Two-Degree-of-Freedom System, State-Space Method

Revision D

By Tom Irvine
Email: [email protected]

August 24, 2018

_______________________________________________________________________

Two-degree-of-freedom System
Consider the damped system in Figure 1.

p2(t)

u2

m2

k2 p1(t)
c2

u1
m1

k1 c1

Figure 1. Two-degree-of-freedom System

Structural dynamics systems can be represented in terms of mass, damping and

stiffness matrices. Each of these matrices may be coupled depending on the model
complexity, degrees-of-freedom, etc. The mass and stiffness matrices in the
assembled equation of motion may be uncoupled using the normal modes for the
undamped system. This approach gives real natural frequencies and real mode shapes.

1
Damping effects can be included in forced response analyses by implicitly assuming
that the damping matrix can be diagonalized into modal damping coefficients by the
undamped modes. But systems with dashpots in general have damping matrices
which cannot be uncoupled in this manner.
The state-space method is useful for modal and forced response analysis of systems
with discrete dashpot damping. This approach yields complex natural frequencies and
mode shapes, with real and imaginary components.

Derivation of Equations of Motion

p2(t) p1(t)
k2 (u 1-u 2) c2 (u1 − u 2 )

u2
u1
m2
m1

k2 (u1-u2) c2 (u1 − u 2 ) k1 u 1 c 2 u1

Figure 2. Free-body Diagrams

Determine the equation of motion for mass 2.

 F = m2 u 2 (1)

m 2 u 2 = p 2 (t) + k 2 ( u1 -u 2 ) + c 2 ( u1 -u 2 ) (2)

m 2 u 2 + c2 ( u 2 -u1 ) + k 2 ( u 2 -u1 ) = p 2 (t) (3)

Determine the equation of motion for mass 1.

2
  F = m1u1 (4)

m1 u1 = p1 (t) − k1u1 − k 2 ( u1 -u 2 ) − c1u1 − c1 ( u1 -u 2 ) (5)

m1 u1 + c1u1 + c2 ( u1 -u 2 ) + k1u1 + k 2 ( u1 -u 2 ) = p1 (t) (6)

m1 u1 + ( c1 + c 2 ) u1 - c 2 u 2 + ( k1 + k 2 ) u1 -k 2 u 2 = p 2 (t) (7)

The equations of motion in matrix form are

 m1 0   u1  c1 + c2 −c2   u1   k1 + k 2 −k 2   u1   p1 (t) 

 0 m   u  +  −c +
c2   u 2   −k 2
=
k 2   u 2   p2 (t) 
(8)
 2 2  2

The equations can be represented as

M u + Cu + K u = P (9)

where

m 0  c + c −c 2  k + k −k 2 
M= 1  , C= 1 2  , K= 1 2
 0 m2   −c 2 c2   −k 2 k 2 

u   p (t) 
u =  1, P =  1 
u 2   p 2 (t) 

The dimensions of the mass, damping and stiffness matrices are (N x N), where N is the
number of degrees-of-freedom. N = 2 for the example in this paper.

Let

3
 u(t)   v(t) 
z(t) =   , z(t) =   (10)
 v(t)   v(t) 

The vector v is defined as the effective velocity vector by the momentum equality.

M v(t) = M u(t) (11)

The equations can be expressed in state-space format as

 C M   v   K 0   u   p(t) 
 M 0   v  +  0 −M   v  =  0  (12)
       

The state-space equations can be simplified as

A z + Bz = F(t) (13)

where

 C M K 0   p(t) 
A=  , B=  , F(t) =  
M 0   0 −M   0 

The dimensions of the A and B matrices are (2N x 2N).

The A and B matrices are not positive definite. The eigenvalues and vectors are either
real or complex, with real and imaginary components. This implies that the modes are
not synchronous. There is a phase lag such that different degrees-of-freedom do not
simultaneously reach their corresponding peaks and valleys.

