Sie sind auf Seite 1von 35

1

Numerical Experimental Approach in Structural Dynamics

3.2 Semi-analytical methods


- when the input force is not known analytical-the input is devided in
components: impulse, step or ramp
analytical methods
3.3 Numerical Methods for solving differential equation with initial
conditions

- Definition of function variation law y which describes the


system dynamics
- y depends on independent variables (previous states)
- Ordinary differential equation
second order differential
equation reduced to differential equation of 1st order.
Ex. Second order:

d2 y
dy

A
(
x
)

B( x)
dx 2
dx

(1)

Can be

dy
z ( x)
dx
dz
A( x) z ( x) B( x) , where
dx
Z(x) is dependent auxiliar function/variable.

(2)

2
Numerical Experimental Approach in Structural Dynamics
Ordinary Differential Equations
reduced to one system
with n differential equation of order 1 written for n function
yi(x), i=1,2,3...n.

dyi
fi ( x, y1 ,... yn ), i 1, 2,...n
,
dx

(3)

f beeing non linear functions subject to independent variables


x, y .
Ec. (3) admits an infinite solution
Particularly is attached
specific conditions (e.g, values
for yi ( x) ).
For initial conditions given.
Cauchy:

y ( x)

dy
f ( x, y), y( x0 ) y0
dx

(4)

3.3.1. Integration and Numerical Differential


b

The meaning of integral

S a f ( x)dx
f ( x)

The area delimited by OX, curve f ( x) and x=a si x=b


Method of trapezoids: area is approximated using
trapezoids defines by range [a, b] and divided in equal
sections.

X0=a

xn=b

ba
h
0
n
x j a h j , j 1, 2...n
Integral is estimated:

(5)

3
Numerical Experimental Approach in Structural Dynamics
n1
h

ST ya yb 2 y j
2
j 1
.

The integral is approximated using subsections and [a b] is


devided in 2n equal section:

ba
x j 1 x j 0,1 j 2n
2n
x j a jh

ST

(6)

h
ya yb 4( y 1 y3 ... y2n1 ) 2( y 2 y4 ... y2n2 )
3

Simpson generalised formula.

3.3.2 Derivatives with finite difference approximation


First order derivative
a)Approximation with regresive difference

f ( x)

f ( xk ) f ( xk 1 )
xk xk 1

b)Approximation with progresive difference

f ( x)

f ( xk ) f ( xk )
xk 1 xk

c)Approximation with central difference difference

f ( x)

f ( xk 1 ) f ( xk 1 )
xk 1 xk 1

Advantages
Representation error reduced due to floating point
accuracy:
.
3.3.3. Euler Method
Principle:

At each iteration, the function is approximated


by adding to the current value a correction value equal
to the product of step and derivative function.
Taylor series for y(x) around xk:

4
Numerical Experimental Approach in Structural Dynamics

x xk
( x xk )2
y( x ) y( x k )
y' ( x k )
y' ' ( x k ) ...
1!
2!
.(8)
If it is known the value of the function calculated in

xk

yk f ( xk )
function f(x,y) enable the calculation of different order derivatives
y q ( x) .
not

y' ( x k ) y' k f ( x k , y k )
not

y' ' ( x k ) y' ' k f ' x ( x k , y k ) f ' y ( x k , y k ) y' k


f ' x ( xk , yk ) f ' y ( xk , yk ) f ( xk , yk )

(9)

where f x, f y are partial derivatives of function f(x,y) with respect


to x and y.

- If it is considered a uniform distribution of points xk


(xk +1 = xk + h) is possible starting from the point x0
corresponding to Cauchy problem and using formula:
h
h2
hp ( p)
y k 1 y k y' k
y' ' k ...
y
...
1!
2!
p! k
(10)
- to determine the numerical solution of differential eq. as a
sequence of points: (x0, y0), (x1, y1), , (xk, yk),
Euler method:
h
y k 1 y k y' k y k h f ( xk , y k )
1!

