Chapter 4 Multiple Degree of Freedom Systems

The Millennium bridge required

many degrees of freedom to model
and design with.

Extending the first 3 chapters to more

then one degree of freedom
The first step in analyzing multiple degrees of freedom (DOF)
is to look at 2 DOF
• DOF: Minimum number of coordinates to specify the position of a system

• Many systems have more than 1 DOF

• Examples of 2 DOF systems

– Car with sprung and unsprung mass (both heave)

– Elastic pendulum (radial and angular)

– Motions of a ship (roll and pitch)

– Airplane roll, pitch and yaw

4.1 Two-Degree-of-Freedom Model (Undamped)

A 2 degree of freedom system used to base much of the

analysis and conceptual development of MDOF systems on.
Free-Body Diagram of each mass

Figure 4.2

k (x -x )
2 2 1

m1 m2
k x
1 1

x x
1 2
Summing forces yields the equations of motion:

m1 x1 (t )  k1 x1 (t )  k2  x2 (t )  x1 (t ) 
(4.1)
m2 x2 (t )  k2  x2 (t )  x1 (t ) 

Rearranging terms:
m1 x1 (t )  (k1  k2 ) x1 (t )  k2 x2 (t )  0
(4.2)
m2 x2 (t )  k2 x1 (t )  k2 x2 (t )  0
Note that it is always the case that
• A 2 Degree-of-Freedom system has
– Two equations of motion!
– Two natural frequencies (as we shall see)!
• Thus some new phenomena arise in going from one to two
degrees of freedom
– Look for these as you proceed through the material
• Two instead of one natural frequency
– Leading to two possible resonance conditions
• The concept of a mode shape arises
The dynamics of a 2 DOF system consists of 2
homogeneous and coupled equations
• Free vibrations, so homogeneous eqs.
• Equations are coupled:
– Both have x1 and x2.
– If only one mass moves, the other follows
– Example: pitch and heave of a car model

• In this case the coupling is due to k2.

– Mathematically and Physically
– If k2 = 0, no coupling occurs and can be solved as two
independent SDOF systems
Initial Conditions

• Two coupled, second -order, ordinary differential equations

with constant coefficients
• Needs 4 constants of integration to solve
• Thus 4 initial conditions on positions and velocities

x1 (0)  x10 , x1 (0)  x10 , x2 (0)  x20 , x2 (0)  x20

Solution by Matrix Methods
The two equations can be written in the form of a
single matrix equation (see pages 272-275 if matrices and vectors are a
struggle for you) :

 x1 (t )   x1 (t )   x1 (t ) 
x(t )    , x(t )   x (t )  , x(t )   x (t ) 
 2 
x (t )  2   2  (4.4), (4.5)

m 0 k  k k2 
M  1  K  1 2 (4.6), (4.9)
k2 
,
0 m2   k2

m1x1(t )  (k1  k 2 )x1(t )  k 2x 2(t )  0

Mx  Kx  0 m2x2(t )  k 2x1(t )  k 2x 2(t )  0
Initial Conditions (two sets needed one for each equation of
motion)

IC’s can also be written in vector form

 x10   x10 
x(0)    , and x(0)   
 x20   x20 
The approach to a Solution:

For 1DOF we assumed the scalar solution aeλt Similarly, now

we assume the vector form:

Let x(t )  ue jt (4.15)

j  1, u  0,  , u unknown
  - 2 M  K  ue jt  0 (4.16)
  - 2 M  K  u  0 (4.17)
This changes the differential equation of motion into
algebraic vector equation:

 M  K u  0
- 2
(4.17)
This is two algebraic equation in 3 uknowns
( 1 vector of two elements and 1 scalar):
 u1 
u =   , and 
u2 
The condition for solution of this matrix equation
requires that the the matrix inverse does not exist:
If the inv  - 2 M  K  exists  u  0 : which is the
static equilibrium position. For motion to occur
u  0   - M  K 
2 1
does not exist
or det  - 2 M  K   0 (4.19)

The determinant results in 1 equation in one unknown ω

(called the characteristic equation)
Back to our specific system: the characteristic equation is
defined as

det  - 2 M  K   0 
  2 m1  k1  k2 k2 
det   0
  k2  m2  k2 
2

m1m2 4  (m1k2  m2 k1  m2 k2 ) 2  k1k2  0

Eq. (4.21) is quadratic in ω2 so four solutions result:

12 and 22  1 and  2

Once ω is known, use equation (4.17) again to calculate
the corresponding vectors u1 and u2
This yields vector equation for each squared frequency:
(12 M  K )u1  0 (4.22)
and
(22 M  K )u 2  0 (4.23)
Each of these matrix equations represents 2 equations in the 2
unknowns components of the vector, but the coefficient matrix is
singular so each matrix equation results in only 1 independent
equation. The following examples clarify this.
Examples 4.1.5 & 4.1.6:calculating u and ω

• m1=9 kg,m2=1kg, k1=24 N/m and k2=3 N/m

• The characteristic equation becomes
ω -6ω +8=(ω -2)(ω -4)=0
4 2 2 2

ω2 = 2 and ω2 =4 or

1,3   2 rad/s, 2,4   2 rad/s

Each value of ω2 yields an expression for u:
Computing the vectors u

 u11 
For  =2, denote u1    then we have
2
1
u12 
(-12 M  K )u1  0 
 27  9(2) 3   u11  0 
 3    
 3  (2)  u12  0 
9u11  3u12  0 and  3u11  u12  0

2 equations, 2 unknowns but DEPENDENT!

(the 2nd equation is -3 times the first)
Only the direction of vectors u can be determined, not
the magnitude as it remains arbitrary

u11 1 1
  u11  u12 results from both equations:
u12 3 3
only the direction, not the magnitude can be determined!
This is because: det(12 M  K )  0.
The magnitude of the vector is arbitrary. To see this suppose
that u1 satisfies
(12 M  K )u1  0, so does au1 , a arbitrary. So
( 12 M  K ) au1  0  ( 12 M  K )u1  0
Likewise for the second value of ω2

 u21 
For  = 4, let u 2    then we have
2
2
u22 
(-12 M  K )u  0 
 27  9(4) 3   u21  0 
 3    
 3  (4)  u22  0 
1
9u21  3u22  0 or u21   u22
3

Note that the other equation is the same

What to do about the magnitude!
Several possibilities, here we just fix one element:

Choose:
 3 1
u12  1  u1   
1
Choose:
 3
1
u22  1  u 2   
1
Thus the solution to the algebraic matrix equation is:

 13 
1,3   2, has mode shape u1   
1
 1 3 
2,4   2, has mode shape u 2   
1
Here we have introduce the name
mode shape to describe the vectors
u1 and u2. The origin of this name comes later
Return now to the time response:
We have computed four solutions:
x(t )  u1e  j1t , u1e j1t , u 2 e  j2t , u 2 e j2t  (4.24)
Since linear, we can combine as:

 j1t j1t  j2t j2t

x(t )  au1e  bu1e  cu 2e  du 2 e
 x(t )   ae  j1t  be j1t  u1   ce  j2t  de j2t  u 2
 A1 sin(1t  1 )u1  A2 sin(2t  2 )u 2 (4.26)

where A1 , A2 , 1 , and 2 are constants of integration

determined by initial conditions.
Physical interpretation of all that math!

