Documente Academic
Documente Profesional
Documente Cultură
(Lecture 1)
Model Validation/Correlation:
Producing major test modes validates the model
Natural frequencies
Mode shapes
Damping information are not available in FE models
Model Updating
Correlation of experimental/analytical
model
Adjust/correct the analytical model
Optimization procedures are used for
updating.
Component Model
Identification
Substructure process
The component model
is incorporated into the
structural assembly
Force Determination
Knowledge of dynamic force is required
Direct force measurement is not possible
Measurement of response + Analytical Model
results the external force
([K ] [M ]){x} = { f }
2
x j ( ) N jrkr
h jk ( ) = = 2 .
f k ( ) r =1 r 2
Curve-fitting the
measured FRF:
Modal Model is obtained
Spatial Model is obtained
Overview of Modal Testing IUST ,Modal Testing Lab ,Dr H Ahmadian
Summary of Measurement
Methods
Basic measurement system:
Single point excitation
1 Only in few
points
2
4 4
3 1
k3 X sin(t ) sin(3t ) =
3
4 4
F sin(t ) cos( ) F cos(t ) sin( )
3
m X + kX + k3 X = F cos( )
2 3
4
cX = F sin( )
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Cubic stiffness nonlinearity
X 1
=
F
2
2
m + k + k3 X + (c )
2 3 2
4
3
keq = k + k3 X 2
4
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Cubic stiffness nonlinearity
X 1
=
F 4c F
k m + i c +
2
X
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Coulomb friction nonlinearity
X 1
=
F 4c F
k m + i c +
2
X
X increasing
X ( )
1
H ( ) = =
F ( ) k m + ic
2
1
k m 2 + id
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Time Domain: Impulse
Response Function
Impedance
F ( ) F ( )
Inertance A( ) 2 X ( )
Inverse is Apparent =
F ( ) F ( )
mass
Re( ) = k m 2
reciprocal plots X ( )
F ( )
Real part Im( ) = c
Imaginary part
X ( )
U 2 +V 2 =
((k m ) + (c ) )
2 2 2 2
1
=
2
4c (( k m ) + ( c ) )
2 2 2 2 2
2c
H ( ) =
1
=
(k m 2 ) id
k + id m 2
(k m ) + d 2
2 2
U=
(k m ) 2
, V =
d
(k m ) + d 2 2 2
(k m ) 2 2
+ d2
2 2
1 1
U + V +
2
=
2d 2d
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
A Demo
Basic Assumptions
The structure is assumed to
be linear
The structure is time invariant
The structure obeys Maxwells
reciprocity
The structure is observable
loose components, or degrees-of-
freedom of motion that are not
measured, are not completely
observable.
Response Model
([K ] [M ]) = [ ( )]
2 1
[ ] ([K ] [M ])[ ] = [ ] [ ( )] [ ]
T 2 T 1
([] [I ]) = [ ] [ ( )] [ ]
2 T 1
[ ( )] = [ ] ([] [I ])[ ]
1 T 2 1
[ ( )] = [ ]([] [I ]) [ ] 2 1 T
jrkr
N N
r A jk
jk ( ) = 2 = 2
r =1 r r =1 r
2 2
1 1.2 0.8
[M ] = [K ] =
kg MN / m
1 0.8 1.2
4e5 1 1 1
[ ]
r =
2
[ ] =
2 e 6 2 1 1
0.5 05 1.2e6 2
11 ( ) = + = .
4e5 2
2 e6 2
8e11 2.4e6 +
2 4
Zero
0.5 05 1.2e6 2
11 ( ) = + = .
4e5 2
2 e6 2
8e11 2.4e6 +
2 4
Poles
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Example (continued) :
0.5 05 8e4
12 ( ) = = .
4e5 2
2 e6 2
8e11 2.4e6 +
2 4
Special cases:
[D ] = [K ],
[D ] = [M ],
[D ] = [K ] + [M ].
