Applications of Eigenvalues & Eigenvectors

Louie L. Yaw

Walla Walla University

Engineering Department

For Linear Algebra Class

November 17, 2014
Outline

1 The eigenvalue/eigenvector problem

2 Principal stresses and directions in stress analysis

3 Fundamental frequencies and mode shapes in vibrations

4 Critical loads and buckled shapes in buckling analysis

5 Applications in electrical area/ feedback and control

6 Principal mass moment of inertia in 3D

1. The eigenvalue/eigenvector problem

Find vector v and scalar λ that satisfies the following:

Kv = λv (1)

where,
K = nxn matrix
v = nx1 vector, an eigenvector
λ= a scalar, an eigenvalue

n vector and scalar pairs will satisfy equation (1). How to find
such pairs is a linear algebra problem.

Equation (1) arises on occasion during the solution of various

types of engineering problems
1. The eigenvalue/eigenvector problem
An example (3x3 matrix ⇒ 3 eigenvalues and 3 eigenvectors):
    
τxx τxy τxz v1 v1
τxy τyy τyz  v2  = τ v2  (2)
τxz τyz τzz v3 v3
    
10 8 0 v1 v1
 8 14 5 v2  = τ v2  (3)
0 5 9 v3 v3
MATLAB solution to eigenvalue problem:
   
−0.64 0.55
τ1 = 2.17, v =  0.62  , τ2 = 9.29, v = −0.05 , (4)
−0.46 −0.83
 
−0.54
τ3 = 21.54, v = −0.78
−0.31
2. Principal stresses and directions in stress analysis

Consider a two dimension problem. A cantilever beam loaded

at its free end.
Deflected Shape

Y P 2

y−distance (in)
X 0

−1
A
−2

−3 Deflection Magnification Factor =10

−4

−5
0 2 4 6 8 10
x−distance (in)
2. Principal stresses and directions in stress analysis

Stress block from point A

y
τyy = 0
τyz = 0 Stress Field τxx (ksi)

τyx = 3.5 1 50
τxz = 0 0.5
A
τzy = 0 τxx = 15 0 0

x −0.5
τxy = 3.5 −1 −50
τzx = 0
0 2 4 6 8 10
τzz = 0 Stress Field τ (ksi)
yy

Units of ksi 1
0.2
z
0.5 0.1
A
0 0
 
τxx τxy τxz −0.5
−1
−0.1

−0.2

τ = τxy τyy τyz  0 2 4 6 8 10

Stress Field τ (ksi)

xy

τxz τyz τzz 1

0.5 −1
A
  0 −2

15 −3.5 0 −0.5
−1
−3

τ = −3.5 0 0 0 2 4 6 8 10

0 0 0
2. Principal stresses and directions in stress analysis

MATLAB solution to eigen problem:

 
0.2166
τ1 = −0.78, v1 = 0.9763 ,
y v1
0.0000
τ1 = 0.78
 
12.5◦
0.0000
τ2 = 0.00, v2 = 0.0000 ,
τ3 = 15.8
x
1.0000
τ2 = 0 v3
 
z
v2
Units of ksi
0.9763
τ3 = 15.8, v3 = −0.2166
0.0000
3. Fundamental frequencies and mode shapes in
vibrations
Numerical Eigen problem solution

2D FEA of cantilever First Five Mode Shapes of Cantilever

x & y dofs at 189 nodes 0

Mode =1, Frequency =1.5351 Hz

(2 x 189 = 378 dofs)

Mode =2, Frequency =8.8765 Hz

⇒ 378 eigenvalues & −5

eigenvectors Mode =3, Frequency =17.8059 Hz

⇒ 378 frequencies & −10

mode shapes Mode =4, Frequency =22.2066 Hz

Eigenvectors are 378th −15

dimensional vectors! Mode =5, Frequency =37.8254 Hz

−20

K̄v = λv, K̄ = 378x378 in

this example! −2 0 2 4 6 8 10 12 14 16

Finer Grid=More accurate

4. Critical loads & buckled shapes in buckling analysis
P P

L θ3
3

k3
θ2
L
3

k2 undef ormed

L
θ1 def ormed
3
small angles assumed
k1
rotation spring

A deformed shape equilibrium analysis results in the following:

    
(k1 + k2 ) −k2 0 θ1 θ
PL  1 
 −k2 (k2 + k3 ) −k 3 θ2  = θ2
3
0 −k3 k3 θ3 θ3

Kθ = λθ
4. Critical loads & buckled shapes in buckling analysis

12

10

For the case of

8
L = 10 inches
k1 = 30 kip-in/rad 6

k2 = 30 kip-in/rad 4

k3 = 30 kip-in/rad 2

Eigenvalue = Pcr3 L
−2 Buckling Mode =1 Buckling Mode =2 Buckling Mode =3

Eigenvector ⇒ shape
−4 Pcr =1.7826 Pcr =13.9946 Pcr =29.2228

−10 0 10 20 30 40 50
5. Applications in electrical engineering
- feedback and control

Outline of conceptual feedback and control

Model dynamic system such as airplane, car, rocket
Mφ̈ + Cφ̇ + Kφ = F(t)
The mathematical model of the system has inherent
eigenvalues and eigenvectors
Eigenvalues describe resonant frequencies where the
system will have its largest, often excessive, response.
We can choose F(t) to reduce the system response at the
resonant frequencies.
5. Applications in electrical engineering
- feedback and control

Perhaps let F(t) = Aφ̇ + Bφ and insert into dynamic

system model
The new system is Mφ̈ + C̄φ̇ + K̄φ = 0
where C̄ = C − A = velocity dependent damping
and K̄ = K − B = displacement dependent forcing
We can adjust A and B so that the eigenvalues of the new
‘barred’ system are different from the original system
By doing so we cause (control) the system to avoid
excessive vibration or instability

R
S
Q
6. Principal mass moment of inertia in 3D

3D kinetics of a rigid body

Inertia tensor (components dependent on X̄ Ȳ Z̄ coordinate
axes orientation)
 
Ixx −Ixy −Ixz
I = −Iyx Iyy −Iyz 
−Izx −Izy Izz

It is possible to orient the axes such that

 
Ix 0 0
I =  0 Iy 0 
0 0 Iz
6. Principal mass moment of inertia in 3D
Ȳ
Z
Z̄
vz
Y
vy

vx

X̄
X

In X̄ Ȳ Z̄ coordinate system In XYZ coordinates

 
Ixx −Ixy −Ixz  
I = −Iyx Iyy −Iyz  Ix 0 0
−Izx −Izy Izz I= 0
 Iy 0
0 0 Iz
Ix , Iy , Iz are eigenvalues
vx , vy , vz are eigenvectors wrt X̄ Ȳ Z̄
Conclusion

A few comments:
Many applications of eigenvalues and eigenvectors in
engineering
K is not always symmetric
eigenvalues are not always positive or real
eigenvectors are orthogonal
eigenvalues are invariant wrt to choice of coordinate axes
Questions.

Kθ =?θ