The free vibration problem is

A z + Bz = 0 (14)
4
Equation (14) represents a homogeneous set of ordinary differential equations with
constant coefficients. The solution can be expressed as

   
z(t) =  exp ( t ) =  u  exp ( t ) =  u  exp ( t ) (15)
 v   u 

   
z(t) =  exp ( t ) =   u  exp ( t ) =   u  exp ( t ) (16)
 v   u 

where λ is a scalar and  is a 2N vector

The corresponding generalized eigenvector problem is

 A  exp ( t ) + B exp ( t ) = 0 (17)

  A + B  = 0 (18)

The solution of this problem yields a set of 2N eigenvalues i , i=1, 2, … , 2N.

It also gives corresponding eigenvectors  i , i=1, 2, … , 2N.

The eigenvalues are found via

det   A + B = 0 (19)

The eigenvalues must either be real for complex conjugate pairs because the
coefficient matrices are real. Real eigenvalues indicate very high damping leading to
overdamped modes.

Furthermore, eigenvalues with complex conjugate pairs have corresponding

eigenvectors which are complex conjugates.

Consider i and j pairings of the generalized eigenvalue problem.

5
   i A + B  i = 0 (20)

 j A + B  j = 0 (21)
 

Premultiply each eigenvalue problem by  T such that

 Tj  i A + B i = 0 (22)

iT  j A + B  j = 0 (23)

A and B are both symmetric matrices thus

iT  i A + B  j = 0 (24)

Subtract equation (24) from (22).

 ( )
iT   j − i A   j = 0

(25)

(  j − i ) iT A  j = 0 (26)

For  j   i ,

iT A  j = 0 , i , j = 1, 2, … , 2N (27)

By substitution, for  j   i ,

iT B  j = 0 , i , j = 1, 2, … , 2N (28)

Equations (27) and (28) are orthogonality relationships.

The complex eigenvector modal matrix is

 =  1  2  2N  (29)

6
The orthogonality relationships yield the following modal diagonal terms a r and b r.

 T A  = diag ( a r ) (30)

 T B  = diag ( b r ) (31)

The eigenvalues are

br
r = − (32)
ar

The complex eigenvalues have the following relationships to the natural frequency r
and damping ratio  r ,

 r =  r + j r = − r r + j r 1 −  r2 (33)

r =  r 2 + r 2 (34)

 r = r r (35)

Frequency Response Function Derivation

The non-homogeneous state-space equation can be expressed as

Az +Bz = P (36)

where

 C M K 0  u  p
A=  , B=  , z =  , P =   (37)
M 0   0 −M  u  0

The 2N complex eigenvectors are

7
     
=  u  =  u  (38)
 v   u 

Now define a generalize coordinate η(t) such that

z =  (39)

A  + B   = P (40)

Premultiply by the transpose of the eigenvector matrix.

T A   + T B   = T P (41)

Recall

 T A  = diag ( a r ) (42)

 T B  = diag ( b r ) (43)

The individual modal equation is thus

a r r + b r r =  Tr P (44)

br 1 T
r + r = r P (45)
ar ar

br
r = − (46)
ar

8
 1 T
r −  r r = r P (47)
ar

Take the Laplace transform

1 
L r −  r r  = L   Tr P  (48)
ar 

1 T ˆ
s ˆ r (s) − r (0) −  r ˆ r (s) =  r P(s) (49)
ar

1 T ˆ
s −  r  ˆ r (s) − r (0) =  r P(s) (50)
ar

For zero initial conditions,

1 T ˆ
s −  r  ˆ r (s) =  r P(s) (51)
ar

1 ˆ
ˆ r (s) =  Tr P(s) (52)
a r s −  r 

2N
1
ˆ (s) = 
 ˆ
 Tr P(s) (53)
r =1 a r s −  r 

Change from the Laplace to the frequency domain with s = j  .