(11)

Algorithm Euler method:


1. Problem statement :
f(x,y) from y=f(x,y), step h, number of points for solution
computing, N:
2. Set the initial condition x0 si y0, with y0=f(x0)
3. Starting point initialization x(1)= x0, y(1)= y0;
4. Euler algorithm:
for k=1 to N-1 do
y(k+1)=y(k)+f(x, y(k))*h
x=x+h
end;
5. The approximated solution is in the vector y[1..n].

5
Numerical Experimental Approach in Structural Dynamics
3.3.4 Runge-Kutta Method
Principle:
Starting from a known point with coordinates (xk, yk), the
next point value yk+1 is calculated applying a linear
combination of corrections Kj, weighted by a series of
coefficientsrj to be determined.
r

y k 1 y k

rj K j
j 1

, r - order

K 1 h f ( xk , y k )
y k 1 y k K 1
Improved accuracy is obtained by calculating the correction to be
applied by yk in the derivative function y (x) in the mid range
(xk +1 / 2 = xk + h / 2). Calculating the derivative at this point
requires an approximation to yk +1 / 2, which is obtained using a
K 1 h f ( xk , y k )

K 2 h f ( xk

K
h
, yk 1 )
2
2

classical Euler step y k 1 y k K 2


Advantages:
- direct methods (methods do not require auxiliary start up points)
- are identical with Taylor series until r rank
- Does not require evaluation of partial derivatives of f (x, y), .
Determination of corrections Kj:
- Product between the step h and some values of f , which
are calculated in neighborhood (xk, yk).
K j h f ( j , j )

j xk j h

j 1, ...,r

j 1

j yk

ij K i
i 1

If

1 0

11 0 ,

and

1 xk

1 yk

K 1 h f ( xk , y k )
K 2 h f ( xk 2 h , y k 21 K 1 )
K 3 h f ( xk 3 h , y k 31 K 1 32 K 2 )
s . a . m. d .

Determination of corections rj, j andji:

6
Numerical Experimental Approach in Structural Dynamics
r

- Assign the Runge Kutta


Taylor Series

y k 1 y k

rj K j
j 1

to coincide with

In practice the particular cases are used:


- Particular cases for RungeKutta
Case r=1 (Runge Kutta of 1st order )
The restriction is applied: 11=1.

yk 1 yk K1 yk h f ( xk , yk ) Euler method.

Case r=2 ( RungeKutta of 2nd order)


Equation:
y k 1 y k 21 K 1 22 K 2
y k 21 h f ( x k , y k ) 22 h f ( x k 2 h , y k 21 h f ( x k , y k ))

y k 1 y k 21 h f ( xk , y k ) 22 h [ f ( xk , y k ) 2 h f ' x ( xk , y k )
21 h f ( xk , y k ) f ' y ( xk , y k )]

From Taylor Series expansions:

y k 1

h2
y k h f ( xk , yk )
f ' x x k , y k f x k , y k f ' y x k , y k
2

Is imposed that the last expressions to coincide:


21 22 1
1
2
1
22 21
2

22 2

If21=0, results: 22=1 si 2=21=1/2,

K 1 h f ( xk , y k )
K 2 h f ( xk

K
h
, yk 1 )
2
2

y k 1 y k K 2 y k h f [ xk

h
h
, y k f ( xk , y k )]
2
2

7
Numerical Experimental Approach in Structural Dynamics
If 22=1/2, results 21=1/2 si 2= 21=1, respectively:
K 1 h f ( xk , y k )
K 2 h f ( x k h, y k K 1 )
y k 1 y k

f ( x k , y k ) f [ x k h , y k h f ( x k , y k )]
1
1
K1 K 2 yk h
2
2
2

Case r=3 (Formula RungeKutta of 3rd order)


- Identicaly as for r=2
- System with 6 eq and 8 unknown variables.
- The formula used:

K 1 h f ( xk , y k )
K 2 h f ( xk h , y k K 1 )
K 3 h f ( xk
y k 1 y k

h
1
1
, yk K1 K 2 )
2
4
4

1
( K1 K 2 4 K 3 )
6

Case r=4 (Formula RungeKutta) This is the Runge-Kutta


method most commonly used in practice and uses a system of
coefficients that leads to the formula:

K 1 h f ( xk , y k )
K
h
, yk 1 )
2
2
K
h
, yk 2 )
2
2
h, yk K 3 )

