Young Engineer IMAC21 2003 Assembled MACL Tcm18-189941

Presentation Topics
Dr. Peter Avitabile

Modal Analysis & Controls Laboratory
Intent
Modal Overview University of Massachusetts Lowell

SDOF Theory
MDOF Theory
IMAC 19
Young Engineer Program
Measurement Definitions
Excitation Considerations
MPE Concepts
Linear Algebra
TUTORIAL:
Basics of Modal Analysis
Dr. Peter Avitabile peter_avitabile@uml.edu Friday, February 07, 2003 1
Intent of Young Engineer Program
The intent of the Young Engineer Program is to expose the new

or young engineer to some of the basic concepts and ideas
concerning analytical and experimental modal analysis.
It is NOT intended to be a detailed treatment of this material.
Rather it is intended to prepare one for some of the in-depth

papers presented at IMAC so that the novice has some
appreciation of the detailed material presented in these papers.
This presentation is intended to identify the basic methodology

and techniques currently employed in this field and to expose
one to the typical modal jargon used in the field.
Intent of Young Engineer Program 1 Dr. Peter Avitabile

Experimental Modal Analysis
A Simple Non-Mathematical Presentation

Dr. Peter Avitabile
Mechanical Engineering - UMASS Lowell
Could you explain and how is it

modal analysis used for solving
dynamic problems?
Illustration by Mike Avitabile Illustration by Mike Avitabile Illustration by Mike Avitabile
Experimental Modal Analysis 1 Dr. Peter Avitabile

Analytical Modal Analysis
Equation of motion [M n ]{&x& n } + [C n ]{x& n } + [K n ]{x n } = {Fn ( t )}
Eigensolution [[K n ] − λ[M n ]]{x n } = {0}

Could you explain modal analysis for me ?
Simple time-frequency response relationship
RESPONSE
increasing rate of oscillation
FORCE
time
frequency

Could you explain modal analysis for me ?
Sine Dwell to Obtain Mode Shape Characteristics
MODE3
MODE 1
MODE 2
MODE 4

Just what are the measurements called FRFs ?
A simple input-
output problem
Magnitude Real
8
1 2 3
0
-3
8
-7
1.0000
-1.0000
Phase Imaginary
Just what are the measurements called FRFs ?
Response at point 3
due to an input at point 3 1 2 3
1 2 3 2
1
h32
1 2 3
3
1
Drive
3
Point
FRF
h33
h31

Why is only one row/column of FRFs needed ?
The third row of the FRF matrix - mode 1
The peak amplitude of the imaginary part of the

FRF is a simple method to determine the mode shape
of the system

The second row of the FRF matrix is similar
The peak amplitude of the imaginary part of the

FRF is a simple method to determine the mode shape
of the system

Any row or
column can be
used to extract
mode shapes
- as long as it is
not the node of
a mode !
? ?

More measurements better defines the shape
MODE # 1
MODE # 2
MODE # 3
DOF # 1
DOF #2
DOF # 3

What’s the difference between shaker and impact ?
1 2 3
h 13
1 2 3 1
3
2
h 23
3
h 33
h 31
h 33 h 32
Theoretically - - - NOTHING ! ! !

What measurements do I actually make ?
Actual time signals

ANALOG SIGNALS
INPUT OUTPUT
ANTIALIASING FILTERS
AUTORANGE ANALYZER
Analog anti-alias filter
ADC DIGITIZES SIGNALS
INPUT OUTPUT
Digitized time signals

APPLY WINDOWS
INPUT
OUTPUT
Windowed time signals
COMPUTE FFT
Compute FFT of signal

LINEAR SPECTRA
LINEAR LINEAR
INPUT OUTPUT
SPECTRUM SPECTRUM
AVERAGING OF SAMPLES
COMPUTATION OF AVERAGED
INPUT/OUTPUT/CROSS POWER SPECTRA
INPUT
POWER
SPECTRUM
CROSS
POWER
SPECTRUM
OUTPUT
POWER
SPECTRUM
Average auto/cross spectra
COMPUTATION OF FRF AND COHERENCE
FREQUENCY RESPONSE FUNCTION COHERENCE FUNCTION

Compute FRF and Coherence

What’s most important in impact testing ?
Hammers and Tips

40
COHERENCE
dB Mag
FRF
INPUT POWER SPECTRUM
-60
0Hz 800Hz
40 COHERENCE
FRF
dB Mag
-60
0Hz 200Hz

What’s most important in impact testing ?
Leakage and Windows
ACTUAL TIME SIGNAL
SAMPLED SIGNAL
WINDOW WEIGHTING
WINDOWED TIME SIGNAL

What’s most important in shaker testing ?
AUTORANGING AVERAGING WITH WINDOW
Random
with
1 2 3 4
Hanning
Different
excitation
AUTORANGING AVERAGING
techniques are
Burst
Random
available for
1 2 3 4
obtaining good
measurements
Sine
Chirp
1 2 3 4

How do I get mode shapes from the FRFs ?
MODE 2
1 4
3 6
MODE 1
5
4
1
3
6

The FRF is made up

from each individual
mode contribution
which is determined
a ij1
from the
ζ
frequency,
1
a ij2 ζ
2
a ij3 ζ3
damping,
ω
1
ω2 ω3
residue

MDOF
SDOF
The task for the modal test engineer is to

determine the parameters that make up the pieces
of the frequency response function

HOW MANY POINTS ???
Mathematical
routines help to
determine the
RESIDUAL
EFFECTS
basic parameters
that make up
RESIDUAL
EFFECTS
the FRF
HOW MANY MODES ???

What is operating data ?
Why and How Do Structures Vibrate?
INPUT TIME FORCE

f(t) y(t)
FFT IFT
INPUT SPECTRUM OUTPUT SPECTRUM
f(j ω) h(j ω) y(j ω)

If I make measurements on a structure at an
operating frequency, sometimes I get some
deformation shapes that look pretty funky .
Maybe they are just noise?
Is that possible ???