• Each of the TWO masses is oscillating at TWO natural

frequencies ω1and ω2
• The relative magnitude of each sine term, and hence of the
magnitude of oscillation of m1 and m2 is determined by the
value of A1u1 and A2u2
• The vectors u1 and u2 are called mode shapes because the
describe the relative magnitude of oscillation between the two
masses
What is a mode shape?
• First note that A1, A2, Φ1 and Φ2 are determined by
the initial conditions
• Choose them so that A2 = Φ1 = Φ2 =0
• Then:  x1 (t )  u11 
x(t )     A1   sin 1t  A1u1 sin 1t
 x2 (t )  u12 
• Thus each mass oscillates at (one) frequency w1 with
magnitudes proportional to u1 the 1st mode shape
A graphic look at mode shapes:
If IC’s correspond to mode 1 or 2, then the response is purely in
mode 1 or mode 2.
Mode 1: x1 x2

k2
k1
 13 
m1 m2 u1   
1
x1=A/3 x2=A

Mode 2: x1 x2

k1 k2
 1 3 
m1 m2 u2   
1
x1=-A/3 x2=A
Example 4.1.7 given the initial conditions compute
the time response
1  0
consider x(0)=   mm, x(0)   
0 0
 A1
  A2 
 x1 (t )   3 sin 2t  1  3 sin  2t  2  
 x (t )    
 1
 2   A sin 2t    A sin  2t    
 1 2 2

 A1
  A2 
 x1 (t )   3 2 cos 2t  1  3 2 cos  2t  2  
 x (t )    
 1 
 2   A 2 cos 2t    A 2 cos  2t    
 1 2 2

At t = 0 we have

 A1 
1mm  sin1  A2
sin2 
   3 3
 A1 sin1  A2 sin2  
 0  

 A1 
2 cos1  2 cos2 
A2
0 
   3 3 
   A 2 cos   2A cos  
0
 1 1 2 2 
 4 equations in 4 unknowns:

3  A1 sin 1   A2 sin 2 

0  A1 sin 1   A2 sin 2 
0  A1 2 cos 1   A2 2 cos 2 
0  A1 2 cos 1   A2 2 cos 2 

Yields:
A1  1.5 mm, A2  1.5 mm, 1  2  rad
2
The final solution is:
x1 (t )  0.5cos 2t  0.5cos 2t mm (4.34)
x2 (t )  1.5cos 2t  1.5cos 2t mm
These initial conditions gives a response that is a combination of modes.
Both harmonic, but their summation is not.

Figure 4.3a x1(t) Figure 4.3b x2(t)

Solution as a sum of modes

x(t )  a1u1 cos 1t  a2u2 cos 2t

Determines how the second

Determines how the first frequency contributes to the
frequency contributes to the response
response
Things to note
• Two degrees of freedom implies two natural frequencies
• Each mass oscillates at with these two frequencies present
in the response and beats could result
• Frequencies are not those of two component systems
k1 k2
1  2   1.63, 2  2   1.732
m1 m2

• The above is not the most efficient way to calculate

frequencies as the following describes
Some matrix and vector reminders

a b 1 1  d b 
A  A   c a 
 c c  ad  cb  
xT x  x12  x22
 m1 0 
M    x T
Mx  m x
1 1
2
 m x
2 2
2

 0 m 2

M  0  xT Mx  0 for every value of x except 0

Then M is said to be positive definite

4.2 Eigenvalues and Natural Frequencies

• Can connect the vibration problem with the algebraic

eigenvalue problem developed in math
• This will give us some powerful computational skills
• And some powerful theory
• All the codes have eigen-solvers so these painful calculations
can be automated
Some matrix results to help us use available
computational tools:
A matrix M is defined to be symmetric if
M  MT
A symmetric matrix M is positive definite if
xT Mx  0 for all nonzero vectors x
A symmetric positive definite matrix M can be factored

M  LL T

Here L is upper triangular, called a Cholesky matrix

If the matrix L is diagonal, it defines the matrix square
root

The matrix square root is the matrix M 1/2 such that

M 1/2 M 1/2  M
If M is diagonal, then the matrix square root is just the root
of the diagonal elements:
 m1 0 
LM 1/2
  (4.35)
 0 m2 
A change of coordinates is introduced to capitalize on
existing mathematics
For a diagonal, positive definite matrix M:

 m1 0 1  1 m1 0  1/2
 1
m1
0 
M   , M  1 
, M   
0 m2  0 m2   0 1
m2 

Let x(t )  M 1/2q(t ) and multiply by M 1/2 :
M 1/2 MM 1/2 q(t )  M 1/2 KM 1/2 q(t )  0 (4.38)
I identity K symmetric

or q(t )  Kq(t )  0 where K  M 1/2 KM 1/2

k
K is called the mass normalized stiffness and is similar to the scalar
m
used extensively in single degree of freedom analysis. The key here is that
K is a SYMMETRIC matrix allowing the use of many nice properties and
computational tools
How the vibration problem relates to the real symmetric
eigenvalue problem
Assume q(t )  ve jt in q(t )  Kq(t )  0
 2 ve jt  Kve jt  0, v  0 or
Kv   2 v  Kv   v v0
vibration problem real symmetric
eigenvalue problem
(4.40) (4.41)

Note that the martrix K contains the same type of information

as does n2 in the single degree of freedom case.
Properties of the n x n Real Symmetric Matrix
• There are n eigenvalues and they are all real valued
• There are n eigenvectors and they are all real valued
• The eigenvalues are all positive if and only if the matrix is
positive definite
• The set of eigenvectors can be chosen to be orthogonal
• The set of eigenvectors are linearly independent
• The matrix is similar to a diagonal matrix
• Numerical schemes to compute the eigenvalues and
eigenvectors of symmetric matrix are faster and more efficient
Square n x n Matrix Review
• Let aik be the ikth element of A then A is symmetric if aik =
aki denoted AT=A
• A is positive definite if xTAx > 0 for all nonzero x (also
implies each λi > 0)
• The stiffness matrix is usually symmetric and positive
semi definite (could have a zero eigenvalue)
• The mass matrix is positive definite and symmetric (and
so far, its diagonal)
Normal and orthogonal vectors

n
x    , y    , inner product is x y   xi yi
T

 x1   y1  i 1
   
   
 xn   yn 
x orthogonal to y if xT y  0 

x is normal if x x  1
T

if a the set of vectores is is both orthogonal and normal it
is called an orthonormal set

n
The norm of x is x  x x  T
i
x 2

i 1
(4.43)
Normalizing any vector can be done by dividing it by its
norm:
x
has norm of 1 (4.44)
T
x x
To see this compute
T T
x x x x x
  T 1
x x T
x x T T
x x x x
Examples 4.2.2 through 4.2.4

1/2 1/2  1 0   27 3   1 0
K  M KM 
3 3
  3 3   0 1 
 0 1   
 3 1
so K    which is symmetric.
 1 3 
3- -1 
det(K   I )  det      6  8  0
2

 -1 3- 
which has roots: 1  2  12 and 2  4  22
( K  1 I ) v1  0 
3  2 1   v11  0 
 1 3  2  v   0  
   12   
1
v11  v12  0  v1    
1
v1   2 (1  1)  1    1
2

1 1
v1  1
2  The first normalized eigenvector
Likewise the second normalized eigenvector is computed
and shown to be orthogonal to the first, so that the set is
orthonormal

1 1 1
v2   1 , v1 v 2  2 (1  1)  0
T

2 
1
v1 v1  (1  1)  1
T

2
1
v 2 v 2  (1  (1)(1))  1
T

2
 v i are orthonormal
Modes u and Eigenvectors v are different but related:

u1  v1 and u 2  v 2
1/2 1/2
xM quM v (4.37)

Note
 3 0   1  1
M u1  
1/2
  3 
   v1
0 1   1  1
This orthonormal set of vectors is used to form an
Orthogonal Matrix
P   v1 v2  called a matrix of eigenvectors (normalized)
 v T
1 v1 v1T v 2  1 0 
P P T
T
T   I
 v 2 v1 v 2 v 2  0 1 
P is called an orthogonal matrix
PT KP  PT  Kv1 Kv 2   PT  1 v1 2 v 2 
1 v1T v1 2 v1T v 2  1 0 
 T 
    diag( 2
,  2) 
2

1 v 2 v1 2 v 2 v 2   0 2 
T 1

P is also called a modal matrix

Example 4.2.4 compute P and show that it is an
orthogonal matrix
From the previous example:

1 1 1 
P   v1 v1   1 1 
2 
1 1 1 1  1 1 
P P
T
1 1 1 1
2 2  
1 1  1 1  1 1  2 0 
      I
2 1  1 1  1 2  0 2 
Example 4.2.5 Compute the square of the frequencies by
matrix manipulation
1 1 1   3 1 1 1 1 
P KP 
T
    
2 1 1  1 3  2 1 1
1 1 1   2 4 
 
2 1 1  2 4 
1 4 0 2 0 12 0 
      2
2 0 8  0 4  0  2

 1  2 rad/s and 2  2 rad/s

In general:

  PT KP  diag  i   diag(i2 ) (4.48)

Example 4.2.6

The equations of motion: Figure 4.4

m1 x1  (k1  k2 ) x1  k2 x2  0
(4.49)
m2 x2  k2 x1  (k2  k3 ) x2  0
In matrix form these become:

 m1 0  k1  k2 k2 
0  x  x0 (4.50)
 m2   k2 k2  k3 
Next substitute numerical values and compute P and Λ
m1  1 kg, m2  4 kg, k1  k3  10 N/m and k2 =2 N/m

1 0  12 2 
M   , K  
 0 4   2 12 
1/2 1/2 12 1
 K  M KM  
 1 12 
12   1 
 det  K   I   det     2
 15  35  0
 1 12   
 1  2.8902 and 2  12.1098
 1  1.7 rad/s and 2  12.1098 rad/s
Next compute the eigenvectors
For 1 equation (4.41 ) becomes:
12 - 2.8902 1   v11 
    0
 1 3 - 2.8902  v21 
 9.1089v11  v21
Normalizing v1 yields
1  v1  v112  v21
2
 v112  (9.1089) 2 v112
 v11  0.1091, and v21  0.9940
 0.1091  0.9940 
v1    , likewise v 2   
0.9940   0.1091 
Next check the value of P to see if it behaves as its
suppose to:
 0.1091 0.9940
P   v1 v2    
0.9940 0.1091 

 0.1091 0.9940 12 1  0.1091 0.9940  2.8402 0 

P KP  
T
      
 0.9940 0.1091 1 3  0.9940 0.1091   0 12.1098 

 0.1091 0.9940  0.1091 0.9940 1 0

P P
T
    
 0.9940 0.1091 0.9940 0.1091   0 1 

Yes!
A note on eigenvectors

In the previous section, we could have chosed v 2 to be

 0.9940  -0.9940
v2    instead of v 2   
 0.1091   0.1091 
because one can always multiple an eigenvector by a constant
and if the constant is -1 the result is still a normalized vector.

Does this make any difference?

No! Try it in the previous example

All of the previous examples can and should be solved by
“hand” to learn the methods

However, they can also be solved on calculators with matrix

functions and with the codes listed in the last section
In fact, for more then two DOF one must use a code to solve
for the natural frequencies and mode shapes.
Next we examine 3 other formulations for solving for modal
data
Matlab commands

• To compute the inverse of the square matrix A: inv(A) or

use A\eye(n) where n is the size of the matrix
• [P,D]=eig(A) computes the eigenvalues and normalized
eigenvectors (watch the order). Stores them in the
eigenvector matrix P and the diagonal matrix D (D=L)
More commands

• To compute the matrix square root use sqrtm(A)

• To compute the Cholesky factor: L= chol(M)
• To compute the norm: norm(x)
• To compute the determinant det(A)
• To enter a matrix:
K=[27 -3;-3 3]; M=[9 0;0 1];
• To multiply: K*inv(chol(M))
An alternate approach to normalizing mode shapes

From equation (4.17)   M  2

 K  u  0, u0

Now scale the mode shapes by computing  such that

1
 i i  i i
     
T
u M u 1 i
uTi ui

w i   i ui is called mass normalized and it satisfies:

i2 Mw i  Kw i  0  i2  wTi Kw i , i  1, 2
(4.53)
There are 3 approaches to computing mode shapes and
frequencies
(i)  Mu  Ku (ii)  u  M Ku (iii)  v  M KM 2 v
2 2 1 2 1 1
2

(i) Is the Generalized Symmetric Eigenvalue Problem

easy for hand computations, inefficient for computers
(ii) Is the Asymmetric Eigenvalue Problem
very expensive computationally
(iii) Is the Symmetric Eigenvalue Problem
the cheapest computationally
 Some Review: Window 4.3
Orthonormal Vectors
similar to the unit vectors of statics and dynamics

x1 and x2 are both normal if xT1 x1  1 and xT2 x2  1

and are orthogonal if x1T x2  0
This is abbreviated by
0 if i  j
xTi x j   ij  
1 if i  j
A set of n vectors xi are set to be orthonorma l if
xTi x j   ij
for all values of i and j .
 4.3 - Modal Analysis
• Physical coordinates are not always the easiest to work in
• Eigenvectors provide a convenient transformation to
modal coordinates
– Modal coordinates are linear combination of physical
coordinates
– Say we have physical coordinates x and want to
transform to some other coordinates u

u1  x 1  3x 2  u1  1 3   x1 
   x 
u 2  x 1  3x 2  2 
u 1 3 2
Review of the Eigenvalue Problem

Start with Mx(t )  Kx  0 where x is a

vector and M and K are matrices. Assume
initial conditions x0 and x 0. Rewrite as
1 1

x  Kx  0 and let

M M
2 2


q
1  21
M x  q  x  M q (coord. trans. #1)
2
 Eigenproblem (cont)
 21
Premultiply by M to get
 21 1  21
  M K M q  q ~
 21
  K q  0
(4.55)
M
M
 q

2

I ~ ~T
K K
• Now we have a symmetric, real matrix
• Guarantees real eigenvalues and distinct, mutually
orthogonal eigenvectors
Eigenvectors = Mode Shapes?

Mode shapes are solutions to M 2u = Ku

in physicalcoordinate s. Eigenvetors are
characteristics of matrices. The two are
related by a simple transforma tion,
but they are not synonymous.
Eigenvectors vs. Mode Shapes

~
The eigenvectors of the symmetricPD matrix K
are orthonorma l, i.e., P T P  I . Are the mode shapes
 21
orthonorma l? Using the transforma tion x  M q,
 21 1
the modes shapesU  M P  P  M U . Now, 2

1 1
P P  U M M U  U T MU  I . Thus,the mode
T T 2 2

shapes are orthogonal only w.r.t. the mass matrix.

 21  21
Similarly,U KU  P M
T T
KM
 P  
~
K
The Matrix of eigenvectors can be used to decouple the
equations of motion
If P orthonormal (unitary), PT P  I  PT  P1
Thus, PT KP    diagonal matrix of eigenvalues.
Back to q  Kq  0. Make the additional coordinate
transformation q  Pr and premultiply by PT
PT Pr  PT K Pr  Ir  r  0 (4.59)

• Now we have decoupled the EOM i.e., we have n

independent 2nd-order systems in modal coordinates r(t)
Writing out equation (4.59) yields
1 0   r1 (t )  12 0   r1 (t )  0 
0 1   r (t )    2    (4.60)
   2   0 2   r2 (t )  0 

r1 (t )  12 r1 (t )  0 (4.62)

r2 (t )  22 r2 (t )  0 (4.63)

We must also transform the initial conditions

 r1 (0)   r10 
r0        P T
q (0)  P T
M 1/2
x(0) (4.64)
 r2 (0)   r20 
 r1 (0)   r10 
r0        P T
q (0)  P T
M 1/2
x(0) (4.65)
 r2 (0)   r20 
This transformation takes the problem from couple
equations in the physical coordinate system in to
decoupled equations in the modal coordinates
2 r1
1)
1
x1 x2
k1 k2
m1 m2 2 r2
2)