K 2 h f ( xk
K 3 h f ( xk
K 4 h f ( xk
y k 1 y k

1
( K1 2 K 2 2 K 3 K 4 )
6

8
Numerical Experimental Approach in Structural Dynamics
Algorithm Runge-Kutta of 4order
1. Problem statement :
f(x,y) from y=f(x,y), step h, number of points for solution
computing, N.
2. Set the initial condition x0 si y0, with y0=f(x0);
3. Starting point initialization x(1)= x0, y(1)= y0;
4. Runge-Kutta:
for i=1 to N-1 do
Determine the coefficients K1-K4 at current point, (x(i), y(i)):
K1=f(x(i), y(i))*h
K2=f(x(i)+h/2, y(i)+K1/2)*h
K3=f(x(i)+h/2, y(i)+K2/2)*h
K4=f(x(i)+h, y(i)+K3)*h
Determine the value of y in the point x(i+1)
end;
5. The approximated solution is found in vectors x[1..n] and
y[1..n].
Example:
System with resort-damping Euler and Runge-Kutta methods:
F
z 2n z n 2 z
m
F
1
1
( s 2 2n s n 2 ) Z G ( s )
2
m
m s 2n s n 2

Fig. 3.1 Displacement Evolution Euler method

9
Numerical Experimental Approach in Structural Dynamics
F=100N, m=1kg, k=10 000N/m, c=20N/m/s, h=0.001

Fig. 3.2 Displacement evolution Euler


Modified Euler Method:

F
z 2n z n 2 z
m
zu
z u

Exact Euler method

z j 1 z j hu j
u j 1 u j hu j

F
z 2n z n 2 z
m
zu
z u
1
z j 1 z j h(u j u j 1 )
2
1
u j 1 u j h(u j u j 1 )
2

Modiffied Euler method

10
Numerical Experimental Approach in Structural Dynamics

Fig. 3.3 Displacement evolution Euler modiffied method


Course IIIrd
3.3.5 Numerical Integration Methods
Types of numerical methods for solving ordinary first-order
equations
A. Direct Numerical Methods (one step methods) calculates
at each step ti the approximated solution yi , using a
recurrence relation expressed as a function of estimated
value in the previous yi 1 and posibly as a function of hi 1 and
ti 1 .

hi 1 ti ti 1

B. Indirect Numerical Methods (multi step methods) calculates


at each step ti the approximated solution yi based on a

11
Numerical Experimental Approach in Structural Dynamics
reccurent relation as a function of calculated values in the
previous step yi 1 yi 1 , ... yi k
-Euler method
-Modified Euler
-Runge Kutta Method (RK1-4, RK-Felberg, RK Verner,
Nystrom)
Both types can be divided in 2 categories:
I Explicit Methods (with predictor) :

yi yi 1 f ( yi 1, ti 1, hi 1)
II. Implicit methods (with corrector):

yi yi 1 f ( yi 1, ti 1, hi 1, yi (ti , h i ))

obtained from a method with predictor

yic yi 1 f ( yi 1, ti 1, hi 1, yip (ti , h i )) .


Advantages of direct and explicit methods
- Less time consuming (unlike indirect and implicit methods)
- Autostart Methods
Disadvantages
The methods fails for some differential equations stiff
stiff differential equation (intraductibil ca noiune de calcul
numeric)
(stiff= dificil, nepenit)
-For a non-linear system with the dynamics expressed as:

dy
f (t , y(t )), y(t0 ) y0 , t0 a, b
dt
the differential equation with initial conditions is (asymptotically)
stable, but the method of integration is unstable (if the method
used is the explicit integration).
-For a linear system with a state-space representation
there are large disproportions between the eigenvalues of the
Jacobian matrix for the first order differential equations associated
with nth order equation.
.

