear ⎢ ⎥
⎢ 0 0 ··· 0 0 0 I ··· 0 0 ⎥
combination of previous inputs and outputs, making this model ⎢ ⎥
very easy to be programmed. ⎢ .. .. . .
.. .. .. .. . . .. .. ⎥
⎣ . . .. . . . . . . ⎦
The 3211 multi-step excitation has been proved to be a reli-
able input signal to generate the training data for a MIMO system. 0 0 ··· 0 0 0 0 ··· I 0
Fig. 4 shows the prescribed velocity and displacement input for
 T
B̃a = B̃0 0 0 · · · 0 I 0 0 · · · 0 ,
the three-DOF system that is found to work well for modeling. In  
order to make the identification process distinguish between the C̃a = A1 A2 · · · Ana−1 Ana B1 B2 · · · Bnb−2 Bnb−1 ,
effects of different inputs in the system response, the input sig-
D̃a = [B0 ] .
nal for each mode should be out of phase [36]. The motion of
the main wing and flap is described using a user-defined func- Once the ROM is obtained, it is first necessary to validate its
tion (UDF) in ANSYS/FLUENT software. The UDF is written in C accuracy. Fig. 5 shows the typical time history of aerodynamic co-
programming language and loaded by the solver dynamically. The efficients for the 3211 multistep input signal. As can be seen that
morphing flaps are applied mesh movement grid by grid to match the ROM results agree very well with those of unsteady CFD calcu-
the desired motion. lations. The steps and peaks of the response curves are captured by
To develop the state-space form for aerodynamic model, a the ROM exactly. This indicates that the ROM is accurate enough
state-vector is defined as follows: to be used in place of the unsteady CFD solution in the aeroelastic
analysis.
 T In order to form the aeroelastic system equations, Eq. (11) is
xa (k) = y (k − 1) ... y (k − na ) u (k − 1) ... u (k − nb + 1)
transformed into the following continuous-time form by bilinear
(10) (Tustin) approximation:

The vector contains (na + nb − 1) state variables. The state-space ẋa (t ) = Aa xa (t ) + Ba ξ (t )

(12)
form for the discrete-time aerodynamic model can be formulated fa (t ) = Ca xa (t ) + Da ξ (t )
as:
The aerodynamic model Eq. (12) can then be coupled with the
xa (k + 1) = Ãa xa (k) + B̃a ξ (k) known structural model Eq. (8) to generate aeroelastic equation in
(11)
fa (k) = C̃a xa (k) + D̃a ξ (k) + fa0 state-space form:

4
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

In Eq. (8), the aerodynamic states xa cannot be measured, so

we cannot rely on a full-state feedback controller. Therefore, an
observer is required for estimating the state-vector, based upon a
measurement of the output. The LQG controller combines the op-
timal regulator and Kalman filter (observer) into an optimal com-
pensator, making it very suitable for active flutter suppression.
By defining the state vector of the aeroelastic system xae =
[ xs xa ] T , the equation of motion with flap deflection input term
can be formulated in state-space form:

ẋae = Aae xae + Bae βc (14)


As + Bs Da Cs qBs Ca
where Aae = is the system matrix,
Ba Cs Aa
 2 
ω Bc  T
Bae = rβ2 ωβ is the input matrix, Bc = 0 0 0 0 0 1 .
α 0
βc is the flap deflection command signal, which is generated by
the controller. The second term in the right hand of Eq. (14) is the
driving moment that is produced by actuating system to deflect
the flap.
In engineering practice, the plunge displacement, the pitch an-
gle and the flap deflection angle are easy to be measured by sen-
sors. These three measurements can be used in the output function
as follows:

ξ = Cae xae (15)

where Cae = [ Cs 0 ].
In order to design the LQG controller, Eq. (14) is rewritten with
noise term as [37]:

ẋae = Aae xae + Bae βc + Fw (16)

where w is the process noise vector which may come from mod-
eling errors, such as neglecting higher-frequency dynamics. The
output Eq. (15) is rewritten as:

Fig. 6. Root locus of wing with plain flap. Y = Cae xae + z (17)
  
