Vibs1 Review Notes PDF

Mechanical Vibrations I
Lecture Notes -1- 06/16/06 12:25 PM


m
k c
x(t) f(t)

Author: Allyn W. Phillips, PhD

University of Cincinnati

UC Course Nr.: 20-MECH-480
Revision: 16-Jun-2006

Copyright 2001-2006
Course Outline
Topics
Review Dynamics
o Units
o Complex Numbers
o Free Body Diagrams
o Absolute vs. Relative Motion
o Differential Equations
o Matrices & Linear Algebra
o Kinematics
Rolling Contact
Coordinate Systems
SDOF / MDOF Concepts
o Newtons Method
Frequency, Damping, Mode Shape & Scaling
o Lagranges Method
Energy Method (SDOF only)
o Eigenvalue / Eigenvector Solutions
Superposition
o Free Vibrations
o Forced Vibrations
Steady-State
Transient
SDOF / MDOF Applications
o SDOF
Estimating Frequency & Damping
Half Power
Log Decrement
Transmissibility
Isolation
o MDOF
Frequency Response Function Concepts
Stiffness / Compliance
Impedance / Mobility
Modal Parameter Estimation
Impedance / FRF Methods
Course Material Sequence
Introduction to Mechanical Vibrations [AWP #1]
Newtons Method [AWP #2]
o Units [TD 0]
o Matrices & Linear Algebra [TMH A]
o Free Body Diagrams [TD 2.2, 5.1]
Single and Multiple Body
o Absolute vs. Relative Motion
Force Balance
Mass absolute acceleration
Damping relative velocity
Stiffness relative displacement
o Linear & Torsional
Rolling Contact
o Equation of Motion [TD 5.1-3]
Single DOF
MCK v. RLC Equivalence [TMH 2.6]
Homogeneous Solution of Equation of Motion [AWP #4]
o Differential Equations [TMH D]
Linear, Constant Coefficient
Harmonic Functions (sin, cos) [TD 1]
Complex Exponentials
o Free Vibrations [TD 2.1, 2.6]
Oscillatory Motion [TD 1]
Harmonic Motion
Periodic Motion
o Vibration Terminology [TD 1]
Frequency
Damping [TD 2.6, 2.8, 2.9]
Viscous
Coulomb
Structural
Mode Shape [TD 5.1]
Modal Scaling
o Eigenvalue / Eigenvector Solutions [TD C]
Superposition
o Effects of Changing
Mass, Stiffness, Damping
o Modal Parameter Estimation
Damping
Log Decrement [TD 2.7, AWP #5]
Lagranges Method [TD 7.23, TMH B, AWP #6]
o Energy Terms
Kinetic
Potential
Dissipative
o Contrast with Newtons Method
Global vs. Local Control Volume
o Equation of Motion
Show Equivalence to Newtons Method
o Single DOF reduces Energy Method [TD 2.3]
Conservative System - only
Steady-State Solution of Equation of Motion [AWP #7]
o Forced Vibrations [TD 3.1]
Harmonic Excitation
o Laplace Transform [TD B]
Converts Differential Equations to Algebraic Equations
Transfer Function
o Fourier Transform [TD 1.2, 13.7, DLB]
Frequency Response Function [AWP #8]
MCK Variations
SDOF Concept Applications [TD 3]
o Harmonic Unbalance Force [TD 3.2, AWP #9]
o Transmissibility [TD 3.5, AWP #10]
o Isolation [TD 3.6, AWP #10]
o Dampers
Mass
Tuned Absorber
o Modal Parameter Estimation [AWP #12]
Frequency
Damping
Half Power [TD 3.10, AWP #8]
Mode Shape
Quadrature
MDOF Systems [AWP #11]
o Equations of Motion
Matrix Solution
o Natural Frequencies
o Mode Shapes
Transient Solution of Equation of Motion [TD 4, AWP #13]
o Forced Vibrations
Impulse Excitation [TD 4.1]
Transient Excitation [TD 4.2]
o Convolution [TMH 2.7]
Properties: F[f(t)*g(t)] = F[f(t)] F[g(t)]
o Numerical Solutions [AWP #14]
Runge-Kutta
MDOF Detailed Overview (provided for reference) [AWP #15]

#1 Introduction to Mechanical Vibrations

All mechanical systems vibrate or undergo oscillatory motion. However, in order to
discuss the behavior of vibrating systems, it is first necessary to establish some
nomenclature. There are several terms used to describe vibrations, among which are:
oscillatory motion, periodic motion, harmonic motion, period, and frequency.

Oscillatory motion is any pattern of motion where the system under observation moves
back and forth across some equilibrium position, but does not necessarily have any
particular repeating pattern.

Periodic motion is a specific form of oscillatory motion where the motion pattern repeats
itself with a uniform time interval. This uniform time interval is referred to as the period
and has units of seconds per cycle. The reciprocal of the period is referred to as the
frequency and has units of cycles per second. This unit combination has been given a
special unit symbol and is referred to as Hertz (Hz).

Harmonic motion is a specific form of periodic motion where the motion pattern can be
describe by either a sine or cosine. This motion is also sometimes referred to as simple
harmonic motion. Because the sine or cosine technically uses angles in radians, the
frequency term expressed in the units radians per second (
sec
rad
). This is sometimes
referred to as the circular frequency. The relationship between the frequency in Hz and
the frequency in
sec
rad
is simply the relationship,
2 rad
cycle
.

Natural frequency is the frequency at which an undamped system will tend to oscillate
due to initial conditions in the absence of any external excitation. Because there is no
damping, the system will oscillate indefinitely.

Damped natural frequency is frequency that a damped system will tend to oscillate due to
initial conditions in the absence of any external excitation. Because there is damping in
the system, the system response will eventually decay to rest.

Resonance is the condition of having an external excitation at the natural frequency of the
system. In general, this is undesirable, potentially producing extremely large system
response.

Degrees-of-freedom is the number of independent coordinates necessary to describe them
configuration (state) of a system. It can also be expressed as the number of coordinates
used to describe the configuration of a system minus the number of independent
constraint equations between those coordinates. For example, a point in space has three
degrees-of-freedom ( , , ) x y z and a rigid body in space has six ( , , , , , )
x y z
x y z .

A decibel is a log scale measure of relative power. Specifically, it is ten times the log
base ten of the power to a reference power,
10
10log
ref
P
dB
P
| |
=
|
|
\ .
. If the quantity being
referenced is not power based, then the quantities are first squared, yielding
2
10
2
10log
ref
V
dB
V
| |
=
|
|
\ .
or
10
20log
ref
V
dB
V
| |
=
|
|
\ .
.

An octave is a frequency range where the upper frequency is twice the lower frequency.

A decade is a frequency range where the upper frequency is ten times the lower
frequency.

Bandwidth and frequency span both terms describe a range of frequencies from low to
high. Bandwidth is particularly associated with the energy of the response of a system
and in this course is used in reference to damping (see section eight.)

Having established a small set of basic nomenclature, it is now possible to provide a brief
overview of the direction of this course. In this course, Mechanical Vibrations I, much of
the effort will be spent upon bringing together the many concepts and techniques
developed in previous courses and applying them to the solution of vibrations problems.

Note that, it is possible to solve simple, but representative, problems utilizing only
information previously learned. To help clarify which courses and knowledge are used,
the following introductory example has been developed.
Introductory Example

For the figure given, assume that the object rolls
without slipping. The objective will be to develop
the equation of motion and determine the natural
frequency of the system. (A more complete
example would involve a general multi-degree of
freedom system, where the objective would be to
determine; the equations of motion, the system
pole, mode shapes, and steady-state response.)

Step 1: Draw the free-body diagram for each independent rigid body in the system. Be
sure to include all external forces and moments and their points of application. (Recall
that the internal force resulting from the spring is due to the actual relative motion of the
ends of the spring. Therefore the coordinate x is measured from the free length of the
spring.) [Mechanics II, Kinematics & Dynamics]

Step 2: Write down all the force and
moment balance equations. [Mechanics
II, Kinematics & Dynamics]

)
f
x
F f k x f mx =

)
0
y
F N mg =

) f cg
cg
M f r f r J + =

Step 3: Identify all relevant coordinate
constraint equations. [Mechanics II, Kinematics & Dynamics]

x r = and therefore x r =

& x r =

Step 4: Choose the desired independent coordinates and eliminate all the dependent
coordinates by substituting the relevant constraint equations. [Mechanics II, Kinematics
& Dynamics]

)
f
x
F f k x f mx =

)
0
y
F N mg =

) f cg
cg
x
M f r f r J
r
+ =

Step 5: Evaluate the set of simultaneous differential equations to yield the equations of
motion. [Mechanics II, Kinematics & Dynamics]
( ) f t
( ) x t
( ) t
k
,
cg
m J
r
Rolls w/o slipping

=
f
k x
f
f
mx
cg
J
N
mg

)
y
F N mg =

)
2
cg
f
cg
J
M f f x
r
| |
=
|
\ .

)
2
&
cg
x cg
J
F M f k x f x mx
r
| | | |
=
| |
\ . \ .

Step 5a: Identify the actual equation of motion.
2
2
cg
J
m x k x f
r
| |
+ + =
|
\ .

Step 6: Solve the simultaneous set of differential equations of motion. The total solution
( ( )
T
x t ) involves the sum of two parts: the particular solution ( ( )
p
x t ) and the
homogeneous solution ( ( )
h
x t ). [Differential Equations]

Step 6a: The particular solution involves knowing the exact form of the forcing function
( ( ) f t ). [Differential Equations]

Step 6b: A second order homogeneous constant coefficient differential equation has
complex exponentials as its solution. [Differential Equations, Numerical Methods]

( )
t
h
x t Ce
= and therefore ( )
t
h
x t C e
= &
2
( )
t
h
x t C e
=

2
2
0
cg t
J
m k Ce
r

| | | |
+ + =
| |
\ . \ .

therefore the system poles (eigenvalues) are:
1,2
2
cg
k
j
J
m
r
=
+

and the homogeneous solution is:

1 2
1 2
( )
t t
h
x t C e C e

= +

Step 6c: Formulate the complete solution ( ( ) ( ) ( )
T p h
x t x t x t = + ) and evaluate the initial
conditions to eliminate the constants of integration. [Differential Equations]

Note: if this example had involved more than one independent dynamic coordinate, the
solution would have additionally required the solution of an eigenvalue/eigenvector
problem. [Differential Equations, Numerical Methods]
Lecture Notes -10- 06/16/06 12:25 PM
#2 - Newtons Method ( -or- d'Alemberts Method )

Force Balance
G
F Mq =
( -or- 0
G
F Mq =
)
Moment Balance
/ P P G P P
M J r Mq = +

( -or-
/
0
P P G P P
M J r Mq =

)

Typical problems with Newtons ( -or- d'Alemberts ) formulation:
Be sure to establish the number of degrees of freedom first and formulate all
terms in only those variables. Clearly identify which degrees of freedom are
relative coordinates versus absolute coordinates. Also, clearly identify what will
be the positive direction of motion for each coordinate. Watch out for
rotational/translational problems. State any constraint relationships that relate
independent and dependent coordinates.
Evaluate the static balance for the problem in order to determine whether the
orientation of the system in the gravitational field will effect the equations of
motion (Are the weights of the objects balanced by an initial static deflection in
the springs?). When in doubt, perform a static force balance to determine the
appropriate constraint equation.
For displacement, velocity and acceleration terms, be sure to develop absolute or
relative displacement, velocity and accelerations of appropriate points as required.
Watch out for 2-D and 3-D vector motions.
Be sure to draw the appropriate free body diagrams for each mass (or combination
of masses) in the system.
o Whenever the system is separated in order to draw a free body diagram,
replace the separation with the appropriate internal forces/moments (equal
and opposite forces/moments on each side of the separation).
o Do not move forces/moments arbitrarily from one mass to another. The
internal forces account for the effects of one mass on another.
Develop one equation of motion for each degree of freedom of the system using
Newton's ( -or- d'Alemberts ) method. Be sure to watch for moving reference
frame issues. Also, check that the units are the same for each term in an equation
(Forces + Moments: NOT!)
If necessary, once the exact equations of motion have been determined, linearize
the equations of motion by neglecting nonlinear terms in the equations of motion.
Note that the linear equations of motion may not adequately describe the original
equations of motion if some of the terms that have been neglected are not
insignificant.

P
G
Lecture Notes -11- 06/16/06 12:25 PM
Newton Example A

In order to solve Newtons Equation,

i
F mx =
,

for the given single degree-of-freedom
system, it is first necessary to draw the
complete free-body diagram. This involves
identifying all internal and external forces
acting upon the degree-of-freedom.
(Recall that the internal force resulting
from the spring is due to the actual relative motion of the ends of the spring. Therefore
the coordinate x is measured from the free length of the spring.)

All the internal and external forces
are combined to yield the actual
equation of motion.

f kx mg cx mx =

collecting all the coordinate based
terms to one side of the equation
and all the externally applied forces
to the other yields

mx cx kx f mg + + =

Since the mg term ends up on the right hand side with the external force, the coordinate
x can be redefined to be from the static equilibrium position. As a result, the mg term
can be eliminated as shown.

Evaluating the static equilibrium equation (i.e. 0 x = , 0 x = & 0 f = .) yields,

kx mg =

Therefore the static deflection ( ) can be evaluated as

mg
k
=

The coordinate system reference can be shifted to the static equilibrium position by
defining
m
k c
x(t) f(t)

k x
m
cx
=
m
mx
mg
f
Lecture Notes -12- 06/16/06 12:25 PM

s
x x =

Since the static deflection ( ) is constant, obviously the velocity and acceleration is
unchanged by the coordinate shift.

s
x x = and
s
x x =

Substituting into the original solution,

( )
s s s
mx cx k x f mg + + + =

But recall that k mg = , therefore

s s s
mx cx kx f + + =

Note: Often the solution solved about the static equilibrium position will eliminate the
mg term when there are springs to carry the load. When this occurs, the mg term will
end up on the right hand side of the equation with the applied external forces.
Lecture Notes -13- 06/16/06 12:25 PM
Newton Example B

Again, in order to solve Newtons Equation,

0
cg
M J L mL = +

,

system, it is first necessary to draw the
complete free-body diagram. This involves
identifying all internal and external forces
acting upon the degree-of-freedom. In this
case, there is no specific external moment applied at point o. Also, recall the parallel
axis theorem,
2
0 cg
J J mL = + . (See above for the general expression.)

All the internal and external moments are
combined to form the actual equation of motion.
Notice that neither the support force T nor the
centripetal acceleration force
2
mL
contribute
to the moment about point o. (However T
would contribute to the moments about the cg.)