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
MDOF Systems with Structurally
Proportional Damping
Response Model
([K ] + i[D] 2 [M ]) = [ ( )]1
[ ]T ([K ] + i[D ] 2 [M ])[ ] = [ ]T [ ( )]1[ ]
([r2 (1 + ir2 )] 2 [I ]) = [ ]T [ ( )]1[ ]
[ ( )]1 = [ ]T ([ r2 (1 + ir2 )] 2 [I ])[ ]1
[ ( )] = [ ]([ r (1 + ir )] [I ]) [ ]T
2 2 2 1
N
jrkr Real Residue
jk ( ) = 2
r =1 r (1 + i 2
r ) 2
Complex Pole
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
MDOF Systems with Viscously
Proportional Damping
Response Model
([K ] + i [C ] [M ]) = [ ( )]
2 1
[ ] ([K ] + i [C ] [M ])[ ] = [ ] [ ( )] [ ]
T 2 T 1
([ ]+ i [2 ] [I ]) = [ ] [ ( )] [ ]
2
r r r
2 T 1
[ ( )] = [ ] ([ ] + i [2 ] [I ])[ ]
1 T 2
r r r
2 1
[ ( )] = [ ]([ ] + i [2 ] [I ]) [ ]
2
r r r
2 1 T
N
jrkr
jk ( ) = 2
r =1 r 2
+ 2 r r
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
MDOF Systems with General
Structural Damping
The equation of motion:
[M ]{&x&(t )}+ ([K ] + i[D ]){x(t )} = { f (t )}
The orthogonality:
[ ] [M ][ ] = [I ], [ ] [K + iD ][ ] = [].
T T
Model 1
m1 = 0.5kg , m2 = 1kg , m3 = 1.5kg
k j = 1e3N / m, j = 1,...,6
Undamped
950 0.464 0.218 1.318
= 3352 , [ ] = 0.536 0.782 0.318
6698 0.635 0.493 0.142
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Example:
Pr oportional
[D ] = 0.05[K ]
950 0.464 0.218 1.318
= (1 + i 0.05) 3352 , [ ] = 0.536 0.782 0.318
6698 0.635 0.493 0.142
Non Pr oportional
d1 = 0.3k1 , d j = 0.0, j = 2,...,6
957(1 + i 0.067)
= 3354(1 + i 0.042) ,
6690(1 + i 0.078)
Almost real modes
0.463( 5.5 ) 0.217(173 ) 1.318(181 )
o o o
([ (1 + i )] [I ]) = [ ] [ ( )] [ ]
2
r
2
r
2 T 1
[ ( )] = [ ] ([ (1 + i )] [I ])[ ]
1 T 2
r
2
r
2 1
[ ( )] = [ ]([ (1 + i )] [I ]) [ ]
2
r
2
r
2 1 T
N
jrkr Complex Residues
jk ( ) = 2
r =1 r (1 + i 2
r ) 2
Complex Poles
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
General Force Vector
In many situations
the system is
excited at several
points.
The solution:
{X } =
N
{ } {F }{ }r
T
r
r =1 (1 + i )
2
r
2
r
2
([ ] [ ]){ }e it = {F }e it
{} + 2
K iD M X
{x(t )} = X ei (t )
Real and imaginary parts:
(([K ] [M ])cos + [D]sin ){X }= {F }
2
T
N
{X } =
r =1 (1 + i )
2
r
2
r
2
512 channel
37 Shakers
{X } = ([K ] [M ] + i[C ]) {F }
2 1
= = +
iX r =1 i r
s r =1 i s r i s *
r
N
u jr ukr u u
jk ( ) = + jr kr
r =1 i sr i s
r
kL2 J 2 kL
2
2
4
2 + z
+
2
= 0.
I0 I0 I0
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Natural Frequencies
kL2 J 2 2
2
4 2 2 +
kL
+ z
= 0.
