Peretz P.

Friedmann
AE544 - 2005
26 CHAPTER 2. INTRODUCTION TO STRUCTURAL DYNAMICS AND A SIMPLE FLUTTER MODEL

the approximate treatment of wing flutter. Subsequent treatments will refine this

approach.

For the sake of simplicity, let us assume that the whole wing of an airplane can

be represented by a typical cross section (in some primitive aeroelastic analyses using

this method, the typical cross section was chosen as the cross section at 75% of the

wing span).

b b undeflected
U centerline
L Kα
x
EA
h, q1
α, q2 MAC
b/2 c.g.
ba bxα
r

Floor of wind tunnel Guide

Kh

Figure 2.2: : Two degree of freedom typical airfoil geometry for simple flutter calcu-

lation.

The geometry of the problem is shown in Fig. 2.2, where

ba elastic axis (E.A.) offset from midchord

bxα cross sectional center of gravity offset from elastic axis (E.A.),

positive aft
 Peretz P. Friedmann
AE544 - 2005
SECTION 2.3 27

Kh , Kα spring constants in bending and torsion respectively

h bending deflection, positive down, also q1 degree of freedom

α pitching (torsional) deflection positive leading edge up, also q2

degree of freedom

r running coordinate shown in Fig. 2.2

The equations of motion which will be described below are linear or small displace-

ment equations, which would be equivalent to a set of nonlinear equations linearized

about a trivial or zero equilibrium position. For this particular problem, it is conve-

nient to derive the equation of motion using Lagrange’s equations.

The kinetic energy per unit length is given by:

! !
1 1
T = ẇ2 dm = (ḣ + rα̇)2 dm
2 chord 2 chord
!
1
= (ḣ2 + 2ḣrα̇ + r2 α̇2 )dm
2 chord
Iα 2 1
= α̇ + mḣ2 + Sαḣα̇ (2.41)
2 2

where
m = mass per unit length of the wing
" 2
chord r dm = Iα = mass moment of inertia about the E.A.
"
chord rdm = Sα = mbxα = static mass moment of wing segment about elastic

axis

The elastic or strain energy is given by

1 1
U = Kh h2 + Kα α2 (2.42)
2 2
 Peretz P. Friedmann
AE544 - 2005
28 CHAPTER 2. INTRODUCTION TO STRUCTURAL DYNAMICS AND A SIMPLE FLUTTER MODEL

Lagrange’s equations are

# $
d ∂T ∂T ∂U
− + = Qr (r = 1, 2) (2.43)
dt ∂ q̇r ∂qr ∂qr

where q1 = h, q2 = α and Qr is a generalized nonconservative force. Combining the

last three equations, one has

# $ ⎫
d ∂T ∂U ⎪
⎪
⎪
+ = Qh ⎪
⎬
dt ∂ ḣ ∂h
# $ (2.44)
d ∂T ∂U ⎪
⎪
⎪
⎪
+ = Qα ⎭
dt ∂ α̇ ∂α
or ⎫
⎪
⎪
mḧ + Sα α̈ + Kh h = Qh ⎪
⎬
(2.45)
⎪
⎪
⎪
⎭
Sα ḧ + Iαα̈ + Kα α = Qα
Furthermore, using the definition of the natural frequencies for the system, one has

Kh = mωh2 , Kα = Iαωα2 (2.46)

The generalized loads Qh and Qα can be replaced by the aerodynamic lift and the

aerodynamic moment

Qh = −L , Qα = My (2.47)

In order to proceed one has to make certain assumptions regarding the calculation

of the aerodynamic forces. The simplest assumption one can make is the so-called

quasi-static aerodynamic assumption.

The aerodynamic loads can be written as

L = qcCLααef f (2.48)

MA = qc2CM α̇α̇ (2.49)

where q = ρU 2 /2 = dynamic pressure, ρ=density of air, and U = velocity of flight.

 Peretz P. Friedmann
AE544 - 2005
SECTION 2.3 29

Equations (2.48) and (2.49) also contain the assumption of incompressible flow,

i.e. low speed flight

∂CL
CLα = lift curve slope =
∂α
CL = lift coefficient

αef f = effective angle of attack

c = 2b = chord

CM α̇ = damping in pitch

Note that both L and M represent force and moment per unit length, respectively.