ẋs As + qBs Da Cs qBs Ca xs where z is the measurement noise vector. Both w and z are as-
= (13) sumed to be white noises.
ẋa Ba Cs Aa xa
The objective function for the optimal control problem is the
According to the stability criterion, the state-dynamics matrix
sum of transient energy and control energy as follows:
influences all the properties of the system such as stability, per-
  
formance and robustness [37]. Calculating eigenvalues of the state-
dynamics matrix in Eq. (13) at different dynamic pressure q, the J= xae T Qxae +βc R βc dt (18)
critical flutter speed can be obtained by the root-locus method.
Repeat the aforementioned procedure at every Mach number, we where Q is a square, symmetric matrix called the state weighting
can get the whole flutter boundary for the wing model. matrix. R is the control cost coefficient. The optimal control prob-
Fig. 6 is the typical root-locus plot of wing with plain flap for lem consists of solving for the feedback gain matrix K, such that
flutter prediction. Fig. 6(a) is the case of Ma = 0.5. When the air- the objective function given by Eq. (18) is minimized. And the con-
speed v ∗ is low, all eigenvalues are in the left-half plane, and trol signal is calculated as:
the wing is stable. With the increasing airspeed, the eigenvalue of
plunge mode moves rightwards. When the eigenvalue has crossed βc = − R −1 Bae
T
Pxae = −Kxae (19)
over into the right-half plane, the wing becomes unstable, and
where P is symmetric matrix. When the control time is infinite,
the critical flutter speed is v ∗f = 0.5942. In Fig. 6(b), the locus of
matrix P is the solution to the following Riccati equation:
plunge mode is always in the left-half plane, and the pitch mode
will make the wing loss its stability. These two sub-figures indi- T
PAae + Aae P − PBae R −1 Bae
T
P+Q=0 (20)
cate that the unstable mode of the aeroelastic system is different
for different Mach numbers. Combine Eq. (16) and Eq. (17) into an aeroelastic plant. The
plant output vector can be used by Kalman filter to estimate the
3.2. Implementation of the LQG controller state vector. The estimated state vector is calculated as:

The active flap deflection may change the unsteady flow field x̂˙ ae = (Aae − Bae K − LCae + LDae K)x̂ae + LY (21)
around the wing, thus affect the motion of the wing. The third
equation of Eq. (6) contains the dynamics characteristics of flap where L is the gain matrix of Kalman filter. After the estimated
with the actuating system. When a proper controller is used, the state vector is obtained, the command input signal for the aeroe-
flap will deflect to follow the command signal. lastic plant can be calculated by the linear optimal regular as:

5
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

Fig. 7. Closed-loop aeroelastic system with LQG controller.

Fig. 8. Partial view of C-type mesh about the NACA 0012 airfoil.

βc = −Kx̂ae (22)
The whole block diagram of the closed-loop system is shown in
Fig. 7. The stability of the closed-loop system can also be analyzed
by root-locus method. Fig. 9. CFD validation for the test case of NACA 0012 airfoil.

4. Results and discussion

grid) and G3 (fine grid) contain 16100 and 37620 cells, respec-
4.1. Validation of unsteady aerodynamics calculation tively.
The lift and moment coefficient curves are compared with
To validate the Euler equation solver and mesh grid, this sub- the published wind tunnel experiment data, as shown in Fig. 9.
section compared the CFD results with published experimental The CFD results show good agreement with experiment data. The
data [38] for the unsteady pitch NACA 0012 airfoil. The aerody- medium grid (G2) generates a maximum relative difference of 0.7%
namic calculations were performed at the free stream with Mach in lift coefficient and 3.8% in moment coefficient when compared
number M ∞ = 0.755, Re = 5.5 × 106 . The airfoil is forced to os- with the fine grid results. This indicates that the medium grid can
cillate in the pitch DOF at a reduced frequency k = 0.0814. The guarantee sufficient accuracy and will be used in the following cal-
instantaneous angle of attack (AOA) α (t ) is described by the fol- culation.
lowing equation:
4.2. Open-loop flutter boundary of wing with plain flap
α (t ) = α0 + α1 sin(ωt ) (23)

