II.

THEORETICAL DEVELOPMENT

A. A FLUTTER THEORY FOR FORWARD FLIGHT

Shipman and Wood, using the theory described in the previous section derived an
equation analogous to Theodorsen and Loewy that accounted for the effects of forward
flight and a shed skewed helical wake. However, evaluation of some of the integrals
required for the ∆F and ∆W terms is extremely difficult. If the view is taken that the
Theodorsen method, which neglects the effect of the shed wakes, is just a first-order
approximation to the rotary-wing flutter problem, then use of the Theodorsen lift
deficiency function should yield suitable first-order results. Therefore, this thesis will
rely on Theodorsen’s lift deficiency function, while following the methods of Shipman
and Wood in modeling forward flight.
1. Mode Shapes and Natural Frequencies

Blade bending frequencies and mode shapes used in this analysis were determined
by a simplified method by assuming a uniform stiffness and mass distribution along the
blade. A more exact and detailed analysis would be required to account for such details
such as local changes in stiffness and mass distribution due to blade features such as
doublers near the blade root and outboard blade balance weights near the tip. While these
details would be desired for an actual blade design, they can be viewed as second order
effects and not necessary for the first order flutter analysis.

For the simplified model, the geometric and inertial properties of the subject blade
are averaged between 20% and 90% of its length. The Fourier-based solution of the
pinned-free uniform beam from Young and Felgar [Ref. 11] are applied to determine the
non-rotating mode shapes:

1
yn =
2
( cosh(r) + cos(r) − A n ( sinh(r) + sin(r) ) )
where An = 1.000777, 1.000001 and 1.0 for the first three bending modes. Then the
method given by Den Hartog [Ref. 12] for determining non-rotating natural frequencies
is applied:

9
 EI xx
ωn = a n
µR 4

where an = 15.4182, 49.9649 and 104.2477 for the first three bending modes, and µ is the
mass per unit length.

Given the natural frequencies, the rotational velocity of the rotor head, and the
non-rotating mode shapes, Yntema [Ref. 13] is called on to supply the rotating natural
frequencies. Figures 5 through 7, taken from Yntema’s report, compares rotating and
non-rotating mode shapes for the first three bending modes of a pinned-free beam to
validate the assumption that the non-rotating mode shape is a close approximation of the
rotating mode shape.

Figure 5. Rotating and Non-Rotating Mode shapes for the First Bending Mode

10
Figure 6. Rotating and Non-Rotating Mode shapes for the Second Bending Mode

Figure 7. Rotating and Non-Rotating Mode Shapes for the Third Bending Mode

11
 Yntema notes that an exact value for the nth bending frequency of a beam rotating
at any rotational speed, Ω, can be found if the nth natural bending mode shape is known
for this value of rotational speed. He obtains his frequency equation by equating the
kinetic energy at zero displacement to the potential energy of both the bending and
centrifugal forces at maximum displacement for vibration perpendicular to the plane of
rotation.
l l
∫ EIy′′n 2 dx ∫ T y′
2
dx
0 1 n
ω2
Rn = 0
l
+ l
(1)
∫ my ∫ my
2 2
n dx n dx
0 0

where n refers to mode of vibration and

l
T1 = ∫ ( η + e ) m dη
x

He goes on to point out that while the rotating mode shape is unknown, a close
approximation to the rotating natural frequency can be obtained by making use of
Rayleigh’s Principle, and using the non-rotating mode shape in equation (1). The report
states that the non-rotating mode shape is consistent with the constraints of the system (in
this case a pinned-free beam). If the nth mode of the non-rotating mode shape, yn, is
substituted into equation (1), the first term becomes exactly the square of the bending
frequency of the non-rotating beam. By denoting the ratio in the second term by Kn, a
Southwell coefficient, the form of the frequency equation becomes:

ω2 = ω2 + K n Ω2 (2)
Rn NR n

To account for blade offset, e, subdivide Kn into two independent parts:

K n = K 0n + K1n e (3)

where K0n is referred to as the zero-offset Southwell coefficient and K1n is referred to as
the offset-correction factor for the Southwell coefficient. As is frequently done, it is
convenient to write the square of the non-rotating frequency in terms of a non-rotating-
beam frequency coefficient, an, and the mass and stiffness of the beam as:

EI0
ω2NR n = a n2 (4)
m 0l4
12
Combining equations (2), (3) and (4) yield:

(
ω2R n = ω2NR n + K 0n + K1n e Ω 2)
Yntema’s report [Ref. 13] gives charts that provide an, K0n and K1n which, in
conjunction with the mass and stiffness of the beam at the root, the length of the beam,
the hinge offset, and the rotational speed, permit rapid estimation of the first three
bending frequencies of rotating beams with hinged or cantilevered root-end support. As
previously noted, once the frequencies have been found, the rotating beam mode shapes
can then be approximated by non-rotating mode shapes which are defined by the Fourier-
based solutions contained in Young and Felgar [Ref. 11].