But if I make a measurement at an operating
frequency and its close to a mode, I can easily
see what appears to be one of the modes
MODE 1 CONTRIBUTION MODE 2 CONTRIBUTION

And if I make a measurement at an operating
frequency and its close to another mode, I can
easily see what appears to be one of the modes

I think I just answered my own question !!!
I think I’m starting to understand this now !!!

The modes of the structure act like filters
which amplify and attenuate input excitations
on a frequency basis
OUTPUT SPECTRUM
y(j ω)
f(j ω )
INPUT SPECTRUM

So what good is modal analysis ?
EXPERIMENTAL FINITE
MODAL ELEMENT
TESTING MODELING
MODAL PERFORM
The dynamic
model can be
PARAMETER EIGEN
ESTIMATION SOLUTION
DEVELOP MASS RIB used for studies

to determine the
MODAL STIFFNER
MODEL
Repeat
SPRING effect of
structural
until STRUCTURAL
desired
characteristics CHANGES No
are REQUIRED DASHPOT
changes of the
obtained
DONE
Yes
USE SDM
TO EVALUATE
mass, damping
STRUCTURAL
CHANGES STRUCTURAL and stiffness
DYNAMIC
MODIFICATIONS

So what good is modal analysis ?
Simulation, Prediction, Correlation, … to name a few
FREQUENCY FINITE
RESPONSE CORRECTIONS ELEMENT
MEASUREMENTS MODEL
PARAMETER EIGENVALUE
ESTIMATION SOLVER
MODAL MODEL MODAL

PARAMETERS VALIDATION PARAMETERS
SYNTHESIS
OF A
DYNAMIC MODAL MODEL
STRUCTURAL FORCED
MASS, DAMPING, REAL WORLD
DYNAMICS RESPONSE
STIFFNESS CHANGES FORCES
MODIFICATION SIMULATION
MODIFIED
STRUCTURAL
MODAL
RESPONSE
DATA

Single Degree of Freedom Overview
x(t) f(t)
m
k c
100 ζ=0.1% T = 2 π/ω n

ζ=1%
X1
ζ=2%
X2
ζ=5%
10
ζ=10%
ζ=20% 0
-90 ζ=20%
ω/ω n ζ=10%
ζ=5% t1 t2
ζ=2%
1
ζ=1%
ζ=0.1%
-180
h (s) =
ms 2 + cs + k
ω/ω n
SDOF Overview 1 Dr. Peter Avitabile

SDOF Definitions
Assumptions
• lumped mass
• stiffness proportional
x(t)
to displacement
m
• damping proportional to
velocity
k c
• linear time invariant
• 2nd order differential
equations

SDOF Equations
Equation of Motion
d2x dx
m 2 + c + kx = f ( t ) or m &x& + cx& + kx = f ( t )
dt dt
Characteristic Equation
ms 2 + cs + k = 0
Roots or poles of the characteristic equation

2
c  c  + k
s1, 2 =− ±  
2m  2m  m

SDOF Definitions
Poles expressed as jω
s1, 2 = − ζω n ± (ζωn )2 −ωn 2 = − σ ± jωd

POLE
ωd
Damping Factor σ = ζωn
Natural Frequency ωn = k
m
σ
ζωn
% Critical Damping ζ= c
cc
Critical Damping c c = 2mωn
Damped Natural CONJUGATE
ωd = ωn 1−ζ 2
Frequency

SDOF - Harmonic Excitation
100 ζ=0.1% 0
ζ=1%
ζ=2%
ζ=5%
10
ζ=10%
ζ=20% -90 ζ=20%

ζ=10%
ζ=5%
1
ζ=2%
ζ=1%
ζ=0.1%
-180
ω/ωn
ω/ω n
x 1  2ζβ 
= φ=tan −1
δst (1 − β 2 )2 +(2ζβ )2 2
1 − β 

SDOF - Damping Approximations
MAG
T = 2 π/ω n
X1
X2
0.707
MAG
t1 t2
ω1 ωn ω2
1 ωn x1
Q= = δ = ln ≈ 2πζ
2ζ ω2 − ω1 x2

SDOF - Laplace Domain
Equation of Motion in Laplace Domain

(ms 2 +cs+k)x (s) = f (s) with b(s ) = (ms 2 +cs+k)
System Characteristic Equation

b(s) x (s) = f (s) and x (s) = b −1 (s)f (s) = h (s)f (s)
System Transfer Function

1
h (s) =
ms 2 + cs + k

SDOF - Transfer Function
Polynomial Form 1
h (s) =
ms 2 + cs + k
1/ m
Pole-Zero Form h (s) =
(s − p1 )(s − p1* )
a1 a1*
Partial Fraction Form h (s) = +
(s − p1 ) (s − p1* )
1 −ζωt
Exponential Form h(t) = e sin ωd t
mωd

SDOF - Transfer Function & Residues
Residue
a1 =
h (s)(s − p1 ) s→p1
1
=
2 jmωd
related to
Source: Vibrant Technology
mode shapes

SDOF - Frequency Response Function
a1 a1*
h ( jω) = h (s) s = jω
= +
( jω − p1 ) ( jω − p1* )

Bode Plot Coincident-Quadrature Plot
0.707 MAG
Nyquist Plot

DYNAMIC COMPLIANCE DISPLACEMENT / FORCE
MOBILITY VELOCITY / FORCE
INERTANCE ACCELERATION / FORCE
DYNAMIC STIFFNESS FORCE / DISPLACEMENT
MECHANICAL IMPEDANCE FORCE / VELOCITY
DYNAMIC MASS FORCE / ACCELERATION

Multiple Degree of Freedom Overview
[B(s )]−1 = [H(s )] = Adj[B(s )] = [A(s )]
det[B(s )] det[B(s )]
p1 f1 p2 f2 p3 f3
m1 m2 m3
k1 c1 k2 c2 k3 c3 R1
D1
MODE 1 MODE 2 MODE 3

