You are on page 1of 22

MATLAB Programming –

Eigenvalue Problems and Mechanical Vibration

A ⋅ x = λx ( A − λI ) ⋅ x = 0

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
A Coupled Mass Vibration Problem

EOM:
mx1 + k1 x1 − k 2 ( x2 − x1 ) = 0
mx2 + k1 x2 + k 2 ( x2 − x1 ) = 0

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Vibration Solutions – harmonic response

⎡ x1 ⎤ ⎡ c1 ⎤
Trial solution: ⎢ x ⎥ = ⎢c ⎥ cos(ωt + ϕ )
⎣ 2⎦ ⎣ 2⎦

Matrix representation of EOM:

⎡− ω 2 + (k1 + k 2 ) / m − k2 / m ⎤ ⎡ c1 ⎤
⎢ ⎥⋅⎢ ⎥ = 0
⎣ − k2 / m − ω + (k1 + k 2 ) / m ⎦ ⎣c2 ⎦
2

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Vibrational Frequencies and Mode Shapes
Characteristic Equation (Determinant = 0 ):

k1 + k 2 − mω = ± k 2 2

k1 ⎡ c1 ⎤ ⎡1⎤
ω = 2
1 ⎢c ⎥ = A1 ⎢1⎥
m ⎣ 2 ⎦1 ⎣⎦
k1 + 2k 2 ⎡ c1 ⎤ ⎡1⎤
ω = 2
2 ⎢c ⎥ = A2 ⎢− 1⎥
m ⎣ 2 ⎦2 ⎣ ⎦

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Vibrations as a general class of “Eigenvalue Problems”
Recast EOM:
⎡− ω 2 + (k1 + k 2 ) / m − k2 / m ⎤ ⎡ c1 ⎤
⎢ ⎥⋅⎢ ⎥ = 0
⎣ − k2 / m − ω + (k1 + k 2 ) / m ⎦ ⎣c2 ⎦
2

As:

⎡(k1 + k 2 ) / m − k 2 / m ⎤ ⎡ c1 ⎤ ⎡ω 2 0 ⎤ ⎡ c1 ⎤
⎢ −k /m ⎥ ⋅⎢ ⎥ = ⎢ ⋅ ⎥
2⎥ ⎢
⎣ 2 (k1 + k 2 ) / m⎦ ⎣c2 ⎦ ⎣ 0 ω ⎦ ⎣ c2 ⎦
⎡(k1 + k 2 ) / m − k 2 / m ⎤ ⎡ c1 ⎤ 2 ⎡1 0⎤ ⎡ c1 ⎤
⎢ −k /m ⎥ ⋅⎢ ⎥ =ω ⎢ ⎥ ⋅⎢ ⎥
⎣ 2 (k1 + k 2 ) / m⎦ ⎣c2 ⎦ ⎣0 1 ⎦ ⎣c2 ⎦
A ⋅ x = λI ⋅ x
( A − λI ) ⋅ x = 0
Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Eigenvalue equation, Eigenvalues, Eigenvectors
Eigenvalue equation:
A ⋅ x = λx ( A − λI ) ⋅ x = 0

Eigenvalues (angular frequencies of the vibration):

λ =ω 2

Eigenvectors (mode shape of the vibration):

⎡ c1 ⎤
x=⎢ ⎥
⎣c2 ⎦

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Solving Eigenvalue Problem in MATLAB
Look at the problem numerically:
m = 1kg k1 = 1N / m k 2 = 2 N / m

Simple m-file:

m=1;
k1=1;
k2=2;
A=[(k1+k2)/m -k2/m; -k2/m (k1+k2)/m]
[X,L]=eig(A);
X
L
Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
MATLAB output of simple vibration problem
eigenvector 1 eigenvector 2
X=

-0.7071 -0.7071
-0.7071 0.7071

L=
eigenvalue 1
1.0000 0 eigenvalue 2
0 5.0000
Ok, we get the same results as solving
the characteristics equation… so what is the big deal?
Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
A more complex vibrations problem

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
EOM for a more complex problem

EOM can be gotten from free body diagrams of


each mass

m1 x1 + (k1 + k 2 + k5 ) x1 − k 2 x2 − k5 x3 = 0
m2 x2 + (k 2 + k3 + k 6 ) x2 − k 2 x1 − k3 x3 − k6 x4 = 0
m3 x3 + (k3 + k 4 + k5 ) x3 − k5 x1 − k3 x2 − k 4 x4 = 0
m4 x4 + (k 4 + k6 ) x4 − k6 x2 − k 4 x3 = 0

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Characteristic equation of more complex problem:

⎡ 2 k1 + k 2 + k5 k2 k5 ⎤
⎢− ω − m1 m1 m1
0⎥
⎢ ⎥ ⎡c ⎤
⎢ k2 k +k +k k3 k6 ⎥ ⎢ 1⎥
−ω2 − 2 3 6
⎢ m2 m2 m2 m2 ⎥ ⎢c 2 ⎥
⎢ k5 k3 k +k +k k4 ⎥ ⋅ ⎢c ⎥ = 0
⎢ −ω2 − 2 4 5 ⎥ ⎢ 3⎥
⎢ m3 m3 m3 m3 ⎥ ⎣c 4 ⎦
⎢ k6 k4 k +k ⎥
⎢ 0 −ω2 − 4 6 ⎥
⎣ m4 m4 m4 ⎦

Solving this and find roots is no longer so simple!