The next problem is to determine the mode shapes and frequencies for the
uniform beam blade model in the torsional case. For this, Den Hartog [Ref. 13] provides
a relatively simple solution. Given that the torsional stiffness along the beam is constant,
the torsional mode shapes are given by

 1 
  n − 2  πr 
y n = sin    
 R 
 
 

The non-rotating torsional natural frequencies are given by

 1 GJ
ωNRn =  n −  π ' 2
 2  IαR

Bramwell [Ref. 14], gives the exact solution for rotating natural frequencies in the
torsional case as

ω2R n = ω2NR n + Ω 2

2. Normal Mode Analysis

With the rotor blade free vibration problem solved, the bending deflection may be
defined as

13
 N
h ( x, t ) = ∑ f n ( x ) q n ( t )
n =1

where fn(x) is the characteristic function (mode shape) for the nth vertical bending mode
of the rotor blade. The quantities qn(t) can be considered as weighting functions for each
mode that contributes to the deflection. They are called the normal coordinates since
they can be shown to reduce the kinetic and potential energy expressions to sums of
squares of the coordinates with no cross product terms.

The corresponding torsional deflection of the rotor blade can be written in terms
of the blade torsion modes as:
N
α ( x, t ) = ∑ Fn ( x ) q n ( t )
n =1

where Fn(x) is the characteristic function of the nth torsional mode of the rotor blade and
qn(t) is the corresponding normal coordinate. Consider the five-degree of freedom (DOF)
case with three bending modes and two torsion modes. The bending and torsional
deflections can be written as:

h ( x, t ) = h1 ( t ) f1 ( x ) + h 2 ( t ) f 2 ( x ) + h 3 ( t ) f 3 ( x )

and

α ( x, t ) = α1 ( t ) F1 ( x ) + α 2 ( t ) F2 ( x )

where
f1(x) = 1st vertical bending mode
f2(x) = 2nd vertical bending mode
f3(x) = 3rd vertical bending mode
F1(x) = 1st torsion mode
F2(x) = 2nd torsion mode
Now all the tools are in place to begin the detailed flutter analysis, which is the
combination of Lagrange’s equation with Theodorsen’s work to provide the equations of
motion.

14
 3. Lagrange and The Equations of Motion

Lagrange’s equation is given as:

d  ∂T  ∂T ∂U ∂D
 − + + = Qn
dt  ∂q n  ∂q n ∂q n ∂q n

where T ≡ kinetic energy, U ≡ potential energy, D ≡ dissipation function, and Qn ≡

generalized force. For the five DOF case,

1 2 1 1 1 1
T= M1h1 + M 2 h 22 + M 3h 32 + Iα1 α 12 + Iα2 α 22
2 2 2 2 2
  
+Sα11 h1α 1 + Sα12 h1α 2 + Sα21 h 2α 1
+Sα22 h 2α 2 + Sα31 h 3α 1 + Sα32 h 3α 2

1 1 1
U= M1ωh21 h12 + M 2ω2h 2 h 22 + M 3ω2h3 h 32
2 2 2
1 1
+ Iα1 ωα2 1 α12 + Iα2 ωα2 2 α 22
2 2

M1g h1 ωh21 h 12 M 2g h 2 ωh2 2 h 22 M 3g h3 ωh23 h 32

D= + +
2ω 2ω 2ω
Iα1 g α1 ωα2 1 α 12 Iα2 g α2 ωα2 2 α 22
+ +
2ω 2ω

and the generalized forces including aerodynamic terms are defined as:

Q h1 = πρω2 ( A11h1 + A12 h 2 + A13h 3 + A14α1 + A15α 2 )