R2
D2
R3
\  \  \  D3
 M {&p&} +  C {p& } +  K {p} = [U ]T {F}

      F1 F2F3
 \   \   \ 
MDOF Overview 1 Dr. Peter Avitabile

MDOF Definitions
Assumptions
f2 x2
• lumped mass m2
• stiffness proportional
k2 c2
to displacement
• damping proportional to f1 x1
m1
velocity
• linear time invariant
k1 c1
• 2nd order differential
equations

MDOF Equations
Equation of Motion - Force Balance

m1&x&1+(c1 + c 2 )x& 1−c 2 x& 2 +(k1 + k 2 )x1−k 2 x 2 =f1 (t )
m 2 &x& 2 −c 2 x& 1+c 2 x& 2 −k 2 x1 +k 2 x 2 =f 2 (t )
Matrix Formulation
m1   &x&1  Matrices and
    Linear Algebra
 m 2  &x& 2 
are important !!!
(c1 + c 2 ) − c 2   x& 1 
+  x& 
 − c 2 c 2  2 
(k1 + k 2 ) − k 2   x1   f1 ( t ) 
+   = 
 − k 2 k 2  2 x f
2  ( t )

MDOF Equations
Equation of Motion
[M ]{&x&}+[C]{x& }+[K ]{x}={F( t )}
Eigensolution
[[K ]−λ[M ]]{x}=0
Frequencies (eigenvalues) and
Mode Shapes (eigenvectors)
\   ω1
2

 = 

Ω2
 
ω22  and [U ] = [{u1} {u 2 } L]
 \   \ 

Modal Space Transformation
Modal transformation
 p1 
 
{x} = [U ]{p} = [{u1} {u 2 } L]p 2 
M
 
Projection operation
[U ]T [M ][U ]{&p&} + [U ]T [C][U ]{p& } + [U ]T [K ][U ]{p} = [U ]T {F}
Modal equations (uncoupled)
 p1  {u1} {F}
T
 m1  &p&1   c1  p& 1   k1
 m2 &p& + c2 p& + k2 p ={u }T {F}
  2    2    2   2 
 \  M   \  M   \  M   M 


Diagonal Matrices -
Modal Mass Modal Damping Modal Stiffness
\  \  \ 
 M {&p&} +  C {p& } +  K {p} = [U ]T {F}
     
 \   \   \ 
Highly coupled system
p1 f1 p2 f2 p3 f3
transformed into
m1 m2 m3
k1 c1 k2 c2 k3 c3
simple system MODE 1 MODE 2 MODE 3

.. .
[M]{x} + [C]{x} + [K]{x} = {F(t)}
= Σ
p1 f1
m1
{x} = [U]{p} = [{u } {u } {u } ]{p} k1 c1

1 2 3
MODE 1
+ p2 f2
MODAL SPACE m2
k2 c2
.. . T +
MODE 2
[ M ]{p} + [ C ]{p} + [ K ]{p} = [U] {F(t)} p3 f3

m3
{u }p + {u }p + {u }p k3 c3
1 1 2 2 3 3
MODE 3

MDOF - Laplace Domain
Laplace Domain Equation of Motion

[[M]s 2
+ [C]s + [K ]]{x (s )} = 0 ⇒ [B(s )]{x (s )} = 0
System Characteristic (Homogeneous) Equation

[[M]s +[C]s+[K ]] = 0
2
⇒ p k = − σ k ± jωdk
Damping Frequency

MDOF - Transfer Function
System Equation
[B(s )]{x (s )} = {F(s )} ⇒ [H(s )] = [B(s )] ={x (s )}
−1
{F(s )}
System Transfer Function
Adj[B(s )] [A(s )]
[B(s )]
−1
= [H(s )] = =
det[B(s )] det[B(s )]
[A(s )] Residue Matrix Mode Shapes
det[B(s )] Characteristic Equation Poles

MDOF - Residue Matrix and Mode Shapes
Transfer Function evaluated at one pole

qk
[H(s )]s=s = {u k } {u k }T
s−p k
k
can be expanded for all modes

q k {u k }{u k } q k {u }{u } * T
m T *
[H(s )] = ∑ + k k
k =1 (s−p k ) (s−p*k )

MDOF - Residue Matrix and Mode Shapes
Residues are related to mode shapes as
[A(s )]k = q k {u k }{u k }

T
 a11k a12 k a13k L  u1k u1k u1k u 2 k u1k u 3k L

a a 22 k a 23k L u u u 2k u 2k u 2 k u 3k L
 21k =q k  2 k 1k 
a 31k a 32 k a 33k L  u 3k u1k u 3k u 2 k u 3k u 3k L
 M O  M O
 M M  M M

MDOF - Drive Point FRF
a ij1 a *ij1
h ij ( jω ) = +
( jω − p 1 ) ( jω − p*1 )
a ij 2 a *ij 2
+ +
( jω − p 2 ) ( jω − p*2 )
a ij 3 a *ij 3
+ +
( jω − p 3 ) ( jω − p*3 )
*
q1 u i 1 u j 1 q1 u i 1 u j 1
h ij ( jω ) = +
( jω − p 1 ) ( jω − p*1 )
*
q 2u i 2u j 2 q 2u i 2u j 2
+ +
( jω − p 2 ) ( jω − p*2 )
*
q 3u i 3u j 3 q 3u i 3u j 3
+ +
( jω − p 3 ) ( jω − p*3 )

MDOF - FRF using Residues or Mode Shapes
a ij1 a*ij1
h ij ( jω ) = +
( jω − p1 ) ( jω − p*1 )
R1
a ij 2 a*ij 2
D1 + + + L
( jω − p 2 ) ( jω − p*2 )
R2
D2
R3
D3
F1 F2F3
a ij1
ζ1
a ij2 ζ2
* * *
q1u i1u j1 qu u
1 i 1 j1
a ij3 ζ3
h ij ( jω ) = +
( jω − p1 ) ( jω − p ) *
1
ω1 ω2 ω3
* * *
q 2u i 2u j 2 qu u
2 i2 j2
+ + + L
( jω − p 2 ) ( jω − p ) *
2