12
Numerical Experimental Approach in Structural Dynamics

fi
dx
A x(t ), A J
|1i
dt
x j j n

Large differences between the modules i


polynomial of matrix A.
min min i
min min i
- max max i ,
hi ti 1 ti
max
2
3
10 (10 )

of characteristic

min

- Stiffness ratio
min kt emint - slowest transient

max kt e

- fastest transient
Significant difficulties can occur when standard numerical
techniques are applied to approximate the solution of a differential
equation when the exact solution contains terms of the form
et, where is a complex number with negative real part.
max

Avoiding these situations:


- Reduce integration step (but a decrease could cause
excessive accumulation of rounding errors or discretization
errors).
- Use predictor - corrector methods (indirect method) with an
integration step in some relationship with time constants
-

Met. Adams Bashforth (predictor)


Met. Adams-Moulton (corector)
Met. Adams predictor- corector

Ex. of stiff systems


Problems involving rapidly decaying solutions:
- Motion study of spring and damping systems having large
values (physical stiffness)
- Control Systems
- Chemical Kinetics
Ex. spring-mass-damping:
mz cz kz 0,
m 1, c 1001, k 1000

,
x(0) x0 , x(0) 0
For a linear representation of form x Ax f (t ) :
1
0
0
x0
A
,
f
(
t
)

,
x
(0)

0
0
1000 1001

13
Numerical Experimental Approach in Structural Dynamics
With eigenvalues:
min 1, max 1000

max
1000!!!
min
The solution is:
1 1000t 1000 t
x(t ) x0 (
e

e ) x0et
999
999
- the term numerical computation very sensitive to step size
Examples of solving and plot the solution of second and
higher order equations using Matlab (from P.Howard, 2007)

-eval()- evaluates string (character array )and y is a symbolic


object.
-vectorize()-converts symbolic objects into string
2. Solve and plot

14
Numerical Experimental Approach in Structural Dynamics

15
Numerical Experimental Approach in Structural Dynamics

2. Example of Simple Harmonic Excitation

State 1 (equilibrium, springs unstressed)

k
c

F sin t

State 2 (at time t during vibration)


u

T T F sin t mu

ku cu F sin t mu

ku cu mu F sin t

(*)

Constitutive equations - try combination with sin and cos

u A sin t B cos t
u A cos t B sin t
u 2 A sin t 2 B cos t

F sin t

16
Numerical Experimental Approach in Structural Dynamics
Substitute in (*):
k ( A sin t B cos t ) c(A cos t B sin t ) m(2 A sin t 2 B cos t )
F sin t

Group sin, cos terms:


(kA cB m2 A F )sin t (kB cA m2 NB) cos t 0
cA

Fc

kB cA m2 NB 0 B k m2 B (k m2 )2 (c)2

kA cB m2 A F

kA

(c)2 A
k m2

m2 A F A F

k m2
(k m2 )2 (c)2

Substitute in :
u A sin t B cos t
k m2

uF

(k m2 )2 (c) 2
Rewrite:

uF
u

(k m 2 ) 2 (c) 2

( (k m 2 ) 2 (c) 2 ) 2

(k m 2 ) sin t c cos t

(k m 2 )2 (c) 2

(k m 2 ) 2 (c) 2

(k m ) (c)
c
2 2

2 2

( (k m ) (c) )

sin(t )
cos( )=

(k m 2 )
( (k m2 ) 2 (c) 2 ) 2

c
(k m2 )

Steady state response:


u

cos t

2 2

sin( )

Fc

(k m 2 ) sin t c cos t

tg

sin t

F
2 2

(k m ) (c)

sin(t )

sin(t ) 1

Amplitude of steady state response:

17
Numerical Experimental Approach in Structural Dynamics

(k m2 )2 (c)2
3.4. Multiple Degree of Freedom Systems (m-DOF)

- generalised coordinates >=2


3.4.1 Category of m-DOF Systems
1. Systems with lumped-parameters consisting of only a
few degrees of freedom
2. Systems - structures which are described by using the
Finite Element Method (FEM) Cartesian (x,y,z)
coordinates . The displacement coordinates at nodes or
grid points usually with rotations about them, known as
global coordinates.
The basic method for setting up the eq. of motion
Lagrange method.
Other methods for setting up the differential eq. of motion for
undamping systems
1. The Newtons second law
2. The DAlembert principle generating the static stiffness (or
flexibility) matrix and adding the inertia forces
3. Lagrange eq. the most common for s-DOF and m-DOF
d T
T U
( )

0, i 1, 2,...n
(1)
dt qi
qi qi
where,
T total kinetic energy in the system
U Potential energy in the system
qi - number of generalised coordinates (degrees of freedom).