( )
2
sin
cg
mgL J mL = +

collecting all the coordinate based terms to one side of the equation and all the externally
applied moments to the other yields

( )
2
sin 0
cg
J mL mgL + + =

This time, since the mg term ends up on the left hand side (multiplied by the coordinate),
it is not possible to eliminate the mg term from the solution. So the exact equation of
motion remains,

( )
2
sin 0
cg
J mL mgL + + =

However, because the sin mgL term is non-linear in the desired coordinate, it is
possible to reduce the solution to a linear approximation by using the first order
McClaurin Series expansion, sin for 1 , as follows.

( )
2
0
cg
J mL mgL + + =

,
cg
m J
L

=
cg
J

mg
T
mL

2
mL

Lecture Notes -14- 06/16/06 12:25 PM
#3 - The Eigenvalue Problem

In many areas of engineering, a mathematical concept referred to as the eigenvalue
problem arises. While the general application of eigenvalue methods has been used to
solve a wide variety of mathematical, scientific and engineering problems, in this course,
we will focus upon its application to the area of mechanical vibrations. Because of this,
we will not examine the eigenvalue/eigenvector problem in detail, but only its most basic
characteristics.

Mathematically, the eigenvalue problem can be expressed as:

| |{ } { } A x x =

In other words, given a matrix
| |
A , the object is to find a vector { } x that when multiplied
times the matrix
| | A does not change direction, but only magnitude (and possibly sense.)
If this vector can be found, it is referred to as an eigenvector of
| |
A , and the scalar ,
which describes the change in magnitude and sense, is referred to as an eigenvalue of
| |
A . (Note that in general, there will be as many eigenvalue/eigenvector pairs as the size
of the matrix. Also, it is only defined for square matrices.)

The solution is essentially a two-step process. First identify the eigenvalues of the matrix,
and then solve for their associated eigenvectors.

Solving for the eigenvalues involves manipulating the equation form as follows.

| |{ } { } { } 0 A x x =
| |{ } | |{ } { } 0 A x I x =
| | | | { } { } 0 A I x ( =

At this point, we recognize the form as a null space solution problem. There are three
ways in which the equation can be satisfied, two of them, | | | | | | 0 A I ( =

and
{ } { } 0 x = , are trivial solutions and are rejected. The third is that the matrix | | | | A I (

is singular (or rank deficient) for certain values of . These values of are the
eigenvalues of the matrix
| |
A . To solve for the s, recognize that a singular (or rank
deficient) matrix has a determinant of zero.

| | | | 0 A I ( =

Lecture Notes -15- 06/16/06 12:25 PM
Expanding this determinant results in a polynomial, known as the characteristic
polynomial, whose roots are the eigenvalues .

After identifying the eigenvalues, go back and solve the linear equations,
| | | | { } { } 0 A I x ( =

, for each eigenvalue . However there is a problem, since the
coefficient matrix is singular (or rank deficient) when evaluated at the eigenvalues, we
cannot simply solve the equation set. In general, the matrix will be rank deficient by one,
indicating that one variable is free, so the most straight-forward solution approach is to
simply set one of the vector { } x elements equal to one and solve as normal. (We will not
get into the issues involved in the repeated root problem. In other words, where two or
more of the eigenvalues are equal.)
Lecture Notes -16- 06/16/06 12:25 PM
Eigenvalue/Eigenvector Example

Given the following matrix, find its associated eigenvalues and eigenvectors.

| |
4 1
2 5
A
(
=
(

First identify the eigenvalues,

| | | |
4 1 1 0 4 1
0
2 5 0 1 2 5
A I

( ( (
= = =
( ( (

4 1
(4 )(5 ) ( 1)( 2) 0
2 5
(
= =
(

2
9 18 0 + =

Which has the roots,
1
3 = and
2
6 = .

Next identify the eigenvector associated with each eigenvalue.

| | | | { } { } 0 A I x ( =

1 1
2 2
4 1 1 0 4 1 0
2 5 0 1 2 5 0
x x
x x
( ( ( (
= =
` ` ` ( ( ( (

) ) )

For the first eigenvalue,
1
3 = ,

1 1
2 1
4 1 0
2 5 0
x
x
(
=
` `
(

) )

1 1
2 2
4 3 1 1 1 0
2 5 3 2 2 0
x x
x x
( (
= =
` ` `
( (

) ) )

Notice the rank deficient nature of the resulting coefficient matrix.

Set one element of the vector { } x equal to one (unity) to solve. For this example, we will
use the first element. (Note that we could actually set it equal to any arbitrary value
except zero and still solve. Occasionally an eigenvector will contain a zero, but we cannot
arbitrarily pick that value. Finally, if after picking a value, the resulting equations cannot
be solved, simply try again picking a different element.)

Letting
1
1 x = ,
Lecture Notes -17- 06/16/06 12:25 PM

2
1 1 1 0
2 2 0 x
(
=
` `
(
) )

We can solve either equation for
2
x . Lets pick the first for simplicity.

| |
2
2
1
1 1 1 0 x
x

= =
`
)

Which yields,
2
1 x = . Therefore, the first eigenvalue/eigenvector pair is:

1
3 = , { }
1
1
1
x

=
`
)

Repeating the process for the second eigenvalue,
2
6 = , yields,

1 2
2 2
4 1 0
2 5 0
x
x
(
=
` `
(

) )

1 1
2 2
4 6 1 2 1 0
2 5 6 2 1 0
x x
x x
( (
= =
` ` `
( (

) ) )

Again letting
1
1 x = , and solving for
2
x , yields,

2
1 2 1 0
2 1 0 x
(
=
` `
(

) )

| |
2
2
1
2 1 2 0 x
x

= =
`
)

Which yields,
2
2 x = . Therefore, the second eigenvalue/eigenvector pair is:

2
6 = , { }
2
1
2
x

=
`
)

Lecture Notes -18- 06/16/06 12:25 PM
Final Notes

Looking at the form of the eigenvalue problem one more time, it can be shown that it can
actually be used to solve certain first order differential equations directly, those of the
form ' y Ay = . Assuming that the vector { } x is of the form { } { } ( )
t
x t X e
= , then the
eigenvalue problem can be written as.

| |{ } { } A x x =
| |{ } { } ( ) ( ) A x t x t =
| |{ } { }
t t
A X e X e

=
| |{ } { } A X X =

Which is really no different than the example problem solved above.

There are a few issues, which will be covered later, applying this method to the second
order differential equations that arise in structural dynamics, specifically the need to use
state-space expansion to convert the set second order differential equations to a larger
system of first order differential equations.

Finally, when using MATLAB to solve eigenvalue problems, while the eigenvalues
should be the same, the eigenvector will in general be scaled differently. MATLAB uses
unity vector length for its scaling.

Lecture Notes -19- 06/16/06 12:25 PM
#4 - Homogeneous Solution of the Equation of Motion

For the single degree of
freedom (SDOF) system
shown, the equation of motion
is:

( ) ( ) ( ) ( ) mx t cx t k x t f t + + =

This differential equation can
be solved by any of several
methods, including: Laplace Transforms, Fourier Transforms, or by an assumed solution.
In this case, assuming a solution appropriate for a second-order, constant coefficient
differential equation.

The general multi-degree of freedom problem can be formulated in matrix form as:

| |{ } | |{ } | |{ } { } ( ) ( ) ( ) ( ) M x t C x t K x t f t + + =

Using a solution appropriate for a linear, constant matrix coefficient, differential
equation, (namely complex exponentials), we have:

{ } { } ( )
st
x t X e = & { } { } ( )
st
f t F e = NOTE: that X, F and s are complex scalars.

By focusing on the homogeneous solution (i.e. 0 f = ), we first evaluate the derivatives of
the assumed solution,{ } { } ( )
st
x t X e = :

{ } { } { } ( ) ( )
st
d
dt
x t x t X se = =
{ } { } { }
2
2
2
( ) ( )
st
d
dt
x t x t X s e = =

By substituting the assumed velocity and acceleration into the original matrix expression
and collecting terms, we have: (NOTE:
st
e is non zero for all time and so may be
eliminated.)

| | | | | | { } { }
2
0 M s C s K X ( + + =

We should recognize this as a form of eigenvalue problem (or a null space solution
involving s.) Also, recognize that the solution { } { } 0 X = is the trivial solution and does
not contribute any useful information. Therefore, since the system matrix becomes rank
deficient for certain values of s, the problem reduces to a polynomial root solver. (The
determinant of the system matrix is zero.)

m
k
c
x(t)
f(t)
Lecture Notes -20- 06/16/06 12:25 PM

| | | | | |
2
0 M s C s K + + =

This determinant results in a polynomial whose roots are the poles of the system. This is
known as the characteristic polynomial. Often the determinant notation is dropped (but
still written equal to the scalar zero implying a determinant) and the matrix characteristic
polynomial is expressed as,

| | | | | |
2
0 M s C s K + + =

Returning to the scalar (1 DOF) problem as initially presented, we have:

2
0 ms cs k + + =

Since the determinant of a scalar is simply the scalar itself, the characteristic polynomial
is simply:

2
0 ms cs k + + =

Since this is a quadratic equation, there are two roots, namely:

2
1,2
4
2
c c mk
m

=

This results in the homogeneous (transient) solutions form of:

1 2
1 2
( )
t t
x t Ae A e

= +

The characteristic behavior of this system solution is determined by the relationship of m,
c, & k.

2
4 c mk >
Two distinct real roots.
2
4 c mk =

Two equal real roots.
2
4 c mk <
One pair of complex
conjugate roots.

For the condition of
2
4 c mk > , no oscillation occurs. This is called over-damped. Since
no oscillation occurs, this is not particularly interesting from a vibration point of view
and we will not investigate it any further.

For the condition of
2
4 c mk = , this is the boundary between oscillatory and non-
oscillatory behavior. This is called critically-damped. We will return to this later.

Lecture Notes -21- 06/16/06 12:25 PM
The interesting oscillatory behavior occurs when
2
4 c mk < . This is called under-damped.
When a system is under-damped, the system will respond to initial conditions with
harmonic oscillatory behavior.

First, we recognize the roots of the characteristic equation as the system poles (
r
). For
the given SDOF system, the roots have the following form:

2
1,2
2 2
r
c k c
j
m m m

| |
= =
|
\ .

Examining the two parts (real and imaginary) of the poles, we can write:

r r r
j =
Where,
2
r
c
m

= is the damping for mode r (for single degree-of-freedom systems, it is
often written as
d
) and
2
2
r
k c
m m
| |
=
|
\ .
is the damped natural frequency for mode r (for single
degree-of-freedom system, it is often written as
d
)

Since
1
and
2
are complex conjugates, the homogeneous solution for the SDOF
problem can be written as:

* *
( )
t t
x t Xe X e

= +

If we let 0 c , then
r
k
m
. This is called the undamped natural frequency, written
as
r
(for single degree-of-freedom system, it is often written as
n
.) This is the
frequency at which the system will oscillate if there is no damping in the system.

Returning to the critically damped condition, where
2
4 c mk = , we can now define a few
more vibration parameters.

Defining the critical damping as:
2
c
c mk =
Allows us to define the damping ratio as:
c
c
c
=
We can now express the damping and damped natural frequency as functions of the
undamped natural frequency and damping ratio, as follows:
Lecture Notes -22- 06/16/06 12:25 PM

2
r r r
c
m
= =
2
2
1
2
r r r
k c
m m

| |
= =
|
\ .

The single degree of freedom equation can now be written as:

2
( )
( ) 2 ( ) ( )
r r r
f t
x t x t x t
m
+ + =

While each of these parameters has been developed from a single degree of freedom (1
DOF) point of view, the concepts are equally applicable to the multi-degree of freedom
case, as well. For this, we need to use a nomenclature that reflects the fact that there is
more than one system pole (with its conjugate). We do this by using the subscript r
(indicating resonance) with each of the modal parameters, as follows.

r r r
j = + is the system pole for mode r, where
r
is the damping and
r
is the
damped natural frequency for mode r. From this, we define
2 2
r r r
= + as the
undamped natural frequency for mode r. Then, the damping ratio is simply defined as,
r
r
r
. The general homogeneous solution then takes on the following form.

{ } { } { }
*
*
( )
t
r t
r
r
r
r
x t X e X e
= +

The outgrowth of this approach is that it is no longer necessary to know the actual mass,
stiffness and damping of the system in order to identify its modal characteristics. This is a
significant advantage when applied to experimental modal analysis because, as will be
shown later, the system poles (
r
s) can be obtained directly from measured Frequency
Response Functions.
Lecture Notes -23- 06/16/06 12:25 PM
Homogeneous Example

Given a single degree-of-freedom mass/spring/damper system (MCK) with 5kg mass,
20
N s
m
damping, and 1000

N
m
stiffness: identify the pole of the system (
r
), the
undamped natural frequency (
r
), and the damping ratio (
r
).

To calculate the system pole, evaluate the characteristic equation,

2
0 ms cs k + + =
2
5 20 1000 0 s s + + =

which has the following roots. Note that the roots of the characteristic equation are the
system poles.

2
20 1000 20
2 5 5 2 5
N s N N s
m m m
r
j
kg kg kg

| |
=
|

\ .

( )
2
2
2
2 200 2
rad rad rad
s s r s
j =
2 14
rad
s r
j =

The pole gives the damping (
r
) and the damped natural frequency (
r
) directly. Recall,
that to express the answer in units of Hertz, it is necessary to divide by 2 .

2
rad
s r
= -or-
1
0.318 Hz Hz
=
14
rad
s r
= -or-
7
2.228 Hz Hz
=

For this problem, the undamped natural frequency (
r
) may be calculated from either
the magnitude of the system pole, or by evaluating the characteristic equation for 0 c = .
Both approaches yield the same solution.

2
2 2 14 200 14.14
rad rad rad
s s r r s
j = = = = -or- 2.251Hz
2
2
1000
200 14.14
5
N
m
rad rad
s r s
k
m kg
= = = = -or- 2.251Hz

Again, for this problem, the damping ratio may be calculated by either of two
approaches, either directly from the system pole or from the systems mechanical
properties (MCK).

Lecture Notes -24- 06/16/06 12:25 PM
2
0.1414
14.14
rad
s r
r
rad
s r

= = =
-or- 14.14%
20
0.1414
2 2 5 1000
N s
m
r
N
c m
c c
c mk kg

= = = =
-or- 14.14%

Recognize one important point in the solution method. The direct (MCK) solutions are
only valid for a single degree-of-freedom (SDOF) system while the solution methods
based upon the system pole are valid for a system with any number of degrees-of-
freedom (MDOF).

Lecture Notes -25- 06/16/06 12:25 PM
#5 Estimating Modal Parameters Time Domain

There are many techniques, both simple and advanced, for estimating modal parameters
(frequency, damping, mode shape, and scaling) from measured experimental data. This
section will focus upon the single degree-of-freedom time domain technique known as
log decrement.