Time / Frequency / Modal Representation
PHYSICAL TIME FREQUENCY
ANALYTICAL
MODAL
p1 f1
m1
k1 c1
MODE 1
MODE 1 + + +
p2 f2
m2
k2 c2
MODE 2 MODE 2
+ + p3
+ f3
m3
k3 c3
MODE 3
MODE 3

Overview Analytical and Experimental Modal Analysis
TRANSFER LAPLACE
FUNCTION DOMAIN
[B(s)] = [M]s2 + [C]s + [K]
[B(s)] -1 = [H(s)]
qk u j {u k} [U] [ A(s) ]
det [B(s)]
ω ζ FINITE ANALYTICAL
[U] ELEMENT MODEL
ω MODEL REDUCTION
T
A
MODAL [K - λ M]{X} = 0 N
PARAMETER
ESTIMATION H(j ω)
LARGE DOF
MISMATCH
N U A
X j(t)
Xj (j ω ) FFT
H(j ω) =
Fi (j ω) MODAL
Fi (t)
TEST
EXPERIMENTAL
MODAL MODEL
EXPANSION

INPUT SYSTEM OUTPUT
u(t) v(t)
H ACTUAL
n(t) m(t)
Σ Σ NOISE
x(t) y(t) MEASURED
-10
-20
-30
1.0000 -40
-50
-60
-1.0000
dB
-70
- 80
- 90
-100
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 15.9375
Measurement Definitions 1 Dr. Peter Avitabile

Actual time signals

ANALOG SIGNALS
INPUT OUTPUT
ANTIALIASING FILTERS
AUTORANGE ANALYZER
Analog anti-alias filter
ADC DIGITIZES SIGNALS
INPUT OUTPUT
Digitized time signals

APPLY WINDOWS
INPUT
OUTPUT
Windowed time signals
COMPUTE FFT
Compute FFT of signal

LINEAR SPECTRA
LINEAR LINEAR
INPUT OUTPUT
SPECTRUM SPECTRUM
AVERAGING OF SAMPLES
COMPUTATION OF AVERAGED
INPUT/OUTPUT/CROSS POWER SPECTRA
INPUT
POWER
SPECTRUM
CROSS
POWER
SPECTRUM
OUTPUT
POWER
SPECTRUM
Average auto/cross spectra
COMPUTATION OF FRF AND COHERENCE
FREQUENCY RESPONSE FUNCTION COHERENCE FUNCTION

Compute FRF and Coherence

Leakage
F
ACTUAL
DATA
CAPTURED R
E
DATA
T Periodic Signal
T
RECONTRUCTED
Q
I
DATA
M U
Non-Periodic Signal E
ACTUAL
E
DATA
CAPTURED N
C
DATA
Y
T
RECONTRUCTED
DATA
Leakage due to
T
signal distortion

Windows
0
-10 Time weighting functions

-20
are applied to minimize
-30
the effects of leakage
-40
Rectangular
-50
-60
dB
-70 Hanning
- 80
Flat Top
and many others
- 90
-100
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 15.9375
Windows DO NOT eliminate leakage !!!

Windows
Special windows are

used for impact testing
Force window
Exponential Window

Measurements - Linear Spectra
x(t) h(t) y(t) TIME
INPUT SYSTEM OUTPUT FFT & IFT

Sx(f) H(f) Sy(f) FREQUENCY
x(t) - time domain input to the system

y(t) - time domain output to the system
Sx(f) - linear Fourier spectrum of x(t)
Sy(f) - linear Fourier spectrum of y(t)
H(f) - system transfer function

h(t) - system impulse response

+∞ +∞
x ( t )= ∫ Sx (f )e j2 πft df Sx (f )= ∫ x ( t )e − j2 πft dt
−∞ −∞
+∞ +∞
y( t )= ∫ S y (f )e j2 πft
df S y (f )= ∫ y( t )e − j2 πft dt
−∞ −∞
+∞ +∞
h ( t )= ∫ H(f )e j2 πft
df H(f )= ∫ h ( t )e − j2 πft dt
−∞ −∞
Note: Sx and Sy are complex valued functions

Measurements - Power Spectra
Rxx(t) Ryx(t) Ryy(t) TIME
INPUT SYSTEM OUTPUT FFT & IFT

Gxx(f) Gxy(f) Gyy(f) FREQUENCY
Rxx(t) - autocorrelation of the input signal x(t)
Ryy(t) - autocorrelation of the output signal y(t)
Ryx(t) - cross correlation of y(t) and x(t)
Gxx(f) - autopower spectrum of x(t) G xx ( f ) = S x ( f ) • S*x ( f )
G yy(f) - autopower spectrum of y(t) G yy ( f ) = S y ( f ) • S*y ( f )
G yx(f) - cross power spectrum of y(t) and x(t) G yx ( f ) = S y ( f ) • S*x ( f )