where α0 = 0.016◦ and α1 = 2.51◦ are the mean AOA and the am- The entire flutter boundary of wing with plain flap is obtained
plitude of pith oscillation, respectively. The non-dimensional angu- by calculating critical flutter speed at Mach numbers from 0.4 to
lar frequency can be calculated by: 0.9. The results are compared with those obtained by Ref. [31], as
shown in Fig. 10.
kv
ω= (24) In the subsonic region (Ma ≤ 0.8), reasonably good agreement
b is achieved. It indicates that the ROM-based calculation in the cur-
where v is the velocity of free stream. rent study is accurate and reliable. In the transonic region (Ma >
The typical computational mesh used in CFD calculations is de- 0.8), a discrepancy between those two sets of results is found at
picted in Fig. 8. A grid study should be conducted to examine the some Mach numbers. A possible explanation is that the shock wave
sensitivity of CFD results to grid size. To this end, three sets of changes its location on the wing at different Mach numbers. This
mesh grids (G1, G2 and G3) with different mesh resolutions in the makes the calculation of aerodynamic coefficients are sensitive to
flow domain are used for comparison. The first grid (G1) is the the flow condition, and more discussions can be found in the lit-
coarse grid which has 9600 cells. The other two grids, G2 (medium erature [31].

6
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

with three kinds of morphing flaps have similar flutter mecha-

nism but the root locus of plunge mode has different positions.
With the increasing order of flap, the locus of plunge mode moves
towards upper left (as indicated by the dashed arrow). This move-
ment leads to a result that the flutter speed and flutter frequency
are increased simultaneously. When the airspeed is vf ∗ = 0.5942,
the wing with plain flap is critical stable, and the eigenvalue of
plunge mode is just on the imaginary axis. Under the same flow
condition, the eigenvalues of the three wings with morphing flap
are still in the left-half plane and the eigenvalue of the 4th order
morphing flap is farther away from the imaginary axis than those
of the 2nd and 3rd order morphing flaps. The locus of wings with
morphing flaps intersects the imaginary axis at a higher position
than that of the wing with plain flap.
In the case of Ma = 0.85, it is the locus of pitch mode that will
enter the right-half plane, as shown in Fig. 12(b). The locus of pitch
mode moves towards lower left (as indicated by the dashed ar-
row) with the increasing order of flap camber. When the airspeed
is vf ∗ = 0.6633, the wing with plain flap is critical stable, whereas
the eigenvalues of wings with morphing flap are still in the left-
half plane. The eigenvalue of wing with the 4th order morphing
flap is farthest away from the imaginary axis. This indicates the
flutter speed is increased with the increasing order of flap camber.
The root locus of wings with morphing flaps cross over the imag-
inary axis at a lower position, so the flutter frequencies are lower
than that of the wing with plain flap.
In Fig. 12(c), the locus plot at Ma = 0.875 is more complex
than those at Ma = 0.5 and 0.85. When the airspeed approaches
vf ∗ = 0.7977, the eigenvalue of plunge mode of the wing with
plain flap is on the verge of crossing into the right-half plane. Un-
der the same flow condition, the locus of wings with morphing
flaps has already entered the right-half plane, so the flutter speed
is decreased. And the unstable mode changes from plunge mode
to pitch mode with the increasing flap camber order.
Fig. 13 shows an overview of the flutter boundary of wings
with different kinds of flaps. The effect of flap camber order on the
flutter characteristics can be divided into three regions: a). When
