Eigenvalues
Eigenvalues
Eigenvalues
Mathematics (Seminar)
Professor : Dr.Farahmand
Alireza Mohammad jafari
Student ID:8519415105
Eigenvalue Problems
Engineering Problems involving vibrations,
elasticity, oscillating systems, etc.,
Determine the eigenvalues for n homogenous linear
equations in n unknowns
[ A]{ x } { x }
Non-homogeneous system
[ A] [ I ]{ x} {0}
: eigenvalue;
homogeneous system
x : eigenvector
Mathematical Background
A I x
a11
a
21
a22
a31
a32
an 1
an 2
a12
a1 n x 1
0
0
a23
a2 n x 2
a33
a3 n x 3 0
0
0
an 3
ann x n
a13
det( A I ) 0
( 1 ) N N C N 1 N 1 C N 2 N 2 C 2 2 C 1 C 0 0
The root of fn( ) = 0 are the solutions for the eigenvalues
Mass-Spring System
Equilibrium positions
d 2 x1
kx 1 k ( x 2 x 1 )
m1
2
dt
2
m d x 2 k ( x x ) kx
2
1
2
2 dt 2
Mass-Spring System
2
let x 1 X 1 sin t , x 2 X 2 sin t ;
Tp
d 2 x1
m1 dt 2 k ( 2 x 2 x 1 ) 0
2
m d x2 k( x 2 x ) 0
2
1
2
dt 2
2k
k
2
X 1
X2 0
m1
m1
2k
k
X 1
2 X 2 0
m2
m2
Homogeneous system
2 k / m1 2
k / m2
k / m2
2 k / m2 2
X1
0
0
X2
Polynomial Method
m1 = m2 = 40 kg, k = 200 N/m
Characteristic equation det[ ] = 0
( 2 k / m1 2 )
k / m2
10 2
2
k / m2
( 2 k / m2 )
5
5
10 2
4 20 2 75 0 ( 2 15 )( 2 5 ) 0
Tp = 2.81 s
X 1 = X2
Buckling of Column
2
EI
dx
M Py
d2y
P
2
p
y
2
EI
dx
y ( 0 ) y( L ) 0
Curvature:
d2y
dx 2
p
y ; y( 0 ) y( L ) 0
2
dx
Eigenvalue problem
y A sin px B cos px
Buckling loads
y( 0 ) B 0
pL n , n 1, 2,
y( L) A sin pL 0
n 2 2 EI
P p EI
L2
2
Fundamental mode: n = 1
2 EI
Pcritical
L2
Euler formula
Buckling
Modes
n
L
2 2
n
EI
2
P p EI
L2
p
Polynomial Method
ODE
d2y M
2
p
y ; y( 0 ) y( L ) 0
2
EI
dx
Finite-difference method
yi 1 2 yi yi 1
2
2 2
p
y
(
2
h
p ) yi yi 1 0
i
i 1
2
hi
2 h2 p 2
1
2 h2 p 2
1
0
0
0
1
2 h2 p 2
1
0
1
0
2 h 2 p 2
y1
0
y
0
2
y3 0
0
yn
Which
Scheme?
Order of
Errors?
th-order polynomial
Characteristic
equation:
2 2 (2n)
n
det ( 2 h p ) 0
Polynomial Method
One interior node (h = L/2)
L
2 2 2
( 2 h2 p 2 ) y1 0 p
( a 10 %)
h
L
pexact
2
pexact ,
L L
0
1 y1
2 2 2
(
2
h
p ) 10
2 2
2 h p y2
0
ph 1, 3 p
3 3 3
,
( a 4.5 %, 17.3 %)
L
L
Polynomial Method
Three interior nodes (h = L/4)
2 h2 p 2
1
0
2 3
pexact ,
,
L L L
y1
1
0
0
2 h2 p 2
1 y2 0
0
1
2 h 2 p 2 y3
( 2 h2 p 2 ) 3 2( 2 h2 p 2 ) 0
p
ph 2 , 2 2
4 2 2 4 2 4 2 2
,
,
( a 2.6 %, 10.0 %, 21.6 %)
L
L
L
Power Method
If is the eigenvalue of A, then
Ax x
A x A( Ax ) A( x ) Ax x
2
A 3 x A( A 2 x ) A( 2 x ) 2 Ax 3 x
A 4 x A( A 3 x ) A( 3 x ) 3 Ax 4 x
A m x m x
m : eigenvalue of A m ; x : eigenvector
Power Method
Power method for finding eigenvalues
1. Start with an initial guess for x
2. Calculate w = Ax
3. Largest value (magnitude) in w is the
estimate of eigenvalue
4. Get next x by rescaling w (to avoid the
computation of very large matrix An )
5. Continue until converged
Power method also gives you eigenvectors
Power Method
Start with initial guess z = x0
(k1 ) w k( 1 )
w ( 1 ) Az ( 1 )
z( 2 )
w ( 1 ) Az ( 1 )
(1) (1)
k
k
w ( 2 ) Az ( 2 )
z
(3)
rescaling
(k2 ) w k( 2 )
w ( 2 ) Az ( 2 )
(2) (2)
k
k
k is the
dominant
eigenvalue
If 1 2 3 n , then
Power Method
1. Initial guess z (1) x0 1,1,...,1
normalize z by biggest wk
Calculate Az (2) w(2) z (3) ; normalize z by biggest wk
M
... Calculate Az ( k ) w( k ) z ( k 1)
Az ( 1 )
2 8 10
A 8 3 4
10 4 7
1
(1)
z x 0 1
1
2 8 10 1
20
0.9524
8 3 4 1 15 21 0.7143
21
10 4 7 1
1
.