lim 1
R xx (τ)=E[ x ( t ), x ( t + τ)]=
T → ∞TT∫ x ( t )x ( t + τ)dt
+∞
G xx (f )= ∫ R xx (τ)e − j2 πft dτ=Sx (f )•S*x (f )
−∞
lim 1
R yy (τ)=E[ y( t ), y( t + τ)]=
T → ∞TT∫ y( t )y( t + τ)dt
+∞
G yy (f )= ∫ R yy (τ)e − j2 πft dτ=S y (f )•S*y (f )
−∞
lim 1
R yx (τ)=E[ y( t ), x ( t + τ)]=
T → ∞TT∫ y( t )x ( t + τ)dt
+∞
G yx (f )= ∫ R yx (τ)e − j2 πft dτ=S y (f )•S*x (f )
−∞
Measurements - Derived Relationships
S y =HSx
H1 formulation
- susceptible to noise on the input
- underestimates the actual H of the system
S y •S*x G yx
S y •S*x =HSx •S*x H= =
Sx •Sx G xx
*
Other
H2 formulation formulations
- susceptible to noise on the output
- overestimates the actual H of the system for H exist
S y •S*y G yy
S y •S*y =HSx •S*y H= =
Sx •S y G xy
*
COHERENCE
(S y •S*x )(Sx •S*y ) G yx / G xx H1
γ 2
xy = = =
(Sx •S*x )(S y •S*y ) G yy / G xy H2
Measurements - Noise
H=G uv /G uu
INPUT SYSTEM OUTPUT
  u(t) v(t)
ACTUAL
H
 1 
H1 =H 
1+G nn  n(t) m(t)
 G uu  Σ Σ NOISE
 G mm 
H 2 =H1+ 
x(t) y(t) MEASURED
 G vv 

Measurements - Auto Power Spectrum
x(t)
INPUT FORCE OUTPUT RESPONSE
G xx (f) yy
AVERAGED INPUT AVERAGED OUTPUT
POWER SPECTRUM POWER SPECTRUM

Measurements - Cross Power Spectrum
AVERAGED INPUT AVERAGED OUTPUT
POWER SPECTRUM POWER SPECTRUM
G xx (f) G yy (f)
AVERAGED CROSS
POWER SPECTRUM
G yx (f)
Measurements - Frequency Response Function
AVERAGED INPUT AVERAGED CROSS AVERAGED OUTPUT
POWER SPECTRUM POWER SPECTRUM POWER SPECTRUM
G xx (f) G yx (f) G yy (f)
FREQUENCY RESPONSE FUNCTION
H(f)

Measurements - FRF & Coherence
Coherence
1
Real
0Hz AVG: 5 200Hz
COHERENCE
Freq Resp
40
dB Mag
-60
0Hz AVG: 5 200Hz
FREQUENCY RESPONSE FUNCTION

Excitation Considerations
1 2 3
1
h 13
1 2 3
2
1
3
2
h 23
3
h 33
h 31
h 32 h 33
Excitation Considerations 1 Dr. Peter Avitabile

Excitation Considerations - Impact
The force spectrum can be customized to some extent
through the use of hammer tips with various hardnesses.
CH1 Time
CH1 Time
7
1.8
Real
Real
-3
-200
-976.5625us 123.9624ms
-976.5625us 123.9624ms
CH1 Pwr Spec
CH1 Pwr Spec
-20
-10
dB Mag
dB Mag
-70
-110
0Hz 6.4kHz
0Hz 6.4kHz
CH1 Time
CH1 Time
3.5
8
Real
Real
-1.5
-2
-976.5625us 123.9624ms
-976.5625us 123.9624ms
CH1 Pwr Spec
CH1 Pwr Spec
-20
-10
dB Mag
dB Mag
-120
-110
0Hz 6.4kHz
0Hz 6.4kHz

Excitation Considerations - Impact/Exponential
The excitation must be sufficient to excite all the modes
of interest over the desired frequency range.
40
COHERENCE
dB Mag
FRF
-60
0Hz 800Hz
40 COHERENCE
FRF
dB Mag
-60
0Hz 200Hz

Excitation Considerations - Impact/Exponential
The response due to impact excitation may need an
exponential window if leakage is a concern.
ACTUAL TIME SIGNAL
SAMPLED SIGNAL
WINDOW WEIGHTING
WINDOWED TIME SIGNAL

Excitation Considerations - Shaker Excitation
AUTORANGING AVERAGING WITH WINDOW
Random Leakage is a serious concern

with
Hanning
Accurate FRFs are necessary
1 2 3 4
Burst
Random Special excitation
techniques can be
1 2 3 4
AUTORANGING AVERAGING used which will result

Sine in leakage free
Chirp measurements without
1 2 3 4
the use of a window
as well as other techniques

Excitation Considerations - MIMO
Multiple referenced FRFs are

obtained from MIMO test
Energy is distributed
better throughout the
structure making
better measurements
possible
Ref#2 Ref#3
Ref#1

Large or
complicated
structures
require
special
attention

[G XF ]=[H][G FF ] Measurements are

developed in a
 H11 H12 L H1, Ni similar fashion to
H  the single input
H 22 L H 2, Ni
[H]= 21  single output case
 M M M  but using a matrix
H H No , 2 L H No, Ni  formulation
 No,1
where
[H ]=[G XF ][G FF ]−1 No - number of outputs
Ni - number of inputs

Measurements on the same structure can show

tremendously different modal densities depending
on the location of the measurement
Source: Michigan Technological University Dynamic Systems Laboratory

Modal Parameter Estimation Concepts
[A k ] [A*k ]
lower j
[A k ] [A*k ] upper [A k ] [A*k ]
[H(s )]= ∑ + +∑ + +∑ +
terms (s−s k ) (s−s k ) k =i (s−s k ) (s−s k ) terms (s−s k ) (s−s k )
* * *
SYSTEM EXCITATION/RESPONSE
PEAK PICK SDOF POLYNOMIAL
MULTIPLE REFERENCE FRF MATRIX DEVELOPMENT
INPUT FORCE
RESIDUAL COMPENSATION
LOCAL CURVEFITTING
INPUT FORCE IFT
GLOBAL CURVEFITTING
INPUT FORCE
POLYREFERENCE CUVREFITTING
COMPLEX EXPONENTIAL MDOF POLYNOMIAL
Modal Parameter Estimation Concepts 1 Dr. Peter Avitabile

Parameter Estimation Concepts
NO COMPENSATION
y=mx
Y
COMPENSATION Y
Y
y=mx+b
X
WHICH DATA ???