Pentru un sistem cu m-DOF, ecuaia Lagrange poate fi scris conform cu (2), atunci
cnd energia cinetic nu depinde de elongaia maselor i amortizarea este neglijat:

d T
U
( )
Qi , i 1, 2,...n
dt qi
qi

(2)

- Matriceal methods based on:


- (mass and stiffness matrices) using global coordinates

18
Numerical Experimental Approach in Structural Dynamics
- Modal coordinates.
- Transformation of generalized coordinates in modal
coordinates and modal matrix highlighting eigenvectors
associated eigenvalues of the system.
Example: For 2-DOF illustrated in Fig. 1 determine eq. of
motion.

- Fig. 1 2 DOF

m1 : F1 m1 z1 k1 z1 k2 ( z2 z1 ) 0
m2 : F2 m2 z2 k2 ( z2 z1 ) 0
m1 0 z1 k1 k2
0 m z k

2 2
2

(3)

k2 z1 F1

k2 z2 F2

The eq. of motion from stiffness matrix


If remove the masses- Fig 1 c):

k1 k2 k2 z1 F1
k
z F
k

2
2 2 2
.
F1 m1z1 k1 k2 k2 z1
F m z k
k2 z2
2
2 2
2

(4)

Eq. Of motion can be obtained by regarding inertia forces as


additional external forces acting on the stiffness matrix.
3.4.2 Eq. Of motion using Lagrange
It is defined the generalised coordinates
q1 z1 Q1 F1
.
q2 z2
Q2 F2

19
Numerical Experimental Approach in Structural Dynamics
1
1
(5)
T m1z12 m2 z22
2
2
1
1
U k1z12 k2 ( z2 z1 ) 2
2
2
1
1
1
k1z12 k2 z22 k2 z1z2 k2 z12
(6)
2
2
2
1
1
(k1z12 k2 z22 ) k2 z22 k2 z1z2
2
2
d T
U
( )
Qi
dt qi
qi
The matrix form is obtained:
m1 : F1 m1 z1 k1 z1 k2 ( z2 z1 ) 0

m2 : F2 m2 z2 k2 ( z2 z1 ) 0
m1 0 z1 k1 k2
0 m z k

2 2
2

k2 z1 F1

k2 z2 F2

3.4.3 Matrix representation for m-DOF


- For larger multi-DOF, the aim is to transform the basic
mass and stiffness elements to eq of motion in global
coordinates.
- Displacements expressed as global coordinate

r [ X m ] z r [ X m ] z,

(7)

Unde:
r vector of displacement of the individual mass
z vector of displacement of global coordinate
1
1
1
T m1r12 m2 r22 ...
T rT m r
(8)
2
2
2

m - diagonal matrices with masses.If one substitute (7) in (8),


it is obtained:
T

1 T
z [ X m ]T m [ X m ]z
2

(9)

M
M- global coordinates mass matrix.
The potential energy stored in single spring of stiffness.
If one would express om global coordinates:

20
Numerical Experimental Approach in Structural Dynamics
s [ X s ] z ,
(10)
We will obtained

1 T
z X s T KX s z
2

(11)

K
Obs:
Global coordinates describe translation and rotation
movements of nodes and indicates only for point considered.
Modal Coordinates
Define amplitude deviation known as mode= mod,
caracteristica.
Two classes of modes:
I.
First class Pattern Mode are used with normal
modes (in FEM) method
II. The second class Normal modes
Matricea sistemului s fie diagonal separnd gradele
de libertate astfel nct
serie de sisteme cu un sDOF.
-The normal modes imply eigenvalues and eigenvectors
determination.
Thus changing a single coordinated movement of the
whole system modal affects a part of the structure

3.4.3.1 Transformation from global coordinates to


modal coordinates
Eq. Of motion:
[M ] z [M ]z F - global coordinates

[ M ] q [ K ]q Q - modal coordinates
Define the transform

21
Numerical Experimental Approach in Structural Dynamics
z [ X ] q

z [ X ] q zT qT [ X ]T
zT qT [ X ]T

(12)