Log Decrement

The log decrement method is based entirely upon the characteristics of a single degree-
of-freedom free decay response.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
-4
-3
-2
-1
0
1
2
3
4
Time (s ec )
A
m
p
l
i
t
u
d
e

Recalling the form of solution, ( )
r
t
x t Xe
= , it is possible to estimate the system pole

(
r r r
j = + ) directly from the observed response.

Estimating the damped natural frequency (
r
) is easy. First estimate the period of one
oscillation ( ), the reciprocal of the period,
1
, is the damped natural frequency (

r
) in
Hz.

2 rad
s r

= -or-
1
Hz

Estimating the damping (
r
) requires a little more effort. The development proceeds as
follows.

Begin by evaluating the position, ( ) x t , for two time instants one cycle apart.
Lecture Notes -26- 06/16/06 12:25 PM
1 1 1
1
( )
r r r
t t j t
x t Xe Xe e

= =
2 2 2
2
( )
r r r
t t j t
x t Xe Xe e

= =

Next evaluate the log decrement ( ) as:

1 1
2 2
1
2
( )
ln ln
( )
r r
r r
t j t
t j t
x t Xe e
x t Xe e

| | | |
= =
| |
\ . \ .

However, since the points are defined one cycle apart,
2 1
t t = + . Substituting into the
above expression yields,

( ) ( )
1 1
1 1
ln
r r
r r
t j t
t j t
Xe e
Xe e

+ +
| |
=
|
\ .

Next, recall that cos sin
j
e j
= + . Hence,
( )
( )
( )
1
1 1 1 1
2
2 2
r
r r r r r
j t
j t j t j t j t j
e e e e e e

+
+ +
= = = =

So the expression for the log decrement can be reduced to

( )
( )
( ) ( )
1
1 1
1
ln ln ln
r
r r r
r
t
t t
r
t
e
e e e
e

+
+
| |
= = = =
|
\ .

Recalling that
r r r
= ,
2
1
r r r
= and
2
r

= . Substituting into the above
expression yields,

2 2
2 2
2
1 1
r r r
r r r r r
r
r r r

= = = = =

Therefore the log decrement equals,

2
2
1
r
r

If the damping ratio is small (eg.
0.1 < -or- 10%), then the above
expression can be simplified to:

2
r

(Error less than 1% when 0.1 < )
Multiple Cycle Solution
If time points multiple (n) periods apart
are used, the expression for the log
decrement is modified as follows:
0
2
1 ( ) 2
ln
( )
1
r
n
r
x t
n x t

| |
= =
|
\ .

where: n is the number of cycles.

Which simplifies to:
0
1 ( )
ln 2
( )
r
n
x t
n x t

| |
=
|
\ .

Lecture Notes -27- 06/16/06 12:25 PM
Log Decrement Example

Using the figure given above, estimate both the damped natural frequency and the
damping ratio.

First, select two convenient time points to estimate the period ( ). Between 0 t =
seconds and 9 t = seconds, there are 11 cycles.

9
0.8182
11
sec
cycle
sec
cycles
= =

Therefore the damped natural frequency is

11
1.222
9
r
cycles
Hz
sec
= = -or- 7.679
rad
s

Second, select two convenient time points to estimate the log decrement ( ). Between
1.8 t seconds and 6.7 t seconds, there are 6 cycles with ( 1.8) 3 x t and
( 6.7) 2 x t .

0 0
1 ( ) 1 ( 1.8) 1 3
ln ln ln 0.0676
( ) 6 ( 6.7) 6 2
n n
x t x t
n x t x t
| | | | | |
= = = =
| | |
\ .
\ . \ .

Since the log decrement ( ) is so small, first try the simplified approximation for
damping ratio ( ).

0.0676
0.0108
2 2

= = -or- 1.08%

(Error check: 0.1 < so the approximation is adequate.)

Just for completeness, since 0.1 < ,
2
1 1 , therefore

2
7.679
1
r
rad
s r r
r
= =
-or- 1.222Hz

0.0108 7.679 0.0829
rad rad
s s r r r
= = = -or- 0.0132Hz

Lecture Notes -28- 06/16/06 12:25 PM
#6 - Lagranges Method

Lagranges Equation

i
i i i i
d T T U D
F
dt q q q q
| |
+ + =
|

\ .

Where:
T = Kinetic Energy
U = Potential Energy
D = Dissipative Energy

i
F = Externally applied force/moment

i
q = Generalized coordinate

Generalized Coordinates
, , , , ,
x y z
q x y z =

Kinetic Energy
2 2
1 1
1 1
2 2
j
m
N
N
i i
i i
T M x J
= =
= +

Potential Energy
( )
2
1 1
1
2
k m
N N
i i
i i
U K x M g x
= =
= +

Dissipative Energy
2
1
1
2
c
N
i
i
D C x
=
=

Lecture Notes -29- 06/16/06 12:25 PM
Lagranges Method

Typical problems with Lagrange formulations:
Be sure to establish number of degrees of freedom first and formulate all energy
terms in only those variables. Clearly identify which degrees of freedom are
relative coordinates versus absolute coordinates. Watch out for
rotational/translational problems.
For kinetic energy terms, be sure to formulate absolute velocities before taking
derivatives. Watch out for 2-D and 3-D vector motions.
For potential energy terms, be sure that the actual deflection, described by relative
and/or absolute coordinates, in spring elements is described. Watch out for 2-D
and 3-D vector motions.
There should be only one total kinetic energy equation, one total potential energy
and one total dissipative energy equation for the system. The kinetic, potential and
dissipative energy equations should involve only the N generalized coordinates
and the constants (mass, damping, stiffness) of the system.
Apply the Lagrange Equation once for each generalized coordinate. For N degrees
of freedom, N generalized coordinates will yield N equations of motion.
If necessary, linearize the equations of motion by neglecting nonlinear terms in
the equations of motion. Note that the linear equations of motion may not
adequately describe the original equations of motion.

NOTE: For the single degree-of-freedom (only one coordinate), conservative (no
damping), un-forced problem (no external forces), there is a simplified method called the
Energy Method.
( ) 0
d
T U
dt
+ =
Lecture Notes -30- 06/16/06 12:25 PM
Lagrange Example A

In order to solve the Lagrange Equation,

i
i i i i
d T T U D
F
dt q q q q
| |
+ + =
|

\ .

,

system, it is first necessary to formulate
the energy equations: kinetic, potential and
dissipative.

First, formulating the kinetic energy yields,
2
1
2
T mx =

Second, formulating the potential energy yields,
2
1
2
U kx mgx = +

Finally, formulating the dissipative energy yields,
2
1
2
D cx =

Next, it is necessary to evaluate each of the Lagrange terms,

( ) ( )
2
1
2
d T d d
mx mx mx
dt x dt x dt
| | | |
= = =
| |

\ . \ .

( )
2
1
2
0
T
mx
x x

= =

( )
2
1
2
U
kx mgx kx mg
x x

= + = +

( )
2
1
2
D
cx cx
x x

= =

Which are combined to form the actual equation of motion.

0 mx kx mg cx f + + + =
mx cx kx f mg + + =

x can be redefined to be from the static equilibrium position. As a result, the mg term is
eliminated as shown.

mx cx kx f + + =
m
k c
x(t) f(t)
Lecture Notes -31- 06/16/06 12:25 PM
Lagrange Example B

Again, in order to solve the Lagrange
Equation,

i
i i i i
d T T U D
F
dt q q q q
| |
+ + =
|

\ .

,

system, it is necessary to formulate the energy
equations: kinetic, potential and dissipative.

First, formulating the kinetic energy yields,
( )
2
2
1 1
2 2 cg
T J m L = +

Second, formulating the potential energy yields, ( ) 1 cos U mgL =

Finally, formulating the dissipative energy yields, 0 D =

Next, it is necessary to evaluate each of the Lagrange terms,

( )
( )
( )
2
2 2 2
1 1
2 2 cg cg cg
d T d d
J m L J mL J mL
dt dt dt

| | | |
= + = + = +
| |

\ . \ .

( )
( )
2
2
1 1
2 2
0
cg
T
J m L

= + =

( ) ( )
1 cos sin
U
mgL mgL

= =

( ) 0 0
D

= =

Which are combined to form the actual equation of motion.

2
0 sin 0 0
cg
J mL mgL + + + =

This time, since the mg term ends up on the left hand side (multiplied by the coordinate),
it is not possible to eliminate the mg term from the solution. So the exact equation of
motion remains,

( )
2
sin 0
cg
J mL mgL + + =

,
cg
m J
L
Lecture Notes -32- 06/16/06 12:25 PM
However, because the sin mgL term is non-linear in the desired coordinate, it is
possible to reduce the solution to a linear approximation by using sin as follows.

( )
2
0
cg
J mL mgL + + =

Lecture Notes -33- 06/16/06 12:25 PM
Energy Example A

In order to solve the Energy Equation,

( ) 0
d
T U
dt
+ = ,

for the given single degree-of-freedom system, it
is first necessary to formulate the energy
equations: kinetic and potential.

Formulating the kinetic energy yields,
2
1
2
T mx = ,
and formulating the potential energy yields,
2
1
2
U kx mgx = + .

Evaluating the Energy equation yields,

( ) ( )
2 2
1 1
2 2
0
d d
T U mx kx mgx mxx kxx mgx
dt dt
+ = + + = + + =

Observe the odd form of the equation. There is an extra velocity term, x , that must be
canceled out to yield the actual equation of motion.

0 mx kx mg + + =
mx kx mg + =

x can be redefined to be from the static equilibrium position. As a result, the mg term is
eliminated as shown.

0 mx kx + =

NOTE: Comparing the solution generated by Lagrange Method to that of the Energy
Method
*
shows that the Energy Method is not simply a single degree-of-freedom
Lagrange Solution, but rather a related, by not equivalent technique. Recognize also that
the application of the Energy Method is extremely limited, i.e. single degree-of-freedom,
conservative and un-forced. The Lagrange Method has no such limitations and is in fact a
completely general solution method, comparable in usefulness to Newtons Method.

*
Note that the Energy Method is presented for completeness. You may encounter it in your studies of other
reference material, but because it is not a general technique, it will not be used in this course.
m
k
x(t)
Lecture Notes -34- 06/16/06 12:25 PM
#7 - Steady-State Solution of Equation of Motion

The general solution to the equation of motion involves both the homogeneous and the
particular solution. While the form of the homogeneous solution is dictated by the nature
of the governing second order constant matrix coefficient differential equation, the form
of the particular solution is controlled by the form of the forcing function (excitation).

Mathematically, there are infinite possibilities, but for practical applications, the form can
be restricted to harmonic functions. By limiting the form of the excitation to harmonic
functions, the concept of steady-state solution can be defined. For vibration purposes, the
steady-state solution is the resulting response after all the initial condition transients have
decayed. When solving the response for the steady-state, the solution is most easily
formulated in either the Laplace Domain (s) or the Frequency Domain ( ). (NOTE: it
will be shown later that the Frequency Domain solution can be developed from the
Laplace Domain by letting s j = .)

Working with a representative single
degree-of freedom system (SDOF), the
equation of motion is:

( ) ( ) ( ) ( ) mx t cx t k x t f t + + =

For a harmonic excitation, the forcing
function can be represented as:

( )
st
f t Fe =

For the assumed solution, the response will
have the compatible form,

( )
st
x t Xe = , ( )
st
x t sXe = and
2
( )
st
x t s Xe =
where both F & X are complex scalars.

Substituting, the above solution forms into the differential equation of motion produces a
solution that must be valid for every value of s. This results in the following Laplace
Domain solution.

2 st st st st
ms Xe csXe kXe Fe + + =

By collecting common terms, the expression reduces to:

( )
2 st st
ms cs k Xe Fe + + =

m
k
c
x(t) f(t)
Lecture Notes -35- 06/16/06 12:25 PM
At this point the response/excitation ratio
can be calculated. This ratio is called the
Transfer Function. It is normally written
as ( ) H s .

2
1
( ) ( )
X
H s s
F ms cs k
=
+ +

While the Transfer Function is a
convenient mathematical model for
defining the input-output relationship,
well suited to controls applications, it is
less appealing for vibrations purposes
primarily because it a continuous
mathematical expression. However,
because the complex surface defined by
the ( ) H s expression is analytic, any slice
through the surface contains all the
information necessary to reconstruct the
entire surface. This recognition leads to
the use of the Fourier Transform, that while its mathematical background is different, it is
effectively the Laplace Surface evaluated at s j = . (Another argument for the
equivalence of the Laplace Transform and the Fourier Transform lies in the fact that both
transforms start with the same Time Domain function, i.e.
1 1
( ) ( ( )) ( ( )) g t L G s F G

= = .)

Evaluating the Transfer Function at s j = yields another input-output form known as
the Frequency Response Function (FRF).

2
1
( ) ( )
X
H
F m j c k

=
+ +

By contrast with the Transfer Function,
the Frequency Response Function is well
suited for vibration applications,
particularly experimental applications.
While there is no discrete equivalent to the
Laplace Transform, there is a discrete
Fourier Transform that yields essentially
the same information as the continuous
integral transform. This is particularly
important when working with discrete,
sampled time data.

By plotting both the Transfer Function and the Fourier Transform (over a limited range),
the equivalence of both transforms can be shown.
Multi-Degree of Freedom
Transfer Function
The same basic development is
applicable to the multi-degree of freedom
system, as well.

Starting from the equation
| |{ } | |{ } | |{ } { } ( ) ( ) ( ) ( ) M x t C x t K x t f t + + =
and combining with
{ } { } ( )
st
x t X e = & { } { } ( )
st
f t F e =
yields the matrix equation
| | | | | | { } { }
2 st st
M s C s K X e F e ( + + =

By recognizing the basic definition of the
Transfer Function, the expression
becomes,
| | | | | |
1
2
( ) H s M s C s K

( = + +

Frequency Response Function
Just as the Transfer Function has a multi-
degree of freedom form, so does the
Frequency Response Function.

By evaluating the multi-degree of
freedom (matrix) Transfer Function at
s j = , the multi-degree of freedom
Frequency Response Function results.
| | | | | |
1
2
( ) H M j C K

( = + +

Lecture Notes -36- 06/16/06 12:25 PM

One interesting aspect of the difference, however, is that while the Transfer Function
requires the plotting of two surfaces, the Frequency Response Function requires only two
curves, clearly a much more convenient representation.

Note that in addition to displacement per unit force input-output relationships, other
steady-state input-output relationships can be developed. Some of these alternative
relationships will be explored later as steady-state applications.