I0 I0 I0
2
2 2
2 2
1, 2 = 2
0 + 2
0 +
2 4
2
kL J
02 = , =
I0 I0
( i1J z / L2 )
(k I / L ) i 0
2
1 0
2
( i2 J z / L2 )
(k I / L ) 1 0
2
2 0
2
I 0 / L2 0 &x& c J z / L2 x& k 0 x f x
2
+ & + = ,
0 I 0 / L &y& J z / L2 c y 0 k y f y
(k I 0 / L + ic ) (iJ z / L )
2 2 2 1
[ ( )] =
(i J z / L2
) (k 2
I 0 / L2
+ ic )
xx ( ) = yy ( ) Loss of Reciprocity
xy ( ) = yx ( )
Coupling Effect
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
FRF of the Rotating Structure
with External Damping
Complex Mode
Shapes
due to significant
imaginary part
xr cos(t ) sin(t ) x
= y
r
y sin( t ) cos( t )
1 = 02 + ( z / 2) 2 z / 2 2 = 02 + ( z / 2)2 + z / 2
Equation of Motion in Rotating Coordinates
I0 / L 2
0 &x&r 0 2 z I 0 / L2 + J z / L2 x&r
+ &
0 I 0 / L2 &y&r 2 z I 0 / L2 J z / L2 0 yr
2z I 0 / L2 + J 2z / L2 + k x c 2 + k y s 2 cs (k y k x ) xr 0
+ 2 = .
cs (k y k x ) z I 0 / L + J z / L + k x c + k y s yr 0
2 2 2 2 2
1 = 02 + ( z / 2)2 z / 2 + z 2 = 02 + ( z / 2) 2 + z / 2 z
Note: Eigenvectors remain unchanged
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Analysis using rotating frame
Fxr cos(t ) sin(t ) Fx
= .
Fyr sin(t ) cos(t ) Fy
I0 / L2
0 &x& cE + cI J z / L 2
x&
2
+ &
0 I 0 / L y J z / L
&& 2
cE + c I y
kx z cI x 0
+ = ,
z c I k y y 0
At super critical speeds the real parts of
eigen-values may become positive,
i.e. unstable system
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Dynamic Analysis of General
Rotor-Stator Systems
The rotating machines and their modal
testing is much more complex
Non-symmetric bearing support
Fixed/Rotating observation frame
Non-axisymmetric rotors
Internal/External damping
These lead to:
Time-varying modal properties
Response harmonies not present in the excitation
Instabilites (negative modal damping)
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Dynamic Analysis of General
Rotor-Stator Systems
Equation of motion of rotating systems
are prone:
to lose the symmetry
to generate complex eigen-values/vectors
from velocity/displacement related non-
symmetry
to include time varying coefficients as
appose to conventional Linear Time
Invariant (LTI) systems
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Dynamic Analysis of General
Rotor-Stator Systems
System Type Stationary Coord. Rotating Coord.
Forwards Whirl
Backwards Whirl
1 3 1
[M ] = , [K ] = ,
1 1 3
0 0.5 1 0.5
[C ] = C + (1 C )
0 . 5 0 0 .5 1
1
[M ] = , [C ] = 0,
1
0 1 3 1
[K ] = K + (1 K )
1 0 1 3
Symmetric Skew-symmetric
Rank 2 matrices
Overview of Modal Testing IUST ,Modal Testing Lab ,Dr H Ahmadian
Extracting real modes
Since V is skew symmetric,
{X }e it
= [ ( )]{F }e it
n =1
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Periodic Vibration
To derive FRF from periodic vibration
signals:
Determine the Fourier Series components
of the input force and the relevant
response
Both series contain components at the
same set of discrete frequencies
The FRF can be defined at the same set of
frequency points by computing the ratio of
response to input components.
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Transient Vibration
Analysis via Fourier Transform
+
1
f (t )e
it
F ( ) = dt
2
X ( ) = H ( ) F ( )
+
H ( ) F ( )e
it
x (t ) = d
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Transient Vibration
Response via time domain (superposition)
+
x (t ) = h(t ) f ( )d
1
Let f (t ) = (0) F ( ) =
2
+
1
2
it
Then x (t ) = H ( ) e d = h ( t )
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Transient Vibration
To derive FRF from transient vibration
signals:
Calculation of the Fourier Transforms of
both excitation and response signals
Computing the ratio of both signals at the
same frequency
In practice it is common to compute a
DFT of the signals.