[ X ] - Square matrix if vectors z and q have the same


number of elements
Applying (12) are obtained the matrices in modal coordinates

[ M ] [ X ]T [ M ] [ X ]

z [ X ] q [ K ] [ X ]T [ K ] [ X ]
[Q] [ X ]T F

(13)

3.4.3.2 Normal undamping modes


In general there is coupling between modes of a system
- there is a form of transformation of coordinates which allows
total decoupling of the modal coordinates
matrix [ X ] is built of its columns which are
eigenvectors associated with eigenvalues (normal modes
transformation).
The number of eigenvalues = nr. Degrees of freedom
Eigenvalues
- i R and eigenvectors for undamping systems
- i C and eigenvectors for systems with proportional
damping
- .
Advantages of normal modes for undamping systems:
Eq. are transformed into state-space model, mass and
stiffness matrix are diagonal matrices
- Therefore systems can be divided into several type of sDOF (uncoupled systems).
3.4.3.3 Damped normal modes
It is considered damping in the system.

22
Numerical Experimental Approach in Structural Dynamics

Fig. 2 Damping m-DOF


Eq. Lagrange
(14)
d T
D U
( )

Fi , i 1, 2,...n
dt qi
qi qi
Where D Dissipation energy calculated as T and U.

1
1
D c1z12 c2 ( z2 z1 )2
2
2

(15)

m1 : F1 m1 z1 k1 z1 c1 z1 c2 ( z2 z1 ) k2 ( z2 z1 ) 0
m2 : F2 m2 z2 c2 ( z2 z1 ) k2 ( z2 z1 ) 0
m1 0 z1 (c1 c2 ) c2 z1 k1 k2
0 m z c
c2 z2 k2

2 2
2

(16)

k2 z1 F1

k2 z2 F2

The eq. (16) became:

m1 0 z1 (c1 c2 ) c2 z1 k1 k2 k2 z1 F1
0 m z c
z k
z F
c
k

2 2
2
2 2
2
2 2 2
m11 m12
z1 c11 c12
z1 k11 k12
z1 F1
m
z c
z k
z F
m22
c22
k22
21
2
2
1
2
21



2 2




C
- Damping matrix C is a square, symmetric marix
comparable to the mass and stiffness matrices.
There are 2 ways of solving eq (17) : for particular forms of
applied forces.
Mz Cz kz F

23
Numerical Experimental Approach in Structural Dynamics
1. Solution by the normal mode summation method
2. Direct solution of global eq. of motion.
- transformation does not always make de C diagonal,
together with mass and stiffness the solution is to use
complex eigenvalues and eigenvectors.-> extra
complication ussualy avoided except the cases:
(1) Rotating systems such as helicopter rotors, where
centrifugal forces can lead to damping coupling between
the modes.
(2) Systems such as vehicle suspensions where discrete
hydraulic sampers may couple the modes

III. Serii Fourier i Transformata Fourier

x(t ) x(t T ), t R semnal periodic

Co(sinusoida) 0

SLIT

Co(sinusoida) 0
Amplitudine i faz diferit

24
Numerical Experimental Approach in Structural Dynamics
- Excitaii periodice: asociate cu mainile rotative,
propulsoare, rotorul elicopterelor, etc.
- Semnal periodic de intrare (for) descompunere n serie
Fourier
- Amplitudinea rspunsului este determinat dac o
component din seria Fourier coincide n frecven cu
frecvena de rezonan.

(1)

frecven fundamental
not

0
n

x(t ) a0 an cos(nt ) bn sin(nt ), t R

(2)

i 1

III.1 Determinarea coeficienilor dezvoltrii n serie Fourier


trigonometric
Determinarea constantei an :
(1)/ * cos(nt )

(3)
Observaii:
Funciile cos(nt ) i sin(nt ) formeaz un set ortogonal de funcii :
Produsul scalar a dou funcii:

(4)

25
Numerical Experimental Approach in Structural Dynamics

(5)
T
2
T

x(t )cos(nt )

an
T
2

Determinarea constantei bn :

(1) / sin(nt )

(6)
Conform (5)

T
2
T

x(t )sin(nt )

bn
T
2

(7)
irul a0 , a1, a2 ,...b0 , b1, b2 ,... constituie o alt reprezentare a
semnalului x(t). Semnalul original poate fi reconstitui perfect pe
baza coeficienilor .