1
Physical Coordinates.
Coupled equations Modal Coordinates.
1
x  M Pr
2 Uncoupled equations
Modal Transforms to SDOF

1
T
• The modal transformation P M 2

transforms our 2 DOF into 2 SDOF systems

• This allows us to solve the two decoupled SDOF systems
independently using the methods of chapter 1
• Then we can recombine using the inverse transformation
to obtain the solution in terms of the physical coordinates.
The free response is calculated for each mode
independently using the formulas from chapter 1:
ri 0
ri (t )  sin i t  ri 0 cos i t , i  1, 2
i
or (see Window 4.3 for a reminder)
r 2
i ri 0
ri (t )  r  2 i0
sin(i t  tan 1
), i  1, 2
i0
 i
2
ri 0
Note, the above assumes neither frequency is zero
Once the solution in modal coordinates is determined (ri)
then the response in Physical Coordinates is computed:

• With n DOFs these transformations are:

x(t)  S r(t)
n 1 n  n n 1
1

S M P
2
where
n  n nxn nxn
(where n = 2 in the previous slides)
Steps in solving via modal analysis (Window 4.5)
Example 4.3.1 via MATLAB (see text for hand
calculations)
9 0   27 3 0 0
M   , K   , x(0)    , x(0)   
0 1   3 3  1  0
• Follow steps in Window 4.5 (page 337)
1 1 1
~
1) Calculate M 2
2) Calculate K  M 2
KM 2

» Minv2 = inv(sqrt(M)) »Kt =Minv2*K*Minv2

Minv2 = Kt =
0.3333 0 3 -1
0 1.0000 -1 3
Example 4.3.1 solved using MATLAB as a calcuator

% 3) Calculate the symmetric eigenvalue problem for K tilde [P,D] = eig(Kt);

[lambda,I]=sort(diag(D)); % just sorts smallest to largest
P=P(:,I); % reorder eigenvectors to match eigenvalues
»lambda =
2
4
P=
-0.7071 -0.7071
-0.7071 0.7071
 Example 4.3.1 (cont)
% 4) Calculate S = M^(-1/2) * P and Sinv = P^T * M^(1/2)
S = Minv2 * P;
Sinv = inv(S);
% 5) Calculate the modal initial conditions
r0 = Sinv * x0;
rdot0 = Sinv * v0;
 Example 4.3.1 (cont)

% 6) Find the free response in modal coordinates

tmax = 10;
numt = 1000;
t = linspace(0,tmax,numt);
[T,W]=meshgrid(t,lambda.^(1/2));
% Use Tony's trick
R0 = r0(:,ones(numt,1));
RDOT0 = rdot0(:,ones(numt,1));
r = RDOT0./W.*sin(W.*T) + R0.*cos(W.*T);
% 7) Transform back to physical space
x = S*r;
 Example 4.3.1 (cont)
% Plot results
figure
subplot(2,1,1)
plot(t,r(1,:),'-',t,r(2,:),'--')
title('free response in modal coordinates')
xlabel('time (sec)')
legend('r_1','r_2')
subplot(2,1,2)
plot(t,x(1,:),'-',t,x(2,:),'--')
title('free response in physical coordinates')
xlabel('time (sec)')
legend('x_1','x_2')
 Modal and Physical Responses
Free response in modal coordinates
Modal Coordinates:
4
Independent r1
oscillators 2 r2

1  2  1  2 0

2 -2
 T1 
1
-4
 2  4.44sec, 0 1 2 3 4 5 6 7 8 9 10

2  4  1  2  sec  sec

 T2   sec
Free response in physical coordinates
Physical 4
Coordinates: x1
Coupled oscillators 2
x2
0
Note IC
-2

-4
0 1 2 3 4 5 6 7 8 9 10
Time (s)
Section 4.4 More then 2 Degrees of Freedom

Fig 4.8

Extending previous section to

any number of degrees of
freedom
Fig 4.7
A FBD of the system of figure 4.8 yields the n equations
of motion o the form:

mi xi  ki  xi  xi 1   ki 1  xi 1  xi   0, i  1, 2,3 n (4.83)

Writing all n of these equations and casting them in matrix

form yields:
Mx(t )  Kx(t )  0, (4.80)

where:
the relevant matrices and vectors are:

 k1  k2 k2 0 0 
 m1 0  k 0 
0
0   2 k 2  k3  k3
m2 0
M  , K  0  k3  (4.83)
   
   kn 1  kn kn 
0 0 mn 
 0 0  kn kn 

 x1 (t )   x1 (t ) 
 x (t )   x (t ) 
x(t )   2  , x(t )   2 
   
   
 n 
x (t )  n 
x (t )
For such systems as figure 4.7 and 4.8 the process stays
the same…just more modal equations result:

Process stays the same as section 4.3

r1(t )  12r1(t )  0
r2(t )  22r2(t )  0
r3(t )  32r3(t )  0
 Just get more modal
rn (t )  n2rn (t )  0 equations, one for each
degree of freedom (n is
the number of dof)

See example 4.4.2 for details

The Mode Summation Approach

• Based on the idea that any possible time response is just a

linear combination of the eigenvectors
Starting with q(t )  Kq(t )  0 (4.88)

 v
n n
let q(t )=  qi (t )   ai e
 j i t j i t
 bi e i
i 1 i 1

 two linearly independent solutions for each term.

n
can also write this as q(t )   di sin i t  i  v i (4.92)
i 1
Mode Summation Approach (cont)

Find the constants d i and i from the I.C.

n n
q(0)  
i
d i sini vi
1
and q (0)  
1
i
di i cos i vi

Assumingeigenvectors normalizedsuch that vTj vi  ij

 
n n
vTj q(0) = vTj  d i sini vi =  d i sini vTj vi  d j sin j
i 1 i 1

Similarly for the initial velocities, vTj q (0) = d j  j cos  j

Mode Summation Approach (cont)

Solve for di and i from the two IC equations

vTi q(0)  T
i i q (0)
v
di  and i  tan 1

sin i T
v i q(0)
IMPORTANT NOTE about q(0)=0
if you just crank it through the above expressions
you might conclude that di  0, i.e., the trivial soln.
Be careful with q(0) = 0 as well.
Mode Summation Approach for zero initial displacement

If q(0) = 0, the return to

n
q(0)   di sin i v i
i 1

and realize that i  0 instead of di  0.

The compute di from the velocity expression
v q(0)=i di cos i
T
i
Mode Summation Approach with rigid body modes (ω1 =
0)
if 1  0,
q1 (t )  (a1e j 0t
 b1e j 0t
)vi  (a1  b1 )vi
does not give two linearly independent solutions.
Now we must use the expansion
n
q(t )  (a1  b1t )v1   (ai e
 j i t j i t
bi e )vi
i 2

and adjust calculation of the constants from the

initial conditions accordingly.
Note that the underline term is a translational motion
Example 4.3.1 solved by the mode summation method

3 0 1  1 1
From before, we have M 1/2    and V =  1 1 
 0 1  2  
1/2  3 1/2 0
Appropriate IC are q 0 =M x 0 =   , q 0 =M v 0 =  
0 0
 
  
 v T
q (0)  v T
q (0)    2
i  tan 1 i T i  tan 1 i i   1   
v i q(0) 0 2     
 2 
3 2 
T
v q(0)   
d 2
di  i
  1   
sin i  d 2  3 2 
 2
Example 4.3.1 constructing the summation of modes
 q1 (t )  3 2    1  1 3 2    1  1
 q (t )   2 sin  2t  2   1  2 sin  2t  2  1
 2    2     2  
the first mode the second mode

Transforming back to the physical coordinates yields:

1/2 3 2    1  13 0   1 3 2    1  13 0  1

x(t )  M q  sin  2t      sin  2t     
2  2  2  0 1   1 2  2  2  0 1   1 
3 2    1  1 3 2    1  1
 sin  2t      
2  3 2  1 2  3 2  1 
sin 2t
2  2 
Example 4.3.1 a comparison of the two solution
methods shows they yield identical results
Steps for Computing the Response By Mode Summation
1. Write the equations of motion in matrix form, identify M
and K
2. Calculate M -1/2 (or L)
~  12
3. Calculate K  M KM  12

4. Compute the eigenvalue problem for the matrix

~
K and get i2 and vi
5. Transform the initial conditions to q(t)

q(0)  M x(0) and q(0)  M 2 x(0)

1 1
2
Summary of Mode Summation Continued
6. Calculate the modal expansion coefficients and phase
constants
  v T
q(0)  vTi q(0)
i  tan  T
1 i i ,
 di 
 v i q (0)  sini

7. Assemble the time response for q

n
q(t )   d i sini t  i vi
i1

8. Transform the solution to physical coordinates

n
x(t )  M  12
q(t )   di sin i t  i  ui
i 1
Nodes of a Mode Shape

• Examination of the mode shapes in Example 4.4.3 shows

that the third entry of the second mode shape is zero!
• Zero elements in a mode shape are called nodes.
• A node of a mode means there is no motion of the mass or
(coordinate) corresponding to that entry at the frequency
associated with that mode.
The second mode shape of Example 4.4.3 has a node
• Note that for more then 2 DOF, a mode shape may have a
zero valued entry
• This is called a node of a mode.
 0.2887 
 0.2887 
u2   
 0 
 
node  0.2887 

They make great mounting points in machines

A rigid body mode is the mode associated with a zero
frequency
Fig 4.11

• Note that the system in Fig 4.12 is not constrained and can
move as a rigid body
• Physically if this system is displaced we would expect it to move
off the page whilst the two masses oscillate back and forth
Example 4.4.4 Rigid body motion
The free body diagram of figure 4.11 yields
m1 x1  k ( x2  x1 ) and m2 x2  k ( x2  x1 )
 m1 0   x1   1 1  x1  0
    k     
0 m2   x2   1 1   x2  0
Solve for the free response given:

m1  1 kg, m2  4 kg, k  400 N subject to

0.01
x0    m and v 0  0
 0 
Following the steps of Window 4.5
1 0 
1. M 1/2  1 
0 
 2
1 0  1 0 
1/2 1/2    1 1    400 200 
2. K  M KM  400   
0 1   1 1  0 1   200 100 
 2  2
  4   2  
3. det  K   I   100 det      100   2
 5   0
  2 1    
 1  0 and 2  5  1  0, 2  2.236 rad/s

Indicates a rigid body motion

Now calculate the eigenvectors and note in particular that
they cannot be zero even if the eigenvalue is zero
 4  0 2   v11  0 
  0  100         4v11  2v21  0
 2 1  0  v21  0 
1  0.4472
 v1    or after normalizing v1   
 
2  0.8944 
 0.8944  0.4472 0.8944
Likewise: v 2    P 
 0.4472   0.8944 0.4472 
As a check note that

PT P  I and PT KP  diag 0 5
5. Calculate the matrix of mode shapes

1/2 1 0  0.4472 0.8944  0.4472 0.8944

SM P     
 0 1/ 2  0.8944 0.4472   0.4472 0.2236 
1  0.4472 1.7889 
S  
 0.8944 0.8944 
7. Calculate the modal initial conditions:
1  0.4472 1.7889  0.01  0.004472 
r (0)  S x0       
 0.8944 0.8944  0   0.008944 
r (0)  S 1x0  0
7. Now compute the solution in modal coordinates and
note what happens to the first mode.
Since ω1 = 0 the first modal equation is
r1  (0)r1  0
 r1 (t )  a  bt Rigid body translation

And the second modal equation is

r2 (t )  5r2 (t )  0
 r2 (t )  a2 cos 5t Oscillation
Applying the modal initial conditions to these two
solution forms yields:
r1 (0)  a  0.004472
r1 (0)  b  0.0
 r1 (t )  0.0042
as in the past problems the initial conditions for r2 yield
r2 (t )  0.0089 cos 5t
 0.0042 
 r (t )   
 0.0089 cos 5t 
8. Transform the modal solution to the physical
coordinate system
0.4472 0.8944  0.0045 
x(t )  Sr (t )     
0.4472 0.2236   0.0089 cos 5t 
 x1 (t )   2.012  7.60 cos 5t  3
 x(t )        10 m
 x2 (t )   2.012  1.990 cos 5t 

Each mass is moved a constant distance and then oscillates at a single

frequency.
Order the frequencies
• It is convention to call the lowest frequency ω1 so that ω1 < ω2 <
ω3 < …
• Order the modes (or eigenvectors) accordingly
• It really does not make a difference in computing the time
response
• However:
– When we measuring frequencies, they appear lowest to highest
– Physically the frequencies respond with the highest energy in
the lowest mode (important in flutter calculations, run up in
rotating machines, etc.)
The system of Example 4.1.5 solved by Mode Summation
From Example 4.1.6 we have:
1   1 
1  2, u1   3  , 2  2, u 2   3 
 1   1 

Use the following initial conditions and note that only one mode
should be excited (why?)
1  0 
x(0)   3  , x(0)   
 1  0 
Transform coordinates
9 0  1 3 0 1 1/ 3 0
M   M 
2
 and M  
2

 0 1   0 1   0 1 

Thus the initial conditions become

1  3 0   1  1
q(0)  M 2 x(0)     3   
0 1   1  1
1 3 0  0 0
q(0)  M x(0)  
2
    
0 1  0  0
Transform Mode Shapes to Eigenvectors
1  3 0   1  1
v1  M 2 u1     3   
0 1   1  1 eigenvectors

1  3 0   1   1
v 2  M 2u 2     3   
0 1   1   1 

Note that unlike the mode shapes, the eigenvectors are orthogonal:

 1  1  2
Note that v v 2  1 1    0, but u1 u 2   1 1  3    0
T T
1  3 
1  1  3
1 1 1  1
Normalizing yields: v1   and v 2   
2 1 21
From Equation (4.92):
2 2
q(t )  
i
d i sin(i t  i )vi
1
 q (t )  
i
d i i cos(i t  i )vi
1

Set t=0 and multiply by v1:

2
q (0)  
i
d i i cos i vi
1

0 T 1 1 T 1   1
 v1    d 1 2 cos 1v1
T
   d 2 2 cos 2 v1  
0 2 1 2  1

 0  d1 cos 1  1   / 2
Or directly from Eq. (4.97)
From the initial displacement:
v1T q(0) 1 1 2
d1   1 1   
sin( / 2) 2 1 2 (4.98)
vT2 q(0) 1 1
d2    1 1   0
sin( / 2) 2 1

Eigenvector 1
Eigenvector 2
2
thus q(t )   di sin(i t  i ) v i
i 1

1 1 1
 2 cos( 2t ) 1  cos( 2t ) 1
2  
Transforming Back to Physical Coordinates:
1 1 0  1
x(t )  M q(t )  
2 3  cos( 2t )  
 0 1  1
1 
 cos 2t 
 3
 
 cos 2t 

1
 x1 (t )  cos 2t and x2 (t )  cos 2t
3

So, the initial conditions generated motion only in the

first mode (as expected)
Alternate Path to Symmetric Single-Matrix Eigenproblem

• Square root of matrix conceptually easy, but

computationally expensive
 21 1  21
  M K M q  q ~
 21
  K q  0
M M q 2

• More efficient to decompose M into product of upper and

lower triangular matrices (Cholesky decomposition)
Cholesky Decomposition

Let M  U TU where U is upper triangular

Introduce the coordinate transformation
1
Ux  q  x  U q  U U x  K x  0
T

1
q U q
premultiply by U T to get
I q  U T K U 1q  q  K q  0
1 T 1 T T
note that: U K U   U  K U
T T T
  U T K U 1
Cholesky (cont)
• Is this really faster? Let’s ask MATLAB