Y
X
X

Parameter Extraction Considerations
HOW MANY POINTS ???
• ORDER OF THE MODEL
• AMOUNT OF DATA TO
RESIDUAL
BE USED
EFFECTS RESIDUAL
EFFECTS
• COMPENSATION FOR
RESIDUALS
HOW MANY MODES ???
The test engineer identifies these items

NOT THE SOFTWARE !!!
HOW MANY POINTS ???
lower
[Ak ]
[H( s) ] = ∑ ( s − s ) + s − s [A ] *
k
terms k( ) *
k
j
[Ak ]
∑ ( s − s ) + (s − s )
[A ]*
k
RESIDUAL
* EFFECTS
k=i k k RESIDUAL
EFFECTS
upper
[Ak ]
∑ ( s − s ) + (s − s )
[A ]
*
k
*
terms k k
HOW MANY MODES ???
[H(s )] = lower residuals + ∑ [Ak ]

+
[A*k ] j
+ upper
k =i (s−s k ) (s−s k )
*

The basic equations can be cast in either the

time or frequency domain
a1 a1* 1 −σt
h (s)= + h ( t )= e sin ωd t
(s − p1 ) (s − p1* ) mωd

MODAL PARAMETER ESTIMATION MODELS
Time representation
h ij ( n ) ( t ) + a1h ij ( n −1) ( t ) + L + a 2 n h ij ( n − 2 N ) ( t ) = 0
Frequency representation
[( jω) 2N
+ a 1 ( jω) 2 N −1 + L + a 2 N ]h ij ( jω) =
[( jω) 2M
+ b1 ( jω) 2 M −1 + L + b 2 M ]

The FRF matrix contains

redundant information
regarding the system
frequency, damping and
mode shapes
Multiple referenced data

can be used to obtain
better estimates of
modal parameters

Selection of Bands
Select bands for possible SDOF or MDOF

extraction for frequency domain technique.
Residuals ??? Complex ???

Mode Determination Tools
Summation MIF
A variety of tools assist in the determination

and selection of modes in the structure
1 Point Each From Panels 1,2, and 3 (37,49,241)
4
10
3
10
2
10
1
10
CMIF
0
10
-1
10
-2
10
0 50 100 150 200 250 300 350 400 450 500
Frequency (Hz)
CMIF Stability Diagram

Modal Extraction Methods
Peak Picking Circle Fitting SDOF Polynomial
A multitude of techniques exist
IFT
Complex Exponential
MDOF Polynomial Methods

Model Validation
Synthesis
Validation tools exist
to assure that an
accurate model has
been extracted from
measured data
1.2
0.8
0.6
0.4
0.2
0
S6
S5
S4
S3
S2
S1
MAC

Linear Algebra Concepts
 {v1} 
T
s1
  T
 {v 2 } 
s2
[A] = [{u1} {u 2 } {u 3} L]
 s3  {v 3 }T 
  
 O  M 
x . . . x 
0 x . . . 
 
[U]= . 0 x . . 
. . 0 x .
  [A]nm{X}m ={B}n
 0 . . 0 x 
[A]nm =[V]nn [S]nm [U]Tmm
[A ]−1=[Adjo int[A ]] {X}m =[A ]gnm{B}n =[[V ]nn [S]nm [U ]Tmm ] {B}n
g
Det[A ]
{X}m =[[U ]mm [S]gnm [V]Tnn ]{B}n
Linear Algebra Concepts 1 Dr. Peter Avitabile

Linear Algebra
The analytical treatment of structural dynamic

systems naturally results in algebraic equations
that are best suited to be represented through
the use of matrices
Some common matrix representations and linear

algebra concepts are presented in this section

Linear Algebra
Common analytical and experimental equations

needing linear algebra techniques
[G ] = [H][G ]
yf ff [H ] = [G yf ][G ff ] −1
[M ]{&x&}+[C]{x& }+[K ]{x}={F( t )} [[K ]−λ[M ]]{x}=0
[B(s )]{x (s )} = {F(s )} [B(s )]−1 = [H(s )] = Adj[B(s )]

det[B(s )]
O 
or [H(s )] = [U]  S  [L]T
 O

Matrix Notation
A matrix [A] can be described using row,column as

 a 11 a 12 a 13 a 14 
a a 22 a 23 a 24 
 21 
[A ] = a 31 a 32 a 33 a 34 
a a 42 a 43 a 44 
 41 
a 51 a 52 a 53 a 54 
( row , column )
[A]T -Transpose - interchange rows & columns
[A]H - Hermitian - conjugate transpose

Matrix Notation
A matrix [A] can have some special forms

Square Diagonal Triangular
 a 11 a 12 a 13 a 14 a 15  a 11  a 11 a 12 a 13 a 14 a 15 
a   0 a 25 
a 22 a 23 a 24 a 25  
a 22
 
a 22 a 23 a 24

 21 
[A ] = a 31 a 32 a 33 a 34 a 35  [A ] =  a 33  [A ] =  0 0 a 33 a 34 a 35 
  0
a a 42 a 43 a 44 a 45  
a 44
 0 0 a 44 a 45 
 41   
a 51 a 52 a 53 a 54 a 55   a 55   0 0 0 0 a 55 
Symmetric Toeplitz Vandermonde

 a 11 a 12 a 13 a 14 a 15 
a 5 a6 a7 a8 a9 
a a 25  a 1 a1 a 12 
 12
a 22 a 23 a 24
 a5 a6 a7 a8   
 4 
[A ] = a13 a 23 a 33 a 34 a 35  [A ] = a 3 a4 a5 a6 a7  [A ] = 
1 a2 a 22 
a a 45  a 1 a 32 
a6 
a 24 a 34 a 44 a3
 14   2
a3 a4 a5
  
a 15 a 25 a 35 a 45 a 55  1 a4 a 24 
 a 1 a2 a3 a4 a 5 

Matrix Manipulation
A matrix [C] can be computed from [A] & [B] as

 b11 b12 
 a 11 a 12 a 13 a 14 a 15  b 21 b 22   c11 c12 
 
a a 22 a 23 a 24 
a 25  b 31 b 32  = c 21 c 22 
 21    
a 31 a 32 a 33 a 34 a 35  b 41

b 42
  31
c c 32 
 b 51 b 52 
c 21 = a 21b11 + a 22 b 21 + a 23 b 31 + a 24 b 41 + a 25 b 51
c ij = ∑ a ik b kj
k