26
Numerical Experimental Approach in Structural Dynamics
Reprezentarea sub form trigonometric
greu interpretabil
datorita prezenei sinusoidei i cosinusoidei.

Forma armonic mai intuitiv


face s apar o singur
component sinusoidal pt fiecare frecven.
III.2 Forma armonic
Se definesc amplitudinea i faza cosinusoidelor n coordonate
polare:

An an2 bn2
b
n arctg n
an
an An cos(n )

(8)

bn An sin(n )
n

x(t ) a0 an cos(nt ) bn sin(nt ), t R


i 1

x(t ) An cos n cos(nt ) An sin n sin(nt ), t R


i 1

x(t ) An cos(n nt )

(9)

n 0

irul de coeficieni A0 , A1, A2 ,... spectrul de amplitudini ale


semnalului x(t), iar 0 ,1,2 ,... spectrul de faze.

Exemplu 1:
Fig. 1 : 2 vectori care se rotesc fiecare de lungime an , respectiv
bn i proieciile lor an cos(n0t ), bn sin(n0t ) .
d n rezultanta celor 2 vectori la frecvena n0 :

d n an2 bn2 .

27
Numerical Experimental Approach in Structural Dynamics

Fig. 1 Componente Fourier vectori care se rotesc

n arctg

an
bn

Reprezentarea n serie sin n funcie de modul i faz:

an cos(n0t ) bn sin(n0t ) dn sin(n0t n )


n

x(t ) a0 an cos(n0t ) bn sin(n0t ), t R


i 1

x(t ) a0 d n sin(n0t n )
n 1

Reprezentarea n serie cos n funcie de modul i faz

n arctg

bn
an

x(t ) a0 d n cos(n0t n )
n 1

Utiliznd identitatea sin A cos( A

d n sin(n0t n ) cos(n0t n )
2
Din Fig. 1 n n

d n sin(n0t n )
d n cos(n0t n )

28
Numerical Experimental Approach in Structural Dynamics

III.3 Forma complex


- Reprezint cea mai general descompunere n serie
Fourier

e jx e jx
e jx e jx
(10)
cos( x)
,sin x j
2
2

(10) A
x(t ) An cos(n nt ) n (e j (n nt ) e j (n nt )
2
n 0
n 0

An j (n nt )
A
(e
) n (e j (n nt ) )
2
2
n 1
n

A0

(11)

Se introduce notaia

(12)

Seria (11) devine:

x(t )

Ance jnt dezvoltarea n serie Fourier armonic

III.4 Transformata Fourier


Definiie: Fie x(t) un semnal de modul integrabil:

(13)
Se definete Transformata Fourier a semnalului x(t) :

29
Numerical Experimental Approach in Structural Dynamics

X ( ) F x(t ) ( )

x(t )e jt dt

(14)

Semnalul original x(t) poate fi recuperat din transformata sa prin


aplicarea operatorului invers:

1
x(t ) F X ( ) (t )
2
1

X ()e

Ec. (14) i (15) utilizate n vibraii aleatoare

Exemplu 2

jt

(15)

30
Numerical Experimental Approach in Structural Dynamics

Semnificaia X ( ) este aceeai cu cea a coeficienilor Anc

X ( ) X ( ) e j ( )

III.4 Transformata Fourier discret

31
Numerical Experimental Approach in Structural Dynamics
- Transformata Fourier a semnalului x(t) se poate calcula dac se
cunoate x(t).
- n analiza vibraiilor sunt cunoscute seturi de date discretizate
la intervale constante de timp
n

x(t ) a0 an cos(n0t ) bn sin(n0t )