M M M M U U
1 1 T
2 2

»M = [9 0 ; 0 1]; »M = [9 0 ; 0 1];
• sqrtm
»flops(0);requires
sqrtm(M);aflops
singular value decomposition
»flops(0); chol(M); flops
ans = 65 ans = 5
(SVD), whereas Cholesky requires only simple
operations
• sqrtm requires a singular value decomposition (SVD),
whereas Cholesky requires only simple operations
1
Note that M =U for diagonal M
2
Section 4.5 Systems with Viscous Damping

Extending the first 4 sections to included

the effects of viscous damping (dashpots)
Viscous Damping in MDOF Systems
• Two basic choices for including damping
– Modal Damping
• Attribute some amount to each mode based on
experience, i.e., an artful guess or
• Estimate damping due to viscoelasticity using some
approximation method
– Model the damping mechanism directly (hard and still
an area of research-good for physicists but engineers
need models that are correct enough).
Modal Damping Method

Solve the undamped vibration problem following Window 4.5

Mx(t )  Kx(t )  0  Ir(t )  r(t )  0
Here the mode shapes and eigenvectors are real valued and
form orthonormal sets, even for repeated natural frequencies
~ 1 1
(known because K  M 2
KM 2
is symmetric)
Modal Damping (cont)
• Decouple system based on M and K, i.e., use the
“undamped” modes
• Attribute some zi (zeta) to each mode of the decoupled
system (a guess. Not known beforehand. Can be tested
with gross data like x): here 2
di  i 1   i
ri  2 i i ri  i2ri  0 (4.106)

sindi t  i 
 i i t
 ri (t )  Ai e (4.107)

Alternately: ri (t )  e  i i t (Ai sindi t  Bi cos di t )

Transform Back to Get Physical Solution
• Use modal transform to obtain modal initial conditions and
compute Ai and Fi:
1 1
1
r(0)  S x(0)  P M T 2
x(0)  P M
T 2
x0
1 1
1
r(0)  S x(0)  P MT 2
x(0)  P M
T 2
x 0

• With r(t) known, use the inverse transform to recover the

physical solution:
1 1
x(t )  M 2
q(t )  M 2
Pr(t )  Sr(t )
Modal Damping by Mode Summation
• Can also use mode summation approach
• Again, modes are from undamped system
• The higher the frequency, the smaller the effect (because of the
exponential term). So just few first modes are enough.

sindi t  i vi
n
q(t )  
i
di e
1
 i i t
where

1 1
M 2
KM 2
vi  i2 vi , and di  i 1   i2
vTi q(0)  T
di i q(0)
v
di  and i  tan 1

sini vTi q (0) +  i i vTi q(0)

Compute q(t), Transform back
• To get the proper initial conditions use:
1 1
q(0)  M 2
x(0), and q(0)  M 2
x(0)
• Use the above to compute q(t) and then:
1
x(t )  M 2
q(t )

the response in physical coordinates.

Example
Consider:
9 0   6 2
0 4 x   2 2  x  0
   
Subject to initial conditions: x  1  0
x0   
0 0
  0
Experiments do not give C. They provide zeta (in modal
coordinates) by the half power method.
Compute the solution assuming modal damping of:
 1  0.01 and  2  0.1
Compute the modal decomposition
L =sqrt(M)

3 0 ~ 1 1  0.667  0.333 
L , K  L KL   
0 2  0.333 0.500 
~ 0.615   0.788 
K v  v  1  0.240 , v1   , and 2  0.947 , v 2   
0 .788   0.615 
0.615  0.788 
P  
0.788 0.615 

Compute the modal initial conditions:

10.205  0.263  1  1.846 

S L P     r0  S x0   
0.394 0.308    2.365 
r0  0
Compute the modal solutions:

 1  0.01, 2  0.1,
1  0.49,d 1  0.49,2  0.963 ,d 2  0.958
Using eq (4.108) and (4.109) yields
r1 (t )  4.208e0.004896t sin(0.49t  1.561)
r2 (t )  3.346e0.096t sin(0.958t  1.471)

Then use x(t)=Sr(t)

 0.205 0.263  r1 (t ) 
x(t )  Sr (t )     r (t ) 
 0.394 0.308  2 
x1 (t )  0.863e0.004896t sin(0.49t  1.561)  0.88e0.096t sin(0.958t  1.471)
x2 (t )  1.658e0.004896t sin(0.49t  1.561)  1.029e0.096t sin(0.958t  1.471)

So, first separate solutions in the modal coordinates were

found and then the modes were assembled by the use of S.
The response in the physical coordinates is therefore a
combination of the modal responses just as in the undamped
case. See page 357 for an additional example.
Lumped Damping models
• In some cases (FEM, machine modeling), the damping
matrix is determined directly from the equations of motion.
• Then our analysis must start with:

Mx(t )  Cx(t )  Kx(t )  0,

subject to x0 and x0
Generic Example: • If the damping
mechanisms are
known then
• Sum forces to find
the equations of
motion

Fig 4.15

Free Body Diagram:

m1 x1  c1 x1  c2 ( x2  x1 )
 k1 x1  k2 ( x2  x1 )
m2 x2  c2 ( x2  x1 )
 k2 ( x2  x1 )
Matrix form of Equations of Motion:

m1 0   x1(t ) c1  c 2  c 2   x1(t )

         
 0 m2  x 2(t )   c 2 c 2  x 2(t )
k 1  k 2  k 2   x 1(t ) 0
   
  k2 k 2  x 2(t ) 0

The C and K matrices have the same form.

It follows from the system itself that consisted damping and stiffness
elements in a similar manner.
A Question of matrix decoupling

• Can we decouple the system with the same coordinate

transformations as before?
Mx  Cx  Kx  0
1 1
I r  P
T
M
CM
2
P r  r  0

2

diagonal ?
• In general, these can not be decoupled since K and C can
not be diagonalized simultaneously
A Little Matrix Theory
• Two symmetric matrices have the same eigenvectors
if and only if the matrices commute

• Define C  M1 / 2
C M 1 / 2

• Transform the damped equations of motion into:

~
(t )  C q(t )  Kˆq(t )  0
Iq
• Let P be the matrix of eigenvectors vi of K and   P KP
T

T ~
Then P C P will be diagonal if and only if transformed
K and C have same eigenvectors, i.e.

Cvi  i vi for all i, so CK  KC

More Matrix Stuff and Normal Mode Systems
~~ ~~
CK  KC 
1 1 1 1 1 1 1 1
M CM
2 2
M 2
KM 2
M 2
KM 2
M 2
CM 2

1 1 1 1 1 1
M CM KM
2 2
M 2
KM CM 2

 CM 1K  KM 1C -1
Happens if and only if CM K is symmetric

• This does not require a matrix square root to check

• This informs us explicitly whether or not the equations of motion
can be decoupled
• If true, such systems are called “normal mode” systems or said
to possess “classical normal modes”
Proportional Damping
• It turns out that CM-1K = symmetric is a necessary and
sufficient condition for C to be diagonalizable by the
eigenvectors of the “undamped” system, i.e., those based on
M,K
• Best known example is “proportional”damping.
• The coefficients are obtained through experiments or just
by guess.
C   M   K  linear combination of M and K .
CM 1 K   M   K  M 1 K   K   KM 1 K
both symmetric
Proportional Damping (cont)

Write the system as Mx   M   K  x  Kx  0

q  M 2 x  Iq   I   K  q  Kq  0
1

q  Pr  Ir   I     r  r  0
diagonal!
Thus, the damping ratios in the decoupled system are
  i
2 ii        i 
2

2i
i
2 (4.124)
Generalized Proportional Damping

For any value of n up to the number of degrees of freedom:

n
C   i 1 K i 1

i 1
For example for n = 2 we get the previous proportional
damping formulation:
C   0 K  1 K   I   K
0
Section 4.6 Modal Analysis of the Forced
Response