-1
-0.5
0
0.5
1
-10
-5
0
5
10
0
0.2
0.4
0.6
0.8
1
Sigma [rad/sec]
Omega [rad/sec]
M
a
g
n
i
t
u
d
e
-1
-0.5
0
0.5
1
-10
-5
0
5
10
-4
-2
0
2
4
Sigma [rad/sec]
Omega [rad/sec]
P
h
a
s
e

[
r
a
d
]
-10 -8 -6 -4 -2 0 2 4 6 8 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Omega [rad/sec]
M
a
g
n
i
t
u
d
e

-10 -8 -6 -4 -2 0 2 4 6 8 10
-4
-3
-2
-1
0
1
2
3
4
Omega [rad/sec]
P
h
a
s
e

[
r
a
d
]
Lecture Notes -37- 06/16/06 12:25 PM
#8 - The Frequency Response Function (FRF)

In general, for most vibrations problems, the input/output relationship known as the
Frequency Response Function (FRF) will be used. This function describes the
input/output relationship on a frequency-by-frequency basis.

Although not generally referred to as a Frequency Response Function, the effect of a
graphic equalizer on the sound from a stereo system can be described by the frequency
response function. In this case, the input is the raw audio and the output is the modified
audio. By adjusting the amplification or attenuation (and relative phasing) of the various
bands, the equalizer can give the resulting audio a variety of sound characteristics, from
flat response to enhancing male or female voices to concert hall sound effects.

Recall that the Frequency Response Function can be expressed by evaluating the Transfer
Function at s j = .

2
1
( ) ( )
X
H
F m j c k

=
+ +

By contrast with the Transfer Function,
the Frequency Response Function is well
suited for vibration applications,
particularly experimental applications.
While there is no discrete equivalent to the
Laplace Transform, there is a discrete
Fourier Transform that yields essentially
the same information as the continuous
integral transform. This is particularly
important when working with discrete,
sampled time data.

Input-Output Model

For the frequency response function, the following pictorial model is helpful. In fact,
when used for determining the steady state
response of a system, its application is
extremely simple. It is merely a
multiplication in the frequency domain.

( ) ( ) ( ) X H F =

Frequency Response Function
Just as the Transfer Function has a multi-
degree of freedom form, so does the
Frequency Response Function.

By evaluating the multi-degree of
freedom (matrix) Transfer Function at
s j = , the multi-degree of freedom
Frequency Response Function results.
| | | | | |
1
2
( ) H M j C K

( = + +

( ) H
( ) F ( ) X
Lecture Notes -38- 06/16/06 12:25 PM
Complex Exponential vs. Sine/Cosine Relationship

The relationship between the complex exponential form and the sine/cosine form of the
can be developed as follows:

Starting with the complex exponential form for the forcing function,

*
( )
j t j t
f t Fe F e

= +

and recalling the Euler identity,

cos sin
i
e i
= +

the relationship is easily developed. (Remember that F is complex.) Expanding the
expression for ( ) f t ,

( ) ( )(cos sin ) ( )(cos sin )
R I R I
f t F jF t j t F jF t j t = + + +

Collecting real and imaginary terms yields,

( ) ( cos sin ) ( cos sin ) ( cos sin ) ( cos sin )
R I I R R I I R
f t F t F t j F t F t F t F t j F t F t = + + + +

Observing that the imaginary terms cancel and that the real terms are equivalent yields,

( ) 2 cos 2 sin
R I
f t F t F t =

Comparing with a traditional differential equations solution of ( ) cos sin f t A t B t = +
shows that 2
R
A F = and 2
I
B F = . Despite the use of complex exponentials in the
development, clearly the resulting time domain waveform is real.

FRF Characteristics

This figure presents a typical
single degree-of-freedom
frequency response function.
The FRF is present two sided
(both positive and negative
frequency) to clearly show that
the FRF for positive and
negative frequencies are
complex conjugates.

*
( ) ( ) H H =
-20 -15 -10 -5 0 5 10 15 20
10
-5
10
-4
10
-3
10
-2
Frequency [Hz]
M
a
g
n
i
t
u
d
e

[
m
/
N
]
-90
0
90
180
m = 5 kg ; c = 20 Ns/m ; k = 1000 N/m
P
h
a
s
e

[
d
e
g
]
Lecture Notes -39- 06/16/06 12:25 PM
Because of this complex conjugate relationship, usually only the positive frequency part
is plotted.

Predicted Response
*

It is important to remember that although the FRF is defined for all frequencies, the
steady-state response (output) for a linear system will always be at the same frequency as
the excitation (input). This prediction comes by manipulating the definition of the
frequency response function, as shown.

( ) ( ) ( ) X H F =

For a single degree-of-freedom system, the predicted steady-state response has the
following form.

2
( )
( )
F
X
m j c k

=
+ +

Peak Response

Another important aspect of the frequency response function involves the frequency of
maximum response. It will be designated by
p
.

Starting with the single degree-of-freedom frequency response function,

2
1
( ) H
m j c k

=
+ +

it is possible to calculate the frequency of maximum (peak) response. Formulating the
magnitude of the frequency response yields,

( ) ( )
2
2
2
1
( ) H
k m c

=
+

In order to identify the frequency of maximum response, it is necessary to evaluate the
derivative of the frequency response magnitude with respect to frequency as follows.

*
TODO: Show example F(w), show resulting X(w). (FRF continuous, F&X discrete) Express F(w) =
CONST . Show units of FRF = 1/k, compliance. Work example. Show MDOF FRF for comparison.
Lecture Notes -40- 06/16/06 12:25 PM
( )
( ) ( )
2
2
2
1
( )
d d
H
d d
k m c

| |
|
=
|
| +
\ .

( ) ( ) ( )
( )
1
2
2 2
2
( )
d d
H k m c
d d

| |
| = +
|
\ .

( ) ( ) ( )
( )
( ) ( )
( )
3
2 2
2 2 2
2 2
1
( )
2
d d
H k m c k m c
d d

= + +

( ) ( ) ( )
( )
( ) ( ) ( ) ( )
3
2
2 2
2 2 2
1
( ) 2 2
2
d d d
H k m c k m k m c c
d d d

| |
= + +
|
\ .

( ) ( ) ( )
( )
( )( ) ( )
( )
3
2
2 2
2 2
1
( ) 2 2 2
2
d
H k m c k m m c c
d

= + +

In order to satisfy the maximum (or minimum), the derivative of the magnitude of the
frequency response function,
( )
( ) 0
p
d
H
d

= , must equal zero. Since only the second

term in the above relation can be equal to zero (except at = ), the solution can be
reduced to solving the following equation.

( )( ) ( ) ( )
2
2 2 2 0
p p p
k m m c c + =

( ) ( )
2 2 2
2 0
p p
km m c + =

By dividing by the mass,
2
m , the expression can be simplified by recognizing that
2
k
m
= , 2
c
m
= and 0
p
= cannot be the peak response for all possible MCK
combinations.

2
2
2 0
p
k c
m m
| | | |
+ =
| |
\ . \ .

( ) ( )
2
2 2
2 2 0
p
+ =

2 2 2 2
2 0
p
=

( )
2 2 2
1 2
p
=
Lecture Notes -41- 06/16/06 12:25 PM

Therefore, the frequency of maximum (peak) response is as follows.

2
1 2
p
=

For comparison, recognize that the damped natural frequency is
2
1
r r
= .

Resonant Response

Starting with the definition of resonance, = , (i.e. the response of the system to an
excitation at the undamped natural frequency of the system.), the amplitude of the system
response can be calculated.

Starting with the single degree-of-freedom frequency response function,

2
1
( ) H
m j c k

=
+ +

from the definition of resonance, = , (and
2
k m = ) the resonant response, ( ) H ,
equals

1 1 1 1 1
( )
2
2 2
c
H
k
j c j c j k j km
j

= = = = =

If the damping is small (i.e. 0.1 < ), then the peak response can be approximated as

1
( ) ( )
2
p
H H
j k
=

Half Power Method

After defining the resonant frequency, it is possible to estimate the damping, , from the
shape of the frequency response function in the vicinity of the resonance. Again,
assuming that the damping is small (i.e. 0.1 < ), the half-power bandwidth points can
be calculated and then damping can be estimated from the bandwidth and the resonance
location.

Starting with the definition of the half-
power bandwidth points,

L

H

p

( )
2
p
H

( )
p
H
Lecture Notes -42- 06/16/06 12:25 PM
1
2
( ) ( ) ( )
p L H
H H H = =

and recognizing that for a lightly damped system,
p
, then expanding the single
degree-of-freedom frequency response yields,

2
1 1 1
2 2 j k k m j c
=
+

Manipulating the expression yields a form from which the half-power bandwidth points
may be calculated.

2 2
2 2
1 1 1 1
2 2
1 1 2
c
j j
k

= =
+ +

2
2
2 2 1 2 j

= +

Squaring both sides yields

2
2 2 2 4 2
2 2 2
2 2 2 4 2
8 1 4 1 2 4

| |
= + = + +
|

\ .

( ) ( )
4 2
2 2
4 2 1 8 0

| | | |
+ + =
| |

\ . \ .

Solving for
2
| |
|
\ .
yields,

( ) ( ) ( )
2
2 2 2
2
4 2 4 2 4 1 8
2

| |
=
|
\ .

2
2 2
1 2 2 1

| |
= +
|
\ .

Assuming is small (i.e. 1 ), then
2
0 , and the half-power bandwidth points are
approximated by,

2
1 2

| |

|
\ .

Lecture Notes -43- 06/16/06 12:25 PM

Taking the difference between the points yields,

2 2
2
4
H L H L H L

+ | || |

| |

\ .\ .

Using the approximation, 2
H L
+ , allows reducing the expression to,

2
H L

| |

|
\ .

Therefore the damping factor can be approximated as,

2
H L

This approximation is also often written as
2
c
f
f

or
2
c
BW
f
.

While this method appears particularly simple, in practice, there are several difficulties
utilizing it effectively. Specifically, when using real sampled data, the ability to
determine the actual magnitude and location of ( )
p
H is limited by the measured data.
The actual sampled data will not in general include the exact frequency (
p
) and so both
the location and magnitude will be in error. Since the magnitude of ( )
p
H is used to
estimate the half-power bandwidth points, the location of
L
and
H
will be in error. In
general, the errors tend to be the following: under estimating the magnitude of ( )
p
H ,
under estimating the magnitude of the half-power points, estimating
L
too low, and
estimating
H
too high. The compound effect of these errors is that generally the
damping factor ( ) is estimated too high.

Lecture Notes -44- 06/16/06 12:25 PM
FRF Example A

Given a single degree-of-freedom (SDOF) system with 5 m kg = , 20
N s
m
c

= &
1000
N
m
k = , assuming a harmonic excitation of magnitude 20N @ 10Hz , determine the
response of the system to the given force.
Solution

First, identify the actual steady-state operating frequency.

10Hz = -or- 20 62.832
rad rad
s s
=

Then, evaluate the SDOF frequency response function.

2
1
( ) H
m j c k

=
+ +

( )
2
1
(20 )
5 20 20 20 1000
rad
s
rad rad N s N
s s m m
H
kg j

=
+ +

( )
1
(20 )
18, 739 1, 257
rad
s
N
m
H
j
=
+

( )
6
(20 ) 53.125 3.563 10
rad m
s N
H j

=

Finally, determine the response in the frequency domain.

( ) ( ) ( ) X H F =
( )
6
(20 ) 53.125 3.563 10 20
rad m
s N
X j N

=
( )
3
(20 ) 1.063 0.071 10
rad
s
X j m

= -or- 1.06 179.16 mm -or-
1.06 3.075rad mm

Therefore, the time domain response of the system to the given forcing function, ( ) f t , is
expressed by
*
:

( ) 20cos(20 ) f t t N =
( ) 1.06cos(20 3.075) x t t mm =

*
TODO: Need to show equivalence of two sided FRF solution.
Lecture Notes -45- 06/16/06 12:25 PM
#9 - Rotating Unbalance

For many situations, the source of excitation is an unbalance in a rotating machine part.
This internal (self) excitation is a function of the operation of the mechanism. However,
the techniques developed are applicable to this condition, as well.

By defining the position of the unbalance mass (
e
m ) as ( ) ( ) x t e t + and the mass of the
system moving in translation only as
e
m m , we
can write the equation of motion for this
representative single degree-of-freedom system
as:

( ( ) ( )) ( ) ( ) ( ) ( ) 0
e e
m x t e t m m x t cx t k x t + + + + =

Rearranging terms yields:

( ) ( ) ( ) ( )
e
mx t cx t k x t m e t + + =

Comparing this solution with the previously
developed equation of motion shows that the expression ( )
e
m e t is equivalent in form to
the original external forcing function ( ) f t . As such we can use the same harmonic
function assumed solution form to convert this to a steady-state solution.

By using both,
( )
st
x t Xe = , ( )
st
x t sXe = &
2
( )
st
x t s Xe =
and
( )
st
e t Ee = , ( )
st
e t sEe = &
2
( )
st
e t s Ee =
where both X & E are complex scalars.

Substituting, the above solution forms into the differential equation of motion produces a
solution that must be valid for every value of s. This results in the following Laplace
Domain solution.

2 2 st st st st
e
ms Xe csXe kXe m s Ee + + =

By collecting common terms, the expression reduces to:

( )
2 2 st st
e
ms cs k Xe m Es e + + =

Again, comparison with the Laplace Domain steady-state solution shows that
2
e
F m Es . Calculation of the effective Transfer Function yields:
m
k
c
x(t)
me e
Lecture Notes -46- 06/16/06 12:25 PM

2
2
( ) ( )
e
X s
H s s
m E ms cs k
= =
+ +

Or expressed in the Frequency Domain as:

2
2
( ) ( )
e
X
H
m E m j c k

= =
+ +

In the Frequency Domain, the effective forcing function is
2
e
F m E . As can be seen
from the form, the magnitude of the force is linearly proportional to both the unbalance
mass and eccentricity, however, it is proportional to the square of the rotational frequency
. Because of this, the unbalance is normally expressed in mass times length units (
e
m e )
and hence the above frequency response form (displacement / unbalance).

-1
-0.5
0
0.5
1
-10
-5
0
5
10
0
2
4
6
8
10
Sigma [rad/sec]
Omega [rad/sec]
M
a
g
n
i
t
u
d
e
-1
-0.5
0
0.5
1
-10
-5
0
5
10
-4
-2
0
2
4
Sigma [rad/sec]
Omega [rad/sec]
P
h
a
s
e

[
r
a
d
]

-10 -8 -6 -4 -2 0 2 4 6 8 10
0
1
2
3
4
5
6
7
8
9
Omega [rad/sec]
M
a
g
n
i
t
u
d
e
-10 -8 -6 -4 -2 0 2 4 6 8 10
-4
-3
-2
-1
0
1
2
3
4
Omega [rad/sec]
P
h
a
s
e

[
r
a
d
]
Lecture Notes -47- 06/16/06 12:25 PM
Rotating Example A

A machine of 100 kg mass has a 20 kg rotor with 5 mm eccentricity. The mounting
springs have 85
kN
m
k = and the damping is negligible. The operating speed is 600 rpm
and the unit is constrained to move vertically. (a) Determine the dynamic amplitude of
the machine. (b) Redesign the mounting so that the dynamic amplitude is reduced to one
half of the original value, but maintaining the same natural frequency.