Q h 2 = πρω2 ( A 21h1 + A 22 h 2 + A 23h 3 + A 24α1 + A 25α 2 )
Q h3 = πρω2 ( A 31h1 + A 32 h 2 + A 33h 3 + A 34α1 + A 35α 2 )
Qα1 = πρω2 ( A 41h1 + A 42 h 2 + A 43h 3 + A 44α1 + A 45α 2 )
Qα2 = πρω2 ( A 51h1 + A 52 h 2 + A 53h 3 + A 54α1 + A 55α 2 )

The expressions for aerodynamic terms that couple the modes together and incorporate
Theodorsen’s lift deficiency function are given as

15
 l
A11 = ∫ b 2 f1 ( x )  L h dx
2
(5)
0

l
A12 = ∫ b 2f1 ( x ) f 2 ( x ) L h dx
0

l
A13 = ∫ b 2f1 ( x ) f 3 ( x ) L h dx
0

l  1  
A14 = ∫ b3f1 ( x ) F1 ( x )  Lα −  + a  L h  dx
0
 2  

l  1  
A15 = ∫ b3f1 ( x ) F2 ( x )  Lα −  + a  L h  dx
0
 2  

l
A 21 = ∫ b 2f 2 ( x ) f1 ( x ) L h dx = A12
0

l
A 22 = ∫ b 2 f 2 ( x )  L h dx
2

l
A 23 = ∫ b 2f 2 ( x ) f 3 ( x ) L h dx
0

l  1  
A 24 = ∫ b3f 2 ( x ) F1 ( x )  Lα −  + a  L h  dx
0
 2  

l  1  
A 25 = ∫ b3f 2 ( x ) F2 ( x )  Lα −  + a  L h  dx
0
 2  

l
A 31 = ∫ b 2f 3 ( x ) f1 ( x ) L h dx = A13
0

l
A 32 = ∫ b 2f 3 ( x ) f 2 ( x ) L h dx = A 23
0

l
A 33 = ∫ b 2 f 3 ( x )  L h dx
2

l  1  
A 34 = ∫ b3f 3 ( x ) F1 ( x )  Lα −  + a  L h  dx
0
 2  

l  1  
A 35 = ∫ b3f 3 ( x ) F2 ( x )  Lα −  + a  L h  dx
0
 2  

16
 l  1  
A 41 = ∫ b3F1 ( x ) f1 ( x )  M h −  + a  L h  dx
0
 2  

l  1  
A 42 = ∫ b3F1 ( x ) f 2 ( x )  M h −  + a  L h  dx
0
 2  

l  1  
A 43 = ∫ b3F1 ( x ) f 3 ( x )  M h −  + a  L h  dx
0
 2  
2 
2
l 1  1 
A 44 = ∫ b  F1 ( x )   M α −  + a  ( Lα + M h ) +  + a  L h  dx
4
0
 2  2  
l  1  1 
2

A 45 = ∫ b 4 F1 ( x ) F2 ( x )  M α −  + a  ( Lα + M h ) +  + a  L h  dx
0
 2  2  

l  1  
A 51 = ∫ b3F2 ( x ) f1 ( x )  M h −  + a  L h  dx
0
 2  

l  1  
A 52 = ∫ b3F2 ( x ) f 2 ( x )  M h −  + a  L h  dx
0
 2  

l  1  
A 53 = ∫ b3F2 ( x ) f 3 ( x )  M h −  + a  L h  dx
0
 2  

l  1  1 
2

A 54 = ∫ b F2 ( x ) F1 ( x )  M α −  + a  ( Lα + M h ) +  + a  L h  dx = A 45
4
0
 2  2  

2 
2
l 1  1 
A 55 = ∫ b  F2 ( x )   M α −  + a 
4
( Lα + M h ) +  + a  Lh  dx
0
 2  2  

The generalized masses of the three bending modes and two torsion modes can be
written as:
l
M1 = ∫ m ( x )  f1 ( x )  dx
2

l
M 2 = ∫ m ( x ) f 2 ( x )  dx
2

l
M 3 = ∫ m ( x ) f 3 ( x )  dx
2

17
 l
Iα1 = ∫ Iα ( x )  F1 ( x )  dx
2

l
Iα2 = ∫ Iα ( x )  F2 ( x )  dx
2

The coupled static unbalance terms are defined as:

l
Sα11 = ∫ Sα ( x ) f1 ( x ) F1 ( x ) dx
0

l
Sα12 = ∫ Sα ( x ) f1 ( x ) F2 ( x ) dx
0

l
Sα21 = ∫ Sα ( x ) f 2 ( x ) F1 ( x ) dx
0

l
Sα22 = ∫ Sα ( x ) f 2 ( x ) F2 ( x ) dx
0

l
Sα31 = ∫ Sα ( x ) f 3 ( x ) F1 ( x ) dx
0

l
Sα32 = ∫ Sα ( x ) f 3 ( x ) F2 ( x ) dx
0

First it is noted that the kinetic energy equation is only a function of the derivative
of the generalized displacement ( h n or α n ). Thus, Lagrange’s equation reduces to:

d  ∂T  ∂U ∂D
 + + = Qn
dt  ∂q n  ∂q n ∂q n

Applying Lagrange’s equation to each of the five DOFs yields the following five
equations:

M1ωh21 g h1 
M1
h1 + Sα11 α
1 + Sα12 α
 2 + M1ω2h1 h1 + h1 = Q h1
ω

M 2ω2h 2 g h 2 
M 2
h 2 + Sα21 α
1 + Sα22 α
 2 + M 2ωh2 2 h 2 + h 2 = Qh2
ω

 M 3ωh23 g h3 
M 3h 3 + Sα31 α
1 + Sα32 α
 2 + M 3ωh3 h 3 +
2
h 3 = Q h3
ω

18
    Iα1 ωα2 1 g α1
Iα1 α
1 + Sα11 h1 + Sα21 h 2 + Sα31 h 3 + Iα1 ωα1 α1 +
2
α 1 = Qα1
ω

   Iα2 ωα2 2 g α2
Iα2 α
 2 + Sα12 h1 + Sα22 h 2 + Sα32 h 3 + Iα2 ωα2 α 2 +
2
α 2 = Qα2
ω

If simple harmonic motion is assumed, that is: 

h n = −ω2 h n , h n = iωh n , α
 n = −ω2αn , and

α n = iωα n , and the expressions for Qhn and Qαn are substituted into the equations of

motion, the results are:

  ωh1  
2

(
 πρA11 + M1 − M1 1 + ig h1  )
  h1 + ( πρA12 ) h 2 + ( πρA13 ) h 3
  ω  

( ) (
+ πρA14 + Sα11 α1 + πρA15 + Sα12 α 2 = 0 )
  ωh 2  
2

( )
( πρA 21 ) h1 +  πρA 22 + M 2 − M 2 1 + ig h2    h 2 + ( πρA 23 ) h 3
  ω  

( ) (
+ πρA 24 + Sα21 α1 + πρA 25 + Sα22 α 2 = 0 )
  ωh3  
2

( πρA31 ) h1 + ( πρA32 ) h 2 +  πρA33 + M3 − M3 1 + ig h3    h 3( )

  ω  

( ) (
+ πρA 34 + Sα31 α1 + πρA 35 + Sα32 α 2 = 0 )
( πρA 41 ) ( ) (
+ Sα11 h1 + πρA 42 + Sα21 h 2 + πρA 43 + Sα31 h 3 )
  ωα1  
2

(
+  πρA 44 + Iα1 − Iα1 1 + ig α1  )
  α1 + πρA 45α 2 = 0
  ω  

( πρA 51 ) ( ) (
+ Sα12 h1 + πρA 52 + Sα22 h 2 + πρA 53 + Sα32 h 3 )
  ωα2  
2

+πρA 54α1 +  πρA 55 + Iα2 − Iα2 1 + ig α2  (   α2 = 0 )

  ω  

19
 The five equations to the flutter problem can be written in matrix form as
(6)

[ πρA11 + M1 
 
 ωh1  
2
 πρA12 πρA13 πρA14 + Sα11 πρA15 + Sα12 
( )
 −M1 1 + igh1   
ω

    
 [πρA22 + M2 
 
 πρA21  ωh2  
2
πρA23 πρA24 + Sα21 πρA25 + Sα22 
 (
−M2 1 + igh2 )   
  ω   
 h 
 [πρA33 + M3  1
   h2 
 ωh3  
2
πρA31 πρA32 πρA34 + Sα31 πρA35 + Sα32   h3  = 0