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Random Vibration
Neither excitation nor response signal
can be subject to a valid Fourier
Transform:
Violation of Dirichlet Conditions
Finite number of isolated min and max
Finite number of points of finite discontinuity
Here we assume the random signals to
be ergodic
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Random Vibration
Time Signal
f (t )
Autocorrelation Function
+
R ff ( ) = f (t ) f (t + )dt
Power Spectral Density
+
1
R
i
S ff ( ) = ( )e d
2
ff
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Random Vibration
Time Signal
Autocorrelation
Sinusoidal Signal
Autocorrelation
Random Signal
Autocorrelation
Noisy Signal
Autocorrelation
u = t +
The Auto/Power Spectral Density function is
real and even.
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Random Vibration
Cross Correlation / Spectral Densities
+ +
1
x(t ) f (t + )dt R
i
Rxf ( ) = S xf ( ) = ( )e d
2
xf
+
R ff ( ) =
f (t ) f (t + )dt S ff ( ) = F * ( ) F ( )
+
Rxf ( ) = + = ( ) F ( )
*
x ( t ) f ( t ) dt S xf ( ) X
+
Rxx ( ) = x(t ) x(t + )dt S ( ) = X ( ) X ( )
*
xx
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Random Vibration
To derive FRF from random vibration signals:
X ( )
H ( ) =
F ( )
X * ( ) X ( ) S xx ( )
H1 ( ) = * =
X ( ) F ( ) S xf ( )
F ( ) X ( ) S fx ( )
*
H 2 ( ) = * =
F ( ) F ( ) S ff ( )
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Complete/ Incomplete Models
It is not possible to measure the
response at all DOF or all modes of
structure (N by N)
Different incomplete models:
Reduced size (from N to n) by deleting
some DOFs
Number of modes are a reduced as well
(from N to m, usually m<n)
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Incomplete Response Models
m< N
Ajk
jk ( ) =
r =1
2
r
+ ir2 2
r r
[ ] 2
r m m
[ ]nm
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Incomplete Response Models
p
(
[K ] r [M ] {r } = {0},
2
)
(
[K ] r [M ]
2 { r }
)
+
[K ]
2
r 2 [M ]
[M ] r {r } = {0},
p p p p
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian
Eigenvalue Sensitivity (MDOF)
Multiply by {r }T
{r }
{r } (T
[K ] [M ]
2
) +
p
r
{r }
T [K ]
2
r 2 [M ]
[M ] r {r } = {0},
p p p
[K ] 2 [M ]
{r }
T
r {r }
2
p p
results = r
p {r } [M ]{r }
T
Starting from :
(
){ }
[K ] r2 [M ] r + [K ] 2
r [M ] r2
[M ]
{r } = {0},
p p p p
{r } N
and taking = j { j }
p j =1
j r
[K ] r2 2 [M ]
( )
[K ] [M ] rj { j }+
N
2
[M ] r {r } = {0}
p p p
r
j =1
jr
[K ] r2 2 [M ]
( )
{s } [K ] r2 [M ] rj { j }+ {s }
N
T T
[M ] r {r } = {0}
j =1 p p p
j r
[K ] 2 [M ]
(
rs + {s }
2 2
) T
r {r } = {0}
p p
s r
p p
s r
[K ] 2 [M ]
{s } T
r {r }
rs = p p
(r2 s2 )
[K ] 2 [M ]
{s }
T
r {r }
{r } N
p p
= {s }
p s =1
sr
(r s )
2 2
( ) 1
=
k (k + ic m )
2 2
( ) i
=
c (k + ic m )
2 2
( ) 2
=
m (k + ic m )
2 2
{ x ( ) A ( )} = { x ( )} [Z ( )][ ( )]A
T T
j j
p p p
[ ( )] [Z ( )]
= [ ( )] [ ( )]
p p
[ ( )] [K ] [C ] 2 [M ]
= [ ( )] + i [ ( )]
p p p p
Theoretical Basis IUST ,Modal Testing Lab ,Dr H Ahmadian