Solution

Given the following parameters:

100 m kg = , 20
e
m kg = , 5 e mm = , 85
kN
m
k = ,
0
N s
m
c

& 600rpm =

(a) Starting with the frequency response function
previously developed,

2
2
( ) ( )
e
X
H
m e m j c k

= =
+ +

Solve for the dynamic amplitude X by rearranging the expression and evaluating its
magnitude,

2
2
( )
e
m e
X
m j c k

=
+ +

The additional parameters needed are the unbalance and the rotational frequency in
rad
s
.

3
20 5 10 0.1
e
m e kg m kg m
= =
min
600
2 20 62.83
60
rad rad
s s
s
rpm
= = =

Evaluating the dynamic amplitude with 0 c = yields,

2 2
2 2
(0.1 )(20 )
85, 000 100 (20 )
rad
s e
N rad
m s
m e kg m
X
k m kg

= =

2
2
394.78
0.001274
309, 784
kg m
s
kg
s
X m
= =
or 1.274mm
m
k
c
x(t)
me e
Lecture Notes -48- 06/16/06 12:25 PM

(b) There are two options for reducing the dynamic amplitude by half and yet keeping the
same natural frequency, increasing mass and increasing damping.

1
2
0.6372
reduced original
X X mm = =
85, 000
29.15
100
N
m
rad
s
k
m kg
= = =

Option 1: increasing mass.

Manipulating the expression for dynamic amplitude

2 2 2
2 2 2 2
( ) ( )
e e e
k
m
m e m e m e
X
k m m m

= = =

yields an expression for the increased mass.

2 2
2 2 2 2
(0.1 )(20 )
200
( ) 0.6372 ((29.15 ) (20 ) )
rad
s e
rad rad
s s
m e kg m
m kg
X mm

= = =

However, keeping the same implies a corresponding increase in k . Therefore,

2 2
200 (29.15 ) 170, 000
rad N
s m
k m kg = = = or 170
kN
m

Notice how the solution makes sense intuitively. The operating speed is above the natural
frequency of the system. In that region, the frequency response function magnitude
(
e
X
m e
) approaches
1
m
, so it would be expected that to reduce the dynamic amplitude by
half would require approximately doubling the mass.

Option 2: increasing damping.

Again, manipulating the expression for dynamic amplitude

2
2
e
m e
X
m j c k

=
+ +

2 2
e
X m j c k m e + + =
2
2
e
m e
m j c k
X
+ + =
Lecture Notes -49- 06/16/06 12:25 PM
( )
2
2
2
2 2
2
( )
e
m e
k m c
X
+ =
( )
2
2
2
2 2
2
( )
e
m e
c k m
X
=

yields an expression for the necessary damping

( )
2
2
2
2
2
2
e
m e
k m
X
c

=
( )
( )
2
2
2
2
2
2
0.1 (20 )
85, 000 100 (20 )
(0.006372 )
(20 )
rad
s
N rad
m s
rad
s
kg m
kg
m
c

=
( )
2 2
4
2 2
2
4 4
2 2
2 2
3
2
9 9
9 2
155.85 10
309, 784
383.87 10 95.966 10
406.0 10
8,540
3,947.8 3, 947.8
kg m
kg s
kg kg
s
kg s s
s
rad rad
s s
m
c
= = =

or more familiarly, 8, 540
N s
m
c

= . This can be expressed as a fraction of critical damping
(
c
c
c
= ) by recalling the expression for the critical damping value, 2
c
c km = .

8, 540
1.465
2 2 85, 000 100
N s
m
N
m
c
km kg

= = =
or 146.5%

Notice that to reduce the dynamic response of this system utilizing only damping requires
an over-damped condition of nearly 150%.

For this problem, the mass addition is most likely to be the easiest and cheapest practical
solution.

Lecture Notes -50- 06/16/06 12:25 PM
#10 - Steady-State Applications

In this section, two different but related steady-state applications will be developed. The
first is the force transmissibility relationship. For transmissibility, the interest is in
determining the ratio of the force transmitted to the base ( ( )
T
F ) to the applied force
( ( ) F ) as a function of frequency. The second application is vibration isolation. For
isolation, the interest is in determining the ratio of the response of the system ( ( ) X ) to
the motion of the base ( ( ) Y ) as a function of frequency. (This is sometimes referred to
as response ratio.) Just as in the earlier development of the Frequency Response Function
(FRF), for both applications, the focus will be upon harmonic excitation.

Transmissibility

For transmissibility, we want to determine the ( )
T
F
F
relationship. We begin with the
equation of motion for our representative single degree-of -freedom system. (NOTE: we
assume negligible motion of the base.)

( ) ( ) ( ) ( ) mx t cx t k x t f t + + =

In addition, we need the restraining force at
the other end of the spring/damper.

( ) ( ) ( )
T
cx t k x t f t + =

Applying the same transformations
developed earlier, namely that the forcing
functions and responses can be express as
complex exponentials, and collecting terms,
we get:

( )
2 st st
ms cs k Xe Fe + + =
and
( )
st st
T
cs k Xe F e + =

Finally, by taking the ratio of the second equation to the first, we get the desired
transmissibility relationship.

2
( )
T
F cs k
s
F ms cs k
+
=
+ +

m
k
c
x(t) f(t)
f
T
(t)
Lecture Notes -51- 06/16/06 12:25 PM

2
( )
T
F j c k
F m j c k

+
=
+ +

Isolation

For isolation, we want to determine the ( )
X
Y
relationship. We begin with the equation
of motion for our representative single degree-of freedom system. (NOTE: this time the
base motion is NOT negligible, however there is no explicitly applied external force.)
Because there is motion for both the structure
and the base, we need to use the difference in
motion across the spring/damper.

( ) ( ) ( ) ( ) ( ) ( ) ( ) 0 mx t c x t y t k x t y t + + =

Transforming the equation as before, we get:

( ) ( )
2
0
st st st st st
ms Xe c sXe sYe k Xe Ye + + =

Collecting terms, and moving the base motion to
the other side of the equation yields,

2
( ) ( )
st st
ms cs k Xe cs k Ye + + = +

Finally, by cross dividing the relevant terms, we get the desired isolation relationship.

2
( )
X cs k
s
Y ms cs k
+
=
+ +


2
( )
X j c k
Y m j c k

+
=
+ +

m
k
c
x(t)
y(t)
Lecture Notes -52- 06/16/06 12:25 PM
Conclusion

Comparing the expression for transmissibility to the expression for isolation, we find that
they are identical. This interesting result implies that the process of determining the force
transmitted to the base is equivalent to determining the base motion transmitted to the
system.

-1
-0.5
0
0.5
1
-10
-5
0
5
10
0
2
4
6
8
10
Sigma [rad/sec]
Omega [rad/sec]
M
a
g
n
i
t
u
d
e
-1
-0.5
0
0.5
1
-10
-5
0
5
10
-4
-2
0
2
4
Sigma [rad/sec]
Omega [rad/sec]
P
h
a
s
e

[
r
a
d
]
-10 -8 -6 -4 -2 0 2 4 6 8 10
0
1
2
3
4
5
6
7
8
9
10
Omega [rad/sec]
M
a
g
n
i
t
u
d
e

-10 -8 -6 -4 -2 0 2 4 6 8 10
-4
-3
-2
-1
0
1
2
3
4
Omega [rad/sec]
P
h
a
s
e

[
r
a
d
]

Lecture Notes -53- 06/16/06 12:25 PM
Transmissibility Example A

A vertical single-cylinder diesel engine of 500 kg mass is mounted on springs with
200
kN
m
k = and dampers with 0.2 = . The rotating parts are well balanced. The mass of
the equivalent reciprocating parts is 10 kg and the stroke is 200 mm. Find the dynamic
amplitude of the vertical motion, the transmissibility, and the force transmitted to the
foundation, if the engine is operated at (a) 200 rpm; (b) 600 rpm.

Solution

Given the following parameters:

500 m kg = , 10
R
m kg = , 200 stroke mm = ,
200
kN
m
k = & 0.2 =

(a) 200 rpm

Starting with the single degree-of-freedom
frequency response function previously developed,

2
1
( ) ( )
X
H
F m j c k

= =
+ +

Solve for the dynamic amplitude X by rearranging the expression and evaluating its
magnitude,

2
( )
( )
F
X
m j c k

=
+ +

The additional parameters needed are the equivalent force, the damping and the operating
frequency expressed in
rad
s
. From the mechanical unbalance development, the equivalent
force can be expressed as
2
eq e
F m e = . Recognizing that the reciprocating mass is the
effective mass,
e R
m m = , and that the effective eccentricity is half the stroke,
2
stroke
e = , yields,

min
200 20
2 20.94
60 3
rad rad
s s
s
rpm
= = =
2 2
10 0.1 (20.94 ) 438.7
rad
s eq e
F m e kg m N = = =
2 2 0.2 200, 000 500 4000
N N s
m m
c km kg

= = =
m
k c
x(t)
stroke
Lecture Notes -54- 06/16/06 12:25 PM

Evaluating the dynamic amplitude yields,

2 2
438.7
200, 000 500 (20.94 ) (20.94 )(4000 )
eq
N rad rad N s
m s s m
F
N
X
k m j c kg j

= =
+ +

( ) 2
438.7
0.0051
19, 324 83, 776
kg
s
N
X m
j
= =
+
or 5.1mm

Evaluating the transmissibility yields,

2 2
200, 000 (20.94 )(4000 )
200, 000 500 (20.94 ) (20.94 )(4000 )
N rad N s
m s m
N rad rad N s
m s s m
k j c j
TR
k m j c kg j
+ +
= =
+ +

( )
( )
200, 000 83, 776
2.52
19, 324 83, 776
N
m
N
m
j
TR
j
+
= =
+

From the definition of transmissibility, namely
T
F
TR
F
= , the force transmitted to the
foundation can be determined.

2.52 438.7 1106
T eq
F TR F TR F N N = = = =

Notice that due to the characteristics of the system, the force transmitted to the
foundation is approximately 2.5 times the force applied to the system.

(b) 600 rpm

Again solving for the dynamic amplitude X using the rearranged expression and
evaluating its magnitude,

2
( )
( )
F
X
m j c k

=
+ +

The additional parameters needed are the equivalent force, the damping and the operating
frequency expressed in
rad
s
. From the mechanical unbalance development, the equivalent
force can be expressed as
2
eq e
F m e = . Recognizing that the reciprocating mass is the
effective mass,
e R
m m = , and that the effective eccentricity is half the stroke,
2
stroke
e = , yields,

min
600
2 20 62.83
60
rad rad
s s
s
rpm
= = =
Lecture Notes -55- 06/16/06 12:25 PM
2 2
10 0.1 (62.83 ) 3, 948
rad
s eq e
F m e kg m N = = =
2 2 0.2 200, 000 500 4000
N N s
m m
c km kg

= = =

Evaluating the dynamic amplitude yields,

2 2
3, 948
200, 000 500 (62.83 ) (62.83 )(4000 )
eq
N rad rad N s
m s s m
F
N
X
k m j c kg j

= =
+ +

( ) 2
6 3
3, 948
0.0022
1.774 10 251.3 10
kg
s
N
X m
j
= =
+
or 2.2mm

Evaluating the transmissibility yields,

2 2
200, 000 (20.94 )(4000 )
200, 000 500 (62.83 ) (62.83 )(4000 )
N rad N s
m s m
N rad rad N s
m s s m
k j c j
TR
k m j c kg j
+ +
= =
+ +

( )
( )
3
6 3
200, 000 251.3 10
0.179
1.774 10 251.3 10
N
m
N
m
j
TR
j
+
= =
+

From the definition of transmissibility, namely
T
F
TR
F
= , the force transmitted to the
foundation can be determined.

0.179 3,948 708
T eq
F TR F TR F N N = = = =

Notice that although the equivalent applied force is 900% larger than for the first case,
the force actually transmitted to the foundation is about 30% smaller.

Lecture Notes -56- 06/16/06 12:25 PM
#11 - Multiple Degree of Freedom Systems

MDOF Example

Starting with the simple two degree-of-freedom
system shown, the objective will be to determine
the equations of motion and the natural
frequency and mode shapes.

The equations of motion can be found by either
of two techniques, Newtons Method or
Lagranges Method.

Equations of Motion Newtons
Method

As has been shown, Newtons Method involves
identifying all forces, both internal and external,
acting on each body and then expressing those forces in terms of the given coordinates.
Therefore, the first action must be the development of the free body diagrams for each
body.

The given two degree-of-freedom
system has the two free body
diagrams, show at right.

Evaluating Newtons equation,
i
i
f ma =

, for each body yields a set
of equations describing the motion of
the system.

( )
1 1 1 2 1 2 1 1 1
f k x k x x m g m x + =
( )
2 2 2 1 2 2 2
f k x x m g m x + =

Collecting and rearranging the terms
into standard form yields,

( )
1 1 1 2 1 2 2 1 1
m x k k x k x f m g + + = +
2 2 2 2 2 1 2 2
m x k x k x f m g + = +
1
( ) x t
1
( ) f t
2
( ) x t
2
( ) f t
1
k
2
k
1
m
2
m
( )
2 2 1
k x x
2
m
1 1
k x
1
m
( )
2 1 2
k x x
=
1
m
1 1
m x
=
2
m
2 2
m x
body #1
body #2
2
m g
1
m g
2
f
1
f
Lecture Notes -57- 06/16/06 12:25 PM

Or expressed in matrix form as,

1 1 1 2 2 1 1 1
2 2 2 2 2 2 2
0
0
m x k k k x f m g
m x k k x f m g
+ ( (
+ = +
` ` ` `
( (
) ) ) )

Recall that when the gravitational term (
i
m g ) ends up on the force side of the equation,
the associated coordinate (
i
x ) can be redefined from the static equilibrium position
thereby eliminating the
i
m g term from the equation. In this case, both terms are
eliminated.

1 1 1 2 2 1 1
2 2 2 2 2 2
0
0
m x k k k x f
m x k k x f
+ ( (
+ =
` ` `
( (
) ) )

Equations of Motion Lagranges Method

As has been shown, Lagranges Method involves identifying the kinetic, potential and
dissipative energies for the entire system, then evaluating the Lagrange Equation for each
degree of freedom.

i
i i i i
d T T U D
F
dt q q q q
| |
+ + =
|

\ .