 (
−M3 1 + igh3  )
ω
   α1 
     
 πρ A + I  α2 
  44 α1

 πρA41 + Sα11 πρA42 + Sα21 πρA43 + Sα31  ωα1  
2
πρA45 
 (
−Iα1 1 + igα1 )  
  ω   
 
 πρA55 + Iα2 
 
 πρA51 + Sα12 πρA52 + Sα22 πρA53 + Sα32 πρA54  ωα2  
2

 ( )
−Iα2 1 + igα2   
  ω  

4. Eigenvalues

Equation (6) is a complex eigenvalue problem. In order to solve the flutter

problem, it is convenient to rewrite the problem in the form ( A − IZ ) X = 0 by arbitrarily

letting
2
 ωα 
Z =  1  (1 + ig ) ,
 ω 

which results in

A11 − Z A12 A13 A14 A15

A 21 A 22 − Z A 23 A 24 A 25
A 31 A 32 A 33 − Z A 34 A 35 =0
A 41 A 42 A 43 A 44 − Z A 45
A 51 A 52 A 53 A 54 A 55 − Z

20
The determinant elements are defined as:
2 2
 πρA11   ωα   πρA12   ωα1 
A11 =  + 1  1  A12 =    (7)
 M1   ωh1   M1   ωh1 

2 2
 πρA13   ωα1   πρA14 + Sα11   ωα1 
A13 =    A14 =    
 M1   ωh1   M1   ωh1 

2
 πρA15 + Sα12   ωα1 
A15 =   
 M1   ωh1 

2 2
 πρA 21   ωα1   πρA 22   ωα 
A 21 =    A 22 =  + 1  1 
 M 2   ωh 2   M2   ωh 2 

2 2
 πρA 23   ωα1   πρA 24 + Sα21   ωα1 
A 23 =    A 24 =    
 M 2   ωh 2   M2   ωh 2 

2
 πρA 25 + Sα22   ωα1 
A 25 =    
 M2   ωh 2 

2 2
 πρA 31   ωα1   πρA 32   ωα1 
A 31 =    A 32 =   
 M 3   ωh3   M 3   ωh3 

2 2
 πρA 33   ωα   πρA 34 + Sα31   ωα1 
A 33 =  + 1  1  A 34 =   
 M3   ωh3   M3   ωh3 

2
 πρA 35 + Sα32   ωα1 
A 35 =    
 M3   ωh3 

 πρA 41 + Sα11   πρA 42 + Sα21 

A 41 =   A 42 =  
 Iα1  Iα1
   

 πρA 43 + Sα31   πρA 44 

A 43 =   A 44 =  + 1
 Iα1  Iα 
   1 

21
  πρA 45 
A 45 = 
 Iα 
 1 

2 2
 πρA 51 + Sα12  ωα1   πρA 52 + Sα22  ωα1 
A 51 =    A 52 =   
 Iα2  ωα  Iα2  ωα
  2    2 

2 2
 πρA 53 + Sα32  ωα1   πρA 54  ωα1 
A 53 =    A 54 = 
 Iα2  ωα  Iα   ωα 
  2   2  2 

2
 πρA 55  ωα 
A 55 =  + 1 1 
 Iα  ωα
 2  2 

It should be noted that the coefficients of the characteristic equation of the

( A − IZ ) matrix (a quintic in Z) are complex, due to Theodorsen’s lift deficiency

function, and thus the eigenvalues will be complex. The coupled frequency of oscillation
(ω) for each eigenvalue can be found from the real part of Z, since the first torsional
natural frequency is already known:

ωα1
ωi =
Re ( Z )

The damping coefficient required for flutter to exist (g) for each eigenvalue can
be found from the imaginary part of Z:
2
 ω 
g i = Im ( Z )  i 
 ωα
 1 

If g is negative for the reduced frequency chosen, then damping must be

decreased in order to obtain flutter. Negative values of g represent the stable, or non-
flutter, condition. If g is positive, then damping must be increased to be at the flutter
point. Positive values of g represent the unstable, or flutter condition. When a plot of g
is made against 1/k (k being reduce frequency), there will be five curves corresponding to
the variation of each eigenvalue as the reduced frequency varies. Some of these curves
will have only values of g that are negative. These are the non-critical curves that
22
represent stable coupled modes and do not influence the flutter solution. However, at
least one curve will start with a negative value of g and then at some point cross the
abscissa (1/k) to a positive value of g. This curve is called the critical curve, and the
value of 1/k where this curve crosses the abscissa represents the critical flutter speed, or
flutter point. The critical flutter speed is found from the relationship:

ωb
U FL =
k crit

where ω is found from the real part of the eigenvalue relationship described above for the
critical curve evaluated at the reduced frequency that corresponds to the crossover point
(kcrit). Once the unstable mode is identified, results are commonly plotted as g vs. UFL.

23