The effective angle of attack is given by

ḣ
αef f = α + = α + ∆α (2.50)
U

The quasi-static assumption is evident in the instantaneous effective change in

angle of attack due to the motion of the ∆α = ḣ/U. The aerodynamic forces can be

written as:
# $
ḣ
Qh = −L = −qcCLα α+ − LD (t) (2.51)
U

where LD (t) is a force due to some external disturbance such as a gust.

# $
b
Qα = MEA = + ba L + MAC
2
) # $ *
1 ḣ
≈ b( + a) qcCLα α + + L (t) + qc2CM α̇ α̇
D
(2.52)
2 U

The approximate symbol in Eq. (2.52) is due to the assumption that for a sym-

metric airfoil, the part of MAC associated with αef f is approximately zero. The last

term in Eq. (2.52) represents a damping in pitch, that is known to exist, based on

simple physical reasoning.

 Peretz P. Friedmann
AE544 - 2005
30 CHAPTER 2. INTRODUCTION TO STRUCTURAL DYNAMICS AND A SIMPLE FLUTTER MODEL

Combining Eqs. (2.45), (2.51) and (2.52) one has

# $
ḣ
mḧ + Sα α̈ + mωh2 h = −qcCLα α+ − LD (t)
U
+ ,) # $ *
1 ḣ
Sα ḧ + Iαα̈ + Iα ωα2 α = b +a qcCLα α+ + L (t) + qc2CM α̇α̇
D
2 U

which yields the following expressions

ḣ
mḧ + Sα α̈ + mωh2 h + qcCLα + qcCLαα = −LD (t)
U
+ ,
1 ḣ
Sα ḧ + Iαα̈ + Iαωα2 α −b +a − qc2CM α̇α̇
2 U
- + , . + ,
2 1 1
+ Iαωα − b + a qcCLα α = b + a LD (t) = M D (t)
2 2

It is convenient to rewrite these equations in matrix form

⎡ ⎤⎧ ⎫ ⎡ ⎤⎧ ⎫
⎪
⎪
⎪
⎪
⎪
⎪
qcCLα ⎪
⎪ ⎪
⎪ ḣ ⎪
⎪
⎢
⎢
m Sα ⎥
⎥
⎨ ḧ ⎬ ⎢
⎢
0 ⎥⎨
⎥
⎬
⎢ ⎥ +⎢ U ⎥
⎣ ⎦⎪ ⎪ ⎣ + , ⎦⎪ ⎪
⎪ ⎪ 1 qcCLα ⎪ ⎪
Sα Iα ⎪ ⎩ α̈ ⎪
⎭ −b +a −qc2CM α̇ ⎪⎩ α̇ ⎪
⎭
2 U
⎡ ⎤⎧ ⎫ ⎧ ⎫
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎢ mωh2 qcCLα ⎪
⎥⎨ h ⎬⎪ ⎪
⎨ −LD (t) ⎪
⎬
⎢ ⎥
+⎢ + , ⎥ =⎪ (2.53)
⎣ 1 ⎦⎪⎪ ⎪
⎪ ⎪ ⎪
⎪
0 Iα ωα2 −b ⎪
+ a qcCLα ⎩ α ⎭ ⎪⎪ ⎩ M D (t) ⎪
⎭
2
Equation (2.53) represents the general aeroelastic problem in its entirety for the rel-

atively simplified case with which we are dealing.

The following problems related to this aeroelastic system can be posed.

(a) What is the response of the aeroelastic system to external disturbances such

as gusts?

(b) What are the conditions under which the aeroelastic system could become

unstable?

(a) The Dynamic Response Problem. Given LD (t), M D (t) and the initial con-

ditions, these equations, being linear, can be easily solved given the flight condition as
 Peretz P. Friedmann
AE544 - 2005
SECTION 2.3 31

specified by U. For this problem, the relevant quantities are h(t) and α(t). Knowing

these and the properties of the structure one can evaluate the dynamic loads, stresses

and ultimately the fatigue life of the structure.