Fig. 10. Flutter boundary of wing with plain flap.
0.4 ≤ Ma ≤ 0.825, plunge mode is the unsteady mode, the flutter
speed and flutter frequency of the wing will increase with the in-
4.3. Open-loop flutter boundary of wing with morphing flap creasing flap camber order. b). When Ma = 0.85, pitch mode is the
unsteady mode. The flutter speed will increase with the increas-
Turning now to the wing with morphing flap. According to the ing flap camber order, while the flutter frequency will decreased.
equivalent deflection angle definition in Eq. (1) and camber line c). When Ma = 0.875, the flap camber may change the unsteady
description in Eq. (2), the morphing flap may deflect as the de- mode from plunge mode to pitch mode. The flutter speed will
scribed 3211 multi-step input signal to obtain aerodynamic ROM. decrease with the increasing flap camber order, while the flutter
As the resulting output, the typical time history of aerodynamic frequency will increase.
coefficients is shown in Fig. 11. When t < t β 0 = 0.125s, the flap is The detailed values of the flutter boundary are listed in Table 2.
fixed and the wing is forced to move in the pitch and plunge DOF. In the case of Ma = 0.825, the morphing flaps have the largest
The aerodynamic coefficients of wings with different kinds of flaps positive impact on the flutter boundary. The 3rd order morphing
are exactly equal to each other. When t > 0.125s, the flap begins flap increases the flutter speed by 31.16% when compared with
deflecting, and the flap camber affects the aerodynamics dramati- the plain flap. When the flow condition is Ma = 0.875, the mor-
cally. With the same deflection angle, the higher order flap leads to phing flaps will decrease the flutter speed. The worst case is the
larger aerodynamic coefficients. Thus, the four kinds of flaps gener- 4th order morphing flap decreases the flutter speed by 2.9% when
ate different aerodynamic ROMs. The aerodynamic coefficients ob- compared with the plain flap.
tained by ROM (hollow circle) are compared with those calculated
by Euler solver (solid line). It shows that these two results are ex- 4.4. Active flutter suppression
actly the same. The unsteady aerodynamics of morphing flaps can
be calculated by ROM accurately. The controller plays a key role in the closed-loop system be-
The root-locus plot enables a direct insight into the effect of cause it can affect the wing flutter boundary and dynamic re-
flap camber on the wing flutter characteristics. The root-locus plots sponse. For the sake of simplicity, the results of Ma = 0.5 are used
of wings with different kinds of flaps are shown in Fig. 12. for a detailed analysis.
In the Mach number range 0.4 ≤ Ma ≤ 0.825, the camber effect Without a controller, the critical flutter speed of wing with
is identical. Take the case of Ma = 0.5 (Fig. 12(a)) as an example, plain flap is vf ∗ = 0.5942, as shown in Fig. 12(a). When the air-
the wing with plain flap is stable when the airspeed is low. With speed increases to v∗ = 0.6666, all wings with different kinds
the increasing airspeed, the eigenvalue of plunge mode will cross of flaps are unstable because their eigenvalues of plunge mode
over into the right-half plane and flutter will occur. The wings have already entered the right-half plane. Under this condition, the

7
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

Fig. 11. Aerodynamics time history of wing with different kinds of flaps, Ma = 0.4.

Table 2
Non-dimensional flutter speed and frequency of wings with different kinds of flaps.

Ma 0.4 0.5 0.6 0.7 0.8 0.825 0.85 0.875

1st vf ∗ 0.6165 0.5942 0.5689 0.5226 0.4371 0.4300 0.6633 0.7977
ω f /ωα 0.7390 0.7062 0.6704 0.6272 0.5573 0.5470 0.9611 0.8857