It is now an eighth order polynomial….
Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Look at more complex vibration as eigenvalue problem

⎡ k1 + k 2 + k5 k2 k5 ⎤
⎢ − − 0 ⎥
m1 m1 m1
⎢ ⎥ 0 0 0⎤ ⎡ c1 ⎤
⎢ − k2 k 2 + k3 + k 6 k k6 ⎥ ⎡ c1 ⎤ ⎡1
− 3 −
⎢ m2 m2 m2
⎢ ⎥
m2 ⎥ ⎢c2 ⎥
⎢0 1 0 0⎥⎥ ⎢⎢c2 ⎥⎥
⎢ ⎥ ⋅ = ω2 ⎢ ⋅
k k k 2 + k 4 + k5 k ⎢ ⎥ ⎢0 0 1 0⎥ ⎢c3 ⎥
⎢ − 5 − 3 − 4 ⎥ ⎢c3 ⎥ ⎢ ⎥ ⎢ ⎥
⎢ m3 m3 m3 m3 ⎥ c 0 0 1 ⎦ ⎣ c4 ⎦
⎣ 4⎦ ⎣0
⎢ k k k 4 + k6 ⎥
− 6 − 4
⎢ m4 ⎥⎦
0
⎣ m4 m4

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Solving more complex problem by MATLAB

Pick some numerical values:

m1 = 1kg m2 = 2kg m3 = 3kg m4 = 4kg


k1 = 1N / m k 2 = 2 N / m k3 = 3N / m
k 4 = 4 N / m k5 = 5 N / m k6 = 6 N / m

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Solving complex vibration problem by MATLAB

Create another m-file:

m1=1;m2=2;m3=3;m4=4;
k1=1;k2=2;k3=3;k4=4;k5=5;k6=6;
A=[(k1+k2+k5)/m1 -k2/m1 -k5/m1 0;
-k2/m2 (k2+k3+k6)/m2 -k3/m2 -k6/m2;
-k5/m3 -k3/m3 (k3+k4+k5)/m3 -k4/m3;
0 -k6/m4 -k4/m4 (k4+k6)/m4]
[X,L]=eig(A);
X
L

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Characteristic matrix of the eigenvalue problem

A=

8.0000 -2.0000 -5.0000 0


-1.0000 5.5000 -1.5000 -3.0000
-1.6667 -1.0000 4.0000 -1.3333
0 -1.5000 -1.0000 2.5000

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Frequencies and mode shapes of complex problem

X=
0.4483 0.9456 0.6229 -0.1650
0.5156 -0.1826 -0.0411 -0.9006
0.5031 -0.2591 0.5788 0.3164
0.5292 0.0735 -0.5246 0.2481

L=
0.0878 0 0 0
0 9.7562 0 0
0 0 3.4858 0
0 0 0 6.6702

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Multiple element vibration problems –
finite element simulation

Consider the vibrational


modes of a four stories building.
The mass are assumed to be
concentrated at the floors.
The walls constitutes springs.
This can be models as a
1-D system.

Inspired by Aladdin web site

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Write equation of motion for the four floors

m1 x1 = 2k1 ( x2 − x1 )
m2 x2 = 2k1 ( x1 − x2 ) + 2k 2 ( x3 − x2 )
m3 x3 = 2k 2 ( x2 − x3 ) + 2k3 ( x4 − x3 )
m4 x4 = 2k3 ( x3 − x4 ) − 2k 4 x4

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Write down the Mass and Stiffness Matrix

⎡m1 0 0 0⎤
⎢0 m2 0 0 ⎥⎥
M =⎢
⎢0 0 m3 0⎥
⎢ ⎥
⎣0 0 0 m4 ⎦

⎡ 2 k1 − 2 k1 0 0 ⎤
⎢− 2k ( 2 k1 + 2 k 2 ) − 2k 2 0 ⎥
K =⎢ 1 ⎥
⎢ 0 − 2k 2 (2k 2 + 2k3 ) − 2k3 ⎥
⎢ ⎥
⎣ 0 0 − 2k3 (2k3 + 2k 4 )⎦

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Using MATLAB Eig with Mass & Stiffness Matrix Directly

KV = MVD

Eig can also operate on the eigenvalue equation


In this form where:
K is the stiffness matrix, V is the matrix containing
All the eigenvectors, M is the mass matrix, and
D is a diagonal matrix containing the eigenvalues

[V,D]=eig(K,M)

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Mode shape of building oscillation
m1 = 1500 m2 = 3000 m3 = 3000 m4 = 4500
k1 = 400 k 2 = 800 k3 = 1200 k 4 = 1600

3.5

2.5

1.5

1
-0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015

ω 2 = 0.12 ω 2 = 0.59 ω 2 = 1.13 ω 2 = 2.08


1 2 3 4

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].
Mode shape of building oscillation 2
m1 = 1500 m2 = 3000 m3 = 3000 m4 = 4500
k1 = 400 k 2 = 800 k3 = 1200 k 4 = 1600
4

3.5

2.5

1.5

1
-0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02

ω 2 = 0.042 ω 2 = 0.70 ω 2 = 1.93 ω 2 = 2.83


1 2 3 4

Cite as: Peter So, course materials for 2.003J / 1.053J Dynamics and Control I, Fall 2007. MIT OpenCourseWare
(http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

You might also like