Simple Set of Equations
A common form of a set of equations is
[A ] {x} = [b]
Underdetermined # rows < # columns
more unknowns than equations
(optimization solution)
Determined # rows = # columns

equal number of rows and columns
Overdetermined # rows > # columns

more equations than unknowns
(least squares or generalized inverse solution)

Simple Set of Equations
This set of equations has a unique solution

2x − y = 1  2 − 1 0  x  1 
− x + 2 y − 1z = 2 − 1 2 − 1  y  = 2
    
−y+z =3  0 − 1 1   z  3
whereas this set of equations does not

2x − y = 1  2 − 1 0  x  1 
− x + 2 y − 1z = 2 − 1 2 − 1  y  = 2
    
4x − 2 y = 2  4 − 2 0   z  2

Static Decomposition
A matrix [A] can be decomposed and written as
[A ] = [L][U ]
Where [L] and [U] are the lower and upper
diagonal matrices that make up the matrix [A]
x 0 0 0 0  x x x x x
x x 0 0 0  0 x x x x
   
[L] =  x x x 0 0  [U] =  0 0 x x x
x x x x 0  0 0 0 x x
   
 x x x x x   0 0 0 0 x 

Static Decomposition
Once the matrix [A] is written in this form then

the solution for {x} can easily be obtained as
[A ] = [L][U ]
[U ] {X} = [L]−1 [B]
Applications for static decomposition and inverse
of a matrix are plentiful. Common methods are
Gaussian elimination Crout reduction
Gauss-Doolittle reduction Cholesky reduction

Eigenvalue Problems
Many problems require that two

matrices [A] & [B] need to be reduced
[A ]{&x&} + [B]{x} = {Q( t )} [[B] − λ[A ]] {x} = 0
Applications for solution of eigenproblems are
plentiful. Common methods are
Jacobi Givens Householder
Subspace Iteration Lanczos

Singular Valued Decomposition
Any matrix can be decomposed using SVD
[A ] = [U ][S][V ]T
[U] - matrix containing left hand eigenvectors

[S] - diagonal matrix of singular values
[V] - matrix containing right hand eigenvectors

SVD allows this equation to be written as

 {v1} 
T
s1
  T
 {v 2 } 
s2
[A] = [{u1} {u 2 } {u 3} L]
 s3  {v 3 }T 
   
 O  M 
which implies that the matrix [A] can be written in

terms of linearly independent pieces which form
the matrix [A]
[A ] = {u1}s1{v1}T + {u 2 }s 2 {v 2 }T + {u 3}s3 {v3}T + L

Assume a vector and singular value to be

1 
 
u1= 2 and s1 = 1
3 
 
Then the matrix [A1] can be formed to be
1  1 2 3
 
[A1 ] = {u1} s1 {u1} = 2 [1] {1 2 3} = 2 4 6
T
3  3 6 9
 
The size of matrix [A1] is (3x3) but its rank is 1
There is only one linearly independent
piece of information in the matrix
Consider another vector and singular value to be

1
 
u 2=  1  and s2 = 1
− 1
 
Then the matrix [A2] can be formed to be
1  1 1 − 1
 
[A 2 ] = {u 2 } s 2 {u 2 } =  1  [1] {1 1 − 1} =  1 1 − 1
T
− 1 − 1 − 1 1 
 
The size and rank are the same as previous case
Clearly the rows and columns
are linearly related
Now consider a general matrix [A3] to be

2 3 2 
[A 3 ] = 3 5 5  = [A1 ] + [A 2 ]
2 5 10
The characteristics of this matrix are not
obvious at first glance.
Singular valued decomposition can be used to

determine the characteristics of this matrix

The SVD of matrix [A3] is
1  1 0 1   {1 2 3} 

[A] = 2       {1 1 − 1} 
1 0
   1
 
3 − 1
 
0 
   0  {0 0 0} 

or
1  1 0
[A] = 21{1 2 3}T +  1 1{1 1 − 1}T + 00{0 0 0}T
3  − 1 0
     
These are the independent quantities that

make up the matrix which has a rank of 2
Linear Algebra Applications
The basic solid mechanics formulations as well as

the individual elements used to generate a finite
element model are described by matrices
 12 6L − 12 6L 
 6L 4L2 − 6L 2L2 
[k ] = EI3  
L − 12 − 6L 12 − 6L 
 6L 2L2 − 6L 4L 
ν ν
 
i j
θ i
θ j
 156 − 22L 54 13L 
E, I − 22L 4L2 − 13L − 3L2 
[m] = ρAL  
F i L Fj 420  54 − 13L 156 22L 
 13L − 3L2 22L 4L2 

 σ x   C11 C12 C13 C14 C15 C16  ε x 

 σ  C C 22 C 23 C 24 C 25 C 26  ε y 
 y   21  
 σ  C C32 C33 C34 C35 C36  ε z 
{σ} = [C]{ε}⇒ z  = 31  
τ xy  C 41 C 42 C 43 C 44 C 45 C 46 γ xy 

τ xz  C51 C52 C53 C54 C55 C56  γ xz 
    
τ yz  C 61 C 62 C 63 C 64 C 65 C 66  γ yz 

Finite element model development uses individual

elements that are assembled into system matrices

Structural system equations - coupled

[M ]{&x&}+[C]{x& }+[K ]{x}={F( t )}
Eigensolution - eigenvalues & eigenvectors
[[K ]−λ[M ]]{x}=0
Modal space representation
of equations - uncoupled
\  \  \ 
 M {&p&} +  C {p& } +  K {p} = [U ]T {F}
     