2nd vf ∗ 0.6148 0.5988 0.5766 0.5371 0.4750 0.5272 0.6850 0.7895

ω f /ωα 0.7829 0.7387 0.6963 0.6512 0.5768 0.6025 0.9176 0.9147

3rd vf ∗ 0.6136 0.5999 0.5776 0.5367 0.4848 0.5640 0.6883 0.7750

ω f /ωα 0.7887 0.7459 0.7009 0.6561 0.5804 0.6235 0.9130 0.9300

4th vf ∗ 0.6132 0.6001 0.5782 0.5378 0.4904 0.5499 0.6907 0.7747

ω f /ωα 0.7917 0.7491 0.7029 0.6591 0.5801 0.6161 0.9106 0.9313

aeroelastic system is treated as a plant. Then the LQG controller is the four sub-figures, besides the locus of the structural modes, a
designed according to the description in the previous section. In root locus of the controller always stays in the left-half plane.
order to compare the effect of different flaps, the weighting matri- A comparison of the dynamic response for the initial condition
ces in controller are set uniformly as Q = 10E, R = 1 for all kinds reveals the effect of flap camber on control efficiency. When the
of flaps. initial condition is α̇0 = 5◦ /s perturbation in the pitch-rate, the
The root locus of closed- and open-loop systems are plotted in dynamic response of open-loop system is shown in Fig. 15. There
the same speed range. Fig. 14(a) is the case of wing with plain flap. are very slight differences among the three wings with morphing
Compared with the open-loop system, the root locus of closed-loop flaps, so only the result of 2nd morphing flap are plotted in the
system moves towards upper left. Thus the eigenvalue of pitch figure to compare with that of the plain flap. As the airspeed is
mode will enter the left-half plane at higher airspeed. This indi- higher than the critical flutter speed, both of the two wings are
cates that the stability of the wing is enhanced by the controller. unstable. After encountering a perturbation in the pitch-rate, the
The critical flutter speed of the closed-loop system is vfc ∗ = 0.7206, transient responses on all the DOF increase rapidly and tend to be
which is increased by 21.2% than that of the open-loop system. infinite after several periods.
The controller has a similar effect on the flutter speed of the When the controller is actuated, the closed-loop initial response
wings with morphing flaps. As shown in Fig. 14(b), 14(c) and is shown in Fig. 16. With the active deflection of flap, the transient
14(d)f, the locus of the plunge mode also moves towards upper responses on the plunge and pitch DOF decrease immediately and
left. This tendency is the same as that of the wing with plain flap. decay to zero within 0.1s. During the time history of flap deflec-
The higher order of the flap camber, the higher flutter speed will tion, the maximum angles of the plain flap and 2nd morphing flap
be. The critical flutter speeds of wings with 2nd, 3rd and 4th mor- are 2.8◦ and 1.7◦ , respectively. This means that the 2nd morphing
phing flaps are vfc ∗ = 0.7305, 0.7328 and 0.7362, respectively. In flap has a higher control effect than the plain flap in the same set-

8
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

Fig. 13. Flutter boundary of wings with different kinds of flaps.

three kinds of morphing flaps. Active control strategy is utilized to

improve the dynamic aeroelastic characteristics of the wing. The
simulation results have shown the potential aeroelastic benefits of
morphing flap when used to increase the flutter speed.
The following conclusions can be drawn:

1. Under the same deflection angle, the morphing flap can gen-
erate lager aerodynamic force and moment than those of the
plain flap. As the unsteady aerodynamics are related to the
flap camber, the flutter boundary of the wing may change with
the order of the flap camber.
2. When the flap is actuated by the LQG controller, the flutter
speed of the wing can be increased significantly. And the in-
crement obtained by using a morphing flap is larger than that
by using a plain flap.
Fig. 12. Root-locus of wing with different kinds of flaps.
3. When the airspeed is above the open-loop critical flutter
speed, the closed-loop initial response can be suppressed by
tling time. In the results of the 3rd and 4th morphing flaps (not
flap deflection. The morphing flaps show control advantage in
shown in the figure), the settling time almost the same, and the
term of smaller deflection angle when compared with that of
maximum deflection angle is further reduced.
the plain flap.
5. Conclusions
Declaration of competing interest
This study has been one of the first attempts to use the CFD-
based method to thoroughly investigate the dynamic aeroelastic The authors declare that they have no known competing finan-
characteristics of the morphing wing. The flutter boundaries of cial interests or personal relationships that could have appeared to
wings with plain flap are compared with those of wings with influence the work reported in this paper.

9
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

Fig. 14. Root locus of closed-loop wings with different kinds of flaps, Ma = 0.5.

Fig. 15. Open-loop initial response, when v∗ = 0.6666.

Fig. 16. Closed-loop initial response, when v∗ = 0.6666.

Acknowledgements
References

This work is supported by the National Natural Science Foun- [1] T.A. Weisshaar, Morphing aircraft systems: historical perspectives and future
dation of China (Grants No. 11472216 and 11672240). challenges, J. Aircr. 50 (2013) 337–353.