Extending the chapters 2 and 3 to more then one degree of

freedom
Forced Response: the response of an mdof system to a
forcing term
 13 
x1 u1   
1
x2
k1 k2

c2
c1 m1 m2

F1 F2

0 0 0 0  0 
0 0 0 0  0 
Mx  Cx  Kx  BF(t )     (4.126)
0 0 0 0  0 
  
0 0 0 1   F4 (t ) 
Assume C diagonalizable for now, i.e.,
1 1 1
q  Cq  Kq  M 2
F where C  M 2 CM 2
If the system of equations decouple then the methods of
Chapters 2 and 3 can be applied
Decouple the system with the eigenvalues of K
   
Ir  2 ii  r     r  P T
M
1
2
BF
  i 
   
(4.129)
so the i equation would be ri  2 ii ri  i ri  fi (t )
th 2
(4.130)

• Responses to harmonic, periodic, or general forces as in

Chapters 2 and 3
• Note that the modal forcing function is a linear combination of
many physical forces
With the modal equation in hand the general solution is
given
ri (t )  2 ii ri (t )  i2 ri (t )  fi (t ) (4.130)
 ri (t )  di e  iit sin di t  i 
t
1
 e iit  fi (t )e ii sin di  t    d (4.131)
di 0
The applied force is distributed across the all of the
modes except in a special case.
1
f (t )  P M
T 2
BF(t ) for the decoupled EOM.
• An excitation on a single physical DOF may “spread” to all modal DOFs
(one F generates many f’s)
• It is actually possible to drive a MDOF system at one of its natural
frequencies and not experience resonant response (an unusual
circumstance)

Let F(t )  b f (t ), where b is some spatial vector

and f (t ) is any fuction of time. What if b happens
to be related to the i th mode shape by b  M u i?
Example 4.6.1
A 2-dof system

Figure 4.16

9 0   2.7 0.3  27 3  0 

0 1  x   0.3 0.3  x   3 3  x   F (t ) 
       1 

3 0 1/2  13 0
M 1/2
  ,M  
 0 1   0 1 

1/2 1/2  1 3 0   2.7 0.3  1 3 0   0.3 0.1

CM CM        
 0 1   0.3 0.3   0 1   0.1 0.3 
Compute the mass normalized stiffness matrix and its
eigen solution
 1 3 0   27 3  1 3 0   3 1
K  M 1/2 KM 1/2        
 0 1  3 3  0 1   1 3 
From before:
 1  2 1 1
Kv   v   , P  0.707  

 2  4 1 1 
Transform the damping matrix, the forcing function and
write down the modal equations
 1 1  0.3 0.1 1 1 0.2 0 
P CP  0.707 
T
  0.1 0.3  0.707 1 1    0 0.4 
 1 1     
2 0
P KP  
T

 0 4 
1/2  0.2357 0.7071  0 
f (t )  P M BF(t )  
T
  F (t ) 
 0.2357 0.7071  2 

From the above coefficients the modal equations

become (note that the force is distributed to each mode)
r1 (t )  0.2r1 (t )  2r1 (t )  (0.7071)(3) cos 2t  2.1213cos 2t
r2 (t )  0.4r2 (t )  4r2 (t )  (0.7071)(3) cos 2t  2.1213cos 2t
Compute the modal values using the single degree of
freedom formulas
0.2
• The modal damping 1   0.0707
2 2
ratios and damped
0.4
natural frequencies are 2   0.1000
computed using the usual 2(2)
formulas and the
coefficients from the d 1  1 1   12  1.41
terms in the modal
equations: d 2  2 1   22  1.99
Use SDOF formula for the particular solution given in
equation (2.36)
r1 p (t )  1.040 cos(2t  0.1974)
r2 p (t )  2.6516sin(2t )

Now transform back to physical coordinates

1 1.040 cos(2t  0.1974) 

x ss (t )  M 2
P 
 2.6516sin(2 t ) 
 x1 (t )  0.2451cos(2t  0.1974)  0.6249sin 2t

 x2 (t )  0.7354 cos(2t  0.1974)  1.8749sin 2t

Note that the force effects both degrees of freedom even though it is applied to one.
The Frequency Response of each mode is plotted:

• This graph shows 20

the amplitude of R1()/f1())
each mode due R2()/f2())
10
to an input modal
force f1 and f2. Amplitude (dB)
0
• A force applied to
mass # 2 F2 will -10
contribute to both
modal forces! -20

-30
0 1 2 3 4 5
Frequency ()
The frequency response of each degree of freedom is
plotted
10
• This graph shows X1()/F2())
the amplitude of 0 X2()/F2())
each mass due to
an input force on -10
Amplitude (dB)
mass #2.
• Each mass is -20
excited by the
force on mass #2 -30
• Both masses are
effected by both -40
modes
-50
0 1 2 3 4 5
Frequency ω
Resonance for multiple degree of freedom systems can
occur at each of the systems natural frequencies
• Note that the frequency response of the previous example
shows two peaks
Special cases:
• If in the odd case that b is orthogonal to one of the mode
shapes then resonance in that mode may not occur (see
example 4.6.2)
• If the modes are strongly coupled the resonant peaks may
combine (see X1/F2 in the previous slide) and be hard to
notice
Example: Illustrating the effect of the input force
allocation
9 0   27 3 3
Consider: 0 1  x   3 3  x  1 cos 2t
     

Compute the modal equations and discuss resonance.

Solution:
1/21/ 3 0 
M    x  M 1/2q
 0 1
1 0   3 1 1/ 3 0  3 1
0 1  q   1 3  q   0 1  1 cos 2t  1 cos 2t
       
Calculating the natural frequencies and mode shapes yields:

1  2 and 2  2 rad/s
 13   13 
u1    , u 2   
1  1 

The mass normalized eigenvectors are:

1 1 1  1
v1  1 , v 2  1
2  2 

1 1 1
P 1 1 
2 
Transform and compute the modal equations:

 q  Pr yields
1 0   r1   2 0   r1  1  1 1 1
0 1   r   0 4   r    1 1 1 cos 2t 
  2   2 2  
r1  2r1  2 cos 2t
r2  4r2  0

No resonance even though

2  2  , the driving frequency
An example with three masses
x1 x2 x3
k1 k2 k3 k4

c1 m1 c2 m2 c3 m3 c4

F1 F2 F3

m1=m2=m3=2Kg k1=k2=k3= k1=3N/m C=0.02K

 m1 0 0  k1  k2 k2 0 
M   0 m2 0  K   k2 k 2  k3 k3 
 0 0 m3   0 k3 k3  k4 
Solving a system with 3 masses is best done using a code.
Using Matlab we can calculate the eigenvectors and eigenvalues and
hence the mode shapes and natural frequencies.