9
 2N
1
ˆ () = 
 ˆ )
 Tr P( (54)
r =1 a r  j −  r 

Recall

z =  (55)

The Fourier transform equivalent for the frequency domain is

Zˆ ( ) =  ˆ ( ) (56)

2N
ˆ ( ) =  1 ˆ )
Z  r  Tr P( (57)
r =1 a r  j −  r 

Recall

 Û ( ) 
 
Ẑ ( ) =   (58)
 j U ( ) 
 ˆ


    
=  u  =  u  (59)
 v   u 

p̂ ( ) 
P̂ ( ) =   (60)
 0 

2N
ˆ ( ) =  1
U (  u )r (  u )Tr p(
ˆ ) (61)
r =1 a r  j −  r 
10
 ˆ ( ) = H() p(
U ˆ ) (62)

2N
1
H() =  (  u )r (  u )Tr (63)
r =1 a r  j −  r 

The receptance function at coordinate i due to a harmonic force excitation only at

coordinate k is

2N
1
Hik () =  (  u )ir (  u )kr (64)
r =1 a r  j −  r 

The mobility function at coordinate i due to a harmonic force excitation only at

coordinate j is

2N
1
H v , i j() = j  (  u ) jr (  u )kr (65)
r =1 a r  j −  r 

The accelerance function at coordinate i due to a harmonic force excitation only at

coordinate j is

2N
1
Ha , i j() = − 
2
 a  j −   (  u ) jr (  u )kr (66)
r =1 r r

11
References

1. C. Hoen, An Engineering Interpretation of the Complex Eigensolution of Linear

Dynamic Systems, IMAC XXIII – Orlando, Florida, 2005.
2. R. Craig & A. Kurdila, Fundamentals of Structural Dynamics, Second Edition,
Wiley, New Jersey, 2006.

APPENDIX A

Example p2(t)

u2

m2

m1 2.0 lbm k2 p1(t)

c2
m2 1.0 lbm
u1
c1 0.2788 lbf sec/in m1
c2 0.161 lbf sec/in
k1 1500 lbf/in k1 c1

k2 1000 lbf/in

Figure A-1.

The system in Figure A-1 is analyzed via a Matlab script. Note that the mass matrix is
divided by 386 inside the Matlab script to convert lbm to lbf sec^2/in.

Mass Matrix Damping Matrix Stiffness Matrix

2 0 0.4398 -0.161 2500 -1000

0 1 -0.161 0.161 -1000 1000
12
A=

0.4398 -0.1610 0.0052 0

-0.1610 0.1610 0 0.0026
0.0052 0 0 0
0 0.0026 0 0

B=

1.0e+03 *

2.5000 -1.0000 0 0
-1.0000 1.0000 0 0
0 0 -0.0000 0
0 0 0 -0.0000

Eigenvalues

Complex: -14.13 + -396.1i Mag: 396.3 rad/sec f= 63.08 Hz

Complex: -14.13 + 396.1i Mag: 396.3 rad/sec f= 63.08 Hz
Complex: -59.38 + -841.4i Mag: 843.5 rad/sec f= 134.2 Hz
Complex: -59.38 + 841.4i Mag: 843.5 rad/sec f= 134.2 Hz

ModeShapes =

-0.0001 + 0.0015i -0.0001 - 0.0015i -0.0001 + 0.0010i -0.0001 - 0.0010i

-0.0001 + 0.0025i -0.0001 - 0.0025i 0.0001 - 0.0012i 0.0001 + 0.0012i
0.5908 + 0.0052i 0.5908 - 0.0052i 0.8397 - 0.0052i 0.8397 + 0.0052i
0.9963 + 0.0037i 0.9963 - 0.0037i -0.9953 - 0.0047i -0.9953 + 0.0047i

lambda1 = -14.13 + -396.1i

lambda2 = -14.13 + 396.1i
13
lambda3 = -59.38 + -841.4i
lambda4 = -59.38 + 841.4i

Figure A-2.

14
Figure A-3.

Figure A-4.

15