10
Y. Ouyang, Y. Gu, X. Kou et al. Aerospace Science and Technology 110 (2021) 106457

[2] D. Li, S. Zhao, A.D. Ronch, J. Xiang, J. Drofelnik, Y. Li, L. Zhang, Y. Wu, Structures, Structural Dynamics, and Materials Conference 14th
M. Kintscher, H.P. Monner, A review of modelling and analysis of morphing AIAA/ASME/AHS Adaptive Structures Conference 7th, 2006, p. 1837.
wings, Prog. Aerosp. Sci. 100 (2018) 46–62. [20] B.K. Stanford, Static and dynamic aeroelastic tailoring with variable-camber
[3] S. Barbarino, O. Bilgen, R. Ajaj, M. Friswell, D. Inman, A review of morphing control, J. Guid. Control Dyn. 39 (2016) 2522–2534.
aircraft, J. Intel. Mat. Syst. Struct. 22 (2011) 823–877. [21] S.J. Huntley, B.K. Woods, C.B. Allen, Computational analysis of the aerodynamics
[4] W.W. Gilbert, Mission adaptive wing system for tactical aircraft, J. Aircr. 18 of camber morphing, in: AIAA Aviation 2019 Forum, 2019, p. 2914.
(1981) 597–602. [22] D.A. Burdette, J.R. Martins, Design of a transonic wing with an adaptive morph-
[5] S.B. Smith, D.W. Nelson, Determination of the aerodynamic characteristics of ing trailing edge via aerostructural optimization, Aerosp. Sci. Technol. 81 (2018)
the mission adaptive wing, J. Aircr. 27 (1990) 950–958. 192–203.
[6] B. Sanders, F.E. Eastep, E. Forster, Aerodynamic and aeroelastic characteristics of [23] Y. Tian, J. Quan, P. Liu, D. Li, C. Kong, Mechanism/structure/aerodynamic mul-
wings with conformal control surfaces for morphing aircraft, J. Aircr. 40 (2003) tidisciplinary optimization of flexible high-lift devices for transport aircraft,
94–99. Aerosp. Sci. Technol. 93 (2019) 104813.
[7] J. Xie, Z. Yang, S. Guo, Aerodynamic and aeroelastic characteristics of conformal [24] F. Liu, J. Cai, Y. Zhu, H. Tsai, A. Wong, Calculation of wing flutter by a coupled
control surfaces and application in active aeroelastic wing technology, in: 53rd fluid-structure method, J. Aircr. 38 (2001) 334–342.
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Con- [25] R. Bennett, J. Edwards, An overview of recent developments in computational
ference 20th AIAA/ASME/AHS Adaptive Structures Conference 14th AIAA, 2012, aeroelasticity, in: 29th AIAA, Fluid Dynamics Conference, 1998, p. 2421.
p. 1513. [26] T.J. Cowan, A.S. Arena, K.K. Gupta, Development of a discrete-time aerodynamic
[8] J. Xie, Z. Yang, S. Guo, A flexible wing with conformal control surfaces for opti- model for cfd-based aeroelastic analysis, in: 37th AIAA Aerospace Sciences
mum trim of a tailless air vehicle, in: 51st AIAA/ASME/ASCE/AHS/ASC Struc- Meeting and Exhibit, 1999, p. 765.
tures, Structural Dynamics, and Materials Conference 18th AIAA/ASME/AHS [27] W.A. Silva, R.E. Bartels, Development of reduced-order models for aeroelastic
Adaptive Structures Conference 12th, 2010, p. 2713. analysis and flutter prediction using the cfl3d v6. 0 code, J. Fluids Struct. 19
[9] J.N. Kudva, B.P. Sanders, J.L. Pinkerton-Florance, E. Garcia, Darpa/afrl/nasa smart (2004) 729–745.
wing program: final overview, Proc. SPIE 4698 (2002) 37–43. [28] Z. Chen, Y. Zhao, R. Huang, Parametric reduced-order modeling of unsteady