(b) The Dynamic Stability Problem. For this case, it is sufficient to look at

the homogeneous system, i.e. LD (t) = M D (t) = 0. The stability of the system

is determined from the solution of a characteristic equation or eigenvalue problem.

Before proceeding, the following comments are relevant:

(1) Looking at Eq. (2.53), it is clear that the aerodynamic forces provide damping

terms and therefore they effect to a great extent the stability of the system. In

addition, the stiffness matrix (or “spring matrix”) also contains aerodynamic

terms which obviously can further effect system behavior.

(2) Setting the dynamic pressure q = 0 in the homogeneous problem, yields the free

vibration problem for coupled bending and torsion of the cross section. These

equations are inertially coupled due to the offset between the cross sectional

center of mass and the elastic axis.

(3) The structural degrees of freedom participating in this problem are bending and

torsion, therefore this problem is usually denoted the “bending-torsional-flutter

problem”.

First, it is instructive to consider the system of Eqs. (2.53) with the damping

terms neglected.
 Peretz P. Friedmann
AE544 - 2005
32 CHAPTER 2. INTRODUCTION TO STRUCTURAL DYNAMICS AND A SIMPLE FLUTTER MODEL

Denoting by {q} the vector of generalized coordinates, we have

⎧ ⎫
⎪
⎪ ⎪
⎪
⎪
⎨ h ⎪
⎬
= {q}
⎪
⎪ ⎪
⎪
⎪
⎩ ⎪
⎭
α

Eq. (2.53) can be rewritten in a more compact manner as

[M]{q̈} + [G]{q̇} + [K]{q} = {Q} (2.54)

The definition of the matrices [M], [G], [K] are obvious when the terms in Eq. (2.53)

are compared to those in Eq. (2.54). Neglecting the damping and forcing terms, i.e.

[G] = {Q} = 0

and assuming the solution in the form of

{q} = {q}ept (2.55)

results in the following equation

⎡ ⎤
⎢
⎢
(mp2 + Kh ) (Sα p2 + qcCLα) ⎥
⎥
⎢ + , ⎥ {q} =0 (2.56)
⎣ 1 ⎦
2 2
Sα p Iα p + Kα − b + a qcCLα
2
or

[H]{q} = 0 (2.57)

The characteristic equation is obtained from the condition

Det([H]) = 0

which yields the following relation

8 - + , .9
1
(mIα − Sα2 )p4 + IαKh − SαqcCLα + m Kα − b + a qcCLα p2
2
- + , .
1
+Kh Kα − b + a qcCLα = 0 (2.58)
2
 Peretz P. Friedmann
AE544 - 2005
SECTION 2.3 33

which is a quadratic equation for p2

Ap4 + Bp2 + C = 0 (2.59)

where the definitions of A, B and C are clear by comparing Eqs. (2.58) with (2.59).

The roots of Eq. (2.59) are

√
2 −B ± B 2 − 4AC
p = (2.60)
2A
1 2
It is important to notice that A is constant while B and C change as q = ρU
2
increases. For q = 0, we have

B = IαKh + mKα , C = Kh Kα , A = mIα − Sα2

B 2 − 4AC = (IαKh + mKα )2 − 4Kh Kα (mIα − Sα2 )

= Iα2 Kh2 + 2mIαKα Kh + m2 Kα2 − 4Kh Kα mIα + 4Kh Kα Sα2

= (IαKh − mKα )2 + 4Kh Kα Sα2 > 0

and p2 < 0 resulting in two purely harmonic roots corresponding to the free vibration

problem.

For q ̸= 0, A is always positive, and the signs of B and C determine the solution.