(16)

i 1

x(t ) a0 2 ak cos(
i 1

2 kt
2 kt
) bn sin(
)
T
T

(17)

an
2 T
2 kt
x
(
t
)cos(
)
dt

k
T 0
T
2
b
2 T
2 kt
bn x(t )sin(
)dt bk n
T 0
T
2
an

(18)

Considernd:

2 kt
T

cos(

2 kt
2 kt
) j sin(
)
T
T

(19)

Forma general a coeficienilor din (18) poate fi scris:

Xk

j
1 T
x(t )e
T 0

2 kt
T

dt , X k ak jbk

(20)

32
Numerical Experimental Approach in Structural Dynamics
n msurtorile reale, x(t) devine o serie de valori discrete de timp

T
T
, t p p ,
N
N

la intervale egale

N - nr. total de valori din x(t)


T - perioada transformat n secunde.

k
, k 0,1,...N 1
T

Xk

j
1 T
x(t )e
T 0

2 kt
T

j(
1 N 1
x pe
T j 0

N 1

x pe

j(

dt , X k ak jbk

2 k
) p
T

(21)

2 kp
)
N

p 0

DFT - Discrete Fourier Transform a seriei discrete

x p ( x0 , x1, x2 ,...xN 1)
2 kp
N 1
j(
)
N
X ke
IDFT - x p
k 0

Xk

(22)

an
b
j n
2
2

Proprietile Transformatei Fourier


1.

Liniaritatea
Fie x(t) i y(t) dou semnale de modul integrabil i a, b C
Liniaritatea:

F ax(t ) by (t ) ( ) aF x(t )} bF{ y (t ) ( )

2. Deplasarea n timp

(23)

33
Numerical Experimental Approach in Structural Dynamics
Deplasarea n timp cu o cantitate constant t0 a unui semnal
corespunde undei deviaii induse n faza spectrului:

F x(t t0 ) ( ) X ( )e jt0
3. Deplasarea n spectru
Deplasarea spectrului unui semnal cu o frecven constant

0 corespunde nmulirii semnalului n timp cu o cosinusoid


complex:

F 1 X ( 0 ) (t ) x(t )e j0t

4. Schimbarea de scal
O contracie a semnalului cu o constant a corespunde unei
relaxri a spectrului cu aceeai constant i vice-versa

F x(at ) ( )

X( )
a
a

5. Derivarea n timp

dx(t )
F
( ) j X ( )
dt

6. Integrarea n timp

F
7. Simetria

x(t ) () j1 X ()

F x(t ) ( ) X ( )
F x(t ) ( ) 2 x( )

8. Conservarea energiei semnalului

x 2 (t )dt

1
2

X ( ) d

34
Numerical Experimental Approach in Structural Dynamics
9. Convoluia n timp
Spectrul semnalului obinut prin convoluia temporal a dou
semnale se obine ca produsul spectrelor celor dou semnale

z (t ) x(t ) * y (t )

x( ) y(t )d

x(t ) y( )d

Atunci ntre transformatele Fourier ale celor 3 semnale are


loc:

Z ( ) X ( ) Y ( ), R
10.

Convoluia n frecven

Spectrul semnalului obinut prin produsul a dou semnale se


obine prin convoluia spectrelor celor dou semnale

1
1
Z ( )
X ( ) * Y ( )
X (0 )Y ( 0 ) d0
2
2
1

X ( 0 )Y (0 )d0
2

Exemplu 3
Rspunsul unui sistem s-DOF la aplicarea unei fore
periodice-sinusoidale F P sin t (determinarea deplasrii
z).

Fig. 3 S-DOF cu resort si amortizor


Descrierea forei serie Fourier de forma (2)

35
Numerical Experimental Approach in Structural Dynamics
n

F (t ) a0 an cos(n0t ) bn sin(n0t ), t R

n 1
Utiliznd expresia seriei n funcie de amplitudine i faz:

F (t ) a0 d n cos(n0t n ) ,
n 1

bn
an
Ecuaia micrii: F mz cz kz
2
Forma standard cu k m n , c 2mn
d n an2 bn2 ,

n arctg

F / m z 2n z n 2 z
Deplasarea la fora sinusoidal F are forma simplificat:

z C cos t D sin t

Das könnte Ihnen auch gefallen