[10] S. Kota, J. Hetrick, R. Osborn, D. Paul, E. Pendleton, P. Flick, C. Tilmann, De- aerodynamics for hypersonic vehicles, Aerosp. Sci. Technol. 87 (2019) 1–14.
sign and application of compliant mechanisms for morphing aircraft structures, [29] J. Kou, W. Zhang, A hybrid reduced-order framework for complex aeroelastic
Proc. SPIE Int. Soc. Opt. Eng. 5054 (2003) 24–33. simulations, Aerosp. Sci. Technol. 84 (2019) 880–894.
[11] U.K. Kaul, N.T. Nguyen, Drag optimization study of variable camber continuous [30] D. Li, A. Da Ronch, G. Chen, Y. Li, Aeroelastic global structural optimization us-
trailing edge flap (vcctef) using overflow, in: 32nd AIAA Applied Aerodynamics ing an efficient cfd-based reduced order model, Aerosp. Sci. Technol. 94 (2019)
Conference, 2014, p. 2444. 105354.
[12] J. Fincham, M. Friswell, Aerodynamic optimisation of a camber morphing aero- [31] W. Zhang, Z. Ye, Effect of control surface on airfoil flutter in transonic flow,
foil, Aerosp. Sci. Technol. 43 (2015) 245–255. Acta Astronaut. 66 (2010) 999–1007.
[13] M. Arena, A. Concilio, R. Pecora, Aero-servo-elastic design of a morphing wing [32] W. Zhang, Z. Ye, Control law design for transonic aeroservoelasticity, Aerosp.
trailing edge system for enhanced cruise performance, Aerosp. Sci. Technol. 86 Sci. Technol. 11 (2007) 136–145.
(2019) 215–235. [33] E. Forster, B. Sanders, F. Eastep, Modelling and sensitivity analysis of a vari-
[14] J. Ko, A.J. Kurdila, T.W. Strganac, Nonlinear control of a prototypical wing sec- able geometry trailing edge control surface, in: 44th AIAA/ASME/ASCE/AHS/ASC
tion with torsional nonlinearity, J. Guid. Control Dyn. 20 (1997) 1181–1189. Structures, Structural Dynamics, and Materials Conference, 2003, p. 1807.
[15] J.S. Vipperman, R.L. Clark, M. Conner, E.H. Dowell, Experimental active control [34] J. Moran, An Introduction to Theoretical and Computational Aerodynamics,
of a typical section using a trailing-edge flap, J. Aircr. 35 (1998) 224–229. Courier Corporation, 2003.
[16] V. Rao, A. Behal, P. Marzocca, C. Rubillo, Adaptive aeroelastic vibration suppres- [35] J. Hollkamp, S. Batill, Automated parameter identification and order reduction
sion of a supersonic airfoil with flap, Aerosp. Sci. Technol. 10 (2006) 309–315. for discrete time series models, AIAA J. 29 (1991) 96–103.
[17] R. Huang, W. Qian, H. Hu, Y. Zhao, Design of active flutter suppression and [36] T.J. Cowan, A.S. Arena Jr, K.K. Gupta, Accelerating computational fluid dynam-
wind-tunnel tests of a wing model involving a control delay, J. Fluids Struct. ics based aeroelastic predictions using system identification, J. Aircr. 38 (2001)
55 (2015) 409–427. 81–87.
[18] Q. Zhou, Z. Yu, G. Cai, Delayed sub-optimal control for active flutter suppression [37] T. Ashish, Modern Control Design with Matlab and Simulink, Indian Institute of
of a three-dimensional wing, J. Fluids Struct. 80 (2018) 275–287. Technology, Kanpur, India, 2002, John Wiley & Sons.
[19] F. Marques, C. De Marqui, P. Caixeta, Active control of non-linear aeroelastic [38] R. Landon, NACA 0012 oscillatory and transient pitching, Technical Report, Air-
response using conformal control surfaces, in: 47th AIAA/ASME/ASCE/AHS/ASC craft Research Association Ltd, Bedford, United Kingdom, 2000.

11