Next, consider the signs of C and B. For flight at speeds U < UD or q < qD , C is

always positive, since it contains within it the static or divergence case. Thus,

- + , .
1
Kα − b + a qcCLα > 0
2

The quantity B may be either positive or negative

8 - + ,.9
1
B = IαKh + mKα − qcCLα Sα + mb +a (2.61)
2
 Peretz P. Friedmann
AE544 - 2005
34 CHAPTER 2. INTRODUCTION TO STRUCTURAL DYNAMICS AND A SIMPLE FLUTTER MODEL

+ ,
1
From Eq. (2.61), if mb + a + Sα < 0, then B > 0 for all q’s. When B < 0, it
2
implies that
- + ,.
Iα Sα 1
Kα + Kh − qcCLα +b +a <0 (2.62)
m m 2

Consider the two cases B > 0 and B < 0 in detail. When B > 0, then the values

of p2 from Eq. (2.60) are real and negative provided that B 2 − 4AC > 0, hence values

of p are purely imaginary representing neutrally stable oscillations ( since we have

neglected damping, this is the only possibility!)

On the other hand, if B 2 − 4AC < 0, the values of p2 are complex, for example,

p2 = a1 + ib1 which implies

p2 = reiθ = r(cos θ + i sin θ)

where
1 b1
r = (a21 + b21) 2 and θ = tan−1
a1

therefore
1
p1,2 = ±r 2 eiθ/2 (2.63)

hence at least one value of p will have a positive real part indicating unstable motion.

Thus, the condition

B 2 − 4AC = 0 (2.64)

represents the boundary between neutrally stable and unstable motion, and Eq. (2.64)

enables one to compute the dynamic pressure at which dynamic instability or flutter

occurs. Using Eq. (2.61), Eq. (2.58), and Eq. (2.64),

8 - + ,.92
2 1
B − 4AC = IαKh + mKα − qcCLα Sα + mb +a
2
 Peretz P. Friedmann
AE544 - 2005
SECTION 2.3 35

- + , .
1
−4(mIα − Sα2 )Kh Kα − b + a qcCLα
2
- + ,.2
1
= Iα2 Kh2 + m2 Kα2 + q 2c2 CLα
2
Sα + mb +a + 2Iα Kh mKα
2
- + ,. - + ,.
1 1
−2qcCLα Sα + mb + a IαKh − qcCLα Sα + mb + a mKα
2 2
+ ,
1
−4(mIα − Sα2 )Kh Kα + 4(mIα − Sα2 )b + a qcCLα = 0
2

- + 2 ,. 8 + ,
1 1
q 2c2 CLα
2
Sα + mb +a + q 4(mIα − Sα2 )b + a cCLα Kh
2 2
- + ,. 9
1
−2cCLα Sα + mb + a (Iα Kh + mKα) + [Iα2Kh2 + m2 Kα2
2
−4(mIα − Sα2 )Kh Kα + 2IαKh mIα] = 0 (2.65)

or

q 2D + qE + F = 0 (2.66)

where

- + 2 ,.
2 2 1
D = CLα c Sα + mb +a
2
8 + , - + ,. 9
1 1
E = 4(mIα − Sα2 )Kh b + a cCLα−2cCLα Sα + mb + a (IαKh + mKα )
2 2
F = Iα2 Kh2 + m2 Kα2 − 4(mIα − Sα2 )Kh Kα + 2Iα Kh mKα

= (IαKh + mKα )2 − 4(mIα − Sα2 )Kh Kα

and
√
−E ± E 2 − 4DF
qF = (2.67)
2D

For flutter to occur, at least one of the qF ’s determined from Eq. (2.67) must

be real and positive. If Eq. (2.67) yields two real qF ’s, the smaller of the two cor-

responds to the critical flutter condition. If no real qF is obtained, flutter does not

occur. Pines [Pines, S., “An Elementary Explanation of the Flutter Mechanism”,
 Peretz P. Friedmann
AE544 - 2005
36 CHAPTER 2. INTRODUCTION TO STRUCTURAL DYNAMICS AND A SIMPLE FLUTTER MODEL

Proceedings of National Specialists Meeting on Dynamics and Aeroelasticity, Insti-

tute of the Aeronautical Sciences, Ft. Worth, Texas, November 1958, pp.52-58] has

studied this example, and derived a number of results. The most important result is

that for

Sα ≤ 0

i.e. center of mass ahead of the elastic axis, no flutter occurs. Conversely, when

Sα increases in a positive sense, i.e. center of mass behind the elastic axis, qF at

which flutter occurs, is decreased. In practice, mass is often added to a flutter prone

structure, so as to decrease Sα and raise qF . Such a structure is said to be “mass

balanced”.