First, identifying the kinetic energy term,

2 2
1 1
1 1 2 2 2 2
T m x m x = +

And the potential energy term,

( )
2
2
1 1
1 1 2 2 1 1 1 2 2 2 2
U k x k x x m gx m gx = +

Provides all the information necessary to determine the equations of motion. Note that
there are no dissipative terms for this problem, so 0 D = .

Second, evaluating the various partial derivatives for each degree of freedom, yields,

(First degree of freedom)
( ) ( )
2 2
1 1
1 1 2 2 1 1 1 1 2 2
1 1
d T d d
m x m x m x m x
dt x dt x dt
| | | |
= + = =
| |

\ . \ .

Lecture Notes -58- 06/16/06 12:25 PM
( )
( )
( )( )
2
2
1 1
1 1 2 2 1 1 1 2 2 1 1 2 2 1 1 2 2
1 1
1
U
k x k x x m gx m gx k x k x x m g
x x

= + = +

(Second degree of freedom)
( ) ( )
2 2
1 1
1 1 2 2 2 2 2 2 2 2
2 2
d T d d
m x m x m x m x
dt x dt x dt
| | | |
= + = =
| |

\ . \ .

( )
( )
( )
2
2
1 1
1 1 2 2 1 1 1 2 2 2 2 1 2 2 2
2 2
U
k x k x x m gx m gx k x x m g
x x

= + =

Finally, collecting the various terms for each degree of freedom yields the equations of
motion.

( )
1 1 1 2 1 2 2 1 1
m x k k x k x f m g + + = +
2 2 2 2 2 1 2 2
m x k x k x f m g + = +

Or expressed in matrix form as,

1 1 1 2 2 1 1 1
2 2 2 2 2 2 2
0
0
m x k k k x f m g
m x k k x f m g
+ ( (
+ = +
` ` ` `
( (
) ) ) )

Again recall, when the gravitational term (
i
m g ) ends up on the force side of the equation,
then the associated coordinate (
i
x ) can be redefined from the static equilibrium position
thereby eliminating the
i
m g term from the equation. In this case, both terms are
eliminated.

1 1 1 2 2 1 1
2 2 2 2 2 2
0
0
m x k k k x f
m x k k x f
+ ( (
+ =
` ` `
( (
) ) )

Notice that the actual equations of motion identified are identical, regardless of whether
Newtons Method or Lagranges Method was used.

Natural Frequency and Mode Shape

In order to identify the natural frequency and mode shape for a set of simultaneous
differential equations of motion, it is necessary to solve an eigenvalue problem. By
transforming the set of equations to the Laplace Domain ( s ), the differential equations of
motion are transformed into a set of simultaneous algebraic equations.

Lecture Notes -59- 06/16/06 12:25 PM
The transformation to the Laplace Domain can be easily accomplished by using an
assumed solution approach.

( )
st
x t Xe = & ( )
st
f t Fe =

Taking derivatives of the assumed solutions yields,

( )
st
x t sXe = & ( )
2 st
x t s Xe =

Recognize that the vector form of the assumed solution is simply,

{ } { }
st
x X e = & { } { }
st
f F e =
{ } { }
st
x s X e = & { } { }
2 st
x s X e =

Substituting into the equation of motion yields,

1 1 1 2 2 1 1 2
2 2 2 2 2 2
0
0
st st st
m X k k k X F
s e e e
m X k k X F
+ ( (
+ =
` ` `
( (
) ) )

Because 0
st
e for all time, the solution reduces to,

1 1 1 2 2 1 1 2
2 2 2 2 2 2
0
0
m X k k k X F
s
m X k k X F
+ ( (
+ =
` ` `
( (
) ) )

Since the objective is to calculate natural frequency and mode shape, it is necessary to
solve the homogeneous portion of the solution. Collecting terms and setting the force side
of the equation to zero yields,

1 1 2 2 1 2
2 2 2 2
0 0
0 0
m k k k X
s
m k k X
+ ( ( (
+ =
` ` ( ( (
) )

The solution is now in one of the eigenvalue/eigenvector forms. The simplest solution
approach is to recognize that in order for the equation to be valid, either of two conditions
must hold. First, the vector { } X is identically zero (the trivial solution) or second, the
determinant of the coefficient matrix is zero (i.e. the coefficient matrix is singular for
certain values of s .) The second is the only useful solution and will be pursued here.

Setting the determinant of the coefficient matrix to zero yields,

1 1 2 2 2
2 2 2
0
0
0
m k k k
s
m k k
+ ( (
+ =
( (

Lecture Notes -60- 06/16/06 12:25 PM

This determinant is called the characteristic polynomial and the solution effectively
becomes the problem of finding the roots of the characteristic polynomial.

2
1 2 2 1
2
2 2 2
0
0
0
k k k m s
k k m s
+ ( (
+ =
( (

2
1 1 2 2
2
2 2 2
0
m s k k k
k m s k
( + +
=
(
+

( )( ) ( )( )
2 2
1 1 2 2 2 2 2
0 m s k k m s k k k + + + =
( ) ( ) ( )
4 2
1 2 1 2 2 1 2 2 1 2
0 mm s mk m k m k s k k + + + + =

In general, the characteristic polynomial is complete (has all terms), but because this
example problem does not include damping, all the odd powers of s are eliminated. This
makes finding the roots (
r
) of the equation somewhat easier.

( ) ( ) ( )( )
( )
2
1 2 2 1 2 2 1 2 2 1 2 2 1 2 1 2 2
1 2
4
2
r
mk m k m k mk m k m k mm k k
mm
+ + + +
=

At this point, continued manipulation of the arbitrary variables does not contribute to the
understanding of the solution, but only proves pure obstinacy.

Returning to the eigenvalue/eigenvector problem above and substituting some numerical
values should prove more interesting. Let
1
2 m kg = ,
2
1 m kg = ,
1
3
N
m
k = and
2
2
N
m
k = .

1 2
2
2 0 3 2 2 0
0 1 2 2 0
X
s
X
+ ( ( (
+ =
` ` ( ( (
) )

Evaluating the characteristic polynomial yields,

( )( ) ( )( )
2 2
2 5 2 2 2 0 s s + + =
( )
4 2
2 9 6 0 s s + + =

Which has the solution,

( )( )
( )
2
2
9 9 4 2 6
9 33
2 2 4
r

= =
2
2
2
9 5.745
0.814, 3.686
4
rad
r s

Lecture Notes -61- 06/16/06 12:25 PM
Finally, taking the square root of
2
r
yields the four roots of the original quartic equation.

0.814, 3.686 0.902 , 1.920
rad
s r
j j j j

Notice that the roots occur in complex conjugate pairs. This will always be the case for
physically real systems (i.e. mass, stiffness & damping are positive and real.)

Which means the natural frequencies are,

0.902 & 1.920
rad
s r
=

Recall the original development of the general solution to the homogeneous problem. It
was stated then that the solution could be written as pairs of complex conjugate terms in
the following form,

{ } { } { }
( )
*
#
*
*
1
( )
r r
dof
t t
r r
r r
r
x t Q e Q e

=
= +

(Ignore for the moment the
r
Q term. It is simply a scaling term to account for the
arbitrary scaling of { }
r
that results from the eigenvalue solution. In essence, it is the
constant of integration for the differential equation.)

Now that the eigenvalues (natural frequencies) are known, it is possible to solve for the
eigenvectors (mode shapes). Because the solution is a sum of complex conjugate terms, it
is only necessary to solve for the positive frequency vectors.

Again, starting with the above homogeneous equation,

1 2
2
2 0 3 2 2 0
0 1 2 2 0
X
s
X
+ ( ( (
+ =
` ` ( ( (
) )

Substitute the roots (
r
) for s and expand the equations,

( )
2
1 2
2 5 2 0
r
X X + =
( )
2
1 2
2 2 0
r
X X + + =

Since the values of
r
are chosen to make the equation set rank deficient, there is an
arbitrary choice of one value (either
1
X or
2
X .) For this example, choosing
1
1 X = yields
the following equation set,

Lecture Notes -62- 06/16/06 12:25 PM
( )
2
2
2 5
2
r
X
+
=

( )
2
2
2
2
r
X
=
+

Substituting the first root (
1
0.902 j = ) into either equation yields
2
1.686 X = .
Substituting the second root (
2
1.920 j = ) into either equation yields
2
1.186 X = .

Therefore, the natural frequencies and mode shapes for this two degree of freedom
system are,

{ }
1
1
1
0.902
1.686
rad
s
j

= =
`
)

and
{ }
2
2
1
1.920
1.186
rad
s
j

= =
`
)

plus their complex conjugates.

(Final note: if the complete homogeneous (free response) solution was desired, the final
step would be to solve for the
r
Q s by evaluating the initial conditions { } (0) x and
{ } (0) x .)

Lecture Notes -63- 06/16/06 12:25 PM
#12 Estimating Modal Parameters Frequency Domain

There are many techniques, both simple and advanced, for estimating modal parameters
(frequency, damping, mode shape, and scaling) from measured experimental data. This
section will focus upon the single degree-of-freedom frequency domain technique known
as quadrature.

Quadrature

The quadrature method is based entirely upon the characteristics of a single degree-of-
freedom free frequency response function. The quadrature method has many advantages
over the log decrement technique presented in section five. In the frequency domain,
many lightly damped, multi degree-of-freedom systems behave much like a single
degree-of-freedom system in the region around a resonance. This allows the single
degree-of-freedom techniques previously developed, to be used to estimate the modal
parameters for multi degree-of-freedom systems.

Examining the principle characteristic of a lightly damped, single degree-of-freedom
system at resonance: the magnitude peaks, the phase passes through 90 , the imaginary
part peaks, and the real part passes through zero. These characteristics, in addition to the
shape of the frequency response in the vicinity of the resonance, can be used to estimate
the natural frequency and damping.
Specifically, recall the definition of resonance (excitation at the undamped natural
frequency of the system). At resonance, the frequency response for a single degree-of-
freedom system reduces to
1
( ) H
j c
=

. This is purely imaginary, implying that the
response is 90 behind the forcing function. Therefore, the imaginary part of the
frequency response is maximum and the real part is zero. Also, from section eight, the
half-power method may be used to estimate the damping as
2
c
f
f

= .

0 1 2 3 4 5 6 7 8 9 10
10
-4
10
-3
10
-2
10
-1
10
0
10
1
10
2
Frequency [Hz]
M
a
g
n
i
t
u
d
e

[
m
/
N
]
-90
0
90
180
P
h
a
s
e

[
d
e
g
]
-1.5
-1
-0.5
0
0.5
1
1.5
I
m
a
g
i
n
a
r
y

p
a
r
t

[
m
/
N
]
0 1 2 3 4 5 6 7 8 9 10
-1.5
-1
-0.5
0
0.5
1
1.5
Frequency [Hz]
R
e
a
l

p
a
r
t

[
m
/
N
]
Lecture Notes -64- 06/16/06 12:25 PM
Examining a multi degree-of-freedom frequency response, many of the same
characteristics can be observed, in particular, the change of phase through resonance, the
peak in the magnitude and imaginary parts, and the zero crossing in the real part. From
these characteristics, in addition to the shape of the frequency response in the vicinity of
the resonance, the natural frequency and damping can be estimated. But additionally, for
multi degree-of-freedom systems, the mode shape can also be estimated.

While the mathematical form of a multi degree-of-freedom frequency response function
is more complicated than for a single degree-of-freedom, it is still possible in many
situations to reasonably apply single degree-of-freedom theory. Notice that there are
several differences between a single and multi degree-of-freedom frequency response
function. One important difference is that there is now more than one frequency response
function for the system. Additionally, notice that there are now multiple peaks and the
phase is no longer restricted to the range 0 , 180 . As a result, the phase at resonance
can be 90 . (It must still lose phase through resonance, though.)

The reason that there are multiple frequency
response functions is easy to show. Simply recall
the definition of the frequency response function,
( )
( )
( )
X
H
F

. For the given figure, there
are force (input) locations and two response
(output) locations. It is then clear that there are
four possible frequency response functions that
can be measured (or calculated.) Therefore, the
expression for the frequency response must be
extended to express which force/response pair is
being presented. The frequency response
function for multi degree-of-freedom system is
written as
( )
( )
( )
p
pq
q
X
H
F

, where p is
the input degree of freedom and q is the output
degree of freedom. Using this definition, the
multi degree-of-freedom frequency response
0 2 4 6 8 10 12
-180
-90
0
90
180
P
h
a
s
e

[
d
e
g
]
H
1 1
. . H
2 1
. . H
3 1
0 2 4 6 8 10 12
10
-4
10
-3
10
-2
10
-1
10
0
Frequenc y [ Hz ]
M
a
g
n
it
u
d
e

[
m
/
N
]
H
1 1
H
2 1
H
3 1
0 2 4 6 8 10 12
-0. 2
-0. 1
0
0. 1
0. 2
0. 3
Im
a
g

[
m
/
N
]
H
1 1
. . H
2 1
. . H
3 1
0 2 4 6 8 10 12
-0. 2
-0. 1
0
0. 1
0. 2
0. 3
Frequenc y [ Hz ]
R
e
a
l
[
m
/
N
]
H
1 1
H
2 1
H
3 1
1
( ) x t
1
( ) f t
2
( ) x t
2
( ) f t
1
k
2
k
1
m
2
m
1
c
2
c
Lecture Notes -65- 06/16/06 12:25 PM
functions plotted above make more sense. The plots represent a three degree-of-freedom
system. If all the FRFs were plotted, there would be nine FRFs displayed, however only
the three FRFs associated with the first input degree-of-freedom were actually shown,
hence the notation used in the legend (
11 21 31
, , H H H ).

Expanding the region around the third peak allows a more clear observation that the multi
degree-of-freedom frequency response shown looks much like a single degree-of-
freedom in the vicinity of the resonance.

To estimate the modal parameters:
Natural Frequency To identify the natural frequency (
r
), it is first necessary
to identify one of the following:
o The 90 phase change location (frequency)
o The peak imaginary response location (frequency)
o The zero real response location (frequency)
Damping To estimate the damping, take the peak FRF magnitude, multiply by
the square root of two (
1
2
( ) H ), find the locations (frequencies), one above and
one below, that have that FRF magnitude, the locations are called the half-power
points. Divide the difference of the two frequencies by twice the peak frequency
(natural frequency).
2
r
c
f
f

=
Mode Shape To estimate the mode shape, either
o Take the value of the peak imaginary part (at the natural frequency) for
each output (response) location keeping the input (force) location fixed.
OR
o Take the value of the peak imaginary part (at the natural frequency) for
each input (force) location keeping the output (response) location fixed.

For the third mode shown above:

The natural frequency is estimated, using the peak imaginary response, as 9.068 Hz.