0
eigenvalue eigenvector
Example
Current estimate for largest eigenvalue is 21
Rescale w by eigenvalue to get new x
20
0.9524
w
1
x
15 0.7143
max(abs( w )) 21
21
1
.
0
0.9524
0.9524
2.3812
0.7143 21 * 0.7143 1.2382
1.0
1.0
1.6188
Norm
Ax x ( 2.3812 ) 2 ( 1.2382 ) 2 ( 1.6188 ) 2 3.1343
2 8 10 0.9524
17.619
8 3 4 0.7143 13.762
19.381
10 4 7
1.0
w
1
max(abs( w )) 19.381
2 8 10
Ax x 8 3 4
10 4 7
17.619
0.9091
13.762 0.7101
19.381
1
.
0
0.9091
0.9091
0.1203
0.7101 19.381 * 0.7101 0.3594
1.0
1.0
0.4496
Norm
Ax x ( 0.1203 ) 2 ( 0.3594 ) 2 ( 0.4496 ) 2 0.5880
Example
One more iteration
Az ( 3 )
2 8 10 0.9091
17.499
0.9243
10 4 7
1.0
1
.
0
2 8 10
Ax x 8 3 4
10 4 7
0.9243
0.9243
0.0147
0.7080 18.931 * 0.7080 0.1153
1.0
1.0
0.1440
Norm
Ax x ( 0.0147 ) 2 ( 0.1153 ) 2 ( 0.1440 ) 2 0.1851
Az ( 5 )
Az ( 6 )
Az ( 7 )
2 8 10 0.9243
17.513
10 4 7
1.0
2 8 10 0.9181
17.506
10 4 7
1.0
2 8 10 0.9206
17.508
10 4 7
1.0
2 8 10 0.9196
17.507
10 4 7
1.0
0.9181
0.7087
1.0
0.9206
0.7084
1.0
0.9196
0.7085
1.0
0.9200
0.7085
1.0
Norm Ax x
MATLAB
Example:
Power
Method
A=[2 8 10; 8 3 4; 10 4 7]
A =
2
8
10
8
3
4
10
4
7
[z,m] = Power_eig(A,100,0.001);
it
m
z(1)
1.0000
21.0000
2.0000
19.3810
3.0000
18.9312
4.0000
19.0753
5.0000
19.0155
6.0000
19.0396
7.0000
19.0299
8.0000
19.0338
9.0000
19.0322
error =
8.3175e-004
z
z =
0.9199
0.7085
1.0000
m
m =
19.0322
x=eig(A)
x =
-7.7013
0.6686
19.0327
z(2)
z(3)
z(4)
0.9524
0.7143
0.9091
0.7101
0.9243
0.7080
0.9181
0.7087
0.9206
0.7084
0.9196
0.7085
0.9200
0.7085
0.9198
0.7085
0.9199
0.7085
eigenvector
eigenvalue
MATLAB function
z(5)
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
MATLABs Methods
e = eig(A)
gives eigenvalues of A
[V, D] = eig(A)
eigenvectors in V(:,k)
eigenvalues = Dii (diagonal matrix D)
[V, D] = eig(A, B) (more general eigenvalue
w Bx A x
x Aw
1/
smallest
A=[2 8 10; 8 3 4; 10 4 7]
A =
2
8
10
8
3
4
10
4
7
max_it=100; tol=0.001;
[z,m] = InvPower(A,max_it,tol);
L =
1.0000
0
0
4.0000
1.0000
0
5.0000
1.2414
1.0000
U =
2.0000
8.0000
10.0000
0 -29.0000 -36.0000
0
0
1.6897
B =
2
8
10
8
3
4
10
4
7
[B] = [L][U]
A =
2
8
10
8
3
4
10
4
7
1.0000 12.7826 0.3000 1.0000 -0.5333
it =
1
2.0000 0.7123 0.1205 1.0000 -0.8013
it =
[L]
[U]
3.0000
0.6687
0.1167
1.0000 -0.8152
4.0000
0.6686
0.1163
1.0000 -0.8155
it =
it =
z
z =
0.1163
1.0000
-0.8155
m
=
0.6686
x=eig(A)
=
-7.7013
0.6686
19.0327
eigenvector
eigenvalue
MATLAB
function
Smallest
eigenvalue