You are on page 1of 8

Two-Degrees-of-Freedom Systems

One-degree-of-freedom systems allow basic concepts, such as


frequency, damping, initial conditions, resonance and phase, etc.,
to be introduced and appreciated. However, real systems have
infinite number of degrees-of-freedom. On many occasions, these
systems can be approximated as having finite degrees-of-
freedom, the simplest of which has two degrees-of-freedom.
Multi-degrees-of-freedom systems afford a venue for
introducing eigenvectors (modes), matrices and vibration
absorption.

Free Vibration

A mass-spring system: k1 k2 k3
m1 m2

Free-body diagrams

x1 x2
k1x1 m1 k2(x1-x2) k2(x1-x2) m2 k3x2

Equations of motion

m1x1  k1 x1  k 2 ( x1  x2 ) m1x1  (k1  k 2 ) x1  k 2 x2  0


m2 x2  k3 x2  k 2 ( x1  x2 ) or m x  k x  (k  k ) x  0 (1)
2 2 2 1 2 3 2

1
In matrix form as

m1 0   x1  k1  k 2  k 2   x1  0


 0 m  x     k    x   0 (2)
 2  2   2 k 2 k 3  2   
or
Mx  Kx  0

Following the procedure for one-degree-of-freedom systems,


assume the solution in the form of

 x  a 
x   1    1  sin  t
 x2  a2 

This leads to

k  k  k2  2 1
m 0   a1  0
( 1 2      )    
  k 2 k 2  k3 
(3)
 0 m2  a2  0

The condition for a non-trivial solution of (K   2M)a  0 to exist is

k1  k2  m1 2  k2 
det  2
0 (4)
  k2 k2  k3  m2 
or
det(K   2M)  0 (5)
Notice  in equation (5) is the natural frequency (frequencies)
2
This is known in linear algebra as an (generalized) eigenvalue
problem. For matrices of very low orders (22, 33, 44), ‘hand’
calculation is feasible. For higher-orders matrices, suitable
algorithms and computer programs must be used.

For the above simple eigenvalue problem, the determinant


method gives (characteristic equation)

(k1  k2  m1 2 )( k2  k3  m2 2 )  k22  0


or
m1m2 4  [m1 (k2  k3 )  m2 (k1  k2 )] 2  k1k2  k1k3  k2 k3  0

This quadratic equation has two roots of 12 and  22 . Suppose


m1  m , m2  2m and k1  k2  k3  k . The above characteristic
equation become

2m 2 4  6mk 2  3k 2  0
k k
The two analytic solutions are 1  0.7962 m and 2  1.5382 m .

Equation (3) allows the ratio of the two amplitudes to be found as


a1 k 2k  2m 2
 
a2 2k  m 2 k
Substitution of one eigenvalue a time into equation (3) yields
(1) ( 2)
 a1   a1 
For 1 :  a   0.731 For 2 :  a   2.73
 2  2

3
Mode shapes

0.73 1 a1 1

a1 a2 a2
-2.73

Proportionality
A mode shape is represented by a vector, called eigenvector
mathematically. The absolute magnitude of individual elements
of an eigenvector is indefinite  an eigenvector multiplied by a
scalar is still an eigenvector. But the relative proportion of the
individual elements of an eigenvector is fixed. For this reason, an
eigenvector is normally normalized (scaled according to a rule)
and then becomes unique. Normalisation helps to graphically
display a mode, compare modes and facilitate computation.

Orthogonality
≠𝟎 𝒊=𝒌 ≠𝟎 𝒊=𝒌
𝐱 𝒊𝐓 𝐌𝐱 𝒌 = { 𝐱 𝒊𝐓 𝐊𝐱 𝒌 = {
=𝟎 𝒊≠𝒌 =𝟎 𝒊≠𝒌
Recall orthogonality between two vectors.

Normalisaton
(1) Let the maximum element be one and the other elements
xk
x   (k  1, 2, ......, n)
scaled as k
max ix
i

4
(2) Multiply a factor to an eigenvector so that x i Mx k  δ ik
T

(mass-normalisation)
 xi  1 .
2
(3) Scale an eigenvector so that i

A normalised eigenvector is called a normal eigenvector.


The two-degree-of-freedom system may vibrate in any one of
the two modes. In general, the motion of the system is a linear
combination of all modes as

(1) ( 2)
 x1 (t )  0.731  x1 (t )   2.73
   A1   sin( 0.796 t  1 )    A2   sin( 1.538 t  2 )
 x2 (t )  1   x2 (t )  1 

Factors A1 and A2, phase angles 1 and  2 , like in one-degree-of-


freedom systems, should be determined by the initial conditions.

Initial Conditions

For the linear system, given initial conditions as

 x1 (0)  2  x1 (0)  0


     and   
 x2 (0) 4  x2 (0) 0

Substituting them into

 x (t )  0.731  2.73
x(t )   1   A1   sin( 0.796 t  1 )  A2   sin( 1.538 t   2 )
 2 
x (t )  1   1 
leads to
5
2 0.731  2.73
   A1  sin 1  A2  sin  2
 
4  1   1 
0 0.731  2.73
   A 
1 1  cos 1  A 
2 2  cos  2
 
0  1   1 
The four unknowns can be determined and the solution is

 x (t )  2.732  0.732
x(t )   1     cos( 0.796 t )    cos(1.538 t )
 x2 (t ) 3.732  0.268 

x1(t)
0 x2(t)
0 10 20 30

-2

-4

For a system of n degrees-of-freedom, M and K are nn square


matrices. There usually exist n natural frequencies i and n
eigenvectors ψ i , by solving equation (5). The motion can be
written as
n
x(t )   Ai ψ i sin( it  i )
i 1

6
Rotational Systems

1 2 1 2

k1 k2 k3 k11 k2(1-2) k32

J1 J2 J1 J2

 J1 0  1   K1  K 2  K 2  1  0


 0 J        K   
K 2  K 3   2  0
 2   2   2

Coupled Pendulum

a
l

1 0 1  ka2  mgl  ka2  1  0 m m


          
2
ml 
     2  0
2 2
0 1  2   ka ka mgl

Two natural frequencies and modes can be found as


g g a2 k 1 1
1  , 2  2 2 ψ1    , ψ2   
l l l m 1  1

7
Damped Vibration

k1 k2 k3
m1 m2

m1 0   x1   c  c  x1  k1  k2  k2   x1  0


 0 m  x    c c   x     k    x   0
 2  2    2   2 k 2 k 3  2   
 x1 (t )   a1 
It is no longer valid to assume that  x (t )  a  sin t due to
 2   2
presence of damping. In general, the solution can be assumed as
 x1 (t )   a1 
     exp( t ) , where  is complex. This leads to
 2  a2 
x (t )

k  k  k2   c  c 2 m1 0   a1  0
( 1 2         )    
  k2 k 2  k3   c c   0 m2  a2  0
Suppose m1  m , m2  2m and k1  k2  k3  k . The above equation
becomes
2m 24  3cm3  6mk2  2ck  3k 2  0
k c
Introduce a new variable   m
, 
2m
,   z . A new equation
appears as
2 z 4  6z 3  6 z 2  4z  3  0
Two roots at   5% : z1, 2  0.0014  i0.80 ; z3, 4  0.0736  i1.54 ;
k k
eigenvalues: 1, 2  (0.0014  i0.80) m ; 3, 4  (0.0736  i1.54) m .

You might also like