0.707 0 0   0.5 0.707 0.5 

M 1/2   0 0.707 0  P  0.707 0 -0.707 
 0 0 0.707   0.5 -0.707 0.5 

0.354 0.5 0.354 

  1  0.94 2  1.73 3  2.26
U   0.5 0  0.5 
0.354  0.5 0.354   1  0.0094  2  0.017  3  0.0226

The frequency response of each mode computed
separately:
50
r 1()/f1())
40
r 2()/f2())
30 r 3()/f3())
Amplitude (dB)
20

10

-10

-20

-30
0 1 2 3 4 5
Frequency ω
A comparison of the Frequency response between driving
mass #1 and driving mass #2

X1 (w ) / F1 (w ) 40
X1 (X
w())/F (w
)) )
40
/ F2(
X 2 (w ) / F1 (w ) X 2 (Xw(
1
) )/F
/ F2(())
2
w)
20
X 3 (w ) / F1 (w )
20 2 2
X 3 (Xw3(
) )/F
/ F22(())
w)
0 Amplitude (dB)
0
-20
-20
-40

-60 -40

-80 -60
0 1 2 3 4 5
0 1 2 3 4 5
Frequency ω Frequency ω
Computing the forced response via the mode summation
technique
Consider Mx(t )  Kx(t )  F(t )
ˆ  12
q(t )  Kq(t )  M F(t )
Transform:

From eq. (4.92) the homogeneous solution in mode

summation form is

n
q H (t )   di sin i t  i vi (4.134)
i 1
The total solution in mode summation form is:

n
q(t )   [bi sin i t  ci cos i t ]vi  q p (t ) (4.135)
i 1
particular
homogenous

But
q p (t )  M 2 x p (t )
1

n
q(t )    bi sin i t  ci cos it vi  M 2 x p (t )
1
(4.136)
i 1
Next use the initial conditions and orthogonality to
evaluate the constants

v q 0  ci  v M 2 x p (0)
T T 1

i i (4.138)
v q 0  i bi  v M 2 x p (0)
T T 1

i i (4.139)

ci  v q 0  v M 2 x p (0)
T T 1

i i

bi 
1

 v q T
i 0  vT
iM 2
1
x p (0) 
i
Substitution of the constants into Equation (4.136) and
multiplying by M-1/2 yields

n
x(t )    di sin i t  ci cos i t ui  x p (t )
i 1

(4.141)
 Decoupled Forced EOM
r1
w 12
f1
1
x1 x2
k1 k2
m1 m2
F1 r2
w 2
2
1 f2
Physical Co-ordinates.
Modal Co-ordinates.
Coupled equations
Uncoupled equations

x(t )  M 1/2 P r(t )

4.7 Lagrange’s Equations

Defining work, energy and

virtual displacements and
work we will learn an
alternate method of deriving
equations of motion

Generalized coordinates: 2 not 4!

Recall equations (1.63) and (1.64)
Definitions (from Dynamics)
1 1 T
Kinetic Energy: T  mr  r  mr r
2 2
r2

Work Done by a force: W12   F  dr

r1

r0 a reference position then the potential energy is

r0

V ( r )   F  dr
r1
Strain Energy in a Spring
Strain energy (elastic potential energy)
for a spring : F  kx
0 0
1 2
V (x )   F ( )d    kd  kx
x x 2
which is the area under the F (x ) vs x curve

F(x)

x
Strain energy in an elastic material
Example of a bar of cross section
Stress σ
A(x) elongated by force P(x,t)

Slope E P(x,t) P(x,t) +Px(x,t) dx

Strain ε dx +ux(x,t)dx

The variation of dx, denoted  (dx) is given by

 u ( x, t )
 (dx)= dx   ( x, t )dx
x
P ( x, t )
The axial stress is  ( x, t )   E ( x, t ) so P=EAε
A( x)
Strain energy continued
1 1
dV  P( x, t ) (dx)  P( x, t ) ( x, t ) dx
2 2
1
 [ EA( x) ( x, t )] ( x, t ) dx
2
1
 EA( x) 2 ( x, t ) dx
2
Integrating yields the strain energy for axial tension
in a bar element:
1
V E  A( x) 2 ( x, t ) dx
2 0
  u ( x, t ) 
2
1
 E  A( x)   dx
2 0  x 
Virtual Reality (actually: virtual displacement)

A virtual displacement

r
Based on variational math
Small or infinitesimal
changes compatible with
constraints
No time associated with
change
Variation or
Change in:
Consequence of satisfying the constraint:
Constraint: f (r )  c, a constant
 f (r   r )  c
Taylor expansion:
n
 f 
f (r )     xi   c
i 1   xi 
f
 r
r

f
  r  0
r
Virtual work
Suppose the i th mass is acted on by fi with system in
static equilibrium
  Wi  fi   ri  0, 
the principle of virtual work:
n

 F  r  0
i 1
i i

which states that if a system is in equilibrium, the

work done by externally applied forces through a
virtual displacement is zero:   V  0
 V has an critical value
Dynamic Equilibrium
D'Alembert's Principle  move inertia force to left side and
treat dynamics as statics. From Newton's law in terms of
change in momentum:
 F  p   F  p  0
i i

This allows us to use virtual work in the dynamic case:

   Fi  p    r  0

  F  mr    r  0
i
Hamilton’s Principle
d
(r   r )  r   r  r r
dt
1
 r   r   ( r  r)
2
d 1
  r   r   (r   r )   ( r  r ), multiply by m
dt 2
d
  W   m (r   r )   T
dt
d
  T   W   m (r   r )
dt
Integrate this last expression
t2
d
 T   W  dt    m (r   r)dt
t2
t1
t1
dt

  T   W  dt   mr   r
t2
0
t2
t1 t1

path indepence

  T   W  dt  0,
t2
for conservative forces  W   V
t1

   T  V  dt  0, Hamilton's principle
t2

t1
Lagrange’s Equation

Let r  r (q1 , q2 , q3 ...qn , t ), qi called generalized coordinates

W
Let Qi  a generalized force (or moment) (4.143)
 qi
The Lagrange formulation, derived from Hamilton's principle
for determining the equations of motion are
d  T   T V
    Qi (4.144)
dt   qi   qi  qi
The Lagrangian, L
Let L = (T - U), called the Lagrangian
Then (4.145) becomes:

d L  L
   0, i  1, 2, n (1.146)
dt   qi   qi

For the (common) case that the potential

U
0
energy does not depend on the velocity:  qi
Advantages
• Equations contain only scalar quantities
• One equation for each degree of freedom
• Independent of the choice of coordinate system since the
energy does not depend on coordinates
• See examples in Section 4.7 pages 369-377
• Useful in situations where F = ma is not obvious
Example of Generalize Coordinates

x
How many dof?
y
1 What are they?
q1 Are there constraints?
2 m1 (x1,y1)
x12  y12  21 (x 2  x1)2  (y 2  y 1)2  22
q2
m2 (x2,y2)
There are only 2 DOF and one choice is:

q1  1 and q2  2
Example 4.7.3 (also illustrates linear approximation method)

Here G is mass center and e is

the distance to the elastic axis.
k1 Let m denote the mass of the
wing and J the rotational inertia
about G.
G . k2 q (t )
e
x(t) Take the generalized coordinates to be:

q1  x (t ), q2   (t )
Called the pitch and plunge model
Computing the Energies
1 2 1 2
T mxG  J
The Kinetic Energy is 2 2

The relationship between xG and x is

xG (t )  x(t )  e sin  (t )
d
 xG (t )  x(t )  e cos  (t )  x(t )  e cos  (t )
dt
So the kinetic energy is
1 1 2
T  m[ x  e cos  ]  J
2

2 2
Potential Energy and the Lagrangian

1 1
The potential energy is: U  k1x  k 2 2
2

2 2

The Lagrangian is:

L  T U 
1 2 1 2 1 2 1
m  x  e cos    J  k1 x  k2 2
2 2 2 2
Computing Derivatives for Equation (1.146)
L L
  m[ x  e c os  ]
 q1  x
d L
   mx  me  me 2
sin 
dt   x 
L L
  k1 x
 q1  x
Now use the Lagrange equation to get:

mx  me cos   em 2 sin   k1 x  0

Likewise differentiation with respect to q2 = θ yields:
J  me cos x  me2 cos2   me2 2 sin  cos  k2  0
Next Linearize and write in matrix form

Using the small angle approximations:

sin   cos  1

In matrix form this becomes:

 m me   x(t )   k1 0   x(t )  0
 me me2  J   (t )    0     
k2   (t )  0
   

Note that this is a dynamically coupled system

Next consider the Single Spring-Mass System and
compute the equation of motion using the Largranian
approach
1 2 1 2
T  mx , U  kx
2 2
1 2 1 2
L  T  U  mx  kx
2 2
L L
 mx,  kx
x x
d L L
   0  mx  kx  0
dt   x   x