8 8. 2 8. 4 8. 6 8. 8 9 9. 2 9. 4 9. 6 9. 8 10
-180
-90
0
90
180
P
h
a
s
e

[
d
e
g
]
H
1 1
. . H
2 1
. . H
3 1
8 8. 2 8. 4 8. 6 8. 8 9 9. 2 9. 4 9. 6 9. 8 10
10
-5
10
-4
10
-3
10
-2
10
-1
Frequenc y [ Hz ]
M
a
g
n
it
u
d
e

[
m
/
N
]
H
1 1
H
2 1
H
3 1
8 8. 2 8. 4 8. 6 8. 8 9 9. 2 9. 4 9. 6 9. 8 10
-0. 03
-0. 02
-0. 01
0
0. 01
0. 02
0. 03
Im
a
g

[
m
/
N
]
H
1 1
. . H
2 1
. . H
3 1
8 8. 2 8. 4 8. 6 8. 8 9 9. 2 9. 4 9. 6 9. 8 10
-0. 03
-0. 02
-0. 01
0
0. 01
0. 02
0. 03
Frequenc y [ Hz ]
R
e
a
l
[
m
/
N
]
H
1 1
H
2 1
H
3 1
Lecture Notes -66- 06/16/06 12:25 PM
3
9.068Hz =

The peak FRF magnitude equals 0.0204 m/N, therefore the half-power magnitude would
be 0.0144 m/N. So the half-power points are estimated as 9.0383 Hz and 9.0982 Hz,
which result in an estimate of damping ratio of 0.0033.

( )
3
9.0982 9.0383
0.0033
2 2 9.068
c
f Hz Hz
f Hz

= = = -or- 0.33%

The peak imaginary values, for each output (response) location for the first input (force)
location, are estimated as [-0.0164 0.0204 0.0091] m/N.

{ }
m
N 11
-0.0164 H = , { }
m
N 21
0.0204 H = & { }
m
N 31
-0.0091 H =

Therefore the mode shape is estimated as,

{ }
3
0.0164
0.0204
0.0091

=
`

)
-or-
0.80
1.00
0.45

`

)
-or-
1.00
1.24
0.55

`

)

So the modal parameter estimate for the third mode is:

3
9.068Hz = ,
3
0.33% = & { }
3
1.00
1.24
0.55

=
`

)

NOTE: By convention, modes are numbered from lowest frequency to highest frequency.
Lecture Notes -67- 06/16/06 12:25 PM
#13 Transient Solution of Equation of Motion

The last piece of the vibration problem is to solve for the transient solution of the
differential equation of motion. In general, since the excitation (forcing) function can be
literally anything, a closed form solution is not practical and numerical techniques are
used. (The numerical methods will be discussed in the next section.) However, a study of
the general solution for certain forcing functions is important to the basic understanding
of the complete solution for the equation of motion.

Impulse Response

The impulse response is the response
of the system to an impulsive
excitation. An impulse is frequently a
large magnitude, short duration force.

The impulse (
F ) of a force is defined
as ( ) f t dt
( ) F f t dt

If the impulse is unity (
1 F = ) and the duration of the pulse ( ) is allowed to approach

zero ( 0 ) then the force ( ( ) f t ) is referred to as a unit impulse. The unit impulse is
equivalent to the mathematical expression called a unit delta function ( ( ) t ). The delta
function is defined by:

( ) 0 t = when t
0
( ) 1 t dt
when 0 < <

Further, the delta function has a number of interesting and useful properties. Of particular
interest is the result of multiplication by another function.

0
( ) ( ) ( ) f t t dt f

Recalling the impulse/momentum relationship ( ( ) f t dt mdv = ) and integrating both sides
shows that an impulse acting upon a mass results in a change in velocity ( v ) without a
significant change in position. Therefore,

( ) f t
t
Lecture Notes -68- 06/16/06 12:25 PM
F
v
m
=

The free response of a single degree-of-freedom system to initial conditions is given by:

(0) (0)
( ) sin( ) (0) cos( )
r
t r
r r
r
x x
x t e t x t
| |
= +
|
\ .

Assuming the system is initially at rest, then combining the free response with the change
in momentum yields,

( ) sin( ) ( )
r
t
r
r
F
x t e t Fh t
m

= where
1
( ) sin( )
r
t
r
r
h t e t
m

=

The unit impulse response function ( ( ) h t ) is important for the analysis of transients!

Arbitrary Excitation

Having developed the unit impulse
response ( ( ) h t ), it is now possible to
establish the response ( ( ) x t ) of a
system to an arbitrary excitation
( ( ) f t ).

First identify the strength of the
impulse at time (
( ) F f = ) and
its contribution to the response ( ( ) x t ))
at time t
( ( ) ( ) ( ) x t f h t = ).
Recognized that t is constant
for the above expressions, it is
that is variable.

Because the system is linear,
the principal of superposition
applies and it is possible to
add up all the impulse
contributions. Adding up all
the contributions of the individual impulses yields the expression for the response of a
system to an arbitrary excitation, (assuming zero initial conditions.)

( ) f t
t

( ) f t
t =
t
( ) f
( ) ( ) f h t
Lecture Notes -69- 06/16/06 12:25 PM
0
( ) ( ) ( )
t
x t f h t d =

This expression is called the convolution integral. As can be seen from the form, the
forcing function ( ( ) f ) could be any arbitrary function and a general closed form
solution is impractical.
Lecture Notes -70- 06/16/06 12:25 PM
#14 Numerical Solution of Equation of Motion

The general solution of the differential equation of motion with initial conditions and an
arbitrary forcing function is most easily done numerically. While, there are numerous
techniques for solving linear, simultaneous, ordinary differential equations, this section
will focus upon only one method, Eulers Method, both for its simplicity and its
representative character.
Eulers Method

Eulers Method is the simplest first-order method. It can be used to solve differential
equations of the form, ( , ) y f x y = . It is an explicit method, which means that the
derivative ( y ) is an explicit function of the variable x and all previous values of both x
and y .

Eulers Method is very simple. Given
the current value of the function (
n
y ),
the current derivative of the function
(
n
y ) and the step size ( x ), simply
evaluate:

1 n n n
y y y x
+
+

Clearly the accuracy of this method is
determined by the step size x .

State-Space Expansion

One critical issue for the application of numerical integration to the differential equation
of motion is that the equation of motion is second-order, while the integration techniques
require first-order differential equations. This is not however a significant limitation. By a
technique known as state-space expansion, a high-order differential equation can be
converted into a set of simultaneous lower-order differential equations. For the equation
of motion, this means converting the single second-order differential equation into a pair
of simultaneous first-order differential equations.

Starting with the original second-order differential equation.

( ) f t mx cx kx = + +

Augment the equation with the identity relationship 0 mx mx = .

( ) y x
x
n
x
n
y
n
y
x
1 n
y
+
Lecture Notes -71- 06/16/06 12:25 PM
( ) f t mx cx kx = + +
0 mx mx =

Rearranging the equations yields,

( ) mx f t cx kx =
mx mx =

Converting to a vector/matrix equation set yields,

( ) mx f t cx kx
mx mx

=
` `
) )

Substituting the state vector,
x
y
x

=
`
)
, into the above expression yields,

{ } { }
( ) x x f t cx kx
d d
m m m y m y
x x mx dt dt

= = = =
` ` `
) ) )

{ }
( ) f t cx kx
m y
mx

=
`
)

Dividing by m and manipulating the right-hand side similarly yields,

{ }
( ) ( ) ( )
0 0 1 0
f t c k f t c k f t c k
x x x x x
y m m m m m m m m m
x
x x

= = + = +
` ` ` ` ` `
)

) ) ) ) )

{ } { }
( ) ( )
0 1 0 0 1 0
f t c k f t c k
x
y y m m m m m m
x

= + = +
` ` ` ` `
)

) ) ) )

Therefore, the state-space expansion of the original equation becomes,

{ } { }
( )
0 1 0
f t c k
y y m m m

= +
` `

) )
-or-
( )
0 1 0
f t c k
x x
d
m m m
x x dt

= +
` ` ` `
) )

) )

At this point, the original second-order differential equation has been manipulated into
two, simultaneous, first-order differential equations. In this form, it is suitable for use in
Lecture Notes -72- 06/16/06 12:25 PM
the numerical integration scheme presented. (Actually, it can be used with any first-order
numerical integration scheme.)

Numerical Solution

Combining the results of the two preceding sections, allows the development of an
iteration scheme for numerically evaluating the response of the system to an arbitrary
input (force).

Constructing the state vector from the current conditions,

{ }
n
n
n
x
y
x

=
`
)

Evaluating the state equation of motion yields the current derivative,

{ } { }
( )
0 1 0
n
n n
f t c k
y y m m m

= +
` `

) )

Finally, evaluating the numerical integration rule (in this case, Eulers Method) yields,

{ } { } { }
1 n n n
y y t y
+
= + -or-
1
1
n n n
n n n
x x x
t
x x x
+
+

= +
` ` `
) ) )

This provides a simple marching scheme to numerically integrate the equations of motion
applying both initial conditions and an arbitrary force. In general, a more precise
numerical integration rule may be substituted for the last step
*
.

*
TODO: Add 2DOF non-linear example.
Lecture Notes -73- 06/16/06 12:25 PM
#15 Multiple Degree of Freedom Detailed Overview

In general, this section will not be covered during the course Mechanical Vibrations I.
Instead, it is provided as a reference for those who wish to know a little more about
MDOF systems, but who for various reasons may not end up taking the two following
courses, Mechanical Vibrations II & III.
General Solution Approach

The matrix equation of motion for a general multi-degree-of-freedom system can be
written as (Time Domain):

| |{ } | |{ } { } { } ( ) ( ) [ ] ( ) ( ) M x t C x t K x t f t + + =

The solution of this linear, constant matrix coefficient, second order differential equation
follows the solution approach for the simpler single degree of freedom problem. The
solution takes on the following form involving complementary and particular parts:

{ } { } { }
( ) ( ) ( )
c p
x t x t x t = +

{ } ( )
c
x t is the complementary portion of the solution and depends on the system
characteristics and initial conditions. The complementary portion of the solution is
sometimes referred to as the transient portion of the solution.

{ }
( )
p
x t is the particular portion of the solution and depends upon the system
characteristics and harmonic forcing functions. The particular portion of the solution is
sometimes referred to as the steady state portion of the solution.

Frequently, one portion of the solution will be of interest due to the application under
study. If both portions are of interest, the initial conditions must not be applied until the
total solution is formed (both complementary and particular portions).

Complementary (Transient) Solution:

The complementary solution is found by transforming the original differential equation,
in homogeneous form (temporarily removing the forcing function(s)), from the time
domain to the frequency domain (transforming from differential to algebraic equations)
by using Fourier transforms, Laplace transforms or by assuming a solution that is
appropriate for the second order, linear, constant coefficient matrix differential equation.

The system of equations has a complementary solution of the assumed form:
Lecture Notes -74- 06/16/06 12:25 PM

{ } { } ( )
st
c
x t X e =
{ } { } { } ( ) ( )
st
c c
x t s X e s x t = =
{ } { } { }
2 2
( ) ( )
st
c c
x t s X e s x t = =

where:
s j = + = complex valued frequency

Note that the derivative of the response vector is simply the same vector multiplied by the
complex frequency.

Substituting the above assumed relationships into the matrix equation of motion:

| | { } | | { } { } { }
2
[ ] 0
st st st
M s X e C s X e K X e + + =

The non-trivial solution of the above equation leaves:

| | | | | | { } { }
2
0 s M s C K X ( + + =

The above equation can be solved in terms of 2N characteristic values and 2N
characteristic vectors. Due to the underdamped nature of vibration problems, there will
always be N pairs of characteristic values and N pairs of characteristic vectors. The
characteristic values are the complex-valued natural frequencies (modal frequencies)
r
and the characteristic vectors are the complex-valued modal vectors { }
r
. Since the
characteristic values and vectors are often found by placing the previous equation in an
eigenvalue form, the characteristic values are often referred to as eigenvalues and the
characteristic vectors as eigenvectors.

More commonly, the characteristic values and vectors are found by way of rudimentary
mathematic manipulations as follows:

The characteristic values can be found for the above algebraic equation, for non-trivial
solutions of { } X , from the matrix characteristic equation as follows:

| | | | | |
2
0 s M s C K + + =

The above matrix characteristic equation is of model order two (2) with coefficient
matrices of size N ( N N ). Therefore, this characteristic equation will yield 2N modal
frequencies.

Expanding the matrix characteristic equation completely yields a high order ( 2N )
polynomial characteristic equation with scalar coefficients:
Lecture Notes -75- 06/16/06 12:25 PM

2 2 1 2 2 2 1 0
2 2 1 2 2 2 1 0
0
N N N
N N N
s s s s s s

+ + + + + + =

The characteristic values (complex valued modal frequencies (
r r r
j = + )) are found
as the roots of this characteristic, high order ( 2N ), scalar polynomial.

Once the modal frequencies (
r
) have been determined, the characteristic vectors (modal
vectors { }
r
) can be found from the following relationship:

| | | | | | { } { }
2
0 s M s C K X ( + + =

Evaluating at
r
s = :

| | | | | | { } { }
2
0
r r r
M C K ( + + =

Note that this system of linear equations is always rank deficient by at least one since the
equation system is being evaluated at one of the characteristic frequencies (
r
).

This process must be repeated for each modal frequency to determine each modal vector.

Once the modal frequencies (complex valued, in general) and modal vectors (complex
valued, in general) are determined, the final form of the complementary solution can be
formulated as follows:

{ } { } { } ( )
*
* *
1
( )
r r
N
t t
c r r r r
r
x t e e

=
= +

Note that, in the above equation, the unknown coefficients
r
appear in complex
conjugate pairs. This will always be true in the underdamped case. It is not necessary to
assume that the conjugate relationship exists; this will result when the solution method is
followed.

If there is no forcing function (the particular solution is zero), the unknown (complex
valued) coefficients in the above equation (
r
) can be determined by applying the initial
conditions to the above equation and/or the derivative of the above equation. The
derivative of the above equation is shown below.

{ } { } { } ( )
*
* * *
1
( )
r r
N
t t
c r r r r r r
r
x t e e

=
= +

If there is a forcing function, the solution for the unknown (complex-valued) coefficients
must wait until the particular solution has been found. The initial conditions apply to the
Lecture Notes -76- 06/16/06 12:25 PM
complete solution and the unknown coefficients must be found after the particular
solution has been added to the complementary solution.

Particular Solution Approach

The particular solution is found by assuming a solution form for the response consistent
with the forcing function characteristic. Since the forcing function (steady-state) is some
form of harmonic (sine plus cosine terms), the forcing function and associated response
can always be put into the following form (use the Euler identity for sine and cosine).
This approach to solving for the particular solution is known as the method of
undetermined coefficients.

( ) cos( ) sin( )
a a
f t A t B t = +
( )
2 2
a a a a
j t j t j t j t
e e e e
f t A B
j

| | | | + +
= +
| |
\ . \ .