Next, consider the case when B < 0; we have shown previously that B > 0 for

q = 0, thus it will only become negative for sufficiently large values of q. However,

B 2 − 4AC = 0 will occur before B = 0, since A > 0 and C > 0. Hence to determine

flutter, only the case B > 0 has to be considered.

For some cases in absence of damping, dynamic instability occurs when the fre-

quencies coalesce (i.e. merge), this situation is shown in the schematic diagram below

on which, both the undamped case and the damped cases are depicted. The un-

damped case is shown in full lines, while the damped case is illustrated by the broken

lines.

It should also be noted that frequency coalesce is not a stringent requirement for

flutter to occur. In some special cases, such as a jet engine fan blade, flutter can

occur without any noticeable frequency coalesce. Also, in some cases, single degree
 Peretz P. Friedmann
AE544 - 2005
SECTION 2.3 37

of flutter can occur in two degree of freedom systems.

no damping
(ωT ) ω2
with damping

(ωB ) ω1

qF 1 qF 2

Figure 2.3: Flutter boundary without damping (full line) and with damping (dashed

line).

Since for the undamped case, flutter occurs when aerodynamic damping is ne-

glected, it obviously is entirely due to the nonsymmetric stiffness matrix.

When damping is included, it is usually convenient to rewrite Eq. (2.54) in state

variable form, i.e. define

{q̇} = {y1} and {q} = {y2}

Then also,

y2 = {ẏ2} = {y1} (2.68)

Furthermore, Eq. (2.54) can be rewritten as

[M]{ẏ1} + [G]{y1} + [K]{y2} = 0

y1 = {y1} = −[M]−1[G]{y1} − [M]−1[K]{y2}

 Peretz P. Friedmann
AE544 - 2005
38 CHAPTER 2. INTRODUCTION TO STRUCTURAL DYNAMICS AND A SIMPLE FLUTTER MODEL

From Eq. (2.68),

{ẏ2} = [I]{y1}
: ;
y1
Defining the state vector as {y} = , the last two equations can be combined as
y2
⎡ ⎤
: ;
ẏ1 ⎢ −[M]−1[G] −[M]−1 [K] ⎥
⎢ ⎥
{ẏ} = = ⎢ ⎥ {y} (2.69)
ẏ2 ⎣ ⎦
[I] 0

which can be more compactly written as

{ẏ} = [A]{y} (2.70)

Taking the solution in the form of

{ẏ} = {y}ept (2.71)

Equation (2.70) yields the usual standard eigenvalue problem

[A]{y} = p{y} (2.72)

Solution of this eigenvalue problem will result in four eigenvalues which occur in com-

plex conjugate pairs. Recall that this is a parametric eigenvalue problem because the

elements of the matrix [A] are a function of the dynamic pressure q. Stability of the

system is investigated as q is increased and following possibilities can occur:

low q p = −λ1 ± iω1 ; −λ2 ± iω2 stable

higher q p = −λ1 ± iω1 ; ±iω2 neutrally stable

higher q p = −λ1 ± iω1 ; λ2 ± iω2 dynamic instability flutter

For some cases, the difference in the plot of the frequencies versus the flutter

parameter is quite similar with or without damping as shown in the previous figure.

However, this is generally not the case.

 Peretz P. Friedmann
AE544 - 2005
SECTION 2.3 39

Finally, it is important to note that Eqs. (2.53) or Eqs. (2.58) contain also the
d
static aeroelastic or divergence problem. For this case, = 0 or p = 0, thus from
dt
Eq. (2.58)
- + , .
1
Kh Kα − b + a qcCLα = F = 0
2

or
Kα
qD = + , (2.73)
1
b + a cCLα
2
ρUD2 Kα
= + ,
2 1
b + a cCLα
2
Divergence speed is given by
<
= 2Kα
=
UD = = + , (2.74)
> 1
ρb + a cCLα
2
It is important to note that no divergence occurs when the offset between elastic

axis and the aerodynamic center is equal to zero, i.e.

+ ,
1
b +a =0
2