 \   \   \ 

Multiple Input Multiple Output Data Reduction

[G ] = [H][G ]
yx xx [H ] = [G yx ][G xx ]−1
[Gyx] = [H] [Gxx]
RESPONSE FREQUENCY RESPONSE FUNCTIONS FORCE

(MEASURED) (UNKNOWN) (MEASURED)
Matrix inversion can only be performed if the

matrix [Gxx] has linearly independent inputs

Principal Component Analysis using SVD

 {v1} 
T
s1
  T
 {v 2 } 
s2
[G xx ] = [{u1} {u 2 } {0} L]
 0   {0}T 
   
[Gxx]  O  M 
SVD of the input excitation matrix identifies the

rank of the matrix - that is an indication of how
many linearly independent inputs exist

SVD of Multiple Reference FRF Data

 {v1} 
T
s1
  T
 {v 2 } 
s2
[H] = [{u1} {u 2 } {u 3} L]
 s3  {v 3 }T 
   
 O  M 
[H]
FREQUENCY RESPONSE FUNCTIONS
0 50 100 150 200 250 300 350 400 450 500

Frequency (Hz)
SVD of the [H] matrix gives an indication

of how many modes exist in the data
Least Squares or Generalized Inverse for

Modal Parameter Estimation Techniques
[H(s )] = ∑ [
j
Ak ]
+
[A*k ]
k =i (s −s k ) (s −s k)
*
Least squares error minimization of

measured data to an analytical function

Extended analysis and evaluation of systems

[K ][U ] = [M I ][U ][`ω2 ]
[U ][K ][U] = [U] [M ][U][`ω ]
T
I
T
I
2
[K ] = [K ] + [V] [`ω + K ][V]

I S
T 2
S
[ ] [
− [K S ][U ][U ] [M I ] − [K S ][U ][U ] [M I ]
T T
]T
[K I ] = [K S ] + [V ]T [`ω2 + K S ][V ] − [[K S ][U ][V ]] − [[K S ][U ][V ]]T

generally require matrix manipulation of some type

Many other applications exist

Correlation Model Updating
Advanced Data Manipulation
Operating Data Rotating Equipment
Nonlinearities
Modal Parameter Estmation
and the list goes on and on


Young Engineer IMAC21 2003 Assembled MACL Tcm18-189941

Uploaded by

Document Information

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

Young Engineer IMAC21 2003 Assembled MACL Tcm18-189941

Uploaded by

Copyright:

Available Formats

Presentation Topics

Dr. Peter Avitabile

Modal Overview University of Massachusetts Lowell

The intent of the Young Engineer Program is to expose the new

It is NOT intended to be a detailed treatment of this material.

Rather it is intended to prepare one for some of the in-depth

This presentation is intended to identify the basic methodology

Intent of Young Engineer Program 1 Dr. Peter Avitabile

A Simple Non-Mathematical Presentation

Could you explain and how is it

Illustration by Mike Avitabile Illustration by Mike Avitabile Illustration by Mike Avitabile

Experimental Modal Analysis 1 Dr. Peter Avitabile

Equation of motion [M n ]{&x& n } + [C n ]{x& n } + [K n ]{x n } = {Fn ( t )}

Eigensolution [[K n ] − λ[M n ]]{x n } = {0}

Experimental Modal Analysis 2 Dr. Peter Avitabile

Simple time-frequency response relationship

Experimental Modal Analysis 3 Dr. Peter Avitabile

Sine Dwell to Obtain Mode Shape Characteristics

Experimental Modal Analysis 4 Dr. Peter Avitabile

Experimental Modal Analysis 6 Dr. Peter Avitabile

The third row of the FRF matrix - mode 1

The peak amplitude of the imaginary part of the

Experimental Modal Analysis 7 Dr. Peter Avitabile

The second row of the FRF matrix is similar

The peak amplitude of the imaginary part of the

Experimental Modal Analysis 8 Dr. Peter Avitabile

Experimental Modal Analysis 9 Dr. Peter Avitabile

Experimental Modal Analysis 10 Dr. Peter Avitabile

Experimental Modal Analysis 11 Dr. Peter Avitabile

Actual time signals

Digitized time signals

Compute FFT of signal

FREQUENCY RESPONSE FUNCTION COHERENCE FUNCTION

Experimental Modal Analysis 12 Dr. Peter Avitabile

Hammers and Tips

INPUT POWER SPECTRUM

Experimental Modal Analysis 13 Dr. Peter Avitabile

Leakage and Windows

ACTUAL TIME SIGNAL

WINDOWED TIME SIGNAL

Experimental Modal Analysis 14 Dr. Peter Avitabile

AUTORANGING AVERAGING WITH WINDOW

Experimental Modal Analysis 15 Dr. Peter Avitabile

Experimental Modal Analysis 16 Dr. Peter Avitabile

The FRF is made up

Experimental Modal Analysis 17 Dr. Peter Avitabile

The task for the modal test engineer is to

Experimental Modal Analysis 18 Dr. Peter Avitabile

HOW MANY POINTS ???

HOW MANY MODES ???

Experimental Modal Analysis 19 Dr. Peter Avitabile

INPUT TIME FORCE

INPUT SPECTRUM OUTPUT SPECTRUM

f(j ω) h(j ω) y(j ω)

Experimental Modal Analysis 20 Dr. Peter Avitabile

Experimental Modal Analysis 21 Dr. Peter Avitabile

MODE 1 CONTRIBUTION MODE 2 CONTRIBUTION

Experimental Modal Analysis 22 Dr. Peter Avitabile

Experimental Modal Analysis 23 Dr. Peter Avitabile

I think I’m starting to understand this now !!!

Experimental Modal Analysis 24 Dr. Peter Avitabile

Experimental Modal Analysis 25 Dr. Peter Avitabile

DEVELOP MASS RIB used for studies

Experimental Modal Analysis 26 Dr. Peter Avitabile