( )
2 2
a a a a
j t j t j t j t
e e e e
f t A jB

| | | | + +
=
| |
\ . \ .

Collecting like terms:

( )
2 2 2 2
a a
j t j t
A B A B
f t j e j e

| | | |
= + +
| |
\ . \ .

Therefore, representing the characterisitics of the force at
a
with a complex-valued
magnitude yields the general form for the forcing function:

{ } { } { }
*
( )
a a
j t j t
f t F e F e

= +

The response(s) to the previous forcing function(s) will be of the same form:

{ } { } { }
*
( )
a a
j t j t
p
x t X e X e

= +
{ } { } { }
*
( )
a a
j t j t
p a a
x t j X e j X e

=
{ } { } { }
2 2 *
( )
a a
j t j t
p a a
x t X e X e

=

Substituting the above relationships into the matrix equation of motion and collecting like
terms gives the following form for the positive frequency terms (
a
j t
e

). Note that the
portion of the solution involving the negative frequency terms (
a
j t
e

) is the complex
conjugate of the positive frequency and provides no new information.

Lecture Notes -77- 06/16/06 12:25 PM
| |{ } | | { } { }
2
[ ]
a a
M X j C K X F + + =

Note that the above linear, algebraic matrix equation involves N independent equations,
as long as the system is not undamped with the forcing frequency
a
equal to one of the
natural frequencies. Since the complex-valued forcing vector { } F provides N known
pieces of information, this system of equations can be solved for the complex-valued
response vector { } X .

Note also that, if more than one forcing frequency is present, the above particular solution
process must be repeated for each forcing frequency.

An alternative to the traditional method of undetermined coefficients is the frequency
response function approach. In this approach, the forcing function(s) are described, as
above, in the frequency domain. The frequency response function(s) between the forcing
degrees-of-freedom (DOFs) and the response degrees-of-freedom (DOFs) are computed
from:

| | | | | | | |
1
2
( )
a a a
H M j C K

( = + +

The responses caused by the forcing functions can now be found (in the frequency
domain) by:

{ } | |{ } ( ) ( ) ( )
a a a
X H F =

The particular response(s) can now be formulated in the time domain by converting the
frequency domain information back to the time domain using the Euler identities (sine
and cosine).

Note also that, if more than one forcing frequency is present, the above particular solution
process must be repeated for each forcing frequency.

Total Solution

The total solution, therefore, can be formulated as:

{ } { } { }
( ) ( ) ( )
c p
x t x t x t = +

{ } { } { } ( )
{ } { } ( )
*
* * *
1
( )
a a r r
N
j t j t t t
r r r r
r
x t e e X e X e

=
= + + +

Lecture Notes -78- 06/16/06 12:25 PM
Example Solution:

For the following three degree of freedom system, formulate the complete solution for the
following initial conditions and forcing function.

| |
10 0 0
0 14 0
0 0 12
M
(
(
=
(
(

| |
50 30 0
30 55 25
0 25 25
C
(
(
=
(
(

| |
5000 3000 0
3000 5500 2500
0 2500 2500
K
(
(
=
(
(

Initial Conditions: Forcing Function:

{ }
1.0
(0) 0
0
x

=
`

)
{ }
20.0
(0) 0
0
x

=
`

)
{ }
0
( ) 75 sin(30 )
0
f t t

=
`

)

Complementary Solution Results:

1
0.1848 6.0760 j = +
2
1.6417 18.0458 j = +
3
3.6795 26.8767 j = +

{ }
1
1.0000
1.5435
1.8763

=
`

)
{ }
2
1.0000
0.5722
0.9933

=
`

)
{ }
3
1.0000
0.7863
0.3105

=
`

)

Example Solution - Particular Part:

| | | | | | { } { }
2
( ) ( )
p
s M s C K X s F s ( + + =

Formulating the positive frequency portion of the solution at 30
rad
s
( 30 s j = ):

{ }
4000 1500 3000 900 0 0
3000 900 7100 1650 2500 750 (30 ) 37.5
0 2500 750 8300 750 0
p
j j
j j j X j j
j j
+ (

(
+ =
`
(

( +
)

{ }
0.0041 0.0011
(30 ) 0.0036 0.0045
0.0016 0.0009
p
j
X j j
j
+

=
`

+
)

Lecture Notes -79- 06/16/06 12:25 PM

Therefore:

{ }
30 30
0.0041 0.0011 0.0041 0.0011
( ) 0.0036 0.0045 0.0036 0.0045
0.0016 0.0009 0.0016 0.0009
j t j t
p
j j
x t j e j e
j j
+

= + +
` `

+
) )

Or:

{ }
0.0082 0.0022
( ) 0.0072 cos(30 ) 0.0090 sin(30 )
0.0031 0.0018
p
x t t t

= +
` `

) )

Example Solution - Complementary Part:

While the complementary solution of the matrix differential equation for linear vibration
problems always involves complex conjugate pair solutions, it is easier, when finding the
numerical solution using MATLAB, to ignore this and determine the solution without
this assumption. Once you find the solution, a check for complex conjugate pairs serves
as a verification that no other mistakes have occurred.

Therefore:

{ } { } { }
( ) ( ) ( )
c p
x t x t x t = +

{ } { }
6
30 30
1
0.0041 0.0011 0.0041 0.0011
( ) 0.0036 0.0045 0.0036 0.0045
0.0016 0.0009 0.0016 0.0009
r
t j t j t
r r
r
j j
x t e j e j e
j j

=
+

= + + +
` `

+
) )

Rewriting in matrix notation:

Lecture Notes -80- 06/16/06 12:25 PM
{ } { } { } { } { } { } { }
3 5 6 1 2 4
1
2
3
1 2 3 4 5 6
4
5
6
30 30
( )
0.0041 0.0011 0.0041 0.0011
0.0036 0.0045 0.0036 0.0045
0.0016 0.0009 0.0016 0.0009
t t t t t t
j t j t
x t e e e e e e
j j
j e j e
j j

( =
`

)
+

+ + +
` `

+
) )

Taking the time derivative to get velocity:

{ } { } { } { } { } { } { }
3 5 6 1 2 4
1
2
3
1 2 3 4 5 6
1 2 3 4 5 6
4
5
6
30
( )
0.0041 0.0011 0.0041 0.0011
30 0.0036 0.0045 30 0.0036 0.0045
0.0016 0.0009 0.0016 0.000
t t t t t t
j t
x t e e e e e e
j j
j j e j j
j

( =
`

)
+

+ +
`

+
)
30
9
j t
e
j

`

)

Evaluating displacement and velocity at t=0:

{ } { } { } { } { } { }
1
2
3
1 2 3 4 5 6
4
5
6
1.0 0.0082
0 0.0071
0 0.0031

( = +
` ` `

) )

)

{ } { } { } { } { } { }
1
2
3
1 2 3 4 5 6
1 2 3 4 5 6
4
5
6
20.0 0.0639
0 0.2722
0 0.0542

( = +
` ` `

) )

)

Lecture Notes -81- 06/16/06 12:25 PM

| |
1
2
3
4
5
6
1.0082
0.0071
0.0031
20.0639
0.2722
0.0542
A

=
` `

) )

Where:
| |
1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
0.7863 0.5772 1.5435 1.5435 0.5772 0.7863
0.3105 0.9933 1.8763 1.8763 0.9933 0.3105
3.6795 26.8767 1.6417 18.0458 0.1848 6.0760 0.1848 6.0760 1.6417 18.0458 3.6795
A
j j j j j

=
+ + + 26.8767
2.8932 21.1335 0.9394 10.3255 0.2852 9.3783 0.2852 9.3783 0.9394 10.3255 2.8932 21.1335
1.1425 8.3456 1.6307 17.9246 0.3467 11.4003 0.3467 11.4003 1.6307 17.9246 1.1425 8.3456
j
j j j j j j
j j j j j j
+ + +
+ + +
(
(
(
(
(
(
(
(

Therefore:

1
2
3
4
5
6
0.2567 0.2265
0.1890 0.2246
0.0584 0.1902
0.0584 0.1902
0.1890 0.2246
0.2567 0.2265
j
j
j
j
j
j
+

+

+

=
` `

) )

Complementary Solution Approach - Adjoint Matrix

An alternative approach to determining the modal vectors in the complementary solution
involves evaluating the adjoint matrix of the system impedance matrix at the modal
frequencies (
r
). This method has two advantages: 1) Formulating the adjoint matrix
effectively solves the set of linear equations one time rather than one set of linear
equations for each modal frequency; 2) Evaluating the adjoint matrix handles the rank
deficient problem in a uniform way (no assumption of 1.0 for a modal coefficient). The
development is as follows:

| | | | | | { } { }
2
0 s M s C K X ( + + =

Define:

Lecture Notes -82- 06/16/06 12:25 PM
| | | | | | | |
2
( ) B s s M s C K ( = + +

where:

| |
( ) B s = System Impedance Matrix

| || | | |
1
( ) ( ) B s B s I
=

| |
| |
| |
1
( )
( )
( )
A
B s
B s
B s
=

where:
| | ( )
A
B s is the adjoint of matrix | | ( ) B s .

| || | | || | ( ) ( ) ( )
A
B s B s B s I =

Note:
| | ( ) 0
r
B = .

Evaluating at
r
s = gives:

| || | | | ( ) ( ) 0
A
r r
B B =

Using any column of
| | ( )
A
r
B , the
th
i column for example { } ( )
A
r
i
B . Therefore:

| |{ } { } ( ) ( ) 0
A
r r
i
B B =
| | | | | | { } { }
2
0
r r r
M C K ( + + =

Note that the columns of the adjoint matrix
| | ( )
A
r
B are all proportional to the
th
r modal
vector.

When the mass, damping and stiffness matrices are symmetric (when absolute
coordinates are used), the system impedance matrix
| | ( ) B s is symmetric. Therefore, in
this case, the adjoint matrix of
| |
( )
r
B is also symmetric. Thus, in this case, the rows of
the adjoint matrix are also proportional to the modal vector.

Lecture Notes -83- 06/16/06 12:25 PM
Complementary Solution Approach - Eigenvalue/Eigenvector
Method

The homogeneous form of the Laplace domain model can be used as a general
representation of the matrix relationship that yields the system modal characterisitics:

| |{ } | |{ } | |{ } { }
2
0 s M X s C X K X + + =

Generally, this problem is solved using eigenvalue-eigenvector solution methods once the
problem is put in the standard eigenvalue form:

| | | | { } { } 0 A I X ( =

| |{ } { } A X X =
| |{ } | |{ } A X B X =

Undamped Case:

In order to manipulate the system equations into a standard eigenvalue/eigenvector
equation form, one approach is to assume that the undamped case is a reasonable
approximation of the damped case. For many lightly damped situations, this is a
reasonable assumption.

| |{ } | |{ }
2
K X s M X =
where:

2
s =

General Damped Case

| |{ } | |{ } | |{ } { } M x C x K x f + + =

This system of equations can be augmented by the identity shown as follows:

| |{ } | |{ } { } 0 M x M x =

The above two equations can be combined to yield a new system of 2N equations.
Note that all the matrices in the resulting equation are symmetric and the equation is now
in a classical eigenvalue solution form. The notation used in the following equation is
consistent with the notation used in many mathmatics and/or controls textbooks.
Lecture Notes -84- 06/16/06 12:25 PM

| |{ } | |{ } { } A y B y f + =

where:
| |
| | | |
| | | |
0 M
A
M C
(
=
(

| |
| | | |
| | | |
0
0
M
B
K
(
=
(

{ }
{ }
{ }
x
y
x

=
`

)
{ }
{ }
{ }
x
y
x

=
`

)

{ }
{ }
{ }
0
f
f

=
`

)

Forming the homogeneous equation of this system equation:

| |{ } | |{ } { } 0 A y B y + =

The solution of the above equation yields the complex-valued natural frequencies
(eigenvalues) and complex-valued modal vectors (eigenvectors) for the augmented
2N equation system. Note that in this mathematical form, the eigenvalues will be found
directly (not the square of the eigenvalue) and the 2N eigenvectors will be 2N in
length. The exact form of the eigenvectors can be seen from the associated modal matrix
for the 2N equation system. Note that the notation { } is used for an eigenvector in the
2N equation system and that the notation { } is used for an eigenvector of the N
equation system.

MATLAB Solution

For the general, homogeneous system equation, using either an assumed solution or
transform methods, the following equation must be solved:

| |{ } | |{ } { } 0 A y B y + =

MATLAB uses the same matrix terminology but refers to a different eigenvalue equation.

{ } { } A y B y ( ( =

Therefore, using the MATLAB EIG function to solve for the eigenvalues and
eigenvectors requires the following form:

| | ( ) ( ) , , , y EIG A B EIG B A = =

Lecture Notes -85- 06/16/06 12:25 PM

The eigenvalues of this system of equations are the same as for the original mass,
stiffness and damping matrix equation. The eigenvectors of this system of equations yield
the modal vectors of the original mass, stiffness and damping matrix equation through the
modal matrix. The modal matrix for this system is a matrix made up of the 2N
eigenvectors. The modal matrix { } for this damped system can now be assembled.

| | { } { } { } { }
1 2 2 r N
( =

| |
{ } { } { } { }
{ } { } { } { }
1 2 2
1 2 2
1 2 2
r N
r N
r N

(
=
(

Weighted Orthogonality Concept
Proportionally Damped Case

A set of weighted orthogonality relationships are valid for the system matrices
| |
M and
| | K .

{ } | |{ } 0
T
r s
M =
{ } | |{ } 0
T
r s
K =

Modal Scaling - Proportionally Damped Case

{ } | |{ }
T
r
r r
M M =
{ } | |{ }
T
r
r r
K K =
{ } | |{ }
T
r
r r
C C =

General Case

A set of weighted orthogonality relationships are valid for the system matrices
| | A and
| |
B .

{ } | |{ } 0
T
r s
A =
{ } | |{ } 0
T
r s
B =
Lecture Notes -86- 06/16/06 12:25 PM

Modal Scaling - General Case

{ } | |{ }
r
T
A
r r
A M =
{ } | |{ }
r
T
B
r r
B M =

The terms modal A and modal B are modal scaling factors for the general case of system
with damping. Note that modal A and modal B are related by the complex modal
frequency for each mode.

Vibs1 Review Notes PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vibs1 Review Notes PDF

Uploaded by

Copyright:

Available Formats

Mechanical Vibrations I

Lecture Notes -1- 06/16/06 12:25 PM

. The general homogeneous solution then takes on the following form.

damping, and 1000

= , it is possible to estimate the system pole

, is the damped natural frequency (

= , must equal zero. Since only the second

1 F = ) and the duration of the pulse ( ) is allowed to approach

when 0 < <

, into the above expression yields,

You might also like